Little Mathematics Library

00
P. P. KO ROVKIN

INEQUALITIES

Mir Publishers. Moscow

SI
c"
Pv
a
IIOIIYJI$IPHbIE JIEHIZHH HO MATEMATHHE

H. H. ICopoBHAH

HEPABEHCTBA

II3j[ATEJIbGTBO «H AYH A»
LITTLE MATHEMATICS LIBRARY

P. P. Korovkin

INEQUALITIES
Translated from the Russian
by
Sergei Vrubel

MI.R PUBLISHERS
MOSCOW
 First Published 1975

Ha auM uU c,co14 mutce

Q Ha)aTenacTBO uHayxa», 1974

Q English translation, Mir Publishers, t975

 CONTENTS

Preface . . . . . . . . . . . . . . . . . . . . . . . . 6

Chapter 1. Inequalities 7
1.1. The Whole Part of a Number . . . . . . . . . . . . . 7
1.2. The Arithmetic Mean and the Geometric Mean . . . . . 12
1.3. The Number . . . . . . . . . . . . . . . . . . . 19
1.4. The Bernoulli Inequality . . . . . . . . . . . . . 23
1.5. The Mean Power of Numbers . . . . . . . . . . . . . 27

Chapter 2. Uses of Inequalities 32

2.1. The Greatest and the Least Function Values . . . . . 32
2.2. The Holder Inequality . . . . . . . . . . . . . . . . 40
2.3. The Use of Inequalities for Calculation of Limits . . . . 43
2.4. The Use of Inequalities for Approximate Calculation of
Quantities . . . . . . . . . . . . . . . . . . . . 49

Solutions to Exercises . . . . . . . . . . . . . . . . . 58

5
 PREFACE

In the mathematics course of secondary schools students

get acquainted with the properties of inequalities and me-
thods of their solution in elementary cases (inequalities
of the first and the second degree).
In this booklet the author did not pursue the aim of
presenting the basic properties of inequalities and made
an attempt only to familiarize students of senior classes
with some particularly remarkable inequalities playing an
important role in various sections of higher mathematics
and with their use for finding the greatest and the least
values of quantities and for calculating some limits.
The book contains 63 problems, 35 of which are provided
with detailed solutions, composing thus its main subject,
and 28 others are given in Sections 1.1 and 2.1, 2.3, 2.4
as exercises for individual training. At the end of the
book the reader will find the solutions to the given
exercises.
The solution of some difficult problems carried out indi-
vidually will undoubtedly do the reader more good than
the solution of a large number of simple ones.
For this reason we strongly recommend the readers to
perform their own solutions before referring to the solutions
given by the author at the end of the book. However, one
should not be disappointed if the obtained results differ
from those of the patterns. The author considers it as
a positive factor.
When proving the inequalities and solving the given
problems, the author has used only the properties of inequa-
lities and limits actually covered by the curriculum on
mathematics in the secondary school.
P. Korovkin.
6
 CHAPTER 1
Inequalities

The important role of inequalities is determined by their

application in different fields of natural science and engi-
neering. The point is that the values of quantities defined
from various practical problems (e.g. the distance to the
Moon, its speed of rotation, etc.) may be found not exactly,
but only approximately. If x is the found value of a quanti-
ty, and Ox is an error of its measurement, then the real
value y satisfies the inequalities
x - I AxI<y<x+IAx{.
When solving practical problems, it is necessary to take
into account all the errors of the measurements. Moreover,
in accordance with the technical progress and the degree of
complexity of the problem, it becomes necessary to improve
the technique of measurement of quantities. Considerable
errors of measurement become inadmissible in solving
complicated engineering problems (i.e., landing the moon-
car in a specified region of the Moon, landing spaceships
on the Venus and so on).

1.1. The Whole Part of a Number

The whole (or integral) part of the number x (denoted by
[ x ]) is understood to be the greatest integer not exceed-
ing x. It follows from this definition that [x] < x, since
the integral part does not exceed x. On the other hand,
since [x] is the greatest integer, satisfying the latter ine-
quality, then [x] + 1 > x.
Thus, [x] is the integer (whole number) defined by the
inequalities
[x] < x C' [x] +,I ..
7
 For example, from the inequalities
3<n<4, 5< 3 <6, -2<-Y2<-1, 5=5<6
it follows that
17
[n] - 3, 1 3 5, 2, [5] = 5.

The ability to find the integral part of a quantity is an

important factor in approximate calculations. If we have
the skill to find an integral part of a quantity x, then taking
[x] or [x] + I for an approximate value of the quantity x,
we shall make an error whose quantity is not greater than 1,
since
0<x-[x]<[x]+1-[x]=1,
0<[x]+1-x<[x]+1-[x]=1.
Furthermore, the knowledge of the integral part of a quanti-
ty permits to find its value with an accuracy up to 12
The quantity [x] + 2 may be taken for this value.
Yet, it is important to note, that the ability to find the
whole part of a number will permit to define this number
and, with any degree of accuracy. Indeed, since
[Nx] <Nx <[Nx] + 1,
then
[Nx] [Nx] + 1
Gx N N

Thus, the number

[Nx] 1
-)V + -IN-
1
differs from the number x not more than by With large
N, .
N the error will be small. The integral part of a number is
found in the following problems.
Problem 1. Find the integral part of the number

8
 Solution. Let us use the following inequalities
1<1 <1,
0.7<j/Z <0.8,
0.5 < V < 0.6,
0.5 < 4 <0.5,
0.4<I/ <0.5
(which are obtained by extracting roots (evolution) with
an' accuracy to 0.1 in excess or deficiency). Combining them
we get
1 + 0.7 + 0.5 + 0.5 + 0.4 < x <
<1+0.8+0.6 d-0.5+0.5,
that is, 3.1 < x < 3.4, hence, [x] = 3.
In this relation, it is necessary to note that the number
3.25 differs from x not more than by 0.15.
Problem 2. Find the integral part of the number
1 1 1 1

x1000000
Solution. This problem differs from the previous one
only by the number of addends (in the first, there were only
5 addends, while in the second, 1000, 000 addends). This
circumstance makes it practically impossible to get the
solution by the former method.
To solve this problem, let us investigate the sum

and prove that

21/n- 1-2Yn<
Vn
<2yn -2Vn-1.
Indeed, since

Vn
a
and
YnT-1>Yn,
it follows that
21/n+1-21/n < 2 - 1

Thereby proof has been made for the first part of the inequa-
lity (1); its second part is proved in a similar way.
Assuming in the inequalities (1) n = 2, 3, 4, ..., n,
we get
21/3-21/2 < <2y-f-2,
1
21/4-21/3 < <21/3-2Y2,
21/5-21/4< -VT
<21/4-21/3,
1
...........................
2V 1-21/n< 1
<21/n-21/n-1.
Vn
Adding these inequalities, we get
21/n-{-1-21/2 <
< 1/12 }
v3 } ... } Vn <2 j/n-2.
Adding 1 to all parts of the obtained inequalities, we
find
21/n-{-1-2V2+1<
< 1-}- 1 + v3
1+ 1
... T 1

vn
<2y-n-1. (2)

Since 21/2 < 3, and 1/n -{-1 > 1/n, it follows from the
inequalities (2) that
21/n-2<1+
v2
+ r/3 + 1
+...
+ vn < 2 1/n =1. (3)

*9
 Using the inequalities (3) we can easily find the inte-
gral part of the number
Y 2 1 1 1

-/2 + V-3 + JA +
=+ + 1000 000
Thus, taking in the inequalities (3) n =1000 000, we get
2 }V1000 000 - 2 <
<1{ -} V3 -f + <2 1000 000 -1,
1000 000

or
1998 < y < 1999.

Hence, [y] = 1998.

From the inequalities (2) it follows that the number
1998.6 differs from y not more than by 0.4. Thus, we have
40
calculated the number y with an accuracy up to 7_998.4 %
= 0.02 %. The numbers 1998 and 1999 differ from the num-
ber y not more than by unity, and the number 1998.5 differs
not more than by 0.5.
Now let us examine the next problem of somewhat diffe-
rent pattern.
Problem 3. Prove the inequality
1 3 , 5 99 1
X - -2-, 4 6 100 10
Solution. Suppose
2 4 6 100
Y3 7 "' 101
Since
1 2 3 4 5 6 99 100
4<5, 6<7100<101
T<,

it follows that x < y and, consequently,

x2 <
_ 1 2 3 4 5 6 99 100 _ 1
xy
2'3'4 6' 7 101 101
100.

Finding the square root of both members of the inequalities

yields
x< 101
<0.1.

ii
 Exercises
1. Prove the inequalities
2Yn+1-2Ym<Vm+ 1
+...
...+ Vn- <2Yn-2Ym-1.
2. Prove the inequalities

1,800 <
10,000 + 1/10,001 + + x/1,000,000

< 1,800.02.
3. Find [50z], where
z
]/10000
+ 1x/1 0,001 + ... + x/1,000,000
_

Answer. [50z] = 90,000.

4. Prove the following inequality using the method of
mathematical induction
1 3 5 2n-1 < 1

2 ' 4' 6 "' 2n 3n -1

5. Prove the inequality
1 3 5 99 1
2*4'6 "' 100[12
1.2. The Arithmetic Mean and the Geometric Mean
If x1, x2, ..., xn are positive numbers, then the numbers
formed with them
a = x1 x2'}'n ... +an
g = n x1x2 ... X-
are called, respectively, the arithmetic mean and the geo-
metric mean of the numbers x1, x2i . . ., xn. At the begin-
ning of the last century, the French mathematician 0. Cau-
chy has established for these numbers the inequality
g < a,

often used in solving problems. Before proving the inequali-

ty we have to establish the validity of an auxiliary assertion
12
 Theorem 1. If the product n of the positive numbers tj,
x21 ..., xn is equal to 1, then the sum of these numbers is
not less than n:
xlx2f . . ., xn = 1 = x1 + x2 + . . . + xn > n.
Proof. Use the method of mathematical induction'. First
of all check up the validity of the theorem for n = 2, i.e.
show that
x1x2=1=x1+x2>2.
Solving the question, examine the two given cases separa-
tely:
(1) x1 = x2 = 1.

In this case xl + x2 = 2, and the theorem is proved.

(2) 0 < x1 < x2
Here xl < 1, and x2 > 1, since their product is equal to 1.
From the equation
(1 - x1) (x2 - 1) = x2 + x1 - x1x2 - 1
it follows that
x1 + x2 = x1x2 + 1 + (1 - x1) (x2 (4)
The equation (4) has been established without limitations to
the numbers x1 and x2. Yet, taking into account, that
x1x2 = 1, we get
x1-+2=2+(1-x1)(x2-1).
At length, since x1 < 1 < x2, then the last number is
positive and x1 + x2 > 2. Thus, for n = 2 the theorem is
already proved. Notice, that the equation
x1+x2=2
is realized only when xl = x2. But if x1 x2f then
x1+x2>2.
Now, making use of the method of mathematical induc-
tion, assume that the theorem is true for n = k, that is, sup-

1 More detailed information concerning mathematical induction

is published in the book by I. S. Sominsky "The Method of Mathemati
cal Induction", Nauka, Moscow, 1974.
13
pose the inequality
x1+x2+x3+. .. +xh>k
occurs, if x1x2x3. . . xh = 1, and prove the theorem for
n = k + 1, i.e. prove that
X1 +X +x3+ . . . +xh+xh+1>k+ 1,
if x1x2x3 . . . xkxk+l =1, for x1 > 0, X1>01 x3 > 0, ...
xh>0, xk+1>0.
First of all, it is necessary to notice that if
. xkxh+l = 1,
X1X2X3 . .

then there may be two cases:

(1) when all the multipliers x1, x2i x3, ..., xh, xk+1 are
equal, that is
x1 = x2 = x3 ` xh = xk+1,
(2) when not all multipliers are equal.
In the first case every multiplier is equal to unity, and
their sum equals k + 1, that is
x1+x2 +x3+ . .. +xh+xh+1 =k+ 1.

In the second case, among the multipliers of the product

X1X2 . .. xkxh+l, there may be both numbers greater than
unity and numbers less than unity (if all the multipliers
were less than unity, then their product as well would be
less than unity).
For example, suppose x1 < 1, and xk+l > 1. We have
(xlxh+l) X2x3 . . . xh = 1.

Assuming yl = xlxk+lf we get

y1x2x3 . . . xh = I.
Since here the product k of positive numbers is equal to
unity, then (according to the assumption) their sum is not
less than k, that is
y1+x2+x3+. .. +xh>k.
But
x1+x2+ x3+ +xh+xk+1 =
= (yl+x2+x3+. . +xk)+xk+1-y1+xl>
k+xk+l-yl+x1=(k+1)+xh+l-yl+xl-1
14
 Remembering that y1 = xlxk+l we get
XI + X2 + X3 + . . . + xk + xk+1 >
>(k+ 1) +xh+l - xlxk+l+xl - 1 =
=(k+1)+(xk+1-1) (1-x1).
Since xl < 1, and xh+l > 1, then (xh+l
and, hence,
- 1) (1 - x1) > 0
xl+x2+x3+ +xk+xk+1>
>(k+1)+(xh+l-1)(1-xl)>k+1.
Thus the theorem is proved.
Problem 1. Prove, that if x1, x2, x3i ..., x, are positive
numbers then
x2 } S3
} ... + xn' m x! n,

the equality being valid only when

x1=x2=x3= ... =xn.
Solution. Since
x2
x1
X2 x3
... xn-1
xn
xn
x1

then the inequality follows from Theorem 1, the sign of

equality holds only when

xn x1

namely, when x1 = x2 = x3 = . . . = xn.

Problem 2. Prove the inequality
x2+2
V52+1
2.
Solution. We have
X2 2 _ x21 1 _Yx2+1+ 1

-/x2+_j j/x2+1 + /x2+1 /x2+1

Since the product of addends in the right-hand member of the
equality equals unity, then their sum is not less than 2.
The sign of equality holds only for x = 0.
Problem 3. Prove that for a > 1
log a + logo 10 > 2.
15
 Solution. Since log,, 10 -log a = 1, then
log a + toga 10 - log a -f - toga > 2.
Problem 4. Prove the inequality
z2 1
1-{-x4 2
Solution. Divide by x2 the numerator and denominator
of the left-hand member of the inequality:
X2
1+x4 = 1
+xa
a 1
x

Since x2= 1, then -- x2> 2 and, hence,

x X 1
1
X2
2
z2+
Now let us prove the statement made at the beginning
of the section.
Theorem 2. The geometric mean of positive numbers is
not greater than the arithmetic mean of the same numbers.
If the numbers x1, x3i . ., xn are not all equal, then
.

the geometric mean of these numbers is less than their arithmetic

mean.
Proof. From the equality g = v xix2 ... X" it follows
that
x1 x2 2n xn
or x1 x2 =1.
9 9 9 9 9 9

Since the product n of the positive numbers equals 1,

then (Theorem 1) their sum is not less than n, that is

9 9 9

Multiplying both members of the last inequality by g and

dividing by n, we get
a= xi+x2+...+xn
n g
16
Notice, that the equality holds only when ?r = x2 = .
9 9
= = 1, that is x1 = x2 = ... == xn = g. But if the
g x1, x2i ... , xn are not equal, then
numbers
avg.
Problem 5. From all parallelepipeds with the given sum
of the three mutually perpendicular edges, find the parallele-
piped having the greatest volume.
Solution. Suppose m a + b + c is the sum of the
edges and V = abc is the volume of the parallelepiped.
Since
V=3 abc< a +b3 } c m
3 '
3
then V-<_ 27 . The sign of equality holds only when a
= b-= c=7, that is, when the parallelepiped is a cube.
Problem 6. Prove the inequality
n!<(n21)n, (5)
Solution. Using Theorem 2, we get
n
o t= 1.2.3 ... n< 1+2+3+... -i- n n -
(n+ 1) n n+1
2n 2

Raising to the nth power both parts of the last inequality,

we get the inequality (5).
Definition. The number
Ca- a1+a2 +an)a
n
is termed the mean power of numbers a1, a2, ..., an of the
order a. Particularly, the number
_ a1-- a2+... { an
Cr n
is the arithmetic mean of the numbers ar, a2, ..., an, the
number
C2 - (a 1
-{- a22 -F ... -{- an 2

n )
2-0866 17
is named the root-mean-square, and the number

C'1
a11+a21... {
n
_ant)-1n 1 1 1

Q1 a2
+---+ an }

is called the harmonic mean of the numbers a1, a2, ..., an,
Problem 7. Prove that if a1, a2i ..., an are positive
numbers and a < 0 < 3, then
ca < g < c o, (6)
that is, the mean power with a negative exponent does not
exceed the geometric mean, and the mean power with
a positive exponent is not less than the geometric mean.
Solution. From the fact, that the geometric mean of
positive numbers does not exceed the arithmetic mean,
we have
a a ... any
ala2 a n
a1+a2+...+an

Raising both parts of the last inequality to a power 1 and

taking into consideration, that < 0, we get
1

a1+a2+...-f-an a
g a1a2 ... an > ( n ) = ca.

So the first part of the inequality (6) is proved; the second

is proved in a similar way.
p" From the inequality (6) it follows, in particular, that
the harmonic mean c_1 does not exceed the arithmetic
mean c1.
Problem 8. Prove that if a1, a2, ..., an are positive
numbers, then
(a1 + a2 + ... +an) ( a1 -{- ... + n) , n2.
2
Solution. Since c_1 <g <c1i then
-1
n < a1+a2+...+an =Cj.
n
--F-H-...+--
1

a1 a2
1
an
1

it follows from this inequality that

n2 < (at + a2 + ... - an) ( a+ a+ ... + an
18
 Problem 9. Prove the inequality
nala2 ... a, < ai -}- a2 + ... H- an, (7)
where a1 > 0, a2>0, ..., a,, > 0.
Solution. Since the geometric mean does not exceed the
arithmetic mean, then
a1a2 ... an = V n ... ann n+ n
V na1a2
Multiplying both members of this inequp1}ty by n, we shall
get the inequality (7).
From the inequality (7) it follows, that
2a1a2 < a; +a2, 3aja2a3 G a'+ a,+ a3,
4aja2a3a4 < a; + az + a3 + a4,

that is, the doubled product of two positive numbers does not
exceed the sum of their squares, the trebled product of three
numbers does not exceed the sum of their cubes and so on.

1.3. The Number e

The number e plays an important role in mathematics.
We shall come to its determination after carrying out the
solution of a number of problems in which only Theorem 2
is used.
Problem 1. Prove that for any positive numbers a, b,
(a b) the inequality
n+1/abn a+nb
n+1
is true.
Solution. We have
n

n+/abn = n+V abb ... b < a - b b ... }b a { nb

n+1 n+1
n
and that suits the requirement.
Problem 2. Prove that with the increase of the number
n the quantities
_( 1 xn-(1+n)n

and 1 zn1n)n

2* 19
 increase, i.e.
1 n+1
xn < xn+1 = (1 + n+1 )
,

1 \\n+1
zn<Zn+1= (t - n+t )
Solution. Setting in the inequality of the previous prob-
lem a=t, b=t+ 1
, we get

n+1 1+n
t t 1

n
1

Raising both parts of the inequality to the (n + 1)th power,

we shall obtain
(t + 1 ) n < (t + n+ 1) n+1 , that is xn < xn+i
1 1

The second inequality is proved in a similar way.

Problem 3. Prove that
yn=(t+ 1 1n+1
n )
decreases with the increase of the number n, that is
1 )n+2
Yn>yn+1= (t + n+1
Solution. We have
)n+1 n -i-1 )n+1
yn=(t+ 1
n n
= n n+1

( n+1)
1 _ 1
( 1 )fl+1 Zn+1
n+1
(se e designations of Problem 2). Since zn increases with
the increase of the number n, then yn decreases.
In Problems 2 and 3 we have proved that
x1= (t + 1 )1 =2<x2= (t + 2 )2=
=2.25<x3< ... <xn< ...,
y1= (t + 2 )2==4>y2=
(t + 1)'=3.375>Y3> .. > yn > ...

20
On the other hand,
2 =x1<xn= (1 + n )n< n In+t-yn<y1-4.
(1 H-
Thus, the variable xn satisfies two conditions:
(1) xn monotonically increases together with the increase
of the number n;
(2) xn is a limited quantity, 2 < x, < 4.
It is known, that monotonically increasing and restricted
variable has a limit. Hence, there exists a limit of the
variable quantity xn. This limit is marked by the letter e,
that is,
e= Jim x= lim(1+
n-.oo n-.oo
-)
1

n
n.

