Assignment 1 Due Friday, January 13
Assignment 1 Due Friday, January 13
Assignment 1 Due Friday, January 13
This is clearly continuous and equals h(p, 0) = f (p) on the boundary of the disk.
Conversely, suppose g : D → X exists and equals f on the boundary of the disk. Define
homotopy h : S 1 × I → X between f and the constant map by
1
If γ : [0, 1] → X is a path from x to x, let γ ? σ be the path which traces γ from x to x
and then traces σ from x to y. I claim this induces a map P (x, x) → P (x, y). Indeed if
γ1 and γ2 are homotopic with fixed endpoints, it is easy to see that γ1 ? σ and γ2 ? σ are
homotopic.
If γ : [0, 1] → X is a path from x to y, let γ ? σ −1 be the path which traces γ from x to
y and then σ −1 back from y to x. I claim this induces a map P (x, y) → P (x, x). Indeed
if γ1 and γ2 are homotopic paths from x to y with fixed endpoints, it is easy to see that
γ1 ? σ −1 and γ2 ? σ −1 are homotopic.
To finish the argument, it is enough to prove that the maps P (x, x) → P (x, y) → P (x, x)
and P (x, y) → P (x, x) → P (x, y) are both identity maps, for then the individual map
P (x, x) → P (x, y) must be one-to-one and onto.
Consider the first map. It starts with a path γ from x to x, and follows this with the path
σ from x to y, and then with the path σ −1 from y back to x. This is homotopic to the path
which starts with γ from x to x and then follows σ part way from x to y and then follows
σ −1 part way from this point back to x. At the start of the homotopy, we follow σ all the
way from x to y and back. In the middle, we following σ part way and then go back. At
the end, we just stay at x and don’t move along σ at all. So at the end of the homotopy
we just have γ. Hence the map P (x, x) → P (x, y) → P (x, x) maps γ to a path homotopic
to γ and thus equivalent to γ in the set of equivalence classes. Said another way, this map
is the identity map.
Similarly P (x, y) → P (x, x) → P (x, y) is the identity map.
13.4c This wasn’t quite fair, since we skipped the earlier section where p ? q was defined.
By definition, p ? q is the path which first traces p and then traces q. In detail
if 0 ≤ t ≤ 21
p(2t)
(p ? q)(t) =
q(2t − 1) if 21 ≤ t ≤ 1
In the exercise we must assume 0 < s < 1. So in words the exercise says that we get the
same curve up to homotopy by tracing p very fast and q very slow, or p and q at the same
speed, or p very slow and q very fast. We prove this by writing down a homotopy. First
we draw a schematic picture of the idea.
Here are the mathematical details. Since t is being used for the homotopy, let us use u
for the parameter of the paths. In the picture, we divide between the two paths at s on
the right. Let us divide between the two paths at s(t) in the middle. On the left we want
s(0) = 21 and we want s(1) = s. By linearity, s(t) = 12 (1 − t) + st. So the homotopy at t
2
should just be the formula on the right, but replacing s by s(t). Thus
p t if 0 ≤ t ≤ s(t)
h(u, t) = s(t)
p t−s(t) if s(t) ≤ t ≤ 1
1−s(t)
3
square. Given a point p not at the center, draw a line from the center through p and on to
the boundary, and map p to this boundary point. This map is homotopy to the identity
map, since we can push p out gradually to the boundary. But the boundary of the square
is a figure eight, since the two vertical sides map to one circle in the figure eight and the
two horizontal sides map to the other circle in the figure eight.
13.10e Let D be the closed unit disk in the plane, and let U ⊂ D be the open unit disk.
Then U is not a retract of D because a continuous map D → D which is the identity on U
must be the identity on all of D and thus cannot map D to U.
But U is a weak retract; we can map D to the closed disk of radius 12 inside U by shrinking
by a factor of 12 . When we restrict this map to U we obtain the map which shrinks the
open disk by 12 and this map is clearly homotopy to the identity map U → U.
13.10f We’ll give a tricky example. Let X be the comb space illustrated below, and let
A ⊂ X be the vertical line at the left side. Map X to A by mapping (x, y) to (0, y),
squashing the space X horizonally to the line segment on the left. This is a retract because
it leaves points in A fixed.
I claim this map is a deformation retract, that is, the map X → A → X is homotopic
to the identity map. Indeed this map pushes X to the vertical line A, and it is clearly
homotopic to the constant map which sends all of X to (0, 0) because we can just push the
vertical line on the left gradually down to (0, 0). But the identity map from X → X is also
homotopic to this constant map, because we can push the comb gradually down until all
of its points are along the x-axis, and then we can gradually push this x-axis over to (0, 0).
Since homotopy is an equivalent relation on maps from X to X, and since the identity
map X → X and the retract map X → A → X are both homotopic to the constant map
sending everything to (0, 0), they are homotopic to each other.
But the retraction is not a strong deformation retract, because we cannot keep A fixed
as we deform the identity map to the map X → A. Indeed, a homotopy must gradually
deform the point (x, y) to the final point (0, y), and the only way to get there is to push
4
the point down to the x-axis, and then over to the origin, and then up to (0, y). But if A
is fixed, lines very close to A must be almost fixed, and certainly cannot make this long
journey down and around.
We can make this final argument rigorous as follows. Suppose h : X ×I → X is a homotopy
which fixes points in A. Let U be the open subset of X consisting of points (x, y) with 21 < y.
Let p be the point (0, 1) ∈ X. Since h(p, t) = p, there is an open neighborhood Vt of p and
an open interval It about t ∈ I such that h maps Vt × It into U.
The intervals It form an open cover of I, which is compact. Form a finite subcover, and let
V be the intersection of the corresponding It . Then h maps V × I into U. But V contains
points in adjacent spikes of the comb, and so these points are not mapped down to the
x-axis by h and so cannot be sent to the vertical interval A on the left.