| Download
William Stein -- Talk for Mathematics is a long conversation: a celebration of Barry Mazur
\documentclass{beamer}12\usepackage{fix-cm}3\usepackage{soul}4\usepackage{float}5\usepackage{tikz}67\usepackage{color}8\definecolor{dblackcolor}{rgb}{0.0,0.0,0.0}9\definecolor{dbluecolor}{rgb}{.01,.02,0.7}10\definecolor{dredcolor}{rgb}{0.8,0,0}11\definecolor{dgraycolor}{rgb}{0.30,0.3,0.30}12\usepackage{listings}13\lstdefinelanguage{Sage}[]{Python}14{morekeywords={True,False,sage,singular},15sensitive=true}16\lstset{frame=none,17showtabs=False,18showspaces=False,19showstringspaces=False,20commentstyle={\ttfamily\color{dredcolor}},21keywordstyle={\ttfamily\color{dbluecolor}\bfseries},22stringstyle ={\ttfamily\color{dgraycolor}\bfseries},23language = Sage,24basicstyle={\scriptsize \ttfamily},25aboveskip=.3em,26belowskip=.1em27}2829\usepackage{fancybox}30\usepackage{graphicx}31\usepackage{amsmath}32\usepackage{amsfonts}33\usepackage{amssymb}34\usepackage{amsthm}35\usepackage{url}363738\DeclareMathOperator{\Gap}{Gap}39\DeclareMathOperator{\Li}{Li}40\DeclareGraphicsRule{.tif}{png}{.png}{`convert #1 `dirname #1`/`basename #1 .tif`.png}4142\newcommand{\mycaption}[1]{\begin{quote}{\bf Figure: } \large #1\end{quote}}4344\newcommand{\ill}[3]{%45\begin{figure}[H]%46\vspace{-2ex}47\centering%48\includegraphics[width=#2\textwidth]{illustrations/#1}%49\caption{#3}%50\vspace{-2ex}51\end{figure}}5253\newcommand{\illtwo}[4]{%54\begin{figure}[H]\centering%55\includegraphics[width=#3\textwidth]{illustrations/#1}$\qquad$\includegraphics[width=#3\textwidth]{illustrations/#2}%56\caption{#4}%57\end{figure}}5859\newcommand{\illthree}[5]{%60\begin{figure}[H]%61\centering%62\includegraphics[width=#4\textwidth]{illustrations/#1}$\qquad$\includegraphics[width=#4\textwidth]{illustrations/#2}$\qquad$\includegraphics[width=#4\textwidth]{illustrations/#3}%63\caption{#5}%64\end{figure}}6566676869\def\GL{\mathrm{GL}}70\def\PGL{\mathrm{PGL}}71\def\PSL{\mathrm{PSL}}72\def\GSP{\mathrm{GSP}}73\def\Z{\mathrm{Z}}74\def\Q{\mathrm{Q}}75\def\Gal{\mathrm{Gal}}76\def\Hom{\mathrm{Hom}}77\def\Ind{\mathrm{Ind}}78\def\End{\mathrm{End}}79\def\Aut{\mathrm{Aut}}80\def\loc{\mathrm{loc}}81\def\glob{\mathrm{glob}}82\def\Kbar{{\bar K}}83\def\D{{\mathcal D}}84\def\L{{\mathcal L}}85\def\R{{\mathcal R}}86\def\G{{\mathcal G}}87\def\W{{\mathcal W}}88\def\H{{\mathcal H}}89\def\OH{{\mathcal OH}}90919293\newcommand{\RH}{Riemann Hypothesis\index{Riemann Hypothesis}}949596\title{PRIMES}97\author{Barry Mazur}98\date{\today}99100\begin{document}101102\begin{frame}103\titlepage104105{\it (A discussion of `Primes: What is Riemann's Hypothesis?,' the book I'm currently writing with William Stein)}106\end{frame}107108\begin{frame}109\frametitle{William:}110\begin{center}\LARGE111\url{https://vimeo.com/90380011}112\end{center}113\ill{steinbsair}{0.7}{William}114\end{frame}115\begin{frame}116117\frametitle{The impact of the Riemann Hypothesis}118\ill{sarnak}{0.20}{Peter Sarnak}119120\begin{quote}121``The Riemann hypothesis is the central problem and it implies many,122many things. One thing that makes it rather unusual in mathematics123today is that there must be over five hundred papers---somebody should124go and count---which start `Assume the Riemann hypothesis,' and125the conclusion is fantastic. And those [conclusions] would then become126theorems ... With this one solution you would have proven five hundred127theorems or more at once.''128\end{quote}129130\end{frame}131\begin{frame}132\frametitle{An expository challenge}133The approach you take when you try to explain anything depends upon your intended audience(s). In our case we wanted to reach two quite different kinds of readers (at the same time):134\vskip20pt135\begin{itemize} \item High School students who are already keen on mathematics,136\vskip20pt137\item A somewhat older crowd of scientists (e.g., engineers) who have a nonprofessional interest in mathematics.138\end{itemize}139\end{frame}140\begin{frame}\frametitle{\bf\large What {\em sort} of Hypothesis is the \RH{}?}141\begin{center}142\shadowbox{ \begin{minipage}{0.91\textwidth}143\mbox{} \vspace{0.2ex}144145Consider the seemingly innocuous series of questions:146147\begin{quote}148\begin{itemize}149\item How many primes (2, 3, 5, 7, 11, 13, $\ldots$) are there less than 100?150\item How many less than 10,000?151\item How many less than 1,000,000?152\end{itemize}153154More generally, how many primes are there less than any given number $X$?155\end{quote}156157Riemann's Hypothesis tells us that a strikingly158simple-to-describe function is a ``very good approximation'' to the number of159primes less than a given number $X$. We now160see that if we could prove this {\em Hypothesis of Riemann} we would have161the key to a wealth of powerful mathematics. Mathematicians are eager162to find that key.163164165\vspace{1ex}166\end{minipage}}\end{center}167\end{frame}168\begin{frame}\frametitle{An expository frame---and goal}\vskip10pt169\ill{raoulbott}{0.20}{Raoul Bott (1923--2005)\label{fig:bott}}170171Raoul Bott, once172said---giving advice to some young mathematicians---that whenever one173reads a mathematics book or article, or goes to a math lecture, one174should aim to come home with something very specific (it can be small,175but should be {\em specific}) that has application to a wider class of176mathematical problem than was the focus of the text or lecture.177178\end{frame}179180\begin{frame}\frametitle{Setting the frame}181If we182were to suggest some possible {\em specific} items to come home with,183after reading our book, three key phrases -- {\bf prime numbers}, {\bf184square-root accurate}, and {\bf spectrum} -- would head the185list.186187\end{frame}188189\begin{frame}\frametitle{PRIMES: order appearing random }190\vskip10pt191\ill{zagier}{.15}{Don Zagier}192193194\begin{quote}195196``{\bf [Primes]}197\begin{itemize}\item are the most arbitrary and ornery objects studied by mathematicians:198they grow like weeds among the natural numbers, seeming to obey no199other law than that of chance, and nobody can predict where the next200one will sprout. \item exhibit stunning201regularity $\dots$202they obey their laws with almost military precision.''\end{itemize}203\end{quote}204\end{frame}205206\begin{frame}\frametitle{How to nudge readers to feel the orneriness of primes }207208There is something compelling about `physically' hunting for a species of mathematical object, and collecting specimens of it. Our book emphasizes this approach for our readers. Here are some routes that allow you to 'pan' (in different ways) for primes:209210\vskip20pt211212213\centerline{ {\bf Factor trees} and {\bf Sieves}}214215216\vskip10pt217218219\centerline{and}220221222\vskip10pt223224225\centerline{\bf Euclid's Proof of the Infinitude of Primes.}226\end{frame}227228\begin{frame}\frametitle{\bf Factor trees}229230\illtwo{factor_tree_300_a}{factor_tree_300_b}{.47}231232233\end{frame}234235\begin{frame}\frametitle{\bf Sieves}236\includegraphics[width=\textwidth]{illustrations/circled_primes}237\end{frame}238239\begin{frame}\frametitle{\bf The ubiquity of primes}240\vskip10pt241\ill{dulcinea1}{.2}{Don Quixote and ``his'' Dulcinea del Toboso}242\vskip10pt243Numbers are obstreperous things. Don Quixote encountered this when he244requested that the ``bachelor'' compose a poem to his lady Dulcinea del245Toboso, the first letters of each line spelling out her name.246\end{frame}247248\begin{frame}\frametitle{\bf The stubbornness of primes and knights}249The250``bachelor'' found251\vskip10pt252253\begin{quote}254``a great difficulty in their composition because the number of255letters in her name was $17$, and if he made four Castilian stanzas256of four octosyllabic lines each, there would be one letter too many,257and if he made the stanzas of five octosyllabic lines each, the ones258called {\em d{\'e}cimas} or {\em redondillas,} there would be three259letters too few...''260\end{quote}261262``It must fit in, however, you do it,'' pleaded Quixote, not willing to263grant the imperviousness of the number $17$ to division.264\end{frame}265\begin{frame}\frametitle{\bf The Art of asking questions}266\vskip10pt267268{\Huge269270\centerline{Questions anyone might ask}271\vskip10pt272\centerline{\it spawning}273\vskip10pt274\centerline{Questions that shape the field}}275\end{frame}276277\begin{frame}\frametitle{\bf Gaps: an example of a `question anyone might ask'}278\vskip10pt279280\ill{zhang}{0.15}{Yitang Zhang\label{fig:zhang}}281282{\Huge283In celebration of Yitang Zhang's recent result, consider the {\em gaps} between one prime and the next.}284\end{frame}\begin{frame}\frametitle{\bf Twin Primes}285{\Huge286\begin{quote} As of 2014, the largest known twin primes are \vskip10pt287$$3756801695685\cdot 2^{666669} \pm 1$$ \vskip10pt288These enormous primes have $200700$ digits each.289\end{quote}}290\end{frame}291\begin{frame}\frametitle{\bf Gaps of width $k$}292Define293{\Huge $$294\Gap_{k}(X):=295$$296number of pairs of {\em consecutive} primes $(p,q)$ with297$q<X$ that have ``gap $k$'' (i.e., such that their difference $q-p$ is298$k$).299300\vskip20pt301302{\bf NOTE:} $\Gap_{4}(10)=0$.}\end{frame}303\begin{frame}\frametitle{\bf Gap statistics}304305306\begin{table}[H]\centering307\caption{Values of $\Gap_{k}(X)$ \label{tab:gap}}308\vspace{1em}309310{\small311\begin{tabular}{|l|c|c|c|c|c|c|}\hline312$X$ & $\Gap_{2}(X)$ & $\Gap_{4}(X)$& $\Gap_{6}(X)$ & $\Gap_{8}(X)$ &313$\Gap_{100}(X)$ & $\Gap_{252}(X)$\\\hline314315$10$ & 2 & 0 & 0 & 0 & 0 & 0\\\hline316$10^{2}$ & 8 & 7 & 7 & 1 & 0 & 0\\\hline317$10^{3}$ & 35 & 40 & 44 & 15 & 0 & 0\\\hline318$10^{4}$ & 205 & 202 & 299 & 101 & 0 & 0\\\hline319$10^{5}$ & 1224 & 1215 & 1940 & 773 & 0 & 0\\\hline320$10^{6}$ & 8169 & 8143 & 13549 & 5569 & 2 & 0\\\hline321$10^{7}$ & 58980 & 58621 & 99987 & 42352 & 36 & 0\\\hline322$10^{8}$ & 440312 & 440257 & 768752 & 334180 & 878 & 0\\\hline323324\end{tabular}325}326\end{table}327\end{frame}328329330\begin{frame}\frametitle{\bf How many primes are there?}331\vskip20pt332333{\Huge $\pi(X):=$ \# of primes $\le X$}334\vskip10pt335336\ill{prime_pi_25_aspect1}{.8}{Staircase of primes up to 25\label{fig:staircase25}}\end{frame}337\begin{frame}\frametitle{\bf How many primes are there?}338\vskip20pt339\ill{prime_pi_100_aspect1}{.8}{Staircase of primes up to 100\label{fig:staircase100a}}340341\end{frame}342343344\begin{frame}\frametitle{\bf Prime numbers viewed from a distance}345\vskip10pt346{\Huge{\it Pictures of data magically become smooth curves as you telescope to greater and greater ranges.}}347\vskip20pt348349350\illtwo{prime_pi_1000}{prime_pi_10000}{0.4}{Staircases of primes up to 1,000 and 10,000\label{fig:staircases2}}351\end{frame}352353354\begin{frame}\frametitle{\bf Proportion of Primes}\vskip10pt355\ill{proportion_primes_100}{1}{Graph of the proportion of primes up to $X$ for each integer $X\leq 100$}356\end{frame}357358359\begin{frame}\frametitle{\bf Proportion of Primes at greater distance}\vskip10pt360361\illtwo{proportion_primes_1000}{proportion_primes_10000}{0.46}{Proportion of primes for $X$ up to $1{,}000$ (left) and $10{,}000$ (right)}362\end{frame}363364365366367\begin{frame}\frametitle{\bf Gauss}\vskip10pt368369\ill{gauss_tables_half}{.9}{A Letter of Gauss\label{fig:gauss_letter}}\end{frame}370371\begin{frame}\frametitle{\bf Gauss' guess}\vskip10pt372373{\Huge The `probability' that a number $N$ is a prime is proportional to the reciprocal of its number of digits; more precisely the probability is374\vskip10pt $$1/\log(N).$$}\end{frame}375376\begin{frame}377378{\Huge This would lead us to this guess for the approximate value of $\pi(X)$: \vskip10pt379$$\Li(X):=\ \ \ \int_2^XdX/\log(X).$$}\end{frame}380381\begin{frame}\frametitle{\bf Approximating $\pi(X)$}\vskip10pt382383384385\ill{three_plots}{1.0}{Plots of $\Li(X)$ (top), $\pi(X)$ (in the middle), and $X/\log(X)$ (bottom).\label{fig:threeplots}}\end{frame}386387\begin{frame}\frametitle{\bf The Prime Number Theorem}\vskip10pt388\ill{three_plots}{0.5}{Plots of $\Li(X)$ (top), $\pi(X)$ (in the middle), and $X/\log(X)$ (bottom).\label{fig:threeplots}}\vskip10pt389390{\Huge All three graphs {\it tend to $\infty$} at the same rate.}\end{frame}391\begin{frame}\frametitle{\bf Ratios }392\vskip30pt393{\Huge \centerline{\bf PNT:}\vskip10pt394\centerline{ The ratios}395$${\frac{\pi(X)}{Li(X)}}\ \ \ {\rm and}\ \ \ {\frac{\pi(X)}{X/\log(X))}}$$\vskip10pt tend to $1$ as $X$ goes to $\infty$.}396\vskip10pt\end{frame}397\begin{frame}\frametitle{\bf Ratios versus Differences}398\vskip10pt399400{\Huge Much subtler question: what about their differences?401402$$|\Li(X)-\pi(X)|?$$}\end{frame}403404\begin{frame}\frametitle{\bf Riemann's405Hypothesis}406407{\Huge \begin{center}408\shadowbox{ \begin{minipage}{0.9\textwidth}409\mbox{} \vspace{0.2ex}410\begin{center}{\bf\large The {\bf \RH{}} (first formulation)}\end{center}411\medskip412413$\pi(X)$ is approximated by $\Li(X)$, with {\bf essentially square-root} accuracy.414415\vspace{1ex}416\end{minipage}}417\end{center}}418\end{frame}419\begin{frame}\frametitle{\bf More precisely $\dots$}420{\Huge {\bf RH} is equivalent to:421422\vskip10pt423$$|\Li(X) - \pi(X)| \leq \sqrt{X}\log(X)$$424425\vskip10pt426427for all $X\geq 2.01$.}\end{frame}428429430\begin{frame}\frametitle{\bf Square-root accuracy}431432{\Huge {\bf The gold standard for empirical data accuracy} \vskip10pt433434Discussion of random error, and random walks435436437\illtwo{random_walks-1000}{random_walks-1000-mean}{.45}{One Thousand Random Walks\label{fig:random_walks_1000}}}\end{frame}438\begin{frame}\frametitle{\bf {The mystery moves to the error term}}439440{\Huge $${\it Mysterious\ quantity}(X)\ \ = \ \ $$441442$$\ \ =\ \ {\it Simple \ expression}(X) \ + \ $$443444$$ \ + \ {\it Error}(X).$$}\end{frame}445\begin{frame}\frametitle{\bf {Our mystery moves to our error term}}446447{\Huge $${\rm Mystery}\ \ = \ {\rm Simple} \ + \ {\rm Error}.$$448449$${\pi}(X)\ \ = \ {Li}(X) \ - \ \big({Li}(X)-\pi(X)\big).