Section 6.10

Cellular homology of a CW complex

[15/100%]
Show
LaTeX
Problem 6.10.1
?
(a)

Show the inclusion (Dk,Sk−1)→(Dk,Lk−1)(D^k, S^{k-1}) \to (D^k, L^{k-1}) induces an isomorphism in homology.

(b)

Show that the inclusion (Xk,Xk−1)→(Xk,Kk−1)(X^k, X^{k-1}) \to (X^k, K^{k-1}) induces an isomorphism in homology.

Remark.
?

For part (a): show that there is a homotopy equivalence of pairs. For part (b): use a similar argument.

Setup: XkX^k denotes the kk-skeleton of XX, the union of the cells of dimension ≤k\leq k. The kk-cell ejke_j^k is a unit disk (up to homeomorphism) attached to Xk−1X^{k-1} via an attaching map fj:Sk−1→Xk−1f_j : S^{k-1} \to X^{k-1}. Extend Xk−1X^{k-1} to a larger set Kk−1K^{k-1} in XkX^k by attaching Lk−1=Dk∖(int 12Dk)L^{k-1} = D^k \setminus (\mathrm{int}\, \frac{1}{2} D^k) via fjf_j for each kk-cell. There is a corresponding pair (Djk,Ljk−1)(D_j^k, L_j^{k-1}) for each kk-cell; write (Dk,Lk−1)=(⨆j=1ckDjk,⨆j=1ckLjk−1)(D^k, L^{k-1}) = (\bigsqcup_{j=1}^{c_k} D_j^k, \bigsqcup_{j=1}^{c_k} L_j^{k-1}) and (Dk,Sk−1)=(⨆j=1ckDjk,⨆j=1ckSjk−1)(D^k, S^{k-1}) = (\bigsqcup_{j=1}^{c_k} D_j^k, \bigsqcup_{j=1}^{c_k} S_j^{k-1}) for the corresponding disjoint unions.

Problem 6.10.2
?
(a)

Show that if (Y,B)=(⨆j=1NYj,⨆j=1NBj)(Y, B) = (\bigsqcup_{j=1}^N Y_j, \bigsqcup_{j=1}^N B_j) is a disjoint union of path components (Yj,Bj)(Y_j, B_j), then Hp(Y,B)=⨁j=1NHp(Yj,Bj)H_p(Y, B) = \bigoplus_{j=1}^N H_p(Y_j, B_j).

(b)

Show that Hk(Dk,Sk−1)≃⨁j=1ckZH_k(D^k, S^{k-1}) \simeq \bigoplus_{j=1}^{c_k} \mathbb {Z} (notation as in Exercise 6.10.1). Call the generator gjkg_j^k that corresponds to a generator of Hk(Djk,Sjk−1)H_k(D_j^k, S_j^{k-1}).

(c)

Show that the inclusion Hp(Dk,Sk−1)→Hp(Dk,Lk−1)H_p(D^k, S^{k-1}) \to H_p(D^k, L^{k-1}) induces an isomorphism in homology. Conclude that

Hp(Dk,Lk−1)=⨁j=1ckHp(Djk,Ljk−1)≃⨁j=1ckZ. H_p(D^k, L^{k-1}) = \bigoplus _{j=1}^{c_k} H_p(D_j^k, L_j^{k-1}) \simeq \bigoplus _{j=1}^{c_k} \mathbb {Z}.
Problem 6.10.3

Show that the restriction of the characteristic maps to the kk-cells (notation as in Exercise 6.10.1) gives a homeomorphism of pairs (Dk∖int Lk−1,Lk−1∖int Lk−1)→(Xk∖int Kk−1,Kk−1∖int Kk−1)(D^k \setminus \mathrm{int}\, L^{k-1}, L^{k-1} \setminus \mathrm{int}\, L^{k-1}) \to (X^k \setminus \mathrm{int}\, K^{k-1}, K^{k-1} \setminus \mathrm{int}\, K^{k-1}). Conclude that there is an isomorphism in homology.

