%! TEX root = NT.tex % vim: tw=50 % 06/11/2023 10AM \begin{proof}[Proof of (2)] \begin{align*} \prod_{p \le x} \left( 1 - \frac{1}{p} \right)^{-1} &= \prod_{p \le x} \left( 1 + \frac{1}{p} + \frac{1}{p^2} + \cdots \right) \\ &= \sum_{k_1, \ldots, k_r \ge 0} (p_1^{k_1} \cdots p_r^{k_r})^{-1} \end{align*} Every integer $1 \le n \le x$ is a product of \glspl{prime} $\le x$, so \[ \prod_{p \le x} \left( 1 - \frac{1}{p} \right)^{-1} \ge \sum_{1 \le n \le x} \frac{1}{n} \to \infty \] as $x \to \infty$ (harmonic series). \end{proof} \begin{flashcard}[riemann-zeta-func-defn] \begin{definition}[Riemann $\zeta$] \glssymboldefn{zeta_fn}{$\zeta$}{$\zeta$} \cloze{ The \emph{Riemann $\zeta$-function} is \[ \zeta(s) = \sum_{n = 1}^\infty n^{-s} .\] Convention: $s \in \CC$. This defines $\zeta(s)$ whenever this series converges. } \end{definition} \end{flashcard} \vspace{-1em} $\zetafn(s)$ studied by Euler for $s \in \RR$ by Riemann for $s \in \CC$ $\to$ complex analysis. \begin{proposition} If $s \in \CC$, $\Re(s) > 1$, then $\zetafn(s)$ converges absolutely. \end{proposition} \begin{notation*} \glssymboldefn{sigmareal}{$\sigma$}{$\sigma$} Notation: $s = \sigmareal + it$, $\sigmareal + it \in \RR$. \end{notation*} \begin{proof} \[ n^{-s} = \exp(-s\log n) = \exp(-(\sigmareal + it)\log n) \] \[ \implies |n^{-s}| = \exp(-\sigmareal \log n) = n^{-\sigmareal} \] So $\sum_{n = 1}^\infty |n^{-s}| = \sum_{n = 1}^\infty n^{-\sigmareal}$. This converges if and only if $\sigmareal > 1$. \end{proof} Same arugument shows that $\zetafn(s)$ converges uniformly in $\{s \in \CC \st \sigmareal > 1 + \delta\}$, for any $\delta > 0$. A uniform limit of holomorphic functions is holomorphic, so $\zetafn(s)$ is holomorphic in $\{s \in \CC \st \sigmareal > 1\}$. \begin{flashcard}[zeta-prime-product-thm] \begin{theorem} If $s \in \CC$, $\sigmareal > 1$, then \[ \zetafn(s) = \cloze{\prod_{\text{$p$ \gls{prime}}} (1 - p^{-s})^{-1}} .\] More precisely \[ \cloze{\lim_{x \to \infty} \prod_{p \le x} (1 - p^{-s})^{-1}} = \zetafn(s) \] and this limit is non-zero. \end{theorem} \end{flashcard} \begin{proof} Arguing informally, we have \[ \prod_{p} (1 - p^{-s})^{-1} = \prod_p (1 + p^{-s} + p^{-2s} + p^{-3s} + \cdots) = \sum_{n = 1}^\infty n^{-s} .\] By \nameref{FTA}. Arguing rigorously, \[ \prod_{p \le x} (1 - p^{-s})^{-1} = \sum_{k_1, \ldots, k_r} \ge 0 (p_1^{k_1} \cdots p_r^{k_r})^{-s} \] where $p_1, \ldots, p_r$ are the \glspl{prime} $\le x$. \nameref{FTA} implies if $n \in \NN$, then $n^{-s}$ apprears at most once in $\sum_{k_1, \ldots, k_r \ge 0} (p_1^{k_1} \cdots p_r^{k_r})^{-s}$, and exactly once if $n \le x$. So \[ \left|\prod_{p \le x} (1 - p^{-s})^{-1} - \zetafn(s)\right| \le \sum_{n > x} n^{-\sigmareal} \to 0 \] as $x \to \infty$. So \[ \lim_{x \to \infty} p \le x (1 - p^{-s})^{-1} = \zetafn(s) .\] To show $\zetafn(s) \neq 0$, consider \[ \prod_{p \le x} (1 - p^{-s}) \zetafn(s) = \prod_{p > x}(1 - p^{-s})^{-1} = 1 + \sum_{n \in S_x} n^{-s} \] where \[ S_x = \{n \in \NN \st \text{all prime factors $p \divides n$ satisfy $p > x$}\} \subset \{n \in \NN \st n > x\} .