$$}\end{frame}450451452453\begin{frame}\frametitle{\bf That `error term'}454\ill{li-minus-pi-250000}{.9}{$\Li(x)-\pi(x)$ (blue middle), its C{\'e}saro smoothing (red bottom), and455$\sqrt{\frac{2}{\pi}}\cdot \sqrt{x/\log(x)}$ (top), all for $x\leq 250{,}000$\label{fig:li-minus-pi-250000}}\end{frame}456\begin{frame}\frametitle{\bf The tension between data and long-range behavior}457458\ill{li-minus-pi-250000}{.3}459460The wiggly blue curve which seems to be growing nicely `like ${\sqrt X}$' will descend below the $X$-axis, for some value of $X > 10^{14}$.461\vskip20pt462\centerline{\bf Skewes Number}\end{frame}463\begin{frame}\frametitle{\bf The tension between data and long-range behavior}464465\hskip100pt \includegraphics[width=0.4\textwidth]{illustrations/littlewood}466{467$$46810^{14}\ \ \ \le\ \ \ {\rm Skewes\ Number}\ \ \ <\ \ \ 10^{317}469$$} \end{frame}470471\begin{frame}\frametitle{\bf Spectrum}472473\hskip100pt \includegraphics[width=0.13\textwidth]{illustrations/rainbow}474475\vskip10pt476{\Huge From Latin:\vskip10pt ``image," or ``appearance."}\end{frame}477\begin{frame}\frametitle{\bf Spectra and the Fourier transform}478479(The essential miracle of the theory of the Fourier transform:)480\vskip10pt481{\Huge $$ G(t) \ \ \ \leftrightarrow \ \ \ F(s)$$482483\vskip10pt484485Each behaves as if it were the \vskip5pt {\it 'spectral analysis'} of the other.}\end{frame}486\begin{frame}\frametitle{\bf packaging the information given by prime powers}487{\Huge $$ g(t)\ \ =\ \ $$488489\vskip10pt490491$$\ \ \ = \ \ \ -\sum_{p^n}{\frac{\log(p)}{p^{n/2}}}\cos(t\log(p^n).)492$$}\end{frame}493\begin{frame}\frametitle{\bf $p^n \leq 5$}494{\Huge495\ill{phihat_even-5}{1}{Plot of $-\sum_{p^n\leq 5}{\frac{\log(p)}{p^{n/2}}}\cos(t\log(p^n))$ with496arrows pointing to the spectrum of the primes\label{fig:pnsum5}}497}\end{frame}498\begin{frame}\frametitle{\bf $p^n \leq 20$}499{\Huge500\ill{phihat_even-20}{1}{Plot of $-\sum_{p^n\leq 20}{\frac{\log(p)}{p^{n/2}}}\cos(t\log(p^n))$ with arrows pointing to the spectrum of the primes}}\end{frame}501\begin{frame}\frametitle{\bf $p^n \leq 50$}502{\Huge503504\ill{phihat_even-50}{1}{Plot of $-\sum_{p^n\leq 50}{\frac{\log(p)}{p^{n/2}}}\cos(t\log(p^n))$ with arrows pointing to the spectrum of the primes}}\end{frame}505506507\begin{frame}\frametitle{\bf $p^n \leq 500$}508{\Huge \ill{phihat_even-500}{1}{Plot of $-\sum_{p^n\leq509500}{\frac{\log(p)}{p^{n/2}}}\cos(t\log(p^n))$ with arrows510pointing to the spectrum of the primes\label{fig:pnsum500}}}\end{frame}511512\begin{frame}\frametitle{\bf From primes to the Riemann Spectrum}513{\Huge Conditional on RH, $g(t)$ converges to a distribution with singular spikes at the red vertical lines: the Riemann spectrum,514515$$\theta_1, \theta_2, \theta_3,\dots $$} \end{frame}516\begin{frame}\frametitle{\bf From the Riemann Spectrum to primes}517{\Huge518519$$f(s)\ \ = $$520521$$\ \ = \ \ 1+ \sum_{i}\cos(\theta_i\cdot \log(s))).$$} \end{frame}522\begin{frame}\frametitle{\bf From the Riemann Spectrum to primes}523{\Huge524525526527\ill{phi_cos_sum_2_30_1000}{.8}{Illustration of $-\sum_{i=1}^{1000}528\cos(\log(s)\theta_i)$, where $\theta_1 \sim 14.13, \ldots$ are the529first $1000$ contributions to the Riemann spectrum. The spikes530are at the prime powers $p^n$, whose size is proportional to531$\log(p)$.}} \end{frame}532\begin{frame}\frametitle{\bf From the Riemann Spectrum to primes}533{\Huge534535\ill{phi_cos_sum_26_34_1000}{.8}{Illustration of $-\sum_{i=1}^{1000}536\cos(\log(s)\theta_i)$ in the neighborhood of a twin prime. Notice537how the two primes $29$ and $31$ are separated out by the Fourier538series, and how the prime powers $3^3$ and $2^5$ also appear.