Академический Документы
Профессиональный Документы
Культура Документы
Metric spaces
1. Let p be a prime number, and d : Z Z [0, +) be a function defined by
dp (x, y) = p max{mN
: pm |xy}
Prove that dp is a metric on Z and that dp (x, y) max{dp (x, z), dp (z, y)} for every x, y, z Z.
Proof. The symmetry condition is trivial. By p > 0 and m N, we have pm > 0 for every m N
with 0 as the limit. Indeed, d(x, y) can only be zero, if pm |x y for all natural numbers m. This
can only happen if x y = 0, thus d(x, y) = 0 x = y. For the triangle inequality condition,
assume wlog that mx,z := logp (d(x, z)) is not larger than my,z (i.e. d(x, z) d(y, z)). Then pmx,z
divides both x z and z y and therefore also x z + z y = x y, which means that mx,y mx,z
or equivalently d(x, y) d(x, z) = max{d(x, z), d(y, z)}, so in particular d(x, y) d(x, z) + d(y, z)
holds.
2. Let X 6= be a set and F the family of all finite subsets of X. Prove that the function d : F F
[0, +) defined by
d(A, B) = #((A\B) (B\A))
is a metric on X.
Proof. d(A, B) = 0 implies that both A\B and B\A must be empty, thus the sets A and B must
coincide. Conversely d(A, A) = 0. Looking at the Venn diagram for sets A, B, C, it can be seen
that AC BC contains AB (where denotes the symmetric difference).
3. Let (X, d) be a metric space and F X be a finite subset. Prove that F is closed in X.
Proof. Let x X \ F and define r := min{d(x, y) : y F }. As F is finite the minimum exists. The
open ball B(x, r) around x does not contain any point of F , thus X \ F is open and F closed.
4. Let (X, d) be a metric space and 6= Y X be a subset. The distance of a point x X from the
subset Y is a function X [0, +) defined by:
d(x, Y ) = inf{d(x, y) : y Y }.
Verify that the distance function is well defined. Prove that
Y = {x X : d(x, Y ) = 0}.
Proof. The fact that d is bounded from below ensures that for each x X the infimum exists. If the
infimum exists, it is unique, so the function is well-defined. The claim directly follows from Lemma
1.5 in the lecture: Take an x
/ Y and r > 0 such that Br (x) Y = . Then all points outside the
ball, in particular the points of Y , must have distance at least r, so x
/ {x X : d(x, Y ) = 0}. Now
let x
/ {x X : d(x, Y ) = 0}. Then by the triangle inequality any ball B(x, r) with 0 < r < d(x, Y )
will be an open neighbourhood of x that doesnt intersect {x X : d(x, Y ) = 0}. Hence it is closed
and because it contains Y , it is a superset of Y .
5. Let X be the set of all sequences x : N R converging to zero. Prove that the function d : X X
[0, +) defined by
d(xn , yn ) = sup |xn yn |
nN
determine, the distance d(y, X \ U ) differs from Dx := d(x, X \ U ) by at most rx . Without loss of
generality, assume that Dx > rx , otherwise scale rx accordingly. Now
d(y,X\U )d(x,X\U )
rx
1
1
d(x,X\U ) d(y,X\U ) = d(x,X\U )d(y,X\U ) Dx (Dx rx ) =: Nx
Thus dU (x, y) d(x, y) + Nx < rx + Nx . Setting rx such that rx + Nx < Rx , ensures that
Bd (x, rx ) BdU (x, Rx ) and therefore lies in W .
9. Let S 6= be a set and (X, d) a complete metric space. Prove that the metric space of all bounded
functions B(S, X) equipped with the metric D(f, g) = supsS d(f (s), g(s)) is also complete.
Proof. Take a Cauchy sequence (fn ) B(S, X) in the sup-metric. For every > 0 there exists an
n N so that d(fn (s), fm (s)) D(fn , fm ) < 3 for all n, m n . From the definition of D we
know that (fn (s)) is a Cauchy sequence in X for every s S. Since X is complete, for every s S
there exists a f (s) X so that fn (s) f (s). Define the function f : S X so that s 7 f (s)
(well defined due to the uniqueness of the limit). To check that (fn ) converges to f , fix s S. Now
there exists ns N so that d(fn (s), f (s)) < 3 for all n ns . Define N := max{n , ns }. Then for
every n n we have
d(fn (s), f (s)) d(fn (s), fN (s)) + d(fN (s), f (s)) <
2
+ = .