?
Problem 6.10.4

Show that there is a commutative diagram where all maps are isomorphisms,

Hp(Dk,Sk−1)→ϕ∗Hp(Xk,Xk−1)↓(iL)∗↓(iK)∗Hp(Dk,Lk−1)→(ϕL)∗Hp(Xk,Kk−1)↑(jL)∗↑(jK)∗Hp(Dk∖int Lk−1,Lk−1∖int Lk−1)→(ϕ−L)∗Hp(Xk∖int Kk−1,Kk−1∖int Kk−1) \begin{array}{ccc} H_p(D^k, S^{k-1}) & \xrightarrow {\phi _*} & H_p(X^k, X^{k-1}) \\ \downarrow {(i_L)_*} & & \downarrow {(i_K)_*} \\ H_p(D^k, L^{k-1}) & \xrightarrow {(\phi _L)_*} & H_p(X^k, K^{k-1}) \\ \uparrow {(j_L)_*} & & \uparrow {(j_K)_*} \\ H_p(D^k \setminus \mathrm{int}\, L^{k-1}, L^{k-1} \setminus \mathrm{int}\, L^{k-1}) & \xrightarrow {(\phi _{-L})_*} & H_p(X^k \setminus \mathrm{int}\, K^{k-1}, K^{k-1} \setminus \mathrm{int}\, K^{k-1}) \end{array}

where ϕ=⊔ϕj\phi = \sqcup \phi_j is the disjoint union of the characteristic maps of the kk-cells (restricting to the attaching maps on the boundary spheres), the other horizontal maps are induced by restrictions of this map, and the vertical maps are inclusions. Use Exercises 6.10.1--6.10.3 to justify why each map is an isomorphism.

?
Problem 6.10.5

Show that when k>0k > 0, Hk(⋁j=1ckSjk,∗)≃⨁j=1ckHk(Sjk,∗)H_k(\bigvee_{j=1}^{c_k} S_j^k, *) \simeq \bigoplus_{j=1}^{c_k} H_k(S_j^k, *).

?
Remark.
?

Combining the results of Exercises 6.10.3 and 6.10.4 shows that the characteristic maps induce an isomorphism ⨁j=1ckHp(Djk,Sjk−1)→Hp(Xk,Xk−1)\bigoplus_{j=1}^{c_k} H_p(D_j^k, S_j^{k-1}) \to H_p(X^k, X^{k-1}); in particular all terms are 00 unless p=kp = k, and when p=kp = k the generator gjkg_j^k of Hp(Djk,Sjk−1)H_p(D_j^k, S_j^{k-1}) is mapped to a generator xjkx_j^k of Hk(Xk,Xk−1)H_k(X^k, X^{k-1}), which is thus a free abelian group with generators x1k,…,xckkx_1^k, \ldots , x_{c_k}^k.

For k>0k > 0, the quotient space Dk/Sk−1D^k/S^{k-1} is homeomorphic to SkS^k. Starting with the disjoint union (Dk,Sk−1)(D^k, S^{k-1}) and forming the quotient Dk/Sk−1D^k/S^{k-1} gives the space formed from ckc_k copies of SkS^k with all the base points (coming from the Sjk−1S_j^{k-1} terms collapsed to a point) identified to a single base point ∗*; thus the quotient space is ⋁j=1ckSjk\bigvee_{j=1}^{c_k} S_j^k. The map ϕ\phi (of Exercise 6.10.4) induces a homeomorphism ϕˉ\bar{\phi } between Xk/Xk−1X^k/X^{k-1} and this wedge, giving an isomorphism Hp(⋁j=1ckSjk,∗)≃Hp(Xk/Xk−1,∗)H_p(\bigvee_{j=1}^{c_k} S_j^k, *) \simeq H_p(X^k/X^{k-1}, *).

Hint for this exercise: use an inductive argument with the Mayer--Vietoris sequence.

