Two Enrichment Classes on Inequalities

given in 2011
Finbarr Holland
March 30, 2011

LECTURE I

1 Introduction
Inequalities or inequations are the lifeblood of the system of real numbers
R, which contains a subset called the set of positive numbers, which is closed
under addition and multiplication, by which is meant that both the sum and
product of two positive numbers is positive. This enables us to order the real
numbers: given a, b R, we say a is bigger than b if a b is positive.
The theory of inequalities has to do with deciding when one real number
is bigger than or equal to another real number, i.e., when their difference is
positive or zero. In these lecture notes we introduce some symbolism, build
up some simple propositions, illustrate these and introduce special named
inequalities, which keep on recurring.

2 Language of Inequalities
The Law of Trichotomy tells us that every real number is either positive,
zero or negative. More precisely, if x R, then one and only one of the
following possibilities holds; (i) x is positive; (ii) x is zero; (iii) x is negative.
In symbols, either (i) x > 0 (read x is greater than zero); (ii) x = 0 (read
x is zero; (iii) x < 0 (read x is less than zero). We say x is non-negative
and write, x 0 to mean that x is either greater than or equal to zero. If
a, b R, we write a b to mean that a b 0. If a, b, x R, we say x

1
is between a and b if either (i) x > a and x < b or (ii) x < a and x > b.
But note that, at most one of these possibilities can occur, and neither may
occur.
If a < b, we denote the set of points between a and b by (a, b), and call it
the open interval with endpoints a, b. Thus

(a, b) = {x R : a < x < b};

Similarly, we define

[a, b) = {x R : a x < b}; (a, b] = {x R : a < x b}; [a, b] = {x R : a x b}.

The set [a, b] stands for the closed interval with endpoints a, b. Hence, for
instance,
(a, b) [a, b) [a, b], (a, b) (a, b] [a, b] R,
and all the inclusions are proper containments. Sometimes its convenient
to introduce the symbols but thats all they are: they are not real
numbersand use them to denote the negative and positive numbers like so:

(, 0) = {x R : x < 0}, (0, ) = {x R : 0 < x}.

Likewise, R = (, ), and

(, 0] = {x R : x 0}, [0, ) = {x R : 0 x}.

3 Elementary facts
The real numbers carry two binary combining operations, called addition
and multiplication. Its important to know how these relate to the order
structure imposed above. The basic fact is axiomatic:

Axiom 1 Suppose a, b R, a > 0 and b > 0. Then a + b > 0 and ab > 0.

Proposition 1 Suppose a, b, c R. If a < b, then a + c < b + c. Conversely,

if a + c < b + c, then a < b.

Proof. Suppose a < b, so that ba > 0. Then (b+c)(a+c) = b+cac =

b a > 0, which means that b + c > a + c. Conversely, if the latter occurs,
then b a = (b + c) (a + c) > 0, i.e., a < b. 

2
Exercise 1 Suppose x R. Prove that x < 0 x > 0.

Proposition 2 Suppose a, b, c R, a < b and b < c. Then a < c

Proof. By hypothesis, b a > 0 and c b > 0. Hence, by Axiom 1,

(b a) + (c b) > 0. But

c a = c b + (b a) = (c a) + (b a) > 0.

In other words, c > a. 

Exercise 2 Suppose a, b, c R, a b and b c. Prove that a c.

Proposition 3 Suppose a, b, c R and a < b. (i)If c > 0, then ac < bc.

(ii)If c = 0, then ac = bc. (iii) If c < 0, then ac > bc.

Proof. By hypothesis, b a > 0. Consider (i). If c > 0, then, by Axiom

1, (b a)c > 0, i.e., bc ac > 0 bc > ac. Thus, (i) holds. Clearly, (ii)
holds. If c < 0, then c > 0 and so, again, by Axiom 1, (c)(b a) > 0,
equivalently, cb + ca > 0, i.e., ac > bc, which implies (iii). 

Theorem 1 Suppose x R. Then x2 0. The inequality is strict unless

x = 0.

Proof. By Trichotomy, either (i) x > 0; or (ii) x = 0; or (iii) x < 0. Consider

each of these in turn. If (i) holds, then, by Axiom 1, x2 = xx > 0. If (ii)
holds, then x2 = 0x = 0. Finally, if x < 0, then x > 0, whence, again by
Axiom 1, x2 = (x)(x) > 0. In summary, x2 0.
Next, if x2 = 0, and x 6= 0, then, by (i) or (iii), x2 > 0, a contradiction.
Thus, x2 = 0 x = 0. 

Theorem 2 Suppose a, b R. If a 0 and b 0, then a+b 0. Moreover,

a + b > 0 unless a = b = 0.

Proof. The first part is an easy consequence of Axiom 1. If a + b = 0, and a

is positive, say, then a + b > b 0, by Proposition 1, which is absurd. 

Theorem 3 Suppose a, b R. Then ab 0, if and only if either (i) a 0

and b 0 or (ii) a 0 and b 0.

Proof. This is an easy consequence of Axiom 1. 

3
Exercise 3 Suppose a, b R. Prove that ab < 0 if and only if either (i)
a > 0 and b < 0 or (ii) a < 0 and b > 0.
Proposition 4 Suppose a, b, c R, and a > 0. Then the minimum value of
the quadratic ax2 + bx + c, x R, is equal to (4ac b2 )/4a.
Proof. Complete the square on x: x R
b 2 b2 b b2 b 4ac b2
ax2 +bx+c = a(x+ ) +ca 2 = a(x+ )2 +c = a(x+ )2 + .
2a 4a 2a 4a 2a 4a
b 2
By Theorem 1, (x + 2a ) 0, and a > 0 by assumption. Hence, by Theorem
b 2
2, a(x + 2a ) 0. Therefore,
b 2 4ac b2 4ac b2
a(x + ) + ,
2a 4a 4a
with equality iff x = b/2a. Hence, the minimum value of the quadratic is
(4ac b2 )/4a, as stated. 
Exercise 4 Suppose a, b, c R, and a < 0. Prove that the maximum value
of the quadratic ax2 + bx + c, x R, is equal to (4ac b2 )/4a.

4 Examples
Example 1 Describe the set of all real x such that x2 + x 0.
Solution. Let S = {x R : x2 + x 0}. Suppose x S. Then x(x + 1) 0.
Hence, either (i) x 0 and x + 1 0 or (ii) x 0 and x + 1 0. Now
(i) holds iff x 0 and x 1, i.e., 1 x 0, equivalently, x [1, 0].
In addition, (ii) holds iff x 0 and x 1 < 0. But no real number can
be simultaneously positive and negative. Thus, (ii) cant happen. It follows
from (i) that S [1, 0]. On the other hand, if x [1, 0], then 1 x 0,
i.e., x + 1 0 and x 0, so that x(x + 1) 0, i.e., x S. In other words,
[1, 0] S. It follows that, S = [1, 0].
Example 2 x R, x2 x + 1 > 0.
Solution. Complete the square on x:
1 1 1 3 3
x2 x + 1 = (x )2 + 1 = (x )2 + ,
2 4 2 4 4
with equality iff x = 1/2. 

4
Example 3 There is no real x such that x2 + 2 x.
Solution. By definition, x2 + 2 x x2 x + 2 0. But, completing the
square on x, we see that
1 1 1 7 7
x2 x + 2 = (x )2 + 2 = (x )2 + .
2 4 2 4 4
In other words, x2 x + 2 is never negative or zero, i.e., the solution set of
the inequality x2 + 2 x is empty, equivalently, {x R : x2 + 2 x} = .

Example 4 Solve the inequality

x3 + 2 x2 , x R.

Solution. If x satisfies the inequality, then x3 x2 + 2 0. Now (?) x3

x2 + 2 = (x + 1)(x2 2x + 2), and (?) x2 2x + 2 > 0. Hence, x 1.
Conversely, if x 1, then x + 1 0. But, as already noted, x2 2x + 2 > 0,
whence
x3 x2 + 2 = (x + 1)(x2 2x + 2) 0.
It follows that
{x R : x3 + 2 x2 } = (, 1].


Example 5 Suppose a, b R. Prove that

T = {x R : x2 + (a + b)x + ab 0} = (, ) \ (c, d),

where c = max(a, b) = min(a, b), d = min(a, b) = max(a, b).

Proof. Clearly, (?) x T (x + a)(x + b) 0. Hence, by Theorem 3,

x T (i) x + a 0 & x + b 0 or (ii) x + a 0 & x + b 0.

But (i) holds iff x max(a, b) = c, while (ii) holds iff x min(a, b) =
d. Thus, T = [c, ) (, d] = R \ (d, c). 
Example 6 Suppose a, b, c, d R. Then

< ad + bc, if a < c and b > d,
ab + cd
> ad + bc, if a < c and b < d.

5
Proof. This follows from the following identity.
ab + cd = ad + bc + (a c)(b d).

Example 7 Suppose a, b, c, d R. Then

< (a + d)(b + c), if a < c and b < d,
(a + b)(c + d)
> (a + d)(b + c), if a < c and b > d.
Proof. This follows from the following identity.
(a + b)(c + d) = (a + d)(b + c) (a c)(b d).


5 Additional Exercises
1. For what real numbers x is it true that x3 + 2x x2 ?
2. For what real numbers x is it true that
1
x + 2?
x
3. Suppose a, b are positive real numbers. Prove that
2 a+b
1 1 ,
a
+ b
2
and determine when the equality is strict.
4. Suppose a, b, c R. Determine the minimum value of (x a)2 + (x
b)2 + (x c)2 .
5. Suppose a, b, c R. Prove that
8a2 b2 c2 (a2 + b2 )(b2 + c2 )(c2 + a2 ).
When does the equality hold?
6. Suppose a, b, c R. Prove that
ab + bc + ca a2 + b2 + c2 .
When does the equality hold?

6
 LECTURE II

6 The absolute value function

The absolute value of a real number is its numerical
value.
Thus, the absolute
value of 2.147 is 2.147; the absolute value of 2 is 2. Generally, we define
the absolute value of any real number x, which we denote by |x|, as follows:

x, if x 0,
|x| =
x, if x 0.

In other words, |x| is the distance between 0 and x on the number line. More
generally, if a, b R, then |a b| is their distance apart.
Since in this way were assigning a unique real number |x| to any real
number x, were actually defining a function on R, which we call the absolute
value function. Sometimes we denote this by abs, for short, so that abs(x) =
|x|, x R.
Note that | x| = |x| 0, x R, and that |x| = 0 x = 0. Hence, the
graph of abs lies above the horizontal axis, and is the union of two straight
lines that emanate from the origin, one with slope 1 and the other with slope
1, to form a V-shaped curve, that is symmetric with respect to the vertical
axis. Notice that this graph is convex, i.e., the line segment joining any two
points on it lies above the corresponding arc of the curve joining them. Also,

|x| x |x|, x R.

How is abs related to addition and multiplication?

Theorem 4 The function abs is multiplicative and sub-additive on R, by

which we mean the following: If x, y are two real numbers, then

|xy| = |x| |y|, |x + y| |x| + |y|.

Proof hint: Consider separately the cases xy 0 and xy 0. 

Exercise 5 Sketch the graphs of the functions y = |x 1|, y = |x + 2| and

y = |x1|+|x+2|. Describe the solution set of the inequality |x1| |x+2|.

Exercise 6 Write out a proof of Theorem 4.

7
Exercise 7 Prove that

|a b| |x a| + |x b|, a, b, x R.

Exercise 8 Prove that

|x + (1 )y| |x| + (1 )|y|, x, y R, [0, 1].

7 The square function

After a linear function, the next simplest function is the quadratic, and the
simplest one of these is the square of the identity function: x x2 . Clearly,
(x)2 = x2 0, x R, and x2 = 0 x = 0. In addition, the graph of
the square function is similar to that of abs, but is smoother; it is U-shaped.
Apart from that, it too lies above the horizontal axis, is symmetric with
respect to the vertical axis, and convex. However, while the square function
is multiplicative, so that (xy)2 = x2 y 2 , x, y R, it is not sub-additive on
R, i.e., the inequality

(x + y)2 x2 + y 2 , x, y R,

is false (?) in general. Moreover, the opposite inequality is also untrue.

However, using convexity, a simple geometric argument (?) shows that

(x + (1 )y)2 x2 + (1 )y 2 , x, y R, [0, 1].

Heres an analytical proof. Expand the LHS and subtract if from the RHS
to get

x2 + (1 )y 2 [2 x2 + 2(1 )xy + (1 )2 y 2 ] = (1 )(x y)2 ,

which is plainly non-negative. Moreover, if 0 < < 1, the inequality is strict

unless x = y. The case = 1/2 deserves special mention.

Proposition 5 If x, y R, then

x2 + y 2
xy ,
2
with equality iff x = y.

8
Proof. The stated result is equivalent to the statement that (x y)2 0,
with equality iff x = y. 

Corollary 1 If x, y R, then

x2 + y 2
|xy| ,
2
with equality iff x = y.

Proof.
|x|2 + |y|2 x2 + y 2
|xy| = |x| |y| = ,
2 2
with equality iff |x| = |y|, and the latter holds (?) iff x = y. 

Corollary 2 If a > 0, b > 0, then

b2
2ab = min{a2 t2 + : t > 0}.
t2
Proof. For all t > 0,
 2
b2 b b
a t + 2 = (at)2 +
2 2
2(at)( ) = 2ab,
t t t

with equality iff at = b/t, i.e., t2 = b/a. 

Example 8 If a, b, c are real numbers, then

8a2 b2 c2 (a2 + b2 )(b2 + c2 )(c2 + a2 ),

with equality iff |a| = |b| = |c|.

Proof. Since 2|ab| a2 + b2 , with equality iff |a| = |b|, etc., and |a|2 = a2 ,
etc., this is easy. 

Example 9 If a, b, c are real numbers, then

ab + bc + ca a2 + b2 + c2 ,

with equality iff a = b = c.

9
Proof. Three applications of Proposition 5 tells us that

a2 + b 2 b2 + c 2 c 2 + a2
ab , bc , ca ,
2 2 2
with equality holding when a = b, b = c, c = a, respectively. Add these
inequalities to get

a2 + b2 + b2 + c2 + c2 + a2
ab + bc + ca = a2 + b 2 + c 2 ,
2
as desired. Clearly, the equality holds if a = b = c.
Conversely, if, for some triple of real numbers a, b, c,

ab + bc + ca = a2 + b2 + c2 ,

then (?)

0 = a2 + b2 2ab + b2 + c2 2bc + c2 + a2 2ca = (a b)2 + (b c)2 + (c a)2 ,

a sum of three non-negative numbers equal to zero. This can only happen if
each of them is zero, whence a b = b c = c a = 0, i.e., a = b = c. 

8 The Cauchy-Schwarz inequality

This is the name given to an ubiquitous inequality of the following sort, but
not the most general form of it.

Theorem 5 Suppose a1 , a2 , a3 , b1 , b2 , b3 are real numbers. Then

(a1 b1 + a2 b2 + a3 b3 )2 (a21 + a22 + a23 )(b21 + b22 + b23 ).

Let
A = a21 + a22 + a23 , B = b21 + b22 + b23 , C = a1 b1 + a2 b2 + a3 b3 .
The claim is that C 2 AB. If either A or B is zero, then C = 0 (?), and
the result holds. So, suppose AB > 0.
First Proof. Consider the quadratic expression

(a1 tb1 )2 + (a2 tb2 )2 + (a3 tb3 )2 = A 2Ct + Bt2 .

10
As a function of the real number t, this is a quadratic that takes only non-
negative values. In particular, its minimum value is nonnegative. According
to Proposition 4, therefore,
(2C)2 AB C 2
0A = ,
4B B
which proves the claim.
Second Proof. Since the square of every real number is nonnegative,
0 (ai bj bi aj )2 = a2i b2j 2(ai bi )(aj bj ) + b2i a2j , i, j = 1, 2, 3.
Add these nine inequalities, to get
0 AB 2C 2 + BA = 2(AB C 2 ),
which is what we want.
Third Proof. If t > 0, then, by Corollary 2,
b2i
2ai bi a2i t + , i = 1, 2, 3.
t
Add these inequalities to find that
B
2C At + , t > 0.
t
Hence, squaring both sides,
B2
4C 2 A2 t2 + 2AB + ,
t2
whence, once more by Corollary 2 (?),
4C 2 2AB + 2Ab = 4AB, C 2 AB.

Example 10 Suppose a, b, c, x, y are real numbers. Then

|a cos x cos y + b cos x sin y + c sin x| a2 + b2 + c2 .
Proof. Apply Theorem 5 with a1 = a, a2 = b, a3 = c, and
b1 = cos x cos y, b1 = cos x sin y, b3 = sin x,
and observe that
b21 + b22 + b23 = cos2 x(cos2 y + sin2 y) + sin2 x = cos2 x + sin2 x = 1.


11
9 Arithmetic, Geometric and Harmonic Means
If a, b are positive real numbers, we define, their arithmetic mean, A(a, b),
geometric mean, G(a, b), and harmonic mean, H(a, b), respectively, by

a+b 2
A(a, b) = , G(a, b) = ab, H(a, b) = 1 .
2 a
+ 1b

Note that H(a, b) is the reciprocal of A(1/a, 1/b). The corresponding means
of a finite number of positive real numbers can be defined similarly: If a, b, c >
0, then, by definition,
a+b+c
3 3
A(a, b, c) = , G(a, b, c) = abc, , H(a, b, c) = 1 1.
3 a
+ 1b + c

In like manner, if a, b, c, d > 0,

a+b+c+d
4 4
A(a, b, c, d) = , G(a, b, c, d) = abcd, H(a, b, c, d) = 1 ,
4 a
+ + 1c + d1
1
b

and so forth for any set of positive numbers.

Proposition 6

H(a, b) G(a, b) A(a, b), a, b > 0.

There is equality throughout iff a = b.

Proof. Well only deal with the statement that G(a, b) A(a, b). The claim
is that
a+b
ab ,
2

with equality iff a = b. Let x = a, y = b, so that x, y > 0 and x2 =
a, y 2 = b. Hence
x2 + y 2 a+b
ab = a b = xy = ,
2 2
by Corollary 1, with equality iff x = y, i.e., a = b. 
We can extend this result first to four numbers and then to three.

12
Proposition 7

H(a, b, c, d) G(a, b, c, d) A(a, b, c, d), a, b, c, d > 0.

There is equality throughout iff a = b = c = d.

Proof. Again, well only prove that G(a, b, c, d) A(a, b, c, d), with equality
iff a = b = c = d. To see this, let x = ab, y = cd. Then (?)

q
G(a, b, c, d) = ab cd
= G(x, y)
A(x, y)
x+y
= (with equality iff x = y)
2
ab + cd
=
2
a+b
+ c+d
2 2
(with equality iff a = b and c = d)
2
a+b+c+d
=
4
= A(a, b, c, d),

with equality throughout iff x = y, &a = b, &c = d. Equivalently (?), iff

a = b = c = d. 
Proposition 8

G(a, b, c) A(a, b, c), a, b, c > 0.

There is equality throughout iff a = b = c.

Proof. Let d = 3 abc and apply the previous proposition, noting that
4

4
abcd = d3 d = d.

Hence
a+b+c+d
d 4d a+b+c+d 3d a+b+c G(a, b, c) A(a, b, c),
4
as desired. Moreover, there is equality throughout iff a = b = c = d. 
Its clear from the proof of Proposition 7 that the geometric mean of eight

13
numbers doesnt exceed their arithmetic mean. We leave it as an exercise to
show that a proof similar to that used in Proposition 8 can be constructed
to prove first that the geometric mean of seven numbers doesnt exceed their
arithmetic mean, next that the geometric mean of six numbers doesnt exceed
their arithmetic mean, and thence that the geometric mean of five numbers
doesnt exceed their arithmetic mean. Indeed, a similar approach can be used
to prove the following general statement, which tells us that the geometric
mean of any set of positive numbers doesnt exceed their arithmetic mean,
and that both means are equal iff the numbers are all equal to one another.
Theorem 6 Let n be a natural number and let a1 , a2 , . . . , an be positive num-
bers. Then
a1 + a2 + + an
G(a1 , a2 , . . . , an ) = n
a1 a2 an = A(a1 , a2 , . . . , an ),
n
with equality iff a1 = a2 = = an .

Example 11 Suppose a, b, c > 0. Then

a b c
3 + + ,
b c a
with equality iff a = b = c.

Proof. The inequality is equivalent to the statement that 3abc a2 bc +

b2 ca + c2 ab. But

(a2 bc)(b2 ca)(c2 ab) = a3 b3 c3 = (abc)3 ,

and so
p3 a2 bc + b2 ca + c2 ab
abc = (a2 bc)(b2 ca)(c2 ab) ,
3
with equality iff a2 bc = b2 ca = c2 ab, i.e., a = b = c. 
Example 12 Suppose 0 x 1. Then 4/27 is the maximum value of
x(1 x)2 .
Proof. The claim is that x(1 x)2 4/27, whenever 0 x 1, and that
x(1 x)2 = 4/27, for at least one value of x in this range. The inequality
certainly holds if x = 0 or x = 1. So, suppose 0 < x < 1, so that x(1 x)2
is the product of three positive numbers. This suggests that it might be

14
profitable to use G(a, b, c) A(a, b, c), with appropriate choices for a, b, c.
Note that the choice a = x, b = c = 1 x wont do. Why? However, we have
better luck with a = x, b = c = (1 x)/2, because then a + b + c = 1, and
we see that

p3

3

3
3 x + 12 (1 x) + 12 (1 x) 3
4
2
x(1 x) = abc 4 4 = ,
3 3
and there is equality when and only when x = 21 (1 x), i.e., x = 31 . Thus,
x [0, 1], x(1 x)2 4/27, and equality occurs here iff x = 1/3. 
Example 13 Suppose 0 < x. Then 3/2 is the minimum value of 2x+1/8x2 .
Proof. The claim is that
3 1
2x + 2 , x > 0,
2 8x
and that equality holds for some positive x. Observing that 2x + 1/8x2 is
three times the arithmetic mean of the three positive numbers, x, x, 1/8x2 ,
we can apply G(a, b, c) A(a, b, c), to get that
r r
x + x + 8x12
 
1 3 1 3 1 3
2x + 2 = 3 3 x.x. 2 = 3 = .
8x 3 8x 8 2
Moreover, equality holds here iff
1
x= 2
, 8x3 = 1, 2x = 1.
8x
3 1
Thus, 2
2x + 8x2
, x > 0, and there is equality iff x = 1/2. 

10 Exercises
1. Show that for all real a, b, x,

|a cos x + b sin x| a2 + b 2 .

2. Suppose a, b, c > 0. Prove that

abc(a + b + c) a4 + b4 + c4 ,

with equality iff a = b = c. [Hint: Use Example 9 or Theorem 5.]

15
 3. Suppose P, Q are two rectangles with dimensions a, b and c, d respec-
tively. Form the rectangle R with dimensions a + c, b + d. Prove that
p
area of P + area of Q area of R.

4. Suppose a, b, c > 0, and d, e, f > 0. Prove that

3
p p
abc + 3 def 3 (a + d)(b + e)(c + f ),

and that there is equality iff

a b c
= = .
d e f

5. Give a geometric interpretation of the previous exercise.

6. Suppose a, b, c > 0. Prove that H(a, b, c) G(a, b, c), and that the
inequality is strict unless a = b = c.
7. Suppose a, b, c > 0. Prove that
1 1 1
9 (a + b + c)( + + ),
a b c
and that the inequality is strict unless a = b = c.
8. (IMO 1984) Prove that
7
0 xy + yz + zx 2xyz ,
27
where x, y, z > 0 and x + y + z = 1.
9. (IMO 2001) Prove that if a, b, c are positive numbers, then
a b c
+ + 1.
a2 + 8bc b2 + 8ca c2 + 8ab

10. (IMO 2004) Suppose a, b, c > 0 and

1 1 1
(a + b + c)( + + ) < 10.
a b c
Prove that a, b, c are the side lengths of a triangle.

16