Using a graphing calculator, it seems that $\lim _{x\to \infty \:}(1+\frac{x^x}{x!})^{\frac{1}{x}} = e$. How can this be proven?
-
2What about Stirling's approximation? – Henricus V. Feb 28 '16 at 04:48
-
1The leading one has no meaning. Try it with 0, 5, 500... – Fred Daniel Kline Feb 28 '16 at 05:51
-
It makes the limit more interesting though – Yuriy S Feb 28 '16 at 05:54
4 Answers
Using Stirling, we find:
$$ \frac{x^x}{x!} \asymp \sqrt{2 \pi x} e^x$$
Now the best thing to do would be to write logarithm of your expression and take the limit:
$$\lim_{x \to \infty} \frac{1}{x} \log \left(1+\sqrt{2 \pi x} e^x\right)=\lim_{x \to \infty} \frac{1}{x} \log \left(\sqrt{2 \pi x} e^x\right)=\lim_{x \to \infty} \frac{1}{x} \left( \frac{\log 2}{2}+\frac{\log \pi}{2}+\frac{\log x}{2}+x \right)=1$$
So
$$\lim_{x \to \infty} \log \left(1+\frac{x^x}{x!} \right)^{1/x}=1$$
Meaning, the original limit is $e$

- 31,474
Since $n! > (n/e)^n$, for integer $x$, $1+\frac{x^x}{x!} <1+\frac{x^x}{(x/e)^x} =1+e^x $.
Therefore, for integer $x$, $(1+\frac{x^x}{x!})^{\frac{1}{x}} < (1+e^x)^{1/x} =e (1+e^{-x})^{1/x} =e (1+\frac{e^{-x}}{x}) =e +e\frac{e^{-x}}{x} $.
Since $n! < (n/e)^{n+1}$, for integer $x$, $1+\frac{x^x}{x!} \gt 1+\frac{x^x}{(x/e)^{x+1}} =1+\frac{e^{x+1}}{x} $.
Therefore, for integer $x$, $(1+\frac{x^x}{x!})^{\frac{1}{x}} > (1+\frac{e^{x+1}}{x})^{1/x} = e (e^{-x}+\frac{e}{x})^{1/x} = e \left(\frac{e}{x}\right)^{1/x}(1+e^{-x}\frac{x}{e})^{1/x} \gt e +\left(\frac{e}{x}\right)^{1/x} $.
Therefore, for integer $x$, $e \left(\frac{e}{x}\right)^{1/x} <(1+\frac{x^x}{x!})^{\frac{1}{x}} <e +e\frac{e^{-x}}{x} $ and the limits of both sides are $e$.

- 107,799
Using the fact
$$\lim_{n\to \infty} a_n^{1/n}=\lim_{n\to \infty} \frac{a_{n+1}}{a_n} $$
we have
$$\frac{a_{n+1} }{a_n} = \frac{ (n+1)! + (1+n)^{1+n} }{(n+1)!+(1+n)n^n} \sim_{n\sim \infty} \frac{ (1+n)^{1+n} }{n^{n+1} } \longrightarrow_{n\to \infty} e $$

- 47,431
We can proceed without appealing to Stirling's Formula. Rather, we use straightforward arithmetic, elementary inequalities, and evaluation of a Riemann sum.
Note that exploiting the monotonicity of the Gamma function, we can write
$$\lim_{x\to \infty}\left(1+\frac{x^x}{x!}\right)^{1/x}=\lim_{n\to \infty}e^{\frac1n \log\left(1+\frac{n^n}{n!}\right)} $$
Next, we observe that
$$\begin{align} \lim_{n\to \infty}\frac1n \log\left(1+\frac{n^n}{n!}\right)&=\lim_{n\to \infty}\left(\frac1n\log\left(\frac{n^n}{n!}\right)+\frac1n \log\left(1+\frac{n!}{n^n}\right)\right)\\\\ &\lim_{n\to \infty}\left(-\frac1n\sum_{k=1}^n\log(k/n)+\frac1n \log\left(1+\frac{n!}{n^n}\right)\right) \tag 1\\\\ &=-\int_0^1 \log(x)\,dx \tag 2\\\\ &=1 \end{align}$$ Therefore, the continuity of the exponential function guarantees that
$$\bbox[5px,border:2px solid #C0A000]{\lim_{x\to \infty}\left(1+\frac{x^x}{x!}\right)^{1/x}=e}$$
as was to be shown!
In going from $(1)$ to $(2)$ we noted first that $\lim_{n\to \infty}\left(\frac1n\sum_{k=1}^n\log(k/n)\right)$ is the Riemann sum for $\int_0^1 \log(x)\,dx$.
Then, we exploited the inequalities $\log(1+z)\le z$ for $z>-1$, and $n!\le n^n$, for $n\ge 1$, which shows that $$0\le \frac1n\log\left(1+\frac{n!}{n^n}\right)\le \frac1n$$ whereupon applying the squeeze theorem reveals $$\lim_{n\to \infty}\frac1n\log\left(1+\frac{n!}{n^n}\right)=0$$

- 179,405
-
We have $\frac{1}{n}\log(1+n^n/n!) \to 1$ not $0$ so this argument has some flaws. – Kibble Feb 28 '16 at 05:51
-
@Kibble Thanks; good catch. I've edited. Let me know how this works for you. - Mark – Mark Viola Feb 28 '16 at 06:21
-
I fail to get the "exploiting the continuity of the Gamma function" argument, would you be alluding to the monotonicity of the Gamma function? – Did Apr 11 '16 at 21:06
-
@Did Didier, Well, I was really alluding to how is $x!$ is defined? That is, it is defined as the Gamma function, which is continuous at all arguments with non-negative integer real parts. Suppose that one defined a function $f(x)$ so that $f(n) = n!$, but was discontinuous at $n$. I understand your comment and can modify my answer if you think it would help clarify. -Mark – Mark Viola Apr 11 '16 at 21:39
-
I got that, the point is that continuity is irrelevant here since one could imagine a fake function Gamma, continuous but behaving erratically between the integers, then the step reducing the limit on the real numbers x to the same limit on the integers n would fail. – Did Apr 11 '16 at 21:41
-
@Did I edited and replaced continuity with monotonicity. Yes, continuity would be insufficient. Thank you for the catch! – Mark Viola Apr 11 '16 at 21:49