You are on page 1of 26

Contents

1 Problem Solving Tricks 3


1.1 The One-Trick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Cauchy’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Transformation between expressions . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Algebraic Manipulations 5
2.1 Summation Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Geometric Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4.1 Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4.2 cancelling the common factor . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Summation Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.6 Homogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.6.1 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Recurrence 9
3.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4 Inequalites 11
4.1 Cauchy’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Cauchy’s Inequality in Inner product Spaces . . . . . . . . . . . . . . . . . . . . . . 14
4.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.4 Bergstrom’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.4.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Inequalities with Geometric Interpretations 19


5.1 Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Miscellaneous 21
6.1 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

7 Solutions/Hints 23
7.1 Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.1.1 Cauchy-Schwartz Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.1.2 Bergstrom’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
7.1.3 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
7.2 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1
2 CONTENTS
Chapter 1

Problem Solving Tricks

1.1 The One-Trick


1.1.1 Cauchy’s
√ 1
a1 + a2 + · · · + an ≤ n(a21 + a22 + · · · + a2n ) 2
p
x + y = x · 1 + 1 · y ≤ (x2 + 1)(y 2 + 1)

1.2 Transformation between expressions


If a, b, c ≥ 1, consider the following expression
p p p ab + bc + ca
a2 − 1 + b2 − 1 + c 2 − 1 ≤
2
Note that p  p 
a2 b2 = ( a2 − 1)2 + 1 ( b2 − 1)2 + 1

Now we can use the Cauchy’s Inequality with the 1-trick to get
r  p 
p p p
2
a −1·1+1· b −1≤ 2 ( a2 − 1)2 + 1 ( b2 − 1)2 + 1 = ab

repeated multiplication will solve the inequality.

3
4 CHAPTER 1. PROBLEM SOLVING TRICKS
Chapter 2

Algebraic Manipulations

2.1 Summation Notation


• The distributive law of double summation:
n X
X n n
X n
X
ai bj = ai bj
i=1 j=1 i=1 j=1

2.2 Patterns

1 + n2 + n4 = (n2 + 1)2 − n2 = (n2 − n + 1)(n2 + n + 1)


1 2
=
1 + 2 + 3 + ··· + n n(n + 1)

x+1 2
=1+
x−1 x−1
x+n 2n
=1+
x−n x−n

X
a + b = (a + b)(b + c)(c + a) = (ab + bc + ca)(a + b + c) − abc
cyclic
X X X
a − b = (a − b)(b − c)(c − a) = (a2 b + b2 c + c2 a) − (ab2 + bc2 + ca2 ) = a2 b − ab2
cyclic cyclic cyclic

(1 − x)(1 − y)(1 − z) = 1 − (x + y + z) + xy + yz + xz − xyz


(1 − ax)(1 − ay)(1 − az) = 1 − a(x + y + z) + a2 (xy + yz + xz) − a3 xyz
especially useful when x + y + z = 1 as it reduces to xy + yz + xz − xyz

1 1 1
= −
k(k + 1) k k+1
 
1 1 1 1
= −
k(k + m) m k k+m
 
1 1 1 1
= −
k(k + 1)(k + 2) 2 k(k + 1) (k + 1)(k + 2)
 
1 1 1 1
= −
k(k + 1)(k + 2) · · · (k + n) n k(k + 1) · · · (k + n − 1) (k + 1)(k + 2) · · · (k + n)
Note that k can be any expression for example
 
1 1 1 1
= −
(n2 − n + 1)(n2 + n + 1) 2n n(n − 1) + 1 n(n + 1) + 1

5
6 CHAPTER 2. ALGEBRAIC MANIPULATIONS

here we have used  


1 1 1 1
= −
k(k + m) m k k+m
with k = n2 − n + 1 and m = 2n

2.3 Geometric Interpretation


1. whenever you have a polynomial in a problem consider its graphical symmetry. For example
in the problem
12 22 92
+ + · · · +
12 − 10 + 50 22 − 20 + 50 92 − 90 + 50
Notice that the polynomial in the denominator, x2 − 10x + 50 is symmetric in the range
1 . . . 9 and will have the same value for the pairs (1, 9), (2, 8), . . . , (6, 7)

2.4 Numbers
2.4.1 Calculations
1. Do approximate calculations when exact not needed
 2  2
75 37
− + < 62 + 32 = 45
13 13

2.4.2 cancelling the common factor


Given the equation
6(x − 1)
= x2 − 1
x
Solve for x. If you cancel out the common factor x − 1 then you miss out one of the solutions to
the equation. So you should solve for two cases, x = 1 and x 6= 1.

