1. Loop Space and Higher homotopy groups
Let K be a compact Hausdorff space and X be a topological space. The space of all continuous functions from K to X is denoted by C(K, X). Let A be a closed subset of K (and thus A is compact) and U be an open subset of X. Let β(A, U ) be the family of continuous functions f : K → X such that f (A) ⊂ U, i.e.
β(A, U ) = {f ∈ C(K, X) : f (A) ⊂ U }.
The family {β(A, U )} forms a subbase of a topology on C(K, X); this topology is called the compact- open topology on C(K, X).
A pointed topological space is a topological space X together with a point x0 ∈ X. A pointed topological space is denoted by (X, x0). A morphism/map from (X, x0) to (Y, y0) is a continuous map f : X → Y such that f (x0) = y0.
A loop in X with based point x0 is a continuous map c : [0, 1] → X
such that c(0) = c(1) = x0. We may identify c with a morphism/map c : (S1, 1) → (X, x0) between pointed topological spaces. The constant loop at x0 is the constant map ex0 : (S1, 1) → (X, x0) so that ex0(t) = x0 for all t ∈ S1. The space of loops in (X, x0) has a subspace topology induced from C(S1, X). Together with the constant loop ex0, the space of loops is denoted by Ω(X, x0) and called the loop space of X with base point x0. The fundamental group of (X, x0) is defined to be
π1(X, x0) = π0(Ω(X, x0)).
Here π0(Y ) denotes the set of connected components of a topological space Y. We call π1(X, x0) the fundamental group of the pointed space (X, x0). Since Ω(X, x0) is again a pointed space, its fundamental group is denoted by π2(X, x0), i.e. π2(X, x0) = π1(Ω(X, x0)). Inductively, we define the higher homotopy groups of (X, x0) by
πn(X, x0) = πn−1(Ω(X, x0)) for n ≥ 1.
One can show that the construction of higher homotopy groups is functorial. In other words, for each n ≥ 1, πn defined a functor from the category of pointed topological spaces into the category of groups.
The following theorem is the most useful tool in homotopy theory.
Theorem 1.1. Let p : (E, e0) → (B, b0) be a morphism of pointed spaces such that p : E → B is a fiber bundle (or a fiber map). Let F = π−1(b0). For each n ≥ 1, there exists a group homomorphism δn: πn(B, b0) → πn−1(F, x0) such that the following sequence is exact:
(1.1) · · · −−−−→ πn(F, x0) −−−−→ ππn(i) n(E, e0) −−−−→ ππn(p) n(B, b0) −−−−→ πδn n−1(F, x0) −−−−→ · · · . Here i : F → X is the inclusion map.
Example 1.1. Let p : (R, 0) → (S1, 1) be the morphism of pointed spaces corresponding to the map p(t) = e2πit for t ∈ R. Then the fiber of p over 1 is Z. Study the long exact sequence of homotopy groups induced from the fiber map
Z → R → S1.
Example 1.2. Let Sn be the n-dimensional unit sphere in Rn+1, i.e.
Sn+1= {(x0, · · · , xn) ∈ Rn+1: x20+ · · · + x2n= 1}.
The antipodal map A is a continuous map Sn → Sn sending x to −x. Identify Z2 = Z/2Z with the subgroup {I, A} of the automorphism group of Sn. The quotient space Sn/Z2denoted by RPn called the n-dimensional real projective space. Study the long exact sequence of the homotopy groups induced from the fiber map
Z2→ Sn→ RPn.
1
Example 1.3. Let S2n+1 be the 2n + 1 dimensional unit sphere in Cn i.e.
S2n+1 = {z = (z1, · · · , zn) ∈ Cn : |z1|2+ · · · + |zn|2= 1}.
The one dimensional compact torus S1= {z ∈ C : |z| = 1} acts on S2n+1 by S1× S2n+17→ S2n+1, (λ, z) → λz.
The quotient space S2n+1/S1denoted by CPn is called the n-dimensional complex projective space.
Study the long exact sequence of homotopy groups induced from the fiber map S1→ S2n+1 → CPn.
