Stirling's approximation gives an approximate value for the factorial function
or the gamma function
for
. The approximation can most simply be derived for
an integer by approximating the sum over the terms of
the factorial with an integral,
so that
|
(1)
| |||
|
(2)
| |||
|
(3)
| |||
|
(4)
| |||
|
(5)
| |||
|
(6)
|
The equation can also be derived using the integral definition of the factorial,
|
(7)
|
Note that the derivative of the logarithm of the integrand can be written
|
(8)
|
The integrand is sharply peaked with the contribution important only near . Therefore, let
where
, and write
|
(9)
| |||
|
(10)
|
Now,
|
(11)
| |||
|
(12)
| |||
|
(13)
|
so
|
(14)
| |||
|
(15)
| |||
|
(16)
|
Taking the exponential of each side then gives
|
(17)
| |||
|
(18)
|
Plugging into the integral expression for then gives
|
(19)
| |||
|
(20)
|
Evaluating the integral gives
|
(21)
| |||
|
(22)
|
(Wells 1986, p. 45). Taking the logarithm of both sides then gives
|
(23)
| |||
|
(24)
|
This is Stirling's series with only the first term retained and, for large , it reduces to Stirling's approximation
|
(25)
|
Taking successive terms of , where
is the floor function,
gives the sequence 1, 2, 4, 10, 26, 64, 163, 416, 1067, 2755, ... (OEIS A055775).
Stirling's approximation can be extended to the double inequality
|
(26)
|
(Robbins 1955, Feller 1968).
Gosper has noted that a better approximation to (i.e., one which approximates the terms in Stirling's
series instead of truncating them) is given by
|
(27)
|
Considering
a real number so that
, the equation (27) also
gives a much closer approximation to the factorial
of 0,
,
yielding
instead of 0 obtained with the conventional Stirling approximation.