Indexed text attachment
partial-differential-equations-in-higher-categories.tex
Source for Partial Differential Equations in Higher Categories
View paper · Download original file
Lines 1221–1320 of 3921. Source text is displayed literally.
1221\noindent The formula is that of variation of constants and holds for every semigroup; for the wave equation1222written as a first-order system it is \Cref{thm:duhamel-wave}. With $f=\sin x$ and $\varphi=0$ it gives1223$(1-e^{-t})\sin x$. \chk{D9}1224 1225\section*{Notes}1226 1227The source treats diffusion in its Sections 2.3--2.5, 3.1, 3.3 and 3.5. Semigroup theory is in \cite{Paz83}. The1228identity between the rate of entropy and the Fisher information is de Bruijn's, see \cite{CoTh06}; monotonicity1229of divergences under Markov kernels is the subject of \cite{ESCHC}. The image and eigenfunction forms of the kernel1230on an interval are two expressions of a Jacobi theta function.1231 1232\section*{Exercises}1233 1234\begin{exercise}\label{exer:scal}1235Derive the Gaussian from the ansatz $u=t^{-1/2}g(x/\sqrt t)$.1236\end{exercise}1237 1238\begin{exercise}\label{exer:surv}1239A particle diffuses with $k=1$ from $x=1$ on the half-line with an absorbing wall at $0$. Find the probability1240that it has not been absorbed by time $1$.1241\end{exercise}1242 1243\begin{exercise}\label{exer:ent}1244Prove (a) and (b) of \Cref{thm:mono}.1245\end{exercise}1246 1247\begin{exercise}\label{exer:terms}1248Estimate, as functions of $kt/l^2$ and the tolerance $\epsilon$, the number of image shells and of eigenfunctions1249needed in \Cref{thm:kernels}, and compare with the table.1250\end{exercise}1251%% ---- end ch/04-diffusions.tex ----1252%% ---- begin ch/05-fourier.tex ----1253\chapter{Boundary Conditions and Fourier Series}\label{ch:fourier}1254 1255On a bounded interval there is no explicit kernel, and the source turns to Fourier's method: separate the1256variables, solve an eigenvalue problem in $x$, and superpose \src{Ch.~4--5}. Three questions organize the chapter.1257What is a boundary condition? Why sines and cosines? In what sense does the series converge?1258 1259\section{Diagonalizing the generator}1260 1261Let $A=-d^2/dx^2$ on $(0,l)$ with boundary conditions fixed. The diffusion and wave equations are1262$u_t=-kAu$ and $u_{tt}=-c^2Au$. If $AX=\lambda X$, then $e^{-k\lambda t}X$ solves the first and1263$\cos(c\sqrt\lambda\,t)X$, $\sin(c\sqrt\lambda\,t)X$ solve the second. Fourier's method is the statement that $A$ can1264be diagonalized, so that the semigroup and the group of the last two chapters become multiplication operators:1265\[ T_t=e^{-ktA},\qquad\text{wave group}=\begin{pmatrix}\cos ct\sqrt A&\sin(ct\sqrt A)/c\sqrt A\\-c\sqrt A\sin ct\sqrt A&\cos ct\sqrt A\end{pmatrix}. \]1266For Dirichlet conditions the eigenpairs are $(n\pi/l)^2$, $\sin(n\pi x/l)$, $n\ge1$; for Neumann conditions1267$(n\pi/l)^2$, $\cos(n\pi x/l)$, $n\ge0$. \chk{F1}1268 1269\section{A boundary condition is a Lagrangian plane}1270 1271For functions $f,g$ on $[a,b]$, Green's second identity in one dimension reads1272\[ \int_a^b(-f''g+fg'')\,dx=\bigl[-f'g+fg'\bigr]_a^b=:\omega\bigl(\beta(f),\beta(g)\bigr), \]1273where $\beta(f)=(f(a),f'(a),f(b),f'(b))\in\R^4$ is the vector of boundary values of $f$.1274The form $\omega$ is symplectic on the space $\R^4$ of boundary values. A pair of independent linear boundary1275conditions is a plane $V\subset\R^4$.1276 1277\begin{theorem}\label{thm:lagrangian}1278\Srf[source, \S5.3; \cite{EvMa99,KoSc99}] The boundary conditions $V$ are symmetric in the source's sense if and only if1279$V$ is Lagrangian: $\omega|_V=0$. In that case eigenfunctions for distinct eigenvalues are orthogonal and all1280eigenvalues are real (the source's Theorems 1 and 2). The self-adjoint realizations of $-d^2/dx^2$ on $[a,b]$ are in1281bijection with the Lagrangian planes of $(\R^4,\omega)$, a compact manifold of dimension three.1282\end{theorem}1283\begin{proof}1284The first statement is the definition, since the source's condition is $\omega(\beta(f),\beta(g))=0$ for all $f,g$1285satisfying the conditions, and a plane on which a symplectic form on $\R^4$ vanishes is Lagrangian. For1286eigenfunctions $X_1,X_2$ the identity gives $(\lambda_1-\lambda_2)\int X_1X_2=\omega(\beta(X_1),\beta(X_2))=0$.1287\end{proof}1288 1289\begin{center}\small1290\begin{tabular}{@{}lll@{}}1291\toprule1292condition & plane $V$ spanned by & $\omega$ on the two vectors\\1293\midrule1294Dirichlet & $(0,1,0,0)$, $(0,0,0,1)$ & $0$\\1295Neumann & $(1,0,0,0)$, $(0,0,1,0)$ & $0$\\1296Robin & $(1,a_0,0,0)$, $(0,0,1,-a_l)$ & $0$\\1297periodic & $(1,0,1,0)$, $(0,1,0,1)$ & $0$\\1298$X(a)=X(b)$, $X'(a)=2X'(b)$ & $(1,0,1,0)$, $(0,2,0,1)$ & $-1$\\1299\bottomrule1300\end{tabular}1301\end{center}1302\noindent \chk{F3} The last row is the source's example of conditions for which the method fails. The first three1303are \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, with1304Dirichlet and Neumann two of its points and the Robin constant a coordinate on the rest. The periodic condition is1305non-local: it is the graph of the identity between the two ends, which is what it means to glue them.1306 1307\begin{center}1308\begin{tikzpicture}1309\draw[->,ink2] (-2.4,0)--(2.6,0) node[right]{\scriptsize $X(a)$};1310\draw[->,ink2] (0,-2.0)--(0,2.2) node[above]{\scriptsize $X'(a)$};1311\draw[ink2,very thin] (0,0) circle (1.3);1312\draw[very thick,fwd] (0,-1.9)--(0,1.9);1313\draw[very thick,bwd] (-2.2,0)--(2.2,0);1314\draw[thick,accent] (-1.9,-1.1)--(1.9,1.1);1315\draw[thick,accent,densely dashed] (-1.5,1.6)--(1.5,-1.6);1316\node[fwd,left] at (-0.05,1.75) {\scriptsize Dirichlet: $X=0$};1317\node[bwd,below] at (1.45,-0.03) {\scriptsize Neumann: $X'=0$};1318\node[accent,right] at (1.9,1.1) {\scriptsize Robin: $X'=aX$, $a>0$};1319\node[accent,right] at (1.5,-1.6) {\scriptsize $a<0$};1320\foreach \p in {(0,1.3),(1.3,0),(1.125,0.651),(0.889,-0.948)} \fill \p circle (1.6pt);