Complete flattened LaTeX source
\documentclass[11pt,openany]{book}
\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage[sc,osf]{mathpazo}
\linespread{1.05}
\usepackage[scaled=0.9]{helvet}
\usepackage[paperwidth=7in,paperheight=10in,inner=0.95in,outer=0.85in,top=0.95in,bottom=1.0in,headsep=14pt]{geometry}
\usepackage{amsmath,amssymb,amsthm,mathtools,stmaryrd}
\usepackage{microtype}
\usepackage{xcolor}
\definecolor{accent}{HTML}{7A1F2B}
\definecolor{ink2}{HTML}{4A5560}
\definecolor{fwd}{HTML}{1F4E79}
\definecolor{bwd}{HTML}{A8432C}
\usepackage{tikz,tikz-cd}
\usetikzlibrary{arrows.meta,positioning,calc,decorations.markings,shapes.geometric,backgrounds,fit}
\usepackage{booktabs,array,tabularx,longtable}
\usepackage{enumitem}
\setlist{itemsep=2pt,topsep=4pt}
\usepackage{titlesec}
\usepackage{fancyhdr}
\usepackage[font=small,labelfont={bf,sf},margin=1.5em]{caption}
\usepackage[colorlinks=true,linkcolor=accent,citecolor=accent,urlcolor=accent,
pdftitle={Partial Differential Equations in Higher Categories},
pdfauthor={Anonymous},
pdfsubject={A higher-categorical treatment of introductory partial differential equations: waves, diffusions and potentials as functors of domain, symmetry and time},
pdfkeywords={partial differential equations, wave equation, diffusion equation, Laplace equation, boundary conditions, method of images, eigenfunctions, semigroups, discretization, gauge fields, category theory}]{hyperref}
\usepackage[nameinlink,capitalise,noabbrev]{cleveref}
\titleformat{\chapter}[display]{\normalfont\Large\scshape\color{accent}}{\chaptertitlename\ \thechapter}{4pt}{\normalfont\huge\bfseries\color{black}}
\titlespacing*{\chapter}{0pt}{10pt}{26pt}
\titleformat{\section}{\normalfont\large\bfseries}{\thesection}{0.7em}{}
\titleformat{\subsection}{\normalfont\normalsize\bfseries\itshape}{\thesubsection}{0.7em}{}
\titleformat{\part}[display]{\normalfont\centering\Large\scshape\color{accent}}{\partname\ \thepart}{12pt}{\normalfont\Huge\bfseries\color{black}}
\pagestyle{fancy}\fancyhf{}
\fancyhead[LE]{\small\thepage\quad\textsc{\nouppercase{\leftmark}}}
\fancyhead[RO]{\small\textsc{\nouppercase{\rightmark}}\quad\thepage}
\renewcommand{\headrulewidth}{0pt}
\renewcommand{\chaptermark}[1]{\markboth{#1}{#1}}
\renewcommand{\sectionmark}[1]{\markright{\thesection\ \ #1}}
\fancypagestyle{plain}{\fancyhf{}\fancyfoot[C]{\small\thepage}}
\setlength{\headheight}{14pt}
\setcounter{tocdepth}{1}
\setcounter{secnumdepth}{2}
\raggedbottom
\emergencystretch=2em
\tolerance=1200
\hbadness=2000
\theoremstyle{plain}
\newtheorem{theorem}{Theorem}[chapter]
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{principle}[theorem]{Principle}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\newtheorem{construction}[theorem]{Construction}
\newtheorem{problem}[theorem]{Open Problem}
\newtheorem{exercise}{Exercise}[chapter]
\theoremstyle{remark}
\newtheorem{remark}[theorem]{Remark}
\crefname{principle}{Principle}{Principles}
\crefname{construction}{Construction}{Constructions}
\crefname{conjecture}{Conjecture}{Conjectures}
\crefname{exercise}{Exercise}{Exercises}
\crefname{problem}{Open Problem}{Open Problems}
\newcommand{\stag}[2]{{\footnotesize\sffamily\upshape\color{#1}[#2]}\enspace\ignorespaces}
\newcommand{\Scl}[1][]{\stag{ink2}{classical\if\relax\detokenize{#1}\relax\else; #1\fi}}
\newcommand{\Sli}[1][]{\stag{fwd}{literature\if\relax\detokenize{#1}\relax\else; #1\fi}}
\newcommand{\Srf}[1][]{\stag{accent}{reformulation\if\relax\detokenize{#1}\relax\else; cf.\ #1\fi}}
\newcommand{\Sob}{\stag{accent}{observation}}
\newcommand{\Spg}{\stag{bwd}{programme}}
\newcommand{\cat}[1]{\mathsf{#1}}
\newcommand{\Set}{\cat{Set}}
\newcommand{\FinSet}{\cat{FinSet}}
\newcommand{\Meas}{\cat{Meas}}
\newcommand{\Stoch}{\cat{Stoch}}
\newcommand{\FinStoch}{\cat{FinStoch}}
\newcommand{\BorelStoch}{\cat{BorelStoch}}
\newcommand{\FinProb}{\cat{FinProb}}
\newcommand{\FinStat}{\cat{FinStat}}
\newcommand{\Krn}{\cat{Krn}}
\newcommand{\Hilb}{\cat{Hilb}}
\newcommand{\Ban}{\cat{Ban}}
\newcommand{\Src}{\cat{Src}}
\newcommand{\Chan}{\cat{Chan}}
\newcommand{\Dyn}{\cat{Dyn}}
\newcommand{\Cat}{\cat{Cat}}
\newcommand{\CC}{\mathcal{C}}
\newcommand{\FF}{\mathcal{F}}
\newcommand{\GG}{\mathcal{G}}
\newcommand{\II}{\mathcal{I}}
\newcommand{\Sinf}{\mathcal{S}}
\newcommand{\Qlaw}{\mathcal{Q}}
\newcommand{\Fun}{\mathcal{F}}
\newcommand{\Dist}{\Delta}
\newcommand{\Giry}{\mathcal{G}}
\newcommand{\R}{\mathbb{R}}
\newcommand{\Rbar}{\overline{\mathbb{R}}}
\newcommand{\N}{\mathbb{N}}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\Fp}{\mathbb{F}_p}
\newcommand{\E}{\mathbb{E}}
\newcommand{\Prob}{\mathbb{P}}
\DeclareMathOperator*{\argmax}{arg\,max}
\DeclareMathOperator*{\colim}{colim}
\DeclareMathOperator{\Hom}{Hom}
\DeclareMathOperator{\Ext}{Ext}
\DeclareMathOperator{\Fix}{Fix}
\DeclareMathOperator{\supp}{supp}
\DeclareMathOperator{\TV}{TV}
\DeclareMathOperator{\rk}{rk}
\newcommand{\id}{\mathrm{id}}
\newcommand{\op}{\mathrm{op}}
\newcommand{\adj}{\dashv}
\newcommand{\iso}{\cong}
\newcommand{\st}{\mid}
\newcommand{\cpy}{\mathrm{copy}}
\newcommand{\del}{\mathrm{del}}
\newcommand{\KL}[2]{D(#1\,\|\,#2)}
\newcommand{\hb}{h}
\newcommand{\AZ}{A^{\Z}}
\newcommand{\BZ}{B^{\Z}}
\newcommand{\term}[1]{\textbf{\itshape #1}}
\newcommand{\Mfd}{\cat{Mfd}}
\newcommand{\Cart}{\cat{Cart}}
\newcommand{\SmSet}{\cat{SmSet}}
\newcommand{\Dflg}{\cat{Dflg}}
\newcommand{\CinfRing}{C^\infty\cat{Ring}}
\newcommand{\DMfd}{\cat{DMfd}}
\newcommand{\Spc}{\mathcal{S}}
\newcommand{\Sp}{\cat{Sp}}
\newcommand{\Cob}{\cat{Cob}}
\newcommand{\Bord}{\cat{Bord}}
\newcommand{\Vect}{\cat{Vect}}
\newcommand{\Top}{\cat{Top}}
\newcommand{\Para}{\cat{Para}}
\DeclareMathOperator{\Diff}{Diff}
\DeclareMathOperator{\Emb}{Emb}
\DeclareMathOperator{\Imm}{Imm}
\DeclareMathOperator{\Map}{Map}
\DeclareMathOperator{\Sing}{Sing}
\DeclareMathOperator{\Fr}{Fr}
\DeclareMathOperator{\Th}{Th}
\DeclareMathOperator{\Gr}{Gr}
\DeclareMathOperator{\GL}{GL}
\DeclareMathOperator{\ind}{ind}
\DeclareMathOperator{\codim}{codim}
\DeclareMathOperator{\vdim}{vdim}
\DeclareMathOperator{\tr}{tr}
\DeclareMathOperator{\grad}{grad}
\DeclareMathOperator{\Tub}{Tub}
\DeclareMathOperator{\Conf}{Conf}
\DeclareMathOperator{\diag}{diag}
\DeclareMathOperator{\relu}{ReLU}
\newcommand{\Or}{\mathrm{O}}
\newcommand{\SO}{\mathrm{SO}}
\newcommand{\RP}{\mathbb{RP}}
\newcommand{\CP}{\mathbb{CP}}
\newcommand{\C}{\mathbb{C}}
\newcommand{\trans}{\pitchfork}
\newcommand{\pt}{\mathrm{pt}}
\newcommand{\loss}{\mathcal{L}}
\tikzset{
ogbox/.style={draw, thick, rounded corners=2pt, minimum width=1.25cm, minimum height=1.05cm, fill=black!4},
fw/.style={-{Stealth[length=5pt]}, thick, fwd},
bw/.style={-{Stealth[length=5pt]}, thick, bwd, densely dashed},
dec/.style={circle, fill, inner sep=1.7pt},
nat/.style={circle, draw, thick, fill=white, inner sep=1.7pt},
leaf/.style={inner sep=2pt, font=\small},
lab/.style={font=\small, inner sep=1.5pt},
infoset/.style={densely dashed, thick, accent},
st/.style={circle, draw, thick, fill=white, inner sep=1.6pt},
ty/.style={rectangle, draw, thick, fill=black!8, inner sep=2.6pt},
}
\newcommand{\sol}[1]{\par\medskip\noindent\textbf{#1.}\enspace\ignorespaces}
\newcommand{\bookauthor}{Anonymous}
\newcommand{\figcap}[1]{\par\nopagebreak\smallskip\nopagebreak\noindent\begin{minipage}{\linewidth}\centering\small\itshape #1\end{minipage}\par}
\newcommand{\DT}[1]{\cite[#1]{DTHC}}
\newcommand{\pay}[2]{\pi(#1\mid #2)}
\renewcommand{\Game}{\cat{Game}}
\newcommand{\Rep}{\mathrm{Rep}}
\providecommand{\Q}{Q}\renewcommand{\Q}{Q}
\providecommand{\one}{\mathbf{1}}\renewcommand{\one}{\mathbf{1}}
\providecommand{\T}{\mathsf{T}}\renewcommand{\T}{\mathsf{T}}
\newcommand{\chk}[1]{{\footnotesize\sffamily\upshape\color{accent}[#1]}}
\newcommand{\src}[1]{{\footnotesize\sffamily\upshape\color{ink2}[source, #1]}}
\providecommand{\ev}{\mathrm{ev}}
\newcommand{\Lap}{\Delta}
\newcommand{\dd}{\mathrm{d}}
\newcommand{\pd}{\partial}
\newcommand{\DD}{\mathcal{D}}
\newcommand{\HH}{\mathcal{H}}
\newcommand{\OO}{\mathcal{O}}
\newcommand{\EE}{\mathcal{E}}
\newcommand{\PDE}{\cat{PDE}}
\newcommand{\LagRel}{\cat{LagRel}}
\newcommand{\Symp}{\cat{Symp}}
\newcommand{\eps}{\varepsilon}
\DeclareMathOperator{\Sol}{Sol}
\DeclareMathOperator{\Div}{div}
\DeclareMathOperator{\curl}{curl}
\DeclareMathOperator{\sech}{sech}
\DeclareMathOperator{\erf}{erf}
\DeclareMathOperator{\spec}{spec}
\DeclareMathOperator{\WF}{WF}
\DeclareMathOperator{\im}{im}
\DeclareMathOperator{\Char}{Char}
\DeclareMathOperator{\sgn}{sgn}
\newcommand{\Jet}{J^\infty}
\begin{document}
\frontmatter
\begin{titlepage}
\thispagestyle{empty}
\vspace*{1.1in}
\noindent{\Large\scshape\color{accent} Higher Categorical Semantics}\par
\vspace{0.9in}
\noindent{\fontsize{28}{36}\selectfont\bfseries\raggedright Partial Differential\\ Equations in Higher\\ Categories\par}
\vspace{0.35in}
\noindent{\Large\itshape Waves, diffusions and potentials as functors\\[2pt] of domain, symmetry and time}\par
\vfill
\noindent\begin{tikzpicture}[x=1cm,y=1cm]
\draw[black,thick] (0,0) -- (3.2,0);
\draw[fw] (0.4,0) -- (1.6,1.2) -- (2.8,0);
\draw[black!45] (1.6,1.2) -- (1.0,1.8) (1.6,1.2) -- (2.2,1.8);
\fill[accent] (1.6,1.2) circle (1.6pt);
\node[lab] at (1.6,-0.3) {$I$}; \node[lab] at (1.6,1.55) {$D(I)$};
\node[lab] at (4.4,0.7) {$\longmapsto$};
\node[ogbox] (G) at (6.3,0.7) {$\Sol$};
\node[lab] at (9.1,0.7) {$\mathsf{Dom}^{\op}\to\Vect$};
\end{tikzpicture}\par
\vspace{0.5in}
\noindent{\large \bookauthor}\par
\vspace{4pt}
\noindent{\small 2026}
\end{titlepage}
\thispagestyle{empty}
\vspace*{\fill}
\noindent{\small
Version 1.0, 18 November 2026.\par\medskip
\noindent This volume is a treatment, in categorical language, of the material of Walter A.~Strauss, \emph{Partial
Differential Equations: An Introduction} (second edition, Wiley, 2008). It reproduces none of that book's text,
figures, tables or exercises; where a numerical example of the source is recomputed, this is said. An appendix gives
a section-level concordance.\par\medskip
\noindent Every numbered statement carries a status tag: \Scl a classical result; \Sli[source] a result of the cited
literature; \Srf[source] a known result restated in the language of this volume; \Sob an observation of this volume,
proved here; \Spg a proposal. Tags of the form \chk{V6} name a check in the companion script
\texttt{verify-pde.py}; tags of the form \src{\S3.2} point to the source.\par\medskip
\noindent This book was drafted with the assistance of an AI system (Anthropic's Claude), which also wrote and ran the
companion script. The bibliography was assembled from memory, except for the entries marked as checked, and should
be verified before it is relied on.}
\clearpage
\chapter*{Preface}
Strauss's \emph{Partial Differential Equations} \cite{Str08} is an introduction for undergraduates built on three
equations, of waves, of diffusion and of equilibrium, and on a small number of methods: characteristics, reflection,
separation of variables, Green's functions, transforms. It is explicit, physically motivated, and careful to say
what it does not prove.
This volume retells that material with the question of the series in mind: which constructions are
\emph{natural}, in the sense that a map of domains, of equations or of function spaces induces a map of solutions?
A differential equation is treated as an object, its solutions as the values of a functor, and the methods of the
source as instances of functoriality in one of a few directions.
\begin{itemize}[leftmargin=1.4em]
\item \emph{Restriction.} Solutions form a sheaf. For the wave equation it is determined by data on a time slice
(\Cref{thm:causal}); for Laplace's equation it has cohomology on an annulus (\Cref{prop:annulus}).
\item \emph{Reflection.} A boundary condition on a chamber of a reflection group is a character of the group
(\Cref{thm:images}), and in general a Lagrangian plane of boundary values (\Cref{thm:lagrangian}). The method of
images and Fourier's method are exchanged by Pontryagin duality (\Cref{thm:kernels}).
\item \emph{Time.} Waves are a functor from time to symplectic maps, diffusions a functor from time to Markov
kernels, and the source's table of contrasts follows from that (\Cref{ch:diffusions}).
\item \emph{Symmetry.} Eigenspaces are representations (\Cref{thm:rep}); the special functions are what the rotation
groups produce; and domains can have the same spectrum because two permutation representations are isomorphic
(\Cref{thm:sunada}).
\item \emph{Gluing.} Dirichlet-to-Neumann maps and transfer matrices compose as Lagrangian relations
(\Cref{thm:gluing}, \Cref{thm:transfer}).
\item \emph{Discretization.} A scheme is a square that almost commutes; its stability is positivity for diffusion,
causality for waves, and compatibility with the de Rham complex for elliptic problems (\Cref{ch:numerics}).
\item \emph{Morphisms of equations.} The solvable nonlinear examples of the source are images of linear equations
(\Cref{thm:colehopf}, \Cref{thm:ist}).
\end{itemize}
Quantitative statements accompany the qualitative ones throughout: amplification factors, exponents, constants,
eigenvalues, rates, counts of terms. All of them are recomputed by the companion script, which performs 107 checks.
How much of this is ``higher'' is assessed in \Cref{ch:coda}. In brief: most of it is ordinary category theory and
representation theory; cohomology enters in degree one; and gauge fields are the one place where the solution
functor must take values in higher groupoids (\Cref{rem:stack}). Nothing here is a new theorem about partial
differential equations, and no existence theorem is proved. A companion paper \cite{PDEML} applies the framework to
machine learning.
The volume belongs to a series with \cite{GIHC}, on game theory, \cite{ESCHC}, on information theory, \cite{DTHC}, on
differential topology, and \cite{EGHC}, on evolutionary games, and uses the second and third.
\tableofcontents
\mainmatter
\part{Equations}
\chapter{What an Equation Is}\label{ch:equations}
The source begins by solving a few equations and only then asks what kind of thing has been solved
\src{\S1.1--1.2, \S1.6}. We reverse the order, because the answer decides what the rest of the volume can say.
A differential equation will be an object of a category. Its solutions will be the values of a functor. The
statements of the following chapters are then of one form: such-and-such a map of domains, of equations or of
function spaces induces a map of solutions.
\section{Equations in jets}
Let $M\subseteq\R^n$ be open and let $u$ be a real function on $M$. The \term{$k$-jet} of $u$ at $x$ is the list
of its derivatives of order at most $k$ there; jets form a bundle $J^kM\to M$ with coordinates
$(x,u_\alpha)_{|\alpha|\le k}$, and $u$ determines a section $j^ku$.
\begin{definition}\label{def:pde}
A \term{differential equation of order $k$} on $M$ is a subset $\EE\subseteq J^kM$, typically the zero set of a
function $F(x,u,u_i,u_{ij},\dots)$. A \term{solution} is a function $u$ whose jet $j^ku$ takes values in $\EE$.
\end{definition}
Differentiating $F=0$ gives equations of higher order that every solution also satisfies; adjoining all of
them gives the \term{prolongation} $\EE^\infty\subseteq\Jet M$. A point of $\EE^\infty$ over $x$ is a
\term{formal solution} at $x$: a Taylor series that satisfies the equation to all orders.
\begin{theorem}[Equations are coalgebras]\label{thm:jet}
\Sli[\cite{Mar86,KhSc17}] The infinite jet construction is a comonad $\Jet$ on bundles over $M$. Its coalgebras are
the (formally integrable) differential equations in the sense of Vinogradov, and a morphism of coalgebras is a
differential operator that carries solutions to solutions. In synthetic differential geometry $\Jet$ is base change
along the unit $M\to\Im M$ of the infinitesimal shape, and in the cases of interest equations over $M$ are
bundles over $\Im M$.
\end{theorem}
\noindent We use two consequences and none of the machinery. There is a category $\PDE$ whose morphisms are
differential substitutions; the Cole--Hopf and Miura transformations of \Cref{ch:nonlinear} are morphisms in it.
And solutions are functorial: $\EE\mapsto\Sol(\EE)$ sends a morphism of equations to a map of solution sets. Since
solutions restrict to open subsets and glue, $\Sol(\EE)$ is a sheaf on $M$.
\section{Linear equations and their modules}
Let $\DD$ be the ring of linear differential operators on $M$ with smooth coefficients. A linear equation $Pu=0$
is recorded by the left $\DD$-module $\DD/\DD P$, and for any left $\DD$-module $\FF$ of functions
\[ \Sol(P;\FF)=\Hom_\DD(\DD/\DD P,\FF)=\{u\in\FF:Pu=0\}. \]
This is functorial in both variables. In $\FF$ it is covariant: smooth functions sit inside distributions, and
classical solutions inside weak ones (\Cref{ch:distributions}). In the equation it is contravariant: if $QR=SP$ then
$u\mapsto Ru$ maps $\Sol(P)$ to $\Sol(Q)$.
\begin{proposition}[Factorization]\label{prop:factor}
\Srf For operators $P,Q$ there is an exact sequence
\[ 0\longrightarrow\Sol(Q)\longrightarrow\Sol(PQ)\xrightarrow{\ Q\ }\Sol(P). \]
If $P$ and $Q$ commute, then $\Sol(P)+\Sol(Q)\subseteq\Sol(PQ)$, with equality as soon as $Q$ maps $\Sol(P)$ onto
itself: for $u\in\Sol(PQ)$ choose $w\in\Sol(P)$ with $Qw=Qu$, and then $u-w\in\Sol(Q)$.
\end{proposition}
\noindent The wave operator is the case $P=\pd_t-c\pd_x$, $Q=\pd_t+c\pd_x$, checked on polynomials \chk{E5}; d'Alembert's
solution in \Cref{ch:waves} is the equality $\Sol(PQ)=\Sol(P)+\Sol(Q)$.
In a diagram: for commuting $P$ and $Q$ the square of inclusions
\[\begin{tikzcd}[column sep=large]
\Sol(P)\cap\Sol(Q)\arrow[r,hook]\arrow[d,hook] & \Sol(P)\arrow[d,hook]\\
\Sol(Q)\arrow[r,hook] & \Sol(PQ)
\end{tikzcd}\]
is a pullback of vector spaces, and it is also a pushout exactly when $\Sol(P)+\Sol(Q)=\Sol(PQ)$. For the wave
operator the corner is the constants, and d'Alembert's theorem says that the square is bicartesian: a solution is
a right-moving and a left-moving wave, glued along what they have in common.
For constant coefficients $\DD=\R[\pd_1,\dots,\pd_n]$ is a polynomial ring, $P=p(\pd)$, and the module is
$\R[\xi]/(p)$: the equation \emph{is} the hypersurface $V(p)=\{p(\xi)=0\}\subseteq\C^n$. Each point $\xi\in V(p)$ gives the
exponential solution $e^{\xi\cdot x}$.
\begin{theorem}[Fundamental principle]\label{thm:ehrenpreis}
\Sli[\cite{Ehr70,Pal70}] On a convex domain every smooth solution of a constant-coefficient equation $p(\pd)u=0$ is an
integral, over $V(p)$, of exponential solutions with polynomial coefficients.
\end{theorem}
\noindent Separation of variables is thus not a trick that happens to work: for constant coefficients it is the
general solution. The source's Fourier method (\Cref{ch:fourier}) is the special case in which boundary conditions
select a discrete set of points of $V(p)$.
\section{The symbol and the type}
Write a second-order operator as $P=\sum_{i,j}a_{ij}(x)\pd_i\pd_j+{}$lower order, with $a_{ij}$ symmetric. Its
\term{principal symbol} is the quadratic form $\sigma_P(x,\xi)=\sum a_{ij}(x)\xi_i\xi_j$.
\begin{theorem}[Type]\label{thm:type}
\Scl[source, Theorem 1 of \S1.6] Under a change of coordinates $y=\varphi(x)$ the symbol transforms by congruence
with the Jacobian, so $\sigma_P$ is a function on the cotangent bundle and its inertia $(p,q,z)$ at each point is
invariant. The operator is \term{elliptic} at $x$ if $\sigma_P$ is definite, \term{hyperbolic} if it is
non-degenerate with one sign differing from the rest, \term{parabolic} if it is semidefinite with a one-dimensional
kernel. In two variables these are the cases $a_{12}^2<a_{11}a_{22}$, $>$ and $=$.
\end{theorem}
\noindent Invariance was tested on three hundred random symbols and changes of variables, and the classification of
the source's variable-coefficient example by the sign of $1-xy$ was reproduced. \chk{E1, E2}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,0) rectangle (4.2,4.2);
\draw[ink2] (0.7,0)--(0.7,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (2.1,0)--(2.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.5,0)--(3.5,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,0.7)--(-0.07,0.7) node[left,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (0,2.1)--(-0.07,2.1) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.5)--(-0.07,3.5) node[left,inner sep=1.5pt]{\scriptsize $2$};
\fill[bwd!14] (2.333,4.2) -- (2.38,3.85) -- (2.427,3.6) -- (2.473,3.413) -- (2.52,3.267) -- (2.567,3.15) -- (2.613,3.055) -- (2.66,2.975) -- (2.707,2.908) -- (2.753,2.85) -- (2.8,2.8) -- (2.847,2.756) -- (2.893,2.718) -- (2.94,2.683) -- (2.987,2.653) -- (3.033,2.625) -- (3.08,2.6) -- (3.127,2.577) -- (3.173,2.557) -- (3.22,2.538) -- (3.267,2.52) -- (3.313,2.504) -- (3.36,2.489) -- (3.407,2.475) -- (3.453,2.462) -- (3.5,2.45) -- (3.547,2.439) -- (3.593,2.428) -- (3.64,2.418) -- (3.687,2.409) -- (3.733,2.4) -- (3.78,2.392) -- (3.827,2.384) -- (3.873,2.376) -- (3.92,2.369) -- (3.967,2.363) -- (4.013,2.356) -- (4.06,2.35) -- (4.107,2.344) -- (4.153,2.339) -- (4.2,2.333) -- (4.2,4.2) -- cycle;
\fill[bwd!14] (1.867,0) -- (1.82,0.35) -- (1.773,0.6) -- (1.727,0.788) -- (1.68,0.933) -- (1.633,1.05) -- (1.587,1.145) -- (1.54,1.225) -- (1.493,1.292) -- (1.447,1.35) -- (1.4,1.4) -- (1.353,1.444) -- (1.307,1.482) -- (1.26,1.517) -- (1.213,1.547) -- (1.167,1.575) -- (1.12,1.6) -- (1.073,1.623) -- (1.027,1.643) -- (0.98,1.662) -- (0.933,1.68) -- (0.887,1.696) -- (0.84,1.711) -- (0.793,1.725) -- (0.747,1.738) -- (0.7,1.75) -- (0.653,1.761) -- (0.607,1.772) -- (0.56,1.782) -- (0.513,1.791) -- (0.467,1.8) -- (0.42,1.808) -- (0.373,1.816) -- (0.327,1.824) -- (0.28,1.831) -- (0.233,1.838) -- (0.187,1.844) -- (0.14,1.85) -- (0.093,1.856) -- (0.047,1.861) -- (0,1.867) -- (0,0) -- cycle;
\draw[thick,bwd] (2.333,4.2) -- (2.38,3.85) -- (2.427,3.6) -- (2.473,3.413) -- (2.52,3.267) -- (2.567,3.15) -- (2.613,3.055) -- (2.66,2.975) -- (2.707,2.908) -- (2.753,2.85) -- (2.8,2.8) -- (2.847,2.756) -- (2.893,2.718) -- (2.94,2.683) -- (2.987,2.653) -- (3.033,2.625) -- (3.08,2.6) -- (3.127,2.577) -- (3.173,2.557) -- (3.22,2.538) -- (3.267,2.52) -- (3.313,2.504) -- (3.36,2.489) -- (3.407,2.475) -- (3.453,2.462) -- (3.5,2.45) -- (3.547,2.439) -- (3.593,2.428) -- (3.64,2.418) -- (3.687,2.409) -- (3.733,2.4) -- (3.78,2.392) -- (3.827,2.384) -- (3.873,2.376) -- (3.92,2.369) -- (3.967,2.363) -- (4.013,2.356) -- (4.06,2.35) -- (4.107,2.344) -- (4.153,2.339) -- (4.2,2.333);
\draw[thick,bwd] (1.867,0) -- (1.82,0.35) -- (1.773,0.6) -- (1.727,0.788) -- (1.68,0.933) -- (1.633,1.05) -- (1.587,1.145) -- (1.54,1.225) -- (1.493,1.292) -- (1.447,1.35) -- (1.4,1.4) -- (1.353,1.444) -- (1.307,1.482) -- (1.26,1.517) -- (1.213,1.547) -- (1.167,1.575) -- (1.12,1.6) -- (1.073,1.623) -- (1.027,1.643) -- (0.98,1.662) -- (0.933,1.68) -- (0.887,1.696) -- (0.84,1.711) -- (0.793,1.725) -- (0.747,1.738) -- (0.7,1.75) -- (0.653,1.761) -- (0.607,1.772) -- (0.56,1.782) -- (0.513,1.791) -- (0.467,1.8) -- (0.42,1.808) -- (0.373,1.816) -- (0.327,1.824) -- (0.28,1.831) -- (0.233,1.838) -- (0.187,1.844) -- (0.14,1.85) -- (0.093,1.856) -- (0.047,1.861) -- (0,1.867);
\draw[ink2,very thin] (0,2.1)--(4.2,2.1);
\draw[ink2,very thin] (2.1,0)--(2.1,4.2);
\node[] at (3.57,3.71) {\scriptsize elliptic};
\node[] at (0.63,0.49) {\scriptsize elliptic};
\node[] at (1.19,3.08) {\scriptsize hyperbolic};
\node[] at (3.045,1.05) {\scriptsize hyperbolic};
\node[bwd] at (3.885,2.534) {\scriptsize $xy=1$};
\end{scope}
\end{tikzpicture}
\figcap{The type of $yu_{xx}-2u_{xy}+xu_{yy}$ in the plane: hyperbolic where $xy<1$, elliptic where $xy>1$, parabolic on the two branches of $xy=1$. The type is a function on the domain and may change across it.}
\end{center}
The \term{characteristic variety} $\Char(P)=\{(x,\xi):\xi\ne0,\ \sigma_P(x,\xi)=0\}$ is the real part of the
hypersurface above, taken to leading order. It is empty exactly for elliptic operators; for the wave operator it
is the light cone and for the heat operator a line. \chk{E5}
\begin{remark}[Algebra does not see the type]\label{rem:type}
\Sob The homogeneous polynomial solutions of degree $k$ of a non-zero homogeneous second-order operator in $n$
variables with constant coefficients form a space of dimension $\binom{n+k-1}{k}-\binom{n+k-3}{k-2}$, whatever the
type: the operator maps polynomials of degree $k$ onto those of degree $k-2$. \chk{E4} The count is $2$ in the plane
and $2k+1$ in space, for Laplace, wave and degenerate operators alike. Elliptic, hyperbolic and parabolic
equations have the same formal solutions; they differ in which formal solutions converge and in how solutions
depend on data. The type is a statement about the \emph{real} points of the characteristic variety, and its
consequences are analytic. This is why the three archetypes need three chapters, and why the categorical language
must include topology on spaces of functions to say anything about them.
\end{remark}
\begin{example}[Kovalevskaya]\label{ex:kov}
For $u_t=u_{xx}$ the Taylor coefficients in $t$ are determined by those in $x$: $\pd_t^ju=\pd_x^{2j}u$. With
polynomial data $x^k$ the series terminates, in the heat polynomials, which solve the equation exactly. With the
analytic datum $1/(1+x^2)$ the coefficients of $t^j$ at $x=0$ are $(-1)^j(2j)!/j!$, successive ratios are $2(2j+1)$,
and the series diverges for every $t\ne0$. \chk{E6} The Cauchy--Kovalevskaya theorem gives convergent solutions from
analytic data on a non-characteristic surface; $t=0$ is characteristic for the heat equation.
\end{example}
\section{First-order equations are quotients}
A first-order linear equation $a(x,y)u_x+b(x,y)u_y=0$ says that $u$ is constant along the integral curves of the
vector field $X=(a,b)$, its \term{characteristics}.
\begin{proposition}[Solutions are functions on the leaf space]\label{prop:char}
\Srf[source, \S1.2] If $q\colon M\to Q$ is a submersion whose fibres are the characteristics of $X$, then pullback
$q^*\colon C^\infty(Q)\to\Sol(X)$ is an isomorphism. A curve $\Gamma\subset M$ that meets every characteristic exactly
once and transversally is a section of $q$, and restriction $\Sol(X)\to C^\infty(\Gamma)$ is then an isomorphism.
\end{proposition}
\noindent The source's two examples are $q=-3x-4y$ for $4u_x-3u_y=0$ and $q=e^{-x}y$ for $u_x+yu_y=0$. \chk{E3} The
second half of the proposition is the first well-posedness statement of the subject, and its failure is
instructive: the line $y=0$ is itself a characteristic of $u_x+yu_y=0$, every solution is constant on it, and the
problem with data $u(x,0)=\varphi(x)$ has no solution unless $\varphi$ is constant and infinitely many if it is.
Data must be given on a section of the quotient.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,0) rectangle (3.7,3.7);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (1.85,0)--(1.85,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.7,0)--(3.7,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (0,1.85)--(-0.07,1.85) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.7)--(-0.07,3.7) node[left,inner sep=1.5pt]{\scriptsize $2$};
\draw[fwd] plot coordinates {(3.145,3.654) (3.238,3.584) (3.33,3.515) (3.423,3.446) (3.515,3.376) (3.607,3.307) (3.7,3.238)};
\draw[fwd] plot coordinates {(2.22,3.654) (2.312,3.584) (2.405,3.515) (2.498,3.446) (2.59,3.376) (2.683,3.307) (2.775,3.238) (2.868,3.168) (2.96,3.099) (3.052,3.029) (3.145,2.96) (3.238,2.891) (3.33,2.821) (3.423,2.752) (3.515,2.683) (3.607,2.613) (3.7,2.544)};
\draw[fwd] plot coordinates {(1.295,3.654) (1.388,3.584) (1.48,3.515) (1.573,3.446) (1.665,3.376) (1.758,3.307) (1.85,3.238) (1.943,3.168) (2.035,3.099) (2.127,3.029) (2.22,2.96) (2.312,2.891) (2.405,2.821) (2.498,2.752) (2.59,2.683) (2.683,2.613) (2.775,2.544) (2.868,2.474) (2.96,2.405) (3.052,2.336) (3.145,2.266) (3.238,2.197) (3.33,2.127) (3.423,2.058) (3.515,1.989) (3.607,1.919) (3.7,1.85)};
\draw[fwd] plot coordinates {(0.37,3.654) (0.463,3.584) (0.555,3.515) (0.647,3.446) (0.74,3.376) (0.832,3.307) (0.925,3.238) (1.018,3.168) (1.11,3.099) (1.203,3.029) (1.295,2.96) (1.388,2.891) (1.48,2.821) (1.573,2.752) (1.665,2.683) (1.758,2.613) (1.85,2.544) (1.943,2.474) (2.035,2.405) (2.127,2.336) (2.22,2.266) (2.312,2.197) (2.405,2.127) (2.498,2.058) (2.59,1.989) (2.683,1.919) (2.775,1.85) (2.868,1.781) (2.96,1.711) (3.052,1.642) (3.145,1.573) (3.238,1.503) (3.33,1.434) (3.423,1.364) (3.515,1.295) (3.607,1.226) (3.7,1.156)};
\draw[fwd] plot coordinates {(0,3.238) (0.093,3.168) (0.185,3.099) (0.278,3.029) (0.37,2.96) (0.463,2.891) (0.555,2.821) (0.647,2.752) (0.74,2.683) (0.832,2.613) (0.925,2.544) (1.018,2.474) (1.11,2.405) (1.203,2.336) (1.295,2.266) (1.388,2.197) (1.48,2.127) (1.573,2.058) (1.665,1.989) (1.758,1.919) (1.85,1.85) (1.943,1.781) (2.035,1.711) (2.127,1.642) (2.22,1.573) (2.312,1.503) (2.405,1.434) (2.498,1.364) (2.59,1.295) (2.683,1.226) (2.775,1.156) (2.868,1.087) (2.96,1.017) (3.052,0.948) (3.145,0.879) (3.238,0.809) (3.33,0.74) (3.423,0.671) (3.515,0.601) (3.607,0.532) (3.7,0.463)};
\draw[fwd] plot coordinates {(0,2.544) (0.093,2.474) (0.185,2.405) (0.278,2.336) (0.37,2.266) (0.463,2.197) (0.555,2.127) (0.647,2.058) (0.74,1.989) (0.832,1.919) (0.925,1.85) (1.018,1.781) (1.11,1.711) (1.203,1.642) (1.295,1.573) (1.388,1.503) (1.48,1.434) (1.573,1.364) (1.665,1.295) (1.758,1.226) (1.85,1.156) (1.943,1.087) (2.035,1.017) (2.127,0.948) (2.22,0.879) (2.312,0.809) (2.405,0.74) (2.498,0.671) (2.59,0.601) (2.683,0.532) (2.775,0.463) (2.868,0.393) (2.96,0.324) (3.052,0.254) (3.145,0.185) (3.238,0.116) (3.33,0.046)};
\draw[fwd] plot coordinates {(0,1.85) (0.093,1.781) (0.185,1.711) (0.278,1.642) (0.37,1.573) (0.463,1.503) (0.555,1.434) (0.647,1.364) (0.74,1.295) (0.832,1.226) (0.925,1.156) (1.018,1.087) (1.11,1.018) (1.203,0.948) (1.295,0.879) (1.388,0.809) (1.48,0.74) (1.573,0.671) (1.665,0.601) (1.758,0.532) (1.85,0.463) (1.943,0.393) (2.035,0.324) (2.127,0.254) (2.22,0.185) (2.312,0.116) (2.405,0.046)};
\draw[fwd] plot coordinates {(0,1.156) (0.093,1.087) (0.185,1.018) (0.278,0.948) (0.37,0.879) (0.463,0.809) (0.555,0.74) (0.647,0.671) (0.74,0.601) (0.832,0.532) (0.925,0.463) (1.018,0.393) (1.11,0.324) (1.203,0.254) (1.295,0.185) (1.388,0.116) (1.48,0.046)};
\draw[fwd] plot coordinates {(0,0.463) (0.093,0.393) (0.185,0.324) (0.278,0.254) (0.37,0.185) (0.463,0.116) (0.555,0.046)};
\node[] at (1.85,3.931) {\scriptsize $4u_x-3u_y=0$};
\end{scope}
\begin{scope}[shift={(5.6,0)}]
\draw[ink2] (0,0) rectangle (3.7,3.7);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (1.85,0)--(1.85,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.7,0)--(3.7,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (0,1.85)--(-0.07,1.85) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.7)--(-0.07,3.7) node[left,inner sep=1.5pt]{\scriptsize $2$};
\draw[fwd] plot coordinates {(0,1.856) (0.062,1.857) (0.123,1.857) (0.185,1.858) (0.247,1.858) (0.308,1.859) (0.37,1.859) (0.432,1.86) (0.493,1.861) (0.555,1.861) (0.617,1.862) (0.678,1.863) (0.74,1.864) (0.802,1.865) (0.863,1.866) (0.925,1.867) (0.987,1.868) (1.048,1.869) (1.11,1.871) (1.172,1.872) (1.233,1.874) (1.295,1.875) (1.357,1.877) (1.418,1.879) (1.48,1.881) (1.542,1.883) (1.603,1.885) (1.665,1.888) (1.727,1.89) (1.788,1.893) (1.85,1.896) (1.912,1.899) (1.973,1.903) (2.035,1.906) (2.097,1.91) (2.158,1.915) (2.22,1.919) (2.282,1.924) (2.343,1.929) (2.405,1.934) (2.467,1.94) (2.528,1.946) (2.59,1.953) (2.652,1.96) (2.713,1.968) (2.775,1.976) (2.837,1.984) (2.898,1.994) (2.96,2.004) (3.022,2.014) (3.083,2.025) (3.145,2.038) (3.207,2.05) (3.268,2.064) (3.33,2.079) (3.392,2.095) (3.453,2.112) (3.515,2.13) (3.577,2.149) (3.638,2.17) (3.7,2.192)};
\draw[fwd] plot coordinates {(0,1.844) (0.062,1.843) (0.123,1.843) (0.185,1.842) (0.247,1.842) (0.308,1.841) (0.37,1.841) (0.432,1.84) (0.493,1.839) (0.555,1.839) (0.617,1.838) (0.678,1.837) (0.74,1.836) (0.802,1.835) (0.863,1.834) (0.925,1.833) (0.987,1.832) (1.048,1.831) (1.11,1.829) (1.172,1.828) (1.233,1.826) (1.295,1.825) (1.357,1.823) (1.418,1.821) (1.48,1.819) (1.542,1.817) (1.603,1.815) (1.665,1.812) (1.727,1.81) (1.788,1.807) (1.85,1.804) (1.912,1.801) (1.973,1.797) (2.035,1.794) (2.097,1.79) (2.158,1.785) (2.22,1.781) (2.282,1.776) (2.343,1.771) (2.405,1.766) (2.467,1.76) (2.528,1.754) (2.59,1.747) (2.652,1.74) (2.713,1.732) (2.775,1.724) (2.837,1.716) (2.898,1.706) (2.96,1.696) (3.022,1.686) (3.083,1.675) (3.145,1.662) (3.207,1.65) (3.268,1.636) (3.33,1.621) (3.392,1.605) (3.453,1.588) (3.515,1.57) (3.577,1.551) (3.638,1.53) (3.7,1.508)};
\draw[fwd] plot coordinates {(0,1.865) (0.062,1.866) (0.123,1.867) (0.185,1.868) (0.247,1.87) (0.308,1.871) (0.37,1.872) (0.432,1.874) (0.493,1.876) (0.555,1.877) (0.617,1.879) (0.678,1.881) (0.74,1.883) (0.802,1.886) (0.863,1.888) (0.925,1.891) (0.987,1.894) (1.048,1.897) (1.11,1.9) (1.172,1.903) (1.233,1.907) (1.295,1.911) (1.357,1.915) (1.418,1.92) (1.48,1.924) (1.542,1.93) (1.603,1.935) (1.665,1.941) (1.727,1.947) (1.788,1.954) (1.85,1.961) (1.912,1.969) (1.973,1.977) (2.035,1.986) (2.097,1.995) (2.158,2.005) (2.22,2.016) (2.282,2.027) (2.343,2.039) (2.405,2.052) (2.467,2.066) (2.528,2.081) (2.59,2.097) (2.652,2.114) (2.713,2.132) (2.775,2.152) (2.837,2.173) (2.898,2.195) (2.96,2.219) (3.022,2.244) (3.083,2.271) (3.145,2.3) (3.207,2.331) (3.268,2.364) (3.33,2.4) (3.392,2.438) (3.453,2.478) (3.515,2.522) (3.577,2.568) (3.638,2.617) (3.7,2.67)};
\draw[fwd] plot coordinates {(0,1.835) (0.062,1.834) (0.123,1.833) (0.185,1.832) (0.247,1.83) (0.308,1.829) (0.37,1.828) (0.432,1.826) (0.493,1.824) (0.555,1.823) (0.617,1.821) (0.678,1.819) (0.74,1.817) (0.802,1.814) (0.863,1.812) (0.925,1.809) (0.987,1.806) (1.048,1.803) (1.11,1.8) (1.172,1.797) (1.233,1.793) (1.295,1.789) (1.357,1.785) (1.418,1.78) (1.48,1.776) (1.542,1.77) (1.603,1.765) (1.665,1.759) (1.727,1.753) (1.788,1.746) (1.85,1.739) (1.912,1.731) (1.973,1.723) (2.035,1.714) (2.097,1.705) (2.158,1.695) (2.22,1.684) (2.282,1.673) (2.343,1.661) (2.405,1.648) (2.467,1.634) (2.528,1.619) (2.59,1.603) (2.652,1.586) (2.713,1.568) (2.775,1.548) (2.837,1.527) (2.898,1.505) (2.96,1.481) (3.022,1.456) (3.083,1.429) (3.145,1.4) (3.207,1.369) (3.268,1.336) (3.33,1.3) (3.392,1.262) (3.453,1.222) (3.515,1.178) (3.577,1.132) (3.638,1.083) (3.7,1.03)};
\draw[fwd] plot coordinates {(0,1.888) (0.062,1.89) (0.123,1.893) (0.185,1.896) (0.247,1.899) (0.308,1.902) (0.37,1.906) (0.432,1.91) (0.493,1.914) (0.555,1.918) (0.617,1.923) (0.678,1.928) (0.74,1.934) (0.802,1.939) (0.863,1.946) (0.925,1.952) (0.987,1.959) (1.048,1.967) (1.11,1.975) (1.172,1.983) (1.233,1.992) (1.295,2.002) (1.357,2.013) (1.418,2.024) (1.48,2.036) (1.542,2.049) (1.603,2.063) (1.665,2.077) (1.727,2.093) (1.788,2.11) (1.85,2.127) (1.912,2.147) (1.973,2.167) (2.035,2.189) (2.097,2.212) (2.158,2.237) (2.22,2.264) (2.282,2.293) (2.343,2.323) (2.405,2.356) (2.467,2.39) (2.528,2.428) (2.59,2.468) (2.652,2.51) (2.713,2.556) (2.775,2.604) (2.837,2.656) (2.898,2.712) (2.96,2.771) (3.022,2.835) (3.083,2.903) (3.145,2.975) (3.207,3.053) (3.268,3.136) (3.33,3.224) (3.392,3.319) (3.453,3.421) (3.515,3.529) (3.577,3.645)};
\draw[fwd] plot coordinates {(0,1.812) (0.062,1.81) (0.123,1.807) (0.185,1.804) (0.247,1.801) (0.308,1.798) (0.37,1.794) (0.432,1.79) (0.493,1.786) (0.555,1.782) (0.617,1.777) (0.678,1.772) (0.74,1.766) (0.802,1.761) (0.863,1.754) (0.925,1.748) (0.987,1.741) (1.048,1.733) (1.11,1.725) (1.172,1.717) (1.233,1.708) (1.295,1.698) (1.357,1.687) (1.418,1.676) (1.48,1.664) (1.542,1.651) (1.603,1.637) (1.665,1.623) (1.727,1.607) (1.788,1.59) (1.85,1.573) (1.912,1.553) (1.973,1.533) (2.035,1.511) (2.097,1.488) (2.158,1.463) (2.22,1.436) (2.282,1.407) (2.343,1.377) (2.405,1.344) (2.467,1.31) (2.528,1.272) (2.59,1.232) (2.652,1.19) (2.713,1.144) (2.775,1.096) (2.837,1.044) (2.898,0.988) (2.96,0.929) (3.022,0.865) (3.083,0.797) (3.145,0.725) (3.207,0.647) (3.268,0.564) (3.33,0.476) (3.392,0.381) (3.453,0.279) (3.515,0.171) (3.577,0.055)};
\draw[fwd] plot coordinates {(0,1.938) (0.062,1.944) (0.123,1.95) (0.185,1.957) (0.247,1.964) (0.308,1.972) (0.37,1.981) (0.432,1.99) (0.493,1.999) (0.555,2.01) (0.617,2.021) (0.678,2.032) (0.74,2.045) (0.802,2.058) (0.863,2.073) (0.925,2.088) (0.987,2.105) (1.048,2.122) (1.11,2.141) (1.172,2.161) (1.233,2.182) (1.295,2.205) (1.357,2.23) (1.418,2.256) (1.48,2.284) (1.542,2.314) (1.603,2.346) (1.665,2.38) (1.727,2.417) (1.788,2.456) (1.85,2.498) (1.912,2.542) (1.973,2.59) (2.035,2.641) (2.097,2.695) (2.158,2.754) (2.22,2.816) (2.282,2.883) (2.343,2.954) (2.405,3.03) (2.467,3.111) (2.528,3.198) (2.59,3.291) (2.652,3.39) (2.713,3.497) (2.775,3.61)};
\draw[fwd] plot coordinates {(0,1.762) (0.062,1.756) (0.123,1.75) (0.185,1.743) (0.247,1.736) (0.308,1.728) (0.37,1.719) (0.432,1.71) (0.493,1.701) (0.555,1.69) (0.617,1.679) (0.678,1.668) (0.74,1.655) (0.802,1.642) (0.863,1.627) (0.925,1.612) (0.987,1.595) (1.048,1.578) (1.11,1.559) (1.172,1.539) (1.233,1.518) (1.295,1.495) (1.357,1.47) (1.418,1.444) (1.48,1.416) (1.542,1.386) (1.603,1.354) (1.665,1.32) (1.727,1.283) (1.788,1.244) (1.85,1.203) (1.912,1.158) (1.973,1.11) (2.035,1.059) (2.097,1.005) (2.158,0.946) (2.22,0.884) (2.282,0.817) (2.343,0.746) (2.405,0.67) (2.467,0.589) (2.528,0.502) (2.59,0.409) (2.652,0.31) (2.713,0.203) (2.775,0.09)};
\draw[fwd] plot coordinates {(0,2.05) (0.062,2.064) (0.123,2.079) (0.185,2.095) (0.247,2.112) (0.308,2.13) (0.37,2.149) (0.432,2.169) (0.493,2.191) (0.555,2.215) (0.617,2.24) (0.678,2.267) (0.74,2.296) (0.802,2.326) (0.863,2.359) (0.925,2.394) (0.987,2.432) (1.048,2.472) (1.11,2.515) (1.172,2.561) (1.233,2.61) (1.295,2.662) (1.357,2.718) (1.418,2.778) (1.48,2.842) (1.542,2.91) (1.603,2.984) (1.665,3.062) (1.727,3.145) (1.788,3.235) (1.85,3.33) (1.912,3.432) (1.973,3.541) (2.035,3.658)};
\draw[fwd] plot coordinates {(0,1.65) (0.062,1.636) (0.123,1.621) (0.185,1.605) (0.247,1.588) (0.308,1.57) (0.37,1.551) (0.432,1.531) (0.493,1.509) (0.555,1.485) (0.617,1.46) (0.678,1.433) (0.74,1.404) (0.802,1.374) (0.863,1.341) (0.925,1.306) (0.987,1.268) (1.048,1.228) (1.11,1.185) (1.172,1.139) (1.233,1.09) (1.295,1.038) (1.357,0.982) (1.418,0.922) (1.48,0.858) (1.542,0.79) (1.603,0.716) (1.665,0.638) (1.727,0.555) (1.788,0.465) (1.85,0.37) (1.912,0.268) (1.973,0.159) (2.035,0.042)};
\draw[fwd] plot coordinates {(0,2.351) (0.062,2.385) (0.123,2.422) (0.185,2.462) (0.247,2.504) (0.308,2.549) (0.37,2.597) (0.432,2.649) (0.493,2.704) (0.555,2.762) (0.617,2.825) (0.678,2.893) (0.74,2.964) (0.802,3.041) (0.863,3.123) (0.925,3.211) (0.987,3.305) (1.048,3.405) (1.11,3.513) (1.172,3.627)};
\draw[fwd] plot coordinates {(0,1.349) (0.062,1.315) (0.123,1.278) (0.185,1.238) (0.247,1.196) (0.308,1.151) (0.37,1.103) (0.432,1.051) (0.493,0.996) (0.555,0.938) (0.617,0.875) (0.678,0.807) (0.74,0.736) (0.802,0.659) (0.863,0.577) (0.925,0.489) (0.987,0.395) (1.048,0.295) (1.11,0.187) (1.172,0.073)};
\draw[fwd] (0,1.85)--(3.7,1.85);
\node[] at (1.85,3.931) {\scriptsize $u_x+yu_y=0$};
\end{scope}
\end{tikzpicture}
\figcap{Characteristics of the source's two first-order examples: the level sets of $q=-3x-4y$ and of $q=e^{-x}y$. A solution is a function on the space of these curves.}
\end{center}
\section*{Notes}
The geometric theory of differential equations in jet spaces is due to Lie, Cartan, Spencer and Vinogradov; for the
categorical form see \cite{Vin01} and, for the comonad, \cite{Mar86,KhSc17}; a gap in the synthetic treatment
concerning limits of jet bundles is closed in \cite{GKSS26}. $\DD$-modules are treated in \cite{Cou95,HTT08}. The
fundamental principle is in \cite{Ehr70,Pal70}. The classification by type in the source follows the classical
reduction of quadratic forms.
\section*{Exercises}
\begin{exercise}\label{exer:inertia}
Find the type of $u_{xx}-4u_{xy}+4u_{yy}=0$, of $u_{xx}+2u_{xy}+5u_{yy}=0$ and of $u_{xy}=0$.
\end{exercise}
\begin{exercise}\label{exer:quot}
Find a function $q$ whose level sets are the characteristics of $u_x+2xy^2u_y=0$, and hence the general solution.
\end{exercise}
\begin{exercise}\label{exer:harmdim}
Verify \Cref{rem:type} in two variables directly, by exhibiting two independent harmonic polynomials of each degree
$k\ge1$ and two independent solutions of $u_{xx}=0$ of each degree.
\end{exercise}
\begin{exercise}\label{exer:factor}
Let $P=\pd_t-c\pd_x$ and $Q=\pd_t+c\pd_x$ act on smooth functions on the plane. Show that
$\Sol(PQ)=\Sol(P)+\Sol(Q)$.
\end{exercise}
\chapter{Three Archetypes and Well-Posedness}\label{ch:archetypes}
Three equations carry the source: the wave equation, the diffusion equation and Laplace's equation. This
chapter says where they come from, in a form that shows which of their properties are forced, and then
states what it means for a problem to be correctly set.
\section{Conservation and constitution}
Each derivation in the source \src{\S1.3} has two steps. The first is a balance: what is in a region changes only
by what crosses its boundary,
\begin{equation}\label{eq:balance}
\pd_t\rho+\Div j=0 .
\end{equation}
The second is a material law expressing the flux $j$ through the unknown.
\begin{center}\small
\begin{tabular}{@{}llll@{}}
\toprule
phenomenon & conserved & material law & equation\\
\midrule
transport & mass & $j=c\rho$ & $u_t+cu_x=0$\\
diffusion & mass, heat & $j=-k\nabla u$ (Fick, Fourier) & $u_t=k\Lap u$\\
vibration & momentum & stress $=T\nabla u$ (Hooke) & $\rho u_{tt}=T\Lap u$\\
equilibrium & --- & either of the last two, steady & $\Lap u=0$\\
\bottomrule
\end{tabular}
\end{center}
\begin{proposition}[Topological and metric parts]\label{prop:tonti}
\Srf[\cite{Ton75,Bos98}] The balance \eqref{eq:balance} is the statement that a differential form is closed: it
uses only the exterior derivative, which commutes with pullback along every smooth map. The material law is a
Hodge star: it uses a metric, and commutes only with isometries. The Laplacian is the composite
$\Lap=-\dd^*\dd$, with $\dd^*=\pm\star\,\dd\,\star$.
\end{proposition}
\noindent On a graph the same factorization is exact arithmetic: with $D$ the incidence matrix, the Laplacian is
$D^{\T}WD$ for a diagonal matrix $W$ of conductances, and the columns of $D^{\T}D$ sum to zero, which is conservation
of mass. \chk{W3} Two things follow at once. The symmetries of the three archetypes are the isometries, because the
only non-natural ingredient is the metric (\Cref{ch:harmonic,ch:symmetry}). And a discretization should discretize
$\dd$ exactly and $\star$ approximately (\Cref{ch:numerics}).
With signs suppressed, the proposition is the square
\[\begin{tikzcd}[column sep=huge,row sep=large]
\text{potential }u\arrow[r,"\dd"]\arrow[d,dashed,"\Lap"'] & \text{gradient }\dd u\arrow[d,"\star\ \text{(material)}"]\\
\text{source }f & \text{flux }q\arrow[l,"\dd"]
\end{tikzcd}\]
in which the horizontal arrows are natural for all smooth maps and the right-hand one only for isometries. Every
linear equation of the volume is a way round a square of this shape, sometimes with time as one of the variables.
\section{Data as a restriction map}
An equation has many solutions; a problem adds conditions \src{\S1.4}. Initial conditions prescribe $u$, and for
second order in time also $u_t$, on $\{t=0\}$. Boundary conditions prescribe $u$ (Dirichlet), the normal derivative
(Neumann) or a combination $\pd_nu+au$ (Robin) on the boundary of the spatial domain. In every case there is a
linear map
\[ r\colon\Sol(P;X)\longrightarrow Y \]
from solutions in a chosen function space $X$ to data in a chosen space $Y$.
\begin{definition}[Hadamard]\label{def:wellposed}
The problem $(P,r,X,Y)$ is \term{well-posed} if $r$ is an isomorphism of topological vector spaces: surjective
(existence), injective (uniqueness), with continuous inverse (stability).
\end{definition}
\noindent The definition is the source's \src{\S1.5} with one addition that the categorical phrasing makes
unavoidable: stability is a property of a morphism in a category of topological vector spaces, and it depends on
the category. A problem may be well-posed for one pair $(X,Y)$ and ill-posed for another.
\begin{example}[Kernel and cokernel]\label{ex:fredholm}
For the Neumann problem $-\Lap u=f$, $\pd_nu=0$, the map from $u$ to $f$ is not an isomorphism: constants are in
the kernel and a solution exists only if $\int f=0$. It is Fredholm of index zero, with an exact sequence
$0\to\R\to X\to Y\to\R\to0$. On a path with six vertices the rank is $5$ and the system is solvable exactly for
sources of mean zero. \chk{W5} The source's underdetermined and overdetermined problems are those with non-zero
kernel and cokernel; its first example, $u''+u=0$ with $u(0)=u(L)=0$, has a kernel exactly when $\sin L=0$. \chk{W6}
\end{example}
\begin{example}[Hadamard]\label{ex:hadamard}
The functions $u_n(x,y)=n^{-1}e^{-\sqrt n}\sin nx\sinh ny$ are harmonic, vanish on $y=0$, and have normal
derivative $e^{-\sqrt n}\sin nx$ there, which tends to zero with all its derivatives. At $y=1$ their size is
$e^{-\sqrt n}\sinh n/n$.
\begin{center}\small
\begin{tabular}{@{}lccc@{}}
\toprule
$n$ & $25$ & $100$ & $400$\\
\midrule
size of the data & $6.7\times10^{-3}$ & $4.5\times10^{-5}$ & $2.1\times10^{-9}$\\
size of the solution at $y=1$ & $9.7\times10^{6}$ & $6.1\times10^{36}$ & $1.3\times10^{162}$\\
\bottomrule
\end{tabular}
\end{center}
The Cauchy problem for Laplace's equation has at most one solution, and its inverse is not continuous in any
topology of finitely many derivatives. \chk{W1}
\end{example}
\begin{example}[Time's arrow]\label{ex:backward}
For $u_t=u_{xx}$ on $(0,\pi)$ the mode $\sin nx$ evolves by the factor $e^{-n^2t}$. Forwards this is a contraction.
Backwards over a time $0.1$ the modes $n=1,10,30$ are amplified by $1.1$, $2.2\times10^4$ and $1.2\times10^{39}$.
\chk{W2} The backward problem is solvable on a dense set of data and its solution map is unbounded: the heat
flow is a semigroup that is not a group (\Cref{ch:diffusions}).
\end{example}
\section{Uniqueness by energy}
\begin{theorem}[Two energy laws]\label{thm:energy}
\Scl[source, \S2.2--2.3] For the wave equation $u_{tt}=c^2u_{xx}$ on the line, with data of compact support, the energy
$E=\tfrac12\int(u_t^2+c^2u_x^2)\,dx$ is constant. For the diffusion equation $u_t=ku_{xx}$ with Dirichlet or Neumann
conditions, $\tfrac{d}{dt}\tfrac12\int u^2\,dx=-k\int u_x^2\,dx\le0$. In both cases a solution with zero data is zero.
\end{theorem}
\noindent On a grid of twenty points the wave energy drifted by $5.5\times10^{-12}$ in relative terms over five time
units, and the square norm under the diffusion flow decreased at every step. \chk{W4} The two laws are the first
appearance of the dichotomy of \Cref{ch:diffusions}: a conserved quadratic form makes the wave flow a group of
isometries; a decreasing one makes the heat flow a semigroup of contractions.
\section{Which problem for which type}
\begin{center}\small
\begin{tabular}{@{}llll@{}}
\toprule
type & archetype & well-posed problem & solution map\\
\midrule
elliptic & $\Lap u=0$ & values on the boundary & a projection (minimizer)\\
hyperbolic & $u_{tt}=c^2\Lap u$ & $u,u_t$ at one time & a group\\
parabolic & $u_t=k\Lap u$ & $u$ at one time, forwards & a semigroup\\
\bottomrule
\end{tabular}
\end{center}
\begin{center}
\begin{tikzpicture}
\begin{scope}
\draw[very thick,accent,fill=fwd!8] plot[smooth cycle,tension=0.8] coordinates {(0.2,0.3) (1.6,0.0) (2.8,0.6) (2.9,1.9) (1.7,2.5) (0.4,2.0)};
\node at (1.55,1.3) {\scriptsize $\Lap u=0$};
\node[accent] at (1.5,-0.4) {\scriptsize values on all of $\partial D$};
\node at (1.5,3.0) {\scriptsize elliptic};
\end{scope}
\begin{scope}[xshift=4.2cm]
\fill[fwd!8] (0,0) rectangle (3,2.5);
\draw[very thick,accent] (0,0)--(3,0);
\draw[fwd] (1.5,1.6)--(0.3,0) (1.5,1.6)--(2.7,0);
\fill (1.5,1.6) circle (1.4pt);
\draw[->,ink2] (-0.25,0)--(-0.25,2.5) node[above]{\scriptsize $t$};
\node at (1.5,2.1) {\scriptsize $u_{tt}=c^2u_{xx}$};
\node[accent] at (1.5,-0.4) {\scriptsize $u$ and $u_t$ at $t=0$};
\node at (1.5,3.0) {\scriptsize hyperbolic};
\end{scope}
\begin{scope}[xshift=8.4cm]
\fill[fwd!8] (0,0) rectangle (3,2.5);
\draw[very thick,accent] (0,2.5)--(0,0)--(3,0)--(3,2.5);
\draw[->,ink2] (-0.25,0)--(-0.25,2.5) node[above]{\scriptsize $t$};
\draw[->,fwd,thick] (1.5,0.5)--(1.5,1.9);
\node at (1.5,2.15) {\scriptsize $u_t=ku_{xx}$};
\node[accent] at (1.5,-0.4) {\scriptsize $u$ at $t=0$ and on the sides};
\node at (1.5,3.0) {\scriptsize parabolic};
\end{scope}
\end{tikzpicture}
\figcap{Where the data live. Elliptic: on the whole boundary. Hyperbolic: two functions on an initial slice, each point depending on an interval of it. Parabolic: one function on the initial slice and on the sides, in one direction of time only.}
\end{center}
\noindent Mismatched pairings fail. Initial data for Laplace's equation give \Cref{ex:hadamard}; final data for the
heat equation give \Cref{ex:backward}; boundary values for the wave equation on a square give non-uniqueness
(\Cref{exer:dirwave}). The rest of the volume takes the three rows in turn, and \Cref{ch:coda} returns to the
table.
\section*{Notes}
The separation of field equations into topological balance laws and metric constitutive laws is Tonti's
\cite{Ton75}; in the language of differential forms it is the basis of compatible discretization
\cite{Bos98,AFW06}. Well-posedness is Hadamard's notion; his example is in \cite{Had23}. Fredholm theory for
elliptic boundary problems is treated in \cite{Eva10}.
\section*{Exercises}
\begin{exercise}\label{exer:dirwave}
Show that the Dirichlet problem for $u_{tt}=u_{xx}$ on the square $[0,\pi]^2$ in the $(x,t)$-plane does not have
unique solutions.
\end{exercise}
\begin{exercise}\label{exer:neumann}
Let $D$ be the incidence matrix of a connected graph. Show that the range of $D^{\T}D$ is the space of vectors of
mean zero.
\end{exercise}
\begin{exercise}\label{exer:backerr}
An initial temperature on $(0,\pi)$ is to be recovered from a measurement at time $0.1$ whose error is $10^{-10}$
in each mode. Which is the first mode in which the error of the reconstruction exceeds $1$?
\end{exercise}
\begin{exercise}\label{exer:robin-energy}
For $u_t=ku_{xx}$ on $(0,l)$ with the Robin conditions $u_x(0)=a_0u(0)$ and $u_x(l)=-a_lu(l)$, compute
$\tfrac{d}{dt}\tfrac12\int u^2$. For which signs of $a_0,a_l$ is it certain to be non-positive?
\end{exercise}
\part{One Space Dimension}
\chapter{Waves}\label{ch:waves}
The wave equation on the line, $u_{tt}=c^2u_{xx}$, is the one archetype that can be solved in closed form by
looking at it. This chapter extracts four structures from the solution: a decomposition of the solution sheaf,
a causal structure, a symplectic form, and an action of reflection groups. Each reappears later in a setting
where no formula is available.
\section{Two transports glued along the constants}
\begin{theorem}[d'Alembert]\label{thm:dalembert}
\Scl[source, \S2.1] Let $q_\pm(x,t)=x\pm ct$. Every smooth solution on the plane is
$u=f\circ q_++g\circ q_-$, and $(f,g)$ is unique up to $(f+\kappa,g-\kappa)$:
\[ 0\longrightarrow\R\longrightarrow C^\infty(\R)\oplus C^\infty(\R)\xrightarrow{\ q_+^*+q_-^*\ }\Sol\longrightarrow0 \]
is exact. The solution with $u=\varphi$, $u_t=\psi$ at $t=0$ is
$u(x,t)=\tfrac12[\varphi(x+ct)+\varphi(x-ct)]+\tfrac1{2c}\int_{x-ct}^{x+ct}\psi$.
\end{theorem}
\begin{proof}
By \Cref{prop:factor} and \Cref{prop:char}: the two factors are transport operators with leaf spaces $q_\pm$, each
maps the solutions of the other onto themselves, and their common solutions are the constants. The formula is the
solution of two linear equations for $f'$ and $g'$. \chk{V1}
\end{proof}
\begin{corollary}[Parallelogram identity]\label{cor:par}
\Scl A function of the form $f\circ q_++g\circ q_-$ satisfies $u(A)+u(C)=u(B)+u(D)$ for the vertices $A,B,C,D$, in
order, of every parallelogram whose sides are characteristics. Conversely a continuous function with this property
is of that form.
\end{corollary}
\noindent The identity holds to rounding error for exact solutions. \chk{V2} It uses no derivative, so it defines
solutions that are merely continuous, or less: the first weak formulation of the volume. Restricted to a grid with
$c\,\Delta t=\Delta x$ it is an exact numerical scheme (\Cref{ch:numerics}).
\section{Causality}
For an interval $I$ of the $x$-axis let $D(I)$ be its \term{domain of dependence}: the points $(x,t)$ with
$[x-c|t|,x+c|t|]\subseteq I$, a diamond with waist $I$.
\begin{theorem}[Time slices]\label{thm:causal}
\Srf[source, \S2.2] Restriction to the waist is an isomorphism
\[ \Sol(D(I))\ \xrightarrow{\ \sim\ }\ C^\infty(I)\oplus C^\infty(I),\qquad u\mapsto(u,u_t)|_{t=0}, \]
natural for inclusions of intervals. Data outside $I$ do not affect the solution in $D(I)$, and data in $I$ affect
the solution only in the \term{domain of influence} of $I$.
\end{theorem}
\noindent Changing the data outside the interval of dependence of a point left the solution there unchanged to the
last bit; changing them inside did not. \chk{V3} Two remarks. The assignment $I\mapsto\Sol(D(I))$ is a functor on the
poset of intervals, and the theorem says it is determined by a much smaller amount of information than a
general sheaf on the plane would be; this is the property that field theory axiomatizes (\Cref{ch:fields}). And
the theorem has no counterpart for diffusion, where the solution at any point depends on all the data
(\Cref{ch:diffusions}).
\begin{center}
\begin{tikzpicture}[scale=0.78]
\begin{scope}
\fill[fwd!9] (-2.5,0)--(0,2.5)--(2.5,0)--(0,-2.5)--cycle;
\draw[thick,fwd] (-2.5,0)--(0,2.5)--(2.5,0)--(0,-2.5)--cycle;
\fill[fwd!24] (-0.6,0)--(0.4,1)--(1.4,0)--(0.4,-1)--cycle;
\draw[fwd] (-0.6,0)--(0.4,1)--(1.4,0)--(0.4,-1)--cycle;
\draw[->,ink2] (-3.1,0)--(3.3,0) node[right]{\small$x$};
\draw[->,ink2] (0,-2.8)--(0,3.0) node[above]{\small$t$};
\draw[very thick] (-2.5,0)--(2.5,0);
\node[below] at (-1.9,0) {\small$I$};
\node at (0.42,0.36) {\small$D(J)$};
\node at (-1.0,0.95) {\small$D(I)$};
\fill (1.25,1.0) circle (1.6pt) node[above right,inner sep=2pt]{\small$(x,t)$};
\draw[densely dashed] (1.25,1.0)--(0.25,0) (1.25,1.0)--(2.25,0);
\end{scope}
\begin{scope}[xshift=7.6cm,yshift=-1.2cm]
\fill[bwd!12] (-0.6,0)--(-3.0,2.4)--(3.0,2.4)--(0.6,0)--cycle;
\draw[thick,bwd] (-0.6,0)--(-3.0,2.4) (0.6,0)--(3.0,2.4);
\draw[->,ink2] (-3.2,0)--(3.4,0) node[right]{\small$x$};
\draw[->,ink2] (0,-0.3)--(0,3.0) node[above]{\small$t$};
\draw[very thick] (-0.6,0)--(0.6,0);
\node[below] at (0,0) {\small$K$};
\node at (0,1.6) {\small influence of $K$};
\end{scope}
\end{tikzpicture}
\figcap{Left: the diamond $D(I)$, the smaller diamond of a subinterval $J$, and the interval on which the value
at a point depends. Right: the region that data in $K$ can affect. Slopes are $\pm1/c$.}
\end{center}
\section{Energy and the symplectic form}
\begin{proposition}\label{prop:symp}
\Scl For solutions $u,v$ with spatially compact support,
\[ \Omega(u,v)=\int_{\R}(u\,v_t-u_t\,v)\,dx \]
does not depend on $t$. It is a symplectic form on $\Sol$; translations in space and time preserve it; and the
energy of \Cref{thm:energy} is $E(u)=-\tfrac12\Omega(u,\pd_tu)$, the Hamiltonian of time translation.
\end{proposition}
\begin{proof}
$\pd_t(uv_t-u_tv)=c^2(uv_{xx}-u_{xx}v)=c^2\pd_x(uv_x-u_xv)$, which integrates to zero. For the last statement
integrate $u\,u_{tt}=c^2u\,u_{xx}$ by parts.
\end{proof}
\noindent Both quantities were computed at several times and agree to eight digits. \chk{V4, V8} The wave flow is thus
a one-parameter group of linear symplectic maps. This is the whole of ``reversibility'' and ``no loss of
information'' in the source's comparison table.
\section{Reflections}
The reflection $\rho\colon x\mapsto-x$ preserves the equation, so $\Z/2$ acts on $\Sol(\R)$, which splits into even
and odd solutions.
\begin{theorem}[Boundary conditions as characters]\label{thm:images}
\Srf[source, \S3.2] Restriction to $x>0$ is an isomorphism from odd solutions on the line to solutions on the
half-line with the Dirichlet condition at $0$, and from even solutions to solutions with the Neumann condition,
for the wave equation.
More generally, let $W$ be a group generated by reflections of $\R^n$ with fundamental chamber $C$, let the
homogeneous wave, heat or Laplace equation be invariant under $W$, and let $\chi\colon W\to\{\pm1\}$ be a
character. Weak solutions on $\R^n$ with
$u\circ w=\chi(w)u$ restrict isomorphically to solutions on $C$ with the Dirichlet condition on the walls whose
reflection has $\chi=-1$ and the Neumann condition on those with $\chi=+1$.
\end{theorem}
\begin{proof}[Sketch]
An odd function vanishes on the mirror and an even one has vanishing normal derivative there. Conversely,
the extension by $u\circ w=\chi(w)u$ solves the equation in every chamber. At an odd wall the zero trace
matches the two values and the normal derivatives agree. At an even wall the values already agree and
the zero normal derivative makes the fluxes match. Thus no distribution supported on a wall appears
in the equation, as in \Cref{ch:distributions}.
\end{proof}
\noindent For the half-line this gives the source's two-region formula. \chk{V5} For an interval $(0,l)$ the group is
the infinite dihedral group generated by the reflections in $0$ and $l$, with chamber $(0,l)$. The product of the
two reflections is translation by $2l$, on which the Dirichlet--Dirichlet character is $+1$: solutions are
$2l$-periodic in $x$ and therefore $2l/c$-periodic in $t$. The source's formulas, one for each diamond of
space-time according to the number of reflections at each end, are the terms of the sum over the group; the sum
agrees with the Fourier series of \Cref{ch:fourier}. \chk{V6}
\begin{center}
\begin{tikzpicture}[scale=0.82]
\fill[fwd!10] (0,0) rectangle (2,3);
\foreach \x in {-2,0,2,4,6} \draw[ink2] (\x,0)--(\x,3);
\draw[->,ink2] (-2.5,0)--(6.6,0) node[right]{\small$x$};
\foreach \x/\s in {-1/$-$,1/$+$,3/$-$,5/$+$} \node at (\x,3.3) {\small\s};
\foreach \x/\s in {-2/$-l$,0/$0$,2/$l$,4/$2l$,6/$3l$} \node[below] at (\x,0) {\small\s};
\draw[thick,bwd,densely dashed] (0.5,0)--(3.5,3);
\draw[very thick,fwd] (0.5,0)--(2,1.5)--(0.5,3);
\fill (0.5,0) circle (1.6pt);
\end{tikzpicture}
\figcap{Unfolding. A characteristic reflected at a wall of the chamber $(0,l)$ (solid) is a straight characteristic on
the line (dashed), read in a translate of the chamber and weighted by the value of the character there; the signs
shown are those of the Dirichlet--Dirichlet character.}
\end{center}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.08) (0.068,1.111) (0.102,1.141) (0.136,1.172) (0.17,1.202) (0.204,1.233) (0.238,1.263) (0.272,1.293) (0.306,1.324) (0.34,1.354) (0.374,1.385) (0.408,1.415) (0.442,1.446) (0.476,1.476) (0.51,1.507) (0.544,1.537) (0.578,1.567) (0.612,1.598) (0.646,1.628) (0.68,1.659) (0.714,1.689) (0.748,1.72) (0.782,1.75) (0.816,1.78) (0.85,1.811) (0.884,1.841) (0.918,1.872) (0.952,1.902) (0.986,1.933) (1.02,1.963) (1.054,1.95) (1.088,1.937) (1.122,1.924) (1.156,1.911) (1.19,1.898) (1.224,1.885) (1.258,1.872) (1.292,1.859) (1.326,1.846) (1.36,1.833) (1.394,1.82) (1.428,1.807) (1.462,1.793) (1.496,1.78) (1.53,1.767) (1.564,1.754) (1.598,1.741) (1.632,1.728) (1.666,1.715) (1.7,1.702) (1.734,1.689) (1.768,1.676) (1.802,1.663) (1.836,1.65) (1.87,1.637) (1.904,1.624) (1.938,1.611) (1.972,1.598) (2.006,1.585) (2.04,1.572) (2.074,1.559) (2.108,1.546) (2.142,1.533) (2.176,1.52) (2.21,1.507) (2.244,1.493) (2.278,1.48) (2.312,1.467) (2.346,1.454) (2.38,1.441) (2.414,1.428) (2.448,1.415) (2.482,1.402) (2.516,1.389) (2.55,1.376) (2.584,1.363) (2.618,1.35) (2.652,1.337) (2.686,1.324) (2.72,1.311) (2.754,1.298) (2.788,1.285) (2.822,1.272) (2.856,1.259) (2.89,1.246) (2.924,1.233) (2.958,1.22) (2.992,1.207) (3.026,1.193) (3.06,1.18) (3.094,1.167) (3.128,1.154) (3.162,1.141) (3.196,1.128) (3.23,1.115) (3.264,1.102) (3.298,1.089) (3.332,1.076) (3.366,1.063) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=0$};
\end{scope}
\begin{scope}[shift={(3.95,0)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.08) (0.068,1.111) (0.102,1.141) (0.136,1.172) (0.17,1.202) (0.204,1.233) (0.238,1.263) (0.272,1.293) (0.306,1.324) (0.34,1.354) (0.374,1.363) (0.408,1.372) (0.442,1.38) (0.476,1.389) (0.51,1.398) (0.544,1.407) (0.578,1.415) (0.612,1.424) (0.646,1.433) (0.68,1.441) (0.714,1.45) (0.748,1.459) (0.782,1.467) (0.816,1.476) (0.85,1.485) (0.884,1.493) (0.918,1.502) (0.952,1.511) (0.986,1.52) (1.02,1.528) (1.054,1.537) (1.088,1.546) (1.122,1.554) (1.156,1.563) (1.19,1.572) (1.224,1.58) (1.258,1.589) (1.292,1.598) (1.326,1.607) (1.36,1.615) (1.394,1.624) (1.428,1.633) (1.462,1.641) (1.496,1.65) (1.53,1.659) (1.564,1.667) (1.598,1.676) (1.632,1.685) (1.666,1.693) (1.7,1.702) (1.734,1.689) (1.768,1.676) (1.802,1.663) (1.836,1.65) (1.87,1.637) (1.904,1.624) (1.938,1.611) (1.972,1.598) (2.006,1.585) (2.04,1.572) (2.074,1.559) (2.108,1.546) (2.142,1.533) (2.176,1.52) (2.21,1.507) (2.244,1.493) (2.278,1.48) (2.312,1.467) (2.346,1.454) (2.38,1.441) (2.414,1.428) (2.448,1.415) (2.482,1.402) (2.516,1.389) (2.55,1.376) (2.584,1.363) (2.618,1.35) (2.652,1.337) (2.686,1.324) (2.72,1.311) (2.754,1.298) (2.788,1.285) (2.822,1.272) (2.856,1.259) (2.89,1.246) (2.924,1.233) (2.958,1.22) (2.992,1.207) (3.026,1.193) (3.06,1.18) (3.094,1.167) (3.128,1.154) (3.162,1.141) (3.196,1.128) (3.23,1.115) (3.264,1.102) (3.298,1.089) (3.332,1.076) (3.366,1.063) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=0.2$};
\end{scope}
\begin{scope}[shift={(7.9,0)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.037) (0.068,1.024) (0.102,1.011) (0.136,0.998) (0.17,0.985) (0.204,0.972) (0.238,0.959) (0.272,0.946) (0.306,0.933) (0.34,0.92) (0.374,0.928) (0.408,0.937) (0.442,0.946) (0.476,0.954) (0.51,0.963) (0.544,0.972) (0.578,0.98) (0.612,0.989) (0.646,0.998) (0.68,1.007) (0.714,1.015) (0.748,1.024) (0.782,1.033) (0.816,1.041) (0.85,1.05) (0.884,1.059) (0.918,1.067) (0.952,1.076) (0.986,1.085) (1.02,1.093) (1.054,1.102) (1.088,1.111) (1.122,1.12) (1.156,1.128) (1.19,1.137) (1.224,1.146) (1.258,1.154) (1.292,1.163) (1.326,1.172) (1.36,1.18) (1.394,1.189) (1.428,1.198) (1.462,1.207) (1.496,1.215) (1.53,1.224) (1.564,1.233) (1.598,1.241) (1.632,1.25) (1.666,1.259) (1.7,1.267) (1.734,1.276) (1.768,1.285) (1.802,1.293) (1.836,1.302) (1.87,1.311) (1.904,1.32) (1.938,1.328) (1.972,1.337) (2.006,1.346) (2.04,1.354) (2.074,1.363) (2.108,1.372) (2.142,1.38) (2.176,1.389) (2.21,1.398) (2.244,1.407) (2.278,1.415) (2.312,1.424) (2.346,1.433) (2.38,1.441) (2.414,1.428) (2.448,1.415) (2.482,1.402) (2.516,1.389) (2.55,1.376) (2.584,1.363) (2.618,1.35) (2.652,1.337) (2.686,1.324) (2.72,1.311) (2.754,1.298) (2.788,1.285) (2.822,1.272) (2.856,1.259) (2.89,1.246) (2.924,1.233) (2.958,1.22) (2.992,1.207) (3.026,1.193) (3.06,1.18) (3.094,1.167) (3.128,1.154) (3.162,1.141) (3.196,1.128) (3.23,1.115) (3.264,1.102) (3.298,1.089) (3.332,1.076) (3.366,1.063) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=0.4$};
\end{scope}
\begin{scope}[shift={(0,-2.55)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.037) (0.068,1.024) (0.102,1.011) (0.136,0.998) (0.17,0.985) (0.204,0.972) (0.238,0.959) (0.272,0.946) (0.306,0.933) (0.34,0.92) (0.374,0.907) (0.408,0.893) (0.442,0.88) (0.476,0.867) (0.51,0.854) (0.544,0.841) (0.578,0.828) (0.612,0.815) (0.646,0.802) (0.68,0.789) (0.714,0.776) (0.748,0.763) (0.782,0.75) (0.816,0.737) (0.85,0.724) (0.884,0.711) (0.918,0.698) (0.952,0.685) (0.986,0.672) (1.02,0.659) (1.054,0.667) (1.088,0.676) (1.122,0.685) (1.156,0.693) (1.19,0.702) (1.224,0.711) (1.258,0.72) (1.292,0.728) (1.326,0.737) (1.36,0.746) (1.394,0.754) (1.428,0.763) (1.462,0.772) (1.496,0.78) (1.53,0.789) (1.564,0.798) (1.598,0.807) (1.632,0.815) (1.666,0.824) (1.7,0.833) (1.734,0.841) (1.768,0.85) (1.802,0.859) (1.836,0.867) (1.87,0.876) (1.904,0.885) (1.938,0.893) (1.972,0.902) (2.006,0.911) (2.04,0.92) (2.074,0.928) (2.108,0.937) (2.142,0.946) (2.176,0.954) (2.21,0.963) (2.244,0.972) (2.278,0.98) (2.312,0.989) (2.346,0.998) (2.38,1.007) (2.414,1.015) (2.448,1.024) (2.482,1.033) (2.516,1.041) (2.55,1.05) (2.584,1.059) (2.618,1.067) (2.652,1.076) (2.686,1.085) (2.72,1.093) (2.754,1.102) (2.788,1.111) (2.822,1.12) (2.856,1.128) (2.89,1.137) (2.924,1.146) (2.958,1.154) (2.992,1.163) (3.026,1.172) (3.06,1.18) (3.094,1.167) (3.128,1.154) (3.162,1.141) (3.196,1.128) (3.23,1.115) (3.264,1.102) (3.298,1.089) (3.332,1.076) (3.366,1.063) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=0.6$};
\end{scope}
\begin{scope}[shift={(3.95,-2.55)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.037) (0.068,1.024) (0.102,1.011) (0.136,0.998) (0.17,0.985) (0.204,0.972) (0.238,0.959) (0.272,0.946) (0.306,0.933) (0.34,0.92) (0.374,0.907) (0.408,0.893) (0.442,0.88) (0.476,0.867) (0.51,0.854) (0.544,0.841) (0.578,0.828) (0.612,0.815) (0.646,0.802) (0.68,0.789) (0.714,0.776) (0.748,0.763) (0.782,0.75) (0.816,0.737) (0.85,0.724) (0.884,0.711) (0.918,0.698) (0.952,0.685) (0.986,0.672) (1.02,0.659) (1.054,0.646) (1.088,0.633) (1.122,0.62) (1.156,0.607) (1.19,0.593) (1.224,0.58) (1.258,0.567) (1.292,0.554) (1.326,0.541) (1.36,0.528) (1.394,0.515) (1.428,0.502) (1.462,0.489) (1.496,0.476) (1.53,0.463) (1.564,0.45) (1.598,0.437) (1.632,0.424) (1.666,0.411) (1.7,0.398) (1.734,0.407) (1.768,0.415) (1.802,0.424) (1.836,0.433) (1.87,0.441) (1.904,0.45) (1.938,0.459) (1.972,0.467) (2.006,0.476) (2.04,0.485) (2.074,0.493) (2.108,0.502) (2.142,0.511) (2.176,0.52) (2.21,0.528) (2.244,0.537) (2.278,0.546) (2.312,0.554) (2.346,0.563) (2.38,0.572) (2.414,0.58) (2.448,0.589) (2.482,0.598) (2.516,0.607) (2.55,0.615) (2.584,0.624) (2.618,0.633) (2.652,0.641) (2.686,0.65) (2.72,0.659) (2.754,0.667) (2.788,0.676) (2.822,0.685) (2.856,0.693) (2.89,0.702) (2.924,0.711) (2.958,0.72) (2.992,0.728) (3.026,0.737) (3.06,0.746) (3.094,0.776) (3.128,0.807) (3.162,0.837) (3.196,0.867) (3.23,0.898) (3.264,0.928) (3.298,0.959) (3.332,0.989) (3.366,1.02) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=0.8$};
\end{scope}
\begin{scope}[shift={(7.9,-2.55)}]
\draw[ink2,very thin] (0,1.05)--(3.4,1.05);
\draw[thick,fwd] plot coordinates {(0,1.05) (0.034,1.037) (0.068,1.024) (0.102,1.011) (0.136,0.998) (0.17,0.985) (0.204,0.972) (0.238,0.959) (0.272,0.946) (0.306,0.933) (0.34,0.92) (0.374,0.907) (0.408,0.893) (0.442,0.88) (0.476,0.867) (0.51,0.854) (0.544,0.841) (0.578,0.828) (0.612,0.815) (0.646,0.802) (0.68,0.789) (0.714,0.776) (0.748,0.763) (0.782,0.75) (0.816,0.737) (0.85,0.724) (0.884,0.711) (0.918,0.698) (0.952,0.685) (0.986,0.672) (1.02,0.659) (1.054,0.646) (1.088,0.633) (1.122,0.62) (1.156,0.607) (1.19,0.593) (1.224,0.58) (1.258,0.567) (1.292,0.554) (1.326,0.541) (1.36,0.528) (1.394,0.515) (1.428,0.502) (1.462,0.489) (1.496,0.476) (1.53,0.463) (1.564,0.45) (1.598,0.437) (1.632,0.424) (1.666,0.411) (1.7,0.398) (1.734,0.385) (1.768,0.372) (1.802,0.359) (1.836,0.346) (1.87,0.333) (1.904,0.32) (1.938,0.307) (1.972,0.293) (2.006,0.28) (2.04,0.267) (2.074,0.254) (2.108,0.241) (2.142,0.228) (2.176,0.215) (2.21,0.202) (2.244,0.189) (2.278,0.176) (2.312,0.163) (2.346,0.15) (2.38,0.137) (2.414,0.167) (2.448,0.198) (2.482,0.228) (2.516,0.259) (2.55,0.289) (2.584,0.32) (2.618,0.35) (2.652,0.38) (2.686,0.411) (2.72,0.441) (2.754,0.472) (2.788,0.502) (2.822,0.533) (2.856,0.563) (2.89,0.593) (2.924,0.624) (2.958,0.654) (2.992,0.685) (3.026,0.715) (3.06,0.746) (3.094,0.776) (3.128,0.807) (3.162,0.837) (3.196,0.867) (3.23,0.898) (3.264,0.928) (3.298,0.959) (3.332,0.989) (3.366,1.02) (3.4,1.05)};
\fill (0,1.05) circle (1.3pt) (3.4,1.05) circle (1.3pt);
\node[] at (1.7,0.183) {\scriptsize $ct=1$};
\end{scope}
\end{tikzpicture}
\figcap{A string of unit length plucked at $x=0.3$, at six times in the first half period, computed from the sum over images. The corner splits in two, each half is reflected with a change of sign, and at $ct=1$ the shape is the initial one turned through half a turn.}
\end{center}
\begin{remark}[How far the method goes]\label{rem:reflection}
The source remarks that the reflection method cannot be generalized to two or three dimensions and does not work
for diffusion \src{\S3.2}. For the diamond-by-diamond formulas this is so. In the form of \Cref{thm:images} the method
applies to the reflection-invariant wave, heat and Laplace equations (\Cref{ch:diffusions}) and to every
domain that is a chamber of a reflection group: half-spaces, slabs, quadrants, wedges of angle $\pi/m$,
rectangles and boxes, and the equilateral, half-square and half-equilateral triangles. This construction
requires the domain to tile space by reflections.
\end{remark}
\section{Sources}
\begin{theorem}[Duhamel]\label{thm:duhamel-wave}
\Scl[source, \S3.4, Theorem 1] The solution of $u_{tt}-c^2u_{xx}=f$ with zero data at $t=0$ is
$u(x,t)=\frac1{2c}\iint_{\Delta(x,t)}f$, the integral over the triangle of dependence of $(x,t)$.
\end{theorem}
\noindent For $f=\cos x$ this is $\cos x\,(1-\cos ct)/c^2$. \chk{V7} The theorem defines the \term{retarded} operator
$G^+\colon f\mapsto u$, whose value at a point depends only on $f$ in the past cone. Reversing time gives the
advanced operator $G^-$, and the difference $G=G^+-G^-$ maps sources to solutions.
\begin{theorem}[Solutions are radiated]\label{thm:bgp}
\Sli[\cite{BGP07}] With $P=\pd_t^2-c^2\pd_x^2$, the sequence
\[ 0\longrightarrow C_c^\infty\xrightarrow{\ P\ }C_c^\infty\xrightarrow{\ G\ }\Sol_{sc}\longrightarrow0 \]
is exact, where $\Sol_{sc}$ denotes solutions with compactly supported Cauchy data. In these terms, and with
suitable sign conventions, $\Omega(Gf,Gh)=\iint f\,Gh$.
\end{theorem}
\noindent Every solution is the field of some source, and two sources give the same field exactly when they differ
by $P$ of a compactly supported function. The theorem holds on every globally hyperbolic space-time, where there
is no d'Alembert formula; it is the form in which the four structures of this chapter survive.
\section*{Notes}
The exposition follows the source's Chapter 2 and Sections 3.2 and 3.4. For the wave equation on Lorentzian
manifolds, Green operators and the exact sequence, see \cite{BGP07}. Reflection groups and their chambers are
treated in \cite{Hum90}; the triangles that tile the plane by reflection are Lam\'e's, see \cite{McC11}.
\section*{Exercises}
\begin{exercise}\label{exer:par}
Prove the first half of \Cref{cor:par}.
\end{exercise}
\begin{exercise}\label{exer:neuwave}
For the half-line with the Neumann condition and data $u=\varphi$, $u_t=0$, write the solution in the region
$0<x<ct$.
\end{exercise}
\begin{exercise}\label{exer:period}
A string on $(0,l)$ is fixed at $0$ and free at $l$. Show from \Cref{thm:images} that every motion has period
$4l/c$.
\end{exercise}
\begin{exercise}\label{exer:duh}
Verify \Cref{thm:duhamel-wave} for $f(x,t)=\cos x$ by computing the integral.
\end{exercise}
\chapter{Diffusions}\label{ch:diffusions}
The diffusion equation $u_t=ku_{xx}$ has an explicit solution on the line as well, but the structures it carries
are different in kind from those of \Cref{ch:waves}: order where there was a symplectic form, a semigroup where
there was a group, probability where there was mechanics.
\section{A semigroup of Markov kernels}
Let $S(x,t)=(4\pi kt)^{-1/2}e^{-x^2/4kt}$, the source function, and $(T_t\varphi)(x)=\int S(x-y,t)\varphi(y)\,dy$.
\begin{theorem}\label{thm:semigroup}
\Scl[source, \S2.4 and \S3.5] For bounded continuous $\varphi$, $u=T_t\varphi$ is the bounded solution of the
diffusion equation with $u\to\varphi$ as $t\to0$. Moreover
\begin{enumerate}[label=(\alph*)]
\item $T_{t+s}=T_tT_s$ and $T_0=\id$;
\item $T_t\varphi\ge0$ if $\varphi\ge0$, and $T_t1=1$;
\item $\inf\varphi\le T_t\varphi\le\sup\varphi$.
\end{enumerate}
\end{theorem}
\noindent Properties (a) and (b) say that $t\mapsto T_t$ is a functor from the monoid $(\R_{\ge0},+)$, regarded as a
category with one object, to the category $\Stoch$ of Markov kernels of \cite{ESCHC}. Property (c), the maximum
principle on the line, is a consequence of (b): an average lies between the extremes. The kernel solves the
equation, has unit mass and composes correctly to twelve digits; a unit step diffuses into
$\tfrac12(1+\erf(x/\sqrt{4kt}))$. \chk{D1, D2}
\begin{theorem}[Maximum principle]\label{thm:max}
\Scl[source, \S2.3] A solution in a space-time rectangle attains its maximum and its minimum on the initial side or
the lateral sides. Hence the Dirichlet problem has at most one solution and depends continuously on its data in
the uniform norm.
\end{theorem}
\noindent The discrete form is exact: the explicit difference scheme is a substochastic matrix precisely when
$k\Delta t/\Delta x^2\le\tfrac12$, and then it obeys the maximum principle; beyond that it has negative entries and
the sawtooth grows. \chk{D3} Stability of that scheme, the first example in the source's chapter on computation, is
membership in the Markov category (\Cref{ch:numerics}).
\begin{remark}[Generators]
\Scl[\cite{Paz83}] A strongly continuous semigroup of contractions on a Banach space has a closed, densely defined
generator $A=\lim(T_t-1)/t$, and the Hille--Yosida theorem characterizes the operators that arise: those with
$\|(\lambda-A)^{-1}\|\le1/\lambda$ for $\lambda>0$. The diffusion equation is the statement that the generator of
$(T_t)$ is $k\,\pd_x^2$. The resolvent is the Laplace transform of the semigroup (\Cref{ch:distributions}).
\end{remark}
\section{Symmetry and the Gaussian}
The equation is invariant under translations, the reflection $x\mapsto-x$, and the scaling
$(x,t)\mapsto(ax,a^2t)$ \src{\S2.4}. The last is not an isometry; it exists because the equation is of first order
in time and second in space.
\begin{proposition}\label{prop:gauss}
\Scl The source function is, up to a constant factor, the unique solution of the form
$t^{-1/2}g(x/\sqrt t)$: the solution of unit mass fixed by the scaling action. \chk{D7}
\end{proposition}
\noindent Convergence of $T_t\varphi$ to a multiple of $S$ as $t\to\infty$, for integrable $\varphi$, is the statement
that this fixed point attracts; it is the central limit theorem in the language of \cite{ESCHC}.
\section{Irreversibility}
\begin{theorem}[Monotone quantities]\label{thm:mono}
\Scl Along the diffusion flow of a probability density $u$:
\begin{enumerate}[label=(\alph*)]
\item $\frac d{dt}\int u^2=-2k\int u_x^2\le0$;
\item $\frac d{dt}\bigl(-\int u\log u\bigr)=k\int u_x^2/u\ge0$ (entropy increases at the rate of the Fisher
information);
\item for two solutions, every relative entropy $\KL{u}{v}$ and every $f$-divergence is non-increasing.
\end{enumerate}
Moreover $T_t\varphi$ is smooth for every $t>0$, and positive everywhere if $\varphi\ge0$ is not zero.
\end{theorem}
\noindent Statement (c) is the data-processing inequality: it holds for every morphism of $\Stoch$ \cite{ESCHC}. The
invertible morphisms of $\Stoch$ are the deterministic bijections, and $T_t$ is not one of them; that is the
categorical form of irreversibility. Entropy and square norm were tested under a doubly stochastic step on fifty
random densities. \chk{D4} With $k=1$ the indicator function of $[-1,1]$ gives the positive value
$2.3\times10^{-90}$ at $x=10$ after time $0.1$: nothing travels, and everything is everywhere at once. \chk{D8}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3)--(0,0)--(5.2,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-4$};
\draw[ink2] (1.3,0)--(1.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (2.6,0)--(2.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.9,0)--(3.9,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (5.2,0)--(5.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.111)--(-0.07,1.111) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (0,2.222)--(-0.07,2.222) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (2.6,-0.38) {\scriptsize $x$};
\draw[thick,bwd] plot coordinates {(0,0) (0.032,0) (0.065,0) (0.097,0) (0.13,0) (0.163,0) (0.195,0) (0.228,0) (0.26,0) (0.293,0) (0.325,0) (0.357,0) (0.39,0) (0.422,0) (0.455,0) (0.488,0) (0.52,0) (0.553,0) (0.585,0) (0.618,0) (0.65,0) (0.682,0) (0.715,0) (0.747,0) (0.78,0) (0.812,0) (0.845,0) (0.878,0) (0.91,0) (0.943,0) (0.975,0) (1.007,0) (1.04,0) (1.073,0) (1.105,0) (1.137,0) (1.17,0) (1.203,0) (1.235,0) (1.268,0) (1.3,0) (1.333,0) (1.365,0) (1.397,0) (1.43,0) (1.463,0) (1.495,0) (1.528,0) (1.56,0) (1.593,0) (1.625,0) (1.657,0) (1.69,0) (1.722,0) (1.755,0.001) (1.788,0.001) (1.82,0.002) (1.853,0.004) (1.885,0.007) (1.918,0.011) (1.95,0.019) (1.982,0.031) (2.015,0.049) (2.047,0.076) (2.08,0.114) (2.113,0.168) (2.145,0.242) (2.178,0.339) (2.21,0.463) (2.243,0.618) (2.275,0.803) (2.308,1.019) (2.34,1.26) (2.373,1.519) (2.405,1.788) (2.438,2.051) (2.47,2.295) (2.502,2.505) (2.535,2.667) (2.568,2.769) (2.6,2.803) (2.632,2.769) (2.665,2.667) (2.698,2.505) (2.73,2.295) (2.763,2.051) (2.795,1.788) (2.827,1.519) (2.86,1.26) (2.893,1.019) (2.925,0.803) (2.958,0.618) (2.99,0.463) (3.023,0.339) (3.055,0.242) (3.087,0.168) (3.12,0.114) (3.152,0.076) (3.185,0.049) (3.218,0.031) (3.25,0.019) (3.283,0.011) (3.315,0.007) (3.348,0.004) (3.38,0.002) (3.413,0.001) (3.445,0.001) (3.478,0) (3.51,0) (3.543,0) (3.575,0) (3.607,0) (3.64,0) (3.673,0) (3.705,0) (3.738,0) (3.77,0) (3.802,0) (3.835,0) (3.868,0) (3.9,0) (3.933,0) (3.965,0) (3.998,0) (4.03,0) (4.062,0) (4.095,0) (4.127,0) (4.16,0) (4.192,0) (4.225,0) (4.258,0) (4.29,0) (4.323,0) (4.355,0) (4.388,0) (4.42,0) (4.452,0) (4.485,0) (4.518,0) (4.55,0) (4.583,0) (4.615,0) (4.647,0) (4.68,0) (4.713,0) (4.745,0) (4.777,0) (4.81,0) (4.843,0) (4.875,0) (4.907,0) (4.94,0) (4.973,0) (5.005,0) (5.038,0) (5.07,0) (5.103,0) (5.135,0) (5.168,0) (5.2,0)};
\draw[thick,fwd] plot coordinates {(0,0) (0.032,0) (0.065,0) (0.097,0) (0.13,0) (0.163,0) (0.195,0) (0.228,0) (0.26,0) (0.293,0) (0.325,0) (0.357,0) (0.39,0) (0.422,0) (0.455,0) (0.488,0) (0.52,0) (0.553,0) (0.585,0) (0.618,0) (0.65,0) (0.682,0) (0.715,0) (0.747,0) (0.78,0) (0.812,0) (0.845,0) (0.878,0) (0.91,0) (0.943,0) (0.975,0.001) (1.007,0.001) (1.04,0.001) (1.073,0.001) (1.105,0.002) (1.137,0.003) (1.17,0.003) (1.203,0.004) (1.235,0.006) (1.268,0.007) (1.3,0.009) (1.333,0.012) (1.365,0.015) (1.397,0.019) (1.43,0.024) (1.463,0.03) (1.495,0.038) (1.528,0.047) (1.56,0.057) (1.593,0.07) (1.625,0.084) (1.657,0.101) (1.69,0.121) (1.722,0.144) (1.755,0.17) (1.788,0.199) (1.82,0.232) (1.853,0.268) (1.885,0.309) (1.918,0.353) (1.95,0.402) (1.982,0.454) (2.015,0.509) (2.047,0.568) (2.08,0.63) (2.113,0.694) (2.145,0.76) (2.178,0.827) (2.21,0.894) (2.243,0.96) (2.275,1.026) (2.308,1.088) (2.34,1.148) (2.373,1.203) (2.405,1.253) (2.438,1.296) (2.47,1.333) (2.502,1.363) (2.535,1.384) (2.568,1.397) (2.6,1.402) (2.632,1.397) (2.665,1.384) (2.698,1.363) (2.73,1.333) (2.763,1.296) (2.795,1.253) (2.827,1.203) (2.86,1.148) (2.893,1.088) (2.925,1.026) (2.958,0.96) (2.99,0.894) (3.023,0.827) (3.055,0.76) (3.087,0.694) (3.12,0.63) (3.152,0.568) (3.185,0.509) (3.218,0.454) (3.25,0.402) (3.283,0.353) (3.315,0.309) (3.348,0.268) (3.38,0.232) (3.413,0.199) (3.445,0.17) (3.478,0.144) (3.51,0.121) (3.543,0.101) (3.575,0.084) (3.607,0.07) (3.64,0.057) (3.673,0.047) (3.705,0.038) (3.738,0.03) (3.77,0.024) (3.802,0.019) (3.835,0.015) (3.868,0.012) (3.9,0.009) (3.933,0.007) (3.965,0.006) (3.998,0.004) (4.03,0.003) (4.062,0.003) (4.095,0.002) (4.127,0.001) (4.16,0.001) (4.192,0.001) (4.225,0.001) (4.258,0) (4.29,0) (4.323,0) (4.355,0) (4.388,0) (4.42,0) (4.452,0) (4.485,0) (4.518,0) (4.55,0) (4.583,0) (4.615,0) (4.647,0) (4.68,0) (4.713,0) (4.745,0) (4.777,0) (4.81,0) (4.843,0) (4.875,0) (4.907,0) (4.94,0) (4.973,0) (5.005,0) (5.038,0) (5.07,0) (5.103,0) (5.135,0) (5.168,0) (5.2,0)};
\draw[thick,ink2] plot coordinates {(0,0.011) (0.032,0.013) (0.065,0.014) (0.097,0.015) (0.13,0.017) (0.163,0.019) (0.195,0.02) (0.228,0.022) (0.26,0.025) (0.293,0.027) (0.325,0.029) (0.357,0.032) (0.39,0.035) (0.422,0.038) (0.455,0.041) (0.488,0.045) (0.52,0.048) (0.553,0.052) (0.585,0.057) (0.618,0.061) (0.65,0.066) (0.682,0.071) (0.715,0.077) (0.747,0.082) (0.78,0.088) (0.812,0.095) (0.845,0.101) (0.878,0.108) (0.91,0.116) (0.943,0.123) (0.975,0.131) (1.007,0.14) (1.04,0.149) (1.073,0.158) (1.105,0.167) (1.137,0.177) (1.17,0.187) (1.203,0.197) (1.235,0.208) (1.268,0.219) (1.3,0.231) (1.333,0.242) (1.365,0.254) (1.397,0.266) (1.43,0.279) (1.463,0.292) (1.495,0.304) (1.528,0.317) (1.56,0.331) (1.593,0.344) (1.625,0.357) (1.657,0.371) (1.69,0.384) (1.722,0.397) (1.755,0.411) (1.788,0.424) (1.82,0.437) (1.853,0.45) (1.885,0.463) (1.918,0.476) (1.95,0.488) (1.982,0.5) (2.015,0.512) (2.047,0.523) (2.08,0.534) (2.113,0.545) (2.145,0.555) (2.178,0.564) (2.21,0.573) (2.243,0.581) (2.275,0.589) (2.308,0.596) (2.34,0.602) (2.373,0.608) (2.405,0.613) (2.438,0.617) (2.47,0.621) (2.502,0.623) (2.535,0.625) (2.568,0.626) (2.6,0.627) (2.632,0.626) (2.665,0.625) (2.698,0.623) (2.73,0.621) (2.763,0.617) (2.795,0.613) (2.827,0.608) (2.86,0.602) (2.893,0.596) (2.925,0.589) (2.958,0.581) (2.99,0.573) (3.023,0.564) (3.055,0.555) (3.087,0.545) (3.12,0.534) (3.152,0.523) (3.185,0.512) (3.218,0.5) (3.25,0.488) (3.283,0.476) (3.315,0.463) (3.348,0.45) (3.38,0.437) (3.413,0.424) (3.445,0.411) (3.478,0.397) (3.51,0.384) (3.543,0.371) (3.575,0.357) (3.607,0.344) (3.64,0.331) (3.673,0.317) (3.705,0.304) (3.738,0.292) (3.77,0.279) (3.802,0.266) (3.835,0.254) (3.868,0.242) (3.9,0.231) (3.933,0.219) (3.965,0.208) (3.998,0.197) (4.03,0.187) (4.062,0.177) (4.095,0.167) (4.127,0.158) (4.16,0.149) (4.192,0.14) (4.225,0.131) (4.258,0.123) (4.29,0.116) (4.323,0.108) (4.355,0.101) (4.388,0.095) (4.42,0.088) (4.452,0.082) (4.485,0.077) (4.518,0.071) (4.55,0.066) (4.583,0.061) (4.615,0.057) (4.647,0.052) (4.68,0.048) (4.713,0.045) (4.745,0.041) (4.777,0.038) (4.81,0.035) (4.843,0.032) (4.875,0.029) (4.907,0.027) (4.94,0.025) (4.973,0.022) (5.005,0.02) (5.038,0.019) (5.07,0.017) (5.103,0.015) (5.135,0.014) (5.168,0.013) (5.2,0.011)};
\node[bwd,right] at (3.087,2.667) {\scriptsize $t=0.05$};
\node[fwd,right] at (3.185,1.222) {\scriptsize $0.2$};
\node[ink2,right] at (3.9,0.489) {\scriptsize $1$};
\end{scope}
\begin{scope}[shift={(6.6,0)}]
\draw[ink2] (0,3)--(0,0)--(5.2,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-4$};
\draw[ink2] (1.3,0)--(1.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (2.6,0)--(2.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.9,0)--(3.9,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (5.2,0)--(5.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.429)--(-0.07,1.429) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (0,2.857)--(-0.07,2.857) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (2.6,-0.38) {\scriptsize $x$};
\draw[thick,bwd] plot coordinates {(0,0) (0.032,0) (0.065,0) (0.097,0) (0.13,0) (0.163,0) (0.195,0) (0.228,0) (0.26,0) (0.293,0) (0.325,0) (0.357,0) (0.39,0) (0.422,0) (0.455,0) (0.488,0) (0.52,0) (0.553,0) (0.585,0) (0.618,0) (0.65,0) (0.682,0) (0.715,0) (0.747,0) (0.78,0) (0.812,0) (0.845,0) (0.878,0) (0.91,0) (0.943,0) (0.975,0) (1.007,0) (1.04,0) (1.073,0) (1.105,0) (1.137,0) (1.17,0) (1.203,0) (1.235,0) (1.268,0) (1.3,0) (1.333,0) (1.365,0) (1.397,0) (1.43,0) (1.463,0) (1.495,0) (1.528,0) (1.56,0) (1.593,0) (1.625,0) (1.657,0) (1.69,0) (1.722,0) (1.755,0) (1.788,0) (1.82,0) (1.853,0) (1.885,0.001) (1.918,0.001) (1.95,0.002) (1.982,0.004) (2.015,0.006) (2.047,0.01) (2.08,0.016) (2.113,0.025) (2.145,0.038) (2.178,0.057) (2.21,0.083) (2.243,0.117) (2.275,0.163) (2.308,0.221) (2.34,0.294) (2.373,0.383) (2.405,0.49) (2.438,0.613) (2.47,0.753) (2.502,0.908) (2.535,1.074) (2.568,1.249) (2.6,1.429) (2.632,1.608) (2.665,1.783) (2.698,1.95) (2.73,2.104) (2.763,2.244) (2.795,2.367) (2.827,2.474) (2.86,2.563) (2.893,2.636) (2.925,2.695) (2.958,2.74) (2.99,2.775) (3.023,2.8) (3.055,2.819) (3.087,2.832) (3.12,2.841) (3.152,2.847) (3.185,2.851) (3.218,2.853) (3.25,2.855) (3.283,2.856) (3.315,2.856) (3.348,2.857) (3.38,2.857) (3.413,2.857) (3.445,2.857) (3.478,2.857) (3.51,2.857) (3.543,2.857) (3.575,2.857) (3.607,2.857) (3.64,2.857) (3.673,2.857) (3.705,2.857) (3.738,2.857) (3.77,2.857) (3.802,2.857) (3.835,2.857) (3.868,2.857) (3.9,2.857) (3.933,2.857) (3.965,2.857) (3.998,2.857) (4.03,2.857) (4.062,2.857) (4.095,2.857) (4.127,2.857) (4.16,2.857) (4.192,2.857) (4.225,2.857) (4.258,2.857) (4.29,2.857) (4.323,2.857) (4.355,2.857) (4.388,2.857) (4.42,2.857) (4.452,2.857) (4.485,2.857) (4.518,2.857) (4.55,2.857) (4.583,2.857) (4.615,2.857) (4.647,2.857) (4.68,2.857) (4.713,2.857) (4.745,2.857) (4.777,2.857) (4.81,2.857) (4.843,2.857) (4.875,2.857) (4.907,2.857) (4.94,2.857) (4.973,2.857) (5.005,2.857) (5.038,2.857) (5.07,2.857) (5.103,2.857) (5.135,2.857) (5.168,2.857) (5.2,2.857)};
\draw[thick,fwd] plot coordinates {(0,0) (0.032,0) (0.065,0) (0.097,0) (0.13,0) (0.163,0) (0.195,0) (0.228,0) (0.26,0) (0.293,0) (0.325,0) (0.357,0) (0.39,0) (0.422,0) (0.455,0) (0.488,0) (0.52,0) (0.553,0) (0.585,0) (0.618,0) (0.65,0) (0.682,0) (0.715,0) (0.747,0) (0.78,0) (0.812,0) (0.845,0) (0.878,0) (0.91,0) (0.943,0) (0.975,0) (1.007,0) (1.04,0) (1.073,0) (1.105,0) (1.137,0.001) (1.17,0.001) (1.203,0.001) (1.235,0.001) (1.268,0.002) (1.3,0.002) (1.333,0.003) (1.365,0.004) (1.397,0.005) (1.43,0.006) (1.463,0.008) (1.495,0.01) (1.528,0.013) (1.56,0.016) (1.593,0.02) (1.625,0.025) (1.657,0.031) (1.69,0.038) (1.722,0.047) (1.755,0.057) (1.788,0.069) (1.82,0.083) (1.853,0.099) (1.885,0.117) (1.918,0.138) (1.95,0.163) (1.982,0.19) (2.015,0.221) (2.047,0.256) (2.08,0.294) (2.113,0.337) (2.145,0.383) (2.178,0.434) (2.21,0.49) (2.243,0.549) (2.275,0.613) (2.308,0.681) (2.34,0.753) (2.373,0.829) (2.405,0.908) (2.438,0.989) (2.47,1.074) (2.502,1.161) (2.535,1.249) (2.568,1.339) (2.6,1.429) (2.632,1.519) (2.665,1.608) (2.698,1.696) (2.73,1.783) (2.763,1.868) (2.795,1.95) (2.827,2.029) (2.86,2.104) (2.893,2.176) (2.925,2.244) (2.958,2.308) (2.99,2.367) (3.023,2.423) (3.055,2.474) (3.087,2.52) (3.12,2.563) (3.152,2.601) (3.185,2.636) (3.218,2.667) (3.25,2.695) (3.283,2.719) (3.315,2.74) (3.348,2.759) (3.38,2.775) (3.413,2.788) (3.445,2.8) (3.478,2.81) (3.51,2.819) (3.543,2.826) (3.575,2.832) (3.607,2.837) (3.64,2.841) (3.673,2.844) (3.705,2.847) (3.738,2.849) (3.77,2.851) (3.802,2.852) (3.835,2.853) (3.868,2.854) (3.9,2.855) (3.933,2.855) (3.965,2.856) (3.998,2.856) (4.03,2.856) (4.062,2.857) (4.095,2.857) (4.127,2.857) (4.16,2.857) (4.192,2.857) (4.225,2.857) (4.258,2.857) (4.29,2.857) (4.323,2.857) (4.355,2.857) (4.388,2.857) (4.42,2.857) (4.452,2.857) (4.485,2.857) (4.518,2.857) (4.55,2.857) (4.583,2.857) (4.615,2.857) (4.647,2.857) (4.68,2.857) (4.713,2.857) (4.745,2.857) (4.777,2.857) (4.81,2.857) (4.843,2.857) (4.875,2.857) (4.907,2.857) (4.94,2.857) (4.973,2.857) (5.005,2.857) (5.038,2.857) (5.07,2.857) (5.103,2.857) (5.135,2.857) (5.168,2.857) (5.2,2.857)};
\draw[thick,ink2] plot coordinates {(0,0.007) (0.032,0.007) (0.065,0.008) (0.097,0.009) (0.13,0.01) (0.163,0.011) (0.195,0.013) (0.228,0.014) (0.26,0.016) (0.293,0.017) (0.325,0.019) (0.357,0.021) (0.39,0.023) (0.422,0.025) (0.455,0.028) (0.488,0.031) (0.52,0.034) (0.553,0.037) (0.585,0.041) (0.618,0.044) (0.65,0.048) (0.682,0.053) (0.715,0.058) (0.747,0.063) (0.78,0.068) (0.812,0.074) (0.845,0.08) (0.878,0.087) (0.91,0.094) (0.943,0.102) (0.975,0.11) (1.007,0.119) (1.04,0.128) (1.073,0.138) (1.105,0.148) (1.137,0.159) (1.17,0.171) (1.203,0.183) (1.235,0.197) (1.268,0.21) (1.3,0.225) (1.333,0.24) (1.365,0.256) (1.397,0.273) (1.43,0.29) (1.463,0.308) (1.495,0.328) (1.528,0.348) (1.56,0.368) (1.593,0.39) (1.625,0.413) (1.657,0.436) (1.69,0.46) (1.722,0.485) (1.755,0.511) (1.788,0.538) (1.82,0.566) (1.853,0.594) (1.885,0.624) (1.918,0.654) (1.95,0.685) (1.982,0.717) (2.015,0.749) (2.047,0.783) (2.08,0.817) (2.113,0.851) (2.145,0.887) (2.178,0.923) (2.21,0.959) (2.243,0.996) (2.275,1.034) (2.308,1.072) (2.34,1.11) (2.373,1.149) (2.405,1.189) (2.438,1.228) (2.47,1.268) (2.502,1.308) (2.535,1.348) (2.568,1.388) (2.6,1.429) (2.632,1.469) (2.665,1.509) (2.698,1.549) (2.73,1.589) (2.763,1.629) (2.795,1.669) (2.827,1.708) (2.86,1.747) (2.893,1.785) (2.925,1.823) (2.958,1.861) (2.99,1.898) (3.023,1.935) (3.055,1.971) (3.087,2.006) (3.12,2.041) (3.152,2.075) (3.185,2.108) (3.218,2.14) (3.25,2.172) (3.283,2.203) (3.315,2.233) (3.348,2.263) (3.38,2.291) (3.413,2.319) (3.445,2.346) (3.478,2.372) (3.51,2.397) (3.543,2.421) (3.575,2.445) (3.607,2.467) (3.64,2.489) (3.673,2.51) (3.705,2.53) (3.738,2.549) (3.77,2.567) (3.802,2.585) (3.835,2.601) (3.868,2.617) (3.9,2.632) (3.933,2.647) (3.965,2.661) (3.998,2.674) (4.03,2.686) (4.062,2.698) (4.095,2.709) (4.127,2.719) (4.16,2.729) (4.192,2.738) (4.225,2.747) (4.258,2.755) (4.29,2.763) (4.323,2.77) (4.355,2.777) (4.388,2.783) (4.42,2.789) (4.452,2.794) (4.485,2.8) (4.518,2.804) (4.55,2.809) (4.583,2.813) (4.615,2.817) (4.647,2.82) (4.68,2.823) (4.713,2.826) (4.745,2.829) (4.777,2.832) (4.81,2.834) (4.843,2.836) (4.875,2.838) (4.907,2.84) (4.94,2.842) (4.973,2.843) (5.005,2.844) (5.038,2.846) (5.07,2.847) (5.103,2.848) (5.135,2.849) (5.168,2.85) (5.2,2.85)};
\end{scope}
\end{tikzpicture}
\figcap{Left: the source function $S(x,t)$ with $k=1$ at three times; the area under each curve is one. Right: the same kernel applied to a unit step. Every curve on the right is strictly between $0$ and $1$ at every point: the maximum principle, and infinite speed.}
\end{center}
\section{Waves and diffusions compared}
The source closes its second chapter with a table of seven contrasts \src{\S2.5}. All seven follow from three.
\begin{center}\small
\begin{tabular}{@{}>{\raggedright}p{0.25\textwidth}>{\raggedright}p{0.3\textwidth}>{\raggedright\arraybackslash}p{0.3\textwidth}@{}}
\toprule
& waves & diffusions\\
\midrule
time acts by & a group & a semigroup\\
what is preserved & a symplectic form; energy & positivity; total mass\\
target category & linear symplectic maps & Markov kernels\\
\midrule
speed of propagation & finite: \Cref{thm:causal} & infinite: positivity of $T_t$\\
singularities & transported along characteristics & erased: $T_t$ smooths\\
backwards in time & well-posed: inverses exist & ill-posed: \Cref{ex:backward}\\
maximum principle & no & yes: \Cref{thm:max}\\
as $t\to\infty$ & energy constant & decay to the fixed point\\
information & kept & lost: \Cref{thm:mono}(c)\\
\bottomrule
\end{tabular}
\end{center}
\noindent The upper three rows are structure; the lower six are the source's, and each is a consequence. A
functor from time to symplectic maps is automatically invertible and conserves a quadratic form. A functor from
time to Markov kernels is automatically monotone and contracts every divergence. The Schr\"odinger equation
(\Cref{ch:space}) sits between the two: a group, like waves, generated by the operator of diffusion.
\section{Images, again}
\Cref{thm:images} applies without change, since the equation is invariant under reflections.
\begin{theorem}[Kernels on the half-line and the interval]\label{thm:kernels}
\Srf[source, \S3.1] The Dirichlet and Neumann kernels on the half-line are $S(x-y,t)\mp S(x+y,t)$. On $(0,l)$ with
Dirichlet conditions,
\begin{align*} K(x,y,t)&=\sum_{m\in\Z}\bigl[S(x-y-2ml,t)-S(x+y-2ml,t)\bigr]\\
&=\frac2l\sum_{n\ge1}e^{-k(n\pi/l)^2t}\sin\frac{n\pi x}l\sin\frac{n\pi y}l . \end{align*}
\end{theorem}
\noindent The first expression is the sum over the infinite dihedral group weighted by the sign character; the
second is the sum over eigenfunctions of \Cref{ch:fourier}. Their equality is the Poisson summation formula. The
two sums are complementary:
\begin{center}\small
\begin{tabular}{@{}lccc@{}}
\toprule
terms needed for accuracy $10^{-10}$ ($k=0.7$, $l=1$) & $t=0.001$ & $t=0.05$ & $t=1$\\
\midrule
shells of images & $0$ & $1$ & $4$\\
eigenfunctions & $56$ & $8$ & $1$\\
\bottomrule
\end{tabular}
\end{center}
\chk{D6} For short times a particle has not yet met the walls and one or two images suffice; for long times it has
forgotten where it started and one mode suffices. On the half-line the odd kernel applied to the constant $1$
gives $\erf(x/\sqrt{4kt})$, the probability that a Brownian particle started at $x$ has not yet reached the wall;
the even kernel reflects it. \chk{D5} So the reflection method does work for diffusion, in the form of a sum over
the group, and its dual is the method that replaces it in the source.
