For me, a fusion category (over $\mathbb{C}$) is just a tensor $F$ (the associator, with $6$ simple-object labels and $4$ fusion space indices) and a tensor $d$ (the quantum dimensions, with one simple-object label), such that the $F$-tensor fulfils the pentagon equation. (This seems to be the most useful language for physics.)

If furthermore the fusion category is unitary, this means that (one can find a basis for the fusion spaces such that) the $F$-tensor is unitary when interpreted as a linear map between two groups of its indices.

What extra structure/conditions do I have to add in this tensor language to make the (unitary) fusion category pivotal/spherical?

For example, it seems that a unitary fusion category is spherical if the $F$-tensor is symmetric under the full tetrahedral symmetry group (after normalizing with the square roots of the quantum dimensions of four of its simple-object labels). This seems not to be the most general case, however.


2 Answers 2


Any unitary fusion category has a canonical spherical structure. See Example 2.12 of this paper and the references therein. It also follows from a more general result Prop 8.23 in this paper. So you don't need to do anything in the unitary case.

For a non-unitary fusion category, I think you're looking for Bruce Bartlett's explicit description of pivotal structures. There he shows that you can think of a pivotal structure as a choice of a number $\gamma_i$ attached to each simple object such that the condition of Theorem 5.4 holds: $$\gamma_j \gamma_k \mathrm{Id} = \gamma_i T_{jk}^i \text{ whenever $X_i$ is a summand of $X_j \otimes X_k$}.$$ Here $T_{jk}^i$ are the pivotal operators which are defined from the F-matrices by Thm 4.17. (There's a slightly tricky point there which I'm unclear on, which is that the formula in 4.17 only works for a "fair basis" which should translate into some simple condition on how you've gauaged your F-matrix, but it's not obvious to me what that condition is.) The pivotal structure is spherical if all the $\gamma_i$ are $\pm 1$.

  • $\begingroup$ Ah, good to know. But does that mean that being "spherical" cannot even be formulated as an additional condition/structure on the level of $F$-tensors, or that those conditions/structures follow directly from the unitarity and pentagon equation? $\endgroup$
    – Andi Bauer
    Feb 3, 2018 at 17:28
  • $\begingroup$ Deleted some comments and incorporated and expanded their content as a new paragraph in the answer. $\endgroup$ Feb 3, 2018 at 21:08

This is really an answer to the question in the comment of Noah's answer.

Let $F$ be a solution to the pentagon equations for some fusion category $\mathcal C$. Pivotal structures on $\mathcal C$ should be in $1-1$ correspondence with solutions to the polynomial equations

\begin{equation} \epsilon_c^{-1} \epsilon_b \epsilon_a = \sum_{s=1}^{N_{b c^*}^{a^*}}\sum_{t=1}^{N_{c^* a}^{b^*}} F_{abc^*}^{\mathbf 1}\lbrack\begin{smallmatrix}i & c & 1 \\ 1 & a^* & s\end{smallmatrix}\rbrack F_{bc^*a}^{\mathbf 1}\lbrack\begin{smallmatrix}s & a^* & 1 \\ 1 & b^* & t\end{smallmatrix}\rbrack F_{c^* ab}^{\mathbf 1}\lbrack\begin{smallmatrix}t & b^* & 1 \\ 1 & c & i\end{smallmatrix}\rbrack \end{equation}

for all $i \in \lbrace 1,\ldots, N_{ab}^c\rbrace$, where $N_{ab}^c$ is the dimension of $hom(a \otimes b,c)$. The $\epsilon_a$ are referred to as pivotal coefficients. There are various ways in which the pivotal coefficients and quantum dimensions for a given solution can be played around with, given our initial solution $F$ to the pentagon equations. For ease, we'll go ahead and assume that our $F$ matrices use basis choices which agree with those in 1305.2229 (which are pretty standard choices), in which case the left and right quantum dimensions of an object $a$ are related to $(F,\epsilon)$ via

\begin{align} q_l(a) &= \epsilon_a\left ( F_{a^*a a^*}^{a^*}\lbrack\begin{smallmatrix}1 & 1 & 1 \\ 1 & 1 & 1\end{smallmatrix}\rbrack\right)^{-1}, & q_r(a) &= \left (\epsilon_a F_{a a^* a}^{a}\lbrack\begin{smallmatrix}1 & 1 & 1 \\ 1 & 1 & 1\end{smallmatrix}\rbrack\right)^{-1}. \end{align}

  • $\begingroup$ Just to check that I understand, there’s an equation for every quadruple (a,b,c,i) where i indexes a basis for Hom(a b,c)? $\endgroup$ Feb 4, 2018 at 14:37
  • $\begingroup$ Yes, and the answer has been updated. This is (partially) gone through in the proof of Proposition 3.12 of 1305.2229, particularly on pages 33-37 to go from a ribbon category $\mathcal C$ to the equations above, and then pages 41-43 to go from a set of coefficients back to a pivotal structure. $\endgroup$ Feb 4, 2018 at 20:25
  • $\begingroup$ One more question, in this version it seems like the sphericality condition isn't exactly $\epsilon=\pm 1$ because the F-matrices in the formulas for the dimensions are not the same F-matrix. But presumably you can always gauge the F-matrices in order to make those particular two equal for every a? $\endgroup$ Feb 5, 2018 at 14:19
  • 1
    $\begingroup$ Correct. Sphericality at this point is just that $\forall a, q_l(a)=q_r(a)$. The presence of the F-matrix entries here correspond to the specific choice of birth/death maps being used, as stated in (31). To obtain the condition mentioned on the pivotal coefficients, DHW choose new birth/death maps in (55) and (56) on page 43. Here (so no one has to go back and look) $u_a$ is the $F_{a a^* a}^a$ entry mentioned above. $\endgroup$ Feb 5, 2018 at 23:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.