Section 6.16

Proof of homotopy invariance of homology

[12/100%]
Show
LaTeX
Problem 6.16.1

Suppose that f,g:X→Yf, g : X \to Y are homotopic maps, via a homotopy FF with F(x,1)=g(x)F(x,1) = g(x), F(x,0)=f(x)F(x,0) = f(x). To show that f#,g#f_\# , g_\# are chain homotopic (and hence f∗=g∗f_* = g_* in homology), we first look at dimension 00. Since these maps are linear, it suffices to look at a singular 00-simplex σ\sigma; the induced maps are f#(σ)=fσf_\# (\sigma ) = f\sigma and g#(σ)=gσg_\# (\sigma ) = g\sigma. Suppose σ(0)=p\sigma (0) = p. Let γ:[0,1]=Δ1→Y\gamma : [0,1] = \Delta_1 \to Y be the singular 11-simplex defined by the composition [0,1]→{p}×[0,1]→FY[0,1] \to \left\{ p\right\} \times [0,1] \xrightarrow {F} Y.

Define H0(σ)=γH_0(\sigma ) = \gamma and show that H0H_0 provides a chain homotopy between fσf\sigma and gσg\sigma.

?
Problem 6.16.2

We now look at dimension 11, needing to define H1:S1(X)→S2(Y)H_1 : S_1(X) \to S_2(Y) so that, for a singular 11-simplex σ:I→X\sigma : I \to X, we have the formula g#σ−f#σ=∂2H1(σ)+H0∂1(σ)g_\# \sigma - f_\# \sigma = \partial_2 H_1(\sigma ) + H_0 \partial_1(\sigma ) (with H0H_0 as in Exercise 6.16.1).

We first use the homotopy FF to define a map K=F(σ×id):I×I→YK = F(\sigma \times \mathrm{id}) : I \times I \to Y, where σ×id:I×I→X×I\sigma \times \mathrm{id} : I \times I \to X \times I. Note that Ki0=fσKi_0 = f\sigma, Ki1=gσKi_1 = g\sigma, where i0(s)=(s,0)i_0(s) = (s,0), i1(s)=(s,1)i_1(s) = (s,1). If τ1(0)=σ(1)\tau_1(0) = \sigma (1), τ0(0)=σ(0)\tau_0(0) = \sigma (0) are the singular 00-simplices coming from restricting σ\sigma to its end points, then ∂1σ=τ1−τ0\partial_1 \sigma = \tau_1 - \tau_0, and H0(τ1)(y)=K(1,y)H_0(\tau_1)(y) = K(1,y), H0(τ0)(y)=K(0,y)H_0(\tau_0)(y) = K(0,y).

Constructing H_1.Constructing H_1.

Motivated by Figure 6.6, define two singular 22-simplices α,β\alpha , \beta in I×I=Δ1×II \times I = \Delta_1 \times I with positive orientation on α\alpha and negative orientation on β\beta as depicted via the arrows on their boundaries, and define H1(σ)=Kβ−Kα=K#(β−α)H_1(\sigma ) = K\beta - K\alpha = K_\# (\beta - \alpha ).

Show that with this definition of H1H_1, we have the required formula g#(σ)−f#(σ)=∂2H1(σ)+H0∂1(σ)g_\# (\sigma ) - f_\# (\sigma ) = \partial_2 H_1(\sigma ) + H_0 \partial_1(\sigma ).

?
Problem 6.16.3

Use Exercises 6.16.1 and 6.16.2 to show that if f,g:X→Yf, g : X \to Y are homotopic maps, then the induced maps f#,g#f_\# , g_\# are chain homotopic in dimensions 00 and 11. Deduce that f∗=g∗f_* = g_* in homology.

