Homework 2

Exercise 2.1. Give a careful proof of De Morgan’s identities: for X , {Ai }i ∈I sets, the following two equalities
hold:
[ \
Xr Ai = (X r Ai )
i ∈I i ∈I
\ [
Xr Ai = (X r Ai ).
i ∈I i ∈I
Solution. Suppose x ∈ X r i ∈I Ai . This means that x ∈ X but x < i ∈I Ai . The latter means that x < Ai for
S S
every i ∈ I . We conclude that x ∈ X r Ai for every i ∈ I , i. e. x ∈ i ∈I (X r Ai ).
T
Conversely, if x ∈ i ∈I (X r Ai ), then this means that x ∈ X r Ai for every i ∈ I . That is, x ∈ X and x < Ai
T
for all i ∈ I . The latter is equivalent to x < i ∈I Ai . So x ∈ X r i ∈I Ai .
S S
Similarly, suppose x ∈ X r i ∈I Ai . Then x ∈ X but x < i ∈I Ai . So there is some j ∈ I such that x < A j .
T T
For this j we then have x ∈ X r A j ⊂ i ∈I (X r Ai ).
S
Conversely, if x ∈ i ∈I (X r Ai ), then there is some j ∈ I such that x ∈ X r A j , i. e. x ∈ X but x < A j . This
S
implies that x < i ∈I Ai because A j ⊃ i ∈I Ai . Hence, x ∈ X r i ∈I Ai .
T T T

Exercise 2.2. Let X be any set. Show that there is a bijection of the power set of X (the set of subsets of X )

P(X ) = {A : A ⊂ X }

and the set of maps from X to {0, 1},

Maps(X , {0, 1}).
Solution. Define a map f : Maps(X , {0, 1}) P(X ) by setting f (χ ) = {x ∈ X : χ (x ) = 1} for a function
χ ∈ Maps(X , {0, 1}). Conversely, given a subset A ⊂ X define a function χA : X {0, 1} by

0
 if x < A
χA (x ) = 
1
 if x ∈ A.

Define д : P(X ) Maps(X , {0, 1}) by д(A) = χA . We check that f and д are inverse functions:

f (д(A)) = f (χA ) = {x ∈ X : χA (x ) = 1} = {x ∈ X : x ∈ A} = A
0
 if x < f (χ ), i. e. χ (x ) = 0
д( f (χ ))(x ) = 
1
 if x ∈ f (χ ), i. e. χ (x ) = 1,

for A ⊂ X and χ ∈ Maps(X , {0, 1}). We conclude that f and д are inverse bijections Maps(X , {0, 1})  P(X ).

Exercise 2.3. Let X be a set. Complete the proof that the discrete metric d discrete : X × X [0. + ∞) on X ,
defined as
 0 if x = y

d discrete (x, y) = 
 1 if x , y,

is indeed a metric.
Then, show that every subset U ⊂ X is an open set with respect to the discrete metric.
Solution. We prove the triangle inequality for d discrete . So suppose, x, y, z ∈ X . There are three cases: x = y = z,
or exactly two points of x, y, z are equal, or x, y and z are distinct points. In the first case, we have

0 = d discrete (x, y) + d discrete (y, z) ≥ d discrete (x, z) = 0.

1
In the second case, if say x = y but y , z, we have

1 = d discrete (x, y) + d discrete (y, z) ≥ d discrete (x, z) = 1

and if x , y but x = z, then

1 = d discrete (x, y) + d discrete (y, z) ≥ d discrete (x, z) = 0.

Finally, in the third case, we have

2 = d discrete (x, y) + d discrete (y, z) ≥ d discrete (x, z) = 1.

Now, let U ⊂ X be a subset and let x ∈ U . Observe that Bddiscrete (x, 1/2) = {x } ⊂ U . So U is a neighborhood
of x. But since x was arbitrary we conclude that U is a neighborhood of each of its points, that is, U is open.

Exercise 2.4.
(i) Consider Rn with its Euclidean metric
q
d Eu (x, y) = (x 1 − y1 ) 2 + · · · + (x n − yn ) 2

and its taxi-cab metric

d Ta (x, y) = |x 1 − y1 | + |x 2 − y2 | + · · · + |x n − yn |.

Show that there exist positive constants c 1 , c 2 such that, for any points x, y ∈ Rn ,

c 1d Eu (x, y) ≤ d Ta (x, y) ≤ c 2d Eu (x, y).

Note that c 1 and c 2 are not allowed to depend on x and y above.

Using the above fact (which you may want to call a Lemma), prove that a subset U ⊂ Rn is open with
respect to the d Eu metric if and only if U is open with respect to the d Ta metric. In other words, d Eu and
d Ta have the same open sets, or induce the same topology.1
(ii) More generally, given an integer p ≥ 1, define

X n  1/p
dp (x, y) B  |x i − yi | 
p 
 i=1 

You may assume that dp is a metric. Show using similar methods that dp also induces the same topology
as d Eu .
Solution.
(i) First, let’s assume that d Eu and d Ta are in fact equivalent and using this show that they induce the same
topology on Rn . For this, assume that U ⊂ Rn is open with respect to d Eu and let x ∈ U . There is some
ε > 0 such that BdEu (x, ε ) ⊂ U . Now, suppose that y ∈ BdTa (x, c 1ε). Then d Eu (x, y) ≤ 1/c 1 · d Ta (x, y) < ε
and therefore y ∈ BdEu (x, ε) ⊂ U . So, BdTa (x, c 1ε) ⊂ U . We conclude that U is open with respect to d Ta
as well.
Conversely, if U ⊂ Rn is open with respect to d Ta and x ∈ U , let ε > 0 be small enough such that
BdTa (x, ε) ⊂ U . Suppose y ∈ BdEu (x, ε/c 2 ). Then d Ta (x, y) ≤ c 2d Eu (x, y) < ε and therefore y ∈ BdTa (x, ε).
Hence, BdEu (x, ε/c 2 ) ⊂ U and we conclude that U is open with respect to d Eu as well.
1 Wesay a pair of metric d 1 , d 2 induce the same topology on a set X if they have the same open sets, meaning for any subset U ⊂ X ,
U is open with respect to d 1 if and only if it is open with respect to d 2 .

2
 To prove that d Eu and d Ta are in fact equivalent, note that for a, b ∈ R we have 0 ≤ (a −b) 2 = a 2 −2ab +b 2
and therefore 2ab ≤ a 2 + b 2 . Then, using that 2|x i − yi ||x j − y j | ≥ 0 for all 1 ≤ i, j ≤ n, we can compute
n
X
(x 1 − y1 ) 2 + · · · + (x n − yn ) 2 ≤ (x 1 − y1 ) 2 + · · · + (x n − yn ) 2 + 2|x i − yi ||x j − y j | =
i, j=1
i <j

= [|x 1 − y1 | + · · · + |x n + yn |]2 = d Ta (x, y) 2 =

X n
= (x 1 − y1 ) + · · · + (x n − yn ) +
2 2
2|x i − yi ||x j − y j | ≤
i, j=1
i <j
X n f g
≤ (x 1 − y1 ) 2 + · · · + (x n − yn ) 2 + (x i − yi ) 2 + (x j − y j ) 2 ≤
i, j=1
i <j
f g
≤ n2 (x 1 − y1 ) 2 + · · · + (x n − yn ) 2 .

Taking square roots this implies d Eu (x, y) ≤ d Ta (x, y) ≤ n d Eu (x, y).

(ii) For x, y ∈ Rn let z max = maxi |x i − yi |. Compute:
n
X X n  p
p
|x i − yi | ≤
p
n z max ≤ n  |x i − yi | 
i=1
 i=1 
n n  p
1
X  X
p
|x i − yi |p ≥ z max ≥  |x i − yi | 
i=1
 n i=1 

Combining these and taking p th roots, we conclude that n−1 d Ta (x, y) ≤ dp (x, y) ≤ n1/p d Ta (x, y). So dp
and d Ta are equivalent metrics and by the same argument as in part (i) they induce the same topologies
on Rn . Hence, dp also induces the same topology as d Eu .

Exercise 2.5.
(i) Let x ∈ Q be a rational number. In class, we defined the 2–adic norm
p
 2−n if x , 0 and n ∈ Z is the unique integer such that x = 2n q with p and q odd
|x |2 = 


0 if x = 0

and the 2–adic metric

d2 : Q × Q [0, ∞)
by d 2 (x, y) = |x − y|2 . Prove that (Q, d 2 ) is a metric space. In fact, prove that d 2 satisfies conditions (i)
and (ii) of being a metric, as well as a condition stronger than (iii):

For all x, y, z ∈ Q, d 2 (x, z) ≤ max(d 2 (x, y), d 2 (y, z)).

Metrics satisfying this stronger condition are sometimes called ultrametrics or non-Archimedian metrics.
(ii) Prove that for any x ∈ Q, there is some r > 0 such that the complement Q r Bd2 (x, r ) is an open set.2
In contrast, note that in the Euclidean metric R r (x − r , x + r ) is never open.
Solution.

that Bd (x, r ) is the ball of radius r centered at x in the metric d, which we defined in class. In short, we defined Bd (x, r ) =
2 Recall
{y ∈ X : d (x, y) < r }.

3
 (i) First note that |x − y|2 = 0 if and only if x = y because 2−n , 0 for all n ∈ Z. Also, if |x − y|2 = 2−n ,
p −p
then x − y = 2n q for some odd p and q. So, y − x = 2n q and therefore |y − x |2 = 2−n = |x − y|2 .
Now, suppose a, b ∈ Q are nonzero rational numbers with |a|2 = 2−n and |b |2 = 2−m . This means
p p0
that a = 2n q and b = 2m q 0 for odd integers p, p 0, q, q 0 ∈ Z. Assume without loss of generality that
|a|2 ≥ |b |2 . Then m ≥ n and therefore
np mp n p m−n p n pq + 2
0 " 0# 0 m−n p 0q
a +b = 2 + 2 0 = 2 +2 = 2
q q q q0 qq 0
Because q and q 0 are both odd, the denominator qq 0 will be odd as well. Because of our assumption
that m − n ≥ 0 the numerator pq 0 + 2m−n p 0q is an integer. It follows that pq 0 + 2m−n p 0q = 2` r for ` ≥ 0
and r ∈ Z odd. Consequently, a + b = 2n+` qqr 0 and |a + b |2 = 2−n−` ≤ 2−n = max{|a|2 , |b |2 }.
We conclude that
d 2 (x, z) = |x − z|2 = |x − y + y − z|2 ≤ max{|x − y|2 , |y − z|2 } = max{d 2 (x, y), d 2 (y, z)}
for x, y, z ∈ Q.
(ii) Let x ∈ Q and suppose r > 0. Suppose y ∈ Q r Bd2 (x, r ) and z ∈ Bd2 (y, r ). The ultrametric triangle
inequality from part (i) implies that
r ≤ |x − y|2 = |x − z − y + z|2 ≤ max{|x − z|2 , |y − z|2 }.
For sake of contradiction, suppose that |x −z|2 ≤ |y −z|2 . Then we would have r ≤ |x −y|2 ≤ |y −z|2 < r
which is impossible. So it must be the case that |y − z|2 ≤ |x − z|2 . But then r ≤ |x − y|2 ≤ |x − z|2 and
therefore z ∈ Q r B(x, r ). Since z was arbitrary we conclude that Bd2 (y, r ) ⊂ Q r Bd2 (x, r ) and, because
y ∈ Q r Bd2 (x, r ) was arbitrary as well, it follows that Q r Bd2 (x, r ) is open.
Exercise 2.6. If (M 1 , d 1 ) and (M 2 , d 2 ) are metric spaces, then one can define a distance function d on the
Cartesian product M 1 × M 2 by
d ((m, n), (m 0, n 0 )) = d 1 (m, m 0 ) + d 2 (n, n 0 ).
(i) Show that d defines a metric on M 1 × M 2 , called the (standard) product metric.
(ii) Prove that if U1 ⊂ M 1 is open and U2 ⊂ M 2 is open, then U1 × U2 ⊂ M 1 × M 2 is open. Conversely, is it
true that every open set V ⊂ M 1 × M 2 is of the form U1 × U2 for some U1 , U2 ?
Solution.
(i) Because d 1 (m, m 0 ) and d 2 (n.n 0 ) are both always non-negative, d ((m, n), (m 0, n 0 )) = 0 is equivalent to
d 1 (m, m 0 ) = 0 and d 2 (n, n 0 ) = 0. But this is equivalent to (m, n) = (m 0, n 0 ). Also,
d ((m, n), (m 0, n 0 )) = d 1 (m, m 0 ) + d 2 (n, n 0 ) = d 1 (m 0, m) + d 2 (n 0, n) = d ((m 0, n 0 ), (m, n)).
Finally, suppose that (m, n), (m 0, n 0 ), (m 00, n 00 ) ∈ M 1 × M 2 . Then by the triangle inequalities for d 1 and
d 2 we conclude
d ((m, n), (m 00, n 00 )) = d 1 (m, m 00 ) + d 2 (n, n 00 ) ≤
≤ d 1 (m, m 0 ) + d 1 (m 0, m 00 ) + d 2 (n, n 0 ) + d 2 (n 0, n 00 ) =
= d ((m, n), (m 0, n 0 )) + d ((m 0, n 0 ), (m 00, n 00 )).
(ii) Suppose that U1 ⊂ M 1 and U2 ⊂ M 2 are both open and let (m, n) ∈ U1 × U2 . Let ε 1 > 0 be small enough
such that Bd1 (m, ε 1 ) ⊂ U1 and similarly let ε 2 > 0 be small enough such that Bd2 (n, ε 2 ) ⊂ U2 . Pick any
ε > 0 with ε < min{ε 1 , ε 2 } and suppose (x, y) ∈ Bd ((m, n), ε). Then
d 1 (m, x ) ≤ d 1 (m, x ) + d 2 (n, y) = d ((m, n), (x, y)) < ε < ε 1
d 2 (n, y) ≤ d 1 (m, x ) + d 2 (n, y) = d ((m, n), (x, y)) < ε < ε 2

4
and therefore x ∈ U1 and y ∈ U2 . So, (x, y) ∈ U1 × U2 and we conclude that Bd ((m, n), ε) ⊂ U1 × U2 .
This implies that U1 × U2 is open since (m, n) ∈ U1 × U2 was arbitrary.
The converse isn’t true. For example (0, 1) × (0, 1) ∪ (2, 3) × (2, 3) ⊂ R × R is open but not a cartesian
product of open sets U1 , U2 ⊂ R.