1
$\begingroup$

I want to prove this statement:

$\alpha \geq1$ and $n\geq 2\alpha \log(2\alpha)$ then $n \geq \alpha \log(2n)$

It seemed pretty simple but I couldn't do too much. My first attempt was

$$ \log(n) \geq \log ( 2\alpha \log(2\alpha) ) $$

But this attempt didn't look promising. Could you give me a hint?

$\endgroup$
2
  • $\begingroup$ I don't see $a$ anywhere in the final result or in the other condition. So how are we expected to use $a \ge 1$? $\endgroup$
    – Haris
    Commented Mar 10 at 9:16
  • 1
    $\begingroup$ @Haris Yeah, it was a typo ($\alpha$ ) $\endgroup$
    – xenuti
    Commented Mar 10 at 9:19

2 Answers 2

2
$\begingroup$

If ${a}\geq 1$, then we know that $\log(a)\geq0$ $\Rightarrow n\geq 0$. Substituting $ p = \alpha\log(2\alpha)$. We can clearly see the following: $$n\geq 2p$$ And because the numbers are positive then $$n\geq p$$ EDIT: Considering the change of what we have to prove: Here is the updated variant. If $n\geq 2\alpha\log(2\alpha)$, and all of the terms are positive, we can say that $n = 2\alpha\log(2\alpha) + \psi_1$ Where $\psi_1$ is some number greater than $0$. Plugging this into the second inequality we get what we have to prove: $$2\alpha\log(2\alpha) + \psi_1 = \alpha\log(4\alpha\log(2\alpha) + 2\psi_1) + \psi_2 $$ $${(2\alpha)}^{2\alpha}e^{\psi_1}=(4\alpha \log(2\alpha)+2\psi_1)^{a}e^{\psi_2}$$ $$(2\alpha)^{2\alpha}=((4\alpha\log(2\alpha)+2\psi_1)^{\frac{1}{2}})^{2a}e^{\psi_2-\psi_1}$$ $$e^{\frac{\psi_1-\psi_2}{2a}}=\left(\frac{\sqrt{(4\alpha\log(2\alpha)+2\psi_1)}}{2\alpha}\right) $$ Let $\psi_3=e^{\frac{\psi_1-\psi_2}{2a}}$, and we know that $\psi_3 \gt 0$, so we only have to prove: $$\left(\frac{\sqrt{(4\alpha\log(2\alpha)+2\psi_1)}}{2\alpha}\right) \gt 0$$ And since all of the terms are positive (including $\psi_1$), this will never equal or be less than $0$.

$\endgroup$
5
  • $\begingroup$ I´m sorry. I made a mistake. I have to prove that $n \geq \alpha log(2n)$ $\endgroup$
    – xenuti
    Commented Mar 10 at 9:21
  • $\begingroup$ @Masd I have a question. I think that $\left(\frac{\sqrt{(4\alpha\log(2\alpha)+2\psi_1)}}{2\alpha}\right) > 0$ and $e^{\frac{\psi_1-\psi_2}{2a}} > 0$ does not imply $\psi_2 \ge 0$. You need to prove that $\psi_2 < 0$ and $e^{\frac{\psi_1-\psi_2}{2a}}=\left(\frac{\sqrt{(4\alpha\log(2\alpha)+2\psi_1)}}{2\alpha}\right) > 0$ is impossible. In other words, you need to prove that $$e^{\frac{\psi_1}{2a}}\ge \left(\frac{\sqrt{(4\alpha\log(2\alpha)+2\psi_1)}}{2\alpha}\right).$$ $\endgroup$
    – River Li
    Commented Mar 11 at 0:37
  • $\begingroup$ By definition $\psi_2$ is bigger than 0 $\endgroup$
    – Masd
    Commented Mar 11 at 13:54
  • $\begingroup$ @Masd I think that you have to prove that there always exists $\psi_2\ge 0$ such that $2\alpha\log(2\alpha) + \psi_1 = \alpha\log(4\alpha\log(2\alpha) + 2\psi_1) + \psi_2$; in other words, $2\alpha\log(2\alpha) + \psi_1 = \alpha\log(4\alpha\log(2\alpha) + 2\psi_1) + \psi_2$ implies $\psi_2 \ge 0$ (this is what you want to prove rather than definition). $\endgroup$
    – River Li
    Commented Mar 11 at 23:24
  • $\begingroup$ I was just rephrasing the statement that what we have to prove, is all $\endgroup$
    – Masd
    Commented Mar 11 at 23:36
1
$\begingroup$

Assume, for the sake of contradiction, that there exists $(\alpha, n)$ such that $\alpha \ge 1$ and $n \ge 2\alpha \ln (2\alpha)$ and $n < \alpha \ln (2n)$.

We have $\alpha \ln(2n) > n \ge 2\alpha \ln(2\alpha)$ which results in $n \ge 2\alpha^2$. Thus, we have $n < \alpha \ln(2n) \le \sqrt{n/2}\ln(2n)$ which results in $\sqrt{2n} < \ln(2n)$, or $\mathrm{e}^{\sqrt{2n}} < 2n$. However, using $\mathrm{e}^x \ge 1 + x + \frac12 x^2 + \frac16 x^3$ for all $x > 0$, we have $$\mathrm{e}^{\sqrt{2n}} \ge 1 + \sqrt{2n} + \frac12 \sqrt{2n}^2 + \frac16 \sqrt{2n}^3 > 2n.$$ Contradiction.

$\endgroup$
2
  • $\begingroup$ I don't understand the last inequality. Is it due to $(\frac{1}{2})\sqrt{2n}^{2} = n$? $\endgroup$
    – xenuti
    Commented Mar 11 at 5:23
  • $\begingroup$ @xenuti By AM-GM, we have $\sqrt{2n} + \frac16 \sqrt{2n}^3 \ge \sqrt{2n} + \frac{1}{16} \sqrt{2n}^3 \ge 2\sqrt{ \sqrt{2n} \cdot \frac{1}{16} \sqrt{2n}^3 } = n$. Also, $\frac12 \sqrt{2n}^2 = n$. $\endgroup$
    – River Li
    Commented Mar 11 at 6:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .