A monoidal category has a braiding — natural isomorphisms τ_{A,B}: A ⊗ B → B ⊗ A satisfying the hexagon axioms. What additional condition must hold for the category to be symmetric rather than merely braided?
AThe tensor product ⊗ must be strictly associative, with no associativity isomorphisms needed.
BThe braiding must satisfy τ_{B,A} ∘ τ_{A,B} = id_{A⊗B} — swapping A and B, then swapping back, must return exactly to the original, making the swap self-inverse.
CEvery object must be isomorphic to its tensor-unit dual, so A ⊗ I ≅ A holds strictly rather than up to isomorphism.
DThe braiding must be a natural transformation with respect to all morphisms, not just a natural isomorphism with respect to objects.
A braided monoidal category already has τ_{A,B} and the hexagon coherence axioms. Symmetry adds exactly one condition: τ_{B,A} ∘ τ_{A,B} = id. This says swapping A⊗B to B⊗A and then swapping back gives the identity — the braiding is its own inverse. In a braided-but-not-symmetric category, τ_{B,A} ∘ τ_{A,B} ≠ id in general: a 'positive crossing followed by a negative crossing' leaves a residual twist that is not trivial. The self-inverse condition collapses the distinction between positive and negative crossings, making the swap genuinely commutative.
Question 2 Multiple Choice
Why can symmetric monoidal categories not be used to detect knot topology, while braided monoidal categories can?
ASymmetric monoidal categories have no natural isomorphisms between tensor products, making it impossible to represent strand crossings.
BIn a symmetric monoidal category τ_{B,A} ∘ τ_{A,B} = id, so over-crossings and under-crossings are identified — all crossing types are equivalent, Reidemeister moves are trivially satisfied, and no nontrivial knot invariants can emerge.
CBraided categories contain infinitely many objects, giving enough combinatorial structure to distinguish different knot types.
DSymmetric monoidal categories lack the hexagon axioms necessary to interpret the crossing diagrams used in knot theory.
Knot invariants arise in braided categories precisely because τ_{B,A} ∘ τ_{A,B} ≠ id: a positive crossing (A over B) is genuinely different from a negative crossing (B over A), and their composition is not trivial. This asymmetry allows braided categories to detect Reidemeister move II violations and construct polynomial invariants like the Jones polynomial. In a symmetric category, the self-inverse condition identifies positive and negative crossings — they are the same morphism. Every knot diagram is automatically trivial because any crossing can be undone without cost. The price of full commutativity (τ² = id) is insensitivity to topology.
Question 3 True / False
Most monoidal category can be equipped with a symmetric monoidal structure, as long as the objects form a set rather than a proper class.
TTrue
FFalse
Answer: False
Symmetric monoidal structure is an additional datum that may or may not exist — it cannot always be added. The braid groupoid Br and certain module categories over noncommutative rings are monoidal but do not admit symmetric structures because genuine commutativity of the tensor product would contradict the underlying algebraic structure. Non-commutativity is fundamental in these settings. Even when a symmetric structure exists, it is essentially unique (up to coherent isomorphism) — there is at most one symmetric monoidal structure compatible with a given monoidal structure, not a family of choices.
Question 4 True / False
MacLane's coherence theorem for symmetric monoidal categories guarantees that any two morphisms built from the structural isomorphisms (associators, unitors, braidings) that have the same source and target are equal.
TTrue
FFalse
Answer: True
The coherence theorem for symmetric monoidal categories extends MacLane's original monoidal coherence: in the symmetric case, any well-formed diagram built from associators, unitors, and braidings commutes. Practically, this means you can freely rearrange tensor products of objects in any order and the resulting isomorphisms are canonical — any two ways of going from A⊗(B⊗C) to (C⊗A)⊗B, say, give the same morphism. This is what licenses algebraic geometers and topologists to work with tensor products of sheaves, chain complexes, and spectra informally, without tracking which parenthesization or ordering was used.
Question 5 Short Answer
Explain the difference between a braided and a symmetric monoidal category in geometric terms. Why does the self-inverse condition τ_{B,A} ∘ τ_{A,B} = id matter?
Think about your answer, then reveal below.
Model answer: In a braided monoidal category, τ_{A,B} corresponds to a 'positive crossing' in a braid diagram: strand A passes over strand B. The reverse braiding τ_{B,A} is a separate 'negative crossing': strand B passes over strand A. These are topologically distinct — over-then-under is not the same as no crossing, and braided categories can detect this difference (this is the basis of quantum group knot invariants). In a symmetric monoidal category, the self-inverse condition τ_{B,A} ∘ τ_{A,B} = id collapses this distinction: crossing over and then crossing back is the identity, meaning positive and negative crossings are identified. Geometrically, you cannot tell which strand goes over — there is only one kind of crossing. The category loses knot-sensitivity but gains free commutativity: tensor products can be permuted in any order without tracking which crossing type was used, just like multiplication of commutative scalars.
The self-inverse condition is the categorical encoding of genuine commutativity: A ⊗ B and B ⊗ A are not just isomorphic (braided case) but isomorphic in a way that remembers no directionality (symmetric case). The canonical example is vector spaces over a field: V ⊗ W and W ⊗ V are canonically isomorphic via τ(v ⊗ w) = w ⊗ v, and applying τ twice gives back v ⊗ w — the self-inverse condition holds. No topological residue remains after double-swapping.