nLab
simplicial complex

Contents

Idea and definition

A polyhedron is a space made up of very simple bits ‘glued’ together. The ‘bits’ are simplices of different dimensions. An abstract simplicial complex is a neat combinatorial way of giving the corresponding ‘gluing’ instructions, a bit like to plan of a construction kit!

Definition

A simplicial complex KK (sometimes called an abstract simplicial complex) consists of a set of objects, V(K)V(K), called vertices and a set, S(K)S(K), of finite non-empty subsets of V(K)V(K), called simplices. The simplices satisfy the following conditions: (1) that if σV(K)\sigma \subset V(K) is a simplex and τσ\tau \subset \sigma, τ\tau \ne \emptyset, then τ\tau is also a simplex; (2) every singleton {v}\{v\}, vV(K)v \in V(K), is a simplex.

We say τ\tau is a face of σ\sigma. If σS(K)\sigma \in S(K) has p+1p+1 elements it is said to be a pp-simplex. The set of pp-simplices of KK is denoted by K pK_p. The dimension of KK is the largest pp such that K pK_p is non-empty.

  • Remark: the word ‘polyhedron’ is used here as it is often used by algebraic topologists, as a space described by a simplicial complex. Elsewhere in mathematics, it might mean a finite union of finite intersections of sets in Euclidean space defined by linear inequalities, usually assumed compact, and often with other assumptions as well (e.g., connected or convex). The various usages have a long history, as recounted in detail in Lakatos’s Proofs and Refutations.

Examples

  1. Any (undirected simple) graph gives a simplicial complex. The usual definition of graph is that it is an ordered pair (V,E)(V,E) where VV is a set of vertices, and EE a set of (unordered) pairs of vertices. This is the simplest form of graph; it is undirected, edges do not have a ‘start’ and ‘finish’, (or ‘head’ and ‘tail’) and ‘simple’, in as much as there can be at most one edge between a given pair of vertices. (The case of a ‘multigraph’ where there can be multiple edges between vertices, and perhaps loops at a vertex, does not correspond to a simplicial complex, but does give a simplicial set.) A graph is a 1-dimensional simplicial complex.

  2. Given a space and an open cover, the nerve of the cover is a simplicial complex (see Čech methods and the discussion there). The Vietoris complex is another given by a related method.

  3. Given any two sets XX and YY, and a relation RX×YR\subseteq X\times Y, there are two simplicial complexes that encode information on the relation. These are generalisations of the nerve and the Vietoris complex. They are studied in detail in Dowker's theorem.

  4. If (P,)(P,\leq) is a poset, then the nerve of the associated category has a simple description. The vertices are the points of PP and the simplices are the flags.

  5. Buildings: An important class of simplicial complexes is provided by the notion of building, due to Jacques Tits.

Simplicial complexes v. simplicial sets

Simplicial complexes are, in some sense, special cases of simplicial sets, but only ‘in some sense’.

To get from a simplicial complex to a fairly small simplicial set, you pick a total order on the set of vertices. Without an order on the vertices, you cannot speak of the k thk^{th} face of a simplex, which is an essential feature of a simplicial set! The degeneracies are obtained by repeating an element when listing the vertices of a simplex. If σ={v 0,v 1,,v n}\sigma = \{v_0,v_1,\ldots, v_n\}, with v 0<v 1<<v nv_0\lt v_1\lt \ldots \lt v_n then, for instance, s 0(σ)={v 0,v 0,v 1,,v n}s_0(\sigma) = \{v_0,v_0, v_1,\ldots, v_n\}.

If you do not want to pick an order then you can still form a simplicial set where to each nn-simplex of the original simplicial complex will correspond to (n+1)!(n+1)! simplices of that associated simplicial set. The result is more unwieldy, but can be useful under some circumstances as it defines a functor from the category of simplicial complexes to that of simplicial sets. This is very important when discussion group actions on simplicial complexes and how this transfers to the associated simplicial set.

