Is $\pi(N+2)\sim \sum_{p_{n+2}\in\Bbb P}^{N+2} p_n^{\frac{1}{\log(p_{n+2})}}?$

35 Views Asked by At

Is$$\pi(N+2)\sim \sum_{p_{n+2}\in\Bbb P}^{N+2} p_n^{\frac{1}{\log(p_{n+2})}}=3^{\frac{1}{\log(5)}}+5^{\frac{1}{\log(7)}}+11^{\frac{1}{\log(13)}}+17^{\frac{1}{\log(19)}}+...+N^{\frac{1}{\log(N+2)}}$$ where $N+2\in \Bbb P?$

When $N+2=181$

$\pi(181)=42$ and the summation is approximately equal to $33.7.$ The difference is about $8.$

What is the difference when $N+2=8011 ? $

I think the difference between the sum and the prime counting function will continue to increase slowly but they will be asymptotic.

1

There are 1 best solutions below

0
On BEST ANSWER

First of all, note that the individual terms you have are of the form $e^{\log(p_n)/\log(2+p_n)}$ (I'm writing it this way since it seems like you mean $p_n+2\in\mathbb{P}$, i.e., that you're interested in the twin primes, but you've written $p_{n+2}$; please clarify if I'm wrong!). Now, $\log(x+2)$ $=\log(x)+\log(1+2/x)$ $\approx\log(x)+2/x$, so $\log(x)/\log(x+2)\approx \log(x)/(\log(x)+2/x)$ $\approx 1-\frac2{x\log x}$. Similarly, near $x=1$, $e^x\approx ex$, so each of the individual terms is roughly $e\cdot(1-\frac2{x\log x})$. Since there are $\pi_2(N)$ terms in your sum, you should expect that the value of your sum is roughly $e\pi_2(N)$; best estimates are that this is $\approx e\frac{N}{\log^2(N)}$, whereas $\pi(N)\approx\frac{N}{\log N}$, so the two are off by a logarithmic factor.