3 3
3
In other words, for every s S the difference d(fn (s), f (s)) is bounded by 2
3 . Hence by taking
supremum over all s S, we have D(fn , f ) 2
<
for
all
n
n
.
To
show
that f is bounded,
3
i.e. supr,sS d(f (s), f (r)) < , use the triangle inequality twice and the fact that fN is bounded.
d(f (s), f (r)) d(f (s), fN (s)) + d(fN (s), fN (r)) + d(fN (r), f (r))
2
+ d(fN (s), fN (r)) <
3
10. Prove that a discrete metric space is compact if and only if its underlying set is finite.
Proof. One direction is obvious, as each subset of a finite set is finite. For the other direction, take
a compact space (X, d) with the discrete metric, suppose the underlying set X were infinite and
look at the open cover C = {{x} : x X}. As C is infinite, any finite subcover C 0 would yield an
element y X that is not covered by C 0 which contradicts the compactness of (X, d).
11. Let (X, d) be a metric space and lim xn = x0 in X. Show that the set K = {x0 , x1 , . . . } is compact.
Proof. As x0 lies in K, there must be one open set U0 in any open covering C that contains it. This
U0 must contain an open ball B(x0 , ). As the sequence (xn ) converges to x0 , there is a n0 such
that for all n n0 , we have d(xn , x0 ) < which leaves only finitely many points outside the ball.
The (finite) subset of those, that arent contained in U0 can be covered by finitely many open sets
from C.
12. Let (X, d) be a metric space, and let F X be a closed subset and K X a compact subset of
X. Prove that
F K 6= inf{d(x, y) : x F, y K} = 0.
Proof. The one direction is straight forward. A point x F K satisfies d(x, x) = 0. For the other
direction, take sequences (xn ) F, (yn ) K that satisfy inf nN d(xn , yn ) = 0. For all > 0 we have
an n1 N s.t. for all n n1 , d(xn , yn ) < 2 . K is compact, so (yn ) has a converging subsequence
(yni ) with limit y K. In particular, we find an n2 N such that for all n n2 , d(yni , y) < 2 .
With the triangle inequality, we get d(y, xni ) d(y, yni ) + d(yni , xni ) < . Thus y is a limit point
of F and since F is closed, ist must be contained in F .
13. Let (X1 , d1 ), . . . , (Xn , dn ) be compact metric spaces, and let X := X1 Xn be equipped with
either the D1 or D metric. Prove that X is also compact.
Proof. Take a sequence (xk ) in X and look at the sequences (xk1 ), . . . , (xkn ) in X1 , . . . , Xn . Since
(X1 , d1 ) is compact, it has a convergent subsequence (xk1l ) with index set I1 . The induced sequence
k
(xk2l ) in X2 has a converging subsequence (x2lm ) for an index set I2 I1 . Iterating gives us an
index set In such that the sequences xK
j for K In converge for all j = 1, . . . , n. Thus, with
both metrics, the corresponding subsequence of (xk ) converges. From the lecture we know that this
implies compactness (Chapter 1.9, Lemma).
14. Let E be a normed space and A and B two compact subspaces. Then the Minkowski sum
A + B := {a + b : a A, b B}
is also compact.
Proof. This can again be shown using sequential compactness. Another approach is the following.
Let C be an open cover of A+B. Then the families CA := {U B : U C} and CB := {U A : U C}
are covers of A and B respectively. Those sets are compact, so there exist finite subcovers. Both
families (translated by A or B) together yield a finite subcover of A + B.
15. Let (X, d) be a metric space. Define a function d0 : X X [0, +) by
d0 (x, y) := min{d(x, y), 1}
for every (x, y) X X. Show that d0 is a metric on X, and moreover is equivalent to d, i.e.,
induces the same topology as d.
Proof. Checking the metric properties is straight forward. To see that the topologies coincide, use
that for < 1, Bd (x, ) = Bd0 (x, ).