Simplicial complexes as sheaves on a site

Simplicial sets are essentially (that is, up to equivalence) presheaves on the simplex category of finite nonempty totally ordered sets, whereas simplicial complexes may be regarded as concrete presheaves on the category Fin +Fin_{+} of finite nonempty sets and functions between them. This works as follows: given a simplicial complex, K=(V(K),S(K))K = (V(K), S(K)), define a presheaf K :Fin + opSetK^\sim: Fin_{+}^{op} \to Set whose values are sets of functions ϕ\phi:

K (B)=def{ϕ:BV(K)|ϕ(B)S(K)},K (f:AB)(ϕ)=defϕfK^\sim(B) \stackrel{def}{=} \{\phi: B \to V(K)| \phi(B) \in S(K)\}, \qquad K^\sim(f: A \to B)(\phi) \stackrel{def}{=} \phi \circ f

This defines an evident functor

SimpComplexSet Fin + op:KK SimpComplex \to Set^{Fin_{+}^{op}}: K \mapsto K^\sim

that is full and faithful. The essential image is the subcategory of concrete presheaves, where a presheaf F:Fin + opSetF \colon Fin_{+}^{op} \to Set is concrete if the canonical map

F(B)F(1) hom(1,B)F(B) \to F(1)^{\hom(1, B)}

is an injection. The point is that a morphism of concrete presheaves FGF \to G is uniquely determined from the function F(1)G(1)F(1) \to G(1) between their underlying sets (i.e., the underlying-set functor on concrete presheaves is faithful, so that a concrete presheaf is a set equipped with extra structure – that’s what makes it “concrete”).

(Equivalently but somewhat more elaborately, the category of concrete presheaves is the same as the full subcategory of concrete sheaves on Fin +Fin_{+} with respect to the trivial topology, where the only covering sieve Fhom(,D)F \hookrightarrow hom(-, D) is the maximal sieve.)

From this point of view, it is immediate that simplicial complexes are the separated objects for the Lawvere-Tierney topology on Set Fin + opSet^{Fin_{+}^{op}} whose sheaves are sets, via the sheafification functor

Γ=hom(1,)=ev 1:Set Fin + opSet\Gamma = \hom(\mathbf{1}, -) = ev_1 \colon Set^{Fin_{+}^{op}} \to Set

which has a right adjoint. (See local topos.) It follows from this characterization that the category of simplicial complexes is a quasitopos, and in particular is locally cartesian closed. The category of simplicial sets on the other hand is a topos.

Geometric realisations and Polyhedra

An abstract simplicial complex is a combinatorial gadget that models certain aspects of a spatial configuration. Sometimes it is useful, perhaps even necessary, to produce a topological space from that data in a simplicial complex.

Idea

To each simplicial complex KK, one can associate a topological space called the polyhedron of KK often also called the geometric realisation of KK and denoted |K||K|. (This is essentially a special case of the geometric realisation of a simplicial sets.)

This can be constructed by taking a copy K(σ)K(\sigma) of a standard topological pp-simplex for each pp-simplex of KK and then ‘gluing’ them together according to the face relations encoded in KK.

We therefore first need the definition of a standard pp-simplex

Definition

The standard (topological) pp-simplex is (usually) taken to be the convex hull of the basis vectors e 1,e 2,,e p+1\mathbf{e}_1, \mathbf{e}_2,\ldots, \mathbf{e}_{p+1} in p+1\mathbb{R}^{p+1}.

Intuition

The geometric realisation |K||K| of a simplicial complex, KK is then constructed by taking, for each abstract pp-simplex, σS(K)\sigma\in S(K), a copy, K(σ)K(\sigma) of such a standard topological pp-simplex, and then ‘gluing’ faces together, so whenever τ\tau is a face of σ\sigma we identify K(τ)K(\tau) with the corresponding face of K(σ)K(\sigma). This space is usually denoted Δ p\Delta^p.

Canonical construction

As a set, |K||K| is constructed as follows:

|K||K| is the set of all functions from V(K)V(K) to the closed interval [0,1][0,1] such that

  • if α|K|\alpha \in {|K|}, the set
{vV(K)α(v)0}\{v \in V(K) \mid \alpha(v) \neq 0\}

is a simplex of KK;

  • for each α|K|\alpha \in {|K|},
    vV(K)α(v)=1.\sum_{v \in V(K)} \alpha (v) = 1.

There are two commonly used topologies on the set |K||K|. The first is the metric topology: we put a metric dd on |K||K| by

d(α,β)=( vV(K)(α(v)β(v)) 2) 12.d(\alpha,\beta) = \Big(\sum_{v\in V(K)} (\alpha(v) - \beta(v))^2\Big)^\frac{1}{2}.

|K||K|, when endowed with the metric space topology, will be denoted |K| d|K|_d. Notice that when V(K)V(K) is finite, this gives |K| d|K|_d as a subspace of the metric space #(V(K))\mathbb{R}^{\#(V(K))} (which is usually of much higher dimension than might seem geometrically significant in a given context).

The second topology is the coherent topology: each geometric simplex |s||s| consists of all α|K|\alpha \in {|K|} supported in ss, and is given the subspace topology inherited as a subset of |K| d|K|_d; then the coherent topology on |K||K| is the largest topology for which all inclusions |s||K|{|s|} \hookrightarrow {|K|} are continuous. This topological space is normally denoted just |K||K|, reflecting the fact that the coherent topology is regarded as the default topology to put on the set |K||K|.

Note that if sts \subseteq t is an inclusion of simplices in KK, then there is an induced subspace inclusion |s||t|{|s|} \hookrightarrow {|t|}. The space |K||K| may then be characterized as the colimit in TopTop of the diagram consisting of geometric simplices |s||s| and inclusions between them, so that a function f:|K|Xf: {|K|} \to X is continuous if and only if its restriction to each simplex |s||s| is continuous. In particular, the identity function |K||K| d{|K|} \to {|K|}_d is continuous, so that the coherent topology contains the metric topology (and is often strictly larger).

  • Warning: The geometric realization of a simplicial complex does not preserve products. Indeed, the product of two intervals in the category of simplicial complexes is the tetrahedron!

Triangulable spaces

If a topological space can be described up to homeomorphism as the geometric realization of a simplicial complex, we say it is triangulable, and a triangulation of a space XX is a simplicial complex KK together with a homeomorphism h:|K|Xh: |K| \to X. (This is discussed in a bit more detail in the entry on classical triangulation.

There is another stronger notion of triangulation used by geometric topologists: a piecewise-linear (PL) structure on a topological manifold XX is given by a PL atlas, where the transition functions are piecewise-linear homeomorphisms. (A homeomorphism UVU \to V is piecewise linear if its graph is the intersection of U×VU \times V with a semilinear set SS, meaning that SS is given by a finite Boolean combination of solution sets of linear inequalities.)

Examples from manifold theory

  • All smooth manifolds are triangulable and, in fact, admit PL structures.

  • All topological manifolds in dimensions 2 and 3 admit PL structures, and are in fact smoothable (admit a smooth manifold structure).

  • The E 8E_8 manifold? does not admit a triangulation, much less a PL structure.

  • In dimensions n5n \geq 5, the (n2)(n-2)-fold suspension of the Poincaré sphere is homeomorphic to the nn-sphere, hence is triangulable, but it does not admit a PL structure.

Relation to simplicial sets

The following statement may seem obvious, but it requires careful proof:

  • A space is triangulable if and only if it is homeomorphic to the geometric realization of a simplicial set.

As an important step:

  • The geometric realization of the nerve of a poset is triangulable.

The basic technique is to use subdivision.

References

A standard textbook reference is

  • Edwin Spanier, Algebraic Topology , McGraw-Hill, 1966.

An exposition is in

  • Kenny Erleben, Simplicial complexes (2010) (pdf slides)

That simplicial complexes form a quasitopos of concrete sheaves is discussed in

Revised on October 2, 2013 18:12:04 by Todd Trimble (173.162.178.38)