An Alternate proof of Nyquist-Shannon

101 Views Asked by At

This problem is from Basic Complex Analysis, Part 2A, by Barry Simon.

This problem will provide an alternate proof of the strong from of the Nyquist-Shannon sampling theorem (Theorem 6.6.16 of Part 1), following closely the 1920 proof of Ogura. We'll suppose $f(z)$ is an entire function obeying

$|f(z)| \leq C \exp((\pi - \epsilon)| \Im z|)$

Fix $z_0 \in \mathbb{C} \backslash \mathbb{Z}$. Let

$g(z) = \frac{f(z)}{(z - z_0)\sin(\pi z)}$

and $R_k$ be the rectangle which is the boundary of $\{x + iy\ |\ |x| \le k + \frac{1}{2},\ |y| \le k + \frac{1}{2} \}$ oriented clockwise.

Prove that

$\lim_{k \to \infty} \int_{R_k} |g(z)| d|z| = 0$

My attempt:

Choose $k \in \mathbb{Z}$ large enough such that $z_0$ is in the area bounded by $R_k$. Then let $z = x + iy$. Thus, \begin{align*} \int_{R_k} |g(z)| |dz| &\le \sup_{R_k} \left | \frac{f(z)}{(z - z_0)(\sin(\pi z)} \right | \ell(R_k)\\ & \le \sup_{R_k} \left | \frac{C \exp((\pi - \epsilon)| \Im z|)}{(\sin(\pi z))(k - |z_0|)} \right | 8(k + 1/2)\\ & = 8C \sup_{R_k} \left| e^{-\epsilon |y|} \frac{e^{\pi|y|}}{\sin(\pi z)} \right | \frac{(k + 1/2)}{k - |z_0|} \end{align*}

Then for $y > \frac{1}{2}$ we have, \begin{align*} \left | \frac{e^{\pi|y|}}{\sin(\pi z)} \right | = \left |\frac{2ie^{\pi y}}{e^{\pi i z} - e^{-\pi i z}} \right | \le \frac{2e^{\pi y}}{||e^{\pi i z}| - |e^{-\pi i z}||} = \frac{2e^{\pi y}}{e^{\pi y} - e^{-\pi y}} = \frac{2}{1 - e^{-2 \pi y}} \le \frac{2}{1 - e^{-\pi}} \end{align*} For $y < \frac{1}{2}$ we have, \begin{align*} \left | \frac{e^{\pi|y|}}{\sin(\pi z)} \right | = \left |\frac{2ie^{-\pi y}}{e^{\pi i z} - e^{-\pi i z}} \right | \le \frac{2e^{-\pi y}}{||e^{\pi i z}| - |e^{-\pi i z}||} = \frac{2e^{-\pi y}}{e^{-\pi y} - e^{\pi y}} = \frac{2}{1 - e^{2 \pi y}} \le \frac{2}{1 - e^{-\pi}} \end{align*} For $-\frac{1}{2} \le y \le \frac{1}{2}$ and $z = k + \frac{1}{2} + iy$ we have,

\begin{align*} \left | \frac{e^{\pi|y|}}{\sin(\pi z)} \right | \le \frac{e^{\pi/2}}{|\sin( \pi(k + \frac{1}{2} + iy))|} = \frac{e^{\pi/2}}{|\sin( \frac{\pi}{2} + \pi iy))|} = \frac{e^{\pi/2}}{|\cosh(\pi y ))|} \le \frac{e^{\pi/2}}{|\cosh(\pi/2 ))|} \end{align*} Let $M = \max \left \{ \frac{2}{1 - e^{-\pi}}, \frac{e^{\pi/2}}{|\cosh(\pi/2 ))|} \right \}$. Then $M$ does not depend on $k$, so \begin{align*} \int_{R_k} |g(z)| |dz| \le 8Ce^{-\epsilon |y|} M \frac{(k + 1/2)}{k - |z_0|} \end{align*} Which implies, for the top and bottom \begin{align*} \lim_{k \to \infty} \int_{R_k} |g(z)| |dz| \le \lim_{k \to \infty} 8Ce^{-\epsilon |k+1|} M \frac{(k + 1/2)}{k - |z_0|} = 0 \end{align*} But what about the sides?