5

Assuming PNT $$\pi(x)\sim \frac{x}{\log{x}}$$ How can we show that given any $\epsilon>0$ $$p_n^{1-\epsilon}< n,$$ for all sufficiently large $n$ ($p_n$ denotes the $n^{th}$ prime.)

My work: Setting $x=p_n$ we get $\lim\limits_{n\to\infty}\dfrac{n\log{p_n}}{p_n}=1\Rightarrow 1-\epsilon<\dfrac{n\log{p_n}}{p_n}$, for sufficiently large $n$. Now if I can show that $$\dfrac{n\log{p_n}}{p_n}\le\dfrac{\log n}{\log{p_n}}$$ then the result follows. But I am unable to show the last inequality.

Can someone please help with this? Other approaches are welcome as well. [NOTE: Using this I want to show that $\lim\limits_{n\to\infty}\frac{\log{n}}{\log{p_n}}=1$ so please don't use that, although it can be derived independently]

Thank you

rtybase
  • 16,907
  • 2
    To be more precise, $\implies 1-\epsilon<\frac{n\log p_n}{p_n}$ fot sufficiently large $n.$ – Thomas Andrews Jun 24 '18 at 04:50
  • 1
    It works for $\varepsilon \geq \frac{1}{2}$ https://math.stackexchange.com/questions/206815/is-there-a-way-to-show-that-sqrtp-n-n – rtybase Jun 24 '18 at 08:43

1 Answers1

2

$\color{red}{\text{1. One way}}$

I will use 2 results

  • PNT $$\lim\limits_{n\rightarrow\infty}\frac{\pi(n)\ln{(n)}}{n}=1 \Rightarrow \lim\limits_{n\rightarrow\infty}\frac{n}{\pi(n)\ln{(n)}}=1 \tag{1}$$
  • and $$\lim\limits_{n\rightarrow\infty}\frac{p_n}{n\ln{(n)}}=1 \tag{2}$$

Proposition 1.1 $$\lim\limits_{n\rightarrow\infty} \frac{\ln{(n)}}{\ln{(p_n)}}=1$$

$\{p_n\}$ is a subsequence of $\{n\}$, thus, from $(1)$, $$\lim\limits_{p_n\rightarrow\infty}\frac{\pi(p_n)\ln{(p_n)}}{p_n}=1 \Rightarrow \lim\limits_{n\rightarrow\infty}\frac{\pi(p_n)\ln{(p_n)}}{p_n}=1 \Rightarrow ...$$ because $\pi(p_n)=n$ $$...\lim\limits_{n\rightarrow\infty}\frac{n\ln{(p_n)}}{p_n}=1 \tag{3}$$ Now $$\lim\limits_{n\rightarrow\infty} \frac{\ln{(n)}}{\ln{(p_n)}}= \lim\limits_{n\rightarrow\infty} \left(\frac{n\ln{(n)}}{p_n}\cdot\frac{p_n}{n\ln{(p_n)}}\right)=\\ \lim\limits_{n\rightarrow\infty} \left(\frac{n\ln{(n)}}{p_n}\right)\cdot \lim\limits_{n\rightarrow\infty} \left(\frac{p_n}{n\ln{(p_n)}}\right)\overset{(2)(3)}{=}1$$

Proposition 1.2 For large enough $n$ $$p_n^{1-\varepsilon}<n$$

From $$\lim\limits_{n\rightarrow\infty} \frac{\ln{(n)}}{\ln{(p_n)}}=1$$ using the definition of limit, $\forall\varepsilon >0, \exists N(\varepsilon)\in\mathbb{N}$ s.t. $\forall n> N(\varepsilon)$ $$\left|\frac{\ln{(n)}}{\ln{(p_n)}}-1\right|<\varepsilon \Rightarrow 1-\varepsilon <\frac{\ln{(n)}}{\ln{(p_n)}}< 1+\varepsilon \Rightarrow \\ (1-\varepsilon)\ln{(p_n)} <\ln{(n)}< (1+\varepsilon)\ln{(p_n)} \Rightarrow \\ \ln{(p_n)^{(1-\varepsilon)}} <\ln{(n)}< \ln{(p_n)^{(1+\varepsilon)}} \Rightarrow ...$$ $e^x$ is ascendng, thus $$... p_n^{1-\varepsilon} <n< p_n^{1+\varepsilon} $$


$\color{red}{\text{2. Another way}}$

Using Vallée-Poussin, for large enough $x$ $$\pi(x)>\frac{x}{\ln(x)-(1-\varepsilon)}>\frac{x}{\ln(x)}$$ Let's show that for large enough $x$ we also have $$\frac{x}{\ln(x)}>x^{1-\varepsilon}$$ which is the same as showing $$\frac{x^{\varepsilon}}{\ln{x}}>1$$ for large $x>0$.

Propositions 2.1 Function $f(x)=\frac{x^{\varepsilon}}{\ln{x}}$ is ascending for large $x>0$.

Because $$f'(x)=\frac{x^{\varepsilon-1} (\varepsilon \ln{x}-1)}{\ln^2{x}}>0 \iff \varepsilon \ln{x}-1>0 \Rightarrow x> e^{\frac{1}{\varepsilon}}$$

Propositions 2.2 $\lim\limits_{x\rightarrow \infty}f(x) \rightarrow \infty$

If we assume it is bounded by a large $\alpha>0, \forall x>1$ and we know $\ln{x}$ is ascending $$\frac{x^{\varepsilon}}{\ln{x}} < \alpha \iff 1<x^{\varepsilon}< \alpha \ln{x} \iff \color{red}{0}<\varepsilon<\frac{\ln{\alpha}}{\ln{x}}+\frac{\ln{\ln{x}}}{\ln{x}}\rightarrow \color{red}{0}, x\rightarrow\infty$$ which is a contradiction.

So, for large $x$ we have $$\pi(x)>x^{1-\varepsilon}$$ which means, for large $n$ we have $$n=\pi(p_n)>p_n^{1-\varepsilon}$$

rtybase
  • 16,907
  • Well I already mentioned in the note that I am trying to derive $\log n\sim \log{p_n}$ using my question (and later on to show that $p_n\sim n\log n$). –  Jun 24 '18 at 11:39
  • You also mentioned "Other approaches are welcome as well.". You can also use Vallée Poussin https://en.wikipedia.org/wiki/Prime_number_theorem#Non-asymptotic_bounds_on_the_prime-counting_function – rtybase Jun 24 '18 at 11:57
  • Well of course the effort is appreciated, thanks for help. –  Jun 24 '18 at 12:06
  • 1
    @Vulthuryol is it better now? – rtybase Jun 24 '18 at 13:10
  • how does the PNT implies vallee-poussin's inequality? Can you link a proof? –  Jun 25 '18 at 03:06
  • It's vice versa in fact. Vallée-Poussin was among the first to prove PNT. – rtybase Jun 25 '18 at 09:28
  • From ${\frac {x}{\log x-(1-\varepsilon )}}<\pi(x)<{\frac {x}{\log x-(1+\varepsilon )}}$ it's easy to see ${\frac {1}{1-\frac{(1-\varepsilon )}{\log x}}}<\frac{\pi(x)\log x}{x}<{\frac {1}{1-\frac{(1+\varepsilon )}{\log x}}}$ – rtybase Jun 25 '18 at 09:35
  • I see how this implies PNT, but I am asking about the other way around. I mean my question is to assume PNT then derive this inequality. –  Jun 25 '18 at 09:58
  • PNT is not the best way to approximate $\pi(x)$, I don't know if the other way around is possible at all, using just PNT. – rtybase Jun 25 '18 at 10:08