?
Problem 6.16.4

To extend H0,H1H_0, H_1 to higher-dimensional chains, note that the homotopy FF is used only at the last step and can be separated from the definition: writing ijX:X→X×Ii_j^X : X \to X \times I for ijX(x)=(x,j)i_j^X(x) = (x,j), we have g=Fi1Xg = F i_1^X, f=Fi0Xf = F i_0^X. If we label the maps Hk=HkX,Y:Sk(X)→Sk+1(Y)H_k = H_k^{X,Y} : S_k(X) \to S_{k+1}(Y) already defined (k=0,1k=0,1, Exercises 6.16.1-6.16.2), there are corresponding maps HkX:Sk(X)→Sk+1(X×I)H_k^X : S_k(X) \to S_{k+1}(X \times I) (using the identity homotopy I(x,t)=(x,t)I(x,t) = (x,t) connecting i0Xi_0^X and i1Xi_1^X) so that HkX,YH_k^{X,Y} factors as F#HkXF_\# H_k^X. The term (∂2H1X,Y+H0X,Y∂1)(σ)(\partial_2 H_1^{X,Y} + H_0^{X,Y}\partial_1)(\sigma ) becomes F#(∂H1X+H0X∂1)(σ)F_\# (\partial H_1^X + H_0^X \partial_1)(\sigma ), and since g#=F#(i1X)#g_\# = F_\# (i_1^X)_\#, f#=F#(i0X)#f_\# = F_\# (i_0^X)_\#, we can rewrite g#(σ)−f#(σ)=F#((i1X)#(σ)−(i0X)#(σ))g_\# (\sigma ) - f_\# (\sigma ) = F_\# ((i_1^X)_\# (\sigma ) - (i_0^X)_\# (\sigma )). So the formula follows from the special case

(∂2H1X+H0X∂1)(σ)=((i1X)#−(i0X)#)(σ) (\partial _2 H_1^X + H_0^X \partial _1)(\sigma ) = ((i_1^X)_\# - (i_0^X)_\# )(\sigma )

by composing with F#F_\#.

Let ijΔ1i_j^{\Delta_1} be the corresponding inclusions for Δ1\Delta_1. Using the identity map ι1∈S1(Δ1)\iota_1 \in S_1(\Delta_1), there is a formula (∂2H1Δ1+H0Δ1∂1)(ι1)=((i1X)#−(i0X)#)(ι1)(\partial_2 H_1^{\Delta_1} + H_0^{\Delta_1} \partial_1)(\iota_1) = ((i_1^X)_\# - (i_0^X)_\# )(\iota_1); applying (σ×id)#(\sigma \times \mathrm{id})_\# to both sides, using that H1X(σ)=(σ×id)#(β−α)=(σ×id)#H1Δ1(ι1)H_1^X(\sigma ) = (\sigma \times \mathrm{id})_\# (\beta - \alpha ) = (\sigma \times \mathrm{id})_\# H_1^{\Delta_1}(\iota_1) (Exercise 6.16.2) and that (σ×id)#(\sigma \times \mathrm{id})_\# is a chain map commuting with ∂2\partial_2, the left-hand side becomes (σ×id)#((∂2H1Δ1+H0Δ1∂1)(ι))(\sigma \times \mathrm{id})_\# ((\partial_2 H_1^{\Delta_1} + H_0^{\Delta_1}\partial_1)(\iota )). On the right-hand side, using σ=σ#(ι1)\sigma = \sigma_\# (\iota_1) and ijXσ=(σ×id)ijΔ1i_j^X \sigma = (\sigma \times \mathrm{id}) i_j^{\Delta_1}, we get (ijX)#(σ)=(σ×id)#(ι1)(i_j^X)_\# (\sigma ) = (\sigma \times \mathrm{id})_\# (\iota_1), so the right-hand side becomes (σ×id)#((i1Δ1)#(ι1)−(i0Δ1)#(ι1))(\sigma \times \mathrm{id})_\# ((i_1^{\Delta_1})_\# (\iota_1) - (i_0^{\Delta_1})_\# (\iota_1)).

Verify by an argument analogous to the one given above that

(∂2XH1X+H0X∂1X)(σ)=(i1X)#(σ)−(i0X)#(σ) (\partial _2^X H_1^X + H_0^X \partial _1^X)(\sigma ) = (i_1^X)_\# (\sigma ) - (i_0^X)_\# (\sigma )

follows from the similar formula

(∂2Δ1H1Δ1+H0Δ1∂1Δ1)([e0,e1])=(i1Δ1)#([e0,e1])−(i0Δ1)#([e0,e1]). (\partial _2^{\Delta _1} H_1^{\Delta _1} + H_0^{\Delta _1} \partial _1^{\Delta _1})([e_0,e_1]) = (i_1^{\Delta _1})_\# ([e_0,e_1]) - (i_0^{\Delta _1})_\# ([e_0,e_1]).
?
Problem 6.16.5

Show that, if f:X→Yf : X \to Y is a continuous map, then HiY(f#(σ))=(f×id)#(HiX(σ))H_i^Y(f_\# (\sigma )) = (f \times \mathrm{id})_\# (H_i^X(\sigma )), i=0,1i = 0, 1 (with HiXH_i^X as in Exercises 6.16.1, 6.16.2, and 6.16.4).

