LU - Department of Mathematics

Dr H. Gebran

Math 3301 Exercises of chapter 1 - Solutions Fall 2017/2018

1. Let X be a set and A ⊂ X. Let U ⊂ IR. Let 1A be the characteristic function of A. Show
that 


X if 1 ∈ U and 0 ∈ U

A if 1 ∈ U and 0 ∈
/U
1−1
A (U ) =  c

A if 1 ∈
/ U and 0 ∈ U

∅ if 1 ∈
/ U and 0 ∈
/ U.

Solution.

Case 1: 1 ∈ U and 0 ∈ U . Let x ∈ A, then 1A (x) = 1 ∈ U so x ∈ 1−1 A (U ). This means

that A ⊂ 1A (U ). Similarly if x ∈ A then x ∈ 1A (U ) and therefore Ac ⊂ 1−1
−1 c −1
A (U ). Thus,
X = A ∪ Ac ⊂ 1−1
A (U ). Whence, 1−1
A (U ) = X.

Case 2: 1 ∈ U and 0 ∈/ U . We see as in case 1 that A ⊂ 1−1 −1

A (U ). Conversely, let x ∈ 1A (U ).
Then 1A (x) ∈ U and so 1A (x) 6= 0 since 0 ∈/ U . Therefore 1A (x) = 1 since 1A (x) can take
−1
just two values. Whence, x ∈ A and so 1A (U ) ⊂ A.

Case 3: 1 ∈
/ U and 0 ∈ U . This case is similar to case 2.

Case 4: 1 ∈ / U and 0 ∈/ U . Suppose 1−1 −1

A (U ) 6= ∅ and let x ∈ 1A (U ). Then 1A (x) 6= 0 and
1A (x) 6= 1. This is impossible.

2. Prove the following relations

(a) 1A∩B = 1A 1B .
(b) 1Ac = 1 − 1A .
(c) 1A∪B = 1A + 1B − 1A 1B .

Solution. a) We distinguish between two cases.

Case 1. x ∈ A ∩ B. Then first, 1A∩B (x) = 1. Second, x ∈ A ∩ B implies that x ∈ A and so
1A (x) = 1. Similarly, 1B (x) = 1. Thus, 1A∩B (x) = 1A (x)1B (x) = 1 in this case.
Case 2. x ∈/ A ∩ B. Then first, 1A∩B (x) = 0. Second, since x ∈/ A ∩ B, then either x ∈ /A
or x ∈
/ B. So either 1A (x) = 0 or 1B (x) = 0. Thus, 1A∩B (x) = 1A (x)1B (x) = 0 in this case
too.

b) Here again we distinguish between two cases.

/ Ac , hence
Case 1. x ∈ A. Then first, 1A (x) = 1 and so 1 − 1A (x) = 0. Second, x ∈
1Ac (x) = 0. Hence the equality in this case.
Case 2. x ∈ Ac . In this case both sides are equal to 1.

1
 c) We also distinguish between two cases.
Case 1. x ∈
/ A ∪ B. Then x ∈
/ A and x ∈
/ B. Therefore both sides are equal to zero.
Case 2. x ∈ A ∪ B. Then either x ∈ A or x ∈ B. Here we must distinguish between three
subcases. If x ∈ A but x ∈ / B, then 1A (x) + 1B (x) − 1A (x)1B (x) = 1 + 0 − 0 = 1. If x ∈ B
but x ∈/ A, then 1A (x) + 1B (x) − 1A (x)1B (x) = 0 + 1 − 0 = 1. If x ∈ A and x ∈ B, then
1A (x) + 1B (x) − 1A (x)1B (x) = 1 + 1 − 1 = 1. Thus, in all cases, the two sides coincide.

3. Let (An ) be a sequence of subsets of a set X.

a) Show that if (An ) is increasing, then, lim 1An = 1∪An .

b) Show that if (An ) is decreasing, then lim 1An = 1∩An .

Solution. a) We must distinguish between two cases.

Case 1. x ∈ ∪An . Then first, 1∪An (x) = 1. Second, x ∈ ∪An implies that x ∈ Am for
some positive integer m. But since the sequence (An ) is increasing, then x ∈ An for all
n ≥ m. Therefore 1An (x) = 1 for n ≥ m. This means that lim 1An (x) = 1. Thus both sides
are equal in this case.
Case 2. x ∈ / ∪An . Then first, 1∪An (x) = 1. Second, x ∈ / ∪An implies that x ∈/ An for
every n. This implies that 1An (x) = 0 for all n and so lim 1An (x) = 0. Thus both sides are
equal in this case too.

b) Use a similar reasoning as in part a) (or use part a) and De Morgan’s law).

4. Let (An ) be a sequence of sets.

a) Suppose that (An ) is increasing. Show that lim inf An = lim sup An = ∪n An .

b) Suppose that (An ) is decreasing. Show that lim inf An = lim sup An = ∩n An .

Solution. a) Since (An ) is increasing, we have ∩k≥n Ak = An . Therefore lim inf An =

∪n An . But lim inf An ⊂ lim sup An ⊂ ∪An . Hence the equality.

b) Since (An ) is decreasing, we have ∪k≥n Ak = An . Therefore lim sup An = ∩n An . But

∩An ⊂ lim inf An ⊂ lim sup An . Hence the equality.

5. Let (Ai )i∈I and (Bi )j∈J be two partitions of a set E. Show that the collection
(Ai ∩ Bj )(i,j)∈I×J
is also a partition of E.

Solution. Let Ci,j = Ai ∩ Bj . We have to prove that i) ∪i,j Ci,j = E and that ii)
Ci,j ∩ Ck,l = ∅ if (i, j) 6= (k, l).
S S S S S S S
For i), we can write i,j Ci,j = i j Ai ∩ Bj = i (Ai ∩ j Bj ) = i Ai ∩ E = i Ai = E.
Alternatively, we can say: let x ∈ E. Since (Ai ) is a partition of E, then x ∈ Ai for some
i ∈ I. Also, since (Bj ) is a partition of E, then x ∈ Bj for some j ∈ J. Thus x ∈ Ai ∩ Bj
for some couple (i, j) ∈ I × J.
For ii) If (i, j) 6= (k, l), then i 6= k or j 6= l. Now Ci,j ∩ Ck,l = Ai ∩ Bj ∩ Ak ∩ Bl . If i 6= k then
Ai ∩ Ak = ∅ and so Ci,j ∩ Ck,l = ∅. If j 6= l, then Bj ∩ Bl = ∅ and so again Ci,j ∩ Ck,l = ∅.

2
6. Let (An ) be an arbitrary sequence of sets.

a) Construct a sequence of sets (Bn ) such that

(i) Bn ⊂ Bn+1 for every n ∈ IN∗ (the sequence is increasing).

N N
An for every N ∈ IN∗ .
S S
(ii) Bn =
n=1 n=1
∞
S ∞
S
(iii) Bn = An .
n=1 n=1

b) Construct a sequence of sets (Bn ) such that

(i) The Bn are pairwise disjoint.

N N
An for every N ∈ IN∗ .
S S
(ii) Bn =
n=1 n=1
∞
S ∞
S
(iii) Bn = An .
n=1 n=1

Hint. Let B1 = A1 and Bn = An \(A1 ∪ · · · ∪ An−1 ) for n ≥ 2.

Solution.

a) Let Bn = A1 ∪ · · · ∪ An . It is clear that (Bn ) is increasing and that An ⊂ Bn . Conversely,

let x ∈ Bn for some n then x ∈ Ak for some k ≤ n.

b) i) Let m < n. Since Bm = Am \(A1 ∪· · ·∪Am−1 ) and Bn = An \(A1 ∪· · ·∪Am ∪· · ·∪An−1 ),

then x ∈ Bm ⇒ x ∈ Am ⇒ x ∈ / Bn .
ii) and iii) It is clear that Bn ⊂ An therefore ∪N N ∞ ∞
n=1 Bn ⊂ ∪n=1 An and ∪n=1 Bn ⊂ ∪n=1 An .
Conversely, let x ∈ An for some n and let n0 = min{m | x ∈ Am } ≤ n. If n0 = 1, then
x ∈ A1 = B1 . If n0 > 1 then x ∈ An0 \(A1 ∪ · · · ∪ An0 −1 ) = x ∈ Bn0 .

Remark. Note that we can also write Bn = An \(B1 ∪· · ·∪Bn−1 ) because B1 ∪· · ·∪Bn−1 =
A1 ∪ · · · ∪ An−1 .

7. Let (X, d) be a metric space. For a subset E of X, let δ(E) = sup{d(x, y) | x, y ∈ E} be the
diameter of E. Show that if A and B are two subsets of X with A ∩ B 6= ∅, then

δ(A ∪ B) ≤ δ(A) + δ(B).

Solution. Let x, y ∈ A ∪ B. We distinguish between 4 cases.

Case 1. x, y ∈ A. Then d(x, y) ≤ δ(A) ≤ δ(A) + δ(B).

Case 2. x, y ∈ B. Then d(x, y) ≤ δ(B) ≤ δ(A) + δ(B).

