10
$\begingroup$

[Cross-posted from this Math SE question.]

In Higher Algebra, Section 3.3 Lurie constructs the $\infty$-operads $\newcommand{\Mod}{\mathrm{Mod}}\newcommand{\cO}{\mathcal{O}}\newcommand{\cC}{\mathcal{C}}\Mod^{\cO}(\cC)^\otimes$ and $\Mod_A^{\cO}(\cC)^\otimes$ for $\cO^\otimes$ a unital $\infty$-operad, $\cC^\otimes \to \cO^\otimes$ a fibration of (generalized) $\infty$-operads, and $\newcommand{\Alg}{\mathrm{Alg}} A \in \Alg_{/ \cO}(\cC)$ an algebra object. The construction is rather involved: First he constructs a simplicial set $\widetilde{\Mod}^{\cO}(\cC)^\otimes$ together with a map $q\colon \widetilde{\Mod}^{\cO}(\cC)^\otimes \to \cO^\otimes$ such that for all maps of simplicial sets $X \to \cO^\otimes$ there is a canonical bijection $$\newcommand{\Fun}{\mathrm{Fun}} \Fun_{\cO^\otimes}(X, \widetilde{\Mod}^{\cO}(\cC)^\otimes) \cong \Fun_{\Fun(\{1\}, \cO^\otimes)}(X \times_{\Fun(\{0\}, \cO^\otimes)} \mathcal{K}_{\cO}, \cC^\otimes) $$ where $\mathcal{K}_{\cO} \subseteq \Fun(\Delta^1, \cO^\otimes)$ is the full subcategory of semi-inert morphisms in $\cO^\otimes$. He then picks out the full simplicial subset $\overline{\Mod}^{\cO}(\cC)^\otimes \subseteq \widetilde{\Mod}^{\cO}(\cC)^\otimes$ spanned by all vertices $\bar{v}$ such that $\bar{v}$ determines a functor $$ \{v\} \times_{\cO^\otimes} \mathcal{K}_{\cO} \to \cC^\otimes $$ which carries inert morphisms to inert morphisms (here a morphism in $\mathcal{K}_{\cO}$ are inert if its image under the evaluation maps $e_0, e_1\colon \mathcal{K}_{\cO} \to \cO^\otimes$ at the vertices of $\Delta^1$ is inert).

This is followed by around a page of observations, a similar definition for algebras, and finally the definition we care about: In the setting of the first sentence of this question, he defines $$ \Mod^{\cO}(\cC)^\otimes := \overline{\Mod}^{\cO}(\cC)^\otimes \times_{{}^\mathrm{p} \Alg_{/ \cO}(\cC)}(\cO^\otimes \times \Alg_{/ \cO}(\cC)) $$
and further $$ \Mod^{\cO}_A(\cC)^\otimes := \Mod^{\cO}(\cC)^\otimes \times_{\Alg_{/ \cO}(\cC)} \{A\} $$ (I won't describe all maps and constructions involved for brevity's sake).

This definition is, to put it mildly, a bit convoluted. In the introductory paragraph to the section, however, Lurie states that "the construction of $\Mod_A^{\cO}(\cC)^\otimes$ as a simplicial set with a map to $N(\mathrm{Fin}_*)$ is fairly straightforward; the bulk of our work will be in proving that $\Mod^\cO_A(\cC)^\otimes$ is actually an $\infty$-operad." I am thus led to wonder what the "straightforward construction" of $\Mod^\cO_A(\cC)^\otimes$ is supposed to be: Am I to "identify" $\Mod^\cO(\cC)^\otimes$ with $\overline{\Mod}^{\cO}(\cC)^\otimes$? After all, the map $\cO^\otimes \times \Alg_{/ \cO}(\cC) \to {}^\mathrm{p} \Alg_{/ \cO}(\cC)$ I am pulling back along in the definition of the former is a categorical equivalence, so I have a categorical equivalence $\Mod^\cO(\cC)^\otimes \to \overline{\Mod}^{\cO}(\cC)^\otimes$. Even so, the construction of $\overline{\Mod}^{\cO}(\cC)^\otimes$ in and of itself is a little complicated in its own right and not exactly pretty.

My question, thus: Assuming that I don't care about proving that $\Mod^\cO(\cC)^\otimes$ is an $\infty$-operad (or even an $\infty$-category), is there a construction of it that is simpler and/or more concise than the one of $\overline{\Mod}^\cO(\cC)^\otimes$ presented in Higher Algebra, possibly under additional simplifying assumptions?

