Published on

Geometric Topology 19

Authors

Geometric Topology 19

Lifting Loops and Homotopies

Theorem: Let Y=IY = I. Fix x0∈Xx_0 \in X and z∈p−1(x0)z \in p^{-1}(x_0). Let α,β:I→X\alpha,\beta : I \rightarrow X be equivalent loops based at x0x_0, meaning α≊β\alpha \approxeq \beta.

  1. α\alpha has a unique lift α~z:I→Z\tilde{\alpha}_z : I \rightarrow Z (a path, occasionally a loop) such that α~(0)=z\tilde{\alpha}(0) = z.
  2. If F:I2→XF : I^2 \rightarrow X is the path-homotopy realizing α≊β\alpha \approxeq \beta, it has a unique lift F~:I2→Z\tilde{F}: I^2 \rightarrow Z such that F~0=α~z\tilde{F}_0 = \tilde{\alpha}_z (the bottom edge of I2I^2).

Proof:

  1. One can partition II into closed intervals [aj,aj+1][a_j,a_{j+1}] such that α([aj,aj+1])\alpha([a_j,a_{j+1}]) lies in an open set UU covered by sheets. α~\tilde{\alpha} is defined on each in succession by bijectivity.
  2. This follows from a similar argument by dividing the square I2I^2 into rectangles RjkR_{jk}, each mapped into a UU, then working upwards from F~0\tilde{F}_0 to define F~\tilde{F} on each column. ■\blacksquare

F~\tilde{F} must be constant on the vertical lines s=0,1s=0,1 so F~(0,t)=z\tilde{F}(0,t) = z and F~1=β~z\tilde{F}_1 = \tilde{\beta}_z. Thus:

β~z(1)=F~(1,1)=F~(1,0)=α~z(1).\tilde{\beta}_z(1) = \tilde{F}(1,1) = \tilde{F}(1,0) = \tilde{\alpha}_z(1).

Corollary: The endpoint α~z(1)\tilde{\alpha}_z(1) of ZZ depends only on the class [α][\alpha] of α\alpha in π1(X,x0)\pi_1(X,x_0), so fixing zz determines a map π1(X,x0)→p−1(x0)\pi_1(X,x_0) \rightarrow p^{-1}(x_0).

Groups Acting on Sets

Let GG be a group (finite or infinite, with identity element ee), and Ω\Omega any set.

Definition: A right action of GG on Ω\Omega is a mapping:

Ω×G→Ω,(z,g)↦z⋅g,\Omega \times G \rightarrow \Omega, \\ (z,g) \mapsto z \cdot g,

such that:

  1. z⋅e=zz \cdot e = z,
  2. z⋅(gh)=(z⋅g)⋅hz \cdot (gh) = (z \cdot g)\cdot h for all z∈Ωz \in \Omega and g,h∈Gg, h \in G.

If Ω\Omega is finite of size nn, this simply means that there is a group homomorphism G→SnG \rightarrow S_n.

Examples:

  • Let Ω=R3={(x1,x2,x3)}, G=SO(3)={A∈R3,3:ATA=I,det⁡A=1}\Omega = \mathbb{R}^3 = \{(x_1,x_2,x_3)\}, \ G = SO(3) = \{A \in \mathbb{R}^{3,3}: A^TA = I, \det A = 1\}. GG acts on Ω\Omega on the right. It also acts on S2S^2 on the right. Let CC be a cube centred at the origin. The subgroup of GG mapping CC onto CC is known to be isomorphic to S4S_4; it contains 22-cycles, 33-cycles, 44-cycles (how do each of these rotate the cube?).

Lemma: Fix z∈Ωz \in \Omega. Then Gz={g∈G:z⋅g=z}G_z = \{g \in G: z \cdot g = z\} is a subgroup of GG, called the stabilizer of zz, and g↦z⋅gg \mapsto z \cdot g identifies the set {Gzg:g∈G}\{G_z g : g \in G\} of right cosets with Ω\Omega. Note that GG acts on this set of cosets on the right, so we can dispense with Ω\Omega.

Action of π1\pi_1 on a Fibre

Let p:Z→Xp : Z \rightarrow X be a covering map. Fix x0∈Xx_0 \in X. Set:

G=π1(X,x0),Ω=p−1(x0), the ’fibre’ over  x0.G = \pi_1 (X,x_0), \\ \Omega = p^{-1}(x_0), \ \text{the 'fibre' over } \ x_0.

Theorem:

  1. Setting z⋅[α]=α~z(1)z \cdot [\alpha] = \tilde{\alpha}_z(1) defines a right action of GG on Ω\Omega.
  2. If z∈Ωz \in \Omega, the induced homomorphism p∗:π1(Z,z)→π1(X,x0)p_{*} : \pi_1 (Z,z) \rightarrow \pi_1(X,x_0) is injective.
  3. The stabilizer {[α]:z⋅[α]=z}\{[\alpha] : z \cdot [\alpha] = z\} is precisely the subgroup p∗(π1(Z,z))p_{*}(\pi_1 (Z,z)) of GG.