Poisson summation is itself a naturality statement. For the quotient $\R\to\R/2l\Z$ and the dual inclusion of the
lattice $\tfrac\pi l\Z$ in the line of frequencies, the square
\[\begin{tikzcd}[column sep=7em,row sep=large]
\mathcal S(\R)\arrow[r,"\text{sum over }2l\Z"]\arrow[d,"\text{Fourier transform}"'] & C^\infty(\R/2l\Z)\arrow[d,"\text{Fourier coefficients}"]\\
\mathcal S(\R)\arrow[r,"\text{restrict to }\frac\pi l\Z"'] & s(\Z)
\end{tikzcd}\]
commutes, up to the constant $1/2l$ fixed by the normalizations: summing over a subgroup on one side is
restricting to its annihilator on the other. Going round the top with the Gaussian $S(\cdot,t)$ gives the sum over
images, and round the bottom the sum over eigenfunctions, since the transform of $S(\cdot,t)$ is $e^{-k\xi^2t}$.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\fill[fwd!10] (4.552,0) rectangle (6.448,3.2);
\draw[ink2,very thin] (0.759,0)--(0.759,3.2);
\node[] at (0.759,-0.257) {\scriptsize $-2$};
\draw[ink2,very thin] (2.655,0)--(2.655,3.2);
\node[] at (2.655,-0.257) {\scriptsize $-1$};
\draw[ink2,very thin] (4.552,0)--(4.552,3.2);
\node[] at (4.552,-0.257) {\scriptsize $0$};
\draw[ink2,very thin] (6.448,0)--(6.448,3.2);
\node[] at (6.448,-0.257) {\scriptsize $1$};
\draw[ink2,very thin] (8.345,0)--(8.345,3.2);
\node[] at (8.345,-0.257) {\scriptsize $2$};
\draw[ink2,very thin] (10.241,0)--(10.241,3.2);
\node[] at (10.241,-0.257) {\scriptsize $3$};
\draw[ink2] (0,1.6)--(11,1.6);
\draw[thick,fwd] plot coordinates {(0,0.461) (0.016,0.428) (0.031,0.397) (0.047,0.369) (0.063,0.343) (0.079,0.319) (0.094,0.299) (0.11,0.281) (0.126,0.266) (0.141,0.255) (0.157,0.246) (0.173,0.241) (0.189,0.24) (0.204,0.242) (0.22,0.247) (0.236,0.256) (0.251,0.268) (0.267,0.283) (0.283,0.301) (0.299,0.323) (0.314,0.347) (0.33,0.374) (0.346,0.404) (0.361,0.437) (0.377,0.471) (0.393,0.508) (0.409,0.547) (0.424,0.587) (0.44,0.629) (0.456,0.673) (0.471,0.717) (0.487,0.763) (0.503,0.809) (0.519,0.857) (0.534,0.904) (0.55,0.953) (0.566,1.001) (0.581,1.05) (0.597,1.099) (0.613,1.148) (0.629,1.197) (0.644,1.246) (0.66,1.295) (0.676,1.344) (0.691,1.392) (0.707,1.441) (0.723,1.49) (0.739,1.538) (0.754,1.587) (0.77,1.635) (0.786,1.684) (0.801,1.732) (0.817,1.781) (0.833,1.829) (0.849,1.878) (0.864,1.927) (0.88,1.976) (0.896,2.025) (0.911,2.074) (0.927,2.123) (0.943,2.172) (0.959,2.221) (0.974,2.269) (0.99,2.317) (1.006,2.365) (1.021,2.412) (1.037,2.458) (1.053,2.503) (1.069,2.547) (1.084,2.59) (1.1,2.631) (1.116,2.671) (1.131,2.709) (1.147,2.745) (1.163,2.778) (1.179,2.81) (1.194,2.838) (1.21,2.864) (1.226,2.887) (1.241,2.907) (1.257,2.924) (1.273,2.938) (1.289,2.949) (1.304,2.956) (1.32,2.96) (1.336,2.96) (1.351,2.957) (1.367,2.95) (1.383,2.941) (1.399,2.928) (1.414,2.912) (1.43,2.893) (1.446,2.871) (1.461,2.846) (1.477,2.819) (1.493,2.789) (1.509,2.758) (1.524,2.724) (1.54,2.689) (1.556,2.652) (1.571,2.614) (1.587,2.575) (1.603,2.535) (1.619,2.495) (1.634,2.454) (1.65,2.413) (1.666,2.372) (1.681,2.332) (1.697,2.292) (1.713,2.252) (1.729,2.213) (1.744,2.175) (1.76,2.138) (1.776,2.103) (1.791,2.068) (1.807,2.035) (1.823,2.003) (1.839,1.973) (1.854,1.944) (1.87,1.916) (1.886,1.89) (1.901,1.866) (1.917,1.843) (1.933,1.821) (1.949,1.801) (1.964,1.782) (1.98,1.765) (1.996,1.749) (2.011,1.734) (2.027,1.72) (2.043,1.707) (2.059,1.696) (2.074,1.686) (2.09,1.676) (2.106,1.667) (2.121,1.66) (2.137,1.653) (2.153,1.646) (2.169,1.641) (2.184,1.636) (2.2,1.631) (2.216,1.627) (2.231,1.624) (2.247,1.62) (2.263,1.618) (2.279,1.615) (2.294,1.613) (2.31,1.611) (2.326,1.61) (2.341,1.608) (2.357,1.607) (2.373,1.606) (2.389,1.605) (2.404,1.604) (2.42,1.604) (2.436,1.603) (2.451,1.603) (2.467,1.602) (2.483,1.602) (2.499,1.601) (2.514,1.601) (2.53,1.601) (2.546,1.601) (2.561,1.601) (2.577,1.601) (2.593,1.6) (2.609,1.6) (2.624,1.6) (2.64,1.6) (2.656,1.6) (2.671,1.6) (2.687,1.6) (2.703,1.6) (2.719,1.6) (2.734,1.599) (2.75,1.599) (2.766,1.599) (2.781,1.599) (2.797,1.599) (2.813,1.598) (2.829,1.598) (2.844,1.598) (2.86,1.597) (2.876,1.597) (2.891,1.596) (2.907,1.596) (2.923,1.595) (2.939,1.594) (2.954,1.593) (2.97,1.592) (2.986,1.59) (3.001,1.589) (3.017,1.587) (3.033,1.585) (3.049,1.582) (3.064,1.579) (3.08,1.576) (3.096,1.573) (3.111,1.569) (3.127,1.564) (3.143,1.559) (3.159,1.553) (3.174,1.547) (3.19,1.54) (3.206,1.532) (3.221,1.523) (3.237,1.514) (3.253,1.503) (3.269,1.492) (3.284,1.479) (3.3,1.465) (3.316,1.45) (3.331,1.434) (3.347,1.417) (3.363,1.398) (3.379,1.378) (3.394,1.356) (3.41,1.333) (3.426,1.308) (3.441,1.282) (3.457,1.254) (3.473,1.225) (3.489,1.195) (3.504,1.163) (3.52,1.129) (3.536,1.095) (3.551,1.059) (3.567,1.022) (3.583,0.984) (3.599,0.945) (3.614,0.906) (3.63,0.866) (3.646,0.825) (3.661,0.784) (3.677,0.743) (3.693,0.702) (3.709,0.662) (3.724,0.622) (3.74,0.583) (3.756,0.545) (3.771,0.509) (3.787,0.474) (3.803,0.44) (3.819,0.409) (3.834,0.379) (3.85,0.352) (3.866,0.328) (3.881,0.306) (3.897,0.287) (3.913,0.271) (3.929,0.259) (3.944,0.249) (3.96,0.243) (3.976,0.24) (3.991,0.241) (4.007,0.245) (4.023,0.252) (4.039,0.263) (4.054,0.277) (4.07,0.294) (4.086,0.314) (4.101,0.337) (4.117,0.364) (4.133,0.393) (4.149,0.424) (4.164,0.458) (4.18,0.494) (4.196,0.532) (4.211,0.572) (4.227,0.613) (4.243,0.656) (4.259,0.7) (4.274,0.745) (4.29,0.792) (4.306,0.839) (4.321,0.886) (4.337,0.934) (4.353,0.983) (4.369,1.031) (4.384,1.08) (4.4,1.129) (4.416,1.178) (4.431,1.227) (4.447,1.276) (4.463,1.325) (4.479,1.374) (4.494,1.423) (4.51,1.471) (4.526,1.52) (4.541,1.568) (4.557,1.617) (4.573,1.665) (4.589,1.714) (4.604,1.762) (4.62,1.811) (4.636,1.86) (4.651,1.909) (4.667,1.957) (4.683,2.006) (4.699,2.055) (4.714,2.104) (4.73,2.153) (4.746,2.202) (4.761,2.251) (4.777,2.299) (4.793,2.347) (4.809,2.394) (4.824,2.44) (4.84,2.486) (4.856,2.53) (4.871,2.574) (4.887,2.616) (4.903,2.656) (4.919,2.695) (4.934,2.731) (4.95,2.766) (4.966,2.798) (4.981,2.828) (4.997,2.855) (5.013,2.879) (5.029,2.9) (5.044,2.918) (5.06,2.933) (5.076,2.945) (5.091,2.954) (5.107,2.959) (5.123,2.96) (5.139,2.958) (5.154,2.953) (5.17,2.945) (5.186,2.933) (5.201,2.918) (5.217,2.9) (5.233,2.879) (5.249,2.856) (5.264,2.829) (5.28,2.801) (5.296,2.77) (5.311,2.737) (5.327,2.702) (5.343,2.666) (5.359,2.629) (5.374,2.59) (5.39,2.55) (5.406,2.51) (5.421,2.469) (5.437,2.429) (5.453,2.388) (5.469,2.347) (5.484,2.307) (5.5,2.267) (5.516,2.228) (5.531,2.19) (5.547,2.152) (5.563,2.116) (5.579,2.081) (5.594,2.047) (5.61,2.015) (5.626,1.984) (5.641,1.955) (5.657,1.926) (5.673,1.9) (5.689,1.875) (5.704,1.851) (5.72,1.829) (5.736,1.808) (5.751,1.789) (5.767,1.771) (5.783,1.754) (5.799,1.739) (5.814,1.725) (5.83,1.712) (5.846,1.7) (5.861,1.689) (5.877,1.68) (5.893,1.671) (5.909,1.663) (5.924,1.655) (5.94,1.649) (5.956,1.643) (5.971,1.637) (5.987,1.633) (6.003,1.629) (6.019,1.625) (6.034,1.622) (6.05,1.619) (6.066,1.616) (6.081,1.614) (6.097,1.612) (6.113,1.61) (6.129,1.609) (6.144,1.608) (6.16,1.606) (6.176,1.605) (6.191,1.605) (6.207,1.604) (6.223,1.603) (6.239,1.603) (6.254,1.602) (6.27,1.602) (6.286,1.602) (6.301,1.601) (6.317,1.601) (6.333,1.601) (6.349,1.601) (6.364,1.601) (6.38,1.6) (6.396,1.6) (6.411,1.6) (6.427,1.6) (6.443,1.6) (6.459,1.6) (6.474,1.6) (6.49,1.6) (6.506,1.6) (6.521,1.6) (6.537,1.599) (6.553,1.599) (6.569,1.599) (6.584,1.599) (6.6,1.599) (6.616,1.598) (6.631,1.598) (6.647,1.598) (6.663,1.597) (6.679,1.597) (6.694,1.596) (6.71,1.595) (6.726,1.594) (6.741,1.593) (6.757,1.592) (6.773,1.591) (6.789,1.589) (6.804,1.587) (6.82,1.585) (6.836,1.583) (6.851,1.58) (6.867,1.577) (6.883,1.574) (6.899,1.57) (6.914,1.566) (6.93,1.561) (6.946,1.555) (6.961,1.549) (6.977,1.543) (6.993,1.535) (7.009,1.527) (7.024,1.518) (7.04,1.507) (7.056,1.496) (7.071,1.484) (7.087,1.471) (7.103,1.456) (7.119,1.44) (7.134,1.423) (7.15,1.405) (7.166,1.385) (7.181,1.364) (7.197,1.342) (7.213,1.318) (7.229,1.292) (7.244,1.265) (7.26,1.236) (7.276,1.206) (7.291,1.175) (7.307,1.142) (7.323,1.108) (7.339,1.073) (7.354,1.036) (7.37,0.999) (7.386,0.96) (7.401,0.921) (7.417,0.881) (7.433,0.84) (7.449,0.8) (7.464,0.759) (7.48,0.718) (7.496,0.677) (7.511,0.637) (7.527,0.598) (7.543,0.56) (7.559,0.522) (7.574,0.487) (7.59,0.453) (7.606,0.42) (7.621,0.39) (7.637,0.362) (7.653,0.337) (7.669,0.314) (7.684,0.294) (7.7,0.277) (7.716,0.263) (7.731,0.252) (7.747,0.245) (7.763,0.241) (7.779,0.24) (7.794,0.243) (7.81,0.249) (7.826,0.258) (7.841,0.271) (7.857,0.287) (7.873,0.306) (7.889,0.328) (7.904,0.353) (7.92,0.381) (7.936,0.412) (7.951,0.445) (7.967,0.48) (7.983,0.517) (7.999,0.556) (8.014,0.597) (8.03,0.64) (8.046,0.683) (8.061,0.728) (8.077,0.774) (8.093,0.821) (8.109,0.868) (8.124,0.916) (8.14,0.964) (8.156,1.013) (8.171,1.062) (8.187,1.111) (8.203,1.16) (8.219,1.209) (8.234,1.258) (8.25,1.307) (8.266,1.355) (8.281,1.404) (8.297,1.453) (8.313,1.501) (8.329,1.55) (8.344,1.598) (8.36,1.647) (8.376,1.695) (8.391,1.744) (8.407,1.793) (8.423,1.841) (8.439,1.89) (8.454,1.939) (8.47,1.988) (8.486,2.037) (8.501,2.086) (8.517,2.135) (8.533,2.184) (8.549,2.232) (8.564,2.281) (8.58,2.329) (8.596,2.376) (8.611,2.423) (8.627,2.469) (8.643,2.514) (8.659,2.557) (8.674,2.6) (8.69,2.641) (8.706,2.68) (8.721,2.718) (8.737,2.753) (8.753,2.786) (8.769,2.817) (8.784,2.845) (8.8,2.87) (8.816,2.892) (8.831,2.912) (8.847,2.928) (8.863,2.941) (8.879,2.951) (8.894,2.957) (8.91,2.96) (8.926,2.959) (8.941,2.956) (8.957,2.948) (8.973,2.938) (8.989,2.924) (9.004,2.907) (9.02,2.888) (9.036,2.865) (9.051,2.84) (9.067,2.812) (9.083,2.782) (9.099,2.75) (9.114,2.716) (9.13,2.68) (9.146,2.643) (9.161,2.605) (9.177,2.565) (9.193,2.525) (9.209,2.485) (9.224,2.444) (9.24,2.403) (9.256,2.362) (9.271,2.322) (9.287,2.282) (9.303,2.243) (9.319,2.204) (9.334,2.166) (9.35,2.13) (9.366,2.094) (9.381,2.06) (9.397,2.027) (9.413,1.996) (9.429,1.966) (9.444,1.937) (9.46,1.91) (9.476,1.884) (9.491,1.86) (9.507,1.837) (9.523,1.816) (9.539,1.796) (9.554,1.778) (9.57,1.761) (9.586,1.745) (9.601,1.73) (9.617,1.717) (9.633,1.705) (9.649,1.693) (9.664,1.683) (9.68,1.674) (9.696,1.665) (9.711,1.658) (9.727,1.651) (9.743,1.645) (9.759,1.639) (9.774,1.635) (9.79,1.63) (9.806,1.626) (9.821,1.623) (9.837,1.62) (9.853,1.617) (9.869,1.615) (9.884,1.613) (9.9,1.611) (9.916,1.609) (9.931,1.608) (9.947,1.607) (9.963,1.606) (9.979,1.605) (9.994,1.604) (10.01,1.603) (10.026,1.603) (10.041,1.602) (10.057,1.602) (10.073,1.602) (10.089,1.601) (10.104,1.601) (10.12,1.601) (10.136,1.601) (10.151,1.601) (10.167,1.6) (10.183,1.6) (10.199,1.6) (10.214,1.6) (10.23,1.6) (10.246,1.6) (10.261,1.6) (10.277,1.6) (10.293,1.6) (10.309,1.6) (10.324,1.599) (10.34,1.599) (10.356,1.599) (10.371,1.599) (10.387,1.599) (10.403,1.598) (10.419,1.598) (10.434,1.598) (10.45,1.597) (10.466,1.597) (10.481,1.596) (10.497,1.595) (10.513,1.595) (10.529,1.594) (10.544,1.593) (10.56,1.591) (10.576,1.59) (10.591,1.588) (10.607,1.586) (10.623,1.584) (10.639,1.581) (10.654,1.579) (10.67,1.575) (10.686,1.572) (10.701,1.568) (10.717,1.563) (10.733,1.558) (10.749,1.552) (10.764,1.545) (10.78,1.538) (10.796,1.53) (10.811,1.521) (10.827,1.511) (10.843,1.501) (10.859,1.489) (10.874,1.476) (10.89,1.462) (10.906,1.447) (10.921,1.43) (10.937,1.412) (10.953,1.393) (10.969,1.372) (10.984,1.35) (11,1.327)};
\node[fwd] at (1.328,3.057) {\scriptsize $+$};
\node[bwd] at (0.19,0.143) {\scriptsize $-$};
\node[fwd] at (5.121,3.057) {\scriptsize $+$};
\node[bwd] at (3.983,0.143) {\scriptsize $-$};
\node[fwd] at (8.914,3.057) {\scriptsize $+$};
\node[bwd] at (7.776,0.143) {\scriptsize $-$};
\node[right] at (11,1.771) {\scriptsize $x$};
\end{scope}
\end{tikzpicture}
\figcap{The Dirichlet kernel of the interval $(0,1)$, shaded, as a sum over images, for a source at $y=0.3$ and $kt=0.014$: positive copies at $y+2m$, negative ones at $-y+2m$. The sum is odd about every integer, so it vanishes at both ends of the interval.}
\end{center}
\section{Sources}
\begin{theorem}[Duhamel]\label{thm:duhamel-heat}
\Scl[source, \S3.3] The solution of $u_t-ku_{xx}=f$ with $u(\cdot,0)=\varphi$ is
$u(t)=T_t\varphi+\int_0^tT_{t-s}f(s)\,ds$.
\end{theorem}
\noindent The formula is that of variation of constants and holds for every semigroup; for the wave equation
written as a first-order system it is \Cref{thm:duhamel-wave}. With $f=\sin x$ and $\varphi=0$ it gives
$(1-e^{-t})\sin x$. \chk{D9}
\section*{Notes}
The source treats diffusion in its Sections 2.3--2.5, 3.1, 3.3 and 3.5. Semigroup theory is in \cite{Paz83}. The
identity between the rate of entropy and the Fisher information is de Bruijn's, see \cite{CoTh06}; monotonicity
of divergences under Markov kernels is the subject of \cite{ESCHC}. The image and eigenfunction forms of the kernel
on an interval are two expressions of a Jacobi theta function.
\section*{Exercises}
\begin{exercise}\label{exer:scal}
Derive the Gaussian from the ansatz $u=t^{-1/2}g(x/\sqrt t)$.
\end{exercise}
\begin{exercise}\label{exer:surv}
A particle diffuses with $k=1$ from $x=1$ on the half-line with an absorbing wall at $0$. Find the probability
that it has not been absorbed by time $1$.
\end{exercise}
\begin{exercise}\label{exer:ent}
Prove (a) and (b) of \Cref{thm:mono}.
\end{exercise}
\begin{exercise}\label{exer:terms}
Estimate, as functions of $kt/l^2$ and the tolerance $\epsilon$, the number of image shells and of eigenfunctions
needed in \Cref{thm:kernels}, and compare with the table.
\end{exercise}
\chapter{Boundary Conditions and Fourier Series}\label{ch:fourier}
On a bounded interval there is no explicit kernel, and the source turns to Fourier's method: separate the
variables, solve an eigenvalue problem in $x$, and superpose \src{Ch.~4--5}. Three questions organize the chapter.
What is a boundary condition? Why sines and cosines? In what sense does the series converge?
\section{Diagonalizing the generator}
Let $A=-d^2/dx^2$ on $(0,l)$ with boundary conditions fixed. The diffusion and wave equations are
$u_t=-kAu$ and $u_{tt}=-c^2Au$. If $AX=\lambda X$, then $e^{-k\lambda t}X$ solves the first and
$\cos(c\sqrt\lambda\,t)X$, $\sin(c\sqrt\lambda\,t)X$ solve the second. Fourier's method is the statement that $A$ can
be diagonalized, so that the semigroup and the group of the last two chapters become multiplication operators:
\[ T_t=e^{-ktA},\qquad\text{wave group}=\begin{pmatrix}\cos ct\sqrt A&\sin(ct\sqrt A)/c\sqrt A\\-c\sqrt A\sin ct\sqrt A&\cos ct\sqrt A\end{pmatrix}. \]
For Dirichlet conditions the eigenpairs are $(n\pi/l)^2$, $\sin(n\pi x/l)$, $n\ge1$; for Neumann conditions
$(n\pi/l)^2$, $\cos(n\pi x/l)$, $n\ge0$. \chk{F1}
\section{A boundary condition is a Lagrangian plane}
For functions $f,g$ on $[a,b]$, Green's second identity in one dimension reads
\[ \int_a^b(-f''g+fg'')\,dx=\bigl[-f'g+fg'\bigr]_a^b=:\omega\bigl(\beta(f),\beta(g)\bigr), \]
where $\beta(f)=(f(a),f'(a),f(b),f'(b))\in\R^4$ is the vector of boundary values of $f$.
The form $\omega$ is symplectic on the space $\R^4$ of boundary values. A pair of independent linear boundary
conditions is a plane $V\subset\R^4$.
\begin{theorem}\label{thm:lagrangian}
\Srf[source, \S5.3; \cite{EvMa99,KoSc99}] The boundary conditions $V$ are symmetric in the source's sense if and only if
$V$ is Lagrangian: $\omega|_V=0$. In that case eigenfunctions for distinct eigenvalues are orthogonal and all
eigenvalues are real (the source's Theorems 1 and 2). The self-adjoint realizations of $-d^2/dx^2$ on $[a,b]$ are in
bijection with the Lagrangian planes of $(\R^4,\omega)$, a compact manifold of dimension three.
\end{theorem}
\begin{proof}
The first statement is the definition, since the source's condition is $\omega(\beta(f),\beta(g))=0$ for all $f,g$
satisfying the conditions, and a plane on which a symplectic form on $\R^4$ vanishes is Lagrangian. For
eigenfunctions $X_1,X_2$ the identity gives $(\lambda_1-\lambda_2)\int X_1X_2=\omega(\beta(X_1),\beta(X_2))=0$.
\end{proof}
\begin{center}\small
\begin{tabular}{@{}lll@{}}
\toprule
condition & plane $V$ spanned by & $\omega$ on the two vectors\\
\midrule
Dirichlet & $(0,1,0,0)$, $(0,0,0,1)$ & $0$\\
Neumann & $(1,0,0,0)$, $(0,0,1,0)$ & $0$\\
Robin & $(1,a_0,0,0)$, $(0,0,1,-a_l)$ & $0$\\
periodic & $(1,0,1,0)$, $(0,1,0,1)$ & $0$\\
$X(a)=X(b)$, $X'(a)=2X'(b)$ & $(1,0,1,0)$, $(0,2,0,1)$ & $-1$\\
\bottomrule
\end{tabular}
\end{center}
\noindent \chk{F3} The last row is the source's example of conditions for which the method fails. The first three
are \emph{local}: $V$ is a sum of a line at $a$ and a line at $b$, and the lines at one endpoint form a circle, with
Dirichlet and Neumann two of its points and the Robin constant a coordinate on the rest. The periodic condition is
non-local: it is the graph of the identity between the two ends, which is what it means to glue them.
\begin{center}
\begin{tikzpicture}
\draw[->,ink2] (-2.4,0)--(2.6,0) node[right]{\scriptsize $X(a)$};
\draw[->,ink2] (0,-2.0)--(0,2.2) node[above]{\scriptsize $X'(a)$};
\draw[ink2,very thin] (0,0) circle (1.3);
\draw[very thick,fwd] (0,-1.9)--(0,1.9);
\draw[very thick,bwd] (-2.2,0)--(2.2,0);
\draw[thick,accent] (-1.9,-1.1)--(1.9,1.1);
\draw[thick,accent,densely dashed] (-1.5,1.6)--(1.5,-1.6);
\node[fwd,left] at (-0.05,1.75) {\scriptsize Dirichlet: $X=0$};
\node[bwd,below] at (1.45,-0.03) {\scriptsize Neumann: $X'=0$};
\node[accent,right] at (1.9,1.1) {\scriptsize Robin: $X'=aX$, $a>0$};
\node[accent,right] at (1.5,-1.6) {\scriptsize $a<0$};
\foreach \p in {(0,1.3),(1.3,0),(1.125,0.651),(0.889,-0.948)} \fill \p circle (1.6pt);
\end{tikzpicture}
\figcap{The boundary conditions at one end of an interval are the lines through the origin in the plane of $(X,X')$, a circle's worth. The Dirichlet and Neumann conditions are two points of it and the Robin conditions fill in the rest; a separated condition for the interval is a point of the product of two such circles.}
\end{center}
\begin{remark}[Intervals compose]\label{rem:compose}
For fixed $\lambda$ the solutions of $-X''=\lambda X$ on $[a,b]$ have boundary values forming a Lagrangian plane
in $\R^4$, the graph of the transfer matrix from $(X(a),X'(a))$ to $(X(b),X'(b))$, which has determinant one. Joining
$[a,b]$ to $[b,c]$ multiplies transfer matrices, that is, composes Lagrangian relations. An eigenvalue is a
$\lambda$ for which the plane of the equation meets the plane of the boundary conditions. This compositional
picture returns for networks in \Cref{ch:green} and for scattering in \Cref{ch:fields}.
\end{remark}
\[\begin{tikzcd}[column sep=4.2em]
B_a\arrow[r,hook] & V_a\arrow[r,"M_{[a,b]}(\lambda)"]\arrow[rr,bend right=22,"M_{[a,c]}(\lambda)"'] & V_b\arrow[r,"M_{[b,c]}(\lambda)"] & V_c & B_c\arrow[l,hook']
\end{tikzcd}\]
Here $V_x=\R^2$ holds the pair $(X(x),X'(x))$ and a separated boundary condition is a line $B$ in it. The number
$\lambda$ is an eigenvalue exactly when the line $M_{[a,c]}(\lambda)B_a$ is the line $B_c$: the shooting method, read as
a composite of relations.
\paragraph{The Robin condition.} With $X'-a_0X=0$ at $0$ and $X'+a_lX=0$ at $l$, the quadratic form of $A$ is
$\int X'^2+a_0X(0)^2+a_lX(l)^2$, which is positive when both constants are positive (radiation) and can be
negative when they are negative (absorption). The source finds a zero eigenvalue exactly when
$a_0+a_l=-a_0a_ll$ \src{\S4.3}. For $a_0=a_l=-1$ that is $l=2$: there is one negative eigenvalue for $l<2$ and two for
$l>2$; at $l=1$ it is $-2.3821$. \chk{F2} A negative eigenvalue $-\gamma^2$ means that the diffusion equation has
the growing solution $e^{k\gamma^2t}X(x)$: the boundary feeds the interior faster than diffusion can spread it.
\section{Why sines and cosines}
Take periodic conditions on $(-l,l)$: the domain is the circle $\mathbb T=\R/2l\Z$, a compact abelian group. The
operator $A$ commutes with translations, so its eigenspaces are representations of $\mathbb T$, and the
irreducible ones are the characters $e^{in\pi x/l}$, $n\in\Z$.
\begin{theorem}[Fourier series as duality]\label{thm:pontryagin}
\Scl The characters of $\mathbb T$ form the dual group $\Z$, and the map $f\mapsto(\hat f(n))_{n\in\Z}$ is a unitary
isomorphism $L^2(\mathbb T)\to\ell^2(\Z)$ that carries every translation-invariant operator to a multiplication
operator; $A$ becomes multiplication by $(n\pi/l)^2$.
\end{theorem}
\noindent The reflection $x\mapsto-x$ acts on $\mathbb T$ and exchanges $n$ and $-n$. Its odd part is spanned by the
sines and its even part by the cosines, and by \Cref{thm:images} these are the Dirichlet and Neumann problems on
$(0,l)$: the interval is the circle of twice its length modulo reflection. The sine and cosine series of the
source \src{\S5.1--5.2} are the two isotypic components of the full Fourier series. On the finite group $\Z/n$ the
whole statement is a computation: the characters diagonalize the periodic difference Laplacian with eigenvalues
$4\sin^2(\pi k/n)$, and their even and odd parts under reflection are the eigenvectors of the Neumann and Dirichlet
matrices. \chk{F7} For the function $x$ on $(0,l)$ Parseval's identity gives $\sum n^{-2}=\pi^2/6$. \chk{F4}
\section{Completeness, and in which category}
The source distinguishes uniform, mean-square and pointwise convergence and proves a theorem for each
\src{\S5.4}.
\begin{theorem}\label{thm:complete}
\Scl[source, \S5.4, Theorems 2--6] For a symmetric boundary condition: \emph{(a)} the eigenfunction expansion of every
$f\in L^2$ converges to $f$ in mean square, and Parseval's equality holds; \emph{(b)} if $f$ is $C^2$ and satisfies the
boundary conditions the convergence is uniform; \emph{(c)} for the classical series, if $f$ and $f'$ are piecewise
continuous the series converges at each point to the mean of the one-sided limits. The partial sum is the best
approximation in mean square among combinations of the first $N$ eigenfunctions.
\end{theorem}
\noindent Statement (a) says that the coefficient map is an isomorphism in the category of Hilbert spaces. It is not
an isomorphism of the Banach space of continuous functions onto a sequence space: there are continuous functions
whose Fourier series diverge at a point. Completeness, like well-posedness (\Cref{def:wellposed}), is a
statement about a morphism in a specified category. The Gibbs phenomenon is the visible form of the difference:
near a unit jump the partial sums overshoot to $\tfrac1\pi\int_0^\pi\frac{\sin\theta}{\theta}d\theta=0.58949$ where the
function is $\tfrac12$, an excess of $8.95$ per cent of the jump that does not decrease with $N$ (the source's ``9
percent''); the peaks at $N=51$, $201$, $801$ are $0.58955$, $0.58949$, $0.58949$. \chk{F5} The series converges in
$L^2$ and not uniformly.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,4)--(0,0)--(8.6,0);
\draw[ink2] (1.544,0)--(1.544,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.749,0)--(3.749,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (5.954,0)--(5.954,-0.07) node[below,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (8.159,0)--(8.159,-0.07) node[below,inner sep=1.5pt]{\scriptsize $1.5$};
\draw[ink2] (0,0.667)--(-0.07,0.667) node[left,inner sep=1.5pt]{\scriptsize $-0.5$};
\draw[ink2] (0,2)--(-0.07,2) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.333)--(-0.07,3.333) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\node[below] at (4.3,-0.38) {\scriptsize $x$};
\draw[ink2,densely dashed] (0,3.333)--(1.544,3.333);
\draw[ink2,densely dashed] (1.544,3.333)--(8.6,3.333);
\draw[ink2,densely dashed] (0,0.667)--(1.544,0.667);
\draw[bwd,densely dotted] (0,3.572)--(8.6,3.572);
\draw[ink2] plot coordinates {(0,0.54) (0.029,0.526) (0.057,0.513) (0.086,0.5) (0.115,0.488) (0.143,0.477) (0.172,0.466) (0.201,0.457) (0.229,0.448) (0.258,0.441) (0.287,0.435) (0.315,0.43) (0.344,0.427) (0.373,0.425) (0.401,0.425) (0.43,0.427) (0.459,0.43) (0.487,0.435) (0.516,0.442) (0.545,0.451) (0.573,0.463) (0.602,0.476) (0.631,0.491) (0.659,0.509) (0.688,0.528) (0.717,0.55) (0.745,0.574) (0.774,0.601) (0.803,0.629) (0.831,0.66) (0.86,0.693) (0.889,0.729) (0.917,0.766) (0.946,0.806) (0.975,0.848) (1.003,0.892) (1.032,0.938) (1.061,0.986) (1.089,1.035) (1.118,1.087) (1.147,1.14) (1.175,1.195) (1.204,1.252) (1.233,1.31) (1.261,1.369) (1.29,1.429) (1.319,1.491) (1.347,1.554) (1.376,1.617) (1.405,1.682) (1.433,1.747) (1.462,1.812) (1.491,1.878) (1.519,1.944) (1.548,2.01) (1.577,2.076) (1.605,2.142) (1.634,2.208) (1.663,2.273) (1.691,2.338) (1.72,2.402) (1.749,2.466) (1.777,2.528) (1.806,2.589) (1.835,2.649) (1.863,2.708) (1.892,2.766) (1.921,2.822) (1.949,2.876) (1.978,2.929) (2.007,2.98) (2.035,3.029) (2.064,3.077) (2.093,3.122) (2.121,3.165) (2.15,3.206) (2.179,3.245) (2.207,3.282) (2.236,3.317) (2.265,3.349) (2.293,3.38) (2.322,3.408) (2.351,3.433) (2.379,3.457) (2.408,3.478) (2.437,3.497) (2.465,3.514) (2.494,3.529) (2.523,3.541) (2.551,3.552) (2.58,3.56) (2.609,3.567) (2.637,3.571) (2.666,3.574) (2.695,3.575) (2.723,3.574) (2.752,3.572) (2.781,3.569) (2.809,3.563) (2.838,3.557) (2.867,3.549) (2.895,3.54) (2.924,3.531) (2.953,3.52) (2.981,3.508) (3.01,3.496) (3.039,3.483) (3.067,3.469) (3.096,3.455) (3.125,3.441) (3.153,3.427) (3.182,3.412) (3.211,3.397) (3.239,3.383) (3.268,3.368) (3.297,3.354) (3.325,3.34) (3.354,3.326) (3.383,3.313) (3.411,3.3) (3.44,3.288) (3.469,3.276) (3.497,3.265) (3.526,3.255) (3.555,3.245) (3.583,3.237) (3.612,3.229) (3.641,3.222) (3.669,3.216) (3.698,3.21) (3.727,3.206) (3.755,3.202) (3.784,3.2) (3.813,3.198) (3.841,3.197) (3.87,3.198) (3.899,3.199) (3.927,3.2) (3.956,3.203) (3.985,3.206) (4.013,3.21) (4.042,3.215) (4.071,3.22) (4.099,3.226) (4.128,3.233) (4.157,3.24) (4.185,3.247) (4.214,3.255) (4.243,3.263) (4.271,3.272) (4.3,3.281) (4.329,3.29) (4.357,3.299) (4.386,3.308) (4.415,3.317) (4.443,3.326) (4.472,3.335) (4.501,3.344) (4.529,3.352) (4.558,3.361) (4.587,3.369) (4.615,3.376) (4.644,3.384) (4.673,3.391) (4.701,3.397) (4.73,3.403) (4.759,3.409) (4.787,3.413) (4.816,3.418) (4.845,3.422) (4.873,3.425) (4.902,3.427) (4.931,3.429) (4.959,3.431) (4.988,3.431) (5.017,3.431) (5.045,3.431) (5.074,3.43) (5.103,3.428) (5.131,3.426) (5.16,3.423) (5.189,3.42) (5.217,3.416) (5.246,3.412) (5.275,3.407) (5.303,3.402) (5.332,3.397) (5.361,3.391) (5.389,3.385) (5.418,3.378) (5.447,3.372) (5.475,3.365) (5.504,3.358) (5.533,3.351) (5.561,3.344) (5.59,3.337) (5.619,3.331) (5.647,3.324) (5.676,3.317) (5.705,3.31) (5.733,3.304) (5.762,3.298) (5.791,3.292) (5.819,3.287) (5.848,3.281) (5.877,3.277) (5.905,3.272) (5.934,3.268) (5.963,3.264) (5.991,3.261) (6.02,3.259) (6.049,3.256) (6.077,3.255) (6.106,3.253) (6.135,3.253) (6.163,3.253) (6.192,3.253) (6.221,3.254) (6.249,3.255) (6.278,3.257) (6.307,3.259) (6.335,3.261) (6.364,3.264) (6.393,3.268) (6.421,3.272) (6.45,3.276) (6.479,3.28) (6.507,3.285) (6.536,3.29) (6.565,3.295) (6.593,3.301) (6.622,3.307) (6.651,3.312) (6.679,3.318) (6.708,3.324) (6.737,3.33) (6.765,3.336) (6.794,3.342) (6.823,3.348) (6.851,3.354) (6.88,3.359) (6.909,3.365) (6.937,3.37) (6.966,3.375) (6.995,3.379) (7.023,3.384) (7.052,3.388) (7.081,3.391) (7.109,3.395) (7.138,3.397) (7.167,3.4) (7.195,3.402) (7.224,3.404) (7.253,3.405) (7.281,3.406) (7.31,3.406) (7.339,3.406) (7.367,3.405) (7.396,3.404) (7.425,3.403) (7.453,3.401) (7.482,3.399) (7.511,3.396) (7.539,3.393) (7.568,3.39) (7.597,3.386) (7.625,3.382) (7.654,3.378) (7.683,3.373) (7.711,3.368) (7.74,3.363) (7.769,3.358) (7.797,3.353) (7.826,3.347) (7.855,3.342) (7.883,3.336) (7.912,3.331) (7.941,3.325) (7.969,3.32) (7.998,3.314) (8.027,3.309) (8.055,3.304) (8.084,3.299) (8.113,3.294) (8.141,3.29) (8.17,3.286) (8.199,3.282) (8.227,3.278) (8.256,3.275) (8.285,3.272) (8.313,3.27) (8.342,3.267) (8.371,3.266) (8.399,3.264) (8.428,3.264) (8.457,3.263) (8.485,3.263) (8.514,3.264) (8.543,3.264) (8.571,3.266) (8.6,3.267)};
\draw[thick,fwd] plot coordinates {(0,0.703) (0.014,0.697) (0.029,0.691) (0.043,0.684) (0.057,0.676) (0.072,0.668) (0.086,0.659) (0.1,0.651) (0.115,0.643) (0.129,0.635) (0.143,0.628) (0.158,0.623) (0.172,0.618) (0.186,0.615) (0.201,0.613) (0.215,0.612) (0.229,0.613) (0.244,0.616) (0.258,0.62) (0.272,0.626) (0.287,0.633) (0.301,0.641) (0.315,0.65) (0.33,0.66) (0.344,0.67) (0.358,0.681) (0.373,0.691) (0.387,0.7) (0.401,0.709) (0.416,0.717) (0.43,0.724) (0.444,0.729) (0.459,0.732) (0.473,0.734) (0.487,0.734) (0.502,0.731) (0.516,0.727) (0.53,0.721) (0.545,0.713) (0.559,0.704) (0.573,0.693) (0.588,0.681) (0.602,0.668) (0.616,0.655) (0.631,0.642) (0.645,0.629) (0.659,0.617) (0.674,0.606) (0.688,0.596) (0.702,0.588) (0.717,0.583) (0.731,0.579) (0.745,0.578) (0.76,0.579) (0.774,0.584) (0.788,0.59) (0.803,0.6) (0.817,0.611) (0.831,0.625) (0.846,0.64) (0.86,0.657) (0.874,0.675) (0.889,0.694) (0.903,0.712) (0.917,0.73) (0.932,0.747) (0.946,0.762) (0.96,0.775) (0.975,0.785) (0.989,0.792) (1.003,0.796) (1.018,0.796) (1.032,0.792) (1.046,0.784) (1.061,0.772) (1.075,0.756) (1.089,0.737) (1.104,0.713) (1.118,0.687) (1.132,0.659) (1.147,0.629) (1.161,0.598) (1.175,0.567) (1.19,0.537) (1.204,0.508) (1.218,0.482) (1.233,0.461) (1.247,0.444) (1.261,0.432) (1.276,0.428) (1.29,0.431) (1.304,0.443) (1.319,0.463) (1.333,0.494) (1.347,0.534) (1.362,0.585) (1.376,0.647) (1.39,0.719) (1.405,0.801) (1.419,0.893) (1.433,0.995) (1.448,1.106) (1.462,1.224) (1.476,1.35) (1.491,1.482) (1.505,1.618) (1.519,1.758) (1.534,1.901) (1.548,2.044) (1.562,2.187) (1.577,2.328) (1.591,2.466) (1.605,2.6) (1.62,2.728) (1.634,2.85) (1.648,2.963) (1.663,3.069) (1.677,3.165) (1.691,3.251) (1.706,3.327) (1.72,3.393) (1.734,3.447) (1.749,3.492) (1.763,3.526) (1.777,3.551) (1.792,3.566) (1.806,3.572) (1.82,3.57) (1.835,3.561) (1.849,3.547) (1.863,3.526) (1.878,3.502) (1.892,3.475) (1.906,3.445) (1.921,3.414) (1.935,3.383) (1.949,3.352) (1.964,3.323) (1.978,3.296) (1.992,3.272) (2.007,3.251) (2.021,3.233) (2.035,3.22) (2.05,3.21) (2.064,3.205) (2.078,3.203) (2.093,3.206) (2.107,3.212) (2.121,3.221) (2.136,3.233) (2.15,3.247) (2.164,3.263) (2.179,3.281) (2.193,3.299) (2.207,3.318) (2.222,3.336) (2.236,3.353) (2.25,3.369) (2.265,3.384) (2.279,3.396) (2.293,3.406) (2.308,3.414) (2.322,3.419) (2.336,3.422) (2.351,3.422) (2.365,3.419) (2.379,3.414) (2.394,3.407) (2.408,3.398) (2.422,3.387) (2.437,3.376) (2.451,3.363) (2.465,3.35) (2.48,3.337) (2.494,3.324) (2.508,3.312) (2.523,3.3) (2.537,3.29) (2.551,3.282) (2.566,3.275) (2.58,3.27) (2.594,3.267) (2.609,3.266) (2.623,3.267) (2.637,3.27) (2.652,3.274) (2.666,3.28) (2.68,3.288) (2.695,3.296) (2.709,3.305) (2.723,3.315) (2.738,3.326) (2.752,3.336) (2.766,3.346) (2.781,3.355) (2.795,3.364) (2.809,3.371) (2.824,3.377) (2.838,3.382) (2.852,3.386) (2.867,3.387) (2.881,3.388) (2.895,3.386) (2.91,3.384) (2.924,3.379) (2.938,3.374) (2.953,3.368) (2.967,3.36) (2.981,3.352) (2.996,3.344) (3.01,3.336) (3.024,3.327) (3.039,3.319) (3.053,3.312) (3.067,3.305) (3.082,3.299) (3.096,3.294) (3.11,3.291) (3.125,3.289) (3.139,3.288) (3.153,3.288) (3.168,3.29) (3.182,3.292) (3.196,3.296) (3.211,3.301) (3.225,3.307) (3.239,3.314) (3.254,3.32) (3.268,3.328) (3.282,3.335) (3.297,3.342) (3.311,3.349) (3.325,3.355) (3.34,3.36) (3.354,3.365) (3.368,3.368) (3.383,3.371) (3.397,3.373) (3.411,3.373) (3.426,3.372) (3.44,3.37) (3.454,3.367) (3.469,3.364) (3.483,3.359) (3.497,3.354) (3.512,3.348) (3.526,3.342) (3.54,3.335) (3.555,3.329) (3.569,3.323) (3.583,3.317) (3.598,3.312) (3.612,3.308) (3.626,3.304) (3.641,3.301) (3.655,3.299) (3.669,3.298) (3.684,3.298) (3.698,3.3) (3.712,3.302) (3.727,3.305) (3.741,3.308) (3.755,3.313) (3.77,3.317) (3.784,3.323) (3.798,3.328) (3.813,3.334) (3.827,3.339) (3.841,3.345) (3.856,3.35) (3.87,3.354) (3.884,3.358) (3.899,3.361) (3.913,3.363) (3.927,3.364) (3.942,3.365) (3.956,3.364) (3.97,3.363) (3.985,3.361) (3.999,3.358) (4.013,3.354) (4.028,3.35) (4.042,3.346) (4.056,3.341) (4.071,3.336) (4.085,3.331) (4.099,3.326) (4.114,3.321) (4.128,3.317) (4.142,3.313) (4.157,3.31) (4.171,3.307) (4.185,3.306) (4.2,3.305) (4.214,3.305) (4.228,3.305) (4.243,3.307) (4.257,3.309) (4.271,3.312) (4.286,3.316) (4.3,3.32) (4.314,3.324) (4.329,3.329) (4.343,3.333) (4.357,3.338) (4.372,3.342) (4.386,3.346) (4.4,3.35) (4.415,3.353) (4.429,3.356) (4.443,3.358) (4.458,3.359) (4.472,3.36) (4.486,3.36) (4.501,3.359) (4.515,3.357) (4.529,3.354) (4.544,3.352) (4.558,3.348) (4.572,3.344) (4.587,3.34) (4.601,3.336) (4.615,3.332) (4.63,3.327) (4.644,3.323) (4.658,3.32) (4.673,3.316) (4.687,3.314) (4.701,3.311) (4.716,3.31) (4.73,3.309) (4.744,3.309) (4.759,3.309) (4.773,3.31) (4.787,3.312) (4.802,3.315) (4.816,3.318) (4.83,3.321) (4.845,3.325) (4.859,3.329) (4.873,3.333) (4.888,3.337) (4.902,3.34) (4.916,3.344) (4.931,3.347) (4.945,3.35) (4.959,3.353) (4.974,3.355) (4.988,3.356) (5.002,3.356) (5.017,3.356) (5.031,3.356) (5.045,3.354) (5.06,3.352) (5.074,3.35) (5.088,3.347) (5.103,3.343) (5.117,3.34) (5.131,3.336) (5.146,3.332) (5.16,3.329) (5.174,3.325) (5.189,3.322) (5.203,3.319) (5.217,3.316) (5.232,3.314) (5.246,3.313) (5.26,3.312) (5.275,3.312) (5.289,3.312) (5.303,3.313) (5.318,3.314) (5.332,3.316) (5.346,3.319) (5.361,3.322) (5.375,3.325) (5.389,3.329) (5.404,3.332) (5.418,3.336) (5.432,3.339) (5.447,3.343) (5.461,3.346) (5.475,3.348) (5.49,3.35) (5.504,3.352) (5.518,3.353) (5.533,3.354) (5.547,3.354) (5.561,3.353) (5.576,3.352) (5.59,3.351) (5.604,3.348) (5.619,3.346) (5.633,3.343) (5.647,3.34) (5.662,3.336) (5.676,3.333) (5.69,3.33) (5.705,3.326) (5.719,3.323) (5.733,3.321) (5.748,3.318) (5.762,3.316) (5.776,3.315) (5.791,3.314) (5.805,3.314) (5.819,3.314) (5.834,3.315) (5.848,3.316) (5.862,3.318) (5.877,3.32) (5.891,3.322) (5.905,3.325) (5.92,3.328) (5.934,3.332) (5.948,3.335) (5.963,3.338) (5.977,3.341) (5.991,3.344) (6.006,3.347) (6.02,3.349) (6.034,3.35) (6.049,3.352) (6.063,3.352) (6.077,3.352) (6.092,3.352) (6.106,3.351) (6.12,3.35) (6.135,3.348) (6.149,3.345) (6.163,3.343) (6.178,3.34) (6.192,3.337) (6.206,3.334) (6.221,3.33) (6.235,3.327) (6.249,3.324) (6.264,3.322) (6.278,3.32) (6.292,3.318) (6.307,3.316) (6.321,3.315) (6.335,3.315) (6.35,3.315) (6.364,3.316) (6.378,3.317) (6.393,3.318) (6.407,3.32) (6.421,3.323) (6.436,3.325) (6.45,3.328) (6.464,3.331) (6.479,3.334) (6.493,3.337) (6.507,3.34) (6.522,3.343) (6.536,3.345) (6.55,3.347) (6.565,3.349) (6.579,3.35) (6.593,3.351) (6.608,3.351) (6.622,3.351) (6.636,3.35) (6.651,3.349) (6.665,3.347) (6.679,3.345) (6.694,3.343) (6.708,3.34) (6.722,3.337) (6.737,3.334) (6.751,3.331) (6.765,3.328) (6.78,3.325) (6.794,3.323) (6.808,3.321) (6.823,3.319) (6.837,3.317) (6.851,3.316) (6.866,3.316) (6.88,3.316) (6.894,3.316) (6.909,3.317) (6.923,3.319) (6.937,3.321) (6.952,3.323) (6.966,3.325) (6.98,3.328) (6.995,3.331) (7.009,3.334) (7.023,3.337) (7.038,3.34) (7.052,3.342) (7.066,3.345) (7.081,3.347) (7.095,3.348) (7.109,3.349) (7.124,3.35) (7.138,3.35) (7.152,3.35) (7.167,3.35) (7.181,3.348) (7.195,3.347) (7.21,3.345) (7.224,3.343) (7.238,3.34) (7.253,3.337) (7.267,3.334) (7.281,3.332) (7.296,3.329) (7.31,3.326) (7.324,3.324) (7.339,3.321) (7.353,3.32) (7.367,3.318) (7.382,3.317) (7.396,3.317) (7.41,3.317) (7.425,3.317) (7.439,3.318) (7.453,3.319) (7.468,3.321) (7.482,3.323) (7.496,3.325) (7.511,3.328) (7.525,3.331) (7.539,3.333) (7.554,3.336) (7.568,3.339) (7.582,3.341) (7.597,3.344) (7.611,3.346) (7.625,3.347) (7.64,3.349) (7.654,3.35) (7.668,3.35) (7.683,3.35) (7.697,3.349) (7.711,3.348) (7.726,3.347) (7.74,3.345) (7.754,3.343) (7.769,3.34) (7.783,3.338) (7.797,3.335) (7.812,3.332) (7.826,3.329) (7.84,3.327) (7.855,3.324) (7.869,3.322) (7.883,3.32) (7.898,3.319) (7.912,3.318) (7.926,3.317) (7.941,3.317) (7.955,3.317) (7.969,3.318) (7.984,3.319) (7.998,3.321) (8.012,3.323) (8.027,3.325) (8.041,3.328) (8.055,3.33) (8.07,3.333) (8.084,3.336) (8.098,3.338) (8.113,3.341) (8.127,3.343) (8.141,3.345) (8.156,3.347) (8.17,3.348) (8.184,3.349) (8.199,3.35) (8.213,3.35) (8.227,3.349) (8.242,3.348) (8.256,3.347) (8.27,3.345) (8.285,3.343) (8.299,3.341) (8.313,3.338) (8.328,3.335) (8.342,3.333) (8.356,3.33) (8.371,3.327) (8.385,3.325) (8.399,3.323) (8.414,3.321) (8.428,3.319) (8.442,3.318) (8.457,3.317) (8.471,3.317) (8.485,3.317) (8.5,3.318) (8.514,3.319) (8.528,3.321) (8.543,3.322) (8.557,3.325) (8.571,3.327) (8.586,3.33) (8.6,3.332)};
\node[bwd,left] at (8.6,3.76) {\scriptsize $0.58949$};
\node[ink2] at (7.056,2.96) {\scriptsize $N=11$};
\node[fwd] at (3.396,3.653) {\scriptsize $N=51$};
\end{scope}
\end{tikzpicture}
\figcap{Partial sums of the Fourier series of a function with a unit jump at $0$. The first peak moves towards the jump as $N$ grows and its height does not decrease: it is $0.58955$ for $N=51$ and tends to $0.58949$.}
\end{center}
Smoothness of $f$ is decay of $\hat f$: the coefficients of a function with a jump fall like $1/n$, with a corner
like $1/n^2$, and for an analytic function geometrically. \chk{F6} This is the dictionary by which regularity of
solutions is read off from the multipliers $e^{-k\lambda_nt}$ (smoothing) and $\cos c\sqrt{\lambda_n}t$ (no
smoothing).
\section{Inhomogeneous boundary values}
If the boundary values are not zero the solutions no longer form a vector space but an affine space over the
solutions of the homogeneous problem. The source's method of shifting the data \src{\S5.6} chooses an origin: any
function $w$ with the right boundary values, after which $u-w$ solves a homogeneous problem with a source. For
$u_t=u_{xx}$ with $u(0)=0$, $u(1)=1$ the natural choice is the steady state $w=x$, and the solution from zero
initial data is $x+\sum_n\frac{2(-1)^n}{n\pi}e^{-n^2\pi^2t}\sin n\pi x$. \chk{F8}
\section*{Notes}
Boundary conditions as Lagrangian subspaces go back to von Neumann's extension theory; for ordinary
differential operators see \cite{EvMa99}, and for networks of intervals \cite{KoSc99}. Fourier analysis on locally
compact abelian groups is in \cite{Rud62}. The convergence theorems and the Gibbs phenomenon are treated as in the
source; for a continuous function with a divergent Fourier series see \cite{Kat04}.
\section*{Exercises}
\begin{exercise}\label{exer:lagr}
Verify the last column of the table of boundary conditions.
\end{exercise}
\begin{exercise}\label{exer:dn}
Find the eigenvalues and eigenfunctions for the Dirichlet condition at $0$ and the Neumann condition at $l$, and
recover \Cref{exer:period}.
\end{exercise}
\begin{exercise}\label{exer:robin0}
Derive the criterion $a_0+a_l=-a_0a_ll$ for a zero eigenvalue of the Robin problem.
\end{exercise}
\begin{exercise}\label{exer:gibbs}
Given $\int_0^\pi\frac{\sin\theta}\theta d\theta=1.85194$, find the overshoot of the partial sums of a function
with a jump of height $J$.
\end{exercise}
\part{Space}
\chapter{Harmonic Functions}\label{ch:harmonic}
Laplace's equation has no time, and its theory is a theory of how a function on a domain is controlled by its
boundary values and by the shape of the domain \src{Ch.~6}. Harmonic functions form a sheaf; this chapter
describes its symmetries, its global sections in terms of the boundary, and its cohomology.
\section{The sheaf and its two principles}
For $U$ open in $\R^n$ let $\HH(U)$ be the space of harmonic functions on $U$. Restrictions of harmonic functions
are harmonic and local harmonic functions glue, so $\HH$ is a sheaf, the solution sheaf of $\Lap$.
\begin{theorem}\label{thm:mvp}
\Scl[source, \S6.1 and \S6.3] \emph{(Mean value.)} A continuous function is harmonic if and only if its value at each
point equals its average over every sphere centred there and contained in the domain. \emph{(Maximum
principle.)} A harmonic function on a bounded connected domain, continuous up to the boundary, attains its
maximum and minimum only on the boundary unless it is constant. Hence the Dirichlet problem has at most one
solution, and it depends continuously on the boundary values in the uniform norm.
\end{theorem}
\noindent For a harmonic cubic the circle average equals the central value to twelve digits; for $x^2$ it exceeds it
by $r^2/2$. On a grid, a discrete harmonic function with random boundary values stays strictly between their
extremes. \chk{H1, H2}
\section{Naturality}
\begin{theorem}[Symmetries of the Laplacian]\label{thm:conformal}
\Scl
\begin{enumerate}[label=(\alph*)]
\item $\Lap(u\circ g)=(\Lap u)\circ g$ for every rigid motion $g$, and the second-order operators with this property
are $a\Lap+b$.
\item In the plane, $\Lap(u\circ f)=|f'|^2(\Lap u)\circ f$ for every holomorphic $f$. Hence $\HH$ is a sheaf on
Riemann surfaces: harmonicity is conformally invariant in dimension two.
\item In dimension $n\ge3$ the Kelvin transform $u\mapsto|x|^{2-n}u(x/|x|^2)$ preserves harmonic functions.
\end{enumerate}
\end{theorem}
\noindent Statements (a) and (b) are polynomial identities, verified exactly for a rotation with rational entries
and for $f(z)=z^2$; a shear fails (a). \chk{H4} By \Cref{prop:tonti} this is the expected answer: the Laplacian is
built from $\dd$, which is natural for everything, and a Hodge star, which is natural for isometries, and in the
plane the star on $1$-forms happens to depend only on the conformal class. The source's computation of the
Laplacian in polar coordinates \src{\S6.1} is (a) used to choose coordinates adapted to the rotation group; the
separated solutions $r^n\cos n\theta$, $r^n\sin n\theta$ are the real and imaginary parts of $z^n$, and (b) explains
why conformal mapping solves plane potential problems.
\section{From the boundary: Poisson and Brownian motion}
\begin{theorem}[Poisson's formula]\label{thm:poisson}
\Scl[source, \S6.3, Theorem 1] For continuous $h$ on the circle of radius $a$,
\[ u(r,\theta)=\frac{a^2-r^2}{2\pi}\int_0^{2\pi}\frac{h(\phi)\,d\phi}{a^2-2ar\cos(\theta-\phi)+r^2} \]
is the harmonic function in the disk with boundary values $h$.
\end{theorem}
\noindent The kernel is positive and has total mass one. \chk{H3} So the solution operator of the Dirichlet problem
is a Markov kernel from the domain to its boundary: to each interior point $x$ it assigns a probability measure
on the boundary, the \term{harmonic measure} seen from $x$. The mean value property, the maximum principle and
the continuity of the solution in the data are all properties of Markov kernels.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3)--(0,0)--(5.6,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-\pi$};
\draw[ink2] (2.8,0)--(2.8,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (5.6,0)--(5.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $\pi$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.395)--(-0.07,1.395) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,2.791)--(-0.07,2.791) node[left,inner sep=1.5pt]{\scriptsize $2$};
\node[below] at (2.8,-0.38) {\scriptsize $\theta$};
\draw[thick,ink2] plot coordinates {(0,0.12) (0.023,0.12) (0.047,0.12) (0.07,0.12) (0.093,0.12) (0.117,0.12) (0.14,0.12) (0.163,0.12) (0.187,0.121) (0.21,0.121) (0.233,0.121) (0.257,0.121) (0.28,0.122) (0.303,0.122) (0.327,0.122) (0.35,0.123) (0.373,0.123) (0.397,0.124) (0.42,0.124) (0.443,0.125) (0.467,0.126) (0.49,0.126) (0.513,0.127) (0.537,0.128) (0.56,0.128) (0.583,0.129) (0.607,0.13) (0.63,0.131) (0.653,0.132) (0.677,0.132) (0.7,0.133) (0.723,0.134) (0.747,0.135) (0.77,0.137) (0.793,0.138) (0.817,0.139) (0.84,0.14) (0.863,0.141) (0.887,0.143) (0.91,0.144) (0.933,0.145) (0.957,0.147) (0.98,0.148) (1.003,0.15) (1.027,0.151) (1.05,0.153) (1.073,0.155) (1.097,0.157) (1.12,0.158) (1.143,0.16) (1.167,0.162) (1.19,0.164) (1.213,0.166) (1.237,0.169) (1.26,0.171) (1.283,0.173) (1.307,0.175) (1.33,0.178) (1.353,0.18) (1.377,0.183) (1.4,0.185) (1.423,0.188) (1.447,0.191) (1.47,0.194) (1.493,0.197) (1.517,0.2) (1.54,0.203) (1.563,0.206) (1.587,0.209) (1.61,0.213) (1.633,0.216) (1.657,0.22) (1.68,0.223) (1.703,0.227) (1.727,0.231) (1.75,0.235) (1.773,0.239) (1.797,0.243) (1.82,0.247) (1.843,0.251) (1.867,0.256) (1.89,0.26) (1.913,0.265) (1.937,0.269) (1.96,0.274) (1.983,0.279) (2.007,0.284) (2.03,0.289) (2.053,0.294) (2.077,0.299) (2.1,0.304) (2.123,0.309) (2.147,0.314) (2.17,0.319) (2.193,0.324) (2.217,0.329) (2.24,0.334) (2.263,0.339) (2.287,0.344) (2.31,0.349) (2.333,0.354) (2.357,0.359) (2.38,0.364) (2.403,0.368) (2.427,0.373) (2.45,0.377) (2.473,0.381) (2.497,0.385) (2.52,0.389) (2.543,0.393) (2.567,0.396) (2.59,0.399) (2.613,0.402) (2.637,0.404) (2.66,0.406) (2.683,0.408) (2.707,0.41) (2.73,0.411) (2.753,0.412) (2.777,0.412) (2.8,0.412) (2.823,0.412) (2.847,0.412) (2.87,0.411) (2.893,0.41) (2.917,0.408) (2.94,0.406) (2.963,0.404) (2.987,0.402) (3.01,0.399) (3.033,0.396) (3.057,0.393) (3.08,0.389) (3.103,0.385) (3.127,0.381) (3.15,0.377) (3.173,0.373) (3.197,0.368) (3.22,0.364) (3.243,0.359) (3.267,0.354) (3.29,0.349) (3.313,0.344) (3.337,0.339) (3.36,0.334) (3.383,0.329) (3.407,0.324) (3.43,0.319) (3.453,0.314) (3.477,0.309) (3.5,0.304) (3.523,0.299) (3.547,0.294) (3.57,0.289) (3.593,0.284) (3.617,0.279) (3.64,0.274) (3.663,0.269) (3.687,0.265) (3.71,0.26) (3.733,0.256) (3.757,0.251) (3.78,0.247) (3.803,0.243) (3.827,0.239) (3.85,0.235) (3.873,0.231) (3.897,0.227) (3.92,0.223) (3.943,0.22) (3.967,0.216) (3.99,0.213) (4.013,0.209) (4.037,0.206) (4.06,0.203) (4.083,0.2) (4.107,0.197) (4.13,0.194) (4.153,0.191) (4.177,0.188) (4.2,0.185) (4.223,0.183) (4.247,0.18) (4.27,0.178) (4.293,0.175) (4.317,0.173) (4.34,0.171) (4.363,0.169) (4.387,0.166) (4.41,0.164) (4.433,0.162) (4.457,0.16) (4.48,0.158) (4.503,0.157) (4.527,0.155) (4.55,0.153) (4.573,0.151) (4.597,0.15) (4.62,0.148) (4.643,0.147) (4.667,0.145) (4.69,0.144) (4.713,0.143) (4.737,0.141) (4.76,0.14) (4.783,0.139) (4.807,0.138) (4.83,0.137) (4.853,0.135) (4.877,0.134) (4.9,0.133) (4.923,0.132) (4.947,0.132) (4.97,0.131) (4.993,0.13) (5.017,0.129) (5.04,0.128) (5.063,0.128) (5.087,0.127) (5.11,0.126) (5.133,0.126) (5.157,0.125) (5.18,0.124) (5.203,0.124) (5.227,0.123) (5.25,0.123) (5.273,0.122) (5.297,0.122) (5.32,0.122) (5.343,0.121) (5.367,0.121) (5.39,0.121) (5.413,0.121) (5.437,0.12) (5.46,0.12) (5.483,0.12) (5.507,0.12) (5.53,0.12) (5.553,0.12) (5.577,0.12) (5.6,0.12)};
\draw[thick,fwd] plot coordinates {(0,0.056) (0.023,0.056) (0.047,0.056) (0.07,0.056) (0.093,0.056) (0.117,0.056) (0.14,0.056) (0.163,0.056) (0.187,0.056) (0.21,0.056) (0.233,0.056) (0.257,0.057) (0.28,0.057) (0.303,0.057) (0.327,0.057) (0.35,0.058) (0.373,0.058) (0.397,0.058) (0.42,0.059) (0.443,0.059) (0.467,0.059) (0.49,0.06) (0.513,0.06) (0.537,0.061) (0.56,0.061) (0.583,0.061) (0.607,0.062) (0.63,0.063) (0.653,0.063) (0.677,0.064) (0.7,0.064) (0.723,0.065) (0.747,0.066) (0.77,0.066) (0.793,0.067) (0.817,0.068) (0.84,0.069) (0.863,0.07) (0.887,0.071) (0.91,0.072) (0.933,0.073) (0.957,0.074) (0.98,0.075) (1.003,0.076) (1.027,0.077) (1.05,0.078) (1.073,0.079) (1.097,0.081) (1.12,0.082) (1.143,0.084) (1.167,0.085) (1.19,0.087) (1.213,0.088) (1.237,0.09) (1.26,0.092) (1.283,0.094) (1.307,0.096) (1.33,0.098) (1.353,0.1) (1.377,0.102) (1.4,0.105) (1.423,0.107) (1.447,0.11) (1.47,0.112) (1.493,0.115) (1.517,0.118) (1.54,0.121) (1.563,0.125) (1.587,0.128) (1.61,0.132) (1.633,0.135) (1.657,0.139) (1.68,0.144) (1.703,0.148) (1.727,0.153) (1.75,0.158) (1.773,0.163) (1.797,0.169) (1.82,0.174) (1.843,0.181) (1.867,0.187) (1.89,0.194) (1.913,0.201) (1.937,0.209) (1.96,0.217) (1.983,0.226) (2.007,0.235) (2.03,0.245) (2.053,0.255) (2.077,0.266) (2.1,0.278) (2.123,0.29) (2.147,0.304) (2.17,0.318) (2.193,0.333) (2.217,0.348) (2.24,0.365) (2.263,0.383) (2.287,0.402) (2.31,0.422) (2.333,0.443) (2.357,0.465) (2.38,0.489) (2.403,0.513) (2.427,0.539) (2.45,0.565) (2.473,0.593) (2.497,0.621) (2.52,0.65) (2.543,0.679) (2.567,0.708) (2.59,0.736) (2.613,0.763) (2.637,0.789) (2.66,0.813) (2.683,0.835) (2.707,0.853) (2.73,0.868) (2.753,0.879) (2.777,0.886) (2.8,0.888) (2.823,0.886) (2.847,0.879) (2.87,0.868) (2.893,0.853) (2.917,0.835) (2.94,0.813) (2.963,0.789) (2.987,0.763) (3.01,0.736) (3.033,0.708) (3.057,0.679) (3.08,0.65) (3.103,0.621) (3.127,0.593) (3.15,0.565) (3.173,0.539) (3.197,0.513) (3.22,0.489) (3.243,0.465) (3.267,0.443) (3.29,0.422) (3.313,0.402) (3.337,0.383) (3.36,0.365) (3.383,0.348) (3.407,0.333) (3.43,0.318) (3.453,0.304) (3.477,0.29) (3.5,0.278) (3.523,0.266) (3.547,0.255) (3.57,0.245) (3.593,0.235) (3.617,0.226) (3.64,0.217) (3.663,0.209) (3.687,0.201) (3.71,0.194) (3.733,0.187) (3.757,0.181) (3.78,0.174) (3.803,0.169) (3.827,0.163) (3.85,0.158) (3.873,0.153) (3.897,0.148) (3.92,0.144) (3.943,0.139) (3.967,0.135) (3.99,0.132) (4.013,0.128) (4.037,0.125) (4.06,0.121) (4.083,0.118) (4.107,0.115) (4.13,0.112) (4.153,0.11) (4.177,0.107) (4.2,0.105) (4.223,0.102) (4.247,0.1) (4.27,0.098) (4.293,0.096) (4.317,0.094) (4.34,0.092) (4.363,0.09) (4.387,0.088) (4.41,0.087) (4.433,0.085) (4.457,0.084) (4.48,0.082) (4.503,0.081) (4.527,0.079) (4.55,0.078) (4.573,0.077) (4.597,0.076) (4.62,0.075) (4.643,0.074) (4.667,0.073) (4.69,0.072) (4.713,0.071) (4.737,0.07) (4.76,0.069) (4.783,0.068) (4.807,0.067) (4.83,0.066) (4.853,0.066) (4.877,0.065) (4.9,0.064) (4.923,0.064) (4.947,0.063) (4.97,0.063) (4.993,0.062) (5.017,0.061) (5.04,0.061) (5.063,0.061) (5.087,0.06) (5.11,0.06) (5.133,0.059) (5.157,0.059) (5.18,0.059) (5.203,0.058) (5.227,0.058) (5.25,0.058) (5.273,0.057) (5.297,0.057) (5.32,0.057) (5.343,0.057) (5.367,0.056) (5.39,0.056) (5.413,0.056) (5.437,0.056) (5.46,0.056) (5.483,0.056) (5.507,0.056) (5.53,0.056) (5.553,0.056) (5.577,0.056) (5.6,0.056)};
\draw[thick,bwd] plot coordinates {(0,0.018) (0.023,0.018) (0.047,0.018) (0.07,0.018) (0.093,0.018) (0.117,0.018) (0.14,0.018) (0.163,0.018) (0.187,0.018) (0.21,0.018) (0.233,0.018) (0.257,0.018) (0.28,0.018) (0.303,0.019) (0.327,0.019) (0.35,0.019) (0.373,0.019) (0.397,0.019) (0.42,0.019) (0.443,0.019) (0.467,0.019) (0.49,0.019) (0.513,0.02) (0.537,0.02) (0.56,0.02) (0.583,0.02) (0.607,0.02) (0.63,0.02) (0.653,0.021) (0.677,0.021) (0.7,0.021) (0.723,0.021) (0.747,0.022) (0.77,0.022) (0.793,0.022) (0.817,0.022) (0.84,0.023) (0.863,0.023) (0.887,0.023) (0.91,0.024) (0.933,0.024) (0.957,0.024) (0.98,0.025) (1.003,0.025) (1.027,0.026) (1.05,0.026) (1.073,0.026) (1.097,0.027) (1.12,0.027) (1.143,0.028) (1.167,0.028) (1.19,0.029) (1.213,0.03) (1.237,0.03) (1.26,0.031) (1.283,0.032) (1.307,0.032) (1.33,0.033) (1.353,0.034) (1.377,0.035) (1.4,0.036) (1.423,0.037) (1.447,0.038) (1.47,0.039) (1.493,0.04) (1.517,0.041) (1.54,0.042) (1.563,0.044) (1.587,0.045) (1.61,0.046) (1.633,0.048) (1.657,0.05) (1.68,0.051) (1.703,0.053) (1.727,0.055) (1.75,0.057) (1.773,0.06) (1.797,0.062) (1.82,0.065) (1.843,0.068) (1.867,0.071) (1.89,0.074) (1.913,0.077) (1.937,0.081) (1.96,0.085) (1.983,0.09) (2.007,0.094) (2.03,0.1) (2.053,0.105) (2.077,0.112) (2.1,0.118) (2.123,0.126) (2.147,0.134) (2.17,0.143) (2.193,0.154) (2.217,0.165) (2.24,0.178) (2.263,0.192) (2.287,0.208) (2.31,0.226) (2.333,0.246) (2.357,0.27) (2.38,0.297) (2.403,0.328) (2.427,0.364) (2.45,0.406) (2.473,0.455) (2.497,0.514) (2.52,0.583) (2.543,0.666) (2.567,0.766) (2.59,0.887) (2.613,1.033) (2.637,1.209) (2.66,1.419) (2.683,1.664) (2.707,1.937) (2.73,2.222) (2.753,2.482) (2.777,2.67) (2.8,2.739) (2.823,2.67) (2.847,2.482) (2.87,2.222) (2.893,1.937) (2.917,1.664) (2.94,1.419) (2.963,1.209) (2.987,1.033) (3.01,0.887) (3.033,0.766) (3.057,0.666) (3.08,0.583) (3.103,0.514) (3.127,0.455) (3.15,0.406) (3.173,0.364) (3.197,0.328) (3.22,0.297) (3.243,0.27) (3.267,0.246) (3.29,0.226) (3.313,0.208) (3.337,0.192) (3.36,0.178) (3.383,0.165) (3.407,0.154) (3.43,0.143) (3.453,0.134) (3.477,0.126) (3.5,0.118) (3.523,0.112) (3.547,0.105) (3.57,0.1) (3.593,0.094) (3.617,0.09) (3.64,0.085) (3.663,0.081) (3.687,0.077) (3.71,0.074) (3.733,0.071) (3.757,0.068) (3.78,0.065) (3.803,0.062) (3.827,0.06) (3.85,0.057) (3.873,0.055) (3.897,0.053) (3.92,0.051) (3.943,0.05) (3.967,0.048) (3.99,0.046) (4.013,0.045) (4.037,0.044) (4.06,0.042) (4.083,0.041) (4.107,0.04) (4.13,0.039) (4.153,0.038) (4.177,0.037) (4.2,0.036) (4.223,0.035) (4.247,0.034) (4.27,0.033) (4.293,0.032) (4.317,0.032) (4.34,0.031) (4.363,0.03) (4.387,0.03) (4.41,0.029) (4.433,0.028) (4.457,0.028) (4.48,0.027) (4.503,0.027) (4.527,0.026) (4.55,0.026) (4.573,0.026) (4.597,0.025) (4.62,0.025) (4.643,0.024) (4.667,0.024) (4.69,0.024) (4.713,0.023) (4.737,0.023) (4.76,0.023) (4.783,0.022) (4.807,0.022) (4.83,0.022) (4.853,0.022) (4.877,0.021) (4.9,0.021) (4.923,0.021) (4.947,0.021) (4.97,0.02) (4.993,0.02) (5.017,0.02) (5.04,0.02) (5.063,0.02) (5.087,0.02) (5.11,0.019) (5.133,0.019) (5.157,0.019) (5.18,0.019) (5.203,0.019) (5.227,0.019) (5.25,0.019) (5.273,0.019) (5.297,0.019) (5.32,0.018) (5.343,0.018) (5.367,0.018) (5.39,0.018) (5.413,0.018) (5.437,0.018) (5.46,0.018) (5.483,0.018) (5.507,0.018) (5.53,0.018) (5.553,0.018) (5.577,0.018) (5.6,0.018)};
\node[bwd,right] at (3.112,2.721) {\scriptsize $r=0.85$};
\node[fwd,right] at (3.468,1.005) {\scriptsize $0.6$};
\node[ink2] at (4.583,0.335) {\scriptsize $0.3$};
\end{scope}
\begin{scope}[shift={(9.2,1.5)}]
\fill[fwd!10,even odd rule] (0,0) circle (1.65) (0,0) circle (0.55);
\draw[thick] (0,0) circle (1.65) (0,0) circle (0.55);
\draw[thick,bwd,-{Stealth[length=5pt]}] (1.1,0) arc (0:350:1.1);
\node[bwd] at (0,1.32) {\scriptsize $\gamma$};
\node at (0,-2.05) {\scriptsize $\oint_\gamma\star\,\dd\log r=2\pi$};
\end{scope}
\end{tikzpicture}
\figcap{Left: the Poisson kernel of the unit disk at three radii; as $r\to1$ it concentrates at the boundary point nearest to the observer. Right: on an annulus the conjugate differential of $\log r$ has a non-zero period around the hole, so $\log r$ is not the real part of a holomorphic function (\Cref{prop:annulus}).}
\end{center}
\begin{theorem}[Kakutani]\label{thm:kakutani}
\Scl[source, \S6.1, Example 4; \cite{Kak44}] The harmonic measure from $x$ is the distribution of the point at which
Brownian motion started at $x$ first reaches the boundary: $u(x)=\E\,h(B_\tau)$.
\end{theorem}
\noindent On a $7\times7$ grid the probability that a random walk leaves through the top side solves the five-point
equation, agrees with the limit of the walk to ten digits, and equals $\tfrac14$ at the centre. \chk{H7} Composition
of kernels gives the consistency of harmonic measures for nested domains: to exit the large domain, first exit
the small one. In the language of \cite{ESCHC}, the Dirichlet problem is a morphism in $\Stoch$, and the heat
semigroup of \Cref{ch:diffusions} and the Poisson kernel are two views of one process.
\section{The annulus has cohomology}
In the plane, a harmonic function $u$ on a disk is the real part of a holomorphic function: it has a
\term{harmonic conjugate} $v$ with $\dd v=\star\,\dd u=-u_y\,dx+u_x\,dy$.
\begin{proposition}\label{prop:annulus}
\Scl On a plane domain $U$ the form $\star\,\dd u$ is closed for harmonic $u$, and $u$ has a harmonic conjugate on
$U$ if and only if its class in $H^1(U;\R)$ vanishes. On an annulus this class is the period
$\oint\star\,\dd u$ around the hole. The period of $\log r$ is $2\pi$; the periods of $r^{\pm n}\cos n\theta$ and
$r^{\pm n}\sin n\theta$ are zero.
\end{proposition}
\noindent The source's solution of the Dirichlet problem in an annulus \src{\S6.4} has, besides the terms that
appear for the disk, the functions $r^{-n}\cos n\theta$, $r^{-n}\sin n\theta$ and $\log r$. The first two families
are the real parts of $z^{-n}$, holomorphic in the annulus. The last is not the real part of anything: it
generates the cokernel of $\mathrm{Re}\colon\OO(U)\to\HH(U)$, which is $H^1(U;\R)=\R$. The three periods were
computed numerically as $2\pi$, $0$, $0$. \chk{H5} This is the first cohomology class of the volume that is not
zero, and its discrete counterpart is computed in \Cref{ch:numerics}.
\section{Corners}
\begin{proposition}\label{prop:wedge}
\Scl[source, \S6.4] In a wedge of opening $\beta$ the harmonic functions vanishing on both edges are spanned near
the vertex by $r^{n\pi/\beta}\sin(n\pi\theta/\beta)$. The gradient of the first behaves like $r^{\pi/\beta-1}$: it is
bounded if $\beta\le\pi$ and unbounded if $\beta>\pi$.
\end{proposition}
\noindent At a re-entrant corner of angle $3\pi/2$ the exponent is $\tfrac23$ and the gradient grows like $r^{-1/3}$.
\chk{H6} Interior regularity of harmonic functions is unconditional; regularity at the boundary is a property of the
boundary. The exponent $\pi/\beta$ is the reason that finite element methods lose accuracy on an L-shaped domain
(\Cref{ch:numerics}), and it is again conformal naturality: $z\mapsto z^{\pi/\beta}$ opens the wedge to a
half-plane.
\section*{Notes}
The source's Chapter 6 is followed throughout. For conformal invariance and the Kelvin transform see
\cite{ABR01}; for Brownian motion and harmonic measure \cite{Kak44,MoPe10}; for corner singularities
\cite{Gri85}. The Dirichlet problem in rectangles and cubes \src{\S6.2} is separation of variables, taken up in
\Cref{ch:symmetry}.
\section*{Exercises}
\begin{exercise}\label{exer:mv}
For a quadratic polynomial $u$ in the plane, show that the average over a circle of radius $r$ exceeds the value
at the centre by $\tfrac{r^2}4\Lap u$.
\end{exercise}
\begin{exercise}\label{exer:conf}
Verify \Cref{thm:conformal}(b) for $f(z)=z^2$ and $u=x^2$.
\end{exercise}
\begin{exercise}\label{exer:periods}
Compute $\oint\star\,\dd u$ around the unit circle for $u=\log r$ and for $u=\mathrm{Re}(1/z)$.
\end{exercise}
\begin{exercise}\label{exer:corner}
Compare the behaviour of the gradient of the first singular function at the re-entrant corner of an L-shaped
room and at one of its convex corners.
\end{exercise}
\chapter{Green's Identities and Green's Functions}\label{ch:green}
The source derives its remaining general facts about harmonic functions from two integral identities and then
inverts the Laplacian by a kernel \src{Ch.~7}. Both steps have a structural reading, and the second has a
compositional one: boundary-value problems can be glued.
\section{Stokes' theorem, twice}
\begin{proposition}[Green's identities]\label{prop:green}
\Scl[source, \S7.1--7.2] For functions on a bounded domain $D$ with smooth boundary,
\begin{align*}
\int_Dv\,\Lap u+\int_D\nabla v\cdot\nabla u&=\int_{\pd D}v\,\pd_nu,\\
\int_D(u\,\Lap v-v\,\Lap u)&=\int_{\pd D}(u\,\pd_nv-v\,\pd_nu).
\end{align*}
\end{proposition}
\noindent Both are Stokes' theorem for the form $v\star\dd u$. On a grid the first is the factorization of
\Cref{prop:tonti}: the Dirichlet Laplacian is $D^{\T}D$. \chk{G1} The right-hand side of the second is a symplectic
form on the space of pairs (boundary values, normal derivatives), the same form as in \Cref{thm:lagrangian}.
\begin{corollary}[The Dirichlet-to-Neumann map is symmetric]\label{cor:dtn}
\Srf The pairs $(u|_{\pd D},\pd_nu|_{\pd D})$ for $u$ harmonic in $D$ form a Lagrangian subspace of the boundary data.
It is the graph of the operator $\Lambda_D\colon h\mapsto\pd_nu$, where $u$ is the harmonic function with boundary
values $h$; hence $\Lambda_D$ is symmetric, and by the first identity non-negative.
\end{corollary}
\section{Dirichlet's principle}
\begin{theorem}\label{thm:dirichlet}
\Scl[source, \S7.1] Among functions with given boundary values, the harmonic one has the least energy
$E(w)=\tfrac12\int_D|\nabla w|^2$. More precisely, if $u$ is harmonic and $z$ vanishes on the boundary, then
$E(u+z)=E(u)+E(z)$.
\end{theorem}
\noindent The identity is the Pythagorean theorem for the energy inner product: harmonic functions are orthogonal to
functions vanishing on the boundary. \chk{G2} So the solution operator of the Dirichlet problem is an orthogonal
projection. The same holds in any subspace, which is the Rayleigh--Ritz or Galerkin method \src{\S7.1, Exercise
7}: the approximate solution in a trial space is the projection of the true one, and its error is orthogonal to
the trial space. \chk{G7} Finite elements are this statement with a particular trial space (\Cref{ch:numerics}).
\section{The kernel of the inverse}
\begin{definition}[source, \S7.3]
The \term{Green's function} of $D$ with pole $x_0$ is the function $G(x,x_0)$ that is harmonic in $D$ away from
$x_0$, vanishes on $\pd D$, and differs from $-1/(4\pi|x-x_0|)$ by a function harmonic near $x_0$ (in three
dimensions).
\end{definition}
\begin{theorem}\label{thm:greenfn}
\Scl[source, \S7.3] \emph{(a)} The solution of $\Lap u=f$ in $D$, $u=h$ on $\pd D$ is
$u(x_0)=\int_{\pd D}h\,\pd_nG+\int_DfG$. \emph{(b)} $G(x,x_0)=G(x_0,x)$. \emph{(c)} $G\le0$ in $D$, and
$\pd_nG\ge0$ on $\pd D$ with total mass one.
\end{theorem}
\noindent In operator terms, $-G$ is the kernel of the inverse of $-\Lap$ with Dirichlet conditions; (b) says the
inverse is symmetric because the operator is; (c) says the inverse preserves positivity, which is the maximum
principle, and identifies $\pd_nG$ with the harmonic measure of \Cref{thm:poisson}. On eight grid points the
inverse of the Dirichlet Laplacian is the symmetric positive matrix $\min(i,j)(9-\max(i,j))/9$, the discrete form
of the kernel $x(1-y)$ for $x<y$ on the unit interval. \chk{G3, G8}
\section{Half-space and ball}
\begin{theorem}\label{thm:images-green}
\Scl[source, \S7.4] For the half-space $\{z>0\}$, $G(x,x_0)=-\frac1{4\pi|x-x_0|}+\frac1{4\pi|x-x_0^*|}$ with
$x_0^*$ the mirror image of $x_0$. For the ball of radius $a$,
$G(x,x_0)=-\frac1{4\pi|x-x_0|}+\frac{a}{|x_0|}\frac1{4\pi|x-x_0^*|}$ with $x_0^*=a^2x_0/|x_0|^2$, and
$\pd_nG=\dfrac{a^2-|x_0|^2}{4\pi a|x-x_0|^3}$ on the sphere.
\end{theorem}
\noindent The first is \Cref{thm:images} for the reflection in a plane with the sign character. The second is the
same after a conformal change: inversion in the sphere exchanges the ball with its exterior, and the factor
$a/|x_0|$ is the weight of the Kelvin transform of \Cref{thm:conformal}(c). Both kernels vanish on the boundary and
have normal derivatives of total mass one; the Poisson kernel of the ball returns the value $r_0$ of the harmonic
function $z$ at height $r_0$. \chk{G4, G5}
\begin{center}
\begin{tikzpicture}
\begin{scope}
\fill[fwd!10] (-2.2,0) rectangle (2.4,1.9);
\draw[very thick] (-2.2,0)--(2.4,0);
\fill[fwd] (0,1.1) circle (2pt) node[above left]{\scriptsize $x_0\ (+)$};
\fill[bwd] (0,-1.1) circle (2pt) node[below left]{\scriptsize $x_0^*\ (-)$};
\draw[densely dashed,ink2] (0,1.1)--(0,-1.1);
\fill (1.5,0) circle (1.4pt) node[below right]{\scriptsize $x$};
\draw[fwd] (0,1.1)--(1.5,0);
\draw[bwd] (0,-1.1)--(1.5,0);
\node at (0.1,-2.0) {\scriptsize half-space: equal distances on the wall};
\end{scope}
\begin{scope}[xshift=6.6cm]
\fill[fwd!10] (0,0) circle (1.4);
\draw[very thick] (0,0) circle (1.4);
\draw[densely dashed,ink2] (0,0)--(3.1,0);
\fill (0,0) circle (1pt) node[below left]{\scriptsize $0$};
\fill[fwd] (0.7,0) circle (2pt) node[below]{\scriptsize $x_0$};
\fill[bwd] (2.8,0) circle (2pt) node[below]{\scriptsize $x_0^*$};
\fill (0.7,1.2124) circle (1.4pt) node[above]{\scriptsize $x$};
\draw[fwd] (0.7,0)--(0.7,1.2124);
\draw[bwd] (2.8,0)--(0.7,1.2124);
\node at (1.2,-2.0) {\scriptsize ball: $|x-x_0^*|/|x-x_0|=a/|x_0|$ on the sphere};
\end{scope}
\end{tikzpicture}
\figcap{The two image constructions. In the half-space the image is the mirror point with the opposite charge. In the ball of radius $a$ it is the inverse point $x_0^*=a^2x_0/|x_0|^2$ with charge $-a/|x_0|$: on the sphere the ratio of the two distances is constant (here $a=1.4$, $|x_0|=0.7$, ratio $2$), which is what makes the sum vanish there.}
\end{center}
\section{Gluing}
Let a \term{network} be a finite graph with positive conductances on its edges and a chosen set of
\term{terminals}; a potential is harmonic if Kirchhoff's law holds at every non-terminal vertex. Eliminating the
interior vertices from the Laplacian (a Schur complement, or Kron reduction) gives the matrix $\Lambda$ that sends
terminal potentials to terminal currents: the discrete Dirichlet-to-Neumann map.
\begin{theorem}[Black-boxing is a functor]\label{thm:gluing}
\Sli[\cite{BaFo18}] Networks with terminals are the morphisms of a category in which composition glues terminals.
Sending a network to the graph of its Dirichlet-to-Neumann map is a functor to the category $\LagRel$ of
symplectic vector spaces and Lagrangian relations. In particular the response of a glued network is determined
by the responses of its parts.
\end{theorem}
\noindent For ten pairs of random networks glued along two shared terminals, the response of the whole agreed with
the composite of the responses of the parts to ten digits; conductances $2$ and $3$ in series give $6/5$. \chk{G6}
The continuous statement is the same with \Cref{cor:dtn} in place of the matrix: if $D=D_1\cup_\Sigma D_2$, the
Lagrangian subspace of $D$ is the composite of those of $D_1$ and $D_2$ over the interface $\Sigma$. Domain
decomposition methods in numerical analysis are this functoriality used as an algorithm: solve on the pieces,
then solve an equation on the interface.
\begin{center}
\begin{tikzpicture}
\tikzset{ntm/.style={circle,fill,inner sep=1.8pt},nin/.style={circle,draw,thick,fill=white,inner sep=1.6pt}}
\begin{scope}
\node[ntm] (a) at (0,0.6) {}; \node[nin] (b) at (1,1.3) {}; \node[nin] (c) at (1,-0.1) {};
\node[ntm,accent] (s1) at (2.1,1.2) {}; \node[ntm,accent] (s2) at (2.1,0) {};
\draw[thick] (a)--(b)--(c)--(a) (b)--(s1) (c)--(s2) (b)--(s2);
\node at (1,-0.8) {\scriptsize $D_1$};
\end{scope}
\begin{scope}[xshift=3.3cm]
\node[ntm,accent] (s1) at (0,1.2) {}; \node[ntm,accent] (s2) at (0,0) {};
\node[nin] (d) at (1.1,0.6) {}; \node[ntm] (e) at (2.1,1.2) {}; \node[ntm] (f) at (2.1,0) {};
\draw[thick] (s1)--(d)--(s2) (d)--(e)--(f)--(d);
\node at (1,-0.8) {\scriptsize $D_2$};
\end{scope}
\draw[->,thick,ink2] (6.1,0.6)--(7.0,0.6);
\begin{scope}[xshift=7.7cm]
\node[ntm] (a) at (0,0.6) {}; \node[nin] (b) at (1,1.3) {}; \node[nin] (c) at (1,-0.1) {};
\node[nin,accent] (s1) at (2.1,1.2) {}; \node[nin,accent] (s2) at (2.1,0) {};
\node[nin] (d) at (3.2,0.6) {}; \node[ntm] (e) at (4.2,1.2) {}; \node[ntm] (f) at (4.2,0) {};
\draw[thick] (a)--(b)--(c)--(a) (b)--(s1) (c)--(s2) (b)--(s2) (s1)--(d)--(s2) (d)--(e)--(f)--(d);
\node at (2.1,-0.8) {\scriptsize $D_1\cup_\Sigma D_2$};
\end{scope}
\end{tikzpicture}
\figcap{Gluing two networks along the shared terminals $\Sigma$ (in colour). Filled nodes are terminals, open ones are interior. After gluing, the nodes of $\Sigma$ are interior, and Kirchhoff's law there is the condition that the currents from the two sides cancel.}
\end{center}
\[\begin{tikzcd}[column sep=2.2em,row sep=2.2em]
& & L_{D}\arrow[dl]\arrow[dr] & & \\
& L_{D_1}\arrow[dl]\arrow[dr] & & L_{D_2}\arrow[dl]\arrow[dr] & \\
V_{\pd_1} & & V_\Sigma & & V_{\pd_2}
\end{tikzcd}\]
Each $V$ is a space of pairs (potential, current) on a piece of boundary, with the sign of the current reversed on
one side of $\Sigma$ so that what leaves $D_1$ enters $D_2$. The top object is the image of the pullback: pairs of
solutions on the two parts that agree on the interface in value and in flux, which is to say solutions on the
whole.
\begin{remark}
\Cref{thm:lagrangian}, \Cref{cor:dtn} and \Cref{thm:gluing} are one statement at three levels: boundary
conditions, solutions and parts of a domain are all Lagrangian relations on spaces of boundary data, and a
boundary-value problem is solved by composing them. An elliptic problem is a morphism from the empty boundary
to the boundary data, and its solution is the composite with the boundary condition. This is as close as the
elliptic theory comes to the time-evolution functors of the hyperbolic and parabolic theories.
\end{remark}
\section*{Notes}
The source's Chapter 7 is followed for the identities, Dirichlet's principle and Green's functions. Black-boxing
of passive linear networks as a functor to Lagrangian relations is due to Baez and Fong \cite{BaFo18}; Kron
reduction is surveyed in \cite{DoBu13}. For Dirichlet-to-Neumann operators and domain decomposition see
\cite{QuVa99}.
\section*{Exercises}
\begin{exercise}\label{exer:dtnsym}
Deduce the symmetry of $\Lambda_D$ from Green's second identity.
\end{exercise}
\begin{exercise}\label{exer:series}
Compute the Kron reduction of two conductances $g_1,g_2$ in series onto the two end terminals.
\end{exercise}
\begin{exercise}\label{exer:green1d}
Show that $-d^2/dx^2$ on $(0,1)$ with Dirichlet conditions has the inverse kernel $x(1-y)$ for $x<y$, and expand
it in eigenfunctions.
\end{exercise}
\begin{exercise}\label{exer:kelvin}
Show that for $|x|=a$ one has $|x-x_0^*|=(a/|x_0|)\,|x-x_0|$, and deduce that the Green's function of the ball
vanishes on the sphere.
\end{exercise}
\chapter{Waves and Quanta in Space}\label{ch:space}
In three dimensions the wave equation keeps its causal structure and gains a property it did not have on the
line; the diffusion kernel becomes a product; and a third evolution equation, Schr\"odinger's, turns out to be
both of the others at once \src{Ch.~9}.
\section{Cones}
\begin{theorem}[Energy and causality]\label{thm:cone}
\Scl[source, \S9.1] For $u_{tt}=c^2\Lap u$ in space the energy $\tfrac12\int(u_t^2+c^2|\nabla u|^2)$ is conserved, and
the value of $u$ at $(x_0,t_0)$ depends only on the data in the ball $|x-x_0|\le ct_0$.
\end{theorem}
\noindent The proof integrates the energy identity over a truncated cone. The cone is the characteristic variety
of \Cref{ch:equations}, $\tau^2=c^2|\xi|^2$, seen in space-time. As in \Cref{thm:causal}, restriction to a spacelike
surface is an isomorphism from solutions in its domain of dependence to pairs of data \src{\S9.3, Theorem 2}.
A function $f(t-\gamma(x))$ of a phase solves the equation for every profile $f$ exactly when the phase satisfies
the eikonal equation $|\nabla\gamma|=1/c$; these are the surfaces the source calls characteristic. A plane phase
with $|\nabla\gamma|=1/c$ passes the test and one with the wrong slope fails it. \chk{S1}
\section{Kirchhoff, Huygens and descent}
\begin{theorem}[Kirchhoff]\label{thm:kirchhoff}
\Scl[source, \S9.2] In three dimensions the solution with data $u=\varphi$, $u_t=\psi$ is
\[ u(x,t)=\frac1{4\pi c^2t}\int_{|y-x|=ct}\psi\,dS+\pd_t\Bigl[\frac1{4\pi c^2t}\int_{|y-x|=ct}\varphi\,dS\Bigr]. \]
It depends only on the data \emph{on} the sphere of radius $ct$.
\end{theorem}
\noindent This is \term{Huygens's principle}: a disturbance of bounded extent is heard for a bounded time. The
reason is visible for radial solutions: $(ru)_{tt}=c^2(ru)_{rr}$, so $ru$ solves the one-dimensional equation and
$u=(F(r-ct)+G(r+ct))/r$. \chk{S1}
In two dimensions the source obtains the solution by \term{descent}: a function of $(x,y)$ is a function of
$(x,y,z)$ that does not depend on $z$.
\begin{proposition}[Descent loses sharpness]\label{prop:descent}
\Srf[source, \S9.2] Solutions in the plane are the solutions in space invariant under translation in $z$. The
pushforward of the uniform measure on a sphere of radius $R$ under projection to the plane has the density
$\bigl(2\pi R\sqrt{R^2-\rho^2}\bigr)^{-1}$ on the disk, so
\[ u(x,t)=\frac1{2\pi c}\int_{|y-x|<ct}\frac{\psi(y)\,dy}{\sqrt{c^2t^2-|y-x|^2}}+\cdots \]
depends on the data in the whole disk, with a tail of order $1/t$.
\end{proposition}
\noindent For a source in the unit ball heard at distance $5$ with $c=1$, the three-dimensional signal is zero at
$t=3$, non-zero at $t=5$ and exactly zero again at $t=6.5$; the two-dimensional one never stops, and $t\,u(t)$ is
$1.29$, $1.002$, $1.0000$ times its limit at $t=8$, $80$, $800$. \chk{S2} Sharp propagation holds in odd dimensions
from three upwards and fails in even ones; the source's remark that three is the best of all possible dimensions
is the remark that it is the smallest of these. The structural point is that the property is not inherited by
invariants under a non-compact group: the plane is the quotient of space by a line, and the shadow of a sphere
is a disk.
\begin{center}
\begin{tikzpicture}[scale=0.8]
\begin{scope}
\draw[->,ink2] (0,0)--(3.4,0) node[right]{\small$r$};
\draw[->,ink2] (0,0)--(0,3.4) node[above]{\small$t$};
\draw[very thick,fwd] (0,0)--(3.1,3.1);
\draw[densely dashed,ink2] (2,0)--(2,3.2);
\fill[bwd] (2,2) circle (2pt);
\node at (1.6,-0.55) {\small three dimensions};
\end{scope}
\begin{scope}[xshift=6.4cm]
\fill[fwd!18] (0,0)--(3.1,3.1)--(0,3.1)--cycle;
\draw[->,ink2] (0,0)--(3.4,0) node[right]{\small$r$};
\draw[->,ink2] (0,0)--(0,3.4) node[above]{\small$t$};
\draw[very thick,fwd] (0,0)--(3.1,3.1);
\draw[densely dashed,ink2] (2,0)--(2,3.2);
\draw[very thick,bwd] (2,2)--(2,3.1);
\node at (1.6,-0.55) {\small two dimensions};
\end{scope}
\end{tikzpicture}
\figcap{The support of the response to an impulse at the origin. A listener at fixed distance (dashed) hears an
instant in three dimensions and a tail that never ends in two.}
\end{center}
\section{Singularities travel along rays}
\begin{theorem}[Propagation of singularities]\label{thm:wf}
\Sli[\cite{Hor71,DuHo72}] For a solution $u$ of $Pu=0$ with $P$ of real principal type, the wave front set
$\WF(u)\subseteq T^*M$ is contained in $\Char(P)$ and is a union of bicharacteristics, the integral curves of the
Hamiltonian vector field of the principal symbol.
\end{theorem}
\noindent For the wave equation the bicharacteristics project to light rays, and the theorem contains the source's
statement that only characteristic surfaces can carry singularities \src{\S9.3, Theorem 3}. For an elliptic
operator $\Char(P)$ is empty and solutions are smooth. For the heat equation the characteristic directions are
$dt$ and singularities do not propagate at finite speed at all. The wave front set lives on the cotangent
bundle; that singularities have a direction as well as a position is the starting point of microlocal analysis,
where the solution sheaf is studied on $T^*M$ \cite{KaSc90}.
\section{Diffusion, and Schr\"odinger's equation}
In $n$ dimensions the diffusion kernel is the product of one-dimensional kernels \src{\S9.4}: the semigroup on
$L^2(\R^n)=L^2(\R)^{\otimes n}$ is $T_t\otimes\cdots\otimes T_t$. Everything in \Cref{ch:diffusions} holds unchanged.
The free Schr\"odinger equation $iu_t=-\tfrac12\Lap u$ is formally the diffusion equation at imaginary time.
\begin{theorem}[One semigroup, two boundary values]\label{thm:wick}
\Scl[\cite{ReSi75}] Let $H$ be a non-negative self-adjoint operator. Then $z\mapsto e^{-zH}$ is a holomorphic
semigroup of contractions on $\mathrm{Re}\,z>0$, strongly continuous up to the imaginary axis. Its restriction to
$z=t>0$ is the heat semigroup; its restriction to $z=it$ is the unitary group solving $iu_t=Hu$.
\end{theorem}
\noindent The kernel $(2\pi it)^{-1/2}e^{ix^2/2t}$ solves the equation and equals the diffusion kernel evaluated at
time $it$. A Gaussian packet keeps its norm exactly while its mean square width grows as $(1+t^2)/2$. \chk{S3}
In the table of \Cref{ch:diffusions} Schr\"odinger's equation takes the first row from waves (a group, here
unitary) and the generator from diffusions; it has infinite speed of propagation and no maximum principle, and it
neither transports singularities sharply nor erases them, but disperses them. For the harmonic oscillator
$-v''+x^2v=\lambda v$ the eigenvalues are $1,3,5,\dots$ \chk{S5}
\section{The hydrogen atom}
\begin{theorem}\label{thm:hydrogen}
\Scl[source, \S9.5 and \S10.7] The equation $-\Lap v-\tfrac2rv=\lambda v$ in space has the square-integrable
solutions $v=w(r)e^{-r/n}Y(\theta,\phi)$ with eigenvalues $\lambda=-1/n^2$, $n=1,2,\dots$; the eigenvalue $-1/n^2$ has
multiplicity $n^2$.
\end{theorem}
\noindent The radial series terminates exactly when the exponent is the reciprocal of an integer, and the
polynomials for $n=1,2,3$ are $1$, $1-\tfrac12r$, $1-\tfrac23r+\tfrac2{27}r^2$, as in the source's table. \chk{S4} The
multiplicity is $\sum_{l<n}(2l+1)$: rotational symmetry accounts for the factor $2l+1$ (\Cref{ch:symmetry}) and does
not explain why different $l$ share an eigenvalue.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3.3)--(0,0)--(8.2,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (2.05,0)--(2.05,-0.07) node[below,inner sep=1.5pt]{\scriptsize $5$};
\draw[ink2] (4.1,0)--(4.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $10$};
\draw[ink2] (6.15,0)--(6.15,-0.07) node[below,inner sep=1.5pt]{\scriptsize $15$};
\draw[ink2] (8.2,0)--(8.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $20$};
\draw[ink2] (0,0.635)--(-0.07,0.635) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.904)--(-0.07,1.904) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (0,3.173)--(-0.07,3.173) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (4.1,-0.38) {\scriptsize $r$};
\draw[ink2,very thin] (0,0.635)--(8.2,0.635);
\draw[thick,ink2] plot coordinates {(0,3.173) (0.041,2.932) (0.082,2.713) (0.123,2.515) (0.164,2.336) (0.205,2.174) (0.246,2.028) (0.287,1.895) (0.328,1.775) (0.369,1.667) (0.41,1.568) (0.451,1.48) (0.492,1.399) (0.533,1.326) (0.574,1.261) (0.615,1.201) (0.656,1.147) (0.697,1.098) (0.738,1.054) (0.779,1.014) (0.82,0.978) (0.861,0.945) (0.902,0.916) (0.943,0.889) (0.984,0.865) (1.025,0.843) (1.066,0.823) (1.107,0.805) (1.148,0.789) (1.189,0.774) (1.23,0.761) (1.271,0.749) (1.312,0.738) (1.353,0.728) (1.394,0.719) (1.435,0.711) (1.476,0.704) (1.517,0.697) (1.558,0.691) (1.599,0.686) (1.64,0.681) (1.681,0.677) (1.722,0.673) (1.763,0.669) (1.804,0.666) (1.845,0.663) (1.886,0.66) (1.927,0.658) (1.968,0.656) (2.009,0.654) (2.05,0.652) (2.091,0.65) (2.132,0.649) (2.173,0.647) (2.214,0.646) (2.255,0.645) (2.296,0.644) (2.337,0.643) (2.378,0.642) (2.419,0.642) (2.46,0.641) (2.501,0.64) (2.542,0.64) (2.583,0.639) (2.624,0.639) (2.665,0.638) (2.706,0.638) (2.747,0.638) (2.788,0.637) (2.829,0.637) (2.87,0.637) (2.911,0.637) (2.952,0.637) (2.993,0.636) (3.034,0.636) (3.075,0.636) (3.116,0.636) (3.157,0.636) (3.198,0.636) (3.239,0.636) (3.28,0.635) (3.321,0.635) (3.362,0.635) (3.403,0.635) (3.444,0.635) (3.485,0.635) (3.526,0.635) (3.567,0.635) (3.608,0.635) (3.649,0.635) (3.69,0.635) (3.731,0.635) (3.772,0.635) (3.813,0.635) (3.854,0.635) (3.895,0.635) (3.936,0.635) (3.977,0.635) (4.018,0.635) (4.059,0.635) (4.1,0.635) (4.141,0.635) (4.182,0.635) (4.223,0.635) (4.264,0.635) (4.305,0.635) (4.346,0.635) (4.387,0.635) (4.428,0.635) (4.469,0.635) (4.51,0.635) (4.551,0.635) (4.592,0.635) (4.633,0.635) (4.674,0.635) (4.715,0.635) (4.756,0.635) (4.797,0.635) (4.838,0.635) (4.879,0.635) (4.92,0.635) (4.961,0.635) (5.002,0.635) (5.043,0.635) (5.084,0.635) (5.125,0.635) (5.166,0.635) (5.207,0.635) (5.248,0.635) (5.289,0.635) (5.33,0.635) (5.371,0.635) (5.412,0.635) (5.453,0.635) (5.494,0.635) (5.535,0.635) (5.576,0.635) (5.617,0.635) (5.658,0.635) (5.699,0.635) (5.74,0.635) (5.781,0.635) (5.822,0.635) (5.863,0.635) (5.904,0.635) (5.945,0.635) (5.986,0.635) (6.027,0.635) (6.068,0.635) (6.109,0.635) (6.15,0.635) (6.191,0.635) (6.232,0.635) (6.273,0.635) (6.314,0.635) (6.355,0.635) (6.396,0.635) (6.437,0.635) (6.478,0.635) (6.519,0.635) (6.56,0.635) (6.601,0.635) (6.642,0.635) (6.683,0.635) (6.724,0.635) (6.765,0.635) (6.806,0.635) (6.847,0.635) (6.888,0.635) (6.929,0.635) (6.97,0.635) (7.011,0.635) (7.052,0.635) (7.093,0.635) (7.134,0.635) (7.175,0.635) (7.216,0.635) (7.257,0.635) (7.298,0.635) (7.339,0.635) (7.38,0.635) (7.421,0.635) (7.462,0.635) (7.503,0.635) (7.544,0.635) (7.585,0.635) (7.626,0.635) (7.667,0.635) (7.708,0.635) (7.749,0.635) (7.79,0.635) (7.831,0.635) (7.872,0.635) (7.913,0.635) (7.954,0.635) (7.995,0.635) (8.036,0.635) (8.077,0.635) (8.118,0.635) (8.159,0.635) (8.2,0.635)};
\draw[thick,fwd] plot coordinates {(0,3.173) (0.041,2.929) (0.082,2.702) (0.123,2.492) (0.164,2.297) (0.205,2.117) (0.246,1.951) (0.287,1.797) (0.328,1.656) (0.369,1.525) (0.41,1.404) (0.451,1.294) (0.492,1.192) (0.533,1.098) (0.574,1.013) (0.615,0.934) (0.656,0.863) (0.697,0.797) (0.738,0.738) (0.779,0.684) (0.82,0.635) (0.861,0.59) (0.902,0.55) (0.943,0.514) (0.984,0.482) (1.025,0.453) (1.066,0.427) (1.107,0.404) (1.148,0.384) (1.189,0.367) (1.23,0.351) (1.271,0.338) (1.312,0.327) (1.353,0.318) (1.394,0.31) (1.435,0.304) (1.476,0.299) (1.517,0.295) (1.558,0.293) (1.599,0.292) (1.64,0.291) (1.681,0.291) (1.722,0.293) (1.763,0.295) (1.804,0.297) (1.845,0.3) (1.886,0.304) (1.927,0.308) (1.968,0.312) (2.009,0.317) (2.05,0.322) (2.091,0.327) (2.132,0.333) (2.173,0.339) (2.214,0.345) (2.255,0.351) (2.296,0.357) (2.337,0.363) (2.378,0.369) (2.419,0.376) (2.46,0.382) (2.501,0.388) (2.542,0.394) (2.583,0.401) (2.624,0.407) (2.665,0.413) (2.706,0.419) (2.747,0.425) (2.788,0.431) (2.829,0.437) (2.87,0.443) (2.911,0.449) (2.952,0.454) (2.993,0.46) (3.034,0.465) (3.075,0.47) (3.116,0.476) (3.157,0.481) (3.198,0.486) (3.239,0.49) (3.28,0.495) (3.321,0.5) (3.362,0.504) (3.403,0.509) (3.444,0.513) (3.485,0.517) (3.526,0.521) (3.567,0.525) (3.608,0.529) (3.649,0.532) (3.69,0.536) (3.731,0.539) (3.772,0.543) (3.813,0.546) (3.854,0.549) (3.895,0.552) (3.936,0.555) (3.977,0.558) (4.018,0.561) (4.059,0.564) (4.1,0.566) (4.141,0.569) (4.182,0.571) (4.223,0.574) (4.264,0.576) (4.305,0.578) (4.346,0.58) (4.387,0.582) (4.428,0.584) (4.469,0.586) (4.51,0.588) (4.551,0.59) (4.592,0.591) (4.633,0.593) (4.674,0.595) (4.715,0.596) (4.756,0.598) (4.797,0.599) (4.838,0.601) (4.879,0.602) (4.92,0.603) (4.961,0.604) (5.002,0.606) (5.043,0.607) (5.084,0.608) (5.125,0.609) (5.166,0.61) (5.207,0.611) (5.248,0.612) (5.289,0.613) (5.33,0.614) (5.371,0.614) (5.412,0.615) (5.453,0.616) (5.494,0.617) (5.535,0.618) (5.576,0.618) (5.617,0.619) (5.658,0.62) (5.699,0.62) (5.74,0.621) (5.781,0.621) (5.822,0.622) (5.863,0.622) (5.904,0.623) (5.945,0.623) (5.986,0.624) (6.027,0.624) (6.068,0.625) (6.109,0.625) (6.15,0.625) (6.191,0.626) (6.232,0.626) (6.273,0.627) (6.314,0.627) (6.355,0.627) (6.396,0.628) (6.437,0.628) (6.478,0.628) (6.519,0.628) (6.56,0.629) (6.601,0.629) (6.642,0.629) (6.683,0.629) (6.724,0.63) (6.765,0.63) (6.806,0.63) (6.847,0.63) (6.888,0.63) (6.929,0.631) (6.97,0.631) (7.011,0.631) (7.052,0.631) (7.093,0.631) (7.134,0.631) (7.175,0.631) (7.216,0.632) (7.257,0.632) (7.298,0.632) (7.339,0.632) (7.38,0.632) (7.421,0.632) (7.462,0.632) (7.503,0.632) (7.544,0.633) (7.585,0.633) (7.626,0.633) (7.667,0.633) (7.708,0.633) (7.749,0.633) (7.79,0.633) (7.831,0.633) (7.872,0.633) (7.913,0.633) (7.954,0.633) (7.995,0.633) (8.036,0.633) (8.077,0.633) (8.118,0.633) (8.159,0.634) (8.2,0.634)};
\draw[thick,bwd] plot coordinates {(0,3.173) (0.041,2.928) (0.082,2.7) (0.123,2.487) (0.164,2.29) (0.205,2.107) (0.246,1.937) (0.287,1.78) (0.328,1.634) (0.369,1.5) (0.41,1.376) (0.451,1.261) (0.492,1.156) (0.533,1.06) (0.574,0.972) (0.615,0.891) (0.656,0.818) (0.697,0.751) (0.738,0.69) (0.779,0.636) (0.82,0.586) (0.861,0.542) (0.902,0.503) (0.943,0.468) (0.984,0.437) (1.025,0.41) (1.066,0.386) (1.107,0.366) (1.148,0.349) (1.189,0.335) (1.23,0.323) (1.271,0.314) (1.312,0.307) (1.353,0.302) (1.394,0.299) (1.435,0.298) (1.476,0.298) (1.517,0.3) (1.558,0.303) (1.599,0.307) (1.64,0.312) (1.681,0.319) (1.722,0.326) (1.763,0.334) (1.804,0.342) (1.845,0.351) (1.886,0.361) (1.927,0.371) (1.968,0.382) (2.009,0.393) (2.05,0.404) (2.091,0.415) (2.132,0.427) (2.173,0.438) (2.214,0.45) (2.255,0.462) (2.296,0.474) (2.337,0.485) (2.378,0.497) (2.419,0.509) (2.46,0.52) (2.501,0.531) (2.542,0.543) (2.583,0.554) (2.624,0.565) (2.665,0.575) (2.706,0.586) (2.747,0.596) (2.788,0.606) (2.829,0.616) (2.87,0.625) (2.911,0.635) (2.952,0.644) (2.993,0.653) (3.034,0.661) (3.075,0.669) (3.116,0.677) (3.157,0.685) (3.198,0.692) (3.239,0.7) (3.28,0.706) (3.321,0.713) (3.362,0.719) (3.403,0.726) (3.444,0.731) (3.485,0.737) (3.526,0.742) (3.567,0.747) (3.608,0.752) (3.649,0.757) (3.69,0.761) (3.731,0.765) (3.772,0.769) (3.813,0.773) (3.854,0.776) (3.895,0.779) (3.936,0.782) (3.977,0.785) (4.018,0.788) (4.059,0.79) (4.1,0.792) (4.141,0.794) (4.182,0.796) (4.223,0.798) (4.264,0.799) (4.305,0.801) (4.346,0.802) (4.387,0.803) (4.428,0.804) (4.469,0.805) (4.51,0.805) (4.551,0.806) (4.592,0.806) (4.633,0.806) (4.674,0.806) (4.715,0.807) (4.756,0.806) (4.797,0.806) (4.838,0.806) (4.879,0.806) (4.92,0.805) (4.961,0.805) (5.002,0.804) (5.043,0.803) (5.084,0.802) (5.125,0.802) (5.166,0.801) (5.207,0.8) (5.248,0.799) (5.289,0.797) (5.33,0.796) (5.371,0.795) (5.412,0.794) (5.453,0.792) (5.494,0.791) (5.535,0.79) (5.576,0.788) (5.617,0.787) (5.658,0.785) (5.699,0.784) (5.74,0.782) (5.781,0.781) (5.822,0.779) (5.863,0.777) (5.904,0.776) (5.945,0.774) (5.986,0.773) (6.027,0.771) (6.068,0.769) (6.109,0.767) (6.15,0.766) (6.191,0.764) (6.232,0.762) (6.273,0.761) (6.314,0.759) (6.355,0.757) (6.396,0.755) (6.437,0.754) (6.478,0.752) (6.519,0.75) (6.56,0.749) (6.601,0.747) (6.642,0.745) (6.683,0.743) (6.724,0.742) (6.765,0.74) (6.806,0.738) (6.847,0.737) (6.888,0.735) (6.929,0.733) (6.97,0.732) (7.011,0.73) (7.052,0.729) (7.093,0.727) (7.134,0.726) (7.175,0.724) (7.216,0.722) (7.257,0.721) (7.298,0.719) (7.339,0.718) (7.38,0.716) (7.421,0.715) (7.462,0.714) (7.503,0.712) (7.544,0.711) (7.585,0.709) (7.626,0.708) (7.667,0.707) (7.708,0.705) (7.749,0.704) (7.79,0.703) (7.831,0.701) (7.872,0.7) (7.913,0.699) (7.954,0.698) (7.995,0.696) (8.036,0.695) (8.077,0.694) (8.118,0.693) (8.159,0.692) (8.2,0.69)};
\fill[fwd] (0.82,0.635) circle (1.6pt);
\fill[bwd] (0.78,0.635) circle (1.6pt);
\fill[bwd] (2.91,0.635) circle (1.6pt);
\node[ink2,right] at (0.574,2.031) {\scriptsize $n=1$};
\node[fwd,right] at (1.886,0.178) {\scriptsize $n=2$};
\node[bwd,right] at (4.51,1.142) {\scriptsize $n=3$};
\end{scope}
\end{tikzpicture}
\figcap{The radial functions $w_n(r)e^{-r/n}$ of hydrogen for $n=1,2,3$ and angular momentum zero. The $n$-th has $n-1$ nodes (marked), and its tail is longer by the factor $n$: the eigenvalue $-1/n^2$ is read off the rate of decay.}
\end{center}
\begin{remark}[The hidden symmetry]\label{rem:so4}
\Sli[\cite{Foc35}] The bound states of the Coulomb problem carry an action of $\mathrm{SO}(4)$, and those of energy
$-1/n^2$ form its irreducible representation on harmonic polynomials of degree $n-1$ in four variables, whose
dimension is $n^2$ by \Cref{rem:type}. \chk{E4} The ``accidental'' degeneracy is forced by a symmetry that is not a
symmetry of space.
\end{remark}
\section*{Notes}
The source's Chapter 9 is followed. For Huygens's principle in general dimension see \cite{Eva10}; for wave front
sets and propagation \cite{Hor71,DuHo72}; for holomorphic semigroups \cite{ReSi75}. The $\mathrm{SO}(4)$ symmetry
of the hydrogen atom is Fock's \cite{Foc35}.
\section*{Exercises}
\begin{exercise}\label{exer:radial}
Show that the radial solutions of the wave equation in space are $(F(r-ct)+G(r+ct))/r$.
\end{exercise}
\begin{exercise}\label{exer:shadow}
Derive the density in \Cref{prop:descent} and hence the two-dimensional formula from Kirchhoff's.
\end{exercise}
\begin{exercise}\label{exer:packet}
For $iu_t=-\tfrac12u_{xx}$ with $u(x,0)=e^{-x^2/2}$, show that $u=(1+it)^{-1/2}e^{-x^2/2(1+it)}$ and that the mean
square width of $|u|^2$ is $(1+t^2)/2$.
\end{exercise}
\begin{exercise}\label{exer:hyd}
Find the radial polynomial for $n=3$ from the recursion $\tfrac{k(k+1)}2a_k=(\tfrac k3-1)a_{k-1}$, and count the
states with eigenvalue $-\tfrac19$.
\end{exercise}
\chapter{Symmetry and Special Functions}\label{ch:symmetry}
The source returns to Fourier's method for domains in the plane and in space, and meets Bessel and Legendre
functions \src{Ch.~10}. The organizing fact is one sentence: the Laplacian commutes with the symmetries of the
domain, so its eigenspaces are representations of the symmetry group.
\section{Eigenspaces are representations}
\begin{theorem}\label{thm:rep}
\Srf[source, \S10.1] Let $G$ be a group of isometries of a bounded domain $D$ preserving the boundary condition. Then
$G$ acts on each eigenspace of the Laplacian. Consequently $L^2(D)$ is the orthogonal sum of its isotypic
components, each preserved by the Laplacian; and an eigenvalue whose eigenspace contains an irreducible
representation of dimension $d$ has multiplicity at least $d$.
\end{theorem}
\noindent The Laplacian of a square grid commutes with the quarter turn and its second eigenvalue is double, the
symmetry group of the square having a two-dimensional irreducible representation; on a $5\times6$ grid, with only
reflections, it is simple. \chk{Y6}
\begin{theorem}[Products]\label{thm:product}
\Scl[source, \S11.5, Theorem 2] $L^2(D_1\times D_2)=L^2(D_1)\otimes L^2(D_2)$ and
$\Lap_{D_1\times D_2}=\Lap_{D_1}\otimes1+1\otimes\Lap_{D_2}$. The eigenfunctions of a product are the products of
eigenfunctions, and its eigenvalues the sums. \chk{Y4}
\end{theorem}
\noindent Separation of variables is therefore two different things. In Cartesian coordinates it is the tensor
product. In polar and spherical coordinates it is the decomposition of \Cref{thm:rep} under a rotation group,
followed by an ordinary differential equation in the radius. The special functions are what the second kind
produces.
\section{The drum}
For the disk of radius $a$ the group is $\mathrm O(2)$. The irreducible representations of the rotations are the
characters $e^{in\theta}$; in the $n$-th isotypic component the eigenvalue equation becomes Bessel's equation
$R''+\tfrac1rR'+(\lambda-\tfrac{n^2}{r^2})R=0$, with the solutions $J_n(\sqrt\lambda r)$ regular at the centre.
\begin{theorem}\label{thm:drum}
\Scl[source, \S10.2 and \S10.5] The Dirichlet eigenvalues of the disk are $(z_{nm}/a)^2$, with $z_{nm}$ the $m$-th
positive zero of $J_n$, and eigenfunctions $J_n(z_{nm}r/a)\cos n\theta$, $J_n(z_{nm}r/a)\sin n\theta$. Eigenvalues with
$n\ge1$ are double.
\end{theorem}
\noindent The first zeros are $2.4048$ ($n=0$), $3.8317$ ($n=1$), $5.1356$ ($n=2$), $5.5201$ ($n=0$), so the
frequencies of a drum are in the ratios $1:1.593:2.136:2.295$ and not $1:2:3:4$. \chk{Y3}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3.2)--(0,0)--(6.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (1.28,0)--(1.28,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (2.56,0)--(2.56,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (3.84,0)--(3.84,-0.07) node[below,inner sep=1.5pt]{\scriptsize $6$};
\draw[ink2] (5.12,0)--(5.12,-0.07) node[below,inner sep=1.5pt]{\scriptsize $8$};
\draw[ink2] (6.4,0)--(6.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $10$};
\draw[ink2] (0,0.96)--(-0.07,0.96) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,2.027)--(-0.07,2.027) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (0,3.093)--(-0.07,3.093) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (3.2,-0.38) {\scriptsize $x$};
\draw[ink2,very thin] (0,0.96)--(6.4,0.96);
\draw[thick,fwd] plot coordinates {(0,3.093) (0.04,3.091) (0.08,3.085) (0.12,3.075) (0.16,3.06) (0.2,3.042) (0.24,3.019) (0.28,2.992) (0.32,2.962) (0.36,2.928) (0.4,2.89) (0.44,2.849) (0.48,2.804) (0.52,2.756) (0.56,2.704) (0.6,2.65) (0.64,2.592) (0.68,2.532) (0.72,2.47) (0.76,2.405) (0.8,2.338) (0.84,2.269) (0.88,2.198) (0.92,2.126) (0.96,2.052) (1,1.977) (1.04,1.901) (1.08,1.824) (1.12,1.747) (1.16,1.67) (1.2,1.592) (1.24,1.515) (1.28,1.438) (1.32,1.361) (1.36,1.285) (1.4,1.21) (1.44,1.137) (1.48,1.064) (1.52,0.993) (1.56,0.924) (1.6,0.857) (1.64,0.792) (1.68,0.729) (1.72,0.668) (1.76,0.61) (1.8,0.554) (1.84,0.502) (1.88,0.452) (1.92,0.405) (1.96,0.362) (2,0.321) (2.04,0.284) (2.08,0.25) (2.12,0.22) (2.16,0.193) (2.2,0.169) (2.24,0.149) (2.28,0.132) (2.32,0.119) (2.36,0.11) (2.4,0.104) (2.44,0.101) (2.48,0.102) (2.52,0.106) (2.56,0.113) (2.6,0.123) (2.64,0.137) (2.68,0.153) (2.72,0.172) (2.76,0.194) (2.8,0.219) (2.84,0.247) (2.88,0.276) (2.92,0.308) (2.96,0.342) (3,0.378) (3.04,0.416) (3.08,0.455) (3.12,0.496) (3.16,0.538) (3.2,0.581) (3.24,0.625) (3.28,0.67) (3.32,0.716) (3.36,0.761) (3.4,0.808) (3.44,0.854) (3.48,0.9) (3.52,0.945) (3.56,0.991) (3.6,1.035) (3.64,1.079) (3.68,1.122) (3.72,1.164) (3.76,1.204) (3.8,1.244) (3.84,1.281) (3.88,1.317) (3.92,1.352) (3.96,1.384) (4,1.415) (4.04,1.443) (4.08,1.469) (4.12,1.493) (4.16,1.515) (4.2,1.534) (4.24,1.551) (4.28,1.566) (4.32,1.578) (4.36,1.587) (4.4,1.594) (4.44,1.598) (4.48,1.6) (4.52,1.6) (4.56,1.596) (4.6,1.591) (4.64,1.583) (4.68,1.573) (4.72,1.56) (4.76,1.545) (4.8,1.528) (4.84,1.509) (4.88,1.488) (4.92,1.465) (4.96,1.44) (5,1.414) (5.04,1.386) (5.08,1.357) (5.12,1.326) (5.16,1.294) (5.2,1.261) (5.24,1.228) (5.28,1.193) (5.32,1.158) (5.36,1.122) (5.4,1.086) (5.44,1.049) (5.48,1.013) (5.52,0.977) (5.56,0.94) (5.6,0.905) (5.64,0.869) (5.68,0.835) (5.72,0.8) (5.76,0.767) (5.8,0.735) (5.84,0.704) (5.88,0.674) (5.92,0.646) (5.96,0.618) (6,0.593) (6.04,0.569) (6.08,0.546) (6.12,0.526) (6.16,0.507) (6.2,0.49) (6.24,0.475) (6.28,0.462) (6.32,0.451) (6.36,0.442) (6.4,0.435)};
\draw[thick,bwd] plot coordinates {(0,0.96) (0.04,1.027) (0.08,1.093) (0.12,1.159) (0.16,1.225) (0.2,1.289) (0.24,1.353) (0.28,1.416) (0.32,1.477) (0.36,1.537) (0.4,1.595) (0.44,1.651) (0.48,1.705) (0.52,1.757) (0.56,1.807) (0.6,1.854) (0.64,1.899) (0.68,1.941) (0.72,1.98) (0.76,2.016) (0.8,2.049) (0.84,2.079) (0.88,2.106) (0.92,2.13) (0.96,2.15) (1,2.167) (1.04,2.181) (1.08,2.191) (1.12,2.198) (1.16,2.201) (1.2,2.201) (1.24,2.197) (1.28,2.19) (1.32,2.18) (1.36,2.167) (1.4,2.15) (1.44,2.13) (1.48,2.107) (1.52,2.081) (1.56,2.052) (1.6,2.02) (1.64,1.986) (1.68,1.949) (1.72,1.91) (1.76,1.869) (1.8,1.825) (1.84,1.78) (1.88,1.732) (1.92,1.683) (1.96,1.633) (2,1.581) (2.04,1.528) (2.08,1.474) (2.12,1.42) (2.16,1.365) (2.2,1.309) (2.24,1.253) (2.28,1.197) (2.32,1.141) (2.36,1.086) (2.4,1.031) (2.44,0.977) (2.48,0.923) (2.52,0.87) (2.56,0.819) (2.6,0.769) (2.64,0.72) (2.68,0.673) (2.72,0.628) (2.76,0.585) (2.8,0.543) (2.84,0.504) (2.88,0.467) (2.92,0.432) (2.96,0.4) (3,0.37) (3.04,0.343) (3.08,0.319) (3.12,0.297) (3.16,0.278) (3.2,0.261) (3.24,0.248) (3.28,0.237) (3.32,0.229) (3.36,0.224) (3.4,0.222) (3.44,0.222) (3.48,0.226) (3.52,0.232) (3.56,0.24) (3.6,0.252) (3.64,0.265) (3.68,0.282) (3.72,0.3) (3.76,0.321) (3.8,0.345) (3.84,0.37) (3.88,0.397) (3.92,0.426) (3.96,0.457) (4,0.489) (4.04,0.523) (4.08,0.558) (4.12,0.594) (4.16,0.632) (4.2,0.67) (4.24,0.709) (4.28,0.749) (4.32,0.789) (4.36,0.829) (4.4,0.869) (4.44,0.91) (4.48,0.95) (4.52,0.99) (4.56,1.029) (4.6,1.068) (4.64,1.106) (4.68,1.144) (4.72,1.18) (4.76,1.215) (4.8,1.249) (4.84,1.281) (4.88,1.312) (4.92,1.341) (4.96,1.369) (5,1.395) (5.04,1.419) (5.08,1.441) (5.12,1.461) (5.16,1.478) (5.2,1.494) (5.24,1.508) (5.28,1.519) (5.32,1.529) (5.36,1.536) (5.4,1.54) (5.44,1.543) (5.48,1.543) (5.52,1.541) (5.56,1.537) (5.6,1.53) (5.64,1.521) (5.68,1.511) (5.72,1.498) (5.76,1.483) (5.8,1.467) (5.84,1.448) (5.88,1.428) (5.92,1.406) (5.96,1.383) (6,1.358) (6.04,1.332) (6.08,1.304) (6.12,1.275) (6.16,1.246) (6.2,1.215) (6.24,1.184) (6.28,1.152) (6.32,1.119) (6.36,1.086) (6.4,1.053)};
\draw[thick,ink2] plot coordinates {(0,0.96) (0.04,0.961) (0.08,0.964) (0.12,0.969) (0.16,0.977) (0.2,0.986) (0.24,0.997) (0.28,1.01) (0.32,1.025) (0.36,1.042) (0.4,1.061) (0.44,1.081) (0.48,1.103) (0.52,1.127) (0.56,1.151) (0.6,1.178) (0.64,1.205) (0.68,1.234) (0.72,1.263) (0.76,1.294) (0.8,1.325) (0.84,1.357) (0.88,1.389) (0.92,1.422) (0.96,1.455) (1,1.488) (1.04,1.521) (1.08,1.554) (1.12,1.587) (1.16,1.62) (1.2,1.651) (1.24,1.682) (1.28,1.713) (1.32,1.742) (1.36,1.77) (1.4,1.798) (1.44,1.823) (1.48,1.848) (1.52,1.871) (1.56,1.892) (1.6,1.912) (1.64,1.929) (1.68,1.945) (1.72,1.959) (1.76,1.971) (1.8,1.981) (1.84,1.988) (1.88,1.994) (1.92,1.997) (1.96,1.998) (2,1.996) (2.04,1.993) (2.08,1.986) (2.12,1.978) (2.16,1.967) (2.2,1.954) (2.24,1.938) (2.28,1.921) (2.32,1.901) (2.36,1.878) (2.4,1.854) (2.44,1.828) (2.48,1.799) (2.52,1.769) (2.56,1.737) (2.6,1.703) (2.64,1.667) (2.68,1.63) (2.72,1.591) (2.76,1.551) (2.8,1.51) (2.84,1.468) (2.88,1.425) (2.92,1.381) (2.96,1.336) (3,1.29) (3.04,1.244) (3.08,1.198) (3.12,1.152) (3.16,1.106) (3.2,1.059) (3.24,1.013) (3.28,0.968) (3.32,0.923) (3.36,0.878) (3.4,0.835) (3.44,0.792) (3.48,0.75) (3.52,0.71) (3.56,0.671) (3.6,0.633) (3.64,0.597) (3.68,0.562) (3.72,0.529) (3.76,0.498) (3.8,0.469) (3.84,0.442) (3.88,0.417) (3.92,0.394) (3.96,0.373) (4,0.355) (4.04,0.339) (4.08,0.325) (4.12,0.313) (4.16,0.304) (4.2,0.297) (4.24,0.293) (4.28,0.291) (4.32,0.292) (4.36,0.295) (4.4,0.3) (4.44,0.307) (4.48,0.317) (4.52,0.329) (4.56,0.343) (4.6,0.359) (4.64,0.377) (4.68,0.398) (4.72,0.42) (4.76,0.443) (4.8,0.469) (4.84,0.496) (4.88,0.524) (4.92,0.554) (4.96,0.585) (5,0.617) (5.04,0.65) (5.08,0.684) (5.12,0.719) (5.16,0.754) (5.2,0.79) (5.24,0.826) (5.28,0.863) (5.32,0.899) (5.36,0.935) (5.4,0.972) (5.44,1.008) (5.48,1.043) (5.52,1.078) (5.56,1.112) (5.6,1.146) (5.64,1.178) (5.68,1.21) (5.72,1.24) (5.76,1.269) (5.8,1.297) (5.84,1.323) (5.88,1.348) (5.92,1.371) (5.96,1.392) (6,1.412) (6.04,1.43) (6.08,1.446) (6.12,1.46) (6.16,1.472) (6.2,1.483) (6.24,1.491) (6.28,1.497) (6.32,1.501) (6.36,1.503) (6.4,1.503)};
\fill[fwd] (1.539,0.96) circle (1.6pt);
\fill[fwd] (3.533,0.96) circle (1.6pt);
\fill[bwd] (2.452,0.96) circle (1.6pt);
\fill[ink2] (3.287,0.96) circle (1.6pt);
\node[fwd,right] at (0.576,2.987) {\scriptsize $J_0$};
\node[bwd,right] at (1.216,2.411) {\scriptsize $J_1$};
\node[ink2,right] at (2.048,2.197) {\scriptsize $J_2$};
\end{scope}
\begin{scope}[shift={(8.3,2.45)}]
\foreach \i/\lab in {0/{$1$},1/{$1.593$}} {\draw[thick] (\i*1.9,0) circle (0.68); \node at (\i*1.9,-0.95) {\scriptsize \lab};}
\draw[very thick,bwd] (1.9,-0.68)--(1.9,0.68);
\begin{scope}[yshift=-2.2cm]
\foreach \i/\lab in {0/{$2.136$},1/{$2.295$}} {\draw[thick] (\i*1.9,0) circle (0.68); \node at (\i*1.9,-0.95) {\scriptsize \lab};}
\draw[very thick,bwd] (0,-0.68)--(0,0.68) (-0.68,0)--(0.68,0);
\draw[very thick,bwd] (1.9,0) circle (0.2963);
\end{scope}
\end{scope}
\end{tikzpicture}
\figcap{Left: the Bessel functions $J_0,J_1,J_2$ with the four smallest zeros $2.405$, $3.832$, $5.136$, $5.520$. Right: the nodal sets of the corresponding modes of a circular drum, labelled by frequency relative to the fundamental. The angular index counts nodal diameters and is the weight of the rotation group; the last mode is the second radial one, with a nodal circle at $2.405/5.520$ of the radius.}
\end{center}
\begin{remark}[What a Bessel function is]\label{rem:bessel}
\Scl The plane wave $e^{iy}$ is an eigenfunction of the Laplacian with eigenvalue $-1$. Its component in the $n$-th
character of the rotation group is $J_n(r)e^{in\theta}$:
\[ e^{ir\sin\theta}=\sum_{n\in\Z}J_n(r)e^{in\theta},\qquad J_n(r)=\frac1{2\pi}\int_0^{2\pi}e^{i(r\sin\tau-n\tau)}\,d\tau . \]
Since the projection onto an isotypic component commutes with the Laplacian, $J_n(r)e^{in\theta}$ is again an
eigenfunction, which is Bessel's equation. The companion script computes $J_n$ from this integral. Bessel
functions are matrix coefficients of the group of rigid motions of the plane \cite{Vil68}.
\end{remark}
\section{The sphere}
Let $\HH_l$ be the space of harmonic homogeneous polynomials of degree $l$ in three variables.
\begin{theorem}[Spherical harmonics]\label{thm:sphharm}
\Scl[source, \S10.3, Theorem 1] $\dim\HH_l=2l+1$; $\HH_l$ is an irreducible representation of $\mathrm{SO}(3)$; its
restriction to the unit sphere is the eigenspace of the spherical Laplacian with eigenvalue $-l(l+1)$; and
$L^2(S^2)$ is the orthogonal sum of these spaces.
\end{theorem}
\begin{proof}[Proof of the first and third statements]
$\Lap$ maps polynomials of degree $l$ onto those of degree $l-2$, so $\dim\HH_l=\binom{l+2}2-\binom l2=2l+1$. With
$E=\sum x_i\pd_i$ the Euler operator and $L_1,L_2,L_3$ the rotation fields, there is the operator identity
$r^2\Lap=E(E+1)-L^2$, where $L^2=-(L_1^2+L_2^2+L_3^2)$. On $\HH_l$ the left side vanishes and $E=l$, so $L^2=l(l+1)$;
and $-L^2$ is the spherical Laplacian.
\end{proof}
\noindent The dimension count, the identity, the commutation relations $[L_1,L_2]=-L_3$ and the commutation of the
$L_i$ with $\Lap$ are checked in exact arithmetic. \chk{Y1} The last statement of the theorem is the Peter--Weyl
theorem for the homogeneous space $S^2=\mathrm{SO}(3)/\mathrm{SO}(2)$. In each $\HH_l$ there is one function
invariant under rotations about the axis, the \term{zonal harmonic} $P_l(\cos\theta)$; the Legendre polynomials
satisfy $((1-s^2)P_l')'+l(l+1)P_l=0$, $P_l(1)=1$, and are orthogonal on $[-1,1]$ with square norm $2/(2l+1)$.
\chk{Y2} For the ball the radial equation is Bessel's with half-integer order \src{\S10.3}.
\section{Angular momentum}
The operators $L_i$ generate the rotations, and in quantum mechanics are the components of angular momentum up to
a factor \src{\S10.7}. The algebra they span is $\mathfrak{so}(3)$; $L^2$ is its Casimir element, which acts on an
irreducible representation by a scalar; the scalar is $l(l+1)$ and the dimension $2l+1$. That is all of the
angular analysis of a rotationally symmetric problem, the hydrogen atom of \Cref{thm:hydrogen} included: the
potential enters only the radial equation.
\section{Nodes}
\begin{theorem}[Courant]\label{thm:courant}
\Scl[source, \S10.4, Theorem 1] The first eigenfunction has no zeros in the domain. The zero set of the $n$-th
eigenfunction divides the domain into at most $n$ pieces; in one dimension, into exactly $n$.
\end{theorem}
\noindent On a path of twelve points the $k$-th eigenvector changes sign $k-1$ times. On a $5\times6$ grid the first
eight eigenvectors have $1,2,2,4,3,3,6,4$ nodal domains. The square has the eigenvalues $2,5,5,8,10,10$ of the
source's list. \chk{Y5} Symmetry forces some nodes: by \Cref{thm:images} an eigenfunction that transforms by the sign
character under a reflection vanishes on the mirror, and the Dirichlet eigenfunctions of a chamber of a
reflection group are exactly the antisymmetric eigenfunctions of the whole.
\begin{center}
\begin{tikzpicture}[scale=0.62]
\foreach \i/\lab in {0/{$\sin x\sin2y$},1/{$\sin2x\sin y$},2/{sum},3/{difference}}{
\begin{scope}[xshift=\i*4.1cm]
\draw[thick] (0,0) rectangle (3,3);
\node at (1.5,-0.5) {\small\lab};
\end{scope}}
\draw[very thick,bwd] (0,1.5)--(3,1.5);
\draw[very thick,bwd] (5.6,0)--(5.6,3);
\draw[very thick,bwd] (8.2,3)--(11.2,0);
\draw[very thick,bwd] (12.3,0)--(15.3,3);
\end{tikzpicture}
\figcap{Nodal lines in the eigenspace of eigenvalue $5$ on the square of side $\pi$. The eigenspace is the
two-dimensional representation of the symmetry group of the square, and the node depends on the vector chosen in
it: $a\sin x\sin2y+b\sin2x\sin y=2\sin x\sin y\,(a\cos y+b\cos x)$.}
\end{center}
\section*{Notes}
The source's Chapter 10 is followed. For special functions as matrix coefficients of group representations see
\cite{Vil68}; for spherical harmonics \cite{StWe71}; for nodal domains \cite{CoHi53}, and for their discrete
analogue \cite{DGLS01}.
\section*{Exercises}
\begin{exercise}\label{exer:sq5}
For the square of side $\pi$ with Dirichlet conditions, describe the action of the symmetries of the square on
the eigenspace of $\lambda=5$. What is the multiplicity of $\lambda=50$, and is that eigenspace irreducible?
\end{exercise}
\begin{exercise}\label{exer:dimh}
Compute $\dim\HH_l$ in $n$ variables and check the cases $n=2,3,4$ against \Cref{rem:type}.
\end{exercise}
\begin{exercise}\label{exer:anger}
Deduce Bessel's equation for $J_n$ from \Cref{rem:bessel}.
\end{exercise}
\begin{exercise}\label{exer:drumf}
List the four lowest frequencies of a circular drum of radius $a$ with wave speed $c$, with their multiplicities.
\end{exercise}
\part{Operators}
\chapter{Eigenvalues}\label{ch:eigen}
For a general domain there are no formulas for the eigenvalues, and the source turns to what can be said without
them: variational characterizations, completeness, and asymptotics \src{Ch.~11}. We add the question the
asymptotics provoke, whether the spectrum determines the domain, because its answer is a statement about
representations.
\section{Minimax, and what it makes functorial}
\begin{theorem}[Minimax]\label{thm:minimax}
\Scl[source, \S11.1--11.2 and \S11.6] For the Dirichlet problem on a bounded domain $D$,
\[ \lambda_n(D)=\min_{\substack{W\subset H^1_0(D)\\\dim W=n}}\ \max_{w\in W\setminus0}\ \frac{\int|\nabla w|^2}{\int w^2}. \]
In particular the Rayleigh--Ritz values in any $n$-dimensional trial space are upper bounds for
$\lambda_1,\dots,\lambda_n$.
\end{theorem}
\noindent On $(0,1)$ the trial function $x(1-x)$ gives $10$, and adding $x^2(1-x)^2$ gives $9.86975$, against
$\pi^2=9.86960$. \chk{L1} In finite dimensions the theorem is the interlacing of the eigenvalues of a matrix with
those of its compressions. \chk{L2}
The formula depends on the domain only through the space of trial functions, and it is monotone in that space.
\begin{corollary}[Monotonicity]\label{cor:mono}
\Scl[source, \S11.6, Theorems 4 and 5] If $D\subseteq D'$ then $\lambda_n(D)\ge\lambda_n(D')$ for every $n$:
Dirichlet eigenvalues are a functor from domains and inclusions to decreasing sequences. For one domain the
Neumann eigenvalues satisfy $\mu_n\le\lambda_n$. The Neumann eigenvalues are \emph{not} monotone in the domain.
\end{corollary}
\begin{proof}
Extension by zero embeds $H^1_0(D)$ isometrically in $H^1_0(D')$, so the minimum is over a larger family. For the
second statement $H^1_0(D)\subseteq H^1(D)$. For the third, a rectangle of sides $1.3$ and $0.1$ fits inside the unit
square along its diagonal, and its first non-zero Neumann eigenvalue $\pi^2/1.69=5.84$ is smaller than that of the
square, $\pi^2=9.87$.
\end{proof}
\noindent An L-shaped region in a $6\times6$ grid has every Dirichlet eigenvalue above the corresponding one of the
square. \chk{L3} The asymmetry has the same source as in \Cref{thm:images}: Dirichlet conditions are compatible
with extension by zero and Neumann conditions with nothing of the kind. The spaces $H^1(D)$ are related by
restriction, which does not preserve the Rayleigh quotient.
\section{Completeness}
\begin{theorem}[Spectral theorem for the Laplacian]\label{thm:spectral}
\Scl[source, \S11.3--11.5] On a bounded domain, with Dirichlet, Neumann or Robin conditions, there is an orthonormal
basis of $L^2(D)$ consisting of eigenfunctions, with eigenvalues $\lambda_n\to\infty$.
\end{theorem}
\noindent The source proves this from the minimax principle. In general the spectral theorem says that a
self-adjoint operator is unitarily equivalent to multiplication by a real function on a measure space, and its
functional calculus $f\mapsto f(A)$ is a homomorphism of algebras. Every operator met so far is a function of one
operator: the heat semigroup $e^{-tA}$, the wave group $\cos(t\sqrt A)$, the resolvent $(s+A)^{-1}$, the Green's
operator $A^{-1}$, the Schr\"odinger group $e^{-itA}$. In finite dimensions this is the resolution of the identity
by eigenvectors. \chk{L9}
\section{How many eigenvalues}
\begin{theorem}[Weyl]\label{thm:weyl}
\Scl[source, \S11.6, Theorem 1] For the Dirichlet problem, $\lambda_n/n\to4\pi/A$ in the plane, with $A$ the area, and
$\lambda_n^{3/2}/n\to6\pi^2/V$ in space, with $V$ the volume.
\end{theorem}
\noindent The source's proof brackets the domain between unions of rectangles and uses \Cref{cor:mono}. For a
$1\times\sqrt2$ rectangle the count $N(\lambda)$ of eigenvalues below $\lambda$ is $208$, $869$, $3531$ at
$\lambda=2000$, $8000$, $32000$, and the two-term expression $A\lambda/4\pi-P\sqrt\lambda/4\pi$, with $P$ the
perimeter, gives $207.9$, $865.9$, $3532.5$. \chk{L4}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,4)--(0,0)--(8.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (2.1,0)--(2.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $100$};
\draw[ink2] (4.2,0)--(4.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $200$};
\draw[ink2] (6.3,0)--(6.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $300$};
\draw[ink2] (8.4,0)--(8.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $400$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,0.889)--(-0.07,0.889) node[left,inner sep=1.5pt]{\scriptsize $10$};
\draw[ink2] (0,1.778)--(-0.07,1.778) node[left,inner sep=1.5pt]{\scriptsize $20$};
\draw[ink2] (0,2.667)--(-0.07,2.667) node[left,inner sep=1.5pt]{\scriptsize $30$};
\draw[ink2] (0,3.556)--(-0.07,3.556) node[left,inner sep=1.5pt]{\scriptsize $40$};
\node[below] at (4.2,-0.38) {\scriptsize $\lambda$};
\node[above,inner sep=1pt] at (0,4.05) {\scriptsize $N(\lambda)$};
\draw[thick,fwd] plot coordinates {(0,0) (0.311,0) (0.311,0.089) (0.622,0.089) (0.622,0.178) (0.933,0.178) (0.933,0.267) (1.14,0.267) (1.14,0.356) (1.244,0.356) (1.244,0.444) (1.762,0.444) (1.762,0.533) (1.865,0.533) (1.865,0.622) (1.969,0.622) (1.969,0.711) (2.28,0.711) (2.28,0.8) (2.487,0.8) (2.487,0.889) (2.798,0.889) (2.798,0.978) (2.798,0.978) (2.798,1.067) (3.42,1.067) (3.42,1.156) (3.42,1.156) (3.42,1.244) (3.523,1.244) (3.523,1.333) (3.731,1.333) (3.731,1.422) (3.938,1.422) (3.938,1.511) (4.249,1.511) (4.249,1.6) (4.456,1.6) (4.456,1.689) (4.56,1.689) (4.56,1.778) (4.974,1.778) (4.974,1.867) (5.285,1.867) (5.285,1.956) (5.285,1.956) (5.285,2.044) (5.596,2.044) (5.596,2.133) (5.596,2.133) (5.596,2.222) (5.907,2.222) (5.907,2.311) (5.907,2.311) (5.907,2.4) (6.114,2.4) (6.114,2.489) (6.84,2.489) (6.84,2.578) (6.84,2.578) (6.84,2.667) (6.943,2.667) (6.943,2.756) (7.047,2.756) (7.047,2.844) (7.461,2.844) (7.461,2.933) (7.565,2.933) (7.565,3.022) (7.772,3.022) (7.772,3.111) (7.876,3.111) (7.876,3.2) (8.394,3.2) (8.394,3.289) (8.394,3.289) (8.394,3.378) (8.4,3.378)};
\draw[ink2,densely dashed] plot coordinates {(0,0) (0.21,0.1) (0.42,0.2) (0.63,0.3) (0.84,0.4) (1.05,0.5) (1.26,0.6) (1.47,0.7) (1.68,0.8) (1.89,0.9) (2.1,1) (2.31,1.1) (2.52,1.2) (2.73,1.3) (2.94,1.4) (3.15,1.501) (3.36,1.601) (3.57,1.701) (3.78,1.801) (3.99,1.901) (4.2,2.001) (4.41,2.101) (4.62,2.201) (4.83,2.301) (5.04,2.401) (5.25,2.501) (5.46,2.601) (5.67,2.701) (5.88,2.801) (6.09,2.901) (6.3,3.001) (6.51,3.101) (6.72,3.201) (6.93,3.301) (7.14,3.401) (7.35,3.501) (7.56,3.601) (7.77,3.701) (7.98,3.801) (8.19,3.901)};
\draw[thick,bwd] plot coordinates {(0,0) (0.105,0) (0.21,0) (0.315,0.018) (0.42,0.047) (0.525,0.079) (0.63,0.113) (0.735,0.148) (0.84,0.184) (0.945,0.221) (1.05,0.259) (1.155,0.297) (1.26,0.336) (1.365,0.375) (1.47,0.414) (1.575,0.454) (1.68,0.495) (1.785,0.535) (1.89,0.576) (1.995,0.617) (2.1,0.659) (2.205,0.7) (2.31,0.742) (2.415,0.784) (2.52,0.826) (2.625,0.869) (2.73,0.911) (2.835,0.954) (2.94,0.996) (3.045,1.039) (3.15,1.082) (3.255,1.125) (3.36,1.169) (3.465,1.212) (3.57,1.255) (3.675,1.299) (3.78,1.342) (3.885,1.386) (3.99,1.43) (4.095,1.474) (4.2,1.518) (4.305,1.562) (4.41,1.606) (4.515,1.65) (4.62,1.694) (4.725,1.738) (4.83,1.783) (4.935,1.827) (5.04,1.872) (5.145,1.916) (5.25,1.961) (5.355,2.005) (5.46,2.05) (5.565,2.095) (5.67,2.14) (5.775,2.185) (5.88,2.229) (5.985,2.274) (6.09,2.319) (6.195,2.364) (6.3,2.409) (6.405,2.455) (6.51,2.5) (6.615,2.545) (6.72,2.59) (6.825,2.635) (6.93,2.681) (7.035,2.726) (7.14,2.771) (7.245,2.817) (7.35,2.862) (7.455,2.908) (7.56,2.953) (7.665,2.999) (7.77,3.044) (7.875,3.09) (7.98,3.136) (8.085,3.181) (8.19,3.227) (8.295,3.273) (8.4,3.318)};
\node[ink2,left] at (6.09,3.511) {\scriptsize $A\lambda/4\pi$};
\node[bwd] at (6.3,1.156) {\scriptsize $A\lambda/4\pi-P\sqrt\lambda/4\pi$};
\end{scope}
\end{tikzpicture}
\figcap{The counting function of the Dirichlet eigenvalues of the $1\times\sqrt2$ rectangle (steps, 38 eigenvalues below $400$), Weyl's leading term (dashed), and the two-term expression with the perimeter correction. The area is heard first and the perimeter second.}
\end{center}
\section{The trace of the semigroup}
The function $Z(t)=\sum_ne^{-\lambda_nt}$ is the trace of the heat semigroup of \Cref{ch:diffusions}:
$Z(t)=\int_DK(x,x,t)\,dx$.
\begin{theorem}[Hearing area, perimeter and holes]\label{thm:heattrace}
\Sli[\cite{Kac66,McSi67}] For a plane domain with smooth boundary and $h$ holes,
$Z(t)=\dfrac A{4\pi t}-\dfrac P{8\sqrt{\pi t}}+\dfrac{1-h}6+o(1)$ as $t\to0$. For a polygon the constant term is
$\sum_i(\pi^2-\theta_i^2)/24\pi\theta_i$ over the interior angles $\theta_i$.
\end{theorem}
\begin{proposition}[The rectangle, exactly]\label{prop:rect}
\Scl For an $a\times b$ rectangle, $Z(t)=\dfrac{ab}{4\pi t}-\dfrac{2(a+b)}{8\sqrt{\pi t}}+\dfrac14+O(e^{-c/t})$.
\end{proposition}
\begin{proof}
By \Cref{thm:product}, $Z=z_az_b$ with $z_a(t)=\sum_{m\ge1}e^{-\pi^2m^2t/a^2}$, and Poisson summation gives
\[ z_a(t)=\frac a{2\sqrt{\pi t}}-\frac12+\frac a{\sqrt{\pi t}}\sum_{k\ge1}e^{-(2ak)^2/4t}. \qedhere \]
\end{proof}
\noindent The two sides of the last identity agree to fourteen digits, and for the rectangle the three-term
expression is exact to $10^{-15}$ at $t=0.02$. \chk{L5, L6} The identity is \Cref{thm:kernels} on the diagonal: the
left side is the sum over eigenfunctions and the right side the sum over images, now indexed by the closed
billiard paths of lengths $2ak$. It is the smallest instance of a trace formula, spectrum on one side and
closed geodesics on the other. The constant $\tfrac14$ is four corners at $\tfrac1{16}$ each; in the smooth case
the constant $(1-h)/6$ is the Euler characteristic divided by six. A topological invariant can be heard.
\section{Can one hear the shape?}
\begin{theorem}[Sunada]\label{thm:sunada}
\Sli[\cite{Sun85}] Let a finite group $G$ act freely by isometries on a compact manifold $M$, and let $H_1,H_2$ be
subgroups that meet every conjugacy class of $G$ in the same number of elements. Then $M/H_1$ and $M/H_2$ have the
same Laplace spectrum.
\end{theorem}
\begin{proof}
The condition says that the permutation representations $\Q[G/H_1]$ and $\Q[G/H_2]$ are isomorphic. For any
representation $V$ of $G$ the invariants are $V^{H_i}=\Hom_G(\Q[G/H_i],V)$, naturally in $V$. Apply this to each
eigenspace of the Laplacian on $M$: the eigenspaces of $M/H_1$ and $M/H_2$ have equal dimensions.
\end{proof}
\noindent Isospectrality is here an isomorphism in a category of representations that is not induced by an
isometry. Gordon, Webb and Wolpert used the method to produce plane domains with the same Dirichlet spectrum that
are not congruent \cite{GWW92}; the answer to Kac's question is no.
\[\begin{tikzcd}[column sep=small,row sep=1.6em]
& M\arrow[dl]\arrow[dr] & \\
M/H_1\arrow[dr] & & M/H_2\arrow[dl]\\
& M/G &
\end{tikzcd}\]
Both middle spaces are covered by $M$ and cover $M/G$; no map between them is asserted, and in general there is
none. What relates them is an isomorphism one level up, between the functors $V\mapsto V^{H_1}$ and
$V\mapsto V^{H_2}$ on representations of $G$.
The smallest case can be computed in full. In the group of order $32$ of affine maps $x\mapsto ux+a$ of $\Z/8$,
the subgroups $\{(0,1),(0,3),(0,5),(0,7)\}$ and $\{(0,1),(4,3),(4,5),(0,7)\}$ meet each of the eleven conjugacy
classes equally and are not conjugate. The two quotient graphs on eight vertices, for the generators
$(0,3),(0,5),(1,7)$, have the same characteristic polynomial and are not isomorphic: their loops are distributed
differently. \chk{L8} Without any group, an exhaustive search finds no two connected graphs on at most five vertices
with the same Laplacian spectrum and different degree sequences, and exactly two such spectra on six vertices;
the first is shared by a graph with degrees $(2,2,2,2,2,4)$ and one with degrees $(1,2,2,3,3,3)$. \chk{L7}
\begin{center}
\begin{tikzpicture}[scale=0.95,every node/.style={st}]
\begin{scope}
\node (a) at (0,0) {}; \node (b) at (1,0.65) {}; \node (c) at (1,-0.65) {};
\node (d) at (-1,0.65) {}; \node (e) at (-1,-0.65) {}; \node (f) at (-2,0) {};
\draw[thick] (a)--(b)--(c)--(a) (a)--(d)--(f)--(e)--(a);
\end{scope}
\begin{scope}[xshift=6cm]
\node (p) at (0,0.9) {}; \node (q) at (0,-0.9) {};
\node (x) at (-1,0) {}; \node (y) at (0,0) {}; \node (z) at (1,0) {}; \node (w) at (2,0) {};
\draw[thick] (p)--(x)--(q) (p)--(y)--(q) (p)--(z)--(q) (z)--(w);
\end{scope}
\end{tikzpicture}
\figcap{Two graphs with the Laplacian characteristic polynomial $x^6-14x^5+73x^4-176x^3+192x^2-72x$. One can hear
the number of spanning trees, which is $72/6=12$ for both ($3\cdot4$ on the left, $2^2\cdot3$ on the right), and
cannot hear the degrees.}
\end{center}
\section*{Notes}
The source's Chapter 11 is followed for the variational principles, completeness and Weyl's law. The heat trace
expansion is in \cite{Kac66,McSi67}. Sunada's theorem is \cite{Sun85}, and the plane domains are in \cite{GWW92}; the
group of order $32$ is Gassmann's example, see \cite{Bro88}.
\section*{Exercises}
\begin{exercise}\label{exer:rq}
Compute the Rayleigh quotient of $x(1-x)$ on $(0,1)$.
\end{exercise}
\begin{exercise}\label{exer:bracket}
Use \Cref{cor:mono} to bound the first Dirichlet eigenvalue of the unit disk between those of an inscribed and a
circumscribed square, and compare with $z_{01}^2$.
\end{exercise}
\begin{exercise}\label{exer:weylrect}
Derive the two-term count for an $a\times b$ rectangle by counting lattice points in a quarter of an ellipse.
\end{exercise}
\begin{exercise}\label{exer:corners}
Verify that the constant term of \Cref{thm:heattrace} for a polygon gives $\tfrac14$ for a rectangle, and find it
for an equilateral triangle.
\end{exercise}
\chapter{Distributions and Transforms}\label{ch:distributions}
Several objects used so far were not functions: the initial state of the source function, the pole of a Green's
function, a solution with a corner. The source legitimizes them late \src{Ch.~12}. In the language of
\Cref{ch:equations} the step is a change of the second argument of the solution functor.
\section{Distributions}
A \term{test function} on an open set $U$ is a smooth function of compact support, and a \term{distribution} is a
continuous linear functional on test functions \src{\S12.1}. Distributions restrict to smaller open sets and glue,
so they form a sheaf $\DD'$, and differential operators act on them by duality: $(\pd f,\varphi)=-(f,\pd\varphi)$.
So $\DD'$ is a sheaf of modules over the differential operators that contains the smooth functions, and
$\Sol(P;\DD')$ is defined: these are the \term{weak solutions}.
\begin{example}
The delta function is the limit of the source functions, $(S(\cdot,t),\varphi)=\varphi(0)+t\varphi''(0)+O(t^2)$ for
$k=1$. The derivative of the step function is $\delta$, and the second derivative of $|x|$ is $2\delta$.
\chk{T1, T2} The function $\tfrac12[\varphi(x+ct)+\varphi(x-ct)]$ is a weak solution of the wave equation for every
locally integrable $\varphi$; this is \Cref{cor:par} stated in $\DD'$.
\end{example}
\begin{theorem}[Kernels]\label{thm:kernel}
\Scl[\cite{Sch57}] Continuous linear maps from test functions on $X$ to distributions on $Y$ correspond bijectively
to distributions on $X\times Y$.
\end{theorem}
\noindent This is the tensor--hom adjunction for the spaces in question; it holds because they are nuclear. It is
the reason that every solution operator in this volume is given by a kernel: the source function, the Green's
function, the Poisson kernel, the Kirchhoff kernel. One limit of the theory should be recorded now. There is no
associative product on $\DD'$ extending the product of continuous functions \cite{Sch54}. Linear equations with
smooth coefficients act on distributions; non-linear ones do not, and \Cref{ch:nonlinear} must find other notions
of weak solution.
\section{Fundamental solutions}
A \term{fundamental solution} of $P$ is a distribution $E$ with $PE=\delta$; then $u=E*f$ solves $Pu=f$ for $f$ of
compact support \src{\S12.2}.
\begin{theorem}\label{thm:fund}
\Scl[\cite{Mal56,Ehr54}] Every non-zero differential operator with constant coefficients has a fundamental solution.
\end{theorem}
\begin{center}\small
\begin{tabular}{@{}lll@{}}
\toprule
operator & fundamental solution & singular support\\
\midrule
$\Lap$ in $\R^3$ & $-1/4\pi r$ & the origin\\
$\pd_t-k\pd_x^2$ & $S(x,t)$ for $t>0$, else $0$ & the origin\\
$\pd_t^2-c^2\pd_x^2$ & $1/2c$ for $|x|<ct$, else $0$ & the cone $|x|=ct$\\
$\pd_t^2-c^2\Lap$ in $\R^3$ & $\delta(r-ct)/4\pi c\,r$ & the cone $r=ct$\\
\bottomrule
\end{tabular}
\end{center}
\noindent The first line is checked against a test function. \chk{T3} The last column is the type of
\Cref{thm:type} once more: an elliptic or parabolic fundamental solution is smooth away from the pole, so that all
solutions are smooth where the right-hand side is; a hyperbolic one is singular along the characteristic cone,
which is \Cref{thm:wf}. The last line is Huygens's principle: the support, and not only the singular support, is
the cone.
\section{The Fourier transform}
\begin{theorem}\label{thm:fourier}
\Scl[source, \S12.3] The Fourier transform $F(k)=\int f(x)e^{-ikx}dx$ is an isomorphism of the Schwartz space with
itself and of the tempered distributions with themselves; it is unitary on $L^2$ up to the factor $2\pi$; and it
exchanges $\pd_x$ with multiplication by $ik$ and multiplication by $x$ with $i\pd_k$.
\end{theorem}
\noindent These properties are checked numerically. \chk{T4} Structurally: the line is its own Pontryagin dual, as
the circle and the integers were dual in \Cref{thm:pontryagin}; and the transform is an automorphism of the Weyl
algebra generated by $x$ and $\pd$ with $[\pd,x]=1$, since $[ik,i\pd_k]=1$. An equation with constant coefficients
becomes an equation of order zero, $p(ik)\hat u=\hat f$, and solving it is division by a polynomial. The zero set
of that polynomial is the characteristic variety of \Cref{ch:equations}, and the difficulty of the division there
is the whole difficulty.
The duality is functorial for homomorphisms of groups: the dual of the quotient $\R\to\R/2l\Z$ is the inclusion of
the lattice $\tfrac\pi l\Z\subset\R$. Applied to a function this is the Poisson summation formula, and applied to the
heat kernel it is \Cref{thm:kernels} and \Cref{prop:rect}. The image method and Fourier's method are dual in the
literal sense.
\section{Source functions}
With the transform the source functions are computed and not guessed \src{\S12.4}. For diffusion,
$\hat S_t=-k^2\hat S$, $\hat S(k,0)=1$, so $\hat S=e^{-k^2t}$ and $S$ is the Gaussian. \chk{T5} For Laplace's equation
in the half-plane $y>0$ the same computation gives $\hat u(k,y)=e^{-|k|y}\hat h(k)$ and the Poisson kernel
\[ P_y(x)=\frac1\pi\frac{y}{x^2+y^2}=\frac1\pi\,\mathrm{Im}\,\frac1{x-iy}. \]
Two things are visible in this formula. First, $P_y*P_{y'}=P_{y+y'}$: the Dirichlet problem in a half-space is
also a semigroup, in the variable normal to the boundary, generated by $-\sqrt{-\pd_x^2}$; it is a Markov
semigroup, the Cauchy process, and harmonic measure is its kernel. Second, $\delta$ is the boundary value of a
holomorphic function: $P_y\to\delta$ as $y\to0$, with error of order $y$. \chk{T7}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3)--(0,0)--(5,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-3$};
\draw[ink2] (2.5,0)--(2.5,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (5,0)--(5,-0.07) node[below,inner sep=1.5pt]{\scriptsize $3$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,0.882)--(-0.07,0.882) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,1.765)--(-0.07,1.765) node[left,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,2.647)--(-0.07,2.647) node[left,inner sep=1.5pt]{\scriptsize $3$};
\node[below] at (2.5,-0.38) {\scriptsize $x$};
\draw[thick,ink2] plot coordinates {(0,0.028) (0.017,0.028) (0.033,0.029) (0.05,0.029) (0.067,0.029) (0.083,0.03) (0.1,0.03) (0.117,0.031) (0.133,0.031) (0.15,0.031) (0.167,0.032) (0.183,0.032) (0.2,0.033) (0.217,0.033) (0.233,0.033) (0.25,0.034) (0.267,0.034) (0.283,0.035) (0.3,0.035) (0.317,0.036) (0.333,0.036) (0.35,0.037) (0.367,0.037) (0.383,0.038) (0.4,0.038) (0.417,0.039) (0.433,0.039) (0.45,0.04) (0.467,0.04) (0.483,0.041) (0.5,0.042) (0.517,0.042) (0.533,0.043) (0.55,0.043) (0.567,0.044) (0.583,0.045) (0.6,0.045) (0.617,0.046) (0.633,0.047) (0.65,0.047) (0.667,0.048) (0.683,0.049) (0.7,0.05) (0.717,0.05) (0.733,0.051) (0.75,0.052) (0.767,0.053) (0.783,0.054) (0.8,0.054) (0.817,0.055) (0.833,0.056) (0.85,0.057) (0.867,0.058) (0.883,0.059) (0.9,0.06) (0.917,0.061) (0.933,0.062) (0.95,0.063) (0.967,0.064) (0.983,0.065) (1,0.066) (1.017,0.067) (1.033,0.069) (1.05,0.07) (1.067,0.071) (1.083,0.072) (1.1,0.073) (1.117,0.075) (1.133,0.076) (1.15,0.077) (1.167,0.079) (1.183,0.08) (1.2,0.082) (1.217,0.083) (1.233,0.085) (1.25,0.086) (1.267,0.088) (1.283,0.09) (1.3,0.091) (1.317,0.093) (1.333,0.095) (1.35,0.097) (1.367,0.099) (1.383,0.1) (1.4,0.102) (1.417,0.104) (1.433,0.106) (1.45,0.109) (1.467,0.111) (1.483,0.113) (1.5,0.115) (1.517,0.117) (1.533,0.12) (1.55,0.122) (1.567,0.125) (1.583,0.127) (1.6,0.13) (1.617,0.132) (1.633,0.135) (1.65,0.138) (1.667,0.14) (1.683,0.143) (1.7,0.146) (1.717,0.149) (1.733,0.152) (1.75,0.155) (1.767,0.158) (1.783,0.161) (1.8,0.165) (1.817,0.168) (1.833,0.171) (1.85,0.175) (1.867,0.178) (1.883,0.181) (1.9,0.185) (1.917,0.188) (1.933,0.192) (1.95,0.196) (1.967,0.199) (1.983,0.203) (2,0.207) (2.017,0.21) (2.033,0.214) (2.05,0.217) (2.067,0.221) (2.083,0.225) (2.1,0.228) (2.117,0.232) (2.133,0.235) (2.15,0.239) (2.167,0.242) (2.183,0.245) (2.2,0.249) (2.217,0.252) (2.233,0.255) (2.25,0.258) (2.267,0.26) (2.283,0.263) (2.3,0.266) (2.317,0.268) (2.333,0.27) (2.35,0.272) (2.367,0.274) (2.383,0.275) (2.4,0.277) (2.417,0.278) (2.433,0.279) (2.45,0.28) (2.467,0.28) (2.483,0.281) (2.5,0.281) (2.517,0.281) (2.533,0.28) (2.55,0.28) (2.567,0.279) (2.583,0.278) (2.6,0.277) (2.617,0.275) (2.633,0.274) (2.65,0.272) (2.667,0.27) (2.683,0.268) (2.7,0.266) (2.717,0.263) (2.733,0.26) (2.75,0.258) (2.767,0.255) (2.783,0.252) (2.8,0.249) (2.817,0.245) (2.833,0.242) (2.85,0.239) (2.867,0.235) (2.883,0.232) (2.9,0.228) (2.917,0.225) (2.933,0.221) (2.95,0.217) (2.967,0.214) (2.983,0.21) (3,0.207) (3.017,0.203) (3.033,0.199) (3.05,0.196) (3.067,0.192) (3.083,0.188) (3.1,0.185) (3.117,0.181) (3.133,0.178) (3.15,0.175) (3.167,0.171) (3.183,0.168) (3.2,0.165) (3.217,0.161) (3.233,0.158) (3.25,0.155) (3.267,0.152) (3.283,0.149) (3.3,0.146) (3.317,0.143) (3.333,0.14) (3.35,0.138) (3.367,0.135) (3.383,0.132) (3.4,0.13) (3.417,0.127) (3.433,0.125) (3.45,0.122) (3.467,0.12) (3.483,0.117) (3.5,0.115) (3.517,0.113) (3.533,0.111) (3.55,0.109) (3.567,0.106) (3.583,0.104) (3.6,0.102) (3.617,0.1) (3.633,0.099) (3.65,0.097) (3.667,0.095) (3.683,0.093) (3.7,0.091) (3.717,0.09) (3.733,0.088) (3.75,0.086) (3.767,0.085) (3.783,0.083) (3.8,0.082) (3.817,0.08) (3.833,0.079) (3.85,0.077) (3.867,0.076) (3.883,0.075) (3.9,0.073) (3.917,0.072) (3.933,0.071) (3.95,0.07) (3.967,0.069) (3.983,0.067) (4,0.066) (4.017,0.065) (4.033,0.064) (4.05,0.063) (4.067,0.062) (4.083,0.061) (4.1,0.06) (4.117,0.059) (4.133,0.058) (4.15,0.057) (4.167,0.056) (4.183,0.055) (4.2,0.054) (4.217,0.054) (4.233,0.053) (4.25,0.052) (4.267,0.051) (4.283,0.05) (4.3,0.05) (4.317,0.049) (4.333,0.048) (4.35,0.047) (4.367,0.047) (4.383,0.046) (4.4,0.045) (4.417,0.045) (4.433,0.044) (4.45,0.043) (4.467,0.043) (4.483,0.042) (4.5,0.042) (4.517,0.041) (4.533,0.04) (4.55,0.04) (4.567,0.039) (4.583,0.039) (4.6,0.038) (4.617,0.038) (4.633,0.037) (4.65,0.037) (4.667,0.036) (4.683,0.036) (4.7,0.035) (4.717,0.035) (4.733,0.034) (4.75,0.034) (4.767,0.033) (4.783,0.033) (4.8,0.033) (4.817,0.032) (4.833,0.032) (4.85,0.031) (4.867,0.031) (4.883,0.031) (4.9,0.03) (4.917,0.03) (4.933,0.029) (4.95,0.029) (4.967,0.029) (4.983,0.028) (5,0.028)};
\draw[thick,fwd] plot coordinates {(0,0.009) (0.017,0.009) (0.033,0.01) (0.05,0.01) (0.067,0.01) (0.083,0.01) (0.1,0.01) (0.117,0.01) (0.133,0.01) (0.15,0.01) (0.167,0.011) (0.183,0.011) (0.2,0.011) (0.217,0.011) (0.233,0.011) (0.25,0.011) (0.267,0.012) (0.283,0.012) (0.3,0.012) (0.317,0.012) (0.333,0.012) (0.35,0.012) (0.367,0.013) (0.383,0.013) (0.4,0.013) (0.417,0.013) (0.433,0.014) (0.45,0.014) (0.467,0.014) (0.483,0.014) (0.5,0.014) (0.517,0.015) (0.533,0.015) (0.55,0.015) (0.567,0.015) (0.583,0.016) (0.6,0.016) (0.617,0.016) (0.633,0.016) (0.65,0.017) (0.667,0.017) (0.683,0.017) (0.7,0.018) (0.717,0.018) (0.733,0.018) (0.75,0.019) (0.767,0.019) (0.783,0.019) (0.8,0.02) (0.817,0.02) (0.833,0.021) (0.85,0.021) (0.867,0.021) (0.883,0.022) (0.9,0.022) (0.917,0.023) (0.933,0.023) (0.95,0.024) (0.967,0.024) (0.983,0.025) (1,0.025) (1.017,0.026) (1.033,0.026) (1.05,0.027) (1.067,0.028) (1.083,0.028) (1.1,0.029) (1.117,0.03) (1.133,0.03) (1.15,0.031) (1.167,0.032) (1.183,0.033) (1.2,0.033) (1.217,0.034) (1.233,0.035) (1.25,0.036) (1.267,0.037) (1.283,0.038) (1.3,0.039) (1.317,0.04) (1.333,0.041) (1.35,0.042) (1.367,0.043) (1.383,0.045) (1.4,0.046) (1.417,0.047) (1.433,0.049) (1.45,0.05) (1.467,0.052) (1.483,0.053) (1.5,0.055) (1.517,0.057) (1.533,0.059) (1.55,0.061) (1.567,0.063) (1.583,0.065) (1.6,0.067) (1.617,0.069) (1.633,0.072) (1.65,0.075) (1.667,0.077) (1.683,0.08) (1.7,0.083) (1.717,0.087) (1.733,0.09) (1.75,0.094) (1.767,0.097) (1.783,0.102) (1.8,0.106) (1.817,0.111) (1.833,0.115) (1.85,0.121) (1.867,0.126) (1.883,0.132) (1.9,0.138) (1.917,0.145) (1.933,0.153) (1.95,0.16) (1.967,0.169) (1.983,0.178) (2,0.187) (2.017,0.198) (2.033,0.209) (2.05,0.221) (2.067,0.234) (2.083,0.248) (2.1,0.263) (2.117,0.279) (2.133,0.297) (2.15,0.316) (2.167,0.337) (2.183,0.359) (2.2,0.384) (2.217,0.41) (2.233,0.438) (2.25,0.468) (2.267,0.5) (2.283,0.535) (2.3,0.571) (2.317,0.609) (2.333,0.648) (2.35,0.688) (2.367,0.729) (2.383,0.769) (2.4,0.807) (2.417,0.843) (2.433,0.874) (2.45,0.9) (2.467,0.92) (2.483,0.932) (2.5,0.936) (2.517,0.932) (2.533,0.92) (2.55,0.9) (2.567,0.874) (2.583,0.843) (2.6,0.807) (2.617,0.769) (2.633,0.729) (2.65,0.688) (2.667,0.648) (2.683,0.609) (2.7,0.571) (2.717,0.535) (2.733,0.5) (2.75,0.468) (2.767,0.438) (2.783,0.41) (2.8,0.384) (2.817,0.359) (2.833,0.337) (2.85,0.316) (2.867,0.297) (2.883,0.279) (2.9,0.263) (2.917,0.248) (2.933,0.234) (2.95,0.221) (2.967,0.209) (2.983,0.198) (3,0.187) (3.017,0.178) (3.033,0.169) (3.05,0.16) (3.067,0.153) (3.083,0.145) (3.1,0.138) (3.117,0.132) (3.133,0.126) (3.15,0.121) (3.167,0.115) (3.183,0.111) (3.2,0.106) (3.217,0.102) (3.233,0.097) (3.25,0.094) (3.267,0.09) (3.283,0.087) (3.3,0.083) (3.317,0.08) (3.333,0.077) (3.35,0.075) (3.367,0.072) (3.383,0.069) (3.4,0.067) (3.417,0.065) (3.433,0.063) (3.45,0.061) (3.467,0.059) (3.483,0.057) (3.5,0.055) (3.517,0.053) (3.533,0.052) (3.55,0.05) (3.567,0.049) (3.583,0.047) (3.6,0.046) (3.617,0.045) (3.633,0.043) (3.65,0.042) (3.667,0.041) (3.683,0.04) (3.7,0.039) (3.717,0.038) (3.733,0.037) (3.75,0.036) (3.767,0.035) (3.783,0.034) (3.8,0.033) (3.817,0.033) (3.833,0.032) (3.85,0.031) (3.867,0.03) (3.883,0.03) (3.9,0.029) (3.917,0.028) (3.933,0.028) (3.95,0.027) (3.967,0.026) (3.983,0.026) (4,0.025) (4.017,0.025) (4.033,0.024) (4.05,0.024) (4.067,0.023) (4.083,0.023) (4.1,0.022) (4.117,0.022) (4.133,0.021) (4.15,0.021) (4.167,0.021) (4.183,0.02) (4.2,0.02) (4.217,0.019) (4.233,0.019) (4.25,0.019) (4.267,0.018) (4.283,0.018) (4.3,0.018) (4.317,0.017) (4.333,0.017) (4.35,0.017) (4.367,0.016) (4.383,0.016) (4.4,0.016) (4.417,0.016) (4.433,0.015) (4.45,0.015) (4.467,0.015) (4.483,0.015) (4.5,0.014) (4.517,0.014) (4.533,0.014) (4.55,0.014) (4.567,0.014) (4.583,0.013) (4.6,0.013) (4.617,0.013) (4.633,0.013) (4.65,0.012) (4.667,0.012) (4.683,0.012) (4.7,0.012) (4.717,0.012) (4.733,0.012) (4.75,0.011) (4.767,0.011) (4.783,0.011) (4.8,0.011) (4.817,0.011) (4.833,0.011) (4.85,0.01) (4.867,0.01) (4.883,0.01) (4.9,0.01) (4.917,0.01) (4.933,0.01) (4.95,0.01) (4.967,0.01) (4.983,0.009) (5,0.009)};
\draw[thick,bwd] plot coordinates {(0,0.003) (0.017,0.003) (0.033,0.003) (0.05,0.003) (0.067,0.003) (0.083,0.003) (0.1,0.003) (0.117,0.003) (0.133,0.003) (0.15,0.004) (0.167,0.004) (0.183,0.004) (0.2,0.004) (0.217,0.004) (0.233,0.004) (0.25,0.004) (0.267,0.004) (0.283,0.004) (0.3,0.004) (0.317,0.004) (0.333,0.004) (0.35,0.004) (0.367,0.004) (0.383,0.004) (0.4,0.004) (0.417,0.004) (0.433,0.005) (0.45,0.005) (0.467,0.005) (0.483,0.005) (0.5,0.005) (0.517,0.005) (0.533,0.005) (0.55,0.005) (0.567,0.005) (0.583,0.005) (0.6,0.005) (0.617,0.005) (0.633,0.006) (0.65,0.006) (0.667,0.006) (0.683,0.006) (0.7,0.006) (0.717,0.006) (0.733,0.006) (0.75,0.006) (0.767,0.006) (0.783,0.007) (0.8,0.007) (0.817,0.007) (0.833,0.007) (0.85,0.007) (0.867,0.007) (0.883,0.007) (0.9,0.008) (0.917,0.008) (0.933,0.008) (0.95,0.008) (0.967,0.008) (0.983,0.008) (1,0.009) (1.017,0.009) (1.033,0.009) (1.05,0.009) (1.067,0.009) (1.083,0.01) (1.1,0.01) (1.117,0.01) (1.133,0.01) (1.15,0.011) (1.167,0.011) (1.183,0.011) (1.2,0.011) (1.217,0.012) (1.233,0.012) (1.25,0.012) (1.267,0.013) (1.283,0.013) (1.3,0.013) (1.317,0.014) (1.333,0.014) (1.35,0.015) (1.367,0.015) (1.383,0.016) (1.4,0.016) (1.417,0.017) (1.433,0.017) (1.45,0.018) (1.467,0.018) (1.483,0.019) (1.5,0.019) (1.517,0.02) (1.533,0.021) (1.55,0.021) (1.567,0.022) (1.583,0.023) (1.6,0.024) (1.617,0.025) (1.633,0.026) (1.65,0.027) (1.667,0.028) (1.683,0.029) (1.7,0.03) (1.717,0.031) (1.733,0.033) (1.75,0.034) (1.767,0.036) (1.783,0.037) (1.8,0.039) (1.817,0.041) (1.833,0.043) (1.85,0.045) (1.867,0.048) (1.883,0.05) (1.9,0.053) (1.917,0.056) (1.933,0.059) (1.95,0.063) (1.967,0.067) (1.983,0.071) (2,0.076) (2.017,0.081) (2.033,0.087) (2.05,0.093) (2.067,0.1) (2.083,0.108) (2.1,0.117) (2.117,0.127) (2.133,0.138) (2.15,0.151) (2.167,0.165) (2.183,0.182) (2.2,0.201) (2.217,0.224) (2.233,0.25) (2.25,0.281) (2.267,0.318) (2.283,0.362) (2.3,0.415) (2.317,0.481) (2.333,0.562) (2.35,0.662) (2.367,0.789) (2.383,0.949) (2.4,1.151) (2.417,1.404) (2.433,1.713) (2.45,2.065) (2.467,2.421) (2.483,2.701) (2.5,2.809) (2.517,2.701) (2.533,2.421) (2.55,2.065) (2.567,1.713) (2.583,1.404) (2.6,1.151) (2.617,0.949) (2.633,0.789) (2.65,0.662) (2.667,0.562) (2.683,0.481) (2.7,0.415) (2.717,0.362) (2.733,0.318) (2.75,0.281) (2.767,0.25) (2.783,0.224) (2.8,0.201) (2.817,0.182) (2.833,0.165) (2.85,0.151) (2.867,0.138) (2.883,0.127) (2.9,0.117) (2.917,0.108) (2.933,0.1) (2.95,0.093) (2.967,0.087) (2.983,0.081) (3,0.076) (3.017,0.071) (3.033,0.067) (3.05,0.063) (3.067,0.059) (3.083,0.056) (3.1,0.053) (3.117,0.05) (3.133,0.048) (3.15,0.045) (3.167,0.043) (3.183,0.041) (3.2,0.039) (3.217,0.037) (3.233,0.036) (3.25,0.034) (3.267,0.033) (3.283,0.031) (3.3,0.03) (3.317,0.029) (3.333,0.028) (3.35,0.027) (3.367,0.026) (3.383,0.025) (3.4,0.024) (3.417,0.023) (3.433,0.022) (3.45,0.021) (3.467,0.021) (3.483,0.02) (3.5,0.019) (3.517,0.019) (3.533,0.018) (3.55,0.018) (3.567,0.017) (3.583,0.017) (3.6,0.016) (3.617,0.016) (3.633,0.015) (3.65,0.015) (3.667,0.014) (3.683,0.014) (3.7,0.013) (3.717,0.013) (3.733,0.013) (3.75,0.012) (3.767,0.012) (3.783,0.012) (3.8,0.011) (3.817,0.011) (3.833,0.011) (3.85,0.011) (3.867,0.01) (3.883,0.01) (3.9,0.01) (3.917,0.01) (3.933,0.009) (3.95,0.009) (3.967,0.009) (3.983,0.009) (4,0.009) (4.017,0.008) (4.033,0.008) (4.05,0.008) (4.067,0.008) (4.083,0.008) (4.1,0.008) (4.117,0.007) (4.133,0.007) (4.15,0.007) (4.167,0.007) (4.183,0.007) (4.2,0.007) (4.217,0.007) (4.233,0.006) (4.25,0.006) (4.267,0.006) (4.283,0.006) (4.3,0.006) (4.317,0.006) (4.333,0.006) (4.35,0.006) (4.367,0.006) (4.383,0.005) (4.4,0.005) (4.417,0.005) (4.433,0.005) (4.45,0.005) (4.467,0.005) (4.483,0.005) (4.5,0.005) (4.517,0.005) (4.533,0.005) (4.55,0.005) (4.567,0.005) (4.583,0.004) (4.6,0.004) (4.617,0.004) (4.633,0.004) (4.65,0.004) (4.667,0.004) (4.683,0.004) (4.7,0.004) (4.717,0.004) (4.733,0.004) (4.75,0.004) (4.767,0.004) (4.783,0.004) (4.8,0.004) (4.817,0.004) (4.833,0.004) (4.85,0.004) (4.867,0.003) (4.883,0.003) (4.9,0.003) (4.917,0.003) (4.933,0.003) (4.95,0.003) (4.967,0.003) (4.983,0.003) (5,0.003)};
\node[bwd,right] at (2.708,2.735) {\scriptsize $y=0.1$};
\node[fwd,right] at (2.958,0.882) {\scriptsize $0.3$};
\node[ink2] at (4,0.247) {\scriptsize $1$};
\end{scope}
\begin{scope}[shift={(6.6,0)}]
\draw[ink2] (0,3)--(0,0)--(5,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (1.667,0)--(1.667,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (3.333,0)--(3.333,-0.07) node[below,inner sep=1.5pt]{\scriptsize $8$};
\draw[ink2] (5,0)--(5,-0.07) node[below,inner sep=1.5pt]{\scriptsize $12$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.429)--(-0.07,1.429) node[left,inner sep=1.5pt]{\scriptsize $0.5$};
\draw[ink2] (0,2.857)--(-0.07,2.857) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (2.5,-0.38) {\scriptsize $\xi$};
\draw[thick,ink2] plot coordinates {(0,2.857) (0.062,2.459) (0.125,2.117) (0.188,1.822) (0.25,1.568) (0.312,1.35) (0.375,1.162) (0.438,1) (0.5,0.861) (0.562,0.741) (0.625,0.638) (0.688,0.549) (0.75,0.472) (0.812,0.406) (0.875,0.35) (0.938,0.301) (1,0.259) (1.062,0.223) (1.125,0.192) (1.188,0.165) (1.25,0.142) (1.312,0.122) (1.375,0.105) (1.438,0.091) (1.5,0.078) (1.562,0.067) (1.625,0.058) (1.688,0.05) (1.75,0.043) (1.812,0.037) (1.875,0.032) (1.938,0.027) (2,0.024) (2.062,0.02) (2.125,0.017) (2.188,0.015) (2.25,0.013) (2.312,0.011) (2.375,0.01) (2.438,0.008) (2.5,0.007) (2.562,0.006) (2.625,0.005) (2.688,0.005) (2.75,0.004) (2.812,0.003) (2.875,0.003) (2.938,0.002) (3,0.002) (3.062,0.002) (3.125,0.002) (3.188,0.001) (3.25,0.001) (3.312,0.001) (3.375,0.001) (3.438,0.001) (3.5,0.001) (3.562,0.001) (3.625,0) (3.688,0) (3.75,0) (3.812,0) (3.875,0) (3.938,0) (4,0) (4.062,0) (4.125,0) (4.188,0) (4.25,0) (4.312,0) (4.375,0) (4.438,0) (4.5,0) (4.562,0) (4.625,0) (4.688,0) (4.75,0) (4.812,0) (4.875,0) (4.938,0) (5,0)};
\draw[thick,fwd] plot coordinates {(0,2.857) (0.062,2.731) (0.125,2.611) (0.188,2.496) (0.25,2.386) (0.312,2.281) (0.375,2.181) (0.438,2.085) (0.5,1.993) (0.562,1.906) (0.625,1.822) (0.688,1.742) (0.75,1.665) (0.812,1.592) (0.875,1.522) (0.938,1.455) (1,1.391) (1.062,1.33) (1.125,1.271) (1.188,1.215) (1.25,1.162) (1.312,1.111) (1.375,1.062) (1.438,1.015) (1.5,0.97) (1.562,0.928) (1.625,0.887) (1.688,0.848) (1.75,0.81) (1.812,0.775) (1.875,0.741) (1.938,0.708) (2,0.677) (2.062,0.647) (2.125,0.619) (2.188,0.591) (2.25,0.565) (2.312,0.541) (2.375,0.517) (2.438,0.494) (2.5,0.472) (2.562,0.452) (2.625,0.432) (2.688,0.413) (2.75,0.394) (2.812,0.377) (2.875,0.361) (2.938,0.345) (3,0.33) (3.062,0.315) (3.125,0.301) (3.188,0.288) (3.25,0.275) (3.312,0.263) (3.375,0.252) (3.438,0.24) (3.5,0.23) (3.562,0.22) (3.625,0.21) (3.688,0.201) (3.75,0.192) (3.812,0.184) (3.875,0.175) (3.938,0.168) (4,0.16) (4.062,0.153) (4.125,0.147) (4.188,0.14) (4.25,0.134) (4.312,0.128) (4.375,0.122) (4.438,0.117) (4.5,0.112) (4.562,0.107) (4.625,0.102) (4.688,0.098) (4.75,0.093) (4.812,0.089) (4.875,0.085) (4.938,0.082) (5,0.078)};
\draw[thick,bwd] plot coordinates {(0,2.857) (0.062,2.815) (0.125,2.773) (0.188,2.731) (0.25,2.691) (0.312,2.651) (0.375,2.611) (0.438,2.572) (0.5,2.534) (0.562,2.496) (0.625,2.459) (0.688,2.423) (0.75,2.386) (0.812,2.351) (0.875,2.316) (0.938,2.281) (1,2.248) (1.062,2.214) (1.125,2.181) (1.188,2.149) (1.25,2.117) (1.312,2.085) (1.375,2.054) (1.438,2.023) (1.5,1.993) (1.562,1.964) (1.625,1.934) (1.688,1.906) (1.75,1.877) (1.812,1.849) (1.875,1.822) (1.938,1.795) (2,1.768) (2.062,1.742) (2.125,1.716) (2.188,1.69) (2.25,1.665) (2.312,1.64) (2.375,1.616) (2.438,1.592) (2.5,1.568) (2.562,1.545) (2.625,1.522) (2.688,1.499) (2.75,1.477) (2.812,1.455) (2.875,1.433) (2.938,1.412) (3,1.391) (3.062,1.37) (3.125,1.35) (3.188,1.33) (3.25,1.31) (3.312,1.29) (3.375,1.271) (3.438,1.252) (3.5,1.233) (3.562,1.215) (3.625,1.197) (3.688,1.179) (3.75,1.162) (3.812,1.144) (3.875,1.127) (3.938,1.111) (4,1.094) (4.062,1.078) (4.125,1.062) (4.188,1.046) (4.25,1.03) (4.312,1.015) (4.375,1) (4.438,0.985) (4.5,0.97) (4.562,0.956) (4.625,0.942) (4.688,0.928) (4.75,0.914) (4.812,0.9) (4.875,0.887) (4.938,0.874) (5,0.861)};
\node[bwd] at (3.958,1.429) {\scriptsize $y=0.1$};
\node[fwd] at (2.167,0.943) {\scriptsize $0.3$};
\node[ink2] at (0.917,0.571) {\scriptsize $1$};
\end{scope}
\end{tikzpicture}
\figcap{Left: the Poisson kernel $P_y(x)=y/\pi(x^2+y^2)$ of the half-plane, tending to the delta function as $y\to0$. Right: its Fourier transform $e^{-y|\xi|}$, tending to the constant $1$. Convergence of distributions on one side is pointwise convergence of multipliers on the other.}
\end{center}
\begin{remark}[Hyperfunctions]\label{rem:hyper}
\Sli[\cite{Sat59}] Sato defines a generalized function on the line as a pair of holomorphic functions above and
below it, modulo functions holomorphic across it: the space is $\OO(\C\setminus\R)/\OO(\C)$, the first
cohomology of the plane with support in the line and coefficients in $\OO$. Distributions embed in it, $\delta$ as
the class of $-1/2\pi iz$. This is the second place in the volume, after \Cref{prop:annulus}, where a cohomology
group is the object of study and not a tool; in several variables it is a higher cohomology group, and the theory
of linear equations built on it is derived from the start \cite{KaSc90}.
\end{remark}
\section{The Laplace transform}
\begin{proposition}\label{prop:resolvent}
\Scl[source, \S12.5; \cite{Paz83}] For a contraction semigroup with generator $A$ and $s>0$,
$(s-A)^{-1}=\int_0^\infty e^{-st}T_t\,dt$.
\end{proposition}
\noindent So the Laplace transform in time, the source's last technique, turns an evolution problem into a family
of stationary ones, and the three presentations of one object, semigroup, generator and resolvent, determine one
another. \chk{T6} Together with \Cref{thm:wick} this closes a circle: elliptic problems $(s-A)u=f$, parabolic
problems $u_t=Au$ and, for $A\le0$ self-adjoint, hyperbolic problems $u_{tt}=Au$ are three functions of one
operator.
\section*{Notes}
The source's Chapter 12 is followed. For distributions and the kernel theorem see \cite{Sch57,Hor83}; for the
impossibility of multiplication \cite{Sch54}; for fundamental solutions \cite{Mal56,Ehr54}; for hyperfunctions
\cite{Sat59,KaSc90}.
\section*{Exercises}
\begin{exercise}\label{exer:absx}
Show that $|x|''=2\delta$ in the sense of distributions.
\end{exercise}
\begin{exercise}\label{exer:fund1}
Find fundamental solutions of $d^2/dx^2$ and of $-d^2/dx^2+1$ on the line.
\end{exercise}
\begin{exercise}\label{exer:fts}
Derive the source function of the diffusion equation from $\hat S=e^{-k^2t}$.
\end{exercise}
\begin{exercise}\label{exer:poisson-sg}
Show that the Poisson kernels of the half-plane form a semigroup under convolution.
\end{exercise}
\chapter{Discretization}\label{ch:numerics}
The source devotes a chapter to computation and opens it with a warning: a scheme that looks reasonable can
produce nonsense \src{\S8.1}. The categorical view of a numerical method is that it is a functor on a small
category of grids that approximately commutes with the solution functor; the warning is that approximate
commutation of one step says nothing about many.
\section{Squares that almost commute}
Let $E(t)$ be the solution operator of a well-posed linear evolution problem, $R_h$ the restriction of functions to
a grid of mesh $h$, and $E_h$ one step of a scheme with time step $\Delta t$ tied to $h$.
\[ \begin{tikzcd}[column sep=large]
X\ar[r,"E(\Delta t)"]\ar[d,"R_h"'] & X\ar[d,"R_h"]\\
X_h\ar[r,"E_h"'] & X_h
\end{tikzcd} \]
The scheme is \term{consistent} of order $p$ if the square commutes up to $O(\Delta t\,h^p)$ on smooth functions;
\term{stable} if the powers $E_h^n$ are bounded uniformly for $n\Delta t\le T$; \term{convergent} if
$E_h^nR_h-R_hE(n\Delta t)\to0$.
\begin{theorem}[Lax equivalence]\label{thm:lax}
\Scl[\cite{LaRi56}] For a consistent scheme, stability is equivalent to convergence.
\end{theorem}
\begin{proof}[Idea]
$E_h^nR_h-R_hE^n=\sum_{j<n}E_h^{n-1-j}(E_hR_h-R_hE)E^j$: the defect of one square, transported by the remaining
steps. Stability is exactly what keeps the transport from amplifying it.
\end{proof}
\noindent The centred three-level scheme for the heat equation is consistent of second order and its amplification
factor exceeds one for every mesh ratio, $1.48$ at ratio $0.1$: it converges for no data but the trivial. \chk{N10}
\section{Diffusion}
For the explicit scheme with $s=k\Delta t/\Delta x^2$ the amplification factor of the mode $e^{ikx}$ is
$1-2s(1-\cos k\Delta x)$ \src{\S8.2}.
\begin{theorem}\label{thm:explicit}
\Scl[source, \S8.2] The explicit scheme is stable if and only if $s\le\tfrac12$, which is also the condition for its
matrix to have non-negative entries.
\end{theorem}
\noindent At $s=0.52$ the computed growth per step equals the predicted $1.0684$; at $s=0.5$ nothing grows.
\chk{N1} The second clause is \Cref{thm:max}: stability, here, is that the scheme is a morphism of the Markov
category, and the source's observation that ``none of the coefficients are negative'' predicts the right
condition for that reason. The error is of second order in $\Delta x$, and of fourth order at the special ratio
$s=\tfrac16$, where the leading error terms in time and space cancel: halving the mesh divides the error by
$4.06$ and by $16.1$. \chk{N2} The Crank--Nicolson scheme, the Cayley transform of the generator, has amplification
factor of modulus at most one for every $s$ and is of second order in time. \chk{N3}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3.6)--(0,0)--(6.6,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.3,0)--(3.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $\pi/2$};
\draw[ink2] (6.6,0)--(6.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $\pi$};
\draw[ink2] (0,0.785)--(-0.07,0.785) node[left,inner sep=1.5pt]{\scriptsize $-1$};
\draw[ink2] (0,2.095)--(-0.07,2.095) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.404)--(-0.07,3.404) node[left,inner sep=1.5pt]{\scriptsize $1$};
\node[below] at (3.3,-0.38) {\scriptsize $k\,\Delta x$};
\node[above,inner sep=1pt] at (0,3.65) {\scriptsize $\xi$};
\fill[fwd!8] (0,0.785) rectangle (6.6,3.404);
\draw[ink2,very thin] (0,2.095)--(6.6,2.095);
\draw[ink2] (0,0.785)--(6.6,0.785);
\draw[ink2] (0,3.404)--(6.6,3.404);
\draw[thick,ink2] plot coordinates {(0,3.404) (0.082,3.403) (0.165,3.402) (0.247,3.399) (0.33,3.396) (0.412,3.391) (0.495,3.386) (0.578,3.379) (0.66,3.372) (0.742,3.363) (0.825,3.354) (0.907,3.344) (0.99,3.332) (1.072,3.32) (1.155,3.307) (1.237,3.293) (1.32,3.279) (1.402,3.263) (1.485,3.247) (1.568,3.23) (1.65,3.212) (1.732,3.193) (1.815,3.174) (1.897,3.154) (1.98,3.134) (2.062,3.113) (2.145,3.091) (2.228,3.069) (2.31,3.046) (2.393,3.023) (2.475,3) (2.558,2.976) (2.64,2.951) (2.723,2.927) (2.805,2.902) (2.887,2.877) (2.97,2.851) (3.052,2.826) (3.135,2.8) (3.217,2.775) (3.3,2.749) (3.382,2.723) (3.465,2.698) (3.547,2.672) (3.63,2.647) (3.712,2.621) (3.795,2.596) (3.877,2.571) (3.96,2.547) (4.042,2.523) (4.125,2.499) (4.207,2.475) (4.29,2.452) (4.373,2.429) (4.455,2.407) (4.538,2.385) (4.62,2.364) (4.702,2.344) (4.785,2.324) (4.867,2.305) (4.95,2.286) (5.032,2.268) (5.115,2.251) (5.197,2.235) (5.28,2.22) (5.362,2.205) (5.445,2.191) (5.527,2.178) (5.61,2.166) (5.692,2.155) (5.775,2.144) (5.858,2.135) (5.94,2.127) (6.022,2.119) (6.105,2.113) (6.187,2.107) (6.27,2.103) (6.353,2.099) (6.435,2.097) (6.518,2.095) (6.6,2.095)};
\draw[thick,fwd] plot coordinates {(0,3.404) (0.082,3.403) (0.165,3.4) (0.247,3.395) (0.33,3.388) (0.412,3.378) (0.495,3.367) (0.578,3.354) (0.66,3.34) (0.742,3.323) (0.825,3.304) (0.907,3.283) (0.99,3.261) (1.072,3.237) (1.155,3.211) (1.237,3.183) (1.32,3.154) (1.402,3.123) (1.485,3.09) (1.568,3.056) (1.65,3.02) (1.732,2.983) (1.815,2.945) (1.897,2.905) (1.98,2.864) (2.062,2.822) (2.145,2.779) (2.228,2.734) (2.31,2.689) (2.393,2.643) (2.475,2.596) (2.558,2.548) (2.64,2.499) (2.723,2.45) (2.805,2.4) (2.887,2.35) (2.97,2.299) (3.052,2.248) (3.135,2.197) (3.217,2.146) (3.3,2.095) (3.382,2.043) (3.465,1.992) (3.547,1.941) (3.63,1.89) (3.712,1.839) (3.795,1.789) (3.877,1.739) (3.96,1.69) (4.042,1.641) (4.125,1.594) (4.207,1.546) (4.29,1.5) (4.373,1.455) (4.455,1.411) (4.538,1.367) (4.62,1.325) (4.702,1.284) (4.785,1.244) (4.867,1.206) (4.95,1.169) (5.032,1.133) (5.115,1.099) (5.197,1.066) (5.28,1.035) (5.362,1.006) (5.445,0.978) (5.527,0.952) (5.61,0.928) (5.692,0.906) (5.775,0.885) (5.858,0.866) (5.94,0.85) (6.022,0.835) (6.105,0.822) (6.187,0.811) (6.27,0.802) (6.353,0.795) (6.435,0.789) (6.518,0.786) (6.6,0.785)};
\draw[thick,bwd] plot coordinates {(0,3.404) (0.082,3.402) (0.165,3.399) (0.247,3.393) (0.33,3.384) (0.412,3.373) (0.495,3.36) (0.578,3.345) (0.66,3.327) (0.742,3.307) (0.825,3.284) (0.907,3.259) (0.99,3.232) (1.072,3.203) (1.155,3.172) (1.237,3.139) (1.32,3.104) (1.402,3.066) (1.485,3.027) (1.568,2.986) (1.65,2.944) (1.732,2.899) (1.815,2.853) (1.897,2.805) (1.98,2.756) (2.062,2.705) (2.145,2.654) (2.228,2.6) (2.31,2.546) (2.393,2.49) (2.475,2.434) (2.558,2.376) (2.64,2.318) (2.723,2.259) (2.805,2.199) (2.887,2.139) (2.97,2.078) (3.052,2.017) (3.135,1.956) (3.217,1.894) (3.3,1.833) (3.382,1.771) (3.465,1.709) (3.547,1.648) (3.63,1.587) (3.712,1.526) (3.795,1.466) (3.877,1.406) (3.96,1.347) (4.042,1.289) (4.125,1.232) (4.207,1.175) (4.29,1.12) (4.373,1.065) (4.455,1.012) (4.538,0.96) (4.62,0.909) (4.702,0.86) (4.785,0.813) (4.867,0.766) (4.95,0.722) (5.032,0.679) (5.115,0.638) (5.197,0.599) (5.28,0.562) (5.362,0.527) (5.445,0.493) (5.527,0.462) (5.61,0.433) (5.692,0.406) (5.775,0.381) (5.858,0.359) (5.94,0.339) (6.022,0.321) (6.105,0.305) (6.187,0.292) (6.27,0.281) (6.353,0.273) (6.435,0.267) (6.518,0.263) (6.6,0.262)};
\draw[thick,accent,densely dashed] plot coordinates {(0,3.404) (0.082,3.4) (0.165,3.388) (0.247,3.368) (0.33,3.341) (0.412,3.307) (0.495,3.267) (0.578,3.221) (0.66,3.17) (0.742,3.116) (0.825,3.058) (0.907,2.997) (0.99,2.935) (1.072,2.872) (1.155,2.808) (1.237,2.744) (1.32,2.68) (1.402,2.617) (1.485,2.555) (1.568,2.495) (1.65,2.436) (1.732,2.38) (1.815,2.325) (1.897,2.272) (1.98,2.221) (2.062,2.172) (2.145,2.125) (2.228,2.08) (2.31,2.037) (2.393,1.996) (2.475,1.957) (2.558,1.92) (2.64,1.885) (2.723,1.851) (2.805,1.819) (2.887,1.789) (2.97,1.76) (3.052,1.732) (3.135,1.706) (3.217,1.682) (3.3,1.658) (3.382,1.636) (3.465,1.615) (3.547,1.595) (3.63,1.576) (3.712,1.558) (3.795,1.541) (3.877,1.524) (3.96,1.509) (4.042,1.495) (4.125,1.481) (4.207,1.468) (4.29,1.455) (4.373,1.444) (4.455,1.433) (4.538,1.422) (4.62,1.412) (4.702,1.403) (4.785,1.394) (4.867,1.386) (4.95,1.379) (5.032,1.371) (5.115,1.365) (5.197,1.358) (5.28,1.352) (5.362,1.347) (5.445,1.342) (5.527,1.337) (5.61,1.333) (5.692,1.329) (5.775,1.326) (5.858,1.322) (5.94,1.32) (6.022,1.317) (6.105,1.315) (6.187,1.313) (6.27,1.312) (6.353,1.311) (6.435,1.31) (6.518,1.309) (6.6,1.309)};
\node[ink2,right] at (6.6,2.095) {\scriptsize $s=0.25$};
\node[fwd,right] at (6.6,0.785) {\scriptsize $s=0.5$};
\node[bwd,right] at (6.6,0.262) {\scriptsize $s=0.6$};
\node[accent,right] at (6.6,1.309) {\scriptsize Crank--Nicolson, $s=2$};
\end{scope}
\end{tikzpicture}
\figcap{The factor by which one step of the explicit scheme multiplies the mode $e^{ikx}$, for three values of $s=k\,\Delta t/\Delta x^2$, and of the Crank--Nicolson scheme at $s=2$. Stability is confinement to the band $|\xi|\le1$; at $s=0.6$ the shortest waves leave it.}
\end{center}
\section{Waves}
\begin{theorem}[Courant--Friedrichs--Lewy]\label{thm:cfl}
\Scl[source, \S8.3; \cite{CFL28}] The centred scheme for the wave equation is stable if and only if
$c\,\Delta t\le\Delta x$. If $c\,\Delta t>\Delta x$ the scheme cannot converge, whatever its other properties: its
domain of dependence is strictly inside the true one.
