1. Fundamental Group of S1.
Let p : R → S1 be the map x 7→ e2πix. For any open interval (a, b) with b − a < 2π, p : (a, b) → S1 is a homeomorphism onto its image p(a, b). Moreover if U is any proper open subset of S1, then p−1(U ) is a disjoint union of open subsets {Ui} of R such that p : Ui→ U are homeomorphisms for all i. This is very easy to see from the map p. This motivates the definition of covering space.
Definition 1.1. We say that a continuous map π : E → X is a covering space of X if for each x ∈ X, there exists an open neighborhood U of x so that p−1(U ) is a disjoint union of open subsets {Ui} of E with the property that p : Ui → U is homeomorphisms for all i.
This definition says that p : R → S1 given by x 7→ e2πix is a covering space of S1. Proposition 1.1. (Lifting Lemma) Let σ : [0, 1] → S1 be a path in S1 with initial point 1. Then there exists a unique path σ0 in R with initial point 0 such that p ◦ σ0 = σ.
Proof. Let us prove the uniqueness first. Suppose σ00 is another path with initial point 1 such that p ◦ σ00 = σ. Then e2πi(σ0(t)−σ00(t)) = 1 for all t ∈ [0, 1]. Hence the image of σ0− σ00 lies in Z. Since both σ0 and σ00 are continuous map, so is σ0− σ00 Since [0, 1] is connected, by the continuity of σ0− σ00, the image of σ0− σ00 is also connected. Hence σ0− σ00 must be a constant function. Since σ0(0) = σ00(0) = 0, we see that σ0(t) − σ00(t) = 0 for all t ∈ [0, 1]
and thus σ0= σ00 on [0, 1].
Since [0, 1] is compact and σ is continuous on [0, 1], σ is uniformly continuous. Hence there exists δ > 0 so that |f (t) − f (t0)| < 1/2 for |t − t0| < δ. We can choose a partition 0 = t0 < · · · < tn= 1 such that |ti− ti+1| < δ/2 for 0 ≤ i ≤ n − 1.
Let B(a, ) be the open ball in C centered at a of radius . Consider the open subset Ui = B(ti, 1/2) ∩ S1. Then for each i, p−1(Ui) =S
αUi,α, where {Ui,α} is a family of disjoint open subsets of R so that p : Ui,α → Ui is a homeomorphism for all α. For i = 0, we choose U0,α0 so that 0 ∈ U0,α0. We see that σ|[t0,t1] ⊂ U0. We define σ10 : [t0, t1] → R by σ10 = p|−1U
0,α0 ◦ σ|[t0,t1]. Then σ01 is a path in R with initial point 0 so that p ◦ σ10 = σ|[t0,t1]. Now we shall construct the path inductively. Assume that we have already construct a path σi0 : [0, ti] → R such that p ◦ σi0 = σ|[0,ti]. We choose Ui,αi such that σi0(ti) ∈ Ui,αi. We define σi+10 : [0, ti+1] → R by
σ0i+1= σi0? (p|−1U
i,αi ◦ σ|[ti,ti+1]).
Then p ◦ σi+10 = σ|[0,ti]? σ|[ti,ti+1] = σ|[0,ti+1]. Here ? means the concatenation of curves.
Then we have already construct a curve σ0 : [0, 1] → R such that p ◦ σ0 = σ. In fact, the construction does not depend on the choice of partition of [0, 1] by the uniqueness proved above.
We call σ0 the lift of σ.
Proposition 1.2. (Covering Homotopy Lemma) Let τ be another path in S1 with initial point 1. Assume that
F : σ ' τ rel(0, 1).
Then there is a unique F0: I × I → R such that
F : σ0 ' τ0 rel(0, 1) and φ ◦ F0 = F.
1
2
Proof. The proof of the uniqueness of F0 is similar as above. We use the connectivity of I × I.
The interval I ×I is compact and F is continuous on I ×I. Hence F is uniformly continuous on I × I. Let Ui be as above for all i. For each point s ∈ I, we can find a neighborhood Ns of s and a partition {ti} of [0, 1] such that f (N × [ti, ti+1]) ⊂ Ui for all i.
Let FNs,1 : Ns× [0, t1] → R be the map defined by FN0s,1 = p|−1U
0,α0 ◦ F |Ns×[0,t1]. Then p ◦ FN0
s,1 = F |Ns×[0,t1]. We can construct a map inductively FN0
s : N × [0, 1] → R such that p ◦ FN0s = F |Ns×[0,1].
Let Ns0 be another open neighborhood of s. Then we can construct a map FN0 0
s such that p◦FN0 0
s = F |N0
s×[0,1]. We know that Ns0∩Nsis an open neighborhood of s. By the uniqueness, we know that FN0
s|Ns∩N0
s = FN0
s|Ns∩N0
s. We can use this to construct a continuous map FN0s∪N0
s : (Ns∪ Ns0) × [0, 1] → R so that p ◦ FN0s∪Ns0 = F |(Ns∪Ns0)×[0,1].
Since {Ns : s ∈ [0, 1]} forms an open cover of [0, 1] and [0, 1] is compact, we can find s1, · · · , sn such that {Nsi : 1 ≤ i ≤ n} forms a finite open subcover of [0, 1]. Let Ns1 be the open neighborhood of s1 containing 0 and denote F10 = FNs1. Let Ns2 be an open neighborhood so that Ns2∩ Ns1 6= ∅. Let F20 = FNs1∪Ns2. Inductively assume that Nsj is an open set such that Nsj∩ (Sj−1
i=1Nsi) 6= ∅ with Fj−10 is constructed. We can define Fj0 to be F(N0
s1∪···Nsj−1)∪Nsj. Then we set F0 = Fn0. It is easy to see that p ◦ F0 = F. Moreover if F is a homotopy between σ and τ, then F0 is a homotopy between their lifts σ0 and τ0. This
completes the proof of our assertion.
This proposition immediately implies that:
Corollary 1.1. The end point of σ0 depends only on the homotopy class of σ.
Theorem 1.1. The fundamental group π1(S1, 1) is isomorphic to Z.
Proof. Let σ be a loop of S1 at 1 and σ0 be its lift in R. Since σ is a loop, σ(1) = e2πiσ0(1) = 1.
Hence σ0(1) ∈ Z. For each [σ] ∈ π1(S1, 1), we choose a representative of σ and define χ : π1(S1, 1) → Z
by χ([σ]) = σ0(1). Then χ is a well-defined map by Corollary 1.1.
Let [σ] and [τ ] in π1(S1, 1). Choose representatives σ and τ of [σ] and [τ ] respectively.
Denote σ0(1) = m and τ0(1) = n. Then σ0(1) + τ0(1) = m + n. Let τ00 be the path from m to m + n defined by τ00= τ0+ m. We know that σ0? τ00 is the lift of σ ? τ with the initial point 0 by the uniqueness of the lift. Hence
χ([σ][τ ]) = χ([σ ? τ ]) = (σ0? τ00)(1) = m + n = χ([σ]) + χ([τ ]).
Therefore χ : π1(S1, 1) → Z is a group homomorphism.
Given n ∈ Z, define σ0(t) = nt for t ∈ [0, 1] and σ(t) = e2πint. Then χ([σ]) = n. Hence χ is surjective. If [σ] ∈ ker χ, then σ0(1) = 0, i.e. σ0 is a loop at 0. Then we know that σ0 is homotopic to the constant loop e0 at 0. Hence σ is homotopy to the constant loop e1= e2πie0 at 1. Hence ker χ is trivial. We complete the proof. Proposition 1.3. Let (X1, x1) and (X2, x2) be two pointed spaces. Then there is a natural identification:
π1(X1× X2, (x1, x2)) ∼= π1(X1, x1) × π1(X2, x2).
3
Proof. Let p1 and p2 be the projections of X1 × X2 to X1 and X2 respectively. Then p1
and p2 induce group homomorphisms:
(pi)∗ : π1(X1× X2, (x1, x2)) → π1(Xi, xi), i = 1, 2.
Hence we obtain a group homomorphism
((p1)∗, (p2)∗) : π1(X1× X2, (x1, x2)) → π1(X1, x1) × π1(X2, x2).
This group homomorphism is in fact an isomorphism. Let us construct its inverse.
Let σ1 and σ2 be loops at x1 and x2 respectively. Then we define a loop σ1× σ2 at (x1, x2) by setting
(σ1× σ2)(t) = (σ1(t), σ2(t)).
For any given [σi] ∈ π1(Xi, xi), i = 1, 2, we choose a representative σi of [σi] for each i.
Then we consider the map
π1(X1, x1) × π1(X2, x2) → π1(X1× X2, (x1, x2)) given by ([σ1], [σ2]) 7→ [σ1× σ2]. Then this is the required inverse map.
If we take T = S1× S1, then we obtain immediately that
π1(T, (1, 1)) ∼= Z2.