B 00
Proof. Remark: T 0 is called the lower limit topology or right half-open interval topology; the
topological space is called Sorgenfrey line. T 00 is called K-topology.
First, check the two basis properties for the three families. This is straightforward. The only thing
one has to watch out for is the second property for B 00 . There one needs to do a (simple) case
analysis.
0
T
S T : Let U T . Then x 0 U (ax , bx ) U : x (ax , bx ). We can rewrite any interval (a, b)
as a<c<b [c, b), so (ax , bx ) T , thus for all y (ax , bx ), so in particular for the previous x, there
exist By B 0 : y By (ax , bx ). Thus U T 0 .
T T 00 : Obvious.
T 00 6 T 0 : Use R \ K. It is open in T 00 but has nonempty intersection with every open set in T 00
that also contains zero.
T 0 6 T 00 : Take for instance [2, ). It is open in T 00 but any basis set in B 00 that contains 2 will
also contain numbers smaller than 2.
4. Let X = R and B = {(a, b) : a, b Q, a < b}. Prove that B is a basis for a topology on X, and
that it generates the standard topology on X.
Proof. Let x R. Let n be some integer larger than x. Then n < x < n. Now let (a, b) and
(c, d) be basis sets that contain x. Then (max(a, c), min(b, d)) lies in B because max(a, c) < x and
x < min(b, d). Finally we need to check that the generated topology coincides with the standard
topology. The one inclusion is obvious. For the other one, take a set U that is open in the standard
topology. Then there are real numbers a < b such that x (a, b) for all x U . There exists an N
1
such that N > xa
. which can be rearranged to 1 < xN aN , so there must be another integer
M between aN and xN . This again can be rearranged to a < M
N < x. Doing the same thing for x
M0
,
)
(a,
b).
Hence
U
is
open
in
the
topology
generated by B.
and b, we get x ( M
N N0
5. Let A be a subspace of X and let B be a subspace of Y . We equip A and B with the subspace
topologies. Prove that the product topology on A B is the same as the topology A B inherits
as a subspace of X Y .
Proof. Denote with TAB,prod the product topology on A B and with TAB,sub the subspace
topology induced by X Y .
TAB,prod TAB,sub : Let UA UB TA TB . The open sets in A and B respectively give us
open sets VA and VB in the spaces X and Y by the definition of the subspace topology and if we
intersect VA VB with A B, we obtain UA UB . Thus the basis of TAB,prod is contained in
TAB,sub , and the same holds for the topology generated by it.
TAB,sub TAB,prod : Let U TAB,sub . Then there exists a set V TXY such that V AB =
U . By the construction of the product topology, for all v V we have a basis set VX VY TX TY
with v VX VY V . In particular this holds for all v U with v UA UB := VX VY AB
U , so U TAB,prod .
6. Let Y be a subspace of X. Prove that if A is closed in Y , and Y is closed in X, then A is closed in
X.
Proof. Denote by UZ (x) the family of sets which contain x and are open in Z and by AZ Z the
closure of A in Z. Recall the following criterion from the lecture (a theorem from Chapter 2.3):
x Z lies in AZ U UZ (x) : U A 6= .
Let x AX , then every element of UX (x) must have non-empty intersection with Y , because A Y .
This means that x Y X which equals Y because Y is closed in X. Also, all sets in UY (x) originate
from a set in UX (x), so they all have non-empty intersection with A. Thus x AY which coincides
with A because A is closed in Y . A AX is trivial, so we deduce A = AX .
7. Let X be a Hausdorff space. Prove that a sequence in X converges to at most one point in X.
Proof. Denote by U(x) the family of open sets which contain x. Assume we have a converging
sequence (xn ) with distinct limit points x and y. Then by definition for all U U(x), V U(y)
there exist nU , nV N such that for all n nU , nV , we have xn U and xn V respectively.
X is Hausdorff, so there exist disjoint neighbourhoods U 0 U(x) and V 0 U(y) and for n
max{nU 0 , nV 0 } we have xn U 0 V 0 = . Contradiction.