Problem 6.10.6

Any element yy of the direct sum ⨁j=1ckHk(Sjk,∗)\bigoplus_{j=1}^{c_k} H_k(S_j^k, *) of Exercise 6.10.5 is expressible as ∑j=1cknjyj\sum_{j=1}^{c_k} n_j y_j, where yjy_j is a chosen generator of Hk(Sjk,∗)H_k(S_j^k, *). Let AjA_j represent the spheres other than the jjth one in ⋁j=1ckSjk\bigvee_{j=1}^{c_k} S_j^k; the quotient ⋁j=1ckSjk/Aj\bigvee_{j=1}^{c_k} S_j^k / A_j is homeomorphic to SjkS_j^k, and under the quotient map PjP_j collapsing AjA_j, yy maps to njyjn_j y_j. This uses the fact that the composition Sjk→⋁j=1ckSjk→⋁j=1ckSjk/Aj=SjkS_j^k \to \bigvee_{j=1}^{c_k} S_j^k \to \bigvee_{j=1}^{c_k} S_j^k/A_j = S_j^k is the identity, and that the composition Spk→⋁j=1ckSjk→⋁j=1ckSjk/Aj=SjkS_p^k \to \bigvee_{j=1}^{c_k} S_j^k \to \bigvee_{j=1}^{c_k} S_j^k/A_j = S_j^k maps SpkS_p^k to the base point ∗* when p≠jp \neq j.

Verify that the map Spk→SjkS_p^k \to S_j^k, p≠jp \neq j, which sends the whole sphere to the base point, induces the zero map in homology in dimension k>0k > 0 (this is the fact used above to conclude that the identity induces the identity map while the constant map induces the zero map, giving the algebraic description of PjP_j).

?
Problem 6.10.7

Let Aj⊂DkA_j \subset D^k, Bj⊂XkB_j \subset X^k denote the subcomplexes consisting of everything except the jjth kk-cell (notation as in Exercise 6.10.1). We can form quotient spaces Dk/AjD^k/A_j, Xk/BjX^k/B_j by collapsing AjA_j and BjB_j to a point. Show that the characteristic map Djk→Dk→XkD_j^k \to D^k \to X^k induces homeomorphisms between the quotient spaces Djk/Sjk−1→Dk/Aj→Xk/BjD_j^k/S_j^{k-1} \to D^k/A_j \to X^k/B_j.

?
Problem 6.10.8

Each sphere SjkS_j^k (as in Exercise 6.10.5) has a closed neighborhood Lˉjk−1\bar{L}_j^{k-1} about the base point ∗*, coming from the image of Ljk−1L_j^{k-1} under the collapsing map. The union of all of these in ⋁j=1ckSjk\bigvee_{j=1}^{c_k} S_j^k gives a closed neighborhood Lˉk−1\bar{L}^{k-1} of ∗*. Similarly, Kk−1K^{k-1} (as in Exercise 6.10.1) provides a closed neighborhood Kˉk−1\bar{K}^{k-1} of the base point ∗∈Xk/Xk−1* \in X^k/X^{k-1}; the homeomorphism ϕˉ\bar{\phi } (Exercise 6.10.5) sends Lˉk−1\bar{L}^{k-1} to Kˉk−1\bar{K}^{k-1}.

Show that (Dk/Sk−1,Lˉk−1)(D^k/S^{k-1}, \bar{L}^{k-1}) is homotopy equivalent to (Dk/Sk−1,∗)(D^k/S^{k-1}, *), and (Xk/Xk−1,Kˉk−1)(X^k/X^{k-1}, \bar{K}^{k-1}) to (Xk/Xk−1,∗)(X^k/X^{k-1}, *).

?
Problem 6.10.9

Show that the restriction of the quotient map gives a homeomorphism

(Xk∖int Kk−1,Kk−1∖int Kk−1)→(Xk/Xk−1∖int Kˉk−1,Kˉk−1∖int Kˉk−1). (X^k \setminus \mathrm{int}\, K^{k-1}, K^{k-1} \setminus \mathrm{int}\, K^{k-1}) \to (X^k/X^{k-1} \setminus \mathrm{int}\, \bar{K}^{k-1}, \bar{K}^{k-1} \setminus \mathrm{int}\, \bar{K}^{k-1}).

(notation as in Exercise 6.10.8).