?
Problem 6.16.6

We now extend HiX(σ)H_i^X(\sigma ) to i≥2i \geq 2: first define it for X=ΔiX = \Delta_i and σ=[e0,…,ei]\sigma = [e_0, \ldots , e_i], showing it satisfies ∂i+1Hi([e0,…,ei])+Hi−1∂([e0,…,ei])=(i1)#([e0,…,ei])−(i0)#([e0,…,ei])\partial_{i+1} H_i([e_0, \ldots , e_i]) + H_{i-1}\partial ([e_0, \ldots , e_i]) = (i_1)_\# ([e_0,\ldots ,e_i]) - (i_0)_\# ([e_0,\ldots ,e_i]), then extend via HiX(σ)=(σ×id)#HiΔi([e0,…,ei])H_i^X(\sigma ) = (\sigma \times \mathrm{id})_\# H_i^{\Delta_i}([e_0,\ldots ,e_i]) (satisfying the analogous formula for XX and the same naturality as Exercise 6.16.5). We extend HiH_i inductively, having already defined HjH_j for j<ij < i on Δj\Delta_j and extended it over any XX via HjX(σ)=(σ×id)#HjΔj([e0,…,ej])H_j^X(\sigma ) = (\sigma \times \mathrm{id})_\# H_j^{\Delta_j}([e_0,\ldots ,e_j]); in particular Hi−1H_{i-1} is already defined on the face maps Fk:Δi−1→ΔiF_k : \Delta_{i-1} \to \Delta_i, giving Hi−1Δi∂i([e0,…,ei])H_{i-1}^{\Delta_i} \partial_i([e_0,\ldots ,e_i]). Write vj=(ej,0)v_j = (e_j, 0) and wj=(ej,1)w_j = (e_j, 1) (points of Δi×I\Delta_i \times I). The general definition of Hi([e0,…,ei])H_i([e_0,\ldots ,e_i]) in any dimension is

Hi([e0,…,ei])=∑j=0i(−1)j[v0,…,vj,wj,…,wi]. H_i([e_0, \ldots , e_i]) = \sum _{j=0}^i (-1)^j [v_0, \ldots , v_j, w_j, \ldots , w_i].
?
(a)

Check that this corresponds to the definition given for i=0,1i = 0, 1 above (Exercises 6.16.1, 6.16.2).

(b)

Check that our naturality property gives

Hi−1([e0,…,ek^,…,ei])={∑j=0i−1(−1)j[v0,…,vj,wj,…,wk^,…,wi]if k>j,∑j=0i−1(−1)j[v0,…,vk^,…,vj+1,wj+1,…,wi]if k≤j. H_{i-1}([e_0, \ldots , \widehat{e_k}, \ldots , e_i]) = \begin{cases} \sum _{j=0}^{i-1} (-1)^j [v_0, \ldots , v_j, w_j, \ldots , \widehat{w_k}, \ldots , w_i] & \text{if } k > j, \\ \sum _{j=0}^{i-1} (-1)^j [v_0, \ldots , \widehat{v_k}, \ldots , v_{j+1}, w_{j+1}, \ldots , w_i] & \text{if } k \leq j. \end{cases}
Problem 6.16.7

Show that

∂i+1Hi([e0,…,ei])=∑j,k=0,j>ki(−1)j+k[v0,…,vk^,…,vj,wj,…,wi] \partial _{i+1} H_i([e_0,\ldots ,e_i]) = \sum _{j,k=0, j>k}^i (-1)^{j+k} [v_0, \ldots , \widehat{v_k}, \ldots , v_j, w_j, \ldots , w_i] +∑j=0i[v0,…,vj−1,wj,…,wi]−∑j=0i[v0,…,vj,wj+1,…,wi] + \sum _{j=0}^i [v_0, \ldots , v_{j-1}, w_j, \ldots , w_i] - \sum _{j=0}^i [v_0, \ldots , v_j, w_{j+1}, \ldots , w_i] +∑j,k=0,j<ki(−1)j+k+1[v0,…,vj,wj,…,wk^,…,wi]. + \sum _{j,k=0,j<k}^i (-1)^{j+k+1}[v_0, \ldots , v_j, w_j, \ldots , \widehat{w_k}, \ldots , w_i].
?
Remark.
?

