Every paper I've found claims this is true but nobody actually proves it.

Let $(C,\otimes)$ be a strict monoidal category with duals and for every object $a\in C$ an evaluation $\epsilon_a:a \otimes a^* \longrightarrow \mathbb{F}$ and a coevaluation $\tilde{\epsilon}_a:\mathbb{F}\longrightarrow a\otimes a^*$ (i.e. $C$ is strict pivotal). These maps give a notion of (left and right but assume they are equal) trace of endomorphisms. Let $f \in$ End$(a)$, then we can define its trace as the endomorphism of the field given by the composition $$\mathbb{F}\xrightarrow{\tilde{\epsilon}_a} a\otimes a^*\xrightarrow{f\otimes id}a\otimes a^* \xrightarrow{\epsilon_a} \mathbb{F}.$$ Now the claim is that if $f\in$End$(a), g \in$End$(b)$ then $$tr(f\otimes g) = tr(f)\cdot tr(g)$$ But I have no idea how to show this. I know that this is true for the usual matrix trace but in this general context I don't know how to approach it. Any advice would be greatly appreciated.


So to clarify, this is true in a spherical category (when you assume that left and right traces are equal, this is the definition of a spherical category).

The best way (IMO) to prove this is through graphical calculus. Here is a picture (which I hope is not too horrible).

pictorial proof of the equality

The only non-obvious equality is the first: it follows from the monoidality of the duality functor $(X \otimes Y)^* \cong Y^* \otimes X^*$ (with the corresponding isomorphisms for evaluation and coevalutation). Then the definition of the trace of $g$ is used, equality of left and right traces, and again the definition of the trace of $g$.

To prove this equality without the graphical calculus would be a feat into itself, but it should be possible to convert the picture above to a symbolic computation. But be prepared to write things like $$\operatorname{ev}_{Y \otimes X} \circ (\operatorname{id}_{(Y \otimes X)^*} \otimes (g \otimes f)) \circ \operatorname{\widetilde{coev}}_{Y \otimes X} = (\operatorname{ev}_X) \circ (\operatorname{id}_{X^*} \otimes \operatorname{ev}_Y \otimes \operatorname{id}_{X^*}) \circ (\operatorname{id} \otimes \operatorname{id} \otimes g \otimes f) \circ \dots$$ and this is just the beginning of the first equality!

  • $\begingroup$ Absolutely wonderful answer! And yes, you're right, I'm in fact working with spherical categories, which is why I added to assume equality of left and right traces. Thank you very much! $\endgroup$ – Nephry Sep 30 '15 at 19:22

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.