2. Singular homology The standard q-simplex is the subset of Rq+1 defined by
∆q = {(t0, t1, · · · , tq) :
q
X
i=0
ti= 1, ti≥ 0}.
It is equipped with the subspace topology induced from the Euclidean topology on Rq+1. Let X be a topological space. A singular q-simplex in X is a continuous map
σ : ∆q→ X.
The space of singular q-simplices is denoted by
Sq = {σ : ∆q → X : σ is continuous}.
A singular q-chain over a ring R1is a function
α : Sq → R
so that {σ ∈ Sq : α(σ) 6= 0R} is a finite set. Here 0Ris the additive identity of the ring R. The space of singular q-chains in X is denoted either by Sq(X, R) or by Cq(X, R). We define an R-module structure on Cq(X, R) by
(α + β)(σ) = α(σ) + β(σ), (aα)(σ) = aα(σ), σ ∈ F ,
where α, β ∈ Cq(X, R) and a ∈ R. The 0 element of Cq(X, R) is the function 0 : Sq→ R defined by 0(σ) = 0R.
When q ∈ Z with q < 0, we set Sq(X, R) = 0 the trivial R-module.
For each σ : ∆q → X, we define a q-chain δσ: Sq → R by δσ(τ ) = δστ1R. Remark. We denote α(σ) by ασ.
Lemma 2.1. The set {δσ : σ ∈ Sq} forms an R-basis for Cq(X, R) for q ≥ 0.
Proof. Show that {δσ : σ ∈ Sq} is linearly independent over R and that α = P
σ∈Sqασδσ for any
α ∈ Cq(X, R).
Lemma 2.2. Show that the map
δ : Sq → Cq(X, R), σ 7→ δσ
is injective.
We identify σ with its image δσ in Cq(X, R). An element α ∈ Cq(X, R) can be rewritten as α = X
σ∈F
ασσ.
2
For each 0 ≤ i ≤ q, we define a map Fqi: Rq → Rq+1by
Fqi(x0, x1, · · · , xq−1) = (x0, x1, · · · , xi−1, 0, xi, · · · , xq−1)
for (x0, x1, · · · , xq) ∈ Rq+1. Then Fqi are continuous for 0 ≤ i ≤ q. For each 0 ≤ i ≤ q, we define the i-th face map fqi : ∆q−1→ ∆q to be the restriction of Fqi to the standard q − 1 simplex fqi = Fqi|∆q−1. The continuities of the family of functions {fqi : 0 ≤ i ≤ q} follow from the fact that Fqi are continuous.
The i-th face of a singular q-simplex σ on X is a singular q − 1-simplex σ(i): ∆q−1→ X
1R is assumed to be commutative with multiplicative identity 1R. 2Thus 0 element in Cq(X, R) can be represented as 0 =P
σ∈F0Rσ.
defined by σ(i)= σ ◦ fqi.
The boundary of a singular q-simplex σ on X is a q − 1-chain ∂qσ on X defined by
∂qσ =
q
X
i=0
(−1)iσ(i).
In general, the boundary of a singular q-chain α is defined to be the q − 1-chain
∂qα = X
σ∈Sq
ασ(∂qσ).
When q < 0, we set ∂q= 0 to be the zero map.
Lemma 2.3. The maps ∂q : Cq(X, R) → Cq−1(X, R) is R-linear for all q ∈ Z such that
∂q◦ ∂q+1= 0.
Let Zq(X, R) = ker ∂q and Bq(X, R) = Im ∂q+1. Elements of Zq(X, R) are called q-cycles and elements of Bq(X, R) are called q-boundaries. Since ∂q◦ ∂q+1= 0, Bq(X, R) is an R-submodule of Zq(X, R). The quotient R-module
Hq(X, R) = Zq(X, R)/Bq(X, R)
is called the q-th singular homological module of X with coefficients in R. We set H∗(X, R) =M
q∈Z
Hq(X, R) called the singular homology theory of X over R. We also denote
C∗(X, R) =M
q∈Z
Cq(X, R).
Example 2.1. Let X = {x} be a topological space with a single point. Then Hq(X, R) =
(R if q = 0 0 otherwise.