?
Problem 6.10.10
?
(a)

Suppose k>0k > 0. Show that there is a commutative diagram where all maps are isomorphisms, horizontal maps induced by inclusions and vertical maps induced by quotient maps,

Hk(Xk,Xk−1)→Hk(Xk,Kk−1)←Hk(Xk∖int Kk−1,Kk−1∖int Kk−1)↓↓↓Hk(Xk/Xk−1,∗)→Hk(Xk/Xk−1,Kˉk−1)←Hk(Xk/Xk−1∖int Kˉk−1,Kˉk−1∖int Kˉk−1) \begin{array}{ccccc} H_k(X^k, X^{k-1}) & \to & H_k(X^k, K^{k-1}) & \leftarrow & H_k(X^k \setminus \mathrm{int}\, K^{k-1}, K^{k-1} \setminus \mathrm{int}\, K^{k-1}) \\ \downarrow & & \downarrow & & \downarrow \\ H_k(X^k/X^{k-1}, *) & \to & H_k(X^k/X^{k-1}, \bar{K}^{k-1}) & \leftarrow & H_k(X^k/X^{k-1} \setminus \mathrm{int}\, \bar{K}^{k-1}, \bar{K}^{k-1} \setminus \mathrm{int}\, \bar{K}^{k-1}) \end{array}
(b)

Deduce a similar result for (Dk,Sk−1)(D^k, S^{k-1}) in place of (Xk,Xk−1)(X^k, X^{k-1}).

(c)

Show that there is a commutative diagram, where all maps are isomorphisms,

Hk(Dk,Sk−1)→Hk(Xk,Xk−1)↓↓Hk(Dk/Sk−1,∗)→Hk(Xk/Xk−1,∗)↘↓Hk(⋁j=1ckSjk,∗) \begin{array}{ccc} H_k(D^k, S^{k-1}) & \to & H_k(X^k, X^{k-1}) \\ \downarrow & & \downarrow \\ H_k(D^k/S^{k-1}, *) & \to & H_k(X^k/X^{k-1}, *) \\ & \searrow & \downarrow \\ & & H_k(\bigvee _{j=1}^{c_k} S_j^k, *) \end{array}
Problem 6.10.11

Define a generator yjky_j^k of Hk(⋁j=1ckSjk,∗)H_k(\bigvee_{j=1}^{c_k} S_j^k, *) coming from the image of the generator gjk∈Hk(Dk,Sk−1)g_j^k \in H_k(D^k, S^{k-1}) (using the commutative diagram of Exercise 6.10.10(c), this is the image of xjkx_j^k via the vertical maps on the right). If PjP_j collapses all but the jjth sphere of ⋁j=1ckSjk\bigvee_{j=1}^{c_k} S_j^k, then (Pj)∗(yjk)=yjk(P_j)_*(y_j^k) = y_j^k and (Pm)∗(yjk)=0(P_m)_*(y_j^k) = 0, m≠jm \neq j, by Exercise 6.10.6.

Define the cellular chain groups Ckc(X)=Hk(Xk,Xk−1)C_k^c(X) = H_k(X^k, X^{k-1}), k≥0k \geq 0, where X−1=∅X^{-1} = \emptyset, so C0c(X)=H0(X0)C_0^c(X) = H_0(X^0); this is a free abelian group with generators xjkx_j^k, one for each kk-cell. Define the boundary homomorphism ∂kc\partial_k^c to be the composition

Hk(Xk,Xk−1)→∂Hk−1(Xk−1)→j∗Hk−1(Xk−1,Xk−2). H_k(X^k, X^{k-1}) \xrightarrow {\partial } H_{k-1}(X^{k-1}) \xrightarrow {j_*} H_{k-1}(X^{k-1}, X^{k-2}).

(When k=1k = 1, this is just ∂1c=∂:H1(X1,X0)→H0(X0)\partial_1^c = \partial : H_1(X^1, X^0) \to H_0(X^0); as usual ∂0c=0\partial_0^c = 0.)

Show that ∂kc∂k+1c=0\partial_k^c \partial_{k+1}^c = 0.

