%! TEX root = AN.tex % vim: tw=50 % 17/02/2025 09AM \subsubsection*{The Brun-Titchmarsh Theorem} \begin{fcthm}[Brun-Titchmarsh Theorem] \label{thm:bt} Assuming: - $x \ge 0$, $y \ge 2$ - $\eps > 0$ and $y$ is large in terms of $\eps$ Then: \[ \pi(x + y) - \pi(x) \le \frac{(2 + \eps)y}{\log y} .\] \end{fcthm} \begin{remark*} We expect \[ \pi(x + y) - \pi(x) = \frac{(n + o(1))y}{\log x} \] in a wide range of $y$ (e.g. $y \ge x^\eps$ for some $\eps > 0$ fixed). The prime number theorem gives this for $y \vgg x$. The \nameref{thm:bt} gives an upper bound of the expected order for $y \ge x^\eps$. \end{remark*} \begin{proof} Apply the \nameref{thm:ss} with $A = [x, x + y] \cap \Nbb$, $\mathcal{P} = \Pbb$. Note that for any $d \ge 1$, \[ |\{a \in A : a \equiv 0 \pmod{d}\} = \frac{y}{d} + O(1) .\] Hence, the \gls{sievehyp} holds with $g(d) = \frac{1}{d}$, $R_d = O(1)$. Now, the function $h$ in \nameref{thm:ss} is given on primes by \[ h(p) = \frac{g(p)}{1 - g(p)} = \frac{1}{p - 1} = \frac{1}{\varphi(p)} ,\] where $\varphi$ is the Euler totient function. In general, \[ h(d) = \mob(d)^2 \cdot \frac{1}{\varphi(d)} \] (since $\varphi$ is multiplicative). Now, for any $z \ge 2$, \nameref{thm:ss} yields \[ S(A, \Pbb, z) \le \frac{y}{\sum_{d \le z} \frac{\mob(d)^2}{\varphi(d)}} + O \left( \sum_{d \le z^2} \tauf_3(d) \right) .\] By Problem 11 on \es{1}, the error term is $O(z^2 (\log z)^2)$. Take $z = y^{\half - \frac{\eps}{10}}$. Then \[ z^2(\log z)^2 \vll (y^{\half - \frac{\eps}{10}})^2 (\log y)^2 \vll y^{1 - \frac{\eps}{20}} \] (for $y \ge y_0(\eps)$). We estimate \begin{align*} \sum_{d \le z} \frac{\mob(d)^2}{\varphi(d)} &= \sum_{d \le z} \frac{\mob(d)^2}{d} \frac{d}{\varphi(d)} \\ &= \sum_{d \le z} \frac{\mob(d)^2}{d} \cdot \prod_{p \mid d} \ub{\left( 1 + \frac{1}{p} + \frac{1}{p^2} + \cdots \right)}_{= \frac{p}{p - 1} = \frac{p}{\varphi(p)}} \\ &\ge \sum_{n \le z} \frac{1}{n} \end{align*} since any $n \le z$ has at least one representation as \[ n = d p_1^{a_1} \cdots p_k^{a_k} ,\] where $d \le z$ is square-free and $p_i \mid d$ are primes and $a_1 \ge 0$. We have proved \[ \sum_{n \le z} \frac{1}{n} = \log z + O(1) \ge \left( 1 - \frac{1}{10} \right) \log z \] for $z \ge z_0(\eps)$. Putting everything together gives us \begin{align*} \pi(x + y) - \pi(x) &\le S(A, \Pbb, z) + z \\ &\le \frac{y}{\left( 1 - \frac{\eps}{10} \right) \log z} + z + y^{1 - \frac{\eps}{20}} \\ &\le \frac{(2 + \eps) y}{\log y} \end{align*} for $y \ge y_0(\eps)$. \end{proof} \newpage \section{The Riemann Zeta Function} Recall that the Riemann zeta function is defined by \[ \zeta(s) = \sum_{n = 1}^{\infty} \frac{1}{n^s} \] for $\Re(s) > 1$. Remarkable properties: \begin{enumerate}[(1)] \item $\zeta(s)$ extends meromorphically to $\Cbb$. \item A functional equation relating $\zeta(s) \leftrightarrow \zeta(1 - s)$. \item All the (non-trivial) zeroes appear to be on the line $\Re(s) = \half$ (Riemann hypothesis). \item $\zeta(s)$ closely relates to the distribution of primes. \end{enumerate} \begin{notation*} $f(x) \asymp g(x)$ means $f(x) \vll g(x) \vll f(x)$. $\asymp_\sigma$ means that the constant in $\vll$ can depend on $\sigma$. \end{notation*} \begin{fclemma}[] Assuming: - $\sigma > 1$ - $t \in \Rbb$ Then: $|\zeta(\sigma + it)| \asymp_\sigma 1$. \end{fclemma} \begin{proof} By the Euler product formula, \[ \zeta(\sigma + it) = \prod_p (1 - p^{-\sigma - it})^{-1} ,\] hence \[ |\zeta(\sigma + it)| = \prod_p |1 - p^{-\sigma - it}|^{-1} .\] By the triangle inequality, \[ 1 - p^{-\sigma} \le |1 - p^{-\sigma - it}| \le 1 + |p^{-\sigma - it}| = 1 + p^{-\sigma} .\] Hence, \[ \prod_p (1 - p^{-\sigma})^{-1} \ge |\zeta(\sigma + it)| \ge \prod_p (1 = p^{-\sigma})^{-1} .\] Note that these products converge if and only if $\sum_p p^{-\sigma}$ converges, and this sum converges by the comparison test. \end{proof} \begin{fclemma}[Polynomial growth of $\zeta$ in half-planes] \phantom{} \begin{cenum}[(i)] \item $f(\zeta)$ extends to a meromorphic function on $\Cbb$, with the only pole being a simple pole which is at $s = 1$. \item Let $k \ge 0$ be an integer. Then, for $\Re(s) \ge -k$ and $|s - 1| \ge \frac{1}{10}$ we have $|\zeta(s)| \vll_k |s|^{k + 2} + 1$. \end{cenum} \end{fclemma}