As the quantity xn increases reaching its limit, then xn is

smaller than its limit, that is
xn= (1 + n )n<e. (8)

It is not difficult to check that e < 3. Indeed, if the num-

ber n is high, then
1
xn<yn<y5 (1 ± 5) = 2.985984.
Hence,
e = lim xn <, 2.985984 < 3.
n-.oo

In mathematics, the number e together with the number at

is of great significance. It is used, for instance, as the base
of logarithms, known as natural logarithms. The logarithm
of the number N at the base e is symbolically denoted by
In N (reads: logarithm natural N).
It is common knowledge that the numbers e and at are
irrational. Each of them is calculated with an accuracy of
up to 808 signs after the decimal point, and
e = 2.7182818285490 . . . .

Now, let us show that the limit of the variable yn also

equals e. Indeed,
limy"=lim(1+n)n+1 .1im(1+1)n(1
n
e-1 =e.
21
Since y,ti diminishes coming close to the number e (Pro-
blem 2), then
i 1) 1 n+i
>e. (9)

Problem 4. Prove the inequality

n! > (en y. (10)

Solution. We shall prove the inequality (10) using the

method of mathematical induction. The inequality is easily
checked for n = 1. Actually,
11=1> e

Assume, that the inequality (10) is true for n = k, that is

k!>(e)k.

Multiplying both members of the last inequality by k+ 1,

we get
(k+1)!>(k)k(k+1)=( keg `k+1 e1
(k+ 1) k! = k
J +k)
)k
Since, according to the inequality (8) (1 + < e, then
(k+1 )k+1
(k+1)!> (k71 )k+1 e
k

e e e

that is the inequality (9) is proved for n = k + 1. Thus

the inequality (9) is proved to be true for all values of n.
Since e < 3, it follows from the inequality (9) that
n! > ( 3 )n.
By means of the last inequality, it is easy to prove that
300! > 100300
Indeed, setting in it n = 300, we get
3001 >
( 330) 300 _ 100300.

22
The inequality
n!<e( n+1 )n+1
e
is proved completely the same way as it is done with the
inequality of Problem 4.

1.4. The Bernoulli Inequality

In this section, making use of Theorem 2 we shall prove
the Bernoulli inequality which is of individual interest
and is often used in solving problems.
Theorem 3. If x > -1 and 0 < a < 1, then
(1 +x)a + ax. (11)
However if a < 0 or a > 1, then
(1+x)01 >1+ax. (12)

The sign of equality in (11) and (12) holds only when x = 0.

Proof. Suppose that a is a rational number, bearing in
mind that 0 < a < 1. Let a = , where m and n are
positive integers, 1 < m < n. Sincen
according to the condi-
tion, 1 + x > 0, then
M
(1+x)a=(1+x) n =V(1+x)m,1n-m _
=V(1+x)(1+x) ... 1

m n-m

n
m (I ±x)+n-m n+mx
n n
1
+m
n
X. 1 + ax.
The sign of equality occurs only when all multipliers stand-
ing under the root sign are identical, i.e., when I + x = 1,
x = 0. But if x 0, then
(1 + x)a < 1 + ax.
Thus, we have proved the first part of the theorem conside-
ring the case, when a is a rational number.
23
 Assume now, that a is an irrational number, 0 < a < 1.
Let r1, r2, . . ., r, . . . be the sequence of rational numbers,
having for a limit the number a. Bear in mind that 0 <
< r < 1. From the inequalities
(1 + x) rr <1 + r,,x, x > -1, n = 1, 2, 3, ...,
already proved by us for the case when the exponent is
a rational number, it follows that
(1-J x)a=- Jim (1+x)'n<lim(1+rrx)=1+ax.
rn-.a rn-.a
Thus the inequality (11) is proved for irrational values of a
as well. What we still have to prove is that for irrational
values of a when x* 0 and 0 < a< 1
(1 + x)a < I -r- ax,
i.e., that when x 0 in (11), the sign of equality does not
hold. For this reason, take a rational number r such that
a < r < 1. Obviously, we have
(1+x)a=[(14_x)r]r.
a < 1, then as it has already been proved
Since 0

(1 -Fx)r + X.
Hence,
('J+x)a<(1 + a x)r.
If x 0, then (1 -F- x) r < I +r a x- 1 -;- ax, that is
(1 +x)a <1 + ax.
Thus the first part of the theorem is proved completely.
Now, move on to proving the second. part of the theorem.
If 1 + ax < 0, then the inequality (12) is obvious, since
its left part is not negative, and its right part is negative.
If 1 -+- ax > 0, ax > -1, then let us consider both
cases separately.
Suppose a > 1; then by virtue of the first part of the
theorem proved above we have
1

(1+ax)a <1+ a ax=1+x.

24
Here the sign of equality holds only when x = 0. Raising
both parts of the last inequality to the power a we get
I -- ax <' (1 + x)a.
Now let us suppose a < 0. If 1 + ax < 0, then the ine-
quality (12) is obvious. But if I + ax > 0, then select the
positive integer n, so that the inequality - n < I would
be valid. By virtue of the first part of the theorem we get
a
(1 { x) n <1-a
x,

(1+a1--x
>1 { a
n
x
n
2
the latter'inequalit.y is true, since 1 > 1- n2 x2) . Raising
both parts of the latter inequality to the nth power we get
(1 +n x)n>1+nn x=1+ax.
Notice, that the equality is possible only when x = 0.
Thus, the theorem is proved completely.
Problem 1. Prove, that if 0 > a > -1, then
(n+1)a+1_na+1 na+1(n-1)a+1
<na< +1 (13)

Solution. Since 0< a +1 <_ 1, then according to the

inequality (11) we have
(1 + `a+1 < 1+ an 1
n
(1-1)a+'<1-a+1
n n
Multiplying these inequalities by na+1 we obtain
(n+1)a+1<na+1+(a+1)na,
(n-1)a+1 < na+1 - (a-{-1) na.
The inequalities (13) easily follow from these inequa-
lities.
25
 Problem 2. Prove, that if 0 > a > -1, then
(n+1)a+1-ma+1
a+1
-}-na<na+1 (+1)a+i
<ma+(m+1)a+ ... (14)
Solution. Setting in the inequalities (13)n= m,
m+1, ..., n, we get
(m+1)1+a-m1+a m1+a-(m-1)1+a
m<a<
1+a 1+a
(m+1)1+a-mi+a
(m+2)1+0'-(m+1)i+a <(m-1)a<
1+a 1+a
(m+3)1+a-(m+2)1+a < (m+2)'< (m+2)1+a-(m+1)1+a
1+a 1+a
.................................
(n+1)1+a-n1+a
1+a <naG n1+a-(n-1)1+a
1+a
Adding these inequalities we shall get the inequality (14).
Problem 3. Find the integral part of the number
x + + 1 1 1

35 3s ... + 31,000,000

33
Solution. Setting in the inequality (14) m = 4, n =
=1,000,000, a = - 3 , we get
2 2 2 2

1,000,013 -43 <x< 1,000,002 3 -33

that is
- 2.4 3 <x<2. 1,000,000 3 - 2 ,3 3
2 2 2 2
.
2.1,000,0013

Since
2 2

. 1,000,001 3 > 2 . 1,000,000 3 =2.10,000=15,000,

2 2 VT-6 - V54 < 4, 2 V9 > 2 V8 = 3,
then
15,000 -4 < x < 15,000 - 3, that is 14,996 < x < 14,997.
From these inequalities it follows that [xl - 14,996.
26
 1.5. The Mean Power of Numbers
In Sec. 1.2 before Problem 7 we have already named the
number
i
as { a2 + ... -{- an a
Ca== n )
the mean power of order a of the positive numbers a1i
a2, ..., an. In the same problem, it has been proved, that
Ca <C g, if a< 0< P.
Here, should be proved the validity of the inequality
Ca < co any time when a < P. In other words, the mean
power of order a is monotonically increasing together
with a.
Theorem 4. If a1, a2, ..., an are positive numbers and
a< then ca < cs, and ca = cg, only when a1 = a2 =
= an.
Proof. For the case, when the numbers a and have
different signs the theorem has been proved above (refer
to Problem 7, Sec. 1.2 and the definition prior to it). Thus,
we have to prove the theorem only for the case when a
and P have the same signs.
Assume, that 0 < a < 3, and let
1
k_ca_ (a1 a2-f-...+an\a
n J

Dividing co by k, we get
/a2
CO CO
(a1
is )
f1

h k)
(1

I
... an
k \I31 1
Ca I n

Now, supposing
d1=(k )a, d2k )a, ..., do k )a,
we obtain

CO _ da d2 -{- ... - do 1
(15)
k n )
27
 Since

( di+d2--...+dn)
n

_((a)a+(a)a+-_)ay 1

k (

n = k ca= a
Ca=-1r

di+d2-L ...+dn = 1, di+d2+ ... +dn= n.

n
Suppose
di = 1 +x1, d2 = 1 +x2, ...,
d,, = I +xn.
From the equality d1 + d2 .+.. + do n it follows that
x1d-x2+. . . -j-xn = 0.
On the basis of Theorem 3 (notice, that a > 1) we have

do' = (1 + x1) > 1 +A xi,

d2 = (1 -F x2)" -I+ x2,

................
a
(*)

do = (1 + xn) " > 1-}- a x,t .

Adding these inequalities, we get

d1 +d2 +... } do >n+ a (Xi+X2+....+x.)=n. (16)

From the inequalities (15) and (16) it follows that
Co _(n)s=1, co>k=ca.
28
 It is necessary to note that c13 = k = ca only when the
signs of equality occur everywhere in (*), that is when
x1 = x2 = ... = x,, - 0 (Theorem 3). In this case d1 =
= d2 = ... = dn, = 1 and, hence, a1 a2 = ... = a =
= k. But if the numbers a1, a2, ..., a, are not identical,
then
Ca > Ca.

Thus Theorem 4 is proved regarding the case when 0 <

<a<P.
If a < P < 0, then 0 < a < I. Reasoning the same way
as before, we get in (*) and (16) the opposite signs of ine-
qualities. But since < 0, then from the inequality
f3

da -{-d2 +...+do
_<1

it follows that
rs f3 p
1

1
CO p d1 { d z } ... - F d na
k n

that is
CO>k=Ca.

Thus, Theorem 4 is proved completely.

