Mathematics
Mathematics
Mathematics
Problem 1
Solution courtesy of Tara Aida
We start by stating the homotopy lifting lemma which we’ve already gone over in class.
Now, we want to generalize this lemma to all continuous maps into B (not just homotopy maps from
I ⇥ I ! B).
Here’s the idea. Suppose we have a continuous map f mapping into some space B. We can imagine
that there exists a covering map p : E ! B. A natural question one may want to ask is whether we can lift
the map f up to the space E; when we can, is the lift unique? (yes). This lemma, gives us sufficient and
necessary conditions for when this is true.
In particular, recall that the induced group homomorphisms f⇤ and p⇤ both map into ⇡1 (B, b0 ). The lemma
tells us that we can uniquely lift f when f⇤ (⇡1 (Y, y0 )) (which is a group) is a subgroup of p⇤ (⇡1 (E, e0 )).
Okay, now that we understand the statement, let’s prove the lemma.
Proof. We start with the easier direction. Suppose that there exists a unique lift of f , f˜. Since p f˜ = f
and f˜ : Y ! E:
f⇤ (⇡1 (Y, y0 )) = p⇤ (f˜⇤ (⇡1 (Y, y0 ))) ⇢ p⇤ (⇡1 (E, e0 ))
Now, we prove the opposite direction. We have three spaces to keep track of: Y, B, E. Given y1 2 Y choose
a path ↵ that maps between y0 and y1 . Applying f , we get a path in B: f ↵, which starts at b0 and ends
at f (y1 ). Finally, we may lift this path up to one in E beginning at e0 . Call the lift and define f˜(y1 ) = (1).
To complete the proof, we now need to prove now that f˜ is unique, well-defined and continuous:
1. Well-defined. Suppose ↵, are two paths in Y from y0 to y1 . We want to show that the lift of f ↵
and the lift of f starting at e0 also have the same endpoints in E.
1
Let refer to the lift of f ↵. Let refer to the lift of f (this begins at the endpoint of ).
Then, ⇤ is a lifting of the loop f (↵ ⇤ ). Since f⇤ (⇡1 (Y, y0 )) ⇢ p⇤ (⇡1 (E, e0 )) we may conclude that
the homotopy class of f (↵ ⇤ ) is in the image of p⇤ . Now, recall Theorem 54.6:
Theorem 54.6. Let p : E ! B be a covering map. Let p(e0 ) = b0 . Let H = p⇤ (⇡1 (E, e0 )).
Then, if f is a loop in B based at b0 , then [f ] 2 H IFF f lifts to a loop in E based at e0 .
Thus, we may conclude that ⇤ is not just any path in E, but a loop! This means that we
have that the lifts of f (this is ) and f ↵ (this is ), both starting at e0 , also must have the same
end points, as desired.
2. Continuity. We want to prove that f˜ is continuous at y1 . Let N be a neighborhood of f˜(y1 ). To
prove continuity at this point is to find W a neighborhood of y1 s.t. f˜(W ) ⇢ N .
f is continuous at y1 and Y is locally path connected. Thus, there exists a path-connected neigh-
borhood W of y1 s.t. f (W ) ⇢ U . Claim: this is the W we’re looking for (aka, f˜(W ) ⇢ N ).
Using the fact that f˜ is well-defined, we will show that given any y 2 W , f˜(y) is in V0 . Suppose
is a map contained in W from y1 to y 2 W . Then, if we take the lift of f (↵ ⇤ ) to a path starting
at e0 , it will have f˜(y) as endpoint. Okay, next we note that f is in U so = p0 1 f is a lift of
f ˜
to E which starts at f (y1 ) (here, p0 is simply p restricted to V0 , a homeomorphism). Further, we
have that is a lift of ↵ starting at e0 and ending at f˜(y1 ). Thus, we may concatenate to get: ⇤ a
lift of f (↵ ⇤ ) which must start at e0 and end at (1) 2 V0 . Since we noted that the lift of f (↵ ⇤ )
must end at f˜(y), we may conclude that f˜(y) is in V0 as desired.
3. Unique. Suppose f˜, as defined in the lemma, exists. We will show that it must be unique by showing
the value of f˜ on any y1 2 Y is unique. Suppose we have a path ↵ in Y from y0 to y1 . We could lift
it to a path in E (with starting point e0 ) by first composing it with f to get a path in B and then
lifting up to E.
Cool! So in addition to being a nice generalization of the homotopy lemma we already saw, this lemma is
also useful in constructing the theory of two covering maps being equivalent. See section 79 of Munkres for
more on this. See problems #1 and #2 on this pset for simple applications of the general lifting lemma,
where it is used to show that every continuous map f : S n ! S 1 is nullhomotopic (for n > 1) and every
continuous map f : P 2 ! S 1 is nullhomotopic.
2
Problem 2
Draw two non-trivial finite and one non-trivial infinite covering spaces of the following graph:
3
Solutions here are courtesy of Je↵ Cai.
Munkres p. 348, #5
Proof. The path f is pictured below.
1
1 Page 483 Problem 1
Show that if n > 1, every continuous map f : S n ! S 1 is nulhomotopic.
Recall the following lifting property of covering spaces: if f : Z ! X is a continuous map, Z is path-connected
and locally path-connected, p : Y ! Z is a covering map, f (z) = p(y), and f § (º1 (Z , z)) Ω p § (º1 (Y , y)), then
there exists a unique map g : Z ! Y with g (z) = y such that the following diagram commutes
X
g
p
Z Y
f
Now let f be the map in question and p : R ! S 1 be the standard covering, then since S n is path-connected
and locally path-connected, and º1 (R) = º1 (S n ) = 0 (since n > 1), we have a unique lift g as above. How-
ever, since R is contractible, idR is homotopic to c g (z) where c g (z) : R ! R is the constant map at 0, and thus
f = p ± g = p ± idR ± g is homotopic to p ± c g (z) ± g = c f (z) , hence f is nulhomotopic. ⇤
We remark that RP 2 is path connected because it is the continuous image of path connected space, as we see
from the covering map p : S 2 ! RP 2 . It is locally path connected because it is locally homeomorphic to R2 as
it is a topological 2-manifold (this is discussed in Munkres), and R2 is locally path-connected. Now since f § is
a map º1 (RP 2 ) ª
= Z/2 ! º1 (S 1 ) ª
= Z, and the only homomorphisms between these groups are trivial, we have
§ 2
f (º1 (RP )) = {0} Ω {0} = º§ (º1 (R)), as º1 (R) = 0. We may therefore apply the lifting lemma, so there exists a
map g such that
R
g
º
f
RP 2 S1
commutes. Hence by the same argument as in the above problem, f is nulhomotopic. ⇤
2.2
Find a continuous map of the torus into S 1 that is not nulhomotopic.
Let f : S 1 £ S 1 be projection onto the left coordinate, and assume that f is nulhomotopic. Then, if g : S 1 !
S 1 £ S 1 is given by µ 7! (µ, 0), f ± g = idS 1 , and thus idS 1 is nulhomotopic, but this is a contradiction. ⇤
1
on C (X ). C (X ) is not locally simply connected because every open subset U contained in the neighbor-
hood V := B 1/2 (0, 0, 0) \ C (X ) is homeomorphic to X £ [0, t ) for some t < 1/2. Thus º1 (U ) = º1 (X £ [0, t )) =
º1 (X ) £ º1 ([0, t )) = º1 (X ) since [0, t ) is contractible. But º1 (X ) is nontrivial as X is a wedge of circles. ⇤