In Formal category theory: adjointness for 2-categories, Gray defines a tensor product of 2-categories, now more commonly known as the lax Gray tensor product, which I will denote by $\otimes_l$. For 2-categories $\mathcal A, \mathcal B, \mathcal C$, there are natural isomorphisms between

  1. 2-functors $\mathcal A \otimes_l \mathcal B \to \mathcal C$
  2. 2-functors $\mathcal A \to [\mathcal B, \mathcal C]_{\mathrm{lax}}$ into the 2-category of 2-functors and lax natural transformations
  3. 2-functors $\mathcal B \to [\mathcal A, \mathcal C]_{\mathrm{oplax}}$ into the 2-category of 2-functors and oplax natural transformations

Gray proves that this tensor product, together with the terminal 2-category and the above correspondences, define a nonsymmetric monoidal category that is closed on both sides (Theorem 1,4.14 ibid.).

2-categories, 2-functors, 2-natural transformations, and modifications form a 3-category $2\text{-}\mathbf{Cat}$, and so we might hope that the lax Gray tensor product moreover defines a closed nonsymmetric 3-dimensional monoidal structure on $2\text{-}\mathbf{Cat}$. Such a statement is not as intimidating as it might at first seem, since everything is strict: since the lax Gray tensor product does define a one-dimensional monoidal structure, to show this monoidal structure is indeed 3-dimensional, it suffices to show that the lax Gray tensor product is 3-functorial and that the associator and unitor for this monoidal structure are 3-natural.

However, most references on the Gray tensor product appear only to treat it 1-categorically. From a cursor search, I was not able to find a reference for even the 2-functoriality of the Gray tensor product.

Does there exist a reference for the 3-functoriality of the lax Gray tensor product, and for the 3-naturality of the associator and unitor for the corresponding monoidal structure, or may they be derived straightforwardly from the existing literature? Alternatively, is there an obstruction to 3-functoriality or 3-naturality?

I would be interested in knowing about partial references – e.g. for the 2-functoriality of the lax Gray tensor product, and 2-naturality of the associator and unitors – if no references for the 3-functoriality may be found. I would also be interested in the same questions for the pseudo Gray tensor product, though I am primarily interested in the lax notion.


1 Answer 1


The lax Gray tensor product is not a two-variable 2-functor.

If it were, we'd have a functor $Fun(A,B) \times Fun(C,D) \to Fun(A \otimes_l C, B \otimes_l D)$ extending the set map $Ob Fun(A, B) \times Ob Fun(C, D) \to Ob Fun(A \otimes_l C, B \otimes_l D)$. Consider the case when $A = C = [0]$ is a point and $B = D = [1]$ is an arrow. Then we'd have a functor $[1] \times [1] \to [1] \otimes_l [1]$ extending the set map which is the identity on objects. That is, we'd have a commuting square living in the walking lax commutative square in an identity-on-objects way. There is no such functor.

There is a map in the other direction, reflecting the fact that the identity functor is a lax monoidal functor from the lax Gray tensor to the cartesian product. But the identity is not a lax monoidal functor from the cartesian product to the lax Gray tensor.

Morally, $n$-functors "don't shift category number". The lax Gray tensor product is too busy shifting category number (e.g. $m$-category $\otimes_l$ $n$-category = $(m+n)$-category) to be $n$-functorial.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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