\end{theorem}
\noindent The source writes the condition as $s=c^2\Delta t^2/\Delta x^2\le1$ and gives the same explanation. With
$c\,\Delta t=2\Delta x$ the scheme returns bit-identical values at $(0,1)$ for two data sets that differ only near
$x=1$, at every refinement, while the exact solutions differ by $1/2e$. \chk{N5} This is \Cref{thm:causal} read as a
constraint: a discretization must respect the causal structure of the solution functor, in the sense that the
discrete domains of dependence contain the continuous ones. At equality the scheme is the parallelogram identity
of \Cref{cor:par} and reproduces exact solutions to rounding error; at $s=1.1$ the sawtooth grows by the predicted
factor $1.8633$ per step. \chk{N4} A scheme that preserves the symplectic form of \Cref{prop:symp} is constructed in
\Cref{ch:fields}.
\begin{center}
\begin{tikzpicture}
\begin{scope}[xshift=0cm]
\fill[ink2] (-2,0) circle (1pt);
\fill[ink2] (-1.5,0) circle (1pt);
\fill[ink2] (-1,0) circle (1pt);
\fill[ink2] (-0.5,0) circle (1pt);
\fill[ink2] (0,0) circle (1pt);
\fill[ink2] (0.5,0) circle (1pt);
\fill[ink2] (1,0) circle (1pt);
\fill[ink2] (1.5,0) circle (1pt);
\fill[ink2] (2,0) circle (1pt);
\fill[ink2] (-2,0.6) circle (1pt);
\fill[ink2] (-1.5,0.6) circle (1pt);
\fill[ink2] (-1,0.6) circle (1pt);
\fill[ink2] (-0.5,0.6) circle (1pt);
\fill[ink2] (0,0.6) circle (1pt);
\fill[ink2] (0.5,0.6) circle (1pt);
\fill[ink2] (1,0.6) circle (1pt);
\fill[ink2] (1.5,0.6) circle (1pt);
\fill[ink2] (2,0.6) circle (1pt);
\fill[ink2] (-2,1.2) circle (1pt);
\fill[ink2] (-1.5,1.2) circle (1pt);
\fill[ink2] (-1,1.2) circle (1pt);
\fill[ink2] (-0.5,1.2) circle (1pt);
\fill[ink2] (0,1.2) circle (1pt);
\fill[ink2] (0.5,1.2) circle (1pt);
\fill[ink2] (1,1.2) circle (1pt);
\fill[ink2] (1.5,1.2) circle (1pt);
\fill[ink2] (2,1.2) circle (1pt);
\fill[ink2] (-2,1.8) circle (1pt);
\fill[ink2] (-1.5,1.8) circle (1pt);
\fill[ink2] (-1,1.8) circle (1pt);
\fill[ink2] (-0.5,1.8) circle (1pt);
\fill[ink2] (0,1.8) circle (1pt);
\fill[ink2] (0.5,1.8) circle (1pt);
\fill[ink2] (1,1.8) circle (1pt);
\fill[ink2] (1.5,1.8) circle (1pt);
\fill[ink2] (2,1.8) circle (1pt);
\fill[fwd!20,opacity=0.6] (0,1.8)--(-1.5,0)--(1.5,0)--cycle;
\draw[thick,fwd] (-1.5,0)--(0,1.8)--(1.5,0);
\draw[thick,bwd,densely dashed] (-0.9,0)--(0,1.8)--(0.9,0);
\draw[very thick,bwd] (-0.9,-0.12)--(0.9,-0.12);
\fill (0,1.8) circle (1.6pt);
\node at (0,-0.6) {\scriptsize $c\,\Delta t<\Delta x$: stable};
\end{scope}
\begin{scope}[xshift=6cm]
\fill[ink2] (-2,0) circle (1pt);
\fill[ink2] (-1.5,0) circle (1pt);
\fill[ink2] (-1,0) circle (1pt);
\fill[ink2] (-0.5,0) circle (1pt);
\fill[ink2] (0,0) circle (1pt);
\fill[ink2] (0.5,0) circle (1pt);
\fill[ink2] (1,0) circle (1pt);
\fill[ink2] (1.5,0) circle (1pt);
\fill[ink2] (2,0) circle (1pt);
\fill[ink2] (-2,0.6) circle (1pt);
\fill[ink2] (-1.5,0.6) circle (1pt);
\fill[ink2] (-1,0.6) circle (1pt);
\fill[ink2] (-0.5,0.6) circle (1pt);
\fill[ink2] (0,0.6) circle (1pt);
\fill[ink2] (0.5,0.6) circle (1pt);
\fill[ink2] (1,0.6) circle (1pt);
\fill[ink2] (1.5,0.6) circle (1pt);
\fill[ink2] (2,0.6) circle (1pt);
\fill[ink2] (-2,1.2) circle (1pt);
\fill[ink2] (-1.5,1.2) circle (1pt);
\fill[ink2] (-1,1.2) circle (1pt);
\fill[ink2] (-0.5,1.2) circle (1pt);
\fill[ink2] (0,1.2) circle (1pt);
\fill[ink2] (0.5,1.2) circle (1pt);
\fill[ink2] (1,1.2) circle (1pt);
\fill[ink2] (1.5,1.2) circle (1pt);
\fill[ink2] (2,1.2) circle (1pt);
\fill[ink2] (-2,1.8) circle (1pt);
\fill[ink2] (-1.5,1.8) circle (1pt);
\fill[ink2] (-1,1.8) circle (1pt);
\fill[ink2] (-0.5,1.8) circle (1pt);
\fill[ink2] (0,1.8) circle (1pt);
\fill[ink2] (0.5,1.8) circle (1pt);
\fill[ink2] (1,1.8) circle (1pt);
\fill[ink2] (1.5,1.8) circle (1pt);
\fill[ink2] (2,1.8) circle (1pt);
\fill[fwd!20,opacity=0.6] (0,1.8)--(-1.5,0)--(1.5,0)--cycle;
\draw[thick,fwd] (-1.5,0)--(0,1.8)--(1.5,0);
\draw[thick,bwd,densely dashed] (-2,0)--(0,1.8)--(2,0);
\draw[very thick,bwd] (-2,-0.12)--(2,-0.12);
\fill (0,1.8) circle (1.6pt);
\node at (0,-0.6) {\scriptsize $c\,\Delta t>\Delta x$: cannot converge};
\end{scope}
\end{tikzpicture}
\figcap{The Courant--Friedrichs--Lewy condition. The value at the top grid point is computed from the grid points in the shaded triangle; the solution there depends on the data in the interval cut out by the characteristics (dashed). If that interval is not inside the triangle, the scheme never sees data on which the answer depends.}
\end{center}
\section{Laplace's equation}
The five-point scheme says that each value is the mean of its four neighbours: the discrete mean value property,
with the random walk of \Cref{thm:kakutani} behind it \src{\S8.4}. The classical iterations solve it by running a
process to equilibrium.
\begin{proposition}\label{prop:iter}
\Scl[\cite{You50}] On an $N\times N$ grid the Jacobi iteration contracts the error by $\cos(\pi/N)$ per sweep, the
Gauss--Seidel iteration by $\cos^2(\pi/N)$, and successive over-relaxation with the optimal parameter
$\omega=2/(1+\sin(\pi/N))$ by $\omega-1$.
\end{proposition}
\noindent For $N=16$ a random error is reduced by $10^{-6}$ in $585$, $298$ and $41$ sweeps. \chk{N6} Jacobi iteration is
the explicit diffusion scheme at $s=\tfrac14$ run to its steady state, as the source notes; Gauss--Seidel is ``about
twice as fast'', exactly so asymptotically; and the choice of $\omega$, which the source leaves as an art, has for
this model problem the closed form above.
\section{Finite elements}
By \Cref{thm:dirichlet} the Galerkin solution in a space of trial functions is the energy projection of the true
solution \src{\S8.5}. For piecewise linear functions on a uniform mesh of the interval the stiffness matrix is the
difference Laplacian divided by $h$; the Galerkin solution is exact at the nodes, because the Green's function
$x(1-y)$ is itself piecewise linear; and the energy error halves with $h$. \chk{N7} On a uniform triangulation of
the square, linear elements give the five-point scheme. \chk{N8} The method earns its keep on domains where
differences do not fit, and there the corner exponents of \Cref{prop:wedge} limit the rate.
\section{Discretize the complex}
\Cref{prop:tonti} suggests discretizing the two factors of the Laplacian separately: the exterior derivative
exactly, on cochains of a mesh, and the Hodge star approximately.
\begin{theorem}[Finite element exterior calculus]\label{thm:feec}
\Sli[\cite{AFW06}] On a triangulated domain the Whitney forms are a finite-dimensional subcomplex of the de Rham
complex, isomorphic to simplicial cochains, and there are projections onto it that are bounded and commute with
$\dd$. Consequently the discrete Hodge Laplacian is stable and convergent in every degree, and its harmonic forms
have the dimensions of the cohomology of the domain.
\end{theorem}
\noindent Stability here is a consequence of naturality: the discretization is a map of complexes. On a
triangulated annulus with $12$ vertices, $24$ edges and $12$ triangles, $\dd\dd=0$ holds in integer arithmetic and
the Betti numbers are $(1,1,0)$. \chk{N9} The single harmonic $1$-cochain is the discrete image of the class of
$\dd\theta$, whose conjugate is $\log r$ (\Cref{prop:annulus}); a scheme that does not respect the complex can
compute a spurious number of such modes.
\section*{Notes}
The source's Chapter 8 is followed for the schemes and their stability analysis. The equivalence theorem is
\cite{LaRi56}; the stability condition for waves is \cite{CFL28}; the theory of over-relaxation is Young's
\cite{You50}. Finite element exterior calculus is due to Arnold, Falk and Winther \cite{AFW06}.
\section*{Exercises}
\begin{exercise}\label{exer:amp}
Derive the amplification factor of the explicit diffusion scheme and the condition $s\le\tfrac12$.
\end{exercise}
\begin{exercise}\label{exer:sixth}
Show that the explicit scheme has truncation error of fourth order in $\Delta x$ when $s=\tfrac16$.
\end{exercise}
\begin{exercise}\label{exer:cflroots}
For the wave scheme the amplification factors satisfy $\xi^2-2(1+p)\xi+1=0$ with $p=s(\cos k\Delta x-1)$. Show
that both roots have modulus one for all $k$ exactly when $s\le1$.
\end{exercise}
\begin{exercise}\label{exer:betti}
Compute the Euler characteristic of the triangulated annulus of the text and deduce its first Betti number.
\end{exercise}
\part{Physics and Nonlinearity}
\chapter{Fields}\label{ch:fields}
The source's chapter on physics \src{Ch.~13} is where the equations stop being scalar. Its topics are
electromagnetism, fluids, scattering and the equations of particles. Each exhibits a structure that the scalar
theory only hinted at, and one of them is the first object of this volume that is higher-categorical in earnest.
\section{Maxwell's equations}
In the source's units the equations are
\[ \pd_tE=c\,\curl B-4\pi J,\qquad\pd_tB=-c\,\curl E,\qquad\Div E=4\pi\rho,\qquad\Div B=0. \]
On space-time let $F=E\wedge dt+B$ be the field $2$-form. Up to constants and signs fixed by conventions, the
second and fourth equations are $\dd F=0$ and the first and third are $\dd\star F=\star J$: two topological equations and one metric relation, the pattern of
\Cref{prop:tonti}. Conservation of charge, $\pd_t\rho+\Div J=0$, is $\dd\dd=0$ applied to $\star F$. (The source's text
gives this equation with the opposite sign of $\Div J$; its Exercise 1 of \S13.1 has it as here.) In vacuum each
component satisfies the wave equation, and the transverse plane wave $E=(0,f(x-ct),0)$, $B=(0,0,f(x-ct))$ is a
solution. \chk{P2}
Since $\dd F=0$, locally $F=\dd A$ for a potential $A$, determined up to $A\mapsto A+\dd\chi$.
\[\begin{tikzcd}[column sep=3.6em,row sep=2.4em]
\underset{\text{gauge }\chi}{\Omega^0}\arrow[r,"\dd"] & \underset{\text{potential }A}{\Omega^1}\arrow[r,"\dd"] & \underset{\text{field }F}{\Omega^2}\arrow[r,"\dd"]\arrow[d,"\star"] & \Omega^3\\
\Omega^4 & \underset{\text{current }\star J}{\Omega^3}\arrow[l,"\dd"'] & \Omega^2\arrow[l,"\dd"'] &
\end{tikzcd}\]
The top row is the de Rham complex of space-time and is natural for every smooth map; the bottom row is the same
complex read from degree two; the one vertical arrow is the only place where the metric, and with it the speed of
light, enters.
\begin{proposition}[Fields form a groupoid]\label{prop:gauge}
\Srf Potentials on a space-time $M$ are the objects of a groupoid whose morphisms are gauge transformations. The
field strength is a functor from it to closed $2$-forms. Flat potentials modulo gauge are classified by
$H^1(M;\R)$ modulo integral classes, their holonomies; and a closed $2$-form is the field strength of a potential on
a line bundle exactly when its periods are integral.
\end{proposition}
\noindent On a cubical decomposition of the $3$-torus, $\curl\,\mathrm{grad}=0$ and $\Div\,\curl=0$ hold in integer
arithmetic, the field strength of a random potential is unchanged by a random gauge transformation, and the Betti
numbers are $(1,3,3,1)$: there are three independent closed potentials that are not gradients, the discrete
Aharonov--Bohm fields. \chk{P1}
\begin{remark}[Where the theory becomes higher]\label{rem:stack}
\Sli[\cite{Fre00,HoSi05}] Gauge-equivalent potentials are the same field, but in more than one way: a
gauge transformation is data. So the fields on $M$ are not a set but a groupoid, and the assignment
$M\mapsto\{\text{line bundles with connection on }M\}$ is not a sheaf of sets but a stack; its isomorphism classes
form the differential cohomology group $\hat H^2(M)$, an extension of integral classes by forms. For a field
whose potential is a $2$-form, gauge transformations have gauge transformations of their own and the fields form a
$2$-groupoid. The solution functor of \Cref{ch:equations} takes values in sets only for equations without gauge
symmetry; for Maxwell's equations and their relatives it takes values in higher groupoids. Nothing earlier in the
volume needed this, and nothing in the source's treatment, which works with $E$ and $B$, sees it.
\end{remark}
\section{Fluids and sound}
The Euler equations of a compressible fluid, $\pd_t\rho+\Div(\rho v)=0$ and
$\pd_tv+(v\cdot\nabla)v=-\rho^{-1}\nabla p(\rho)$, are a conservation law and a constitutive law again
\src{\S13.2}. Their linearization at a state of rest is the wave equation with $c^2=p'(\rho_0)$: in one dimension the
linearized system has characteristic speeds $\pm c$, numerically $\pm1.2272$ for $p=\rho^{1.4}$ at $\rho_0=1.2$.
\chk{P3} Linearization is functorial: the linearized equation at a solution is the tangent space of the solution
set there, and acoustics is the tangent of fluid dynamics at rest. For an incompressible fluid Arnold showed that
the motions are the geodesics of the group of volume-preserving diffeomorphisms \cite{Arn66}.
\section{Scattering}
A wave meets an obstacle and is partly reflected. For a string made of two materials with speeds $c_1,c_2$ the
reflected and transmitted amplitudes are $(c_2-c_1)/(c_2+c_1)$ and $2c_2/(c_2+c_1)$ \src{\S13.3}; the solution and its
slope are continuous at the junction and the energy fluxes balance. \chk{P4}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\fill[fwd!8] (4.364,0) rectangle (9.6,1.7);
\draw[ink2] (0,1.7)--(0,0)--(9.6,0);
\draw[ink2] (0.873,0)--(0.873,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-4$};
\draw[ink2] (2.618,0)--(2.618,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (4.364,0)--(4.364,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (6.109,0)--(6.109,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (7.855,0)--(7.855,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (9.6,0)--(9.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $6$};
\draw[ink2] (0,0.236)--(-0.07,0.236) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.181)--(-0.07,1.181) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[thick,fwd] plot coordinates {(0,0.236) (0.022,0.236) (0.044,0.236) (0.065,0.236) (0.087,0.236) (0.109,0.236) (0.131,0.236) (0.153,0.236) (0.175,0.236) (0.196,0.236) (0.218,0.236) (0.24,0.236) (0.262,0.236) (0.284,0.236) (0.305,0.236) (0.327,0.236) (0.349,0.236) (0.371,0.236) (0.393,0.236) (0.415,0.236) (0.436,0.236) (0.458,0.236) (0.48,0.236) (0.502,0.236) (0.524,0.236) (0.545,0.236) (0.567,0.236) (0.589,0.236) (0.611,0.236) (0.633,0.236) (0.655,0.236) (0.676,0.236) (0.698,0.236) (0.72,0.236) (0.742,0.236) (0.764,0.236) (0.785,0.236) (0.807,0.236) (0.829,0.236) (0.851,0.236) (0.873,0.236) (0.895,0.236) (0.916,0.236) (0.938,0.236) (0.96,0.236) (0.982,0.236) (1.004,0.236) (1.025,0.236) (1.047,0.236) (1.069,0.236) (1.091,0.236) (1.113,0.236) (1.135,0.236) (1.156,0.236) (1.178,0.236) (1.2,0.236) (1.222,0.236) (1.244,0.236) (1.265,0.236) (1.287,0.236) (1.309,0.236) (1.331,0.236) (1.353,0.236) (1.375,0.236) (1.396,0.236) (1.418,0.236) (1.44,0.236) (1.462,0.236) (1.484,0.236) (1.505,0.236) (1.527,0.236) (1.549,0.236) (1.571,0.236) (1.593,0.237) (1.615,0.237) (1.636,0.237) (1.658,0.237) (1.68,0.238) (1.702,0.238) (1.724,0.239) (1.745,0.24) (1.767,0.241) (1.789,0.242) (1.811,0.244) (1.833,0.247) (1.855,0.25) (1.876,0.253) (1.898,0.258) (1.92,0.263) (1.942,0.27) (1.964,0.278) (1.985,0.287) (2.007,0.298) (2.029,0.311) (2.051,0.326) (2.073,0.344) (2.095,0.364) (2.116,0.387) (2.138,0.412) (2.16,0.44) (2.182,0.472) (2.204,0.506) (2.225,0.543) (2.247,0.582) (2.269,0.624) (2.291,0.669) (2.313,0.714) (2.335,0.761) (2.356,0.809) (2.378,0.857) (2.4,0.904) (2.422,0.949) (2.444,0.992) (2.465,1.033) (2.487,1.07) (2.509,1.102) (2.531,1.13) (2.553,1.151) (2.575,1.168) (2.596,1.177) (2.618,1.181) (2.64,1.177) (2.662,1.168) (2.684,1.151) (2.705,1.13) (2.727,1.102) (2.749,1.07) (2.771,1.033) (2.793,0.992) (2.815,0.949) (2.836,0.904) (2.858,0.857) (2.88,0.809) (2.902,0.761) (2.924,0.714) (2.945,0.669) (2.967,0.624) (2.989,0.582) (3.011,0.543) (3.033,0.506) (3.055,0.472) (3.076,0.44) (3.098,0.412) (3.12,0.387) (3.142,0.364) (3.164,0.344) (3.185,0.326) (3.207,0.311) (3.229,0.298) (3.251,0.287) (3.273,0.278) (3.295,0.27) (3.316,0.263) (3.338,0.258) (3.36,0.253) (3.382,0.25) (3.404,0.247) (3.425,0.244) (3.447,0.242) (3.469,0.241) (3.491,0.24) (3.513,0.239) (3.535,0.238) (3.556,0.238) (3.578,0.237) (3.6,0.237) (3.622,0.237) (3.644,0.237) (3.665,0.236) (3.687,0.236) (3.709,0.236) (3.731,0.236) (3.753,0.236) (3.775,0.236) (3.796,0.236) (3.818,0.236) (3.84,0.236) (3.862,0.236) (3.884,0.236) (3.905,0.236) (3.927,0.236) (3.949,0.236) (3.971,0.236) (3.993,0.236) (4.015,0.236) (4.036,0.236) (4.058,0.236) (4.08,0.236) (4.102,0.236) (4.124,0.236) (4.145,0.236) (4.167,0.236) (4.189,0.236) (4.211,0.236) (4.233,0.236) (4.255,0.236) (4.276,0.236) (4.298,0.236) (4.32,0.236) (4.342,0.236) (4.364,0.236) (4.385,0.236) (4.407,0.236) (4.429,0.236) (4.451,0.236) (4.473,0.236) (4.495,0.236) (4.516,0.236) (4.538,0.236) (4.56,0.236) (4.582,0.236) (4.604,0.236) (4.625,0.236) (4.647,0.236) (4.669,0.236) (4.691,0.236) (4.713,0.236) (4.735,0.236) (4.756,0.236) (4.778,0.236) (4.8,0.236) (4.822,0.236) (4.844,0.236) (4.865,0.236) (4.887,0.236) (4.909,0.236) (4.931,0.236) (4.953,0.236) (4.975,0.236) (4.996,0.236) (5.018,0.236) (5.04,0.236) (5.062,0.236) (5.084,0.236) (5.105,0.236) (5.127,0.236) (5.149,0.236) (5.171,0.236) (5.193,0.236) (5.215,0.236) (5.236,0.236) (5.258,0.236) (5.28,0.236) (5.302,0.236) (5.324,0.236) (5.345,0.236) (5.367,0.236) (5.389,0.236) (5.411,0.236) (5.433,0.236) (5.455,0.236) (5.476,0.236) (5.498,0.236) (5.52,0.236) (5.542,0.236) (5.564,0.236) (5.585,0.236) (5.607,0.236) (5.629,0.236) (5.651,0.236) (5.673,0.236) (5.695,0.236) (5.716,0.236) (5.738,0.236) (5.76,0.236) (5.782,0.236) (5.804,0.236) (5.825,0.236) (5.847,0.236) (5.869,0.236) (5.891,0.236) (5.913,0.236) (5.935,0.236) (5.956,0.236) (5.978,0.236) (6,0.236) (6.022,0.236) (6.044,0.236) (6.065,0.236) (6.087,0.236) (6.109,0.236) (6.131,0.236) (6.153,0.236) (6.175,0.236) (6.196,0.236) (6.218,0.236) (6.24,0.236) (6.262,0.236) (6.284,0.236) (6.305,0.236) (6.327,0.236) (6.349,0.236) (6.371,0.236) (6.393,0.236) (6.415,0.236) (6.436,0.236) (6.458,0.236) (6.48,0.236) (6.502,0.236) (6.524,0.236) (6.545,0.236) (6.567,0.236) (6.589,0.236) (6.611,0.236) (6.633,0.236) (6.655,0.236) (6.676,0.236) (6.698,0.236) (6.72,0.236) (6.742,0.236) (6.764,0.236) (6.785,0.236) (6.807,0.236) (6.829,0.236) (6.851,0.236) (6.873,0.236) (6.895,0.236) (6.916,0.236) (6.938,0.236) (6.96,0.236) (6.982,0.236) (7.004,0.236) (7.025,0.236) (7.047,0.236) (7.069,0.236) (7.091,0.236) (7.113,0.236) (7.135,0.236) (7.156,0.236) (7.178,0.236) (7.2,0.236) (7.222,0.236) (7.244,0.236) (7.265,0.236) (7.287,0.236) (7.309,0.236) (7.331,0.236) (7.353,0.236) (7.375,0.236) (7.396,0.236) (7.418,0.236) (7.44,0.236) (7.462,0.236) (7.484,0.236) (7.505,0.236) (7.527,0.236) (7.549,0.236) (7.571,0.236) (7.593,0.236) (7.615,0.236) (7.636,0.236) (7.658,0.236) (7.68,0.236) (7.702,0.236) (7.724,0.236) (7.745,0.236) (7.767,0.236) (7.789,0.236) (7.811,0.236) (7.833,0.236) (7.855,0.236) (7.876,0.236) (7.898,0.236) (7.92,0.236) (7.942,0.236) (7.964,0.236) (7.985,0.236) (8.007,0.236) (8.029,0.236) (8.051,0.236) (8.073,0.236) (8.095,0.236) (8.116,0.236) (8.138,0.236) (8.16,0.236) (8.182,0.236) (8.204,0.236) (8.225,0.236) (8.247,0.236) (8.269,0.236) (8.291,0.236) (8.313,0.236) (8.335,0.236) (8.356,0.236) (8.378,0.236) (8.4,0.236) (8.422,0.236) (8.444,0.236) (8.465,0.236) (8.487,0.236) (8.509,0.236) (8.531,0.236) (8.553,0.236) (8.575,0.236) (8.596,0.236) (8.618,0.236) (8.64,0.236) (8.662,0.236) (8.684,0.236) (8.705,0.236) (8.727,0.236) (8.749,0.236) (8.771,0.236) (8.793,0.236) (8.815,0.236) (8.836,0.236) (8.858,0.236) (8.88,0.236) (8.902,0.236) (8.924,0.236) (8.945,0.236) (8.967,0.236) (8.989,0.236) (9.011,0.236) (9.033,0.236) (9.055,0.236) (9.076,0.236) (9.098,0.236) (9.12,0.236) (9.142,0.236) (9.164,0.236) (9.185,0.236) (9.207,0.236) (9.229,0.236) (9.251,0.236) (9.273,0.236) (9.295,0.236) (9.316,0.236) (9.338,0.236) (9.36,0.236) (9.382,0.236) (9.404,0.236) (9.425,0.236) (9.447,0.236) (9.469,0.236) (9.491,0.236) (9.513,0.236) (9.535,0.236) (9.556,0.236) (9.578,0.236) (9.6,0.236)};
\node[right] at (0.087,1.464) {\scriptsize before};
\node[above] at (2.618,1.417) {\scriptsize incident, $1$};
\node[] at (6.982,1.417) {\scriptsize $c_2=2c_1$};
\end{scope}
\begin{scope}[shift={(0,-2.45)}]
\fill[fwd!8] (4.364,0) rectangle (9.6,1.7);
\draw[ink2] (0,1.7)--(0,0)--(9.6,0);
\draw[ink2] (0.873,0)--(0.873,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-4$};
\draw[ink2] (2.618,0)--(2.618,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (4.364,0)--(4.364,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (6.109,0)--(6.109,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (7.855,0)--(7.855,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (9.6,0)--(9.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $6$};
\draw[ink2] (0,0.236)--(-0.07,0.236) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.181)--(-0.07,1.181) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[thick,fwd] plot coordinates {(0,0.236) (0.022,0.236) (0.044,0.236) (0.065,0.236) (0.087,0.236) (0.109,0.236) (0.131,0.236) (0.153,0.236) (0.175,0.236) (0.196,0.236) (0.218,0.236) (0.24,0.236) (0.262,0.236) (0.284,0.236) (0.305,0.236) (0.327,0.236) (0.349,0.236) (0.371,0.236) (0.393,0.236) (0.415,0.236) (0.436,0.236) (0.458,0.236) (0.48,0.236) (0.502,0.236) (0.524,0.236) (0.545,0.236) (0.567,0.236) (0.589,0.236) (0.611,0.236) (0.633,0.236) (0.655,0.236) (0.676,0.236) (0.698,0.236) (0.72,0.236) (0.742,0.236) (0.764,0.236) (0.785,0.236) (0.807,0.236) (0.829,0.236) (0.851,0.236) (0.873,0.236) (0.895,0.236) (0.916,0.236) (0.938,0.236) (0.96,0.236) (0.982,0.236) (1.004,0.236) (1.025,0.236) (1.047,0.236) (1.069,0.236) (1.091,0.236) (1.113,0.236) (1.135,0.236) (1.156,0.236) (1.178,0.236) (1.2,0.236) (1.222,0.236) (1.244,0.236) (1.265,0.236) (1.287,0.236) (1.309,0.236) (1.331,0.236) (1.353,0.236) (1.375,0.236) (1.396,0.236) (1.418,0.236) (1.44,0.236) (1.462,0.236) (1.484,0.236) (1.505,0.236) (1.527,0.236) (1.549,0.236) (1.571,0.236) (1.593,0.236) (1.615,0.236) (1.636,0.236) (1.658,0.236) (1.68,0.237) (1.702,0.237) (1.724,0.237) (1.745,0.237) (1.767,0.238) (1.789,0.238) (1.811,0.239) (1.833,0.24) (1.855,0.241) (1.876,0.242) (1.898,0.243) (1.92,0.245) (1.942,0.247) (1.964,0.25) (1.985,0.253) (2.007,0.257) (2.029,0.261) (2.051,0.266) (2.073,0.272) (2.095,0.279) (2.116,0.286) (2.138,0.295) (2.16,0.304) (2.182,0.315) (2.204,0.326) (2.225,0.338) (2.247,0.352) (2.269,0.366) (2.291,0.38) (2.313,0.396) (2.335,0.411) (2.356,0.427) (2.378,0.443) (2.4,0.459) (2.422,0.474) (2.444,0.488) (2.465,0.502) (2.487,0.514) (2.509,0.525) (2.531,0.534) (2.553,0.541) (2.575,0.547) (2.596,0.55) (2.618,0.551) (2.64,0.55) (2.662,0.547) (2.684,0.541) (2.705,0.534) (2.727,0.525) (2.749,0.514) (2.771,0.502) (2.793,0.488) (2.815,0.474) (2.836,0.459) (2.858,0.443) (2.88,0.427) (2.902,0.411) (2.924,0.396) (2.945,0.38) (2.967,0.366) (2.989,0.352) (3.011,0.338) (3.033,0.326) (3.055,0.315) (3.076,0.304) (3.098,0.295) (3.12,0.286) (3.142,0.279) (3.164,0.272) (3.185,0.266) (3.207,0.261) (3.229,0.257) (3.251,0.253) (3.273,0.25) (3.295,0.247) (3.316,0.245) (3.338,0.243) (3.36,0.242) (3.382,0.241) (3.404,0.24) (3.425,0.239) (3.447,0.238) (3.469,0.238) (3.491,0.237) (3.513,0.237) (3.535,0.237) (3.556,0.237) (3.578,0.236) (3.6,0.236) (3.622,0.236) (3.644,0.236) (3.665,0.236) (3.687,0.236) (3.709,0.236) (3.731,0.236) (3.753,0.236) (3.775,0.236) (3.796,0.236) (3.818,0.236) (3.84,0.236) (3.862,0.236) (3.884,0.236) (3.905,0.236) (3.927,0.236) (3.949,0.236) (3.971,0.236) (3.993,0.236) (4.015,0.236) (4.036,0.236) (4.058,0.236) (4.08,0.236) (4.102,0.236) (4.124,0.236) (4.145,0.236) (4.167,0.236) (4.189,0.236) (4.211,0.236) (4.233,0.236) (4.255,0.236) (4.276,0.236) (4.298,0.236) (4.32,0.236) (4.342,0.236) (4.364,0.236) (4.385,0.236) (4.407,0.236) (4.429,0.236) (4.451,0.236) (4.473,0.236) (4.495,0.236) (4.516,0.236) (4.538,0.236) (4.56,0.236) (4.582,0.236) (4.604,0.236) (4.625,0.236) (4.647,0.236) (4.669,0.236) (4.691,0.236) (4.713,0.236) (4.735,0.236) (4.756,0.236) (4.778,0.236) (4.8,0.236) (4.822,0.236) (4.844,0.236) (4.865,0.236) (4.887,0.236) (4.909,0.236) (4.931,0.236) (4.953,0.236) (4.975,0.236) (4.996,0.236) (5.018,0.236) (5.04,0.236) (5.062,0.236) (5.084,0.236) (5.105,0.236) (5.127,0.236) (5.149,0.236) (5.171,0.236) (5.193,0.236) (5.215,0.236) (5.236,0.236) (5.258,0.236) (5.28,0.236) (5.302,0.236) (5.324,0.236) (5.345,0.236) (5.367,0.236) (5.389,0.236) (5.411,0.236) (5.433,0.236) (5.455,0.236) (5.476,0.236) (5.498,0.236) (5.52,0.236) (5.542,0.236) (5.564,0.236) (5.585,0.236) (5.607,0.236) (5.629,0.236) (5.651,0.236) (5.673,0.236) (5.695,0.236) (5.716,0.236) (5.738,0.236) (5.76,0.237) (5.782,0.237) (5.804,0.237) (5.825,0.237) (5.847,0.237) (5.869,0.237) (5.891,0.237) (5.913,0.237) (5.935,0.238) (5.956,0.238) (5.978,0.238) (6,0.238) (6.022,0.239) (6.044,0.239) (6.065,0.24) (6.087,0.24) (6.109,0.241) (6.131,0.242) (6.153,0.243) (6.175,0.243) (6.196,0.244) (6.218,0.246) (6.24,0.247) (6.262,0.248) (6.284,0.25) (6.305,0.252) (6.327,0.254) (6.349,0.256) (6.371,0.259) (6.393,0.262) (6.415,0.265) (6.436,0.268) (6.458,0.272) (6.48,0.276) (6.502,0.281) (6.524,0.286) (6.545,0.291) (6.567,0.297) (6.589,0.304) (6.611,0.311) (6.633,0.319) (6.655,0.327) (6.676,0.336) (6.698,0.346) (6.72,0.357) (6.742,0.368) (6.764,0.38) (6.785,0.393) (6.807,0.407) (6.829,0.421) (6.851,0.437) (6.873,0.453) (6.895,0.471) (6.916,0.489) (6.938,0.508) (6.96,0.529) (6.982,0.55) (7.004,0.572) (7.025,0.596) (7.047,0.62) (7.069,0.645) (7.091,0.671) (7.113,0.698) (7.135,0.725) (7.156,0.754) (7.178,0.783) (7.2,0.813) (7.222,0.843) (7.244,0.874) (7.265,0.905) (7.287,0.936) (7.309,0.968) (7.331,1) (7.353,1.032) (7.375,1.063) (7.396,1.095) (7.418,1.126) (7.44,1.157) (7.462,1.187) (7.484,1.216) (7.505,1.244) (7.527,1.272) (7.549,1.298) (7.571,1.324) (7.593,1.347) (7.615,1.37) (7.636,1.391) (7.658,1.41) (7.68,1.427) (7.702,1.443) (7.724,1.457) (7.745,1.468) (7.767,1.478) (7.789,1.486) (7.811,1.491) (7.833,1.494) (7.855,1.495) (7.876,1.494) (7.898,1.491) (7.92,1.486) (7.942,1.478) (7.964,1.468) (7.985,1.457) (8.007,1.443) (8.029,1.427) (8.051,1.41) (8.073,1.391) (8.095,1.37) (8.116,1.347) (8.138,1.324) (8.16,1.298) (8.182,1.272) (8.204,1.244) (8.225,1.216) (8.247,1.187) (8.269,1.157) (8.291,1.126) (8.313,1.095) (8.335,1.063) (8.356,1.032) (8.378,1) (8.4,0.968) (8.422,0.936) (8.444,0.905) (8.465,0.874) (8.487,0.843) (8.509,0.813) (8.531,0.783) (8.553,0.754) (8.575,0.725) (8.596,0.698) (8.618,0.671) (8.64,0.645) (8.662,0.62) (8.684,0.596) (8.705,0.572) (8.727,0.55) (8.749,0.529) (8.771,0.508) (8.793,0.489) (8.815,0.471) (8.836,0.453) (8.858,0.437) (8.88,0.421) (8.902,0.407) (8.924,0.393) (8.945,0.38) (8.967,0.368) (8.989,0.357) (9.011,0.346) (9.033,0.336) (9.055,0.327) (9.076,0.319) (9.098,0.311) (9.12,0.304) (9.142,0.297) (9.164,0.291) (9.185,0.286) (9.207,0.281) (9.229,0.276) (9.251,0.272) (9.273,0.268) (9.295,0.265) (9.316,0.262) (9.338,0.259) (9.36,0.256) (9.382,0.254) (9.404,0.252) (9.425,0.25) (9.447,0.248) (9.469,0.247) (9.491,0.246) (9.513,0.244) (9.535,0.243) (9.556,0.243) (9.578,0.242) (9.6,0.241)};
\node[right] at (0.087,1.464) {\scriptsize after};
\node[above] at (2.618,0.661) {\scriptsize reflected, $\tfrac13$};
\node[] at (7.855,1.53) {\scriptsize };
\node[left] at (6.284,1.369) {\scriptsize transmitted, $\tfrac43$};
\end{scope}
\end{tikzpicture}
\figcap{A pulse on a string whose speed doubles at $x=0$ (shaded side). The reflected pulse has amplitude $(c_2-c_1)/(c_2+c_1)=\frac13$ and the transmitted one $2c_2/(c_2+c_1)=\frac43$, stretched by the ratio of speeds. The displacement and its slope are continuous at the junction throughout.}
\end{center}
For the equation $-\psi''+V\psi=k^2\psi$ with $V$ of compact support, let $M_V(k)$ be the matrix taking
$(\psi,\psi')$ on the left of the support to $(\psi,\psi')$ on the right.
\begin{theorem}[Transfer is a functor]\label{thm:transfer}
\Srf[source, \S13.4, Theorem 1] $M_V(k)$ has determinant one, and if $V$ is $V_1$ followed by $V_2$ then
$M_V=M_{V_2}M_{V_1}$. The reflection and transmission coefficients computed from $M_V$ satisfy $|R|^2+|T|^2=1$.
\end{theorem}
\begin{proof}
The Wronskian of two solutions is constant, which is the first statement, and it is the symplectic form of
\Cref{thm:lagrangian} on boundary values. Composition is uniqueness of solutions of the initial value problem. The
identity is the constancy of the Wronskian of $\psi$ and $\bar\psi$, the source's argument.
\end{proof}
\noindent For a potential of four constant pieces the product of the four matrices agreed with a direct numerical
integration across the whole to eight digits, and the identity held to twelve, at three energies. \chk{P5} This is
\Cref{rem:compose} with the eigenvalue parameter in the continuous spectrum: an interval with a potential is a
symplectic map between boundary values, and concatenation is composition.
\[\begin{tikzcd}[column sep=5em]
V_{x_0}\arrow[r,"M_{V_1}(k)"]\arrow[rr,bend right=20,"M_V(k)"'] & V_{x_1}\arrow[r,"M_{V_2}(k)"] & V_{x_2}
\end{tikzcd}\]
When $V\to0$ at infinity the operator $-d^2/dx^2+V$ has, besides finitely many eigenvalues, the \term{continuous
spectrum} $[0,\infty)$, with generalized eigenfunctions that are bounded and not square-integrable \src{\S13.4}. The
spectral theorem of \Cref{thm:spectral} then requires the distributions of \Cref{ch:distributions}: the
eigenfunction expansion is an integral, and for $V=0$ it is the Fourier transform.
\section{Equations of particles}
The Klein--Gordon equation $u_{tt}-c^2\Lap u+m^2u=0$ has plane waves with $\omega^2=c^2k^2+m^2$; its group velocity
is below $c$ and its phase velocity above, with product $c^2$. \chk{P6}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3.1)--(0,0)--(4.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (1.1,0)--(1.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (2.2,0)--(2.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (3.3,0)--(3.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $3$};
\draw[ink2] (4.4,0)--(4.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,0.721)--(-0.07,0.721) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,1.442)--(-0.07,1.442) node[left,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,2.163)--(-0.07,2.163) node[left,inner sep=1.5pt]{\scriptsize $3$};
\draw[ink2] (0,2.884)--(-0.07,2.884) node[left,inner sep=1.5pt]{\scriptsize $4$};
\node[below] at (2.2,-0.38) {\scriptsize $k$};
\node[above,inner sep=1pt] at (0,3.15) {\scriptsize $\omega$};
\draw[thick,fwd] plot coordinates {(0,0.721) (0.055,0.722) (0.11,0.725) (0.165,0.729) (0.22,0.735) (0.275,0.743) (0.33,0.753) (0.385,0.764) (0.44,0.776) (0.495,0.791) (0.55,0.806) (0.605,0.823) (0.66,0.841) (0.715,0.86) (0.77,0.88) (0.825,0.901) (0.88,0.923) (0.935,0.946) (0.99,0.97) (1.045,0.994) (1.1,1.02) (1.155,1.045) (1.21,1.072) (1.265,1.099) (1.32,1.126) (1.375,1.154) (1.43,1.182) (1.485,1.211) (1.54,1.24) (1.595,1.27) (1.65,1.3) (1.705,1.33) (1.76,1.36) (1.815,1.391) (1.87,1.422) (1.925,1.453) (1.98,1.484) (2.035,1.516) (2.09,1.548) (2.145,1.58) (2.2,1.612) (2.255,1.644) (2.31,1.677) (2.365,1.709) (2.42,1.742) (2.475,1.775) (2.53,1.808) (2.585,1.841) (2.64,1.874) (2.695,1.908) (2.75,1.941) (2.805,1.975) (2.86,2.008) (2.915,2.042) (2.97,2.076) (3.025,2.11) (3.08,2.143) (3.135,2.177) (3.19,2.212) (3.245,2.246) (3.3,2.28) (3.355,2.314) (3.41,2.348) (3.465,2.383) (3.52,2.417) (3.575,2.451) (3.63,2.486) (3.685,2.52) (3.74,2.555) (3.795,2.59) (3.85,2.624) (3.905,2.659) (3.96,2.694) (4.015,2.728) (4.07,2.763) (4.125,2.798) (4.18,2.833) (4.235,2.868) (4.29,2.903) (4.345,2.938) (4.4,2.972)};
\draw[ink2,densely dashed] plot coordinates {(0,0) (2.2,1.442) (4.4,2.884)};
\node[fwd] at (1.32,1.586) {\scriptsize $\omega^2=k^2+1$};
\node[ink2] at (3.52,1.802) {\scriptsize $\omega=k$};
\end{scope}
\begin{scope}[shift={(6.2,0)}]
\draw[ink2] (0,3.1)--(0,0)--(4.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (1.1,0)--(1.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (2.2,0)--(2.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (3.3,0)--(3.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $3$};
\draw[ink2] (4.4,0)--(4.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.033)--(-0.07,1.033) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,2.067)--(-0.07,2.067) node[left,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,3.1)--(-0.07,3.1) node[left,inner sep=1.5pt]{\scriptsize $3$};
\node[below] at (2.2,-0.38) {\scriptsize $k$};
\draw[thick,fwd] plot coordinates {(0,0) (0.055,0.052) (0.11,0.103) (0.165,0.153) (0.22,0.203) (0.275,0.251) (0.33,0.297) (0.385,0.341) (0.44,0.384) (0.495,0.424) (0.55,0.462) (0.605,0.498) (0.66,0.532) (0.715,0.563) (0.77,0.593) (0.825,0.62) (0.88,0.646) (0.935,0.669) (0.99,0.691) (1.045,0.712) (1.1,0.731) (1.155,0.748) (1.21,0.765) (1.265,0.78) (1.32,0.794) (1.375,0.807) (1.43,0.819) (1.485,0.83) (1.54,0.841) (1.595,0.851) (1.65,0.86) (1.705,0.868) (1.76,0.876) (1.815,0.884) (1.87,0.891) (1.925,0.897) (1.98,0.903) (2.035,0.909) (2.09,0.914) (2.145,0.919) (2.2,0.924) (2.255,0.929) (2.31,0.933) (2.365,0.937) (2.42,0.941) (2.475,0.944) (2.53,0.948) (2.585,0.951) (2.64,0.954) (2.695,0.957) (2.75,0.959) (2.805,0.962) (2.86,0.964) (2.915,0.967) (2.97,0.969) (3.025,0.971) (3.08,0.973) (3.135,0.975) (3.19,0.977) (3.245,0.979) (3.3,0.98) (3.355,0.982) (3.41,0.983) (3.465,0.985) (3.52,0.986) (3.575,0.988) (3.63,0.989) (3.685,0.99) (3.74,0.991) (3.795,0.992) (3.85,0.994) (3.905,0.995) (3.96,0.996) (4.015,0.997) (4.07,0.998) (4.125,0.998) (4.18,0.999) (4.235,1) (4.29,1.001) (4.345,1.002) (4.4,1.002)};
\draw[thick,bwd] plot coordinates {(0.408,2.974) (0.441,2.776) (0.475,2.608) (0.508,2.464) (0.542,2.339) (0.575,2.23) (0.609,2.134) (0.642,2.049) (0.676,1.974) (0.71,1.906) (0.743,1.846) (0.777,1.792) (0.81,1.742) (0.844,1.698) (0.877,1.657) (0.911,1.62) (0.944,1.586) (0.978,1.555) (1.011,1.527) (1.045,1.5) (1.079,1.476) (1.112,1.453) (1.146,1.433) (1.179,1.413) (1.213,1.395) (1.246,1.378) (1.28,1.363) (1.313,1.348) (1.347,1.334) (1.381,1.321) (1.414,1.309) (1.448,1.298) (1.481,1.287) (1.515,1.277) (1.548,1.268) (1.582,1.259) (1.615,1.25) (1.649,1.242) (1.682,1.235) (1.716,1.227) (1.75,1.221) (1.783,1.214) (1.817,1.208) (1.85,1.202) (1.884,1.197) (1.917,1.191) (1.951,1.186) (1.984,1.181) (2.018,1.177) (2.052,1.173) (2.085,1.168) (2.119,1.164) (2.152,1.16) (2.186,1.157) (2.219,1.153) (2.253,1.15) (2.286,1.147) (2.32,1.144) (2.353,1.141) (2.387,1.138) (2.421,1.135) (2.454,1.132) (2.488,1.13) (2.521,1.127) (2.555,1.125) (2.588,1.123) (2.622,1.121) (2.655,1.118) (2.689,1.116) (2.722,1.114) (2.756,1.113) (2.79,1.111) (2.823,1.109) (2.857,1.107) (2.89,1.106) (2.924,1.104) (2.957,1.102) (2.991,1.101) (3.024,1.1) (3.058,1.098) (3.092,1.097) (3.125,1.095) (3.159,1.094) (3.192,1.093) (3.226,1.092) (3.259,1.091) (3.293,1.089) (3.326,1.088) (3.36,1.087) (3.394,1.086) (3.427,1.085) (3.461,1.084) (3.494,1.083) (3.528,1.082) (3.561,1.082) (3.595,1.081) (3.628,1.08) (3.662,1.079) (3.695,1.078) (3.729,1.077) (3.763,1.077) (3.796,1.076) (3.83,1.075) (3.863,1.074) (3.897,1.074) (3.93,1.073) (3.964,1.072) (3.997,1.072) (4.031,1.071) (4.064,1.071) (4.098,1.07) (4.132,1.069) (4.165,1.069) (4.199,1.068) (4.232,1.068) (4.266,1.067) (4.299,1.067) (4.333,1.066) (4.366,1.066) (4.4,1.065)};
\draw[ink2,densely dashed] (0,1.033)--(4.4,1.033);
\node[fwd] at (2.86,0.641) {\scriptsize group velocity};
\node[bwd] at (2.09,1.808) {\scriptsize phase velocity};
\end{scope}
\end{tikzpicture}
\figcap{The Klein--Gordon equation with $c=m=1$. Left: the dispersion relation, a hyperbola asymptotic to the light cone. Right: the group velocity $k/\omega$ and the phase velocity $\omega/k$; their product is $1$ at every wave number.}
\end{center}
\begin{proposition}[Dirac]\label{prop:dirac}
\Scl[source, \S13.5] The source's four matrices satisfy
$\gamma^\mu\gamma^\nu+\gamma^\nu\gamma^\mu=2\eta^{\mu\nu}$ with $\eta=\mathrm{diag}(1,-1,-1,-1)$. Hence
$(\gamma^\mu\pd_\mu)^2=\pd_t^2-\Lap$, and every component of a solution of the Dirac equation solves the Klein--Gordon
equation.
\end{proposition}
\noindent \chk{P7} The wave operator has no first-order square root with scalar coefficients, since its symbol is an
indefinite quadratic form. It has one with coefficients in the Clifford algebra of that form, which is the
universal algebra in which the form becomes a square. This is \Cref{prop:factor} for the hyperbolic operator in
more than one space dimension.
\section{Fields as functors}
The four structures of \Cref{ch:waves} are available for the Klein--Gordon equation on any globally hyperbolic
space-time.
\begin{theorem}[Locally covariant fields]\label{thm:lcqft}
\Sli[\cite{BFV03,BGP07}] The assignment $M\mapsto(\Sol_{sc}(M),\Omega)$ is a functor from globally hyperbolic
space-times and causal isometric embeddings to symplectic vector spaces. Solutions on causally separated regions
are $\Omega$-orthogonal, and an embedding whose image contains a Cauchy surface induces an isomorphism.
Composing with the functor from symplectic spaces to algebras of canonical commutation relations gives a quantum
field theory.
\end{theorem}
\noindent So the free quantum field is the solution functor of a hyperbolic equation, followed by quantization: the
exact sequence of \Cref{thm:bgp}, the causality of \Cref{thm:causal} and the symplectic form of \Cref{prop:symp} are
its axioms. A numerical scheme can respect this structure: the leapfrog step for a discretized Klein--Gordon
equation satisfies $M^{\T}JM=J$ to rounding error, and the Euler step does not. \chk{P8} In the formalism of
factorization algebras the same data are organized by open sets of space-time, and the solution space is
replaced by its derived version, a complex whose zeroth cohomology is $\Sol$ \cite{CoGw17}.
\section*{Notes}
The source's Chapter 13 is followed for the equations themselves. For differential forms in electromagnetism see
\cite{Bos98}; for differential cohomology and gauge fields \cite{Fre00,HoSi05}; for the geometry of the Euler
equations \cite{Arn66}; for scattering theory \cite{ReSi75}; for locally covariant field theory
\cite{BFV03,BGP07} and factorization algebras \cite{CoGw17}.
\section*{Exercises}
\begin{exercise}\label{exer:charge}
Derive the conservation of charge from the source's form of Maxwell's equations.
\end{exercise}
\begin{exercise}\label{exer:string}
Verify the energy balance $R^2+(c_1/c_2)T^2=1$ for the string of two materials.
\end{exercise}
\begin{exercise}\label{exer:delta}
Find the transfer matrix of the potential $V=g\,\delta(x)$ and the transmission probability $|T|^2$.
\end{exercise}
\begin{exercise}\label{exer:kg}
Show that the group and phase velocities of the Klein--Gordon equation have product $c^2$.
\end{exercise}
\chapter{Nonlinear Equations}\label{ch:nonlinear}
The source ends with a sample of nonlinear phenomena: shocks, solitons, variational problems, bifurcation and
water waves \src{Ch.~14}. Linear structure is gone, and with it most of the functors of the previous chapters.
What remains, and what explains the two solvable examples, are morphisms of equations in the sense of
\Cref{thm:jet}.
\section{Shocks}
For $u_t+a(u)u_x=0$ the solution is constant along characteristics, which are straight lines of slope determined
by the data, as in \Cref{prop:char} \src{\S14.1}. When they cross the quotient is no longer a manifold and the
smooth solution ends.
\begin{proposition}\label{prop:break}
\Scl For $u_t+uu_x=0$ with data $\varphi$, the smooth solution exists exactly until $t_b=-1/\min\varphi'$.
\end{proposition}
\noindent For $\varphi=e^{-x^2}$, $t_b=\sqrt{e/2}=1.1658$. \chk{X1} The source's explicit solution for $\varphi=x^2$ is
verified, with its domain $1+4tx\ge0$. \chk{X2}
\begin{center}
\begin{tikzpicture}[xscale=1.25,yscale=1.25]
\draw[->,ink2] (-2.6,0)--(4.2,0) node[right]{\small$x$};
\draw[->,ink2] (-2.4,0)--(-2.4,2.3) node[above]{\small$t$};
\foreach \x in {-2.25,-2,...,2.76}{
\pgfmathsetmacro{\s}{exp(-(\x)*(\x))}
\draw[fwd] (\x,0)--({\x+2*\s},2);}
\draw[densely dashed,bwd] (-2.4,1.166)--(4.1,1.166) node[above left]{\small$t_b$};
\end{tikzpicture}
\figcap{Characteristics of $u_t+uu_x=0$ for the data $e^{-x^2}$: straight lines $x=x_0+\varphi(x_0)t$. They first
meet at $t_b=1.166$; above that level the map from $(x_0,t)$ to $(x,t)$ is no longer injective.}
\end{center}
After $t_b$ one asks for weak solutions. Distributions are not available, since $u^2$ must be formed
(\Cref{ch:distributions}); instead the equation is written as a conservation law $u_t+A(u)_x=0$ and tested
against smooth functions. A piecewise smooth weak solution has jumps moving with the Rankine--Hugoniot speed
$s=(A(u^+)-A(u^-))/(u^+-u^-)$.
\begin{proposition}[Weak solutions belong to the conservation form]\label{prop:consform}
\Scl The equations $u_t+(\tfrac12u^2)_x=0$ and $(u^2)_t+(\tfrac23u^3)_x=0$ have the same smooth solutions and
different weak ones: a jump from $1$ to $0$ moves at speed $\tfrac12$ under the first and $\tfrac23$ under the
second.
\end{proposition}
\noindent An isomorphism of equations on smooth solutions need not extend to weak solutions; the weak solution
functor is defined on balance laws, in the sense of \Cref{prop:tonti}, and not on the ideals they generate. The
same distinction decides numerical methods: for the step from $1$ to $0$ a scheme in conservation form places the
shock at $1.005$ after two time units, against the exact $1$, and the equally consistent non-conservative scheme
leaves it at the origin. \chk{X3}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,2.7)--(0,0)--(4.6,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (2.3,0)--(2.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (4.6,0)--(4.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.35)--(-0.07,1.35) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,2.7)--(-0.07,2.7) node[left,inner sep=1.5pt]{\scriptsize $2$};
\node[below] at (2.3,-0.38) {\scriptsize $x$};
\node[above,inner sep=1pt] at (0,2.75) {\scriptsize $t$};
\draw[fwd] (0.287,0)--(2.587,2.7);
\draw[fwd] (0.575,0)--(2.875,2.7);
\draw[fwd] (0.862,0)--(3.162,2.7);
\draw[fwd] (1.15,0)--(3.45,2.7);
\draw[fwd] (1.438,0)--(3.162,2.025);
\draw[fwd] (1.725,0)--(2.875,1.35);
\draw[fwd] (2.012,0)--(2.587,0.675);
\draw[fwd] (2.587,0)--(2.587,0.675);
\draw[fwd] (2.875,0)--(2.875,1.35);
\draw[fwd] (3.162,0)--(3.162,2.025);
\draw[fwd] (3.45,0)--(3.45,2.7);
\draw[fwd] (3.737,0)--(3.737,2.7);
\draw[fwd] (4.025,0)--(4.025,2.7);
\draw[fwd] (4.312,0)--(4.312,2.7);
\draw[very thick,bwd] (2.3,0)--(3.45,2.7);
\node[] at (2.3,3.038) {\scriptsize shock: $u=1$ behind, $0$ ahead};
\end{scope}
\begin{scope}[shift={(6.4,0)}]
\draw[ink2] (0,2.7)--(0,0)--(4.6,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-2$};
\draw[ink2] (2.3,0)--(2.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (4.6,0)--(4.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $2$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.35)--(-0.07,1.35) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,2.7)--(-0.07,2.7) node[left,inner sep=1.5pt]{\scriptsize $2$};
\node[below] at (2.3,-0.38) {\scriptsize $x$};
\node[above,inner sep=1pt] at (0,2.75) {\scriptsize $t$};
\draw[fwd] (0.287,0)--(0.287,2.7);
\draw[fwd] (0.575,0)--(0.575,2.7);
\draw[fwd] (0.862,0)--(0.862,2.7);
\draw[fwd] (1.15,0)--(1.15,2.7);
\draw[fwd] (1.438,0)--(1.438,2.7);
\draw[fwd] (1.725,0)--(1.725,2.7);
\draw[fwd] (2.012,0)--(2.012,2.7);
\draw[fwd] (2.3,0)--(2.3,2.7);
\draw[fwd] (2.3,0)--(4.6,2.7);
\draw[fwd] (2.587,0)--(4.6,2.363);
\draw[fwd] (2.875,0)--(4.6,2.025);
\draw[fwd] (3.162,0)--(4.6,1.688);
\draw[fwd] (3.45,0)--(4.6,1.35);
\draw[fwd] (3.737,0)--(4.6,1.013);
\draw[fwd] (4.025,0)--(4.6,0.675);
\draw[fwd] (4.312,0)--(4.6,0.338);
\draw[bwd] (2.3,0)--(2.76,2.7);
\draw[bwd] (2.3,0)--(3.22,2.7);
\draw[bwd] (2.3,0)--(3.68,2.7);
\draw[bwd] (2.3,0)--(4.14,2.7);
\node[] at (2.3,3.038) {\scriptsize fan: $u=0$ behind, $1$ ahead};
\end{scope}
\end{tikzpicture}
\figcap{The two Riemann problems for $u_t+uu_x=0$. Left: characteristics run into the shock, which moves at the Rankine--Hugoniot speed $\frac12$. Right: characteristics leave a gap, which the entropy solution fills with the fan $u=x/t$; a shock placed there would have characteristics coming out of it and is the solution that vanishing viscosity rejects.}
\end{center}
Weak solutions are not unique: the step from $0$ to $1$ admits both a stationary-looking shock and the fan $x/t$.
The \term{entropy condition} $a(u^-)>s>a(u^+)$ selects the fan, as does every conservative monotone scheme.
\chk{X3}
\begin{theorem}[Vanishing viscosity]\label{thm:viscosity}
\Scl[\cite{Hop50,Ole57,Kru70}] As $\eps\to0$ the solutions of $u_t+uu_x=\eps u_{xx}$ converge to a weak solution of
the inviscid equation satisfying the entropy condition, and that solution is unique.
\end{theorem}
\section{A morphism of equations}
\begin{theorem}[Cole--Hopf]\label{thm:colehopf}
\Scl[\cite{Hop50,Col51}] If $\varphi>0$ solves the heat equation $\varphi_t=\eps\varphi_{xx}$, then
$u=-2\eps\varphi_x/\varphi$ solves $u_t+uu_x=\eps u_{xx}$, and every solution arises so, from $\varphi$ unique up to a
constant factor.
\end{theorem}
\noindent In the category $\PDE$ this is a morphism from the heat equation to the viscous Burgers equation, the
quotient by scaling. For three superposed exponential solutions of the heat equation the residual of the image is
$10^{-12}$. The image of $1+e^{-(x-t/2)/2\eps}$ is the viscous profile $(1+e^{(x-t/2)/2\eps})^{-1}$, which travels at
the Rankine--Hugoniot speed $\tfrac12$ and steepens to the shock as $\eps\to0$. \chk{X4} A nonlinear equation is
solved here by exhibiting it as the image of a linear one; the theorem above is then a statement about the heat
semigroup.
\[\begin{tikzcd}[column sep=7em,row sep=large]
\varphi(\cdot,0)\arrow[r,"\text{heat semigroup}"]\arrow[d,"-2\eps\,\pd_x\log"'] & \varphi(\cdot,t)\arrow[d,"-2\eps\,\pd_x\log"]\\
u(\cdot,0)\arrow[r,"\text{Burgers flow}"'] & u(\cdot,t)
\end{tikzcd}\]
\section{Solitons}
The Korteweg--de Vries equation $u_t+u_{xxx}+6uu_x=0$ has the travelling waves
$\tfrac c2\sech^2\bigl(\tfrac{\sqrt c}2(x-ct)\bigr)$, of mass $2\sqrt c$ \src{\S14.2}. \chk{X5} Two of them pass through
each other and emerge unchanged except for a shift: for the exact two-soliton solution with wave numbers
$k_1=1.6$, $k_2=1$ the fast one is advanced by $\tfrac2{k_1}\log\tfrac{k_1+k_2}{k_1-k_2}=1.8329$ and the slow one
retarded by $2.9327$. \chk{X6}
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,3.2)--(0,0)--(4.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-8$};
\draw[ink2] (1.1,0)--(1.1,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-4$};
\draw[ink2] (2.2,0)--(2.2,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (3.3,0)--(3.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (4.4,0)--(4.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $8$};
\draw[ink2] (0,0)--(-0.07,0) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,1.455)--(-0.07,1.455) node[left,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (0,2.909)--(-0.07,2.909) node[left,inner sep=1.5pt]{\scriptsize $2$};
\node[below] at (2.2,-0.38) {\scriptsize $x$};
\draw[thick,ink2] plot coordinates {(0,0.001) (0.022,0.001) (0.044,0.001) (0.066,0.001) (0.088,0.001) (0.11,0.001) (0.132,0.002) (0.154,0.002) (0.176,0.002) (0.198,0.002) (0.22,0.002) (0.242,0.002) (0.264,0.003) (0.286,0.003) (0.308,0.003) (0.33,0.003) (0.352,0.004) (0.374,0.004) (0.396,0.004) (0.418,0.004) (0.44,0.005) (0.462,0.005) (0.484,0.006) (0.506,0.006) (0.528,0.007) (0.55,0.007) (0.572,0.008) (0.594,0.008) (0.616,0.009) (0.638,0.01) (0.66,0.011) (0.682,0.012) (0.704,0.013) (0.726,0.014) (0.748,0.015) (0.77,0.016) (0.792,0.017) (0.814,0.019) (0.836,0.02) (0.858,0.022) (0.88,0.024) (0.902,0.025) (0.924,0.028) (0.946,0.03) (0.968,0.032) (0.99,0.035) (1.012,0.038) (1.034,0.041) (1.056,0.044) (1.078,0.048) (1.1,0.051) (1.122,0.055) (1.144,0.06) (1.166,0.065) (1.188,0.07) (1.21,0.075) (1.232,0.081) (1.254,0.088) (1.276,0.094) (1.298,0.102) (1.32,0.109) (1.342,0.118) (1.364,0.127) (1.386,0.136) (1.408,0.146) (1.43,0.157) (1.452,0.169) (1.474,0.181) (1.496,0.194) (1.518,0.207) (1.54,0.222) (1.562,0.237) (1.584,0.253) (1.606,0.27) (1.628,0.287) (1.65,0.305) (1.672,0.324) (1.694,0.344) (1.716,0.364) (1.738,0.385) (1.76,0.407) (1.782,0.428) (1.804,0.45) (1.826,0.473) (1.848,0.495) (1.87,0.518) (1.892,0.54) (1.914,0.561) (1.936,0.582) (1.958,0.603) (1.98,0.622) (2.002,0.641) (2.024,0.658) (2.046,0.673) (2.068,0.687) (2.09,0.699) (2.112,0.709) (2.134,0.717) (2.156,0.723) (2.178,0.726) (2.2,0.727) (2.222,0.726) (2.244,0.723) (2.266,0.717) (2.288,0.709) (2.31,0.699) (2.332,0.687) (2.354,0.673) (2.376,0.658) (2.398,0.641) (2.42,0.622) (2.442,0.603) (2.464,0.582) (2.486,0.561) (2.508,0.54) (2.53,0.518) (2.552,0.495) (2.574,0.473) (2.596,0.45) (2.618,0.428) (2.64,0.407) (2.662,0.385) (2.684,0.364) (2.706,0.344) (2.728,0.324) (2.75,0.305) (2.772,0.287) (2.794,0.27) (2.816,0.253) (2.838,0.237) (2.86,0.222) (2.882,0.207) (2.904,0.194) (2.926,0.181) (2.948,0.169) (2.97,0.157) (2.992,0.146) (3.014,0.136) (3.036,0.127) (3.058,0.118) (3.08,0.109) (3.102,0.102) (3.124,0.094) (3.146,0.088) (3.168,0.081) (3.19,0.075) (3.212,0.07) (3.234,0.065) (3.256,0.06) (3.278,0.055) (3.3,0.051) (3.322,0.048) (3.344,0.044) (3.366,0.041) (3.388,0.038) (3.41,0.035) (3.432,0.032) (3.454,0.03) (3.476,0.028) (3.498,0.025) (3.52,0.024) (3.542,0.022) (3.564,0.02) (3.586,0.019) (3.608,0.017) (3.63,0.016) (3.652,0.015) (3.674,0.014) (3.696,0.013) (3.718,0.012) (3.74,0.011) (3.762,0.01) (3.784,0.009) (3.806,0.008) (3.828,0.008) (3.85,0.007) (3.872,0.007) (3.894,0.006) (3.916,0.006) (3.938,0.005) (3.96,0.005) (3.982,0.004) (4.004,0.004) (4.026,0.004) (4.048,0.004) (4.07,0.003) (4.092,0.003) (4.114,0.003) (4.136,0.003) (4.158,0.002) (4.18,0.002) (4.202,0.002) (4.224,0.002) (4.246,0.002) (4.268,0.002) (4.29,0.001) (4.312,0.001) (4.334,0.001) (4.356,0.001) (4.378,0.001) (4.4,0.001)};
\draw[thick,fwd] plot coordinates {(0,0) (0.022,0) (0.044,0) (0.066,0) (0.088,0) (0.11,0) (0.132,0) (0.154,0) (0.176,0) (0.198,0) (0.22,0) (0.242,0) (0.264,0) (0.286,0) (0.308,0) (0.33,0) (0.352,0) (0.374,0) (0.396,0.001) (0.418,0.001) (0.44,0.001) (0.462,0.001) (0.484,0.001) (0.506,0.001) (0.528,0.001) (0.55,0.001) (0.572,0.001) (0.594,0.002) (0.616,0.002) (0.638,0.002) (0.66,0.002) (0.682,0.002) (0.704,0.003) (0.726,0.003) (0.748,0.003) (0.77,0.004) (0.792,0.004) (0.814,0.005) (0.836,0.005) (0.858,0.006) (0.88,0.007) (0.902,0.007) (0.924,0.008) (0.946,0.009) (0.968,0.01) (0.99,0.011) (1.012,0.013) (1.034,0.014) (1.056,0.016) (1.078,0.018) (1.1,0.02) (1.122,0.023) (1.144,0.025) (1.166,0.028) (1.188,0.032) (1.21,0.035) (1.232,0.04) (1.254,0.044) (1.276,0.049) (1.298,0.055) (1.32,0.062) (1.342,0.069) (1.364,0.077) (1.386,0.086) (1.408,0.096) (1.43,0.107) (1.452,0.119) (1.474,0.133) (1.496,0.148) (1.518,0.164) (1.54,0.183) (1.562,0.203) (1.584,0.226) (1.606,0.25) (1.628,0.277) (1.65,0.307) (1.672,0.339) (1.694,0.374) (1.716,0.412) (1.738,0.453) (1.76,0.497) (1.782,0.544) (1.804,0.594) (1.826,0.647) (1.848,0.703) (1.87,0.761) (1.892,0.822) (1.914,0.884) (1.936,0.947) (1.958,1.01) (1.98,1.073) (2.002,1.134) (2.024,1.193) (2.046,1.248) (2.068,1.299) (2.09,1.344) (2.112,1.383) (2.134,1.413) (2.156,1.436) (2.178,1.45) (2.2,1.455) (2.222,1.45) (2.244,1.436) (2.266,1.413) (2.288,1.383) (2.31,1.344) (2.332,1.299) (2.354,1.248) (2.376,1.193) (2.398,1.134) (2.42,1.073) (2.442,1.01) (2.464,0.947) (2.486,0.884) (2.508,0.822) (2.53,0.761) (2.552,0.703) (2.574,0.647) (2.596,0.594) (2.618,0.544) (2.64,0.497) (2.662,0.453) (2.684,0.412) (2.706,0.374) (2.728,0.339) (2.75,0.307) (2.772,0.277) (2.794,0.25) (2.816,0.226) (2.838,0.203) (2.86,0.183) (2.882,0.164) (2.904,0.148) (2.926,0.133) (2.948,0.119) (2.97,0.107) (2.992,0.096) (3.014,0.086) (3.036,0.077) (3.058,0.069) (3.08,0.062) (3.102,0.055) (3.124,0.049) (3.146,0.044) (3.168,0.04) (3.19,0.035) (3.212,0.032) (3.234,0.028) (3.256,0.025) (3.278,0.023) (3.3,0.02) (3.322,0.018) (3.344,0.016) (3.366,0.014) (3.388,0.013) (3.41,0.011) (3.432,0.01) (3.454,0.009) (3.476,0.008) (3.498,0.007) (3.52,0.007) (3.542,0.006) (3.564,0.005) (3.586,0.005) (3.608,0.004) (3.63,0.004) (3.652,0.003) (3.674,0.003) (3.696,0.003) (3.718,0.002) (3.74,0.002) (3.762,0.002) (3.784,0.002) (3.806,0.002) (3.828,0.001) (3.85,0.001) (3.872,0.001) (3.894,0.001) (3.916,0.001) (3.938,0.001) (3.96,0.001) (3.982,0.001) (4.004,0.001) (4.026,0) (4.048,0) (4.07,0) (4.092,0) (4.114,0) (4.136,0) (4.158,0) (4.18,0) (4.202,0) (4.224,0) (4.246,0) (4.268,0) (4.29,0) (4.312,0) (4.334,0) (4.356,0) (4.378,0) (4.4,0)};
\draw[thick,bwd] plot coordinates {(0,0) (0.022,0) (0.044,0) (0.066,0) (0.088,0) (0.11,0) (0.132,0) (0.154,0) (0.176,0) (0.198,0) (0.22,0) (0.242,0) (0.264,0) (0.286,0) (0.308,0) (0.33,0) (0.352,0) (0.374,0) (0.396,0) (0.418,0) (0.44,0) (0.462,0) (0.484,0) (0.506,0) (0.528,0) (0.55,0) (0.572,0) (0.594,0) (0.616,0) (0.638,0) (0.66,0) (0.682,0) (0.704,0) (0.726,0) (0.748,0) (0.77,0) (0.792,0) (0.814,0) (0.836,0.001) (0.858,0.001) (0.88,0.001) (0.902,0.001) (0.924,0.001) (0.946,0.001) (0.968,0.001) (0.99,0.002) (1.012,0.002) (1.034,0.002) (1.056,0.003) (1.078,0.003) (1.1,0.004) (1.122,0.005) (1.144,0.005) (1.166,0.006) (1.188,0.007) (1.21,0.009) (1.232,0.01) (1.254,0.012) (1.276,0.014) (1.298,0.016) (1.32,0.019) (1.342,0.023) (1.364,0.027) (1.386,0.031) (1.408,0.036) (1.43,0.043) (1.452,0.05) (1.474,0.059) (1.496,0.069) (1.518,0.08) (1.54,0.094) (1.562,0.11) (1.584,0.129) (1.606,0.151) (1.628,0.176) (1.65,0.206) (1.672,0.24) (1.694,0.279) (1.716,0.325) (1.738,0.378) (1.76,0.438) (1.782,0.507) (1.804,0.586) (1.826,0.675) (1.848,0.775) (1.87,0.887) (1.892,1.012) (1.914,1.149) (1.936,1.298) (1.958,1.457) (1.98,1.626) (2.002,1.802) (2.024,1.981) (2.046,2.158) (2.068,2.33) (2.09,2.489) (2.112,2.63) (2.134,2.748) (2.156,2.836) (2.178,2.891) (2.2,2.909) (2.222,2.891) (2.244,2.836) (2.266,2.748) (2.288,2.63) (2.31,2.489) (2.332,2.33) (2.354,2.158) (2.376,1.981) (2.398,1.802) (2.42,1.626) (2.442,1.457) (2.464,1.298) (2.486,1.149) (2.508,1.012) (2.53,0.887) (2.552,0.775) (2.574,0.675) (2.596,0.586) (2.618,0.507) (2.64,0.438) (2.662,0.378) (2.684,0.325) (2.706,0.279) (2.728,0.24) (2.75,0.206) (2.772,0.176) (2.794,0.151) (2.816,0.129) (2.838,0.11) (2.86,0.094) (2.882,0.08) (2.904,0.069) (2.926,0.059) (2.948,0.05) (2.97,0.043) (2.992,0.036) (3.014,0.031) (3.036,0.027) (3.058,0.023) (3.08,0.019) (3.102,0.016) (3.124,0.014) (3.146,0.012) (3.168,0.01) (3.19,0.009) (3.212,0.007) (3.234,0.006) (3.256,0.005) (3.278,0.005) (3.3,0.004) (3.322,0.003) (3.344,0.003) (3.366,0.002) (3.388,0.002) (3.41,0.002) (3.432,0.001) (3.454,0.001) (3.476,0.001) (3.498,0.001) (3.52,0.001) (3.542,0.001) (3.564,0.001) (3.586,0) (3.608,0) (3.63,0) (3.652,0) (3.674,0) (3.696,0) (3.718,0) (3.74,0) (3.762,0) (3.784,0) (3.806,0) (3.828,0) (3.85,0) (3.872,0) (3.894,0) (3.916,0) (3.938,0) (3.96,0) (3.982,0) (4.004,0) (4.026,0) (4.048,0) (4.07,0) (4.092,0) (4.114,0) (4.136,0) (4.158,0) (4.18,0) (4.202,0) (4.224,0) (4.246,0) (4.268,0) (4.29,0) (4.312,0) (4.334,0) (4.356,0) (4.378,0) (4.4,0)};
\node[bwd,right] at (2.448,2.836) {\scriptsize $c=4$};
\node[fwd,right] at (2.64,1.309) {\scriptsize $2$};
\node[ink2] at (3.108,0.436) {\scriptsize $1$};
\end{scope}
\begin{scope}[shift={(6,0)}]
\draw[ink2] (0,0)--(6.6,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-16$};
\draw[ink2] (1.65,0)--(1.65,-0.07) node[below,inner sep=1.5pt]{\scriptsize $-8$};
\draw[ink2] (3.3,0)--(3.3,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (4.95,0)--(4.95,-0.07) node[below,inner sep=1.5pt]{\scriptsize $8$};
\draw[ink2] (6.6,0)--(6.6,-0.07) node[below,inner sep=1.5pt]{\scriptsize $16$};
\draw[thick,fwd] plot coordinates {(0,0.071) (0.017,0.071) (0.033,0.071) (0.05,0.071) (0.066,0.071) (0.083,0.071) (0.099,0.071) (0.116,0.071) (0.132,0.072) (0.149,0.072) (0.165,0.072) (0.182,0.072) (0.198,0.072) (0.214,0.072) (0.231,0.073) (0.247,0.073) (0.264,0.073) (0.28,0.074) (0.297,0.074) (0.313,0.075) (0.33,0.075) (0.346,0.076) (0.363,0.077) (0.379,0.077) (0.396,0.078) (0.412,0.079) (0.429,0.081) (0.446,0.082) (0.462,0.083) (0.479,0.085) (0.495,0.087) (0.512,0.089) (0.528,0.092) (0.545,0.095) (0.561,0.098) (0.578,0.102) (0.594,0.106) (0.611,0.111) (0.627,0.116) (0.644,0.122) (0.66,0.129) (0.676,0.136) (0.693,0.145) (0.709,0.155) (0.726,0.166) (0.742,0.178) (0.759,0.191) (0.775,0.207) (0.792,0.224) (0.808,0.242) (0.825,0.263) (0.841,0.286) (0.858,0.311) (0.874,0.339) (0.891,0.369) (0.907,0.401) (0.924,0.436) (0.94,0.473) (0.957,0.513) (0.973,0.554) (0.99,0.597) (1.006,0.641) (1.023,0.685) (1.039,0.729) (1.056,0.772) (1.072,0.813) (1.089,0.852) (1.105,0.886) (1.122,0.916) (1.138,0.94) (1.155,0.958) (1.171,0.969) (1.188,0.972) (1.204,0.968) (1.221,0.958) (1.237,0.94) (1.254,0.916) (1.27,0.886) (1.287,0.851) (1.303,0.813) (1.32,0.772) (1.337,0.729) (1.353,0.685) (1.369,0.64) (1.386,0.597) (1.403,0.554) (1.419,0.513) (1.435,0.473) (1.452,0.436) (1.469,0.401) (1.485,0.369) (1.502,0.339) (1.518,0.312) (1.535,0.287) (1.551,0.264) (1.567,0.243) (1.584,0.224) (1.6,0.208) (1.617,0.192) (1.633,0.179) (1.65,0.167) (1.666,0.156) (1.683,0.147) (1.7,0.138) (1.716,0.131) (1.732,0.124) (1.749,0.118) (1.766,0.113) (1.782,0.109) (1.798,0.105) (1.815,0.101) (1.832,0.098) (1.848,0.096) (1.864,0.093) (1.881,0.091) (1.897,0.09) (1.914,0.088) (1.93,0.087) (1.947,0.086) (1.963,0.086) (1.98,0.085) (1.996,0.085) (2.013,0.084) (2.03,0.084) (2.046,0.084) (2.062,0.085) (2.079,0.085) (2.095,0.085) (2.112,0.086) (2.128,0.087) (2.145,0.088) (2.161,0.088) (2.178,0.09) (2.195,0.091) (2.211,0.092) (2.228,0.094) (2.244,0.095) (2.26,0.097) (2.277,0.099) (2.293,0.101) (2.31,0.103) (2.326,0.106) (2.343,0.108) (2.359,0.111) (2.376,0.114) (2.392,0.118) (2.409,0.121) (2.425,0.125) (2.442,0.129) (2.458,0.134) (2.475,0.138) (2.491,0.143) (2.508,0.149) (2.524,0.154) (2.541,0.16) (2.558,0.167) (2.574,0.173) (2.591,0.181) (2.607,0.188) (2.623,0.196) (2.64,0.204) (2.656,0.213) (2.673,0.222) (2.689,0.231) (2.706,0.241) (2.722,0.251) (2.739,0.261) (2.755,0.271) (2.772,0.282) (2.788,0.293) (2.805,0.303) (2.821,0.314) (2.838,0.325) (2.854,0.336) (2.871,0.346) (2.887,0.356) (2.904,0.366) (2.92,0.375) (2.937,0.384) (2.954,0.392) (2.97,0.399) (2.986,0.406) (3.003,0.411) (3.019,0.416) (3.036,0.419) (3.052,0.421) (3.069,0.423) (3.086,0.423) (3.102,0.422) (3.118,0.42) (3.135,0.417) (3.151,0.413) (3.168,0.407) (3.184,0.401) (3.201,0.394) (3.217,0.386) (3.234,0.378) (3.25,0.369) (3.267,0.359) (3.283,0.349) (3.3,0.339) (3.316,0.328) (3.333,0.318) (3.349,0.307) (3.366,0.296) (3.382,0.285) (3.399,0.274) (3.415,0.264) (3.432,0.254) (3.448,0.244) (3.465,0.234) (3.481,0.225) (3.498,0.215) (3.514,0.207) (3.531,0.198) (3.547,0.19) (3.564,0.183) (3.58,0.176) (3.597,0.169) (3.613,0.162) (3.63,0.156) (3.646,0.15) (3.663,0.145) (3.679,0.14) (3.696,0.135) (3.712,0.13) (3.729,0.126) (3.745,0.122) (3.762,0.119) (3.778,0.115) (3.795,0.112) (3.811,0.109) (3.828,0.106) (3.845,0.103) (3.861,0.101) (3.877,0.099) (3.894,0.097) (3.91,0.095) (3.927,0.093) (3.943,0.091) (3.96,0.09) (3.977,0.088) (3.993,0.087) (4.01,0.086) (4.026,0.085) (4.043,0.083) (4.059,0.083) (4.075,0.082) (4.092,0.081) (4.109,0.08) (4.125,0.079) (4.141,0.079) (4.158,0.078) (4.174,0.077) (4.191,0.077) (4.207,0.076) (4.224,0.076) (4.24,0.076) (4.257,0.075) (4.273,0.075) (4.29,0.074) (4.306,0.074) (4.323,0.074) (4.339,0.074) (4.356,0.073) (4.372,0.073) (4.389,0.073) (4.405,0.073) (4.422,0.073) (4.438,0.072) (4.455,0.072) (4.471,0.072) (4.488,0.072) (4.505,0.072) (4.521,0.072) (4.537,0.072) (4.554,0.072) (4.571,0.072) (4.587,0.071) (4.603,0.071) (4.62,0.071) (4.636,0.071) (4.653,0.071) (4.67,0.071) (4.686,0.071) (4.702,0.071) (4.719,0.071) (4.736,0.071) (4.752,0.071) (4.768,0.071) (4.785,0.071) (4.801,0.071) (4.818,0.071) (4.835,0.071) (4.851,0.071) (4.867,0.071) (4.884,0.071) (4.901,0.071) (4.917,0.071) (4.934,0.071) (4.95,0.071) (4.966,0.071) (4.983,0.071) (4.999,0.071) (5.016,0.071) (5.032,0.071) (5.049,0.071) (5.065,0.071) (5.082,0.071) (5.098,0.071) (5.115,0.071) (5.131,0.071) (5.148,0.071) (5.164,0.071) (5.181,0.071) (5.197,0.071) (5.214,0.071) (5.23,0.071) (5.247,0.071) (5.263,0.071) (5.28,0.071) (5.296,0.071) (5.313,0.071) (5.329,0.071) (5.346,0.071) (5.362,0.071) (5.379,0.071) (5.395,0.071) (5.412,0.071) (5.428,0.071) (5.445,0.071) (5.462,0.071) (5.478,0.071) (5.494,0.071) (5.511,0.071) (5.527,0.071) (5.544,0.071) (5.561,0.071) (5.577,0.07) (5.593,0.07) (5.61,0.07) (5.627,0.07) (5.643,0.07) (5.659,0.07) (5.676,0.07) (5.692,0.07) (5.709,0.07) (5.726,0.07) (5.742,0.07) (5.758,0.07) (5.775,0.07) (5.791,0.07) (5.808,0.07) (5.824,0.07) (5.841,0.07) (5.857,0.07) (5.874,0.07) (5.89,0.07) (5.907,0.07) (5.923,0.07) (5.94,0.07) (5.956,0.07) (5.973,0.07) (5.989,0.07) (6.006,0.07) (6.022,0.07) (6.039,0.07) (6.055,0.07) (6.072,0.07) (6.088,0.07) (6.105,0.07) (6.121,0.07) (6.138,0.07) (6.154,0.07) (6.171,0.07) (6.188,0.07) (6.204,0.07) (6.22,0.07) (6.237,0.07) (6.253,0.07) (6.27,0.07) (6.286,0.07) (6.303,0.07) (6.319,0.07) (6.336,0.07) (6.353,0.07) (6.369,0.07) (6.385,0.07) (6.402,0.07) (6.418,0.07) (6.435,0.07) (6.452,0.07) (6.468,0.07) (6.484,0.07) (6.501,0.07) (6.518,0.07) (6.534,0.07) (6.55,0.07) (6.567,0.07) (6.583,0.07) (6.6,0.07)};
\node[right] at (0,0.268) {\scriptsize $t=-4$};
\draw[thick,fwd] plot coordinates {(0,0.881) (0.017,0.881) (0.033,0.881) (0.05,0.881) (0.066,0.881) (0.083,0.881) (0.099,0.881) (0.116,0.881) (0.132,0.881) (0.149,0.881) (0.165,0.881) (0.182,0.881) (0.198,0.881) (0.214,0.881) (0.231,0.881) (0.247,0.881) (0.264,0.881) (0.28,0.881) (0.297,0.881) (0.313,0.881) (0.33,0.881) (0.346,0.881) (0.363,0.881) (0.379,0.881) (0.396,0.881) (0.412,0.881) (0.429,0.881) (0.446,0.881) (0.462,0.881) (0.479,0.881) (0.495,0.881) (0.512,0.881) (0.528,0.881) (0.545,0.881) (0.561,0.881) (0.578,0.881) (0.594,0.881) (0.611,0.881) (0.627,0.881) (0.644,0.881) (0.66,0.881) (0.676,0.881) (0.693,0.881) (0.709,0.881) (0.726,0.881) (0.742,0.881) (0.759,0.881) (0.775,0.881) (0.792,0.881) (0.808,0.881) (0.825,0.881) (0.841,0.881) (0.858,0.881) (0.874,0.881) (0.891,0.881) (0.907,0.881) (0.924,0.881) (0.94,0.881) (0.957,0.881) (0.973,0.881) (0.99,0.882) (1.006,0.882) (1.023,0.882) (1.039,0.882) (1.056,0.882) (1.072,0.882) (1.089,0.882) (1.105,0.882) (1.122,0.882) (1.138,0.882) (1.155,0.882) (1.171,0.882) (1.188,0.883) (1.204,0.883) (1.221,0.883) (1.237,0.883) (1.254,0.883) (1.27,0.884) (1.287,0.884) (1.303,0.884) (1.32,0.885) (1.337,0.885) (1.353,0.886) (1.369,0.886) (1.386,0.887) (1.403,0.887) (1.419,0.888) (1.435,0.889) (1.452,0.89) (1.469,0.891) (1.485,0.893) (1.502,0.894) (1.518,0.896) (1.535,0.898) (1.551,0.9) (1.567,0.902) (1.584,0.905) (1.6,0.908) (1.617,0.911) (1.633,0.915) (1.65,0.92) (1.666,0.925) (1.683,0.93) (1.7,0.936) (1.716,0.944) (1.732,0.951) (1.749,0.96) (1.766,0.97) (1.782,0.981) (1.798,0.994) (1.815,1.008) (1.832,1.023) (1.848,1.041) (1.864,1.06) (1.881,1.081) (1.897,1.104) (1.914,1.129) (1.93,1.156) (1.947,1.186) (1.963,1.219) (1.98,1.253) (1.996,1.29) (2.013,1.329) (2.03,1.369) (2.046,1.411) (2.062,1.454) (2.079,1.497) (2.095,1.539) (2.112,1.581) (2.128,1.62) (2.145,1.657) (2.161,1.689) (2.178,1.718) (2.195,1.74) (2.211,1.757) (2.228,1.766) (2.244,1.769) (2.26,1.765) (2.277,1.754) (2.293,1.737) (2.31,1.713) (2.326,1.685) (2.343,1.652) (2.359,1.615) (2.376,1.576) (2.392,1.534) (2.409,1.492) (2.425,1.45) (2.442,1.408) (2.458,1.368) (2.475,1.329) (2.491,1.291) (2.508,1.256) (2.524,1.223) (2.541,1.193) (2.558,1.165) (2.574,1.139) (2.591,1.116) (2.607,1.094) (2.623,1.076) (2.64,1.059) (2.656,1.044) (2.673,1.031) (2.689,1.019) (2.706,1.009) (2.722,1.001) (2.739,0.994) (2.755,0.988) (2.772,0.983) (2.788,0.979) (2.805,0.977) (2.821,0.975) (2.838,0.974) (2.854,0.974) (2.871,0.975) (2.887,0.976) (2.904,0.978) (2.92,0.981) (2.937,0.984) (2.954,0.988) (2.97,0.993) (2.986,0.998) (3.003,1.003) (3.019,1.009) (3.036,1.016) (3.052,1.023) (3.069,1.031) (3.086,1.039) (3.102,1.048) (3.118,1.057) (3.135,1.066) (3.151,1.076) (3.168,1.086) (3.184,1.096) (3.201,1.107) (3.217,1.117) (3.234,1.128) (3.25,1.138) (3.267,1.149) (3.283,1.159) (3.3,1.169) (3.316,1.179) (3.333,1.188) (3.349,1.196) (3.366,1.204) (3.382,1.211) (3.399,1.218) (3.415,1.223) (3.432,1.227) (3.448,1.231) (3.465,1.233) (3.481,1.234) (3.498,1.234) (3.514,1.233) (3.531,1.231) (3.547,1.228) (3.564,1.224) (3.58,1.219) (3.597,1.213) (3.613,1.205) (3.63,1.198) (3.646,1.189) (3.663,1.18) (3.679,1.17) (3.696,1.16) (3.712,1.15) (3.729,1.139) (3.745,1.129) (3.762,1.118) (3.778,1.107) (3.795,1.096) (3.811,1.085) (3.828,1.075) (3.845,1.065) (3.861,1.054) (3.877,1.045) (3.894,1.035) (3.91,1.026) (3.927,1.018) (3.943,1.009) (3.96,1.001) (3.977,0.994) (3.993,0.986) (4.01,0.979) (4.026,0.973) (4.043,0.967) (4.059,0.961) (4.075,0.956) (4.092,0.95) (4.109,0.946) (4.125,0.941) (4.141,0.937) (4.158,0.933) (4.174,0.929) (4.191,0.926) (4.207,0.923) (4.224,0.92) (4.24,0.917) (4.257,0.914) (4.273,0.912) (4.29,0.909) (4.306,0.907) (4.323,0.905) (4.339,0.904) (4.356,0.902) (4.372,0.9) (4.389,0.899) (4.405,0.898) (4.422,0.896) (4.438,0.895) (4.455,0.894) (4.471,0.893) (4.488,0.892) (4.505,0.891) (4.521,0.891) (4.537,0.89) (4.554,0.889) (4.571,0.889) (4.587,0.888) (4.603,0.888) (4.62,0.887) (4.636,0.887) (4.653,0.886) (4.67,0.886) (4.686,0.885) (4.702,0.885) (4.719,0.885) (4.736,0.885) (4.752,0.884) (4.768,0.884) (4.785,0.884) (4.801,0.884) (4.818,0.883) (4.835,0.883) (4.851,0.883) (4.867,0.883) (4.884,0.883) (4.901,0.883) (4.917,0.883) (4.934,0.882) (4.95,0.882) (4.966,0.882) (4.983,0.882) (4.999,0.882) (5.016,0.882) (5.032,0.882) (5.049,0.882) (5.065,0.882) (5.082,0.882) (5.098,0.882) (5.115,0.882) (5.131,0.882) (5.148,0.882) (5.164,0.882) (5.181,0.882) (5.197,0.882) (5.214,0.881) (5.23,0.881) (5.247,0.881) (5.263,0.881) (5.28,0.881) (5.296,0.881) (5.313,0.881) (5.329,0.881) (5.346,0.881) (5.362,0.881) (5.379,0.881) (5.395,0.881) (5.412,0.881) (5.428,0.881) (5.445,0.881) (5.462,0.881) (5.478,0.881) (5.494,0.881) (5.511,0.881) (5.527,0.881) (5.544,0.881) (5.561,0.881) (5.577,0.881) (5.593,0.881) (5.61,0.881) (5.627,0.881) (5.643,0.881) (5.659,0.881) (5.676,0.881) (5.692,0.881) (5.709,0.881) (5.726,0.881) (5.742,0.881) (5.758,0.881) (5.775,0.881) (5.791,0.881) (5.808,0.881) (5.824,0.881) (5.841,0.881) (5.857,0.881) (5.874,0.881) (5.89,0.881) (5.907,0.881) (5.923,0.881) (5.94,0.881) (5.956,0.881) (5.973,0.881) (5.989,0.881) (6.006,0.881) (6.022,0.881) (6.039,0.881) (6.055,0.881) (6.072,0.881) (6.088,0.881) (6.105,0.881) (6.121,0.881) (6.138,0.881) (6.154,0.881) (6.171,0.881) (6.188,0.881) (6.204,0.881) (6.22,0.881) (6.237,0.881) (6.253,0.881) (6.27,0.881) (6.286,0.881) (6.303,0.881) (6.319,0.881) (6.336,0.881) (6.353,0.881) (6.369,0.881) (6.385,0.881) (6.402,0.881) (6.418,0.881) (6.435,0.881) (6.452,0.881) (6.468,0.881) (6.484,0.881) (6.501,0.881) (6.518,0.881) (6.534,0.881) (6.55,0.881) (6.567,0.881) (6.583,0.881) (6.6,0.881)};
\node[right] at (0,1.079) {\scriptsize $t=-2$};
\draw[thick,fwd] plot coordinates {(0,1.692) (0.017,1.692) (0.033,1.692) (0.05,1.692) (0.066,1.692) (0.083,1.692) (0.099,1.692) (0.116,1.692) (0.132,1.692) (0.149,1.692) (0.165,1.692) (0.182,1.692) (0.198,1.692) (0.214,1.692) (0.231,1.692) (0.247,1.692) (0.264,1.692) (0.28,1.692) (0.297,1.692) (0.313,1.692) (0.33,1.692) (0.346,1.692) (0.363,1.692) (0.379,1.692) (0.396,1.692) (0.412,1.692) (0.429,1.692) (0.446,1.692) (0.462,1.692) (0.479,1.692) (0.495,1.692) (0.512,1.692) (0.528,1.692) (0.545,1.692) (0.561,1.692) (0.578,1.692) (0.594,1.692) (0.611,1.692) (0.627,1.692) (0.644,1.692) (0.66,1.692) (0.676,1.692) (0.693,1.692) (0.709,1.692) (0.726,1.692) (0.742,1.692) (0.759,1.692) (0.775,1.692) (0.792,1.692) (0.808,1.692) (0.825,1.692) (0.841,1.692) (0.858,1.692) (0.874,1.692) (0.891,1.692) (0.907,1.692) (0.924,1.692) (0.94,1.692) (0.957,1.692) (0.973,1.692) (0.99,1.692) (1.006,1.692) (1.023,1.692) (1.039,1.692) (1.056,1.692) (1.072,1.692) (1.089,1.692) (1.105,1.692) (1.122,1.692) (1.138,1.692) (1.155,1.692) (1.171,1.692) (1.188,1.692) (1.204,1.692) (1.221,1.692) (1.237,1.692) (1.254,1.692) (1.27,1.692) (1.287,1.692) (1.303,1.692) (1.32,1.692) (1.337,1.692) (1.353,1.692) (1.369,1.692) (1.386,1.692) (1.403,1.692) (1.419,1.692) (1.435,1.692) (1.452,1.692) (1.469,1.692) (1.485,1.692) (1.502,1.692) (1.518,1.692) (1.535,1.692) (1.551,1.692) (1.567,1.692) (1.584,1.692) (1.6,1.692) (1.617,1.692) (1.633,1.692) (1.65,1.692) (1.666,1.692) (1.683,1.692) (1.7,1.692) (1.716,1.692) (1.732,1.693) (1.749,1.693) (1.766,1.693) (1.782,1.693) (1.798,1.693) (1.815,1.693) (1.832,1.693) (1.848,1.693) (1.864,1.693) (1.881,1.693) (1.897,1.693) (1.914,1.694) (1.93,1.694) (1.947,1.694) (1.963,1.694) (1.98,1.694) (1.996,1.694) (2.013,1.695) (2.03,1.695) (2.046,1.695) (2.062,1.696) (2.079,1.696) (2.095,1.696) (2.112,1.697) (2.128,1.697) (2.145,1.697) (2.161,1.698) (2.178,1.698) (2.195,1.699) (2.211,1.7) (2.228,1.7) (2.244,1.701) (2.26,1.702) (2.277,1.703) (2.293,1.704) (2.31,1.705) (2.326,1.706) (2.343,1.707) (2.359,1.709) (2.376,1.71) (2.392,1.712) (2.409,1.714) (2.425,1.715) (2.442,1.718) (2.458,1.72) (2.475,1.722) (2.491,1.725) (2.508,1.728) (2.524,1.731) (2.541,1.735) (2.558,1.739) (2.574,1.743) (2.591,1.748) (2.607,1.753) (2.623,1.758) (2.64,1.764) (2.656,1.771) (2.673,1.778) (2.689,1.785) (2.706,1.794) (2.722,1.803) (2.739,1.812) (2.755,1.823) (2.772,1.834) (2.788,1.846) (2.805,1.86) (2.821,1.874) (2.838,1.889) (2.854,1.905) (2.871,1.922) (2.887,1.94) (2.904,1.959) (2.92,1.98) (2.937,2.001) (2.954,2.023) (2.97,2.046) (2.986,2.07) (3.003,2.094) (3.019,2.118) (3.036,2.143) (3.052,2.168) (3.069,2.193) (3.086,2.217) (3.102,2.24) (3.118,2.262) (3.135,2.283) (3.151,2.302) (3.168,2.319) (3.184,2.334) (3.201,2.347) (3.217,2.357) (3.234,2.364) (3.25,2.369) (3.267,2.371) (3.283,2.371) (3.3,2.368) (3.316,2.363) (3.333,2.355) (3.349,2.346) (3.366,2.336) (3.382,2.324) (3.399,2.312) (3.415,2.299) (3.432,2.286) (3.448,2.273) (3.465,2.26) (3.481,2.247) (3.498,2.235) (3.514,2.224) (3.531,2.215) (3.547,2.206) (3.564,2.198) (3.58,2.191) (3.597,2.186) (3.613,2.182) (3.63,2.178) (3.646,2.176) (3.663,2.175) (3.679,2.174) (3.696,2.174) (3.712,2.174) (3.729,2.175) (3.745,2.176) (3.762,2.177) (3.778,2.177) (3.795,2.178) (3.811,2.178) (3.828,2.177) (3.845,2.176) (3.861,2.173) (3.877,2.17) (3.894,2.166) (3.91,2.161) (3.927,2.155) (3.943,2.147) (3.96,2.139) (3.977,2.13) (3.993,2.12) (4.01,2.108) (4.026,2.096) (4.043,2.084) (4.059,2.07) (4.075,2.057) (4.092,2.043) (4.109,2.028) (4.125,2.014) (4.141,1.999) (4.158,1.984) (4.174,1.97) (4.191,1.956) (4.207,1.942) (4.224,1.928) (4.24,1.914) (4.257,1.901) (4.273,1.889) (4.29,1.877) (4.306,1.865) (4.323,1.854) (4.339,1.844) (4.356,1.834) (4.372,1.825) (4.389,1.816) (4.405,1.807) (4.422,1.799) (4.438,1.792) (4.455,1.785) (4.471,1.778) (4.488,1.772) (4.505,1.766) (4.521,1.761) (4.537,1.756) (4.554,1.751) (4.571,1.747) (4.587,1.743) (4.603,1.739) (4.62,1.735) (4.636,1.732) (4.653,1.729) (4.67,1.726) (4.686,1.724) (4.702,1.721) (4.719,1.719) (4.736,1.717) (4.752,1.715) (4.768,1.713) (4.785,1.712) (4.801,1.71) (4.818,1.709) (4.835,1.707) (4.851,1.706) (4.867,1.705) (4.884,1.704) (4.901,1.703) (4.917,1.702) (4.934,1.702) (4.95,1.701) (4.966,1.7) (4.983,1.699) (4.999,1.699) (5.016,1.698) (5.032,1.698) (5.049,1.697) (5.065,1.697) (5.082,1.697) (5.098,1.696) (5.115,1.696) (5.131,1.696) (5.148,1.695) (5.164,1.695) (5.181,1.695) (5.197,1.694) (5.214,1.694) (5.23,1.694) (5.247,1.694) (5.263,1.694) (5.28,1.694) (5.296,1.693) (5.313,1.693) (5.329,1.693) (5.346,1.693) (5.362,1.693) (5.379,1.693) (5.395,1.693) (5.412,1.693) (5.428,1.693) (5.445,1.693) (5.462,1.693) (5.478,1.692) (5.494,1.692) (5.511,1.692) (5.527,1.692) (5.544,1.692) (5.561,1.692) (5.577,1.692) (5.593,1.692) (5.61,1.692) (5.627,1.692) (5.643,1.692) (5.659,1.692) (5.676,1.692) (5.692,1.692) (5.709,1.692) (5.726,1.692) (5.742,1.692) (5.758,1.692) (5.775,1.692) (5.791,1.692) (5.808,1.692) (5.824,1.692) (5.841,1.692) (5.857,1.692) (5.874,1.692) (5.89,1.692) (5.907,1.692) (5.923,1.692) (5.94,1.692) (5.956,1.692) (5.973,1.692) (5.989,1.692) (6.006,1.692) (6.022,1.692) (6.039,1.692) (6.055,1.692) (6.072,1.692) (6.088,1.692) (6.105,1.692) (6.121,1.692) (6.138,1.692) (6.154,1.692) (6.171,1.692) (6.188,1.692) (6.204,1.692) (6.22,1.692) (6.237,1.692) (6.253,1.692) (6.27,1.692) (6.286,1.692) (6.303,1.692) (6.319,1.692) (6.336,1.692) (6.353,1.692) (6.369,1.692) (6.385,1.692) (6.402,1.692) (6.418,1.692) (6.435,1.692) (6.452,1.692) (6.468,1.692) (6.484,1.692) (6.501,1.692) (6.518,1.692) (6.534,1.692) (6.55,1.692) (6.567,1.692) (6.583,1.692) (6.6,1.692)};
\node[right] at (0,1.889) {\scriptsize $t=0$};
\draw[thick,fwd] plot coordinates {(0,2.502) (0.017,2.502) (0.033,2.502) (0.05,2.502) (0.066,2.502) (0.083,2.502) (0.099,2.502) (0.116,2.502) (0.132,2.502) (0.149,2.502) (0.165,2.502) (0.182,2.502) (0.198,2.502) (0.214,2.502) (0.231,2.502) (0.247,2.502) (0.264,2.502) (0.28,2.502) (0.297,2.502) (0.313,2.502) (0.33,2.502) (0.346,2.502) (0.363,2.502) (0.379,2.502) (0.396,2.502) (0.412,2.502) (0.429,2.502) (0.446,2.502) (0.462,2.502) (0.479,2.502) (0.495,2.502) (0.512,2.502) (0.528,2.502) (0.545,2.502) (0.561,2.502) (0.578,2.502) (0.594,2.502) (0.611,2.502) (0.627,2.502) (0.644,2.502) (0.66,2.502) (0.676,2.502) (0.693,2.502) (0.709,2.502) (0.726,2.502) (0.742,2.502) (0.759,2.502) (0.775,2.502) (0.792,2.502) (0.808,2.502) (0.825,2.502) (0.841,2.502) (0.858,2.502) (0.874,2.502) (0.891,2.502) (0.907,2.502) (0.924,2.502) (0.94,2.502) (0.957,2.502) (0.973,2.502) (0.99,2.502) (1.006,2.502) (1.023,2.502) (1.039,2.502) (1.056,2.502) (1.072,2.502) (1.089,2.502) (1.105,2.502) (1.122,2.502) (1.138,2.502) (1.155,2.502) (1.171,2.502) (1.188,2.502) (1.204,2.502) (1.221,2.502) (1.237,2.502) (1.254,2.502) (1.27,2.502) (1.287,2.502) (1.303,2.502) (1.32,2.502) (1.337,2.502) (1.353,2.502) (1.369,2.502) (1.386,2.502) (1.403,2.502) (1.419,2.502) (1.435,2.502) (1.452,2.502) (1.469,2.502) (1.485,2.502) (1.502,2.502) (1.518,2.502) (1.535,2.502) (1.551,2.502) (1.567,2.503) (1.584,2.503) (1.6,2.503) (1.617,2.503) (1.633,2.503) (1.65,2.503) (1.666,2.503) (1.683,2.503) (1.7,2.503) (1.716,2.503) (1.732,2.503) (1.749,2.503) (1.766,2.503) (1.782,2.503) (1.798,2.503) (1.815,2.503) (1.832,2.503) (1.848,2.503) (1.864,2.503) (1.881,2.503) (1.897,2.503) (1.914,2.503) (1.93,2.503) (1.947,2.503) (1.963,2.503) (1.98,2.503) (1.996,2.503) (2.013,2.503) (2.03,2.503) (2.046,2.503) (2.062,2.503) (2.079,2.503) (2.095,2.503) (2.112,2.503) (2.128,2.503) (2.145,2.503) (2.161,2.503) (2.178,2.503) (2.195,2.503) (2.211,2.503) (2.228,2.504) (2.244,2.504) (2.26,2.504) (2.277,2.504) (2.293,2.504) (2.31,2.504) (2.326,2.504) (2.343,2.504) (2.359,2.504) (2.376,2.505) (2.392,2.505) (2.409,2.505) (2.425,2.505) (2.442,2.505) (2.458,2.506) (2.475,2.506) (2.491,2.506) (2.508,2.507) (2.524,2.507) (2.541,2.507) (2.558,2.508) (2.574,2.508) (2.591,2.509) (2.607,2.509) (2.623,2.51) (2.64,2.51) (2.656,2.511) (2.673,2.511) (2.689,2.512) (2.706,2.513) (2.722,2.514) (2.739,2.515) (2.755,2.516) (2.772,2.517) (2.788,2.518) (2.805,2.519) (2.821,2.521) (2.838,2.522) (2.854,2.524) (2.871,2.526) (2.887,2.527) (2.904,2.529) (2.92,2.532) (2.937,2.534) (2.954,2.536) (2.97,2.539) (2.986,2.542) (3.003,2.545) (3.019,2.548) (3.036,2.552) (3.052,2.556) (3.069,2.56) (3.086,2.564) (3.102,2.569) (3.118,2.574) (3.135,2.579) (3.151,2.584) (3.168,2.59) (3.184,2.597) (3.201,2.603) (3.217,2.61) (3.234,2.618) (3.25,2.626) (3.267,2.634) (3.283,2.642) (3.3,2.651) (3.316,2.66) (3.333,2.67) (3.349,2.68) (3.366,2.69) (3.382,2.701) (3.399,2.711) (3.415,2.722) (3.432,2.733) (3.448,2.744) (3.465,2.755) (3.481,2.766) (3.498,2.777) (3.514,2.787) (3.531,2.797) (3.547,2.807) (3.564,2.816) (3.58,2.824) (3.597,2.832) (3.613,2.839) (3.63,2.845) (3.646,2.85) (3.663,2.855) (3.679,2.858) (3.696,2.86) (3.712,2.861) (3.729,2.86) (3.745,2.859) (3.762,2.857) (3.778,2.853) (3.795,2.849) (3.811,2.844) (3.828,2.837) (3.845,2.83) (3.861,2.823) (3.877,2.815) (3.894,2.806) (3.91,2.797) (3.927,2.788) (3.943,2.779) (3.96,2.769) (3.977,2.76) (3.993,2.75) (4.01,2.741) (4.026,2.733) (4.043,2.724) (4.059,2.716) (4.075,2.709) (4.092,2.702) (4.109,2.696) (4.125,2.691) (4.141,2.687) (4.158,2.683) (4.174,2.681) (4.191,2.679) (4.207,2.679) (4.224,2.68) (4.24,2.682) (4.257,2.685) (4.273,2.69) (4.29,2.697) (4.306,2.705) (4.323,2.715) (4.339,2.726) (4.356,2.74) (4.372,2.756) (4.389,2.774) (4.405,2.794) (4.422,2.816) (4.438,2.841) (4.455,2.868) (4.471,2.898) (4.488,2.929) (4.505,2.963) (4.521,2.998) (4.537,3.035) (4.554,3.073) (4.571,3.111) (4.587,3.149) (4.603,3.187) (4.62,3.222) (4.636,3.256) (4.653,3.286) (4.67,3.312) (4.686,3.333) (4.702,3.35) (4.719,3.36) (4.736,3.364) (4.752,3.362) (4.768,3.353) (4.785,3.339) (4.801,3.319) (4.818,3.293) (4.835,3.264) (4.851,3.23) (4.867,3.194) (4.884,3.155) (4.901,3.116) (4.917,3.075) (4.934,3.035) (4.95,2.995) (4.966,2.956) (4.983,2.919) (4.999,2.884) (5.016,2.85) (5.032,2.818) (5.049,2.789) (5.065,2.762) (5.082,2.736) (5.098,2.713) (5.115,2.692) (5.131,2.673) (5.148,2.655) (5.164,2.64) (5.181,2.625) (5.197,2.612) (5.214,2.601) (5.23,2.59) (5.247,2.581) (5.263,2.573) (5.28,2.565) (5.296,2.558) (5.313,2.552) (5.329,2.547) (5.346,2.542) (5.362,2.538) (5.379,2.534) (5.395,2.531) (5.412,2.528) (5.428,2.525) (5.445,2.523) (5.462,2.52) (5.478,2.519) (5.494,2.517) (5.511,2.515) (5.527,2.514) (5.544,2.513) (5.561,2.512) (5.577,2.511) (5.593,2.51) (5.61,2.509) (5.627,2.508) (5.643,2.508) (5.659,2.507) (5.676,2.507) (5.692,2.506) (5.709,2.506) (5.726,2.506) (5.742,2.505) (5.758,2.505) (5.775,2.505) (5.791,2.505) (5.808,2.504) (5.824,2.504) (5.841,2.504) (5.857,2.504) (5.874,2.504) (5.89,2.504) (5.907,2.503) (5.923,2.503) (5.94,2.503) (5.956,2.503) (5.973,2.503) (5.989,2.503) (6.006,2.503) (6.022,2.503) (6.039,2.503) (6.055,2.503) (6.072,2.503) (6.088,2.503) (6.105,2.503) (6.121,2.503) (6.138,2.503) (6.154,2.503) (6.171,2.503) (6.188,2.503) (6.204,2.503) (6.22,2.503) (6.237,2.503) (6.253,2.503) (6.27,2.503) (6.286,2.503) (6.303,2.503) (6.319,2.503) (6.336,2.503) (6.353,2.503) (6.369,2.503) (6.385,2.503) (6.402,2.503) (6.418,2.503) (6.435,2.503) (6.452,2.503) (6.468,2.503) (6.484,2.503) (6.501,2.502) (6.518,2.502) (6.534,2.502) (6.55,2.502) (6.567,2.502) (6.583,2.502) (6.6,2.502)};
\node[right] at (0,2.7) {\scriptsize $t=2$};
\draw[thick,fwd] plot coordinates {(0,3.313) (0.017,3.313) (0.033,3.313) (0.05,3.313) (0.066,3.313) (0.083,3.313) (0.099,3.313) (0.116,3.313) (0.132,3.313) (0.149,3.313) (0.165,3.313) (0.182,3.313) (0.198,3.313) (0.214,3.313) (0.231,3.313) (0.247,3.313) (0.264,3.313) (0.28,3.313) (0.297,3.313) (0.313,3.313) (0.33,3.313) (0.346,3.313) (0.363,3.313) (0.379,3.313) (0.396,3.313) (0.412,3.313) (0.429,3.313) (0.446,3.313) (0.462,3.313) (0.479,3.313) (0.495,3.313) (0.512,3.313) (0.528,3.313) (0.545,3.313) (0.561,3.313) (0.578,3.313) (0.594,3.313) (0.611,3.313) (0.627,3.313) (0.644,3.313) (0.66,3.313) (0.676,3.313) (0.693,3.313) (0.709,3.313) (0.726,3.313) (0.742,3.313) (0.759,3.313) (0.775,3.313) (0.792,3.313) (0.808,3.313) (0.825,3.313) (0.841,3.313) (0.858,3.313) (0.874,3.313) (0.891,3.313) (0.907,3.313) (0.924,3.313) (0.94,3.313) (0.957,3.313) (0.973,3.313) (0.99,3.313) (1.006,3.313) (1.023,3.313) (1.039,3.313) (1.056,3.313) (1.072,3.313) (1.089,3.313) (1.105,3.313) (1.122,3.313) (1.138,3.313) (1.155,3.313) (1.171,3.313) (1.188,3.313) (1.204,3.313) (1.221,3.313) (1.237,3.313) (1.254,3.313) (1.27,3.313) (1.287,3.313) (1.303,3.313) (1.32,3.313) (1.337,3.313) (1.353,3.313) (1.369,3.313) (1.386,3.313) (1.403,3.313) (1.419,3.313) (1.435,3.313) (1.452,3.313) (1.469,3.313) (1.485,3.313) (1.502,3.313) (1.518,3.313) (1.535,3.313) (1.551,3.313) (1.567,3.313) (1.584,3.313) (1.6,3.313) (1.617,3.313) (1.633,3.313) (1.65,3.313) (1.666,3.313) (1.683,3.313) (1.7,3.313) (1.716,3.313) (1.732,3.313) (1.749,3.313) (1.766,3.313) (1.782,3.313) (1.798,3.313) (1.815,3.313) (1.832,3.313) (1.848,3.313) (1.864,3.313) (1.881,3.313) (1.897,3.313) (1.914,3.313) (1.93,3.313) (1.947,3.313) (1.963,3.313) (1.98,3.313) (1.996,3.313) (2.013,3.313) (2.03,3.313) (2.046,3.313) (2.062,3.313) (2.079,3.313) (2.095,3.313) (2.112,3.313) (2.128,3.313) (2.145,3.313) (2.161,3.313) (2.178,3.313) (2.195,3.313) (2.211,3.313) (2.228,3.313) (2.244,3.313) (2.26,3.313) (2.277,3.313) (2.293,3.313) (2.31,3.313) (2.326,3.313) (2.343,3.313) (2.359,3.313) (2.376,3.313) (2.392,3.313) (2.409,3.313) (2.425,3.313) (2.442,3.314) (2.458,3.314) (2.475,3.314) (2.491,3.314) (2.508,3.314) (2.524,3.314) (2.541,3.314) (2.558,3.314) (2.574,3.314) (2.591,3.314) (2.607,3.314) (2.623,3.314) (2.64,3.314) (2.656,3.314) (2.673,3.314) (2.689,3.314) (2.706,3.315) (2.722,3.315) (2.739,3.315) (2.755,3.315) (2.772,3.315) (2.788,3.315) (2.805,3.315) (2.821,3.316) (2.838,3.316) (2.854,3.316) (2.871,3.316) (2.887,3.317) (2.904,3.317) (2.92,3.317) (2.937,3.318) (2.954,3.318) (2.97,3.318) (2.986,3.319) (3.003,3.319) (3.019,3.32) (3.036,3.32) (3.052,3.321) (3.069,3.321) (3.086,3.322) (3.102,3.323) (3.118,3.324) (3.135,3.325) (3.151,3.325) (3.168,3.326) (3.184,3.328) (3.201,3.329) (3.217,3.33) (3.234,3.331) (3.25,3.333) (3.267,3.334) (3.283,3.336) (3.3,3.338) (3.316,3.34) (3.333,3.342) (3.349,3.344) (3.366,3.347) (3.382,3.35) (3.399,3.352) (3.415,3.356) (3.432,3.359) (3.448,3.362) (3.465,3.366) (3.481,3.37) (3.498,3.375) (3.514,3.379) (3.531,3.384) (3.547,3.389) (3.564,3.395) (3.58,3.401) (3.597,3.407) (3.613,3.414) (3.63,3.421) (3.646,3.428) (3.663,3.436) (3.679,3.444) (3.696,3.452) (3.712,3.461) (3.729,3.47) (3.745,3.48) (3.762,3.49) (3.778,3.5) (3.795,3.51) (3.811,3.521) (3.828,3.531) (3.845,3.542) (3.861,3.553) (3.877,3.564) (3.894,3.575) (3.91,3.585) (3.927,3.595) (3.943,3.605) (3.96,3.615) (3.977,3.624) (3.993,3.632) (4.01,3.639) (4.026,3.646) (4.043,3.652) (4.059,3.657) (4.075,3.661) (4.092,3.663) (4.109,3.665) (4.125,3.666) (4.141,3.665) (4.158,3.663) (4.174,3.661) (4.191,3.657) (4.207,3.652) (4.224,3.646) (4.24,3.639) (4.257,3.632) (4.273,3.624) (4.29,3.615) (4.306,3.605) (4.323,3.595) (4.339,3.585) (4.356,3.575) (4.372,3.564) (4.389,3.553) (4.405,3.542) (4.422,3.532) (4.438,3.521) (4.455,3.51) (4.471,3.5) (4.488,3.49) (4.505,3.48) (4.521,3.471) (4.537,3.461) (4.554,3.453) (4.571,3.444) (4.587,3.436) (4.603,3.428) (4.62,3.421) (4.636,3.414) (4.653,3.408) (4.67,3.401) (4.686,3.396) (4.702,3.39) (4.719,3.385) (4.736,3.38) (4.752,3.376) (4.768,3.372) (4.785,3.368) (4.801,3.364) (4.818,3.361) (4.835,3.358) (4.851,3.355) (4.867,3.353) (4.884,3.35) (4.901,3.348) (4.917,3.346) (4.934,3.345) (4.95,3.344) (4.966,3.342) (4.983,3.342) (4.999,3.341) (5.016,3.341) (5.032,3.34) (5.049,3.34) (5.065,3.341) (5.082,3.342) (5.098,3.343) (5.115,3.344) (5.131,3.346) (5.148,3.348) (5.164,3.35) (5.181,3.353) (5.197,3.357) (5.214,3.361) (5.23,3.366) (5.247,3.372) (5.263,3.378) (5.28,3.385) (5.296,3.394) (5.313,3.403) (5.329,3.414) (5.346,3.426) (5.362,3.439) (5.379,3.454) (5.395,3.471) (5.412,3.489) (5.428,3.51) (5.445,3.533) (5.462,3.558) (5.478,3.586) (5.494,3.616) (5.511,3.648) (5.527,3.683) (5.544,3.72) (5.561,3.759) (5.577,3.8) (5.593,3.843) (5.61,3.887) (5.627,3.931) (5.643,3.975) (5.659,4.018) (5.676,4.058) (5.692,4.096) (5.709,4.13) (5.726,4.159) (5.742,4.183) (5.758,4.2) (5.775,4.21) (5.791,4.213) (5.808,4.209) (5.824,4.198) (5.841,4.18) (5.857,4.155) (5.874,4.125) (5.89,4.09) (5.907,4.052) (5.923,4.011) (5.94,3.968) (5.956,3.924) (5.973,3.88) (5.989,3.836) (6.006,3.793) (6.022,3.752) (6.039,3.713) (6.055,3.676) (6.072,3.641) (6.088,3.609) (6.105,3.58) (6.121,3.552) (6.138,3.527) (6.154,3.504) (6.171,3.484) (6.188,3.465) (6.204,3.448) (6.22,3.433) (6.237,3.42) (6.253,3.408) (6.27,3.397) (6.286,3.387) (6.303,3.379) (6.319,3.371) (6.336,3.364) (6.353,3.358) (6.369,3.353) (6.385,3.348) (6.402,3.344) (6.418,3.341) (6.435,3.337) (6.452,3.335) (6.468,3.332) (6.484,3.33) (6.501,3.328) (6.518,3.326) (6.534,3.325) (6.55,3.323) (6.567,3.322) (6.583,3.321) (6.6,3.32)};
\node[right] at (0,3.51) {\scriptsize $t=4$};
\end{scope}
\end{tikzpicture}
\figcap{Left: solitary waves $\frac c2\,\mathrm{sech}^2\bigl(\frac{\sqrt c}2x\bigr)$ of the Korteweg--de Vries equation; taller ones are narrower and faster. Right: the exact two-soliton solution with $k_1=1.6$, $k_2=1$ at five times, each curve raised for clarity. The taller wave overtakes the shorter; both emerge with their shapes, the taller advanced by $1.83$ and the shorter retarded by $2.93$ relative to free motion.}
\end{center}
\begin{theorem}[Inverse scattering]\label{thm:ist}
\Sli[\cite{GGKM67,Lax68}; source, \S14.2, Theorem 1] Let $u(\cdot,t)$ solve the equation and let
$L(t)=-\pd_x^2-u(\cdot,t)$. Then $L(t)$ is unitarily equivalent to $L(0)$: its eigenvalues $-\kappa_n^2$ do not depend
on $t$, and its scattering data evolve by $T_t=0$, $R_t=8ik^3R$, $\dot c_n=4\kappa_n^3c_n$.
\end{theorem}
\noindent For the two-soliton potential the two eigenvalues are $-0.6400$ and $-0.2500$, that is $-k_i^2/4$, at
$t=0$ and at $t=1.5$. \chk{X7} The theorem is a second morphism of equations, deeper than the first: the map from
potentials to scattering data of \Cref{thm:transfer} carries the nonlinear flow to a linear one with constant
coefficients, as the Fourier transform does for linear equations. The conserved eigenvalues are the solitons,
and the factors $e^{4\kappa^3t}$ are their motion.
\[\begin{tikzcd}[column sep=7em,row sep=large]
u(\cdot,0)\arrow[r,"\text{scattering}"]\arrow[d,"\text{KdV flow}"'] & (R,\kappa_n,c_n)(0)\arrow[d,"{R\mapsto e^{8ik^3t}R,\ c_n\mapsto e^{4\kappa_n^3t}c_n}"]\\
u(\cdot,t)\arrow[r,"\text{scattering}"'] & (R,\kappa_n,c_n)(t)
\end{tikzcd}\]
\section{Variational problems}
A function that minimizes $\int\mathcal L(x,u,\nabla u)\,dx$ satisfies the Euler--Lagrange equation
$\pd\mathcal L/\pd u=\Div(\pd\mathcal L/\pd\nabla u)$ \src{\S14.3}. Dirichlet's principle is the case
$\mathcal L=\tfrac12|\nabla u|^2$. For the area of a surface of revolution the equation is $uu''=1+u'^2$, solved by the
catenary, with first integral $u/\sqrt{1+u'^2}$. \chk{X8}
\begin{theorem}[Noether]\label{thm:noether}
\Scl[\cite{Noe18,Olv93}] A one-parameter group of symmetries of the action gives a conservation law of the
Euler--Lagrange equation, and conversely.
\end{theorem}
\noindent Time translation gives the energy of \Cref{thm:energy}. The symplectic form of \Cref{prop:symp} is of the
same origin: it is the integral over a time slice of a closed form on space-time built from two variations. In
the jet-space language of \Cref{ch:equations}, conservation laws are cohomology classes of degree $n-1$ in the
horizontal complex of the equation \cite{And92}, and the Euler--Lagrange equation is the critical locus of a
function, which in the presence of gauge symmetry (\Cref{rem:stack}) must be taken in the derived sense
\cite{CoGw17}.
\section{Bifurcation}
\begin{theorem}[Buckling]\label{thm:buckling}
\Scl[source, \S14.4; \cite{CrRa71}] For $u''+\lambda\sin u=0$ on $(0,l)$ with $u'=0$ at both ends, the trivial
solution is the only one for $\lambda\le\pi^2/l^2$. At each $\lambda_n=n^2\pi^2/l^2$ a pair of branches of non-trivial
solutions leaves it, and on the first branch the amplitude $A=u(0)$ satisfies
\[ \lambda/\lambda_1=\bigl(\tfrac2\pi K(\sin\tfrac A2)\bigr)^2=1+\tfrac18A^2+\cdots, \]
with $K$ the complete elliptic integral.