8. Show that if A is closed in X, and B is closed in Y , then A B is closed in X Y .
Proof. X \ A and Y \ B are open and so is (X \ A) Y and X (Y \ B). Thus (X Y ) \ (A B) =
((X \ A) Y )(X (Y \ B)) is also open. Alternatively, use A B = AB from exercise 10.
9. Show that if U is open in X and A is closed in X, then U \A is open in X, and A\U is closed in X.
Proof. U \ A = U (X \ U ) and X \ (A \ U ) = (X \ A) U are the finite intersection resp. union
of open sets and therefore open.
10. Let A X and B Y . Show that A B = A B in X Y .
Proof. Wlog only look at the limit points (for isolated points there is no problem).
Let (an , bn ) be a convergent sequence in A B with limit point (a, b). Let V TX and
W TY with a V and b W , then V W TXY and (a, b) V W . Thus there
exists an nV,W N such that all subsequent elements (an , bn ) lie in V W . In particular
an V, bn W for all n nV,W which makes a and b limit points. A and B are closed,
therefore a A and b B.
Let an a in A and bn b in B. Take an arbitrary U TXY with (a, b) U . Then for
all u U there is a basis set V W TXY such that u V W U . The open sets V
and W give us numbers nV , nW N such that an V, bn W for all subsequent n. Define
nU := max(nv , nW ), then n nU : (an .bn ) U . We deduce that (a, b) is a limit point.
11. Show that the product of two Hausdorff spaces is Hausdorff.
Proof. Let (a, b) 6= (c, d) be points in X Y , where X, Y are Hausdorff spaces. Then a 6= c or
b 6= d. Wlog assume a 6= c. Then there exist disjoint open neighbourhoods A and C for a and c in
X. This gives us disjoint open neighbourhoods A Y and C Y of (a, b) and (c, d).
12. Show that X is Hausdorff if and only if the diagonal := {(x, x) : x X} is closed in X X.
Proof. X is Hausdorff. For every pair a, b X with a 6= b there there exist disjoint open
neighbourhoods A 3 a and B 3 b. Every element (a, b) of X \ has a basis set A B with
(a, b) A B X \ . X \ is open. is closed.
13. If A X, let us define the boundary of A by the equation:
A := A X\A.
Show that:
(a) A A = and A = A A;
(b) A = if and only if A is both open and closed;
(c) A is open if and only if A = A\A.
Proof.
(a) x A. There is an open neighbourhood U A of x. There is an open neighbourhood U
of x with U X \ A = . x
/ X \ A. This means A = X \ X \ A. From this we deduce two
therefore A A = A (X \ A) A = A.
(b) A is open and closed if and only if A = A = A and using (a), this corresponds to A = .
Subtracting A from the second equation (and using the first statement) of (a), we get that
A = A (A \ A) = A \ (A \ A), so A \ A = .
14. Let f : X Y be a continuous map, A X and A0 denotes the set of all limit points of A. If
x A0 is it true that f (x) f (A)0 , i.e., is f (x) a limit point of f (A)?
Proof. No, in general this is not true. Isolated points can cause a problem. Consider for instance
the constant map f : R [0, 1] {2} given by f (x) = 2. It is obviously continuous but 2 is no limit
point.
15. Let f : A B and g : C D be continuous functions. Consider the map f g : A C B D
defined by
(f g)(a, c) := (f (a), g(c))
for every (a, c) A C. Prove that f g is also continuous.
Proof. Let U TBD . Then for all u U there is a set Bu Du TB TD with u Bu Du U .
Define Au := f 1 (Bu ) and Cu := g 1 (Du ). Au Cu is an open neighbourhood for any v
(f g)1 (u) due to the fact that f and g are continuous. Thus for every v (f g)1 (U ) there is
a set A(f g)(v) A(f g)(v) TA TC with v A(f g)(v) C(f g)(v) (f g)1 (U ) which proves
our claim.
16. Let X and Y be topological spaces with Y being Hausdorff. Let A X and f : A Y be a
continuous map. Prove that if f can be extended to a continuous function g : A Y , then g is
uniquely determined by f .
Proof. From exercise 2.14 we know that a convergent sequence an in A with limit point a yields
a convergent sequence g(an ) with limit point g(a). In exercise 2.7 we learned that the limit of a
convergent sequence is unique in Hausdorff spaces, so g(a) = limn f (an ) is uniquely defined.
