1

Find $\displaystyle \lim_{x\to\infty} \frac {(x!)^{\frac 1 x}}{x}$

I have no idea how to solve it, I can approximate it to be in $(0,1)$ by squeezing but getting to the solution $(\frac 1 e)$ seems like it would require a lot more. Is this an identity?

Note: no integrals nor gamma function.

shinzou
  • 3,981

5 Answers5

4

While waiting for someone to find a duplicate a reminder of Stirling formula

$$x\to\infty\implies x!\sim \sqrt{2\pi x}\left(\frac{x}{e}\right)^x$$

So we have, when $x\to\infty$,

$$\frac{(x!)^{\frac{1}{x}}}{x}\sim \frac{1}{e}\left(\sqrt{2\pi}\right)^{\frac{1}{x}}x^{\frac{1}{2x}}\sim \frac{1}{e}$$

marwalix
  • 16,773
2

Note this $$ \left( \frac{x!}{x^x} \right)^{1/x} = (a_x)^{1/x} $$

where $a_x = \frac{x!}{x^x}$ and then use the fact that

$$ \lim_{x\to \infty} (a_x)^{1/x} = \lim_{x\to \infty} \frac{a_{x+1}}{a_x} $$

and the evaluation of limit will become easy

$$ \lim_{x\to \infty} \frac{a_{x+1}}{a_x} = \lim_{x\to \infty} \frac{1}{(1+1/x)^x} = \frac{1}{e}. $$

science
  • 2,900
2

Just for variety, here's an alternative solution, based on the first simple observation in Terry Tao's notes on Stirling's formula.

We'll show that $$ \frac{n^n}{n!} \le e^n \le 2(n+1)\frac{n^n}{n!} \tag{$\ast$} $$ Taking $n$th roots and squeezing yields the desired limit. (The standard Riemann sum method — as, e.g., in the first stage in Tim Gowers' "Removing the magic from Stirling's formula" — gives the similar bounds $e\frac{n^n}{n!} \le e^n \le en\frac{n^n}{n!}$.)

Tao's observation gives us the lower bound in ($\ast$): $$ \frac{n^n}{n!} \le \sum_{k=0}^\infty \frac{n^k}{k!} = e^n $$ Now let's consider how much we lost there by throwing away all terms of the series except the one with $k=n$. First, $$ \frac{n^k}{k!} \le \frac{n^{k+1}}{(k+1)!} \iff k\le n $$ and so the terms in the series $\sum_{k=0}^\infty \frac{n^k}{k!}$ increase from $k=0$ to $k=n$ and decrease thereafter. In particular, for a longish head of the series we get $$ \sum_{k=0}^{2n-1} \frac{n^k}{k!} \le 2n\cdot\frac{n^n}{n!} \tag{1} $$ In the remaining tail, we have $k\ge 2n$, and so $$ \frac{n^{k+1}}{(k+1)!} = \frac{n}{k+1}\cdot\frac{n^k}{k!} \le \frac12\cdot\frac{n^k}{k!} $$ so each term in the tail is less than half the preceding one. Thus we can bound them by the geometric series $1+\frac12+\frac14+\dotsb$, getting $$ \sum_{k=2n}^\infty \frac{n^k}{k!} \le 2\frac{n^{2n}}{(2n)!} \le 2\frac{n^n}{n!} $$ which when combined with (1) gives the upper bound in ($\ast$).

2

Note that $$ \frac{(n+1)!}{n!}=n+1 $$ and $$ \frac{(n+1)^{n+1}}{n^n}=\left(1+\frac1n\right)^{n}(n+1) $$ Therefore, $$ \left.\frac{(n+1)!}{(n+1)^{n+1}}\middle/\frac{n!}{n^n}\right.=\left(1+\frac1n\right)^{-n} $$ Since $\lim\limits_{n\to\infty}\left(1+\frac1n\right)^{-n}=e^{-1}$, Stolz-Cesaro applied to the logarithms says that $$ \lim_{n\to\infty}\frac{(n!)^{1/n}}{n}=\lim_{n\to\infty}\left(\frac{n!}{n^n}\right)^{1/n}=e^{-1} $$

robjohn
  • 345,667
1

Let us first consider the limit as $L$. Then, applying natural logarithm, we get,

$$\ln(L)=\lim_{x\to\infty}\frac{1}{x}\sum_{i=1}^x\left(\ln\left(\frac{i}{x}\right)\right)=\int\limits_0^1\ln(x)\,dx$$

Now, since the function $f(x)=\ln(x)$ is Riemann-integrable on $(0,1)$, we see, by the Riemann Sum argument that $L$ exists finitely. Next, we start to evaluate it.

$$L=\int\limits_0^1\ln (x)\,\mathrm dx=\left[x\ln(x)-x\right]_0^1\\ \implies \ln(L)=\left(\ln(1)-1\right)-\left(\lim_{x\to 0^+}(x\ln(x)-x)\right)=(-1)\\ \implies \ln(L)=(-1)\implies L=e^{-1}=\frac{1}{e}$$

  • 1
    This is nice but should be stated in another way. By saying "take the limit as $L$" we are assuming that the limit exists. It exists quite trivially (since $\log x$ is a Riemann-integrable function over $(0,1)$), but a better choice for wording would have been "since the following Riemann sum argument shows that [...] we have that the limit equals $\frac{1}{e}$". – Jack D'Aurizio Apr 12 '15 at 18:43
  • 1
    @JackD'Aurizio, I have edited my answer. Can you please verify it for me? I'm actually still learning Calculus, so I'm not familiar with many terms, like the term "Riemann integrable", etc. If you find any more errors, it'd be kind of you if you could notify me. Thanks. :) – Prasun Biswas Apr 12 '15 at 19:02