Generalized Fourier Transform: Schwartz Space

The Schwartz space \(\mathcal{S}\) is the topological vector space of functions \(f : \mathbb{R}_n \to \mathbb{C}\) such that:

  • \(f(x)\) infinitely differentiable

  • any derivitive tends to zero fast than any power of x:

    \[\forall m, n \ge 0 \implies \lim_{x \to \infty} |x|^n | f^{(m)} (x) | = 0\]

Properties without proof:

  • the Schwartz space is closed under differentiation and multiplication by polynomials.

  • Schwartz class functions are bounded and decay faster than any polynomial

  • Schwartz class functions are integrable

Lemma

Fourier transform of Schwartz class is bounded

Proof:

\[\begin{split}\because | \int_{-\infty}^{+\infty} e^{-2 \pi i s x} f(x) dx| & \le \int_{-\infty}^{+\infty} | e^{-2 \pi i s x} | \cdot | f(x) | dx \\ & = \int_{-\infty}^{+\infty} | f(x) | dx \\ & = \| f \|_1\end{split}\]
\[\therefore | \mathcal{F} f (s) | \le \| f \|_1\]

Differentiation Formulas

First statement:

\[f \in \mathcal{S} \implies \mathcal{F} f^{(n)} = (2 \pi i s)^n \cdot \mathcal{F} f\]

Second statement

\[f \in \mathcal{S} \implies \frac{\partial^n \mathcal{F} f}{\partial s^n} = \mathcal{F} ((-2 \pi i x)^n f(x))\]

Proof of first statement:

\[ \begin{align}\begin{aligned}& \because f \in \mathcal{S}\\& \therefore f' \in \mathcal{S}\end{aligned}\end{align} \]
\[\begin{split}\therefore (\mathcal{F} f') (s) & = \int_{-\infty}^{+\infty} e^{-2 \pi i s x} f' (x) dx \\ & = e^{-2 \pi i s x} \cdot f(x) |_{-\infty}^{+\infty} - \int_{-\infty}^{+\infty} f(x) d e^{-2 \pi i s x} \\ & = - \int_{-\infty}^{+\infty} (-2 \pi i s) \cdot e^{-2 \pi i s x} f(x) dx \\ & = 2 \pi i s \cdot \mathcal{F} f (s)\end{split}\]

By induction it follows that:

\[\mathcal{F} f^{(n)} = (2 \pi i s)^n \cdot \mathcal{F} f\]

Proof of first statement:

\[ \begin{align}\begin{aligned}& \because f \in \mathcal{S}\\& \therefore x f(x) \in \mathcal{S}\end{aligned}\end{align} \]
\[\begin{split}\therefore -2 \pi i \cdot (\mathcal{F} (x f(x))) (s) & = \int_{-\infty}^{+\infty} -2 \pi i \cdot e^{-2 \pi i s x} \cdot x f(x) dx \\ & = \int_{-\infty}^{+\infty} \frac{\partial}{\partial s} e^{-2 \pi i s x} \cdot f(x) dx \\ & = \frac{\partial \mathcal{F} f (s)}{\partial s}\end{split}\]

By induction it follows that:

\[\frac{\partial^n \mathcal{F} f}{\partial s^n} = \mathcal{F} ((-2 \pi i x)^n f(x))\]
\[\tag*{$\blacksquare$}\]

Closure under Fourier Transform

\[f \in \mathcal{S} \implies \mathcal{F} f \in \mathcal{S}\]

Proof:

\[\begin{split}(2 \pi i s)^n \cdot \frac{\partial^m}{\partial s^m} \mathcal{F} f(s) & = (2 \pi i s)^n \cdot \mathcal{F} ((-2 \pi i x)^m f(x)) (s) \\ & = \mathcal{F} (\frac{\partial^n}{\partial x^n} (-2 \pi i x)^m f(x)) (s) \\ & = (-2 \pi i)^m \cdot \mathcal{F} (\frac{\partial^n}{\partial x^n} x^m f(x)) (s)\end{split}\]
\[ \begin{align}\begin{aligned}\therefore |(2 \pi s)^n| \cdot |\frac{\partial^m}{\partial s^m} \mathcal{F} f(s)| = |(-2 \pi)^m| \cdot |\mathcal{F} (\frac{\partial^n}{\partial x^n} x^m f(x)) (s)|\\\therefore |s^n| \cdot |\frac{\partial^m}{\partial s^m} \mathcal{F} f(s)| = (2 \pi)^{m - n} \cdot |\mathcal{F} (\frac{\partial^n}{\partial x^n} x^m f(x)) (s)|\end{aligned}\end{align} \]
\[\because f \in \mathcal{S}\]
\[\therefore \frac{\partial^n}{\partial x^n} x^m f(x) \in \mathcal{S}\]
\[\therefore |s^n| \cdot |\frac{\partial^m}{\partial s^m} \mathcal{F} f(s)| \le (2 \pi)^{m - n} \cdot \| \frac{\partial^n}{\partial x^n} x^m f \|\]
\[\therefore \mathcal{F} f \in \mathcal{S}\]

also:

\[\mathcal{F}^{-1} f \in \mathcal{S}\]

Parseval’s Theorem of Fourier Transform

\[f, g \in \mathcal{S} \implies \int_{-\infty}^{+\infty} F (s) \bar{G} (s) ds = \int_{-\infty}^{+\infty} f (x) \bar{g} (x) dx =\]

where

\[F = \mathcal{F} f, G = \mathcal{F} g\]

Proof:

\[ \begin{align}\begin{aligned}\because g(x) = \int_{-\infty}^{+\infty} e^{2 \pi i s x} G(s) ds\\\therefore \bar{g}(x) = \int_{-\infty}^{+\infty} e^{-2 \pi i s x} \bar{G}(s) ds\end{aligned}\end{align} \]
\[\begin{split}\therefore \int_{-\infty}^{+\infty} f (x) \bar{g} (x) dx & = \int_{-\infty}^{+\infty} \left( f (x) \int_{-\infty}^{+\infty} e^{-2 \pi i s x} \bar{G}(s) ds \right) dx \\ & = \int_{-\infty}^{+\infty} \int_{-\infty}^{+\infty} e^{-2 \pi i s x} f(x) \bar{G} (s) ds dx\end{split}\]

Since everything converges, integrations above are interchangeable.

\[\begin{split}\therefore \int_{-\infty}^{+\infty} f (x) \bar{g} (x) dx & = \int_{-\infty}^{+\infty} \int_{-\infty}^{+\infty} e^{-2 \pi i s x} f(x) \bar{G} (s) dx ds \\ & = \int_{-\infty}^{+\infty} \left( \int_{-\infty}^{+\infty} e^{-2 \pi i s x} f(x) dx \right) \bar{G} (s) ds \\ & = \int_{-\infty}^{+\infty} F(x) \bar{G} (s) ds\end{split}\]

Inference:

\[f, g \in \mathcal{S} \implies \int_{-\infty}^{+\infty} | F (s) |^2 ds = \int_{-\infty}^{+\infty} | f (x) |^2 dx\]

Back to Fourier Transform.