# Functoriality in Low Dimensions. Note Quote.

Let CW be the category of CW-complexes and cellular maps, let CW0 be the full subcategory of path connected CW-complexes and let CW1 be the full subcategory of simply connected CW-complexes. Let HoCW denote the category of CW-complexes and homotopy classes of cellular maps. Let HoCWn denote the category of CW-complexes and rel n-skeleton homotopy classes of cellular maps. Dimension n = 1: It is straightforward to define a covariant truncation functor

t<n = t<1 : CW0 → HoCW together with a natural transformation

emb1 : t<1 → t<∞,

where t<∞ : CW0 → HoCW is the natural “inclusion-followed-by-quotient” functor given by t<∞(K) = K for objects K and t<∞(f) = [f] for morphisms f, such that for all objects K, emb1∗ : H0(t<1K) → H0(t<∞K) is an isomorphism and Hr(t<1K) = 0 for r ≥ 1. The details are as follows: For a path connected CW-complex K, set t<1(K) = k0, where k0 is a 0-cell of K. Let emb1(K) : t<1(K) = k0 → t<∞(K) = K be the inclusion of k0 in K. Then emb1∗ is an isomorphism on H0 as K is path connected. Clearly Hr(t<1K) = 0 for r ≥ 1. Let f : K → L be a cellular map between objects of CW0. The morphism t<1(f) : t<1(K) = k0 → l0 = t<1(L) is the homotopy class of the unique map from a point to a point. In particular, t<1(idK) = [idk0] and for a cellular map g : L → P we have t<1(gf) = t<1(g) ◦ t<1(f), so that t<1 is indeed a functor. To show that emb1 is a natural transformation, we need to see that

that is

commutes in HoCW. This is where we need the functor t<1 to have values only in HoCW, not in CW, because the square need certainly not commute in CW. (The points k0 and l0 do not know anything about f, so l0 need not be the image of k0 under f.) Since L is path connected, there is a path ω : I → L from l0 = ω(0) to f (k0) = ω(1). Then H : {k0} × I → L, H(k0, t) = ω(t), defines a homotopy from

k0 → l0 → L to k0 → K →f L.

Dimension n = 2: We will define a covariant truncation functor t<n = t<2 : CW1 → HoCW

together with a natural transformation
emb2 : t<2 → t<∞,

where t<∞ : CW1 → HoCW is as above (only restricted to simply connected spaces), such that for all objects K, emb2∗ : Hr(t<2K) → Hr(t<∞K) is an isomorphism for r = 0, 1, and Hr(t<2K) = 0 for r ≥ 2. For a simply connected CW-complex K, set t<2(K) = k0, where k0 is a 0-cell of K. Let emb2(K) : t<2(K) = k0 → t<∞(K) = K be the inclusion as in the case n = 1. It follows that emb2∗ is an isomorphism both on H0 as K is path connected and on H1 as H1(k0) = 0 = H1(K), while trivially Hr(t<2K) = 0 for r ≥ 2. On a cellular map f, t<2(f) is defined as in the case n = 1. As in the case n = 1, this yields a functor and emb2 is a natural transformation.

# Diffeomorphism Diffeology via Leaves of Lie Foliation

The notion of diffeological space is due to Jean-Marie Souriau.

Let M be a set. Any set map α: U ⊂ Rn → M defined on an open set U of some Rn, n ≥ 0, will be called a plot on M. The name plot is chosen instead of chart to avoid some confusion with the usual notion of chart in a manifold. When possible, a plot α with domain U will be simply denoted by αU.

A diffeology of class C on the set M is any collection P of plots α: Uα ⊂ Rnα → M, nα ≥ 0, verifying the following axioms:

1. (1)  Any constant map c: Rn → M, n ≥ 0, belongs to P;
2. (2)  Let α ∈ P be defined on U ⊂ Rn and let h: V ⊂ Rm → U ⊂ Rn beany C map; then α ◦ h ∈ P;
3. (3)  Let α: U ⊂ Rn → M be a plot. If any t ∈ U has a neighbourhood Ut such that α|Ut belongs to P then α ∈ P.

Usually, a diffeology P on the set M is defined by means of a generating set, that is by giving any set G of plots (which is implicitly supposed to contain all constant maps) and taking the least diffeology containing it. Explicitly, the diffeology ⟨G⟩ generated by G is the set of plots α: U → M such that any point t ∈ U has a neighbourhood Ut where α can be written as γ ◦ h for some C map h and some γ ∈ G.

A finite dimensional manifold M is endowed with the diffeology generated by the charts U ⊂ Rn → M, n = dimM, of any atlas.

Basic constructions. A map F : (M, P) → (N, Q) between diffeological spaces is differentiable if F ◦ α ∈ Q for all α ∈ P. A diffeomorphism is a differentiable map with a differentiable inverse.

Let (M,P) be a diffeological space and F : M → N a map of sets. The final diffeology FP on N is that generated by the plots F ◦ α, α ∈ P. A particular case is the quotient diffeology associated to an equivalence relation on M.

Analogously, let (N, Q) be a diffeological space and F : M → N a map of sets. The initial diffeology FQ on M is that generated by the plots α in M such that F ◦ α ∈ Q. A particular case is the induced diffeology on any subset M ⊂ N.

Finally, let D(M,N) be the space of differentiable maps between two diffeological spaces (M,P) and (N,Q). We define the functional diffeology on it by taking as a generating set all plots α: U → D(M,N) such that the associated map α~ : U × M → N given by α~ (t, x) = α(t)(x) is differentiable.

Diffeological groups.

Definition 2.3. A diffeological group is a diffeological space (G,P) endowed with a group structure such that the division map δ : G × G → G, δ(x, y) = xy−1, is differentiable.

A typical example of diffeological group is the diffeomorphism group of a finite dimensional manifold M, endowed with the diffeology induced by D(M,M). It is proven that the diffeomorphism group of the space of leaves of a Lie foliation is a diffeological group too.

Both constructions verify the usual universal properties.

Let (M, P), (N, Q) be two diffeological spaces. We can endow the cartesian product M × N with the product diffeology P × Q generated by the plots α × β, α ∈ P, β ∈ Q.