0

I am given the sequence $(x_n)_{n \ge 2}$ such that:

$$x_n = \sqrt[n]{1 + \sum\limits_{k=2}^n(k-1)(k-1)!}$$

And I am asked to find $\lim\limits_{n \to \infty} \dfrac{x_n}{n}$ . While trying to solve this, I first made the following observation:

$$\begin{align}\sum\limits_{k = 2}^n (k-1)(k-1)! &= \\ &=\sum\limits_{k = 2}^n(k-1+1)(k-1)! - (k-1)! \\ &=\sum\limits_{k=2}^nk(k-1)! - (k-1)! \\ &=\sum\limits_{k=2}^nk!-(k-1)! \\ &=2!-1!+3!-2!+4!-3!+5!-4!+...n!-(n-1)! \\ &= \boxed{\color{#D04}{-1 + n!}} \end{align}$$

So, $x_n$ becomes:

$$ \begin{align}x_n &= \sqrt[n]{1-1+n!} \\ x_n &= \color{#a90}{\sqrt[n]{n!}} \end{align}$$

So the limit I have to find is:

$$\lim\limits_{n \to \infty} \dfrac{x_n}{n} = \lim\limits_{n \to \infty} \dfrac{\sqrt[n]{n!}}{n} = \lim\limits_{n \to \infty} \sqrt[n]{\dfrac{n!}{n^n}}$$

And this is where I got stuck, kinda. I could find the limit by using the Stirling Approximation:

$$n! \approx \sqrt{2 \pi n} \bigg (\dfrac{n}{e} \bigg )^n$$

By replacing the $n!$ term in my limit with this approximation, I eventually found that

$$\lim\limits_{n \to \infty} \dfrac{x_n}{n} = \dfrac{1}{e}$$

Which seems to be the right result, according to my textbook.

What I don't like about this is that use of the Stirling Approximation. It always feels a bit off using an approximation to get an exact result. My question is this: is there any other way of arriving at this answer without using the Stirling Approximation? In other words, can I find the following limit:

$$\lim\limits_{n \to \infty} \sqrt[n]{\dfrac{n!}{n^n}}$$

Without using an approximation?

  • See also https://math.stackexchange.com/q/1617252/42969 – Martin R Dec 01 '19 at 12:55
  • Also note that it is, $n! \color{red}{\sim} \sqrt{2 \pi n} \bigg (\dfrac{n}{e} \bigg )^n$. It's not an approximation, but an asymptote, which gets closer to value as the value of n, as $n \rightarrow \infty$. So its use is justifiable. – user712576 Dec 01 '19 at 13:02
  • @user712576 So if I'd use it on an exam, would I get maximum points? –  Dec 01 '19 at 21:51
  • You might wanna use the version with the error term: $\sqrt{2\pi} n^{n+1/2}e^{-n}\lt n! \lt \sqrt{2\pi} n^{n+1/2}e^{-n}(1+\dfrac{1}{4n})$. – Divide1918 Dec 02 '19 at 03:34

0 Answers0