2
$\begingroup$

I'm having a hard time finding the constants/witnesses $C_0$ and $k_0$ that show $c^n \in O(n!)$. That is $|c^n| \leq C_0|n!|$ for $n > k_0$ ($c > 1$).

To be clear, I understand we can prove the Big-O relation through calculus, but I'm specifically interested in finding the constants/witnesses.

I'm really stuck here on how to solve this, since I can try solving for the points of intersection but am not sure how to. Also not sure of what calculus methods can help me find such constants.

Kindly please do help me here. Maybe somehow your solution comes out similar to Determining the witnesses (constants $C_0$ and $k_0$) when showing $(log_b n)^c$ is $O(n^d)$ (b > 1 and c,d are positive) (which works too but of course isn't necessary lol).

$\endgroup$
1

2 Answers 2

2
$\begingroup$

Calculus is completely unnecessary here and so is Stirling's approximation. Let's do $c = 2$ as a warmup. How do we compare $2^n$ and $n!$? Well, observe that if $n \ge 2$ then

$$\begin{align*} n! &= 2 \times 3 \times 4 \times \dots \times n \\ &\ge 2 \times 2 \times \dots \cdot 2 \\ &\ge 2^{n-1}. \end{align*}$$

This gives that $2^n \le 2 \times n!$, so we can take $k_0 = 2, C_0 = 2$ in this case. Hopefully it is intuitively clear that this argument generalizes but let's do $c = 3$ just to be sure. How do we compare $3^n$ and $n!$? Well, if $n \ge 3$ then

$$\begin{align*} n! &= 2 \times 3 \times 4 \times \dots \times n \\ &\ge 2 \times 3 \times \dots \cdot 3 \\ &\ge 2 \times 3^{n-2}. \end{align*}$$

This gives that $3^n \le \frac{9}{2} n!$, so we can take $k_0 = 3, C_0 = \frac{9}{2}$ in this case.

The point is that, by definition, once $n$ goes past any given positive integer $m$, $n!$ is always increasing by at least a factor of $m$. So it always grows at least as fast as (a constant times) $m^n$ from that point on. This follows directly from the definition and nothing even slightly advanced is required, one should be able to "see" it. The generalization of the above argument to an arbitrary positive integer $m$ is that if $n \ge m$ then

$$\begin{align*} n! &= 2 \times 3 \times 4 \times \dots \times n \\ &\ge 2 \times 3 \times \dots (m-1) \times m \times m \times \dots \times m \\ &\ge m! m^{n-m} \end{align*}$$

which gives that $\boxed{ m^n \le \frac{m^m}{m!} n! }$. If we now set $m = \lceil c \rceil$ to be the smallest positive integer larger than $c$, then $c^n \le m^n$, so we can take $k_0 = m, C_0 = \frac{m^m}{m!} = \frac{\lceil c \rceil^{\lceil c \rceil}}{\lceil c \rceil!}$. Of course by increasing $k_0$ we can decrease $C_0$ arbitrarily but we don't need this.

$\endgroup$
3
  • $\begingroup$ Wait but how is $3^{n-2}$ for example growing as fast as $3^n$ ? $\endgroup$
    – Bob Marley
    Commented Jul 4 at 22:39
  • $\begingroup$ Also when you say "once $n!$ goes past any given positive integer $m$", you mean once the terms in $n!$ go past $m$, right? $\endgroup$
    – Bob Marley
    Commented Jul 4 at 22:44
  • $\begingroup$ @Bob: I mean up to a multiplicative constant (which is after all the goal of this discussion). And yes, that's what I mean, sorry if that was unclear. $\endgroup$ Commented Jul 4 at 22:51
-1
$\begingroup$

Actually, in such cases the Stirling approximation formula may be helpful.

$$n! = \sqrt{2 \pi n} \cdot n^{n}e^{-n}e^{\frac{\theta}{12n}}$$

For some $\theta = \theta(n) \in (0,1)$

$\endgroup$
1
  • $\begingroup$ This is vastly overkill. $\endgroup$ Commented Jun 28 at 18:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .