On the Exponential Inequalities and the Exponential Function

Author(s): V. Ramaswami Aiyar

Source: The Mathematical Gazette, Vol. 4, No. 61 (Jan., 1907), pp. 8-12
Published by: Mathematical Association
Stable URL: http://www.jstor.org/stable/3604264
Accessed: 22-01-2016 04:09 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Mathematical Association is collaborating with JSTOR to digitize, preserve and extend access to The Mathematical Gazette.

http://www.jstor.org

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions
8 THE MATHEMATICAL GAZETTE.

ON THE EXPONENT'IAL INEQUALITIES AND THE

EXPONENTIAL FUNCTION.
? 1. THEOREM. If a be any positive quantity not equal to 1, and x, y, z be
any three rational quantities in descending order of magnitude, then
ax(y - z) + a(z- x) + a(x -) > 0.
Proof. Let p, q, r be any three integers in descending order of magnitude.
Since (a- 1)2 >0, we have a2 - a > a - . Hence in general
aM _ lm- >1 m- a -2;
from which we have,
-
(aP _- ap-l) > (Cp-l - ap-2) > . > (aq+l aq)
> (aq - aq-l) > ... > (ar+l - ar)
The arithmetic mean of the terms in the first row should be greater than
the arithmetic mean of the terms in the second row, since each of the terms
in the first row is greater than each of the terms in the second row.
Therefore
aP -aq aq ar
p-Cq q -r

or - r)+aq(r -p)+ ar(p-q)>0 ........... ...........(A)

a(q
Let n be any positive integer. In the above, for a, we can write an.
Doing so, and dividing throughout by n, we get
- +a -
a(
\n n/)
+a 7in aP)
n \n n/
>0 ..................... (B)

(A) proves the theorem when x, ?/, z are integers, and (B) in any other case.
? 2. In the theorem of the above section, for x > y > z, write
first, > 1 > 0;
secondly, 1 > m > 0;
and thirdly, 1> 0 > >m.
And the result is,
in the first case, a" - 1 > m(a - 1);
in the second case, aM"-1 < (a- 1);
in the third case, a _- 1 > m(a- 1).
Hence, if a is positive and not equal to 1, a - 1 (a - 1), according as
m(m - 1) 0.
? 3. In the theorem of the first article, for x > y> z, write
first, x>y>O;
secondly, x > 0 > y;
thirdly, 0 > x > y.
ax - 1 asv- 1
And in each case the result reduces to - > . Consequently,
x Y
a'- 1 ay- 1
if a is positive and = 1, and x > y, a- x > -a .
y
? 4. Let x and y be two positive quantities, x=l=y. In the result of
? 2, for a write, first, x/y, and then y/x, and reducing, we get the result,
mxm '(x - - m -
) m my"-l'(x y)
according as m(m - 1) ?0.

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions
 EXPONENTIAL INEQUALITIES AND FUNCTION. 9

CoR.1. If >y>0,
xm ~ m-1
9nxm--1 =
x-y
according as m(m- 1) ' 0.
COR.2. If a is any positive quantity not equal to 1,
ma-l(a - 1) am - 1,

according as m(m - 1) 0.

? 5. Let a and b be positive, a =]=b; and let p and q be any two positive
rational numbers. In the inequality of ? 1, for a write ; and for x>y>z
write 1> P >0. We get
p+q
p

a p q a\P+q
or b +* + >

Hence
Hence pa+qb , ,Wpl
p+q (>(apbq1p+q

pa+ qb+... +tk (a t)P++ +t,

p+q+...t
if p, q,... t be positive and rational, and a, b, ... k positive but not all equal.

? 6. THEOREM. ax is a compact function of the rational variable x.

Proof. Let p, q, x, y, p' q' be any six rational numbers in descending
order of magnitude. Now the inequality in the first section can be written
ax - ay a _- a'
- > if x>y>z.
x-y y-Z
-
aP aq aq - aa a - ay ay - aP' aP' - aq'
Hence, > -> -> >
p-q q-x x-y y-p' p'-q'
Suppressing the 2nd and 4th terms in the above relation and multiplying
by x-y, we get
ap - aq ap - aq'
P-q (x-y) > aX-a (x-y).> ,
p-q a' -
In this inequality the extreme terms can be made as small as we please by
making x-y small enough. Hence the middle term ax-ay can also be
made as small as we please by making x-y sufficiently small; which proves
the theorem.
A simpler inequality suited to the same purpose can now be stated; that
is, if p+ 1, p, x, y, q + 1, q be in descending order of magnitude,
(aP+1 _ aP)(x -y) > ax - ay > (aq+l - aq)(x -y),
a being any positive quantity =1=1.
? 7. From the above inequality we could define ax for irrational values of
x and establish the Theory of Irrational Indices; that is, extend the Theory
of Indices to irrational values, or to real values in general, of the exponent.

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions
10 THE MATHEMATICAL GAZETTE.

Thus completely defined, az becomes a continuous function of x, which, if

a>l, increases from 0 to + oo as x increases from - o to + o, and if a< 1,
decreases from + co to 0 as x increases from - oo to + c .*
? 8. It follows from the above that every positive quantity x has a definite
logarithm with respect to any base a, if a is positive and not equal to 1. It
follows further that if a > , log ax increases from - o to + o as x increases
from 0 to oo; and if a < 1, loga x decreases from + oo to - oo as x increases
from 0 to o.
ax - 1
? 9. In this section we shall consider the graph of the function -- as x
decreases from a positive to a negative value. x
We have already observed that ax is a continuous function of x. Hence,
ax-1 ?
by the Theory of Continuity, is a continuous function of x. From
ax - 1
the inequality of ? 3, a decreases constantly as x decreases. Hence
x
we see that the graph is a Continuous Curve, with a sudden decrease of
ax - 1
value, possibly, as x passes through the value 0 at which becomes
xx
a -1 a- -
indeterminate. The identity - a . , however, shows that there
x -X
is no sudden change of value as x passes through the value 0, since
aax- 1
L a= 1. Hence it follows that has a definite finite limit when x = 0.
2=0 x
We shall represent this limit by A and shall determine its value presently.
We notice that A should lie between the values of a- , when x= +1,
x
and when x= -1 ; that is, between a- 1 and 1 - .We further notice that
ax~1 a 1
if x is numerically less than 1, -- lies between a- 1 and 1 --.
X a

? 10. We now proceed to give a demonstration of the Exponential

Theorem. We have
a ={1 + (an- 1)}", n being any positive integer,

= 1+X(an- )}

n an - 1
( x
= + * An) where A=a~-1
n
=-1+ x-(+1
-)+ + ...+x~ .....................(1)
where X7 n(n- 1)... (n-r+l) (x'

It may be observed that the extreme values of ax follow as follows: If x> 1, we

have ax>l+x(a-1). Consequently, if a>1, L aX=cc. Hence L ax=O, if a>1;
L ax=0, if a<1; and L a =oo, if a<1.
.2-=Z0 X=--o

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions
 EXPONENTIAL INEQUALITIES AND FUNCTION. 11

We shall suppose n numerically greater than x to start with; so that

x
x is less than 1. It follows that X, or a- 1 lies between a- 1
sn numerically x
n
and 1-, however much n be hereafter increased. Choose any positive

quantity K so as to be numerically greater than both a- 1 and 1 - a. Using

modular notation for numerical values, it follows that IX.I or a

x
n
is always less than K. Hence from (2),
.Kr where<'
I\X 1<x'".K" where x'=Ixt([ ................................
(3)
r

From this we see that each term in (1) after the first is numerically less
2
than the corresponding term in the series 1 + -iKX+ .....................(4)
This is a convergent series of positive terms.
We shall denote by Rp the remainder after p terms of this series. Going
back to (1), we shall denote by R'p, the remainder after the first p terms
therein, that is R'p,nXp+ Xp++ ... + X. It follows that R'p,n 1Rp,
however great n may become, since the numerical value of each of the terms
constituting R'p,, is less than that of the correspondingterm in Rp, in which
the terms are all positive. We thus have
a1= l + X,X2+ +Xp-l+R'p,n .........................(5)
where R'p,n is numerically not greater than Rp, however great n may be.
Now making n infinite, we see from (2) that
.xx ax-- I
L Xr=-. A where X=L .
n=xan |r- r=0 X
We thus obtain
xX xX2,2 XP-lXp-1
ax=l+ + ~+' .+ p
-1
where IR'p is less than Rp,the remainder after p terms in the series
XIK / X!2K2
-2. ......... (6)
+ +q- + .....................................(6)
Here p may be any integer whatever. If we increase p, Rp steadily
decreases and becomes infinitely small as p becomes infinitely great. There-
fore also R'pbecomes infinitely small as p becomes infinitely great. That is,
-we obtain the expansion
xX 2X2 az- 1
a==1 + 2A2+...(to infinity) where X=L X
. ........(7)
1_1 \ z=O

Writing in particular A for x, we get

1 1 1
aA=1=+- +- +...=e; .'. a=e\ or X=log,a. ................(8)
al++_+..

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions
12 THE MATHEMATICAL GAZETTE.

1-
We thus find that L
X
=logea, and get the expansion
x=0

1log a + 2(lo ,)2(9)

aX 1+ +.. . . -.........................(9),
In particular,

e x=1+ + ..................................(10)
.

ao - 1
? 11. Since L l
=ogea, it follows from the discussiont in ? 9, that ifx
x=O X
be any positive quantity
ax- 1 a--_ 1
-> log? a >-
,x -x

COR. 1. a- > log > 1

ex_- I e-"- 1
COR. 2. -- > 1> ;
X -X

or if x is any positive quantity

ex> 1+x; and e- > -x.
COR. 4. If x > y,
ax -a"
a log a > > ay log a.
x-y
For, x -y is positive, and therefore,
ax-y - 1 a-(x-Y)- 1
> log a >
x-y -(x-y)
which reduces to the above.
COR. 5. If > y,
ex - eY
ex > > ey.
x-y
COR. 6. If X>y>0,
1 < logx-logy <
x x-?/y
For, since x > y > 0, and e > 1, log x > log y.
Substituting log x, logy for x, y in Cor. 5, we get
x> x-y
X-C
>Y
log x- logy
whence, by taking reciprocals, we get the Logarithmic Inequality above.
It will be observed that Corollaries 4, 5, and 6 give the differential coeffi-
cients of ax, ex and log x even as the Power Inequality (Cor. 1, ? 4) gives
the differential coefficient of xm.
I hope these notes will be found useful by readers, and should this be the
case, and I be allowed some further space in the Gazette, I hope to give some
additional notes.
Gooty, India, V. RAMASWAMI AIYAR.
25th June, 1906.

This content downloaded from 178.250.250.21 on Fri, 22 Jan 2016 04:09:51 UTC
All use subject to JSTOR Terms and Conditions