Formerly publishing as Yuri Ximenes Martins.
Introduction
The search for an axiomatic and synthetic foundation for continuum physics and differential geometry—originating in Hilbert’s Sixth Problem [13 ] and Lawvere’s program of categorical dynamics and axiomatic cohesion [21 , 22 , 23 ] —has found a powerful modern realization in higher topos theory [26 , 38 ] . In Lawvere’s insight, a category of spaces H \mathbf{H} H is distinguished from an arbitrary category of sheaves by a string of adjoint functors relating it to a base category of discrete or homotopy types, capturing the dialectical tension between discrete points and continuous cohesion. When lifted to the setting of ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topoi by Schreiber [38 , 39 , 40 ] , an adjoint quadruple of geometric functors
Π ⊣ L C s t ⊣ Γ ⊣ c o D i s c : H ⟶ ∞ G r p d \Pi\dashv\mathrm{LCst} \dashv\Gamma\dashv\mathrm{coDisc} : \mathbf{H} \longrightarrow\infty\mathbf{Grpd} Π ⊣ LCst ⊣ Γ ⊣ coDisc : H ⟶ ∞ Grpd
induces an adjoint triple of internal modalities on H \mathbf{H} H —the shape modality ∫ = L C s t ∘ Π \int= \mathrm{LCst} \circ\Pi ∫ = LCst ∘ Π , the flat modality ♭ = L C s t ∘ Γ \flat= \mathrm{LCst} \circ\Gamma ♭ = LCst ∘ Γ , and the sharp modality ♯ = c o D i s c ∘ Γ \sharp= \mathrm{coDisc} \circ\Gamma ♯ = coDisc ∘ Γ :
∫ ⊣ ♭ ⊣ ♯ : H ⟶ H . \int\dashv\flat\dashv\sharp: \mathbf{H} \longrightarrow\mathbf{H}. ∫ ⊣ ♭ ⊣ ♯ : H ⟶ H .
Remarkably, this purely categorical structure suffices to internalize the kinematic (and, with linear/infinitesimal modalities, dynamical) structures of classical and quantum gauge field theories [10 , 19 , 45 ] :
(i) The shape modality ∫ \int ∫ governs geometric realization, covering spaces, and the classification of principal bundles;
(ii) The flat modality ♭ \flat ♭ and its homotopy kernel, the flat de Rham modality ♭ d R \flat_{\mathrm{dR}} ♭ dR , govern constant coefficients, local systems, and flat differential forms;
(iii) The interaction ∫ ⊣ ♭ \int\dashv\flat ∫ ⊣ ♭ produces universal Maurer–Cartan forms θ : G → ♭ d R B H G \theta: G \to\flat_{\mathrm{dR}}B H G θ : G → ♭ dR B H G and, via action groupoids and homotopy pullbacks with closed differential forms, the moduli stack of principal connections B c o n G B_{\mathrm{con}}G B con G ;
(iv) The full triple ∫ ⊣ ♭ ⊣ ♯ \int\dashv\flat\dashv\sharp ∫ ⊣ ♭ ⊣ ♯ modifies the naive internal hom [ X , B c o n G ] H [X, B_{\mathrm{con}}G]_{\mathbf{H}} [ X , B con G ] H to yield the true internal moduli stack of gauge connections G C o n n ( X ) ∈ H \mathrm{GConn}(X) \in\mathbf{H} GConn ( X ) ∈ H whose morphisms are gauge transformations rather than smooth concordances.
While the full ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categorical machinery is required for arbitrary higher gauge theories (such as B B B -fields, string structures, and M M M -theory fluxes [14 ] ), ordinary gauge theories of Yang–Mills, Chern–Simons, and Einstein–Hilbert–Palatini type [46 , 31 , 24 , 32 ] live naturally within the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical truncation, where hom-spaces are 1-groupoids (capturing gauge configurations and gauge transformations between them). Working explicitly in the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical setting of cohesive 2-topoi offers one advantage:
Concrete 2-categorical control: Every non-strict ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (bicategory in which all 2-cells are invertible) can be strictified up to equivalence, allowing diagrams, limits, colimits, and Kan extensions to be manipulated via concrete 2-categorical algebra [43 , 20 ] rather than infinite coherence towers. This could help the reader interested in the subject to have a first overview of the theory.
Furthermore, when generalizing existing theories or results, it is more natural to start to reason in some particular concrete limit. The present work is the initial step of a program with such kind of scope, and the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical truncation is the perfect concrete limit. Indeed, as initially described in [33 ] , a natural strategy concerning the generalization of ordinary gauge theories (which are about Lie algebra valued entities) is to consider their analogues inside some sort of Algebra-Valued Geometry , where other flavors of algebras play the role of “entity of infinitesimal symmetries” (for initial applications of this strategy, see [31 ] and [24 ] ).
In turn, in the same way as ordinary gauge theory admits a cohesive ∞ \infty ∞ -topoi presentation, one should expect to have an incarnation of Algebra-Valued Geometry in some sort of “Algebra-Valued Topoi”, having cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi as a particular case. Therefore, the present work not only reviews the theory of cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi, but does so in a way that makes it natural to understand the origin and motivation for those “Algebra-Valued Topoi”, whose theory will appear in subsequent works.
More precisely, in this paper, we provide a rigorous, self-contained review of cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi, the geometric structures existing on them, and how they model intrinsic cohomology and its differential refinement. Beyond serving as a pedagogical survey and dictionary for the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical truncation of [26 , 27 , 38 ] , our exposition in Section 3 3 3 formalizes the 2 2 2 -functorial comparison between the internal and external incarnations of delooping via Yoneda 2 2 2 -embeddings, and establishes their 2 2 2 -natural equivalence (Theorem 3.1 3.1 3.1 ). While this equivalence is mathematically straightforward in the cartesian setting due to objectwise limit computation and left-exact stackification, isolating its precise 2 2 2 -categorical mechanics is essential for understanding where and why it breaks down in our non-cartesian Algebra-Valued Topoi theory. Furthermore, we clarify the exact ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation boundaries of higher deloopings B H k G B_{\mathrm{H}}^{k}G B H k G , emphasizing that internal quotient deloopings are non-trivial only for k = 1 k=1 k = 1 , whereas higher deloopings ( k ≥ 2 ) (k \ge2) ( k ≥ 2 ) require internal groupoids or the ambient ∞ \infty ∞ -topos. Finally, in Section 4.3.3 4.3.3 4.3.3 , we analyze extensions of the abstract Maurer–Cartan morphism to non-groupal objects (Proposition 4.19 4.19 4.19 , Corollary 4.20 4.20 4.20 , and Conjecture 4.21 4.21 4.21 ), laying the groundwork for differential cohomology with non-groupal coefficients.
Organization of the paper
In Section 2 2 2 , we review the ( n , r ) (n,r) ( n , r ) -categorical hierarchy, ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf and Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi via stackification at coverages, internal groups and effective internal groupoids, symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories, and algebras over the little k k k -cubes ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads E k E_k E k , culminating in the 2 2 2 -truncated Dunn–Lurie additivity and stabilization theorems at k = 3 k=3 k = 3 . In Section 3 3 3 , we study delooping in cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmoi. We construct both the internal delooping B i n t B^{\mathrm{int}} B int and the external delooping B e x t B^{\mathrm{ext}} B ext (via transportation along Yoneda 2 2 2 -embeddings), prove their 2 2 2 -natural equivalence, characterize k k k -tuply groupal objects as groupal E k E_k E k -algebras via the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical Eckmann–Hilton argument, establish the looping–delooping equivalence B H : Grp ( τ ≤ 0 H ) ≃ Conn ( ∗ / H ) B_{\mathrm{H}} : \operatorname{Grp}(\tau_{\le0}\mathrm{H}) \simeq\operatorname{Conn}(* /\mathrm{H}) B H : Grp ( τ ≤ 0 H ) ≃ Conn ( ∗ / H ) in any Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, and clarify the truncation behavior of higher deloopings B H k G B_{\mathrm{H}}^{k}G B H k G . In Section 4 4 4 , we introduce cohesive modalities ( ∫ ⊣ ♭ ⊣ ♯ ) (\int\dashv\flat\dashv\sharp) ( ∫ ⊣ ♭ ⊣ ♯ ) , formalize principal k k k -bundles and their topological realization, construct the flat de Rham modality ♭ d R \flat_{\mathrm{dR}} ♭ dR , closed forms Ω c l k ( − ; A ) \Omega_{\mathrm{cl}}^{k}(-;A) Ω cl k ( − ; A ) , universal Maurer–Cartan forms θ k \theta_k θ k , nonabelian and higher connection moduli stacks B c o n k G B_{\mathrm{con}}^{k}G B con k G , and the sharp-adjusted internal moduli objects GConn k ( X ) \operatorname{GConn}^{k}(X) GConn k ( X ) , illustrated with a concrete worked example in S m o o t h G r p 2 \mathbf{SmoothGrp}_{2} SmoothGrp 2 . Finally, in Section 4.3 4.3 4.3 , we develop intrinsic, modal, and differential ( 2 , 1 ) (2,1) ( 2 , 1 ) -cohomology, investigate extensions of the Maurer–Cartan form to arbitrary pointed objects, and discuss internal classifying objects.
(2,1)-Topoi
In this section we recall the fundamentals of ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos theory. We begin with the ( n , r ) (n,r) ( n , r ) -categorical hierarchy and the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical context, followed by ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf categories and Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi defined via localizations at coverages. Next, we examine internal group and groupoid objects and the precise scope of effectiveness of internal groupoids in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos. Finally, we review symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories and algebraic objects modeled as algebras over ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads, specifically the little k k k -cubes operads E k E_k E k .
(2,1)-Context
For a given integer k ≥ 0 k \ge0 k ≥ 0 , let ( n , r ) C a t k + 1 (n,r)\mathrm{Cat}_{k+1} ( n , r ) Cat k + 1 denote the ( k + 1 ) (k+1) ( k + 1 ) -category of $(n,r)− c a t e g o r i e s , w h e r e $ 0 ≤ r ≤ n ≤ k -categories, where \$0 \le r \le n \le k − c a t e g or i es , w h er e $0 ≤ r ≤ n ≤ k (or more generally n , r ≤ k n,r \le k n , r ≤ k with the convention below) [1 , 26 ] . Recall that an n n n -category has only trivial (identity) j j j -morphisms for j > n j>n j > n , and an ( n , r ) (n,r) ( n , r ) -category is an n n n -category in which all j j j -morphisms for j > r j>r j > r are invertible.
If n ≤ r n \le r n ≤ r , the invertibility condition for j > r j>r j > r is vacuous since all j j j -morphisms with j > r ≥ n j>r \ge n j > r ≥ n are already trivial (identities); thus we write simply n C a t k + 1 n\mathrm{Cat}_{k+1} n Cat k + 1 for ( n , n ) C a t k + 1 (n,n)\mathrm{Cat}_{k+1} ( n , n ) Cat k + 1 .
If r = 0 r=0 r = 0 , ( n , 0 ) C a t k + 1 (n,0)\mathrm{Cat}_{k+1} ( n , 0 ) Cat k + 1 coincides with the ( k + 1 ) (k+1) ( k + 1 ) -category of n n n -groupoids, denoted n G r p d k + 1 n\mathrm{Grpd}_{k+1} n Grpd k + 1 .
If n = 0 n=0 n = 0 , then ( 0 , r ) C a t k + 1 (0,r)\mathrm{Cat}_{k+1} ( 0 , r ) Cat k + 1 is equivalent to S e t k + 1 \mathrm{Set}_{k+1} Set k + 1 , i.e., the category of sets trivially regarded as a ( k + 1 ) (k+1) ( k + 1 ) -category.
For every k , n , r k,n,r k , n , r , we have canonical ( k + 1 ) (k+1) ( k + 1 ) -embeddings
disc n , n + 1 k : ( n , r ) C a t k + 1 ↪ ( n + 1 , r ) C a t k + 1 , \operatorname{disc}_{n,n+1}^{k} : (n,r)\mathrm{Cat}_{k+1} \hookrightarrow(n+1,r)\mathrm{Cat}_{k+1}, disc n , n + 1 k : ( n , r ) Cat k + 1 ↪ ( n + 1 , r ) Cat k + 1 ,
which regard an ( n , r ) (n,r) ( n , r ) -category as a discrete ( n + 1 , r ) (n+1,r) ( n + 1 , r ) -category (adding only identity ( n + 1 ) (n+1) ( n + 1 ) -morphisms), as well as full subcategory inclusions ( n , r ) C a t k + 1 ↪ ( n , r + 1 ) C a t k + 1 (n,r)\mathrm{Cat}_{k+1} \hookrightarrow(n,r+1)\mathrm{Cat}_{k+1} ( n , r ) Cat k + 1 ↪ ( n , r + 1 ) Cat k + 1 . The discrete embedding d i s c n − 1 , n k \mathrm{disc}^{k}_{n-1,n} disc n − 1 , n k admits a left adjoint
τ ≤ n − 1 k : ( n , r ) C a t k + 1 ⟶ ( n − 1 , r ) C a t k + 1 \tau^{k}_{\leq n-1} : (n,r)\mathrm{Cat}_{k+1} \longrightarrow(n-1,r)\mathrm{Cat}_{k+1} τ ≤ n − 1 k : ( n , r ) Cat k + 1 ⟶ ( n − 1 , r ) Cat k + 1
given by truncation at ( n − 1 ) (n-1) ( n − 1 ) -morphisms (identifying ( n − 1 ) (n-1) ( n − 1 ) -morphisms that are connected by an invertible n n n -morphism), whereas the inclusion ( n , r ) C a t k + 1 ↪ ( n , r + 1 ) C a t k + 1 (n,r)\mathrm{Cat}_{k+1} \hookrightarrow(n,r+1)\mathrm{Cat}_{k+1} ( n , r ) Cat k + 1 ↪ ( n , r + 1 ) Cat k + 1 admits a right adjoint given by the core construction c o r e r + 1 , r k \mathrm{core}^{k}_{r+1,r} core r + 1 , r k , which discards non-invertible ( r + 1 ) (r+1) ( r + 1 ) -morphisms [26 , 36 ] . Because an ( n , r ) (n,r) ( n , r ) -category requires all j j j -morphisms for j > r j>r j > r to be invertible and possesses only identity morphisms in dimensions j > n j>n j > n , the invertibility condition is vacuous for r ≥ n r \geq n r ≥ n . Consequently, ( n , r ) C a t ≡ ( n , n ) C a t (n,r)\mathrm{Cat} \equiv(n,n)\mathrm{Cat} ( n , r ) Cat ≡ ( n , n ) Cat for all r ≥ n r \geq n r ≥ n , and the distinct classes of ( n , r ) (n,r) ( n , r ) -categories for n ≤ 2 n \leq2 n ≤ 2 form the lower-triangular hierarchy 0 ≤ r ≤ n ≤ 2 0 \leq r \leq n \leq2 0 ≤ r ≤ n ≤ 2 displayed below (both as a lower-triangular grid and, equivalently, as a diamond separating the groupoid and category branches):
Figure 2.1.
Since our primary goal is the synthetic formalization and generalization of ordinary gauge theories, in what follows we focus on the k = 1 k=1 k = 1 case—that is, the 2-categorical setting, with special emphasis on the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical context. Higher values of k k k are central to higher gauge theories (e.g., 2-groups, 3-groups, and string/fivebrane structures [2 , 10 ] ), which lie beyond the scope of this paper. Diagram (2.1) summarizes the fundamental adjunctions in the k = 1 k=1 k = 1 case.
(3) For n = 2 n=2 n = 2 , weak 2 2 2 -categories are naturally presented by Bénabou’s bicategories [3 ] . By the Mac Lane–Paré coherence and strictification theorem [30 , 20 ] , every bicategory is biequivalent to a strict 2 2 2 -category; in particular, every weak ( 2 , 1 ) (2,1) ( 2 , 1 ) -category is equivalent to a strict ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (i.e., a Grpd -enriched category).1
(4) For n > 2 n>2 n > 2 , there is no global strictification theorem: weak n n n -categories are not in general equivalent to strict n n n -categories (for n = 3 n=3 n = 3 , they are equivalent only to semi-strict Gray-categories [12 , 41 ] ).
(5) The fully weak ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categorical case is intimately tied to abstract homotopy theory, with combinatorial models (such as quasi-categories, complete Segal spaces, and simplicial categories) extensively developed by Joyal, Rezk, Lurie, and others [16 , 34 , 26 ] . Presentations of weak ( 2 , 1 ) (2,1) ( 2 , 1 ) -category theory can thus be systematically obtained by 2 2 2 -truncating combinatorial models of ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories.
Remark 2.3. In the strict setting, for every k k k and every m ≤ n m\le n m ≤ n , r ≤ k r\le k r ≤ k , there are also forgetful ( k + 1 ) (k+1) ( k + 1 ) -functors u n , n − m k : ( n , r ) C a t k + 1 → ( n − m , r ) C a t k + 1 u^{k}_{n,n-m}:(n,r)\mathrm{Cat}_{k+1}\to(n-m,r)\mathrm{Cat}_{k+1} u n , n − m k : ( n , r ) Cat k + 1 → ( n − m , r ) Cat k + 1 that discard i i i -morphisms for n − m < i ≤ n n-m<i\le n n − m < i ≤ n . These forgetful functors are generally ill-defined in the weak setting because composition of ( n − 1 ) (n-1) ( n − 1 ) -morphisms is associative only up to coherent n n n -isomorphisms, so one cannot simply discard higher-order cells. An important exception occurs when n = m n=m n = m , where
u n , 0 k = O b : ( n , r ) C a t k + 1 ⟶ S e t k + 1 u^{k}_{n,0}=\mathrm{Ob}:(n,r)\mathrm{Cat}_{k+1}\longrightarrow\mathrm{Set}_{k+1} u n , 0 k = Ob : ( n , r ) Cat k + 1 ⟶ Set k + 1
assigns to an ( n , r ) (n,r) ( n , r ) -category C \mathbf{C} C its underlying set of objects O b ( C ) \mathrm{Ob}(\mathbf{C}) Ob ( C ) . There is a canonical ( k + 1 ) (k+1) ( k + 1 ) -natural transformation π : u n , 0 k ⇒ τ ≤ 0 k \pi:u^{k}_{n,0}\Rightarrow\tau^{k}_{\leq0} π : u n , 0 k ⇒ τ ≤ 0 k : for every C \mathbf{C} C , π C : O b ( C ) ↠ I s o ( C ) = τ ≤ 0 C \pi_{\mathbf{C}}:\mathrm{Ob}(\mathbf{C})\twoheadrightarrow\mathrm{Iso}(\mathbf{C})=\tau_{\leq0}\mathbf{C} π C : Ob ( C ) ↠ Iso ( C ) = τ ≤ 0 C is the quotient projection sending each object to its equivalence class.
(2,1)-Topos
Recall that if C \mathbf{C} C is a small 1 1 1 -category and A \mathbf{A} A is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, we can form the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category ( 2 , 1 ) P S h v A ( C ) (2,1)\mathrm{PShv}_{\mathbf{A}}(\mathbf{C}) ( 2 , 1 ) PShv A ( C ) of 2 2 2 -functors (pseudofunctors) F : C 2 o p → A F:\mathbf{C}^{\mathrm{op}}_{2}\to\mathbf{A} F : C 2 op → A , pseudonatural transformations, and modifications [43 ] . Varying C \mathbf{C} C contravariantly via precomposition f ∗ f^{*} f ∗ yields a 2 2 2 -functor
( 2 , 1 ) P S h v A ( − ) : C a t 2 o p ⟶ ( 2 , 1 ) C a t 2 , (2,1)\mathrm{PShv}_{\mathbf{A}}(-):\mathrm{Cat}^{\mathrm{op}}_{2}\longrightarrow(2,1)\mathrm{Cat}_{2}, ( 2 , 1 ) PShv A ( − ) : Cat 2 op ⟶ ( 2 , 1 ) Cat 2 ,
which factors through ( n , r ) C a t 2 ⊂ 2 C a t 2 (n,r)\mathrm{Cat}_{2}\subset2\mathrm{Cat}_{2} ( n , r ) Cat 2 ⊂ 2 Cat 2 whenever A \mathbf{A} A is an ( n , r ) (n,r) ( n , r ) -category for 0 ≤ r ≤ n ≤ 2 0\le r\le n\le2 0 ≤ r ≤ n ≤ 2 . When A \mathbf{A} A is complete and cocomplete, precomposition f ∗ : ( 2 , 1 ) P S h v A ( D ) → ( 2 , 1 ) P S h v A ( C ) f^{*}:(2,1)\mathrm{PShv}_{\mathbf{A}}(\mathbf{D})\to(2,1)\mathrm{PShv}_{\mathbf{A}}(\mathbf{C}) f ∗ : ( 2 , 1 ) PShv A ( D ) → ( 2 , 1 ) PShv A ( C ) along f : C → D f:\mathbf{C}\to\mathbf{D} f : C → D admits both a left 2 2 2 -adjoint f ! = Lan f o p f_{!}=\operatorname{Lan}_{f^{\mathrm{op}}} f ! = Lan f op (left Kan extension along f o p f^{\mathrm{op}} f op ) and a right 2 2 2 -adjoint f ∗ = Ran f o p f_{*}=\operatorname{Ran}_{f^{\mathrm{op}}} f ∗ = Ran f op (right Kan extension along f o p f^{\mathrm{op}} f op ); in particular, f ↦ f ! f\mapsto f_{!} f ↦ f ! defines a covariant 2 2 2 -functor ( 2 , 1 ) P S h v A c o v ( − ) : C a t 2 → ( 2 , 1 ) C a t 2 (2,1)\mathrm{PShv}^{\mathrm{cov}}_{\mathbf{A}}(-):\mathrm{Cat}_{2}\to(2,1)\mathrm{Cat}_{2} ( 2 , 1 ) PShv A cov ( − ) : Cat 2 → ( 2 , 1 ) Cat 2 . We call ( 2 , 1 ) P S h v A ( C ) (2,1)\mathrm{PShv}_{\mathbf{A}}(\mathbf{C}) ( 2 , 1 ) PShv A ( C ) the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category of A \mathbf{A} A -valued ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaves (or prestack s) on C \mathbf{C} C . In this work we focus primarily on the canonical case A = G r p d 2 \mathbf{A}=\mathbf{Grpd}_{2} A = Grpd 2 .
We say that a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -category if it is 2 2 2 -equivalent to ( 2 , 1 ) P S h v G r p d 2 ( C ) (2,1)\mathrm{PShv}_{\mathbf{Grpd}_{2}}(\mathbf{C}) ( 2 , 1 ) PShv Grpd 2 ( C ) for some small 1 1 1 -category C \mathbf{C} C .2 A geometric morphism f : H → H ′ f:\mathbf{H}\to\mathbf{H}' f : H → H ′ between two ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories is an adjoint pair of 2 2 2 -functors ( f ∗ ⊣ f ∗ ) (f^{*}\dashv f_{*}) ( f ∗ ⊣ f ∗ ) , with direct image f ∗ : H → H ′ f_{*}:\mathbf{H}\to\mathbf{H}' f ∗ : H → H ′ and inverse image f ∗ : H ′ → H f^{*}:\mathbf{H}'\to\mathbf{H} f ∗ : H ′ → H , such that f ∗ f^{*} f ∗ is left exact, i.e., preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits (first diagram below). A geometric transformation ξ : f ⇒ g \xi:f\Rightarrow g ξ : f ⇒ g between geometric morphisms f , g : H → H ′ f,g:\mathbf{H}\to\mathbf{H}' f , g : H → H ′ is a mate pair ( ξ ∗ , ξ ∗ ) (\xi_{*},\xi^{*}) ( ξ ∗ , ξ ∗ ) of 2 2 2 -natural transformations ξ ∗ : f ∗ ⇒ g ∗ \xi_{*}:f_{*}\Rightarrow g_{*} ξ ∗ : f ∗ ⇒ g ∗ and ξ ∗ : g ∗ ⇒ f ∗ \xi^{*}:g^{*}\Rightarrow f^{*} ξ ∗ : g ∗ ⇒ f ∗ (where our convention for 2 2 2 -cells in the 2 2 2 -category of topoi follows the direction of ξ ∗ : f ∗ ⇒ g ∗ \xi_{*}:f_{*}\Rightarrow g_{*} ξ ∗ : f ∗ ⇒ g ∗ , with mate ξ ∗ : g ∗ ⇒ f ∗ \xi^{*}:g^{*}\Rightarrow f^{*} ξ ∗ : g ∗ ⇒ f ∗ ; second diagram below). This defines the 2 2 2 -category ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathrm{PShvTopos}_{2} ( 2 , 1 ) PShvTopos 2 of ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi, geometric morphisms, and geometric transformations [44 , 26 ] .
H → f ∗ H ′ f ∗ ↑ ↓ f ∗ H → f ∗ H ′ \begin{CD}
\mathbf{H} @>{f^*}>> \mathbf{H}'\\
@A{f_*}AA @VV{f_*}V\\
\mathbf{H} @>{f^*}>> \mathbf{H}'
\end{CD} H f ∗ ⏐ ↑ H f ∗ f ∗ H ′ ↓ ⏐ f ∗ H ′
H → f ∗ H ′ g ∗ ↑ ↓ g ∗ H → f ∗ H ′ \begin{CD}
\mathbf{H} @>{f^*}>> \mathbf{H}'\\
@A{g^*}AA @VV{g^*}V\\
\mathbf{H} @>{f^*}>> \mathbf{H}'
\end{CD} H g ∗ ⏐ ↑ H f ∗ f ∗ H ′ ↓ ⏐ g ∗ H ′
The 2-category ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathbf{PShvTopos}_2 ( 2 , 1 ) PShvTopos 2 is a full sub-2-category of the 2-category ( 2 , 1 ) P T o p o s 2 (2,1)\mathbf{PTopos}_2 ( 2 , 1 ) PTopos 2 of (infinitary) ( 2 , 1 ) (2,1) ( 2 , 1 ) -pretopoi with geometric morphisms as 1-morphisms, i.e., ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories that are 2-exact (effective internal groupoids and universal colimits) and infinitary 2-extensive (possessing disjoint, universal small coproducts) [44 , 35 ] . Between ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathbf{PShvTopos}_2 ( 2 , 1 ) PShvTopos 2 and ( 2 , 1 ) P T o p o s 2 (2,1)\mathbf{PTopos}_2 ( 2 , 1 ) PTopos 2 lies the full sub-2-category ( 2 , 1 ) S h v T o p o s 2 (2,1)\mathbf{ShvTopos}_2 ( 2 , 1 ) ShvTopos 2 of Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi, which are ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories 2-equivalent to full sub-2-categories H 0 ⊂ H \mathbf{H}_0 \subset\mathbf{H} H 0 ⊂ H of a ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf category H ≃ ( 2 , 1 ) P S h v G r p d 2 ( C ) \mathbf{H} \simeq(2,1)\mathbf{PShvGrpd}_2(\mathbf{C}) H ≃ ( 2 , 1 ) PShvGrpd 2 ( C ) such that:
H 0 \mathbf{H}_0 H 0 is reflective : the full inclusion 2-functor ι : H 0 ↪ H \iota: \mathbf{H}_0 \hookrightarrow\mathbf{H} ι : H 0 ↪ H admits a left 2-adjoint L 0 : H → H 0 L_0 : \mathbf{H} \to\mathbf{H}_0 L 0 : H → H 0 (the stackification or sheafification functor);
H 0 \mathbf{H}_0 H 0 is left exact : the left adjoint L 0 L_0 L 0 preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits.
Remark 2.4. The 2-category ( 2 , 1 ) P T o p o s 2 (2,1)\mathbf{PTopos}_2 ( 2 , 1 ) PTopos 2 (and its full sub-2-categories ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathbf{PShvTopos}_2 ( 2 , 1 ) PShvTopos 2 , ( 2 , 1 ) S h v T o p o s 2 (2,1)\mathbf{ShvTopos}_2 ( 2 , 1 ) ShvTopos 2 ) is not a sub-2-category of ( 2 , 1 ) C a t 2 (2,1)\mathbf{Cat}_2 ( 2 , 1 ) Cat 2 with respect to arbitrary 2-functors. However, there are two dual identity-on-objects forgetful 2-functors:
U t o p o s = ( − ) ∗ : ( 2 , 1 ) P T o p o s 2 → ( 2 , 1 ) C a t 2 , U t o p o s ∗ = ( − ) ∗ : ( 2 , 1 ) P T o p o s 2 o p → ( 2 , 1 ) C a t 2 , U_{\mathrm{topos}} = (-)_* : (2,1)\mathbf{PTopos}_2 \to(2,1)\mathbf{Cat}_2,\qquad U_{\mathrm{topos}}^* = (-)^* : (2,1)\mathbf{PTopos}_2^{\mathrm{op}} \to(2,1)\mathbf{Cat}_2, U topos = ( − ) ∗ : ( 2 , 1 ) PTopos 2 → ( 2 , 1 ) Cat 2 , U topos ∗ = ( − ) ∗ : ( 2 , 1 ) PTopos 2 op → ( 2 , 1 ) Cat 2 ,
where U t o p o s U_{\mathrm{topos}} U topos sends a geometric morphism f = ( f ∗ , f ∗ ) f = (f^*,f_*) f = ( f ∗ , f ∗ ) to its direct image 2-functor f ∗ : H → H ′ f_* : \mathbf{H} \to\mathbf{H}' f ∗ : H → H ′ (which preserves all ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits), while U t o p o s ∗ U_{\mathrm{topos}}^* U topos ∗ sends f f f to its inverse image 2-functor f ∗ : H ′ → H f^* : \mathbf{H}' \to\mathbf{H} f ∗ : H ′ → H (which preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and all ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits).
By the 2-categorical Giraud Theorem (due to Giraud [11 ] and Street [44 ] , and corresponding to the 2-truncation of the Rezk–Lurie ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -Giraud theorem [35 , 26 ] ), the 2-category ( 2 , 1 ) S h v T o p o s 2 (2,1)\mathbf{ShvTopos}_2 ( 2 , 1 ) ShvTopos 2 is 2-equivalent to the 2-category S t k 2 \mathbf{Stk}_2 Stk 2 of stack ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories. Explicitly, if C \mathbf{C} C is a small 1-category endowed with a Grothendieck coverage (or pretopology) J J J , making ( C , J ) (\mathbf{C},J) ( C , J ) a site,3 the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category of stacks S t k ( C , J ) \mathbf{Stk}(\mathbf{C},J) Stk ( C , J ) is obtained as the left-exact reflective localization of ( 2 , 1 ) P S h v G r p d 2 ( C ) (2,1)\mathbf{PShvGrpd}_2(\mathbf{C}) ( 2 , 1 ) PShvGrpd 2 ( C ) at the class W J W_J W J of local J J J -equivalences :
S t k ( C , J ) ≃ ( 2 , 1 ) P S h v G r p d 2 ( C ) [ W J − 1 ] . (1) \mathbf{Stk}(\mathbf{C},J) \simeq(2,1)\mathbf{PShvGrpd}_2(\mathbf{C})[W_J^{-1}]. \tag*{(1)} Stk ( C , J ) ≃ ( 2 , 1 ) PShvGrpd 2 ( C ) [ W J − 1 ] . (1)
Let S i t e 2 \mathbf{Site}_2 Site 2 denote the 2-category of small sites and cover-preserving functors. It comes equipped with a faithful forgetful 2-functor U s i t e : S i t e 2 → C a t 2 U_{\mathrm{site}} : \mathbf{Site}_2 \to\mathbf{Cat}_2 U site : Site 2 → Cat 2 sending ( C , J ) ↦ C (\mathbf{C},J) \mapsto\mathbf{C} ( C , J ) ↦ C . For a general coverage J J J , the underlying category C \mathbf{C} C need not have all pullbacks. However, constructing the localization at W J W_J W J via Čech descent requires pullbacks along covering families. Let C a t 2 , p b ⊂ C a t 2 \mathbf{Cat}_{2,\mathrm{pb}} \subset\mathbf{Cat}_2 Cat 2 , pb ⊂ Cat 2 be the sub-2-category of small 1-categories with pullbacks and pullback-preserving functors, and let S i t e 2 , p b ⊂ S i t e 2 \mathbf{Site}_{2,\mathrm{pb}} \subset\mathbf{Site}_2 Site 2 , pb ⊂ Site 2 be the pullback sub-2-category of sites with pullbacks and cover- and pullback-preserving functors:
C a t 2 , p b → C a t 2 U s i t e ↑ ↑ U s i t e S i t e 2 , p b → S i t e 2 \begin{CD}
\mathbf{Cat}_{2,\mathrm{pb}} @>>> \mathbf{Cat}_2\\
@A{U_{\mathrm{site}}}AA @AA{U_{\mathrm{site}}}A\\
\mathbf{Site}_{2,\mathrm{pb}} @>>> \mathbf{Site}_2
\end{CD} Cat 2 , pb U site ⏐ ↑ Site 2 , pb Cat 2 ⏐ ↑ U site Site 2
Restricted to S i t e 2 , p b \mathbf{Site}_{2,\mathrm{pb}} Site 2 , pb , the assignment ( C , J ) ↦ S t k ( C , J ) (\mathbf{C},J) \mapsto\mathbf{Stk}(\mathbf{C},J) ( C , J ) ↦ Stk ( C , J ) is 2-functorial: for any morphism f : ( C , J ) → ( D , K ) f : (\mathbf{C},J) \to(\mathbf{D},K) f : ( C , J ) → ( D , K ) in S i t e 2 , p b \mathbf{Site}_{2,\mathrm{pb}} Site 2 , pb , precomposition f ∗ : ( 2 , 1 ) P S h v G r p d 2 ( D ) → ( 2 , 1 ) P S h v G r p d 2 ( C ) f^* : (2,1)\mathbf{PShvGrpd}_2(\mathbf{D}) \to(2,1)\mathbf{PShvGrpd}_2(\mathbf{C}) f ∗ : ( 2 , 1 ) PShvGrpd 2 ( D ) → ( 2 , 1 ) PShvGrpd 2 ( C ) preserves local equivalences (f ∗ ( W K ) ⊂ W J f^*(W_K) \subset W_J f ∗ ( W K ) ⊂ W J ) and stacks, descending via the universal property of localization to the inverse image functor f s ∗ : S t k ( D , K ) → S t k ( C , J ) f_s^* : \mathbf{Stk}(\mathbf{D},K) \to\mathbf{Stk}(\mathbf{C},J) f s ∗ : Stk ( D , K ) → Stk ( C , J ) of the induced geometric morphism ( f s ∗ ⊣ f s ∗ ) (f_s^* \dashv f_{s*}) ( f s ∗ ⊣ f s ∗ ) satisfying f s ∗ ∘ L K ≃ L J ∘ f ∗ f_s^* \circ L_K \simeq L_J \circ f^* f s ∗ ∘ L K ≃ L J ∘ f ∗ . Thus, passing to inverse image 2-functors via U t o p o s ∗ = ( − ) ∗ U_{\mathrm{topos}}^* = (-)^* U topos ∗ = ( − ) ∗ on Stk 2 o p \operatorname{Stk}_{2}^{\mathrm{op}} Stk 2 op and ( 2 , 1 ) PShvTopos 2 o p (2,1)\operatorname{PShvTopos}_{2}^{\mathrm{op}} ( 2 , 1 ) PShvTopos 2 op , we obtain a contravariant 2-functor U t o p o s ∗ ∘ Stk : Site 2 , p b o p → Stk 2 o p → ( 2 , 1 ) Cat 2 U_{\mathrm{topos}}^{*}\circ\operatorname{Stk}:\operatorname{Site}_{2,\mathrm{pb}}^{\mathrm{op}}\to\operatorname{Stk}_{2}^{\mathrm{op}}\to(2,1)\operatorname{Cat}_{2} U topos ∗ ∘ Stk : Site 2 , pb op → Stk 2 op → ( 2 , 1 ) Cat 2 , and the stackification localization extends to a 2-natural transformation L : ( 2 , 1 ) PShvGrpd 2 ∘ U s i t e ⇒ U t o p o s ∗ ∘ Stk L:(2,1)\operatorname{PShvGrpd}_{2}\circ U_{\mathrm{site}}\Rightarrow U_{\mathrm{topos}}^{*}\circ\operatorname{Stk} L : ( 2 , 1 ) PShvGrpd 2 ∘ U site ⇒ U topos ∗ ∘ Stk , as depicted below:
( 2 , 1 ) Cat 2 ← U t o p o s ∗ ( 2 , 1 ) PShvTopos 2 o p U t o p o s ∗ ↑ ↑ ( 2 , 1 ) PShvGrpd 2 ( − ) Stk 2 o p → Stk Cat 2 , p b o p → Cat 2 o p Stk ↑ ↑ U s i t e ↑ U s i t e Site 2 , p b o p → Site 2 o p (2) \begin{CD}
(2,1)\operatorname{Cat}_{2} @<{U_{\mathrm{topos}}^{*}}<< (2,1)\operatorname{PShvTopos}_{2}^{\mathrm{op}} \\
@A{U_{\mathrm{topos}}^{*}}AA @AA{(2,1)\operatorname{PShvGrpd}_{2}(-)}A \\
\operatorname{Stk}_{2}^{\mathrm{op}} @>{\operatorname{Stk}}>> \operatorname{Cat}_{2,\mathrm{pb}}^{\mathrm{op}} @>{}>> \operatorname{Cat}_{2}^{\mathrm{op}} \\
@A{\operatorname{Stk}}AA @AA{U_{\mathrm{site}}}A @AA{U_{\mathrm{site}}}A \\
\operatorname{Site}_{2,\mathrm{pb}}^{\mathrm{op}} @>{}>> \operatorname{Site}_{2}^{\mathrm{op}} \tag*{(2)}
\end{CD} ( 2 , 1 ) Cat 2 U topos ∗ ⏐ ↑ Stk 2 op Stk ⏐ ↑ Site 2 , pb op U topos ∗ Stk ( 2 , 1 ) PShvTopos 2 op ⏐ ↑ ( 2 , 1 ) PShvGrpd 2 ( − ) Cat 2 , pb op ⏐ ↑ U site Site 2 op Cat 2 op ⏐ ↑ U site (2)
Furthermore, any section 2-functor Site : Cat 2 , p b → Site 2 , p b \operatorname{Site}:\operatorname{Cat}_{2,\mathrm{pb}}\to\operatorname{Site}_{2,\mathrm{pb}} Site : Cat 2 , pb → Site 2 , pb endowing categories with a canonical coverage can be characterized by universal properties of the induced composite localization transformation L Site : ( 2 , 1 ) PShvGrpd 2 ( − ) ⇒ U t o p o s ∗ ∘ Stk ∘ Site L_{\operatorname{Site}}:(2,1)\operatorname{PShvGrpd}_{2}(-)\Rightarrow U_{\mathrm{topos}}^{*}\circ\operatorname{Stk}\circ\operatorname{Site} L Site : ( 2 , 1 ) PShvGrpd 2 ( − ) ⇒ U topos ∗ ∘ Stk ∘ Site :
( 2 , 1 ) Cat 2 ← U t o p o s ∗ ( 2 , 1 ) PShvTopos 2 o p U t o p o s ∗ ↑ ↑ ( 2 , 1 ) PShvGrpd 2 ( − ) Stk 2 o p → Stk Cat 2 , p b o p → Cat 2 o p Stk ↑ ↑ Site ↑ U s i t e Site 2 , p b o p → Site 2 o p Cat 2 , p b o p → L Site ( 2 , 1 ) Cat 2 U t o p o s ∗ ∘ Stk ∘ Site ↑ ↑ ( 2 , 1 ) PShvGrpd 2 ( − ) ( 2 , 1 ) PShvGrpd 2 → \begin{CD}
(2,1)\operatorname{Cat}_{2} @<{U_{\mathrm{topos}}^{*}}<< (2,1)\operatorname{PShvTopos}_{2}^{\mathrm{op}} \\
@A{U_{\mathrm{topos}}^{*}}AA @AA{(2,1)\operatorname{PShvGrpd}_{2}(-)}A \\
\operatorname{Stk}_{2}^{\mathrm{op}} @>{\operatorname{Stk}}>> \operatorname{Cat}_{2,\mathrm{pb}}^{\mathrm{op}} @>{}>> \operatorname{Cat}_{2}^{\mathrm{op}} \\
@A{\operatorname{Stk}}AA @AA{\operatorname{Site}}A @AA{U_{\mathrm{site}}}A \\
\operatorname{Site}_{2,\mathrm{pb}}^{\mathrm{op}} @>{}>> \operatorname{Site}_{2}^{\mathrm{op}}
\end{CD}
\qquad
\begin{CD}
\operatorname{Cat}_{2,\mathrm{pb}}^{\mathrm{op}} @>{L_{\operatorname{Site}}}>> (2,1)\operatorname{Cat}_{2} \\
@A{U_{\mathrm{topos}}^{*}\circ\operatorname{Stk}\circ\operatorname{Site}}AA @AA{(2,1)\operatorname{PShvGrpd}_{2}(-)}A \\
(2,1)\operatorname{PShvGrpd}_{2} @>{}>>
\end{CD} ( 2 , 1 ) Cat 2 U topos ∗ ⏐ ↑ Stk 2 op Stk ⏐ ↑ Site 2 , pb op U topos ∗ Stk ( 2 , 1 ) PShvTopos 2 op ⏐ ↑ ( 2 , 1 ) PShvGrpd 2 ( − ) Cat 2 , pb op ⏐ ↑ Site Site 2 op Cat 2 op ⏐ ↑ U site Cat 2 , pb op U topos ∗ ∘ Stk ∘ Site ⏐ ↑ ( 2 , 1 ) PShvGrpd 2 L Site ( 2 , 1 ) Cat 2 ⏐ ↑ ( 2 , 1 ) PShvGrpd 2 ( − )
For example:
The trivial coverage Site 0 ( C ) \operatorname{Site}_{0}(\mathbf{C}) Site 0 ( C ) , whose covering families consist solely of identity morphisms, is the universal coverage for which L Site 0 L_{\operatorname{Site}_{0}} L Site 0 is a 2-natural equivalence. Consequently, every ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf category ( 2 , 1 ) PShvGrpd 2 ( C ) (2,1)\operatorname{PShvGrpd}_{2}(\mathbf{C}) ( 2 , 1 ) PShvGrpd 2 ( C ) is canonically a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos equipped with the trivial coverage.
A subcanonical coverage Site s u b ( C ) \operatorname{Site}_{\mathrm{sub}}(\mathbf{C}) Site sub ( C ) is one for which the localization L Site L_{\operatorname{Site}} L Site restricts to an equivalence on all representable ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaves y C ( X ) y_{\mathbf{C}}(X) y C ( X ) . Explicitly, if C \mathbf{C} C has coproducts that are universal (preserved by pullbacks), a coverage J J J on C \mathbf{C} C is subcanonical if and only if for every covering family { U i → X } i ∈ I \{U_i\to X\}_{i\in I} { U i → X } i ∈ I in J J J , the induced Čech diagram
∐ i , j U i × X U j ⇉ ∐ i U i ⟶ X \coprod_{i,j} U_i\times_X U_j \rightrightarrows\coprod_i U_i \longrightarrow X i , j ∐ U i × X U j ⇉ i ∐ U i ⟶ X
is a colimit (coequalizer) diagram in C \mathbf{C} C .
Remark 2.5. Recall that in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H with ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks, a 1-morphism f : U → X f:U\to X f : U → X is an effective epimorphism if the diagram
U × X U → pr 1 → pr 2 U → f X U\times_X U \mathrel{\substack{\xrightarrow{\operatorname{pr}_{1}}\\[-0.4em]\xrightarrow[\operatorname{pr}_{2}]{{}}}} U \xrightarrow{f} X U × X U pr 1 pr 2 U f X
(together with the 2-cell witnessing the homotopy pullback) exhibits X X X as the ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimit (homotopy coequalizer) of the kernel pair [26 ] . If C \mathbf{C} C is a 1-category with pullbacks and universal coproducts (for example, an extensive or cartesian closed 1-category with coproducts), pullbacks distribute over coproducts; hence, setting U = ∐ i U i U=\coprod_i U_i U = ∐ i U i , a coverage J J J is subcanonical if and only if the summed covering morphisms U → X U\to X U → X are effective epimorphisms.
Internal Groupoids
Let C a t 2 , c ⊂ C a t 2 \mathbf{Cat}_{2,c} \subset\mathbf{Cat}_2 Cat 2 , c ⊂ Cat 2 be the sub-2-category of cartesian 1-categories (categories with finite products, i.e., binary products and a terminal object ∗ \ast ∗ ) and cartesian (finite-product-preserving) functors. Let C a t 2 , c p = C a t 2 , c ∩ C a t 2 , p b \mathbf{Cat}_{2,cp} = \mathbf{Cat}_{2,c} \cap\mathbf{Cat}_{2,pb} Cat 2 , c p = Cat 2 , c ∩ Cat 2 , p b be the sub-2-category of categories with all finite limits (pullbacks and terminal object) and left-exact functors.
Recall that in every C ∈ C a t 2 , c \mathbf{C} \in\mathbf{Cat}_{2,c} C ∈ Cat 2 , c there is an internal notion of group object . A group object is an object G ∈ C G \in\mathbf{C} G ∈ C equipped with multiplication m : G × G → G m : G \times G \to G m : G × G → G , unit e : ∗ → G e : \ast\to G e : ∗ → G , and inversion inv : G → G \operatorname{inv} : G \to G inv : G → G satisfying the standard associative, unital, and inverse commutative diagrams [28 ] . Homomorphisms of group objects form a category Grp ( C ) \operatorname{Grp}(\mathbf{C}) Grp ( C ) equipped with a faithful forgetful functor to C \mathbf{C} C (written with a hooked arrow by abuse of notation). Denoting by Grp 2 ( C ) = disc 1 , 2 ( Grp ( C ) ) \operatorname{Grp}_2(\mathbf{C}) = \operatorname{disc}_{1,2}(\operatorname{Grp}(\mathbf{C})) Grp 2 ( C ) = disc 1 , 2 ( Grp ( C )) the corresponding discrete ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, and noting that cartesian functors preserve group objects, we obtain a 2-functor
Grp 2 : C a t 2 , c ⟶ ( 2 , 1 ) C a t 2 . \operatorname{Grp}_2 : \mathbf{Cat}_{2,c} \longrightarrow(2,1)\mathbf{Cat}_2. Grp 2 : Cat 2 , c ⟶ ( 2 , 1 ) Cat 2 .
Similarly, in every C ∈ C a t 2 , c p \mathbf{C} \in\mathbf{Cat}_{2,cp} C ∈ Cat 2 , c p there is an internal notion of groupoid object (or internal groupoid ) G = ( G 1 ⇉ G 0 ) \mathbf{G} = (G_1 \rightrightarrows G_0) G = ( G 1 ⇉ G 0 ) , consisting of an object of objects G 0 G_0 G 0 , an object of morphisms G 1 G_1 G 1 , source and target morphisms s , t : G 1 → G 0 s,t : G_1 \to G_0 s , t : G 1 → G 0 , composition ∘ : G 1 × s , G 0 , t G 1 → G 1 \circ: G_1 \mathbin{\times_{s,G_0,t}} G_1 \to G_1 ∘ : G 1 × s , G 0 , t G 1 → G 1 , identity id : G 0 → G 1 \operatorname{id} : G_0 \to G_1 id : G 0 → G 1 , and inversion inv : G 1 → G 1 \operatorname{inv} : G_1 \to G_1 inv : G 1 → G 1 satisfying associativity, unitality, and invertibility equations [15 ] . Together with internal functors and internal natural transformations , internal groupoids in C \mathbf{C} C assemble into a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category Grpd 2 ( C ) \operatorname{Grpd}_2(\mathbf{C}) Grpd 2 ( C ) . Every object X ∈ C X \in\mathbf{C} X ∈ C defines a discrete internal groupoid E X = ( X ⇉ X ) E_X = (X \rightrightarrows X) E X = ( X ⇉ X ) with ( E X ) 0 = ( E X ) 1 = X (E_X)_0 = (E_X)_1 = X ( E X ) 0 = ( E X ) 1 = X and s = t = id X s = t = \operatorname{id}_X s = t = id X , while every group object G ∈ Grp ( C ) G \in\operatorname{Grp}(\mathbf{C}) G ∈ Grp ( C ) defines a one-object internal groupoid B C ( G ) = ( G ⇉ ∗ ) B_{\mathbf{C}}(G) = (G \rightrightarrows\ast) B C ( G ) = ( G ⇉ ∗ ) with object of objects ∗ \ast ∗ . Thus, for each C ∈ C a t 2 , c p \mathbf{C} \in\mathbf{Cat}_{2,cp} C ∈ Cat 2 , c p , we have canonical inclusions of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories (where Grp 2 ( C ) \operatorname{Grp}_2(\mathbf{C}) Grp 2 ( C ) embeds both via discrete groupoids on the underlying object and via one-object delooping groupoids; below we display the underlying-object inclusion):
Grp 2 ( C ) ⟶ disc 1 , 2 ( C ) ⟶ Grpd 2 ( C ) . \operatorname{Grp}_2(\mathbf{C}) \longrightarrow\operatorname{disc}_{1,2}(\mathbf{C}) \longrightarrow\operatorname{Grpd}_2(\mathbf{C}). Grp 2 ( C ) ⟶ disc 1 , 2 ( C ) ⟶ Grpd 2 ( C ) .
Since left-exact functors preserve pullbacks and hence internal groupoids, Grpd 2 : C a t 2 , c p ⟶ ( 2 , 1 ) C a t 2 \operatorname{Grpd}_2 : \mathbf{Cat}_{2,cp} \longrightarrow(2,1)\mathbf{Cat}_2 Grpd 2 : Cat 2 , c p ⟶ ( 2 , 1 ) Cat 2 is a 2-functor, and we have a sequence of 2-natural transformations:
Grp 2 ⟹ disc 1 , 2 ⟹ Grpd 2 . \operatorname{Grp}_2 \Longrightarrow\operatorname{disc}_{1,2} \Longrightarrow\operatorname{Grpd}_2. Grp 2 ⟹ disc 1 , 2 ⟹ Grpd 2 .
Now let C a t 2 , c c p ⊂ C a t 2 , c p \mathbf{Cat}_{2,ccp} \subset\mathbf{Cat}_{2,cp} Cat 2 , cc p ⊂ Cat 2 , c p denote the sub-2-category of 1-categories that additionally possess coequalizers, with functors preserving finite limits and coequalizers. Fixing C ∈ C a t 2 , c c p \mathbf{C} \in\mathbf{Cat}_{2,ccp} C ∈ Cat 2 , cc p , to every internal groupoid G ∈ Grpd 2 ( C ) \mathbf{G} \in\operatorname{Grpd}_2(\mathbf{C}) G ∈ Grpd 2 ( C ) we associate its quotient morphism coeq C ( G ) : G 0 ↠ G 0 / G \operatorname{coeq}_{\mathbf{C}}(\mathbf{G}) : G_0 \twoheadrightarrow G_0/G coeq C ( G ) : G 0 ↠ G 0 / G , obtained by taking the coequalizer of its source and target morphisms:
G 1 → s G 0 → coeq C ( G ) cod ( coeq C ( G ) ) ≃ G 0 / G t ↓ \begin{CD}
G_1 @>{s}>> G_0 @>{\operatorname{coeq}_{\mathbf{C}}(\mathbf{G})}>> \operatorname{cod}(\operatorname{coeq}_{\mathbf{C}}(\mathbf{G})) \simeq G_0/G \\
@V{t}VV
\end{CD} G 1 t ↓ ⏐ s G 0 coeq C ( G ) cod ( coeq C ( G )) ≃ G 0 / G
This construction extends 2-functorially to internal functors and internal natural transformations (which collapse to identities on the 1-categorical quotient in disc 1 , 2 ( C ) \operatorname{disc}_{1,2}(\mathbf{C}) disc 1 , 2 ( C ) ), defining a 2-functor coeq C : Grpd 2 ( C ) → Arrow 2 , c c p ( C ) \operatorname{coeq}_{\mathbf{C}} : \operatorname{Grpd}_2(\mathbf{C}) \to\operatorname{Arrow}_{2,ccp}(\mathbf{C}) coeq C : Grpd 2 ( C ) → Arrow 2 , cc p ( C ) , where Arrow 2 , c c p ( C ) = Arrow 2 ( C ) \operatorname{Arrow}_{2,ccp}(\mathbf{C}) = \operatorname{Arrow}_2(\mathbf{C}) Arrow 2 , cc p ( C ) = Arrow 2 ( C ) is the arrow category viewed as a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (here and below, the subscript c c p ccp cc p on Grpd 2 , c c p \operatorname{Grpd}_{2,ccp} Grpd 2 , cc p , disc 1 , 2 , c c p \operatorname{disc}_{1,2,ccp} disc 1 , 2 , cc p , Grpd 2 , c c p \operatorname{Grpd}_{2,ccp} Grpd 2 , cc p , Arrow 2 , c c p \operatorname{Arrow}_{2,ccp} Arrow 2 , cc p , and Eff 2 , c c p \operatorname{Eff}_{2,ccp} Eff 2 , cc p indicates the restriction of the corresponding 2-functors to the domain 2-category C a t 2 , c c p \mathbf{Cat}_{2,ccp} Cat 2 , cc p ). Varying C ∈ C a t 2 , c c p \mathbf{C} \in\mathbf{Cat}_{2,ccp} C ∈ Cat 2 , cc p yields a 2-natural transformation coeq \operatorname{coeq} coeq fitting into the (non-commutative) diagram below, where cod : Arrow 2 , c c p ⟹ disc 1 , 2 , c c p \operatorname{cod} : \operatorname{Arrow}_{2,ccp} \Longrightarrow\operatorname{disc}_{1,2,ccp} cod : Arrow 2 , cc p ⟹ disc 1 , 2 , cc p is the codomain projection transformation:
Grp 2 , c c p → disc 1 , 2 , c c p → Grpd 2 , c c p ↙ cod coeq ↓ Arrow 2 , c c p \begin{CD}
\operatorname{Grp}_{2,ccp} @>{}>> \operatorname{disc}_{1,2,ccp} @>{}>> \operatorname{Grpd}_{2,ccp} \\
& \mathrel{\swarrow}^{\operatorname{cod}} & @V{\operatorname{coeq}}VV \\
&& \operatorname{Arrow}_{2,ccp}
\end{CD} Grp 2 , cc p ↙ cod coeq ↓ ⏐ Arrow 2 , cc p disc 1 , 2 , cc p Grpd 2 , cc p
Because every coequalizer morphism is a regular (and hence effective) epimorphism, coeq factors objectwise through the full sub-2-category E f f 2 , c c p ( C ) ⊂ A r r o w 2 , c c p ( C ) \mathrm{Eff}_{2,ccp}(\mathcal{C}) \subset\mathrm{Arrow}_{2,ccp}(\mathcal{C}) Eff 2 , cc p ( C ) ⊂ Arrow 2 , cc p ( C ) of effective epimorphisms:
A r r o w 2 , c c p ← ⊂ E f f 2 , c c p cod ↓ ↓ coeq d i s c 1 , 2 , c c p ← coeq G r p d 2 , c c p ↑ ↑ G r p 2 , c c p → d i s c 1 , 2 , c c p \begin{CD}
\mathrm{Arrow}_{2,ccp} @<{\subset}<< \mathrm{Eff}_{2,ccp} \\
@V{\operatorname{cod}}VV @VV{\operatorname{coeq}}V \\
\mathrm{disc}_{1,2,ccp} @<<{\operatorname{coeq}}< \mathrm{Grpd}_{2,ccp} \\
@AAA @AAA \\
\mathrm{Grp}_{2,ccp} @>>> \mathrm{disc}_{1,2,ccp}
\end{CD} Arrow 2 , cc p cod ↓ ⏐ disc 1 , 2 , cc p ⏐ ↑ Grp 2 , cc p ⊂ coeq Eff 2 , cc p ↓ ⏐ coeq Grpd 2 , cc p ⏐ ↑ disc 1 , 2 , cc p
For an internal groupoid G ∈ G r p d 2 , c c p ( C ) \mathbf{G} \in\mathbf{Grpd}_{2,ccp}(\mathcal{C}) G ∈ Grpd 2 , cc p ( C ) , we call the codomain cod ( coeq ( G ) ) ≃ G 0 / G ∈ C \operatorname{cod}(\operatorname{coeq}(\mathbf{G})) \simeq G_0/G \in\mathcal{C} cod ( coeq ( G )) ≃ G 0 / G ∈ C its quotient object (or representing object ). Of fundamental importance are those internal groupoids that can be reconstructed from their quotient morphism—the effective internal groupoids :
Given C ∈ C a t 2 , c p \mathcal{C} \in\mathbf{Cat}_{2,cp} C ∈ Cat 2 , c p and any morphism f : U → X f : U \to X f : U → X in C \mathcal{C} C , we associate to f f f its Čech groupoid C ˇ ( f ) ∈ G r p d 2 ( C ) \check{\mathcal{C}}(f) \in\mathbf{Grpd}_2(\mathcal{C}) C ˇ ( f ) ∈ Grpd 2 ( C ) , defined by C ˇ ( f ) 0 = U \check{\mathcal{C}}(f)_0 = U C ˇ ( f ) 0 = U , C ˇ ( f ) 1 = U × X U = pb ( f , f ) \check{\mathcal{C}}(f)_1 = U \times_X U = \operatorname{pb}(f,f) C ˇ ( f ) 1 = U × X U = pb ( f , f ) , with source, target, and composition induced by the universal property of the pullback:
U × X U → f ^ 1 U f ^ 2 ↓ ↓ f U → f X U × X U → f ^ 1 U f ^ 2 ↓ \begin{CD}
U \times_X U @>{\widehat{f}_1}>> U \\
@V{\widehat{f}_2}VV @VV{f}V \\
U @>>{f}> X
\end{CD}
\qquad
\begin{CD}
U \times_X U @>{\widehat{f}_1}>> U \\
@V{\widehat{f}_2}VV
\end{CD} U × X U f 2 ↓ ⏐ U f 1 f U ↓ ⏐ f X U × X U f 2 ↓ ⏐ f 1 U
This assignment extends to a 2-functor C ˇ : A r r o w 2 , c p ( C ) → G r p d 2 ( C ) \check{\mathcal{C}} : \mathbf{Arrow}_{2,cp}(\mathcal{C}) \to\mathbf{Grpd}_2(\mathcal{C}) C ˇ : Arrow 2 , c p ( C ) → Grpd 2 ( C ) and a 2-natural transformation C ˇ : A r r o w 2 , c p ⇒ G r p d 2 \check{\mathcal{C}} : \mathbf{Arrow}_{2,cp} \Rightarrow\mathbf{Grpd}_2 C ˇ : Arrow 2 , c p ⇒ Grpd 2 . By Remark 2.5, a morphism f : U → X f : U \to X f : U → X is an effective epimorphism if and only if coeq ( C ˇ ( f ) ) ≅ f \operatorname{coeq}(\check{\mathcal{C}}(f)) \cong f coeq ( C ˇ ( f )) ≅ f . Consequently, the composite coeq ∘ C ˇ \operatorname{coeq} \circ\check{\mathcal{C}} coeq ∘ C ˇ is naturally isomorphic to the identity when restricted to E f f 2 , c c p \mathrm{Eff}_{2,ccp} Eff 2 , cc p :
G r p 2 , c c p ⟹ d i s c 1 , 2 , c c p ⟹ G r p d 2 , c c p ⇄ C ˇ coeq E f f 2 , c c p ⟹ A r r o w 2 , c c p \mathrm{Grp}_{2,ccp} \Longrightarrow\mathrm{disc}_{1,2,ccp} \Longrightarrow\mathbf{Grpd}_{2,ccp}
\quad
\mathop{\rightleftarrows}^{\operatorname{coeq}}_{\check{\mathcal{C}}}
\quad
\mathrm{Eff}_{2,ccp} \Longrightarrow\mathrm{Arrow}_{2,ccp} Grp 2 , cc p ⟹ disc 1 , 2 , cc p ⟹ Grpd 2 , cc p ⇄ C ˇ coeq Eff 2 , cc p ⟹ Arrow 2 , cc p
Conversely, the opposite composition C ˇ ∘ coeq \check{\mathcal{C}} \circ\operatorname{coeq} C ˇ ∘ coeq takes an internal groupoid G \mathbf{G} G , forms its quotient morphism G 0 ↠ cod ( coeq ( G ) ) G_0 \twoheadrightarrow\operatorname{cod}(\operatorname{coeq}(\mathbf{G})) G 0 ↠ cod ( coeq ( G )) , and takes the Čech groupoid of that morphism. An internal groupoid G \mathbf{G} G is called effective if the canonical comparison functor G → C ˇ ( coeq ( G ) ) \mathbf{G} \to\check{\mathcal{C}}(\operatorname{coeq}(\mathbf{G})) G → C ˇ ( coeq ( G )) is an equivalence in G r p d 2 ( C ) \mathbf{Grpd}_2(\mathcal{C}) Grpd 2 ( C ) .
The discussion above extends from 1-categories C a t 2 , c c p \mathbf{Cat}_{2,ccp} Cat 2 , cc p to ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories ( 2 , 1 ) C a t 2 , c c p (2,1)\mathbf{Cat}_{2,ccp} ( 2 , 1 ) Cat 2 , cc p , replacing 1-limits by ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits (homotopy pullbacks / iso-comma objects, even when H \mathbf{H} H is strictified to a G r p d \mathbf{Grpd} Grpd -enriched category as in Remark 2.2 (3)) and 1-coequalizers by ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits (codescent objects / homotopy coequalizers of groupoid nerves, which preserve non-trivial 2-cells in the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical arrow category A r r o w 2 , c c p ( H ) \mathbf{Arrow}_{2,ccp}(\mathbf{H}) Arrow 2 , cc p ( H ) rather than collapsing them to identities as in a 1-category C \mathcal{C} C ) [44 , 35 , 26 ] . In any such ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H ∈ ( 2 , 1 ) C a t 2 , c c p \mathbf{H} \in(2,1)\mathbf{Cat}_{2,ccp} H ∈ ( 2 , 1 ) Cat 2 , cc p , C ˇ \check{\mathcal{C}} C ˇ and coeq \operatorname{coeq} coeq restrict to an equivalence of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories between effective epimorphisms E f f 2 , c c p ( H ) \mathrm{Eff}_{2,ccp}(\mathbf{H}) Eff 2 , cc p ( H ) and the full sub-2-category G r p d 2 , c c p e f f ( H ) ⊂ G r p d 2 , c c p ( H ) \mathbf{Grpd}^{\mathrm{eff}}_{2,ccp}(\mathbf{H}) \subset\mathbf{Grpd}_{2,ccp}(\mathbf{H}) Grpd 2 , cc p eff ( H ) ⊂ Grpd 2 , cc p ( H ) of effective internal groupoids (whereas for a 1-category C \mathcal{C} C , E f f 2 , c c p ( C ) \mathrm{Eff}_{2,ccp}(\mathcal{C}) Eff 2 , cc p ( C ) is 0-truncated and thus equivalent to the 1-truncation τ ≤ 1 G r p d 2 , c c p e f f ( C ) \tau_{\leq1}\mathbf{Grpd}^{\mathrm{eff}}_{2,ccp}(\mathcal{C}) τ ≤ 1 Grpd 2 , cc p eff ( C ) ).
In a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H , mapping spaces H ( U , X ) \mathbf{H}(U,X) H ( U , X ) are 1-groupoids (i.e., 1-truncated spaces); an object Y ∈ H Y \in\mathbf{H} Y ∈ H is called 0-truncated (or discrete ) if H ( U , Y ) \mathbf{H}(U,Y) H ( U , Y ) is a 0-groupoid (a set) for all U ∈ H U \in\mathbf{H} U ∈ H , forming the full reflective 1-subcategory Disc ( H ) = τ ≤ 0 H ⊂ H \operatorname{Disc}(\mathbf{H}) = \tau_{\leq0}\mathbf{H} \subset\mathbf{H} Disc ( H ) = τ ≤ 0 H ⊂ H . Because the homotopy pullback U × X U U \times_X U U × X U of any morphism f : U → X f : U \to X f : U → X in a 0-truncated domain U U U has 0-truncated fibers (in particular, for any pointed object ∗ → X * \to X ∗ → X , the loop object Ω X = ∗ × X ∗ \Omega X = * \times_X * Ω X = ∗ × X ∗ is automatically 0-truncated), the Čech groupoid C ˇ ( f ) \check{\mathcal{C}}(f) C ˇ ( f ) of any morphism f : U → X f : U \to X f : U → X in H \mathbf{H} H has 0-truncated diagonal fibers ( s , t ) : C ˇ ( f ) 1 → U × U (s,t) : \check{\mathcal{C}}(f)_1 \to U \times U ( s , t ) : C ˇ ( f ) 1 → U × U . By the 2-truncation of the Giraud–Rezk–Lurie theorem [26 Theorem 6.1.0.6, Proposition 6.1.2.11] (see also [44 , 35 ] ), every Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H satisfies: (1) H is locally presentable as a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, and therefore possesses all small ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits; hence ( 2 , 1 ) S h v T o p o s 2 ⊂ ( 2 , 1 ) C a t 2 , c c p (2,1)\mathbf{ShvTopos}_{2} \subset(2,1)\mathbf{Cat}_{2,\mathrm{ccp}} ( 2 , 1 ) ShvTopos 2 ⊂ ( 2 , 1 ) Cat 2 , ccp on underlying categories;
(2) Every internal groupoid G = ( G 1 ⇉ G 0 ) ∈ G r p d 2 ( H ) \mathbf{G}=(G_{1}\rightrightarrows G_{0})\in\mathbf{Grpd}_{2}(\mathbf{H}) G = ( G 1 ⇉ G 0 ) ∈ Grpd 2 ( H ) with 0 0 0 -truncated morphism fibers ( s , t ) : G 1 → G 0 × G 0 (s,t):G_{1}\to G_{0}\times G_{0} ( s , t ) : G 1 → G 0 × G 0 —and in particular every internal groupoid in Disc ( H ) = τ ≤ 0 H \operatorname{Disc}(\mathbf{H})=\tau_{\leq0}\mathbf{H} Disc ( H ) = τ ≤ 0 H —is effective in H \mathbf{H} H :
G r p d 2 ( τ ≤ 0 H ) ⊂ G r p d 2 e f f ( H ) ≃ E f f 2 , c c p ( H ) , \mathbf{Grpd}_{2}(\tau_{\leq0}\mathbf{H})\subset\mathbf{Grpd}_{2}^{\mathrm{eff}}(\mathbf{H})\simeq\mathbf{Eff}_{2,\mathrm{ccp}}(\mathbf{H}), Grpd 2 ( τ ≤ 0 H ) ⊂ Grpd 2 eff ( H ) ≃ Eff 2 , ccp ( H ) ,
while an arbitrary internal groupoid G ∈ G r p d 2 ( H ) \mathbf{G}\in\mathbf{Grpd}_{2}(\mathbf{H}) G ∈ Grpd 2 ( H ) with 1 1 1 -truncated morphism fibers is effective as a 2 2 2 -stack in the ambient ( 3 , 1 ) (3,1) ( 3 , 1 ) -topos 2 S t k ( C , J ) ≃ τ ≤ 2 H ∞ 2\mathbf{Stk}(\mathcal{C},J)\simeq\tau_{\leq2}\mathbf{H}_{\infty} 2 Stk ( C , J ) ≃ τ ≤ 2 H ∞ (where H ∞ \mathbf{H}_{\infty} H ∞ is the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos of which H ≃ τ ≤ 1 H ∞ \mathbf{H}\simeq\tau_{\leq1}\mathbf{H}_{\infty} H ≃ τ ≤ 1 H ∞ is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation), with cod ( coeq H ( G ) ) ∈ H \operatorname{cod}(\operatorname{coeq}_{\mathbf{H}}(\mathbf{G}))\in\mathbf{H} cod ( coeq H ( G )) ∈ H computing its 1 1 1 -truncation.
Monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -Categories
Recall that a monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H equipped with a tensor 2 2 2 -bifunctor ⊗ : H × H → H \otimes:\mathbf{H}\times\mathbf{H}\to\mathbf{H} ⊗ : H × H → H , a unit object 1 : ∗ 2 → H 1:\ast_{2}\to\mathbf{H} 1 : ∗ 2 → H , and coherent associator and unitor 2 2 2 -natural equivalences satisfying the pentagon and triangle coherence conditions up to coherent 2 2 2 -isomorphisms [7 , 37 ] . Equivalently, this is the 2 2 2 -truncation of a monoidal ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -category in the sense of Lurie [27 ] . A monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category is braided or symmetric if it is further equipped with a coherent braiding X ⊗ Y ⟶ ∼ Y ⊗ X X\otimes Y\overset{\sim}{\longrightarrow}Y\otimes X X ⊗ Y ⟶ ∼ Y ⊗ X (satisfying the hexagon identities, and for symmetric, being involutive up to coherent 2 2 2 -isomorphism).
Let ( 2 , 1 ) M o n 2 ⊂ ( 2 , 1 ) C a t 2 (2,1)\mathbf{Mon}_{2}\subset(2,1)\mathbf{Cat}_{2} ( 2 , 1 ) Mon 2 ⊂ ( 2 , 1 ) Cat 2 denote the 2 2 2 -category of monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories, strong monoidal 2 2 2 -functors, and monoidal 2 2 2 -natural transformations, and let ( 2 , 1 ) S y m M o n 2 ⊂ ( 2 , 1 ) M o n 2 (2,1)\mathbf{SymMon}_{2}\subset(2,1)\mathbf{Mon}_{2} ( 2 , 1 ) SymMon 2 ⊂ ( 2 , 1 ) Mon 2 denote the sub-2 2 2 -category of symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories.
Example 2.6. For our purposes, the fundamental examples of symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories are the cartesian monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories, in which the tensor product ⊗ = × \otimes=\times ⊗ = × is given by the categorical ( 2 , 1 ) (2,1) ( 2 , 1 ) -product and the unit is the terminal object ∗ \ast ∗ . In particular, every Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H carries a canonical cartesian symmetric monoidal structure. More precisely, choosing products gives a canonical inclusion 2 2 2 -functor ( 2 , 1 ) C a t 2 , c ↪ ( 2 , 1 ) S y m M o n 2 (2,1)\mathbf{Cat}_{2,c}\hookrightarrow(2,1)\mathbf{SymMon}_{2} ( 2 , 1 ) Cat 2 , c ↪ ( 2 , 1 ) SymMon 2 .
In any monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category ( H , ⊗ , 1 ) (\mathbf{H},\otimes,1) ( H , ⊗ , 1 ) , we have the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category M o n 2 ( H ) \mathbf{Mon}_{2}(\mathbf{H}) Mon 2 ( H ) of monoid objects (or pseudomonoids [7 ] ) in H \mathbf{H} H . An object of M o n 2 ( H ) \mathbf{Mon}_{2}(\mathbf{H}) Mon 2 ( H ) is an object X ∈ H X\in\mathbf{H} X ∈ H equipped with multiplication μ : X ⊗ X → X \mu:X\otimes X\to X μ : X ⊗ X → X , unit e : 1 → X e:1\to X e : 1 → X , and associativity and unitality 2 2 2 -isomorphisms satisfying the Mac Lane pentagon and triangle equations. This assignment defines a 2 2 2 -functor M o n 2 : ( 2 , 1 ) M o n 2 → ( 2 , 1 ) C a t 2 \mathbf{Mon}_{2}:(2,1)\mathbf{Mon}_{2}\to(2,1)\mathbf{Cat}_{2} Mon 2 : ( 2 , 1 ) Mon 2 → ( 2 , 1 ) Cat 2 .
If ( H , ⊗ , 1 ) (\mathbf{H},\otimes,1) ( H , ⊗ , 1 ) is symmetric monoidal, we further obtain ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories B r M o n 2 ( H ) \mathbf{BrMon}_{2}(\mathbf{H}) BrMon 2 ( H ) and C o m M o n 2 ( H ) \mathbf{ComMon}_{2}(\mathbf{H}) ComMon 2 ( H ) of braided monoid objects and commutative (or symmetric) monoid objects in H \mathbf{H} H , linked by forgetful functors:
C o m M o n 2 ( H ) ⟶ B r M o n 2 ( H ) → f o r g e t M o n 2 ( H ) → f o r g e t H . \mathbf{ComMon}_{2}(\mathbf{H})\longrightarrow\mathbf{BrMon}_{2}(\mathbf{H})\xrightarrow{\mathrm{forget}}\mathbf{Mon}_{2}(\mathbf{H})\xrightarrow{\mathrm{forget}}\mathbf{H}. ComMon 2 ( H ) ⟶ BrMon 2 ( H ) forget Mon 2 ( H ) forget H .
Note that in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, being commutative is a property of a braided monoid (the braiding 2 2 2 -cell γ X , X : μ ∘ τ ⇒ μ \gamma_{X,X}:\mu\circ\tau\Rightarrow\mu γ X , X : μ ∘ τ ⇒ μ satisfies γ X , X = id \gamma_{X,X}=\operatorname{id} γ X , X = id ), whereas being braided on a monoid or being a monoid on an object involves extra data (1 1 1 -morphisms and 2 2 2 -cells), though we write hooked arrows by abuse of notation to indicate faithful/subcategory-like forgetful functors.
When H \mathbf{H} H is a cartesian monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, a monoid object G ∈ M o n 2 ( H ) G\in\mathbf{Mon}_{2}(\mathbf{H}) G ∈ Mon 2 ( H ) is a group object (or 2 2 2 -group / gr -object) if the shear morphism ( p 1 , μ ) : G × G → G × G (p_{1},\mu):G\times G\to G\times G ( p 1 , μ ) : G × G → G × G is an equivalence in H \mathbf{H} H (equivalently, if τ ≤ 0 H ( X , G ) \tau_{\leq0}\mathbf{H}(X,G) τ ≤ 0 H ( X , G ) is a group for every X ∈ H X\in\mathbf{H} X ∈ H ). Thus, the forgetful 2 2 2 -functor U : G r p 2 ( H ) ↪ M o n 2 ( H ) U:\mathbf{Grp}_{2}(\mathbf{H})\hookrightarrow\mathbf{Mon}_{2}(\mathbf{H}) U : Grp 2 ( H ) ↪ Mon 2 ( H ) is fully faithful. Defining the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories of braided group objects B r G r p 2 ( H ) \mathbf{BrGrp}_{2}(\mathbf{H}) BrGrp 2 ( H ) and commutative group objects C o m G r p 2 ( H ) \mathbf{ComGrp}_{2}(\mathbf{H}) ComGrp 2 ( H ) as the homotopy pullbacks in the diagram below, we obtain a commutative ladder of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories:
C o m G r p 2 ( H ) ⟶ B r G r p 2 ( H ) ⟶ G r p 2 ( H ) ⟶ H ⟶ G r p d 2 ( H ) \mathbf{ComGrp}_{2}(\mathbf{H})\longrightarrow\mathbf{BrGrp}_{2}(\mathbf{H})\longrightarrow\mathbf{Grp}_{2}(\mathbf{H})\longrightarrow\mathbf{H}\longrightarrow\mathbf{Grpd}_{2}(\mathbf{H}) ComGrp 2 ( H ) ⟶ BrGrp 2 ( H ) ⟶ Grp 2 ( H ) ⟶ H ⟶ Grpd 2 ( H )
↓ ↓ ↓ ∥ C o m M o n 2 ( H ) ⟶ B r M o n 2 ( H ) ⟶ M o n 2 ( H ) ⟶ H \begin{array}{ccccc}
\downarrow& & \downarrow& & \downarrow& & \Vert\\
\mathbf{ComMon}_{2}(\mathbf{H})\longrightarrow\mathbf{BrMon}_{2}(\mathbf{H})\longrightarrow\mathbf{Mon}_{2}(\mathbf{H})\longrightarrow\mathbf{H}
\end{array} ↓ ComMon 2 ( H ) ⟶ BrMon 2 ( H ) ⟶ Mon 2 ( H ) ⟶ H ↓ ↓ ∥
This construction is 2-functorial in H ∈ ( 2 , 1 ) C a t 2 , c p \mathbf{H} \in(2,1)\mathrm{Cat}_{2,\mathrm{cp}} H ∈ ( 2 , 1 ) Cat 2 , cp , extending to a diagram of 2-natural transformations:
C o m G r p 2 → B r G r p 2 → G r p 2 → i d → G r p d 2 ↓ ↓ ↓ ∥ C o m M o n 2 → B r M o n 2 → M o n 2 → i d \begin{CD}
\mathrm{ComGrp}_2 @>>> \mathrm{BrGrp}_2 @>>> \mathrm{Grp}_2 @>>> \mathrm{id} @>>> \mathbf{Grpd}_2 \\
@VVV @VVV @VVV @| \\
\mathrm{ComMon}_2 @>>> \mathrm{BrMon}_2 @>>> \mathrm{Mon}_2 @>>> \mathrm{id}
\end{CD} ComGrp 2 ↓ ⏐ ComMon 2 BrGrp 2 ↓ ⏐ BrMon 2 Grp 2 ↓ ⏐ Mon 2 id id Grpd 2
Algebras Over (2,1)-Operads
In a symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H , monoids, braided monoids, and commutative monoids are special cases of a unified notion: algebras over a ( 2 , 1 ) (2,1) ( 2 , 1 ) -operad [27 ] . Let ( 2 , 1 ) O p e r 2 (2,1)\mathrm{Oper}_2 ( 2 , 1 ) Oper 2 denote the 2-category of ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads (i.e., 2-truncated ∞ \infty ∞ -operads, or G r p d \mathbf{Grpd} Grpd -enriched operads), and let ( 2 , 1 ) S y m O p e r 2 ⊂ ( 2 , 1 ) O p e r 2 (2,1)\mathrm{SymOper}_2 \subset(2,1)\mathrm{Oper}_2 ( 2 , 1 ) SymOper 2 ⊂ ( 2 , 1 ) Oper 2 be the sub-2-category of symmetric ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads, which is equipped with a symmetric monoidal structure ⊙ \odot ⊙ given by the 2-truncated Boardman–Vogt tensor product [4 , 27 ] .
The terminal object of ( 2 , 1 ) S y m O p e r 2 (2,1)\mathrm{SymOper}_2 ( 2 , 1 ) SymOper 2 is the commutative operad E ∞ \mathbb{E}_{\infty} E ∞ , which is filtered by the sequence of little k k k -cubes ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads E k \mathbb{E}_k E k (for k ≥ 0 k \ge0 k ≥ 0 ), obtained by 2-truncating the topological little k k k -cubes operads:
E 0 ⟶ E 1 ⟶ E 2 ⟶ ⋯ ⟶ E k − 1 ⟶ E k ⟶ E k + 1 ⟶ ⋯ ⟶ E ∞ = lim k E k . (3) \mathbb{E}_0 \longrightarrow\mathbb{E}_1 \longrightarrow\mathbb{E}_2 \longrightarrow\cdots\longrightarrow\mathbb{E}_{k-1} \longrightarrow\mathbb{E}_k \longrightarrow\mathbb{E}_{k+1} \longrightarrow\cdots\longrightarrow\mathbb{E}_{\infty} = \lim_{k} \mathbb{E}_k. \tag*{(3)} E 0 ⟶ E 1 ⟶ E 2 ⟶ ⋯ ⟶ E k − 1 ⟶ E k ⟶ E k + 1 ⟶ ⋯ ⟶ E ∞ = k lim E k . (3)
A foundational property of the operads E k \mathbb{E}_k E k is Dunn–Lurie additivity:
Theorem 2.7 (Dunn–Lurie Additivity [8 , 27 ] ). For every k , l ≥ 0 k,l \ge0 k , l ≥ 0 , the Boardman–Vogt tensor product of E k \mathbb{E}_k E k and E l \mathbb{E}_l E l satisfies
E k ⊙ E l ≃ E k + l . \mathbb{E}_k \odot\mathbb{E}_l \simeq\mathbb{E}_{k+l}. E k ⊙ E l ≃ E k + l .
Proof. This is the direct 2-truncation of Lurie’s ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categorical generalization of Dunn’s additivity theorem; see [27 Theorem 5.1.2.2] . □
For any ( 2 , 1 ) (2,1) ( 2 , 1 ) -operad O \mathcal{O} O and any symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H , we have the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category O A l g 2 ( H ) \mathcal{O}\mathrm{Alg}_2(\mathbf{H}) O Alg 2 ( H ) of O \mathcal{O} O -algebra objects in H \mathbf{H} H . Equivalently, associating to H \mathbf{H} H its endomorphism ( 2 , 1 ) (2,1) ( 2 , 1 ) -operad E ( H ) \mathcal{E}(\mathbf{H}) E ( H ) , O A l g 2 ( H ) \mathcal{O}\mathrm{Alg}_2(\mathbf{H}) O Alg 2 ( H ) is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category of ( 2 , 1 ) (2,1) ( 2 , 1 ) -operad morphisms O → E ( H ) \mathcal{O} \to\mathcal{E}(\mathbf{H}) O → E ( H ) . Every morphism of ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads ϕ : O → O ′ \phi:\mathcal{O}\to\mathcal{O}' ϕ : O → O ′ induces a forgetful 2-functor ϕ ∗ : O ′ A l g 2 ( H ) → O A l g 2 ( H ) \phi^*:\mathcal{O}'\mathrm{Alg}_2(\mathbf{H})\to\mathcal{O}\mathrm{Alg}_2(\mathbf{H}) ϕ ∗ : O ′ Alg 2 ( H ) → O Alg 2 ( H ) , which admits a left 2-adjoint ϕ ∗ \phi_* ϕ ∗ whenever H \mathbf{H} H is presentable and the tensor product preserves colimits in each variable:
O A l g 2 ( H ) ← ϕ ∗ O ′ A l g 2 ( H ) ϕ ∗ ↑ ↓ ϕ ∗ \begin{CD}
\mathcal{O}\mathrm{Alg}_2(\mathbf{H}) @<{\phi_*}<< \mathcal{O}'\mathrm{Alg}_2(\mathbf{H}) \\
@A{\phi^*}AA @VV{\phi^*}V
\end{CD} O Alg 2 ( H ) ϕ ∗ ⏐ ↑ ϕ ∗ O ′ Alg 2 ( H ) ↓ ⏐ ϕ ∗
Example 2.8. Recall that E 0 \mathbb{E}_0 E 0 -algebras in H \mathbf{H} H are pointed objects 1 → X \mathbb{1}\to X 1 → X in H \mathbf{H} H (or, forgetting the basepoint as in our convention for D e l 0 2 ( H ) \mathrm{Del}_0^2(\mathbf{H}) Del 0 2 ( H ) below, objects of H \mathbf{H} H ), E 1 \mathbb{E}_1 E 1 -algebras are associative monoid objects, E 2 \mathbb{E}_2 E 2 -algebras are braided monoid objects, and E 3 \mathbb{E}_3 E 3 -algebras in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category are symmetric/commutative monoid objects [17 , 1 , 27 ] :
E 0 A l g 2 u n p t ( H ) ≃ H (or E 0 A l g 2 ( H ) ≃ 1 / H ) , \mathbb{E}_0\mathrm{Alg}_2^{\mathrm{unpt}}(\mathbf{H}) \simeq\mathbf{H}
\qquad\text{(or }\mathbb{E}_0\mathrm{Alg}_2(\mathbf{H}) \simeq\mathbb{1}/\mathbf{H}\text{)}, E 0 Alg 2 unpt ( H ) ≃ H (or E 0 Alg 2 ( H ) ≃ 1/ H ) ,
E 1 A l g 2 ( H ) ≃ M o n 2 ( H ) , \mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) \simeq\mathrm{Mon}_2(\mathbf{H}), E 1 Alg 2 ( H ) ≃ Mon 2 ( H ) ,
E 2 A l g 2 ( H ) ≃ B r M o n 2 ( H ) , \mathbb{E}_2\mathrm{Alg}_2(\mathbf{H}) \simeq\mathrm{BrMon}_2(\mathbf{H}), E 2 Alg 2 ( H ) ≃ BrMon 2 ( H ) ,
E 3 A l g 2 ( H ) ≃ C o m M o n 2 ( H ) . \mathbb{E}_3\mathrm{Alg}_2(\mathbf{H}) \simeq\mathrm{ComMon}_2(\mathbf{H}). E 3 Alg 2 ( H ) ≃ ComMon 2 ( H ) .
Hence, the filtration (3) induces a commutative ladder of forgetful 2-functors:
E 3 A l g 2 ( H ) → E 2 A l g 2 ( H ) → E 1 A l g 2 ( H ) → H ≃ ↓ ≃ ↓ ≃ ↓ ∥ C o m M o n 2 ( H ) → B r M o n 2 ( H ) → M o n 2 ( H ) → H (4) \begin{CD}
\mathbb{E}_3\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_2\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbf{H} \\
@V{\simeq}VV @V{\simeq}VV @V{\simeq}VV @| \\
\mathrm{ComMon}_2(\mathbf{H}) @>>> \mathrm{BrMon}_2(\mathbf{H}) @>>> \mathrm{Mon}_2(\mathbf{H}) @>>> \mathbf{H} \tag*{(4)}
\end{CD} E 3 Alg 2 ( H ) ≃ ↓ ⏐ ComMon 2 ( H ) E 2 Alg 2 ( H ) ≃ ↓ ⏐ BrMon 2 ( H ) E 1 Alg 2 ( H ) ≃ ↓ ⏐ Mon 2 ( H ) H H (4)
Now suppose H \mathbf{H} H is a cartesian monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category. Given a ( 2 , 1 ) (2,1) ( 2 , 1 ) -operad O \mathcal{O} O equipped with a morphism ι : E 1 → O \iota:\mathbb{E}_1 \to\mathcal{O} ι : E 1 → O , we define the full sub-2 2 2 -category O A l g 2 , ι g r p ( H ) ⊂ O A l g 2 ( H ) \mathcal{O}\mathrm{Alg}^{\mathrm{grp}}_{2,\iota}(\mathbf{H}) \subset\mathcal{O}\mathrm{Alg}_2(\mathbf{H}) O Alg 2 , ι grp ( H ) ⊂ O Alg 2 ( H ) of groupal (or grouplike ) O \mathcal{O} O -algebra objects relative to ι \iota ι as the homotopy pullback of G r p 2 ( H ) ↪ M o n 2 ( H ) \mathrm{Grp}_2(\mathbf{H}) \hookrightarrow\mathrm{Mon}_2(\mathbf{H}) Grp 2 ( H ) ↪ Mon 2 ( H ) along ι ∗ : O A l g 2 ( H ) → E 1 A l g 2 ( H ) ≃ M o n 2 ( H ) \iota^*:\mathcal{O}\mathrm{Alg}_2(\mathbf{H}) \to\mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) \simeq\mathrm{Mon}_2(\mathbf{H}) ι ∗ : O Alg 2 ( H ) → E 1 Alg 2 ( H ) ≃ Mon 2 ( H ) :
O A l g 2 ( H ) → ι ∗ E 1 A l g 2 ( H ) → H ↑ ≃ ↓ ↓ ∥ O A l g 2 , ι g r p ( H ) → ι ∗ G r p 2 ( H ) → H ↑ ↑ M o n 2 ( H ) → H \begin{CD}
\mathcal{O}\mathrm{Alg}_2(\mathbf{H}) @>{\iota^*}>> \mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbf{H}\\
@AAA @V{\simeq}VV @VV{\Vert}V\\
\mathcal{O}\mathrm{Alg}^{\mathrm{grp}}_{2,\iota}(\mathbf{H}) @>{\iota^*}>> \mathrm{Grp}_2(\mathbf{H}) @>>> \mathbf{H}\\
@. @AAA @AAA\\
@. \mathrm{Mon}_2(\mathbf{H}) @>>> \mathbf{H}
\end{CD} O Alg 2 ( H ) ⏐ ↑ O Alg 2 , ι grp ( H ) ι ∗ ι ∗ E 1 Alg 2 ( H ) ≃ ↓ ⏐ Grp 2 ( H ) ⏐ ↑ Mon 2 ( H ) H ↓ ⏐ ∥ H ⏐ ↑ H
By direct inspection, the groupal E k \mathbb{E}_k E k -algebras for k = 1 , 2 , 3 k=1,2,3 k = 1 , 2 , 3 (relative to the canonical inclusion E 1 → E k \mathbb{E}_1 \to\mathbb{E}_k E 1 → E k ) are precisely group objects, braided group objects, and commutative group objects in H \mathbf{H} H , extending diagram (4) to:
E 3 A l g 2 g r p ( H ) → E 2 A l g 2 g r p ( H ) → E 1 A l g 2 g r p ( H ) ≃ ↓ ≃ ↓ ≃ ↓ C o m G r p 2 ( H ) → B r G r p 2 ( H ) → G r p 2 ( H ) ↓ ↓ ↓ C o m M o n 2 ( H ) → B r M o n 2 ( H ) → M o n 2 ( H ) → H ≃ ↑ ≃ ↑ ≃ ↑ ↓ ∥ E 3 A l g 2 ( H ) → E 2 A l g 2 ( H ) → E 1 A l g 2 ( H ) → H \begin{CD}
\mathbb{E}_3\mathrm{Alg}^{\mathrm{grp}}_2(\mathbf{H}) @>>> \mathbb{E}_2\mathrm{Alg}^{\mathrm{grp}}_2(\mathbf{H}) @>>> \mathbb{E}_1\mathrm{Alg}^{\mathrm{grp}}_2(\mathbf{H})\\
@V{\simeq}VV @V{\simeq}VV @V{\simeq}VV\\
\mathrm{ComGrp}_2(\mathbf{H}) @>>> \mathrm{BrGrp}_2(\mathbf{H}) @>>> \mathrm{Grp}_2(\mathbf{H})\\
@VVV @VVV @VVV\\
\mathrm{ComMon}_2(\mathbf{H}) @>>> \mathrm{BrMon}_2(\mathbf{H}) @>>> \mathrm{Mon}_2(\mathbf{H}) @>>> \mathbf{H}\\
@A{\simeq}AA @A{\simeq}AA @A{\simeq}AA @VV{\Vert}V\\
\mathbb{E}_3\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_2\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbf{H}
\end{CD} E 3 Alg 2 grp ( H ) ≃ ↓ ⏐ ComGrp 2 ( H ) ↓ ⏐ ComMon 2 ( H ) ≃ ⏐ ↑ E 3 Alg 2 ( H ) E 2 Alg 2 grp ( H ) ≃ ↓ ⏐ BrGrp 2 ( H ) ↓ ⏐ BrMon 2 ( H ) ≃ ⏐ ↑ E 2 Alg 2 ( H ) E 1 Alg 2 grp ( H ) ≃ ↓ ⏐ Grp 2 ( H ) ↓ ⏐ Mon 2 ( H ) ≃ ⏐ ↑ E 1 Alg 2 ( H ) H ↓ ⏐ ∥ H
Finally, although the sequence (3) of ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads continues beyond E 3 \mathbb{E}_3 E 3 , the Baez–Dolan Stabilization Theorem guarantees that in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, E k \mathbb{E}_k E k -algebras stabilize at k = 3 k=3 k = 3 :
Theorem 2.9 (Stabilization of E k \mathbb{E}_k E k -Algebras in ( 2 , 1 ) (2,1) ( 2 , 1 ) -Categories [1 , 27 ] ). For every symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H and every k ≥ 3 k \ge3 k ≥ 3 , the canonical forgetful functor E k A l g 2 ( H ) → E 3 A l g 2 ( H ) ≃ C o m M o n 2 ( H ) \mathbb{E}_k\mathrm{Alg}_2(\mathbf{H}) \to\mathbb{E}_3\mathrm{Alg}_2(\mathbf{H}) \simeq\mathrm{ComMon}_2(\mathbf{H}) E k Alg 2 ( H ) → E 3 Alg 2 ( H ) ≃ ComMon 2 ( H ) is an equivalence of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories:
\begin{CD}
\mathbb{E}_k\mathrm{Alg}_2(\mathbf{H}) @>>> \cdots@>>> \mathbb{E}_3\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_2\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbb{E}_1\mathrm{Alg}_2(\mathbf{H}) @>>> \mathbf{H}\\
@. @. @V{\simeq}VV @V{\simeq}VV @V{\simeq}VV @VV{\Vert}V\\
@. @. \mathrm{ComMon}_2(\mathbf{H}) @>>> \mathrm{BrMon}_2(\mathbf{H}) @>>> \mathrm{Mon}_2(\mathbf{H}) @>>> \mathbf{H}
\end{CD}
Proof. Since H \mathbf{H} H is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, its hom-spaces are 1 1 1 -groupoids (i.e., 1 1 1 -truncated spaces). By [27 Corollary 5.1.1.7] , in a symmetric monoidal ∞ \infty ∞ -category whose mapping spaces are ( n − 1 ) (n-1) ( n − 1 ) -truncated, the forgetful functor E k A l g ( H ) → E ∞ A l g ( H ) \mathbb{E}_k\mathrm{Alg}(\mathbf{H}) \to\mathbb{E}_\infty\mathrm{Alg}(\mathbf{H}) E k Alg ( H ) → E ∞ Alg ( H ) is an equivalence for all k ≥ n + 1 k \ge n+1 k ≥ n + 1 . Setting n = 2 n=2 n = 2 yields stabilization at k = 3 k=3 k = 3 . □ \square □
Remark 2.10. Of course, Theorem 2.9 does not imply that the only algebraic structures in a symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category are monoids, braided monoids, and commutative monoids (one may consider algebras over arbitrary operads or Lawvere theories, such as Lie algebras, Poisson algebras, or racks). Rather, it asserts that within the E k \mathbb{E}_k E k -hierarchy of weakening commutativity, there is no further weakening beyond k = 3 k=3 k = 3 in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category. In general, in a symmetric monoidal ( n , 1 ) (n,1) ( n , 1 ) -category, the sequence of E k \mathbb{E}_k E k -algebras stabilizes for k ≥ n + 1 k \ge n+1 k ≥ n + 1 .
Delooping
In this section we examine delooping groupoids and delooping objects in ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi. In our broader program, deloopings are required to formulate an internal and parameterized definition of principal G G G -bundles and nonabelian 1-cohomology. As we demonstrate in § 3.1 –§ 3.3 , the delooping construction can be realized either internally (via internal groupoid objects in a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos) or externally (by transporting the canonical delooping along Yoneda 2-embeddings on presheaf categories and stackifying), producing 2-naturally equivalent internal groupoids. The internal formulation has the advantage of making sense in any cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos even when it is not a ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, whereas the external formulation directly bridges point-set/site-level algebraic symmetries with sheafified stacks and serves as the blueprint for plain and genuine delooping systems in algebra-valued cosmoi. We then study the representing delooping objects B H G B_HG B H G , higher iterated deloopings B H k G B_H^kG B H k G , their characterization via the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical Eckmann–Hilton argument and E k E_k E k -algebras, and the looping–delooping equivalence on connected pointed objects.
Internal Delooping
Let ( 2 , 1 ) C a r t 2 ⊂ ( 2 , 1 ) C a t 2 , c p ⊂ ( 2 , 1 ) C a t 2 (2,1)\mathbf{Cart}_2 \subset(2,1)\mathbf{Cat}_{2,cp} \subset(2,1)\mathbf{Cat}_2 ( 2 , 1 ) Cart 2 ⊂ ( 2 , 1 ) Cat 2 , c p ⊂ ( 2 , 1 ) Cat 2 denote the 2-category of cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmoi, defined as the sub-2-category of complete and cocomplete ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories with 2-functors preserving finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits, and let ( 2 , 1 ) C a r t 2 , c o e q ⊂ ( 2 , 1 ) C a t 2 , c c p (2,1)\mathbf{Cart}_{2,coeq} \subset(2,1)\mathbf{Cat}_{2,ccp} ( 2 , 1 ) Cart 2 , coe q ⊂ ( 2 , 1 ) Cat 2 , cc p denote the subcategory whose 2-functors additionally preserve ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits (homotopy coequalizers). Let
G r p 2 , G r p d 2 : ( 2 , 1 ) C a r t 2 ⟶ ( 2 , 1 ) C a t 2 \mathbf{Grp}_2,\mathbf{Grpd}_2 : (2,1)\mathbf{Cart}_2 \longrightarrow(2,1)\mathbf{Cat}_2 Grp 2 , Grpd 2 : ( 2 , 1 ) Cart 2 ⟶ ( 2 , 1 ) Cat 2
denote the restrictions to ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 of the internal group and internal groupoid 2-functors from § 2.3 .
The internal delooping is the 2-natural transformation B i n t : G r p 2 ⇒ G r p d 2 \mathbf{B}^{\mathrm{int}} : \mathbf{Grp}_2 \Rightarrow\mathbf{Grpd}_2 B int : Grp 2 ⇒ Grpd 2 which assigns to every H ∈ ( 2 , 1 ) C a r t 2 \mathbf{H} \in(2,1)\mathbf{Cart}_2 H ∈ ( 2 , 1 ) Cart 2 and every group object G ∈ G r p 2 ( H ) G \in\mathbf{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) the one-object internal groupoid B H i n t ( G ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G) \in\mathbf{Grpd}_2(\mathbf{H}) B H int ( G ) ∈ Grpd 2 ( H ) defined by
B H i n t ( G ) 0 = ∗ H , B H i n t ( G ) 1 = G , \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)_0 = *_\mathbf{H}, \qquad\mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)_1 = G, B H int ( G ) 0 = ∗ H , B H int ( G ) 1 = G ,
where ∗ H *_\mathbf{H} ∗ H is the terminal object of H \mathbf{H} H , the source and target morphisms s , t : G → ∗ H s,t : G \to*_\mathbf{H} s , t : G → ∗ H are the unique terminal morphisms, and composition ∘ : G × ∗ H G ≃ G × G → m G \circ: G \times_{*_\mathbf{H}} G \simeq G \times G \xrightarrow{m} G ∘ : G × ∗ H G ≃ G × G m G , unit e : ∗ H → G e : *_\mathbf{H} \to G e : ∗ H → G , and inversion inv : G → G \operatorname{inv} : G \to G inv : G → G are given by the group structure of G G G :
( 2 , 1 ) C a t 2 G r p 2 ⟹ B i n t ( 2 , 1 ) C a r t 2 G r p d 2 \begin{gathered}
(2,1)\mathbf{Cat}_2\\
\mathbf{Grp}_2 \underset{(2,1)\mathbf{Cart}_2}{\overset{\mathbf{B}^{\mathrm{int}}}{\Longrightarrow}} \mathbf{Grpd}_2
\end{gathered} ( 2 , 1 ) Cat 2 Grp 2 ( 2 , 1 ) Cart 2 ⟹ B int Grpd 2
Notice that an arbitrary ( 2 , 1 ) (2,1) ( 2 , 1 ) -pretopos need not belong to ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 , as it may lack arbitrary small ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits or ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits. However, every ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and every Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos is locally presentable [26 ] , and therefore is complete and cocomplete; moreover, direct image functors f ∗ f_* f ∗ of geometric morphisms are right 2-adjoints and thus preserve all ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits (so U t o p o s ( f ) = f ∗ U_{\mathrm{topos}}(f)=f_* U topos ( f ) = f ∗ lands in ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 , which suffices for G r p 2 \mathbf{Grp}_2 Grp 2 , G r p d 2 \mathbf{Grpd}_2 Grpd 2 , and B i n t \mathbf{B}^{\mathrm{int}} B int ), while inverse image functors f ∗ f^* f ∗ preserve both finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and all ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits (so U t o p o s ∗ ( f ) = f ∗ U^*_{\mathrm{topos}}(f)=f^* U topos ∗ ( f ) = f ∗ lands in ( 2 , 1 ) C a r t 2 , c o e q ⊂ ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_{2,coeq} \subset(2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 , coe q ⊂ ( 2 , 1 ) Cart 2 , which additionally commutes with taking quotient delooping objects B H = cod ∘ coeq ∘ B H i n t B_\mathbf{H}=\operatorname{cod}\circ\operatorname{coeq}\circ\mathbf{B}^{\mathrm{int}}_\mathbf{H} B H = cod ∘ coeq ∘ B H int and stackification localizations L J L_J L J ). Consequently, both forgetful 2-functors U t o p o s : ( 2 , 1 ) P T o p o s 2 → ( 2 , 1 ) C a t 2 U_{\mathrm{topos}} : (2,1)\mathbf{PTopos}_2 \to(2,1)\mathbf{Cat}_2 U topos : ( 2 , 1 ) PTopos 2 → ( 2 , 1 ) Cat 2 and U t o p o s ∗ : ( 2 , 1 ) P T o p o s 2 2 → ( 2 , 1 ) C a t 2 U^*_{\mathrm{topos}} : (2,1)\mathbf{PTopos}^{2}_{2} \to(2,1)\mathbf{Cat}_2 U topos ∗ : ( 2 , 1 ) PTopos 2 2 → ( 2 , 1 ) Cat 2 from Remark 2.4 factor on ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathbf{PShvTopos}_2 ( 2 , 1 ) PShvTopos 2 and ( 2 , 1 ) S h v T o p o s 2 (2,1)\mathbf{ShvTopos}_2 ( 2 , 1 ) ShvTopos 2 through ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 (via U P S h v U_{\mathbf{PShv}} U PShv , U S h v U_{\mathbf{Shv}} U Shv and U P S h v ∗ U^{*}_{\mathbf{PShv}} U PShv ∗ , U S h v ∗ U^{*}_{\mathbf{Shv}} U Shv ∗ , respectively):4 ^{4} 4
( 2 , 1 ) S h v T o p o s 2 o p → ( 2 , 1 ) P S h v T o p o s 2 o p → ( 2 , 1 ) C a r t 2 ↓ ↓ ( 2 , 1 ) P T o p o s 2 o p → ( 2 , 1 ) C a t 2 \begin{CD}
(2,1)\mathbf{ShvTopos}_2^{\mathrm{op}} @>>> (2,1)\mathbf{PShvTopos}_2^{\mathrm{op}} @>>> (2,1)\mathbf{Cart}_2 \\
@. @VVV @VVV \\
@. (2,1)\mathbf{PTopos}_2^{\mathrm{op}} @>>> (2,1)\mathbf{Cat}_2
\end{CD} ( 2 , 1 ) ShvTopos 2 op ( 2 , 1 ) PShvTopos 2 op ↓ ⏐ ( 2 , 1 ) PTopos 2 op ( 2 , 1 ) Cart 2 ↓ ⏐ ( 2 , 1 ) Cat 2
Whiskering B i n t \mathbf{B}^{\mathrm{int}} B int along U P S h v ∗ U^{*}_{\mathbf{PShv}} U PShv ∗ and U S h v ∗ U^{*}_{\mathbf{Shv}} U Shv ∗ (or covariantly along U P S h v U_{\mathbf{PShv}} U PShv and U S h v U_{\mathbf{Shv}} U Shv ) immediately defines the internal delooping transformations B P S h v i n t \mathbf{B}^{\mathrm{int}}_{\mathbf{PShv}} B PShv int and B S h v i n t \mathbf{B}^{\mathrm{int}}_{\mathbf{Shv}} B Shv int for ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf topoi and Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi:
( 2 , 1 ) S h v T o p o s 2 o p → ( 2 , 1 ) P S h v T o p o s 2 o p → ( 2 , 1 ) C a r t 2 ↓ ↓ ↓ C a t 2 → C a t 2 → C a t 2 (5) \begin{CD}
(2,1)\mathbf{ShvTopos}_2^{\mathrm{op}} @>>> (2,1)\mathbf{PShvTopos}_2^{\mathrm{op}} @>>> (2,1)\mathbf{Cart}_2 \\
@VVV @VVV @VVV \\
\mathbf{Cat}_2 @>>> \mathbf{Cat}_2 @>>> \mathbf{Cat}_2 \tag*{(5)}
\end{CD} ( 2 , 1 ) ShvTopos 2 op ↓ ⏐ Cat 2 ( 2 , 1 ) PShvTopos 2 op ↓ ⏐ Cat 2 ( 2 , 1 ) Cart 2 ↓ ⏐ Cat 2 (5)
External Delooping
The construction in Section 3.1 3.1 3.1 is internal in the sense that it operates directly on group objects inside H \mathbf{H} H without reference to a presenting site ( C , J ) (\mathbf{C},\mathcal{J}) ( C , J ) . We now construct an external presentation of delooping by acting on presheaves of groups and groupoids via the 2 2 2 -categorical Yoneda embedding.
Consider the two-variable presheaf 2 2 2 -functor ( 2 , 1 ) P S h v ( − ) ( − ) : C a t 2 o p × ( 2 , 1 ) C a t 2 ⟶ ( 2 , 1 ) C a t 2 (2,1)\mathbf{PShv}_{(-)}(-) : \mathbf{Cat}_2^{\mathrm{op}} \times(2,1)\mathbf{Cat}_2 \longrightarrow(2,1)\mathbf{Cat}_2 ( 2 , 1 ) PShv ( − ) ( − ) : Cat 2 op × ( 2 , 1 ) Cat 2 ⟶ ( 2 , 1 ) Cat 2 , ( C , A ) ⟼ ( 2 , 1 ) P S h v A ( C ) (\mathbf{C},\mathbf{A}) \longmapsto(2,1)\mathbf{PShv}_{\mathbf{A}}(\mathbf{C}) ( C , A ) ⟼ ( 2 , 1 ) PShv A ( C ) , together with its covariant realization ( 2 , 1 ) P S h v ( − ) c o v : C a t 2 , c p × ( 2 , 1 ) C a r t 2 , f i l t ⟶ ( 2 , 1 ) C a r t 2 (2,1)\mathbf{PShv}^{\mathrm{cov}}_{(-)} : \mathbf{Cat}_{2,\mathrm{cp}} \times(2,1)\mathbf{Cart}_{2,\mathrm{filt}} \longrightarrow(2,1)\mathbf{Cart}_2 ( 2 , 1 ) PShv ( − ) cov : Cat 2 , cp × ( 2 , 1 ) Cart 2 , filt ⟶ ( 2 , 1 ) Cart 2 acting on 1 1 1 -categories with finite 1 1 1 -limits C ∈ C a t 2 , c p \mathbf{C} \in\mathbf{Cat}_{2,\mathrm{cp}} C ∈ Cat 2 , cp via left-exact left Kan extension f ! = Lan f o p f_{!}=\operatorname{Lan}_{f^{\mathrm{op}}} f ! = Lan f op (where ( 2 , 1 ) C a r t 2 , f i l t ⊂ ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_{2,\mathrm{filt}} \subset(2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 , filt ⊂ ( 2 , 1 ) Cart 2 denotes cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmoi—such as G r p d 2 \mathbf{Grpd}_2 Grpd 2 , S e t 2 \mathbf{Set}_2 Set 2 , and all Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi—in which filtered ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits commute with finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits; indeed, for any finite-limit-preserving functor f : C → D f:\mathbf{C}\to\mathbf{D} f : C → D in C a t 2 , c p \mathbf{Cat}_{2,\mathrm{cp}} Cat 2 , cp and d ∈ D d\in\mathbf{D} d ∈ D , the comma category ( d ↓ f ) (d\mathbin{\downarrow}f) ( d ↓ f ) has all finite limits and is therefore cofiltered, so the indexing category ( f o p ↓ d ) ≃ ( d ↓ f ) o p (f^{\mathrm{op}}\mathbin{\downarrow}d)\simeq(d\mathbin{\downarrow}f)^{\mathrm{op}} ( f op ↓ d ) ≃ ( d ↓ f ) op for the pointwise ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimit ( Lan f o p F ) ( d ) (\operatorname{Lan}_{f^{\mathrm{op}}}F)(d) ( Lan f op F ) ( d ) is filtered, ensuring that f ! = Lan f o p f_{!}=\operatorname{Lan}_{f^{\mathrm{op}}} f ! = Lan f op preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and hence internal groupoids). Recall the 2 2 2 -categorical Yoneda embedding [43 ] , which assigns to every small ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (and in particular every C ∈ C a t 2 , c p \mathbf{C}\in\mathbf{Cat}_{2,\mathrm{cp}} C ∈ Cat 2 , cp ) the fully faithful 2 2 2 -functor
y C : C 2 ↪ ( 2 , 1 ) P S h v G r p d 2 ( C ) , X ⟼ C 2 ( − , X ) . y_{\mathbf{C}}:\mathbf{C}_2\hookrightarrow(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{C}),\qquad X\longmapsto\mathbf{C}_2(-,X). y C : C 2 ↪ ( 2 , 1 ) PShv Grpd 2 ( C ) , X ⟼ C 2 ( − , X ) .
Observe that G r p d 2 ≃ G r p d 2 ( S e t 2 ) \mathbf{Grpd}_2\simeq\mathbf{Grpd}_2(\mathbf{Set}_2) Grpd 2 ≃ Grpd 2 ( Set 2 ) , where S e t 2 = disc 1 , 2 ( S e t ) ∈ ( 2 , 1 ) C a r t 2 \mathbf{Set}_2=\operatorname{disc}_{1,2}(\mathbf{Set})\in(2,1)\mathbf{Cart}_2 Set 2 = disc 1 , 2 ( Set ) ∈ ( 2 , 1 ) Cart 2 , and for f : C → D f:\mathbf{C}\to\mathbf{D} f : C → D in C a t 2 , c p \mathbf{Cat}_{2,\mathrm{cp}} Cat 2 , cp we have f ! ∘ y C ≃ y D ∘ f f_{!}\circ y_{\mathbf{C}}\simeq y_{\mathbf{D}}\circ f f ! ∘ y C ≃ y D ∘ f . Denoting by id × { G r p d 2 } : C a t 2 , c p × ∗ 2 → C a t 2 , c p × ( 2 , 1 ) C a r t 2 \operatorname{id}\times\{\mathbf{Grpd}_2\}:\mathbf{Cat}_{2,\mathrm{cp}}\times\ast_2\to\mathbf{Cat}_{2,\mathrm{cp}}\times(2,1)\mathbf{Cart}_2 id × { Grpd 2 } : Cat 2 , cp × ∗ 2 → Cat 2 , cp × ( 2 , 1 ) Cart 2 the constant inclusion selecting the object G r p d 2 ∈ ( 2 , 1 ) C a r t 2 \mathbf{Grpd}_2\in(2,1)\mathbf{Cart}_2 Grpd 2 ∈ ( 2 , 1 ) Cart 2 , the assignment C ↦ y C \mathbf{C}\mapsto y_{\mathbf{C}} C ↦ y C (extended to internal groupoids G = ( G 1 ⇒ G 0 ) \mathbf{G}=(G_1\Rightarrow G_0) G = ( G 1 ⇒ G 0 ) in C \mathbf{C} C via G r p d 2 ( C ) ↪ ( 2 , 1 ) P S h v G r p d 2 ( S e t 2 ) ( C ) ↪ ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) ( C ) \mathbf{Grpd}_2(\mathbf{C})\hookrightarrow(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2(\mathbf{Set}_2)}(\mathbf{C})\hookrightarrow(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2(\mathbf{Grpd}_2)}(\mathbf{C}) Grpd 2 ( C ) ↪ ( 2 , 1 ) PShv Grpd 2 ( Set 2 ) ( C ) ↪ ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ( C ) induced by S e t 2 ↪ G r p d 2 \mathbf{Set}_2\hookrightarrow\mathbf{Grpd}_2 Set 2 ↪ Grpd 2 ) is 2 2 2 -natural in C ∈ C a t 2 , c p C \in\mathbf{Cat}_{2,cp} C ∈ Cat 2 , c p , defining a 2-natural transformation y G r p d y^{\mathbf{Grpd}} y Grpd —the groupoid Yoneda 2-embedding —as in the diagrams below (where colors are used as visual aids to trace the corresponding 2-functors):
C a t 2 , c p × ∗ 2 ← ≃ C a t 2 , c p ↓ id × { G r p d 2 } ↓ G r p d 2 C a t 2 , c p × ( 2 , 1 ) C a r t 2 → id × G r p d 2 C a t 2 , c p × ( 2 , 1 ) C a t 2 ↓ G r p d 2 C a t 2 , c p × ( 2 , 1 ) C a t 2 → ( 2 , 1 ) P S h v ( − ) c o v ( − ) ( 2 , 1 ) C a t 2 (6) \begin{aligned}
\mathbf{Cat}_{2,cp}\times*_2 &\xleftarrow{\simeq} \mathbf{Cat}_{2,cp}\\
\downarrow^{\operatorname{id}\times\{\mathbf{Grpd}_2\}} &\qquad\downarrow^{\mathbf{Grpd}_2}\\
\mathbf{Cat}_{2,cp}\times(2,1)\mathbf{Cart}_2 &\xrightarrow{\operatorname{id}\times\mathbf{Grpd}_2} \mathbf{Cat}_{2,cp}\times(2,1)\mathbf{Cat}_2\\
&\qquad\qquad\qquad\qquad\downarrow^{\mathbf{Grpd}_2}\\
\mathbf{Cat}_{2,cp}\times(2,1)\mathbf{Cat}_2 &\xrightarrow{(2,1)\mathbf{PShv}^{\mathrm{cov}}_{(-)}(-)} (2,1)\mathbf{Cat}_2 \tag*{(6)}
\end{aligned} Cat 2 , c p × ∗ 2 ↓ id × { Grpd 2 } Cat 2 , c p × ( 2 , 1 ) Cart 2 Cat 2 , c p × ( 2 , 1 ) Cat 2 ≃ Cat 2 , c p ↓ Grpd 2 id × Grpd 2 Cat 2 , c p × ( 2 , 1 ) Cat 2 ↓ Grpd 2 ( 2 , 1 ) PShv ( − ) cov ( − ) ( 2 , 1 ) Cat 2 (6)
Next, recall from Section 2.3 that on ( 2 , 1 ) C a r t 2 (2,1)\mathbf{Cart}_2 ( 2 , 1 ) Cart 2 (and on C a t 2 , c p \mathbf{Cat}_{2,cp} Cat 2 , c p ) we have the sequence of inclusion 2-natural transformations between G r p 2 \mathbf{Grp}_2 Grp 2 , disc 1 , 2 \operatorname{disc}_{1,2} disc 1 , 2 (or id ( 2 , 1 ) C a r t 2 \operatorname{id}_{(2,1)\mathbf{Cart}_2} id ( 2 , 1 ) Cart 2 ), and G r p d 2 \mathbf{Grpd}_2 Grpd 2 . We denote these 2-functors and their Yoneda embeddings by G r p 2 \mathbf{Grp}_2 Grp 2 , disc 1 , 2 \operatorname{disc}_{1,2} disc 1 , 2 , and G r p d 2 \mathbf{Grpd}_2 Grpd 2 (colored blue, green, and red below, with their intermediate inclusions colored cyan and orange):
( 2 , 1 ) C a r t 2 G r p 2 ⟹ id ⟹ G r p d 2 ↓ ↓ ↓ ( 2 , 1 ) C a t 2 \begin{aligned}
(2,1)\mathbf{Cart}_2 &\\
\mathbf{Grp}_2 &\Longrightarrow\operatorname{id} \Longrightarrow\mathbf{Grpd}_2\\
&\downarrow\qquad\downarrow\qquad\downarrow\\
(2,1)\mathbf{Cat}_2 &
\end{aligned} ( 2 , 1 ) Cart 2 Grp 2 ( 2 , 1 ) Cat 2 ⟹ id ⟹ Grpd 2 ↓ ↓ ↓
Replacing the groupoid 2-functor G r p d 2 \mathbf{Grpd}_2 Grpd 2 in diagram (6) with disc 1 , 2 \operatorname{disc}_{1,2} disc 1 , 2 (where id ( G r p d 2 ) = G r p d 2 \operatorname{id}(\mathbf{Grpd}_2)=\mathbf{Grpd}_2 id ( Grpd 2 ) = Grpd 2 ) or G r p 2 \mathbf{Grp}_2 Grp 2 , the Yoneda 2-embedding factors through the canonical inclusions, yielding the discrete Yoneda 2-embedding y d i s c y^{\mathrm{disc}} y disc and the group Yoneda 2-embedding y G r p y^{\mathrm{Grp}} y Grp :
G r p 2 → disc 1 , 2 → G r p d 2 ↓ y G r p ↓ y d i s c ↓ y G r p d ( 2 , 1 ) P S h v G r p 2 c o v ( G r p d 2 ) ( − ) → ( 2 , 1 ) P S h v G r p d 2 c o v ( − ) → ( 2 , 1 ) P S h v G r p d 2 c o v ( G r p d 2 ) ( − ) \begin{aligned}
\mathbf{Grp}_2 &\xrightarrow{} \operatorname{disc}_{1,2} \xrightarrow{} \mathbf{Grpd}_2\\
\downarrow^{y^{\mathrm{Grp}}} &\downarrow^{y^{\mathrm{disc}}} \downarrow^{y^{\mathbf{Grpd}}}\\
(2,1)\mathbf{PShv}^{\mathrm{cov}}_{\mathbf{Grp}_2}(\mathbf{Grpd}_2)(-) &\xrightarrow{} (2,1)\mathbf{PShv}^{\mathrm{cov}}_{\mathbf{Grpd}_2}(-) \xrightarrow{} (2,1)\mathbf{PShv}^{\mathrm{cov}}_{\mathbf{Grpd}_2}(\mathbf{Grpd}_2)(-)
\end{aligned} Grp 2 ↓ y Grp ( 2 , 1 ) PShv Grp 2 cov ( Grpd 2 ) ( − ) disc 1 , 2 Grpd 2 ↓ y disc ↓ y Grpd ( 2 , 1 ) PShv Grpd 2 cov ( − ) ( 2 , 1 ) PShv Grpd 2 cov ( Grpd 2 ) ( − )
Now, taking the delooping transformation B G r p d 2 : G r p 2 ( G r p d 2 ) → G r p d 2 ( G r p d 2 ) \mathbf{B}_{\mathbf{Grpd}_2}:\mathbf{Grp}_2(\mathbf{Grpd}_2)\to\mathbf{Grpd}_2(\mathbf{Grpd}_2) B Grpd 2 : Grp 2 ( Grpd 2 ) → Grpd 2 ( Grpd 2 ) (which restricts on 0-truncated objects S e t 2 ⊂ G r p d 2 \mathbf{Set}_2\subset\mathbf{Grpd}_2 Set 2 ⊂ Grpd 2 to B S e t 2 : G r p 2 ( S e t 2 ) → G r p d 2 ( S e t 2 ) \mathbf{B}_{\mathbf{Set}_2}:\mathbf{Grp}_2(\mathbf{Set}_2)\to\mathbf{Grpd}_2(\mathbf{Set}_2) B Set 2 : Grp 2 ( Set 2 ) → Grpd 2 ( Set 2 ) ) and applying ( 2 , 1 ) P S h v ( − ) ( C ) (2,1)\mathbf{PShv}_{(-)}(C) ( 2 , 1 ) PShv ( − ) ( C ) yields a 2-natural transformation (both contravariantly on C a t 2 o p \mathbf{Cat}_2^{\mathrm{op}} Cat 2 op via precomposition f ∗ f^* f ∗ and covariantly on C a t 2 , c p \mathbf{Cat}_{2,cp} Cat 2 , c p via left-exact left Kan extension f ! f_! f ! )
B P S h v e x t : ( 2 , 1 ) P S h v G r p 2 ( G r p d 2 ) ( − ) ⟹ ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) ( − ) , \mathbf{B}^{\mathrm{ext}}_{\mathbf{PShv}}:(2,1)\mathbf{PShv}_{\mathbf{Grp}_2}(\mathbf{Grpd}_2)(-)\Longrightarrow(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{Grpd}_2)(-), B PShv ext : ( 2 , 1 ) PShv Grp 2 ( Grpd 2 ) ( − ) ⟹ ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ( − ) ,
compatible with the Yoneda 2-embeddings y G r p , y d i s c , y G r p d y^{\mathrm{Grp}},y^{\mathrm{disc}},y^{\mathbf{Grpd}} y Grp , y disc , y Grpd above. This is the external delooping for ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi. For Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi Stk ( C , J ) \operatorname{Stk}(C,J) Stk ( C , J ) , the external delooping B S h v e x t \mathbf{B}^{\mathrm{ext}}_{\mathrm{Shv}} B Shv ext is obtained by composing B P S h v e x t \mathbf{B}^{\mathrm{ext}}_{\mathrm{PShv}} B PShv ext with the stackification localization transformation L L L from (2) .
Equivalence
The equivalence between the internal and external delooping constructions rests on the fundamental property that internal algebraic or categorical structures in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf category ( 2 , 1 ) P S h v A ( C ) (2,1)\mathbf{PShv}_{A}(C) ( 2 , 1 ) PShv A ( C ) are computed objectwise in A A A . Formally, because the 2-functor ( 2 , 1 ) P S h v ( − ) ( C ) : ( 2 , 1 ) C a r t 2 → ( 2 , 1 ) C a r t 2 (2,1)\mathbf{PShv}_{(-)}(C):(2,1)\mathbf{Cart}_2\to(2,1)\mathbf{Cart}_2 ( 2 , 1 ) PShv ( − ) ( C ) : ( 2 , 1 ) Cart 2 → ( 2 , 1 ) Cart 2 preserves ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits (and both precomposition f ∗ f^* f ∗ along f : C → D f:C\to D f : C → D in C a t 2 o p \mathbf{Cat}_2^{\mathrm{op}} Cat 2 op , which is the inverse image functor U P S h v ∗ U^*_{\mathrm{PShv}} U PShv ∗ of the geometric morphism ( f ∗ ⊣ f ∗ ) : ( 2 , 1 ) P S h v ( C ) → ( 2 , 1 ) P S h v ( D ) (f^*\dashv f_*):(2,1)\mathbf{PShv}(C)\to(2,1)\mathbf{PShv}(D) ( f ∗ ⊣ f ∗ ) : ( 2 , 1 ) PShv ( C ) → ( 2 , 1 ) PShv ( D ) , and left Kan extension f ! f_! f ! along f ∈ C a t 2 , c p f\in\mathbf{Cat}_{2,cp} f ∈ Cat 2 , c p , which is the inverse image functor of the opposite geometric morphism ( f ! ⊣ f ∗ ) : ( 2 , 1 ) P S h v ( D ) → ( 2 , 1 ) P S h v ( C ) (f_!\dashv f^*):(2,1)\mathbf{PShv}(D)\to(2,1)\mathbf{PShv}(C) ( f ! ⊣ f ∗ ) : ( 2 , 1 ) PShv ( D ) → ( 2 , 1 ) PShv ( C ) , preserve finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits), it commutes up to canonical 2-natural equivalence with G r p d 2 \mathbf{Grpd}_2 Grpd 2 , id ( 2 , 1 ) C a r t 2 \operatorname{id}_{(2,1)}\mathbf{Cart}_2 id ( 2 , 1 ) Cart 2 , and G r p 2 \mathbf{Grp}_2 Grp 2 :
C a t 2 o p × ( 2 , 1 ) C a t 2 → ( 2 , 1 ) P S h v ( − ) ( − ) ( 2 , 1 ) C a t 2 → ↑ id × G r p d 2 → ↑ G r p d 2 C a t 2 o p × ( 2 , 1 ) C a r t 2 → ( 2 , 1 ) P S h v ( − ) ( − ) ( 2 , 1 ) C a r t 2 → ↑ id × { G r p d 2 } C a t 2 o p × ∗ 2 ← ≃ C a t 2 o p → ( 2 , 1 ) P S h v G r p d 2 ( 2 , 1 ) P S h v T o p o s 2 o p (7) \begin{aligned}
\mathbf{Cat}_2^{\mathrm{op}}\times(2,1)\mathbf{Cat}_2
&\xrightarrow{(2,1)\mathbf{PShv}_{(-)}(-)}
(2,1)\mathbf{Cat}_2\\
&\mathrel{\phantom{\xrightarrow{}}}\mathord{\uparrow}^{\operatorname{id}\times\mathbf{Grpd}_2}
\mathrel{\phantom{\xrightarrow{}}}\mathord{\uparrow}^{\mathbf{Grpd}_2}\\
\mathbf{Cat}_2^{\mathrm{op}}\times(2,1)\mathbf{Cart}_2
&\xrightarrow{(2,1)\mathbf{PShv}_{(-)}(-)}
(2,1)\mathbf{Cart}_2\\
&\mathrel{\phantom{\xrightarrow{}}}\mathord{\uparrow}^{\operatorname{id}\times\{\mathbf{Grpd}_2\}}
\\
\mathbf{Cat}_2^{\mathrm{op}}\times*_2
&\xleftarrow{\simeq}
\mathbf{Cat}_2^{\mathrm{op}}
\xrightarrow{(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}}
(2,1)\mathbf{PShvTopos}_2^{\mathrm{op}} \tag*{(7)}
\end{aligned} Cat 2 op × ( 2 , 1 ) Cat 2 Cat 2 op × ( 2 , 1 ) Cart 2 Cat 2 op × ∗ 2 ( 2 , 1 ) PShv ( − ) ( − ) ( 2 , 1 ) Cat 2 ↑ id × Grpd 2 ↑ Grpd 2 ( 2 , 1 ) PShv ( − ) ( − ) ( 2 , 1 ) Cart 2 ↑ id × { Grpd 2 } ≃ Cat 2 op ( 2 , 1 ) PShv Grpd 2 ( 2 , 1 ) PShvTopos 2 op (7)
( 2 , 1 ) C a r t 2 ← ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) → ≃ G r p d 2 ∘ U P S h v ∗ ( 2 , 1 ) P S h v T o p o s 2 o p \begin{aligned}
(2,1)\mathbf{Cart}_2
&\mathrel{\substack{\xleftarrow{\ (2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{Grpd}_2)\ }\\[-2pt]\xrightarrow[\simeq]{} }}
\mathbf{Grpd}_2\circ U^*_{\mathbf{PShv}}\\
&\hspace{4em}\mathbf{(2,1)\mathbf{PShvTopos}_2^{\mathrm{op}}}
\end{aligned} ( 2 , 1 ) Cart 2 ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ≃ Grpd 2 ∘ U PShv ∗ ( 2 , 1 ) PShvTopos 2 op
and analogously (both contravariantly via f ∗ f^* f ∗ on C a t 2 o p \mathbf{Cat}_2^{\mathrm{op}} Cat 2 op and covariantly via f ! f_! f ! on C a t 2 , c p \mathbf{Cat}_{2,\mathrm{cp}} Cat 2 , cp ) with G r p d 2 \mathbf{Grpd}_2 Grpd 2 replaced by id \operatorname{id} id or G r p 2 \mathbf{Grp}_2 Grp 2 .
Pasting each Yoneda diagram (6) with the corresponding limit commutation diagram (7) , and comparing with the definition (5) of the internal delooping B P S h v i n t \mathbf{B}^{\mathrm{int}}_{\mathbf{PShv}} B PShv int , we see that B P S h v i n t \mathbf{B}^{\mathrm{int}}_{\mathbf{PShv}} B PShv int is precisely the transport of the external delooping B P S h v e x t \mathbf{B}^{\mathrm{ext}}_{\mathbf{PShv}} B PShv ext along the canonical 2 2 2 -equivalences of (7) . Objectwise, for each C ∈ C a t 2 , c p \mathbf{C}\in\mathbf{Cat}_{2,\mathrm{cp}} C ∈ Cat 2 , cp , this is expressed by the 2 2 2 -commutativity of the following diagram (and for a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos Stk ( C , J ) \operatorname{Stk}(\mathbf{C},J) Stk ( C , J ) , by appending the exact localization functor L J L_J L J from (2) , which preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and all ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits, and therefore commutes 2 2 2 -naturally with G r p 2 \mathbf{Grp}_2 Grp 2 , G r p d 2 \mathbf{Grpd}_2 Grpd 2 , B \mathbf{B} B , and quotient delooping objects B = cod ∘ coeq ∘ B B=\operatorname{cod}\circ\operatorname{coeq}\circ\mathbf{B} B = cod ∘ coeq ∘ B ):
G r p 2 ( C ) → y C ( 2 , 1 ) P S h v G r p 2 ( G r p d 2 ) ( C ) → ≃ G r p 2 ( ( 2 , 1 ) P S h v G r p d 2 ( C ) ) ↓ B C ↓ disc 1 , 2 ( C ) ↓ B P S h v , C e x t G r p d 2 ( C ) → y C ( 2 , 1 ) P S h v G r p d 2 ( C ) → ≃ ( 2 , 1 ) P S h v G r p d 2 ( C ) ↓ ↓ ↓ B P S h v , C i n t G r p d 2 ( C ) → y C ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) ( C ) → ≃ G r p d 2 ( ( 2 , 1 ) P S h v G r p d 2 ( C ) ) (8) \begin{aligned}
\mathbf{Grp}_2(\mathbf{C})
&\xrightarrow{y_{\mathbf C}}
(2,1)\mathbf{PShv}_{\mathbf{Grp}_2(\mathbf{Grpd}_2)}(\mathbf{C})
\xrightarrow{\simeq}
\mathbf{Grp}_2\bigl((2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{C})\bigr)\\
\downarrow{\mathbf{B}_{\mathbf C}}
&\qquad\downarrow{\operatorname{disc}_{1,2}(\mathbf C)}
\qquad\qquad\downarrow{\mathbf{B}^{\mathrm{ext}}_{\mathbf{PShv},\mathbf C}}\\
\mathbf{Grpd}_2(\mathbf{C})
&\xrightarrow{y_{\mathbf C}}
(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{C})
\xrightarrow{\simeq}
(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{C})\\
\downarrow
&\qquad\downarrow
\qquad\qquad\downarrow{\mathbf{B}^{\mathrm{int}}_{\mathbf{PShv},\mathbf C}}\\
\mathbf{Grpd}_2(\mathbf C)
&\xrightarrow{y_{\mathbf C}}
(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2(\mathbf{Grpd}_2)}(\mathbf C)
\xrightarrow{\simeq}
\mathbf{Grpd}_2\bigl((2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf C)\bigr) \tag*{(8)}
\end{aligned} Grp 2 ( C ) ↓ B C Grpd 2 ( C ) ↓ Grpd 2 ( C ) y C ( 2 , 1 ) PShv Grp 2 ( Grpd 2 ) ( C ) ≃ Grp 2 ( ( 2 , 1 ) PShv Grpd 2 ( C ) ) ↓ disc 1 , 2 ( C ) ↓ B PShv , C ext y C ( 2 , 1 ) PShv Grpd 2 ( C ) ≃ ( 2 , 1 ) PShv Grpd 2 ( C ) ↓ ↓ B PShv , C int y C ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ( C ) ≃ Grpd 2 ( ( 2 , 1 ) PShv Grpd 2 ( C ) ) (8)
We summarize these results in the following theorem:
Theorem 3.1 (Equivalence of Internal and External Delooping). We have:
(1) For every cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H ∈ ( 2 , 1 ) C a r t 2 \mathbf H\in(2,1)\mathbf{Cart}_2 H ∈ ( 2 , 1 ) Cart 2 , a 2 2 2 -functor
B H i n t : G r p 2 ( H ) ⟶ G r p d 2 ( H ) , \mathbf{B}^{\mathrm{int}}_{\mathbf H}:\mathbf{Grp}_2(\mathbf H)\longrightarrow\mathbf{Grpd}_2(\mathbf H), B H int : Grp 2 ( H ) ⟶ Grpd 2 ( H ) , the internal delooping:
(2) For every ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H ≃ ( 2 , 1 ) P S h v G r p d 2 ( C ) \mathbf H\simeq(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf C) H ≃ ( 2 , 1 ) PShv Grpd 2 ( C ) , a 2 2 2 -functor
B H e x t : ( 2 , 1 ) P S h v G r p 2 ( G r p d 2 ) ( C ) ⟶ ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) ( C ) , \mathbf{B}^{\mathrm{ext}}_{\mathbf H}:(2,1)\mathbf{PShv}_{\mathbf{Grp}_2(\mathbf{Grpd}_2)}(\mathbf C)\longrightarrow(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2(\mathbf{Grpd}_2)}(\mathbf C), B H ext : ( 2 , 1 ) PShv Grp 2 ( Grpd 2 ) ( C ) ⟶ ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ( C ) , the external delooping.
Furthermore, under the canonical equivalences
G r p 2 ( H ) ≃ ( 2 , 1 ) P S h v G r p 2 ( G r p d 2 ) ( C ) and G r p d 2 ( H ) ≃ ( 2 , 1 ) P S h v G r p d 2 ( G r p d 2 ) ( C ) , \mathbf{Grp}_2(\mathbf H)\simeq(2,1)\mathbf{PShv}_{\mathbf{Grp}_2(\mathbf{Grpd}_2)}(\mathbf C)
\qquad\text{and}\qquad
\mathbf{Grpd}_2(\mathbf H)\simeq(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2(\mathbf{Grpd}_2)}(\mathbf C), Grp 2 ( H ) ≃ ( 2 , 1 ) PShv Grp 2 ( Grpd 2 ) ( C ) and Grpd 2 ( H ) ≃ ( 2 , 1 ) PShv Grpd 2 ( Grpd 2 ) ( C ) , *the 2 2 2 -functors B H i n t \mathbf{B}^{\mathrm{int}}_{\mathbf H} B H int and B H e x t \mathbf{B}^{\mathrm{ext}}_{\mathbf H} B H ext are 2 2 2 -naturally equivalent. Moreover, if H 0 = Stk ( C , J ) ⊂ H \mathbf H_0=\operatorname{Stk}(\mathbf C,J)\subset\mathbf H H 0 = Stk ( C , J ) ⊂ H is a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos with left-exact stackification L J : H → H 0 L_J:\mathbf H\to\mathbf H_0 L J : H → H 0 , then both deloopings descend along L J L_J L J and remain 2 2 2 -naturally equivalent on H 0 \mathbf H_0 H 0 .
:::`
Proof. In ( 2 , 1 ) P S h v G r p d 2 ( C ) (2,1)\mathrm{PShv}_{\mathrm{Grpd}_2}(\mathcal{C}) ( 2 , 1 ) PShv Grpd 2 ( C ) , finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits are computed objectwise in G r p d 2 \mathrm{Grpd}_2 Grpd 2 . Given a group object G ∈ G r p 2 ( ( 2 , 1 ) P S h v G r p d 2 ( C ) ) G \in\mathrm{Grp}_2((2,1)\mathrm{PShv}_{\mathrm{Grpd}_2}(\mathcal{C})) G ∈ Grp 2 (( 2 , 1 ) PShv Grpd 2 ( C )) , evaluating at each U ∈ C U \in\mathcal{C} U ∈ C yields a group object G ( U ) ∈ G r p 2 ( G r p d 2 ) G(U) \in\mathrm{Grp}_2(\mathrm{Grpd}_2) G ( U ) ∈ Grp 2 ( Grpd 2 ) whose one-object delooping groupoid in G r p d 2 ( G r p d 2 ) \mathrm{Grpd}_2(\mathrm{Grpd}_2) Grpd 2 ( Grpd 2 ) is given by B G r p d 2 ( G ( U ) ) = ( G ( U ) ⇒ ∗ ) \mathbf{B}\mathrm{Grpd}_2(G(U)) = (G(U) \Rightarrow*) B Grpd 2 ( G ( U )) = ( G ( U ) ⇒ ∗ ) . By definition of B H e x t \mathbf{B}^{\mathrm{ext}}_{\mathcal{H}} B H ext , the presheaf U ↦ B G r p d 2 ( G ( U ) ) U \mapsto\mathbf{B}\mathrm{Grpd}_2(G(U)) U ↦ B Grpd 2 ( G ( U )) is B H e x t ( G ) \mathbf{B}^{\mathrm{ext}}_{\mathcal{H}}(G) B H ext ( G ) . Since the terminal presheaf ∗ * ∗ has ∗ ( U ) = ∗ *(U) = * ∗ ( U ) = ∗ , the internal groupoid associated to B H e x t ( G ) \mathbf{B}^{\mathrm{ext}}_{\mathcal{H}}(G) B H ext ( G ) has object of objects ∗ * ∗ and object of morphisms G G G , which is literally B H i n t ( G ) \mathbf{B}^{\mathrm{int}}_{\mathcal{H}}(G) B H int ( G ) . Finally, since L J : H → H 0 L_J : \mathcal{H} \to\mathcal{H}_0 L J : H → H 0 preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits, it preserves terminal objects and products, hence commutes with both G r p 2 ( − ) \mathrm{Grp}_2(-) Grp 2 ( − ) , G r p d 2 ( − ) \mathrm{Grpd}_2(-) Grpd 2 ( − ) , and the construction G ↦ ( G ⇒ ∗ ) G \mapsto(G \Rightarrow*) G ↦ ( G ⇒ ∗ ) . □
Remark 3.2 (Foundational Role and Bifurcation in Non-Cartesian Algebra-Valued Cosmoi). While the equivalence between internal and external delooping in Theorem 3.1 is mathematically immediate in the cartesian setting (following directly from objectwise limits and left-exact stackification), its explicit 2-functorial formalization serves a vital purpose for higher geometry and mathematical physics: it pinpoints the exact structural axioms responsible for the coincidence of internal and external delooping in G r p d 2 \mathrm{Grpd}_2 Grpd 2 -valued topoi. Specifically, ( 2 , 1 ) P S h v G r p d 2 ( C ) (2,1)\mathrm{PShv}_{\mathrm{Grpd}_2}(\mathcal{C}) ( 2 , 1 ) PShv Grpd 2 ( C ) is cartesian monoidal ( ⊗ = × ) (\otimes= \times) ( ⊗ = × ) , its monoidal unit is the terminal presheaf ∗ H ( U ) = ∗ *_{\mathcal{H}}(U) = * ∗ H ( U ) = ∗ , and finite pullbacks over ∗ H *_{\mathcal{H}} ∗ H coincide with binary products G × ∗ H G ≃ G × G G \times_{*_{\mathcal{H}}} G \simeq G \times G G × ∗ H G ≃ G × G .
To see how this connects directly to physics, recall from [31 , 24 ] that in an algebra-valued gauge theory (such as algebra-valued Yang–Mills or Einstein–Hilbert–Palatini gravity), one replaces the Lie algebra g \mathfrak{g} g of a Lie group G G G with a general (associative or non-associative) algebra A A A equipped with a bilinear product ⋅ A : A × A → A \cdot_A : A \times A \to A ⋅ A : A × A → A . Locally over a chart U U U , a connection is an A A A -valued 1-form ω ∈ Ω 1 ( U ; A ) \omega\in\Omega^1(U; A) ω ∈ Ω 1 ( U ; A ) whose curvature F ω = d ω + ω ∧ A ω F_\omega= \mathrm{d}\omega+ \omega\wedge_A \omega F ω = d ω + ω ∧ A ω (and higher polynomial invariants) is governed by the algebraic multiplication ⋅ A \cdot_A ⋅ A rather than a Lie bracket integrated to an invertible group G G G . Passing from local infinitesimal data to global finite symmetries encounters two structural departures from ordinary gauge theory: (i) because a general algebra A A A (or its multiplicative monoid M M M ) lacks global group inverses g − 1 g^{-1} g − 1 , the classical Maurer–Cartan formula g − 1 d g g^{-1}\mathrm{d}g g − 1 d g is unavailable (motivating the non-groupal de Rham unit η d R \eta^{\mathrm{dR}} η dR developed in § 5.3 ); and (ii) in the sequel to this work [33 ] , when the base cosmos G r p d 2 \mathrm{Grpd}_2 Grpd 2 is replaced by a non-cartesian symmetric monoidal Bénabou 2-cosmos ( A , ⊗ A , I A ) (\mathcal{A}, \otimes_{\mathcal{A}}, I_{\mathcal{A}}) ( A , ⊗ A , I A ) (e.g., differential graded algebras, O O O -algebras, or modules over a ring stack) equipped with the Day convolution or objectwise tensor product, the monoidal unit I A I_{\mathcal{A}} I A is no longer the terminal object ∗ * ∗ , and B i n t \mathbf{B}^{\mathrm{int}} B int (defined via internal ⊗ A \otimes_{\mathcal{A}} ⊗ A -categories over I A I_{\mathcal{A}} I A ) bifurcates from B e x t \mathbf{B}^{\mathrm{ext}} B ext (defined via Yoneda transport of point-set delooping) ) ) , giving rise to distinct plain and genuine delooping systems.
Delooping Objects
3.4. Delooping Objects. By Theorem 3.1 , to every group object G ∈ G r p 2 ( H ) G \in\mathrm{Grp}_2(\mathcal{H}) G ∈ Grp 2 ( H ) in a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathcal{H} H there corresponds an internal delooping groupoid B H i n t ( G ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathcal{H}}(G) \in\mathrm{Grpd}_2(\mathcal{H}) B H int ( G ) ∈ Grpd 2 ( H ) . Taking its quotient object in H \mathcal{H} H via the homotopy coequalizer functor cod ∘ coeq : G r p d 2 ( H ) → H \operatorname{cod} \circ\operatorname{coeq} : \mathrm{Grpd}_2(\mathcal{H}) \to\mathcal{H} cod ∘ coeq : Grpd 2 ( H ) → H from § 2.3 , we obtain the representing object
B H i n t G : = cod ( coeq ( B H i n t ( G ) ) ) = colim ( G ⇒ ∗ ) ∈ H . B^{\mathrm{int}}_{\mathcal{H}}G := \operatorname{cod}(\operatorname{coeq}(B^{\mathrm{int}}_{\mathcal{H}}(G))) = \operatorname{colim}(G \Rightarrow*) \in\mathcal{H}. B H int G := cod ( coeq ( B H int ( G ))) = colim ( G ⇒ ∗ ) ∈ H .
We call B H i n t G B^{\mathrm{int}}_{\mathcal{H}}G B H int G the internal delooping object (or classifying object ) of G G G .
If H \mathcal{H} H is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos or a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, we also have the external delooping groupoid B H e x t ( G ) \mathbf{B}^{\mathrm{ext}}_{\mathcal{H}}(G) B H ext ( G ) and its representing object, the external delooping object
B H e x t G : = cod ( coeq ( B H e x t ( G ) ) ) ∈ H . B^{\mathrm{ext}}_{\mathcal{H}}G := \operatorname{cod}(\operatorname{coeq}(B^{\mathrm{ext}}_{\mathcal{H}}(G))) \in\mathcal{H}. B H ext G := cod ( coeq ( B H ext ( G ))) ∈ H .
By Theorem 3.1 , the canonical 2-natural equivalence B H i n t ( G ) ≃ B H e x t ( G ) B^{\mathrm{int}}_{\mathcal{H}}(G) \simeq B^{\mathrm{ext}}_{\mathcal{H}}(G) B H int ( G ) ≃ B H ext ( G ) in G r p d 2 ( H ) \mathrm{Grpd}_2(\mathcal{H}) Grpd 2 ( H ) induces an equivalence of delooping objects in H \mathcal{H} H :
B H i n t G ≃ B H e x t G , B^{\mathrm{int}}_{\mathcal{H}}G \simeq B^{\mathrm{ext}}_{\mathcal{H}}G, B H int G ≃ B H ext G ,
which we henceforth denote simply by B G BG B G (or B G BG B G when H \mathcal{H} H is clear from context).
Crucially, because in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathcal{H} H every internal groupoid 0-truncated morphism fibers is effective in H \mathcal{H} H , for any 0-truncated group object G ∈ G r p ( τ ≤ 0 H ) ⊂ G r p 2 ( H ) G \in\mathrm{Grp}(\tau_{\leq0}\mathcal{H}) \subset\mathrm{Grp}_2(\mathcal{H}) G ∈ Grp ( τ ≤ 0 H ) ⊂ Grp 2 ( H ) the delooping groupoid B H i n t ( G ) \mathbf{B}^{\mathrm{int}}_{\mathcal{H}}(G) B H int ( G ) can be losslessly reconstructed from its delooping object B H G B_{\mathcal{H}}G B H G together with its canonical basepoint ∗ → B H G * \to B_{\mathcal{H}}G ∗ → B H G as the Čech groupoid C ˇ ( ∗ → B H G ) \check{\mathcal{C}}(* \to B_{\mathcal{H}}G) C ˇ ( ∗ → B H G ) , whereas for a general 1-truncated 2-group G ∈ G r p 2 ( H ) G \in\mathrm{Grp}_2(\mathcal{H}) G ∈ Grp 2 ( H ) , B H i n t ( G ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathcal{H}}(G) \in\mathrm{Grpd}_2(\mathcal{H}) B H int ( G ) ∈ Grpd 2 ( H ) retains the full 2-group data prior to 1-truncation by cod ∘ coeq \operatorname{cod} \circ\operatorname{coeq} cod ∘ coeq .
Higher Delooping
For every cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathbf{H} H , composing the internal delooping 2 2 2 -functor B H i n t : G r p 2 ( H ) → G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}:\mathbf{Grp}_{2}(\mathbf{H})\to\mathbf{Grpd}_{2}(\mathbf{H}) B H int : Grp 2 ( H ) → Grpd 2 ( H ) with the quotient object 2 2 2 -functor cod ∘ coeq : G r p d 2 ( H ) → H \operatorname{cod}\circ\operatorname{coeq}:\mathbf{Grpd}_{2}(\mathbf{H})\to\mathbf{H} cod ∘ coeq : Grpd 2 ( H ) → H yields the delooping object 2 2 2 -functor
B H = cod ∘ coeq ∘ B H i n t : G r p 2 ( H ) ⟶ H . \mathbf{B}_{\mathbf{H}}=\operatorname{cod}\circ\operatorname{coeq}\circ\mathbf{B}^{\mathrm{int}}_{\mathbf{H}}:\mathbf{Grp}_{2}(\mathbf{H})\longrightarrow\mathbf{H}. B H = cod ∘ coeq ∘ B H int : Grp 2 ( H ) ⟶ H .
It is natural to ask when the delooping construction can be iterated. Observe that a second delooping B H 2 G : = B H ( B H G ) \mathbf{B}^{2}_{\mathbf{H}}G:=\mathbf{B}_{\mathbf{H}}(\mathbf{B}_{\mathbf{H}}G) B H 2 G := B H ( B H G ) (realized losslessly via the internal groupoid B H i n t ( G ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)\in\mathbf{Grpd}_{2}(\mathbf{H}) B H int ( G ) ∈ Grpd 2 ( H ) as explained in Remark 3.3 below, or via the quotient B H G ∈ H \mathbf{B}_{\mathbf{H}}G\in\mathbf{H} B H G ∈ H when G ∈ G r p ( τ ≤ 0 H ) G\in\mathbf{Grp}(\tau_{\leq0}\mathbf{H}) G ∈ Grp ( τ ≤ 0 H ) is 0 0 0 -truncated) is defined if and only if B H i n t ( G ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)\in\mathbf{Grpd}_{2}(\mathbf{H}) B H int ( G ) ∈ Grpd 2 ( H ) (resp. B H G ∈ H \mathbf{B}_{\mathbf{H}}G\in\mathbf{H} B H G ∈ H ) itself admits a lift to a group object in G r p 2 ( G r p d 2 ( H ) ) \mathbf{Grp}_{2}(\mathbf{Grpd}_{2}(\mathbf{H})) Grp 2 ( Grpd 2 ( H )) (resp. G r p 2 ( H ) \mathbf{Grp}_{2}(\mathbf{H}) Grp 2 ( H ) ). That is, let D e l 2 2 ( H ) \mathbf{Del}^{2}_{2}(\mathbf{H}) Del 2 2 ( H ) denote the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (fitting into the lossless homotopy pullback square on the left below, which for 0 0 0 -truncated groups in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos is equivalent to the quotient pullback square on the right) of group objects G G G whose delooping is equipped with a compatible group structure:
D e l 2 2 ( H ) → G r p 2 ( H ) → B H i n t G r p d 2 ( H ) B ~ H ↓ B H i n t ↓ G r p 2 ( G r p d 2 ( H ) ) → U G r p d 2 ( H ) D e l 2 2 ( τ ≤ 0 H ) → G r p ( τ ≤ 0 H ) → B H H B ~ H ↓ U ↓ G r p 2 ( H ) → B H H \begin{CD}
\mathbf{Del}^{2}_{2}(\mathbf{H}) @>>> \mathbf{Grp}_{2}(\mathbf{H}) @>{\mathbf{B}^{\mathrm{int}}_{\mathbf{H}}}>> \mathbf{Grpd}_{2}(\mathbf{H})\\
@V{\widetilde{\mathbf{B}}_{\mathbf{H}}}VV @V{\mathbf{B}^{\mathrm{int}}_{\mathbf{H}}}VV\\
\mathbf{Grp}_{2}(\mathbf{Grpd}_{2}(\mathbf{H})) @>{U}>> \mathbf{Grpd}_{2}(\mathbf{H})
\end{CD}
\qquad
\begin{CD}
\mathbf{Del}^{2}_{2}(\tau_{\leq0}\mathbf{H}) @>>> \mathbf{Grp}(\tau_{\leq0}\mathbf{H}) @>{\mathbf{B}_{\mathbf{H}}}>> \mathbf{H}\\
@V{\widetilde{\mathbf{B}}_{\mathbf{H}}}VV @V{U}VV\\
\mathbf{Grp}_{2}(\mathbf{H}) @>{\mathbf{B}_{\mathbf{H}}}>> \mathbf{H}
\end{CD} Del 2 2 ( H ) B H ↓ ⏐ Grp 2 ( Grpd 2 ( H )) U Grp 2 ( H ) B H int ↓ ⏐ Grpd 2 ( H ) B H int Grpd 2 ( H ) Del 2 2 ( τ ≤ 0 H ) B H ↓ ⏐ Grp 2 ( H ) B H Grp ( τ ≤ 0 H ) U ↓ ⏐ H B H H
As we will prove in § 3.6 , the forgetful functor D e l 2 2 ( H ) → G r p 2 ( H ) \mathbf{Del}^{2}_{2}(\mathbf{H})\to\mathbf{Grp}_{2}(\mathbf{H}) Del 2 2 ( H ) → Grp 2 ( H ) is faithful on 1 1 1 -morphisms (1 1 1 -faithful), while D e l 2 3 ( H ) → D e l 2 2 ( H ) \mathbf{Del}^{3}_{2}(\mathbf{H})\to\mathbf{Del}^{2}_{2}(\mathbf{H}) Del 2 3 ( H ) → Del 2 2 ( H ) is fully faithful and D e l 2 k ( H ) → D e l 2 3 ( H ) \mathbf{Del}^{k}_{2}(\mathbf{H})\to\mathbf{Del}^{3}_{2}(\mathbf{H}) Del 2 k ( H ) → Del 2 3 ( H ) is an equivalence for all k ≥ 3 k\geq3 k ≥ 3 . Recursively, we define ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories D e l 2 k ( H ) \mathbf{Del}^{k}_{2}(\mathbf{H}) Del 2 k ( H ) for k ≥ 0 k\geq0 k ≥ 0 , with forgetful functors D e l 2 k ( H ) → D e l 2 l ( H ) \mathbf{Del}^{k}_{2}(\mathbf{H})\to\mathbf{Del}^{l}_{2}(\mathbf{H}) Del 2 k ( H ) → Del 2 l ( H ) for l ≤ k l\leq k l ≤ k , consisting of objects that are k k k -tuply deloopable, i.e., for which the k k k -fold delooping B H k G \mathbf{B}^{k}_{\mathbf{H}}G B H k G exists. Setting D e l 2 1 ( H ) : = G r p 2 ( H ) \mathbf{Del}^{1}_{2}(\mathbf{H}):=\mathbf{Grp}_{2}(\mathbf{H}) Del 2 1 ( H ) := Grp 2 ( H ) and D e l 2 0 ( H ) : = H \mathbf{Del}^{0}_{2}(\mathbf{H}):=\mathbf{H} Del 2 0 ( H ) := H , we obtain the tower of iterated delooping 2 2 2 -functors:
⋯ ⟶ D e l 2 k + 1 ( H ) ⟶ D e l 2 k ( H ) ⟶ D e l 2 k − 1 ( H ) ⟶ ⋯ ⟶ D e l 2 2 ( H ) ⟶ D e l 2 1 ( H ) ⟶ D e l 2 0 ( H ) \cdots\longrightarrow\mathbf{Del}^{k+1}_{2}(\mathbf{H})\longrightarrow\mathbf{Del}^{k}_{2}(\mathbf{H})\longrightarrow\mathbf{Del}^{k-1}_{2}(\mathbf{H})\longrightarrow\cdots\longrightarrow\mathbf{Del}^{2}_{2}(\mathbf{H})\longrightarrow\mathbf{Del}^{1}_{2}(\mathbf{H})\longrightarrow\mathbf{Del}^{0}_{2}(\mathbf{H}) ⋯ ⟶ Del 2 k + 1 ( H ) ⟶ Del 2 k ( H ) ⟶ Del 2 k − 1 ( H ) ⟶ ⋯ ⟶ Del 2 2 ( H ) ⟶ Del 2 1 ( H ) ⟶ Del 2 0 ( H )
where B m j : D e l 2 m ( H ) → D e l 2 0 ( H ) \mathbf{B}^{j}_{m}:\mathbf{Del}^{m}_{2}(\mathbf{H})\to\mathbf{Del}^{0}_{2}(\mathbf{H}) B m j : Del 2 m ( H ) → Del 2 0 ( H ) denotes the j j j -fold delooping B H j \mathbf{B}^{j}_{\mathbf{H}} B H j (and B m 0 \mathbf{B}^{0}_{m} B m 0 is the underlying object functor). Taking the inverse ( 2 , 1 ) (2,1) ( 2 , 1 ) -limit of the forgetful tower in ( 2 , 1 ) C a t 2 (2,1)\mathbf{Cat}_{2} ( 2 , 1 ) Cat 2 yields the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category
D e l 2 ∞ ( H ) : = lim ⟵ k D e l 2 k ( H ) \mathbf{Del}^{\infty}_{2}(\mathbf{H}):=\lim_{\longleftarrow k}\mathbf{Del}^{k}_{2}(\mathbf{H}) Del 2 ∞ ( H ) := ⟵ k lim Del 2 k ( H )
of infinitely deloopable objects (or connective spectrum objects) in H \mathbf{H} H .
Remark 3.3 (1-Truncation of Quotient Deloopings vs. Higher Ambient Deloopings). In a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H (where every object is 1 1 1 -truncated), one must carefully distinguish between:
The internal delooping groupoid B H i n t ( G ) = ( G ⇒ ∗ ) ∈ G r p d 2 ( H ) \mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)=(G\Rightarrow*)\in\mathbf{Grpd}_{2}(\mathbf{H}) B H int ( G ) = ( G ⇒ ∗ ) ∈ Grpd 2 ( H ) , which is lossless for any group object G ∈ G r p 2 ( H ) G\in\mathbf{Grp}_{2}(\mathbf{H}) G ∈ Grp 2 ( H ) (since Ω ∗ B H i n t ( G ) = G \Omega_{*}\mathbf{B}^{\mathrm{int}}_{\mathbf{H}}(G)=G Ω ∗ B H int ( G ) = G in G r p d 2 ( H ) \mathbf{Grpd}_{2}(\mathbf{H}) Grpd 2 ( H ) ); and
(2) The quotient delooping object B H q u o t G : = cod ( coeq H ( G ⇉ ∗ ) ) ∈ H B^{\mathrm{quot}}_{\mathbf{H}}G := \operatorname{cod}(\operatorname{coeq}_{\mathbf{H}}(G \rightrightarrows*)) \in\mathbf{H} B H quot G := cod ( coeq H ( G ⇉ ∗ )) ∈ H , which reflects B H i n t ( G ) B^{\mathrm{int}}_{\mathbf{H}}(G) B H int ( G ) into the 1-truncated category H \mathbf{H} H .
Because Ω X = ∗ × X ∗ \Omega X = * \times_X * Ω X = ∗ × X ∗ in a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H is always 0-truncated (Ω X ∈ τ ≤ 0 H \Omega X \in\tau_{\leq0}\mathbf{H} Ω X ∈ τ ≤ 0 H ), the quotient B H q u o t G ∈ H B^{\mathrm{quot}}_{\mathbf{H}}G \in\mathbf{H} B H quot G ∈ H satisfies Ω B H q u o t G ≃ τ ≤ 0 G \Omega B^{\mathrm{quot}}_{\mathbf{H}}G \simeq\tau_{\leq0}G Ω B H quot G ≃ τ ≤ 0 G . In particular, if A ∈ ComGrp ( τ ≤ 0 H ) A \in\operatorname{ComGrp}(\tau_{\leq0}\mathbf{H}) A ∈ ComGrp ( τ ≤ 0 H ) is a 0-truncated abelian group sheaf (e.g., U ( 1 ) U(1) U ( 1 ) ), its first delooping B H A ∈ H B_{\mathbf{H}}A \in\mathbf{H} B H A ∈ H is 0-connected (τ ≤ 0 B H A ≃ ∗ \tau_{\leq0}B_{\mathbf{H}}A \simeq* τ ≤ 0 B H A ≃ ∗ ), so taking a second quotient coequalizer inside H \mathbf{H} H would yield cod ( coeq H ( B H A ⇉ ∗ ) ) ≃ τ ≤ 1 K ( A , 2 ) ≃ ∗ H \operatorname{cod}(\operatorname{coeq}_{\mathbf{H}}(B_{\mathbf{H}}A \rightrightarrows*)) \simeq\tau_{\leq1}K(A,2) \simeq*_{\mathbf{H}} cod ( coeq H ( B H A ⇉ ∗ )) ≃ τ ≤ 1 K ( A , 2 ) ≃ ∗ H . Instead, for k ≥ 2 k \geq2 k ≥ 2 , the k k k -fold delooping B H k G B_{\mathbf{H}}^kG B H k G of a k k k -tupl[y] groupal object G G G is losslessly realized as the ( k − 1 ) (k-1) ( k − 1 ) -fold internal groupoid B H k − 1 ( B H G ) ∈ Grpd 2 k − 1 ( H ) B_{\mathbf{H}}^{k-1}(B_{\mathbf{H}}G) \in\operatorname{Grpd}_{2}^{k-1}(\mathbf{H}) B H k − 1 ( B H G ) ∈ Grpd 2 k − 1 ( H ) (equivalently, as the k k k -truncated object B k G B^kG B k G in the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos H ∞ \mathbf{H}_{\infty} H ∞ of which H = τ ≤ 1 H ∞ \mathbf{H}=\tau_{\leq1}\mathbf{H}_{\infty} H = τ ≤ 1 H ∞ is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation). Notice that for k = 2 k=2 k = 2 and A ∈ ComGrp ( τ ≤ 0 H ) A \in\operatorname{ComGrp}(\tau_{\leq0}\mathbf{H}) A ∈ ComGrp ( τ ≤ 0 H ) , B H A ∈ ComGrp 2 ( H ) B_{\mathbf{H}}A \in\operatorname{ComGrp}_{2}(\mathbf{H}) B H A ∈ ComGrp 2 ( H ) is a 1-truncated symmetric 2-group in H \mathbf{H} H and B H 2 A = B H i n t ( B H A ) = ( B H A ⇉ ∗ ) ∈ Grpd 2 ( H ) B_{\mathbf{H}}^2A = B_{\mathbf{H}}^{\mathrm{int}}(B_{\mathbf{H}}A) = (B_{\mathbf{H}}A \rightrightarrows*) \in\operatorname{Grpd}_{2}(\mathbf{H}) B H 2 A = B H int ( B H A ) = ( B H A ⇉ ∗ ) ∈ Grpd 2 ( H ) is a single internal groupoid (a 2-stack) in H \mathbf{H} H classifying A A A -bundle gerbes. Furthermore, since any finite-limit-preserving endofunctor (such as ♭ \flat ♭ , ♯ \sharp ♯ , ♭ d R \flat_{\mathrm{dR}} ♭ dR , or ∫ \int ∫ on a strongly 1-connected topos) and any representable functor H ( X , − ) \mathbf{H}(X,-) H ( X , − ) preserves internal groupoids, they prolong objectwise to Grpd 2 k − 1 ( H ) \operatorname{Grpd}_{2}^{k-1}(\mathbf{H}) Grpd 2 k − 1 ( H ) , with H ( X , B H k G ) : = τ ≤ 1 H ∞ ( X , B k G ) \mathbf{H}(X,B_{\mathbf{H}}^kG):=\tau_{\leq1}\mathbf{H}_{\infty}(X,B^kG) H ( X , B H k G ) := τ ≤ 1 H ∞ ( X , B k G ) denoting the 1-truncated hom-groupoid of ( k − 1 ) (k-1) ( k − 1 ) -bundles. Throughout what follows, whenever k ≥ 2 k\geq2 k ≥ 2 (or when G ∈ Grp 2 ( H ) G\in\operatorname{Grp}_{2}(\mathbf{H}) G ∈ Grp 2 ( H ) is 1-truncated), B H k G B_{\mathbf{H}}^kG B H k G denotes this uncollapsed delooping.
Eckmann–Hilton
To say that the delooping B H G B_{\mathbf{H}}G B H G exists is to say that G G G is a group object in H \mathbf{H} H . In turn, to say that B H 2 G = B H ( B H G ) B_{\mathbf{H}}^2G=B_{\mathbf{H}}(B_{\mathbf{H}}G) B H 2 G = B H ( B H G ) exists (in the lossless sense of Remark 3.3) is to say that both G G G and its first delooping B H i n t ( G ) B_{\mathbf{H}}^{\mathrm{int}}(G) B H int ( G ) carry compatible group structures. Since finite products commute with B H i n t B_{\mathbf{H}}^{\mathrm{int}} B H int (and with B H B_{\mathbf{H}} B H ), a group structure on B H i n t ( G ) B_{\mathbf{H}}^{\mathrm{int}}(G) B H int ( G ) (or on B H G B_{\mathbf{H}}G B H G when Ω B H G ≃ G \Omega B_{\mathbf{H}}G\simeq G Ω B H G ≃ G ) induces a second group structure on G G G that is homomorphic with respect to the first.
More generally, define the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category Grp 2 k ( H ) \operatorname{Grp}_{2}^{k}(\mathbf{H}) Grp 2 k ( H ) of k k k -tuply group objects in H \mathbf{H} H inductively by
Grp 2 0 ( H ) : = H , Grp 2 1 ( H ) : = Grp 2 ( H ) , Grp 2 k ( H ) : = Grp 2 ( Grp 2 k − 1 ( H ) ) ( k ≥ 1 ) . \operatorname{Grp}_{2}^{0}(\mathbf{H}) := \mathbf{H}, \qquad\operatorname{Grp}_{2}^{1}(\mathbf{H}) := \operatorname{Grp}_{2}(\mathbf{H}), \qquad\operatorname{Grp}_{2}^{k}(\mathbf{H}) := \operatorname{Grp}_{2}\bigl(\operatorname{Grp}_{2}^{k-1}(\mathbf{H})\bigr) \quad(k \geq1). Grp 2 0 ( H ) := H , Grp 2 1 ( H ) := Grp 2 ( H ) , Grp 2 k ( H ) := Grp 2 ( Grp 2 k − 1 ( H ) ) ( k ≥ 1 ) .
That is, an object of Grp 2 k ( H ) \operatorname{Grp}_{2}^{k}(\mathbf{H}) Grp 2 k ( H ) is an object of H \mathbf{H} H equipped with k k k pairwise compatible group structures.
Lemma 3.4. For every cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathbf{H} H (with Del 2 k ( H ) \operatorname{Del}_{2}^{k}(\mathbf{H}) Del 2 k ( H ) defined via the lossless internal delooping B H i n t B_{\mathbf{H}}^{\mathrm{int}} B H int of Remark 3.3, or via B H B_{\mathbf{H}} B H on τ ≤ 0 H \tau_{\leq0}\mathbf{H} τ ≤ 0 H in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos) and every k ≥ 0 k\geq0 k ≥ 0 , there is a natural equivalence of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories
Grp 2 k ( H ) ≃ Del 2 k ( H ) . \operatorname{Grp}_{2}^{k}(\mathbf{H}) \simeq\operatorname{Del}_{2}^{k}(\mathbf{H}). Grp 2 k ( H ) ≃ Del 2 k ( H ) .
Proof. We proceed by induction on k k k . For k = 0 , 1 k=0,1 k = 0 , 1 , we have Grp 2 0 ( H ) = H = Del 2 0 ( H ) \operatorname{Grp}_{2}^{0}(\mathbf{H})=\mathbf{H}=\operatorname{Del}_{2}^{0}(\mathbf{H}) Grp 2 0 ( H ) = H = Del 2 0 ( H ) and Grp 2 1 ( H ) = Grp 2 ( H ) = Del 2 1 ( H ) \operatorname{Grp}_{2}^{1}(\mathbf{H})=\operatorname{Grp}_{2}(\mathbf{H})=\operatorname{Del}_{2}^{1}(\mathbf{H}) Grp 2 1 ( H ) = Grp 2 ( H ) = Del 2 1 ( H ) by definition. For the inductive step, observe that B H i n t : Grp 2 ( H ) → Grpd 2 ( H ) B_{\mathbf{H}}^{\mathrm{int}}:\operatorname{Grp}_{2}(\mathbf{H})\to\operatorname{Grpd}_{2}(\mathbf{H}) B H int : Grp 2 ( H ) → Grpd 2 ( H ) (and B H B_{\mathbf{H}} B H ) preserves finite products, and recovering G G G as the automorphism/loop object of the basepoint ∗ → B H i n t ( G ) * \to B_{\mathbf{H}}^{\mathrm{int}}(G) ∗ → B H int ( G ) preserves all ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits. Consequently, a group object structure on the delooping of G G G corresponds to an internal group structure on G G G inside Grp 2 ( H ) \operatorname{Grp}_{2}(\mathbf{H}) Grp 2 ( H ) , i.e., an object of Grp 2 ( Grp 2 ( H ) ) = Grp 2 2 ( H ) \operatorname{Grp}_{2}(\operatorname{Grp}_{2}(\mathbf{H}))=\operatorname{Grp}_{2}^{2}(\mathbf{H}) Grp 2 ( Grp 2 ( H )) = Grp 2 2 ( H ) . Iterating this equivalence k − 1 k-1 k − 1 times yields Del 2 k ( H ) ≃ Grp 2 ( Del 2 k − 1 ( H ) ) ≃ Grp 2 ( Grp 2 k − 1 ( H ) ) = Grp 2 k ( H ) \operatorname{Del}_{2}^{k}(\mathbf{H})\simeq\operatorname{Grp}_{2}(\operatorname{Del}_{2}^{k-1}(\mathbf{H}))\simeq\operatorname{Grp}_{2}(\operatorname{Grp}_{2}^{k-1}(\mathbf{H}))=\operatorname{Grp}_{2}^{k}(\mathbf{H}) Del 2 k ( H ) ≃ Grp 2 ( Del 2 k − 1 ( H )) ≃ Grp 2 ( Grp 2 k − 1 ( H )) = Grp 2 k ( H ) . □
By the ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical Eckmann–Hilton argument [9 , 17 , 1 ] , two compatible group (or monoid) structures on an object G G G in a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -category coincide up to canonical 2-isomorphism, and their compatibility data translates into a braiding on the single underlying multiplication; for k ≥ 3 k\geq3 k ≥ 3 compatible structures, the braiding becomes symmetric and no further data can be added. Let us formalize this via E k E_k E k -algebras.
Let H \mathbf{H} H be a cartesian monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category. From Lemma 3.4 , we have the tower of forgetful functors:
⋯ ⟶ Grp 2 k ( H ) ⟶ ⋯ ⟶ Grp 2 3 ( H ) ⟶ Grp 2 2 ( H ) ⟶ Grp 2 1 ( H ) ⟶ Grp 2 0 ( H ) . (9) \cdots\longrightarrow\operatorname{Grp}_{2}^{k}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\operatorname{Grp}_{2}^{3}(\mathbf{H}) \longrightarrow\operatorname{Grp}_{2}^{2}(\mathbf{H}) \longrightarrow\operatorname{Grp}_{2}^{1}(\mathbf{H}) \longrightarrow\operatorname{Grp}_{2}^{0}(\mathbf{H}). \tag*{(9)} ⋯ ⟶ Grp 2 k ( H ) ⟶ ⋯ ⟶ Grp 2 3 ( H ) ⟶ Grp 2 2 ( H ) ⟶ Grp 2 1 ( H ) ⟶ Grp 2 0 ( H ) . (9)
Define the corresponding tower of k k k -tuply monoid objects Mon 2 k ( H ) \operatorname{Mon}_{2}^{k}(\mathbf{H}) Mon 2 k ( H ) inductively by
Mon 2 0 ( H ) : = H , Mon 2 1 ( H ) : = Mon 2 ( H ) , Mon 2 k ( H ) : = Mon 2 ( Mon 2 k − 1 ( H ) ) ( k ≥ 1 ) . \operatorname{Mon}_{2}^{0}(\mathbf{H}) := \mathbf{H}, \qquad\operatorname{Mon}_{2}^{1}(\mathbf{H}) := \operatorname{Mon}_{2}(\mathbf{H}), \qquad\operatorname{Mon}_{2}^{k}(\mathbf{H}) := \operatorname{Mon}_{2}\bigl(\operatorname{Mon}_{2}^{k-1}(\mathbf{H})\bigr) \quad(k \geq1). Mon 2 0 ( H ) := H , Mon 2 1 ( H ) := Mon 2 ( H ) , Mon 2 k ( H ) := Mon 2 ( Mon 2 k − 1 ( H ) ) ( k ≥ 1 ) .
Since G r p 2 ( C ) ⊂ M o n 2 ( C ) \mathrm{Grp}_{2}(\mathbf{C}) \subset\mathbf{Mon}_{2}(\mathbf{C}) Grp 2 ( C ) ⊂ Mon 2 ( C ) is the full sub-2-category of grouplike monoids for any cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -category C \mathbf{C} C , each G r p 2 k ( H ) \mathrm{Grp}_{2}^{k}(\mathbf{H}) Grp 2 k ( H ) is a full sub-2-category of M o n 2 k ( H ) \mathbf{Mon}_{2}^{k}(\mathbf{H}) Mon 2 k ( H ) fitting into the pullback squares:
⋯ ⟶ M o n 2 k ( H ) ⟶ ⋯ ⟶ M o n 2 3 ( H ) ⟶ M o n 2 2 ( H ) ⟶ M o n 2 1 ( H ) ⟶ M o n 2 0 ( H ) ↑ ↑ ↑ ↑ ∥ ⋯ ⟶ G r p 2 k ( H ) ⟶ ⋯ ⟶ G r p 2 3 ( H ) ⟶ G r p 2 2 ( H ) ⟶ G r p 2 1 ( H ) ⟶ G r p 2 0 ( H ) (10) \begin{aligned}
\cdots&\longrightarrow\mathbf{Mon}_{2}^{k}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{Mon}_{2}^{3}(\mathbf{H}) \longrightarrow\mathbf{Mon}_{2}^{2}(\mathbf{H}) \longrightarrow\mathbf{Mon}_{2}^{1}(\mathbf{H}) \longrightarrow\mathbf{Mon}_{2}^{0}(\mathbf{H}) \\
&\uparrow\qquad\qquad\qquad\qquad\qquad\qquad\uparrow\qquad\qquad\qquad\qquad\uparrow\qquad\qquad\qquad\uparrow\qquad\qquad\qquad\Vert\\
\cdots&\longrightarrow\mathbf{Grp}_{2}^{k}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{Grp}_{2}^{3}(\mathbf{H}) \longrightarrow\mathbf{Grp}_{2}^{2}(\mathbf{H}) \longrightarrow\mathbf{Grp}_{2}^{1}(\mathbf{H}) \longrightarrow\mathbf{Grp}_{2}^{0}(\mathbf{H}) \tag*{(10)}
\end{aligned} ⋯ ⋯ ⟶ Mon 2 k ( H ) ⟶ ⋯ ⟶ Mon 2 3 ( H ) ⟶ Mon 2 2 ( H ) ⟶ Mon 2 1 ( H ) ⟶ Mon 2 0 ( H ) ↑ ↑ ↑ ↑ ∥ ⟶ Grp 2 k ( H ) ⟶ ⋯ ⟶ Grp 2 3 ( H ) ⟶ Grp 2 2 ( H ) ⟶ Grp 2 1 ( H ) ⟶ Grp 2 0 ( H ) (10)
Theorem 3.5 (Eckmann–Hilton / Dunn Identification for k k k -Tuply Monoids). Let H \mathbf{H} H be a cartesian monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category. For every k ≥ 0 k \ge0 k ≥ 0 , there is a 2-natural equivalence
M o n 2 k ( H ) ≃ E k A l g 2 ( H ) \mathbf{Mon}_{2}^{k}(\mathbf{H}) \simeq\mathbf{E}_{k}\mathbf{Alg}_{2}(\mathbf{H}) Mon 2 k ( H ) ≃ E k Alg 2 ( H ) between k k k -tuply monoid objects and E k \mathbf{E}_{k} E k -algebra objects (with E 0 A l g 2 \mathbf{E}_{0}\mathbf{Alg}_{2} E 0 Alg 2 understood in the unpointed convention E 0 A l g 2 ( H ) ≃ H \mathbf{E}_{0}\mathbf{Alg}_{2}(\mathbf{H}) \simeq\mathbf{H} E 0 Alg 2 ( H ) ≃ H for k = 0 k=0 k = 0 ), such that the following square of forgetful 2-functors commutes up to 2-natural equivalence:
E k A l g 2 ( H ) → E k − 1 A l g 2 ( H ) ≃ ↓ ↓ ≃ M o n 2 k ( H ) → M o n 2 k − 1 ( H ) \begin{CD}
\mathbf{E}_{k}\mathbf{Alg}_{2}(\mathbf{H}) @>>> \mathbf{E}_{k-1}\mathbf{Alg}_{2}(\mathbf{H}) \\
@V{\simeq}VV @VV{\simeq}V \\
\mathbf{Mon}_{2}^{k}(\mathbf{H}) @>>> \mathbf{Mon}_{2}^{k-1}(\mathbf{H})
\end{CD} E k Alg 2 ( H ) ≃ ↓ ⏐ Mon 2 k ( H ) E k − 1 Alg 2 ( H ) ↓ ⏐ ≃ Mon 2 k − 1 ( H )
Proof. We proceed by induction on k ≥ 1 k \ge1 k ≥ 1 . For k = 1 k=1 k = 1 , M o n 2 1 ( H ) = M o n 2 ( H ) ≃ E 1 A l g 2 ( H ) \mathbf{Mon}_{2}^{1}(\mathbf{H})=\mathbf{Mon}_{2}(\mathbf{H})\simeq\mathbf{E}_{1}\mathbf{Alg}_{2}(\mathbf{H}) Mon 2 1 ( H ) = Mon 2 ( H ) ≃ E 1 Alg 2 ( H ) by Example 2.8. Suppose the equivalence M o n 2 k − 1 ( H ) ≃ E k − 1 A l g 2 ( H ) \mathbf{Mon}_{2}^{k-1}(\mathbf{H})\simeq\mathbf{E}_{k-1}\mathbf{Alg}_{2}(\mathbf{H}) Mon 2 k − 1 ( H ) ≃ E k − 1 Alg 2 ( H ) holds for k − 1 ≥ 1 k-1 \ge1 k − 1 ≥ 1 . By the universal property of the Boardman–Vogt tensor product ⊙ \odot ⊙ of ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads [27 Theorem 5.1.2.2] , for any symmetric ( 2 , 1 ) (2,1) ( 2 , 1 ) -operads O 1 , O 2 \mathcal{O}_{1},\mathcal{O}_{2} O 1 , O 2 and any symmetric monoidal ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H , there is a canonical equivalence
( O 1 ⊙ O 2 ) A l g 2 ( H ) ≃ O 1 A l g 2 ( O 2 A l g 2 ( H ) ) . (\mathcal{O}_{1}\odot\mathcal{O}_{2})\mathbf{Alg}_{2}(\mathbf{H}) \simeq\mathcal{O}_{1}\mathbf{Alg}_{2}\bigl(\mathcal{O}_{2}\mathbf{Alg}_{2}(\mathbf{H})\bigr). ( O 1 ⊙ O 2 ) Alg 2 ( H ) ≃ O 1 Alg 2 ( O 2 Alg 2 ( H ) ) .
Taking O 1 = E 1 \mathcal{O}_{1}=\mathbf{E}_{1} O 1 = E 1 and O 2 = E k − 1 \mathcal{O}_{2}=\mathbf{E}_{k-1} O 2 = E k − 1 , and applying the Dunn–Lurie Additivity Theorem 2.7 (E 1 ⊙ E k − 1 ≃ E k \mathbf{E}_{1}\odot\mathbf{E}_{k-1}\simeq\mathbf{E}_{k} E 1 ⊙ E k − 1 ≃ E k ), we obtain
E k A l g 2 ( H ) ≃ ( E 1 ⊙ E k − 1 ) A l g 2 ( H ) ≃ E 1 A l g 2 ( E k − 1 A l g 2 ( H ) ) ≃ M o n 2 ( M o n 2 k − 1 ( H ) ) = M o n 2 k ( H ) , \mathbf{E}_{k}\mathbf{Alg}_{2}(\mathbf{H}) \simeq(\mathbf{E}_{1}\odot\mathbf{E}_{k-1})\mathbf{Alg}_{2}(\mathbf{H}) \simeq\mathbf{E}_{1}\mathbf{Alg}_{2}\bigl(\mathbf{E}_{k-1}\mathbf{Alg}_{2}(\mathbf{H})\bigr) \simeq\mathbf{Mon}_{2}\bigl(\mathbf{Mon}_{2}^{k-1}(\mathbf{H})\bigr)=\mathbf{Mon}_{2}^{k}(\mathbf{H}), E k Alg 2 ( H ) ≃ ( E 1 ⊙ E k − 1 ) Alg 2 ( H ) ≃ E 1 Alg 2 ( E k − 1 Alg 2 ( H ) ) ≃ Mon 2 ( Mon 2 k − 1 ( H ) ) = Mon 2 k ( H ) ,
compatibly with the forgetful functors induced by E 0 → E 1 \mathbf{E}_{0}\to\mathbf{E}_{1} E 0 → E 1 (followed by the basepoint-forgetting projection 1 / H → H 1/\mathbf{H}\to\mathbf{H} 1/ H → H at the bottom level k = 0 k=0 k = 0 ). □
Combining Theorem 3.5 with diagram (10) and the definition of groupal E k \mathbf{E}_{k} E k -algebras from § 2.5 , the vertical equivalences restrict to groupal objects:
⋯ ⟶ E k A l g 2 g r p ( H ) ⟶ ⋯ ⟶ E 2 A l g 2 g r p ( H ) ⟶ E 1 A l g 2 g r p ( H ) ≃ ↓ ≃ ↓ ≃ ↓ ⋯ ⟶ G r p 2 k ( H ) ⟶ ⋯ ⟶ G r p 2 2 ( H ) ⟶ G r p 2 1 ( H ) ↓ ↓ ↓ ⋯ ⟶ M o n 2 k ( H ) ⟶ ⋯ ⟶ M o n 2 2 ( H ) ⟶ M o n 2 1 ( H ) ⟶ M o n 2 0 ( H ) ≃ ↑ ≃ ↑ ≃ ↑ ∥ ⋯ ⟶ E k A l g 2 ( H ) ⟶ ⋯ ⟶ E 2 A l g 2 ( H ) ⟶ E 1 A l g 2 ( H ) ⟶ H \begin{aligned}
\cdots&\longrightarrow\mathbf{E}_{k}\mathbf{Alg}_{2}^{\mathrm{grp}}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{E}_{2}\mathbf{Alg}_{2}^{\mathrm{grp}}(\mathbf{H}) \longrightarrow\mathbf{E}_{1}\mathbf{Alg}_{2}^{\mathrm{grp}}(\mathbf{H}) \\
&\simeq\downarrow\qquad\qquad\qquad\qquad\qquad\qquad\simeq\downarrow\qquad\qquad\qquad\qquad\simeq\downarrow\\
\cdots&\longrightarrow\mathbf{Grp}_{2}^{k}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{Grp}_{2}^{2}(\mathbf{H}) \longrightarrow\mathbf{Grp}_{2}^{1}(\mathbf{H}) \\
&\downarrow\qquad\qquad\qquad\qquad\qquad\qquad\downarrow\qquad\qquad\qquad\qquad\downarrow\\
\cdots&\longrightarrow\mathbf{Mon}_{2}^{k}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{Mon}_{2}^{2}(\mathbf{H}) \longrightarrow\mathbf{Mon}_{2}^{1}(\mathbf{H}) \longrightarrow\mathbf{Mon}_{2}^{0}(\mathbf{H}) \\
&\simeq\uparrow\qquad\qquad\qquad\qquad\qquad\qquad\simeq\uparrow\qquad\qquad\qquad\qquad\simeq\uparrow\qquad\qquad\qquad\Vert\\
\cdots&\longrightarrow\mathbf{E}_{k}\mathbf{Alg}_{2}(\mathbf{H}) \longrightarrow\cdots\longrightarrow\mathbf{E}_{2}\mathbf{Alg}_{2}(\mathbf{H}) \longrightarrow\mathbf{E}_{1}\mathbf{Alg}_{2}(\mathbf{H}) \longrightarrow\mathbf{H}
\end{aligned} ⋯ ⋯ ⋯ ⋯ ⟶ E k Alg 2 grp ( H ) ⟶ ⋯ ⟶ E 2 Alg 2 grp ( H ) ⟶ E 1 Alg 2 grp ( H ) ≃↓ ≃↓ ≃↓ ⟶ Grp 2 k ( H ) ⟶ ⋯ ⟶ Grp 2 2 ( H ) ⟶ Grp 2 1 ( H ) ↓ ↓ ↓ ⟶ Mon 2 k ( H ) ⟶ ⋯ ⟶ Mon 2 2 ( H ) ⟶ Mon 2 1 ( H ) ⟶ Mon 2 0 ( H ) ≃↑ ≃↑ ≃↑ ∥ ⟶ E k Alg 2 ( H ) ⟶ ⋯ ⟶ E 2 Alg 2 ( H ) ⟶ E 1 Alg 2 ( H ) ⟶ H
Putting Lemma 3.4 and Theorem 3.5 together yields the complete algebraic characterization of k k k -tuply deloopable objects:
Theorem 3.6. For every cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathbf{H} H and every k ≥ 1 k \ge1 k ≥ 1 , there are 2 2 2 -natural equivalences of ( 2 , 1 ) (2,1) ( 2 , 1 ) -categories
Del 2 k ( H ) ≃ Grp 2 k ( H ) ≃ E k A l g 2 g r p ( H ) . \operatorname{Del}^{k}_{2}(\mathbf{H}) \simeq\operatorname{Grp}^{k}_{2}(\mathbf{H}) \simeq\mathop{\mathrm{E}}^{k}\mathrm{Alg}^{\mathrm{grp}}_{2}(\mathbf{H}). Del 2 k ( H ) ≃ Grp 2 k ( H ) ≃ E k Alg 2 grp ( H ) . Joining Theorem 3.6 with the Stabilization Theorem 2.9 and Example 2.8, we obtain:
Corollary 3.7. Let X ∈ H X \in\mathbf{H} X ∈ H be an object of a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos (or cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos as above). Then:
X X X is deloopable (k = 1 k=1 k = 1 ) if and only if X X X admits the structure of a group object;
X X X is doubly deloopable (k = 2 k=2 k = 2 ) if and only if X X X admits the structure of a braided group object;
X X X is triply deloopable (k = 3 k=3 k = 3 ) if and only if X X X admits the structure of a commutative group object;
X X X is k k k -tuply deloopable for k > 3 k>3 k > 3 if and only if X X X is triply deloopable;
X X X is infinitely deloopable (X ∈ Del 2 ∞ ( H ) X \in\operatorname{Del}^{\infty}_{2}(\mathbf{H}) X ∈ Del 2 ∞ ( H ) ) if and only if X X X admits the structure of a commutative group object.
Looping vs. Delooping
Recall that a based (or pointed ) object in a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -category H \mathbf{H} H with terminal object ∗ * ∗ is a 1 1 1 -morphism x : ∗ → X x:* \to X x : ∗ → X , i.e., an object of the coslice ( 2 , 1 ) (2,1) ( 2 , 1 ) -category ∗ / H */\mathbf{H} ∗ / H . Equivalently, it is an object X X X endowed with a global element. There is a canonical projection 2 2 2 -functor π : ∗ / H → H \pi: */\mathbf{H} \to\mathbf{H} π : ∗ / H → H that forgets the basepoint.
Assuming H \mathbf{H} H has ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks, we define the looping 2 2 2 -functor Ω : ∗ / H → H \Omega: */\mathbf{H} \to\mathbf{H} Ω : ∗ / H → H by taking the homotopy self-intersection (( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback) of the basepoint ∗ → x X * \xrightarrow{x} X ∗ x X along itself, equipped with its universal loop 2 2 2 -isomorphism λ X : x ∘ ! ⇒ x ∘ ! \lambda_X:x\circ! \Rightarrow x\circ! λ X : x ∘ ! ⇒ x ∘ ! :
Ω X → ! ∗ ! ↓ ↓ x ∗ → x X ⇓ λ X \begin{CD}
\Omega X @>{!}>> * \\
@V{!}VV @VV{x}V \\
* @>{x}>> X
\end{CD}
\qquad
\Downarrow\lambda_X Ω X ! ↓ ⏐ ∗ ! x ∗ ↓ ⏐ x X ⇓ λ X
We call Ω X = ∗ × X ∗ \Omega X=* \times_X * Ω X = ∗ × X ∗ the looping object of ( X , x ) (X,x) ( X , x ) . Observe that the pullback diagram above is precisely the diagram defining the Čech groupoid C ˇ H ( x ) \check{\mathrm{C}}_{\mathbf{H}}(x) C ˇ H ( x ) of the basepoint morphism x : ∗ → X x:* \to X x : ∗ → X . Since the object of objects of C ˇ H ( x ) \check{\mathrm{C}}_{\mathbf{H}}(x) C ˇ H ( x ) is C ˇ H ( x ) 0 = ∗ \check{\mathrm{C}}_{\mathbf{H}}(x)_0=* C ˇ H ( x ) 0 = ∗ , the internal groupoid C ˇ H ( x ) = ( Ω X ⇒ ∗ ) \check{\mathrm{C}}_{\mathbf{H}}(x)=(\Omega X \Rightarrow*) C ˇ H ( x ) = ( Ω X ⇒ ∗ ) is a one-object internal groupoid, and hence is the internal delooping B H i n t ( Ω X ) \mathrm{B}^{\mathrm{int}}_{\mathbf{H}}(\Omega X) B H int ( Ω X ) of a canonical group object structure on Ω X \Omega X Ω X (where multiplication is induced by vertical pasting of the loop 2 2 2 -cell λ X \lambda_X λ X in the pullback). Moreover, because H \mathbf{H} H is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category (so its hom-groupoids are 1 1 1 -truncated), the homotopy pullback Ω X = ∗ × X ∗ \Omega X=* \times_X * Ω X = ∗ × X ∗ is automatically 0 0 0 -truncated (Ω X ∈ τ ≤ 0 H \Omega X \in\tau_{\leq0}\mathbf{H} Ω X ∈ τ ≤ 0 H ). Consequently, the looping 2 2 2 -functor Ω \Omega Ω factors canonically through 0 0 0 -truncated group objects Grp ( τ ≤ 0 H ) ⊂ Grp 2 ( H ) \operatorname{Grp}(\tau_{\leq0}\mathbf{H}) \subset\operatorname{Grp}_{2}(\mathbf{H}) Grp ( τ ≤ 0 H ) ⊂ Grp 2 ( H ) :
∗ / H → Ω Grp ( τ ≤ 0 H ) Ω ↓ ↓ H → Grp 2 ( H ) \begin{CD}
*/\mathbf{H} @>{\Omega}>> \operatorname{Grp}(\tau_{\leq0}\mathbf{H}) \\
@V{\Omega}VV @VV{}V \\
\mathbf{H} @>{}>> \operatorname{Grp}_{2}(\mathbf{H})
\end{CD} ∗ / H Ω ↓ ⏐ H Ω Grp ( τ ≤ 0 H ) ↓ ⏐ Grp 2 ( H )
Conversely, for every group object G ∈ Grp 2 ( H ) G \in\operatorname{Grp}_{2}(\mathbf{H}) G ∈ Grp 2 ( H ) in a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos, its internal delooping object B H G = coeq ( G ⇉ ∗ ) \mathrm{B}_{\mathbf{H}}G=\operatorname{coeq}(G \rightrightarrows*) B H G = coeq ( G ⇉ ∗ ) comes equipped with the universal quotient morphism p G : ∗ ↠ B H G p_G:* \twoheadrightarrow\mathrm{B}_{\mathbf{H}}G p G : ∗ ↠ B H G , which is a canonical global element of B H G \mathrm{B}_{\mathbf{H}}G B H G . Thus, B H \mathrm{B}_{\mathbf{H}} B H factors canonically through the pointed category ∗ / H */\mathbf{H} ∗ / H :
G ← p G B H G → ∗ p G ↓ ↑ ∗ ← G Grp 2 ( H ) → B H H B H ↓ ↑ π ∗ / H → H \begin{CD}
G @<{p_G}<< \mathrm{B}_{\mathbf{H}}G @>>> * \\
@V{p_G}VV @. @AA{}A \\
* @<<< G
\end{CD}
\qquad
\begin{CD}
\operatorname{Grp}_{2}(\mathbf{H}) @>{\mathrm{B}_{\mathbf{H}}}>> \mathbf{H} \\
@V{\mathrm{B}_{\mathbf{H}}}VV @AA{\pi}A \\
*/\mathbf{H} @>{}>> \mathbf{H}
\end{CD} G p G ↓ ⏐ ∗ p G B H G G ∗ ⏐ ↑ Grp 2 ( H ) B H ↓ ⏐ ∗ / H B H H ⏐ ↑ π H
Combining both factorizations yields the adjoint pair B H ⊣ Ω B_{\mathrm{H}} \dashv\Omega B H ⊣ Ω between Grp 2 ( H ) \operatorname{Grp}_{2}(\mathrm{H}) Grp 2 ( H ) (or Grp ( τ ≤ 0 H ) \operatorname{Grp}(\tau_{\leq0}\mathrm{H}) Grp ( τ ≤ 0 H ) ) and ∗ / H \ast/\mathrm{H} ∗ / H :
H ← B H Grp 2 ( H ) π ↑ ↓ ∗ / H → Ω H \begin{CD}
\mathrm{H} @<B_{\mathrm{H}}<< \operatorname{Grp}_{2}(\mathrm{H}) \\
@A\pi AA @VVV \\
\ast/\mathrm{H} @>\Omega>> \mathrm{H}
\end{CD} H π ⏐ ↑ ∗ / H B H Ω Grp 2 ( H ) ↓ ⏐ H
When H \mathrm{H} H is a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, every internal groupoid with 0 0 0 -truncated morphism fibers is effective in H \mathrm{H} H . Thus, for every 0 0 0 -truncated group object G ∈ Grp ( τ ≤ 0 H ) G \in\operatorname{Grp}(\tau_{\leq0}\mathrm{H}) G ∈ Grp ( τ ≤ 0 H ) (while for a general 2 2 2 -group G ∈ Grp 2 ( H ) G \in\operatorname{Grp}_{2}(\mathrm{H}) G ∈ Grp 2 ( H ) , Ω B H q u o t G ≃ τ ≤ 0 G \Omega B_{\mathrm{H}}^{\mathrm{quot}}G \simeq\tau_{\leq0}G Ω B H quot G ≃ τ ≤ 0 G unless B H G B_{\mathrm{H}}G B H G is taken in the ambient ( 3 , 1 ) (3,1) ( 3 , 1 ) -topos Grpd 2 ( H ) \operatorname{Grpd}_{2}(\mathrm{H}) Grpd 2 ( H ) as in Remark 3.3), effectiveness means that B H i n t ( G ) → ∼ C ˇ ( ∗ ↠ B H G ) = ( Ω B H G ⇉ ∗ ) B_{\mathrm{H}}^{\mathrm{int}}(G) \xrightarrow{\sim} \check{\mathrm{C}}(\ast\twoheadrightarrow B_{\mathrm{H}}G)=(\Omega B_{\mathrm{H}}G \rightrightarrows\ast) B H int ( G ) ∼ C ˇ ( ∗ ↠ B H G ) = ( Ω B H G ⇉ ∗ ) , so the unit of B H ⊣ Ω B_{\mathrm{H}}\dashv\Omega B H ⊣ Ω is a 2 2 2 -natural equivalence:
G → ∼ Ω B H G , G \xrightarrow{\sim} \Omega B_{\mathrm{H}}G, G ∼ Ω B H G ,
as illustrated by the universal pullback comparison:
\begin{CD}
\Omega B_{\mathrm{H}}G @>>> \ast\\
@VVV @VVV \\
\ast@>{p_G}>> B_{\mathrm{H}}G \\
@AAA @AAA \\
\ast@>{p_G}>> G
\end{CD}
\begin{CD}
G @>{\sim}>> \Omega B_{\mathrm{H}}G @>>> \ast\\
@VVV @VVV @VVV \\
\ast@>>> B_{\mathrm{H}}G
\end{CD}
On the other hand, for a pointed object ( X , x : ∗ → X ) ∈ ∗ / H (X,x:\ast\to X)\in\ast/\mathrm{H} ( X , x : ∗ → X ) ∈ ∗ / H , the counit B H Ω X → X B_{\mathrm{H}}\Omega X\to X B H Ω X → X is an equivalence if and only if the basepoint x : ∗ → X x:\ast\to X x : ∗ → X is an effective epimorphism (so that X X X is the quotient of its Čech groupoid C ˇ ( x ) = B H i n t ( Ω X ) \check{\mathrm{C}}(x)=B_{\mathrm{H}}^{\mathrm{int}}(\Omega X) C ˇ ( x ) = B H int ( Ω X ) ). A pointed object ( X , x ) (X,x) ( X , x ) whose basepoint ∗ ↠ X \ast\twoheadrightarrow X ∗ ↠ X is an effective epimorphism (equivalently, τ ≤ 0 X ≃ ∗ \tau_{\leq0}X\simeq\ast τ ≤ 0 X ≃ ∗ ) is called 0 0 0 -connected (or simply connected ). Thus, for an arbitrary pointed object X ∈ ∗ / H X\in\ast/\mathrm{H} X ∈ ∗ / H and group object G ∈ Grp 2 ( H ) G\in\operatorname{Grp}_{2}(\mathrm{H}) G ∈ Grp 2 ( H ) , B H q u o t B_{\mathrm{H}}^{\mathrm{quot}} B H quot and Ω \Omega Ω inside H \mathrm{H} H fail to be mutually inverse in general for two reasons:
In an arbitrary cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathrm{H} H , internal delooping groupoids B H i n t ( G ) B_{\mathrm{H}}^{\mathrm{int}}(G) B H int ( G ) need not be effective (and even in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathrm{H} H , the 1 1 1 -truncated quotient B H q u o t G ∈ H B_{\mathrm{H}}^{\mathrm{quot}}G\in\mathrm{H} B H quot G ∈ H is effective in H \mathrm{H} H if and only if G ∈ Grp ( τ ≤ 0 H ) G\in\operatorname{Grp}(\tau_{\leq0}\mathrm{H}) G ∈ Grp ( τ ≤ 0 H ) is 0 0 0 -truncated);
Even when all 0 0 0 -truncated groupoids are effective, B H Ω X → X B_{\mathrm{H}}\Omega X\to X B H Ω X → X is an equivalence only on the connected component of the chosen basepoint ∗ → X \ast\to X ∗ → X .
In a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathrm{H} H , restricting to 0 0 0 -truncated group objects Grp ( τ ≤ 0 H ) \operatorname{Grp}(\tau_{\leq0}\mathrm{H}) Grp ( τ ≤ 0 H ) and to the full sub-2 2 2 -category Conn ( ∗ / H ) ⊂ ∗ / H \operatorname{Conn}(\ast/\mathrm{H})\subset\ast/\mathrm{H} Conn ( ∗ / H ) ⊂ ∗ / H of 0 0 0 -connected pointed objects resolves both obstructions:
Theorem 3.8 (Looping–Delooping Equivalence [26 Lemma 7.2.2.11] ). For every Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathrm{H} H , the internal delooping 2 2 2 -functor B H B_{\mathrm{H}} B H factors through connected pointed objects and establishes an equivalence of categories
B H : Grp ( τ ≤ 0 H ) → ∼ Conn ( ∗ / H ) , B_{\mathrm{H}}:\operatorname{Grp}(\tau_{\leq0}\mathrm{H})\xrightarrow{\sim}\operatorname{Conn}(\ast/\mathrm{H}), B H : Grp ( τ ≤ 0 H ) ∼ Conn ( ∗ / H ) , with quasi-inverse given by the looping 2 2 2 -functor Ω : Conn ( ∗ / H ) → ∼ Grp ( τ ≤ 0 H ) \Omega:\operatorname{Conn}(\ast/\mathrm{H})\xrightarrow{\sim}\operatorname{Grp}(\tau_{\leq0}\mathrm{H}) Ω : Conn ( ∗ / H ) ∼ Grp ( τ ≤ 0 H ) (while on all of Grp 2 ( H ) \operatorname{Grp}_{2}(\mathrm{H}) Grp 2 ( H ) , the unquotiented delooping B H i n t : Grp 2 ( H ) → ∼ Conn ( ∗ / Grpd 2 e f f ( H ) ) B_{\mathrm{H}}^{\mathrm{int}}:\operatorname{Grp}_{2}(\mathrm{H})\xrightarrow{\sim}\operatorname{Conn}(\ast/\operatorname{Grpd}_{2}^{\mathrm{eff}}(\mathrm{H})) B H int : Grp 2 ( H ) ∼ Conn ( ∗ / Grpd 2 eff ( H )) is an equivalence onto connected pointed 2 2 2 -stacks).
Proof. This is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation of May’s recognition principle in an ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos (see [26 Lemma 7.2.2.11] ). Explicitly, for any G ∈ Grp ( τ ≤ 0 H ) G\in\operatorname{Grp}(\tau_{\leq0}\mathrm{H}) G ∈ Grp ( τ ≤ 0 H ) , the coequalizer morphism ∗ ↠ B H G \ast\twoheadrightarrow B_{\mathrm{H}}G ∗ ↠ B H G is an effective epimorphism, so ( B H G , p G ) ∈ Conn ( ∗ / H ) (B_{\mathrm{H}}G,p_G)\in\operatorname{Conn}(\ast/\mathrm{H}) ( B H G , p G ) ∈ Conn ( ∗ / H ) , and effectiveness of B H i n t ( G ) B_{\mathrm{H}}^{\mathrm{int}}(G) B H int ( G ) gives G → ∼ Ω B H G G\xrightarrow{\sim}\Omega B_{\mathrm{H}}G G ∼ Ω B H G . Conversely, for ( X , x ) ∈ Conn ( ∗ / H ) (X,x)\in\operatorname{Conn}(\ast/\mathrm{H}) ( X , x ) ∈ Conn ( ∗ / H ) , x : ∗ → X x:\ast\to X x : ∗ → X is an effective epimorphism with 0 0 0 -truncated kernel pair Ω X ∈ Grp ( τ ≤ 0 H ) \Omega X\in\operatorname{Grp}(\tau_{\leq0}\mathrm{H}) Ω X ∈ Grp ( τ ≤ 0 H ) , so X ≃ cod ( coeq ( C ˇ ( x ) ) ) = B H ( Ω X ) X\simeq\operatorname{cod}(\operatorname{coeq}(\check{\mathrm{C}}(x)))=B_{\mathrm{H}}(\Omega X) X ≃ cod ( coeq ( C ˇ ( x ))) = B H ( Ω X ) .
Remark 3.9 (Higher Looping–Delooping Equivalence). In the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos H ∞ \mathbf{H}_{\infty} H ∞ of which H = τ ≤ 1 H ∞ \mathbf{H}=\tau_{\leq1}\mathbf{H}_{\infty} H = τ ≤ 1 H ∞ is the 1-truncated subtopos (or via iterated internal groupoids G r p d 2 k ( H ) \mathbf{Grpd}^{k}_{2}(\mathbf{H}) Grpd 2 k ( H ) as in Remark 3.3), Theorem 3.8 extends by induction to k k k -tuply groupal objects for all k ≥ 1 k\geq1 k ≥ 1 [27 Theorem 5.2.6.10] : defining a pointed object ( X , x ) (X,x) ( X , x ) to be ( k − 1 ) (k-1) ( k − 1 ) -connected if ∗ → X \ast\to X ∗ → X is an effective epimorphism and Ω X \Omega X Ω X is ( k − 2 ) (k-2) ( k − 2 ) -connected, the k k k -fold delooping B k B^{k} B k and k k k -fold looping Ω k \Omega^{k} Ω k establish an equivalence
B k : Del 2 k ( H ) ≃ Grp 2 k ( H ) ≃ E k Alg 2 g r p ( H ) → ≃ Conn ≤ k − 1 k + 1 ( ∗ / H ∞ ) B^{k}\colon\operatorname{Del}^{k}_{2}(\mathbf{H})\simeq\operatorname{Grp}^{k}_{2}(\mathbf{H})\simeq\mathbf{E}_{k}\operatorname{Alg}^{\mathrm{grp}}_{2}(\mathbf{H})\xrightarrow{\ \simeq\ } \operatorname{Conn}^{k+1}_{\leq k-1}(\ast/\mathbf{H}_{\infty}) B k : Del 2 k ( H ) ≃ Grp 2 k ( H ) ≃ E k Alg 2 grp ( H ) ≃ Conn ≤ k − 1 k + 1 ( ∗ / H ∞ )
between groupal E k \mathbf{E}_{k} E k -algebras in H \mathbf{H} H and ( k − 1 ) (k-1) ( k − 1 ) -connected, ( k + 1 ) (k+1) ( k + 1 ) -truncated pointed objects in H ∞ \mathbf{H}_{\infty} H ∞ (note that inside H = τ ≤ 1 H ∞ \mathbf{H}=\tau_{\leq1}\mathbf{H}_{\infty} H = τ ≤ 1 H ∞ , every object is 1-truncated, so a 1-connected object inside H \mathbf{H} H is already contractible).
Cohesive 2-Topoi
Having established the general theory of Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi and delooping objects, we now introduce the axiomatic structure of cohesion [23 , 38 ] . Whereas an arbitrary Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos behaves like a category of generalized spaces on which one can do topology and sheaf theory, a cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos carries an adjoint triple of internal modalities ℜ ⊣ ♭ ⊣ ♯ \Re\dashv\flat\dashv\sharp ℜ ⊣ ♭ ⊣ ♯ that enriches its internal logic with the synthetic contrast between discrete underlying sets of points, constant coefficients, and continuous/smooth geometric cohesion. In § 4.1 we define cohesive geometric morphisms and their induced modalities. In § 4.2 we show how the progressive combination of ℜ \Re ℜ , ♭ \flat ♭ , and ♯ \sharp ♯ internalizes principal bundles, flat and closed differential forms, principal connections, and internal moduli stacks of gauge fields. Finally, in § 5 we synthesize these constructions into intrinsic, modal, and differential ( 2 , 1 ) (2,1) ( 2 , 1 ) -cohomology.
Cohesive Modalities
Let f = ( f ∗ , f ∗ ) : H → H ′ f=(f_{\ast},f^{\ast})\colon\mathbf{H}\to\mathbf{H}' f = ( f ∗ , f ∗ ) : H → H ′ be a geometric morphism between ( 2 , 1 ) (2,1) ( 2 , 1 ) -pretopoi (or Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topoi).
We say that f f f is 1-connected if the inverse image 2-functor f ∗ : H ′ → H f^{\ast}\colon\mathbf{H}'\to\mathbf{H} f ∗ : H ′ → H is fully faithful.
If f ∗ f^{\ast} f ∗ admits a further left 2-adjoint f ! : H → H ′ f_{!}\colon\mathbf{H}\to\mathbf{H}' f ! : H → H ′ , we say that f f f is locally 1-connected (first diagram below).
If f f f is locally 1-connected and f ! f_{!} f ! preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -products, we say that f f f is strongly 1-connected (or product-preserving locally 1-connected ).
Every 1-connected locally 1-connected geometric morphism satisfies f ! f ∗ ≃ i d H ′ f_{!}f^{\ast}\simeq\mathrm{id}_{\mathbf{H}'} f ! f ∗ ≃ id H ′ ; in particular, a locally 1-connected geometric morphism ( f ! , f ∗ , f ∗ ) (f_{!},f^{\ast},f_{\ast}) ( f ! , f ∗ , f ∗ ) is 1-connected if and only if the counit f ! f ∗ ⇒ i d H ′ f_{!}f^{\ast}\Rightarrow\mathrm{id}_{\mathbf{H}'} f ! f ∗ ⇒ id H ′ is an equivalence (and when H ′ = G r p d 2 \mathbf{H}'=\mathbf{Grpd}_{2} H ′ = Grpd 2 , if and only if f ! f_{!} f ! preserves the terminal object ∗ \ast ∗ [15 , 38 ] ). Thus, we have the strict hierarchy of inclusions:
strongly 1-connected ⊂ 1-connected ⊂ locally 1-connected . (11) \text{strongly 1-connected}\subset\text{1-connected}\subset\text{locally 1-connected}. \tag*{(11)} strongly 1-connected ⊂ 1-connected ⊂ locally 1-connected . (11)
Dually:
If the direct image f ∗ : H → H ′ f_{\ast}\colon\mathbf{H}\to\mathbf{H}' f ∗ : H → H ′ admits a further right 2-adjoint f ! : H ′ → H f^{!}\colon\mathbf{H}'\to\mathbf{H} f ! : H ′ → H , we say that f f f is locally local (second diagram below).
If f f f is locally local and f ! f^{!} f ! is fully faithful, we say that f f f is local . By general properties of adjoint triples ( f ∗ ⊣ f ∗ ⊣ f ! ) (f^{\ast}\dashv f_{\ast}\dashv f^{!}) ( f ∗ ⊣ f ∗ ⊣ f ! ) , f ! f^{!} f ! is fully faithful if and only if f ∗ f^{\ast} f ∗ is fully faithful; hence a locally local geometric morphism is local if and only if it is 1-connected [15 ] .
Finally, we say that f f f is locally cohesive (third diagram below) if it is both locally 1-connected and locally local, so that we have an adjoint quadruple f ! ⊣ f ∗ ⊣ f ∗ ⊣ f ! f_{!}\dashv f^{\ast}\dashv f_{\ast}\dashv f^{!} f ! ⊣ f ∗ ⊣ f ∗ ⊣ f ! ; and we say that f f f is cohesive if, in addition, it is strongly 1-connected and local.
Figure 1.
Given a ( 2 , 1 ) (2,1) ( 2 , 1 ) -pretopos H \mathbf{H} H , endo-2 2 2 -functors M : H → H M : \mathbf{H} \to\mathbf{H} M : H → H (typically idempotent monads or comonads) are called internal modalities on H \mathbf{H} H . For X ∈ H X \in\mathbf{H} X ∈ H , we say that M ( X ) ∈ H M(X) \in\mathbf{H} M ( X ) ∈ H represents X X X in the modality M M M . Every adjoint pair ( L ⊣ R ) : C → D (L \dashv R) : \mathbf{C} \to\mathbf{D} ( L ⊣ R ) : C → D between 2 2 2 -categories defines an endofunctor L ∘ R : C → C L \circ R : \mathbf{C} \to\mathbf{C} L ∘ R : C → C (a comonad on C \mathbf{C} C ). In particular, every geometric morphism f : H → H ′ f : \mathbf{H} \to\mathbf{H}' f : H → H ′ induces the f f f -flat modality (a lex idempotent comonad when f f f is 1 1 1 -connected):
♭ f : = f ∗ ∘ f ∗ : H ⟶ H . \flat_f := f^* \circ f_* : \mathbf{H} \longrightarrow\mathbf{H}. ♭ f := f ∗ ∘ f ∗ : H ⟶ H .
Furthermore, an adjoint triple L ′ ⊣ L ⊣ R L' \dashv L \dashv R L ′ ⊣ L ⊣ R induces an adjoint pair of endofunctors ( L ∘ L ′ ) ⊣ ( L ∘ R ) (L \circ L') \dashv(L \circ R) ( L ∘ L ′ ) ⊣ ( L ∘ R ) .
Consequently:
Every locally 1 1 1 -connected geometric morphism ( f ! , f ∗ , f ∗ ) : H → H ′ (f_!, f^*, f_*) : \mathbf{H} \to\mathbf{H}' ( f ! , f ∗ , f ∗ ) : H → H ′ induces an adjoint pair of internal modalities ∫ f ⊣ ♭ f \int_f \dashv\flat_f ∫ f ⊣ ♭ f , where
∫ f : = f ∗ ∘ f ! : H ⟶ H \int_f := f^* \circ f_! : \mathbf{H} \longrightarrow\mathbf{H} ∫ f := f ∗ ∘ f ! : H ⟶ H
is called the f f f -shape modality (an idempotent monad when f f f is 1 1 1 -connected).
Every locally local geometric morphism ( f ∗ , f ∗ , f ! ) : H → H ′ (f^*, f_*, f_!) : \mathbf{H} \to\mathbf{H}' ( f ∗ , f ∗ , f ! ) : H → H ′ induces an adjoint pair of internal modalities ♭ f ⊣ ♯ f \flat_f \dashv\sharp_f ♭ f ⊣ ♯ f , where
♯ f : = f ! ∘ f ∗ : H ⟶ H \sharp_f := f^! \circ f_* : \mathbf{H} \longrightarrow\mathbf{H} ♯ f := f ! ∘ f ∗ : H ⟶ H
is called the f f f -sharp modality (an idempotent monad when f f f is local).
More generally, in a locally cohesive (and in particular a cohesive) geometric morphism ( f ! ⊣ f ∗ ⊣ f ∗ ⊣ f ! ) : H → H ′ (f_! \dashv f^* \dashv f_* \dashv f^!) : \mathbf{H} \to\mathbf{H}' ( f ! ⊣ f ∗ ⊣ f ∗ ⊣ f ! ) : H → H ′ , the adjoint quadruple induces an adjoint triple of internal modalities on H \mathbf{H} H .5
∫ f ⊣ ♭ f ⊣ ♯ f : H ⟶ H . \int_f \dashv\flat_f \dashv\sharp_f : \mathbf{H} \longrightarrow\mathbf{H}. ∫ f ⊣ ♭ f ⊣ ♯ f : H ⟶ H .
Figure 2 (PDF p. 24).
Recall that G r p d 2 ≃ ( 2 , 1 ) P S h v G r p d 2 ( ∗ ) \mathbf{Grpd}_2 \simeq(2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(*) Grpd 2 ≃ ( 2 , 1 ) PShv Grpd 2 ( ∗ ) is the terminal object in the 2 2 2 -category of ( 2 , 1 ) (2,1) ( 2 , 1 ) -presheaf topoi ( 2 , 1 ) P S h v T o p o s 2 (2,1)\mathbf{PShvTopos}_2 ( 2 , 1 ) PShvTopos 2 . Specifically, for every H = ( 2 , 1 ) P S h v G r p d 2 ( C ) \mathbf{H} = (2,1)\mathbf{PShv}_{\mathbf{Grpd}_2}(\mathbf{C}) H = ( 2 , 1 ) PShv Grpd 2 ( C ) , there is an essentially unique terminal geometric morphism
( Γ , Cst ) = ( El , Cst ) : H ⟶ G r p d 2 , (\Gamma, \operatorname{Cst}) = (\operatorname{El}, \operatorname{Cst}) : \mathbf{H} \longrightarrow\mathbf{Grpd}_2, ( Γ , Cst ) = ( El , Cst ) : H ⟶ Grpd 2 ,
whose direct image is the global sections (or global elements ) 2 2 2 -functor Γ = El : = H ( ∗ H , − ) ≃ lim C o p : H → G r p d 2 \Gamma= \operatorname{El} := \mathbf{H}(*_{\mathbf{H}}, -) \simeq\lim_{\mathbf{C}^{\mathrm{op}}} : \mathbf{H} \to\mathbf{Grpd}_2 Γ = El := H ( ∗ H , − ) ≃ lim C op : H → Grpd 2 , and whose inverse image is the constant presheaf 2 2 2 -functor Cst : G r p d 2 → H \operatorname{Cst} : \mathbf{Grpd}_2 \to\mathbf{H} Cst : Grpd 2 → H sending K ∈ G r p d 2 K \in\mathbf{Grpd}_2 K ∈ Grpd 2 to the constant presheaf U ↦ K U \mapsto K U ↦ K . Since the stackification localization L J : H → H 0 = Stk ( C , J ) L_J : \mathbf{H} \to\mathbf{H}_0 = \operatorname{Stk}(\mathbf{C}, J) L J : H → H 0 = Stk ( C , J ) defining a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos is left exact, G r p d 2 \mathbf{Grpd}_2 Grpd 2 remains the terminal object in ( 2 , 1 ) S h v T o p o s 2 (2,1)\mathbf{ShvTopos}_2 ( 2 , 1 ) ShvTopos 2 , with terminal geometric morphism ( Γ , L Cst ) : H 0 → G r p d 2 (\Gamma, L\operatorname{Cst}) : \mathbf{H}_0 \to\mathbf{Grpd}_2 ( Γ , L Cst ) : H 0 → Grpd 2 given by
Γ : = H 0 ( ∗ H 0 , − ) = El ∘ ι , L Cst : = L J ∘ Cst . \Gamma:= \mathbf{H}_0(*_{\mathbf{H}_0}, -) = \operatorname{El} \circ\iota, \qquad L\operatorname{Cst} := L_J \circ\operatorname{Cst}. Γ := H 0 ( ∗ H 0 , − ) = El ∘ ι , L Cst := L J ∘ Cst .
A Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H is called locally $1$-connected , strongly 1 1 1 -connected , locally local , local , locally cohesive , or cohesive (over G r p d 2 \mathbf{Grpd}_2 Grpd 2 ) if its terminal geometric morphism $(\Gamma, L\operatorname{Cst}) : H → G r p d 2 \mathbf{H} \to\mathbf{Grpd}_2 H → Grpd 2 has the corresponding property, with extra adjoints denoted Π ⊣ L C s t ⊣ Γ ⊣ c o D i s c \Pi\dashv\mathrm{LCst} \dashv\Gamma\dashv\mathrm{coDisc} Π ⊣ LCst ⊣ Γ ⊣ coDisc . In a cohesive Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , we work with the canonical modalities
∫ = L C s t ∘ Π ⊣ ♭ = L C s t ∘ Γ ⊣ ♯ = c o D i s c ∘ Γ \int= \mathrm{LCst}\circ\Pi\dashv\flat= \mathrm{LCst}\circ\Gamma\dashv\sharp= \mathrm{coDisc}\circ\Gamma ∫ = LCst ∘ Π ⊣ ♭ = LCst ∘ Γ ⊣ ♯ = coDisc ∘ Γ
induced by the terminal geometric morphism.
Geometric Structures
In the archetypal model for classical gauge theories—the cohesive Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H = S m o o t h G r p d 2 = S t k ( C a r t S p , J o p e n ) \mathbf{H}=\mathbf{SmoothGrpd}_2=\mathbf{Stk}(\mathbf{CartSp},J_{\mathrm{open}}) H = SmoothGrpd 2 = Stk ( CartSp , J open ) of smooth 1 1 1 -groupoids (stacks on the site of Cartesian spaces R n \mathbb{R}^n R n and smooth maps with the open cover topology [38 , 10 ] )—the three cohesive modalities play the following distinct geometric roles:
The shape modality ∫ \int ∫ forgets smooth structure down to fundamental path groupoids (homotopy types) and is used to formalize principal bundles and their topological realization;
The flat modality ♭ \flat ♭ sends a smooth stack to the constant stack on its underlying discrete groupoid of points and is used to formalize flat differential forms and local systems;
The sharp modality ♯ \sharp ♯ sends a smooth stack to the codiscrete smooth stack on its points (into which any smooth plot maps uniquely once it maps on bare points) and is used to adjust internal mapping stacks into differential moduli spaces .
Consequently, their combinations yield:
Shape + Flat (∫ ⊣ ♭ \int\dashv\flat ∫ ⊣ ♭ ): formalizes universal Maurer–Cartan forms , connection forms , and curvature forms on principal bundles;
Shape + Flat + Sharp (∫ ⊣ ♭ ⊣ ♯ \int\dashv\flat\dashv\sharp ∫ ⊣ ♭ ⊣ ♯ ): formalizes the internal differential moduli stacks of connection forms G C o n n k ( X ) ∈ H \mathrm{GConn}^k(X)\in\mathbf{H} GConn k ( X ) ∈ H .
Because this internal moduli stack of connection forms serves as the domain of the action functional exp ( i S ) : G C o n n k ( X ) → B n U ( 1 ) \exp(iS):\mathrm{GConn}^k(X)\to B^nU(1) exp ( i S ) : GConn k ( X ) → B n U ( 1 ) defining a classical gauge theory, the synthetic dictionary above shows that the kinematics of gauge theories can be completely internalized inside any cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos [38 , 39 ] . (Adding infinitesimal/jet modalities further internalizes variational calculus and field dynamics [19 ] ; here we focus on the kinematic and cohomological structures.) For a comprehensive ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categorical treatment, see [38 Sections 3.7–3.9, 5.1–5.2] .
Bundles: ∫ \int ∫
Let T o p 2 = d i s c 1 , 2 ( T o p ) \mathbf{Top}_2=\mathrm{disc}_{1,2}(\mathbf{Top}) Top 2 = disc 1 , 2 ( Top ) denote the 1 1 1 -category of (compactly generated weak Hausdorff) topological spaces and continuous maps viewed as a discrete ( 2 , 1 ) (2,1) ( 2 , 1 ) -category, and let 2 T o p 2 2\mathbf{Top}_2 2 Top 2 denote the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category of topological spaces, continuous maps, and homotopy classes of homotopies between them; its 1 1 1 -truncation τ ≤ 1 ( 2 T o p 2 ) \tau_{\leq1}(2\mathbf{Top}_2) τ ≤ 1 ( 2 Top 2 ) is the classical homotopy 1 1 1 -category H o ( T o p ) \mathrm{Ho}(\mathbf{Top}) Ho ( Top ) . In classical bundle theory, isomorphism classes of principal G G G -bundles over a paracompact space X X X are in natural bijection Iso G ( X ) ≅ [ X , B G ] \operatorname{Iso}_G(X)\cong[X,BG] Iso G ( X ) ≅ [ X , B G ] via pullback of the universal principal G G G -bundle u G : E G → B G u_G:EG\to BG u G : E G → B G .
This universal bundle admits a direct interpretation in terms of internal groupoids. Let G r p d 2 ( T o p 2 ) \mathbf{Grpd}_2(\mathbf{Top}_2) Grpd 2 ( Top 2 ) be the ( 2 , 1 ) (2,1) ( 2 , 1 ) -category of topological groupoids, equipped with the forgetful functor to G r p d 2 ( S e t 2 ) = G r p d 2 \mathbf{Grpd}_2(\mathbf{Set}_2)=\mathbf{Grpd}_2 Grpd 2 ( Set 2 ) = Grpd 2 . There is a classical geometric realization 2 2 2 -functor
∣ − ∣ : G r p d 2 ⟶ 2 T o p 2 ( or ∣ − ∣ T o p 2 : G r p d 2 ( T o p 2 ) ⟶ 2 T o p 2 ) , |-|:\mathbf{Grpd}_2\longrightarrow2\mathbf{Top}_2
\qquad
(\text{or }|-|_{\mathbf{Top}_2}:\mathbf{Grpd}_2(\mathbf{Top}_2)\longrightarrow2\mathbf{Top}_2), ∣ − ∣ : Grpd 2 ⟶ 2 Top 2 ( or ∣ − ∣ Top 2 : Grpd 2 ( Top 2 ) ⟶ 2 Top 2 ) ,
whose essential image lands in homotopy 1 1 1 -types (CW complexes with π 1 ( − ) = 0 \pi_1(-)=0 π 1 ( − ) = 0 , i.e., disjoint unions of Eilenberg–Mac Lane spaces of type K ( π , 1 ) K(\pi,1) K ( π , 1 ) ). To every topological group G ∈ G r p d 2 ( T o p 2 ) G\in\mathbf{Grpd}_2(\mathbf{Top}_2) G ∈ Grpd 2 ( Top 2 ) , one associates the internal action (or pair) groupoid E G = ( G × G ⇉ G ) EG=(G\times G\rightrightarrows G) E G = ( G × G ⇉ G ) (which is equivalent to the terminal groupoid ∗ * ∗ , and should not be confused with the discrete groupoid E x = ( X ⇉ X ) \mathrm{Ex}=(X\rightrightarrows X) Ex = ( X ⇉ X ) of Section § 2.3 ) and the one-object internal delooping groupoid B T o p 2 i n t ( G ) = ( G ⇉ ∗ ) \mathbf{B}^{\mathrm{int}}_{\mathbf{Top}_2}(G)=(G\rightrightarrows*) B Top 2 int ( G ) = ( G ⇉ ∗ ) , linked by the canonical quotient functor μ G : E G → B T o p 2 i n t ( G ) \mu_G:EG\to\mathbf{B}^{\mathrm{int}}_{\mathbf{Top}_2}(G) μ G : E G → B Top 2 int ( G ) (or, on object-of-objects inclusions, d i s c 1 , 2 ( ∗ ) → B T o p 2 i n t ( G ) \mathrm{disc}_{1,2}(*)\to\mathbf{B}^{\mathrm{int}}_{\mathbf{Top}_2}(G) disc 1 , 2 ( ∗ ) → B Top 2 int ( G ) ). Its geometric realization ∣ μ G ∣ |\mu_G| ∣ μ G ∣ in 2 T o p 2 2\mathbf{Top}_2 2 Top 2 is precisely the universal principal G G G -bundle u G : E G → B G u_G:EG\to BG u G : E G → B G .
On the other hand, in a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , every internal groupoid with 0 0 0 -truncated morphism fibers is effective and thus completely determined by its representing quotient object. Hence the universal G G G -bundle is axiomatized internally in any cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathbf{H} H simply by the canonical basepoint morphism ∗ ↠ B H G \ast\twoheadrightarrow B_{\mathbf{H}}G ∗ ↠ B H G , and principal G G G -bundles over X ∈ H X \in\mathbf{H} X ∈ H are axiomatized as homotopy pullbacks of ∗ ↠ B H G \ast\twoheadrightarrow B_{\mathbf{H}}G ∗ ↠ B H G .
P → ∗ π ↓ ↓ p G X → f B H G \begin{CD}
P @>>> \ast\\
@V\pi VV @VV p_G V \\
X @>f>> B_{\mathbf{H}}G
\end{CD} P π ↓ ⏐ X f ∗ ↓ ⏐ p G B H G
Definition 4.1 (Abstract Principal G G G -Bundle). Let H \mathbf{H} H be a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos and let G ∈ Grp 2 ( H ) G \in\operatorname{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) be a group object. A 1 1 1 -morphism f : X → B H G f : X \to B_{\mathbf{H}}G f : X → B H G in H \mathbf{H} H is said to classify an abstract principal G G G -bundle over X X X . The principal G G G -bundle classified by f f f is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback morphism π : P → X \pi: P \to X π : P → X in the diagram above (equipped with its canonical G G G -action inherited from Ω B H G ≃ G \Omega B_{\mathbf{H}}G \simeq G Ω B H G ≃ G ).
To relate an “abstract bundle” P → X P \to X P → X in H \mathbf{H} H to a concrete topological bundle in 2 T o p 2 2\mathbf{Top}_2 2 Top 2 , we require an abstract geometric realization 2 2 2 -functor ∣ − ∣ H : H → 2 T o p 2 \lvert- \rvert_{\mathbf{H}} : \mathbf{H} \to2\mathbf{Top}_2 ∣ − ∣ H : H → 2 Top 2 that sends abstract universal G G G -bundles to concrete universal bundles. Thus, ∣ − ∣ H \lvert- \rvert_{\mathbf{H}} ∣ − ∣ H must preserve group objects, terminal objects, and deloopings objects (which are ( 2 , 1 ) (2,1) ( 2 , 1 ) -coequalizers), and should factor through the classical geometric realization ∣ − ∣ : G r p d 2 → 2 T o p 2 \lvert- \rvert: \mathbf{Grpd}_2 \to2\mathbf{Top}_2 ∣ − ∣ : Grpd 2 → 2 Top 2 via a 2 2 2 -functor F : H → G r p d 2 F : \mathbf{H} \to\mathbf{Grpd}_2 F : H → Grpd 2 :
H → F G r p d 2 → ∣ − ∣ 2 T o p 2 ∣ − ∣ H ↓ (12) \begin{CD}
\mathbf{H} @>F>> \mathbf{Grpd}_2 @>{\lvert-\rvert}>> 2\mathbf{Top}_2 \\
@V{\lvert-\rvert_{\mathbf{H}}}VV \tag*{(12)}
\end{CD} H ∣ − ∣ H ↓ ⏐ F Grpd 2 ∣ − ∣ 2 Top 2 (12)
Since the classical geometric realization ∣ − ∣ : G r p d 2 → 2 T o p 2 \lvert- \rvert: \mathbf{Grpd}_2 \to2\mathbf{Top}_2 ∣ − ∣ : Grpd 2 → 2 Top 2 is left adjoint to the fundamental groupoid 2 2 2 -functor Π 1 : 2 T o p 2 → G r p d 2 \Pi_1 : 2\mathbf{Top}_2 \to\mathbf{Grpd}_2 Π 1 : 2 Top 2 → Grpd 2 and also preserves finite products, ∣ − ∣ \lvert- \rvert ∣ − ∣ preserves both ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits and finite products. Therefore, ∣ − ∣ H = ∣ − ∣ ∘ F \lvert- \rvert_{\mathbf{H}} = \lvert- \rvert\circ F ∣ − ∣ H = ∣ − ∣ ∘ F will preserve group objects, terminal objects, and deloopings provided F : H → G r p d 2 F : \mathbf{H} \to\mathbf{Grpd}_2 F : H → Grpd 2 preserves finite products and ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits.
In a strongly 1 1 1 -connected Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , the extra left adjoint Π : H → G r p d 2 \Pi: \mathbf{H} \to\mathbf{Grpd}_2 Π : H → Grpd 2 to the terminal inverse image L C s t \mathrm{LCst} LCst preserves all ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits (being a left adjoint) and preserves finite products (by definition of strong 1 1 1 -connectivity). This motivates setting F = Π F = \Pi F = Π :
H → Π G r p d 2 → ∣ − ∣ 2 T o p 2 L C s t ↑ \begin{CD}
\mathbf{H} @>{\Pi}>> \mathbf{Grpd}_2 @>{\lvert-\rvert}>> 2\mathbf{Top}_2 \\
@A{\mathrm{LCst}}AA
\end{CD} H LCst ⏐ ↑ Π Grpd 2 ∣ − ∣ 2 Top 2
Note that the internal endofunctor corresponding to Π \Pi Π inside H \mathbf{H} H is the shape modality \Imath= \mathrm{LCst}\circ\Pi .
Definition 4.2 (Topological Realization). Let H \mathbf{H} H be a strongly 1 1 1 -connected Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and let G ∈ Grp 2 ( H ) G \in\operatorname{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) be a group object. The topological realization of an abstract G G G -bundle P → X P \to X P → X classified by f : X → B H G f : X \to B_{\mathbf{H}}G f : X → B H G is the topological principal ∣ G ∣ H \lvert G\rvert_{\mathbf{H}} ∣ G ∣ H -bundle over ∣ X ∣ H \lvert X\rvert_{\mathbf{H}} ∣ X ∣ H classified by
∣ f ∣ H : ∣ X ∣ H ⟶ ∣ B H G ∣ H ≃ B ∣ G ∣ H \lvert f\rvert_{\mathbf{H}} : \lvert X\rvert_{\mathbf{H}} \longrightarrow\lvert B_{\mathbf{H}}G\rvert_{\mathbf{H}} \simeq B\lvert G\rvert_{\mathbf{H}} ∣ f ∣ H : ∣ X ∣ H ⟶ ∣ B H G ∣ H ≃ B ∣ G ∣ H .
Remark 4.3. Observe that, a priori, the topological bundle classified by ∣ f ∣ H \lvert f\rvert_{\mathbf{H}} ∣ f ∣ H (which is the pullback of E ∣ G ∣ H → B ∣ G ∣ H E\lvert G\rvert_{\mathbf{H}} \to B\lvert G\rvert_{\mathbf{H}} E ∣ G ∣ H → B ∣ G ∣ H along ∣ f ∣ H \lvert f\rvert_{\mathbf{H}} ∣ f ∣ H ) need not coincide with ∣ P ∣ H → ∣ X ∣ H \lvert P\rvert_{\mathbf{H}} \to\lvert X\rvert_{\mathbf{H}} ∣ P ∣ H → ∣ X ∣ H for an arbitrary classifying map f f f . Indeed, while ∣ − ∣ : G r p d 2 → 2 T o p 2 \lvert- \rvert: \mathbf{Grpd}_2 \to2\mathbf{Top}_2 ∣ − ∣ : Grpd 2 → 2 Top 2 preserves finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits, the shape functor Π : H → G r p d 2 \Pi: \mathbf{H} \to\mathbf{Grpd}_2 Π : H → Grpd 2 of a strongly 1 1 1 -connected ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos is only guaranteed to preserve finite products, not arbitrary ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks. However, on smooth manifolds and locally trivial fiber bundles (where Π \Pi Π commutes with pullbacks along submersion/fibration projections by the Seifert–van Kampen theorem for Π \Pi Π ), ∣ P ∣ H → ∣ X ∣ H \lvert P\rvert_{\mathbf{H}} \to\lvert X\rvert_{\mathbf{H}} ∣ P ∣ H → ∣ X ∣ H does agree with the pullback bundle [38 Proposition 4.3.13] . A ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos for which Π \Pi Π preserves all finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits is called totally 1 1 1 -connected (or Π \Pi Π -lex).
Replacing group objects with k k k -tuple deloopable objects (with B H k G B_{\mathbf{H}}^{k}G B H k G understood in the lossless sense of Remark 3.3) yields higher bundles (such as bundle gerbes for k = 2 k=2 k = 2 ):
Definition 4.4 (Abstract k k k -Bundles). Let H \mathbf{H} H be a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos and let G ∈ Del 2 k ( H ) G \in\operatorname{Del}_{2}^{k}(\mathbf{H}) G ∈ Del 2 k ( H ) be a k k k -tuply deloopable object (k ≥ 1 k \ge1 k ≥ 1 ). With respect to ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation, for k = 1 k=1 k = 1 , B H G ∈ H B_{\mathbf{H}}G \in\mathbf{H} B H G ∈ H is internal to H \mathbf{H} H and classifies ordinary principal G G G -bundles; for k = 2 k=2 k = 2 and A ∈ ComGrp ( τ ≤ 0 H ) A \in\operatorname{ComGrp}(\tau_{\le0}\mathbf{H}) A ∈ ComGrp ( τ ≤ 0 H ) , B H 2 A = B H i n t ( B H A ) ∈ Grpd 2 ( H ) B_{\mathbf{H}}^{2}A = B_{\mathbf{H}}^{\mathrm{int}}(B_{\mathbf{H}}A) \in\operatorname{Grpd}_{2}(\mathbf{H}) B H 2 A = B H int ( B H A ) ∈ Grpd 2 ( H ) is an internal 2 2 2 -stack in H \mathbf{H} H classifying A A A -bundle gerbes; while for k ≥ 3 k \ge3 k ≥ 3 , B H k G B_{\mathbf{H}}^{k}G B H k G is an object of the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos H ∞ \mathbf{H}_{\infty} H ∞ with 1 1 1 -truncated mapping groupoid H ( X , B H k G ) : = τ ≤ 1 H ∞ ( X , B k G ) \mathbf{H}(X,B_{\mathbf{H}}^{k}G) := \tau_{\le1}\mathbf{H}_{\infty}(X,B^{k}G) H ( X , B H k G ) := τ ≤ 1 H ∞ ( X , B k G ) . An abstract G G G -k k k -bundle over X ∈ H X \in\mathbf{H} X ∈ H is a morphism P k → X P_k \to X P k → X obtained as the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback of ∗ → B H k G * \to B_{\mathbf{H}}^{k}G ∗ → B H k G along a classifying morphism f : X → B H k G f:X \to B_{\mathbf{H}}^{k}G f : X → B H k G :
P k → ∗ ↓ ↓ X → f B H k G \begin{CD}
P_k @>>> * \\
@VVV @VVV \\
X @>{f}>> B_{\mathbf{H}}^{k}G
\end{CD} P k ↓ ⏐ X f ∗ ↓ ⏐ B H k G When H \mathbf{H} H is a strongly 1 1 1 -connected ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, the topological realization of P k → X P_k \to X P k → X is the topological k k k -bundle over ∣ X ∣ H |X|_{\mathbf{H}} ∣ X ∣ H classified by ∣ f ∣ H : ∣ X ∣ H → ∣ B H k G ∣ H ≃ B k ∣ G ∣ H |f|_{\mathbf{H}}:|X|_{\mathbf{H}} \to|B_{\mathbf{H}}^{k}G|_{\mathbf{H}} \simeq B^{k}|G|_{\mathbf{H}} ∣ f ∣ H : ∣ X ∣ H → ∣ B H k G ∣ H ≃ B k ∣ G ∣ H .
Let ( Γ , L C s t ) : H → Grpd 2 (\Gamma,\mathrm{LCst}):\mathbf{H}\to\operatorname{Grpd}_{2} ( Γ , LCst ) : H → Grpd 2 be a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and let ♭ = L C s t ∘ Γ : H → H \flat=\mathrm{LCst}\circ\Gamma:\mathbf{H}\to\mathbf{H} ♭ = LCst ∘ Γ : H → H be its flat modality. The counit of the adjunction L C s t ⊣ Γ \mathrm{LCst}\dashv\Gamma LCst ⊣ Γ defines a canonical 2 2 2 -natural transformation ϵ : ♭ ⇒ id H \epsilon:\flat\Rightarrow\operatorname{id}_{\mathbf{H}} ϵ : ♭ ⇒ id H , assigning to every object A ∈ H A\in\mathbf{H} A ∈ H the canonical inclusion of constant points ϵ A : ♭ A → A \epsilon_A:\flat A\to A ϵ A : ♭ A → A .
Whenever A A A is a pointed object ∗ → A *\to A ∗ → A (and in particular whenever A ∈ Grp 2 ( H ) A\in\operatorname{Grp}_{2}(\mathbf{H}) A ∈ Grp 2 ( H ) or A = B H k G A=B_{\mathbf{H}}^{k}G A = B H k G ), let i A : ♭ d R A → ♭ A i_A:\flat_{\mathrm{dR}}A\to\flat A i A : ♭ dR A → ♭ A denote the homotopy kernel (homotopy fiber over the basepoint) of ϵ A : ♭ A → A \epsilon_A:\flat A\to A ϵ A : ♭ A → A , defined by the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback square:
♭ d R A → i A ♭ A ↓ ↓ ϵ A ∗ H → A \begin{CD}
\flat_{\mathrm{dR}}A @>{i_A}>> \flat A \\
@VVV @VV{\epsilon_A}V \\
*_{\mathbf{H}} @>>> A
\end{CD} ♭ dR A ↓ ⏐ ∗ H i A ♭ A ↓ ⏐ ϵ A A
We call ♭ d R A \flat_{\mathrm{dR}}A ♭ dR A the de Rham refinement (or de Rham coefficient object ) of A A A . Varying A ∈ ∗ / H A\in*/\mathbf{H} A ∈ ∗ / H defines a 2 2 2 -functor
♭ d R : ∗ / H ⟶ ∗ / H \flat_{\mathrm{dR}}:*/\mathbf{H}\longrightarrow*/\mathbf{H} ♭ dR : ∗ / H ⟶ ∗ / H
(or H ⟶ H \mathbf{H}\longrightarrow\mathbf{H} H ⟶ H for canonically pointed coefficients), called the flat de Rham modality [38 ] . The terminology stems from the fact that for a group object G G G , by pasting the pullback square for ♭ d R B G \flat_{\mathrm{dR}}BG ♭ dR B G with the looping pullback G ≃ Ω B G = ∗ × B G ∗ G\simeq\Omega BG=*\times_{BG}* G ≃ Ω B G = ∗ × B G ∗ , ♭ d R B G \flat_{\mathrm{dR}}BG ♭ dR B G is equivalently the homotopy cofiber (quotient action groupoid) G / / ♭ G G//\flat G G //♭ G :
♭ d R B H G ≃ B H ( ♭ G → G ) , \flat_{\mathrm{dR}}B_{\mathbf{H}}G\simeq B_{\mathbf{H}}(\flat G\to G), ♭ dR B H G ≃ B H ( ♭ G → G ) ,
which in SmoothGrpd 2 \operatorname{SmoothGrpd}_{2} SmoothGrpd 2 is weakly equivalent via the Maurer--Cartan map to the stack of flat Lie ( G ) \operatorname{Lie}(G) Lie ( G ) -valued 1 1 1 -forms Ω f l a t 1 ( − ; g ) / / G \Omega_{\mathrm{flat}}^{1}(-;\mathfrak{g})//G Ω flat 1 ( − ; g ) // G . More generally, if A ∈ Del 2 k ( H ) A\in\operatorname{Del}_{2}^{k}(\mathbf{H}) A ∈ Del 2 k ( H ) is k k k -tuply deloopable (k ≥ 1 k\ge1 k ≥ 1 ), then ♭ d R B H k A \flat_{\mathrm{dR}}B_{\mathbf{H}}^{k}A ♭ dR B H k A acts as the moduli stack of flat A A A -valued differential k k k -forms .
To pass from flat forms to closed forms (which serve as curvature targets for non-flat connections), we observe that on any test object X X X , every flat form is a closed form, and locally (on contractible test spaces) every closed k k k -form is the exterior derivative/curvature of a ( k − 1 ) (k-1) ( k − 1 ) -form potential:
Definition 4.5 (Moduli Stack of Closed Differential k k k -Forms [38 Section 3.9.6] ). Let H \mathbf{H} H be a cohesive Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and let A ∈ Del 2 k ( H ) A\in\operatorname{Del}_{2}^{k}(\mathbf{H}) A ∈ Del 2 k ( H ) be a k k k -tuply deloopable object (where Ω c l k ( − ; A ) \Omega_{\mathrm{cl}}^{k}(-;A) Ω cl k ( − ; A ) and ♭ d R B H k A \flat_{\mathrm{dR}}B_{\mathbf{H}}^{k}A ♭ dR B H k A live internal to H \mathbf{H} H for k = 1 k=1 k = 1 , as internal 2 2 2 -stacks in Grpd 2 ( H ) \operatorname{Grpd}_{2}(\mathbf{H}) Grpd 2 ( H ) for k = 2 k=2 k = 2 , and in the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos H ∞ \mathbf{H}_{\infty} H ∞ for k ≥ 3 k\ge3 k ≥ 3 ). A moduli stack of closed A A A -valued k k k -forms (or a differential coefficients specification ) is an object Ω c l k ( − ; A ) \Omega_{\mathrm{cl}}^{k}(-;A) Ω cl k ( − ; A ) equipped with a morphism σ k : Ω c l k ( − ; A ) → ♭ d R B H k A \sigma_k:\Omega_{\mathrm{cl}}^{k}(-;A)\to\flat_{\mathrm{dR}}B_{\mathbf{H}}^{k}A σ k : Ω cl k ( − ; A ) → ♭ dR B H k A such that:
Ω c l k ( − ; A ) \Omega_{\mathrm{cl}}^{k}(-;A) Ω cl k ( − ; A ) is 0 0 0 -truncated (a sheaf of sets) for abelian A A A (or a 1 1 1 -truncated action groupoid Ω c l 1 ( − ; a ) / / A \Omega_{\mathrm{cl}}^{1}(-;\mathfrak{a})//A Ω cl 1 ( − ; a ) // A when twisting by an adjoint action on the linearized/infinitesimal coefficient object a = Lie ( A ) \mathfrak{a}=\operatorname{Lie}(A) a = Lie ( A ) in concrete models such as SmoothGrpd 2 \operatorname{SmoothGrpd}_{2} SmoothGrpd 2 ) and shape-contractible (∫ Ω c l k ( − ; A ) ≃ ∗ \int\Omega_{\mathrm{cl}}^{k}(-;A)\simeq* ∫ Ω cl k ( − ; A ) ≃ ∗ );
(2) The homotopy fiber of σ k : Ω c l k ( − ; A ) → d R B H k A \sigma_k:\Omega^k_{\mathrm{cl}}(-;A)\to{}_{\mathrm{dR}}B^k_{\mathrm{H}}A σ k : Ω cl k ( − ; A ) → dR B H k A over the basepoint classifies ( k − 1 ) (k-1) ( k − 1 ) -form potentials Ω k − 1 ( − ; a ) \Omega^{k-1}(-;\mathfrak{a}) Ω k − 1 ( − ; a ) (via the de Rham differential d : d R B H k − 1 A → Ω c l k ( − ; A ) d:{}_{\mathrm{dR}}B^{k-1}_{\mathrm{H}}A\to\Omega^k_{\mathrm{cl}}(-;A) d : dR B H k − 1 A → Ω cl k ( − ; A ) ), and on every test object U U U in a generating family of contractible test spaces (where H ( U , b B H k A ) ≃ ∗ \mathrm{H}(U,{}_{\mathrm{b}}B^k_{\mathrm{H}}A)\simeq* H ( U , b B H k A ) ≃ ∗ ), pulling back along σ k \sigma_k σ k exhibits local exactness of closed forms (the Poincaré lemma).
In classical differential geometry, a principal connection on a principal G G G -bundle P → X P\to X P → X is specified either globally as a G G G -equivariant vertical g \mathfrak{g} g -valued 1-form ω ∈ Ω 1 ( P ; g ) G \omega\in\Omega^1(P;\mathfrak{g})^G ω ∈ Ω 1 ( P ; g ) G , or locally over a cover { U i → X } \{U_i\to X\} { U i → X } by local gauge potentials A i ∈ Ω 1 ( U i ; g ) A_i\in\Omega^1(U_i;\mathfrak{g}) A i ∈ Ω 1 ( U i ; g ) transforming on overlaps U i ∩ U j U_i\cap U_j U i ∩ U j via
A j = Ad g i j − 1 ( A i ) + g i j ∗ θ G , θ G = g − 1 d g ∈ Ω f l a t 1 ( G ; g ) . (13) A_j=\operatorname{Ad}_{g_{ij}^{-1}}(A_i)+g_{ij}^{*}\theta_G,\qquad\theta_G=g^{-1}dg\in\Omega^1_{\mathrm{flat}}(G;\mathfrak{g}). \tag*{(13)} A j = Ad g ij − 1 ( A i ) + g ij ∗ θ G , θ G = g − 1 d g ∈ Ω flat 1 ( G ; g ) . (13)
Here θ G = g − 1 d g \theta_G=g^{-1}dg θ G = g − 1 d g is the canonical left-invariant Maurer–Cartan form on the Lie group G G G associated with the left action g ⋅ x g\cdot x g ⋅ x (under the convention where gauge transformations act on potentials as A ↦ g − 1 A g + g − 1 d g A\mapsto g^{-1}Ag+g^{-1}dg A ↦ g − 1 A g + g − 1 d g ).
Remarkably, in any Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , every group object G ∈ Grp 2 ( H ) G\in\operatorname{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) carries a canonical abstract Maurer–Cartan morphism
θ G : G ⟶ b d R B H G , \theta_G:G\longrightarrow{}_{\mathrm{b}_{\mathrm{dR}}}B_{\mathbf{H}}G, θ G : G ⟶ b dR B H G ,
constructed purely from the universal property of the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback defining b d R B H G {}_{\mathrm{b}_{\mathrm{dR}}}B_{\mathbf{H}}G b dR B H G . Indeed, since Γ \Gamma Γ is a right adjoint and L C s t \mathrm{LCst} LCst is left exact, ♭ = L C s t ∘ Γ \flat=\mathrm{LCst}\circ\Gamma ♭ = LCst ∘ Γ preserves all finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and, in particular, preserves the terminal object: ♭ ∗ ≃ ∗ \flat*\simeq* ♭ ∗ ≃ ∗ . Applying ♭ \flat ♭ to the basepoint x = p G : ∗ → B G x=p_G:* \to BG x = p G : ∗ → B G gives ♭ x : ∗ ≃ ♭ ∗ → ♭ B G \flat x:* \simeq\flat*\to\flat BG ♭ x : ∗ ≃ ♭ ∗ → ♭ B G . By naturality of ϵ : ♭ ⇒ id H \epsilon:\flat\Rightarrow\operatorname{id}_{\mathbf{H}} ϵ : ♭ ⇒ id H , we have ϵ B G ∘ ♭ x ≃ x ∘ ϵ ∗ = x : ∗ → B G \epsilon_{BG}\circ\flat x\simeq x\circ\epsilon_* =x:* \to BG ϵ B G ∘ ♭ x ≃ x ∘ ϵ ∗ = x : ∗ → B G . Since G ≃ Ω B G = ∗ × B G ∗ G\simeq\Omega BG=*\times_{BG}* G ≃ Ω B G = ∗ × B G ∗ is the homotopy pullback of x : ∗ → B G x:* \to BG x : ∗ → B G along itself equipped with the universal loop 2-cell λ G : x ∘ ! ⇒ x ∘ ! \lambda_G:x\circ!\Rightarrow x\circ! λ G : x ∘ ! ⇒ x ∘ ! , pasting λ G \lambda_G λ G with the naturality 2-isomorphism x ≃ ϵ B G ∘ ♭ x x\simeq\epsilon_{BG}\circ\flat x x ≃ ϵ B G ∘ ♭ x fills the outer square on the right below with a non-trivial 2-isomorphism x ∘ ! ⇒ ϵ B G ∘ ♭ x ∘ ! x\circ!\Rightarrow\epsilon_{BG}\circ\flat x\circ! x ∘ ! ⇒ ϵ B G ∘ ♭ x ∘ ! , inducing the universal Maurer–Cartan morphism θ G : G → b d R B G \theta_G:G\to{}_{\mathrm{b}_{\mathrm{dR}}}BG θ G : G → b dR B G :
b d R B G → ♭ B G ↓ ↓ ϵ B G ∗ → x B G G → ! ♭ ∗ ≃ ∗ θ G ↓ ↓ ♭ x b d R B G → ♭ B G \begin{CD}
{}_{\mathrm{b}_{\mathrm{dR}}}BG @>>> \flat BG\\
@VVV @VV\epsilon_{BG}V\\
* @>{x}>> BG
\end{CD}
\qquad
\begin{CD}
G @>{!}>> \flat*\simeq*\\
@V{\theta_G}VV @VV{\flat x}V\\
{}_{\mathrm{b}_{\mathrm{dR}}}BG @>>> \flat BG
\end{CD} b dR B G ↓ ⏐ ∗ x ♭ B G ↓ ⏐ ϵ B G B G G θ G ↓ ⏐ b dR B G ! ♭ ∗ ≃ ∗ ↓ ⏐ ♭ x ♭ B G
This first Maurer–Cartan morphism θ 1 = θ G : G → b d R B H G \theta_1=\theta_G:G\to{}_{\mathrm{b}_{\mathrm{dR}}}B_{\mathbf{H}}G θ 1 = θ G : G → b dR B H G exists for every (generally nonabelian) group object G G G , without requiring G G G to be braided or B H 2 G B^2_{\mathbf{H}}G B H 2 G to exist. Let us see how θ G \theta_G θ G directly governs nonabelian principal connections:
(1) General Nonabelian Group Objects G ∈ Grp 2 ( H ) G\in\operatorname{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) : For an arbitrary nonabelian group object G G G (such as G = S U ( N ) G=SU(N) G = S U ( N ) or S O ( p , q ) SO(p,q) S O ( p , q ) ), G G G is not braided and B H 2 G B^2_{\mathbf{H}}G B H 2 G does not exist; correspondingly, the curvature F A = d A + 1 2 [ A ∧ A ] F_A=dA+\frac{1}{2}[A\wedge A] F A = d A + 2 1 [ A ∧ A ] of a nonabelian connection is not an invariant closed 2-form on the base X X X , but rather transforms in the adjoint representation (d A F A = 0 d_A F_A=0 d A F A = 0 ) and takes values in the delooping of the inner automorphism 2-group Inn ( G ) = ( G → Ad Aut ( G ) ) \operatorname{Inn}(G)=(G\xrightarrow{\operatorname{Ad}}\operatorname{Aut}(G)) Inn ( G ) = ( G Ad Aut ( G )) [2 , 38 ] . Synthetically in any cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , Inn ( G ) \operatorname{Inn}(G) Inn ( G ) deloops in Grp 2 ( H ) \operatorname{Grp}_2(\mathbf{H}) Grp 2 ( H ) with universal adjoint Maurer–Cartan morphism θ a d j : B H G → b d R B H Inn ( G ) \theta_{\mathrm{adj}}:B_{\mathbf{H}}G\to{}_{\mathrm{b}_{\mathrm{dR}}}B_{\mathbf{H}}\operatorname{Inn}(G) θ adj : B H G → b dR B H Inn ( G ) , and a nonabelian differential 1-form coefficient specification for G G G consists of a closed adjoint curvature 2-form stack σ 2 a d j : Ω c l 2 ( − ; Inn ( G ) ) → b d R B H Inn ( G ) \sigma^{\mathrm{adj}}_2:\Omega^2_{\mathrm{cl}}(-;\operatorname{Inn}(G))\to{}_{\mathrm{b}_{\mathrm{dR}}}B_{\mathbf{H}}\operatorname{Inn}(G) σ 2 adj : Ω cl 2 ( − ; Inn ( G )) → b dR B H Inn ( G ) together with a local 1-form potential stack Ω c o n 1 ( − ; G ) → Ω c l 2 ( − ; INN ( G ) ) \Omega^1_{\mathrm{con}}(-;G) \to\Omega^2_{\mathrm{cl}}(-;\operatorname{INN}(G)) Ω con 1 ( − ; G ) → Ω cl 2 ( − ; INN ( G )) whose (2, 1)-pullback defines the nonabelian moduli stack of principal G G G -connections B c o n G B_{\mathrm{con}}G B con G :
B c o n G → curv Ω c l 2 ( − ; INN ( G ) ) p c o n ↓ ↓ σ 2 a d j B H G → θ a d j ♭ d R B H INN ( G ) (14) \begin{CD}
B_{\mathrm{con}}G @>{\operatorname{curv}}>> \Omega^2_{\mathrm{cl}}(-;\operatorname{INN}(G)) \\
@V{p_{\mathrm{con}}}VV @VV{\sigma^{\mathrm{adj}}_2}V \\
B_{\mathrm{H}}G @>{\theta_{\mathrm{adj}}}>> \flat_{\mathrm{dR}}B_{\mathrm{H}}\operatorname{INN}(G) \tag*{(14)}
\end{CD} B con G p con ↓ ⏐ B H G curv θ adj Ω cl 2 ( − ; INN ( G )) ↓ ⏐ σ 2 adj ♭ dR B H INN ( G ) (14)
such that the homotopy fiber of curv \operatorname{curv} curv over 0 0 0 is ♭ d R B H G ≃ G / / ♭ G → B c o n G \flat_{\mathrm{dR}}B_{\mathrm{H}}G \simeq G//\flat G \to B_{\mathrm{con}}G ♭ dR B H G ≃ G //♭ G → B con G . Concretely, at the level of connection 1-forms in SmoothGrpd 2 _2 2 , the first Maurer–Cartan morphism θ G : G → ♭ d R B H G ≃ G / / ♭ G \theta_G : G \to\flat_{\mathrm{dR}}B_{\mathrm{H}}G \simeq G//\flat G θ G : G → ♭ dR B H G ≃ G //♭ G induces the affine gauge action (13) of G G G on local 1-form potentials Ω 1 ( − ; g ) \Omega^1(-;\mathfrak{g}) Ω 1 ( − ; g ) , presenting B c o n G B_{\mathrm{con}}G B con G as the stackified quotient action groupoid:
B c o n G ≃ L J ( Ω 1 ( − ; g ) / / G ) → p c o n B H G . (15) B_{\mathrm{con}}G \simeq L_J\left(\Omega^1(-;\mathfrak{g})//G\right) \xrightarrow{p_{\mathrm{con}}} B_{\mathrm{H}}G. \tag*{(15)} B con G ≃ L J ( Ω 1 ( − ; g ) // G ) p con B H G . (15)
(2) Braided/Abelian and Higher Deloopable Objects G ∈ Del 2 k + 1 ( H ) G \in\operatorname{Del}^{k+1}_2(\mathbf{H}) G ∈ Del 2 k + 1 ( H ) : When G ∈ Del 2 k ( H ) G \in\operatorname{Del}^{k}_2(\mathbf{H}) G ∈ Del 2 k ( H ) is k k k -tuply deloopable (k ≥ 1 k \ge1 k ≥ 1 ), applying the abstract Maurer–Cartan construction to the group object B H k − 1 G B^{k-1}_{\mathrm{H}}G B H k − 1 G yields the k k k -th universal Maurer–Cartan / universal curvature characteristic morphism :
θ k = θ B k − 1 G : B H k G ⟶ ♭ d R B H k G . \theta_k = \theta_{B^{k-1}G} : B^{k}_{\mathrm{H}}G \longrightarrow\flat_{\mathrm{dR}}B^{k}_{\mathrm{H}}G. θ k = θ B k − 1 G : B H k G ⟶ ♭ dR B H k G .
In particular, for k = 2 k=2 k = 2 , when G ∈ BrGrp 2 ( H ) ≃ Del 2 2 ( H ) G \in\operatorname{BrGrp}_2(\mathbf{H}) \simeq\operatorname{Del}^{2}_2(\mathbf{H}) G ∈ BrGrp 2 ( H ) ≃ Del 2 2 ( H ) is braided/abelian, the adjoint action is trivial so INN ( G ) ≃ B H G \operatorname{INN}(G) \simeq B_{\mathrm{H}}G INN ( G ) ≃ B H G deloops to B H 2 G B^2_{\mathrm{H}}G B H 2 G in the sense of Remark 3.3 (or when B H G B_{\mathrm{H}}G B H G is mapped to an abelian target via a Chern–Weil invariant polynomial); then θ a d j \theta_{\mathrm{adj}} θ adj reduces to θ 2 : B H G → ♭ d R B H 2 G \theta_2 : B_{\mathrm{H}}G \to\flat_{\mathrm{dR}}B^2_{\mathrm{H}}G θ 2 : B H G → ♭ dR B H 2 G , and (14) reduces directly to the (2, 1)-pullback of θ 2 \theta_2 θ 2 along σ 2 : Ω c l 2 ( − ; G ) → ♭ d R B H 2 G \sigma_2 : \Omega^2_{\mathrm{cl}}(-;G) \to\flat_{\mathrm{dR}}B^2_{\mathrm{H}}G σ 2 : Ω cl 2 ( − ; G ) → ♭ dR B H 2 G , as formalized in the next definition.
Definition 4.6 (Moduli Stack of Principal Connections). Let H \mathbf{H} H be a Grothendieck (2,1)-topos. For a general nonabelian group object G ∈ Grp 2 ( H ) G \in\operatorname{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) , the moduli stack of principal G G G -connections B c o n G = B c o n 1 G ∈ H B_{\mathrm{con}}G = B^1_{\mathrm{con}}G \in\mathbf{H} B con G = B con 1 G ∈ H is the (2,1)-pullback (14) over p c o n : B c o n G → B H G p_{\mathrm{con}} : B_{\mathrm{con}}G \to B_{\mathrm{H}}G p con : B con G → B H G (presented on local potentials by (15) ); when G ∈ BrGrp 2 ( H ) ≃ Del 2 2 ( H ) G \in\operatorname{BrGrp}_2(\mathbf{H}) \simeq\operatorname{Del}^{2}_2(\mathbf{H}) G ∈ BrGrp 2 ( H ) ≃ Del 2 2 ( H ) is braided (or equipped with a Chern–Weil curvature characteristic θ 2 : B H G → ♭ d R B H 2 G \theta_2 : B_{\mathrm{H}}G \to\flat_{\mathrm{dR}}B^2_{\mathrm{H}}G θ 2 : B H G → ♭ dR B H 2 G and closed curvature 2-forms Ω c l 2 ( − ; G ) \Omega^2_{\mathrm{cl}}(-;G) Ω cl 2 ( − ; G ) ), (14) simplifies to the (2,1)-pullback:
B c o n G → curv Ω c l 2 ( − ; G ) p c o n ↓ ↓ σ 2 B H G → θ 2 ♭ d R B H 2 G \begin{CD}
B_{\mathrm{con}}G @>{\operatorname{curv}}>> \Omega^2_{\mathrm{cl}}(-;G) \\
@V{p_{\mathrm{con}}}VV @VV{\sigma_2}V \\
B_{\mathrm{H}}G @>{\theta_2}>> \flat_{\mathrm{dR}}B^2_{\mathrm{H}}G
\end{CD} B con G p con ↓ ⏐ B H G curv θ 2 Ω cl 2 ( − ; G ) ↓ ⏐ σ 2 ♭ dR B H 2 G
Given an abstract principal G G G -bundle P → X P \to X P → X classified by f : X → B H G f : X \to B_{\mathrm{H}}G f : X → B H G , an abstract principal G G G -connection on P P P is a lift f ^ : X → B c o n G \hat{f} : X \to B_{\mathrm{con}}G f ^ : X → B con G of f f f along p c o n p_{\mathrm{con}} p con (depicted below for the braided/Chern–Weil case, and with (14) as the right-hand square for general nonabelian G G G ):
B c o n G → curv Ω c l 2 ( − ; G ) f ^ ↑ ↓ σ 2 X → f B H G → θ 2 ♭ d R B H 2 G \begin{CD}
B_{\mathrm{con}}G @>{\operatorname{curv}}>> \Omega^2_{\mathrm{cl}}(-;G) \\
@A{\hat{f}}AA @VV{\sigma_2}V \\
X @>{f}>> B_{\mathrm{H}}G @>{\theta_2}>> \flat_{\mathrm{dR}}B^2_{\mathrm{H}}G
\end{CD} B con G f ^ ⏐ ↑ X curv f Ω cl 2 ( − ; G ) ↓ ⏐ σ 2 B H G θ 2 ♭ dR B H 2 G
More generally, replacing θ 2 \theta_2 θ 2 by θ k + 1 \theta_{k+1} θ k + 1 for a ( k + 1 ) (k+1) ( k + 1 ) -tuply deloopable object G ∈ Del 2 k + 1 ( H ) G \in\operatorname{Del}^{k+1}_2(\mathbf{H}) G ∈ Del 2 k + 1 ( H ) defines connections on higher k k k -bundles:
Definition 4.7 (Moduli Stack of k k k -Connections). Let H \mathbf{H} H be a Grothendieck (2,1)-topos and let G ∈ Del 2 k + 1 ( H ) G \in\operatorname{Del}^{k+1}_2(\mathbf{H}) G ∈ Del 2 k + 1 ( H ) be a ( k + 1 ) (k+1) ( k + 1 ) -tuply deloopable object (k ≥ 0 k \ge0 k ≥ 0 ). At the level of (2,1)-truncation, B c o n G = B c o n 1 G ∈ H B_{\mathrm{con}}G = B^1_{\mathrm{con}}G \in\mathbf{H} B con G = B con 1 G ∈ H is internal to H \mathbf{H} H ; for k = 2 k=2 k = 2 , B c o n 2 G ∈ Grpd 2 ( H ) B^2_{\mathrm{con}}G \in\operatorname{Grpd}_2(\mathbf{H}) B con 2 G ∈ Grpd 2 ( H ) is an internal 2-stack; for k ≥ 3 k \ge3 k ≥ 3 , B c o n k G B^k_{\mathrm{con}}G B con k G is an object of the ambient ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos H ∞ \mathbf{H}_{\infty} H ∞ , with mapping
groupoid H ( X , B c o n G k ) : = τ ≤ 1 H ∞ ( X , B c o n G k ) \mathbf{H}(X,B^k_{\mathrm{con}G}) := \tau_{\leq1}\mathbf{H}_{\infty}(X,B^k_{\mathrm{con}G}) H ( X , B con G k ) := τ ≤ 1 H ∞ ( X , B con G k ) . The moduli stack of G G G -connection k k k -forms B c o n G k B^k_{\mathrm{con}G} B con G k is the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback on the left below; a connection on a G G G -k k k -bundle P k → X P_k \to X P k → X classified by f : X → B H k G f:X\to B^k_{\mathbf{H}}G f : X → B H k G is a lift f ^ : X → B c o n G k \hat{f}:X\to B^k_{\mathrm{con}G} f ^ : X → B con G k as on the right:
B c o n G k → curv k Ω c l k + 1 ( − ; G ) p c o n k ↓ ↓ σ k + 1 B H k G → θ k + 1 b d R B H k + 1 G B c o n G k f ^ ↑ X → f B H k G \begin{CD}
B^k_{\mathrm{con}G} @>{\operatorname{curv}_k}>> \Omega^{k+1}_{\mathrm{cl}}(-;G) \\
@V{p^k_{\mathrm{con}}}VV @VV{\sigma_{k+1}}V \\
B^k_{\mathbf{H}}G @>{\theta_{k+1}}>> \mathbf{b}_{\mathrm{dR}}B^{k+1}_{\mathbf{H}}G
\end{CD}
\qquad
\begin{CD}
B^k_{\mathrm{con}G} \\
@A{\hat{f}}AA \\
X @>{f}>> B^k_{\mathbf{H}}G
\end{CD} B con G k p con k ↓ ⏐ B H k G curv k θ k + 1 Ω cl k + 1 ( − ; G ) ↓ ⏐ σ k + 1 b dR B H k + 1 G B con G k f ^ ⏐ ↑ X f B H k G
In Section 4.2.3, we constructed the classifying object B c o n G k ∈ H B^k_{\mathrm{con}G}\in\mathbf{H} B con G k ∈ H whose external hom-groupoid H ( X , B c o n G k ) ∈ G r p d 2 \mathbf{H}(X,B^k_{\mathrm{con}G})\in\mathbf{Grpd}_2 H ( X , B con G k ) ∈ Grpd 2 is the groupoid of G G G -connections on k k k -bundles over X X X and gauge transformations between them. However, in the study of classical and quantum gauge field theories, action functionals, symplectic structures, and path integrals are defined on the smooth moduli space of connections on X X X , which must exist as an internal object G Conn k ( X ) ∈ H G\operatorname{Conn}^k(X)\in\mathbf{H} G Conn k ( X ) ∈ H rather than merely an external groupoid in G r p d 2 \mathbf{Grpd}_2 Grpd 2 .
Since every Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H is cartesian closed, it possesses an internal hom 2 2 2 -bifunctor [ − , − ] H : H o p × H → H [-,-]_{\mathbf{H}}:\mathbf{H}^{\mathrm{op}}\times\mathbf{H}\to\mathbf{H} [ − , − ] H : H op × H → H characterized by the adjunction H ( U × X , Y ) ≃ H ( U , [ X , Y ] H ) \mathbf{H}(U\times X,Y)\simeq\mathbf{H}(U,[X,Y]_{\mathbf{H}}) H ( U × X , Y ) ≃ H ( U , [ X , Y ] H ) . One might naively attempt to define the internal moduli stack of k k k -connections on X X X simply as the internal hom [ X , B c o n G k ] H [X,B^k_{\mathrm{con}G}]_{\mathbf{H}} [ X , B con G k ] H . However, as analyzed in detail in [38 Sections 5.2.13.4, 6.4.16] , the naive internal hom [ X , B c o n G k ] H [X,B^k_{\mathrm{con}G}]_{\mathbf{H}} [ X , B con G k ] H fails to reproduce the physical gauge groupoid even for G = U ( 1 ) G=U(1) G = U ( 1 ) and k = 1 k=1 k = 1 in S m o o t h G r p d 2 \mathbf{SmoothGrpd}_2 SmoothGrpd 2 !
To see why, evaluate [ X , B c o n U ( 1 ) 1 ] H [X,B^1_{\mathrm{con}U(1)}]_{\mathbf{H}} [ X , B con U ( 1 ) 1 ] H on a test Cartesian space U ∈ C a r t S p U\in\mathbf{CartSp} U ∈ CartSp :
[ X , B c o n U ( 1 ) 1 ] H ( U ) ≃ H ( U × X , B c o n U ( 1 ) 1 ) . [X,B^1_{\mathrm{con}U(1)}]_{\mathbf{H}}(U)\simeq\mathbf{H}(U\times X,B^1_{\mathrm{con}U(1)}). [ X , B con U ( 1 ) 1 ] H ( U ) ≃ H ( U × X , B con U ( 1 ) 1 ) .
Every object of H ( U × X , B c o n U ( 1 ) 1 ) \mathbf{H}(U\times X,B^1_{\mathrm{con}U(1)}) H ( U × X , B con U ( 1 ) 1 ) is a U ( 1 ) U(1) U ( 1 ) -bundle with connection on the product manifold U × X U\times X U × X , which includes not only U U U -parameterized families of connections on X X X (components along T ∗ X T^*X T ∗ X ), but also spurious connection components along T ∗ U T^*U T ∗ U ! Equivalently, a U U U -parameterized family of gauge configurations on X X X should be a smooth family of differential forms on X X X (parameterized smoothly by U U U ), whereas the underlying topological bundle and gauge transformations along U U U may vary only via bare U U U -indexed families of gauge transformations on X X X .
This discrepancy is resolved systematically in any locally local Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H using the sharp modality ♯ = coDisc ∘ Γ \sharp=\operatorname{coDisc}\circ\Gamma ♯ = coDisc ∘ Γ (right adjoint to ♭ \flat ♭ ). Recall that for any object Y ∈ H Y\in\mathbf{H} Y ∈ H , the unit of the adjunction ♭ ⊣ ♯ \flat\dashv\sharp ♭ ⊣ ♯ (or the canonical morphism η Y : Y → ♯ Y \eta_Y:Y\to\sharp Y η Y : Y → ♯ Y ) embeds Y Y Y into its codiscrete version ♯ Y \sharp Y ♯ Y , where U U U -plots U → ♯ Y U\to\sharp Y U → ♯ Y correspond via H ( U , ♯ Y ) ≃ H ( ♭ U , Y ) \mathbf{H}(U,\sharp Y)\simeq\mathbf{H}(\flat U,Y) H ( U , ♯ Y ) ≃ H ( ♭ U , Y ) to bare point-indexed maps Γ ( U ) → Γ ( Y ) \Gamma(U)\to\Gamma(Y) Γ ( U ) → Γ ( Y ) .
To ensure that U U U -plots of the internal moduli stack of connections have smoothly U U U -parameterized local connection k k k -form potentials (without T ∗ U T^*U T ∗ U -legs) while gauge transformations and bundle transitions are parameterized via ♯ \sharp ♯ (i.e., by ♭ U \flat U ♭ U ), we form the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback over the curvature:
Definition 4.8 (Internal Moduli Stack of Connections G Conn k ( X ) G\operatorname{Conn}^k(X) G Conn k ( X ) ) [38 Sections 5.2.13.4, 6.4.16] . Let H = S t k ( C , J ) \mathbf{H}=\mathbf{Stk}(\mathcal{C},J) H = Stk ( C , J ) be a locally local Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos (so that ♭ ⊣ ♯ \flat\dashv\sharp ♭ ⊣ ♯ exists) and let G ∈ D e l k + 1 ( H ) G\in\mathbf{Del}^{k+1}(\mathbf{H}) G ∈ Del k + 1 ( H ) (or G ∈ G r p d 2 ( H ) G\in\mathbf{Grpd}_2(\mathbf{H}) G ∈ Grpd 2 ( H ) for k = 1 k=1 k = 1 ). The internal moduli stack of G G G -k k k -connections on X X X , denoted G Conn k ( X ) G\operatorname{Conn}^k(X) G Conn k ( X ) , is defined synthetically by the curvature ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback square in H \mathbf{H} H :
G Conn k ( X ) → [ X , Ω c l k + 1 ( − ; G ) ] H ↓ ↓ [ X , η k ] H ♯ [ X , B c o n G k ] H → ♯ [ X , curv k ] H ♯ [ X , Ω c l k + 1 ( − ; G ) ] H (16) \begin{CD}
G\operatorname{Conn}^k(X) @>>> [X,\Omega^{k+1}_{\mathrm{cl}}(-;G)]_{\mathbf{H}} \\
@VVV @VV{[X,\eta^k]_{\mathbf{H}}}V \\
\sharp[X,B^k_{\mathrm{con}G}]_{\mathbf{H}} @>{\sharp[X,\operatorname{curv}_k]_{\mathbf{H}}}>> \sharp[X,\Omega^{k+1}_{\mathrm{cl}}(-;G)]_{\mathbf{H}} \tag*{(16)}
\end{CD} G Conn k ( X ) ↓ ⏐ ♯ [ X , B con G k ] H ♯ [ X , curv k ] H [ X , Ω cl k + 1 ( − ; G ) ] H ↓ ⏐ [ X , η k ] H ♯ [ X , Ω cl k + 1 ( − ; G ) ] H (16) At the level of ( 2 , 1 ) (2,1) ( 2 , 1 ) -truncation, G Conn 1 ( X ) ∈ H G\operatorname{Conn}^1(X)\in\mathbf{H} G Conn 1 ( X ) ∈ H is an internal 1 1 1 -stack in H \mathbf{H} H ; for k = 2 k=2 k = 2 , G Conn 2 ( X ) G\operatorname{Conn}^2(X) G Conn 2 ( X ) is an internal 2 2 2 -stack in G r p d 2 ( H ) \mathbf{Grpd}_2(\mathbf{H}) Grpd 2 ( H ) ; for k ≥ 3 k\geq3 k ≥ 3 , it is an object of H ∞ \mathbf{H}_{\infty} H ∞ . In particular, for
k = 1 k = 1 k = 1 where B c o n G ≃ L J ( Ω 1 ( − ; g ) / / G ) B_{\mathrm{con}G} \simeq L_J(\Omega^1(-;\mathfrak{g})//G) B con G ≃ L J ( Ω 1 ( − ; g ) // G ) as in (15) , G C o n n 1 ( X ) G\mathrm{Conn}^1(X) G Conn 1 ( X ) admits the explicit presentation as the stackified quotient action groupoid:
G C o n n 1 ( X ) ≃ L J ( [ X , Ω 1 ( − ; g ) ] H / / # [ X , G ] H ) , (17) G\mathrm{Conn}^1(X) \simeq L_J([X,\Omega^1(-;\mathfrak{g})]_{\mathbf{H}} //_{\#} [X,G]_{\mathbf{H}}), \tag*{(17)} G Conn 1 ( X ) ≃ L J ([ X , Ω 1 ( − ; g ) ] H / / # [ X , G ] H ) , (17)
so that U U U -plots of G C o n n 1 ( X ) G\mathrm{Conn}^1(X) G Conn 1 ( X ) consist of smooth U U U -families of connection 1-forms on X X X (without T ∗ U T^*U T ∗ U -legs) and ♯ \sharp ♯ -parameterized gauge transformations between them.
Proposition 4.9 (Global Sections of the Internal Connection Moduli). Let H \mathbf{H} H be a local Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos. For every X ∈ H X \in\mathbf{H} X ∈ H and G ∈ D e l 2 k + 1 ( H ) G \in\mathrm{Del}_{2}^{k+1}(\mathbf{H}) G ∈ Del 2 k + 1 ( H ) (or G ∈ G r p 2 ( H ) G \in\mathrm{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) for k = 1 k=1 k = 1 ), the global sections groupoid of G C o n n k ( X ) G\mathrm{Conn}^k(X) G Conn k ( X ) recovers the external groupoid of G G G -k k k -connections on X X X :
Γ ( G C o n n k ( X ) ) = H ( ∗ , G C o n n k ( X ) ) ≃ H ( X , B c o n G k ) . \Gamma(G\mathrm{Conn}^k(X)) = \mathbf{H}(*,G\mathrm{Conn}^k(X)) \simeq\mathbf{H}(X,B_{\mathrm{con}G}^k). Γ ( G Conn k ( X )) = H ( ∗ , G Conn k ( X )) ≃ H ( X , B con G k ) .
Proof. Since H \mathbf{H} H is local, the global sections 2-functor Γ = H ( ∗ , − ) : H → G r p 2 \Gamma= \mathbf{H}(*,-):\mathbf{H} \to\mathbf{Grp}_{2} Γ = H ( ∗ , − ) : H → Grp 2 is both a right 2-adjoint to L C s t \mathrm{LCst} LCst and a left 2-adjoint to c o D i s c \mathrm{coDisc} coDisc , so it preserves all ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits, ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits, and stackification localizations (since ∗ H *_{\mathbf{H}} ∗ H is locally contractible/projective on local covers). Moreover, since ♯ = c o D i s c ∘ Γ \sharp= \mathrm{coDisc} \circ\Gamma ♯ = coDisc ∘ Γ with Γ ∘ c o D i s c ≃ i d G r p 2 \Gamma\circ\mathrm{coDisc} \simeq\mathrm{id}_{\mathbf{Grp}_{2}} Γ ∘ coDisc ≃ id Grp 2 , we have Γ ( ♯ Y ) ≃ Γ ( Y ) \Gamma(\sharp Y) \simeq\Gamma(Y) Γ ( ♯ Y ) ≃ Γ ( Y ) naturally in Y ∈ H Y \in\mathbf{H} Y ∈ H , under which Γ ( η Y ♯ ) \Gamma(\eta^{\sharp}_{Y}) Γ ( η Y ♯ ) becomes the identity equivalence. Applying Γ \Gamma Γ to the defining curvature pullback square (16) replaces every ♯ [ X , − ] H \sharp[X,-]_{\mathbf{H}} ♯ [ X , − ] H by Γ ( [ X , − ] H ) ≃ H ( X , − ) \Gamma([X,-]_{\mathbf{H}}) \simeq\mathbf{H}(X,-) Γ ([ X , − ] H ) ≃ H ( X , − ) , under which Γ ( η ♯ ) \Gamma(\eta^{\sharp}) Γ ( η ♯ ) becomes an equivalence. Thus, Γ ( G C o n n k ( X ) ) \Gamma(G\mathrm{Conn}^k(X)) Γ ( G Conn k ( X )) is computed as the pullback along an equivalence, canonically recovering H ( X , B c o n G k ) \mathbf{H}(X,B_{\mathrm{con}G}^k) H ( X , B con G k ) (which also directly matches Γ ( [ X , Ω 1 ] H / / ♯ [ X , G ] H ) ≃ H ( X , Ω 1 ) / / H ( X , G ) ≃ H ( X , B c o n G ) \Gamma([X,\Omega^1]_{\mathbf{H}} //_{\sharp} [X,G]_{\mathbf{H}}) \simeq\mathbf{H}(X,\Omega^1)//\mathbf{H}(X,G) \simeq\mathbf{H}(X,B_{\mathrm{con}G}) Γ ([ X , Ω 1 ] H / / ♯ [ X , G ] H ) ≃ H ( X , Ω 1 ) // H ( X , G ) ≃ H ( X , B con G ) via (17) for k = 1 k=1 k = 1 ). □
Example 4.10 (Concrete Realization in S m o o t h G r p d 2 \mathbf{SmoothGrpd}_2 SmoothGrpd 2 ). Consider the cohesive 2-topos H = S m o o t h G r p d 2 = S t k ( C a r t S p , J o p e n ) \mathbf{H} = \mathbf{SmoothGrpd}_2 = \mathrm{Stk}(\mathrm{CartSp},J_{\mathrm{open}}) H = SmoothGrpd 2 = Stk ( CartSp , J open ) and let G G G be a Lie group with Lie algebra g \mathfrak{g} g , viewed as a 0-truncated group sheaf U ↦ C ∞ ( U , G ) U \mapsto C^\infty(U,G) U ↦ C ∞ ( U , G ) .
Delooping and Flat de Rham Coefficient: The external/internal delooping prestack sends U ∈ C a r t S p U \in\mathrm{CartSp} U ∈ CartSp to the one-object groupoid B e x t G ( U ) = ( C ∞ ( U , G ) ⇒ ∗ ) \mathbf{B}^{\mathrm{ext}}G(U) = (C^\infty(U,G) \Rightarrow*) B ext G ( U ) = ( C ∞ ( U , G ) ⇒ ∗ ) , and its stackification B H G = L J o p e n ( B e x t G ) \mathbf{B}_{\mathbf{H}}G = L_{J_{\mathrm{open}}}(\mathbf{B}^{\mathrm{ext}}G) B H G = L J open ( B ext G ) classifies smooth principal G G G -bundles: for a smooth manifold X X X with good open cover { U i → X } \{U_i \to X\} { U i → X } , H ( X , B H G ) \mathbf{H}(X,\mathbf{B}_{\mathbf{H}}G) H ( X , B H G ) is the groupoid of Čech 1-cocycles g i j : U i ∩ U j → G g_{ij}:U_i \cap U_j \to G g ij : U i ∩ U j → G (g i j g j k = g i k g_{ij}g_{jk}=g_{ik} g ij g j k = g ik ) and coboundaries h i : U i → G h_i:U_i \to G h i : U i → G (g i j ′ = h i − 1 g i j h j g'_{ij}=h_i^{-1}g_{ij}h_j g ij ′ = h i − 1 g ij h j ). Since ♭ G ( U ) = G δ ( U ) \flat G(U)=G^\delta(U) ♭ G ( U ) = G δ ( U ) consists of locally constant maps U → G U \to G U → G , the homotopy fiber ♭ d R B H G ≃ G / G δ \flat_{\mathrm{dR}}\mathbf{B}_{\mathbf{H}}G \simeq G/G^\delta ♭ dR B H G ≃ G / G δ is weakly equivalent via the Maurer–Cartan map g ↦ g − 1 d g g \mapsto g^{-1}dg g ↦ g − 1 d g to the action groupoid Ω f l a t 1 ( − ; g ) / / G \Omega^1_{\mathrm{flat}}(-;\mathfrak{g})//G Ω flat 1 ( − ; g ) // G , where G G G acts by gauge transformations A ↦ Ad g − 1 A + g − 1 d g A \mapsto\operatorname{Ad}_{g^{-1}}A+g^{-1}dg A ↦ Ad g − 1 A + g − 1 d g .
Moduli Stack of Connections B c o n G \mathbf{B}_{\mathrm{con}}G B con G : Dropping the flatness constraint d A + 1 2 [ A ∧ A ] = 0 dA+\frac{1}{2}[A\wedge A]=0 d A + 2 1 [ A ∧ A ] = 0 on U ∈ C a r t S p U \in\mathrm{CartSp} U ∈ CartSp defines B c o n G ( U ) = Ω 1 ( U ; g ) / / C ∞ ( U , G ) \mathbf{B}_{\mathrm{con}}G(U)=\Omega^1(U;\mathfrak{g})//C^\infty(U,G) B con G ( U ) = Ω 1 ( U ; g ) // C ∞ ( U , G ) . By Čech descent over { U i → X } \{U_i \to X\} { U i → X } , an object of H ( X , B c o n G ) \mathbf{H}(X,\mathbf{B}_{\mathrm{con}}G) H ( X , B con G ) is a pair ( g i j , A i ) (g_{ij},A_i) ( g ij , A i ) with A i ∈ Ω 1 ( U i ; g ) A_i \in\Omega^1(U_i;\mathfrak{g}) A i ∈ Ω 1 ( U i ; g ) satisfying A j = Ad g i j − 1 ( A i ) + g i j − 1 d g i j A_j=\operatorname{Ad}_{g_{ij}^{-1}}(A_i)+g_{ij}^{-1}dg_{ij} A j = Ad g ij − 1 ( A i ) + g ij − 1 d g ij on U i ∩ U j U_i\cap U_j U i ∩ U j , i.e., a principal G G G -bundle with connection ( P , ∇ ) → X (P,\nabla)\to X ( P , ∇ ) → X , and a morphism ( g i j , A i ) → ( g i j ′ , A i ′ ) (g_{ij},A_i)\to(g'_{ij},A'_i) ( g ij , A i ) → ( g ij ′ , A i ′ ) is a gauge transformation h i : U i → G h_i:U_i\to G h i : U i → G .
Sharp-Adjusted Internal Moduli G C o n n 1 ( X ) G\mathrm{Conn}^1(X) G Conn 1 ( X ) : Evaluating the pullback (16) on a test space U ∈ C a r t S p U \in\mathrm{CartSp} U ∈ CartSp , an object of G C o n n 1 ( X ) ( U ) G\mathrm{Conn}^1(X)(U) G Conn 1 ( X ) ( U ) is a smoothly U U U -parameterized family of connection 1-forms A i ( u ) ∈ Ω 1 ( U i ; g ) A_i(u) \in\Omega^1(U_i;\mathfrak{g}) A i ( u ) ∈ Ω 1 ( U i ; g ) on T ∗ X T^*X T ∗ X only (no d u du d u -legs along T ∗ U T^*U T ∗ U ), because [ X , Ω 1 ( − ; g ) ] H ( U ) = Ω 1 ( X ; C ∞ ( U , g ) ) [X,\Omega^1(-;\mathfrak{g})]_{\mathbf{H}}(U)=\Omega^1(X;C^\infty(U,\mathfrak{g})) [ X , Ω 1 ( − ; g ) ] H ( U ) = Ω 1 ( X ; C ∞ ( U , g )) , together with a U U U -point-indexed family of transition functions and gauge transformations u ↦ h i ( u ) ∈ C ∞ ( U i , G ) u\mapsto h_i(u)\in C^\infty(U_i,G) u ↦ h i ( u ) ∈ C ∞ ( U i , G ) coming from ♯ [ X , B c o n G ] H ( U ) ≃ H ( U × X , B c o n G ) \,\sharp[X,\mathbf{B}_{\mathrm{con}}G]_{\mathbf{H}}(U)\simeq\mathbf{H}(U\times X,\mathbf{B}_{\mathrm{con}}G) ♯ [ X , B con G ] H ( U ) ≃ H ( U × X , B con G ) .
Geometric Cohomology
We now show how the categorical and geometric structures developed in the preceding sections unify under the umbrella of intrinsic, modal, and differential cohomology internal to a cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos.
Intrinsic Cohomology
A vast array of cohomology theories across mathematics—including abelian sheaf cohomology, singular cohomology, group cohomology, Čech cohomology, and nonabelian 1-cohomology—are manifestations of a single homotopical principle first axiomatized by Brown [5 ] : they are connected components (π 0 = τ ≤ 0 \pi_0=\tau_{\leq0} π 0 = τ ≤ 0 ) of derived mapping spaces into a classifying or Eilenberg–Mac Lane object. Lurie [26, Section 7.2.2, Definition 7.2.2.14] observed that this definition is completely intrinsic to any ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -topos (and indeed any ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -category [38 ] ). In our ( 2 , 1 ) (2,1) ( 2 , 1 ) -categorical context, it takes the following concise form:
Definition 4.11 (Intrinsic Cohomology). Let H \mathbf{H} H be a ( 2 , 1 ) (2,1) ( 2 , 1 ) -category and let A ∈ H A \in\mathbf{H} A ∈ H be an object. For every X ∈ H X \in\mathbf{H} X ∈ H , define the 0 0 0 -th intrinsic cohomology of X X X with coefficients in A A A as the set of equivalence classes of 1 1 1 -morphisms X → A X \to A X → A :
H 0 ( X ; A ) : = τ ≤ 0 H ( X , A ) ∈ S e t . H^{0}(X;A) := \tau_{\leq0}\mathbf{H}(X,A) \in\mathbf{Set}. H 0 ( X ; A ) := τ ≤ 0 H ( X , A ) ∈ Set . Suppose now that H \mathbf{H} H is a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos and that A ∈ Del 2 k ( H ) A \in\operatorname{Del}_{2}^{k}(\mathbf{H}) A ∈ Del 2 k ( H ) is k k k -tuply deloopable (k ≥ 1 k \geq1 k ≥ 1 ; formed internal to H \mathbf{H} H for k = 1 k=1 k = 1 , in Grpd 2 ( H ) \operatorname{Grpd}_{2}(\mathbf{H}) Grpd 2 ( H ) for k = 2 k=2 k = 2 , and in H ∞ \mathbf{H}_{\infty} H ∞ for k ≥ 3 k \geq3 k ≥ 3 , cf. Remark 3.3). Define the k k k -th intrinsic cohomology of X X X with coefficients in A A A as
H k ( X ; A ) : = τ ≤ 0 H ( X , B H k A ) = H 0 ( X ; B H k A ) . H^{k}(X;A) := \tau_{\leq0}\mathbf{H}(X,B_{\mathbf{H}}^{k}A)=H^{0}(X;B_{\mathbf{H}}^{k}A). H k ( X ; A ) := τ ≤ 0 H ( X , B H k A ) = H 0 ( X ; B H k A ) .
Combining Definition 4.11 with Corollary 3.7 immediately yields:
Lemma 4.12. Let H \mathbf{H} H be a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos (or cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos as in Corollary 3.7) and let A ∈ H A \in\mathbf{H} A ∈ H be an object. Then:
(1) A A A admits a group object structure if and only if H k ( X ; A ) H^{k}(X;A) H k ( X ; A ) for k = 0 , 1 k=0,1 k = 0 , 1 is canonically defined via delooping for every X ∈ H X \in\mathbf{H} X ∈ H ;
(2) A A A admits a braided group object structure if and only if H k ( X ; A ) H^{k}(X;A) H k ( X ; A ) for k = 0 , 1 , 2 k=0,1,2 k = 0 , 1 , 2 is canonically defined for every X ∈ H X \in\mathbf{H} X ∈ H ;
(3) A A A admits a commutative group object structure if and only if H k ( X ; A ) H^{k}(X;A) H k ( X ; A ) for k = 0 , 1 , 2 , 3 k=0,1,2,3 k = 0 , 1 , 2 , 3 is canonically defined for every X ∈ H X \in\mathbf{H} X ∈ H ;
(4) A A A admits a commutative group object structure if and only if H k ( X ; A ) H^{k}(X;A) H k ( X ; A ) is canonically defined for every X ∈ H X \in\mathbf{H} X ∈ H and every k ≥ 0 k \geq0 k ≥ 0 .
Proof. Immediate from the characterization of k k k -tuply deloopable objects in Corollary 3.7. □
For a fixed object X ∈ H X \in\mathbf{H} X ∈ H , 0 0 0 -th cohomology defines a 2 2 2 -functor H 0 ( X ; − ) : H → S e t 2 H^{0}(X;-) : \mathbf{H} \to\mathbf{Set}_{2} H 0 ( X ; − ) : H → Set 2 given by the composite
H → H ( X , − ) Grpd 2 → τ ≤ 0 S e t 2 \begin{CD}
\mathbf{H} @>{\mathbf{H}(X,-)}>> \operatorname{Grpd}_{2} @>{\tau_{\leq0}}>> \mathbf{Set}_{2}
\end{CD} H H ( X , − ) Grpd 2 τ ≤ 0 Set 2
Because the representable 2 2 2 -functor H ( X , − ) : H → Grpd 2 \mathbf{H}(X,-) : \mathbf{H} \to\operatorname{Grpd}_{2} H ( X , − ) : H → Grpd 2 preserves all ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and τ ≤ 0 : Grpd 2 → S e t 2 \tau_{\leq0} : \operatorname{Grpd}_{2} \to\mathbf{Set}_{2} τ ≤ 0 : Grpd 2 → Set 2 preserves finite products, the composite H 0 ( X ; − ) : H → S e t 2 H^{0}(X;-) : \mathbf{H} \to\mathbf{Set}_{2} H 0 ( X ; − ) : H → Set 2 is a cartesian (finite-product-preserving) 2 2 2 -functor. Consequently, H 0 ( X ; − ) H^{0}(X;-) H 0 ( X ; − ) preserves group objects, braided group objects, and commutative group objects:
ComGrp 2 ( H ) → H 0 ( X ; − ) ComGrp 2 ( S e t 2 ) ↓ ↓ BrGrp 2 ( H ) → H 0 ( X ; − ) BrGrp 2 ( S e t 2 ) ↓ ↓ Grp 2 ( H ) → H 0 ( X ; − ) Grp 2 ( S e t 2 ) ↓ ↓ H → H ( X , − ) Grpd 2 → τ ≤ 0 S e t 2 \begin{CD}
\operatorname{ComGrp}_{2}(\mathbf{H}) @>{H^{0}(X;-)}>> \operatorname{ComGrp}_{2}(\mathbf{Set}_{2}) \\
@VVV @VVV \\
\operatorname{BrGrp}_{2}(\mathbf{H}) @>{H^{0}(X;-)}>> \operatorname{BrGrp}_{2}(\mathbf{Set}_{2}) \\
@VVV @VVV \\
\operatorname{Grp}_{2}(\mathbf{H}) @>{H^{0}(X;-)}>> \operatorname{Grp}_{2}(\mathbf{Set}_{2}) \\
@VVV @VVV \\
\mathbf{H} @>{\mathbf{H}(X,-)}>> \operatorname{Grpd}_{2} @>{\tau_{\leq0}}>> \mathbf{Set}_{2}
\end{CD} ComGrp 2 ( H ) ↓ ⏐ BrGrp 2 ( H ) ↓ ⏐ Grp 2 ( H ) ↓ ⏐ H H 0 ( X ; − ) H 0 ( X ; − ) H 0 ( X ; − ) H ( X , − ) ComGrp 2 ( Set 2 ) ↓ ⏐ BrGrp 2 ( Set 2 ) ↓ ⏐ Grp 2 ( Set 2 ) ↓ ⏐ Grpd 2 τ ≤ 0 Set 2
Since S e t 2 \mathbf{Set}_{2} Set 2 is a discrete 2 2 2 -category (a 1 1 1 -category), any braided group object in S e t 2 \mathbf{Set}_{2} Set 2 is already an abelian group ( BrGrp 2 ( S e t 2 ) = ComGrp 2 ( S e t 2 ) = A b 2 ) (\operatorname{BrGrp}_{2}(\mathbf{Set}_{2})=\operatorname{ComGrp}_{2}(\mathbf{Set}_{2})=\mathbf{Ab}_{2}) ( BrGrp 2 ( Set 2 ) = ComGrp 2 ( Set 2 ) = Ab 2 ) , and for A ∈ Del 2 k ( H ) A \in\operatorname{Del}_{2}^{k}(\mathbf{H}) A ∈ Del 2 k ( H ) , B H k A B_{\mathbf{H}}^{k}A B H k A retains $0 group structures when k = 1 k = 1 k = 1 on a general group A A A , 1 group structure when A ∈ D e l 2 k + 1 ( H ) A \in\mathrm{Del}_{2}^{k+1}(\mathbf{H}) A ∈ Del 2 k + 1 ( H ) , and at least 2 group structures (hence braided/abelian on H k H^k H k ) when A ∈ D e l 2 k + 2 ( H ) A \in\mathrm{Del}_{2}^{k+2}(\mathbf{H}) A ∈ Del 2 k + 2 ( H ) . More precisely:
Lemma 4.13. Let H \mathbf{H} H be a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos and let A ∈ H A \in\mathbf{H} A ∈ H be an object.
(1) If H 1 ( X ; A ) H^1(X;A) H 1 ( X ; A ) is defined for every X ∈ H X \in\mathbf{H} X ∈ H (so A ∈ Grp 2 ( H ) A \in\operatorname{Grp}_2(\mathbf{H}) A ∈ Grp 2 ( H ) ), then H 0 ( X ; A ) H^0(X;A) H 0 ( X ; A ) is a group and H 1 ( X ; A ) H^1(X;A) H 1 ( X ; A ) is a pointed set; if H 2 ( X ; A ) H^2(X;A) H 2 ( X ; A ) is also defined (so A ∈ BrGrp 2 ( H ) A \in\operatorname{BrGrp}_2(\mathbf{H}) A ∈ BrGrp 2 ( H ) ), then H 0 ( X ; A ) H^0(X;A) H 0 ( X ; A ) is an abelian group and H 1 ( X ; A ) H^1(X;A) H 1 ( X ; A ) is a group;
(2) If H k + 1 ( X ; A ) H^{k+1}(X;A) H k + 1 ( X ; A ) is defined for every X ∈ H X \in\mathbf{H} X ∈ H with k ≥ 1 k \ge1 k ≥ 1 (and in particular if A ∈ ComGrp 2 ( H ) A \in\operatorname{ComGrp}_2(\mathbf{H}) A ∈ ComGrp 2 ( H ) ), then H k ( X ; A ) H^k(X;A) H k ( X ; A ) is an abelian group for all k ≥ 1 k \ge1 k ≥ 1 (since B H k A ∈ BrGrp 2 ( H ) B_{\mathbf{H}}^k A \in\operatorname{BrGrp}_2(\mathbf{H}) B H k A ∈ BrGrp 2 ( H ) for A ∈ ComGrp 2 ( H ) A \in\operatorname{ComGrp}_2(\mathbf{H}) A ∈ ComGrp 2 ( H ) ).
This justifies referring to H 1 ( X ; A ) H^1(X;A) H 1 ( X ; A ) as nonabelian 1-cohomology whenever A ∈ Grp 2 ( H ) A \in\operatorname{Grp}_2(\mathbf{H}) A ∈ Grp 2 ( H ) is not braided (so H k ( X ; A ) H^k(X;A) H k ( X ; A ) is undefined for k > 1 k > 1 k > 1 ).
Modal Cohomologies
Each internal modality F : H → H F : \mathbf{H} \to\mathbf{H} F : H → H on a ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos induces a corresponding modal cohomology theory by twisting the coefficient object:
Definition 4.14 (Modal Cohomology). Let H \mathbf{H} H be a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos and let F : H → H F : \mathbf{H} \to\mathbf{H} F : H → H be an internal modality on H \mathbf{H} H .
(1) For any object A ∈ H A \in\mathbf{H} A ∈ H , define the 0 0 0 -th F F F -cohomology of X ∈ H X \in\mathbf{H} X ∈ H with coefficients in A A A as
H F 0 ( X ; A ) : = H 0 ( X ; F ( A ) ) = τ ≤ 0 H ( X , F ( A ) ) . H_F^0(X;A) := H^0(X;F(A)) = \tau_{\leq0}\mathbf{H}(X,F(A)). H F 0 ( X ; A ) := H 0 ( X ; F ( A )) = τ ≤ 0 H ( X , F ( A )) . (2) If F ( A ) ∈ Del 2 k ( H ) F(A) \in\operatorname{Del}_2^k(\mathbf{H}) F ( A ) ∈ Del 2 k ( H ) is k k k -tuply deloopable, define the k k k -th F F F -cohomology of X ∈ H X \in\mathbf{H} X ∈ H with coefficients in A A A as
H F k ( X ; A ) : = H 0 ( X ; B H k F ( A ) ) . H_F^k(X;A) := H^0(X;B_{\mathbf{H}}^kF(A)). H F k ( X ; A ) := H 0 ( X ; B H k F ( A )) .
Whenever the modality F : H → H F : \mathbf{H} \to\mathbf{H} F : H → H preserves finite products and ( 2 , 1 ) (2,1) ( 2 , 1 ) -coequalizers (for instance, when F = ♭ = LCst ∘ Γ F = \flat= \operatorname{LCst} \circ\Gamma F = ♭ = LCst ∘ Γ in a locally local ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos where Γ \Gamma Γ is a left adjoint to coDisc \operatorname{coDisc} coDisc and LCst \operatorname{LCst} LCst is both a left adjoint and left exact, or when F = ∫ = LCst ∘ Π F = \int= \operatorname{LCst} \circ\Pi F = ∫ = LCst ∘ Π in a strongly 1-connected ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos), F F F commutes with the delooping object construction:
Lemma 4.15. Let F : H → H F : \mathbf{H} \to\mathbf{H} F : H → H be a modality on a cartesian ( 2 , 1 ) (2,1) ( 2 , 1 ) -cosmos H \mathbf{H} H that preserves finite products and ( 2 , 1 ) (2,1) ( 2 , 1 ) -coequalizers. If A ∈ Del 2 k ( H ) A \in\operatorname{Del}_2^k(\mathbf{H}) A ∈ Del 2 k ( H ) is k k k -tuply deloopable, then F ( A ) ∈ Del 2 k ( H ) F(A) \in\operatorname{Del}_2^k(\mathbf{H}) F ( A ) ∈ Del 2 k ( H ) is k k k -tuply deloopable and there is a canonical equivalence
B H k F ( A ) ≃ F ( B H k A ) . B_{\mathbf{H}}^kF(A) \simeq F(B_{\mathbf{H}}^kA). B H k F ( A ) ≃ F ( B H k A ) .
Consequently, H F k ( X ; A ) H_F^k(X;A) H F k ( X ; A ) is defined and satisfies H F k ( X ; A ) ≅ H F 0 ( X ; B H k A ) = H 0 ( X ; F ( B H k A ) ) H_F^k(X;A) \cong H_F^0(X;B_{\mathbf{H}}^kA) = H^0(X;F(B_{\mathbf{H}}^kA)) H F k ( X ; A ) ≅ H F 0 ( X ; B H k A ) = H 0 ( X ; F ( B H k A )) .
Proof. Since F F F preserves finite products, it preserves k k k -tuply group objects, so F ( A ) ∈ Grp 2 k ( H ) ≃ Del 2 k ( H ) F(A) \in\operatorname{Grp}_2^k(\mathbf{H}) \simeq\operatorname{Del}_2^k(\mathbf{H}) F ( A ) ∈ Grp 2 k ( H ) ≃ Del 2 k ( H ) . Since B H G = coeq ( G ⇉ ∗ ) B_{\mathbf{H}}G = \operatorname{coeq}(G \rightrightarrows*) B H G = coeq ( G ⇉ ∗ ) is a ( 2 , 1 ) (2,1) ( 2 , 1 ) -coequalizer and F F F preserves ∗ * ∗ and ( 2 , 1 ) (2,1) ( 2 , 1 ) -coequalizers, induction on k k k gives B H k F ( A ) ≃ F ( B H k A ) B_{\mathbf{H}}^kF(A) \simeq F(B_{\mathbf{H}}^kA) B H k F ( A ) ≃ F ( B H k A ) . □
Specializing to the cohesive modalities on a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H :
The flat modality ♭ \flat ♭ induces flat cohomology H ♭ 0 ( X ; A ) = H 0 ( X ; ♭ A ) H_{\flat}^0(X;A) = H^0(X;\flat A) H ♭ 0 ( X ; A ) = H 0 ( X ; ♭ A ) (measuring local systems / flat bundles), and its refinement ♭ d R \flat_{\mathrm{dR}} ♭ dR induces flat de Rham cohomology H ♭ d R 0 ( X ; A ) = H 0 ( X ; ♭ d R A ) H_{\flat_{\mathrm{dR}}}^0(X;A) = H^0(X;\flat_{\mathrm{dR}}A) H ♭ dR 0 ( X ; A ) = H 0 ( X ; ♭ dR A ) ;
In a locally 1-connected ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, the shape modality ∫ \int ∫ induces shape cohomology H ∫ 0 ( X ; A ) = H 0 ( X ; ∫ A ) H_{\int}^0(X;A) = H^0(X;\int A) H ∫ 0 ( X ; A ) = H 0 ( X ; ∫ A ) ;
In a locally local ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos, the sharp modality ♯ \sharp ♯ induces sharp cohomology H ♯ 0 ( X ; A ) = H 0 ( X ; ♯ A ) H_{\sharp}^0(X;A) = H^0(X;\sharp A) H ♯ 0 ( X ; A ) = H 0 ( X ; ♯ A ) .
Proposition 4.16. Let H \mathbf{H} H be a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and let A ∈ Del 2 k ( H ) A \in\operatorname{Del}_2^k(\mathbf{H}) A ∈ Del 2 k ( H ) be a k k k -tuply deloopable object.
(1) If H \mathbf{H} H is strongly 1-connected, then ∫ A \int A ∫ A is k k k -tuply deloopable and B H k ∫ A ≃ ∫ B H k A B_{\mathbf{H}}^k\int A \simeq\int B_{\mathbf{H}}^kA B H k ∫ A ≃ ∫ B H k A . In particular, H ∫ k ( X ; A ) ≅ H ∫ 0 ( X ; B H k A ) H_{\int}^k(X;A) \cong H_{\int}^0(X;B_{\mathbf{H}}^kA) H ∫ k ( X ; A ) ≅ H ∫ 0 ( X ; B H k A ) .
(2) If H \mathbf{H} H is locally local, then ♭ A \flat A ♭ A is k k k -tuply deloopable, B H k ♭ A ≃ ♭ B H k A B_{\mathbf{H}}^k\flat A \simeq\flat B_{\mathbf{H}}^k A B H k ♭ A ≃ ♭ B H k A , and H ♭ k ( X ; A ) ≅ H ♭ k ( X ; B H k A ) H_{\flat}^k(X; A) \cong H_{\flat}^k(X; B_{\mathbf{H}}^k A) H ♭ k ( X ; A ) ≅ H ♭ k ( X ; B H k A ) , while H d R k ( X ; A ) : = H 0 ( X ; ♭ d R B H k A ) H_{\mathrm{dR}}^k(X; A) := H^0(X; \flat_{\mathrm{dR}}B_{\mathbf{H}}^k A) H dR k ( X ; A ) := H 0 ( X ; ♭ dR B H k A ) classifies flat A A A -valued k k k -forms on X X X .
Proof. In (1), Π \Pi Π preserves ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits and finite products (by strong 1 1 1 -connectivity), and L C s t \mathrm{LCst} LCst preserves ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits (being left adjoint to Γ \Gamma Γ ) and finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits (being an inverse image functor), so R = L C s t ∘ Π \mathcal{R} = \mathrm{LCst} \circ\Pi R = LCst ∘ Π satisfies Lemma 4.15. In (2), Γ \Gamma Γ is left adjoint to c o D i s c \mathrm{coDisc} coDisc (by local locality) and right adjoint to L C s t \mathrm{LCst} LCst , so ♭ = L C s t ∘ Γ \flat= \mathrm{LCst} \circ\Gamma ♭ = LCst ∘ Γ preserves both finite ( 2 , 1 ) (2,1) ( 2 , 1 ) -limits and all ( 2 , 1 ) (2,1) ( 2 , 1 ) -colimits, again satisfying Lemma 4.15. □ \square □
Differential Cohomology
In Section 4.2.3 we constructed the moduli stack of G G G -connections on k k k -bundles B c o n G k ∈ H B_{\mathrm{con}G}^k \in\mathbf{H} B con G k ∈ H whenever G ∈ D e l k + 1 ( H ) G \in\mathrm{Del}_{k+1}(\mathbf{H}) G ∈ Del k + 1 ( H ) is ( k + 1 ) (k+1) ( k + 1 ) -tuply deloopable (as well as B c o n G B_{\mathrm{con}}G B con G for nonabelian group objects G ∈ G r p 2 ( H ) G \in\mathrm{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) ). We now define intrinsic differential cohomology with coefficients in G G G , and investigate whether the assignment G ↦ B c o n k G G \mapsto B_{\mathrm{con}}^k G G ↦ B con k G can be extended to a genuine internal modality ( − ) c o n : ∗ / H → ∗ / H (-)_{\mathrm{con}} : */\mathbf{H} \to*/\mathbf{H} ( − ) con : ∗ / H → ∗ / H defined on arbitrary (non-groupal) pointed objects A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H .
Definition 4.17 (Differential Cohomology). Let H \mathbf{H} H be a Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and let G ∈ D e l k + 1 ( H ) G \in\mathrm{Del}_{k+1}(\mathbf{H}) G ∈ Del k + 1 ( H ) be a ( k + 1 ) (k+1) ( k + 1 ) -tuply deloopable object (k ≥ 0 k \ge0 k ≥ 0 , with B c o n k G B_{\mathrm{con}}^kG B con k G formed internal to H \mathbf{H} H for k = 0 , 1 k = 0,1 k = 0 , 1 , in G r p 2 ( H ) \mathrm{Grp}_2(\mathbf{H}) Grp 2 ( H ) for k = 2 k = 2 k = 2 , and in H ∞ \mathbf{H}_\infty H ∞ for k ≥ 3 k \ge3 k ≥ 3 , cf. Remark 3.3) equipped with closed differential ( k + 1 ) (k+1) ( k + 1 ) -form coefficients σ k + 1 : Ω c l k + 1 ( − ; G ) → ♭ d R B H k G \sigma_{k+1} : \Omega_{\mathrm{cl}}^{k+1}(-;G) \to\flat_{\mathrm{dR}}B_{\mathbf{H}}^kG σ k + 1 : Ω cl k + 1 ( − ; G ) → ♭ dR B H k G (or G ∈ G r p 2 ( H ) G \in\mathrm{Grp}_2(\mathbf{H}) G ∈ Grp 2 ( H ) for k = 1 k = 1 k = 1 via Definition 4.6 and (15) ). The k k k -th differential cohomology of X ∈ H X \in\mathbf{H} X ∈ H with coefficients in G G G is defined as
H d i f f k ( X ; G ) : = H 0 ( X ; B c o n k G ) = τ ≤ 0 H ( X , B c o n k G ) . H_{\mathrm{diff}}^k(X;G) := H^0(X;B_{\mathrm{con}}^kG)=\tau_{\leq0}\mathbf{H}(X,B_{\mathrm{con}}^kG). H diff k ( X ; G ) := H 0 ( X ; B con k G ) = τ ≤ 0 H ( X , B con k G ) . By the defining ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback square of B c o n k G B_{\mathrm{con}}^kG B con k G (Definition 4.7), H d i f f k ( X ; G ) H_{\mathrm{diff}}^k(X;G) H diff k ( X ; G ) comes equipped with the canonical characteristic class map and curvature map :
char k = ( p c o n k ) ∗ : H d i f f k ( X ; G ) ⟶ H k ( X ; G ) , curv k : H d i f f k ( X ; G ) ⟶ H 0 ( X ; Ω c l k + 1 ( − ; G ) ) , \operatorname{char}_k=(p_{\mathrm{con}}^k)_*:H_{\mathrm{diff}}^k(X;G)\longrightarrow H^k(X;G),\qquad\operatorname{curv}_k:H_{\mathrm{diff}}^k(X;G)\longrightarrow H^0(X;\Omega_{\mathrm{cl}}^{k+1}(-;G)), char k = ( p con k ) ∗ : H diff k ( X ; G ) ⟶ H k ( X ; G ) , curv k : H diff k ( X ; G ) ⟶ H 0 ( X ; Ω cl k + 1 ( − ; G )) , fitting into a commutative square over the de Rham characteristic map $(\theta_{k+1})_*:H^k(X;G)\to H_{\mathrm{dR}}^{k+1}(X;G).
