I am trying to prove the following fact

Let functors $K,K':\mathcal{D}^{\text{op}}\to\mathbf{Set}$ have representations $(r,\phi)$ and $(r',\phi')$ respectively. Prove that to each natural transformation $\tau:K\to K'$, there is a unique morphism $h:r\to r'$ of $\mathcal{D}$, such that $$\tau\circ\phi=\phi'\circ\text{Hom}_{\mathcal{D}}(-,h):\text{Hom}_{\mathcal{D}}(-,r)\to \text{Hom}_{\mathcal{D}}(-,r')$$

I think I am in the correct path but I feel I can not phrase exactly the required statement from what I prove:

The composite $\tau\circ \phi$ is a natural transformation $\text{Hom}_{\mathcal{D}}(-,r)\to K'$ in $[\mathcal{D}^{\text{op}},\mathbf{Set}]$. By Yoneda, we have $$\text{Hom}_{[\mathcal{D}^{\text{op}},\mathbf{Set}]}(\text{Hom}_{\mathcal{D}}(-,r),K')\cong K'(r)\cong \text{Hom}_{\mathcal{D}}(r,r')$$ where the last isomorphism holds since $\phi$ is a representation. Therefore, every natural transformation $\text{Hom}_{\mathcal{D}}(-,r)\to K'$ (hence also the composite $\tau\circ \phi$), corresponds to an arrow $r\to r'$ in $\mathcal{D}$.

This looks something like what needs to be proved, but I am not sure how the statement follows, if of course what I am doing is correct. Any help would be appreciated.


The idea it's correct but personally I find more easily to follow the argument when stated in the following way.

More in details it should work like this: by Yoneda lemma we have that every natural transformation of kind $$\tau' \colon \hom_{\mathcal D}(-,r) \Rightarrow \hom_{\mathcal D}(-,r')$$ is of the form $\hom_{\mathcal D}(-,h)$ for some $h \in \hom_{\mathcal D}(r,r')$; now by the hypothesis we have the following natural transformation $$\hom_{\mathcal D}(-,r) \stackrel{\phi}\cong K \stackrel{\tau}{\Rightarrow} K' \stackrel{\phi'^{-1}}{\cong} \hom_{\mathcal D}(-,r')$$ by what we've said above $\phi'^{-1} \circ \tau \circ \phi = \hom_{\mathcal D}(-,h)$ for a certain $h \in \hom_{\mathcal D}(r,r')$.

Now composing on the left both sides of the equation you get that $$\tau \circ \phi = \phi' \circ \hom_{\mathcal D}(-,h)$$ which is exactly what you wanted to prove.

Hope this helps.

Edit: The fact that every natural transformation $$\hom_{\mathcal D}(-,r) \Rightarrow \hom_{\mathcal D}(-,r')$$ is of the form $\hom_{\mathcal D}(-,h)$ for some $h \in \mathcal D(r,r')$ follows in the following way.

By yoneda lemma we have a natural bijection $$[\mathcal D^\text{op},\mathbf{Set}](\hom_{\mathcal D}(-,r),\hom_{\mathcal D}(-,r')) \cong \hom_{\mathcal D}(r,r')$$ the inverse of this bijection is given by the mapping sending every $h \in \hom_{\mathcal D}(r,r')$ in the family of funtions $$\langle g \in \hom_{\mathcal D}(z,r) \mapsto \hom_{\mathcal D}(g,r')(h)=h \circ g \in \hom_{\mathcal D}(z,r')\rangle_{z \in\mathcal D}\ .$$

Since for every $g \in \hom_{\mathcal D}(z,r)$ we have that $$h \circ g = \hom_{\mathcal D}(z,h)(g)$$ we have that mapping above is exactly $\hom_{\mathcal D}(-,h)$.

Summarizing we have that the inverse of the yoneda bijection sends every $h \in \hom_{\mathcal D}(r,r')$ in the natural trasformation $\hom_{\mathcal D}(-,h)$. Since this mapping is a bijection every natural transformation in $[\mathcal D^\text{op},\mathbf{Set}](\hom_{\mathcal D}(-,r),\hom_{\mathcal D}(-,r'))$ is of the form $\hom_{\mathcal D}(-,h)$ for some $h \in \mathcal D(r,r')$.

  • $\begingroup$ Ok, thanks! But I am afraid I don't see exactly how Yoneda proves the statement you mention in your second paragraph: I can see that $$\text{Hom}_{[\mathcal{D}^{op},\mathbf{Set}]}(\text{Hom}_{\mathcal{D}}(-,r), \text{Hom}_{\mathcal{D}}(-,r'))\cong \text{Hom}_{\mathcal{D}}(r,r')$$ by Yoneda, but why is every natural transformation coming from the left-hand side, of the form $\text{Hom}_{\mathcal{D}}(-,h)$? $\endgroup$ – Niels.Remb05 Apr 11 '14 at 8:02
  • $\begingroup$ It follows by the definition of the yoneda isomorphism. In some minutes I'm gonna add that to the answer. :) $\endgroup$ – Giorgio Mossa Apr 11 '14 at 8:16
  • $\begingroup$ @Niels.Remb05 take a look to the edit and see if that solves all your doubts. $\endgroup$ – Giorgio Mossa Apr 11 '14 at 8:33
  • $\begingroup$ Yes, of course. How stupid of me. Thank you very much! $\endgroup$ – Niels.Remb05 Apr 11 '14 at 8:35

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.