Definition 2.1. A complex over a ring R is a sequence of R-modules {Cn: n ∈ Z} together with a sequence of R-linear maps ∂n : Cn → Cn−1 such that ∂n◦ ∂n+1= 0. A complex over R is denoted by C∗= (Cn, ∂n). The n-th homology of C∗is defined to be
Hn(C∗) = ker ∂n/ Im ∂n+1.
Using the same terminology, elements of Zn(C∗) = ker ∂n are called n-cycles while elements of Bn(C∗) = Im ∂n+1are called n-boundaries.
3. Reduced Homology
Theorem 3.1. Let X be a space and {Xα: α ∈ Λ} be the set of all path components of X. For all q ≥ 0,
Hq(X, R) ∼=M
α∈Λ
Hq(Xα, R).
Hence to study the singular homology of a space, we only need to find out the singular homology of its path components. Assume that X is path connected.
The standard 0-simplex is the set ∆0 = {0}. A singular 0-simplex in X is a continuous map σ : ∆0→ X. Then σ is determined by σ(0). The set S0 and X are in one-to-one correspondence:
ι : S0→ X, σ 7→ σ(0).
We identify C0(X, R) with the space of functions α : X → R so that such that {x ∈ X : α(x) 6= 0}
is a finite set. Thus an element of C0(X, R) is represented asP
x∈Xrxx where rx∈ R.
The standard 1-simplex ∆1 is the subset {(t0, t1) ∈ R2 : t0, t1 ≥ 0, t0+ t1 = 1}. A singular simplex in X is a continuous map σ : ∆1→ X. Let h : [0, 1] → ∆1be the map t 7→ (1 − t, t). Then h is a homeomorphism3. We obtain a path σ ◦ h : [0, 1] → X in X. The space of path in X is denoted by P = C([0, 1], X). There is a bijection
S1→ P, σ 7→ σ ◦ h.
We identify C1(X, R) with the space of functions β : P → R so that {c : β(c) 6= 0} is a finite set.
We represent β as a finite sum β =P
c∈Prcc. Here rc ∈ R and c : [0, 1] → X denotes a path. Now we would like to express the boundary map ∂1using the new expression.
The face map f10: ∆0→ ∆1is the map f10(0) = (0, 1) and f11: ∆0→ ∆1is the map f11(0) = (1, 0).
If σ : ∆1→ X is a singular 1-simplex, then
σ(0)(0) = σ(0, 1) = σ ◦ h(1), σ(1)(0) = σ(1, 0) = σ ◦ h(0).
Recall the boundary of an 1-singular simplex σ in X is defined to be the 0-chain
∂1σ = σ(0)− σ(1).
Thus if c : [0, 1] → X is a path corresponding to the singular 1-simplex c ◦ h−1 : ∆1→ X, then by the identification described above, we find ∂1c = c(1) − c(0). If β =P
c∈Prcc, then
∂1β =X
c∈P
rc(c(1) − c(0))
=X
c
rcc(1) −X
c
rcc(0)
= X
x:c(1)=x
rcx − X
c:c(0)=x
rcx.
Let : C0(X, R) → R be the map
X
x∈X
rxx
!
= X
x∈X
rx.
Then is surjective and thus by the first isomorphism, we find C0(X, R)/ ker ∼= R. Observe that
X
x:c(1)=x
rcx
=X
c
rc=
X
x:c(0)=x
rcx
. Since is R-linear, we find
◦ ∂1β =X
c
rc−X
c
rc= 0.
3Check as an exercise
In other words, ◦ ∂1= 0 which gives Im ∂1⊆ ker . In fact, we can prove Lemma 3.1. ker = Im ∂1.
Proof. Let z ∈ ker . Then z = 0. Write z =P
x∈Xrxx. ThenP
x∈Xrx= 0. Denote {x ∈ X : rx6=
0} = {x1, · · · , xn} and rxi = ri for 1 ≤ i ≤ n. Then z = Pn
i=1rix. Choose x0 ∈ X. By the path connectedness of X, for each 1 ≤ i ≤ n, we choose a path βi from x0 to xn. Write β =Pn
i=1riβi. Then
∂β =
n
X
i=1
ri(xi− x0)
=
n
X
i=1
rixi−
n
X
i=1
rix0
=
n
X
i=1
rixi−
n
X
i=1
ri
! x0
= z
Here use the fact that r1+ · · · + rn = 0. We find z = ∂β for some β ∈ C1(X, R). We prove our
assertion.
