Exercises:Saul - Algebraic Topology - 3/Exercise 3.2
Exercises
Exercise 3.2
Suppose that (X,J) is a non-empty path-connected topological space, equipped with a Δ-complex structure. Show, directly from the definitions (Hatcher, of course...) that HΔ0(X)≅Z
- We may assume without proof that the 1-skeleton is path connected.
Notes
- HΔn(X):=Ker(∂0)Im(∂1)
As ∂0 just sends everything to 0 we see Ker(∂0)=Δ0(X) - all the vertices. Thus, essentially, Ker(∂0)≅Z#vertices.
Okay now ∂1:Δ1(X)→Δ0(X), what is its image?
- First of all, for f∈Δ1(X) we see ∂1(X):=∂1(∑α∈I1nασα)=∑α∈I1nα∂1(σα) =∑α∈I1nα(Terminal(σα)−Initial(σα))
- where I1 is the set of 1-simplices involved in X, and nα∈Z with nα≠0 for only finitely many of the α∈I1
This shows us that (sort of anyway) the image is spanned by various Terminal(σα)−Initial(σα) (which are vertices)
Using X1 to denote the 1-skeleton (consistent notation be damned) then Caveat:and this is the informal part for any two vertexes of X, say v0 and v1, there is a path through the "edges" of X1, such that ∂1(that path)=v1−v0
- We can make this a bit better! If (pi)ki=1 is a representation of the path, where pi is an edge, such that the initial vertex of p1 is v0 and the final vertex of pk is v1 we can represent the path by the formal linear combination: ∑ki=1pi.
- Note the pi need not be unique, it could have several loops in it, that wont matter (as the boundaries of cycles are 0)
Now what we can sort of do is... well we want that sum to collapse to a few terms. What we can do is consider each σα which is an edge - plus the path from its endpoint to some fixed point of our choice. The boundaries then will have the same terminal point, subtracting various initials (and some/one (?) will have the same terminal and initial, giving us our "one less")
It would have to be shown this is in bijection with the edges.
Proof
Let us define some notation before we start.
- Xn is the n-skeleton of X. Which made from all the simplices involved in X of dimension ≤n. So X1 is itself a complex made up of the 0 and 1 simplices.
- X(n) is not a complex but rather a set of all the n-dimensional simplices (and only those simplices) involved in X. For example X(1) are all the 1-simplices of X (and not the 0-simplices), and so forth.
- Δn(X):=F(X(n)) is the free abelian group with generators the set of n-simplices involved in X (F(A) denotes the free abelian group generated by the elements a∈A)
- ∂n:Δn(X)→Δn−1(X) is the boundary map.
- ∂0:Δ0(X)→0 is a group homomorphism onto the trivial group. ∂0:x↦0 always; thus Ker(∂0)=Δ0(X)
- HΔn(X):=Ker(∂n)Im(∂n+1)
We also indulge in a few abuses of notation
- c∈Δ0(X) means c=∑αnαvα where it is understood that nα∈Z is non-zero for only finitely many of the α in the implied indexing set. The indexing set for which each α is in can also be used to index X(0), thus vα addresses each element of X(0) - under the identification of v∈X(0) with 1v=v∈F(X(0)).
- d∈Δ1(X) means d=∑αnαeα almost exactly as above but this time with X(1) instead of X(0) in play.
- For e∈X(1) (possibly identified with 1e∈Δ1(X)) we use e(0) for the initial point of the edge and e(1) for the final point of the edge. Thus ∂(e)=e(1)−e(0)
We wish to compute:
- HΔ0(X):=Ker(∂0)Im(∂1)=Δ0(X)Im(∂1)
Proof:
- Define I:Δ0(X)→Z by I:∑αnαvα↦∑αnα, clearly this is a group homomorphism and clearly also it is surjective (as X is non-empty, we have at least one generator of Δ0(X), so from that alone we can get any integer).
- By the first group isomorphism theorem we have: Δ0(X)Ker(I)≅ˉIZ where ¯I:Δ0(X)Ker(I)→Z is the induced group isomorphism from I, thus ¯I:[c]↦I(C). This notation is unambiguous, [c] represents an arbitrary equivalence class of the quotient (as this is first year work I will not elaborate any further. See the page first group isomorphism theorem for more information)
- Suppose that Im(∂1)=Ker(I). Then ¯I would be an isomorphism from HΔ0(X) to Z, and the result would be shown. This it the route we will take.
- To do so we must show: Im(∂1)=Ker(I). This will consist of two steps: Im(∂1)⊆Ker(I) and Ker(I)⊆Im(∂1)
- Showing that Im(∂1)⊆Ker(I), by the implies-subset relation, we need only show ∀c∈Im(∂1)[c∈Ker(I)]
- Let c∈Im(∂1) be given
- By definition of image, c∈Im(∂1)⟺∃d∈Δ1(X)[∂1(d)=c]
- Thus let d∈Δ1(X) be such that ∂1(d)=c.
