C

WIT 3 INEQUALITIES -

Structure
3.1 Introduction
Objectives

32 Inequalities Known to the Ancients

3.3.1 lnequality of the Means
3.3.2 Tnangle inequality

33 Less Ancient Inequalities

3.3.1 Cauchy-Schwarz lnequality
3.3.2 Weierstrass' Inequalities
3.3.3 Tchebychev's Inequalities

3.4 summary

3.1 INTRODUCTION
So far we have discussed equations of various kinds. vow we shall consider some
inequalities; not of the social kind, but between real nuhbers. A mathematical inequality is a
mathematical expression of the condition that of two quantities one is greater than, greater
than o r equal to, less than or less than or equal to the other. An inequality that holds for every
real number is called an absolute inequality. In this unit we shall restrict ourselves to such
inequalities.

We will discuss six famous absolute inequalities. We have divided them into two sections-
those that have been used for centuries and those that were discovered by some famous
nineteenth century European mathematicians. These inequalities have several applications
also. We will discuss a few of them. You may come across some applications in other courses
too, at which time we hope that you will find that you didn't study this unit in vain!

Let us list our unit objectives now.

Objectives
After reading this unit you should be able to prove and apply
the inequalities of the means;
the triangle inequality;
the Cauchy-Schwan.(Bunyakovskii)inequality:
Weierstrass' inequalities:
Tchebychev's inequalities.

Let us discuss the ineqyalities one by one.

3.2 INEQUALITIES KNOWN TO THE ANCIENTS

In this section we shall discuss two inequalities handed down to us by ancient
mathematicians. But first we will give a list of some properties of inequalities you must be
fmiliar with. They are the following:
for a, b, c, d E R
i) aZb,cZO*acZbc
ii) a2be-as-b
 lnequnlities
1 1
iii) a2b~-5-provideda;tO,bzO.
a b
iv) a2b,c2d*a+c>b+d
v) an>-bn,a2O*a2b,wherene N.
We will often use these properties implicitly while proving the inequalities mentioned in the
unit objectives.
Now let us discuss the inequality that relates three averages.

i 3.2.1 Inequality of the Means

An important part of arithmetic that can be traced back to the Babylonians and Pythagoreans
(approximately 6th century B.C.) is the theory of means or averages. The word "average"
comes from the Latin word "havaria", which was the insurance paid to compensate for damage
to goods in transit in the olden days. All of us are familiar with compensate for damage to
goods in transit in the olden days. All of us are familiar with the term "average". In fact, all of
us must have often calculated fie average of a fmite set of numbers by addhg them up and
dividing the sum by the total number of these numbers. But this is only one of Inany types of
averages. We will discuss three of these types here. Let us start with the "usual" average.

Definition: the arithmetic mean (AM) ofn real numbers x,, x2, ..., X, is

+ X2 + ...+-
XI
n
X"
, that is n [A
i = ~xi] .

1 1
1 -1 A

+o 1
- -

For example, the AM of - - and 0 is

2 3
2 -=-
18 3 '

The AM is often used in statistics for studying data.

Another type of average is the geometric mean. This is the best mean to use if we want to find
the mean of any finite set of positive numbers that follow geometric progression. Thus, this
mean is very useful for studying population'growth. Let us see how the geometric mean is
defmed.
Definition: The geometric mean (GM) of n positive real numbers

XI, %, ......,Xn 1s

For example, the GM of 3 and4 is .== f i ,and the GM of 2 , 4 and 8 is

(2x4x8)lB=4. .
Yet another kind of average of numbers is their harmonic mean, which we now define.
Definition: The harmonic mean 0of n non-zero real numbers
Eauations and Incaualiticr
This proof is due to Cauchy]
who you will meet again in See.
6.3.
1 3 - 42
For example, the HM of -2, - and 7 is --
3 - - -1+ 7 + - 1 37 '
The HM is the most appropriate type of average to use when we want to find the average rate
of a set of varying rates. Thus, it is the best average to use for obtaining the average velocity
of a vehicle covering various distances at different speeds..
At this point we would like to make a remark.
Note: We can obtain the AM of any n real numbers. But, we only define the GM of n positive
real numbers; and the HM of n non-zero real numbers.
Now let us look at the three different means together. To do so, we clearly need to restrict
ourselves to positive real numbers. What is the AM of 2,4 and 8? How is it related to their
GM? And, how is their GM related to their HM? The following result answers these questions.
Theorem 1: Let {x,, xz, ...,x, ) be any finite set of positive real numbers, and let A, G and H
denote their arithmetic, geometric and harmonic means, respectively. Then
We will only give a broad outline of the proof here. The inequality A 2 G is first proved by
induction (see Unit 2) for all those integers n that are powers of two That is,
and equality holds iff x, = 3 = ...= x,m.
Now, given any n E N, we can always choose r E N such that 2' > n.
We apply (1) to the 2' numbers x,, x,, ...,x,, A, ...,A, where the number of A's is 2' - n
We get
(with equality iff x, = 3 = ...=xn =A.)
* 2'
, since Cxi
i-I
= nA.
* A 5 G, since A and G are positive real numbers.
Note thatA = G iff x, = 3 = ... =a.
Thus,.the result is true v n E N.
1 1 1
-
Now let us consider the n positive numbers -9 .
XI X2 Xn
Since their AM is greater than or equal to their GM, we get
Thus, A 2 G 2 H, with equality f i x , = x, = ...= x,.
In about 320 ~ bthe.geometer Pappus of Alexandria gave a geometric construction of the
AM, GM and HM of two numbers. His construction is as follows:
Draw a semicircle with a+b as diameter (see Fig. 1). Let its diameter be AC, with mid-point 0.

Flg. 1. Xbe AM, CM and HM of 8 and b arc DO, DB a11d DE, respectively.

Then OA is the radius of the circle. Mark off the point B on AC such that AB = a. Then BC = b.
Draw BD I AC to meet the semicircle in D. Then draw BE I DO, as in Fig. 1.
Then Pappus proved that
DO is the AM of a and b.
DB is the GM of a and b.
DEistheHMofaandb.

"
Since DO 2DB 2 DE, this gives us a geometric proof for Theorem 1, when n = 2.
Now let us apply Theorem 1 to prove some more inequalities.

Example 1: Show that [T~

1= 1
I ] > nn (n!)r, n! denotes fwtorial n and r > 0.

Solution: Let r be a fixed positive real number. Consider the n positive qmber lr, 2', ..., nr'By
Thkorem 1

Since the numbers lr, 2: ...,n'are not equal, their AM is strictly greater thantheir GM. Thus

(1' +2' fn. .+ .'1" 2 (n!)'

We can prove several inequalities, which are particularly useful in mathematics, by using
Theorem 1. We ask you to provdsome of them in the following exercises.

E 1) Show that jab + xy) (ax + by) 2 4abxy, where a, b, x, y are positive real numbers.
Under what conditions on a, b, x and y would the equality hold?
Equationr and lnequalltier E2) For any n E N and positive real numbers x and y, show that

E 3) IsTheorem 1 true if we remove the condition that the numbers are positive? Why?

Now, you know that the inequalitiesin Theorem 1 become equalities when x, = x,= ...= x,,.
When this happens, then xi = A = G = H v i = 1, ..., n. Thus, we see that
ifx,, x,,...,x,are n positive real numbers such that x, + x,+ ... + x, is a constant,

then their arithmetic mean attains its lowest value and their geometric mean attains its

maximumvaluewhenx,=x,= ...- x =A = G

Let us see how to use this fact for obtaining some maximum and minimum values. For
convenience, we shall denote the set of positive real numbers by R". '
Example 2: Find the greatest value of xyz, where x, y, z E R+are subject to the condition
yz+zx+xy= 12.

~ o l h i o nxyz
: has greatest value when (xyz)' = (yz) (ZX) (xy) has greatest value. Since yz+zx+xy
is a constant, we know that the maximum value of (yz) (zx) (xy) is attained when yz = zx = xy,
that is, when x = y = z.

Hence, the maximum value of xyz is 23= 8.

Example 3: If the sum of the sides of a triangle is the constant 2s, prove that the area is
greatest when the triangle is equilateral.

Solution: Let a, b, c be the sides of the triangle, where

a+b+c= 2s,
and let A denote the area of the triangle.

ThenA= J s ( s - a ) ( s - b ) ( s - c ) .

So, A will be greatest when (s - a) (s - b) (s - c) is maximum.

Now (s - a) + (s - b) + (s - c) = s, a constant.

Thus, (s - a) (s - b) (s - c) is maximum when

s-a=s-b=s-c,thatis,a=b=c.
Thus, the area is maximum when the triangle is an equilateral triangle.

Why don't you try these exercises now?

E4) a) Prove that if the sum of two positive numbers is given, their product is greatest
when they are equal.

' b) Is (a) m e if the words 'sum' and 'product' are interchanged? Why? .
E 5) Find the greatest value of (5+x13(5-x14, for -5 < x < 5.
 (Hint: The greatest value of ( 5 + ~(5-x)'
) ~ occurs when the greatest value of

($I3 (y] 4

occurs.)

E 6) When docs a cuboid, with dimensions x, y and z such that x+y+z is fixed, have maximum
volu@e?
E7) Under what conditions on the dimensions, will a cuboid with fixed volume have minimal
surface area?
(Hint: Use the inequality G 2 H.)

You can study other techniques for obtaining maximum values in our course on calculus.

Let us now consider another inequality, which follows from Theorem 1

Theorem 2: If x,, ..., xn E R+such that not all of them are equal, and m E Q, m # 0, m # 1, then

The proof of this result uses Theotem 1. We shall not give it here.
A result that follows from Theorem 2 (and Theorem 1) is that

I If x, + xz + ... + x, = c, a constant, then

n
I
for 0 < m < 1 the maximum value of C
i-1
X; is n'-rn cm,and
n
f o r m < O o r m ~1 the minimum valueof C xisnl-"'
is1
~c rn

These values are attained when x, = x2 = ...= x,

Again, we shall not prove this result in this course. But let us consider an example of its use
for fmding some maximum and minimum values,
1 1 1
;Y
Example 4: Find the least value of + - + - ,where x, y, z E and x + y + z = 27
Z
1 1 1
Solution: - + - + -
X Y = isoftheformxm+ym+zm,wherem=-1<O.Sincex+y+z=27,the
least value is obtained when x = y = z.
Andthenx+y+z=27givesusx=y=z=9.
1 1 1 1 1 1 1
Thus,theminimumvalueof - + - + - is - + - + - = -
x y z 9 9 9 3
Note that we could have also obtained the answer by applying the inequality G 1 H
(c.f Theorem l), exactly on the lines of the solution of E7.

Now for some exercises.

ES) Show that the sum of the mth powers of the first n even numbers is greater than n(n+l)'",
ifm> i.
-

E 10) Let a,, %, a,, E R+and p, q E N such that p > q. Show that
a: +a! +...+a: < n H (a1 + a , +...+a$)
-9
(Hint: Put m = -,xi = a! in Theorem 2.)
P
Equatloor and Ioequalltler So far we have discussed various inequalities related ta the arithmetic, geometric and harmonic
means. Now let us consider an inequality that had its origin in ancient Greek geometry.

