Let $G$ be a group, and $\mathbf{G\text{-}Sets}$ the category whose objects are left G-Sets and whose morphisms are G-Set homomorphisms, that is functions $f:X\to Y$ such that $f(ax) = af(x)$, $a\in G, x\in X$.

Does the forgetful functor $\mathbf{G\text{-}Sets}$ $\to$ $\mathbf{Sets}$ have a left-adjoint?

Are there any other interesting functors between the categories $\mathbf{G\text{-}Sets}$ and $\mathbf{Sets}$?

up vote 6 down vote accepted

Let $S$ be a set. Then $G \times S$ can be equipped with a $G$-action by having $G$ act by left-multiplication on itself and trivially on $S$. This is a left-adjoint to the forgetful functor -- for $T$ an arbitrary $G$-set, one has

$$ \text{Hom}_\text{G-set}(G \times S, T) = \text{Hom}_\text{set}(S, T) $$ by the rule $f \mapsto \left[s \mapsto f(1, s)\right]$. The map the other way is $\phi \mapsto \left[(g, s) \mapsto g\phi(s)\right]$ .

  • Why did you write $\text{Hom}_\text{G-set}(G \times S, T)$? Shouldn't it be $\text{Hom}_\text{G-set}(S, T)$ where $S$ is a set equipped with a certain map $G\times S \to S$? Those are the objects in G-Set -Thanks – iwriteonbananas Apr 21 '15 at 17:18
  • my functor eats $S$ and returns $G \times S$ with the action described in the answer. That's what it means to be a left adjoint. In other words, if I call the functor $F$, the statement is $Hom_{G-set}(F(S), T) = Hom_{set}(S, Forget(T))$. – hunter Apr 21 '15 at 21:52

This is an instance of Kan extensions.

Let $\mathbf G$ be the one object category associated to the group $G$, and $\mathbf 1$ the one associated to the trivial group. Then $\mathbf{G{-}Sets}$ is nothing else than $[\mathbf G,\mathbf{Sets}]$ and the forgetful functor is the functor $$ i^\ast \colon [\mathbf G,\mathbf{Sets}] \to [\mathbf 1,\mathbf{Sets}] \simeq \mathbf{Sets},\quad F \mapsto F \circ i $$ induced by the (unique) functor $i\colon \mathbf 1 \to \mathbf G$. As $\mathbf{Sets}$ is cocomplete, the functor $i^\ast$ admits a left adjoint, usually denoted $i_!$ (see nlab:Kan extension). Note that, as $\mathbf{Sets}$ is also complete, $i^\ast$ also admits a right adjoint, usually denoted $i_\ast$.

Remark that by computing the general formula for left Kan extension in this case, you get back hunter's answer: $$ i_! \colon S \mapsto \coprod_{g\in G}S \simeq G \times S. $$

Here's another way of seeing why this functor has a left-adjoint.

Theorem. Let $\mathsf{S}$ and $\mathsf{T}$ denote Lawvere theories. Then given a morphism $\varphi : \mathsf{S} \leftarrow \mathsf{T}$, the corresponding forgetful functor $$\mathrm{Mod}(\varphi) : \mathrm{Mod}(\mathsf{T}) \leftarrow \mathrm{Mod}(\mathsf{S})$$ has a left-adjoint.


  • The initial Lawvere theory is denoted $\mathsf{Set}$, and its models are (basically) sets.
  • Given a group $G,$ we can build a Lawvere theory $\mathrm{Lath}(G)$ such that the models of $\mathrm{Lath}(G)$ are (basically) precisely the $G$-sets.
  • Write $\varphi$ for the unique morphism $\varphi : \mathrm{Lath}(G) \leftarrow \mathsf{Set}.$

So use the above theorem to conclude that

$$\mathrm{Mod}(\varphi) : \mathrm{Mod}(\mathsf{Set}) \leftarrow \mathrm{Mod}(\mathrm{Lath}(G))$$

has a left-adjoint. In other words, the usual forgetful functor

$$\mathbf{Set} \leftarrow \mathbf{Set}^G$$

has a left-adjoint, where I write $\mathbf{Set}^G$ for the category of $G$-sets, since a $G$-set is the same thing as a functor $\mathbf{Set} \leftarrow G,$ where $G$ is viewed as a category with one object.

Although this uses a lot more technical machinery than is strictly required, I like this viewpoint because it provides a quick heuristic that can help you immediately conclude that a variety of different functors have left-adjoints.

If $H\le G$ is a subgroup let $\mathrm{Res}_H^G:G\textrm{-}\mathsf{Set}\to H\textrm{-}\mathsf{Set}$ be the restriction functor. One can think of $\mathsf{Set}$ as the case where $H=1$ is the trivial subgroup. It has a left adjoint, call it $\mathrm{Ind}_H^G$, so that

$$\hom_{G\textrm{-}\mathsf{Set}}(\mathrm{Ind}_H^G-,-) ~~\cong~ \hom_{H\textrm{-}\mathsf{Set}}(-,\mathrm{Res}_H^G-)$$

are naturally isomorphic functors $H\textrm{-}\mathsf{Set}\times G\textrm{-}\mathsf{Set}\to\mathsf{Set}$.

An explicit construction is $\mathrm{Ind}_H^G\Omega\cong G\times_H \Omega$, which is $G\times\Omega$ (the action of $G$ purely on the first coordinate) modulo the relation $(gh,\omega)\sim(g,h\omega)$ for all $g\in G,h\in H,\omega\in\Omega$.

Does this look familiar? If you've studied representation theory, it should. The linearized version of this fact is called Frobeius reciprocity. Let $G\textrm{-}\mathsf{Rep}$ be the category of linear representations of $G$, and then define the restriction functor $\mathrm{Res}_H^G:G\textrm{-}\mathsf{Rep}\to H\textrm{-}\mathsf{Rep}$. Frobenius reciprocity (the categorical version of it, as opposed to the more standard, decategorified, character-theoretic version of it) states that restriction has a left adjoint called induction $\mathrm{Ind}_H^G$, so $\hom_G(\mathrm{Ind}_H^GV,W)\cong\hom_H(V,\mathrm{Res}_H^GW)$ canonically for all representations $V$ of $G$ and $W$ of $H$.

An explicit construction is $\mathrm{Ind}_H^GV\cong \Bbb C[G]\otimes_{\Bbb C[H]}V$, the linearized analogue of $G\times_H\Omega$!

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.