- We observe now that I(c)=I(∂1(d))
- =I(∂1(∑αnαeα))
- =I(∑αnα∂1(eα))
- =I(∑αnα(eα(1)−eα(0))
- =I(∑αnαeα(1)+∑α(−nα)eα(0))
- =∑αnα+∑α(−nα)
- =∑α(nα−nα)
- =0
- Thus I(c)=0, so c∈Ker(I)
- Since c∈Im(∂1) was arbitrary, we have shown that for all such c that c∈Ker(I). This completes the first step.
- Let c∈Im(∂1) be given
- Showing that Ker(I)⊆Im(∂1). By the implies-subset relation we need only show ∀c∈Ker(I)[c∈Im(∂1)]
- We require two lemmas before we can continue:
- ∀v0,v∈X(0)∃(pi)ki=1⊆X(1)[∂1(∑ki=1pi)=v−v0∈Δ0(X)] (this requires path-connectedness) and
- ∀c∈Δ0(X)∀v0∈X(0)∃d∈Δ1(X)[∂1(d)=c−I(c)v0∈Δ0(X)]
- The proof of these can be found below (at this level of indentation)
- Suppose the lemmas hold. We will prove the statement: ∀c∈Ker(I)[c∈Im(∂1)]
- Let c∈Ker(I) be given
- We wish to show c∈Im(∂1), by definition of image:
- c∈Im(∂1)⟺∃d∈Δ1(X)[∂1(d)=c]
- Choose v0∈X(0) arbitrarily
- As c∈Ker(I)⊆Δ0(X) we see c∈Δ0(X), thus we can apply the second lemma
- Choose d∈Δ1(X) to be the d posited to exist by the second lemma. So we have: ∂1(d)=c−I(c)v0
- now we wish to show our choice of d is such that ∂1(d)=c
- ∂1(d)=c−I(c)v0, we know this already
- But c∈Ker(I), so I(c)=0, thus
- ∂1(d)=c
- As required.
- Thus we have shown our choice of d satisfies ∂1(d)=c
- We wish to show c∈Im(∂1), by definition of image:
- Since c∈Ker(I) was arbitrary we have shown ∀c∈Ker(I)∃d∈Δ1(X)[∂1(d)=c]
- Which as we established earlier is the same as ∀c∈Ker(I)[c∈Im(∂1)] which was itself the same as Ker(I)⊆Im(∂1)
- Let c∈Ker(I) be given
- Thus we have shown that, provided the lemmas hold, Ker(I)⊆Im(∂1)
- Suppose the lemmas hold. We will prove the statement: ∀c∈Ker(I)[c∈Im(∂1)]
- Proof of lemmas:
- Proof that: ∀v0,v∈X(0)∃(pi)ki=1⊆X(1)[∂1(∑ki=1pi)=v−v0∈Δ0(X)]
- Let v0∈X(0) be given
- Let v∈X(0) be given
- By path-connectedness of X1 (the 1-skeleton) we see there is a path through the 1-skeleton, say p, from v0 to v
- Say (pi)ki=1⊆X(1) such that p=∑ki=1pi∈X1 which of course means p=∑ki=1pi∈Δ1(X) too
- Thus: ∂1(∑ki=1pi)=v−v0∈Δ0(X) by the very definition of the path we took.
- Let v∈X(0) be given
- Let v0∈X(0) be given
- ∀c∈Δ0(X)∀v0∈X(0)∃d∈Δ1(X)[∂1(d)=c−I(c)v0∈Δ0(X)]
- Let c∈Δ0(X) be given, we shall write c=∑αnαvα
- Let v0∈X(0) be given
- We already know that for each vα in ∑αnαvα there exists a path p(vα) say from v0 to vα, p(vα):=∑ki=1pi such that ∂1(p(vα))=vα−v0 from the first lemma.
- Choose d:=∑αnαp(vα)∈Δ1(X)
- We must show that this d satisfies the "such that" part of the lemma
- ∂1(d)=∂1(∑αnαp(vα))
- =∑αnα∂1(p(vα))
- =∑αnα(vα−v0)
- =∑αnαvα+(−∑αvα)v0
- =c+(−I(c))v0
- =c−I(c)v0
- As required
- Our choice of d satisfies the requirements of the claim
- Since v0∈X(0) was arbitrary we have shown the claim for all such v0
- Let v0∈X(0) be given
- Since v∈X(0) was arbitrary we have shown the claim for all such v
- Let c∈Δ0(X) be given, we shall write c=∑αnαvα
- Proof that: ∀v0,v∈X(0)∃(pi)ki=1⊆X(1)[∂1(∑ki=1pi)=v−v0∈Δ0(X)]
- We require two lemmas before we can continue:
- Showing that Im(∂1)⊆Ker(I), by the implies-subset relation, we need only show ∀c∈Im(∂1)[c∈Ker(I)]
- We have now established Ker(I)=Im(∂1) thus HΔ0(X)≅Z - as required.
Notes
References