Since the zeroth singular homology group H0(X, R) is defined to be C0(X, R)/ Im ∂1 and ∂1 = ker , we find
H0(X, R) = C0(X, R)/ ker ∼= R when X is path connected. In general by Theorem 3.1, we obtain:
Corollary 3.1. Let X be a space and C be the set of all path components of X. Then H0(X, R) ∼=M
C
R.
We have set Ci(X, R) = 0 when i < 0 and ∂i: Ci(X, R) → Ci−1(X, R) to be the zero map for all i < 0. In fact, we can define a new complex C0= (Ci0, ∂i0) by setting
Ci0=
Ci(X, R) if i ≥ 0
R if i = −1
0 if i < −1.
and ∂i0=
∂i if i > 0
if i = 0 0 if i < 0.
.
The i-th homology of the new complex C0is called the reduced i-th homology of the space X and denoted by
Hi#(X, R) = Hi(C0).
It follows from the definition that Hi#(X, R) =
(Hi(X, R) if i 6= 0
0 if i = 0 and if X is path connected.
When X is an one pointed space, all of its reduced homology modules vanish, i.e.
Hi#(pt, R) = 0, i ∈ Z.
This motivates the definition of acyclic complex.
Definition 3.1. A complex C = (Ci, ∂i) is acyclic if all of its homological modules vanish, i.e.
Hi(C) = 0 for all i ∈ Z.
Notice that Hi(C) = ker ∂i/ Im ∂i+1 and hence C is acyclic if and only if ker ∂i= Im ∂i+1, i.e. the complex C forms an exact sequence of R-modules.
Let C = (Ci, ∂i) be a complex with the property that Ci = 0 for all i < 0. An augmentation of such a complex C over R is an R-epimorphism : C0→ R such that ∂1 = 0. A complex with an augmentation gives us a new complex C0 defined by
Ci0 =
Ci if i ≥ 0 R if i = −1 0 if i < −1.
and ∂0i=
∂i if i > 0
if i = 0 0 if i < 0.
.
The new complex C0 is called the reduced chain complex associated with (C, ). The complex C0 depends on the choice of . The corresponding homology of C0 is called the reduced homology associated with (C, ) and denoted by
Hi#(C) = Hi(C0), i ∈ Z.
Definition 3.2. A chain complex C with augmentation is acyclic if its corresponding reduced chain complex is acyclic.
Hence the chain complex of an one point space with the above given augmentation is acyclic.
Proposition 3.1. A complex C with augmentation is acyclic if Hi(C) =
(0 if i 6= 0 R if i = 0.
Proof. Using the fact that Hi#(C) = Hi(C) for i 6= 0, we see that Hi#(C) = 0 if and only if Hi(C) = 0 for i 6= 0.
Since : C0 → R is surjective, C0/ ker ∼= R. If C0 is acyclic, ker = Im ∂1. In this case, C0/ Im ∂1∼= R. Note that H0(C) = C0/ Im ∂1, we find H0(C) ∼= R. This completes the proof.
4. Category of complexes and the Homological Functor
Let f : X → Y be a continuous map. Given a singular q-simplex σ : ∆q → X in X, the continuous map f ◦ σ : ∆q → Y determine a q-simplex in Y. We define an R-linear map
Cq(f ) : Cq(X, R) → Cq(Y, R), X
σ
rσσ 7→X
σ
rσ(f ◦ σ).
Then we obtain the following diagrams of R-linear maps
· · · −−−−→ Cq+1(X, R) ∂
X
−−−−→ Cq+1 q(X, R) ∂
X
−−−−→ Cq q−1(X, R) −−−−→ · · ·
y Cq+1(f )
y Cq(f )
y Cq−1(f )
y
y
· · · −−−−→ Cq+1(Y, R) ∂
Y
−−−−→ Cq+1 q(Y, R) ∂
Y
−−−−→ Cq q−1(Y, R) −−−−→ · · · .