Further on we shall name the geometric mean by mean
power of the order zero, that is, we shall assume g = co.
Notice, that Theorem 4 is applicable in this case as well,
since (see Problem 7, Sec. 1.2) ca < g = co, if a < 0,
and c5 g-c0, ifI >0.
From the proved theorem it follows, in particular, that
C _1 C CO C C1 1 C21

i.e. the harmonic mean does not exceed the geometric mean,
the geometric mean in its turn does not exceed the arithme-
tic mean, while the arithmetic mean does not exceed the
root-mean-square of positive numbers. For example, if
29
a1= 1, a2 2, a3 = 4, then
(aiI+a2I a31 `-1- =
c_ 1=` 3 ) 1+ 1
3
7 =1.7
12
.

co=Vaja2a3=V1.2.4=2,
c1= 1+3+4 = 7 = 2.3 ...,
1

2
=Y 1+4+16
C2= 3 J 3 V7=2.6
and therefore
c_1=1.7...<2=co<2, 3... =c1<2.6... =c2.
Problem 1. Prove, that x2 + y2 -F- z2 > 12, if
x + y + z = 6.
Solution. Since the arithmetic mean does not exceed
the root-mean-square, then
x -3.1-z f x2-} y2-}-z2 ) 2
l 3 J

that is
x2+y2+z2> (x+3+Z)2
In our problem x2 + y2 + z2' 32 = 12. The sign of equa-
lity holds only when x = y = z = 2.
Problem 2. Prove, that if x, y, z are positive numbers
and x2 + y2 + z2 = 8, then
x3+y3+z3> 16 3

Solution. Since c2 <c3i then

1 1

x2 } y2 } z2 \ 2 ( x3+y3+z3 \ 3
3 3 1

3
I\ 1

In our problem
In
x3±-y3+z3 1 3 8
3 J

30
that is
x3+y3+z3j3' 3t/T=16
3 3.
Problem 3. Prove, that for positive numbers al, a2i
a3, ., an, the following inequalities are true
. .

(at a2 1f ... +an)a Cna ' (aa+a2 -E- ... -}-an), a> 1, (17)
(a1 +a2 + .. +an)a>
>na(a"'+a2-}-...+an, 0<a<1. (18)
Solution. If a> 1, then

C,
( ai-F2--...+an \ a a1+a2+...+an
n 1 n = C1.

The inequality (17) follows easily from this inequality.

The inequality (18) is proved in exactly the same way.
In particular, from the inequalities (17) and (18) it follows
that
(x + y)a < 2a-1(xa + ya), a> 1, x > 0, y > 0,
(x+y)a>2a-1(x"+y"), 0<a<1, x>O, y>O.
Problem 4. Prove, that if x3 + y3 + z3 = 81,
x > 0, y > 0, z > 0, then
x+y --f- z<' 9.
Solution. Since
(x+y+z)3<, 32(x3+y3+z3) =9.81 =729
(the inequality (17)), then
x+y+z729=9.

31
 CHAPTER 2

Uses of Inequalities

The use of inequalities in finding the greatest and the

least function values and in calculating limits of some
sequences will be examined in this chapter. Besides that,
some important inequalities will be demonstrated here as
well.
2.1. The Greatest and the Least Function Values
A great deal of practical problems come to various func-
tions. For example, if x, y, z are the lengths of the edges
of a box with a cover (a parallelepiped), then the area of
the box surface is
S - 2xy + 2yz + 2zx,
and its volume is
V _ xyz.
If the material from which the box is made is expensive,
then, certainly, it is desirable, with the given volume of
the box, to manufacture it with the least consumption of
the material, i.e., so that the area of the box surface should
be the least. We gave a simple example of a problem consi-
dering the maximum and the minimum functions of a great
number of variables. One may encounter similar problems
very often and the most celebrated mathematicians always
pay considerable attention to working out methods of their
solution.
Here, we shall solve a number of such problems, making
use of the inequalities, studied in the first chapter'. First
of all, we shall prove one theorem.
1 Concerning the application of inequalities of the second degree
to solving problems for finding the greatest and the least values see
the book by I.P. Natanson "Simplest Problems for Calculating the
Maximum and Minimum Values", 2nd edition, Gostekhizdat, Moscow,
1952.
32
 Theorem 5. If a > 0, a > 1, x y 0, then the function

xa - ax takes the least value in the point x= (a)

(1-a)(a)a-1.
equal to a
Proof. The theorem is proved very simply for the
case when a=2. Indeed, since
a
x2-ax= (x- a)2

4a2 '
the function has the least value when x = a > 0, this
value being equal to -
4
In case of arbitrary value of a > 1 the theorem is proved
by using the inequality (12), demonstrated in Theorem 3.
Since a > 1, then
(1+z)a+az, z
the equality holding only when z = 0. Assuming here,
that 1 + z = y, we get
ya>1 +a(y-1), ya - ay>I -a, y>0,
the sign of equality holds only when y = 1. Multiplying
both members of the latter inequality by ca, we get
(cy)a - aca'1 (cy) > (1 - a) ca, y > 0.
Assuming
-)
1

aca-1=a, a as-1
x=cy and c= ( a a

we get
a
xa-ax>(1-a)ca=(1-a) (a ) a-1 A

here the equality occurs only when x = c = (a)

a=1
a
Thus, the function
xa - ax, a >1, a>0, x>0,
1

a
takes the least value in the point x= equal to
a
a-1
(1- a) ( a) . The theorem is proved.
3-0888 33
 In particular, the function x2 - dx (a= 2) takes the
1

least value in the point x = (2 )L 1

= a , equal to
z

(1-2) ( a ) 2-1 =-4 . This result is in accordance with

the conclusion, obtained earlier by a different method. The
function X3-27x takes the least value in the point
1 3
3-1
x- (2) 3-1 =3, equal to (1-3) (2) = 54.

Note. Let us mark for the following, that the function

ax -,xa = -(xa - ax),
where a > 1, a > 0, x > 0, takes the
greatest value in the point
X=
I
1 a
equal to
a

Fig. 1
(a-1)(a l
Problem 1. It is required to saw out a beam of the grea-
test durability from a round log (the durability of the beam
is directly proportional to the product of the width of the
beam by the square of its height).
Solution. Suppose AB = x is the width of the beam,
BC = y is its height and AC = d is the diameter of the log
(Fig. 1). Denoting the durability of the beam by P, we get
P = kxy2 = kx (d2 - x2) = k (d2x - x3).
The function d2x - x3 takes the greatest value when
1

X= (d2
3
`
3-1
J
-- d y2=d2-x2= 23 d2,

y= d1/9 Y2=xV2
34
Thus, the beam may have the highest (greatest) durability
if the ratio of its height to its width will be equal to V/2 x
1.4- 5.

Problem 2. Find the greatest value of the function

y = sin x sin 2x.
Solution. Since sin 2x = 2 sin x cos x, then sin x sin 2x=
= 2 cos x sine x = 2 cos x (1 - cost x) = 2 (z - z3), where
z =:cos x and, hence, -1 < z < 1. The 'function z - z3 =
Z (1 - z2) takes a negative value when -1 < z < 0,

Fig. 2

is equal to 0 when z = 0 and takes a -positive value when

0 < z < 1. Therefore, the greatest value of the function
is gained in the interval 0 < z < 1.
It is shown in Theorem 5 that the function z - z3, z > 0,
takes the greatest value in the point

Z=(13) 3-1 - '1

In this point

sin x sin 2x = 2z (1- zz) =

V-3 ( 1 - T3) = 3 i/3
So, the function y sin x sin 2x takes the greatest value
in those points, where z cos x = _ and this value is
3
equal to 0.77. The graph of the function y =
3 _/3
= sin x sin 2x is shown in Fig. 2.
Problem- 3. Find the greatest value of the function
y = cos x cos 2x.
3* 35
 Solution. The function y = cos x cos 2x does not exceed
1, since each of the cofactors cos x and cos 2x does not
exceed 1. But in the points x = 0, ±2n, ±4n, .. .
cos x cos 2x = 1.

Thus, the function y = cos x cos 2x takes the greatest value

of I in the points x = 0, ±2n, ±4n, ...
. The graph of
the function y = cos x cos 2x is drawn in Fig. 3.

Fig. 3

Problem 4. Find the least value of the function

xa + ax,
where a ] 0, a < 0, x >,O.
Solution. Since a < 0, then according to the inequali-
ty (12)
(1 + z)a > 1 + az,
and the sign of equality holds only when z = 0. Assuming
1 -{-z=y, z=y-1, we get
ya>1+a(y-1), Y>0,
the sign of equality occurring only when y = 1. From the last
inequality it follows, that
y17 -ay>1 - a, (cy)"-aca_'(cy)>(1 a) ca.
Assuming a = -aca-1, x = cy, we get
a
) a-1
A
a

a ) a= 1
the equality holding only when x = c (

36
 Thus, the function xa -{- ax takes the least value in the
point
X='
as )a 1'
a
a ) a-1
equal to (1 -a) .

For example, the function

+27x, x> 0,
takes the least value in the point
1

1
-1
27 3 1

3
This value equals
3
1 _1

(14- 31 /
1
27

3
- 3
=4.

Problem 5. Find the optimum dimensions of a cylin-

drical tin having a bottom and a cover (dimensions of a
vessel are considered to be the most profitable, if for a given
volume the least amount of material is required for its
manufacture, that is, the vessel has the least surface
area).
Solution. Let V = nr2h be the volume of the vessel,
where r is the radius, h is the height of the cylinder. The
total surface area of the cylinder is
S = 2nr2 + 2nrh.
Since h = V , then

S = 2nr2 + 2nr r2 = 2nr2 + 2V

r
.

1
Assuming x = T , we get
$=2nx 2+2Vr=2n (X-2 -
a7
The function x 2+ V x, according to the solution of the
previous problem, takes the least value when

X= / V
l2n
-2-1 ; / V
V
Returning back to our previous designations, we find
1 _ 2n V nr2h h
713 ,r3=
2a= 2a 'r-2
h = 2r = d.
Thus, the vessel has the most profitable dimensions, if the
height and diameter of the vessel are equal.
Exercises
6. Find the greatest value of the function x (6 - x)2
when 0<x<6.
Indication. Suppose y = 6 - x.
7. From a square sheet whose side is equal to 2a it is
required to make a box without a cover by cutting out a
square at each vertex and then bending the obtained edges,

H Ej

20
a-2x

Fig. 4

so that the box would be produced with the greatest volume

(Fig. 4). What should the length of the side of the cut-out
squares be?
8. Find the least value of the function
x6,+ 2 -f-
36
 9. Find the least value of the function
xs - 8x2 + 5.
10. Find the greatest value of the function
xa-ax when O<a<1, a>O, x>O.
It. Prove that, when x > 0, the following inequality
is true
3x<8+ 2x.
12. Prove that, when n > 3, the following inequality is
true
rn>n+/n +1.
Indication. Make use of the inequality (8).
13. Find the greatest of the numbers
1, -1/12, V3, V4, y5, ..., fn, ... .