\end{theorem}
\noindent The values of $\lambda/\lambda_1$ at amplitudes $0.2$, $1$, $2$ are $1.0050$, $1.1371$, $1.7660$: the branch
bends to the right, as in the source's figure. \chk{X9} The bifurcation points are the eigenvalues of the
linearization, as they must be by the implicit function theorem; that a branch really leaves each of them follows
from the simplicity of the eigenvalue \cite{CrRa71}, or from the change of the index of the trivial solution,
which is the degree argument of \DT{Ch.~7}. The symmetry $u\mapsto-u$ is why the branches come in pairs.
\begin{center}
\begin{tikzpicture}
\begin{scope}[shift={(0,0)}]
\draw[ink2] (0,4)--(0,0)--(7.4,0);
\draw[ink2] (0,0)--(0,-0.07) node[below,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0.925,0)--(0.925,-0.07) node[below,inner sep=1.5pt]{\scriptsize $1$};
\draw[ink2] (3.7,0)--(3.7,-0.07) node[below,inner sep=1.5pt]{\scriptsize $4$};
\draw[ink2] (7.4,0)--(7.4,-0.07) node[below,inner sep=1.5pt]{\scriptsize $8$};
\draw[ink2] (0,0.067)--(-0.07,0.067) node[left,inner sep=1.5pt]{\scriptsize $-\pi$};
\draw[ink2] (0,2)--(-0.07,2) node[left,inner sep=1.5pt]{\scriptsize $0$};
\draw[ink2] (0,3.933)--(-0.07,3.933) node[left,inner sep=1.5pt]{\scriptsize $\pi$};
\node[below] at (3.7,-0.38) {\scriptsize $\lambda/\lambda_1$};
\node[above,inner sep=1pt] at (0,4.05) {\scriptsize $u(0)$};
\draw[very thick,fwd] (0,2)--(0.925,2);
\draw[thick,fwd,densely dashed] (0.925,2)--(7.4,2);
\draw[very thick,bwd] plot coordinates {(7.22,0.128) (6.752,0.138) (6.356,0.147) (6.016,0.157) (5.718,0.166) (5.455,0.176) (5.22,0.186) (5.008,0.195) (4.816,0.205) (4.64,0.214) (4.479,0.224) (4.33,0.234) (4.192,0.243) (4.064,0.253) (3.944,0.262) (3.832,0.272) (3.727,0.282) (3.628,0.291) (3.535,0.301) (3.447,0.31) (3.363,0.32) (3.284,0.33) (3.209,0.339) (3.138,0.349) (3.069,0.358) (3.004,0.368) (2.942,0.378) (2.883,0.387) (2.826,0.397) (2.771,0.406) (2.719,0.416) (2.669,0.426) (2.62,0.435) (2.574,0.445) (2.529,0.454) (2.486,0.464) (2.444,0.474) (2.404,0.483) (2.366,0.493) (2.328,0.502) (2.292,0.512) (2.257,0.522) (2.223,0.531) (2.19,0.541) (2.158,0.55) (2.128,0.56) (2.098,0.57) (2.069,0.579) (2.041,0.589) (2.013,0.598) (1.987,0.608) (1.961,0.618) (1.936,0.627) (1.911,0.637) (1.888,0.646) (1.865,0.656) (1.842,0.666) (1.82,0.675) (1.799,0.685) (1.778,0.694) (1.758,0.704) (1.738,0.714) (1.719,0.723) (1.701,0.733) (1.682,0.742) (1.664,0.752) (1.647,0.762) (1.63,0.771) (1.614,0.781) (1.597,0.79) (1.582,0.8) (1.566,0.81) (1.551,0.819) (1.536,0.829) (1.522,0.838) (1.508,0.848) (1.494,0.858) (1.481,0.867) (1.467,0.877) (1.455,0.886) (1.442,0.896) (1.43,0.906) (1.418,0.915) (1.406,0.925) (1.394,0.934) (1.383,0.944) (1.372,0.954) (1.361,0.963) (1.351,0.973) (1.34,0.982) (1.33,0.992) (1.32,1.002) (1.31,1.011) (1.301,1.021) (1.292,1.03) (1.282,1.04) (1.273,1.05) (1.265,1.059) (1.256,1.069) (1.248,1.078) (1.239,1.088) (1.231,1.098) (1.223,1.107) (1.216,1.117) (1.208,1.126) (1.2,1.136) (1.193,1.146) (1.186,1.155) (1.179,1.165) (1.172,1.174) (1.165,1.184) (1.159,1.194) (1.152,1.203) (1.146,1.213) (1.14,1.222) (1.134,1.232) (1.128,1.242) (1.122,1.251) (1.116,1.261) (1.11,1.27) (1.105,1.28) (1.1,1.29) (1.094,1.299) (1.089,1.309) (1.084,1.318) (1.079,1.328) (1.074,1.338) (1.069,1.347) (1.065,1.357) (1.06,1.366) (1.056,1.376) (1.051,1.386) (1.047,1.395) (1.043,1.405) (1.039,1.414) (1.035,1.424) (1.031,1.434) (1.027,1.443) (1.023,1.453) (1.02,1.462) (1.016,1.472) (1.013,1.482) (1.009,1.491) (1.006,1.501) (1.003,1.51) (0.999,1.52) (0.996,1.53) (0.993,1.539) (0.99,1.549) (0.987,1.558) (0.985,1.568) (0.982,1.578) (0.979,1.587) (0.977,1.597) (0.974,1.606) (0.972,1.616) (0.969,1.626) (0.967,1.635) (0.965,1.645) (0.963,1.654) (0.96,1.664) (0.958,1.674) (0.956,1.683) (0.954,1.693) (0.953,1.702) (0.951,1.712) (0.949,1.722) (0.947,1.731) (0.946,1.741) (0.944,1.75) (0.943,1.76) (0.941,1.77) (0.94,1.779) (0.939,1.789) (0.938,1.798) (0.936,1.808) (0.935,1.818) (0.934,1.827) (0.933,1.837) (0.932,1.846) (0.931,1.856) (0.931,1.866) (0.93,1.875) (0.929,1.885) (0.928,1.894) (0.928,1.904) (0.927,1.914) (0.927,1.923) (0.926,1.933) (0.926,1.942) (0.926,1.952) (0.925,1.962) (0.925,1.971) (0.925,1.981) (0.925,1.99) (0.925,2) (0.925,2.01) (0.925,2.019) (0.925,2.029) (0.925,2.038) (0.926,2.048) (0.926,2.058) (0.926,2.067) (0.927,2.077) (0.927,2.086) (0.928,2.096) (0.928,2.106) (0.929,2.115) (0.93,2.125) (0.931,2.134) (0.931,2.144) (0.932,2.154) (0.933,2.163) (0.934,2.173) (0.935,2.182) (0.936,2.192) (0.938,2.202) (0.939,2.211) (0.94,2.221) (0.941,2.23) (0.943,2.24) (0.944,2.25) (0.946,2.259) (0.947,2.269) (0.949,2.278) (0.951,2.288) (0.953,2.298) (0.954,2.307) (0.956,2.317) (0.958,2.326) (0.96,2.336) (0.963,2.346) (0.965,2.355) (0.967,2.365) (0.969,2.374) (0.972,2.384) (0.974,2.394) (0.977,2.403) (0.979,2.413) (0.982,2.422) (0.985,2.432) (0.987,2.442) (0.99,2.451) (0.993,2.461) (0.996,2.47) (0.999,2.48) (1.003,2.49) (1.006,2.499) (1.009,2.509) (1.013,2.518) (1.016,2.528) (1.02,2.538) (1.023,2.547) (1.027,2.557) (1.031,2.566) (1.035,2.576) (1.039,2.586) (1.043,2.595) (1.047,2.605) (1.051,2.614) (1.056,2.624) (1.06,2.634) (1.065,2.643) (1.069,2.653) (1.074,2.662) (1.079,2.672) (1.084,2.682) (1.089,2.691) (1.094,2.701) (1.1,2.71) (1.105,2.72) (1.11,2.73) (1.116,2.739) (1.122,2.749) (1.128,2.758) (1.134,2.768) (1.14,2.778) (1.146,2.787) (1.152,2.797) (1.159,2.806) (1.165,2.816) (1.172,2.826) (1.179,2.835) (1.186,2.845) (1.193,2.854) (1.2,2.864) (1.208,2.874) (1.216,2.883) (1.223,2.893) (1.231,2.902) (1.239,2.912) (1.248,2.922) (1.256,2.931) (1.265,2.941) (1.273,2.95) (1.282,2.96) (1.292,2.97) (1.301,2.979) (1.31,2.989) (1.32,2.998) (1.33,3.008) (1.34,3.018) (1.351,3.027) (1.361,3.037) (1.372,3.046) (1.383,3.056) (1.394,3.066) (1.406,3.075) (1.418,3.085) (1.43,3.094) (1.442,3.104) (1.455,3.114) (1.467,3.123) (1.481,3.133) (1.494,3.142) (1.508,3.152) (1.522,3.162) (1.536,3.171) (1.551,3.181) (1.566,3.19) (1.582,3.2) (1.597,3.21) (1.614,3.219) (1.63,3.229) (1.647,3.238) (1.664,3.248) (1.682,3.258) (1.701,3.267) (1.719,3.277) (1.738,3.286) (1.758,3.296) (1.778,3.306) (1.799,3.315) (1.82,3.325) (1.842,3.334) (1.865,3.344) (1.888,3.354) (1.911,3.363) (1.936,3.373) (1.961,3.382) (1.987,3.392) (2.013,3.402) (2.041,3.411) (2.069,3.421) (2.098,3.43) (2.128,3.44) (2.158,3.45) (2.19,3.459) (2.223,3.469) (2.257,3.478) (2.292,3.488) (2.328,3.498) (2.366,3.507) (2.404,3.517) (2.444,3.526) (2.486,3.536) (2.529,3.546) (2.574,3.555) (2.62,3.565) (2.669,3.574) (2.719,3.584) (2.771,3.594) (2.826,3.603) (2.883,3.613) (2.942,3.622) (3.004,3.632) (3.069,3.642) (3.138,3.651) (3.209,3.661) (3.284,3.67) (3.363,3.68) (3.447,3.69) (3.535,3.699) (3.628,3.709) (3.727,3.718) (3.832,3.728) (3.944,3.738) (4.064,3.747) (4.192,3.757) (4.33,3.766) (4.479,3.776) (4.64,3.786) (4.816,3.795) (5.008,3.805) (5.22,3.814) (5.455,3.824) (5.718,3.834) (6.016,3.843) (6.356,3.853) (6.752,3.862) (7.22,3.872)};
\draw[thick,ink2] plot coordinates {(7.369,0.666) (7.281,0.675) (7.196,0.685) (7.113,0.694) (7.033,0.704) (6.954,0.714) (6.877,0.723) (6.802,0.733) (6.729,0.742) (6.658,0.752) (6.588,0.762) (6.52,0.771) (6.454,0.781) (6.389,0.79) (6.326,0.8) (6.264,0.81) (6.204,0.819) (6.145,0.829) (6.088,0.838) (6.031,0.848) (5.976,0.858) (5.923,0.867) (5.87,0.877) (5.819,0.886) (5.768,0.896) (5.719,0.906) (5.671,0.915) (5.624,0.925) (5.577,0.934) (5.532,0.944) (5.488,0.954) (5.445,0.963) (5.403,0.973) (5.361,0.982) (5.32,0.992) (5.281,1.002) (5.242,1.011) (5.204,1.021) (5.166,1.03) (5.13,1.04) (5.094,1.05) (5.059,1.059) (5.024,1.069) (4.99,1.078) (4.957,1.088) (4.925,1.098) (4.893,1.107) (4.862,1.117) (4.832,1.126) (4.802,1.136) (4.773,1.146) (4.744,1.155) (4.716,1.165) (4.688,1.174) (4.661,1.184) (4.635,1.194) (4.609,1.203) (4.584,1.213) (4.559,1.222) (4.534,1.232) (4.511,1.242) (4.487,1.251) (4.464,1.261) (4.442,1.27) (4.42,1.28) (4.398,1.29) (4.377,1.299) (4.356,1.309) (4.336,1.318) (4.316,1.328) (4.297,1.338) (4.278,1.347) (4.259,1.357) (4.241,1.366) (4.223,1.376) (4.205,1.386) (4.188,1.395) (4.172,1.405) (4.155,1.414) (4.139,1.424) (4.123,1.434) (4.108,1.443) (4.093,1.453) (4.078,1.462) (4.064,1.472) (4.05,1.482) (4.036,1.491) (4.023,1.501) (4.01,1.51) (3.997,1.52) (3.985,1.53) (3.973,1.539) (3.961,1.549) (3.95,1.558) (3.938,1.568) (3.927,1.578) (3.917,1.587) (3.906,1.597) (3.896,1.606) (3.887,1.616) (3.877,1.626) (3.868,1.635) (3.859,1.645) (3.85,1.654) (3.842,1.664) (3.833,1.674) (3.826,1.683) (3.818,1.693) (3.81,1.702) (3.803,1.712) (3.796,1.722) (3.79,1.731) (3.783,1.741) (3.777,1.75) (3.771,1.76) (3.766,1.77) (3.76,1.779) (3.755,1.789) (3.75,1.798) (3.745,1.808) (3.741,1.818) (3.737,1.827) (3.733,1.837) (3.729,1.846) (3.725,1.856) (3.722,1.866) (3.719,1.875) (3.716,1.885) (3.714,1.894) (3.711,1.904) (3.709,1.914) (3.707,1.923) (3.706,1.933) (3.704,1.942) (3.703,1.952) (3.702,1.962) (3.701,1.971) (3.7,1.981) (3.7,1.99) (3.7,2) (3.7,2.01) (3.7,2.019) (3.701,2.029) (3.702,2.038) (3.703,2.048) (3.704,2.058) (3.706,2.067) (3.707,2.077) (3.709,2.086) (3.711,2.096) (3.714,2.106) (3.716,2.115) (3.719,2.125) (3.722,2.134) (3.725,2.144) (3.729,2.154) (3.733,2.163) (3.737,2.173) (3.741,2.182) (3.745,2.192) (3.75,2.202) (3.755,2.211) (3.76,2.221) (3.766,2.23) (3.771,2.24) (3.777,2.25) (3.783,2.259) (3.79,2.269) (3.796,2.278) (3.803,2.288) (3.81,2.298) (3.818,2.307) (3.826,2.317) (3.833,2.326) (3.842,2.336) (3.85,2.346) (3.859,2.355) (3.868,2.365) (3.877,2.374) (3.887,2.384) (3.896,2.394) (3.906,2.403) (3.917,2.413) (3.927,2.422) (3.938,2.432) (3.95,2.442) (3.961,2.451) (3.973,2.461) (3.985,2.47) (3.997,2.48) (4.01,2.49) (4.023,2.499) (4.036,2.509) (4.05,2.518) (4.064,2.528) (4.078,2.538) (4.093,2.547) (4.108,2.557) (4.123,2.566) (4.139,2.576) (4.155,2.586) (4.172,2.595) (4.188,2.605) (4.205,2.614) (4.223,2.624) (4.241,2.634) (4.259,2.643) (4.278,2.653) (4.297,2.662) (4.316,2.672) (4.336,2.682) (4.356,2.691) (4.377,2.701) (4.398,2.71) (4.42,2.72) (4.442,2.73) (4.464,2.739) (4.487,2.749) (4.511,2.758) (4.534,2.768) (4.559,2.778) (4.584,2.787) (4.609,2.797) (4.635,2.806) (4.661,2.816) (4.688,2.826) (4.716,2.835) (4.744,2.845) (4.773,2.854) (4.802,2.864) (4.832,2.874) (4.862,2.883) (4.893,2.893) (4.925,2.902) (4.957,2.912) (4.99,2.922) (5.024,2.931) (5.059,2.941) (5.094,2.95) (5.13,2.96) (5.166,2.97) (5.204,2.979) (5.242,2.989) (5.281,2.998) (5.32,3.008) (5.361,3.018) (5.403,3.027) (5.445,3.037) (5.488,3.046) (5.532,3.056) (5.577,3.066) (5.624,3.075) (5.671,3.085) (5.719,3.094) (5.768,3.104) (5.819,3.114) (5.87,3.123) (5.923,3.133) (5.976,3.142) (6.031,3.152) (6.088,3.162) (6.145,3.171) (6.204,3.181) (6.264,3.19) (6.326,3.2) (6.389,3.21) (6.454,3.219) (6.52,3.229) (6.588,3.238) (6.658,3.248) (6.729,3.258) (6.802,3.267) (6.877,3.277) (6.954,3.286) (7.033,3.296) (7.113,3.306) (7.196,3.315) (7.281,3.325) (7.369,3.334)};
\fill[black] (0.93,2.123) circle (1.6pt);
\fill[black] (1.052,2.615) circle (1.6pt);
\fill[black] (1.634,3.231) circle (1.6pt);
\draw[ink2,very thin,densely dotted] (0,3.933)--(7.4,3.933);
\draw[ink2,very thin,densely dotted] (0,0.067)--(7.4,0.067);
\node[bwd] at (5.273,3.415) {\scriptsize first branch};
\node[ink2] at (6.66,2.554) {\scriptsize second};
\end{scope}
\end{tikzpicture}
\figcap{The bifurcation diagram of the buckling problem: the amplitude $u(0)$ against the load. The trivial solution is the only one up to $\lambda_1$ and persists, unstable, beyond it (dashed). The first pair of branches is $\lambda/\lambda_1=(\frac2\pi K(\sin\frac A2))^2$, with the three computed points marked; the second leaves at $4\lambda_1$. All branches bend to the right and approach $\pm\pi$, the rod folded back on itself.}
\end{center}
\section{Water waves}
The motion of water under gravity is a free boundary problem: Laplace's equation for the velocity potential in a
domain whose upper boundary is the unknown \src{\S14.5}. In the formulation of Zakharov and of Craig and Sulem the
unknowns are the surface elevation and the potential on it, and the dynamics is driven by the
Dirichlet-to-Neumann operator of \Cref{cor:dtn} for the moving domain \cite{CrSu93}. Linearizing at still water of
depth $h$ gives the dispersion relation $\omega^2=gk\tanh kh$: waves travel at $\sqrt{gh}$ when they are long and at
$\sqrt{g/k}$ when they are short. \chk{X10} The Korteweg--de Vries equation is the first correction to the
long-wave limit, which is how the soliton was first seen on a canal.
\section*{Notes}
The source's Chapter 14 is followed for the examples. For conservation laws and entropy solutions see
\cite{Lax73,Eva10}; for the Cole--Hopf transformation \cite{Hop50,Col51}; for inverse scattering
\cite{GGKM67,Lax68}; for Noether's theorem and the variational bicomplex \cite{Olv93,And92}; for bifurcation from a
simple eigenvalue \cite{CrRa71}; for the Hamiltonian formulation of water waves \cite{CrSu93}.
\section*{Exercises}
\begin{exercise}\label{exer:break}
Find the breaking time for $u_t+uu_x=0$ with $\varphi=1/(1+x^2)$.
\end{exercise}
\begin{exercise}\label{exer:rh}
Verify the two shock speeds of \Cref{prop:consform}.
\end{exercise}
\begin{exercise}\label{exer:ch}
Prove the first statement of \Cref{thm:colehopf}.
\end{exercise}
\begin{exercise}\label{exer:sol}
Verify that $f=\tfrac c2\sech^2(\tfrac{\sqrt c}2\xi)$ satisfies $-cf+f''+3f^2=0$.
\end{exercise}
\chapter{Assessment}\label{ch:coda}
\section{What the language did}
Eight things, roughly in the order of how much they changed our reading of the source.
\emph{Boundary conditions became objects.} A boundary condition is a character of a reflection group when the
domain is a chamber (\Cref{thm:images}) and a Lagrangian plane of boundary values in general
(\Cref{thm:lagrangian}). The first explains when the method of images works and extends it to diffusion, where
the source says it fails; the second explains the source's ``symmetric'' conditions and puts Dirichlet, Neumann,
Robin and periodic conditions in one three-dimensional family.
\emph{Two methods turned out to be one.} The sum over images and the sum over eigenfunctions are exchanged by
Pontryagin duality (\Cref{thm:kernels}, \Cref{thm:fourier}), and the same exchange computes the heat trace
(\Cref{prop:rect}). The source presents Fourier's method as the replacement for a method that does not
generalize; they are dual.
\emph{The comparison of waves and diffusions was reduced to three lines.} Group or semigroup; symplectic form or
positivity; and the six remaining rows of the source's table follow (\Cref{ch:diffusions}).
\emph{Stability of schemes was identified with structure.} For diffusion it is membership in the Markov category
(\Cref{thm:explicit}); for waves it is respect for the causal structure (\Cref{thm:cfl}); for elliptic problems it is
being a map of complexes (\Cref{thm:feec}).
\emph{Problems could be glued.} Dirichlet-to-Neumann maps, transfer matrices and boundary conditions compose as
Lagrangian relations (\Cref{thm:gluing}, \Cref{thm:transfer}).
\emph{The type was located.} It is invisible to the algebra of formal solutions and lives in the real points of the
characteristic variety (\Cref{rem:type}); it reappears as the singular support of the fundamental solution
(\Cref{thm:fund}) and as the kind of functor that time is.
\emph{Solvable nonlinear equations were explained by morphisms.} Cole--Hopf and the scattering transform are maps
of equations onto linear ones (\Cref{thm:colehopf}, \Cref{thm:ist}).
\emph{Isospectrality became an isomorphism of representations} that is not an isometry (\Cref{thm:sunada}).
\section{What it did not do}
It proved no theorem of analysis. Existence of solutions, in every chapter, rests on estimates that the language
does not supply: the maximum principle, energy inequalities, compactness. The definition of well-posedness as an
isomorphism in a category of topological vector spaces (\Cref{def:wellposed}) says where the difficulty is and
does not reduce it. The machinery cited in \Cref{ch:equations}, the jet comonad and $\DD$-modules, was used only to
say what a morphism of equations is. Microlocal analysis and hyperfunctions, where sheaf theory does real work
for linear equations, are described and not used. The nonlinear chapter treats two integrable examples and one
scalar conservation law; for systems of conservation laws, for the Euler and Navier--Stokes equations and for the
full water wave problem the categorical viewpoint has, so far as we can see, nothing to offer beyond the remarks
made. The observations tagged as such are short and may well be known in other forms.
\section{How much of this is higher}
\begin{center}\small
\begin{tabular}{@{}>{\raggedright}p{0.2\textwidth}>{\raggedright}p{0.4\textwidth}>{\raggedright\arraybackslash}p{0.31\textwidth}@{}}
\toprule
Source & Structure used here & Level\\
\midrule
Ch.~1 & solution functor; symbol; quotient by characteristics & $1$-categorical\\
Ch.~2--3 & group and Markov semigroup; reflection groups; causal functor & $1$-categorical\\
Ch.~4--5 & Lagrangian planes; Pontryagin duality & $1$-categorical; relations compose\\
Ch.~6--7 & sheaf of harmonic functions; Lagrangian relations & cohomology in degree $1$; a category of cospans\\
Ch.~8 & commuting squares; cochain maps & $1$-categorical; chain level\\
Ch.~9 & causal functor; descent; a holomorphic semigroup & $1$-categorical; microlocal sheaves cited\\
Ch.~10 & representations of symmetry groups & $1$-categorical\\
Ch.~11 & monotone functors; trace of a semigroup; permutation representations & $1$-categorical\\
Ch.~12 & sheaf of distributions; hyperfunctions & local cohomology\\
Ch.~13 & gauge groupoids; functorial field theory & stacks; derived solution spaces\\
Ch.~14 & morphisms of equations; conservation laws & $1$-categorical; horizontal cohomology\\
\bottomrule
\end{tabular}
\end{center}
\noindent The subject at the level of the source is ordinary category theory with a good deal of representation
theory. Higher structure enters at four points, and only the last is unavoidable. Cohomology in degree one appears
for the annulus, for comparison of discretizations, and for flat gauge fields. Networks and domains with boundary
compose as cospans, which form a bicategory or double category when the gluing is tracked up to isomorphism.
Hyperfunctions and the solution complexes of $\DD$-modules are derived objects; the cokernel in
\Cref{ex:fredholm} is the first $\operatorname{Ext}$ group of the theory. And gauge fields form stacks, so that
for the equations of \Cref{ch:fields} the solution functor does not take values in sets.
\section{Programme}
\Spg (1) Characterize the boundary-value problems solvable by images as those on quotients of space by reflection
groupoids, including orbifold covers, and decide whether mixed and Robin conditions have a place there. (2)
Formulate the equivalence theorem of \Cref{thm:lax} as a statement about lax natural transformations between
functors on a category of grids, with stability as the condition for the colimit to exist. (3) Extend
\Cref{thm:gluing} from networks to domains with geometry, as a functor from a category of cobordisms to Lagrangian
relations, and relate it to the gluing formulas for determinants of Laplacians. (4) Find a universal property that
selects entropy solutions among weak solutions of a balance law, as \Cref{thm:viscosity} selects them by a limit.
\section*{Notes}
The companion paper \cite{PDEML} applies the framework to machine learning: padding as a boundary condition, the
score in diffusion models as the datum that makes the backward heat equation well-posed, the spectral gap as the
rate of oversmoothing in graph networks, and the type of an equation as the rank of its solution operator. The
volume belongs to a series with \cite{GIHC,ESCHC,DTHC,EGHC}.
\appendix
\chapter{Toolkit}\label{app:toolkit}
General notions used in the text, in the order in which they are needed.
\paragraph{Sheaves.} A sheaf on a space assigns a set, or vector space, to each open subset, with restriction maps,
such that compatible local sections glue uniquely. Solutions of a differential equation form a sheaf, as do smooth
functions and distributions. Its cohomology in degree one measures the failure of local solutions of an auxiliary
problem to be global.
\paragraph{Modules and solutions.} For a ring $\DD$ of differential operators and a left ideal generated by $P$,
maps of left modules $\DD/\DD P\to\FF$ are the elements of $\FF$ killed by $P$. Exactness properties of
$\Hom_\DD(-,\FF)$ turn factorizations of operators into relations between solution spaces.
\paragraph{Quadratic forms.} A real quadratic form has an inertia: the numbers of positive, negative and zero
squares in any diagonalization. It is invariant under change of basis (Sylvester).
\paragraph{Groups, actions, characters.} A group acting on a space acts on functions by $(g\cdot u)(x)=u(g^{-1}x)$.
For a homomorphism $\chi$ to $\{\pm1\}$ the functions with $g\cdot u=\chi(g)u$ form the $\chi$-isotypic part. A
representation is irreducible if it has no invariant subspace; an operator commuting with the group preserves
isotypic parts and acts on each irreducible one by a scalar (Schur).
\paragraph{Semigroups and Markov kernels.} A one-parameter semigroup is a family $T_t$, $t\ge0$, with
$T_{t+s}=T_tT_s$. A Markov kernel assigns to each point a probability measure, measurably; kernels compose, and they
are the morphisms of the category $\Stoch$. A semigroup of Markov kernels is positive and preserves constants.
\paragraph{Symplectic spaces and Lagrangian relations.} A symplectic form is a non-degenerate antisymmetric bilinear
form; a subspace is Lagrangian if the form vanishes on it and it has half the dimension. A Lagrangian relation
from $V$ to $W$ is a Lagrangian subspace of $\overline V\oplus W$, with the sign of the form reversed on $V$; graphs
of symplectic maps are examples, and relations compose.
\paragraph{Duality of abelian groups.} The characters of a locally compact abelian group form its dual group; the
dual of the circle is $\Z$, the line is self-dual, and the dual of a quotient is a subgroup. The Fourier transform
is the isomorphism of $L^2$ of a group with $L^2$ of its dual.
\paragraph{Complexes.} A complex is a sequence of maps with $\dd\dd=0$; its cohomology is kernel modulo image. The
de Rham complex of forms and the cochain complex of a triangulation have the same cohomology, and a map of
complexes induces a map on cohomology.
\paragraph{Groupoids and stacks.} A groupoid is a category in which every morphism is invertible. A stack assigns a
groupoid to each open set, with gluing up to specified isomorphism; fields with gauge symmetry form stacks.
\chapter{Concordance with the Source}\label{app:concordance}
The table lists the sections of Walter A.~Strauss, \emph{Partial Differential Equations: An Introduction}, second
edition \cite{Str08}, with the place in this volume where their content is treated. The source was read from a text
extraction of its PDF: the contents, every statement labelled as a theorem, definition or principle, located by
search, and selected passages of every chapter. It was not read in full, its exercises were consulted only where
the text refers to them, and figures were seen only through the surrounding text, with one page inspected as an
image to confirm a sign.
{\small
\begin{longtable}{@{}>{\raggedright}p{0.34\textwidth}>{\raggedright}p{0.25\textwidth}>{\raggedright\arraybackslash}p{0.32\textwidth}@{}}
\toprule
Source & Here & Form\\
\midrule
\endhead
1.1 What is a PDE? & \Cref{def:pde}, \Cref{thm:jet} & jets; a category of equations\\
1.2 First-order linear equations & \Cref{prop:char} & quotient by the characteristics\\
1.3 Flows, vibrations, diffusions & \Cref{prop:tonti} & balance law plus material law\\
1.4 Initial and boundary conditions & \Cref{ch:archetypes} & data as a restriction map\\
1.5 Well-posed problems & \Cref{def:wellposed}, \Cref{ex:hadamard} & an isomorphism of topological vector spaces\\
1.6 Types of second-order equations & \Cref{thm:type}, \Cref{rem:type} & inertia of the symbol\\
2.1 The wave equation & \Cref{thm:dalembert}, \Cref{cor:par} & two pullbacks; parallelogram identity\\
2.2 Causality and energy & \Cref{thm:causal}, \Cref{prop:symp} & functor on intervals; symplectic form\\
2.3 The diffusion equation & \Cref{thm:max}, \Cref{thm:energy} & as stated\\
2.4 Diffusion on the whole line & \Cref{thm:semigroup}, \Cref{prop:gauss} & Markov semigroup; scaling\\
2.5 Comparison of waves and diffusions & \Cref{ch:diffusions} & three structural rows\\
3.1 Diffusion on the half-line & \Cref{thm:kernels} & images; survival probability\\
3.2 Reflections of waves & \Cref{thm:images}, \Cref{rem:reflection} & characters of reflection groups\\
3.3 Diffusion with a source & \Cref{thm:duhamel-heat} & variation of constants\\
3.4 Waves with a source & \Cref{thm:duhamel-wave}, \Cref{thm:bgp} & Green operators; exact sequence\\
3.5 Diffusion revisited & \Cref{thm:semigroup}, \Cref{thm:mono} & smoothing; monotone quantities\\
4.1--4.2 Separation of variables & \Cref{ch:fourier} & diagonalizing the generator\\
4.3 The Robin condition & \Cref{thm:lagrangian} & a line in the plane of end values\\
5.1--5.2 Coefficients; even, odd, periodic & \Cref{thm:pontryagin} & duality; isotypic components\\
5.3 Orthogonality & \Cref{thm:lagrangian} & Lagrangian planes\\
5.4--5.5 Completeness; Gibbs & \Cref{thm:complete} & isomorphism of Hilbert spaces only\\
5.6 Inhomogeneous boundary conditions & \Cref{ch:fourier} & affine space of solutions\\
6.1 Laplace's equation & \Cref{thm:mvp}, \Cref{thm:conformal} & sheaf; naturality\\
6.2 Rectangles and cubes & \Cref{thm:product} & tensor products\\
6.3 Poisson's formula & \Cref{thm:poisson}, \Cref{thm:kakutani} & Markov kernel to the boundary\\
6.4 Circles, wedges, annuli & \Cref{prop:annulus}, \Cref{prop:wedge} & first cohomology; corner exponents\\
7.1--7.2 Green's identities & \Cref{prop:green}, \Cref{thm:dirichlet}, \Cref{cor:dtn} & Stokes; projection; Lagrangian graph\\
7.3 Green's functions & \Cref{thm:greenfn} & kernel of the inverse\\
7.4 Half-space and sphere & \Cref{thm:images-green} & reflection; inversion\\
8.1 Opportunities and dangers & \Cref{thm:lax} & squares that almost commute\\
8.2 Approximations of diffusions & \Cref{thm:explicit} & stability as positivity\\
8.3 Approximations of waves & \Cref{thm:cfl} & stability as causality\\
8.4 Approximations of Laplace's equation & \Cref{prop:iter} & iterations and their rates\\
8.5 Finite element method & \Cref{ch:numerics}, \Cref{thm:feec} & energy projection; complexes\\
9.1 Energy and causality & \Cref{thm:cone} & cones\\
9.2 The wave equation in space-time & \Cref{thm:kirchhoff}, \Cref{prop:descent} & Huygens; descent\\
9.3 Rays, singularities, sources & \Cref{thm:wf} & wave front sets\\
9.4 Diffusion and Schr\"odinger & \Cref{thm:wick} & one holomorphic semigroup\\
9.5 The hydrogen atom & \Cref{thm:hydrogen}, \Cref{rem:so4} & hidden symmetry\\
10.1 Fourier's method, revisited & \Cref{thm:rep}, \Cref{thm:product} & eigenspaces as representations\\
10.2, 10.5 Drumhead; Bessel functions & \Cref{thm:drum}, \Cref{rem:bessel} & characters of the rotation group\\
10.3, 10.6 Ball; Legendre functions & \Cref{thm:sphharm} & harmonic polynomials\\
10.4 Nodes & \Cref{thm:courant} & as stated\\
10.7 Angular momentum & \Cref{ch:symmetry} & Casimir element\\
11.1--11.2 Minima; computation & \Cref{thm:minimax} & minimax\\
11.3--11.5 Completeness & \Cref{thm:spectral} & functional calculus\\
11.6 Asymptotics & \Cref{cor:mono}, \Cref{thm:weyl} & monotone functor; Weyl's law\\
12.1 Distributions & \Cref{ch:distributions}, \Cref{thm:kernel} & change of solution space; kernels\\
12.2 Green's functions, revisited & \Cref{thm:fund} & fundamental solutions\\
12.3 Fourier transforms & \Cref{thm:fourier} & duality of the line\\
12.4 Source functions & \Cref{ch:distributions} & Poisson semigroup\\
12.5 Laplace transform techniques & \Cref{prop:resolvent} & resolvent\\
13.1 Electromagnetism & \Cref{prop:gauge}, \Cref{rem:stack} & forms; gauge groupoid\\
13.2 Fluids and acoustics & \Cref{ch:fields} & linearization\\
13.3--13.4 Scattering; continuous spectrum & \Cref{thm:transfer} & transfer matrices compose\\
13.5 Equations of elementary particles & \Cref{prop:dirac}, \Cref{thm:lcqft} & Clifford relations; field functor\\
14.1 Shock waves & \Cref{prop:break}, \Cref{prop:consform}, \Cref{thm:colehopf} & weak solutions; a morphism of equations\\
14.2 Solitons & \Cref{thm:ist} & isospectral flow\\
14.3 Calculus of variations & \Cref{thm:noether} & conservation laws\\
14.4 Bifurcation theory & \Cref{thm:buckling} & as stated, with the branch computed\\
14.5 Water waves & \Cref{ch:nonlinear} & dispersion relation only\\
Appendix A.1--A.5 & --- & assumed\\
--- & \Cref{thm:gluing} & added: gluing of boundary problems\\
--- & \Cref{thm:heattrace}, \Cref{prop:rect}, \Cref{thm:sunada} & added: heat trace; isospectrality\\
\bottomrule
\end{longtable}}
\chapter{Computations}\label{app:computations}
The script \texttt{verify-pde.py} that accompanies this volume uses only the Python standard library, fixes its
random seed, and runs in under ten seconds. It performs 107 checks, each printing \texttt{PASS} or \texttt{FAIL}
with a one-sentence statement of what is checked and the computed values, and exits with status zero if and only
if all pass. The checks recompute the numerical statements of the text and test finite or numerical instances of
the structural ones. They are evidence that examples are stated correctly; they do not prove the theorems.
Polynomial identities and ranks are computed in exact rational arithmetic; integrals by Simpson's rule;
derivatives by fourth-order differences; eigenvalues by Jacobi rotations or Sturm bisection.
The curves in the figures were computed from the formulas of the text by a separate plotting script and are
embedded in the source as coordinates. Where a figure shows a number that the text states (the Gibbs peak, the
two-soliton amplitude, the points on the buckling branch), the plotting script asserts agreement with the value
checked here before drawing it.
{\small
\begin{longtable}{@{}p{0.09\textwidth}p{0.11\textwidth}>{\raggedright\arraybackslash}p{0.71\textwidth}@{}}
\toprule
Checks & Chapter & What is computed\\
\midrule
\endhead
E1--E6 & \ref{ch:equations} & invariance of the type under changes of variables; the source's variable-coefficient example; solutions by characteristics; dimensions of polynomial solutions for three types; factorization of the wave operator; heat polynomials and Kovalevskaya's divergent series\\
W1--W6 & \ref{ch:archetypes} & Hadamard's example with sizes; backward amplification; the Laplacian as $D^{\T}D$; the two energy laws; the Neumann alternative; a boundary problem with a kernel\\
V1--V8 & \ref{ch:waves} & d'Alembert's formula; the parallelogram identity; causality; energy; reflection on the half-line; the sum over the dihedral group against the Fourier series; Duhamel's formula; the symplectic form\\
D1--D9 & \ref{ch:diffusions} & the kernel as a Markov semigroup; the diffused step; the explicit scheme as a Markov matrix; entropy and square norm; half-line kernels; the two expansions on an interval with term counts; scaling; positivity at a distance; Duhamel's formula\\
F1--F8 & \ref{ch:fourier} & orthogonality; Robin eigenvalues and the zero criterion; Lagrangian planes; Parseval; the Gibbs constant; decay of coefficients; characters of $\Z/n$ and their even and odd parts; shifted data\\
H1--H7 & \ref{ch:harmonic} & mean values; the discrete maximum principle; Poisson's formula; naturality for rotations and conformal maps; periods on the annulus; the corner exponent; hitting probabilities\\
G1--G8 & \ref{ch:green} & summation by parts; Dirichlet's principle; the discrete Green's function; half-space and ball by images; gluing of networks; Galerkin orthogonality; the kernel on an interval\\
S1--S5 & \ref{ch:space} & radial and plane waves, the eikonal condition; Huygens in three dimensions and the tail in two; the Schr\"odinger kernel and a spreading packet; hydrogen; the oscillator\\
Y1--Y6 & \ref{ch:symmetry} & harmonic polynomials and the rotation algebra; Legendre polynomials; Bessel zeros; spectra of products; nodal counts; multiplicities forced by symmetry\\
L1--L9 & \ref{ch:eigen} & Rayleigh--Ritz bounds; interlacing; monotonicity and its failure for Neumann conditions; Weyl's law; the heat trace of a rectangle; a trace formula; cospectral graphs by exhaustive search; the Sunada pair of order 32; completeness\\
T1--T7 & \ref{ch:distributions} & the delta function as a limit; derivatives by duality; the fundamental solution in space; the Fourier transform; the source function from the transform; resolvent as Laplace transform; boundary values\\
N1--N10 & \ref{ch:numerics} & instability at $s=0.52$; orders two and four; Crank--Nicolson; the exact wave scheme and its instability; causality; three iterations; finite elements in one and two dimensions; Betti numbers of a triangulated annulus; a consistent unstable scheme\\
P1--P8 & \ref{ch:fields} & the cubical $3$-torus and gauge invariance; a plane electromagnetic wave; the sound speed; the string of two materials; transfer matrices; Klein--Gordon; Dirac matrices; a symplectic integrator\\
X1--X10 & \ref{ch:nonlinear} & breaking time; the source's explicit solution; conservative and non-conservative schemes; Cole--Hopf; one and two solitons with phase shifts; isospectrality; the catenary; the buckling branch; water waves\\
\bottomrule
\end{longtable}}
\noindent Three points of contact with the source's own numbers and statements were found. Its Gibbs constant
``0.59'' and ``9 percent'' are $0.58949$ and $8.95$ per cent (F5). Its remark that the reflection method does not work
for diffusion is true of the formulas it gives and not of the sum over the reflection group, which agrees with
the eigenfunction series to $10^{-10}$ (D6). And the sign of the continuity equation in the text of its Section
13.1 differs from that in its Exercise 1 of the same section, which is the correct one.
\chapter{Solutions to the Exercises}\label{app:solutions}
Solutions to all exercises follow, by chapter.
\section*{Chapter 1}
\sol{1.1} With $d=a_{12}^2-a_{11}a_{22}$: $d=4-4=0$, parabolic; $d=1-5<0$, elliptic; $a_{11}=a_{22}=0$, $a_{12}=\tfrac12$,
$d>0$, hyperbolic.
\sol{1.2} Along a characteristic $dy/dx=2xy^2$, so $-1/y=x^2+C$. Take $q=x^2+1/y$; the general solution is
$u=f(x^2+1/y)$ on each of $y>0$ and $y<0$.
\sol{1.3} The real and imaginary parts of $(x+iy)^k$ are harmonic and independent. For $u_{xx}=0$ take $y^k$ and
$xy^{k-1}$. In both cases the space of all homogeneous polynomials has dimension $k+1$ and the operator maps it onto
the space of degree $k-2$, of dimension $k-1$.
\sol{1.4} $\Sol(P)$ consists of the functions $h(x+ct)$, and $Q\,h(x+ct)=2c\,h'(x+ct)$. Every smooth function of one
variable is a derivative, so $Q$ maps $\Sol(P)$ onto itself and \Cref{prop:factor} applies.
\section*{Chapter 2}
\sol{2.1} The functions $\sin nx\sin nt$ solve the equation and vanish on the whole boundary of the square.
\sol{2.2} $D^{\T}D$ is symmetric, so its range is the orthogonal complement of its kernel. If $D^{\T}Dv=0$ then
$|Dv|^2=0$, so $v$ is constant along every edge and, the graph being connected, constant.
\sol{2.3} The error in mode $n$ is $10^{-10}e^{0.1n^2}$, which exceeds $1$ when $n^2>10\ln10^{10}=230.3$: the mode
$n=16$.
\sol{2.4} $\frac d{dt}\frac12\int u^2=k\int uu_{xx}=k[uu_x]_0^l-k\int u_x^2=-ka_lu(l)^2-ka_0u(0)^2-k\int u_x^2$. It is
non-positive for all solutions when $a_0,a_l\ge0$.
\section*{Chapter 3}
\sol{3.1} In coordinates $(\xi,\eta)=(x+ct,x-ct)$ the vertices are $(\xi,\eta)$, $(\xi+h,\eta)$, $(\xi+h,\eta+k)$,
$(\xi,\eta+k)$, and both sums equal $f(\xi)+f(\xi+h)+g(\eta)+g(\eta+k)$.
\sol{3.2} With the even extension, $v(x,t)=\tfrac12[\varphi(ct+x)+\varphi(ct-x)]$ for $0<x<ct$.
\sol{3.3} The character is $-1$ on the reflection in $0$ and $+1$ on the reflection in $l$, hence $-1$ on their
product, the translation by $2l$. So the extended solution satisfies $u(x+2l,t)=-u(x,t)$ and has period $4l$ in $x$;
by \Cref{thm:dalembert} the functions $f$ and $g$ have period $4l$, and $u$ has period $4l/c$ in $t$.
\sol{3.4} $\int_{x-c(t-s)}^{x+c(t-s)}\cos y\,dy=2\cos x\sin c(t-s)$, and
$\frac1{2c}\int_0^t2\cos x\sin c(t-s)\,ds=\cos x\,(1-\cos ct)/c^2$.
\section*{Chapter 4}
\sol{4.1} With $\xi=x/\sqrt t$, the equation becomes $-\tfrac12g-\tfrac12\xi g'=kg''$, that is
$(kg'+\tfrac12\xi g)'=0$. For a solution decaying at infinity $kg'=-\tfrac12\xi g$ and $g=Ce^{-\xi^2/4k}$.
\sol{4.2} By \Cref{thm:kernels} it is $\erf(x/\sqrt{4kt})=\erf(\tfrac12)=0.5205$.
\sol{4.3} (a) $\frac d{dt}\int u^2=2k\int uu_{xx}=-2k\int u_x^2$. (b) Since $\int u_t=0$,
$\frac d{dt}(-\int u\log u)=-\int u_t\log u=-k\int u_{xx}\log u=k\int u_x^2/u$.
\sol{4.4} The $m$-th shell of images is at distance at least $(2m-1)l$, so it is negligible when
$e^{-(2m-1)^2l^2/4kt}<\epsilon$: about $\tfrac12+\sqrt{kt\ln(1/\epsilon)}/l$ shells. The $n$-th mode is negligible when
$e^{-kn^2\pi^2t/l^2}<\epsilon$: about $(l/\pi)\sqrt{\ln(1/\epsilon)/kt}$ modes. For the table these give $0.6$ and
$58$, $1.4$ and $8.2$, $4.5$ and $1.8$.
\section*{Chapter 5}
\sol{5.1} With $\omega(u,v)=-u_4v_3+u_3v_4+u_2v_1-u_1v_2$: for the Robin pair every term vanishes; for $(1,0,1,0)$ and
$(0,2,0,1)$ it is $1-2=-1$.
\sol{5.2} $X=\sin((n+\tfrac12)\pi x/l)$, $\lambda=((n+\tfrac12)\pi/l)^2$, $n\ge0$. All frequencies are odd multiples
of $\pi c/2l$, so every motion has period $4l/c$.
\sol{5.3} For $\lambda=0$, $X=A+Bx$. The conditions give $B=a_0A$ and $B=-a_l(A+Bl)$, which have a non-zero solution
exactly when $a_0=-a_l(1+a_0l)$.
\sol{5.4} $(1.85194/\pi-\tfrac12)J=0.0895\,J$ on each side of the jump.
\section*{Chapter 6}
\sol{6.1} Expand $u$ about the centre. The linear terms average to zero and the quadratic terms
$\tfrac{r^2}2(u_{xx}\cos^2\theta+2u_{xy}\cos\theta\sin\theta+u_{yy}\sin^2\theta)$ average to $\tfrac{r^2}4(u_{xx}+u_{yy})$.
\sol{6.2} $u\circ f=(x^2-y^2)^2$, whose Laplacian is $(12x^2-4y^2)+(12y^2-4x^2)=8(x^2+y^2)$; and
$|f'|^2\Lap u=4(x^2+y^2)\cdot2$.
\sol{6.3} For $\log r$, $\star\,\dd u=(-y\,dx+x\,dy)/r^2=\dd\theta$, with period $2\pi$. The function
$\mathrm{Re}(1/z)$ has the single-valued conjugate $\mathrm{Im}(1/z)$, so its period is $0$.
\sol{6.4} At the re-entrant corner $\beta=3\pi/2$: the exponent is $\tfrac23$ and the gradient grows like $r^{-1/3}$. At
a convex corner $\beta=\pi/2$: the exponent is $2$ and the gradient vanishes like $r$.
\section*{Chapter 7}
\sol{7.1} For harmonic $u,v$ with boundary values $h,h'$, Green's second identity gives
$\int_{\pd D}(h\,\Lambda h'-h'\,\Lambda h)=\int_D(u\Lap v-v\Lap u)=0$.
\sol{7.2} Eliminating the middle vertex from the Laplacian with rows $(g_1,-g_1,0)$, $(-g_1,g_1+g_2,-g_2)$,
$(0,-g_2,g_2)$ gives the conductance $g_1-g_1^2/(g_1+g_2)=g_1g_2/(g_1+g_2)$.
\sol{7.3} The kernel is linear in $x$ on each side of $y$, vanishes at $0$ and $1$, is continuous, and its derivative
drops by $1$ at $x=y$: $x(1-y)$ for $x<y$ and $y(1-x)$ for $x>y$. In eigenfunctions,
$\sum_n2\sin(n\pi x)\sin(n\pi y)/(n\pi)^2$.
\sol{7.4} With $r_0=|x_0|$ and $|x|=a$:
$|x-x_0^*|^2=a^2-2\tfrac{a^2}{r_0^2}x\cdot x_0+\tfrac{a^4}{r_0^2}=\tfrac{a^2}{r_0^2}(r_0^2-2x\cdot x_0+a^2)=\tfrac{a^2}{r_0^2}|x-x_0|^2$.
The two terms of the Green's function then cancel.
\section*{Chapter 8}
\sol{8.1} For radial $u$, $\Lap u=u_{rr}+\tfrac2ru_r=(ru)_{rr}/r$, so $(ru)_{tt}=c^2(ru)_{rr}$ and \Cref{thm:dalembert}
applies to $ru$.
\sol{8.2} On the sphere of radius $R$ the area element over the plane element $\rho\,d\rho\,d\theta$ is
$R\rho\,d\rho\,d\theta/\sqrt{R^2-\rho^2}$ on each hemisphere. Dividing twice this by $4\pi R^2$ gives the density.
With $R=ct$, $\frac1{4\pi c^2t}\int_{\text{sphere}}\psi\,dS=\frac1{2\pi c}\int_{\text{disk}}\psi\,dy/\sqrt{c^2t^2-\rho^2}$.
\sol{8.3} With $a=1+it$: $u_t=i\bigl(-\tfrac1{2a}+\tfrac{x^2}{2a^2}\bigr)u$ and $u_{xx}=\bigl(-\tfrac1a+\tfrac{x^2}{a^2}\bigr)u$,
so $iu_t=-\tfrac12u_{xx}$. Then $|u|^2=|a|^{-1}e^{-x^2\mathrm{Re}(1/a)}$ with $\mathrm{Re}(1/a)=1/(1+t^2)$, a Gaussian of
variance $(1+t^2)/2$.
\sol{8.4} $a_1=-\tfrac23$, $3a_2=(\tfrac23-1)a_1=\tfrac29$ so $a_2=\tfrac2{27}$, and $a_3=0$. The states are those with
$l=0,1,2$: $1+3+5=9$.
\section*{Chapter 9}
\sol{9.1} The eigenspace is spanned by $\sin x\sin2y$ and $\sin2x\sin y$. Reflection in the diagonal exchanges them
and the quarter turn about the centre sends the first to minus the second: the two-dimensional irreducible
representation. For $50=1+49=49+1=25+25$ the multiplicity is $3$. The function $\sin5x\sin5y$ spans an invariant
line, and $\sin x\sin7y\pm\sin7x\sin y$ span two more: the eigenspace is a sum of three one-dimensional
representations, and its multiplicity is arithmetic and not forced by symmetry.
\sol{9.2} $\binom{n+l-1}{l}-\binom{n+l-3}{l-2}$: $2$ for $n=2$, $2l+1$ for $n=3$, $(l+1)^2$ for $n=4$.
\sol{9.3} $e^{iy}$ satisfies $\Lap u=-u$. Projection onto the $n$-th character of the rotation group commutes with
$\Lap$, so $J_n(r)e^{in\theta}$ satisfies the same equation, which in polar coordinates is
$J_n''+\tfrac1rJ_n'+(1-\tfrac{n^2}{r^2})J_n=0$.
\sol{9.4} $cz/2\pi a$ with $z=2.4048$ (simple), $3.8317$ (double), $5.1356$ (double), $5.5201$ (simple).
\section*{Chapter 10}
\sol{10.1} $\int_0^1(1-2x)^2=\tfrac13$ and $\int_0^1x^2(1-x)^2=\tfrac1{30}$; the quotient is $10$.
\sol{10.2} The inscribed square has side $\sqrt2$ and first eigenvalue $2\pi^2/2=9.87$; the circumscribed one has
side $2$ and first eigenvalue $2\pi^2/4=4.93$. So $4.93\le\lambda_1\le9.87$; in fact $\lambda_1=2.4048^2=5.78$.
\sol{10.3} The eigenvalues are $\pi^2(m^2/a^2+n^2/b^2)$ with $m,n\ge1$. The lattice points with $m,n\ge0$ in the
quarter ellipse number about its area, $ab\lambda/4\pi$; those on the two axes, about
$(a+b)\sqrt\lambda/\pi$, are counted with weight one half too many each, giving the correction
$-(a+b)\sqrt\lambda/2\pi=-P\sqrt\lambda/4\pi$.
\sol{10.4} Rectangle: $4\cdot(\pi^2-\pi^2/4)/(24\pi\cdot\pi/2)=\tfrac14$. Equilateral triangle:
$3\cdot(\pi^2-\pi^2/9)/(24\pi\cdot\pi/3)=\tfrac13$.
\section*{Chapter 11}
\sol{11.1} $\int|x|\varphi''=\int_0^\infty x\varphi''-\int_{-\infty}^0x\varphi''=[x\varphi'-\varphi]_0^\infty-[x\varphi'-\varphi]_{-\infty}^0=2\varphi(0)$.
\sol{11.2} $\tfrac12|x|$, by the previous exercise; and $\tfrac12e^{-|x|}$, which satisfies $E''=E$ away from $0$ and
whose derivative drops by $1$ at $0$.
\sol{11.3} $S=\frac1{2\pi}\int e^{-k^2t+ikx}dk$; completing the square, $-t(k-ix/2t)^2-x^2/4t$, gives
$(4\pi t)^{-1/2}e^{-x^2/4t}$.
\sol{11.4} The transform of $P_y$ is $e^{-|k|y}$, and $e^{-|k|y}e^{-|k|y'}=e^{-|k|(y+y')}$.
\section*{Chapter 12}
\sol{12.1} Substituting $u_j^n=\xi^ne^{ijk\Delta x}$ gives $\xi=1-2s(1-\cos k\Delta x)\in[1-4s,1]$, of modulus at most
one for all $k$ exactly when $s\le\tfrac12$.
\sol{12.2} The scheme equals $u_t+\tfrac{\Delta t}2u_{tt}-u_{xx}-\tfrac{\Delta x^2}{12}u_{xxxx}$ up to higher order.
For solutions $u_{tt}=u_{xxxx}$, so the leading error is $(\tfrac{\Delta t}2-\tfrac{\Delta x^2}{12})u_{xxxx}$, which
vanishes when $\Delta t=\Delta x^2/6$.
\sol{12.3} The product of the roots is $1$. They are complex conjugate, hence of modulus one, when $(1+p)^2\le1$,
that is $-2\le p\le0$; otherwise real and distinct with one of modulus greater than one. Since $p$ ranges over
$[-2s,0]$, the condition is $s\le1$.
\sol{12.4} $12-24+12=0=b_0-b_1+b_2$ with $b_0=1$ and $b_2=0$, so $b_1=1$.
\section*{Chapter 13}
\sol{13.1} $\pd_t\rho=\tfrac1{4\pi}\Div\pd_tE=\tfrac1{4\pi}\Div(c\,\curl B-4\pi J)=-\Div J$.
\sol{13.2} $R^2+\tfrac{c_1}{c_2}T^2=\bigl((c_2-c_1)^2+4c_1c_2\bigr)/(c_1+c_2)^2=1$.
\sol{13.3} $\psi$ is continuous and $\psi'$ jumps by $g\psi(0)$: the matrix is $\begin{psmallmatrix}1&0\\g&1\end{psmallmatrix}$.
From $1+R=T$ and $g(1+R)+ik(1-R)=ikT$, $T=2ik/(2ik-g)$ and $|T|^2=4k^2/(4k^2+g^2)$.
\sol{13.4} $\omega\,d\omega=c^2k\,dk$, so $\frac{d\omega}{dk}\cdot\frac\omega k=c^2$.
\section*{Chapter 14}
\sol{14.1} $\varphi'=-2x/(1+x^2)^2$ is least at $x=1/\sqrt3$, where it equals $-9/8\sqrt3$; $t_b=8\sqrt3/9=1.5396$.
\sol{14.2} $[\tfrac12u^2]/[u]=\tfrac12$ and $[\tfrac23u^3]/[u^2]=\tfrac23$ for the jump from $1$ to $0$.
\sol{14.3} Let $w=\log\varphi$, so $w_t=\eps(w_{xx}+w_x^2)$ and $u=-2\eps w_x$. Then
$u_t=-2\eps^2(w_{xxx}+2w_xw_{xx})=\eps u_{xx}-uu_x$.
\sol{14.4} With $\theta=\tfrac{\sqrt c}2\xi$, $(\sech^2)''=4\sech^2-6\sech^4$, so
$f''=\tfrac{c^2}2\sech^2\theta-\tfrac{3c^2}4\sech^4\theta$, while $-cf=-\tfrac{c^2}2\sech^2\theta$ and
$3f^2=\tfrac{3c^2}4\sech^4\theta$.
\backmatter
\chapter{References}
\begingroup\small
\renewcommand{\chapter}[2]{}
\noindent Entries were assembled from memory, except \cite{KhSc17} and \cite{GKSS26}, whose authors, titles and
identifiers were checked online on 17 November 2026. Volume and page numbers in particular should be verified
before they are relied on.
\begin{thebibliography}{GGKM67}\setlength{\itemsep}{1pt}
\newcommand{\bi}[1]{\bibitem[#1]{#1}}
\bi{ABR01} S.~Axler, P.~Bourdon, W.~Ramey, \emph{Harmonic Function Theory}, 2nd ed., Graduate Texts in Mathematics 137, Springer, 2001.
\bi{AFW06} D.~N.~Arnold, R.~S.~Falk, R.~Winther, Finite element exterior calculus, homological techniques, and applications, \emph{Acta Numer.} 15 (2006), 1--155.
\bi{And92} I.~M.~Anderson, Introduction to the variational bicomplex, in \emph{Mathematical Aspects of Classical Field Theory}, Contemp.\ Math.\ 132, American Mathematical Society, 1992, 51--73.
\bi{Arn66} V.~I.~Arnold, Sur la g\'eom\'etrie diff\'erentielle des groupes de Lie de dimension infinie et ses applications \`a l'hydrodynamique des fluides parfaits, \emph{Ann.\ Inst.\ Fourier} 16 (1966), 319--361.
\bi{BaFo18} J.~C.~Baez, B.~Fong, A compositional framework for passive linear networks, \emph{Theory Appl.\ Categ.} 33 (2018), 1158--1222.
\bi{BFV03} R.~Brunetti, K.~Fredenhagen, R.~Verch, The generally covariant locality principle: a new paradigm for local quantum field theory, \emph{Comm.\ Math.\ Phys.} 237 (2003), 31--68.
\bi{BGP07} C.~B\"ar, N.~Ginoux, F.~Pf\"affle, \emph{Wave Equations on Lorentzian Manifolds and Quantization}, European Mathematical Society, 2007.
\bi{Bos98} A.~Bossavit, \emph{Computational Electromagnetism}, Academic Press, 1998.
\bi{Bro88} R.~Brooks, Constructing isospectral manifolds, \emph{Amer.\ Math.\ Monthly} 95 (1988), 823--839.
\bi{CFL28} R.~Courant, K.~Friedrichs, H.~Lewy, \"Uber die partiellen Differenzengleichungen der mathematischen Physik, \emph{Math.\ Ann.} 100 (1928), 32--74.
\bi{CoGw17} K.~Costello, O.~Gwilliam, \emph{Factorization Algebras in Quantum Field Theory}, vol.~1, Cambridge University Press, 2017.
\bi{CoHi53} R.~Courant, D.~Hilbert, \emph{Methods of Mathematical Physics}, vol.~I, Interscience, 1953.
\bi{Col51} J.~D.~Cole, On a quasi-linear parabolic equation occurring in aerodynamics, \emph{Quart.\ Appl.\ Math.} 9 (1951), 225--236.
\bi{CoTh06} T.~M.~Cover, J.~A.~Thomas, \emph{Elements of Information Theory}, 2nd ed., Wiley, 2006.
\bi{Cou95} S.~C.~Coutinho, \emph{A Primer of Algebraic D-Modules}, London Math.\ Soc.\ Student Texts 33, Cambridge University Press, 1995.
\bi{CrRa71} M.~G.~Crandall, P.~H.~Rabinowitz, Bifurcation from simple eigenvalues, \emph{J.~Funct.\ Anal.} 8 (1971), 321--340.
\bi{CrSu93} W.~Craig, C.~Sulem, Numerical simulation of gravity waves, \emph{J.~Comput.\ Phys.} 108 (1993), 73--83.
\bi{DGLS01} E.~B.~Davies, G.~M.~L.~Gladwell, J.~Leydold, P.~F.~Stadler, Discrete nodal domain theorems, \emph{Linear Algebra Appl.} 336 (2001), 51--60.
\bi{DoBu13} F.~D\"orfler, F.~Bullo, Kron reduction of graphs with applications to electrical networks, \emph{IEEE Trans.\ Circuits Syst.\ I} 60 (2013), 150--163.
\bi{DTHC} Anonymous, \emph{Differential Topology in Higher Categories}, Higher Categorical Semantics, 2026. Companion volume.
\bi{DuHo72} J.~J.~Duistermaat, L.~H\"ormander, Fourier integral operators II, \emph{Acta Math.} 128 (1972), 183--269.
\bi{EGHC} Anonymous, \emph{Evolutionary Game Theory in Higher Categories}, Higher Categorical Semantics, 2026. Companion volume.
\bi{Ehr54} L.~Ehrenpreis, Solution of some problems of division I, \emph{Amer.\ J.~Math.} 76 (1954), 883--903.
\bi{Ehr70} L.~Ehrenpreis, \emph{Fourier Analysis in Several Complex Variables}, Wiley-Interscience, 1970.
\bi{ESCHC} Anonymous, \emph{Entropy, Sources, and Channels in Higher Categories}, Higher Categorical Semantics, 2026. Companion volume.
\bi{Eva10} L.~C.~Evans, \emph{Partial Differential Equations}, 2nd ed., Graduate Studies in Mathematics 19, American Mathematical Society, 2010.
\bi{EvMa99} W.~N.~Everitt, L.~Markus, \emph{Boundary Value Problems and Symplectic Algebra for Ordinary Differential and Quasi-Differential Operators}, Math.\ Surveys Monogr.\ 61, American Mathematical Society, 1999.
\bi{Foc35} V.~Fock, Zur Theorie des Wasserstoffatoms, \emph{Z.~Phys.} 98 (1935), 145--154.
\bi{Fre00} D.~S.~Freed, Dirac charge quantization and generalized differential cohomology, in \emph{Surveys in Differential Geometry VII}, International Press, 2000, 129--194.
\bi{GGKM67} C.~S.~Gardner, J.~M.~Greene, M.~D.~Kruskal, R.~M.~Miura, Method for solving the Korteweg--de Vries equation, \emph{Phys.\ Rev.\ Lett.} 19 (1967), 1095--1097.
\bi{GIHC} Anonymous, \emph{Games and Information in Higher Categories}, Higher Categorical Semantics, 2026. Companion volume.
\bi{GKSS26} G.~Giotopoulos, I.~Khavkine, H.~Sati, U.~Schreiber, Synthetic differential jet bundles are reduced, preprint, arXiv:2601.15862, 2026. Checked.
\bi{Gri85} P.~Grisvard, \emph{Elliptic Problems in Nonsmooth Domains}, Pitman, 1985.
\bi{GWW92} C.~Gordon, D.~Webb, S.~Wolpert, Isospectral plane domains and surfaces via Riemannian orbifolds, \emph{Invent.\ Math.} 110 (1992), 1--22.
\bi{Had23} J.~Hadamard, \emph{Lectures on Cauchy's Problem in Linear Partial Differential Equations}, Yale University Press, 1923.
\bi{Hop50} E.~Hopf, The partial differential equation $u_t+uu_x=\mu u_{xx}$, \emph{Comm.\ Pure Appl.\ Math.} 3 (1950), 201--230.
\bi{Hor71} L.~H\"ormander, Fourier integral operators I, \emph{Acta Math.} 127 (1971), 79--183.
\bi{Hor83} L.~H\"ormander, \emph{The Analysis of Linear Partial Differential Operators I}, Springer, 1983.
\bi{HoSi05} M.~J.~Hopkins, I.~M.~Singer, Quadratic functions in geometry, topology, and M-theory, \emph{J.~Differential Geom.} 70 (2005), 329--452.
\bi{HTT08} R.~Hotta, K.~Takeuchi, T.~Tanisaki, \emph{D-Modules, Perverse Sheaves, and Representation Theory}, Progr.\ Math.\ 236, Birkh\"auser, 2008.
\bi{Hum90} J.~E.~Humphreys, \emph{Reflection Groups and Coxeter Groups}, Cambridge University Press, 1990.
\bi{Kac66} M.~Kac, Can one hear the shape of a drum?, \emph{Amer.\ Math.\ Monthly} 73 (1966), no.~4, part II, 1--23.
\bi{Kak44} S.~Kakutani, Two-dimensional Brownian motion and harmonic functions, \emph{Proc.\ Imp.\ Acad.\ Tokyo} 20 (1944), 706--714.
\bi{KaSc90} M.~Kashiwara, P.~Schapira, \emph{Sheaves on Manifolds}, Grundlehren 292, Springer, 1990.
\bi{Kat04} Y.~Katznelson, \emph{An Introduction to Harmonic Analysis}, 3rd ed., Cambridge University Press, 2004.
\bi{KhSc17} I.~Khavkine, U.~Schreiber, Synthetic geometry of differential equations: I. Jets and comonad structure, preprint, arXiv:1701.06238, 2017. Checked.
\bi{KoSc99} V.~Kostrykin, R.~Schrader, Kirchhoff's rule for quantum wires, \emph{J.~Phys.\ A} 32 (1999), 595--630.
\bi{Kru70} S.~N.~Kru\v{z}kov, First order quasilinear equations in several independent variables, \emph{Math.\ USSR-Sb.} 10 (1970), 217--243.
\bi{LaRi56} P.~D.~Lax, R.~D.~Richtmyer, Survey of the stability of linear finite difference equations, \emph{Comm.\ Pure Appl.\ Math.} 9 (1956), 267--293.
\bi{Lax68} P.~D.~Lax, Integrals of nonlinear equations of evolution and solitary waves, \emph{Comm.\ Pure Appl.\ Math.} 21 (1968), 467--490.
\bi{Lax73} P.~D.~Lax, \emph{Hyperbolic Systems of Conservation Laws and the Mathematical Theory of Shock Waves}, SIAM, 1973.
\bi{Mal56} B.~Malgrange, Existence et approximation des solutions des \'equations aux d\'eriv\'ees partielles et des \'equations de convolution, \emph{Ann.\ Inst.\ Fourier} 6 (1956), 271--355.
\bi{Mar86} M.~Marvan, A note on the category of partial differential equations, in \emph{Differential Geometry and its Applications (Brno, 1986)}, 1987, 235--244.
\bi{McC11} B.~J.~McCartin, \emph{Laplacian Eigenstructure of the Equilateral Triangle}, Hikari, 2011.
\bi{McSi67} H.~P.~McKean, I.~M.~Singer, Curvature and the eigenvalues of the Laplacian, \emph{J.~Differential Geom.} 1 (1967), 43--69.
\bi{MoPe10} P.~M\"orters, Y.~Peres, \emph{Brownian Motion}, Cambridge University Press, 2010.
\bi{Noe18} E.~Noether, Invariante Variationsprobleme, \emph{Nachr.\ Ges.\ Wiss.\ G\"ottingen, Math.-Phys.\ Kl.} (1918), 235--257.
\bi{Ole57} O.~A.~Ole\u{\i}nik, Discontinuous solutions of non-linear differential equations, \emph{Uspekhi Mat.\ Nauk} 12 (1957), 3--73.
\bi{Olv93} P.~J.~Olver, \emph{Applications of Lie Groups to Differential Equations}, 2nd ed., Graduate Texts in Mathematics 107, Springer, 1993.
\bi{Pal70} V.~P.~Palamodov, \emph{Linear Differential Operators with Constant Coefficients}, Springer, 1970.
\bi{Paz83} A.~Pazy, \emph{Semigroups of Linear Operators and Applications to Partial Differential Equations}, Springer, 1983.
\bi{PDEML} Anonymous, Partial differential equations in machine learning and AI: naturality, well-posedness, and spectral rates, Higher Categorical Semantics, 2026. Companion paper.
\bi{QuVa99} A.~Quarteroni, A.~Valli, \emph{Domain Decomposition Methods for Partial Differential Equations}, Oxford University Press, 1999.
\bi{ReSi75} M.~Reed, B.~Simon, \emph{Methods of Modern Mathematical Physics}, vols.~II and III, Academic Press, 1975 and 1979.
\bi{Rud62} W.~Rudin, \emph{Fourier Analysis on Groups}, Interscience, 1962.
\bi{Sat59} M.~Sato, Theory of hyperfunctions I, \emph{J.~Fac.\ Sci.\ Univ.\ Tokyo Sect.\ I} 8 (1959), 139--193.
\bi{Sch54} L.~Schwartz, Sur l'impossibilit\'e de la multiplication des distributions, \emph{C.~R.\ Acad.\ Sci.\ Paris} 239 (1954), 847--848.
\bi{Sch57} L.~Schwartz, \emph{Th\'eorie des distributions}, Hermann, 1950--1951; new edition 1966.
\bi{Str08} W.~A.~Strauss, \emph{Partial Differential Equations: An Introduction}, 2nd ed., Wiley, 2008. The source.
\bi{StWe71} E.~M.~Stein, G.~Weiss, \emph{Introduction to Fourier Analysis on Euclidean Spaces}, Princeton University Press, 1971.
\bi{Sun85} T.~Sunada, Riemannian coverings and isospectral manifolds, \emph{Ann.\ of Math.} 121 (1985), 169--186.
\bi{Ton75} E.~Tonti, \emph{On the Formal Structure of Physical Theories}, Istituto di Matematica del Politecnico di Milano, 1975.
\bi{Vil68} N.~Ja.~Vilenkin, \emph{Special Functions and the Theory of Group Representations}, Transl.\ Math.\ Monogr.\ 22, American Mathematical Society, 1968.
\bi{Vin01} A.~M.~Vinogradov, \emph{Cohomological Analysis of Partial Differential Equations and Secondary Calculus}, Transl.\ Math.\ Monogr.\ 204, American Mathematical Society, 2001.
\bi{You50} D.~M.~Young, Iterative methods for solving partial difference equations of elliptic type, \emph{Trans.\ Amer.\ Math.\ Soc.} 76 (1954), 92--111; Ph.D.\ thesis, Harvard University, 1950.
\end{thebibliography}
\endgroup
\end{document}