This means that the definition of lax functor involves a choice of orientation of these structural cells which is not visible for pseudofunctors. The choice is such that the first example below comes out as stated. With the opposite choice one speaks of an oplax functor.
Often the term lax functor is used for -functors whose domain is an ordinary category (regarded as an -category with only trivial higher morphisms), while the codomain is often taken to be a 2-category.
The compositor of the lax functor is the monad product, the unitor is the monad unit.
Similarly, oplax functors are equivalent to comonads in .
In particular, if , then this example reduces to the first one.
Another special case arises when for some monoidal category . Then lax functors are the same as categories enriched in the monoidal category .
It makes sense to ask that a functor is lax and oplax in a compatible way such that yields Frobenius monads.
Some old remarks on this case are in Note on lax functors and bimodules.
This relation between lax-oplax functors and conformal field theory was developed in detail in
A general discussion of lax-oplax functors is in section 2.1 there.
Isn’t it odd not to require any extra condition at all on the coherence morphisms? I would have expected a definition where they are required to be split epi, or require that the codomain be a subobject of the domain. Is there a name for something like that?
Mike Shulman: One could certainly add that as a condition, but I don’t think I’ve ever heard of anyone having a use for it, or giving it a name. The interesting examples listed above (and others) don’t use any such condition.