We will now explain how to embed an arbitrary \(\infty \)-operad into the multimorphism operad of a presheaf category. This is the multiplicative version of the Yoneda embedding, first established by Nikolaus (2016), Proposition 2.7.
Lemma 18.2.1. Let \(D\) be a symmetric monoidal \(\infty \)-category. Then the Hom-functor \[ \Hom _D(-,-)\colon D\catop \times D \to \An \] admits a canonical lax symmetric monoidal structure, where \(\An \) carries the cartesian monoidal structure.
Proof. The twisted arrow construction preserves finite products, hence preserves commutative monoid objects of \(\Cat _{\infty }\). Thus the symmetric monoidal structure on \(D\) induces a symmetric monoidal structure on \(\Tw (D)\). The source and target projections \[ s\colon \Tw (D) \to D\catop , \qquad t\colon \Tw (D) \to D \] are natural in \(D\), and therefore assemble to a symmetric monoidal functor \[ (s,t)\colon \Tw (D) \to D\catop \times D. \] We claim that the induced functor on total categories \((s,t)^{\otimes }\colon \Tw (D)^{\otimes } \to (D\catop \times D)^{\otimes }\) is still a left fibration. Under straightening over \(\Span (\Fin )\), this functor is induced by the natural transformation whose value on a finite set \(I\) is the \(I\)-fold product \[ (s,t)^I\colon \Tw (D)^I \to (D\catop \times D)^I \] of the usual source-target left fibration. Since left fibrations are stable under finite products, Lemma 23.2.2 shows that \((s,t)^{\otimes }\) is a left fibration. The fiber over an object \(\{(x_i,y_i)\}_{i \in I}\) is the anima \(\prod _{i \in I}\Hom _D(x_i,y_i)\), with transport induced by the symmetric monoidal structure on \(D\) and composition in \(D\).
Straightening this left fibration gives a functor \[ H^{\otimes }\colon (D\catop \times D)^{\otimes } \to \An . \] On the fiber over \(\lra {1}\) this is the usual Hom-functor \(\Hom _D(-,-)\colon D\catop \times D \to \An \). Moreover, the description of its fibers shows that \(H^{\otimes }\) preserves finite products: products in \((D\catop \times D)^{\otimes }\) are given by concatenating finite tuples, and \(H^{\otimes }\) sends such a tuple to the corresponding product of hom animae. By the universal property of the cartesian monoidal structure on \(\An \) from Theorem 15.3.11, this finite-product-preserving functor is precisely a lax symmetric monoidal structure on \(\Hom _D(-,-)\). β‘
Proposition 18.2.2. Let \(D\) be a small symmetric monoidal \(\infty \)-category. Then the Yoneda embedding \[ Y\colon D \to \PSh (D) = \Fun (D\catop ,\An ) \] admits a canonical strong symmetric monoidal structure, where \(\PSh (D)\) is equipped with the Day convolution monoidal structure.
Proof. By the universal property of Day convolution, a lax symmetric monoidal structure on \(Y\) is equivalent to a lax symmetric monoidal structure on the associated functor \[ D\catop \times D \to \An , \qquad (d',d) \mapsto \Hom _D(d',d). \] This is provided by Lemma 18.2.1.
It remains to show that this lax structure is strong. Let \(\{d_i\}_{i \in I}\) be a finite collection of objects of \(D\). We must show that the induced map \[ \bigotimes ^{\Day }_{i \in I} Y(d_i) \to Y\left (\bigotimes _{i \in I} d_i\right ) \] is an isomorphism in \(\PSh (D)\). Testing against an arbitrary presheaf \(H\colon D\catop \to \An \), the left-hand side represents natural transformations \[ \Nat \left (\prod _{i \in I} Y(d_i)(-), H\left (\bigotimes _{i \in I} -\right )\right ) \] of functors \((D^I)\catop \to \An \) by the defining property of the Day convolution tensor product. Applying the Yoneda lemma for \(D^I\) identifies this anima with \(H(\bigotimes _{i \in I} d_i)\), which is also \(\Nat (Y(\bigotimes _{i \in I} d_i),H)\). This proves the claim. β‘
Proposition 18.2.3 (Multiplicative Yoneda embedding). Let \(\Oo \) be a small \(\infty \)-operad and put \(D := \Env (\Oo )\). The composite \[ \Oo \xrightarrow {\eta _{\Oo }} \Mm _D \xrightarrow {\Mm _Y} \Mm _{\PSh (D)} \] is fully faithful in the sense that it induces equivalences on all multimorphism animae. Moreover, if \(\Oo \) admits operadic limits indexed by some \(\infty \)-category \(I\), then the induced functor on colors \[ \Oo _{\lra {1}} \to \PSh (D) \] preserves \(I\)-indexed limits.
Proof. Let \(x_1,\dots ,x_n,y\) be colors of \(\Oo \). Since \(Y\) is strong symmetric monoidal by Proposition 18.2.2, we have natural equivalences \begin {align*} \Mm _{\PSh (D)}(Yx_1,\dots ,Yx_n;Yy) &\simeq \Hom _{\PSh (D)}(Yx_1 \otimes ^{\Day } \dots \otimes ^{\Day } Yx_n,Yy) \\ &\simeq \Hom _{\PSh (D)}(Y(x_1 \otimes \dots \otimes x_n),Yy) \\ &\simeq \Hom _D(x_1 \otimes \dots \otimes x_n,y) \\ &\simeq \Oo (x_1,\dots ,x_n;y). \end {align*}
The first and last equivalences are induced by the unit \(\eta _{\Oo }\) and agree with the identifications of Lemma 17.3.17. This proves full faithfulness. For the final claim, let \(y \to F\) be an operadic limit cone in \(\Oo _{\lra {1}}\). By Remark 18.1.2, its image in \(D = \Env (\Oo )\) is an ordinary limit cone. Since the Yoneda embedding preserves limits, its image in \(\PSh (D)\) is again a limit cone. β‘
Corollary 18.2.4. Let \(\Oo \) be a small \(\infty \)-operad that admits finite operadic limits and put \(D := \Env (\Oo )\). Then the color functor \[ \Oo _{\lra {1}} \to \PSh (D) \] induced by the multiplicative Yoneda embedding is fully faithful, preserves finite limits, and has essential image closed under finite limits in \(\PSh (D)\). In particular, by Lemma 14.1.8, we may identify \(\Oo \) with the full suboperad of \(\Mm _{\PSh (D)}\) spanned by this essential image. β‘
Generated from the authoritative LaTeX source.