Indexed text attachment
partial-differential-equations-in-higher-categories.tex
Source for Partial Differential Equations in Higher Categories
View paper · Download original file
Lines 3001–3100 of 3921. Source text is displayed literally.
3001\noindent \chk{P7} The wave operator has no first-order square root with scalar coefficients, since its symbol is an3002indefinite quadratic form. It has one with coefficients in the Clifford algebra of that form, which is the3003universal algebra in which the form becomes a square. This is \Cref{prop:factor} for the hyperbolic operator in3004more than one space dimension.3005 3006\section{Fields as functors}3007 3008The four structures of \Cref{ch:waves} are available for the Klein--Gordon equation on any globally hyperbolic3009space-time.3010 3011\begin{theorem}[Locally covariant fields]\label{thm:lcqft}3012\Sli[\cite{BFV03,BGP07}] The assignment $M\mapsto(\Sol_{sc}(M),\Omega)$ is a functor from globally hyperbolic3013space-times and causal isometric embeddings to symplectic vector spaces. Solutions on causally separated regions3014are $\Omega$-orthogonal, and an embedding whose image contains a Cauchy surface induces an isomorphism.3015Composing with the functor from symplectic spaces to algebras of canonical commutation relations gives a quantum3016field theory.3017\end{theorem}3018 3019\noindent So the free quantum field is the solution functor of a hyperbolic equation, followed by quantization: the3020exact sequence of \Cref{thm:bgp}, the causality of \Cref{thm:causal} and the symplectic form of \Cref{prop:symp} are3021its axioms. A numerical scheme can respect this structure: the leapfrog step for a discretized Klein--Gordon3022equation satisfies $M^{\T}JM=J$ to rounding error, and the Euler step does not. \chk{P8} In the formalism of3023factorization algebras the same data are organized by open sets of space-time, and the solution space is3024replaced by its derived version, a complex whose zeroth cohomology is $\Sol$ \cite{CoGw17}.3025 3026\section*{Notes}3027 3028The source's Chapter 13 is followed for the equations themselves. For differential forms in electromagnetism see3029\cite{Bos98}; for differential cohomology and gauge fields \cite{Fre00,HoSi05}; for the geometry of the Euler3030equations \cite{Arn66}; for scattering theory \cite{ReSi75}; for locally covariant field theory3031\cite{BFV03,BGP07} and factorization algebras \cite{CoGw17}.3032 3033\section*{Exercises}3034 3035\begin{exercise}\label{exer:charge}3036Derive the conservation of charge from the source's form of Maxwell's equations.3037\end{exercise}3038 3039\begin{exercise}\label{exer:string}3040Verify the energy balance $R^2+(c_1/c_2)T^2=1$ for the string of two materials.3041\end{exercise}3042 3043\begin{exercise}\label{exer:delta}3044Find the transfer matrix of the potential $V=g\,\delta(x)$ and the transmission probability $|T|^2$.3045\end{exercise}3046 3047\begin{exercise}\label{exer:kg}3048Show that the group and phase velocities of the Klein--Gordon equation have product $c^2$.3049\end{exercise}3050%% ---- end ch/13-fields.tex ----3051%% ---- begin ch/14-nonlinear.tex ----3052\chapter{Nonlinear Equations}\label{ch:nonlinear}3053 3054The source ends with a sample of nonlinear phenomena: shocks, solitons, variational problems, bifurcation and3055water waves \src{Ch.~14}. Linear structure is gone, and with it most of the functors of the previous chapters.3056What remains, and what explains the two solvable examples, are morphisms of equations in the sense of3057\Cref{thm:jet}.3058 3059\section{Shocks}3060 3061For $u_t+a(u)u_x=0$ the solution is constant along characteristics, which are straight lines of slope determined3062by the data, as in \Cref{prop:char} \src{\S14.1}. When they cross the quotient is no longer a manifold and the3063smooth solution ends.3064 3065\begin{proposition}\label{prop:break}3066\Scl For $u_t+uu_x=0$ with data $\varphi$, the smooth solution exists exactly until $t_b=-1/\min\varphi'$.3067\end{proposition}3068 3069\noindent For $\varphi=e^{-x^2}$, $t_b=\sqrt{e/2}=1.1658$. \chk{X1} The source's explicit solution for $\varphi=x^2$ is3070verified, with its domain $1+4tx\ge0$. \chk{X2}3071 3072\begin{center}3073\begin{tikzpicture}[xscale=1.25,yscale=1.25]3074\draw[->,ink2] (-2.6,0)--(4.2,0) node[right]{\small$x$};3075\draw[->,ink2] (-2.4,0)--(-2.4,2.3) node[above]{\small$t$};3076\foreach \x in {-2.25,-2,...,2.76}{3077 \pgfmathsetmacro{\s}{exp(-(\x)*(\x))}3078 \draw[fwd] (\x,0)--({\x+2*\s},2);}3079\draw[densely dashed,bwd] (-2.4,1.166)--(4.1,1.166) node[above left]{\small$t_b$};3080\end{tikzpicture}3081\figcap{Characteristics of $u_t+uu_x=0$ for the data $e^{-x^2}$: straight lines $x=x_0+\varphi(x_0)t$. They first3082meet at $t_b=1.166$; above that level the map from $(x_0,t)$ to $(x,t)$ is no longer injective.}3083\end{center}3084 3085After $t_b$ one asks for weak solutions. Distributions are not available, since $u^2$ must be formed3086(\Cref{ch:distributions}); instead the equation is written as a conservation law $u_t+A(u)_x=0$ and tested3087against smooth functions. A piecewise smooth weak solution has jumps moving with the Rankine--Hugoniot speed3088$s=(A(u^+)-A(u^-))/(u^+-u^-)$.3089 3090\begin{proposition}[Weak solutions belong to the conservation form]\label{prop:consform}3091\Scl The equations $u_t+(\tfrac12u^2)_x=0$ and $(u^2)_t+(\tfrac23u^3)_x=0$ have the same smooth solutions and3092different weak ones: a jump from $1$ to $0$ moves at speed $\tfrac12$ under the first and $\tfrac23$ under the3093second.3094\end{proposition}3095 3096\noindent An isomorphism of equations on smooth solutions need not extend to weak solutions; the weak solution3097functor is defined on balance laws, in the sense of \Cref{prop:tonti}, and not on the ideals they generate. The3098same distinction decides numerical methods: for the step from $1$ to $0$ a scheme in conservation form places the3099shock at $1.005$ after two time units, against the exact $1$, and the equally consistent non-conservative scheme3100leaves it at the origin. \chk{X3}