OctoInformBeta

OpenAI preprint, September 2026 · 2. Globular models and the weak-equivalence test

§2 (setting): homotopy sets, components and based groups: Fixed boundaries and iterated identities determine positive-degree homotopy groups

Homotopy quotients of cells in a model.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

The homotopy sets of a globular model concern cells with specified boundary data. Their based versions have positive degree, while components concern vertices.

The globe category 𝑮\mathbf G has an object gjg_j for each natural number jj.✦ Its morphisms are identities and, for m<jm<j, the source and target faces from gmg_m to gjg_j.✦ A globular presheaf PP is a contravariant set-valued functor on 𝑮\mathbf G.✦ The underlying globular set of a globular presheaf PP on 𝑮\mathbf G is the family Pj=P(gj)P_j=P(g_j), with source and target maps s,t:Pj+1→Pjs,t:P_{j+1}\to P_j induced by the faces and satisfying (1).✦ A globular map between globular sets is a family of cell maps commuting with their source and target maps.✦ The underlying globular map of a morphism f:P→Qf:P\to Q of globular presheaves is the family of its components at the globes, with the component at gjg_j giving the map on jj-cells. s(s(a))=s(t(a)),t(s(a))=t(t(a))(a∈Pj+2).s(s(a))=s(t(a)),\qquad t(s(a))=t(t(a)) \quad(a\in P_{j+2}).(1)

The representable disk Dj=𝑮(−,gj)D_j=\mathbf G(-,g_j) is the single-globe arity of dimension jj. The source and target face maps σm,j,τm,j:Dm→Dj\sigma_{m,j},\tau_{m,j}:D_m\to D_j of the representable disk DjD_j, for m<jm<j, are the Yoneda images of the globe faces.✦ A table of dimensions describes a finite globular sum SS by starting with one disk and successively attaching a new disk DhD_h along a bb-face.✦ If the previous last disk has dimension ℓ\ell, an attachment requires b<ℓb<\ell and b<hb<h. The realization |S||S| of a finite globular sum SS is the globular presheaf obtained by the disk attachments of its table of dimensions, with its distinguished last globe inclusion λ:Dℓ→|S|\lambda:D_\ell\to |S|; the next attachment S′S' is the pushout in (2).✦ |S′|=|S|∐DbDh,Db→λ∘τb,ℓ|S|,Db→σb,hDh.\begin{gathered} |S'|=|S|\amalg_{D_b}D_h,\\ D_b\xrightarrow{\lambda\circ\tau_{b,\ell}}|S|, \qquad D_b\xrightarrow{\sigma_{b,h}}D_h. \end{gathered}(2) The previous arity inclusion and the attached globe inclusion are the two pushout maps in (2); the latter becomes the new last globe inclusion. The gluing face inclusion is the target bb-face of the previous last globe.✦ The height of a table, and of its arity SS, is the greatest disk dimension occurring in it.

A globular theory CC is a category with finite globular sums as objects and an arity functor ι\iota from their category that fixes objects and sends each elementary gluing to a pushout.✦ The morphisms of a globular theory CC are called operations. A theory morphism between globular theories is a functor fixing arities and their structural maps; identities and compositions of theory morphisms are those of functors.

Two operations Dj→SD_j\to S are parallel when their source and target restrictions agree; in dimension zero every pair is parallel.✦ An admissible pair is a parallel pair a,b:Dj→Sa,b:D_j\to S with the height of SS at most j+1j+1.✦ A filler z:Dj+1→Sz:D_{j+1}\to S of an admissible pair a,b:Dj→Sa,b:D_j\to S is an operation whose source restriction is aa and whose target restriction is bb.✦

A free extension by fillers is a theory morphism adjoining specified fillers for a family of admissible pairs, with the following universal property: a morphism from the old theory to any theory EE, together with chosen fillers of the image pairs in EE, extends uniquely to a morphism from the new theory sending the adjoined fillers to the chosen ones.✦ A cellular presentation of a globular theory CC is a sequence of free extensions by fillers starting at an initial theory, with compatible inclusions of the stages into CC exhibiting CC as their colimit.✦ A Grothendieck coherator is a globular theory CC that admits a cellular presentation and has a filler for every admissible pair.✦

A model XX of CC is a presheaf on CC sending elementary globular sums to pullbacks.✦ The globe inclusion gj↦ιDjg_j\mapsto\iota D_j is the composite of the Yoneda globe inclusion into arities with ι\iota.✦ The underlying globular set of a model XX of CC is the restriction of XX along the globe inclusion; its jj-cell set is Xj=X(ιDj)X_j=X(\iota D_j).✦ A model morphism f:X→Yf:X\to Y induces the globular map with dimensionwise components fj:Xj→Yjf_j:X_j\to Y_j.✦

Fixed Boundaries and Based Cells

Let CC be a Grothendieck coherator and XX a model of CC. For a natural number nn and cells a,b∈Xna,b\in X_n, the set En(a,b)E_n(a,b) of connecting cells consists of the (n+1)(n+1)-cells with source aa and target bb, as in (3). En(a,b)={e∈Xn+1∣s(e)=a,t(e)=b}.E_n(a,b)=\{e\in X_{n+1}\mid s(e)=a,\ t(e)=b\}.(3)✦

Fix a vertex x∈X0x\in X_0. The iterated identity at xx is the family of cells uj(x)∈Xju_j(x)\in X_j defined by (4), where 1a1_a denotes the identity cell on a cell aa. u0(x)=x,uj+1(x)=1uj(x).u_0(x)=x,\qquad u_{j+1}(x)=1_{u_j(x)}.(4)✦ The based loop set Ln+1(X,x)L_{n+1}(X,x) is the set of (n+1)(n+1)-cells whose source and target are both un(x)u_n(x), as in (5).✦ Ln+1(X,x)=En(un(x),un(x)).L_{n+1}(X,x)=E_n(u_n(x),u_n(x)).(5)✦ For a model morphism f:X→Yf:X\to Y, the induced map on based loops is the restriction of fn+1f_{n+1}, with target basepoint f0(x)f_0(x).✦

The Claim

Definition 1. Let CC be a Grothendieck coherator, let X,YX,Y be models of CC, and let f:X→Yf:X\to Y be a model morphism. Cell homotopy ∼j\sim_j on the jj-cells of XX is the joining relation of (6); it satisfies (7). a∼jb⇔there exists h∈Xj+1 with s(h)=a and t(h)=b.a\sim_j b\quad\Longleftrightarrow\quad \text{there exists }h\in X_{j+1}\text{ with }s(h)=a\text{ and }t(h)=b.(6)✦ ∼jis an equivalence relation on Xj(j≥0).\sim_j\ \text{is an equivalence relation on }X_j \qquad(j\geq 0).(7)✦ For each n≥0n\geq0, the cell groupoid 𝒢n(X)\mathcal G_n(X) has the object set (8). Ob⁡𝒢n(X)=Xn.\operatorname{Ob}\mathcal G_n(X)=X_n.(8)✦ For objects a,ba,b, connecting-cell homotopy is the equivalence relation ≃a,b\simeq_{a,b} on En(a,b)E_n(a,b) in (9), whose quotient is the hom set. e≃a,be′⇔e∼n+1e′,𝒢n(X)(a,b)=En(a,b)/≃a,b.e\simeq_{a,b}e'\quad\Longleftrightarrow\quad e\sim_{n+1}e', \qquad \mathcal G_n(X)(a,b)=E_n(a,b)/{\simeq_{a,b}}.(9)✦ The connected component set π0(X)\pi_0(X) is the quotient in (10). π0(X)=X0/∼0.\pi_0(X)=X_0/{\sim_0}.(10)✦ For a vertex x∈X0x\in X_0, the based homotopy group πn+1(X,x)\pi_{n+1}(X,x) in positive degree n+1n+1 is the quotient in (11). πn+1(X,x)=Ln+1(X,x)/∼n+1(n≥0).\pi_{n+1}(X,x)=L_{n+1}(X,x)/{\sim_{n+1}} \qquad(n\geq0).(11)✦ The homotopy class map f*f_* induced by the model morphism f:X→Yf:X\to Y is the group homomorphism of (12). f*:πn+1(X,x)→πn+1(Y,f0(x)),[a]↦[fn+1(a)].f_*:\pi_{n+1}(X,x)\longrightarrow\pi_{n+1}(Y,f_0(x)), \qquad [a]\longmapsto[f_{n+1}(a)].(12)

In the paper, the result reads: parallel rr-cells are homotopic when an (r+1)(r+1)-cell joins them; for parallel (r−1)(r-1)-cells u,vu,v and r≥1r\geq1, πr(X;u,v)\pi_r(X;u,v) is the set of rr-cells from uu to vv modulo this relation; components are vertices modulo joining arrows, and πr(X,x)\pi_r(X,x) uses the iterated identity boundaries at xx; weak equivalences induce a bijection on components and isomorphisms on all these based groups.

What Is Checked

The constructions are stated over models of a coherator, with its cellular presentation and fillers for every admissible pair; model morphisms act on joining cells by their corresponding dimensionwise components.