3.2.2 Triangle Inequality

If you look up any translation of the ancient Greek mathematician Euclid's "Elements", you will
find that Proposition 20 of Book 1 says:
t
"In anytrimgle, two sides taken together in any manner are greater than the remaiping one.".
This result is the basis of the triangle inequality, which is a statement about the absolute value
of numbers.
Recall that absolute value of x E R is defined by

Thus it satisfies the following properties

3 Ixl=I-xI v x c R and a) xSJXV
( XER

You can study the absolute value of real numbers in more detail in our course on cafculus.

Now let us state the triangle inequality.

Theorem 3: Let x,, x,,...,x,,E R. Then

Ix,+x,+ ...+ x,ISIx,I+Ix,I+ ...+Iql.
Moreover, equality holds only when all the non-zero xi's have the same sign.
ProoZ: Let us prove the result for n = 2 first.
Now

(lx, +x21)' = ( ~ , + x , ) ~ , s i r r c ( x ( y~ =X xE ~R

= x: + 2x, x, + x,2
~ ~ x , ~ ~ + 2 ~ x ~ ~ ~ ~ ~ ~ + ~vxX,E~R~ , s i n c e x < ~ x (

=(lx11+1x~l)~.
Now we take the square root on both sides, keeping in mind that I x t 2 0 V X E R We get

I x, + x, 1 5 1 x, I + ( x, I, which is what we wanted to prove.

Notethatifx, c 0 , sayx, =-a,and~2>0sayx,=b,wherea,b>O,then
Ix,+x,I=Ib-a),while)x, (+Ix,)=a+b.T@, whenx,andx,haveoppositesigns

So, Theorem 3 is true for n = 2.

Now, let us prove the result for the general case, by induction. So let n > 2 and assume that
Theorem 3 is true for any n-1 numbers. Now consider
Ix,+$+ ...+\I =I~~+x,+...+x,~)+x,I

~lx,+x,+...+~lI+I~nI
S I X, ( + ( X, 1 + ...+ 1 xWl( + ( X, 1,since the result is true for n - 1 numbers.
 Further, just as we have shown for the case n = 2, strict inequality holds if all the non-zero 5's
don't have the same sign.

Theorem 3 is not only m e for real numbers. In our course on linear algebra we have proved
C,then(z,+z,+...+z,IIlz, (+Iz,I+ ...+I z,I,
thatifz,,~,,...,z , ~
where I z 1 is the modulus of z.
Let us verify Theorem 3 for the numbers - 2 , l , 5 , O .
Since1-2+1+5+OI=141=4,and

we find that strict inequality in Theorem 3 is true in this case. Note that -2 and 1 have opposite
signs.

I Why don't you try some exercises now?

I
E 11) The absolute value of the AM of n numbers is less than or equal to the AM of their
absolute values. True or false. Why?

E 12) Prove or disprove that

( To disprove a statement means to show that it is false. See the appendix of Block 1.)

E 13) Prove or disprove that

v xI J, Y e
I X - Y I ~ ~ I ~ I - I Y R.
.(Hint: W r i t e ( x I = ( ( x - y ) + y I,andalsousethefactthat(xI=I-XIv XE R.)

Now let us discuss some "newer" inequalities.

,-=%

3.3 LESS ANCIENT INEQUALITIES

In this section we shall discuss four important mequalities which are due to some mathematical
giants of the nineteenth century. We start an mequality due to three mathematicians.

I -
3.3.1 Cauchy Schwarz Inequality
I Augustin - Louis Canchy, the famous French mathematician, was responsible for
I
I developments in infinite series, function theory, differential equations, determinants,
t probability and several other areas of mathematics. One of his contributions was result, which
was later generalised by the German mathematician H.A. Schwarz (1848- 1921). We now state
this result, which was also proved independently by the Russian mathematician Bunyakovskii. : C8uchy (1789-1857)

Theorem4 (Cauchy-Schwarz Ipequality): Let a,, a,, ....,a,, b,, b,, ..., b,, E R.

Then

( a , b , + a 2 b 2 + . . . + a n b n ) 2 5 (+a:
a~ + ....+az)(b:+b: + .... +bz),
When a, = cb, y i = 1, .., n,
with equality iff ai = cb, v i = 1, ...,n, where c is a fixed real number. where c is a constant, we say
that the n-tuples (a, . ., an) and
(b, ... b,,) a n proportional.
Proof: To help you understand the proof we shall prove it for n = 3 fust. Then you can try and
generalise it (see El 4).

1 NOW

I
(a: +a: +a:) (b: + b? + hi)-(a,b, + a2b2+ a3b312
Equations and lacqualltfes

When will the equality sign hold? Equality holds iff a,b2- a2b, = 0, %b3- a3b2= 0 and
a,b,- a,b, = 0, that is, a, = cb,, a2 = cb2,a, = cb,, for a fixed real number c.

Thus, we have proved the result for n = 3.

Now, to complete the proof of Theorem 4, why don't you try this exercise?

E 14)Prove Theorem 4 for any n E N.

Let us consider an application of Theorem4 for locating the roots of a polynomial. Before
going further, you may like to keep Unit 3 nearby for easy reference.

Theorem 5: If all the roots of the real polynomial equation

x" + a, x"' + a2xW2+ ... + an= 0 are real, then they lie between -y
n - l Ja: - (z)a2
n-1

and n+e
Ja: - (%Ia2
n-1

Proof: From Theorem 1 of Unit 3, you know that the given equations has n roots. Let x be a
root. If x,, ..., x", are the other roots, then by Theorem 4 of Unit 3,

Also, by Theorem 4 of Unit 3,

:. (a, + X)2 ~ ( n 1)(a:

- -2a2-x2)

by the quadratic formula.

This holds for all x such that

Thus, any root of +e given polynomial equation must lie between the bounds of the given
statement of the theorem.
Before giving an example of the use of Theorem 5, we shall make some related observations.

Remark 1 :Consider the polynomial equation a, + a,x + a2x2+ ... + anxn= 0, where a, E Z.
 d
Then any rational root of this equation is of the form -
an
,where d is a factor of a,.

Remark,Z: For cubic cases, we know from the discriminant (Sec. 3.3.2) when the roots are all
real. An4 then Theorem 5 can be very useful, especially if we know that the roots are rational.

Let us now consider an example of the use of Theorem 5.

Example 5: Solve x3- 23i2+ 167x- 385 = 0.
Solution: The discriminant of this equation (see Sec. 3.3.2) is posltlve. Hence the given
equation has three distinct real roots. 1f the roots are rational they must be integral factors of
- 385. Thus, they musthelong to the set

23 4 23 4
But, by Theorem 5 the roots must lie between - - - f i and - + - f i . Thus, lithey are
3 3 3 3
rational, they can only be 5,7, 11. On substituting these values in the equation, we h d that
they are indeed roots of the given equation. Also, you know that the equation can oiily have 3
roots. ~ence,'thesevalues are the only roots.

You may now like to try to apply Theorem 5 yourself.

Now let us consider another example of the use of Theorem 4. In this example, we shall apply
the Cauchy-Schwarz inequality twice to get an inequality that we want.

Example 6: Let x, y, z E R+such that x2+ y2+ z2= 27. Show that

312 312 312

Solution: Let us first apply Theorem 4 to the two triples of real numbers, (x ,y , z ) and
(x1I2,y1I2,z1I2).We get
( X PXl12 r y + z z s (x3+ y3+ z3)(xtY+z),that is
+ 312 112 312 112 2
)
(x + y 2 + z ~ ( x 3 + y 3 + z 3 ) ( x + y + z )
)

Now let us apply the Cauchy-Schwarz inequality to the triples

(x,y,z)and(l, 1,l).Weget

!x.l+y.l + ~ . l ) ~ S ( x ~ + ~ ~ 12+
+ z 12),thatis
~)(l~+
( x + y + ~ ) ~3 S( ~ ~ + ~ ~ + z ~ ) = 8 1

*x+y+zS9

Thus, by (2)
(X2+y2+z2)S9(x3+Y3+z3)

But x2+ y2-+z2= 27. Thus,

Why don't you try some exercises now?

Equations and Inequalities E16) If a, b, x, y E R such that a2 + b2 = 1 andx2 +y2 = 1, then prove that ax + by S 1.

El 7) Prove that if a,, ...,a, E R", then

a) ( a , + a 2 + ...+ a,)
an

C) (&+A+
...+A)
' ~ n ( a , + a , + ...+ a,)

E18) (Another form of the triangle inequality) If a, b, x, y E R, then show that

(X - y)f 5 JZ?+ JW
(Hint: Write (a - b), + (x - y)2= (a2+ x2)+ (b2+ Y2)- 2(ab + xy), and then apply
Theorem 4 to (a, x) and (b, y).

E 1 9 ) I f i , y , z ~~"suchthatx~+~~+z~=81,then~rovethatx+~+z19.

E2O)Prove or disprove the following generalisationof Theorem 4 :

N , p # l , a n d a ,,..., a,,,b,,b2,,...,~ , E R .
Letp~

Then(a,b,+a,b,+ ...+ anbn)'<(a: +a! +...+ a!)(bp +b,P+...+b;)

The Cauchy-Schwaxz inequality has several applications in physics and mathematics,

especially in the context of inner product spaces.

Fig. 3 : K.rt Tbcodor Now let us consider another useful set of inequalities.
Weierstrasr

3.3.2 Weierstrass' Inequalities

It is generally through that a good mathematician must have started serious mathematical
studies at an early age. But the German-mathematicianWeimtrass (18 15 - 1897) is an
exception to this rule. This outstanding mathematician started serious mathematics at tl.ie age
of forty. He was responsible for making ~ a l y s imore
s rigorous, and is considered to be the
"Wher of modem analysis". He is responsible for the following result.

Theorem 6 (Weierstrass' Inequalities): Let a,, a,, ...,a, be positive real numbers less than 1
and S, = a, + a2 + ...+a,,. Then

1 where it is assumed that S, < 1.

(ii) l + S , S ( l + a , ) ( l + a J ...(l + a n ) <

ProoZ:We prove (i) by induction on n, a principle that we introduced you to in Unit 2.

So, ( i j is true when n = 1.

.
Let us assume that (i) is true for x i = m, where m E N.
We will see if it is also true for n = m+l .
 (l -a,) ... -a;) - S , by ouras+,p~m. ,i~6irisdi i;L !,.d-- I .. - ;sj . . *siiiisu@s5r ~%qliIhbe.

:dp b: r.d,a5 . 4s .-- I rj !h:

Th&,i(,l-al)(l-%) .A (1 -a,,,)(l -amlj ~ (-sm)(i-.a,,)
1 ' ' ,

RRS.=I-(S~+~,,)+S~~,, ,;j~su:rni8
1s.i rC+?

; !), , = 1- S, + sma,, :.<-$

..,
- f<.:,~
.
:,; :y ,
. ,
2
. -. &,.&
...

,., , ..<.f-. >:,& ,.;. 4

!j.,i !iL>.;.
- . ~l-S,l,sinceSma,,>O. . i I '
,

",. -,,:.,-, ..,:, !'-; . ... , .,,. . ::: ', ,.ii$bprl .

S.o,(l-a,) ...(I-a,,,+,)>l-Swl
. -d ; , ~z : ; , d t - ,. . *i . a i , & s j i
(..*.. ; lnj ;
., 6 - : . i . , ~ + , . i. I .
.. . , ,..~ !

.
A . . ;-!.
.
'1 : ~. .. .

-
i ' . '

. . & ~ WQ- %?&=&+

Funher, Since (1 - a,) (1 - p ) t , j , ~ r ? & l & ~,..kvflw@?
we fmd that :
l.:
,.: ... L .;
I
'
.. .<I .
, ., ,.;.;a ;.,,>z ; ,:. :..;::&.- .d . i. .: ,4>;,B,:t
. ,,

1 .

, (3) and ( 4 , takentogether, tell us W t (i) is m e for n = m + l!!Yenc% -. ..indy~fio#~.(i)@Qq~

. py '-" , .*
... . . . ! 1

\ I ~ N.
E

Now, to comple@the proof you can try E21.

E 21) Rove (ii) of Theorem 6.

~ 2 2 ) or 0 c a,, i,,:..,a, < 1,prove that I :j.,:; - o. ci ., - ,..:.> 5: >a$,,r, .: , ,. r$ , i ,::
n n
.
-
.
I P n. - ,
I
I ..
, ;
.
?
.>
. - +:-.-.:.
Is,ai
1

.. [,ni,:?;,:/.
.
f:'!:r, ,

.
fi%i ?j~ijj?;A
,:ii>.?>l~

, i-I
, I
~i.2,.
:
i=l
.* !$ ,. ,
.
,i- y j ; : ~ , . :
.
~-, ;;&,
.' . - ,.';,,rd7f,.,
,.&:'i$;,,.?,f;,
. . . .c. , ... j 2'.,IX
.1

(Hint: 0 < 1-a, c 1.) ,

, 2
. .-. . .
.!
b.1. - I
.
B)+,( ( j ~ . :1 ,. ;[,i.z;:;?cr:
',Ti!
.
! .

E23) Does Theorem 6 hold if ai < 0 or a. > 1 for any i = 1, ...,n? Give reasons for your answer.
. r $ , .:<.a)
~ .i#FlVIi)911$3d$fo,:~)-tq$!ti %:i)iqru~%:t
ir3 ..vo:x bnA

1 : .. ,Fig. 4 : Tchibychev
generalisations
.
are very useful.
,
it$ .
,~;,... .>I... i-.
. i;t , r3iq i ~ k
. ,,..~y:,,;> #.. -
;>.
(1821-1894)

And fmlly, we shall discuss some inequalit&due. ~ ~ . ~ , ~ ~. . , , . : w . t h e w .

o~ $~ leadi%grR9sip
c t . ~ ~ ~t ,
. , X I L . . ~ . , , .

The mathematicuqPafnutii L. Tchebychev (prono+ced Che-bee-chefO ir most,@oyn for his

tremendous work-manalytic number theory and the theory of o r t h o g o n t a $ d ~ & d @ ~ @
here we shall prove and apply some inequalities that are;ng@d afterhim.
.
..
<,>. <:;,...? .j .L,; .+.
.. < ,

Theorem 7 (~iheb~chev's Inequalities): If a,, ....an,b,,,:. b j s R such that . , '

8 a, ,
n
, , " , : ; : : , , ? ., : -; ' : -
/ I _
.. T.9.
9 " : ~ii~ls2

. . n ( a ~ b 2~
' + + + ?nb&!+efl~$ %+..:$.Q + b2 +-- +b ,)
.: fiii ,; . ,. - .,
: ; c I.-,;;,
... . ..
s'
f :l j
!! ii) a, 2 a, 2 ... 1 an,b, Ib, 5.... 5 b,, then,
. ': ..F , . .. ..
. , r 2 .:
. .*
n(a,b, + a b + ...+ a;,bn)z5 (a, + a,.* .:I
2 2.
. . ,,

n.
-:f b 2
(b :%.b,+--. , - n '
; ,< ; .,: i
! :..
,-
- . .:
7.

. . - l i:

Proof: Let us prove (i) for the case n = 3, so that you can understand the proof more easily.
i>?<jtj ci,! :,><>: ~ ( :icj;;~'T
9
-. .. .-. .- - . *
J<,. ;a3bj;: -;>!,9i:g.* k
Now an add albl + +bz + %b, to both sides d simp@ to get ,

.Theproof of (i) f a my a E N is on exactly the sank lines. Let bs prove it by induction on n.

The result is tne for n = 3 (uul, in fact, for p = 1 and 2).
Assume thatitistrpCf&n- 1. Then
(n- l)(albl +...+a,+lbwl)2(al+++ ... +b2+...+b,,)
Also a,b, + Q, 2 a,ba + s b , , since (a, - (b, - bn)2 0-

Adding up the left brnd 8id- d the Gght hand sides of these n inequalities, we get

~d now,tbcompkttt h e k t ofthe theorem, try ~ 2 4 .

Wt:Put 5 - , a , i 1, ...n, and use (i))

Now kt us consider m applicadon of Tchebychev's inequalities.

6' .
&t.rmrlc7:Sbowthllt J

Taking thetsquare root on 'both sides, we get

 I -8

Now you can try some exercises.

E25) Show it$

\

(Hint: First apply Tchebychev's inquality to the n-tuples

E26) If a, b, c, E%+, then show that.

(Hlnt: See if it is possible to apply Tbeomn7 to b+c, cta,a+b and their inversk.)
b

With this ine;1uaIity we come to &e end of this unit. This d-t mean that we've e x h a d
all the inequalities, or even all the important ones. We have just exposedyou to a few
'
elementary ones and some of their wlications. As you study more mathematics you will come
across these and several others.
Now let us quickly go through what we have covered in this unit. t

3.4 SUMMARY .

We have discussed several inequalities and their applications in this unit. Let us list themone
by one,
1) The inequdlity of the means: The AM of any fimie.ct of elements of R+is greater

I) 2)
than or equal to their GM, which is greater than or equal to their HM.
Ifx,, ...,x,,~Rsu~hthatnotall~fthemarequal,andm~Q,mtO,l.then
- 2

3) The triangle inequality :For x,, 3...., x, E R

I

I The inequalit)' is strict in case all the non-zero xis don't have the same sign. .

I 4) Cauchy-Schwarz (or Bunyakovaskii) inequality: If a,, ...; a,,, b,, ...,b, ER;then
t- and Inequalltkr

. , 5; y!,
,137i3.%9x:31 ;:c::pi;*-?t,is j?

--.
with ??!i4b~:?*bd~~~!~$~~1: ...I*.*.I* .:
. ly.L%. :? %v=PQ.

I n .)- ?,<,! -' - ? :

5) Weiaa-' incquhties: For 0 < a,, ....an< l'and S, = ai
irl

6) Tchebychev's i n w t i e s : If (a,, ....a,), (b,, ...,bd) E R" such that

--
3 a, S a , S ..:Sa,, bi9 b,S ...S b,, ,then I I
I
\
I . ' i :
n(albl+a,b2+ ...+ anbd)2(a,+a,+....+an)(bl+bp+...+bn)

n(a,b, + a,b, + ...+ anbn)S (a, + + ...+an)(bl '+bp+ ...+*n) %:

1;

As usual, are suggest &t you go back to the beginning of & d >hddsdt'ifyou'hi&

. this we have cqma tb %e. end of. ?this,course. We

With
,,Z,?[ 3, 01 gj;:?
, , -
hope
, ii yov

;;.3&<>f!.9m.
. . .; ., . I f . Lit, , :.:!.A
have,enj,oied
.. -*,-

i,,!'; :;.~
; ;>
.
> and wgl,
,

.:<:;
,
I;;: ;,

:,.,~!,I;:,;;. . ,cm.:;:
,,.

..,.: 4 ; ; :
1.'..

;L).. . .;
.
. . ..dt:;:t::,:,*>(F:
2ij.-

&;
;

!.&
it

Note that equality tiolds iff ab = xy and ax = by e a = y and b = x.

r s
. P "
E2) a) Applying the inequality A 2 G to the n+ 1 nup.lyrs ".
j2

~
C

. -s +

X, Y, Y, ..:, Y (ntimes), we get

..

Note that equality holds iff x = y.

, , I:

.
,
: i. .' , ,:- *-;*

, %
Y i I ' 1 I ! > 1 ' , k
n n r l
I
I

t
I.
.
~.

*
+
d
C) k t m, n ci N and.m n. Then, by fi),

. b) N~.PO*

NO?,
Hence the result,

Their AIM is

*, : ,,;;,,
r " .

example, let i,y E R such that x i 4.

x + y, since x =5 and y =

E5) Since-5<~<5,5+x>Oand5-x >0.

The volume of the cuboid is xyz.

,

-3 and their HM is

mxiqum, that is,when x = y.

';,, < ;,,>.., ,;, ...;: *.: : ' : * . .. .
+,,:
, :
<

Thus, the cube is a cuboid.wjfh.mU@umum~&,.&the.gj&&

. .. ,C
-
,.

--

h r ~ 1 e , & e t h e & n ~ ~ - 2 , 1 . n d 3 b - . ;; . : :,; ':'.-,:

2

: .: ,.: '..
S'
-1
3

-5.1 fa-k;give

the surface areabf the cuboid is 2(xy + yz +hi.. T ~&

is mininnun, that when

Now, xyz=c, and the HM of x, y and z is

i!!: '>,,
.,u:l
L-'
,i,,

E 7) Let x, y and z be dimensions of the cuboid, where xyz = c, a constant.

. .-

-+I+-
2
1 = -
3
E4) a) Let x, y E R+such that x + =~c, a constant. Then, xy hmaximum when , xy IS
.
%

E 3) No, since the GM of negative numbers is not defintd.,:E~qg, 2 H does not remain
true any longer.

18
5
.., . .

. ,,,,..

. !

.,.: if!,3:jJ.,,.?a.:;:,

minimumwhen x = y = p. Thus the surface area of a cuboid with fixed volume j i

minimum when the cuboid is a cube.
- --.- .
3
. , ,..; . *

..
. 6.; _ .,
:.:,;%i,',,.,,'

.'!>j

1 + --1 + .-1 ' .

-
X
.
".

y X.; ; - I .,.
.
-:.,:
.
.

,'
,A

a @rgk.vdtieof x +y,

< .'
2,
,
..
..
"

2
, 2 -5 .
i

..
.
.

.
.
>
.-.. .. . . .
.
'

.
.
:?.

. ;-
..

,,.','!
18

' .

..:

.
.: ,;, ,a.jI

i'
$,

! :: ;.

.:
:
:

? .
..%..:

*
> , r

.' ''p..:
1:. >

'

:%! 5 >,, j4.>;.,) 2t:,,Liv;.

."
. . :>" .
-.r

i t . ; .
,L I"

!?&
b

;, . ;. ... ,. x
..

xy +yl +a
:
:

,!
,
.,I

.
." ,

.
I

,::;j:

; = : "<

I '
3 : r.

%, , !,,
'.
.<

;liijj
,

. ' ,,~ f , .-.,:.

. $

;
:,,'

,':
.,.

{c{':j. .
. ,

. .
.. .

e.
m.1 I.cq.mlltla E8) We apply Theorem 2 to then numbers 2,4,6, ...,2n.
We get . ~

Ji+fi+...+&
n.
t
Hence the result.

9
E 10)Putrnz -.ThenO~rn< l.Nowwe&pply'lheorem2tothinpositivenumbers
P
a!, at, ....a!. w e get

v
Q a: +a: +...+a! s (nH)l1q (a! + a t +...+a!)q1p

<d7 (a: +&:+...+a:),

1 '9
since - < 1 md < 1.
P . P
- .- '.

E l l ) Letx,,..., x,benmt&em.

'Ihm their AM, A =

XI + X2 +...+X,
n

I+Ix~(+..-+Ix,(by theorem 3.
.
.-.\A\ = I'+
X, x2J...+)L,
I (XI
n
4

:AMoftx, 1.13I ,.-.,I \ I . .

I
Hence the statement is true.
'~12)F a l s e . , F o r ~ l e , ~ e xl,y=-3.
=
..
Thenlx-yl={ 4 1=4,
md( x 1-1 y 1-1-31-2.
E13) I x 1-1 ( ~ - Y ) + Y
IClx-Y l+l Y I. ..,(lo)
l x 1-1 Y 151 x-Y 1
#

Similarly,[ y 1-1 x+(y-x) IS1 il + l y - x l=l x l + l x-y I, i

with equality iffaibj= ajbi v i, j = 1, ...,n, i #j, .

that is, iffa, =cbl \I i = 1, ...,II, where c is a co~utant.

The -of the equation is positive. Thus, the given equation hrrs distinct red
. 2 2
y lie between
roots. ~ h .must -
fi) and j(l + fi).~ftheyarc rational, they have
*
to be factors of 2. Hence, they can be 1,-I, 2 or -2. Of these, 1 and 2, lie within the
given bounds. On substitution we find that they actually arc the mots. Since fhe
equation has only 3 roots, - 1 , l and 2 an its roots. f

~ u sapplyins-.Theorem4
t ot the numbers a, b, Y, we get

(a)We apply the Cauchy-Schwa inquality to the n-tuple&

b) Applying Theorem 4 to the n-tuples (1; 1, ...,1) and (a,, 4, ...,a,,), we get the mdt.

C)ApplyingTheorem4tothen-mples(1,1, ..., l)and(&,& ,...,&)&get

the result.

Also, by Theorem 4
(ab + xYl2s(a2+ x2)(b2+ y2)
3I ab+xy =
/,I-/,. . ..(13)

(12)and(13)*(~)~+(x-y)~~
which gives us the desired result. . . . I

By the Cauchy-Schwarzinequality applied 'to (x, y, z) and (l,1, I), we have

(x+y+z)' ..(14)
. 4
51
 ' ~ ~ a iapplying
n, W r e m 4 to ( A,J;, & ) and (x312';y" ,z3")

E 20) False. For example, take p = 3 and the pairs (1, O), ( 1. 1 -
hen {(1)(1)+(0)(-1)}~i ( 1 ~ + 0 ~ ) ( 1 ~ + ( )- 1 ) ~ I, .,, * *, I *

,.
,.
E21) We use the principk of ind~ctiof"~?$p.,~ >, .,, . r, I , . '

k F,o r n = 1' a = I ,4ndhence,

. . 1 +s,,s 1 +"::r$<.:,
a -\: b k 3 ,
/ .
- 1 *<

'I
< y; , - ver* f * ; >:

Thcn(l +a,)(l +%):..(I + a J > p + s i "-.' t 3 L ' . 7 i C ~ f # :*? ) L , TJzq,\<

* '
.*. (1 +=,)(I +%).,.(I + - a ~ ~ b ' a m , ) $ 1 ~ + ~ ~ ; 4 i $ ~ ~ ~:it3~ g :(li
~l''.A

> l+sm,
L " \ ,
I' ..

Also, (1 +a,) (1 +a2)..;.(1+ g,J -< il . +

-
' ,
-I- " . ;
T.& '.* # V L ? ?&,, I. ,It

.: (1 + a ,)...(l + Q ( 1 +a,,)< s(iiSm)(i~'.m,"i9 . -

Thus, (ii) is true fbrn = m+ 1, mehence t/ n E N.

4. r ' -. <.- .,3 ,
E 22) We apply T h e o m 6 (i) to the n ~ufhbers1 +,,.kla;; ..., 1 - a,.
! .QJ
i 1

1-(1-a++1-%+ ...+ 1-an)4a,% ...a,.

132 Y3J< $ h ,< ' ,%< . , i , *: * .' -, , . ,;; 1 - *: - ,, **&>

(Sir , . ,i
,
r , k y A I , , -

a 1- = a , s n t a i
j=l i-1
r L r
4" rP~ 9::*id(
- ,
jd .+;dfik
i-2 + -d,;=x-. *i,!>+ ,ryT
E23) ~ o . ~ ~ r e x a m ~ l ~ , l e t u s t=-1
a k em i-;- <
a ~d % = 2 . :- ;- ?- '
iF1 j ' . c 41, j . r b ,+dt .
Then 1 - (a, + a2) 1(,I.-
a,) ( 1-
' , . -, , , - , ,
a 0 S2(- 1)=-2,v&ichisfil&.
I '
, i . t .
" '. ,. . __'
... 5xn.Alsob,9 ~ , B ' . : J . I - ~ ~ , , '
I J

EX) Ifweputxi=-4,thenx,sr,s
 E25) Applying T'cheb@hev9$-1iify to (L
1
3
....
1
(1, ,; ..., 2 ,
we get

since T
1
*-
(i-1)i
Vi.

= n {l+(l-;) +...+(A-g}
=n ( l + l - 2 .

1 1
*l+ - +,..+ - s J = = ...(16)
2 n

Again, applying Theorem '7 to [ ~ ~ . and- . * ~ [&-~>--.*~

we get

I E26) Given a, b and c, they can always be ordered. Let us assume that a 5 b 5 c.

1 1 1
Therefore, -2 2 -
a+b a+c b+c'
1 1 1
We apply Theorem 7 (i) to - - -andtoc,b,a.
a+b a+c9b+c
We get

Also, applying T h e o m 7 (ii) to b + c, a+c, a+b and their invmcs, we get

MISCELLANEOUS EXERCISES

This section is optional

As in the previous block, this section contains some extra problems related to the material
covered in this block, Doing them may give you a better understanding of simultaneous linear
equations and inequalities. As before, our solutions to the questions follow the list of
problems.

1) Solve the following linear system, if possible.

2) a) If a + b + c = 0,solve the equation

b) Solve the equation

3) Use Cramer's rule to solve the system

4) A company produces three products, each of which must be processed through three
different departments. In Table 1 we give the number of hours that each unit of each
product must stay for in each department. We also give the weekly capacity of each
department.
Table 1

Department Product No. of hours available

per week

PI p2 p3
1 6 2 2 80
2 7 4 1 60
3 5 5 3 100
-
What combination of the three products will use up the weekly labour availability in all
depamnents?

5) Prove the following identities:

II 6) A fish of species S, consumes 10 gm of food F, and 5 gm. of food F, per day. A fish of ~iscellrneousExercises
I species S, consumes 6 gm. of F,and 4 gm. of F, per day. If a given environment has 2.2
1 kg of F 1 and 1.3 kg of F2 available daily, what population sizes of the species S, and S,
I will consumeexactly all of the available food?
t 7) Does the system
I

x+y+z+w=o
x+3y-2z+w=o

have a non-trivial solution? '

I
8) Obtain a solution of the following system, if it exists.
x+'2y+4z+t=4
2~-~-3t=4

3 ~ - y - Z - 5t=5

9) Show that the following lintar'system has a two-parameter solution set.

2x, +x3-x4+xs=2

x, tx3-x4+xs=1

10) Use Cramer's rule, ifpossible, for solving the following linear systems:

a ) . 3x+y=3

11) If the coordinate axes in a plane are rotated through an angle ,then we can express the old
coordinates (x, y) in t e r n of the new coordinates (x', y') as

y=x'sin0+y'cos8

Use Cramer's rule to write (x', y') in terms of (x, y).

Solutions
1) We apply the Gaussian elimination process. We get the solution set

2 a) Adding the 2nd and 3rd columns to the fmt column of the determinant doesn't change its
value. On d6ing this and using the fact that a + b + c = 0, we get
c b
a
-0
C-X
Equations and Inequalities
1 c
a(-,) o0 ' b -a-c
x-c c-x-b =O,applyingP4twice.

3 (-X) [(b - c - X)(C- b - X)- (a - c)] = 0, expanding along the first column.

On solving this equation, we find that x = 0 or

2x2= 3(a2+ b2+ c2),using the condition that a + b + c = 0.

Thus, the solution set is

b) Using the properties P1 to P5 of determinants, we see that

3) In matrix notation the system is

Here D =
I 3 ' 4
-5
-2
1 =-32~;O.

Thus, we can apply Cramer's rule.

Now

D,=
-1 -2

4) Let x, y and z be the required quantities ofP,, P, and P,. Then we need to solve the
system.

By Gaussian elimination (or Cramer's rule) we get

x=S,y=O,z=25.
 So, the ideal combination is 5 units of P,, 25 units of P,, and none of P,, per week. Mlscellrneous Exerclscr

1 a a2 1 a a2
1 b b2 0 b-a b2-a2 subtracting the f;lrstrow from
=
5, a) 1 c c2 0 c-a c2 -a 2 9 the second and third rows

=
I c+a a+b I
-2p + p p+q ,subtracting the second and third columns from the fmt column.
-2x z + x x+y

a c b
,subtracting the fmst column from the second and third columns.

a b c
,interchanging the second and third columns.

6) Let x and y denote the number of fish of species S, and S2,respectively. We have the
information given in Table 2 below.

Table 2

Food consumed per week (in

Fish
gms)
Spec~es F, F,
s, 10 5

S 6 4

Total food available 2200 1300

per week

Thus, we need to solve the system

10x+6y=2200

solving by any of the methods, we get

Thus, the required sizes are 100 fish of species S, and 200 of species S,.

7) By elimination we find that the system has infrnitely many solutions (x, O,0, - x), where
x E R. Thus, for any x # 0, we would get a non-trivial solution.
Equations and Inequalities 8) By Gaussian elimination, we reach a stage where we get 0 = 817. Hence, the given system
is inconsistent.

9) We apply Gaussian elimination. After a few steps we get the following system
x, = 1

x* +4x4-3x,=o

x3-x,+x,=O.

Thus,x,=l,x,=-4x4+3x,,x3=x4-x,.
So, if x, = s and x, = t, then our solution set is

( ( 1 , - 4 ~ + 3 t , ~ - t , s , t ) I ~R).
,t~

Thus, we have expressed the solutions.interms of the two parameters, s and t.

10)a) In matrix notation, the system is

l3 *
Since D = 5 2 = 1 0, we can apply Cramer's rule.
lI

Thus, the solution is x = 5, y = - 12.

b) Here the coefficient matrix is

Thus, we can apply Crarner's rule.

11) Here we can write the equations as

0
Since D = cOsO( I = ms20+ sin20= 1 + 0,
 Miscellaneous, Exercises
we can apply Cramer's rule. ,

- sin8 cose + y sine, and

,,,e =x