2
$\begingroup$

I want to prove this,

$$2^23^34^45^5\cdots n^n \leq \Big(n+1-(n!)^{\frac{1}{n}}\Big)^{(n(n+1))0.5}$$

I try induction but we can't multiply by $(n+1)^{n+1}$ it's too big .On the other hand i try to study the following function : $$ f(n)=\Big(n+1-(n!)^{\frac{1}{n}}\Big)^{(n(n+1))0.5} )$$

But it reveals nothing.

I try also to approximate the factorial but it's not enought

I prefer hints.

Thanks a lot for sharing your time.

$\endgroup$
  • $\begingroup$ Note: Same setup as mathoverflow.net/questions/344504/an-olympiad-like-inequality with $f(x) = x$, though the general case isn't true. $\endgroup$ – Calvin Lin Nov 13 '19 at 7:17
  • $\begingroup$ $n\ge 3$ is needed. Use bounds for $\sum k\ln k$ and $(n!)^{1/n}$. $\endgroup$ – River Li Nov 13 '19 at 14:43
  • $\begingroup$ With Stirling, you can show $n+1-(n!)^{1/n}>n\left(1-\frac{1}{e}\right)-\varepsilon$ for large enough $n$. Numerically checking for some values of $\varepsilon$, it appears that for large enough $n$, the stronger bound $2^23^34^45^5\cdots n^n \leq \left(n\left(1-\frac{1}{e}\right)-\varepsilon\right)^{n(n+1)/2}$ holds. $\endgroup$ – alex.jordan Nov 13 '19 at 15:46
  • 1
    $\begingroup$ Still any question left ? ;) $\endgroup$ – user90369 Nov 13 '19 at 20:46
2
$\begingroup$

This is a long chain of symbol-pushing, with repeated use of Stirling. Maybe it has mistakes. If it does not, it surely could be cleaned up and streamlined. But at the moment, I think it works. The strategy is to use these Stirling bounds: $$\sqrt{2\pi n}\frac{n^n}{e^n}<n!<\sqrt{2\pi n}\frac{n^n}{e^n}\left(1+\frac{1}{11n}\right)$$ using the appropriate one of these two inequalities to keep the argument going. Eventually there are two sides that have simplified to each having a finite limit, and the lef-side limit is less than the right-side limit.

The left side is: $$P(n)=\frac{(n!)^n}{\prod_{k=1}^{n-1}k!}$$ By Stirling: $$ \begin{align} P(n) &<\frac{\left(n!\right)^n}{\prod_{k=1}^{n-1}\left(\sqrt{2\pi k}\frac{k^k}{e^k}\right)}\\ &=\frac{\left(n!\right)^n}{(2\pi)^{(n-1)/2}\sqrt{(n-1)!}e^{-n(n-1)/2}\prod_{k=1}^{n-1}\left(k^k\right)}\\ &=\frac{n^n\left(n!\right)^n}{(2\pi)^{(n-1)/2}\sqrt{(n-1)!}e^{-n(n-1)/2}\prod_{k=1}^{n}\left(k^k\right)}\\ &=\frac{n^n\left(n!\right)^n}{(2\pi)^{(n-1)/2}\sqrt{(n-1)!}e^{-n(n-1)/2}P(n)}\end{align} $$

Multiply by $P(n)$ and take the square root:

$$ P(n)<\sqrt{\frac{n^n\left(n!\right)^n}{(2\pi)^{(n-1)/2}\sqrt{(n-1)!}e^{-n(n-1)/2}}} $$

So we want to show $$\sqrt{\frac{n^n\left(n!\right)^n}{(2\pi)^{(n-1)/2}\sqrt{(n-1)!}e^{-n(n-1)/2}}}<\left(n+1-(n!)^{1/n}\right)^{n(n+1)/2}$$ which is equivalent to: $$\frac{n\left(n!\right)e^{(n-1)/2}\sqrt[2n]{2\pi n}}{\sqrt{2\pi}\sqrt[2n]{n!}}<\left(n+1-(n!)^{1/n}\right)^{n+1}$$

But since $n!>\sqrt{2\pi n}\frac{n^n}{e^n}$, it suffices to show: $$\frac{n\left(n!\right)e^{(n-1)/2}\sqrt[2n]{2\pi n}}{\sqrt{2\pi}\sqrt[2n]{\sqrt{2\pi n}\frac{n^n}{e^n}}}<\left(n+1-(n!)^{1/n}\right)^{n+1}$$

which is equivalent to: $$\frac{\sqrt{n}\left(n!\right)e^{n/2}\sqrt[4n]{2\pi n}}{\sqrt{2\pi}}<\left(n+1-(n!)^{1/n}\right)^{n+1}$$

Using Stirling to more precision, for large enough $n$, $n!<\sqrt{2\pi n}\frac{n^n}{e^n}\left(1+\frac{1}{11n}\right)$. And it suffices to show:

$$\frac{\sqrt{n}\sqrt{2\pi n}\frac{n^n}{e^n}\left(1+\frac{1}{11n}\right)e^{n/2}\sqrt[4n]{2\pi n}}{\sqrt{2\pi}}<\left(n+1-\left(\sqrt{2\pi n}\frac{n^n}{e^n}\left(1+\frac{1}{11n}\right)\right)^{1/n}\right)^{n+1}$$

which is equivalent to $$n\sqrt[n+1]{\frac{\left(1+\frac{1}{11n}\right)\sqrt[4n]{2\pi n}}{e^{n/2}}}<n+1-\frac{n}{e}\left(\sqrt{2\pi n}\left(1+\frac{1}{11n}\right)\right)^{1/n}$$

It suffices to show $$n\sqrt[n+1]{\frac{\left(1+\frac{1}{11n}\right)\sqrt[4n]{2\pi n}}{e^{n/2}}}<n-\frac{n}{e}\left(\sqrt{2\pi n}\left(1+\frac{1}{11n}\right)\right)^{1/n}$$ which is equivalent to $$\sqrt[n+1]{e^{n/2+1}\left(1+\frac{1}{11n}\right)\sqrt[4n]{2\pi n}}<e-\sqrt[n]{\sqrt{2\pi n}\left(1+\frac{1}{11n}\right)}$$

The left side converges to $\sqrt{e}\approx1.648$ while the right side converges to $e-1\approx1.718$. So this inequality holds for large enough $n$. (It appears from a spreadsheet that it holds for $n\geq56$.)

It remains to check the original inequality for small values of $n$.

$\endgroup$
1
$\begingroup$

For e.g. $~n=2~$ it's not correct, but from a certain value $~n_0~$ already.

Comparing asymptotic approximations.

$\displaystyle \left(n^{-\left({\frac{n^2}{2}+\frac{n}{2}}\right)}\prod\limits_{k=1}^n k^k\right)^\frac{2}{n^2}\approx \left(A e^{-\frac{n^2}{4}}n^{\frac{1}{12}} \right)^\frac{2}{n^2} \approx \frac{1}{\sqrt{e}}$

where $~A~$ is called the Glaisher–Kinkelin constant

$\displaystyle \left(n^{-\left({\frac{n^2}{2}+\frac{n}{2}}\right)}\left(n+1-\sqrt[n]{n!}\right)^{\frac{n(n+1)}{2}}\right)^\frac{2}{n^2}\approx 1-\frac{1}{n}-\frac{\sqrt[n]{n!}}{n}\approx 1-\frac{1}{e}$

It's $~\displaystyle \frac{1}{\sqrt{e}} < 1-\frac{1}{e}~$ .

This means that $~n_0~$ exists so that the claim is correct for $~n\geq n_0~$ .

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.