Chapter 2

Real Numbers

Ah, but my Computations, People say,

Have Squared the Year to human Compass, eh?
If so, by striking from the Calendar
Unborn Tomorrow, and dead Yesterday

Omar Khayyam (1048–1123)

• Mathematical induction is a method of proof used to establish that a given statement is true
for all natural numbers. Let S(n) be a statement about the positive integer n. If
1. S(1) is true and
2. for all k ≥ 1, the truth of S(k) implies the truth of S(k + 1),
then S(n) is true for all n ≥ 1.
Verifying S(1) is true is called the basis step. The assumption that S(k) is true for some k ≥ 1
is called the induction hypothesis. Using the induction hypothesis to prove S(k + 1) is true
is called the induction step. There are variants of mathematical induction used in practice,
for example if one wants to prove a statement not for all natural numbers but only for all
numbers greater than or equal to a certain number b, then
1. Show S(b) is true.
2. Assume S(m) is true for m ≥ b and show that truth of S(m) implies the truth of
S(m + 1).
Another generalization, called strong induction, says that in the inductive step we may assume
not only the statement holds for S(k + 1) but also that it is true for S(m) for all m ≤ k + 1. In

21

A.G. Aksoy, M.A. Khamsi, A Problem Book in Real Analysis, Problem Books in Mathematics,
DOI 10.1007/978-1-4419-1296-1_2, © Springer Science+Business Media, LLC 2010
22 CHAPTER 2. REAL NUMBERS

strong induction it is not necessary to list the basis step, it is clearly true that the statement
holds for all previous cases. The inductive step of a strong induction in this case corresponds
to the basis step in ordinary induction.
• Let A be a nonempty subset of R. The number b is called an upper bound for A if for all
x ∈ A, we have x ≤ b. A number b is called a least upper bound of A if, first, b is an upper
bound for A and, second, b is less than or equal to every upper bound for A. The supremum
of A (also called least upper bound of A) is denoted by sup(A), sup A or lub(A).
If A ⊂ R is not bounded above , we say that sup A is infinite and write sup A = +∞.
• A lower bound for a set A ⊂ R is a number b such that b ≤ x for all x ∈ A. Also b is called
a greatest lower bound if and only if it is a lower bound and for any lower bound c of A,
c ≤ b. The infimum of A (also called greatest lower bound of A) is denoted by inf(A), inf A
or glb(A). If A ⊂ R is not bounded below, we set inf A = −∞.
• Well-Ordering Property: If A is a nonempty subset of N, then there is a smallest element in
A, i.e., there is an a ∈ A such that a ≤ x for every x ∈ A.
• Archimedean Property: If x ∈ Q, then there is an integer n with x < n.
• For each n ∈ N, let In be a nonempty closed interval in R. The family {In : n ∈ N} is called
a nest of intervals if the following conditions hold:
– In+1 ⊂ In for all n ∈ N.
– For each ε > 0, there is some n ∈ N such that |In | < ε.

√
Problem 2.1 Prove that 2 is not rational.

Problem 2.2 Show that two real numbers x and y are equal if and only if ∀ε > 0 it follows
that |x − y| < ε.

Problem 2.3 Use the induction argument to prove that

n(n + 1)
1 + 2 + ··· + n =
2
for all natural numbers n ≥ 1.

Problem 2.4 Use the induction argument to prove that

2n3 + 3n2 + n
12 + 22 + · · · + n2 =
6
for all natural numbers n ≥ 1.
CHAPTER 2. REAL NUMBERS 23

Problem 2.5 Use the induction argument to prove that n3 + 5n is divisible by 6 for all natural
numbers n ≥ 1.

Problem 2.6 Use induction to prove that if 1 + x > 0, then (1 + x)n ≥ 1 + nx for all natural
numbers n ≥ 0. This is known as Bernoulli’s inequality.

Problem 2.7 Consider the Fibonacci numbers {Fn } defined by

F1 = 1 , F2 = 1 , and Fn+2 = Fn+1 + Fn .

Show that √ √
(1 + 5)n − (1 − 5)n
Fn = √ , n = 1, 2, . . . .
2n 5

Problem 2.8 Show by induction that if X is a finite set with n elements, then P(X), the power
set of X (i.e., the set of subsets of X), has 2n elements.

Problem 2.9 Let A be a nonempty subset of R bounded above. Set

B = {−a; a ∈ A} .

Show that B is bounded below and

inf B = − sup A .

Problem 2.10 Let S and T be nonempty bounded subsets of R with S ⊂ T . Prove that

inf T ≤ inf S ≤ sup S ≤ sup T .

Problem 2.11 Let x ∈ R be positive, i.e., x ≥ 0. Show that there exists a ∈ R such that
a2 = x.

Problem 2.12 Let x and y be two real numbers such that x < y. Show that there exists a
rational number r such that x < r < y. (In this case we say Q is dense in R.)Use this result to
conclude that any open nonempty interval (a, b) contains infinitely many rationals.
24 CHAPTER 2. REAL NUMBERS

Problem 2.13 Let x and y be two positive real numbers such that x < y. Show that there
exists a rational number r such that x < r2 < y, without using the square-root function.

Problem 2.14 Let ω ∈ R be an irrational positive number. Set

A = {m + nω : m + nω > 0 and m, n ∈ Z} .

Show that inf A = 0.

Problem 2.15 Show that the Cantor set

 
C = {en }; en = 0 or 1 = {0, 1} × {0, 1} × · · ·

is uncountable.

Problem 2.16 If x ≥ 0 and y ≥ 0, show that

√ x+y
xy ≤ .
2
When do we have equality?

x a
Problem 2.17 Let x, y, a, and b be positive real numbers not equal to 0. Assume that < .
y b
Show that
x x+a a
< < .
y y+b b

Problem 2.18 Let x and y be two real numbers. Show that

|x + y| |x| |y|
≤ + .
1 + |x + y| 1 + |x| 1 + |y|
CHAPTER 2. REAL NUMBERS 25

a
Problem 2.19 Let r ∈ Q ∩ (0, 1). Write r = where a ≥ 1 and b ≥ 1 are coprime natural
b
numbers. Show that there exists a natural number n ≥ 1 such that
1 a 1
≤ < .
n+1 b n
Use this to show that there exist natural numbers n1 , . . . , nk such that
a 1 1
r= = + ··· + .
b n1 nk

Problem 2.20 Let x and y be two different real numbers. Show that there exist a neighborhood
X of x and a neighborhood Y of y such that X ∩ Y = ∅.

Problem 2.21 Show that (a, b) is a neighborhood of any point x ∈ (a, b).

Problem 2.22 (Young Inequality) Prove that for p ∈ (1, ∞), we have xy ≤ p1 xp + 1q y q for
x, y ∈ R+ := {x ∈ R : x ≥ 0}, where q := p−1
p
is the Hölder conjugate of p determined by
1 1
p + q = 1.

Problem 2.23 (Arithmetic and Geometric Means) Prove that for n ∈ N\{0} and xj ∈ R+
for 1 ≤ j ≤ n, one has that 

 n 1
n

n xj ≤ xj .
n
j=1 j=1

Problem 2.24 (Hölder Inequality) For p ∈ (1, ∞) and x = (x1 , x2 , . . . , xn ) ∈ Rn , define

⎛ ⎞1
p

n
|x|p := ⎝ p⎠
|xi | .
j=1

Show that

n
|xj yj | ≤ |x|p |y|q for x, y ∈ Rn .
j=1

Note that in the case of p = q = 2, this reduces to the Cauchy–Schwartz Inequality.

Problem 2.25 (Minkowski Inequality) Show that for all p ∈ (1, ∞), one has |x + y|p ≤
|x|p + |y|p where x, y ∈ Rn .
26 CHAPTER 2. REAL NUMBERS

Problem 2.26 (The Nested Intersection Property) Let {In } be a decreasing
sequence of
nonempty closed intervals in R, i.e., In+1 ⊂ In for all n ≥ 1. Show that In is a nonempty
n≥1
closed interval. When is this intersection a single point?

Problem 2.27 (The Interval Intersection Property) Let {Iα }α∈Γ be a
family of nonempty
closed intervals in R, such that Iα ∩ Iβ = ∅ for any α, β ∈ Γ. Show that Iα is a nonempty
α∈Γ
closed interval.
CHAPTER 2. REAL NUMBERS 27

Solutions

Solution 2.1
√
Assume not. Let r ∈ Q such that r = 2 or r2 = 2. Without loss of generality, we may assume
r ≥ 0. And since r2 = 2, we have r > 0. Since r is rational, there exist two natural numbers n ≥ 1
and m ≥ 1 such that
n
r= .
m
Moreover one may assume that n and m are relatively prime, i.e., the only common divisor is 1.
Since r2 = 2, we get
 n 2 n2
= 2 =2
m m
which implies 2m2 = n2 . Therefore, n2 is even, so it is a multiple of 2. Assume that n is not even,
then n = 2k + 1 for some natural number k. Hence n2 = 4k 2 + 4k + 1 = 2(2k 2 + 2k) + 1. In
other words, if n is not even, then n2 will not be even. Therefore, n is also even. Set n = 2k for
some natural number k. Then n2 = 4k 2 and since n2 = 2m2 , we deduce that m2 = 2k 2 . The same
previous argument will imply that m is also even. So both n and m are even so both are multiples
of 2. This is a contradiction with our assumption that both are relatively prime. Therefore, such
a rational number r does not exist which completes the proof of our statement.

Solution 2.2

This is an if and only if statement, and we need to prove the implications in both directions.
(⇒): If x = y, then |x − y| = 0 and thus |x − y| < ε no matter what ε > 0 is chosen.
:(⇐) We give a proof by contradiction. Assume x = y, then ε0 = |x − y| > 0. However, the
statements
|x − y| = ε0 and |x − y| < ε0
cannot be both true, our assumption is wrong, thus x = y.

Solution 2.3

First note that

1(1 + 1)
1=
2
which implies that the desired identity holds for n = 1. Assume that it holds for n, and let us prove
that it also holds for n + 1. We have

1 + 2 + · · · + (n + 1) = 1 + 2 + · · · + n + (n + 1) .

Using our assumption we get

n(n + 1)
1 + 2 + · · · + (n + 1) = + (n + 1) .
2
Since
n(n + 1) n(n + 1) + 2(n + 1) (n + 1)(n + 2)
+ (n + 1) = = ,
2 2 2
28 CHAPTER 2. REAL NUMBERS

we conclude that the identity is also valid for n + 1. By induction we clearly showed that the above
identity is valid for any natural number n ≥ 1.

Solution 2.4

Since
2+3+1
12 = ,
6
the above identity holds in the case n = 1. Assume that the identity holds for n and let us prove
that it also holds for n + 1. Since

12 + 22 + · · · + (n + 1)2 = 12 + 22 + · · · + n2 + (n + 1)2 ,

the induction assumption implies

2n3 + 3n2 + n
12 + 22 + · · · + (n + 1)2 = + (n + 1)2 .
6
Algebraic manipulations imply

2n3 + 3n2 + n 2n3 + 3n2 + n + 6(n + 1)2

+ (n + 1)2 =
6 6

2n3 + 3n2 + n + 3(n + 1)2 + 3(n + 1)2

=
6

2n3 + 6n2 + 7n + 3 + 3(n + 1)2

= .
6
On the other hand, we have

2(n + 1)3 + 3(n + 1)2 + (n + 1) 2n3 + 6n2 + 6n + 2 + 3(n + 1)2 + (n + 1)

=
6 6

2n3 + 6n2 + 7n + 3 + 3(n + 1)2

=
6
which implies
2(n + 1)3 + 3(n + 1)2 + (n + 1)
12 + 22 + · · · + (n + 1)2 = .
6
So our identity is also valid for n + 1. By induction we clearly showed that the identity is valid for
any natural number n ≥ 1.

Solution 2.5

First take n = 1. Then n3 + 5n = 6 which is a multiple of 6. Assume that n3 + 5n is divisible by

6 and let us prove that (n + 1)3 + 5(n + 1) is divisible by 6. But

(n + 1)3 + 5(n + 1) = n3 + 3n2 + 3n + 1 + 5n + 5 = n3 + 5n + 3(n2 + n) + 6 .

Next note that n2 + n is always even or a multiple of 2. Indeed if n is even, then n2 is also even
and therefore n2 + n is even. Now assume n is odd, then n2 is also odd. Since the sum of two odd
CHAPTER 2. REAL NUMBERS 29

numbers is even we get that n2 + n is even. Hence 3(n2 + n) is a multiple of 6. Our induction
assumption implies that n3 + 5n is a multiple of 6. So n3 + 5n + 3(n2 + n) + 6 is a multiple of 6
which implies (n + 1)3 + 5(n + 1) is a multiple of 6. This completes our proof by induction, i.e.,
n3 + 5n is divisible by 6 (or multiple of 6) for all natural numbers n ≥ 1.

Solution 2.6

It is clear that for n = 0, both sides of the inequality are equal to 1. Now assume that we have
(1 + x)n ≥ 1 + nx and let us prove that (1 + x)n+1 ≥ 1 + (n + 1)x. We have

(1 + x)n+1 = (1 + x)n (1 + x) ≥ (1 + nx)(1 + x)

and (1 + x)n ≥ 1 + nx. Since (1 + nx)(1 + x) = 1 + nx + x + nx2 = 1 + (n + 1)x + nx2 and nx2 ≥ 0,
we get
(1 + x)n+1 ≥ 1 + (n + 1)x + nx2 ≥ 1 + (n + 1)x.
Hence the inequality is also true for n + 1. Therefore, by induction we have (1 + x)n ≥ 1 + nx for
all natural numbers n ≥ 0.

Solution 2.7

The classical induction argument will not work here. The main reason is that in order to reach
Fn+2 one will need to make assumptions about Fn+1 and Fn . Therefore, we will use a strong
induction argument. Indeed, first it is obvious that
√ √
(1 + 5)1 − (1 − 5)1
F1 = F2 = √ .
21 5
Next assume that √ √
(1 + 5)k − (1 − 5)k
Fk = √ , k = 1, 2, . . . , n
2k 5
and let us prove that √ √
5)n+1 − (1 −
(1 + 5)n+1
Fn+1 = √ .
2n+1 5
By the definition of the Fibonacci numbers, we have

Fn+1 = Fn + Fn−1 .

Our induction assumption implies

√ √ √ √
(1 + 5)n − (1 − 5)n (1 + 5)n−1 − (1 − 5)n−1
Fn+1 = √ + √ .
2n 5 2n−1 5
Algebraic manipulations will imply
√ √ √ √
(1 + 5)n − (1 − 5)n + 2(1 + 5)n−1 − 2(1 − 5)n−1
Fn+1 = √
2n 5
or √ √ √ √
2(1 + 5)n + 4(1 + 5)n−1 − 2(1 − 5)n − 4(1 − 5)n−1
Fn+1 = √ .
2n+1 5
30 CHAPTER 2. REAL NUMBERS

Note that √ √ √
(1 + √5)2 = 6 + 2√5 = 2(1 + √5) + 4
(1 − 5)2 = 6 − 2 5 = 2(1 − 5) + 4 .
This easily implies √ √ √
(1 + √5)n+1 = 2(1 + √5)n + 4(1 + √5)n−1
(1 − 5)n+1 = 2(1 − 5)n + 4(1 − 5)n−1 .
From the above equations we get
√ √
(1 + 5)n+1 − (1 − 5)n+1
Fn+1 = √ .
2n+1 5
The induction argument then concludes that
√ √
(1 + 5)n − (1 − 5)n
Fn = √ , n = 1, 2, . . . .
2n 5
Note that the number √
Fn+1 1+ 5
Φ = lim =
n→∞ Fn 2
is known as the golden ratio and is one of the roots of the quadratic equation x2 = x + 1.

Solution 2.8

Note that when n = 0, the set X is the empty set. In this case we have

P(X) = {∅}

with one element. Since 20 = 1, the statement is true when n = 0. Assume that whenever a set
X has n elements, then P(X) has 2n elements. Now let us prove that whenever a set X has n + 1
elements, then P(X) has 2n+1 elements. Indeed, let X be a set with n + 1 elements. Fix a ∈ X
and set Y = X \ {a}. Then Y has n elements. Clearly, we have

P(X) = P(Y ) ∪ Pa (Y )

where  
Pa (Y ) = M ∪ {a}; M ∈ P(Y ) .
The map Ta : P(Y ) → Pa (Y ) defined by

T (M ) = M ∪ {a}

is a bijection. Hence P(Y ) and Pa (Y ) have the same number of elements. Since Y has n elements,
our assumption implies that P(Y ) has 2n elements. Note that P(Y ) and Pa (Y ) have no common
point, i.e.,
P(Y ) ∩ Pa (Y ) = ∅,
so

number of elements of P(X) = number of elements of P(Y ) + number of elements of Pa (Y )

or
number of elements of P(X) = 2n + 2n = 2 · 2n = 2n+1 .
CHAPTER 2. REAL NUMBERS 31

This proves our claim. So by induction we conclude that whenever a set X has n elements, then
P(X) has 2n elements, for any natural number n ≥ 0.

Solution 2.9

Since A is bounded above, there exists m ∈ R such that

∀a ∈ A a ≤ m .

Hence
∀a ∈ A − m ≤ −a
which implies
∀b ∈ B −m≤b.
So B is bounded below. Therefore inf B exists. Let us now complete the proof by showing that
inf B = − sup A. By definition of sup A, we know that

∀ a ∈ A a ≤ sup A .

So
∀a ∈ A − sup A ≤ −a
or
∀b ∈ B − sup A ≤ b .
The definition of inf B implies − sup A ≤ inf B. Next we have

∀b ∈ B inf B ≤ b

which implies
∀b ∈ B − b ≤ − inf B
or
∀ a ∈ A a ≤ − inf B
since A = {−b; b ∈ B}. By the definition of sup A we get sup A ≤ − inf B. Combining this
conclusion with − sup A ≤ inf B, we deduce that

inf B = − sup A .

Note that a similar proof will show that if A is bounded below, then B, defined as above, will be
bounded above. Moreover we will have

sup B = − inf A .

In fact the above proof may be generalized to get the following result: if we set k·A = {k·a; a ∈ A},
for any k ∈ R, then ⎧
⎪
⎪ sup(k · A) = k sup A provided k ≥ 0,
⎨
inf(k · A) = k inf A provided k ≥ 0,
⎪
⎪ sup(k · A) = −k inf A provided k ≤ 0,
⎩
inf(k · A) = −k sup A provided k ≤ 0.
32 CHAPTER 2. REAL NUMBERS

Solution 2.10

It is always the case that inf S ≤ sup S for any bounded nonempty subset of R. So we need only
to prove that inf T ≤ inf S and sup S ≤ sup T . Let x ∈ S. Then x ∈ T since S ⊂ T . So inf T ≤ x
by definition of inf T . This implies that inf T is a lower bound for S because x was taken arbitrary
in S. Since inf S is the greatest lower bound we get
inf T ≤ inf S .
Similarly one can easily show that sup S ≤ sup T .

Solution 2.11

Without loss of generality assume x > 0. Set

A = {a ∈ R; a2 ≤ x} .
Obviously we have 0 ∈ A which means that A is not empty. Next note that A is bounded above.
Indeed, let n ≥ 1 be a natural number such that x ≤ n. We claim that n is an upper bound of A.
Indeed let a ∈ A and assume n < a. In particular, we have 0 < a which implies n2 < a2 . Since
a ∈ A then we have a2 ≤ x which implies n2 < x. But n ≤ n2 which implies n < x, contradiction.
So we must have a ≤ n for any a ∈ A. Since A is bounded above, then sup A exists. Set y = sup A.
Let us prove that y 2 = x which will complete the proof of our problem. Since 0 ∈ A, we get y ≥ 0.
2y + 1
Assume that y 2 < x. So the real number is well defined. Let n ≥ 1 be a natural number
x − y2
such that
2y + 1
≤n,
x − y2
which implies
2y + 1
≤ x − y2 ,
n
or
2y + 1
+ y2 ≤ x .
n
1 1
Since n ≥ 1 we know that 2 ≤ which implies
n n
2y 1 2y + 1
y2 + + 2 ≤ y2 + ≤x,
n n n
or  
1 2
y+ ≤x.
n
1
Hence y + ∈ A, contradicting the fact that y is an upper bound of A. So we must have x ≤ y 2 .
n
Assume that y 2 = x. So we have x < y 2 . In particular we have y > 0. Let n ≥ 1 be a natural
number such that
2y
≤n.
y2 − x
Similar calculations as above will yield
 
1 2
x≤ y− .
n
CHAPTER 2. REAL NUMBERS 33

1
Since y = sup A, there exists a ∈ A such that y − < a. Since
n
1 2y
< 2 ,
y y −x
 2
1 1
we get y − > 0. So y− < a2 which implies x < a2 contradicting the fact that a ∈ A. So
n n
we must have y 2 = x.

Solution 2.12

We have y − x > 0. Since R is Archimedean, there exists a positive integer N ≥ 1 such that
2
N> .
y−x
So N (y − x) > 2 or N x + 2 < N y. Again because R is Archimedean, there exists a unique integer
n such that
n ≤ Nx < n + 1 .
We claim that n + 1 ∈ (N x, N y). Indeed we have

Nx < n + 1 ≤ Nx + 1 < Nx + 2 < Ny ,

n+1 n+1
and since N x < n + 1, we have our conclusion. N ≥ 1 implies that x < < y. Take r =
N N
which completes the proof of the first statement. Next assume that the nonempty interval (a, b)
has finitely many rationals. Then

r∗ = min{r ∈ Q, a < r < b}

exists and is in (a, b), i.e., a < r∗ < b. Using the previous statement, we know that there exists a
rational number q such that a < q < r∗ . Obviously we have q ∈ (a, b) contradicting the definition
of r∗ . So the set {r ∈ Q, a < r < b} is infinite.

Solution 2.13

Set A = {r ∈ Q , 0 ≤ r and x < r2 }. Since A is not empty and bounded below, it has an infimum.
Let m = inf A. Clearly 0 ≤ m. We claim that m2 ≤ x. Assume not. Then x < m2 which implies
m > 0. Let  2 
m −x
ε = min ,m .
2m
It is clear that ε > 0. Then we have 2mε ≤ m2 − x which implies

x ≤ m2 − 2mε < m2 − 2mε + ε2 = (m − ε)2 .

Since m − ε < m, there exists a rational number r ∈ Q such that m − ε < r < m. Note
that r is positive because ε ≤ m. This obviously implies (m − ε)2 < r2 . In particular we have
x < (m − ε)2 < r2 . So r ∈ A which contradicts m = inf A. Hence our claim is valid, that is,
m2 ≤ x. This implies m2 < y. Let
 
y − m2
δ = min ,1 .
2m + 1
34 CHAPTER 2. REAL NUMBERS

Note that δ > 0. Since δ(2m + 1) ≤ y − m2 , and δ ≤ 1, we get

(m + δ)2 = m2 + 2δm + δ 2 ≤ m2 + 2δm + δ = m2 + δ(2m + 1) ≤ y .

Hence, (m + δ)2 ≤ y. Using the characterization of the inf A, we know that there exists r ∈ A such
that m < r < m + δ. So we have r2 < (m + δ)2 ≤ y and since r ∈ A we get

x < r2 < y

which completes the proof of our statement.

Solution 2.14

Since all the elements of A are positive, inf A exists and is positive, i.e., inf A ≥ 0. Assume that
α = inf A > 0. Let us first note that α ∈ A. Assume not, i.e., α ∈ A. Then, there exists x ∈ A
such that α ≤ x < 2α. Since α ∈ A, then we have α < x < 2α. Again by definition of the inf A,
there exists y ∈ A such that α ≤ y < x. Using the same argument as before, we get

α < y < x < 2α .

So x − y < 2α − α = α. Since x − y ∈ A, we have a contradiction with the definition of α. Therefore

α ∈ A. Note that A ∩ (nα, (n + 1)α) = ∅, for any natural number n ≥ 1. Assume not, i.e., there
exists a natural number n ≥ 1 such that A ∩ (nα, (n + 1)α) = ∅. Let x ∈ A ∩ (nα, (n + 1)α). Then
x − nα < α and x − nα ∈ A because of the algebraic properties satisfied by the elements of A. This
will generate another contradiction with the definition of α. Therefore

A = {nα; n = 1, . . . } .

In particular we have ω = n0 α, and 1 = m0 α for some natural numbers n0 ≥ 1 and m0 ≥ 1. This

obviously will imply
ω n0 α n0
= =
1 m0 α m0
or that ω is rational. This is the final contradiction which forces α = 0.

Solution 2.15

Assume not, i.e., C is countable. Hence there exists a map f : N → C which is one-to-one and
onto. Define {xn } as follows:
xn = 1 − f (n)n
where f (n)n is the nth element of the sequence f (n) in C. Clearly {xn } is in C. Also note that
f (n) = {xn } for any n ∈ N. This contradicts the onto property of f .

Solution 2.16

We have
√ √ √
x + y − 2 xy = ( x − y)2 .
Hence
√
x + y − 2 xy ≥ 0
CHAPTER 2. REAL NUMBERS 35

√
which obviously implies x + y ≥ 2 xy or the desired inequality. Moreover we have
√ x+y
xy =
2
√ √ √ √ √
if and only if x + y = 2 xy or ( x − y)2 = 0. This will imply x = y or x = y. So the equality
holds if and only if x = y.

Solution 2.17

First, let us prove the inequality

x x+a
< .
y y+b
We have x(y + b) − y(x + a) = xb − ya = y( xy b − a) = yb( xy − ab ) < 0 because of the assumed
inequality and all the numbers involved are positive. So x(y + b) < y(x + a) or
x x+a
< .
y y+b
The other inequality follows from the same ideas.

Solution 2.18

Note that the fractions involved are always defined since the denominator cannot be 0. First
assume |x + y| ≤ |x|.
Now, since |x+y| ≤ |x|, we get |x+y|+|x||x+y| ≤ |x|+|x||x+y|. So |x+y|(1+|x|) ≤ |x|(1+|x+y|)
which obviously implies
|x + y| |x|
≤ .
1 + |x + y| 1 + |x|
|y|
Since is positive, we get
1 + |y|

|x + y| |x| |y|
≤ + .
1 + |x + y| 1 + |x| 1 + |y|

If |x + y| ≤ |y|, a similar proof will give us the above inequality. Now let us assume max{|x|, |y|} ≤
|x + y|. The triangle inequality gives |x + y| ≤ |x| + |y|. So

|x + y| |x| |y|
≤ + .
1 + |x + y| 1 + |x + y| 1 + |x + y|

Since max{|x|, |y|} ≤ |x + y|, then we have

|x| |y| |x| |y|

+ ≤ + .
1 + |x + y| 1 + |x + y| 1 + |x| 1 + |y|
Hence
|x + y| |x| |y|
≤ + ,
1 + |x + y| 1 + |x| 1 + |y|
which completes the proof of the desired inequality.
36 CHAPTER 2. REAL NUMBERS

Solution 2.19

Using the Archimedean property of the reals, we know that there exists a unique integer m such
that
b
m≤ <m+1.
a
b
Since > 1, we get m ≥ 1. Easy algebra manipulations give
a
1 a 1
< ≤ .
m+1 b m
b 1
If = , then we must have m > 1. In this case take n = m − 1. Otherwise take n = m to get
a m
1 a 1
≤ < .
n+1 b n
To prove the second part we will use the strong induction argument. If a = 1, then the conclusion
is obvious. Assume the conclusion is true for a = 1, . . . , k and let us prove that the conclusion is
also true for a = k + 1. Let b ≥ 1 be a natural number coprime with k + 1 such that k + 1 < b.
Then, the first part implies the existence of a natural number n ≥ 1 such that
1 k+1 1
≤ < .
n+1 b n
1 k+1 1 k+1
If = , then we have nothing to prove. Otherwise assume < . Then
n+1 b n+1 b
k+1 1 1 1
0< − < − <1.
b n+1 n n+1
Set
k+1 1 a∗
r∗ = − = ∗ ∈Q,
b n+1 b
where a∗ and b∗ are natural coprime numbers. Obviously we have a∗ ≤ (k + 1)(n + 1) − b. Since
(k + 1) 1
< , then (k + 1)n < b or (k + 1)(n + 1) − b < k + 1. This implies a∗ < k + 1. Our induction
b n
assumption implies
a∗ 1 1
∗
= + ··· +
b n1 nl
for some natural numbers n1 , . . . , nl . Hence
k+1 1 1 1
= + + ··· + .
b n + 1 n1 nl
Therefore, by induction we have proved our initial claim.

Solution 2.20

Since x = y, then we have

|x − y|
ε= >0.
3
CHAPTER 2. REAL NUMBERS 37

Set
X = (x − ε, x + ε) and Y = (y − ε, y + ε) .
Let us check that X ∩ Y = ∅. Assume not. And let z ∈ X ∩ Y . So

|z − x| < ε and |z − y| < ε .

Hence
2|x − y|
|x − y| = |x − z + z − y| ≤ |x − z| + |z − y| < ε + ε = < |x − y| .
3
This contradiction proves our claim.

Solution 2.21

Let x ∈ (a, b). Since x = a and x = b, we have

1
ε= min(|x − a|, |b − x|) > 0 .
3
Let us show that (x − ε, x + ε) ⊂ (a, b). Indeed, we have
1
x+ε=x+ min(|x − a|, |b − x|) < x + |x − b| = x + b − x = b
3
because x < b. Similarly we have
1
x−ε=x− min(|x − a|, |b − x|) > x − |x − a| = x − (x − a) = a .
3
These two inequalities obviously imply

(x − ε, x + ε) ⊂ (a, b),

which implies that (a, b) is a neighborhood of any point x ∈ (a, b).

Solution 2.22

Suppose x, y ∈ (0, ∞). From the concavity of the logarithim function and the fact that p1 + 1q = 1,
 p q

this implies that log xp + yq ≥ p1 log xp + 1q log y q = log x + log y = log xy. Since the exponential
function is increasing and elog x = x for all x, the claimed inequality follows.

Solution 2.23

We use induction to prove. For n = 1, the above inequality is true. Suppose it is true for some
n ≥ 1. Then  ⎛ ⎞ n
n+1 n+1
 1 
n
1

n+1
xj ≤ ⎝ xj ⎠ (xn+1 ) n+1 .
n
j=1 j=1

To the right-hand side of this inequality, apply the Young Inequality (previous problem) with
⎛ ⎞ n
n+1
1 n
1
x= ⎝ xj ⎠ , y = (xn+1 ) n+1 , and p = 1 + n1 . Then
n
j=1
38 CHAPTER 2. REAL NUMBERS

1  1 
n n+1
1 1 1
xy ≤ xp + y q = xj + (xn+1 ) = xj .
p q n+1 n+1 n+1
j=1 j=1

Solution 2.24

Without loss of generality, take x = 0 and y = 0. From the Young Inequality,

|xj | |yj | 1 |xj |p 1 |yj |q
· ≤ · + · , for 1 ≤ j ≤ n.
|x|p |y|q p |x|pp q |y|qq

Summing this inequality over j yields

#n
j=1 |xj yj | 1 1
≤ + =1
|x|p |y|q p q

and therefore the desired inequality follows.

Solution 2.25

Applying the triangle inequality, we get


n 
n 
n
|x + y|pp = |xj + yj |p−1 · |xj + yj | ≤ |xj + yj |p−1 · |xj | + |xj + yj |p−1 · |yj |.
j=1 j=1 j=1

Thus the Hölder Inequality implies

⎛ ⎞1 ⎛ ⎞1
q q

n n p
|x + y|pp ≤ |x|p ⎝ |xj + yj |p⎠
+ |y|p ⎝ |xj + yj |p⎠
= (|x|p + |y|p ) |x + y|pq .
j=1 j=1

p
If x + y = 0, then the inequality is true. Otherwise, divide both sides of the inequality by |x + y|pq
p− pq p
to get |x + y|p ≤ |x|p + |y|p . Since p − q = 1, the claim follows.

Solution 2.26

Set In = [an , bn ]. Then our assumption on {In } implies

an ≤ an+1 ≤ bn+1 ≤ bn

for all n ≥ 1. Since In ⊂ I1 , both {an } and {bn } are bounded. Set

a∞ = sup{an ; n ≥ 1} and b∞ = inf{bn ; n ≥ 1} .

From the inequalities an ≤ an+1 ≤ bn+1 ≤ bn , for all n ≥ 1, we get

an ≤ a∞ ≤ b∞ ≤ bn

for all n ≥ 1. Indeed, for any n, m ≥ 1, we have an ≤ bm . This follows from the fact that {an } is
increasing and {bn } is decreasing. By definition of the supremum and infimum, we get a∞ ≤ b∞ .
CHAPTER 2. REAL NUMBERS 39

The inequalities an ≤ a∞ and b∞ ≤ bn follow easily. Next let us complete the proof by showing
that

In = [a∞ , b∞ ] .
n≥1

above inequalities we have [a∞ , b∞ ] ⊂ [an , bn ] = In for all n ≥ 1. Hence we are only left
From the

to prove In ⊆ [a∞ , b∞ ]. So let x ∈ In for all n ≥ 1. Let us prove that x ∈ [a∞ , b∞ ]. Since
n≥1
x ∈ In , we get an ≤ x ≤ bn for all n ≥ 1. By definition of the supremum and infimum, we get
a∞ ≤ x ≤ b∞ , or x ∈ [a∞ , b∞ ]. Clearly the intersection is a single point if a∞ = b∞ . This happens
if and only if bn − an goes to 0 as n goes to ∞.

Solution 2.27

Let us set Iα = [aα , bα ]. First let us show that the condition satisfied by the intervals implies
aα ≤ bβ for any α, β ∈ Γ. Indeed, fix α, β ∈ Γ, then consider x ∈ Iα ∩ Iβ because Iα ∩ Iβ = ∅.
Hence aα ≤ x ≤ bβ which implies the desired inequality. This will force the sets {aα } and {bα } to
be bounded. Set
a∞ = sup{aα ; α ∈ Γ} and b∞ = inf{bα ; α ∈ Γ} .


Following the same argument in the previous problem one will easily check that Iα = [a∞ , b∞ ].
α∈Γ
http://www.springer.com/978-1-4419-1295-4