}} \end{frame}539\begin{frame}\frametitle{\bf From the Riemann Spectrum to primes}540{\Huge541542\ill{phi_cos_sum_1010_1026_15000}{.7}{Fourier series from $1,000$ to543$1,030$ using 15,000 of the numbers $\theta_i$. Note the twin544primes $1019$ and $1021$ and that $1024=2^{10}$.}} \end{frame}545\begin{frame}\frametitle{\bf Information and Structure}546547{\Huge The Riemann spectrum holds the key to the position of prime numbers on the number line.548\vspace{1em}549550What even deeper structure of primes can they reveal to us?}\end{frame}551\begin{frame}552\frametitle{Riemann}553\ill{riemann}{.2}{Bernhard Riemann (1826--1866)}554555\ill{riemann_zoom}{1}{From Riemann's 1859 Manuscript\label{fig:riemamn}}556\end{frame}557\begin{frame}558\frametitle{William}559\begin{center}\LARGE560\url{https://vimeo.com/90380011}561\end{center}562\ill{steinbsair}{0.7}{William}563\end{frame}564\end{document}565566567568%\ill{riemann}{.2}{Bernhard Riemann (1826--1866)}569570571\centerline{1853: His theory of trigonometric sums}572\centerline{1859: His number theory}573574\ill{riemann_zoom}{1}{From Riemann's 1859 Manuscript\label{fig:riemamn}}575576577578There are 455,052,512 primes less than ten billion; i.e.,57910,000,000,000 (so we might say that the chances are down to roughly580$1$ in $22$).581582Primes, then, seem to be thinning out. We return to the sifting process583we carried out earlier, and take a look at a few graphs, to get a sense of why584that might be so. There are a $100$ numbers less than or equal to585$100$, a thousand numbers less than or equal to $1000$, etc.: the586shaded graph in Figure~\ref{fig:sieve_2_100} that looks like a regular staircase, each step the587same length as each riser, climbing up at, so to speak, a 45 degree588angle, counts all numbers up to and including~$X$.589590Fol591592593594595596597\end{document}598The recent results of Zhang as sharpened by Maynard (and others) we mentioned above tell us that for at least one even number $k$ among the even numbers $k \le 252$, $\Gap_{k}(X)$ goes to infinity as $X$ goes to infinity. One expects that this happens for {\it all} even numbers $k$. We expect this as well, of course, for $\Gap_{252}(X)$ despite what might be misconstrued as discouragement by the above data.599600\ill{primegapdist}{1}{Frequency histogram showing the distribution of601prime gaps of size $\leq 50$ for all primes up to $10^7$. Six is602the most popular gap in this data. The vertical axis labels603such as ``6e4'' mean $6\cdot 10^4=60{,}000$.604\label{fig:primegapdist}}605606607\ill{primegap_race}{1}{Plots of $\Gap_k(X)$ for $k=2,4,6,8$. Which wins?}608609Here is yet another question that deals with the spacing of prime610numbers that we do not know the answer to:611612{\em Racing Gap $2$, Gap $4$, Gap $6$, and Gap $8$ against each other:}613614\begin{quote}615Challenge: As $X$ tends to infinity which of $\Gap_2(X)$,616$\Gap_4(X)$,617$\Gap_6(X),$ or $\Gap_8(X)$ do you think will grow faster? How618much would you bet on the truth of your guess? \bibnote{%619Hardy and Littlewood give a nice conjectural answer to such620questions about gaps between primes. See Problem {\bf A8} of621Guy's book {\em Unsolved Problems in Number Theory} (2004).622Note that Guy's book discusses counting the number $P_k(X)$ of pairs of623primes up to $X$ that differ by a fixed even number $k$; we have624$P_k(X)\geq \Gap_k(X)$, since for $P_k(X)$ there is no requirement625that the pairs of primes be consecutive.}626\end{quote}627628629\end{document}630%sagemathcloud={"zoom_width":95}631632