Now let us ask whether B c o n k G B_{\mathrm{con}}^kG B con k G can be viewed as the modal cohomology of an internal modality ( − ) c o n : ∗ / H → ∗ / H (-)_{\mathrm{con}} : */\mathbf{H} \to*/\mathbf{H} ( − ) con : ∗ / H → ∗ / H satisfying ( B H k G ) c o n ≃ B c o n k G (B_{\mathbf{H}}^kG)_{\mathrm{con}}\simeq B_{\mathrm{con}}^kG ( B H k G ) con ≃ B con k G . Observe that for k = 0 k=0 k = 0 , B c o n 0 A B_{\mathrm{con}}^0A B con 0 A is defined by the ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback square:
B c o n 0 A → curv 0 Ω c l 1 ( − ; A ) p c o n 0 ↓ ↓ σ 1 A → θ A ♭ d R B H A (18) \begin{CD}
B_{\mathrm{con}}^0A @>{\operatorname{curv}_0}>> \Omega_{\mathrm{cl}}^1(-;A) \\
@V{p_{\mathrm{con}}^0}VV @VV{\sigma_1}V \\
A @>{\theta_A}>> \flat_{\mathrm{dR}}B_{\mathbf{H}}A \tag*{(18)}
\end{CD} B con 0 A p con 0 ↓ ⏐ A curv 0 θ A Ω cl 1 ( − ; A ) ↓ ⏐ σ 1 ♭ dR B H A (18)
Thus, the only place where the group structure of A A A is used in (18) is in the target ♭ d R B H A \flat_{\mathrm{dR}}B_{\mathbf{H}}A ♭ dR B H A of the Maurer–Cartan morphism θ A : A → ♭ d R B H A \theta_A:A\to\flat_{\mathrm{dR}}B_{\mathbf{H}}A θ A : A → ♭ dR B H A ! To extend (18) to an arbitrary pointed object A ∈ ∗ / H A\in*/\mathbf{H} A ∈ ∗ / H , we must reformulate the Maurer–Cartan morphism without assuming that A A A is deloopable.
We construct two canonical reformulations of the Maurer–Cartan morphism for arbitrary pointed objects A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H : the first is valid in any Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos and produces a morphism with domain Ω A \Omega A Ω A , while the second uses the shape modality in a locally 1 1 1 -connected ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos to produce both a looped comparison map with codomain A A A and an unlooped unit morphism η A d R : A → ♭ d R ∫ d R A \eta^{\mathrm{dR}}_A : A \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A η A dR : A → ♭ dR ∫ dR A with domain A A A itself.
(1) The Flat/Looping Extension (θ ~ A ♭ : Ω A → ♭ d R A \widetilde{\theta}^{\flat}_A : \Omega A \to\flat_{\mathrm{dR}}A θ A ♭ : Ω A → ♭ dR A ): Recall from Theorem 3.8 that looping Ω \Omega Ω and delooping B H B_{\mathbf{H}} B H form an equivalence between 0 0 0 -truncated group objects Grp ( τ ≤ 0 H ) \operatorname{Grp}(\tau_{\leq0}\mathbf{H}) Grp ( τ ≤ 0 H ) and 0 0 0 -connected pointed objects Conn ( ∗ / H ) \operatorname{Conn}(* / \mathbf{H}) Conn ( ∗ / H ) (and on all of Grp 2 ( H ) \operatorname{Grp}_2(\mathbf{H}) Grp 2 ( H ) with connected pointed 2 2 2 -stacks Conn ( ∗ / Grp 2 e f f ( H ) ) \operatorname{Conn}(* / \operatorname{Grp}^{\mathrm{eff}}_2(\mathbf{H})) Conn ( ∗ / Grp 2 eff ( H )) ). Applying the looping functor Ω \Omega Ω to the Maurer–Cartan morphism θ G : G → ♭ d R B H G \theta_G : G \to\flat_{\mathrm{dR}}B_{\mathbf{H}}G θ G : G → ♭ dR B H G of a group object G G G yields a morphism Ω θ G : Ω G → Ω ♭ d R B H G \Omega\theta_G : \Omega G \to\Omega\flat_{\mathrm{dR}}B_{\mathbf{H}}G Ω θ G : Ω G → Ω ♭ dR B H G . Now observe that both ♭ d R X = ∗ × X ♭ X \flat_{\mathrm{dR}}X = * \mathbin{\times}_X \flat X ♭ dR X = ∗ × X ♭ X and Ω X = ∗ × X ∗ \Omega X = * \mathbin{\times}_X * Ω X = ∗ × X ∗ are defined as ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks, and ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks commute with one another and with the right-exact/left-exact flat modality ♭ \flat ♭ :
Ω ♭ d R B H G ≃ ♭ d R Ω B H G ≃ ♭ d R G , \Omega\flat_{\mathrm{dR}}B_{\mathbf{H}}G \simeq\flat_{\mathrm{dR}}\Omega B_{\mathbf{H}}G \simeq\flat_{\mathrm{dR}}G, Ω ♭ dR B H G ≃ ♭ dR Ω B H G ≃ ♭ dR G ,
where in the second step we used the equivalence Ω B H G ≃ G \Omega B_{\mathbf{H}}G \simeq G Ω B H G ≃ G . More directly, for any pointed object ( A , x : ∗ → A ) ∈ ∗ / H (A,x : * \to A) \in*/\mathbf{H} ( A , x : ∗ → A ) ∈ ∗ / H , since ♭ ∗ ≃ ∗ \flat* \simeq* ♭ ∗ ≃ ∗ and ϵ A ∘ ♭ x ≃ x \epsilon_A \circ\flat x \simeq x ϵ A ∘ ♭ x ≃ x , the universal property of the pullback ♭ d R A = ∗ × A ♭ A \flat_{\mathrm{dR}}A = * \mathbin{\times}_A \flat A ♭ dR A = ∗ × A ♭ A immediately induces a canonical morphism
θ ~ A ♭ : Ω A = ∗ × A ∗ ⟶ ∗ × A ♭ A = ♭ d R A , (19) \widetilde{\theta}^{\flat}_A : \Omega A = * \mathbin{\times}_A * \longrightarrow* \mathbin{\times}_A \flat A = \flat_{\mathrm{dR}}A, \tag*{(19)} θ A ♭ : Ω A = ∗ × A ∗ ⟶ ∗ × A ♭ A = ♭ dR A , (19)
which makes sense for every pointed object A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H without any group structure.6 Notice carefully the degree shift in the domain of (19) : θ ~ A ♭ \widetilde{\theta}^{\flat}_A θ A ♭ has domain Ω A \Omega A Ω A rather than A A A . Consequently, pulling back a closed 1-form coefficient σ 1 : Ω c l 1 ( − ; Ω A ) → ♭ d R A \sigma_1 : \Omega^1_{\mathrm{cl}}(-;\Omega A) \to\flat_{\mathrm{dR}}A σ 1 : Ω cl 1 ( − ; Ω A ) → ♭ dR A along θ ~ A ♭ : Ω A → ♭ d R A \widetilde{\theta}^{\flat}_A : \Omega A \to\flat_{\mathrm{dR}}A θ A ♭ : Ω A → ♭ dR A produces the 0 0 0 -connection object B c o n 0 ( Ω A ) → Ω A B^0_{\mathrm{con}}(\Omega A) \to\Omega A B con 0 ( Ω A ) → Ω A over Ω A \Omega A Ω A : when A = B H G A = B_{\mathbf{H}}G A = B H G , θ ~ B H G ♭ : G ≃ Ω B H G → ♭ d R B H G \widetilde{\theta}^{\flat}_{B_{\mathbf{H}}G} : G \simeq\Omega B_{\mathbf{H}}G \to\flat_{\mathrm{dR}}B_{\mathbf{H}}G θ B H G ♭ : G ≃ Ω B H G → ♭ dR B H G is literally θ G \theta_G θ G , so this pullback recovers B c o n 0 G → G B^0_{\mathrm{con}}G \to G B con 0 G → G (whereas when A = B H 2 G A = B^2_{\mathbf{H}}G A = B H 2 G for G ∈ BrGrp 2 ( H ) G \in\operatorname{BrGrp}_2(\mathbf{H}) G ∈ BrGrp 2 ( H ) , θ ~ B H 2 G ♭ : B H G ≃ Ω B H 2 G → ♭ d R B H 2 G \widetilde{\theta}^{\flat}_{B^2_{\mathbf{H}}G} : B_{\mathbf{H}}G \simeq\Omega B^2_{\mathbf{H}}G \to\flat_{\mathrm{dR}}B^2_{\mathbf{H}}G θ B H 2 G ♭ : B H G ≃ Ω B H 2 G → ♭ dR B H 2 G is θ 2 \theta_2 θ 2 , whose pullback with Ω c l 2 ( − ; G ) → ♭ d R B H 2 G \Omega^2_{\mathrm{cl}}(-;G) \to\flat_{\mathrm{dR}}B^2_{\mathbf{H}}G Ω cl 2 ( − ; G ) → ♭ dR B H 2 G recovers B c o n 1 G → B H G B^1_{\mathrm{con}}G \to B_{\mathbf{H}}G B con 1 G → B H G ).
(2) The Shape/De Rham Adjunction Extension (η A d R : A → ♭ d R ∫ d R A \eta^{\mathrm{dR}}_A : A \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A η A dR : A → ♭ dR ∫ dR A and θ ~ A ∫ : Ω ∫ d R A → A \widetilde{\theta}^{\int}_{A} : \Omega\int_{\mathrm{dR}} A \to A θ A ∫ : Ω ∫ dR A → A ): Suppose H \mathbf{H} H is 1 1 1 -connected and locally 1 1 1 -connected, so that we have the adjunction ∫ ⊣ ♭ \int\dashv\flat ∫ ⊣ ♭ with unit η A ∫ : A → ∫ A \eta^{\int}_A : A \to\int A η A ∫ : A → ∫ A and ♭ ∫ A ≃ ∫ A \flat\int A \simeq\int A ♭ ∫ A ≃ ∫ A . Dual to the homotopy kernel ♭ d R X = fib ( ♭ X → X ) \flat_{\mathrm{dR}}X = \operatorname{fib}(\flat X \to X) ♭ dR X = fib ( ♭ X → X ) , define for any pointed object ( A , x : ∗ → A ) (A,x : * \to A) ( A , x : ∗ → A ) its de Rham shape ∫ d R A \int_{\mathrm{dR}} A ∫ dR A as the homotopy cofiber ( ( 2 , 1 ) ((2,1) (( 2 , 1 ) -coequalizer / pushout) ) ) of A → ∫ A A \to\int A A → ∫ A :
A → η A ∫ ∫ A ↓ ↓ ∗ H → ∫ d R A \begin{CD}
A @>{\eta^{\int}_A}>> \int A \\
@VVV @VVV \\
*_{\mathbf{H}} @>>> \int_{\mathrm{dR}} A
\end{CD} A ↓ ⏐ ∗ H η A ∫ ∫ A ↓ ⏐ ∫ dR A
By the universal property of pushouts/pullbacks under the adjunction ∫ ⊣ ♭ \int\dashv\flat ∫ ⊣ ♭ , for any pointed objects A , Y ∈ ∗ / H A,Y \in*/\mathbf{H} A , Y ∈ ∗ / H there is a natural adjunction equivalence
( ∗ / H ) ( ∫ d R A , Y ) ≃ ( ∗ / H ) ( A , ♭ d R Y ) , (* /\mathbf{H})\left(\int_{\mathrm{dR}} A,Y\right) \simeq(* /\mathbf{H})\left(A,\flat_{\mathrm{dR}}Y\right), ( ∗ / H ) ( ∫ dR A , Y ) ≃ ( ∗ / H ) ( A , ♭ dR Y ) ,
exhibiting ∫ d R ⊣ ♭ d R \int_{\mathrm{dR}} \dashv\flat_{\mathrm{dR}} ∫ dR ⊣ ♭ dR on ∗ / H */\mathbf{H} ∗ / H . This adjunction provides a universal unit morphism with domain A A A itself for every pointed object A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H :
η A d R : A ⟶ ♭ d R ∫ d R A , (20) \eta^{\mathrm{dR}}_A : A \longrightarrow\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A, \tag*{(20)} η A dR : A ⟶ ♭ dR ∫ dR A , (20)
as well as a universal counit ϵ Y d R : ∫ d R ♭ d R Y → Y \epsilon^{\mathrm{dR}}_Y : \int_{\mathrm{dR}}\flat_{\mathrm{dR}}Y \to Y ϵ Y dR : ∫ dR ♭ dR Y → Y . Moreover, for every pointed object A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H , the canonical fiber–cofiber boundary morphism of the cofiber sequence A → η A ∫ A \xrightarrow{\eta^{\int}_A} A η A ∫ ∫ A → ∫ d R A \int A \to\int_{\mathrm{dR}} A ∫ A → ∫ dR A composed with the projection fib ( η A ∫ ) → A \operatorname{fib}(\eta_A^{\int}) \to A fib ( η A ∫ ) → A defines a companion morphism with codomain A A A :
θ ~ A ∫ : Ω ∫ d R A ⟶ fib ( A → η A ∫ ∫ A ) ⟶ A . (21) \widetilde{\theta}_A^{\int}:\Omega\int_{\mathrm{dR}} A \longrightarrow\operatorname{fib}\left(A \xrightarrow{\eta_A^{\int}} \int A\right) \longrightarrow A. \tag*{(21)} θ A ∫ : Ω ∫ dR A ⟶ fib ( A η A ∫ ∫ A ) ⟶ A . (21)
When A ∈ Grp 2 ( H ) A \in\operatorname{Grp}_2(\mathbf{H}) A ∈ Grp 2 ( H ) is a group object, the Maurer–Cartan morphism θ A : A → ♭ d R B H A \theta_A:A\to\flat_{\mathrm{dR}}BH A θ A : A → ♭ dR B H A transposes uniquely under ∫ d R ⊣ ♭ d R \int_{\mathrm{dR}}\dashv\flat_{\mathrm{dR}} ∫ dR ⊣ ♭ dR to a comparison morphism ϑ A : ∫ d R A → B H A \vartheta_A:\int_{\mathrm{dR}}A\to BH A ϑ A : ∫ dR A → B H A , through which θ A \theta_A θ A factors as θ A = ♭ d R ( ϑ A ) ∘ η A d R : A → ♭ d R ∫ d R A → ♭ d R B H A \theta_A=\flat_{\mathrm{dR}}(\vartheta_A)\circ\eta_A^{\mathrm{dR}}:A\to\flat_{\mathrm{dR}}\int_{\mathrm{dR}}A\to\flat_{\mathrm{dR}}BH A θ A = ♭ dR ( ϑ A ) ∘ η A dR : A → ♭ dR ∫ dR A → ♭ dR B H A ; applying the looping functor Ω \Omega Ω to ϑ A \vartheta_A ϑ A and using Ω B H A ≃ A \Omega BH A\simeq A Ω B H A ≃ A (or the adjunction B H ⊣ Ω BH\dashv\Omega B H ⊣ Ω ) then recovers θ ~ A ∫ ≃ Ω ϑ A : Ω ∫ d R A → Ω B H A ≃ A \widetilde{\theta}_A^{\int}\simeq\Omega\vartheta_A:\Omega\int_{\mathrm{dR}}A\to\Omega BH A\simeq A θ A ∫ ≃ Ω ϑ A : Ω ∫ dR A → Ω B H A ≃ A .
Proposition 4.19 (Adjunction of Extended Maurer–Cartan Constructions). In any 1 1 1 -connected locally 1 1 1 -connected Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , the de Rham shape and flat de Rham functors form an adjoint pair ∫ d R ⊣ ♭ d R : ∗ / H → ∗ / H \int_{\mathrm{dR}}\dashv\flat_{\mathrm{dR}}:\ast/\mathbf{H}\to\ast/\mathbf{H} ∫ dR ⊣ ♭ dR : ∗ / H → ∗ / H induced by ∫ ⊣ ♭ \int\dashv\flat ∫ ⊣ ♭ . Under this adjunction and the loop–delooping adjunction, the universal morphisms θ ~ ♭ \widetilde{\theta}^{\flat} θ ♭ of (19) , η d R \eta^{\mathrm{dR}} η dR of (20) , and θ ~ ∫ \widetilde{\theta}^{\int} θ ∫ of (21) are mutual mates.
Proof. Since H \mathbf{H} H is 1 1 1 -connected and locally 1 1 1 -connected, Γ ∘ L C s t ≃ id Grpd 2 \Gamma\circ\mathrm{LCst}\simeq\operatorname{id}_{\operatorname{Grpd}_2} Γ ∘ LCst ≃ id Grpd 2 , the unit at the terminal object η ∗ ∫ : ∗ → ∼ ∫ ∗ \eta_\ast^{\int}:\ast\xrightarrow{\sim}\int\ast η ∗ ∫ : ∗ ∼ ∫ ∗ is an equivalence, and ϵ ∫ A : ♭ ∫ A → ∼ ∫ A \epsilon_{\int A}:\flat\int A\xrightarrow{\sim}\int A ϵ ∫ A : ♭ ∫ A ∼ ∫ A is a 2 2 2 -natural equivalence. Explicitly, under the unpointed 2 2 2 -adjunction equivalence H ( ∫ A , Y ) ≃ H ( A , ♭ Y ) \mathbf{H}(\int A,Y)\simeq\mathbf{H}(A,\flat Y) H ( ∫ A , Y ) ≃ H ( A , ♭ Y ) sending g : ∫ A → Y g:\int A\to Y g : ∫ A → Y to g ~ = ♭ ( g ) ∘ ϵ ∫ A − 1 ∘ η A ∫ : A → ♭ Y \widetilde{g}=\flat(g)\circ\epsilon_{\int A}^{-1}\circ\eta_A^{\int}:A\to\flat Y g = ♭ ( g ) ∘ ϵ ∫ A − 1 ∘ η A ∫ : A → ♭ Y , naturality of ϵ : ♭ ⇒ id H \epsilon:\flat\Rightarrow\operatorname{id}_{\mathbf{H}} ϵ : ♭ ⇒ id H yields ϵ Y ∘ g ~ = ϵ Y ∘ ♭ ( g ) ∘ ϵ ∫ A − 1 ∘ η A ∫ ≃ g ∘ ϵ ∫ A ∘ ϵ ∫ A − 1 ∘ η A ∫ = g ∘ η A ∫ \epsilon_Y\circ\widetilde{g}=\epsilon_Y\circ\flat(g)\circ\epsilon_{\int A}^{-1}\circ\eta_A^{\int}\simeq g\circ\epsilon_{\int A}\circ\epsilon_{\int A}^{-1}\circ\eta_A^{\int}=g\circ\eta_A^{\int} ϵ Y ∘ g = ϵ Y ∘ ♭ ( g ) ∘ ϵ ∫ A − 1 ∘ η A ∫ ≃ g ∘ ϵ ∫ A ∘ ϵ ∫ A − 1 ∘ η A ∫ = g ∘ η A ∫ ; hence precomposition ( η A ∫ ) ∗ : H ( ∫ A , Y ) → H ( A , Y ) (\eta_A^{\int})^\ast:\mathbf{H}(\int A,Y)\to\mathbf{H}(A,Y) ( η A ∫ ) ∗ : H ( ∫ A , Y ) → H ( A , Y ) is canonically equivalent to postcomposition ( ϵ Y ) ∗ : H ( A , ♭ Y ) → H ( A , Y ) (\epsilon_Y)_\ast:\mathbf{H}(A,\flat Y)\to\mathbf{H}(A,Y) ( ϵ Y ) ∗ : H ( A , ♭ Y ) → H ( A , Y ) . For pointed objects ( A , x A ) , ( Y , y 0 ) ∈ ∗ / H (A,x_A),(Y,y_0)\in\ast/\mathbf{H} ( A , x A ) , ( Y , y 0 ) ∈ ∗ / H , since η ∗ ∫ : ∗ → ∼ ∫ ∗ \eta_\ast^{\int}:\ast\xrightarrow{\sim}\int\ast η ∗ ∫ : ∗ ∼ ∫ ∗ is an equivalence, the pushout ∫ d R A = ∗ ⊔ A ∫ A \int_{\mathrm{dR}}A=\ast\sqcup_A\int A ∫ dR A = ∗ ⊔ A ∫ A in H \mathbf{H} H coincides with the cofiber of η A ∫ : ( A , x A ) → ( ∫ A , ∫ x A ∘ η ∗ ∫ ) \eta_A^{\int}:(A,x_A)\to(\int A,\int x_A\circ\eta_\ast^{\int}) η A ∫ : ( A , x A ) → ( ∫ A , ∫ x A ∘ η ∗ ∫ ) in ∗ / H \ast/\mathbf{H} ∗ / H . Thus, the pointed hom-groupoid ( ∗ / H ) ( ∫ d R A , Y ) (\ast/\mathbf{H})(\int_{\mathrm{dR}}A,Y) ( ∗ / H ) ( ∫ dR A , Y ) is the homotopy fiber { y 0 ∘ ! A } × H ( A , Y ) H ( ∫ A , Y ) \{y_0\circ!_A\}\times_{\mathbf{H}(A,Y)}\mathbf{H}(\int A,Y) { y 0 ∘ ! A } × H ( A , Y ) H ( ∫ A , Y ) of ( η A ∫ ) ∗ (\eta_A^{\int})^\ast ( η A ∫ ) ∗ (whose trivialization along A A A restricts along x A : ∗ → A x_A:\ast\to A x A : ∗ → A to the basepoint datum on ∫ ∗ ≃ ∗ \int\ast\simeq\ast ∫ ∗ ≃ ∗ ), which under H ( ∫ A , Y ) ≃ H ( A , ♭ Y ) \mathbf{H}(\int A,Y)\simeq\mathbf{H}(A,\flat Y) H ( ∫ A , Y ) ≃ H ( A , ♭ Y ) is canonically equivalent to the pointed homotopy pullback ( ∗ / H ) ( A , ∗ × Y ♭ Y ) = ( ∗ / H ) ( A , ♭ d R Y ) (\ast/\mathbf{H})(A,\ast\times_Y\flat Y)=(\ast/\mathbf{H})(A,\flat_{\mathrm{dR}}Y) ( ∗ / H ) ( A , ∗ × Y ♭ Y ) = ( ∗ / H ) ( A , ♭ dR Y ) . This establishes ∫ d R ⊣ ♭ d R \int_{\mathrm{dR}}\dashv\flat_{\mathrm{dR}} ∫ dR ⊣ ♭ dR , with unit η A d R : A → ♭ d R ∫ d R A \eta_A^{\mathrm{dR}}:A\to\flat_{\mathrm{dR}}\int_{\mathrm{dR}}A η A dR : A → ♭ dR ∫ dR A corresponding to id ∫ d R A \operatorname{id}_{\int_{\mathrm{dR}}A} id ∫ dR A . Finally, taking Y = A Y=A Y = A for θ ~ A ♭ : Ω A → ♭ d R A \widetilde{\theta}_A^{\flat}:\Omega A\to\flat_{\mathrm{dR}}A θ A ♭ : Ω A → ♭ dR A , its left transpose under ∫ d R ⊣ ♭ d R \int_{\mathrm{dR}}\dashv\flat_{\mathrm{dR}} ∫ dR ⊣ ♭ dR is ∫ d R Ω A → A \int_{\mathrm{dR}}\Omega A\to A ∫ dR Ω A → A , which on connected pointed objects A = B H G A=BH G A = B H G gives ϑ G : ∫ d R G → B H G \vartheta_G:\int_{\mathrm{dR}}G\to BH G ϑ G : ∫ dR G → B H G satisfying θ G = ♭ d R ( ϑ G ) ∘ η G d R \theta_G=\flat_{\mathrm{dR}}(\vartheta_G)\circ\eta_G^{\mathrm{dR}} θ G = ♭ dR ( ϑ G ) ∘ η G dR . Since ϑ G \vartheta_G ϑ G maps the cofiber sequence G → ∫ d R G G\to\int_{\mathrm{dR}}G G → ∫ dR G to the fiber–cofiber sequence G → ∗ → B H G G\to\ast\to BH G G → ∗ → B H G , taking homotopy fibers (loops) identifies the boundary map θ ~ G ∫ : Ω ∫ d R G → fib ( η G d R ) → G \widetilde{\theta}_G^{\int}:\Omega\int_{\mathrm{dR}}G\to\operatorname{fib}(\eta_G^{\mathrm{dR}})\to G θ G ∫ : Ω ∫ dR G → fib ( η G dR ) → G of (21) with Ω ϑ G : Ω ∫ d R G → Ω B H G ≃ G \Omega\vartheta_G:\Omega\int_{\mathrm{dR}}G\to\Omega BH G\simeq G Ω ϑ G : Ω ∫ dR G → Ω B H G ≃ G . □
Because the unit η A d R : A → ♭ d R ∫ d R A \eta_A^{\mathrm{dR}}:A\to\flat_{\mathrm{dR}}\int_{\mathrm{dR}}A η A dR : A → ♭ dR ∫ dR A of (20) has domain A A A for every pointed object A ∈ ∗ / H A\in\ast/\mathbf{H} A ∈ ∗ / H , Proposition 4.19 immediately allows us to factor the universal Maurer–Cartan morphism θ k + 1 : B H k G → ♭ d R B H k + 1 G \theta_{k+1}:B_H^kG\to\flat_{\mathrm{dR}}B_H^{k+1}G θ k + 1 : B H k G → ♭ dR B H k + 1 G and its connection pullback through the group-independent unit η B H k G d R \eta_{B_H^kG}^{\mathrm{dR}} η B H k G dR :
Corollary 4.20 (Factorization of Connection Pullbacks via η d R \eta^{\mathrm{dR}} η dR ). *In any 1 1 1 -connected locally 1 1 1 -connected Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H , for every G ∈ Del 2 k + 1 ( H ) G\in\operatorname{Del}_{2}^{k+1}(\mathbf{H}) G ∈ Del 2 k + 1 ( H ) (k ≥ 0 k\geq0 k ≥ 0 ), the universal Maurer–Cartan morphism θ k + 1 : B H k G → ♭ d R B H k + 1 G \theta_{k+1}:B_H^kG\to\flat_{\mathrm{dR}}B_H^{k+1}G θ k + 1 : B H k G → ♭ dR B H k + 1 G factors via the unit η B H k G d R \eta_{B_H^kG}^{\mathrm{dR}} η B H k G dR of ∫ d R ⊣ ♭ d R \int_{\mathrm{dR}}\dashv\flat_{\mathrm{dR}} ∫ dR ⊣ ♭ dR as
θ k + 1 ≃ ♭ d R ( ϑ B H k G ) ∘ η B H k G d R : B H k G ⟶ ♭ d R ∫ d R B H k G ⟶ ♭ d R B H k + 1 G . \theta_{k+1}\simeq\flat_{\mathrm{dR}}(\vartheta_{B_H^kG})\circ\eta_{B_H^kG}^{\mathrm{dR}}:B_H^kG\longrightarrow\flat_{\mathrm{dR}}\int_{\mathrm{dR}}B_H^kG\longrightarrow\flat_{\mathrm{dR}}B_H^{k+1}G. θ k + 1 ≃ ♭ dR ( ϑ B H k G ) ∘ η B H k G dR : B H k G ⟶ ♭ dR ∫ dR B H k G ⟶ ♭ dR B H k + 1 G . Consequently, defining the pulled-back closed form coefficient object over ♭ d R ∫ d R B H k G \flat_{\mathrm{dR}}\int_{\mathrm{dR}}B_H^kG ♭ dR ∫ dR B H k G by
Ω ~ c l k + 1 ( − ; G ) : = ♭ d R ∫ d R B H k G × ♭ d R B H k + 1 G Ω c l k + 1 ( − ; G ) , \widetilde{\Omega}_{\mathrm{cl}}^{k+1}(-;G):=\flat_{\mathrm{dR}}\int_{\mathrm{dR}}B_H^kG\times_{\flat_{\mathrm{dR}}B_H^{k+1}G}\Omega_{\mathrm{cl}}^{k+1}(-;G), Ω cl k + 1 ( − ; G ) := ♭ dR ∫ dR B H k G × ♭ dR B H k + 1 G Ω cl k + 1 ( − ; G ) ,
the moduli stack of k k k -connections B c o n k G B^k_{\mathrm{con}}G B con k G is canonically presented by the pasted ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullback along the unit η B H k G d R \eta^{\mathrm{dR}}_{B^k_{\mathrm{H}}G} η B H k G dR }.
B c o n k G → Ω ~ c l k + 1 ( − ; G ) → Ω c l k + 1 ( − ; G ) p c o n k ↓ σ ~ k + 1 ↓ σ k + 1 ↓ B H k G → η B H k G d R ♭ d R ∫ d R B H k G → ♭ d R ( ϑ B H k G ) ♭ d R B H k + 1 G \begin{CD}
B^k_{\mathrm{con}}G @>>> \widetilde{\Omega}^{k+1}_{\mathrm{cl}}(-;G) @>>> \Omega^{k+1}_{\mathrm{cl}}(-;G) \\
@V p^k_{\mathrm{con}} VV @V \widetilde{\sigma}_{k+1} VV @V \sigma_{k+1} VV \\
B^k_{\mathrm{H}}G @>{\eta^{\mathrm{dR}}_{B^k_{\mathrm{H}}G}}>> \flat_{\mathrm{dR}}\int_{\mathrm{dR}} B^k_{\mathrm{H}}G @>{\flat_{\mathrm{dR}}(\vartheta^{B^k_{\mathrm{H}}G})}>> \flat_{\mathrm{dR}}B^{k+1}_{\mathrm{H}}G
\end{CD} B con k G p con k ↓ ⏐ B H k G η B H k G dR Ω cl k + 1 ( − ; G ) σ k + 1 ↓ ⏐ ♭ dR ∫ dR B H k G ♭ dR ( ϑ B H k G ) Ω cl k + 1 ( − ; G ) σ k + 1 ↓ ⏐ ♭ dR B H k + 1 G
Proof. Immediate from Proposition 4.19 , Definition 4.7 , and the pasting law for ( 2 , 1 ) (2,1) ( 2 , 1 ) -pullbacks.
Motivated by Corollary 4.20, pulling back a closed form coefficient morphism σ A : Ω c l ∙ ( − ; A ) → ♭ d R ∫ d R A \sigma_A : \Omega^\bullet_{\mathrm{cl}}(-;A) \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A σ A : Ω cl ∙ ( − ; A ) → ♭ dR ∫ dR A along η A d R \eta^{\mathrm{dR}}_A η A dR produces a pointed connection object ( A ) c o n , ♭ : = A × ♭ d R ∫ d R A Ω c l ∙ ( − ; A ) (A)_{\mathrm{con},\flat} := A \times_{\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A} \Omega^\bullet_{\mathrm{cl}}(-;A) ( A ) con , ♭ := A × ♭ dR ∫ dR A Ω cl ∙ ( − ; A ) directly over A A A for an arbitrary pointed object A ∈ ∗ / H A \in*/\mathbf{H} A ∈ ∗ / H , while dually using the counit ϵ A d R : ∫ d R ♭ d R A → A \epsilon^{\mathrm{dR}}_A : \int_{\mathrm{dR}}\flat_{\mathrm{dR}}A \to A ϵ A dR : ∫ dR ♭ dR A → A (or θ ~ A J : Ω ∫ d R A → A \widetilde{\theta}^{\mathrm{J}}_A : \Omega\int_{\mathrm{dR}}A \to A θ A J : Ω ∫ dR A → A ) produces a companion endofunctor ( A ) c o n , ∫ (A)_{\mathrm{con},\int} ( A ) con , ∫ :
Conjecture 4.21 (Pointed Connection Endofunctors for Non-Groupal Coefficients). In a 1 1 1 -connected locally 1 1 1 -connected Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H equipped with differential form coefficients, there exists a functorial differential form assignment A ↦ ( σ A : Ω c l ∙ ( − ; A ) → ♭ d R ∫ d R A ) A \mapsto(\sigma_A : \Omega^\bullet_{\mathrm{cl}}(-;A) \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A) A ↦ ( σ A : Ω cl ∙ ( − ; A ) → ♭ dR ∫ dR A ) on ∗ / H */\mathbf{H} ∗ / H restricting on A = B H k G A = B^k_{\mathrm{H}}G A = B H k G to σ ~ k + 1 : Ω ~ c l k + 1 ( − ; G ) → ♭ d R ∫ d R B H k G \widetilde{\sigma}_{k+1} : \widetilde{\Omega}^{k+1}_{\mathrm{cl}}(-;G) \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} B^k_{\mathrm{H}}G σ k + 1 : Ω cl k + 1 ( − ; G ) → ♭ dR ∫ dR B H k G of Corollary 4.20, such that the pullback of σ A \sigma_A σ A along η A d R : A → ♭ d R ∫ d R A \eta^{\mathrm{dR}}_A : A \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A η A dR : A → ♭ dR ∫ dR A defines a pointed connection endofunctor ( − ) c o n , ♭ : ∗ / H → ∗ / H (-)_{\mathrm{con},\flat} : */\mathbf{H} \to*/\mathbf{H} ( − ) con , ♭ : ∗ / H → ∗ / H (and dually via ϵ A d R \epsilon^{\mathrm{dR}}_A ϵ A dR and θ ~ A J \widetilde{\theta}^{\mathrm{J}}_A θ A J a companion endofunctor ( − ) c o n , ∫ : ∗ / H → ∗ / H (-)_{\mathrm{con},\int} : */\mathbf{H} \to*/\mathbf{H} ( − ) con , ∫ : ∗ / H → ∗ / H ) satisfying ( B H k G ) c o n , ♭ ≃ B c o n k G (B^k_{\mathrm{H}}G)_{\mathrm{con},\flat} \simeq B^k_{\mathrm{con}}G ( B H k G ) con , ♭ ≃ B con k G for all k ≥ 0 k \ge0 k ≥ 0 , with loop fiber Ω ( A c o n , ♭ ) ≃ fib ( Ω η A d R : Ω A → ♭ d R Ω ∫ d R A ) \Omega(A_{\mathrm{con},\flat}) \simeq\operatorname{fib}(\Omega\eta^{\mathrm{dR}}_A : \Omega A \to\flat_{\mathrm{dR}}\Omega\int_{\mathrm{dR}} A) Ω ( A con , ♭ ) ≃ fib ( Ω η A dR : Ω A → ♭ dR Ω ∫ dR A ) whenever Ω c l ∙ ( − ; A ) \Omega^\bullet_{\mathrm{cl}}(-;A) Ω cl ∙ ( − ; A ) is 0 0 0 -truncated (which on connected deloopings A = B H k G A = B^k_{\mathrm{H}}G A = B H k G with σ ~ k + 1 \widetilde{\sigma}_{k+1} σ k + 1 as in Corollary 4.20 reduces to fib ( θ ~ A J ) ≃ ♭ Ω A \operatorname{fib}(\widetilde{\theta}^{\mathrm{J}}_A) \simeq\flat\Omega A fib ( θ A J ) ≃ ♭Ω A ), while pulling back Ω c l ∙ ( − ; Ω A ) → ♭ d R A \Omega^\bullet_{\mathrm{cl}}(-;\Omega A) \to\flat_{\mathrm{dR}}A Ω cl ∙ ( − ; Ω A ) → ♭ dR A along θ ~ A J : Ω A → ♭ d R A \widetilde{\theta}^{\mathrm{J}}_A : \Omega A \to\flat_{\mathrm{dR}}A θ A J : Ω A → ♭ dR A recovers B c o n 0 ( Ω A ) B^0_{\mathrm{con}}(\Omega A) B con 0 ( Ω A ) .
Example 4.23 (Extended Maurer–Cartan Unit in S m o o t h G r p 2 \mathbf{SmoothGrp}_{2} SmoothGrp 2 ). In H = S m o o t h G r p 2 \mathbf{H} = \mathbf{SmoothGrp}_{2} H = SmoothGrp 2 , let ( A , x 0 : ∗ → A ) (A,x_0 : * \to A) ( A , x 0 : ∗ → A ) be a pointed 0 0 0 -truncated smooth sheaf (for instance, a pointed smooth manifold or a smooth monoid M ∈ Mon 2 ( S m o o t h G r p 2 ) M \in\operatorname{Mon}_2(\mathbf{SmoothGrp}_{2}) M ∈ Mon 2 ( SmoothGrp 2 ) pointed at its identity 1 M 1_M 1 M ). Its shape ∫ A = LCst ( Π 1 ( A ) ) \int A = \operatorname{LCst}(\Pi_1(A)) ∫ A = LCst ( Π 1 ( A )) is the locally constant stack associated to the fundamental 1 1 1 -groupoid Π 1 ( A ) \Pi_1(A) Π 1 ( A ) of smooth paths in A A A up to thin/smooth homotopy, and the unit η A ∫ : A → ∫ A \eta^{\int}_A : A \to\int A η A ∫ : A → ∫ A sends a smooth plot f : U → A f : U \to A f : U → A to the path-component/constant-path family at f f f . Consequently, the de Rham shape ∫ d R A = ∗ ⊔ ∫ A ∫ A \int_{\mathrm{dR}} A = * \sqcup_{\int A} \int A ∫ dR A = ∗ ⊔ ∫ A ∫ A collapses the points of A A A in each plot to the basepoint while retaining the non-trivial smooth path-homotopy loops generated by smooth variation across A A A , and the flat de Rham fiber ♭ d R ∫ d R A = ∗ × ∫ d R A ♭ ∫ d R A \flat_{\mathrm{dR}}\int_{\mathrm{dR}} A = * \times_{\int_{\mathrm{dR}} A} \flat\int_{\mathrm{dR}} A ♭ dR ∫ dR A = ∗ × ∫ dR A ♭ ∫ dR A is the stack of smooth A A A -plots trivialized inside the discrete fundamental groupoid ♭ ∫ d R A \flat\int_{\mathrm{dR}} A ♭ ∫ dR A . The universal unit η A d R : A → ♭ d R ∫ d R A \eta^{\mathrm{dR}}_A : A \to\flat_{\mathrm{dR}}\int_{\mathrm{dR}} A η A dR : A → ♭ dR ∫ dR A of
(20) sends a smooth plot f : U → A f: U \to A f : U → A to the canonical path-parallel-transport discrepancy between f f f and its locally constant flattening ♭ f \flat f ♭ f , without requiring A A A to have inverses or group multiplication. When A = M ∈ Mon 2 ( SmoothGrpd 2 ) A=M\in\operatorname{Mon}_{2}(\operatorname{SmoothGrpd}_{2}) A = M ∈ Mon 2 ( SmoothGrpd 2 ) is a non-groupal smooth monoid (such as ( M n ( R ) , ⋅ , I n ) (M_n(\mathbb{R}),\cdot,I_n) ( M n ( R ) , ⋅ , I n ) ), η M d R : M → ♭ d R ∫ d R M \eta^{\mathrm{dR}}_M:M\to\flat_{\mathrm{dR}}\int_{\mathrm{dR}}M η M dR : M → ♭ dR ∫ dR M retains the full pointed smooth structure of M M M before comparison with the group completion gr ( M ) \operatorname{gr}(M) gr ( M ) of Remark 4.22 (which for ( M n ( R ) , ⋅ ) (M_n(\mathbb{R}),\cdot) ( M n ( R ) , ⋅ ) either localizes to G L n ( R ) GL_n(\mathbb{R}) G L n ( R ) on units or collapses non-invertible matrices).
Classified Structures
Recall that in classical Grothendieck 1-topos theory [29 , 15 ] , a geometric theory T \mathbb{T} T possesses a classifying 1-topos Set [ T ] \operatorname{Set}[\mathbb{T}] Set [ T ] equipped with a universal T \mathbb{T} T -model M T ∈ Mod Set [ T ] ( T ) M_{\mathbb{T}}\in\operatorname{Mod}_{\operatorname{Set}[\mathbb{T}]}(\mathbb{T}) M T ∈ Mod Set [ T ] ( T ) such that for every Grothendieck 1-topos E \mathcal{E} E , pulling back M T M_{\mathbb{T}} M T along geometric morphisms induces an equivalence of categories
Mod E ( T ) ≃ 1 ShvTopos 2 ( E , Set [ T ] ) . \operatorname{Mod}_{\mathcal{E}}(\mathbb{T})\simeq1\operatorname{ShvTopos}_{2}(\mathcal{E},\operatorname{Set}[\mathbb{T}]). Mod E ( T ) ≃ 1 ShvTopos 2 ( E , Set [ T ]) .
Passing to isomorphism classes of models ( τ ≤ 0 ) (\tau_{\leq0}) ( τ ≤ 0 ) , we recognize the right-hand side as the 0-th intrinsic cohomology of E \mathcal{E} E with coefficients in Set [ T ] \operatorname{Set}[\mathbb{T}] Set [ T ] inside the 2-category 1 ShvTopos 2 1\operatorname{ShvTopos}_{2} 1 ShvTopos 2 :
Iso ( Mod E ( T ) ) ≅ τ ≤ 0 ( 1 ShvTopos 2 ( E , Set [ T ] ) ) = H 1 ShvTopos 2 0 ( E ; Set [ T ] ) . \operatorname{Iso}(\operatorname{Mod}_{\mathcal{E}}(\mathbb{T}))\cong\tau_{\leq0}(1\operatorname{ShvTopos}_{2}(\mathcal{E},\operatorname{Set}[\mathbb{T}]))=H^{0}_{1\operatorname{ShvTopos}_{2}}(\mathcal{E};\operatorname{Set}[\mathbb{T}]). Iso ( Mod E ( T )) ≅ τ ≤ 0 ( 1 ShvTopos 2 ( E , Set [ T ])) = H 1 ShvTopos 2 0 ( E ; Set [ T ]) .
Crucially, this classifying paradigm can be reproduced internally within any cohesive ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H (replacing 1 ShvTopos 2 1\operatorname{ShvTopos}_{2} 1 ShvTopos 2 by H \mathbf{H} H , test topoi E \mathcal{E} E by objects X ∈ H X\in\mathbf{H} X ∈ H , and Set [ T ] \operatorname{Set}[\mathbb{T}] Set [ T ] by a classifying stack A ∈ H A\in\mathbf{H} A ∈ H with universal structure E → A E\to A E → A ; see [33 Section 9.3] and [38 Section 5.1.10.1] ). Specifically, given a 2-functor Struct A : H o p → Grpd 2 \operatorname{Struct}_{A}:\mathbf{H}^{\mathrm{op}}\to\operatorname{Grpd}_{2} Struct A : H op → Grpd 2 assigning to each X ∈ H X\in\mathbf{H} X ∈ H the groupoid of geometric A A A -structures over X X X , an object A c l a s s ∈ H A_{\mathrm{class}}\in\mathbf{H} A class ∈ H together with a universal structure u A ∈ Struct A ( A c l a s s ) u_A\in\operatorname{Struct}_{A}(A_{\mathrm{class}}) u A ∈ Struct A ( A class ) classifies Struct A \operatorname{Struct}_{A} Struct A if pulling back u A u_A u A induces a 2-natural equivalence of groupoids
H ( X , A c l a s s ) → ≃ Struct A ( X ) , \mathbf{H}(X,A_{\mathrm{class}})\xrightarrow{\simeq}\operatorname{Struct}_{A}(X), H ( X , A class ) ≃ Struct A ( X ) ,
and hence on equivalence classes a bijection with intrinsic cohomology:
τ ≤ 0 Struct A ( X ) ≅ H 0 ( X ; A c l a s s ) . \tau_{\leq0}\operatorname{Struct}_{A}(X)\cong H^{0}(X;A_{\mathrm{class}}). τ ≤ 0 Struct A ( X ) ≅ H 0 ( X ; A class ) .
Table 1 summarizes the dictionary between the classifying objects constructed in § 3 and § 4 , the corresponding intrinsic/modal/differential cohomologies, and the geometric structures they classify over an object X ∈ H X\in\mathbf{H} X ∈ H .
Table 1. Summary of classifying objects, cohomologies, and classified geometric structures in a cohesive Grothendieck ( 2 , 1 ) (2,1) ( 2 , 1 ) -topos H \mathbf{H} H .
Lemma Group, Brazil
Email address: ximenesyuri@gmx.ie
UTFPR, Curitiba, Brazil
Email address: willianvelasco@utfpr.edu.br