This was a question in a maths contest, where no calculator was allowed. Also, note that only a (>,< or =) relationship is being searched for and not the value of the numbers itself.

Which is larger, $\sqrt[2015]{2015!}$ or $\sqrt[2016]{2016!}$ ?

What I've done: My approach is to divide one number by the other and infer from the result which number is the bigger one;
WolframAlpha gives $\frac{\sqrt[2016]{2016!}}{\sqrt[2015]{2015!}}=1.0049\ldots$, so clearly $\sqrt[2016]{2016!}>\sqrt[2015]{2015!}$

Let $a=\sqrt[2016]{2016!}$ and $b=\sqrt[2015]{2015!}$

$\therefore a=\sqrt[2016]{2016!}={2016!}^{1 \over 2016}=2016^{1 \over 2016}\times2015!^{1\over 2016}=\sqrt[2016]{2016}\cdot \sqrt[2016]{2015!}$
$\therefore b=\sqrt[2015]{2015!}={2015!}^{1 \over 2015}={2015!}^{\frac{2016}{2015}\cdot\frac{1}{2016}}=\sqrt[2016]{2015!^{2016 \over 2015}}$

Hence $$\begin{align} \require{cancel} \frac{a}{b}=\frac{\sqrt[2016]{2016!}}{\sqrt[2015]{2015!}}&=\frac{\sqrt[2016]{2016}\cdot \sqrt[2016]{2015!}}{\sqrt[2016]{2015!^{2016 \over 2015}}}\\ &=\sqrt[2016]{2016}\cdot \sqrt[2016]{2015!^{\frac{-1}{2015}}}= \cancelto{*}{\sqrt[2016]{\frac{2016}{2015!^{2015}}} \quad \text{which appears to be} <1}\\ =\sqrt[2016]{\frac{2016}{2015!^\frac{1}{2015}}}\\ \end{align}$$ That is $\cancelto{*}{\frac{a}{b}<1 \implies a<b}$ which is false as per the result from WA.

*: Correction due to error pinpointed out by Daniel Fischer. But now I'm stuck; how do I infer what value is $\sqrt[2016]{\frac{2016}{2015!^\frac{1}{2015}}}$?

So, where did I go wrong?. How do I proceed now?

  • 6
    $\begingroup$ $$(2015!)^{\frac{-1}{2015}} = \frac{1}{(2015!)^{\frac{1}{2015}}} \neq \frac{1}{(2015!)^{2015}}$$ $\endgroup$ – Daniel Fischer Jan 4 '16 at 10:17
  • 6
    $\begingroup$ It's easier to work with only integer exponents. You have $$\sqrt[n+1]{(n+1)!} > \sqrt[n]{n!} \iff ((n+1)!)^n > (n!)^{n+1}.$$ I think you can work from that. $\endgroup$ – Daniel Fischer Jan 4 '16 at 10:40
  • $\begingroup$ But how do we know that $\sqrt[n+1]{(n+1)!} > \sqrt[n]{n!}$ ? $\endgroup$ – K. Rmth Jan 4 '16 at 10:43
  • 1
    $\begingroup$ $$\frac ab=\sqrt[2015\cdot 2016]{\frac{2016^{2015}}{2015!}}\gt 1$$ $\endgroup$ – mathlove Jan 4 '16 at 10:46

Starting with:


Raise each side to the power of $2015\cdot2016$:


Divide each side by $2015!^{2015}$:


Write it explicitly as:

$$\underbrace{1\cdot2\cdot\ldots\cdot2015}_{2015\text{ terms}}\mid\underbrace{2016\cdot2016\cdot\ldots\cdot2016}_{2015\text{ terms}}$$

Obviously, the RHS is larger.

| cite | improve this answer | |
  • 7
    $\begingroup$ | looks like dividing relation so it is little bit confusing. $\endgroup$ – dust05 Jan 4 '16 at 11:57
  • 4
    $\begingroup$ "Raise each side to the power of $2015⋅2016$" nice trick $\endgroup$ – Apeiron Jan 4 '16 at 12:17
  • 1
    $\begingroup$ @barakmanos Well, how about $\gtreqless$ or $\stackrel{?}{\gtreqless}$ ? $\endgroup$ – dust05 Jan 4 '16 at 12:28
  • 1
    $\begingroup$ @dust05: I wouldn't even know how to make those in LaTex. Perhaps I should just add "vs" in between the expressions... Or if I could make that "$\mid$" longer... $\endgroup$ – barak manos Jan 4 '16 at 13:12
  • 1
    $\begingroup$ @barakmanos Simply begin with the assumption that $\sqrt[2015] {2015} > \sqrt[2016] {2016}$ and then come to the conclusion that $\sqrt[2015] {2015} \not > \sqrt[2016] {2016}$. No fancy notation needed. $\endgroup$ – Axoren Jan 5 '16 at 5:34

Look at the function: $$y(x)=\sqrt[x]{x!}=\left(\Gamma(x+1)\right)^{1/x}$$ It's a monotonically strictly increasing function, as: $$y'(x) = \frac{1}{x}\left(\Gamma(x+1)\right)^{1/x-1}\Gamma'(x+1)>0,\forall x>1$$ $\blacksquare$

As a side note: $$\lim_{x\to\infty}y(x)=\frac{1}{e}x$$ So you can expect the RHS to be lager by a factor of around $1+1/2015$.

| cite | improve this answer | |

$$\log \left(2016!^{1/2016}\right) -\log \left(2015!^{1/2015}\right)=$$ $$=\frac {1}{2016}\log (2016)-\left(\frac {1}{2015}-\frac {1}{2016}\right)\sum_{n=1}^{2015}\log n=$$ $$=\frac {1}{2016}\log (2016)- \frac {1}{(2016)(2015)}\sum_{n=1}^{2015}\log n>$$ $$>\frac {1}{2016}\log (2016)-\frac {1}{(2016)(2015)}\sum_{n=1}^{2015}\log (2015)=$$ $$=\frac {1}{2016}\log (2016)-\frac {1}{2016}\log (2015)\;>0\;.$$

| cite | improve this answer | |

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.