ConceptComplete

Kan Complex

A Kan complex is a simplicial set satisfying the Kan extension condition: every horn can be filled. Kan complexes are the simplicial analogue of topological spaces -- they model homotopy types combinatorially. The singular simplicial set of any topological space is a Kan complex, and conversely, the geometric realization of a Kan complex has the "correct" homotopy type. Kan complexes are also precisely the \infty-groupoids in the theory of quasi-categories.


Horns

Definition1.1Horn

The (n,k)(n,k)-horn Λkn\Lambda^n_k (for n1n \geq 1, 0kn0 \leq k \leq n) is the simplicial subset of Δ[n]\Delta[n] obtained by removing the interior and the kk-th face. Formally:

Λkn=ikdi(Δ[n1])Δ[n]\Lambda^n_k = \bigcup_{i \neq k} d^i(\Delta[n-1]) \subseteq \Delta[n]

where di:Δ[n1]Δ[n]d^i: \Delta[n-1] \to \Delta[n] is the ii-th face inclusion. An mm-simplex φ:[m][n]\varphi: [m] \to [n] belongs to Λkn\Lambda^n_k if and only if φ\varphi is not surjective, or φ\varphi is surjective and its image misses some vertex other than kk.

ExampleThe horns of Delta[2]

The standard 22-simplex Δ[2]\Delta[2] has three horns:

  • Λ02\Lambda^2_0: edges {0,1}\{0,1\} and {0,2}\{0,2\} (the two edges meeting at vertex 00), missing the edge {1,2}\{1,2\} and the interior. This looks like a "V" opening toward the right.

  • Λ12\Lambda^2_1: edges {0,1}\{0,1\} and {1,2}\{1,2\} (the two edges meeting at vertex 11), missing the edge {0,2}\{0,2\} and the interior. This is the "inner horn."

  • Λ22\Lambda^2_2: edges {0,2}\{0,2\} and {1,2}\{1,2\} (the two edges meeting at vertex 22), missing the edge {0,1}\{0,1\} and the interior. This looks like a "V" opening toward the left.

Filling Λ12\Lambda^2_1 means finding a 22-simplex with the given edges -- geometrically, filling in a triangle. This is related to composition in a category.

ExampleHorns of Delta[3]

Δ[3]\Delta[3] has four horns Λ03,Λ13,Λ23,Λ33\Lambda^3_0, \Lambda^3_1, \Lambda^3_2, \Lambda^3_3. Each is obtained by removing one face of the tetrahedron. For instance, Λ13\Lambda^3_1 consists of the faces {0,1,2}\{0,1,2\}, {0,1,3}\{0,1,3\}, and {0,2,3}\{0,2,3\} -- everything except the face {1,2,3}\{1,2,3\}.

Filling a horn in Δ[3]\Delta[3] requires finding a 33-simplex (a solid tetrahedron) with the given faces. In the context of categories, this encodes associativity of composition.


The Kan Condition

Definition1.2Kan complex

A simplicial set XX is a Kan complex if every horn has a filler. That is, for every n1n \geq 1 and 0kn0 \leq k \leq n, every map ΛknX\Lambda^n_k \to X extends to a map Δ[n]X\Delta[n] \to X:

For every ΛknX, there exists Δ[n]X making the diagram commute.\text{For every } \Lambda^n_k \to X, \text{ there exists } \Delta[n] \to X \text{ making the diagram commute.}

Equivalently, the restriction map Hom(Δ[n],X)Hom(Λkn,X)\operatorname{Hom}(\Delta[n], X) \to \operatorname{Hom}(\Lambda^n_k, X) is surjective for all n1n \geq 1 and 0kn0 \leq k \leq n.

Note: the filler is required to exist but is not required to be unique. This non-uniqueness corresponds to the non-uniqueness of homotopies in topology.

ExampleSingular simplicial sets are Kan complexes

For any topological space YY, the singular simplicial set Sing(Y)\operatorname{Sing}(Y) is a Kan complex. Given a horn ΛknSing(Y)\Lambda^n_k \to \operatorname{Sing}(Y), which is a continuous map ΛknY|\Lambda^n_k| \to Y, we need to extend it to ΔnY|\Delta^n| \to Y. Since Λkn|\Lambda^n_k| is a retract of Δn|\Delta^n| (the topological horn can be deformation-retracted to the boundary missing one face, and then extended), such an extension always exists.

More precisely, there is a retraction r:ΔnΛknr: |\Delta^n| \to |\Lambda^n_k|, so frf \circ r gives the desired extension.

ExampleNerves of non-groupoid categories are not Kan

Consider the category [1]=(01)[1] = (0 \to 1) with one non-identity morphism. Its nerve N([1])=Δ[1]N([1]) = \Delta[1]. The horn Λ02Δ[1]\Lambda^2_0 \to \Delta[1] given by the identity id:00\mathrm{id}: 0 \to 0 on edge {0,1}\{0,1\} and the morphism f:01f: 0 \to 1 on edge {0,2}\{0,2\} would require a filler -- a 22-simplex with d0d_0 being a morphism 010 \to 1 and d2d_2 being id0\mathrm{id}_0. This filler does exist. But the horn Λ22\Lambda^2_2 with edge {0,2}=f\{0,2\} = f and {1,2}=id1\{1,2\} = \mathrm{id}_1 would require finding a morphism 010 \to 1 composing correctly, which may not exist in general categories.

In fact, N(C)N(\mathcal{C}) is a Kan complex if and only if C\mathcal{C} is a groupoid (every morphism is invertible). The outer horn fillers require inverses.

ExampleNerves of groupoids are Kan complexes

If G\mathcal{G} is a groupoid, then N(G)N(\mathcal{G}) is a Kan complex. Given any horn ΛknN(G)\Lambda^n_k \to N(\mathcal{G}), the missing face can be filled using compositions and inverses of the given morphisms.

For instance, a Λ02\Lambda^2_0-horn in N(G)N(\mathcal{G}) gives morphisms g:c1c2g: c_1 \to c_2 and h:c0c2h: c_0 \to c_2. The filler requires f:c0c1f: c_0 \to c_1 with gf=hg \circ f = h, which is f=g1hf = g^{-1} \circ h.

A Λ22\Lambda^2_2-horn gives f:c0c1f: c_0 \to c_1 and h:c0c2h: c_0 \to c_2. The filler requires g:c1c2g: c_1 \to c_2 with gf=hg \circ f = h, so g=hf1g = h \circ f^{-1}.


Homotopy Groups

Definition1.3Homotopy groups of a Kan complex

Let XX be a Kan complex with a basepoint xX0x \in X_0. The nn-th homotopy group πn(X,x)\pi_n(X, x) for n1n \geq 1 is defined as follows:

An element of πn(X,x)\pi_n(X, x) is represented by an nn-simplex σXn\sigma \in X_n such that di(σ)=sn1s0(x)d_i(\sigma) = s_{n-1} \cdots s_0(x) (the unique degenerate (n1)(n-1)-simplex at xx) for all 0in0 \leq i \leq n. Two such simplices σ,τ\sigma, \tau represent the same element if they are homotopic relative to the boundary.

For n=1n = 1: an element of π1(X,x)\pi_1(X, x) is a 11-simplex σ\sigma with d0(σ)=d1(σ)=xd_0(\sigma) = d_1(\sigma) = x (a "loop" at xx), up to homotopy.

For n2n \geq 2, πn(X,x)\pi_n(X, x) is abelian. For n=1n = 1, π1(X,x)\pi_1(X, x) is a group (possibly non-abelian).

ExampleHomotopy groups of BG

For a group GG, the nerve N(BG)N(BG) is a Kan complex (since BGBG is a groupoid). Its homotopy groups are:

π0(N(BG))={},π1(N(BG))=G,πn(N(BG))=0 for n2.\pi_0(N(BG)) = \{*\}, \quad \pi_1(N(BG)) = G, \quad \pi_n(N(BG)) = 0 \text{ for } n \geq 2.

So N(BG)N(BG) is a K(G,1)K(G, 1), an Eilenberg--MacLane space. This matches the topological classifying space BGBG.

ExampleHomotopy groups via Sing

For Sing(Sn)\operatorname{Sing}(S^n) (the singular set of the nn-sphere):

πk(Sing(Sn))=πk(Sn)\pi_k(\operatorname{Sing}(S^n)) = \pi_k(S^n)

since Sing\operatorname{Sing} preserves homotopy groups. So πn(Sing(Sn))=Z\pi_n(\operatorname{Sing}(S^n)) = \mathbb{Z}, π3(Sing(S2))=Z\pi_3(\operatorname{Sing}(S^2)) = \mathbb{Z} (Hopf), etc.

The simplicial homotopy groups agree with the topological ones via geometric realization.

ExampleContractible Kan complexes

A Kan complex XX is contractible if π0(X)={}\pi_0(X) = \{*\} and πn(X,x)=0\pi_n(X, x) = 0 for all n1n \geq 1 and all basepoints xx. Equivalently, the map XΔ[0]X \to \Delta[0] is a weak equivalence.

For example, Sing(Rn)\operatorname{Sing}(\mathbb{R}^n) is contractible for all nn, as is Sing(Δn)\operatorname{Sing}(|\Delta^n|).


Kan Fibrations

Definition1.4Kan fibration

A map p:XYp: X \to Y of simplicial sets is a Kan fibration if it has the right lifting property with respect to all horn inclusions ΛknΔ[n]\Lambda^n_k \hookrightarrow \Delta[n]. That is, for every commutative square

ΛknX\Lambda^n_k \to X p\downarrow \quad\quad \downarrow p Δ[n]Y\Delta[n] \to Y

there exists a lift Δ[n]X\Delta[n] \to X making both triangles commute.

A simplicial set XX is a Kan complex if and only if the unique map XΔ[0]X \to \Delta[0] is a Kan fibration.

ExampleTrivial Kan fibrations

A map p:XYp: X \to Y is a trivial Kan fibration (also called an acyclic fibration) if it has the right lifting property with respect to all boundary inclusions Δ[n]Δ[n]\partial\Delta[n] \hookrightarrow \Delta[n].

Trivial Kan fibrations are both Kan fibrations and weak equivalences. If p:XΔ[0]p: X \to \Delta[0] is a trivial Kan fibration, then XX is contractible.

Any simplicial set XX admits a trivial Kan fibration X~X\widetilde{X} \to X where X~\widetilde{X} is a contractible Kan complex. This is the simplicial analogue of a path space.

ExampleFibers of Kan fibrations

If p:XYp: X \to Y is a Kan fibration and yY0y \in Y_0 is a vertex, the fiber p1(y)=X×Y{y}p^{-1}(y) = X \times_Y \{y\} is a Kan complex. The long exact sequence of homotopy groups applies:

πn(p1(y))πn(X)πn(Y)πn1(p1(y))\cdots \to \pi_n(p^{-1}(y)) \to \pi_n(X) \to \pi_n(Y) \to \pi_{n-1}(p^{-1}(y)) \to \cdots

This is the simplicial analogue of the long exact sequence for a fibration in topology.

ExamplePath space fibration

For a Kan complex XX with basepoint xX0x \in X_0, the path space PXPX is defined so that (PX)n(PX)_n consists of (n+1)(n+1)-simplices σ\sigma of XX with d0(σ)=s0n(x)d_0(\sigma) = s_0^n(x) (the degenerate nn-simplex at xx). The evaluation map PXXPX \to X sending σ\sigma to d1d2dn+1(σ)d_1 d_2 \cdots d_{n+1}(\sigma) is a Kan fibration, with fiber the loop space ΩX\Omega X.

This gives πn(X)πn1(ΩX)\pi_n(X) \cong \pi_{n-1}(\Omega X), the fundamental looping-delooping relationship.


Minimal Kan Complexes

Definition1.5Minimal Kan complex

A Kan complex XX is minimal if whenever two simplices σ,τXn\sigma, \tau \in X_n are homotopic relative to their boundary (i.e., di(σ)=di(τ)d_i(\sigma) = d_i(\tau) for all ii and there exists a homotopy fixing the boundary), then σ=τ\sigma = \tau.

Equivalently, XX is minimal if the homotopy relation on simplices with the same boundary is trivial.

ExampleUniqueness of minimal models

Every Kan complex XX contains a minimal sub-Kan complex XminXX_{\min} \subseteq X that is a deformation retract. The inclusion XminXX_{\min} \hookrightarrow X is a weak equivalence.

Moreover, two minimal Kan complexes are weakly equivalent if and only if they are isomorphic. This provides a canonical "normal form" for each homotopy type.

For example, the minimal Kan complex representing K(Z,1)K(\mathbb{Z}, 1) has one 00-simplex, Z\mathbb{Z} many non-degenerate 11-simplices, and its higher simplices are determined by the group structure.


Summary

RemarkKey points

The essential ideas about Kan complexes are:

  1. A Kan complex is a simplicial set where every horn has a filler. This is the combinatorial analogue of a topological space.

  2. Sing(Y)\operatorname{Sing}(Y) is always a Kan complex. The nerve N(C)N(\mathcal{C}) is Kan if and only if C\mathcal{C} is a groupoid.

  3. Homotopy groups πn(X,x)\pi_n(X, x) are well-defined for Kan complexes and match topological homotopy groups via geometric realization.

  4. Kan fibrations are the "good" notion of fibration for simplicial sets, with fibers being Kan complexes and long exact sequences of homotopy groups.

  5. Kan complexes are precisely the \infty-groupoids -- this is one formulation of the homotopy hypothesis.