Here ∂qX and ∂qY are the boundaries maps on Cq(X, R) and on Cq(Y, R) respectively. If σ : ∆q→ X is a singular q-simplex, then ∂qσ =Pq
i=0(−1)iσ(i). Thus Cq−1(f ) ◦ ∂qX(σ) = Cq−1(f )(∂qσ) =
q
X
i=0
(−1)if ◦ σ(i)=
q
X
i=0
(−1)i(f ◦ σ)(i). On the other hand, Cq(f )(σ) = f ◦ σ and hence
∂qYCq(f )(σ) =
q
X
i=0
(−1)i(f ◦ σ)(i). The above two equations imply that
∂qYCq(f ) = Cq−1(f )∂qX.
In other words, the above diagram commutes. This motivates the definition of chain maps or morphisms between chain complexes.
Definition 4.1. Let A = (Ai, ∂iA) and B = (Bi, ∂iB) be two chain complexes over R. A chain map f = {fi} (morphism) from A to B is a sequence of R-linear maps fi: Ai→ Bi such that
∂iB◦ fi= fi−1◦ ∂iA, for all i ∈ Z.
In other words, the following diagram commutes:
· · · −−−−→ Ai+1 ∂
A
−−−−→ Ai+1 i ∂ A
−−−−→ Ai i−1 −−−−→ · · ·
y fi+1
y fi
y fi−1
y
y
· · · −−−−→ Bi+1
∂Bi+1
−−−−→ Bi
∂Bi
−−−−→ Bi−1 −−−−→ · · · .
The identify morphism idA: A → A is defined by the sequence of identity maps {idAi: Ai→ Ai}.
If f : A → B and g : B → C are chain maps, we define their composition g ◦ f to be the sequence of R-linear maps {gi◦ fi: Ai→ Ci}. We leave to the reader to verify that g ◦ f is again a chain map.
Proposition 4.1. The collections of all complexes over R together with chain maps forms a category denoted by CompR. The category CompR is called the category of complexes over R.
Corollary 4.1. Any continuous map f : X → Y between topological spaces X, Y determines a chain map
C∗(f ) : C∗(X, R) → C∗(Y, R),
between the singular chain complexes of X and Y , where C∗(f ) = {Ci(f ) : i ∈ Z} such that the following properties hold.
(1) If idX : X → X is the identity map, then C∗(idX) = idC∗(X,R).
(2) If f : X → Y and Y → Z are continuous maps, then C∗(g ◦ f ) = C∗(g) ◦ C∗(f ).
This corollary implies that C∗: Top → CompR defines a functor from the category of topological spaces into the category of complexes over R.
Lemma 4.1. Let f : A → A0 be a chain map between chain complexes A and B. For each i ∈ Z, we define
Hi(f ) : Hi(A) → Hi(B), [z] 7→ [fi(z)].
Hi(f ) is a well-defined R-linear homomorphism.
Proof. Let z, z0 ∈ ker ∂i so that z − z0 = ∂Ai+1w for some w ∈ Ai+1. Since f is a chain map, fi(z) − fi(z0) = fi(z − z0) = fi∂i+1A w = ∂i+1B fi+1w
Hence fi(z) − fi(z0) ∈ Im ∂i+1B . Thus [fi(z)] = [fi(z0)]. The linearity of Hi(f ) follows from the
linearity of fi.
Lemma 4.2. Let f : A → B and g : B → C be chain maps. Then Hi(g ◦ f ) = Hi(g) ◦ Hi(f ) for any i ∈ Z.
Proof. Routine check.
Corollary 4.2. For each i ∈ Z, the assignment Hi : A 7→ Hi(A) sending a chain complex over R to its i-th homology and Hi : (f : A → B) 7→ Hi(f ) : Hi(A) → Hi(B) sending a chain map to its corresponding induced map defines a functor
Hi: CompR→ ModR
from the category of complexes over R to the category of R-modules ModR.
5. Homotopy and Chain Homotopy