% vim: tw=50 % 04/11/2022 10AM \newpage \mychapter{Inhomogeneous ODEs; Fourier Transforms} \newpage \section{The Dirac Delta Function} \subsection{Definition of $\delta(x)$} Define a generalised function $\delta(x - \xi)$ with the following properties: \begin{equation*} \boxed{ \begin{aligned} \delta(x - \xi) = 0, \quad \forall x \neq \xi \\ \int_{-\infty}^\infty \delta(x - \xi) \dd x = 1 \end{aligned} } \tag{6.1} \end{equation*} \begin{center} \includegraphics[width=0.6\linewidth] {images/9021edf669ca11ed.png} \end{center} This acts as a linear operator $\int \dd x \delta(x - \xi)$ on an arbitrary function $f(x)$ to produce a number $f(\xi)$, that is, \[ \boxed{\int_{-\infty}^\infty \dd x \delta(x - \xi) f(x) = f(\xi)} \tag{6.2} \] provided $f(x)$ is `well-behaved' at $x = \xi$ and $\pm \infty$. \subsubsection*{Notes} \begin{itemize} \item The delta function $\delta(x)$ is classified as a distribution (not a function). See lecture notes of Jozsa and Skinner section 6.1 (optional). \item $\delta(x)$ always appears in an integrand as a linear operator where it is well-defined. \item Represents a unit point source (for example mass, charge) or an impulse. \end{itemize} \subsubsection*{Some limiting approximations} Discrete: \[ \lim_{n \to \infty} \delta_n(x) = \begin{cases} 0 & x > \frac{1}{n} \\ \frac{n}{2} & |x| \le \frac{1}{n} \\ 0 & x < -\frac{1}{n} \end{cases} \] \begin{center} \includegraphics[width=0.6\linewidth] {images/5813f61069cb11ed.png} \end{center} \myskip Continuous: \[ \lim_{\eps \to 0} \delta_\eps(x) = \frac{1}{\eps \sqrt{\pi}} e^{-x^2/\eps^2} \tag{6.3} \] verify (6.2): \begin{align*} \int_{-\infty}^\infty f(x) \delta(x) \dd x &= \lim_{\eps \to 0} \int_{-\infty}^\infty \frac{1}{\eps\sqrt{\pi}} e^{-x^2/\eps^2} f(x) \dd x \\ &= \lim_{\eps \to 0} \int_{-\infty}^\infty \frac{1}{\sqrt{\pi}} e^{-y^2} f(\eps y) \dd y \\ &= \lim_{\eps \to 0} \int_{-\infty}^\infty \dd y \frac{1}{\sqrt{\pi}} (f(0) + \eps y f'(0) + \cdots ) \\ &= f(0) \end{align*} $\forall f$. `well-behaved' at $x = 0$ so that we can take the Taylor expansion, and also need well behaved at $\pm \infty$ so that it doesn't grow faster than $1/e^{-x^2/\eps^2}$ \begin{center} \includegraphics[width=0.6\linewidth] {images/d9d3216269cb11ed.png} \end{center} Further examples: ($\lim n\to \infty$) \[ \delta_n(x) = \frac{\sin nx}{\pi x} = \frac{1}{2\pi} \int_{-n}^n e^{ikx} \dd k \tag{6.4} \] \[ \delta_n(x) = \frac{n}{2} \sech^2 nx \tag{6.5} \] \subsection{Properties of $\delta(x)$} \subsubsection*{Heaviside function $H(x)$} The unit step function, \[ H(x) = \begin{cases} 1 & x \ge 0 \\ 0 & x < 0 \end{cases} \tag{6.6} \] \begin{center} \includegraphics[width=0.6\linewidth] {images/2cf15a3069cc11ed.png} \end{center} is the \emph{integral of $\delta(x)$}. \[ H(x) = \int_{-\infty}^x \delta(x) \dd x \tag{6.7} \] and we can identify $H'(x) = \delta(x)$. \begin{example*} Verify using (6.5) $\delta(x) = \lim_{n \to \infty} \frac{n}{2} \sech^2 nx$. (You will find $\half (\tanh nx + 1)$ is the approximate step function. - also $H(0) = \half$ (alternate definition).) \end{example*} \subsubsection*{Derivative of $\delta(x)$} Define $\delta'(x)$ using integration by parts: \begin{align*} \int_{-\infty}^\infty \delta'(x - \xi) f(x) \dd x &= [\delta(x - \xi) f(x)]_{-\infty}^\infty - \int_{-\infty}^\infty \delta(x - \xi) f'(x) \dd x \\ &= -f'(\xi) \tag{6.8} \end{align*} for all $f(x)$ smooth at $x = \xi$. \begin{example*} Consider Gaussian approximation (6.3) \[ \delta_\eps'(x) = -\frac{2x}{\eps^3 \sqrt{\pi}}e^{-x^2/\eps^2} \] \begin{center} \includegraphics[width=0.6\linewidth] {images/d5b3d26069cc11ed.png} \end{center} \end{example*} \subsubsection*{Sampling property} \[ \int_a^b f(x) \delta(x - \xi) \dd x = \begin{cases} f(\xi) & a < \xi < b \\ 0 & \text{otherwise} \end{cases} \] \subsubsection*{Even property} \[ \int_{-\infty}^\infty f(x) \delta(-(x - \xi)) \dd x = \int_{-\infty}^\infty f(x) \delta(x - \xi) \dd x \tag{6.10} \] \begin{align*} LHS &= \int_\infty^{-\infty} f(\xi - u) \delta(u) (-\dd u) \\ &= \int_{-\infty}^\infty f(\xi - u) \delta(u) \dd u \\ &= f(\xi) \\ &= RHS \end{align*} \subsubsection*{Scaling property} \[ \int_{-\infty}^\infty f(x) \delta(a(x - \xi)) \dd x = \frac{1}{|a|} f(\xi) \tag{6.11} \] Exercise: Show this using $u = ax$ (noting integral limit order with $a < 0$). \begin{hiddenflashcard}[delta-fn-scaling-property] For $a \neq 0$, \[ \delta(ax) = \cloze{\frac{1}{|a|} \delta(x)} \] \end{hiddenflashcard} \subsubsection*{Advanced scaling} Suppose $g(x)$ has $n$ isolated zeros at $x_1, x_2, \dots, x_n$ then (with $g'(x_i) \neq 0$): \[ \delta(g(x)) = \sum_{i = 1}^n \frac{\delta(x - x_i)}{|g'(x_i)|} \tag{6.12} \] (Exercise: Show for $g$ has 1 root at $x = x_i$). \begin{example*} \[ I = \int_{-\infty}^\infty f(x) \delta(x^2 - 1) \dd x \] $x^2 - 1$ has roots $x = \pm 1$ with $g'(x) = 2x$. So \begin{align*} I &= \int_{1 - \eps}^{1 + \eps} f(x) \frac{\delta(x - 1)}{|2x|} \dd x + \int_{-1 - \eps}^{-1 + \eps} f(x) \frac{\delta(x + 1)}{|2x|} \dd x \\ &= \half (f(1) + f(-1)) \end{align*} \end{example*} \subsubsection*{Isolation property} If $g(x)$ is continuous at $x = 0$ then \[ g(x) \delta(x) = g(0) \delta(x) \tag{6.13} \] Exercise: evaluate and show \[ \int_0^\infty \delta'(x^2 - 1) x^2 \dd x = -\frac{1}{4} \] using $u = x^2 - 1$ and note (6.8) and (6.12). \begin{hiddenflashcard}[delta-fn-isolation-property] \prompt{Isolation property of $\delta$ function?} \[ \cloze{g(x)\delta(x) = g(0)\delta(x)} \] \cloze{if $g(x)$ is \fcemph{continuous} at $x = 0$.} \end{hiddenflashcard}