We demonstrate the existence of positive phase space density functions which yield the quantum mechanical marginal distributions of position and momentum.
The problem of expressing quantum mechanical expectation values as averages with respect to nonnegative density functions on phase space, by analogy with classical mechanics, is reexamined in the light of some earlier work on fuzzy phase spaces. It is shown that such phase space representations are possible if ordinary phase space is replaced by a so-called fuzzy phase space, on which the usual marginal distribution functions are redefined to conform to the fact that arbitrarily precise simultaneous measurements on position and momentum are not compatible with quantum mechanics. In the process a generalization of Wigner’s theorem on the nonexistence of phase space representations of quantum mechanics, which also satisfy the standard (classical) marginality conditions in position and momentum, is obtained. It is shown that a (continuous) representation of quantum mechanics exists on a given fuzzy phase space if an only if the corresponding confidence functions for position and momentum measurements satisfy the Heisenberg uncertainty relations.
We address the classical and quantum marginal problems, namely the question of simultaneous realizability through a common probability density in phase space of a given set of compatible probability distributions. We consider only distributions authorized by quantum mechanics, i.e., those corresponding to complete commuting sets of observables. For four-dimensional phase space with position variables q⃗ and momentum variables p⃗, we establish the two following points: (i) given four compatible probabilities for (q1,q2), (q1,p2), (p1,q2), and (p1,p2), there does not always exist a positive phase space density ρ(q⃗,p⃗) reproducing them as marginals; this settles a long standing conjecture; it is achieved by first deriving Bell-type inequalities in phase space which have their own theoretical and experimental interest. (ii) Given instead at most three compatible probabilities, there always exist an associated phase space density ρ(q⃗,p⃗); the solution is not unique and its general form is worked out. These two points constitute our “three marginal theorem.”
Marginal distributions in (bf2N)-dimensional phase space and the quantum (bfN+1) marginal theorem
We study the problem of constructing a probability density in 2N-dimensional phase space which reproduces a given collection of n joint probability distributions as marginals. Only distributions authorized by quantum mechanics, i.e. depending on a (complete) commuting set of N variables, are considered. A diagrammatic or graph theoretic formulation of the problem is developed. We then exactly determine the set of ``admissible'' data, i.e. those types of data for which the problem always admits solutions. This is done in the case where the joint distributions originate from quantum mechanics as well as in the case where this constraint is not imposed. In particular, it is shown that a necessary (but not sufficient) condition for the existence of solutions is nleqN+1. When the data are admissible and the quantum constraint is not imposed, the general solution for the phase space density is determined explicitly. For admissible data of a quantum origin, the general solution is given in certain (but not all) cases. In the remaining cases, only a subset of solutions is obtained.