(Context: I am giving a talk about $\infty$-operads in which I am supposed to introduce, among other things, modules. Giving the full definition would eat a chunk of my time that I already don't have, and any amount of simplification would probably aid understanding in my audience.)


N.b. I have looked at various other sources, e.g. Gepner's An Introduction to Higher Categorical Algebra or Hebestreit's Algebraic and Hermitian $K$-Theory Notes but it seems like everyone either avoids modules altogether or only treats the special case of left/right modules over monoids or modules over commutative monoids.

$\endgroup$
2
  • 1
    $\begingroup$ I definitely sympathize with the difficulties in grokking this section — I think I for one never really got this part. One think to note is that Lurie’s “modules” construction does not specialize either to left or right modules in the case of the associative operad: it specializes to bimodules. This actually makes some sense when you remember that for Lurie, modules over an $O$- algebra are $O$-monoidal, whereas left modules over an $ E_n$ algebra are only $E_{n-1}$ monoidal. I think Lurie says something about the strangeness of his setup in HA—iirc attributing the idea to John Francis. $\endgroup$ Commented Nov 4 at 16:15
  • 1
    $\begingroup$ I think Gepner and Haugseng’s enriched infinity categories work uses a construction of a similar flavor in a “nonsymmetric operad setting”. Perhaps they have a bit more discussion $\endgroup$ Commented Nov 4 at 16:18

1 Answer 1

4
$\begingroup$

$\DeclareMathOperator{\Map}{Map}\newcommand{\cO}{\mathcal{O}}\newcommand{\cC}{\mathcal{C}}$So how you convey the structure of a module depends on how you think about $\cO$-algebras and $\cO$-monoidal structures in the first place. The description you've given above is in terms of the technical scaffolding present in the quasicategorical formulation; I'm going to give slightly more informal descriptions that don't spend time on the coherence data.


Spaces and classical (single-object) operads are the easiest example. For $A$ to be an $\cO$-algebra, it needs action maps $\cO(n) \times_{\Sigma_n} A^n \to A$ that are compatible with operad composition. A module then can be interpreted as having maps $\cO(n+1) \times_{\Sigma_n} (A^n \times M) \to M$ compatible with the action of $\cO$ on $A$.


Let's start with $\cO$. An $\infty$-operad $\cO^\otimes$ has an underlying category with some collection of objects; but rather than just having $\Map_{\cO}(X,Y)$, for any objects $X_1,\dots,X_n$, and $Y$ we have spaces $\Map_{\cO}(X_1,\dots,X_n;Y)$ of "multilinear maps $\oplus X_i \to Y$", with multilinear composition rules.

The fibration $\cC^\otimes \to \cO^\otimes$ expresses a decomposition of $\cC$. For any object $X \in \cO$, we have a category $\cC_X$, and for any map $\alpha: \oplus X_i \to Y$ and objects $x_i \in \cC_{X_i}, y \in \cC_Y$, we have a collection $\Map_\alpha(x_1,\dots,x_n;y)$ of "maps $\alpha_!(x_1,\dots,x_n) \to y$". These have multilinear composition rules; and if our fibration was coCartesian, we legitimately can produce functors $\alpha_!: \cC_{X_1} \times \dots \times \cC_{X_n} \to \cC_Y$.

In these terms, the data of an $\cO$-algebra $A$ can be interpreted as the following:

  • For any X, we have an object $A(X) \in \cC_X$.
  • For any map $\alpha: \oplus X_i \to Y$ in $\cO$, we have a point $A(\alpha)$ in $\Map_\alpha(A(X_1),\dots,A(X_n);A(Y))$, thought of as a map $\alpha_!(A(X_1),\dots,A(X_n)) \to A(Y)$.
  • These respect composition. Roughly, $A(\alpha \circ (\beta_1, \dots, \beta_n)) = A(\alpha) \circ (A(\beta_1),\dots,A(\beta_n))$.

A module $M$ over $A$ carries similar data. It consists of:

  • An object $Y$ of $\cO$.
  • An object $M \in \cC_Y$.
  • For any map $\alpha: X_1 \oplus \dots \oplus Y \oplus \dots \oplus X_n \to Y$ in $\cO$, we have a map $M(\alpha)$ in $\Map_\alpha(A(X_1),\dots, M,\dots,A(X_n);M)$, thought of as a map $\alpha_!(A(X_1),M,\dots,A(X_n)) \to M$.
  • These respect composition. Roughly, $M(\alpha \circ (\beta_1, \dots, \beta_n)) = M(\alpha) \circ (A(\beta_1),\dots,M(\beta_i), \dots,A(\beta_n))$.
$\endgroup$

You must log in to answer this question.

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