7

There is a question asked me to compare $100! \ and \ 50^{100}$. I think more than 2 hours to solve it. I show my work below, but I looking for other Ideas to prove the inequality. Thanks in advance for any hints or new Ideas.

my work : $$50 \times 50 > 1 \times 99 \times 2\\ 50 \times 50 > \ 2 \times 98\\ 50 \times 50 > \ 3 \times 97\\ 50 \times 50 > \ 4 \times 96\\ \vdots \\ 50 \times 50 > \ 48 \times 52=(50+2)(50-2)=50^2-4\\ 50 \times 50 > \ 49 \times 51=(50+1)(50-1)=50^2-1\\ 50 \times 50 \geq 50 \times 50 $$ then multiply them $$50^{49}\times 50^{49}\times 50^2 >1.2.3...50....99.(2.50)\\50^{100}>100!$$

Remark: I use numerical approximation for $50^{100}\sim 7.8886\times 10^{169} $ and $100!=9.33236\times 10^{157}$ But is not interesting( Ithink).

Khosrotash
  • 24,922
  • 3
    What you show is probably the easiest way to show it. It's certainly the way I came up with. Otherwise, Stirling's approximation is useful, but not nearly as elegant. – Brian Tung Mar 04 '21 at 19:50
  • 1
    Log base two might be useful – JP McCarthy Mar 04 '21 at 20:02
  • @JPMcCarthy: Can you elaborate more? or give me the clue? I do not mean for $50^{100}$ this is easy to you $\log$ but I don't know how to use it for $100!$?! – Khosrotash Mar 04 '21 at 20:07
  • Not sure if it works here but worth a try: like this https://math.stackexchange.com/questions/562538/which-is-larger-20-or-240/562552#562552 – JP McCarthy Mar 04 '21 at 20:53
  • Moreover perhaps other ideas in that question. – JP McCarthy Mar 04 '21 at 20:54
  • The $\log_2$ approach does not work... – JP McCarthy Mar 04 '21 at 21:25
  • That's the best way. And it should be clear that both sides have $100$ terms that average near $50$. But $\sqrt[100]{n!}$ would be the geometric mean. And AM.GM says GM < AM. SO that should be a thought that can occur early. Have to tweak as the actual AM is $50.5$. – fleablood Mar 04 '21 at 21:43

3 Answers3

4

Well they both have $100$ factor components.

The LHS the factor components run from $1$ to $100$. The RHS the factors are a constant coomponent of $50$.

$50$ is almost the arithmetic average of $1$ to $100$. By the A.M. G.M inequality that is greater than the geometric mean of $1$ to $100$ which is $\sqrt[100]{1*2*3*....*100}$.

And so we ought to have $\sqrt[100]{100!} < 50$ and the $100! < 50^{100}$.

.... but the arithmetic average is actually $50.5$ so we have $\sqrt[100]{100!} < 50.1$.

But we can probably tweak it not to make such a difference.

.....

And indeed you did tweak it!

$2\times 4 \times ....... \times 98 =98!$ and $\sqrt[97]{98!}$ is the geometry mean of $2,....,98$. And the arithmetic mean of $2,...., 98$ is $50$.

So we have $\sqrt[97]{98!} < 50$ and so $98! < 50^{97}$

Now the tweak: $99\times 100 < 100^2 = 4\times 50^2 < 50^3$ so $100! = 98!\cdot (99\times 100) < 50^{97} \times 50^3= 50^{100}$

The tweak makes sense to look at the edge the difference between geometric mean and arithmetic mean is greatest when the numbers are far apart. In this case we managed to be a product of three terms which wasn't just less than the cube of the AM. It was barely significantly more than the square of the AM.

fleablood
  • 124,253
1

Definitely quicker than 2 hours, plus one more method to have in arsenal (similar to this, by the way). At some point I will use $$\log(1+x)\leq x, \forall x \geq 0$$ $$\log{2}<\frac{3}{4} \iff 16 < e^3$$ (in fact $2.6^3>16 \iff 13^3 > 5^3\cdot 4^2=5\cdot400$) and $$\log{101}<\log{128}=7\log{2}<\frac{21}{4}$$

Let's look at $\sum\limits_{k=1}^{100}\log{k}$. Function $f(x)=\log{x}$ is monotone ascending, thus: $$\sum\limits_{k=1}^{100}\log{k}= \sum\limits_{k=1}^{100}\log{k} \cdot (k+1 - k)\leq \int\limits_{1}^{101}\log{x} dx = x (\log{x}-1) \Big|_{1}^{101}=\\ 101\log{101}-101+1= 101\log{101}-100=\\ 100\log{100}+100\log{\left(\frac{101}{100}\right)}+\log{101}-100<\\ 100\log{100}+100\log{\left(1+\frac{1}{100}\right)}+\frac{21}{4}-100\leq\\ 100\log{100}+1+\frac{21}{4}-100=\\ 100\log{50}+100\log{2}+\frac{25}{4}-100<\\ 100\log{50}+100\cdot \frac{3}{4}+\frac{25}{4}-100<100\log{50}$$

rtybase
  • 16,907
  • VERY cute...very nice. But I have a little illusion in $$\cdot (k+1 - k)\leq \int\limits_{1}^{101}\log{x} dx = x (\log{x}-1) \Big|_{1}^{101}$$ how do you reached to this inequality? – Khosrotash Mar 05 '21 at 06:50
  • @JP McCarthy: why do you delete the post? – Khosrotash Mar 05 '21 at 06:55
  • @Khosrotash that is a lower Riemann sum of a monotone ascending function, I put a link to the Wikipedia article explaining it. Here is another one. – rtybase Mar 05 '21 at 07:54
  • Also here. In a very easy explanation, the sum in the answer is the sum of rectangles on the partition $1<2<3<...<k<k+1<...<101$, below the function's graphic. So, it's smaller than the integral of the function on $[1, 101]$ interval. – rtybase Mar 05 '21 at 07:59
  • @Khosrotash there was an error and when I fixed it the method did not work. – JP McCarthy Mar 05 '21 at 09:24
0

$$\ln (100!)=\sum_{n=1}^{100}\ln n<\sum_{n=1}^{100}\int_n^{n+1}\ln x\,dx=$$ $$=\int_1^{101}\ln x\,dx=$$ $$=[101\ln 101 - 101]-[1\ln 1 -1]=$$ $$=101\ln 101 - 100.$$ Therefore $$\ln (100!)-\ln (50^{100})<101\ln 101 -100-100\ln 50=$$ $$=(100)(1.01\ln 100 +1.01\ln 1.01-1-(\ln 100-\ln 2))=$$ $$=(100)(0.01\ln 100 +1.01\ln 1.01+\ln 2 -1)<0$$ because $\ln 2<0.70,$ and $\ln 100<6,$ and $1.01\ln 1.01<(1.01)(0.01)<0.02.$

  • We could omit the term $n=1$ in the 1st line (because $\ln 1=0$) and then the 4th line would be $101\ln 101-(99+2\ln 2)$ but this was unneeded. And in the last line it would suffice that $\ln 2<0.90.$ – DanielWainfleet Sep 01 '21 at 14:56