Questions: Compact Closed Categories

Question 1

In the compact closed category FinVect_k, what is the dual X* of a vector space X, and what do the evaluation and coevaluation maps do?

• A. X* is the orthogonal complement of X; evaluation and coevaluation define the inner product structure
• B. X* = Hom(X, k) is the dual vector space; evaluation encodes the pairing X* ⊗ X → k, and coevaluation encodes k → X ⊗ X*, with both satisfying triangle identities
• C. X* is X itself (every finite-dimensional space is self-dual); evaluation and coevaluation are both the identity morphism
• D. X* is defined by topological compactness; evaluation maps open sets to closed sets under the duality

Answer: (B)

In FinVect_k, the dual of X is the linear dual Hom(X, k) — linear functionals on X. The evaluation map ε: X* ⊗ X → k sends (f, v) to f(v). The coevaluation map η: k → X ⊗ X* sends 1 to Σᵢ eᵢ ⊗ eᵢ* (the sum over a basis and dual basis). These satisfy the triangle (snake) identities: composing η with ε in the right order recovers the identity on X (and on X*). This structure is the concrete instantiation of compact closure, and the triangle identities are what the abstract definition demands.

Question 2

The triangle (snake) identities in a compact closed category express that:

• A. The tensor product is associative and the swap isomorphism is its own inverse
• B. Composing the coevaluation and evaluation on the appropriate factors yields the identity morphism — a wire bent into a U and then straightened is the same as an unbent wire
• C. Every object is canonically isomorphic to its double dual via a natural transformation
• D. The categorical trace of any identity morphism equals the dimension of the tensor unit

Answer: (B)

The triangle identities state: (idX ⊗ ε) ∘ (η ⊗ idX) = idX and (ε ⊗ idX*) ∘ (idX* ⊗ η) = idX*. In string diagram notation, these become the 'snake equations': drawing a coevaluation curve followed by an evaluation curve on the same wire simplifies to a straight wire. This is the fundamental identity that allows wire-bending in the graphical calculus — you can redirect any input wire into an output wire (or vice versa) using the dual structure, and the snake equations guarantee the manipulation is consistent.

Question 3

In a compact closed category, every object has a dual, and morphisms can be 'bent' in string diagrams — an input wire of type X can be redirected into an output wire of type X* using the coevaluation map.

• T. True
• F. False

Answer: True

This is the defining feature of compact closed categories and the source of their power in the string diagram calculus. The coevaluation η: I → X ⊗ X* allows a new X-wire and X*-wire to be created from nothing (from the monoidal unit), effectively 'bending' an input into an output. The triangle identities ensure this bending is internally consistent — bending and then straightening leaves the morphism unchanged. This diagrammatic flexibility is what makes compact closed categories the natural setting for quantum information protocols like teleportation.

Question 4

The term 'compact' in compact closed categories is related to topological compactness — objects correspond to compact topological spaces, which have finite open cover properties.

• T. True
• F. False

Answer: False

This is the most common misconception about compact closed categories. 'Compact' here is an algebraic notion capturing something like finite-dimensionality — specifically, the existence of well-behaved dual objects with evaluation and coevaluation satisfying triangle identities. It has no relationship to topological compactness (the Heine-Borel property). The connection is to finite-dimensional vector spaces, not to compact topological spaces. The Common Misconceptions section of this topic explicitly flags this confusion.

Question 5

Why does the compact closed structure specifically capture finite-dimensionality, and why do infinite-dimensional vector spaces fail to form compact closed categories in the standard sense?

Model answer: The compact closed structure requires evaluation and coevaluation maps satisfying triangle identities. For finite-dimensional vector spaces, the coevaluation η: k → V ⊗ V* can be defined as η(1) = Σᵢ eᵢ ⊗ eᵢ* using any finite basis — this is a well-defined element of V ⊗ V* because the sum is finite. For an infinite-dimensional space, the analogous sum would be an infinite series of simple tensors, which is not an element of V ⊗ V* (the algebraic tensor product). Even using topological tensor products, the resulting maps fail the triangle identities in the required categorical form. The finite basis is essential: without it, the coevaluation cannot be constructed, and the dual pair structure collapses.

This is why compact closed categories appear in finite-dimensional quantum mechanics (finite-dimensional Hilbert spaces) and linear logic (which restricts resource use to prevent infinite accumulation). The 'compact' condition is a categorical way of saying 'there exists a finite basis that lets you build the dual pairing.' Infinite-dimensional spaces require much more careful treatment — they can have duals in weaker senses (conjugate-linear duals, continuous duals), but these do not satisfy the strict compact closed conditions.