14
$\begingroup$

Let $f: \mathcal{C} \rightarrow \mathcal{D}$ be a colimit-preserving functor between locally presentable categories. Assume that $f$ induces an equivalence between the groupoids underlying $\mathcal{C}$ and $\mathcal{D}$. Is $f$ necessarily an equivalence? What if $\mathcal{C}$ and $\mathcal{D}$ are assumed to be linear/dg/stable?

$\endgroup$
8
  • $\begingroup$ From the point of view of 1-category theory, this is equivalent to the following: say that $f: A \to B$ is a $\kappa$-colimit preserving functor between small categories with $\kappa$-small colimits. Is it true that if it induced an equivalence at the level of the underling 1-groupoid, then it was an equivalence itself? $\endgroup$ Commented Sep 17, 2019 at 16:05
  • $\begingroup$ What is the "underlying" groupoid? All isomorphisms? $\endgroup$
    – fosco
    Commented Sep 17, 2019 at 20:21
  • $\begingroup$ Yes. That's what I mean. $\endgroup$ Commented Sep 17, 2019 at 21:22
  • 1
    $\begingroup$ I edited the question a bit since the infinities were probably orthogonal to the problem. I think your reformulation may be ultimately what I am interested in, but now it is not so clear to me how it is equivalent to the original thing. It seems that one needs some compatibility between "being an equivalence at the level of groupoids" and "ind-competion/passing to $\kappa$ compact objects"... $\endgroup$ Commented Sep 18, 2019 at 5:18
  • 1
    $\begingroup$ @PeterLeFanuLumsdaine: right now I hate automorphisms. $\endgroup$ Commented Sep 18, 2019 at 8:33

1 Answer 1

21
$\begingroup$

The answer in general is no.

Let $\mathcal C$ be the category of sets, let $\mathcal D$ be the category of pointed sets (with basepoint-preserving maps), and let $f: \mathcal C \to \mathcal D$ be the functor which adds a disjoint basepoint. Then $f$ is an equivalence on underlying groupoids, but not an equivalence of categories. Moreover, the forgetful functor is a right adjoint to $f$.

Of course, since any locally presentable 1-category is also presentable as an $\infty$-category, this answers the question in the $\infty$-categorical case, too. But for a more "intrinsically $\infty$-categorical" example, let $\mathcal C$ be the $\infty$-category of spaces, let $\mathcal D$ be the $\infty$-category of pointed spaces with disjoint basepoint (i.e. pointed spaces such that the connected component of the basepoint is contractible; morphisms are basepoint-preserving maps), and let $f: \mathcal C \to \mathcal D$ be the functor which adds a disjoint basepoint. Again, the forgetful functor is a right adjoint to $f$, and $f$ is an equivalence on object spaces. It's perhaps surprising that $\mathcal D$ is presentable, but it is: colimits are formed by taking the colimit in the category of pointed spaces and then collasping the connected component of the basepoint to be contractible; from this presentability is easy to verify.


The answer in the stable case is yes.

Assume that $\mathcal C$ and $\mathcal D$ are stable (or even just additive with suspensions), and that $f$ is exact (or even just preserves finite direct sums and suspensions).

First, $f$ is faithful on the homotopy category. For if $f(\phi) = 0$, then $f\begin{pmatrix} 1 & \phi \\ 0 & 1 \end{pmatrix} = f\begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}$. Since $f$ is faithful on isomorphisms, we get $\begin{pmatrix} 1 & \phi \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}$, so that $\phi = 0$.

Now, $f$ is full on the homotopy category. For given a map $\phi$, because $f$ is full on isomorphisms, we have $f\begin{pmatrix} a & b \\ c & d \end{pmatrix} = \begin{pmatrix} 1 & \phi \\ 0 & 1 \end{pmatrix}$ for some $\begin{pmatrix} a & b \\ c & d \end{pmatrix}$, so that $f(b) = \phi$.

Thus $f$ is an equivalence on the homotopy category (since it is essentially surjective too). In this context, this implies that $f$ is an equivalence, as $\pi_n(Hom(X,Y)) = \pi_0(Hom(\Sigma^n X, Y))$.

This argument (without the bit about suspensions) also works for any additive ordinary category (i.e. any additive category with discrete hom-spaces).


Here's a special case where the answer is yes.

Proposition: Let $f: L \to M$ be a sup-preserving map between complete lattices. Suppose that $f$ is a bijection. Then $f$ is an isomorphism.

Proof: We want to show that for $x,y \in L$, we have $x \leq y \Leftrightarrow f(x) \leq f(y)$. The forward implication follows from $f$ being sup-preserving (and hence order-preserving). For the reverse implication, suppose that $f(x) \leq f(y)$. We have $f(\sup(x,y)) = \sup(f(x),f(y)) = f(y)$. Because $f$ is a bijection, we have $\sup(x,y) = y$, i.e. $x \leq y$.


Two observations for the general case: Let $f_!: \mathcal C \to \mathcal D$ be a left adjoint between locally presentable categories which induces an equivalence $\iota f_!: \iota \mathcal C \to \iota \mathcal D$ of spaces of objects, as in the question statement.

Claim: $f_!$ is conservative.

Proof: We may factor $f_!$ as a localization $l_!: \mathcal C \to L\mathcal C$ followed by a conservative left adjoint $g_!: L\mathcal C \to \mathcal D$. Because $l_!$ is a localization, its right adjoint $l^\ast$ is fully faithful. But this means that on object spaces, $\iota f_!$ factors through a retract $\iota L \mathcal C$ of $\iota \mathcal C$. Since $\iota f_!$ is an equivalence, $\iota l_!$ must be an equivalence, and $\iota l^\ast$ must also be an equivalence. Because $l^\ast$ is fully faithful, this implies that $l^\ast$ must be an equivalence, so that $l_!$ is also an equivalence. Thus $f_!$ may be identified with the conservative functor $g_!$.

Corollary: $f_!$ is faithful.

Proof: This is a general fact: a conservative left adjoint between cocomplete categories is faithful, for if $f_!(\phi) = f_!(\psi)$ for some $\phi,\psi: C^\to_\to C'$, then the map $f_! C' \to coeq(f_!\phi,f_!\psi)$ is an isomorphism. Since $f_!$ is conservative and preserves colimits, the map $C' \to coeq(\phi,\psi)$ is an isomorphism, i.e. $\phi = \psi$.

For the faithfulness statement, I'm not 100% sure how this translates into the $\infty$-categorical world.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .