The Function log x

Let x > 0 . Then the function f ( t ) = 1 t is defined and continuous on the interval between t = 1 and t = x . We define a function for x > 0 , which we shall call log x , by

log x = 1 x 1 t d t

Obviously this function has one property which we have attributed to that rather indefinite function which we have previously written as log x : namely, (\log x)' = \frac{1}{x}. The next basic property we prove is

log ( a b ) = log a + log b

But for any fixed a , consider log ( a x ) as a function of x . Then

(\log (ax))' = \frac{1}{ax} \cdot a = \frac{1}{x} = (\log x)'

Therefore log ( a x ) = log x + c , c some constant. For x = 1 , we have

log a = log 1 + c

Since

log 1 = 0 ( = 1 1 1 t d t ) ,

c = log a , and log ( a x ) = log x + log a . This is what we wished to prove.

Now let a > 1 ; then log a a 1 , for this is the area of the larger rectangle in the diagram below.

Illustration for The Function log x

The smaller rectangle has area ( a 1 ) 1 a log a . Thus we have for a > 1 ,

a 1 log a a 1 a = 1 1 a .

This is an equality when a = 1 , so is true whenever a 1 . Now if a < 1 , then 1 a > 1 , and we have

1 a 1 log 1 a 1 1 1 / a

by the above. Now

log a + log 1 a = log 1 = 0 ,

so

log 1 a = log a .

This gives

1 a 1 log a 1 a ,

or, reversing signs

\tag{*} a - 1 \geqq \log a \geqq 1 - \frac{1}{a}

Thus we have proved the inequality (*) for all a > 0 .

Let us use this inequality to derive some properties of log x .
Using a = 1 + 1 n , n a positive integer, we find

1 n log ( 1 + 1 n ) 1 1 1 + 1 n = 1 n n + 1 = 1 n + 1

Since n ( log a ) = log ( a n ) (prove by induction), we have

1 = n n n log ( 1 + 1 n ) = log ( 1 + 1 n ) n n n + 1

Now as n becomes large, n n + 1 approaches 1 . Therefore log ( 1 + 1 n ) n is between 1 and a quantity which approaches 1 , and hence must itself approach 1 . We write this as

lim n log ( 1 + 1 n ) n = 1.

Next consider the expression log ( 2 n ) = n log 2 n ( 1 1 2 ) = n 2 . As n becomes large, n 2 also becomes large, and consequently log ( 2 n ) becomes as large as we wish. Similarly, log ( 2 n ) n 2 , and as n becomes arbitrarily large, log ( 2 n ) becomes arbitrarily small (i.e., far negative). Thus we see that log x takes on arbitrarily large and arbitrarily small values.

log x is continuous at every x > 0 , since it is differentiable. Moreover, it is strictly increasing. If x 1 < 1 , x 2 1 , it is clear that log x 1 < 0 log x 2 . If 1 x 1 < x 2 , then the area under y = 1 / x from 1 to x 1 is evidently less than that from 1 to x 2 , so again log x 1 < log x 2 . Finally, if x 1 < x 2 < 1 , the area from x 1 to 1 is greater than that from x 2 to 1 ; but the logarithms are these areas taken with a negative sign, giving once more log x 1 < log x 2 . Thus we have proved the assertion.