Sımplısyal dağılımlar, çokyüzlüler ve kuantum bağlamsallığı
Serkan Doğan
Abstract
Quantum contextuality is the phenomenon, established by the theorems of Bell and of Kochen and Specker, that the predictions of quantum mechanics cannot be reproduced by any assignment of measurement outcomes that is independent of the measurement context. The sheaf-theoretic approach of Abramsky and Brandenburger formulates contextuality as an obstruction to gluing local probability distributions into a global one. The theory of simplicial distributions carries this program further by modeling both measurements and outcomes as simplicial sets: a simplicial distribution is a simplicial map p: X →D(Y ) from a measurement space X to the probability monad applied to an outcome space Y . This thesis develops the theory of simplicial distributions as a polytope-theoretic framework for studying quantum contextuality. The simplicial distributions on a fixed scenario form a convex polytope in standard form. Its vertices encode the extremal probabilistic models, and its convex structure captures contextuality: a distribution is noncontextual precisely when it lies in the convex hull of the deterministic distributions, and quantum states give points of this polytope through the Born rule. The central tool is the support criterion: a point of a standard form polytope is a vertex if and only if its support is inclusion-minimal, which reduces the study of vertices to a combinatorial problem about supports and carrier faces. The main original contribution concerns collapsing maps, which identify a simplex of the measurement space with a degenerate one. For the pullback π∗along a collapsing map π: X →X, the equality π∗(carr(p)) = carr(π∗(p)) of carrier faces is proved. As corollaries, being a vertex, being deterministic, being strongly contextual, and being contextual are all preserved and reflected under pullback, and the homotopical collapsing results in the literature are recovered by a uniform polyhedral argument. These tools are then applied to one-dimensional scenarios, where the measurement space is a graph. It is proved that removing a leaf edge carries vertices to vertices; consequently every vertex of the distribution polytope of a tree is deterministic, so tree scenarios admit no contextuality. The vertices of the loop scenario with outcomes in Zd are classified: they are in bijection with cycles in Zd, and are in general contextual. Finally, two equivalent necessary criteria for extremality are given, one graph-theoretic and one algebraic: the support graphs of a vertex contain no cycle, equivalently the associated incidence matrices do not admit −2 as an eigenvalue.