Proofs:

  1. It is obvious that z⋅[ε]=zz \cdot [\varepsilon] = z. To prove that z⋅([α][β])=(z⋅[α])⋅[β]z \cdot ([\alpha][\beta]) = (z \cdot [\alpha]) \cdot [\beta], observe that αβ~z=α~zβ~z′\tilde{\alpha \beta}_z = \tilde{\alpha}_z \tilde{\beta}_{z'}, where z′=α~z(1)z' = \tilde{\alpha}_z(1) (since α~zβ~z′\tilde{\alpha}_z \tilde{\beta}_{z'} is a lift of αβ\alpha \beta from zz).
  2. If γ\gamma is a loop at zz such that p∘γ≊Fεp \circ \gamma \approxeq_F \varepsilon, then the unique lift F~\tilde{F} gives γ≊εz\gamma \approxeq \varepsilon_z.
  3. If α~z=z\tilde{\alpha}_z = z, then α~z\tilde{\alpha}_z is a loop and:
    [α]=[p∘α~z]=p∗[α~z].[\alpha] = [p \circ \tilde{\alpha}_z] = p_{*}[\tilde{\alpha}_z].
    Conversely, if [α]=p∗[γ][\alpha] = p_{*}[\gamma] with γ\gamma a loop based at zz, then γ=α~z\gamma = \tilde{\alpha}_z and z⋅[γ]=zz \cdot [\gamma] = z. ■\blacksquare

Fundamental Group of the Circle

Let p:Z→Xp : Z \rightarrow X be a covering map, fix x0∈Xx_0 \in X and z∈p−1(x0)z \in p^{-1}(x_0). Set G=π1(X,x0)G = \pi_1(X,x_0) and Ω=p−1(x0)\Omega = p^{-1}(x_0) as before.

Lemma: The map G→ΩG \rightarrow \Omega defined by [α]↦z⋅[α][\alpha] \mapsto z \cdot [\alpha] is:

  1. surjective if ZZ if path-connected,
  2. bijective if ZZ is simply-connected (meaning path-connected and π1(Z)={e}\pi_1(Z) = \{e\}).

Proof of Surjectivity: Given z′∈Ωz' \in \Omega, choose a path σ\sigma from zz to z′z' in ZZ. Then α=p∘σ\alpha = p \circ \sigma is a loop based at x0x_0. By uniqueness, α~z=σ\tilde{\alpha}_z = \sigma, so α~z(1)=z′\tilde{\alpha}_z(1) = z'. Proof of Injectivity: If z⋅[α]=z⋅[β]z \cdot [\alpha] = z \cdot [\beta] then z⋅g=zz \cdot g = z where g=[α][β]−1g = [\alpha][\beta]^{-1}. But the stabilizer of zz is trivial by part (3) of the previous theorem, so [α]=[β][\alpha] = [\beta].

We can apply the lemma to X=S1⊂CX = S^1 \subset \mathbb{C}, x0=1,Z=R,p(z)=e2πisx_0 = 1, Z = \mathbb{R}, p(z) = e^{2\pi i s}. Then Ω=Z\Omega = \mathbb{Z}. The path s↦sns \mapsto sn in R\mathbb{R} projects to the loop αn:s↦exp⁡(2πisn)\alpha_n : s \mapsto \exp (2 \pi i sn) in S1S^1, and 0⋅[αn]=n0 \cdot [\alpha_n] = n.

Corollary: π1(S1,1)\pi_1 (S^1,1) is isomorphic to (Z,+)(\mathbb{Z},+), i.e. the infinite cyclic group F1F_1.

Double Coverings

We previously defined RPn\mathbb{RP}^n to be the quotient of SnS^n obtained by identifying each point v=(x0,x1,...,xn)v = (x_0,x_1,...,x_n) with −v=(−x0,−x1,...,−xn)-v=(-x_0,-x_1,...,-x_n), its antipodal point.

Recall that RP2\mathbb{RP}^2 is homeomorphic to the surface PP originally defined from a square. The square model showed that PP is also a quotient of a circular disk DD in which opposite points on the boundary ∂D\partial D are identified. Now deform DD into the southern hemisphere of S2S^2 and retain ∂D\partial D as its equator. Then any equivalence class {v,−v}\{v,-v\} (defining a point on RP2\mathbb{RP}^2) has a unique representative in PP.