17. Let {X }J be an indexed family of topological spaces. Assume that the topology on each X is
given by a basis B . Prove that the family
Y
B : B B
J
X , while
X .
Q
Proof.
Let U Tbox , then for all u U there exists a set J U with the property that u
Q
For all , we have u U and a basis set B such that u B U , so
J
QU U . Q
u J B J U U . The same line of reasoning works for the product topology.
Q
18. Let {X }J be an indexed family of Hausdorff spaces. Prove that J X is a Hausdorff space
in both the box and product topology.
Proof. If two points (x ) and (y ) are different, then there must be an index J such that
x 6= y . Use the same argument as in 11. to prove that the claim (for both topologies).
19. Let p : X Y be a continuous map. Show that if there is a continuous map f : Y X such that
p f = idY , then p is a quotient map.
Proof. For every y Y the element x := f (y) is in p1 (y), so p is surjective. Now let p1 (U ) TX .
Then U = (p f )1 (U ) = f 1 p1 (U ) is open because f is continuous. Together with the fact
that p is continuous, this yields the claim.
20. Let X be a topological space and A X a subspace. A continuous map r : X A is a retraction
if r|A = idA , i.e., r(a) = a for every a A. Show that a retraction is a quotient map.
Proof. Use the previous exercise with p := r and f : A , X, the embedding of A into X.
Proof. We can write X as the disjoint union of A, A and (X \ A). If C A is empty, then
X and B
Proof. A and B dont coincide with X and Y , so we find points x X \ A and y Y \ B and can
construct sets {x} Y and X {y} that are disjoint from A B. The union ({x} Y ) (X {y}) is
connected as it is the union of connected sets with a point (x, y) in common (see Exercise 3.1).
6. Let p : X Y be a quotient map, p1 ({y}) be connected for every y Y , and Y be connected.
Show that X is connected.
Proof. Assume U and V form a separation of X. We know that p(U ) p(U ) = Y because p
is surjective. Assume there exists a y p(U ) p(V ). Denote the fibre p1 ({y}) with C. C is
connected and the sets C U and C V are open in C and C = (C U ) t (C V ) which gives us
a contradiction. This means that p(U ) and p(V ) are disjoint but also that p1 (p(U )) = U . p is a
quotient map, so if p1 (p(U )) is open, then so is p(U ). The same holds for p(V ). Thus p(U ) and
p(V ) form a separation of Y , which cannot be true.
7. Let Y be a subspace of X, and both connected. Show that if A and B form a separation of X\Y ,
then Y A and Y B are connected
Proof. Let us work with the definition of connectedness using limit points, which was introduced in
Lemma 1 of the third chapter. Assume that Y A is not connected. Then we find non-empty U, V
with the property that U V 0 = , U 0 V = and Y A = U V . Y is connected, so by the second
lemma of Chapter 3 either Y U or Y V . Say Y U . Then V A A0 , hence V B 0 = and
V 0 B = . If we now look at the union U B, we see that V (U B)0 = V (U 0 B 0 ) = and
V 0 (U B) = . So V and U B form a separation of X = Y A B = V U B. Contradiction.
The same line of reasoning proves that Y B must be connected.
8. Show that Rd and R are not homeomorphic if d > 1.
16. Let f : X Y be a continuous map with Y compact Hausdorff. Show that f is continuous if and
only if the graph of f
Gf := {(x, f (x)) : x X}
is a closed subset of X Y .
Proof.
Let (x, y) (X Y ) \ Gf . Then y 6= f (x) and therefore we can find disjoint open neighbourhoods U 3 y, V 3 f (x). Define W := f 1 (V ). Then f (W ) V Y \ U and thus
W U Gf = . W U is open, so we found an open neighbourhood of (x, y) in (X Y ) \ Gf .
Now let Gf be closed and let V be an open set in Y . Then the intersection Gf X (Y \ V )
is also closed. If we apply Exercise 3.15, we can deduce that 1 (Gf X (Y \ V )) = {x X :
f (x)
/ V } = X \ f 1 (V ) is closed and equivalently, that f 1 (V ) is open.
10