Let $\mathcal{D} \underset{G}{\overset{F}{\leftrightarrows}}\mathcal{C}$ with $F\dashv G$ and $G$ fully faithful. Write $T=GF$. $\epsilon$ denotes the counit of the adjunction.
The comparison functor $K:\mathcal{D}\rightarrow TAlg$ for this adjunction is given by $K(D)=(GD,G\epsilon_D)$ and $K(f:A\rightarrow B) = G(f):G(A)\rightarrow G(B)$.
I want to prove that $K$ is an equivalence.

My attempt:
If I prove that $K$ is fully faithful and essentially surjective on objects, I'm done.
It was easy to see that $K$ inherits the fullness property of $G$. Similarly, using faithfulness of $G$ I was able to prove straight away that $K$ is faithful.
To prove that $K$ is essentially surjective: Let $(X,h)$ be a T-algebra. We want to show that there is $D\in\mathcal{D}$ such that $K(D)\overset{\phi}{\cong} (X,h)$.
Well, we can write $h:T(X)\rightarrow X$, so setting $D=T(X)$ and $\phi=h$ seems like a good bet. But I don't see why $h$ should be an isomorphism. I was thinking on trying to find $h^{-1}$ using the fullness of $G$, but failed.

Any hints?

  • 1
    $\begingroup$ If $(X,h)$ is a $T$-algebra, then $h\eta_X=1$. So to prove that $h$ is an iso it is enough prove that $\eta_Xh=1$ as well. $\endgroup$ – Arnaud D. Apr 5 '18 at 15:43
  • $\begingroup$ @ArnaudD. Ok, great! Then I did the following: $\eta_X h = \mu_X T\eta_X=(..easy..)=id_{TX}$ where in the first equality I use that $\eta_X$ is a morphism of T-algebras. But I am not sure why this is true.. $\endgroup$ – Soap Apr 5 '18 at 17:39
  • 1
    $\begingroup$ Well, in general it's not true... Use instead the fact that if $\mu$ is an isomorphism, then $T(\eta)=\eta_T$ to prove that $h$ must be an iso in $\mathcal{C}$. $\endgroup$ – Arnaud D. Apr 5 '18 at 18:30
  • $\begingroup$ @ArnaudD. Yes, I was able to show that $\mu $ is an isomorphism and that this implies $T\eta=\eta_T$. Now I've been trying to use that to show that $\eta_X h =1$ but I don't see how $\endgroup$ – Soap Apr 6 '18 at 15:39
  • 1
    $\begingroup$ Ok, new hint : use the naturality of $\eta$. $\endgroup$ – Arnaud D. Apr 6 '18 at 16:04

If the right adjoint $G$ is fully faithful, then the counit $\epsilon$ is an isomorphism, and as a consequence, so is the multiplication $\mu=G(\epsilon_F)$, which means that the monad $T$ is idempotent. Since $\mu T(\eta)=1=\mu\eta_T$, we then have $T(\eta)=\eta_T=\mu^{-1}$.

Now an algebra for an idempotent monad is the same thing as an object $X$ such that $\eta_X$ is an isomorphism. Indeed, if $(X,h:TX\to X)$ is an algebra, then $h\eta_X=1$, and moreover $$\eta_Xh=T(h)\eta_{TX}=T(h)T(\eta_X)=T(h\eta_X)=1$$ since $\eta:Id\Rightarrow T$ is a natural transformation. Conversely, if $\eta_X$ is an isomorphism, then $h=\eta_X^{-1}:TX\to X$ makes $X$ an algebra, since $h\eta_X=1$ holds trivially and $$h T(h)=hT(\eta_X^{-1})=hT(\eta_X)^{-1}=h\mu.$$ Moreover, any map $f:X\to Y$ between two objects such that $\eta_X$ and $\eta_Y$ are isomorphisms is an algebra map, since we always have $\eta_Yf=T(f)\eta_X$ and thus also $f\eta_X^{-1}=\eta_Y^{-1}T(f)$.

So the category of $T$-algebra is equivalent (and even isomorphic) to the full subcategory $\mathcal{C}'$ of objects $X$ such that $\eta_X$ is an isomorphism. Since $\eta_{FY}=F(\epsilon_Y)^{-1}$ for all object $Y$ of $\mathcal{D}$ (by the triangle identity), the original adjunction restricts to an adjunction $\mathcal{D} \underset{G}{\overset{F}{\leftrightarrows}}\mathcal{C}'$, where now all components of the unit and counit are isomorphisms. Thus $G$ restricts to an equivalence between $\mathcal{D}$ and $\mathcal{C}'$, which is in fact the comparison functor $\mathcal{D}\to \mathcal{C}^T$.


There's an almost-one liner for that: the counit of $F\dashv G$ is an isomorphism, since $G$ is fully faithful, so $FG\cong 1$, and $K$ is such that $U^TK=G$, $KF=F^T$ if $F^T\dashv U^T$ denotes the free-forgetful adjunction generated by $T$.

If you put these things, together, it's pretty easy to show that $FU^T$ is an inverse for $K$:

  1. $FU^TK = FG\cong 1$, because $G$ is fully faithful.
  2. $KFU^T=F^T U^T \cong 1$, because the monad $T$ is idempotent ($T$ is idempotent iff $U^T$ is fully faithful, if and only if the counit $\epsilon^T : F^TU^T\Rightarrow 1$ is invertible, if and only if the multiplication of $T$ is an isomorphism).

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.