% vim: tw=50 % 01/11/2022 12PM \subsubsection*{3.1.3 Harmonic Oscillator} \begin{center} \includegraphics[width=0.6\linewidth] {images/6753780059de11ed.png} \end{center} \begin{flashcard}[harmonic-oscillator-potential] \prompt{Potential for quantum harmonic oscillator?} \[ U(x) = \cloze{\half kx^2} \] \end{flashcard} $k \in \RR$ elastic constant. $\omega = \sqrt{\frac{k}{m}}$. Classical mechanics: Newton 2 is $\ddot{x}(t) = -\omega^2 x(t)$. \[ \implies x(t) = A \sin \omega t + B \cos\omega t\] with $T = \frac{2\pi}{\omega}$ period oscillations. \\ Quantum mechanics: \[ -\frac{\hbar^2}{2m} \chi''(x) + \half m\omega^2 x^2 \chi(x) = E \chi(x) \tag{1} \] We know: \begin{itemize} \item Discrete eigenvalues \item even / odd eigenfunctions \end{itemize} Change of variables: \[ \xi^2 \equiv \frac{m\omega}{\hbar}x^2 \] \[ \eps \equiv \frac{2E}{\hbar\omega} \] Plug into (1) \[ -\dfrac[2]{\chi}{\xi}(\xi) + \xi^2 \chi(\xi) = \eps \chi(\xi) \tag{2} \] Solve it by starting from a particular solution \[ \eps = 1 \quad \left( E_0 = \frac{\hbar\omega}{2} \right) \] ansatz: \[ \chi_0(\xi) = e^{-\xi^2 / 2} \tag{3} \] Plug (3) into (2) with $\eps = 1$ works. We found one eigenvalues $E_0 = \frac{\hbar\omega}{2}$, $\chi_0(x) = A e^{-\frac{m\omega}{2\hbar} x^2}$ To find other eigenfunction of $\hat{H}$ take general form \[ \xi(\xi) = f(\xi) e^{-\xi^2 / x} \tag{4} \] Plug (4) into (2) \[ -\dfrac[2]{f}{\xi} + 2\xi \dfrac{f}{\xi} + (1 - \eps) f = 0 \tag{5} \] Use power series method ($\xi = 0$ regular point) \[ f(\xi) = \sum_{n = 0}^\infty a_n \xi^n \tag{6} \] $a_n \in \RR$. Clearly \[ \xi \dfrac{f}{\xi} = \sum_{n = 0}^\infty na_n \xi^n \tag{7} \] \[ \dfrac[2]{f}{\xi} = \sum_{n = 0}^\infty n(n - 1)a_n \xi^{n - 2} = \sum_{n = 0}^\infty (n + 1)(n + 2) a_{n + 2} \xi^n \] Plug (6)-(8) into (5): \begin{align*} \sum_{n = 0}^\infty [(n + 1)(n + 2)a_{n + 2} - 2na_n + (\eps - 1) a_n]\xi^n &= 0 \\ \implies a_{n + 2} &= \frac{(2n - \eps + 1)}{(n + 1)(n + 2)}a_n \end{align*} Because of parity of eigenfunction: \begin{itemize} \item Either $a_n = 0$ for odd $n$ ($f(\xi) = f(-\xi)$) even eigenfunction \item or $a_n = 0$ for even $n$, ($f(\xi) = -f(-\xi)$) odd eigenfunction. \end{itemize} \begin{proposition*} If series (6) does \emph{not} terminate then eigenfunction of $\hat{H}$ would \emph{not} be normalisable. \end{proposition*} \begin{proof} Suppose that the series in (6) does \emph{not} terminate. Hence can look at asymptotic behaviour of series. Take (0) \[ \frac{a_{n + 2}}{a_n} \to \frac{2}{n} \] as $n \to \infty$. This is same asymptotic behaviour as \[ g(\xi) = e^{\xi^2} = \sum_{m = 0}^\infty \frac{\xi^{2m}}{m!} = \sum_{m = 0}^\infty b_m \xi^m \] where \[ b_m = \begin{cases} \frac{1}{m!} & \text{for $m$ even} \\ 0 & \text{for $m$ odd} \end{cases} \] asymptotic behaviour of $g(\xi)$ \[ \frac{b_{n + 2}}{b_n} = \frac{\left( \frac{m}{2} \right)!}{ \left( \frac{m}{2} + 1 \right)!} = \frac{2}{m + 2} \to \frac{2}{m} \] as $m \to \infty$. So if $e^{\xi^2/2}$ and $f(\xi)$ have same asymptotic behaviour \[ \chi(\xi) \sim e^{\xi^2} e^{-\xi^2/2} = e^{\xi^2/2} \to \infty \] \end{proof} \myskip Given that the series (6) terminates then there exists $N $such that \[ a_{N + 2} = 0 \tag{10} \] with $a_N \neq 0$. Plug (10) into (9) \[ a_{N + 2} = \frac{(2N - \eps + 1)}{(N + 1)(N + 2)}a_N = 0 \] \[ \implies 2N - \eps + 1 = 0 \] Plugging in definition of $\eps$ \[ \implies E_N = \left( N + \half \right) \hbar\omega \] eigenvalues $N = 0$, $E_0 = \frac{\hbar\omega}{2}$ \[ E_{N + 1} - E_n = \hbar \omega \] eigenfunction $\chi_N(\xi) = f_N(\xi) e^{-\xi^2/2}$ \[ \chi_N(-\xi) = (-1)^N \chi_N(\xi) \] Hermite polynomials are defined with recursive relation \[ f_N(\xi) = (-1)^N e^{\xi^2} \dfrac[N]{}{\xi} (e^{-\xi^2}) \] \begin{center} \begin{tabular}{c|c|c} $N$ & $E_N$ & $f_N(\xi)$ \\ \hline $0$ & $\frac{\hbar\omega}{2}$ & $1$ \\ $1$ & $\frac{3\hbar\omega}{2}$ & $\xi$ \\ $2$ & $\frac{5\hbar\omega}{2}$ & $(1 - 2\xi^2)$ \\ $3$ & $\frac{7\hbar\omega}{2}$ & $(\xi - \frac{2}{3}\xi^3)$ \end{tabular} \end{center}