?
Remark.
?

Use the exactness of the long exact sequence of the pair (Xk,Xk−1)(X^k, X^{k-1}).

Problem 6.10.12

For S2×S2S^2 \times S^2, there are cells in dimensions 0,2,40, 2, 4, and the cellular chain complex is 0→Z→0→Z⊕Z→0→Z0 \to \mathbb {Z} \to 0 \to \mathbb {Z} \oplus \mathbb {Z} \to 0 \to \mathbb {Z} in low dimensions 00--55. Since all boundary maps are 00, the homology groups are the chain groups, giving

Hkc(S2×S2)≃{Zk=0,4,Z⊕Zk=2,0otherwise. H_k^c(S^2 \times S^2) \simeq \begin{cases} \mathbb {Z} & k = 0, 4, \\ \mathbb {Z} \oplus \mathbb {Z} & k = 2, \\ 0 & \text{otherwise}. \end{cases}

Show, by a similar computation with the cellular chain complex of CPn\mathbb {CP}^n, that

Hkc(CPn)={Z0≤k≤2n, k even,0otherwise. H_k^c(\mathbb {CP}^n) = \begin{cases} \mathbb {Z} & 0 \leq k \leq 2n, \, k \text{ even}, \\ 0 & \text{otherwise}. \end{cases}
?
Problem 6.10.13

Compute the cellular homology of S1×S2S^1 \times S^2, using the cellular chain complex (Ckc(X),∂kc)(C^c_k(X), \partial_k^c) of Exercise 6.10.11.

?
Remark.
?

Worked model (the cellular homology of S1×RP2S^1 \times \mathbb {RP}^2): the chain complex is C3c=Z(x3)→C2c=Z(x12)⊕Z(x22)→C1c=Z(x11)⊕Z(x21)→C0c=Z→0C_3^c = \mathbb {Z}(x^3) \to C_2^c = \mathbb {Z}(x_1^2) \oplus \mathbb {Z}(x_2^2) \to C_1^c = \mathbb {Z}(x_1^1) \oplus \mathbb {Z}(x_2^1) \to C_0^c = \mathbb {Z} \to 0, where x11x_1^1 is the generator from the 11-cell forming the circle S1×yS^1 \times y, x21x_2^1 is the generator from the 11-cell forming the circle x×S1⊂RP2x \times S^1 \subset \mathbb {RP}^2, x12x_1^2 is the generator from the 22-cell forming the torus, x22x_2^2 is the generator from the 22-cell used in forming x×RP2x \times \mathbb {RP}^2, and x3x^3 is the generator for the 33-cell. Here ∂1c=0\partial_1^c = 0, and from the methods for two-dimensional complexes, ∂2c(x12)=0\partial_2^c(x_1^2) = 0, ∂2c(x22)=2x21\partial_2^c(x_2^2) = 2x_2^1. To compute ∂3c\partial_3^c, look at the attaching map of the 33-cell, f:S2→X2f : S^2 \to X^2; writing S2=∂(D1×D2)=S0×D2∪D1×S1S^2 = \partial (D^1 \times D^2) = S^0 \times D^2 \cup D^1 \times S^1, this map sends each copy of {±1}×D2\left\{ \pm 1\right\} \times D^2 via the characteristic map of the 22-cell in RP2\mathbb {RP}^2, and sends D1×S1D^1 \times S^1 via the characteristic map of the first circle crossed with the attaching map of the 22-cell in RP2\mathbb {RP}^2 (a degree-22 map). The two copies of D2D^2 are mapped the same way but inherit different orientations in S2S^2, so they contribute 00 to ∂3(x3)\partial_3(x^3) (equivalently, they have opposite local degrees at the two points in the inverse image of a point); D1×S1D^1 \times S^1 maps with a degree-22 map to the torus, giving ∂3c(x3)=2x12\partial_3^c(x^3) = 2x_1^2. Thus ker⁡(∂3)=0\ker (\partial_3) = 0, so H3c(S1×RP2)=0H_3^c(S^1 \times \mathbb {RP}^2) = 0; ker⁡(∂2c)\ker (\partial_2^c) is generated by x12x_1^2 and im(∂3)\mathrm{im}(\partial_3) by 2x122x_1^2, so H2(S1×RP2)≃Z2H_2(S^1 \times \mathbb {RP}^2) \simeq \mathbb {Z}_2; im(∂2c)\mathrm{im}(\partial_2^c) is generated by 2x212x_2^1, so H1(S1×RP2)=Z⊕Z2H_1(S^1 \times \mathbb {RP}^2) = \mathbb {Z} \oplus \mathbb {Z}_2; and H0(S1×RP2)=ZH_0(S^1 \times \mathbb {RP}^2) = \mathbb {Z}, all other homology vanishing.

Problem 6.10.14

Compute the cellular homology of RP2×RP2\mathbb {RP}^2 \times \mathbb {RP}^2, using the cellular chain complex (Ckc(X),∂kc)(C^c_k(X), \partial_k^c) of Exercise 6.10.11.

?
Problem 6.10.15

Compute the cellular homology of RPn\mathbb {RP}^n, using the cellular chain complex (Ckc(X),∂kc)(C^c_k(X), \partial_k^c) of Exercise 6.10.11.

?
Remark.
?

Worked model (the cellular homology of RP3\mathbb {RP}^3 and RP4\mathbb {RP}^4): the cellular chain complex of RP3\mathbb {RP}^3 is 0→Z→Z→Z→Z→00 \to \mathbb {Z} \to \mathbb {Z} \to \mathbb {Z} \to \mathbb {Z} \to 0. The 22-skeleton is RP2\mathbb {RP}^2, with ∂2c(x2)=2x1\partial_2^c(x^2) = 2x^1, ∂1c(x1)=0\partial_1^c(x^1) = 0; since im(∂3c)⊂ker⁡(∂2c)=0\mathrm{im}(\partial_3^c) \subset \ker (\partial_2^c) = 0, we must have ∂3c=0\partial_3^c = 0, giving

Hkc(RP3)={Zk=0,3,Z2k=1,0otherwise. H_k^c(\mathbb {RP}^3) = \begin{cases} \mathbb {Z} & k = 0, 3, \\ \mathbb {Z}_2 & k = 1, \\ 0 & \text{otherwise}. \end{cases}

This can also be seen using the fact (Exercise 6.9.15) that the degree of the map S2→RP2→RP2/S1=S2S^2 \to \mathbb {RP}^2 \to \mathbb {RP}^2/S^1 = S^2 is 1+(−1)3=01 + (-1)^3 = 0. For higher real projective spaces algebra alone no longer determines the top boundary map, but Exercise 6.9.15 still applies: the attaching map of the 44-cell of RP4\mathbb {RP}^4 is the standard double covering S3→RP3S^3 \to \mathbb {RP}^3, composed with collapsing RP2⊂RP3\mathbb {RP}^2 \subset \mathbb {RP}^3 to a point to get S3S^3; each hemisphere covers the 33-sphere exactly once, and the two hemisphere maps are related by f(−x)=f(x)f(-x) = f(x), i.e. f=fAf = fA with A(x)=−xA(x) = -x. Since A:S3→S3A : S^3 \to S^3 has degree 11, the local degree is the same at both points in the inverse image of a point, so the generator of S3S^3 is sent to twice the generator, giving ∂c(x4)=±2x3\partial^c(x^4) = \pm 2x^3. Hence H4(RP4)=0H_4(\mathbb {RP}^4) = 0, H3(RP4)=Z2H_3(\mathbb {RP}^4) = \mathbb {Z}_2, and since the 33-skeleton of RP4\mathbb {RP}^4 agrees with RP3\mathbb {RP}^3, the lower-dimensional homology agrees too, giving

Hk(RP4)={Zk=0,Z2k=1,3,0otherwise. H_k(\mathbb {RP}^4) = \begin{cases} \mathbb {Z} & k = 0, \\ \mathbb {Z}_2 & k = 1, 3, \\ 0 & \text{otherwise}. \end{cases}