I'm reading Remark 5.1.20(a) from Leinster (p. 119). He claims the following (in my words):

Fix a small category $\mathbf I$, a category $\mathscr A$, and a functor $D:\mathbf I\to\mathscr A$. Let $(p_I:L\to D(I))_{I\in\mathbf I}$ be a limit cone of $D$. Then for any $A\in\mathscr A$, there is a bijection $$\{\text{arrows } A\to L \text{ in }\mathscr A\}\leftrightarrow\{\text{cones on } D \text{ with vertex } A\}$$ given by assigning to $g:A\to L$ the cone $(p_I\circ g:A\to D(I))_{I\in\mathbf I}$. Leinster claims that the bijectivity of this assignment follows from the definition of limit. I'm trying to verify this in detail.

Surjectivity. I believe the statement is based on the assumption that a limit cone of $D$ exists. In such a case, any cone on $D$ with vertex $A$ gives rise to an arrow $A\to L$ (such that some properties hold, but this is not relevant here). So the assignment is surjective. Am I right here? Added after I've written the below passage on injectivity: actually this is not clear as well. We get an arrow from $A$ to a some limit $L'$ which need not a priori be equal to $L$...

Injectivity. This is where I don't understand what's going on. First of all, Leinster says that for now we do not know that a limit is unique and that it will follow from the claim being proved. So at this point we cannot use the uniqueness of a limit, right? To prove injectivity, we must prove that if $(f_I:A\to D(I))$ and $(h_I:A\to D(I))$ are two cones on $D$ with vertex $A$, then the arrows $\overline f$ and $\overline h$ from the definition of limit (5.1.19(b)) are equal. But a priori they are not even arrows to the same $L$ if we don't know that $L$ is unique. And the arrows $p_I$ from the definition of limit cone of $(f_I)$ need not be equal to the arrows $p_I'$ from the definition of limit cone of $(h_I)$. And even if we know that $L$ and $p_I$ are unique (which we don't), I still wasn't able to prove $\overline h=\overline f$. How to do that?

  • $\begingroup$ You only need the uniqueness of a limiting arrow $A → L$ of cone vertices, making a given cone $C$ with vertex $A$ commute with the limit cone. $\endgroup$ – k.stm Jul 22 '19 at 7:35

You have to be careful : "L is a limiting cone over $D$" is a statement about $L$ ! You have supposed that $L$ is a limiting cone and so you have to use the properties of $L$ that follow from the definition of a limiting cone. You don't need to care about the possibility of other limiting cones for the moment.

We have that $\{d_i : L \to D(i)\}$ is a limiting cone over $D$ if for any cone $\{h_i : C \to D(i)\}$ there is a unique morphism $h : C \to L$ such that $ d_i \circ h = h_i$. So here in the definition we have existence and unicity which should mean that there is a hidden bijection. What is the map realising the bijection ? Well it is the map that you described but I will write it in this way : $$ Hom_{\mathcal A}(C,L) \to Nat_{Fun(I,A)}(\delta(C),D) $$ $$ h \mapsto \{d_i \circ h, i \in I\} $$ Giving a cone with vertex $C$ is the same thing as giving a natural transformation between the constant functor $\delta(C) : i\mapsto C$ (which sends maps $i \to j$ in $I$ to the identity of $C$) and the diagram $D$.

The existence part of the definition gives surjectivity : given a cone $\{h_i : C \to D(i)\}$ there is a preimage $h: C \to L$ of the map just described, this is surjectivity. The unicity part is the injectivity of this map : Given two maps $h : C \to L$ and $f : C\to L$ such that $d_i \circ f = d_i \circ g$ for any $i \in I$, i.e. given two elements of $Hom_{\mathcal A}(C,L)$ which induce the same cone over $D$, then $h = g$ by unicity.

Then the unicity of $L$ follows : given $\{ d_i' : L'\to D(i)\}$ another limiting cone by definition there are two maps $f :L \to L'$ and $h : L' \to L$ such that $d_i \circ h = d_i'$ and $d_i' \circ f = d_i$, but then, $h\circ f : L \to L$ in $Hom_{\mathcal A}(L,L)$ will be sent to $\{d_i \circ h \circ f = d_i, i \in I\}$ meaning $h\circ f = id_L$ by the injectivity of the map. Similarly $f\circ h = id_{L'}$. Meaning that $L$ and $L'$ are isomorphic in $\mathcal A$, and they are actually isomorphic in a unique way as cones over $D$ meaning $f$ and $h$ are the unique maps that make the diagrams comute.

| cite | improve this answer | |
  • $\begingroup$ In the last paragraph (when you are saying "by the bijectivity of the map"), are you using surjectivity at all? I can only see that injectivity is used. $\endgroup$ – user634426 Jul 22 '19 at 21:14
  • $\begingroup$ I’m using surjectivity in defining the maps $f$ and $h$, then I’m using injectivity to compare the compositions to the identity which have the same image under the map. $\endgroup$ – jeanmfischer Jul 22 '19 at 21:21
  • $\begingroup$ I changed the word I used to be more precise. $\endgroup$ – jeanmfischer Jul 22 '19 at 21:25

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.