2.5 Summation Notation


Definition 2.5.1. Let P(x, y, z) be a three variable function of x, y, z then
X
P(x, y, z) = P(x, y, z) + P(y, z, x) + P(z, x, y)
cyclic

and
X
P(x, y, z) = P(x, y, z) + P(x, z, y) + P(y, x, z) + P(y, z, x) + P(z, x, y) + P(z, y, x)
sym

2.6 Homogeneity
Definition 2.6.1. A function or an expression f (a1 , a2 , . . . , an ) is said to be homogeneous of
degree k if and only if there exists real number k such that for every t > 0 we have

tk · f (a1 , a2 , . . . , an ) = f (ta1 , ta2 , . . . , tan )

Consider the inequality


a b c
√ +√ +√ ≥1
a2 + 8bc b2 + 8ca c2 + 8ab
which we can rewrite as X a
f (a, b, c) = √ −1
cyc a2 + 8bc
note that
f (a, b, c) = t0 f (a, b, c) = f (ta, tb, tc)
here f is a homogeneous expression of degree 0.
2.6. HOMOGENEITY 7

2.6.1 Advantages
Assume that
a+b+c=m
where m > 0 then
a b c
+ + =1
m m m
Now let a0 = a
m, b0 = b
m, c0 = c
m. Then

a0 + b0 + c0 = 1

But the homogeneity of degree 0 tells us that


a b c
f (a0 , b0 , c0 ) = f ( + + ) = f (a, b, c)
m m m
here we have used t = 1/m. Hence if we prove that f (a0 , b0 , c0 ) ≥ 0 then this is equivalent to
f (a, b, c) ≥ 0 Thus we can assume without loss of generality that

a+b+c=1

We can also assume other things, like


• a = 1: Just take t = 1/a

• abc = 1: If abc = m take t = 1/ 3 m

• ab + bc + ca = 1:
• a2 + b2 + c2 = 1:
Note that the number on the right-hand side of these assumptions doesn’t have to be 1.
8 CHAPTER 2. ALGEBRAIC MANIPULATIONS
Chapter 3

Recurrence

3.1 Problems
• Solve the recurrence relation
Tn = mTn−1 + c

with T0 = v0
Solution:

Tn = mTn−1 + c
c c
= m(Tn−1 + )−
m−1 m−1
c c
Tn + = m(Tn−1 + )
m−1 m−1
c c
Setting Un = Tn + m−1 and U0 = v0 + m−1 gives the solution
 
c
Un = mn
m−1

putting back the value of Tn


 
n c
Un = m v0 +
m−1
 
c n c
Tn + = m v0 +
m−1 m−1
 
n c c
Tn = m v0 + −
m−1 m−1
n c n
Tn = v0 m + (m − 1)
m−1
1 − mn
 
n
Tn = v0 m + c
1−m

9
10 CHAPTER 3. RECURRENCE
Chapter 4

Inequalites

4.1 Cauchy’s Inequality


Theorem 4.1.1. Let (a1 , a2 , . . . , an ) and (b1 , b2 , . . . , bn ) be two sequences of real numbers, then

(a1 b1 + · · · + an bn )2 ≤ (a21 + · · · + a2n )(b21 + · · · + b2n )

or in summation notation !2 ! !
n
X n
X n
X
ai bi ≤ a2i b2i
i=1 i=1 i=1

with equality if and only if the sequences (a1 , a2 , . . . , an ) and (b1, b2, . . . , bn) are proportional,
i.e., there is a constant λ such that ak = λbk for each k ∈ {1, 2, ..., n}.

4.1.1 Proofs
Proof 1
Titu Andreescu and Bogdan Enescu give an elegant – and memorable – proof of the Cauchy-
Schwarz inequality among the gems in their Mathematical Olympiad Treasures (Birkhauser,
2003). Nominally, the proof is inductive, but what I like so much about it is that the induction
step comes as close to being “computation free” as one can imagine.

