On twin prime distribution and associated biases

Abstract

A modified totient function ($\phi_2$) is seen to play a significant role in the study of the twin prime distribution. The function is defined as $\phi_2(n):=\#\{a\le n ~\vert ~\textrm{$a(a+2)$ is coprime to $n$}\}$ and is shown here to have following product form: $\phi_2(n) = n (1-\frac{\theta_n}{2}) \prod_{p>2,~p\vert n}(1-\frac 2 p)$, where $p$ denotes a prime and $\theta_n = 0$ or $1$ for odd or even $n$ respectively. Using this function it is proved for a given $n$ that there always exists a number $m > n$ so that $(p, m(m + 2)) = 1$ for every prime $p \le n$. We also establish a Legendre-type formula for the twin prime counting function in the following form: $\pi_2(x) - \pi_2(\sqrt{x}) = \sum_{ab\vert P(\sqrt{x})}\mu(ab) \left[\frac{x-l_{a,b}}{ab}\right]$, where $P(z)=\prod_{p\le z}p$ and $a$ is always odd. Here $l_{a,b}$ is the lowest positive integer so that $a\vert l_{a,b}$ and $b\vert (l_{a,b}+2)$. In the latter part of this work, we discussion three different types of biases in the distribution of twin primes. The first two biases are similar to the biases in primes as reported by Chebyshev, and Oliver and Soundararajan. Our third reported bias is on the difference ($D$) between (the first members of) two consecutive twin primes; it is observed that $D\pm1$ is more likely to be a prime than an odd composite number.