Case 3. x ∈ A, y ∈ B. Let z ∈ A ∩ B. Then d(x, y) ≤ d(x, z) + d(z, y) ≤ δ(A) + δ(B).

Case 4. x ∈ B, y ∈ A. Similar to case 3.

So in all cases, d(x, y) ≤ δ(A) + δ(B) and therefore δ(A ∪ B) ≤ δ(A) + δ(B).

3
 8. Let I be an interval of IR. Show that the length of I is equal to its diameter.

Solution. Let `(I) denote the length of I.

Case 1. I is unbounded. Then both `(I) and δ(I) are infinite.

Case 2. I is bounded, so I is of the form ]a, b[, ]a, b], [a, b[, [a, b]. But all these intervals
have the same length and the same diameter. So it is enough to consider the case I =]a, b[
for example. Let first x, y ∈]a, b[, then a − b < x − y < b − a and so |x − y| < b − a so
taking the sup we get, δ(I) ≤ b − a = `(I). Next, let ε > 0 be given, and let x = a + 2ε and
y = b − 2ε . Then |x − y| = b − a − ε. Consequently, δ(I) ≥ |x − y| = b − a − ε = `(I) − ε.
Since ε is arbitrary, we conclude that δ(I) ≥ `(I). Hence the equality.

n
Pnof IR and let {Ii }i=1 be a finite family of open intervals that cover I.
9. Let I be an interval
Show that `(I) ≤ i=1 `(In ). Hint. Use induction and the connectedness of I.

Solution. The proposition is true for n = 1. Assume that it holds for some n. Let
I ⊂ I1 ∪ · · · ∪ In ∪ In+1 . We distinguish between two cases.

Case 1. In+1 ∩ (I1 ∪ · · · ∪ In ) = ∅. Since I P is connected, then either I ⊂ In+1 or

n+1
I ⊂ I1 ∪ · · · ∪ In . In the first case, `(I) ≤ `(In+1 ) ≤ i=1 `(Ii ). In theP
second case, it follows
from the induction assumption that `(I) ≤ i=1 `(Ii ) and so `(I) ≤ n+1
Pn
i=1 `(Ii ).

Case 2. In+1 ∩ (I1 ∪ · · · ∪ In ) 6= ∅. Then In+1 ∩ Ik 6= ∅ for some k = 1, . . . n. Now recall

that a union of two connected sets having a common point is also connected. Therefore,
In+1 ∪ Ik is connected and therefore is an interval J. Then I ⊂ ∪i6=k,n+1 Ii ∪ J so that I
is covered by n open intervals. Note that by the previous two exercises, `(J) = δ(J) ≤
δ(In+1 ) + δ(Ik ) = `(In+1 ) + `(Ik ). Therefore, by the induction

X X n+1
X
`(I) ≤ `(Ii ) + `(J) ≤ `(Ii ) + `(In+1 ) + `(Ik ) = `(Ii ).
i6=k,n+1 i6=k,n+1 i=1

interval of IR and let {In }∞

10. Let I be an P n=1 be a sequence of open intervals that cover I. Show
that `(I) ≤ ∞ n=1 `(I n ). Hint. Prove first that `(I) = sup{`(K)|K is a compact interval ⊂
I}.

Solution. Let us prove first that `(I) = sup{`(K)|K is a compact interval ⊂ I}. It is
clear that sup `(K) ≤ `(I) because `(K) ≤ `(I). Now we distinguish between two cases.

Case 1. I is bounded. So it is of the form ]a, b[, ]a, b], [a, b[, [a, b]. Let ε > 0 be given and
let Kε = [a + ε, b − ε]. Then `(Kε ) = `(I) − 2ε. Therefore

sup `(K) ≥ `(Kε ) = `(I) − 2ε.

Since ε was arbitrary, we conclude that sup `(K) ≥ `(I). Hence the equality.

Case 2. I is unbounded. So it is of the form |a, ∞[, ] − ∞, a|, or IR. Suppose for example
that I =]a, ∞[ (the other cases being similar). For ε > 0 and n ∈ IN, set Kε,n = [a + ε, n].
Then
sup `(K) ≥ `(Kε,n ) = n − a − ε

4
 for all n ∈ N . This means that sup `(K) = +∞ and so sup `(K) = `(I) = +∞.

Now let K be a compact interval of I. Then K ⊂ ∪n In and so by compactness K ⊂ ∪n∈J In

for some finite subset J of IN. By the previous exercise,

X ∞
X
`(K) ≤ `(In ) ≤ `(In ),
n∈J n=1

and so
∞
X
`(I) = sup `(K) ≤ `(In ).
n=1

11. Let f : [α, β] → IR be a bounded function. The oscillation of f on [α, β] is

ω = sup f (x) − inf f (x).

x∈[α,β] x∈[α,β]

Show that the oscillation of f is also equal to sup{|f (x) − f (y)| | x, y ∈ [α, β]}.

Solution. Let M = supx∈[α,β] f (x) and m = inf x∈[α,β] f (x). Then for every x, y ∈ [α, β],
we have f (x) − f (y) ≤ M − m and f (y) − f (x) ≤ M − m. Therefore |f (x) − f (y)| ≤ M − m
and so
sup |f (x) − f (y)| ≤ M − m.
x,y∈[α,β]

Now let ε > 0. By the properties of the sup and inf there exit x1 , x2 ∈ [α, β] such that
M − ε < f (x1 ) and f (x2 ) < m + ε. Therefore

(M − m) − 2ε < f (x1 ) − f (x2 ) ≤ sup |f (x) − f (y)|.

x,y∈[α,β]

Since ε is arbitrary we get M − m ≤ sup |f (x) − f (y)|. Hence the equality.

x,y∈[α,β]

12. Show that inf(−A) = − sup(A) and sup(−A) = − inf(A).

Solution. Let x ∈ A, then −x ∈ −A so −x ≥ inf(−A), and therefore x ≤ − inf(−A).

Since this is true for all x ∈ A, we conclude that sup A ≤ − inf(−A), that is inf(−A) ≤
− sup A. Let now x ∈ −A, then −x ∈ A so −x ≤ sup A and so x ≥ − sup A. Therefore
inf(−A) ≥ − sup A. Hence the equality. The other equality is obtained similarly or can be
deduced from the first one by noting that −(−A) = A.

13. Let {xn } be a sequence in IR. Show that lim inf(−xn ) = − lim sup(xn ) and that lim sup(−xn ) =
− lim inf(xn ).

Solution. We go back to the definition and use the previous exercise.

lim inf(−xn ) = sup inf (−xk ) = sup(− sup xk ) = − inf sup xk = − lim sup xn .
n≥1 k≥n n≥1 k≥n

The other identity is obtained similarly.

5
14. Show that (whenever the sums are defined)

lim inf xn + lim inf yn ≤ lim inf(xn + yn ) and lim sup(xn + yn ) ≤ lim sup xn + lim sup yn .

Show that equality holds if one of the sequences is convergent.

Solution. Note first that inf k≥n xk ≤ xk for all k ≥ n. Therefore inf k≥n xk + inf k≥n yk ≤
xk + yk . Therefore inf k≥n xk + inf k≥n yk ≤ inf k≥n (xk + yk ). Now take the limit as n → ∞.
The other inequality is obtained in a similar way or deduced from the first one by using the
previous exercise. Now suppose that xn converges to `. By what we proved,

lim inf(xn +yn )−lim(xn ) = lim inf(xn +yn )+lim(−xn ) = lim inf(xn +yn )+lim inf(−xn ) ≤ lim inf(yn )

and so lim inf(xn + yn ) ≤ lim(xn ) + lim inf(yn ).

15. Let X be a compact metric space with no isolated points. Show that X is uncountable.
Conclude that IR is uncountable.

Hint. Show first that given any nonempty open set U of X and any point x ∈ X, there
exists a nonempty open set V ⊂ U such that x ∈
/ V̄ .

Solution. Choose a point y ∈ U different from x; this is possible if x ∈ U because x is

not an isolated point of X and it is possible if x ∈
/ U because U is not empty. Now choose
disjoint open sets W1 and W2 about x and y respectively. Then, the set V := U ∩ W2 is the
desired open set: it is contained in U , it is nonempty because it contains y and its closure
does not contain x.
Suppose now by contradiction that X is countable, so that X = {x1 , x2 , . . .}. Applying
the claim above to the nonempty open set U = X and x = x1 , we get the existence of a
nonempty open set V1 ⊂ X such that x1 ∈ / V¯1 . In general, given Vn−1 open and nonempty,
choose Vn to be a nonempty set such that Vn ⊂ Vn−1 and xn ∈ / V¯n . This procedure gives a
decreasing sequence of nonempty and closed sets Vn . Compactness of X implies that ∩V¯n is
¯
not empty. Let a be a point in this intersection. Then a is equal to some element xm . Then
a ∈ ∩V¯n but a = xm ∈/ V¯m by construction. Contradiction.
Since IR is compact and and has no isolated points, it follows that it is uncountable. But
IR differs from IR by only two elements. Therefore IR is also uncountable. We can also say
that [0,1] is compact and has no isolated points. Therefore it is uncountable. It follows that
IR is uncountable (since it contains [0,1]).