Pictures of complete positivity in arbitrary dimension
Two fundamental contributions to categorical quantum mechanics are presented. First, we generalize the CP-construction, that turns any dagger compact category into one with completely positive maps, to arbitrary dimension. Second, we axiomatize when a given category is the result of this construction.
💡 Research Summary
The paper makes two fundamental contributions to the categorical foundations of quantum theory. First, it extends the well‑known CP‑construction (often called the CPM‑construction) beyond the usual setting of finite‑dimensional dagger compact categories, providing a formulation that works for arbitrary dimensions, including infinite‑dimensional Hilbert spaces and more general dagger compact categories. The authors achieve this by introducing an “environment structure” – a family of discarding maps ⊥ₓ : X → I that are compatible with the dagger – together with a purification axiom stating that every mixed morphism can be obtained by discarding an ancillary system after a pure morphism. Using these ingredients they define, for any object A, a CP‑object as A⊗A* and construct the completely positive morphism associated with a map f : A → B as
CP(f) = (id_B ⊗ ε_A) ∘ (f ⊗ f†) ∘ (η_B ⊗ id_{A*}),
where η and ε are the unit and counit of the duality. This definition reproduces the Kraus representation in a purely categorical way and does not rely on trace‑class operators, making it valid in infinite dimensions. They prove a representation theorem: the category CP(C) built from these morphisms is dagger‑compact, monoidal, and isomorphic to the traditional CPM(C) when C is finite‑dimensional, establishing that the construction is a genuine generalisation.
The second major contribution is an axiomatic characterisation of those categories that arise as a CP‑construction of some underlying dagger compact category. The authors identify three minimal axioms: (1) dagger‑compactness, providing the basic tensor and †‑structure; (2) causality (or normalisation of discarding), requiring that discarding after preparing the unit yields the identity on the monoidal unit; and (3) purification, demanding that every morphism factors through a pure morphism followed by discarding an environment. They show that any category D satisfying these axioms possesses a canonical environment structure, that purification is essentially unique, and that there exists a strong monoidal dagger functor establishing an equivalence D ≅ CPM(C) for a suitable C. The proof proceeds by constructing C from the “pure” subcategory of D, then demonstrating that the CP‑construction on C recovers D.
To illustrate the theory, the paper examines the category of (possibly infinite‑dimensional) Hilbert spaces with bounded linear maps, and the category of C∗‑algebras with completely positive maps. In both cases the generalized CP‑construction coincides with the standard notion of completely positive trace‑preserving (CPTP) maps, and Stinespring dilation emerges naturally from the purification axiom. This shows that the abstract framework captures the familiar operational structures of quantum information even when the underlying spaces are not finite‑dimensional.
Overall, the work removes the dimensional restriction that has limited the applicability of categorical quantum mechanics, opening the door to rigorous categorical treatments of continuous‑variable quantum optics, quantum field theory, and other infinite‑dimensional settings. Moreover, the axiomatic characterisation provides a clear criterion for when a given categorical model can be interpreted as a theory of completely positive processes, which is valuable for exploring generalized probabilistic theories, higher‑order quantum operations, and categorical notions of entropy. Future research directions suggested include extending the framework to “environment‑free” categories, incorporating higher‑order transformations, and studying categorical analogues of thermodynamic concepts within this generalized CP setting.
Comments & Academic Discussion
Loading comments...
Leave a Comment