Algebraic $K$-theory of a symmetric monoidal category $C$ is defined in two steps: first take geometric realization, then group complete: $K(C) = \Omega B |C|$.

In HAK II, Grayson (following Quillen) lifts the group completion step to the categorical level, defining a symmetric monoidal category $C^{-1} C$ such that $K(C) = |C^{-1} C|$. So $C^{-1} C$ is a "pre group completion" in the sense that it realizes to group completion. But what is the significance of $C^{-1}C$ qua symmetric monoidal category, before geometric realization?

It's not the group completion at the categorical level. Although every object $c \in C$ has an "inverse" $c^{-1}$, and although there is an obvious map $I \to c \otimes c^{-1}$ (where $I$ is the monoidal unit), this map is not an isomorphism. In fact, there need not be a map in the other direction $c\otimes c^{-1} \to I$. Only upon geometric realization does the map $I \to c \otimes c^{-1}$ become invertible. So there remains the

Question: What is the universal property of $C^{-1} C$ as a symmetric monoidal category? EDIT What is the relationship between the construction $C \mapsto C^{-1} C$ and group completion at the level of categories?

Note that the nonexistence of a morphism $c \otimes c^{-1} \to I$ rules out something else: the free symmetric monoidal category with duals for objects on $C$ would also realize to $K(C)$, but that category would have both unit and counit maps, whereas $C^{-1}C$ has only unit maps.

It's probably worth recalling the construction. An object of $C^{-1} C$ (by which I really mean $Iso(C)^{-1} C$ in Grayson/Quillen's notation) is a pair of objects of $C$, which we may write $c_1^{-1} \otimes c_2$. A morphism $c_1^{-1} \otimes c_2 \to d_1^{-1} \otimes d_2$ is an equivalence class of triples $(c,f_1,f_2)$ where $c \in C$, $f_1: c_1 \otimes c \to d_1$ is an isomorphism in $C$, and $f_2: c_2 \otimes c \to d_2$ is a morphism in $C$; these are modded out by the diagonal action of $Aut_C(c)$. Defining composition can be done by following your nose.

Grayson/Quillen's construction actually applies more generally to any symmetric monoidal category $S$ acting on a category $C$, producing a category $S^{-1}C$, but I'd be happy with understanding the case $S = Iso(C)$, and perhaps even with the restriction that $C$ be a groupoid.

  • 7
    $\begingroup$ Thomason's paper "Beware the phony multiplication on Quillen's $\mathcal{A}^{-1}\mathcal{A}$" explains some things that look like they work, but are subtly wrong. $\endgroup$ Commented Oct 20, 2017 at 17:51
  • 1
    $\begingroup$ If I understand correctly, the universal property is that of the quotient of $C\times C$ by the diagonal action of $S=Iso(C)$; generally, given an action of a monoidal category $S$ on a category $C$, Grayson defines the quotient as a universal pair $(q,\alpha)$ where $q:C\to Q$ is a functor and $\alpha$ is an isomorphism between $q\circ\text{(action)}:S\times C\to C\to Q$ and $q\circ\text{(projection)}:S\times C\to C\to Q$. $\endgroup$ Commented Oct 20, 2017 at 19:33
  • $\begingroup$ @მამუკაჯიბლაძე Right... I suppose in that sense it's pretty straightforward. And this generalizes the usual set-level construction of the group completion of a commutative monoid. I suppose what's puzzling me is the fact that this construction coincides with group completion for sets and for spaces, but not for categories. Maybe I should formulate my question as follows: What is the relationship between this construction and group completion at the level of categories? $\endgroup$
    – Tim Campion
    Commented Oct 20, 2017 at 20:34

1 Answer 1


It is the classifying category for the left action of $C$ on its product $C \times C$.

Let me elaborate on this a bit further. Let $C$ be an $E_\infty$-monoid, for example represented by a symmetric monoidal groupoid (which is the case for Quillen's applications of e.g. $Proj(R)$). We can then consider the symmetric monoidal $\infty$-category $B C$ given by a single object and mapping space equal to $C$.There's a natural functor $C^\circlearrowleft : B C \rightarrow Spc$, essentially given by the functor corepresented by the single object. Here $Spc$ means the $\infty$-category of spaces.

Then we can consider the functor $C^\circlearrowleft \times C^\circlearrowleft : B C \rightarrow Spc$, and take its unstraightening, i.e. Grothendieck construction. This is given by the functor $C^{-1} C \rightarrow B C$ which sends the arrow $s+ : (n,m) \rightarrow (s+n,s+m)$ in $C^{-1} C$ to the arrow $* \xrightarrow{s} *$ in $B C$.

An almost trivial consequence of this is that the realization of $C^{-1} C$ is just the ($\infty$-categorical) colimit $(C^\circlearrowleft \times C^\circlearrowleft)_{hC}$.

This construction exists for an arbitrary $E_n$-monoid, but needs Quillen's additional assumptions that we started with a symmetric monoidal groupoid on which addition by an object is faithful to make sure that $C^{-1} C$ is actually (equivalent to) a 1-category.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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