\] Then \[ \left|\prod_{p \le x} (1 - p^{-s}) \zetafn(s)\right| \ge 1 - \sum_{n > x} n^{-\sigmareal} .\] Since $\sigmareal > 1$, $\sum_{n > x} n^{-\sigmareal} \to 0$ as $X \to \infty$, so we can choose $x$ such that \[ 1 - \sum_{n > x} n^{-\sigmareal} > 0. \] Then we deduce that \[ \left| \prod_{p \le x} (1 - p^{-s}) \zetafn(s) \right| \neq 0 \implies \zetafn(s) \neq 0. \qedhere \] \end{proof} \subsubsection*{Non-examinable discussion of $\zeta(s)$} \begin{itemize} \item Meromorphic continuation: $\zetafn(s)$ admits a unique function on $\CC$, with a simple polt at $s = 1$, and no other poles. \item \glssymboldefn{Gamma_fn}{$\Gamma$}{$\Gamma$} \glssymboldefn{xi_fn}{$\xi$}{$\xi$} Functional equation: we define $\xi(s) = \pi^{-s/2} \Gamma(s/2) \zeta(s)$, where $\Gamma(s)$ is the \emph{Gamma function}, a meromorphic function in $\CC$ defined for $\sigmareal > 0$ by the integral \[ \Gamma(s) = \int_{y = 0}^\infty e^{-y} y^s \frac{\dd y}{y} .\] Then $\xi(s) = \xi(1 - s)$. \item Trivial zeroes: $\xifn(s)$ is meromorphic with simple poles at $s = 0$, $s = 1$ and no other poles. $\Gammafn(s)$ has simple poles at $s = 0, -1, -2, \ldots$ and no other poles. $\Gammafn(s/2)$ has simple poles at $s = 0, -2, -4, \ldots$. Since $\xifn$ is holomorphic at $s = -2, -4, -6, \ldots$ but $\Gammafn(s/2)$ has a pole, $\Gammafn(s)$ must vanish, whenever $s$ is a negative even integer (these are the trivial zeroes). Picture of $\zetafn(s)$: \begin{center} \includegraphics[width=0.6\linewidth]{images/604e3340e5f24ed5.png} \end{center} \item Critical strip: this is the region $\{s \in \CC \st \sigmareal \in [0, 1]\}$. All non-trivial zeroes of $\zetafn(s)$ lie in the critical strip. Fact: their location is closely related to the distribution of \glspl{prime}. For example, the ``hard part'' in the proof of \nameref{pnt} (\cref{pnt}) is the non-existence of zeroes of $\zetafn(s)$ with $\sigmareal = 1$. \begin{conjecture}[Riemann Hypothesis] \vspace{0.5em} If $s \in \CC$ is a non-trivial zero of $\zetafn(s)$, then $\sigmareal = \half$. \end{conjecture} \vspace{-1em} As stated in the first lecture, this is equivalent to the bound \[ |\pc(x) - \li(x)| \le \sqrt{x} \log x \] for any $x \ge 3$. Recall \nameref{pnt} says \[ \left| \frac{\pc(x)}{\li(x)} - 1 \right| \to 0 \] as $x \to \infty$. \end{itemize} \textbf{This is now the end of the non-examinable content.} \begin{flashcard}[dirichlet-series-defn] \begin{definition}[Dirichlet series] \glsnoundefn{d_series}{Dirichlet series}{N/A} \cloze{A Dirichlet series is one of the form \[ \sum_{n = 1}^\infty a_n n^{-s} \qquad a_n \in \CC \]} \end{definition} \end{flashcard} \vspace{-1em} \begin{example*} If $a_n = 1 ~\forall n \in \NN$, this is just $\zetafn(s)$. If $N \in \NN$ is odd, then the \gls{d_series} \[ \sum_{n = 1}^\infty \legendre{n}{N} n^{-s} \] plays a rule in the proof of \cref{pnt_modulo} analogous to the role of $\zetafn(s)$ in the proof of \cref{pnt}. \end{example*} \begin{remark*} If $A, B > 0$ and $|a_n| \le An^B$ for all $n \ge 1$, then $\sum_{n = 1}^\infty a_n n^{-s}$ converges absolutely whe $\sigma > 1 + B$. \end{remark*}