Exercises:Saul - Algebraic Topology - 9/Exercise 9.7
Exercises
Exercise 9.7
Show that the cone on the real projective plane, i.e. C(RP2), is not a topological 3-manifold with boundary.
I might have two solutions for this, I accidentally answered the wrong question initially.
Solution 1
- I struggled to focus writing this up, however the idea should be clear, if the presentation isn't up to my usual standard please forgive me. Aditionally:
- ⟨G⟩ means the "normal subgroup generated by G", fortunately where it is used here G is trivial, thus is a normal subgroup itself so ⟨0⟩=0, thus the question of "normal subgroup closure" never pops up in this solution.
Before we start, I want to note that we make heavy use of the fundamental group, and I may write this without a base-point, this is because for a path-connected space the fundamental group based anywhere is isomorphic to any other fundamental group based elsewhere of the space. Please know that I know this.
First we consider C(RP2) as a topological space and work out some fundamental group isomorphisms.
- Immediately, as any topological cone is contractible space we see:
- π1(C(RP2))≅0 (where 0 denotes the trivial group, we do not use 1 even though fundamental groups are usually not Abelian)
Let p∈C(RP2) be the apex of the cone, i.e. the image of RP2×{1} under the the cone's quotient map, some people may identify X×{0} as the apex, we do not.
Consider C(RP2)−{p}, this is easily seen to be homeomorphic to RP2×[0,1)
- Note that RP2×[0,1) has RP2 as a strong deformation retraction (by slowly "projecting down" along [0,1)), and thus:
- π1(C(RP2)−{p})≅π1(RP2)≅Z2Z
We summarise:
- π1(C(RP2))≅0 and
- π1(C(RP2)−{p})≅π1(RP2)≅Z2Z
We now use the "fact" it is a manifold to tackle the claim: π1(C(RP2))≅π1(C(RP2)−{p}) - this will make use of the Seifert-Van Kampen theorem.
- Let x∈C(RP2) be an arbitrary point that is not the implicit basepoint for π1
- suppose that x is in the interior of the manifold[Note 1], then:
- there exists a homeomorphism: φ:U→B3 which takes an open neighbourhood of x onto the open unit ball at the origin.
- Suppose WLOG that φ(x)=0 (we can do this as if this wasn't the case we can use that translations are homeomorphisms in Euclidean space to get a new chart, φ′:z↦φ(z)−φ(x) instead, this is also a homeomorphism and has the property that φ′(x)=0)
- Note that homeomorphic spaces have isomorphic fundamental groups so π1(U)≅π1(B3)
- Note also given a homeomorphism all subspaces of the domain are homeomorphic to their image under the homeomorphism itself so
- φ|U−{x}:(U−{x})→(B3−{0}) is a homeomorphism
- We now use the Seifert-Van kampen theorem to see:
- π1(C(RP2))≅π1(C(RP2)−{x})∗π1(U)⟨π1((C(RP2)−{x})∩U)⟩[Note 2] which by homeomorphic spaces have isomorphic fundamental groups:
- ≅π1(C(RP2)−{x})∗π1(B3)⟨π1(U−{x})⟩ (by replacing spaces with homeomorphic ones)
- ≅π1(C(RP2)−{x})∗π1(B3)⟨π1(B3−{0})⟩ (by replacing spaces with homeomorphic ones)
- ≅π1(C(RP2)−{x})∗00 as B3 is clearly contractible, and B3−{0} strongly deformation retracts to the 2-sphere which has trivial fundamental group (easily seen as any loop can be "pulled" to its base point in S2)
- ≅π1(C(RP2)−{x})
- π1(C(RP2))≅π1(C(RP2)−{x})∗π1(U)⟨π1((C(RP2)−{x})∩U)⟩[Note 2] which by homeomorphic spaces have isomorphic fundamental groups:
- We conclude:
- π1(C(RP2))≅π1(C(RP2)−{x}) - provided x is an interior point of the 3-manifold C(RP2)
- Suppose WLOG that φ(x)=0 (we can do this as if this wasn't the case we can use that translations are homeomorphisms in Euclidean space to get a new chart, φ′:z↦φ(z)−φ(x) instead, this is also a homeomorphism and has the property that φ′(x)=0)
- there exists a homeomorphism: φ:U→B3 which takes an open neighbourhood of x onto the open unit ball at the origin.
- now suppose x∈∂(C(RP2)) - that is x is a boundary point of our 3-manifold with boundary[Note 3]
- We see that there exists a homeomorphism: φ:U→H3 where U is an open neighbourhood of x and H3 denotes the 3-dimensional half space:
- H3:={(x1,x2,x3)∈R3 | ∥(x1,x2,x3)∥<1∧x3≥0}=B3∩{(x1,x2,x3)∈R3 | x3≥0}
- Such that (φ(x))3=0 (i.e. on the x3=0 plane in R3)
- Again WLOG we can suppose φ(x)=0 - for the same reasoning as the previous case[Note 4]
- As before we note that (U−{x})≅(H3−{0}) by the restriction of φ to the left hand side of the ≅
- We use the Seifert-Van Kapmen theorem again:
- π1(C(RP2))≅π1(C(RP2)−{x})∗π1(U)⟨π1((C(RP2)−{x})∩U)⟩[Note 5] which by homeomorphic spaces have isomorphic fundamental groups:
- ≅π1(C(RP2)−{x})∗π1(H3)⟨π1(U−{x})⟩ (by replacing spaces with homeomorphic ones)
- ≅π1(C(RP2)−{x})∗π1(H3)⟨π1(H3−{0})⟩ (by replacing spaces with homeomorphic ones)
- ≅π1(C(RP2)−{x})∗00 as H3 and H3−{0} are both clearly contractible. (To convince yourself of this take the origin for H3, you can just "straight-line" pull everything onto it. For H3−{0} pick the point (0,0,1) to pull back on, again straight line onto this point)
- ≅π1(C(RP2)−{x})
- π1(C(RP2))≅π1(C(RP2)−{x})∗π1(U)⟨π1((C(RP2)−{x})∩U)⟩[Note 5] which by homeomorphic spaces have isomorphic fundamental groups:
- We conclude:
- π1(C(RP2))≅π1(C(RP2)−{x}) - provided x is a boundary point of the 3-manifold with boundaryC(RP2)
- Again WLOG we can suppose φ(x)=0 - for the same reasoning as the previous case[Note 4]
- We see that there exists a homeomorphism: φ:U→H3 where U is an open neighbourhood of x and H3 denotes the 3-dimensional half space:
- Thus in either case:
- π1(C(RP2))≅π1(C(RP2)−{x})
- suppose that x is in the interior of the manifold[Note 1], then:
- Since x∈C(RP2) was arbitrary (and not the implicit base-point) we see π1(C(RP2))≅π1(C(RP2)−{x}) holds for all x
- Next we combine everything to see:
- 0≅π1(C(RP2))≅π1(C(RP2)−{p})≅π1(RP2)≅Z2Z
- Thus: 0≅Z2Z - obviously a contradiction!
- 0≅π1(C(RP2))≅π1(C(RP2)−{p})≅π1(RP2)≅Z2Z
- Thus C(RP2) cannot be a 3-manifold with boundary
- In fact it cannot be a 3-manifold without boundary either, as per the first half where we assumed x was a point interior to the manifold.
This completes the proof.
Notes
- Jump up ↑ Not to be confused with a topological interior point
- Jump up ↑ I take it as obvious that:
- C(RP2)−{x} is an open set in C(RP2) - clearly the complement is closed!
- U is open by definition
- Jump up ↑ Not to be confused with a topological boundary point
- Jump up ↑ Restated:
- we can do this as if this wasn't the case we can use that translations are homeomorphisms in Euclidean space to get a new chart, φ′:z↦φ(z)−φ(x) instead, this is also a homeomorphism and has the property that φ′(x)=0
- Jump up ↑ I take it as obvious that:
- C(RP2)−{x} is an open set in C(RP2) - clearly the complement is closed!
- U is open by definition
References