Using Exercise 6.16.6(b),

Hi−1∂([e0,…,ei])=Hi−1(∑k=0i(−1)k[e0,…,ek^,…,ei]) H_{i-1}\partial ([e_0,\ldots ,e_i]) = H_{i-1}\left(\sum _{k=0}^i (-1)^k [e_0,\ldots ,\widehat{e_k},\ldots ,e_i]\right) =∑k=0i(∑j=0,j<ki−1(−1)j+k[v0,…,vj,wj,…,wk^,…,wi]+∑j=0,j≥ki−1(−1)j+k[v0,…,vk^,…,vj+1,wj+1,…,wi]) = \sum _{k=0}^i\left(\sum _{j=0,j<k}^{i-1}(-1)^{j+k}[v_0,\ldots ,v_j,w_j,\ldots ,\widehat{w_k},\ldots ,w_i] + \sum _{j=0,j\geq k}^{i-1}(-1)^{j+k}[v_0,\ldots ,\widehat{v_k},\ldots ,v_{j+1},w_{j+1},\ldots ,w_i]\right) =∑j,k=0,j<ki(−1)k+j[v0,…,vj,wj,…,wk^,…,wi]+∑j,k=0,j>ki(−1)k+j+1[v0,…,vk^,…,vj,wj,…,wi]. = \sum _{j,k=0,j<k}^i (-1)^{k+j}[v_0,\ldots ,v_j,w_j,\ldots ,\widehat{w_k},\ldots ,w_i] + \sum _{j,k=0,j>k}^i (-1)^{k+j+1}[v_0,\ldots ,\widehat{v_k},\ldots ,v_j,w_j,\ldots ,w_i].

Now compute ∂i+1Hi([e0,…,ei])\partial_{i+1} H_i([e_0,\ldots ,e_i]) directly from the formula of Exercise 6.16.6, and compare with this expression for Hi−1∂H_{i-1}\partial.

Problem 6.16.8

Show that

∑j=0i[v0,…,vj−1,wj,…,wi]−∑j=0i[v0,…,vj,wj+1,…,wi]=(i1)#([e0,…,ei])−(i0)#([e0,…,ei]) \sum _{j=0}^i [v_0, \ldots , v_{j-1}, w_j, \ldots , w_i] - \sum _{j=0}^i [v_0, \ldots , v_j, w_{j+1}, \ldots , w_i] = (i_1)_\# ([e_0,\ldots ,e_i]) - (i_0)_\# ([e_0,\ldots ,e_i])

(notation as in Exercises 6.16.6-6.16.7).

?
Problem 6.16.9

Combine the calculations of Exercises 6.16.7 and 6.16.8 to show that

(∂i+1Hi+Hi−1∂i)([e0,…,ei])=((i1)#−(i0)#)([e0,…,ei]). (\partial _{i+1} H_i + H_{i-1}\partial _i)([e_0,\ldots ,e_i]) = ((i_1)_\# - (i_0)_\# )([e_0,\ldots ,e_i]).
?
Problem 6.16.10

Show that if we define HiX(σ)=(σ×id)#Hi([e0,…,ei])H_i^X(\sigma ) = (\sigma \times \mathrm{id})_\# H_i([e_0,\ldots ,e_i]) (with HiH_i as in Exercises 6.16.6-6.16.9) and extend linearly to chains, then this satisfies the condition

∂i+1XHiX+Hi−1X∂iX=(i1X)#−(i0X)#. \partial _{i+1}^X H_i^X + H_{i-1}^X \partial _i^X = (i_1^X)_\# - (i_0^X)_\# .

Moreover, show that this satisfies the naturality property that when f:X→Yf : X \to Y, then (f×id)#HiX=HiYf#(f \times \mathrm{id})_\# H_i^X = H_i^Y f_\#.

?
Problem 6.16.11

The formula of Exercise 6.16.10 shows that (i0X)#(i_0^X)_\# and (i1X)#(i_1^X)_\# are chain-homotopic chain maps, with chain homotopy HiXH_i^X; hence they induce the same map in homology, (i0X)∗=(i1X)∗(i_0^X)_* = (i_1^X)_*. Applying this to homotopic maps f0∼f1f_0 \sim f_1 via a homotopy F:X×I→YF : X \times I \to Y: (fj)∗=F∗(ij)∗(f_j)_* = F_*(i_j)_*, so (f0)∗=(f1)∗(f_0)_* = (f_1)_*, proving the homotopy property in the absolute case.

Prove that if f:X→Yf : X \to Y is a homotopy equivalence, then f∗:Hk(X)→Hk(Y)f_* : H_k(X) \to H_k(Y) is an isomorphism.

?
Problem 6.16.12
?
(a)

Adapt Exercises 6.16.1--6.16.11 to prove the relative case: if f,g:(X,A)→(Y,B)f, g : (X,A) \to (Y,B) are homotopic, then f∗=g∗f_* = g_*.

(b)

Show that if f:(X,A)→(Y,B)f : (X,A) \to (Y,B) is a homotopy equivalence, then f∗f_* is an isomorphism.