14. Prove the inequality

rn<1+ 2vn
15. Prove the inequality
(1 + al) (1 + a2) . . (1 + an)
.

>1+al+a2-}-...+an,
if the numbers ai are of the same sign and are not less
than -1.
16. Prove the inequality
(aib1 + a2b2 + ... + anbn)2
(a,+a2+...+a')(b;+bz+... +bn). (19)
Indication. First prove, that the polynomial
(a1x - b1)2 -I- (a2x - b2)2 + . . . + (a,,x - b,,)'
x2(a2+a2+ +an) -
- 2x (a1b1 + a2b2 + ... + a,, b,,) +
+ (b i + b z -{- . . . + bn )
cannot have two different real roots.
17. Using the inequality (19), prove, that the arithmetic
mean is not greater than the root-mean-square.
39
 18. Prove the inequality
_n <Vn 1-Vn-1.
19. Using the inequality of Exercise 18, prove the inequa-
lity

vZ v3 vn
20. Find the greatest value of the functions
x3
x4+5 , x0-0.6x10.

Answer. 4 ° 315 ; 0.4.

21. At what value of a is the least value of the function
V-+ x2 equal to 2.5?
Answer. a = 8.

2.2. The Holder Inequality

In Theorem 7, by means of Theorems 5 and 6, the Holder
inequality is proved. This inequality will find application
in solving problems.
Theorem 6. If p > 1, 1 + 4 =1, x>0, y > 0, then
xy < P -{ 4 (20)
Proof. By virtue of Theorem 5, if a < 1, a > 0, x > 0,
then
a
xa-ax> (1-a) - a
a
a-1

Assuming in this inequality that a = p, a = py, we get

P
P
XP-(Py)x>(1-P) (py )' 1=(1-P)yP-'. (21)

Since P -}- 4 =1, then

q
=1- P`p-
, q= ppq , P-1=
40 p
Putting these values into the inequality (21), we get
xn-pyx> - 9 y4.
Dividing all the members of the latter inequality by p and
transposing the negative members to the opposite side, we
get the inequality (20).
Theorem 7. If a1, a2f ..., an, b1, b2i ..., bn are posi-
tive numbers, and p and q satisfy the conditions of Theorem 6,
then
ajb1 + a2b2 + ... + anbn
1 1

(a' + a2 -I- ... + an) p (b? + b2 + ... + bn) q (22)

Proof. Suppose
ai+a2+ ... +an=Ap, bi I bz+ ... + bn=B4.
Then the right member of the inequality (22) will be equal to
1 1

(Ar) P (B4) 4 = AB.

Now suppose
a1 = Act, a2 = Act, ... , an = Acn,
b1= Bd1i b2 Bd2, ... , bn -{- Bdn.
Since
AP=api +a2+ ... + an
=APcPj +APcP2 +...+Apcn=Ap(ci+cP2 +...4cn),
then
Ci-f-CZ+...+cn=1.
In a similar way, it is checked that
dI+d2+...+dn=1.
Now using the inequality (20), we get

a1b1= AB (c1d1) <AB (P + 91 )

a2b2<AB (p + 92) ,
(')

GAB c -}- d"

From these inequalities it follows, that
a1b1 + a2b2 -}- ... + anb,,
ci+c2+...+cn di+d2+... do
l
<AB + 4 /
=AB(p-} q)=AB
(let us recall that

1
F-q=1,
7

p
di d2+...+d9n=1).
Thus, it is proved that the left-hand member of the inequa-
lity (22) does not exceed AB, that is, does not exceed the
right-hand member.
It is not difficult to mark the case when the sign of equali-
ty is valid in (22). Indeed, the sign of equality holds in (21)
only when
1 1 9

(refer to Theorem 6). Just in the same way, the equality

sign will be valid in each line of (*) only when
9 9 4

ct=d1p , c2=d2P , ..., Cn=dnp

i.e., when
p 9 p P
C1 = d1, C2 = d2, ..., Cn = dn.
o

Finally, multiplying these equalities by APB9 we get

B9 (Ac1)p = AP (Bd1)9, that is, B9ap = Apbi,
a1 A a2 an Ap
b1
= BQ ' b2
= Ap
B9
, .,b= B9

Thus, in (22) the sign of equality is valid if

ap a2
b1b2=...=b;
an
 Note. Taking in the inequality (22) p = 2, q = 2, we get
the inequality (19) (refer to Exercise 16):
ajb1+ a2b2 + ... -{- anbn
G (a; + az -{- ... + an) (bi + bz + ... + bn).
2.3. The Use of Inequalities for Calculation of Limits -

In the following problems, the limits of quite complicated

sequences are calculated by means of previously proved
inequalities.
Problem 1. Prove the inequality
+
<ln(1-+ -n)<n. (23)

In (1 -{- denotes the logarithm from. (1 + with

n)
base e (see pp. 21-22). n)
Solution. Combining the inequalities (8) and (9), we get
(1-f-n)n<e<(1 } n
Finding the logarithm of these inequalities with base e,
we finally get
nln(1+ n)<lne=1<(n+1)ln(1-{ n
n--1 <ln(1-} n)<n'
Problem 2. Assuming
z1 =1-{ 2 , z2= 2+ 3 -J--

z3=3+4+5+n,
Zy+5-{-1+1+8,
1 1 1 1

find lira Zn.

n-0
Solution. Substituting n-1 for n in the first memi?e of
the inequality (23), we get
<In(11n n
From this inequality and the second member of the inequality
(23) it follows that
In
nn 1 < -1i<ln nn1y (24)

Now, using the inequality (24), we write the inequa-

lities
nn1 nn-1
In < n<In
nn
In n+2 < n+1 <ln 1

1 n+2
n+2<n-+.2<lnn+1
In n+3
................
2 1
In 2n < 2n < In 2n2n
1

Adding them and taking into consideration that the sum

of logarithms is equal to the logarithm of the product,
we get
(n+1) (n+2) (n+3) ... (2n+1)
n + n+1 + .. .
1 1
In n (n+1) (n+2) ... 2n <
1 n (n+1) (n-{-2) ... 2n
" + 2n < In (n-1) n (n+1) ... (2n-1)
that is
In
2nn 1 < n -} n+1 -}- ... -+ <In n2n1
(25)
2n
Since
2nn 1
= 2 -}- n , then

lim In 2nn 1 = lim In (2 + -' In 2.

n-00 n-.oo
n2 -n 2
Exactly in the same way from 1
2 -}
n -1 it fol-
lows,that
n2n
lim In 1
In 2.
n-.oo
Thus, the extreme terms of the inequalities (25) have the
same limits. I. eice, the mean term has also the sam9
44
limit, that is
1im (n -+- n+1
n-.. +...+2 )=1imZn In 2.

Problem 3. Taking x1= 1, x2 = 1- 3 , x3 = 1- 2 + 3 , . .

.., xn= 1
+ 1
1
1 1 1
L(-1) n_ 1
n
calculate lim xn.
n-.oo
Solution. We have
+2n1

x2n=1 1
+ 1
5
6-+-
1 2n

=(1 } 2 } 3 } 4+ 5+ +-...-+ 2n11+2n

6
-22-4 } 6 } ...+2n
=(1+3+3+4- 5+6+...+2n1 1+2n)
-(1+2+6+...+n)= n+1 + n+2 -} ...+2n
In the previous problem, we have supposed that

Zn=n+n+1-+-... +
1 1 1
2n.

1
Therefore, x2n = Zn - . But lira zn =1n 2 (refer to the
n-w
previous problem). Thus,
Ilm x2n = llm (Zn - n) = In 2.
n-.oo n-.oo
It is necessary to note also, that x2n+1 =x2n+ 2n+1 , and,
hence,
lira x2n+1= lira (x2n + 2n+1 ) -1n 2.
n-.oo n-+oo

Thus,
lim x, = In 2.
n-.oo

Note. The numbers xl = al, x2 = al + a2, x3 = a1 +

a2 + c3, . . ., xn = al + a2 +. . . + an are termed
45
partial sums of the series
a1+a2+as+..,an-1--. .

The series is said to be convergent, if the sequence of its

partial sums has a finite limit. In this case the number
S = lim xn is called the sum of the series.
n-1
From Problem 3, it follows that the series
1-2+3-4--5-T+---+2n1 1

converges and its sum equals In 2.

Problem 4. The series

is called harmonic series. Prove that the harmonic series

diverges.
Solution. According to the inequality (23)
nn1 .
n>ln
Assuming n =l, 2, 3, ..., n, write n inequalities
1> ln 2 ,
1
1

2>In2
3>ln34

11 >in nn1 .
Adding them, we get
xn -}-
4
+ 4
+...+ 4 (n+1)=1n(n+1).
2 n 1.2.3 ... n
It follows from this inequality that
lira xn>lim In(n-}-1)=oo;
n-ao n-00
hence, the harmonic series diverges.
46
 Problem 5. Prove that the series
1
} Za 3a ... { na -}- (26)
converges at any a> 1.
Solution. The sequence of partial sums of this series
x1 = 1,
x2 = +
2a
x3=1+2a+3- 1

X4=1+ -{-
V } 4a 1

. .................
1
xn=1-} 2a -} 3a +...+na1
is monotonically increasing, that is
xl <x2 <x3<x4<. . . <xn <. . .
On the other hand, it is known that monotonically in
creasing limited sequence of numbers has a finite limit.
Therefore, if we prove that the sequence of numbers xn is,
limited, then the convergence of the series (26) will be proved.
as well. Suppose
yen=1-
1
2"-
V
1 1
- na +5a - 1 1
na ...
1

2n-1)a
- 1
(2n)a
Since
1 1 1
5a)-...
1

_
- (
(2n-2)0'
1 1

(2n-1)a) (2n)a
1

then (the numbers in each bracket are positive)

Y2n<1.
On the other hand,
1
y2n=1-2a + 1 1
na -}-5asa--.
1 1
-+-(2n-1)a
1 _ 1 _
3a (2n)a
_(1-}-2a + -(2n1)a
3a+4+5a+3a+...-+
(2n)a
47
 i i
-2 (2a+4ai
+ 6aI +...± (2n)a) =1
1

1+ 1
2a 3a 4a
1 1 1
5a + 2a --
1
... 1

+ (2n)0' )
(2n -1)a
` 2
2a }
1

2a+3a
-} ... noc )
1 1

Since x,, =1 + 2a + 2a
+ ... + na , then
2
Y2n=x2n-2a xn.

Now, since x2n > xn, y2n < 1, then

2 2a-2
1>Y2n>xn- 22 xn= 2a xn-

Hence, it follows that

2a
xn 2a-2

that is, the numbers xn are limited when a> 1. Thus,

it is proved that the series (26) converges and its sum is
a
not greater than 2a
2.
For example, if a = 2, then
_ 212+32+...+ 1n2<22-2=2,
xn+ 1 22

n-+oo

In the course of higher mathematics it is proved that

2

5=1+"n+ --n2...+n2+... (27)

Exercises
22. Find the sum of the series
_1 1
S= 1- 1
22 + 32 -
1 1
... -f- (- 1) nz
`!2 --
Indication. Use the equality (27).
,2
Answer. S = 12 .
48
 23. Prove the inequalities
na+1
a T1 < I+2a+3a+ ... +nr4< a+
n+1)
24. Assuming
xn = 1 4- 2a -i- 3a -+- ... + na,
prove that
xn
lim -(X+ 1
= a+1 1 a > 0.
n . oo

25. Prove the inequality

(a ibici -{- a2b2ca -I-... + anbncn)3

<(a1-f- az ... + an) (b1 -E- bz ... bn) (c1 -1 cz + ... + cn),
if ak, bk, Ch are positive numbers.
Indication. Use the inequality (7) and the method
given in (22).
4 1 1
26. Assuming xn = n
-}
n -+- 1
-{
n -{- 2
-}- ... -} hn, whe-
re k is a positive integral number, prove that
lira xn= Ink.
n-. 00

Indication. Use the method of solving Problem 2 of the

present section.

2.4. The Use of Inequalities

for Approximate Calculation of Quantities
At the very beginning of Chapter 1, we have paid attention
to the fact that practical problems require, as a rule, an
approximate calculation of quantities and, as well, an abili-
ty to treat such approximately calculated quantities.
A more accurate estimation of such quantities will certainly
permit to decrease errors in solving problems.
In the present section, we are going to return to an appro-
ximate calculation of numbers of the form

S'n, k =
1

ka + (k+1)a -}- + na , 0 < a < 1, k<n.

1/z 4-0866 49
 In Sec. 1.1 we have succeeded in finding the number Sn, h
with an accuracy of up to 0.4 for k = 1, n = 1,000,000
and a = 2 (refer to Problem 2). In the same section (see
Exercises 2 and 3), for n = 106 and k = 10,000, we were
able to find the number Sn, h already with an accuracy of
up to 0.01. The comparison of these two examples shows,
that the indicated method of their solution yields much
better results of calculation for greater values of k.
In Sec. 1.4 (Problem 3) we found the integral part of the
number Sn, k, for k = 4, n = 106 and a = 1 . Thus, this
number was also calculated with an accuracy of up to 0.5.
However, we could not find the integral part of the number
Sn, i for a = 1 and n = 106 because the method of calcu-
lation of such quantities, indicated in Chapter 1, did not
permit doing it. In this section, we shall improve the method
of calculation of the quantity Sn, 1. This improvement will
make it possible to find similar quantities with a higher
degree of accuracy quite easily.
Lemma 1. If x1 > x2 > x3 > . . . > xn, then
0<A=x1-x2+x3-x4+...+(-1)n_ixn<x1.

Proof. The number of positive terms in the written

algebraic sum is not less than the number of negative terms.
Besides this, the preceding positive terms are greater than
the following negative term. This proves that their algebraic
sum is positive, A > 0. On the other hand, since
A =x1- (x2-x3+xa-.. . + (-1)n 2xn)
and the quantity in brackets is positive too, then A < x1.
Thus, the lemma is proved.
Lemma 2. If 0 < a < 1, then the following inequalities
are true
(2n+1)1_0'-(n+1)1-"
1-a < (n+1)a + (n+2)a + ...
1

(2n)1-ani_a
... + 1
(2n)0'
< 1- a (28)
Proof. The inequality (28) follows from the inequality (14)
(see Sec. 1.4, Problem 2) when substituting n + I for m,
2n for n and -a for a.
50
 Theorem 8. True is the equality
1 1 1

2a 3a na
2a 1 1
} .
2-2' - (n+1)a (n i 2)a
2a 1
1
{
1
3a-...-(2n)a (29)
2-2a 2a

Proof. We have

Sn, t= 1 } 2a
1 1
3a } ... F n } 1
(n+1)a
1

1 1
1
+ (2n)a 1

- L (n.11)a
+ (n+2)a + ... + (2n)7 I.

Adding and subtracting from the right-hand member of

the equality the number
2 [ 2a -} a2 -} 2a }- ... + 1
"] ,
( 2n )

1 1 1 _1
Sn, =1-r2a-{ 3a-4a+...-(2n)"+

+2
L -2a
-{- 2a }
6
a } ... }
(2n)6'
]-

The numbers of the first square brackets have a common

factor 2a . Taking it out of the brackets, we get

Sn. 2a } 2a (2n)a +
+ 2a (1 { 2a } 3a . } 2a)

- [ (n+1)a + (n+2)a + ... {

(2n)a ] .
4* 51
Since in round brackets there is the number Sn,1, then
(n+1)a -}- (n+2)a + ... + (2n)a -
L1-2+3-
-. 1 1
. .
1

(2n)aJ
11
-
2a
(2a - 1) S, 1 = 2 2 Sn, 1
Hence, after multiplying by 2a and dividing by 2 - 2a,
we get the equality (29).
The equality (29) is of interest because it brings the calcu-
lation of the quantity Sn,1 to the computation of the quanti-
t Y S 2n,
,n, n+1 the quantity 1 1
f 3a 1- 1

2a
-' - (2n)a
The first of these quantities for great it is calculated with
a high degree of accuracy by means of the inequality (28).
Concerning the second quantity, we know from Lemma 1,
that it is less than zero and greater than - 2a . But
2--2(x
if we find the sum of the first four summands of the latter
quantity, then the remaining quantity (the error) will be
less than zero and greater than - 1 2a
5a 2-2a
In the following problems we shall perform the calcula-
tion of this quantity with a higher degree of accuracy as well.
Problem 1. Find the sum
A=1 + V2
1 -}- -{- ... -{-
v3 v106
accurate to 0.002.
Solution. By virtue of Theorem 8
A
-/2 1 1

2-_V2 v106 -{-1 + x/106 -j- 2 + .

-(1-
. _/2.106

-/2
2_v2 1

v2
-...-1
1

1/3v2.106
I 1
l
__(V
+ \
(
106+1
+ _V2.106

+1) 11- -VI 2-j- _V3-...- 1/2.106

= (V Y+ 1) (B-C),
52
where
B= ]-
1

V106 + +
-V1G6+ 1

C-1 I-
2
- 106

i/3 - ...
1 1 1
+ v2.106
v2
According to Lemma 2, the number B satisfies the inequa-
lities
2(1/2.101;+1-y106+1)<B< 2(1/2.106-y106).
The extreme numbers of the inequalities differ from each
other by less than 3.10-4. Indeed,
2 (1/106±1-1!106)-2 (1/2.106-{-1-1/2.100)
_ 12 2
106-{-1+ -/106 -/2.106-{ 1-{- -VT-106; 1/106
1 - 1/2 -1 1
1,000 ! 3.10 4.
1 66 1/2
Thus, the middle number will differ from the number B
by less than 2.10-4. Calculating the first number and
subtracting from it 2.10-4, we get
B = 828.4269 ± A t,
I Ar I< 2.10_4.

Now, proceed to calculating the number C. Let ni be an

odd number. Estimate the quantity

na -m }
1 1
D 1 { 1
1 -V W, 72 2n

For this reason, it is necessary to notice, that

Vk±I-Vk-1= 1/k-71 +1/k-1
2

and
2 2
E
1/md-1+Vili
2
+1/m-}-3-- 1/m+1 1/m -H- 4 -}- 1/m d- 2

53
 2

V2n71-V2n-1
±V M+V;W-+3-Y-iW -+I If ;W-+4+
+-Vm+2+ ... -V'2n+1 =Vm-
-Vm-1+V2n-I/2n+1.
Thus, the number E is quite easily calculated. Subtracting
the quantity D from the quantity E, we get
E-D=( Vm+1-2 Vm-1 - Vm 1 -
2 1
+ .. .
(Vm+2+Vm - Vm-}-1
2
- ( V2n+1-,1/2n--1 - V2
1

Demonstrate, that all the numbers in the brackets are posi-

tive and monotonically decreasing. Indeed,
2 2Vm-(Vm 71 +Vm-1)
Vm f 1+Vm-1 Vm Vm (Vm-f-1+V m-1) =
2m-2Vm2-1
Vm(Vm+1+Vm-1)(2Vm+V;a I1+Vm-1)
2

Vm (V'n t +V'n-1) (2Vm +V 771+1/m-1) X

X (m+Vm2__ 1)
Hence, it is proved, that such numbers are positive and
monotonically decreasing with the increase of m. According
to Lemma I
0<E-D<
2
Vm(V 7-}-1+V m-1) (2Vm + V m-}-1+V7 --1) X
X (m +Vnt2-1)
We shall not make a great mistake, substituting m for the
numbers m + I and m - 1 in the denominator. Here, we get
2
0<E-D< 1

1/-m

54
Taking m = 9 we get
O<E-D<8 8 s<0.0006.
This proves, that when m=9 and n= 106
E-D=0.0003±A2, A2 1<0.0003,
D= E-0.0003 ± 02= j/9 -V 8+ V2.106-
-)/ 2.106 -1-0.0003±A2-0.1710±A2i
Now let us return to the quantity C. We have
C = 1 - -2
-1- 1
-V3

-{- V,3
1
-V4
1

i/4
'

{
V5-
1

1/5
1/6
1

i/6
{-

{-
i i/ {- D =

+ V7- +0.1710±A2=

= 1 - -- 1- -f- .y7 +
- 2 11 + 2) +
- V=3 /5 v6
+0. 1710 2
4

2
3y2
4
{
v3
3
-{- v5- -I y7
5 6 7
'I_

+ 0.1710 t A2.
Thus, for the calculation of the number C with an accuracy
of up to 3.10-4 it will be required to find only 5 roots and
to produce a number of arithmetic operations. Using the
tables and carrying out necessary calculations, we find
C = 0.6035 ± A 2.
Taking into consideration the found quantities B and C,
and returning to the quantity A, we get
A = (12 + 1) (B - C) _ (12 + 1) (827.8226±A3) _
+ 1) .827.8226 ± 2.5A3,
where
I 2.5A3 I < 2.5 (1 Al I I A2 I)<2.5.5.10-4<2.10-3.
-I-

Thus, the calculation with an accuracy of up to 2.10-3

will be
A = (V2 + 1) 827.8226 = 1998.539.
55
 Problem 2. Calculate the number
A=1-r- 1
2 -}
1

y s+
... + 1
1012
4
with an accuracy of up to unity.
Solution. By virtue of Theorem 8
A_ 12 1 1 1

2-y 2
(
\ y11012+1 + 4012+2 +. 1 2.1012
V2 /1-
The first term can be easily found and with a high degree
y2
1 + 1/1

3
_
yj2-1
2 .1012 1.

of accuracy by means of the inequalities (28). By virtue of

these inequalities the first term can he substituted by the
number

23
3 3

y2
(2.1012)4-(1012)4

4 4
109 (V8-1) 109.
2-y 2 1_? 8

By virtue of Lemma 1 the sum

y2 1

2-/2 (1-_V_2 VY
1 i l ---...-21012
is positive and is not greater than the first term. Since the
term is less than two, then
3.109-2<A< 3.109.
The extreme numbers differ from each other by 2, and from
the number A by less than 2. The middle number 3.109
differs from A by less than unity. Substituting this number,
we get
A = 1333333332.3 ± A, I A I < 1.
Notice that the accuracy of calculating the number A, con-
taining a trillion of addends, is extremely high. The relative
error is less than
100 : 1333333332.3 < 0.0000001 % .
56
 Exercises
27. Calculate (with an accuracy of up to unity) the sum
1 { v22--3 F ... 106.

Answer. 14,999.
28. Show that the equality
nl-a
{
2a
1 1
3a ... 1

na
= 1-a C

is true, where P,, is an infinitely small quantity, urn Nn = 0,

n-.oo
and

C
_ 2a
T- 2 - I
1
1 1

3M4(Z
1
... (-1) n { .,..
n-1 1 l
 SOLUTIONS TO EXERCISES

1. Setting in the inequalities (1) (p. 9) n = m,

m+1,...n:
21/m+1-21%m< 1 <22Vm-1,
/m
21/m+2-21/m+1 < 1/m+1
<2j/m+121/m,
2j/j/m+2< +2
<2j/m j-2-2j/m } 1,

21/ n+1-2Vn<
V-n
<2j/n-2j/n-1.
Adding these inequalities we get
2j/ n+1-2j/m< + +
m 1/md 1
+
Vm+2
+ ... +
ti n
_ <21/n-2
2. Taking in the inequalities of Exercise I m = 10,000,
n = 1,000,000, we obtain
211,000,001 -2 1/10,000<
10,000 +
1

10,001 + + 1/1,000,000 <

< 2 j/ 1,000,000- 21/9,999.
Since
2 V 1,000,001 > 2 j/ 1,000,000 = 2,000, 2Y10,000=200,
21/9,999 = 1/39,996 > 199.98
(the last inequality can be easily checked, extracting the
square root with an accuracy of up to 0.01), then
2,000-200=1,800<
1/10,000 +
-L-
+ 1/10,001 + 1/1,000,000 <
< 2,000 -199.98 = 1800.02.
58
 3. Multiplying the inequalities of Exercise 2 by 50, we
shall get in our designation
90,000 < 50z < 90,001;
hence
[50z] = 90,000.
4. For n = 1, it is obivous, that the inequality is true

3.1+1
2 1 1 _ 2. 1

Assuming now that the inequality is true for n k

2.4.6
3 5
1
... 2k+1
2k
< 1
V3k+1 (a)

prove that it is true for n k + 1, that is, prove that

1 3 5 2k-1 2k+1 1

2.4.6 ... 2k - 2k+2 G i/3k+4 (h)

Multiplying the inequality (a) by 2k+2 , we get

13 5 2k -1 2k+2 1 2k +1
2 ' 4 ' 6 ... 2k 2k+2< 3k+1 2k+2'
What is left is to prove the inequality
I 2k+1 1

-j/3k+1 2k-f-2 < i/3k+4

Multiplying it by (2k + 2) V 3k -+1 1/3k + 4 and squaring
both parts of the obtained inequality, we get
(2k + 1)2 (3k + 4) < (2k + 2)2 (2k + 1),
or
12k3 + 28k2 -{- 19k + 4 < 12k3 + 28k2 + 20k + 4.
The latter inequality is obvious, since k > 1.
This proves that the inequality
1 3 2n-1 ` 1
-T* T "' 2n ' -j/3n+1
is true for all n.
5. Assuming in the inequality of Exercise 4 that n = 50,
we get
1 3 99 < = < 1 1 1 1

T4 ''' 100 x/3.50 +1 151 17144 = 12

5* 59
 6. Assuming y = 6 - x, x = 6 - y, we shall bring the
problem to finding the greatest value of the function
(6 - y) y2 = 6y2 - y3
when 0 < y < 6. Assuming then, that y2 = z, we shall get
the function
3

6z - Z2,
whose greatest value (refer to note on p. 34) is equal to
3

0.5.43=32
H-1 32
7-
(-)
and is obtained in the point
1

3_1
Z= 32 =42.
C -2- /

The function 6y2 - y3 takes the greatest value in the

point y = I/ z = 4, and this value equals 32.
The function x (6 - x)2 attains the greatest value of 32
in the point x=6-y =6-4 =2.
7. The volume of a box (see Fig. 4, p. 38) equals
V -x(2a-2x)2 =4x(a-x)2, 0<x<a.
Assuming y = a - x, y2 = z, we get
3

V=4 (az-z2).
3

The greatest value of the function az - z2 is obtained

in the point
1

z (23 1 2 1 =\3 )2

Therefore,
y= /z= 3 , x=a-y=a- 32a = a3 .
2a

60
 Thus, the volume of a box will be the greatest, if the
length of the side of the cut-out square is that of the
side of the given square.
8. The least value of the function x6 + 8x2 -F 5 equals 5
and is obtained when x = 0.
9. Assuming y = x2, we shall bring the problem to finding
the least value of the function
y3 - 8y + 5
for positive values of y.
In Theorem 5, we have proved that the least value of the
function y3 - 8y is equal to
3 3
1= 8z 321/6
(1-3) (83 )3 2 3- 9

32
The least value of the function y3-8y+5 is equal to
32
-
9 + 5 = - 3.6 ... .

10. Assuming y= xa we get the function

y-aya (Q y-ya), a >0, a >1.
By virtue of Theorem 5, the greatest value of the function
a y-ya is
1 1

1 a
(a-1)
1 a a

1aa (a
Multiplying the last quantity by a, we shall find the greatest
1 1
value of the function a a y - ya which is, hence, equal to

(1-a) aaa )all a

a + a 1 1

a
a-1
= ( 1-a ) ( aa)
61
 It. The function i/x - 2x, x> 0, a = 4 , a = 2, bas the
greatest value, equal to
4

4-1
3 1 3
4/ 1

`4f
I

,0 the following inequality is true

Therefore, for all x>
x - 2x < 8 , or Vx < 8 -I- 2x.
12. Write down the inequality (8) in the form of
(nn1)n<e,
If n > 3 > e, then
<nnn=nn+i.
(n+1)n<enn<3nn
Raising both members of the latter inequality to the power of
1
n (n+ 1)' we get
n+Vn -{-1 < rn.

13. Since 1 < j/ 2 = 38 < V9 = then V3 is the

greatest of the numbers 1, Y2, V /-l On the other hand,
in the previous problem we have shown that the sequence
of the numbers 1Y3, VT' ... , Vn, ... decreases. Hence,
V3 is the greatest of the numbers 1, 1/2, V3, ...
. . . , Vn
14. Suppose r n = 1 + (Xn, an > 0. Raising to a power
of n we get
n=(1+a)- n
r +a
n)n 12
.(12
Assuming that n 2, 2 taking Theorem 3 as the
basis, we get
n
n>(1-{ 2 an)=1+nan+42an.
2

62
Hence, it follows that
n> n2 an, an<
n
4
, an< V2

ic
, /n=1++
an<1--?_
vn .

Note. Using Newton's binomial, it is easy to check that

rn<+ n
Indeed,

(4+J/-n n
)n=1+nf//,?+n(n-1) 2-}-...>1+
2 n
+ n (n -1) 2
2 n = n.
Hence, it follows that

inn<1+Y'rnn
15. When n = 1 and a1 > -1, the inequality is obvious
1+a1>1+ a1.
Let us assume, that the inequality is true for n = k,
that is
(1 + a1) (1 + a2) . . . (1 + ak)
>1+a1+a2+...+ak.
Multiplying both members of the inequality by (1 + ak+1),
we get
(1 + a1) (1 + a2) ... (1 + ak) (1 + ak+1)
>(1+a1+a2+.. +ak)(1+ak+1)
= 1 +a1+... +ak+ak+1 +alak+1+
+ a2ah+l + . . .+ akak+-1
Since the numbers a1i a2, ..., ak, ak+l are of the same
sign, then
alak+l + a2ah+l + . + aka k+ l > 0
.

and, therefore,
(1 + a1) (1 + a2) ... (1 + a,,) (1 + ak+1)
>1 + a1+ a2 + ... + ak +ak+1,
that is, the inequality is proved also for n = k + 1.
63
 This finally proves the inequality to be true
(1 + ai) (1 + a2) ... (1 + an)
+a1+a2+... ma,,
for all n.
16. If the polynomial (a1x - b1)2 + (a2x - b2)2 + .. .
. + (an,x - bn)2 has a true root x - x1, that is
. + (anx1 - bn)2 = 0,
(a1x1 - b1)2 + (a2x1 - b2)2 + . .

then every number a1x1 - b1, a2x1 - b2, ..., anx1 bn is

equal to zero, that is,
0 = a .1x1 - b1 = a2x1 - b2 = ... = a,,xl - bn,
b1 b2 b,,
x1 =
a1
= a,,
a2
Thus we proved that the polynomial
(alx -
b1)2 + (a2x - b2)2 + ... + (anx - bn)`
x2 (a2 + a2 + . . + an) - .

2x (alb1 + a2b2 + ... + anbn) +

+ (b2 + b2 T . . + bn) .

cannot have two different true roots and, therefore,

(albs + a2b2 + ... + anbn)2 -
- (a, + + a 'n) (b i + . . . -4-- b n< 0.
From this follows the inequality (19)
(albs + a2b2 + + anbn)2
(a1 + a2 ++ an) (bi + b2 + bn)
Notice, that the sign of equality holds only when the
polynomial under consideration has a true root, i.e. when
at
b 1 = b2
a2
... an
bn
17. Using the inequality (19), we get
C1
a1+a2-}-... +an 12 =
n /

an
(
a1

Vn Vn
1
+ 1
I2
2
Vn Vn
< n -{- n + .. .
2
(n+n+...+n ) =
)
n
ai - a2 + ... -- an = C2
n 2

64
Hence, it follows that cl < c2 (the arithmetic mean does
not exceed the root-mean-square).
18. From the inequality
(y-n+ 1 + 2=n + 1 + 21/;2 -1+n-1=
=2n+2Yn2- 1<2n+2Vn2=4n
it follows that
Vn +1+Vtn-1 <2Vn,
1 1 _
2Vn < 1/;i-{ 1-{ Vn-1
Vn+1- n-1 Vn-I-1-Vn-i
(V-n+ 1 + n-9) (V;-Fl- n-9) 2

Multiplying by 2, we get

Vn
1
<Vn+i-vn-1.
19. Setting in the inequality of Exercise 18n = 2, 3, ... , V n
<v3-1,

Vn
- <j n+1-Vn-1.
Combining the written inequalities, we get
V2+V3+...+
V-n
<Vn+1+Vn-V2-1.
1

Adding 1 to both parts of the inequality, we finally get

1
V2
-}
V3
- V4 {
1

V5
-- ... 1

Vn
<
-
<1/n+1+Vn_

05
 Note. It was proved in Sec. 2.1 that
1-F-
jr2 + V3-
-E-...+ >2Yn+1-2V2+1.
V--
The numbers + V n - V2 and 2 Vn + 1 -
- 2 V2 + 1 differ from each other less than by 0.42. Each
of these numbers could be taken for an approximate value
of the sum
+
1/2
+ r3 + ... + Vn-
=zn.
Let us notice without proving, that the number Yn + 1 +
+ V -n - Y Y differs less from the number zn, than the
number 2Vn+1-2V2+1.
20. The function x4+5 takes a negative value when
x < 0. Therefore, the greatest value of the function is
obtained for positive values of x.
Since
x3 1
+5 1 x+x3)
5 (5 =

then the greatest value of the function is reached in the

same point in which the function x + x-3 takes the least
value. It follows from Problem 5 4 Sec. 2.1 that the least
value of this function is equal to
-3
3-1 3

( l
(1-4-3)
3 =4(45)4'
The greatest value of the function -+5 is equal to
1 _ 154 15 _ 3
3
TO- = 20V 15 4V15
1 a
( 151
To find the greatest value of the function x6- 0.6x10, we
get y = x6. It is clear that y> 0. The function
10 10 10
y-0.6y6 =0.6 (s y-y6 )
66
takes the greatest value (see the note on p. 34) equal to
10
6
10
10 6

0.6 (6 - 1) o =0.4.

A.snming in this exercise that y= 2

we get
- a - 1

l' :z 2 = y 4 + ay.
1

The least value of the iur tion y 4+ay, as it follows

from Problem 4 Sec. 2.1. k l- equal to
,.

(14-
4) (4a)5 ('ia;'
1

Assuming 4(4a)5=2.5, we get

1

(4a)5 = 2, 4a = 32, a = 8.

=(1 2z--32-+-4252-} fiz

}

-2(22-x[42+62+ ...)=
1 1 1

0+ 22 -} 12-x-42 + 52+ 1 + ...

22
+ -22 +- 32 -x- ... ) _
1 1 R2 a2
=Z 1
(1+Zz+32+...)=2.
6 -12
(we have used the equality (27)).
23. Since a > 0, then a + 1 > 1 and, hence,
1
(1+ n )1+a>1+1+a
n

67
Multiplying these inequalities by nl+a, we get
(n+1)'+a> nl+a+(1 +a)na,
(n-1)1+a>n1+a-(l +a)na

From these inequalities it follows that

nl+a-(n-1)1+a (n+1)1+a-n1+a

1+a < na< 1+a

Write these inequalities for the values n = 1, 2, 3, ... , n:
1 21+a-1
1+a <1< 1-f-a
31+a-21+a
21+a-1
1+a < 2(< 1+a
n1+a-(n- 1)1+a < (n+ 1)1+a-nl+a
1+a <na 1+a
Adding them, we get
+na<(n+1)1+a-1
nl+a
T a < 1-{ - 2a - }- 3a -+-
-
1-}-a
(n+1)l+a
1+a
24. It follows from the inequalities of Exercise 23 that
1 l+a
-'i
1+2a 3a } ...+na n
<
1

1--a < n1+a +

The left-hand member of the latter inequalities is a constant
number 1a , and the right-hand member tends to a limit
equal to 1+a when n tends to infinity. Hence, the mean
,

member of the inequalities tends to the same limit as well,

that is
lim 1+2a+ 3a+...+na
_ 1

n-.oo nl+a 17a

25. Let us introduce the designations

B3=b;+b2± ... --bn,

C3=C3+C3+... +Cry
68
 ai an
x1= A , x2= A
a2
, e xn= A
bn
y1= B , y2= B , yn= B
Z1= C
Cl , Z2= C , .. . C2
, Zn= CnC
On the basis of the inequalities (7) we have
xi + A + Z11-
ajb1c1= ABCxiylzl <ABC 3

a2b2c2 = ABCx2y2z2 < ABC x2+Y + z2

3

+yln zn
anbncn = ABCxnynzn <ABC X34
3
Adding the written inequalities we get
aibjcj .+ a2b2c2 '_+ _
. .. + anbncn <
<ABC (xi+x2+3 +xn + Y i b4+...+y n zi+z2+ ... _ I z ll
3 3 ).
Taking into consideration the introduced designations, it is
easy to calculate that
3 3 3 al+a2--...+4 A3

y3+y2+ ... +yn = 1 , zi+Z3+ ... +Zn = 1.

Hence,
aibicj +a2b2c2 } ... }anbncn < ABC ; 3 } 3 {- 3) = ABC.
Raising both members of the inequality to a cube, we finally
obtain
(albicl -F- a2b2c2 +... + anbnCn)3 A3B3C3 =
_ (al + a2 + ... + an) (bi + b2 + ... + bn) (ci + cz + ... + Cn).
26. Write down the inequalities (24) for different values of n
In nn1 < 3 <ln n n 1

In n+1 < n+1 <_ In n n 1

In
kknl <kn<In kn

69
Adding these inequalities, we get
(n-{-1) (n+2) ... (kn+ 1) 1 1 1
In <
n(n+1)...kn n + n-}-1 + +kn<
n
<1n I n-1 n+1 kn 1

n kn-1
that is
In knn 1
1n (k -{- n <
CIn kn
< n+ 1

n -F-1
1

+n-{-2+ +kn
1

n-1
k
In (k+ n 1

If n tends to infinity, then In (k -E n) tends to In k and

k
In (Jr j
n
1 ) tends to the same limit as well. Therefore,
lim (n + n+1 + ... +kn)=1nk.
n- o0

++10
27. By virtue of Theorem 6
1 1

_ V2
2-V2
(
I
y1106 -f- 1
+ V106+ 2
1 + ... + 1
2.108

2-V2
(1- 1
+T3 1 1

3p/-2 V2.106 /

The second addend is negative but greater than- V3

2-V2
- 1.9. The first addend, according to the inequalities (28),
satisfies the inequalities
2 (2.106+1-V106-1) V
2-V2 <
jl2
+ ... + V2 108 <
1 1 1

C 2-V2 V108-{-1 + 108-{-2

2
(2.106-Y106) = 15,000.
70
Since the extreme terms of the latter inequalities differ
from each other very slightly (less than 0.1), then
15,000-2<1+--+ ... + 10e <15,000.

106

The mean number 14,999 differs from ' s1k less than by 1.
k=1
28. By virtue of Theorem 6

2-2 a
[(n+1)a +(n+2)a + .. .
+ (2n)a J
1
-
- 2 -2
2a
a [1- 2a -} 3a
1
- ... - (21n)

a ]=An-Bn,
where

2-2" (n+1)a (n+2)a (2n)]'

Bn - 2- 2a [ 1-
2a 1

2a
+ 3a
1
- ... - (2n)a
1
] .

The number Bn is a partial sum of the series

2a (_1)k_1
2-2a ka
k- 1
This series is sign-alternating with monotonically decrea-
sing (by absolute value) terms. Its remainder (by absolute
value) is not greater than the absolute value of the first
term of the remainder, that is, the number .
2a q
2 _ 2a na

Since this number tends to zero when n-} cc, then the
series converges and
a
lim Bn =
n.oo k_1 2-2a (-1)k ak = C,
2

71
that is yn = Bn - C is an infinitesimally small value. Now,
using the inequalities (28), we get
2a
2-2a [(2n-1)
1-a
-(n+1) 1-a ]<A,<
n1 a
< 2a
2_2a [(2n)1-a- n1-a] = 1-a
Since the difference between the extreme terms of the ine-
qualities tends to zero when n - - oo, then 6n = An - n1 -a
1--a
is an infinitesimally small value.
Thus,
1 -I 2a + ... +4-=A-B=
n1-a -Sn-G n1-a C
1-a 1-a 1

where Nn = Sn + y is an infinitesimally small value.

TO THE READER

Mir Publishers welcome your comments on the con-

tent, translation and design of this book.
We would also be pleased to receive any proposals
you care to make about our future publications.
Our address is:
USSR, 129820, Moscow 1-110, GSP Pervy Rizhsky Pereulok, 2
Mir Publishers

Printed in the Union of Soviet Socialist Republics

The present booklet contains some particularly interesting
inequalities playing an important role in various sections
of higher mathematics. These inequalities are used for finding
the greatest and the least values as well as for calculating
the limits. The booklet contains 63 problems and most of them
are provided with detailed solutions.
The book is intended for students of senior classes of secondary
schools.

llir Publishers
Ni -oscow