Let's explain why the group SO(3)SO(3) of rotations fixing the origin is homeomorphic to RP3\mathbb{RP}^3. Any non-identity rotation is determined by an axis (i.e. a unit vector vv), and an angle θ∈[0,π]\theta \in [0,\pi]. Map this to the point (θπ)v(\frac{\theta}{\pi})v in R3\mathbb{R}^3. The only ambiguity is that an angle π\pi around vv is the same rotation as π\pi around −v-v. So we are identifying antipodal points of the boundary S2=∂BS^2 = \partial B of the solid unit ball BB. Since BB is just a 33-dimensional analogue of DD above, the resulting space is RP3\mathbb{RP}^3 by the same argument.

Torus to Klein Bottle in Squares

Represent the torus TT by the square I2I^2 with boundary aba−1b−1aba^{-1}b^{-1}, giving qT:I2→Tq_T : I^2 \rightarrow T. Represent the Klein bottle KK by I2I^2 with boundary cd−1c−1d−1cd^{-1}c^{-1}d^{-1}, giving qK:I2→Kq_K : I^2 \rightarrow K.

Lemma: The mapping p:T→Kp : T \rightarrow K defined by:

p(qT(s,t))={qK(2s,t) if  s≤12,qK(2s−1,1−t) if  s≥12.p(q_T(s,t)) = \begin{cases} q_K(2s,t) & \ \text{if } \ s \leq \frac{1}{2}, \\ q_K(2s-1,1-t) & \ \text{if } \ s \geq \frac{1}{2}. \end{cases}

is a 2:1 covering map (the quotient map P→P^\mathscr{P} \rightarrow \hat{\mathscr{P}} is denoted by qq rather than π\pi).

If we regard a,ca,c as loops in T,KT,K respectively, then p∘a=ccp \circ a = cc, whilst p∘b=dp \circ b = d.

TT to KK in Letters and Words

Fix z=qT(0,1)z = q_T(0,1) as a basepoint z∈Tz \in T, and x0=qK(0,1)∈Kx_0 = q_K(0,1) \in K. The interior of qK(I2)q_K(I^2) allows us to define a homotopy cd−1c−1d−1≊ε\boxed{cd^{-1}c^{-1}d^{-1}} \approxeq \varepsilon with basepint x0x_0.

Abusing notation to identify loops with classes in π1(K,x0)\pi_1 (K,x_0), write cd−1c−1d−1=ecd^{-1}c^{-1}d^{-1} = e:

  ⟹  cd−1=dc, dc−1=c−1d−1,  ⟹  c2d−1=cdc=d−1c2.\implies cd^{-1} = dc, \ dc^{-1} = c^{-1}d^{-1},\\ \implies c^2d^{-1} = cdc = d^{-1}c^2.

If we now set a=c2a=c^2 and b=d−1b=d^{-1} (justified since p∗:π1(T,x0)→π1(K,x0)p_{*} : \pi_1 (T,x_0) \rightarrow \pi_1(K,x_0) is injective) we recover the relation ab=baab=ba or aba−1b−1=eaba^{-1}b^{-1} = e in π1(T,z)\pi_1(T,z).

We shall see that π1(K,x0)\pi_1 (K,x_0) is the group generated by cc and bb subject only to the relation cbc−1b=ecbc^{-1}b = e, induced by the boxed homotopy that shrinks the loop to a point. It follows that π1(T,z)\pi_1 (T,z) is isomorphic (via p∗p_{*}) to the abelian subgroup generated by a=c2a=c^2 and bb. Each element of π1(T,z)\pi_1(T,z) can be written uniquely as ambna^mb^n for some m,n∈Zm,n \in \mathbb{Z}. Converting to additive notation:

π1(T,x0)≅Z⊕Z≅Z2.\pi_1(T,x_0) \cong \mathbb{Z} \oplus \mathbb{Z} \cong \mathbb{Z}^2.

Summary

Topologically the 22-torus T=T2T = T^2 is the product S1×S1S^1 \times S^1, and there is a 'combined' covering map R2→T\mathbb{R}^2 \rightarrow T defined by:

(s,t)↦(e2πis,e2πit).(s,t) \mapsto (e^{2 \pi i s}, e^{2 \pi i t}).

A theorem about π1\pi_1 of a product of spaces confirms the isomorphism π1(T)≅Z×Z\pi_1(T) \cong \mathbb{Z} \times \mathbb{Z}.

Covering Spaces:

R2→T→2:1K,\mathbb{R^2} \rightarrow T \xrightarrow{2:1} K,

Fundamental Groups:

{e}→π1(T,z)=⟨c2,b:c2b=bc2⟩→index 2π1(K,x0)=⟨c,b:cb=b−1c⟩.\{e\} \rightarrow \pi_1(T,z) = \langle c^2, b: c^2b= bc^2 \rangle \xrightarrow{\text{index } 2} \pi_1(K,x_0) = \langle c, b: cb = b^{-1}c \rangle.