1
$\begingroup$

It's well known that endofunctors on $Set$ have an unique strength. A strength for a functor $T : Set \to Set$ is a natural transformation $t_{A,B} : A \times T B \to T (A \times B)$ such that certain diagrams commute.

I want an expression for the strength morphism. It seems that I can come up with a morphism with the correct source and target. Fixing $A$ and $B$:

$$A \times TB \to (A \times TB) \times Hom(A \times B,A \times B) \to \int^{C,D} (C\times TD) \times Hom(C\times D,A\times B) \to T(A \times B)$$

The middle morphism is the injection, and the last one is given by an isomorphism that can be proven using the coYoneda lemma. Does the family of these morphisms form a strength for $T$? That $T$ is small in some sense can be assumed.

The isomorphism between $\int^{C,D} (C\times TD) \times Hom(C\times D,A\times B)$ and $T (A \times B)$ is related to the left unit of Day convolution, and it is independent on the choice of $T$.

I wonder, in general, if there's a relation between Day convolution unit axiom and functors' strengths.

$\endgroup$

1 Answer 1

5
$\begingroup$

The OP's description of the strength is correct, but there's a simpler one: just think of it as corresponding to the composite

$$A \stackrel{coev_A}{\to} \hom(B, A \times B) \to \hom(T(B), T(A \times B))$$

under the adjunction $(- \times T(B)) \dashv \hom(T(B), -)$. Here $coev$ denotes the unit of the adjunction $(- \times B) \dashv \hom(B, -)$, and the second arrow comes from $T$'s being a ($Set$-enriched) functor.

Since we are dealing with functors on $Set$, we may as well make this very concrete and describe this in terms of sets and elements (we remember sets and elements, don't we? (-: ). One can work out as an exercise that the strength on $T$ according to the description above is

$$\langle a \in A, y \in T(B)\rangle \mapsto T(b \in B \mapsto \langle a, b\rangle)(y)$$

where on the right we are applying the functor $T$ to the indicated function of type $B \to A \times B$ induced by an element $a \in A$.

(The categorical definition above generalizes so that given a closed monoidal category $V$, the structure of a $V$-monoidal strength on a functor $T: V \to V$ is equivalent to the structure of a $V$-enrichment on $T$. This doesn't require dealing with any colimits involved in coends; it just uses $V$ being closed monoidal.)


To connect this description with the OP's description, it help to be describe concretely what the co-Yoneda isomorphism is doing, again in terms of sets and elements. The co-Yoneda isomorphism

$$\int^C \hom(C, A) \times T(C) \stackrel{\sim}{\to} T(A)$$

takes a typical element given by an equivalence class $[f: C \to A, x \in T(C)]$ to $T(f)(x) \in T(A)$.

There are actually two co-Yoneda isomorphisms implicit in the OP's description:

$$\int^{C, D} \hom(1, C) \times T(D) \times \hom(C \times D, A \times B) \cong \int^D T(D) \times \hom(1 \times D, A \times B)$$

$$\int^D T(D) \times \hom(D, A \times B) \cong T(A \times B)$$

Putting these together, the composite co-Yoneda isomorphism can be described as

$$[c \in C, y \in T(D), g: C \times D \to A \times B] \qquad \mapsto \qquad T(g(c, -))(y)$$

Now setting $A = C, B = D$, and $g = id_{A \times B}$, we easily calculate the OP's composite as the function

$$\langle a \in A, y \in T(B) \rangle \mapsto T(id_{A \times B}(a, -))(y)$$

where $id_{A \times B}(a, -)$ is just another way of writing $b \mapsto \langle a, b\rangle$. Hence the two descriptions are equivalent.

$\endgroup$
2
  • $\begingroup$ Any tip on how I could prove that the via-coend expression is actually the strength, i.e. that the diagrams commute? I'm more interested in this path because of its relation with Day convolution. $\endgroup$
    – user46784
    Feb 10, 2014 at 18:37
  • $\begingroup$ @user46784 Here you go... $\endgroup$
    – Todd Trimble
    Feb 10, 2014 at 23:02

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.