PlanetMath (more info)
 Math for the people, by the people. Sponsor PlanetMath
Encyclopedia | Requests | Forums | Docs | Wiki | Random | RSS  
Login
create new user
name:
pass:
forget your password?
Main Menu
Owner confidence rating: High Entry average rating: No information on entry rating
[parent] asymptotic bounds for factorial (Result)

Stirling's formula furnishes the following asymptotic bound for the factorial:

$\displaystyle n! = \sqrt{2\pi} \, n^{n+1/2} e^{-n} \, \bigl( 1+ \Theta(\tfrac 1 n) \bigr)\,, \quad n \to \infty\,.$    

The following less precise bounds are useful for counting purposes in computer science:
$\displaystyle n!$ $\displaystyle = o(n^n)$ (1)
$\displaystyle n!$ $\displaystyle = \omega(n^{\lambda n})$   for every $ 0 \leq \lambda < 1$ (2)
$\displaystyle \log n!$ $\displaystyle = \Theta(n \log n)$ (3)

($o$ , $\omega$ , and $\Theta$ are Landau notation.)
Proof. [Proof of the upper bound of equation (1)] We must show that for every $\varepsilon > 0$ , we have $$ n! \leq \varepsilon \, n^n \quad \text{for large enough $n$.} $$

To see this, write

$\displaystyle n!$ $\displaystyle = \underbrace{\sqrt{2\pi} \, \frac{\sqrt{n}}{e^n} \, \bigl(1 + O(\tfrac 1 n) \bigr)} \, n^n\,,$    

and observe that the middle quantity tends to $0$ as $n \to \infty$ , so it can be made smaller than any fixed $\varepsilon$ by taking $n$ large. $ \qedsymbol$
Proof. [Proof of the lower bound of equation (2)] We are to show that for every $C > 0$ , we have $$ n! \geq C \, n^{\lambda n} \quad \text{for large enough $n$.} $$

We write:

$\displaystyle n!$ $\displaystyle = \underbrace{\sqrt{2\pi n} \, \bigl( 1+O(\tfrac 1 n) \bigr) \left(\frac{n}{e^{1/(1-\lambda)}}\right)^{(1-\lambda) n}} \, n^{\lambda n}\,,$    

and the middle expression can be made to be $\geq C$ by taking $n$ to be large. $ \qedsymbol$
Proof. [Proof of bound for $\log n!$ , equation (3)] Showing $\log n! = O (n \log n)$ is simple:
$\displaystyle n!$ $\displaystyle = n(n-1)\dotsm(2)(1) \leq n^n$    
$\displaystyle \log n!$ $\displaystyle \leq n \log n$    

for all $n \geq 1$ .

To show $\log n! = \Omega (n \log n)$ , we can take equation (2) with the constants $C = 1$ and $\lambda = \tfrac 1 2$ . Then

$\displaystyle \log n! \geq \tfrac 1 2 \, n \log n$    

for large enough $n$ . In fact, it may be checked that $n \geq 1$ suffices. $ \qedsymbol$

Bibliography

1
Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, Clifford Stein. Introduction to Algorithms, second edition. MIT Press, 2001.
2
Michael Spivak. Calculus, third edition. Publish or Perish, 1994.




"asymptotic bounds for factorial" is owned by stevecheng.
(view preamble | get metadata)

View style:

See Also: Stirling's approximation, growth of exponential function


This object's parent.
Log in to rate this entry.
(view current ratings)

Cross-references: simple, expression, lower bound, fixed, equation, upper bound, Landau notation, computer, useful, factorial, bound, Stirling's formula
There is 1 reference to this entry.

This is version 4 of asymptotic bounds for factorial, born on 2006-05-09, modified 2006-05-10.
Object id is 7910, canonical name is AsymptoticBoundsForFactorial.
Accessed 6740 times total.

Classification:
AMS MSC41A60 (Approximations and expansions :: Asymptotic approximations, asymptotic expansions )
 68Q25 (Computer science :: Theory of computing :: Analysis of algorithms and problem complexity)

Pending Errata and Addenda
None.
[ View all 1 ]
Discussion
Style: Expand: Order:
forum policy
binomial estimate? by Algeboy on 2006-05-11 12:43:13
It could be helpful to add a word about the resulting asymptotic estimates of nPr and nCr following from Sterling's formula. Of course this is widely available on other sites so it is not a priority.
[ reply | up ]

Interact
post | correct | update request | add example | add (any)