nLab fusion category



Monoidal categories

monoidal categories

With braiding

With duals for objects

With duals for morphisms

With traces

Closed structure

Special sorts of products



Internal monoids



In higher category theory




A fusion category is a rigid semisimple linear (Vect-enriched) monoidal category (“tensor category”), with only finitely many isomorphism classes of simple objects, such that the endomorphisms of the unit object form just the ground field kk.

Often one also assumes a braiding and speaks of a braided fusion category.


The name “fusion category” comes from the central examples of structures whose canonical tensor product is called a “fusion product”, notably representations of loop groups and of Hopf algebras and of vertex operator algebras.

Simple examples:


(graded vector spaces)

the category of finite-dimensional GG-graded 𝕂\mathbb{K}-vector spaces

Vect G fdim Vect_{G}^{fdim}

is a fusion category, with tensor product given by the tensor product of vector spaces and the binary operation of the group:

V gW h(VW) gh. V_g \otimes W_h \;\; \coloneqq \;\; (V \otimes W)_{g \cdot h} \,.

More generally, for

the above construction but with associator multiplied by the 3-cocycle applied to the GG-degrees of the 3 factors is again a fusion category

Vect G,ω fdim. Vect_{G, \omega}^{fdim} \,.

(Etingof, Nikshych & Ostrik 2005, item 1. on p. 584)


For 𝕂\mathbb{K} a field and GG a finite group (or finite super-group), whose order is relatively prime to the characteristic of 𝕂\mathbb{K}, then the category of representations Rep(G,𝕂)Rep(G, \mathbb{K}) is a fusion category.

(Etingof, Nikshych & Ostrik 2005, item 2. on p. 584)


Relation to weak Hopf algebras

Under Tannaka duality, every fusion category CC arises as the representation category of a weak Hopf algebra (Ostrik). However, this does not mean that every fusion category admits a fiber functor to the category of vector spaces Vect=kMod\text{Vect}= k-Mod.

Given any multi-fusion category CC, one can always construct a fiber functor F:CRModF:C\to RMod for RR the algebra spanned by a basis of orthogonal idempotents {v i} iI\{v_i\}_{i\in I} for II the equivalence classes of simple objects of CC. This functor is referred to in some sources as a generalized fiber functor. The endomorphisms of this functor then give a weak Hopf algebra that represents CC. In Hayashi 1999 (see there for the relevant definitions), this is computed as a coend, where one has that CRep(A)C\cong Rep(A) for A=coend(F *F:C op×CBmd(E))A= \text{coend}(F^*\otimes F: C^{op} \times C \to Bmd(E)), where E=R˙RE=\dot R\otimes R is equipped with a coalgebra structure

Δ(λ˙μ)= νIλ˙νν˙μ \Delta(\dot\lambda \mu) = \sum_{\nu\in I} \dot\lambda \nu\otimes \dot\nu \mu
ϵ(λ˙μ)=δ λ,μ \epsilon (\dot\lambda \mu) = \delta_{\lambda,\mu}

It is important to note that, generally speaking, CC may admit other fiber functor to different module categories RModRMod, as is the case for fusion categories of the form Rep(H)Rep(H) for HH a Hopf algebra, which admits both the fiber functor described above, as well as a fiber functor to Vect\text{Vect}.

Relation to pivotal and spherical categories

Fusion categories were first systematically studied by Etingof, Nikshych & Ostrik 2005. This paper listed many examples and proved many properties of fusion categories. One of the important conjectures made in that paper was the following:

Conjecture (Etingof, Nikshych, and Ostrik)

Every fusion category admits a pivotal structure.

Providing a certain condition is satisfied, a pivotal structure on a fusion category can be shown to correspond to a ‘twisted’ monoidal natural endotransformation of the identity functor on the category, where the twisting is given by the pivotal symbols.

Relation to extended 3d TQFT

Given the data of a fusion category one can build a 3d extended TQFT by various means. This is explained by the fact, see below, that fusion categories are (probably precisely) the fully dualizable objects in the 3-category MonCatMonCat of monoidal categories. By the homotopy hypothesis this explains how they induce 3d TQFTs.


Write MonCat bimMonCat_{bim} for the (infinity,3)-category which has as


With its natural tensor product, MonCatMonCat is a symmetric monoidal (infinity,3)-category.


A monoidal category which is fusion is fully dualizable in the (infinity,3)-category MonCat bimMonCat_{bim}, def. .

This is due to (Douglas & Schommer-Pries & Snyder 13).


Via the cobordism theorem prop. implies that fusion categories encode extended TQFTs on 3-dimensional framed cobordisms, while their O(3)O(3)-homotopy fixed points encode extended 3d TQFTs on general (not framed) cobordisms.

These 3d TQFTs hence arise from similar algebraic data as the Turaev-Viro model and the Reshetikhin-Turaev construction, however there are various slight differences. See (Douglas & Schommer-Pries & Snyder 13, p. 5).


Here are three things such that it’d be awesome if they were sorted out on this page:

  1. Kuperberg’s theorem saying that abelian semisimple implies linear over some field. Finite, connected, semisimple, rigid tensor categories are linear

  2. Some correct version of the claim that abelian semisimple is the same as idempotent complete and nondegenerate. Math Overflow question

  3. Good notation distinguishing simple versus absolutely simple? (is End(V)=kEnd(V) = k or just VV has no nontrivial proper subobjects).

Together 1 and 2 let you go between the two different obvious notions of semisimple which seem a bit muddled here when I clicked through the links.



Original articles:

Further review:

On the Tannaka duality to weak Hopf algebras:

The relation to 3d TQFT clarified via the cobordism hypothesis:

and for the case of modular tensor categories:

Discussion in terms of skein relations:

See also:

  • Math Overflow, Why are fusion categories interesting? [MO:q/6180]

Further work on their classification using finite groups:

  • Agustina Czenky: Diagramatics for cyclic pointed fusion categories [arXiv:2404.08084]

On a notion of fusion 2-categories:

On fusion categories invertible with respect to the Deligne tensor product:

Anyonic topological order in terms of braided fusion categories

Claim and status

In condensed matter theory it is folklore that species of anyonic topological order correspond to braided unitary fusion categories/modular tensor categories.

The origin of the claim is:

Early accounts re-stating this claim (without attribution):

Further discussion (mostly review and mostly without attribution):

Emphasis that the expected description of anyons by braided fusion categories had remained folklore, together with a list of minimal assumptions that would need to be shown:

An argument that the statement at least for SU(2)-anyons does follow from an enhancement of the K-theory classification of topological phases of matter to interacting topological order:

In string/M-theory

Arguments realizing such anyonic topological order in the worldvolume-field theory on M5-branes:

Via KK-compactification on closed 3-manifolds (Seifert manifolds) analogous to the 3d-3d correspondence (which instead uses hyperbolic 3-manifolds):

Via 3-brane defects:

Further discussion

Relation to ZX-calculus:

On detection of topological order by observing modular transformations on the ground state:

See also:

Last revised on July 4, 2024 at 16:52:01. See the history of this page for a list of all contributions to it.