Proof. The proof begins with a simple lemma that is useful in its own right.
Lemma 4.1.2. For real a and b and for x > 0, and y > 0, one has

(a + b)2 a2 b2
≤ +
x+y x y

Just by expansion and factorization one finds that it is simply a restatement of (ay −bx)2 ≥ 0
but the inequality is self-generalizing. For example, if we replace b by b + c and replace y by
y + z then we find
(a + b + c)2 a2 (b + c)2 a2 b2 c2
≤ + ≤ + +
x+y+y x y+z x y z
Repeating this step then gives the general relationship

(a1 + a2 + · · · + an )2 a2 a2 a2
≤ 1 + 2 + ··· + n
x1 + x2 + · · · + xn x1 x2 xn
from which we can see Cauchy-Schwarz inequality at a glance. One simply sets

ak = αk βk and xk = βk2

Proof 2

11
12 CHAPTER 4. INEQUALITES

Proof. Using the distributive law of double summation:


n X
X n n
X n
X
ai bj = ai bj
i=1 j=1 i=1 j=1

n X
X n n X
X n n X
X n n X
X n
(ai bj − aj bi )2 = a2i b2j + a2j b2i − 2ai bj aj bi
i=1 j=1 i=1 j=1 i=1 j=1 i=1 j=1
 
n
X n
X n
X n
X Xn n
X
= a2i b2j + b2i a2j − 2  ai bi aj bj 
i=1 j=1 i=1 j=1 i=1 j=1
n n n n n n
!
X X X X X X
= a2i b2i + b2i a2i −2 ai bi a i bi
i=1 i=1 i=1 i=1 i=1 i=1
n n n
!2
X X X
=2 a2i b2i −2 ai bi
i=1 i=1 i=1

Since the left hand side is greater or equal to zero we have the desired result. If ak = λbk for
each k ∈ {1, 2, ..., n} then
ai bj − aj bi = λbi bj − λbj bi = 0
and equality will hold.

Proof 3
If n ∈ N and a1 , . . . , an and b1 , . . . , bn are real numbers, then

(a1 b1 + · · · + an bn )2 ≤ (a21 + · · · + a2n )(b21 + · · · + b2n )

Moreover, if not all of the bj = 0, then equality holds if and only if a1 = sb1 , . . . an = sbn for
some s ∈ R

Proof. We define a function F such that

F (t) = (a1 − tb1 )2 + · · · + (an − tbn )2

Note that F (t) ≥ 0 for all t ∈ R. Expanding the squares we get

F (t) = At2 − 2Bt + C ≥ 0

where
A = b21 + · · · + b2n , B = a1 b1 + · · · + an bn C = a21 + · · · + a2n

Since the quadratic function F (t) is nonnegative it cannot have two distinct real roots and
the determinant
∆ = (−2B)2 − 4AC = 4(B 2 − AC) ≤ 0
. Thus we must have B 2 ≤ AC which proves our assertion. Note that F (t) = 0 only when
all an − tbn ’s are equal to 0.

Proof 4
We will now use induction to prove Cauchy’s inequality:

Proof. Consider the Cauchy’s inequality for n = 2 which says

(a1 b1 + a2 b2 )2 ≤ (a21 + a22 )(b21 + b22 )

after cancellations and collecting terms to one side we get

0 ≤ (a1 b2 )2 − 2(a1 b2 )(a2 b1 ) + (a2 b1 )2 = (a1 b2 − a2 b1 )2

the nonnegativity of this term confirms the truth of the inequality for n = 2
4.1. CAUCHY’S INEQUALITY 13

(a1 b1 + · · · + an bn )2 ≤ (a21 + · · · + a2n )(b21 + · · · + b2n )


(a1 b1 + · · · + an bn ) + an+1 bn+1 ≤ (a21 + · · · + a2n )1/2 (b21 + · · · + b2n )1/2 + an+1 bn+1
Now for n = 2 Cauch’s inequality is of the form:
αβ + cd ≤ (α2 + c2 )1/2 (β 2 + d2 )1/2
Taking α = (a21 + · · · + a2n )1/2 , β = (b21 + · · · + b2n )1/2 , c = an+1 and d = bn+1 we get
a1 b1 + · · · + an bn + an+1 bn+1 ≤ (a21 + · · · + a2n )1/2 (b21 + · · · + b2n )1/2 + an+1 bn+1
≤ (a21 + · · · + a2n + a2n+1 )(b21 + · · · + b2n + b2n+1 )
Which completes the induction step.

Proof 5
Suppose that we have two sequences such that

X ∞
X
a2k < ∞ and b2k < ∞
k=1 k=1

We know that

0 ≤ (x − y)2 = x2 − 2xy + y 2
which gives us

1 2 1 2
xy ≤ x + y
2 2
Substituting x = |ak | and y = |bk | and summing over all k we get
∞ ∞ ∞
X 1X 2 1X 2
|ak bk | ≤ ak + bk
2 2
k=1 k=1 k=1

The additive bound (1.6) has two terms on the right-hand side, and Cauchy’s inequality has
just one. Thus, as a first step, we might look for a way to combine the two terms of the
additive bound, and a natural way to implement this idea is to normalize the sequences {ak }
and {bk } so that each of the right-hand sums is equal to one.

! 12 ∞
! 12
X X
2 2
âk = ak / ak and b̂k = bk / bk
k=1 k=1

which are normalized in the sense that


∞ ∞ ∞
( !)
X X X
â2k = a2k / a2k =1
k=1 k=1 k=1

and
∞ ∞ ∞
( !)
X X X
b̂2k = b2k / b2k =1
k=1 k=1 k=1

Substituting in the inequality


∞ ∞ ∞
X 1X 2 1X 2
âk b̂k ≤ âk + âk = 1
2 2
k=1 k=1 k=1

In terms of the original values


∞ 


! 12   ∞
! 12 
X X  X 
ak / a2k bk / b2k ≤1
  
k=1 k=1 k=1

which give us the Cauchy’s inequality


∞ ∞
! 12 ∞
! 21
X X X
ak bk ≤ a2k b2k
k=1 k=1 k=1
14 CHAPTER 4. INEQUALITES

4.2 Cauchy’s Inequality in Inner product Spaces


It is an interesting fact that in every inner product space one has the inequality:
1 1
hv, wi ≤ hv, vi 2 hw, wi 2
This reduces to the standard form of Cauchy’s Inequality in Rn if we use the standard inner product
on Rn
Theorem 4.2.1. For any real inner product space (V, h·, ·i), one has for all v and w in V that
1 1
hv, wi ≤ hv, vi 2 hw, wi 2
moreover, for nonzero vectors v and w, one has
1 1
hv, wi = hv, vi 2 hw, wi 2 if and only if v = λw
for a nonzero constant λ
Proof. We use the same strategy we used with in the proof of the infinte case. The analogy with
(x − y)2 ≥ 0 is
hv − w, v − wi ≥ 0
expanding this we get
1 1
hv, vi + hw, wi
hv, wi ≤
2 2
Using normalization we want to replace v and w with related terms that reduce the right side
fo the bound to 1. Since our inner product version of Cauchy’s inequality holds trivially if either v
or w is equal to zero, we may assume that hv, vi and hw, wi are both nonzero and define normalized
variables (unit vectors)
v w
v̂ = 1 and ŵ = 1
hv, vi 2 hw, wi 2
Substitution gives us
hv̂, ŵi ≤ 1
or
v w
h 1 , 1 i ≤ 1
hv, vi 2 hw, wi 2
which simplifies to
1 1
hv, wi ≤ hv, vi 2 hw, wi 2
If equality holds in the above equation then
hv̂, ŵi = 1
and we have:
hv̂ − ŵ, v̂ − ŵi = hv̂, v̂i − 2hv̂, ŵi + hŵ, ŵi = 2 − 2 = 0
which gives us v̂ − ŵ = 0 or v̂ = ŵ or
1
1 w hv, vi 2
v = hv, vi 2 · 1 = 1 ·w =λ·w
hw, wi 2 hw, wi 2

In the standard book proof like the one given below the normalization is performed at the
beginning so it appears magical.

Proof.
√ We rescale u and v to new vectors which have the same√length; namely √ ( v · v)u and
( u · u)v. We take the difference of these two vectors: w = ( u · u)v − ( v · v)u. We now
compute the dot product of this vector with itself:
√ √ √ √
0 ≤ w · w = (( u · u)v − ( v · v)u) · (( u · u)v − ( v · v)u)
√ √
= (u · u)(v · v) − 2 u · u v · v(u · v) + (v · v)(u · u)
√ √
= 2(u · u)(v · v) − 2 u · u v · v(u · v)
√ √
2 u · u v · v(u · v) ≤ 2(u · u)(v · v)
√ √
(u · v) ≤ u · u v · v
4.3. PROBLEMS 15

4.3 Problems
Prove using Cauchy-Schwartz

1. √ 1
a1 + a2 + · · · + an ≤ n(a21 + a22 + · · · + a2n ) 2

2. s r r r
 
1 1 1 b+c−a c+a−b a+b−c
(a + b + c) + + ≥ + +
a b c a b c

3.
n n
! 21 n
! 21
2 4
X X X
ak ≤ |ak | 3 |ak | 3

k=1 k=1 k=1

4. Suppose that pj ≥ 0 for all j = 1, 2, . . . , n and p1 + p2 + · · · + pn = 1. If aj and bj are


nonnegative real numbers that satisfy the termwise bound 1 ≤ aj bj for all j = 1, 2, . . . , n
then X n n
 X 
1≤ pj aj p j bj
j=1 j=1

5. !4 !2
n
X n
X n
X n
X
ak bk ac ≤ a2k b4k c4k
k=1 k=1 k=1 k=1

6. !2
n
X n
X n
X n
X
ak bk ac ≤ a2k b2k c2k
k=1 k=1 k=1 k=1

7. Show that for all positive x, y, z one has


 1/2  1/2  1/2
x+y x+z y+z
S= + + ≤ 61/2
x+y+z x+y+z x+y+z

8. Show that for all positive x, y, z one has


 2
y2 z2

x
x+y+z ≤2 + +
y+z x+z x+y

9. If a1 , a2 , . . . , a2019 are positive real numbers such that a1 + a2 + · · · + a2019 = 2019, then find
the minimum value of the following expression
1 1 1
+ + ··· +
a1 a2 a2019

10. If a, b, c ≥ 1, prove that


p p p ab + bc + ca
a2 − 1 + b2 − 1 + c2 − 1 ≤
2
Generalize!

11.
(a1 + a2 + · · · + an )2 a2 a2 a2
≤ 1 + 2 + ··· + n
x1 + x2 + · · · + xn x1 x2 xn

12. Note that f (x) = cos(βx) satisfies the identity f 2 (x) = 21 (1 + f (2x)), and show that if pk ≥ 0
for 1 ≤ k ≤ n and p1 + p1 + · · · + pn = 1 then
n
X 1
g(x) = pk cos(βk x) satisfies g 2 (x) ≤ (1 + g(2x))
2
k=1
16 CHAPTER 4. INEQUALITES

13. Suppose that pk > 0 for 1 ≤ k ≤ n and p1 + p1 + · · · + pn = 1. Show that one has the bound
n  2
X 1
pk + ≥ n3 + 2n + 1/n,
pk
k=1

and determine the necessary and sufficient conditions for equality to hold.
14. Prove that for all rea, x, y, α and β one has

(5αx+αy + βx + 3βy)2
≤ (5α2 + 2αy + 3β 2 )(5x2 + 2xy + 3y 2 )

More precisely, show that the bound is an immediate corollary of the Cauchy-Schwarz in-
equality provided that one designs a special inner product h·, ·i for the job.
15. For 0 ≤ x < 1 show that
∞ ∞
! 21
X
k 1 X
ak x ≤ √ a2k
k=0
1 − x2 k=0

16. verify that


n n
! 21
X ak √ X
< 2 a2k
k
k=1 k=1

17. verify that


n n
! 21
a 1
√ k
X X
< (log 1) 2 a2k
k=1
n+k k=1

18. verify that


n    2 1 n
! 12
X n 2n X
ak < a2k
k n
k=0 k=0

19. Many problems of mathematical analysis depend on the discovery of bounds which are
stronger than those one finds with the direct appli- cation of Cauchy’s inequality. To illustrate
the kind of opportunity one might miss, show that for any real numbers aj , j = 1, 2 . . . , n,
one has the bound
n n n
X 2 X 2 X
j
a2j


aj

+
(−1) aj ≤ (n + 2)
j=1 j=1 j=1

Here the direct application of Cauchy’s inequality gives a bound with 2n instead of the value
n + 2, so for large n one does better by a factor of nearly two.
20. Let h·, ·i denote an inner product on the vector space V and suppose that x1 , x2 , . . . , xn and
y1 , y2 , . . . , yn are sequences of elements of V . Prove that one has the following vector analog
of Cauchy’s inequality:
n
X n
X  21  X
n  21
hxj , yj i ≤ hxj , xj i hyj , yj i
j=1 j=1 i=1

21. Show that for each rectangular array {cjk : 1 ≤ j ≤ m, 1 ≤ k ≤ n}, and each pair of sequences
{xj : 1 ≤ j ≤ m} and {yj : 1 ≤ k ≤ n}, we have the bound
 12  X  12
m X
n
√ m n
X X
2 2

cjk xj yk ≤ RC
|xj | |yk |
j=1 k=1 j=1 k=1

where R and C are the row sum and column sum maxima defined by
n
X m
X
R = max |cjk | and C = max |cjk |
j k
k=1 j=1

Note that in the extreme case when n = m, cjk = 0 for j 6= k and cjk = 1 for 1 ≤ j ≤ n,
Schur’s Lemma simple recovers Cauchy’s inequality.
4.4. BERGSTROM’S INEQUALITY 17

22. Let v and w be elements of the inner product space (V, h·, ·i) and consider the the quadratic
polynomial defined for t ∈ R by

p(t) = hv + tw, +v + twi

Observe that this polynomial is nonnegative and use what you know about the solution of the
quadratic equation to prove the inner product version of Cauchy’s inequality. Also, examine
the steps of your proof to establish the conditions under which the case of equality can apply.
23. Show that if {ajk : 1 ≤ j ≤ m, 1 ≤ k ≤ n} is an array of real numbers then one has
Xm X n 2 Xm X n 2  X m X n 2 m X
X n
m ajk + n ajk ≤ ajk + mn a2jk
j=1 k=1 j=1 k=1 j=1 k=1 j=1 k=1

Moreover, show that equality holds here if and only if there exist αj and βk such that
ajk = αj + βk for all 1 ≤ j ≤ m and 1 ≤ k ≤ n
24. Show that for any nonnegative aij , bij , cij with 1 ≤ i, j, k ≤ n one has the triple product
inequaity
n X n  21  X
n  12  X
n  12
X 1 1 1
aij bij cij ≤
2 2 2
aij bjk cki
i,j,k=1 i,j=1 j,k=1 k,i=1

25. Show that


p q q
(x1 + y1 )2 + (x2 + y2 )2 + · · · + (xn + yn )2 ≤ x21 + x22 + · · · + x2n + y12 + y22 + · · · + yn2

4.4 Bergstrom’s Inequality


The inequality below is often considered as a special form (Engel’s form, after A. Engel) of Cauchy-
Schwarz (Cauchy-Bunyakowski-Schwarz) inequality, but due to its numerous applications won a
designation of its own. It was referred to in the classic book Inequalities by E. F. Beckenbach and
R. Bellman, (Springer, 1961). In recent years it became better known as Titu’s Inequality in honor
of Titu Andreescu.
Theorem 4.4.1. If xk ∈ R and ak > 0, k ∈ {1, 2, . . . , n}, then

a21 a2 a2 (a1 + a2 + · · · + an )2
+ 2 + ··· + n ≥
x1 x2 xn x1 + x2 + · · · + xn
The equality is only attained for
a1 a2 an
= = ··· =
x1 x2 xn

4.4.1 Problems
1. If S = x1 + x2 + · · · + xn , and Sk = S − xk , where xk > 0, k = 1, . . . , n prove that
n
X S n2

Sk n−1
k=1

with equality if and only if x1 = x2 = · · · = xn


2. If for k = 1, 2, 3, we have ak + k > 0 and a1 + 2a2 + 3a3 ≤ 4, then
1 2 3
+ + ≥2
a1 + 1 a2 + 2 a3 + 3

3. If a1 , a2 , . . . , an > 1, and σ ∈ Sn (symmetric group), then


n
X a2k
≥ 4n
aσ(k) − 1
k=1

4. For a, b, c > 0,
a b c 3
+ + ≥
b+c c+a a+b 2
18 CHAPTER 4. INEQUALITES

5. For a, b, c > 0 such that abc = 1. Prove that


1 1 1
+ + ≥1
a3 (b + c) b3 (c + a) c3 (a + b)

6. For a, b, c > 0 such that a2 + b2 + c2 = 3. Prove that


1 1 1
+ + ≥1
1 + 2ab 1 + 2bc 1 + 2ca)
Chapter 5

Inequalities with Geometric


Interpretations

5.1 Triangle
If you choose a square with sides that have a length equal to a1 + a2 + · · · + an where ai > 0 then
you can draw a grid in which each pair of values represent the sides of a rectangle. This gives us
an opportunity to create inequalites involving these values.
√ √
1. If we take the non-hypotenues sides of a right triangle as a and b then we have the
inequality √ √ √
a+b≤ a+ b

2. The quadratic
p mean is defined as the arithmetic mean of the squares, which for two numbers
a and b is (a2 + b2 )/2. If we take a square with sides a + b then we can use the triangle
inequality twice to show that the √
for positive numbers a and b, the root mean square lies
between the arithmetic mean and 2 times the arithmetic mean
 r 2
a + b2 √ a+b
  
a+b
≤ ≤ 2
2 2 2

19
20 CHAPTER 5. INEQUALITIES WITH GEOMETRIC INTERPRETATIONS
Chapter 6

Miscellaneous

6.1 Problems
1. A round-robin tournament has n teams, and each team plays at most once in a round. Deter-
mine the minimum number of rounds f (n) needed to complete the tournament. (Romanian
Olympiad 1978).

21
22 CHAPTER 6. MISCELLANEOUS
Chapter 7

Solutions/Hints

7.1 Inequalities
7.1.1 Cauchy-Schwartz Inequality
Solve using Cauchy-Schwartz

1. √ 1
a1 + a2 + · · · + an ≤ n(a21 + a22 + · · · + a2n ) 2

2. s r r r
 
1 1 1 b+c−a c+a−b a+b−c
(a + b + c) + + ≥ + +
a b c a b c
Solution:
r r r s
1√ 1√ 1√
 
1 1 1
b+c−a+ c+a−b+ a + b − c ≤ (a + b + c) + +
a b c a b c

3.
n n
! 21 n
! 21
2 4
X X X
ak ≤ |ak | 3 |ak | 3

k=1 k=1 k=1

4. Suppose that pj ≥ 0 for all j = 1, 2, . . . , n and p1 + p2 + · · · + pn = 1. If aj and bj are


nonnegative real numbers that satisfy the termwise bound 1 ≤ aj bj for all j = 1, 2, . . . , n
then X n n
 X 
1≤ pj aj p j bj
j=1 j=1

5. !4 !2
n
X n
X n
X n
X
ak bk ac ≤ a2k b4k c4k
k=1 k=1 k=1 k=1

6. !2
n
X n
X n
X n
X
ak bk ac ≤ a2k b2k c2k
k=1 k=1 k=1 k=1

7. Show that for all positive x, y, z one has


 1/2  1/2  1/2
x+y x+z y+z
S= + + ≤ 61/2
x+y+z x+y+z x+y+z

8. Show that for all positive x, y, z one has


 2
y2 z2

x
x+y+z ≤2 + +
y+z x+z x+y

23
24 CHAPTER 7. SOLUTIONS/HINTS

9. If a1 , a2 , . . . , a2019 are positive real numbers such that a1 + a2 + · · · + a2019 = 2019, then find
the minimum value of the following expression
1 1 1
+ + ··· +
a1 a2 a2019

10. If a, b, c ≥ 1, prove that


p p p ab + bc + ca
a2 − 1 + b2 − 1 + c2 − 1 ≤
2
Generalize! √
Solution: We denote x = a2 − 1, etc. Obviously x, y, z ≥ 0 and the task becomes to prove
Xp
2(x + y + z) ≤ (x2 + 1)(y 2 + 1)
cycl
p
By the Cauchy-Schwarz inequlity, (x2 + 1)(y 2 + 1) ≥ x + y, with equlity only when xy = 1.
Applying this term-by-term
√ yields the required inequality. Equality holds only if x = y =
z = 1, i.e., a = b = c = 2. For a generalization,
n q n
X 1X
a2i − 1 ≤ ai ai+1
i=1
2 i=1

where an+1 = a1 . For n odd, the equalilty is only attained when all ai = 2; for n even,
whenever x1 x2 = x2 x3 = . . . = xn x1 = 1
11.
(a1 + a2 + · · · + an )2 a2 a2 a2
≤ 1 + 2 + ··· + n
x1 + x2 + · · · + xn x1 x2 xn

12. Note that f (x) = cos(βx) satisfies the identity f 2 (x) = 21 (1 + f (2x)), and show that if pk ≥ 0
for 1 ≤ k ≤ n and p1 + p1 + · · · + pn = 1 then
n
X 1
g(x) = pk cos(βk x) satisfies g 2 (x) ≤ (1 + g(2x))
2
k=1

13. Suppose that pk > 0 for 1 ≤ k ≤ n and p1 + p1 + · · · + pn = 1. Show that one has the bound
n  2
X 1
pk + ≥ n3 + 2n + 1/n,
pk
k=1

and determine the necessary and sufficient conditions for equality to hold.
14. Prove that for all rea, x, y, α and β one has

(5αx+αy + βx + 3βy)2
≤ (5α2 + 2αy + 3β 2 )(5x2 + 2xy + 3y 2 )

More precisely, show that the bound is an immediate corollary of the Cauchy-Schwarz in-
equality provided that one designs a special inner product h·, ·i for the job.
15. For 0 ≤ x < 1 show that
∞ ∞
! 21
X
k 1 X
ak x ≤ √ a2k
k=0
1 − x2 k=0

16. verify that


n n
! 21
X ak √ X
< 2 a2k
k
k=1 k=1

17. verify that


n n
! 21
a 1
√ k
X X
< (log 1) 2 a2k
k=1
n+k k=1
7.1. INEQUALITIES 25

18. verify that


n     12 n
! 12
X n 2n X
ak < a2k
k n
k=0 k=0

19. Many problems of mathematical analysis depend on the discovery of bounds which are
stronger than those one finds with the direct appli- cation of Cauchy’s inequality. To illustrate
the kind of opportunity one might miss, show that for any real numbers aj , j = 1, 2 . . . , n,
one has the bound
n n n
X 2 X
2
X
j
a2j


aj

+
(−1) aj ≤ (n + 2)
j=1 j=1 j=1

Here the direct application of Cauchy’s inequality gives a bound with 2n instead of the value
n + 2, so for large n one does better by a factor of nearly two.

20. Let h·, ·i denote an inner product on the vector space V and suppose that x1 , x2 , . . . , xn and
y1 , y2 , . . . , yn are sequences of elements of V . Prove that one has the following vector analog
of Cauchy’s inequality:

n
X n
X  21  X
n  12
hxj , yj i ≤ hxj , xj i hyj , yj i
j=1 j=1 i=1

21. Show that for each rectangular array {cjk : 1 ≤ j ≤ m, 1 ≤ k ≤ n}, and each pair of sequences
{xj : 1 ≤ j ≤ m} and {yj : 1 ≤ k ≤ n}, we have the bound
 12  X  12
m X
n
√ m n
X X
2 2

c x y
jk j k
≤ RC |x j | |y k |
j=1 k=1 j=1 k=1

where R and C are the row sum and column sum maxima defined by
n
X m
X
R = max |cjk | and C = max |cjk |
j k
k=1 j=1

Note that in the extreme case when n = m, cjk = 0 for j 6= k and cjk = 1 for 1 ≤ j ≤ n,
Schur’s Lemma simple recovers Cauchy’s inequality.

22. Let v and w be elements of the inner product space (V, h·, ·i) and consider the the quadratic
polynomial defined for t ∈ R by

p(t) = hv + tw, +v + twi

Observe that this polynomial is nonnegative and use what you know about the solution of the
quadratic equation to prove the inner product version of Cauchy’s inequality. Also, examine
the steps of your proof to establish the conditions under which the case of equality can apply.

23. Show that if {ajk : 1 ≤ j ≤ m, 1 ≤ k ≤ n} is an array of real numbers then one has
m X
X n 2 m X
X n 2 m X
X n 2 m X
X n
m ajk +n ajk ≤ ajk + mn a2jk
j=1 k=1 j=1 k=1 j=1 k=1 j=1 k=1

Moreover, show that equality holds here if and only if there exist αj and βk such that
ajk = αj + βk for all 1 ≤ j ≤ m and 1 ≤ k ≤ n

24. Show that for any nonnegative aij , bij , cij with 1 ≤ i, j, k ≤ n one has the triple product
inequaity
n n
X  21  X
n  12  X
n  12
X 1 1 1
aij2 bij2 cij2 ≤ aij bjk cki
i,j,k=1 i,j=1 j,k=1 k,i=1
26 CHAPTER 7. SOLUTIONS/HINTS

7.1.2 Bergstrom’s Inequality


1. If S = x1 + x2 + · · · + xn , and Sk = S − xk , where xk > 0, k = 1, . . . , n prove that
n
X S n2

Sk n−1
k=1

with equality if and only if x1 = x2 = · · · = xn


2. If for k = 1, 2, 3, we have ak + k > 0 and a1 + 2a2 + 3a3 ≤ 4, then
1 2 3
+ + ≥2
a1 + 1 a2 + 2 a3 + 3

3. If a1 , a2 , . . . , an > 1, and σ ∈ Sn (symmetric group), then


n
X a2k
≥ 4n
aσ(k) − 1
k=1

4. For a, b, c > 0,
a b c 3
+ + ≥
b+c c+a a+b 2
5. For a, b, c > 0 such that abc = 1. Prove that
1 1 1
+ + ≥1
a3 (b + c) b3 (c + a) c3 (a + b)

6. For a, b, c > 0 such that a2 + b2 + c2 = 3. Prove that


1 1 1
+ + ≥1
1 + 2ab 1 + 2bc 1 + 2ca)

7.1.3 Miscellaneous
1. A round-robin tournament has n teams, and each team plays at most once in a round. Deter-
mine the minimum number of rounds f (n) needed to complete the tournament. (Romanian
Olympiad 1978). Answer: see Documents/TournamentScheduling.html

You might also like