Fourier transform with different measure

1.4k Views Asked by At

I am interested in the following question: can we generalize Fourier theory to different $L^2(\mu)$ spaces, not just the Lebesgue measure? For example on $\mathbb{R}$, defining the Fourier transform of $f$ as $$ \mathcal{F}[f](\omega)=\int_\mathbb{R}f(x)e^{-i\omega x}d\mu(x). $$

Is there literature on this subject? Do any properties hold with this (in particular Plancherel’s theorem)? I have tried to look this idea up online but found nothing.

I am a beginner when it comes to Fourier theory, I apologize if it is a simple question.

1

There are 1 best solutions below

10
On BEST ANSWER

Yes, in fact one can define it on measures on topological groups (besides integrable functions with respect to Haar measures). In any event, lets just focus on $\mathbb{R}^n$. There, the Fourier transform is well define on the space of complex (or finite real valued) Borel measures $\mathcal{M}(\mathbb{R}^n)$. There is also an inversion formula that shows that the Fourier transform a one to one function on $\mathcal{M}$. The Fourier transform in this setting is often used in probability theory under the name of characteristic function.

Definition: Let $\mu$ be a Complex measure on $(\mathbb{R}^d,\mathscr{B}(\mathbb{R}^d))$. The Fourier transform or characteristic function of $\mu$ is the function defined as $$\widehat{\mu}({\bf t})=\int\exp(i{\bf t}\cdot{\bf x})\mu(d{\bf x})$$


When $\mu\ll\lambda_d$ and $\mu=f\cdot\lambda_d$, then the Fourier transform of $\mu$ is closely related to the Fourier transform of $f$. and $\hat{\mu}(-2\pi i t)=\hat{f}(t)$.


Here are a few important results on uniqueness:

Theorem: Suppose that $\mu$ and $\nu$ are complex measures (measures of finite variation) on $\mathscr{B}(\mathbb{R}^d)$. Then, $\mu=\nu$ iff $\widehat{\mu}=\widehat{\nu}$.

Theorem: Let $\mu$ be a complex Borel measure on $\mathbb{R}$ and let $\widehat{\mu}$ be its characteristic function. Then

  1. For any $-\infty<a<b <\infty$, $$ \begin{align} \mu((a,b))+\frac12\mu(\{a,b\})=\frac{1}{2\pi} \lim_{T\rightarrow\infty}\int^T_{-T}\int^b_a e^{-iyt}\widehat{\mu}(t)\,dy\,dt \tag{1}\label{inversion} \end{align} $$

$$ \mu(\{a\})=\lim_{T\rightarrow\infty}\frac{1}{2T}\int^T_{-T} e^{-iat}\widehat{\mu}(t)\,\tag{2}\label{two} $$ 2. If $f\in\mathcal{L}_1(\mathbb{R},\lambda)$ then $$\begin{align} f(x)=\frac{1}{2\pi}\lim_{T\rightarrow\infty}\int^T_{-T}e^{-iyt}\widehat{f}(-t/2\pi)\,dt\tag{3}\label{inversion3} \qquad\text{a.s.} \end{align} $$ 3. If $\widehat{\mu}\in\mathcal{L}_1$, then $\mu\ll\lambda$ and $$\begin{align} \frac{d\mu}{d\lambda}(y)=\frac{1}{2\pi}\int e^{-ity}\widehat{\mu}(t)\,dt\qquad\text{a.s.}\tag{4}\label{density-prob} \end{align} $$

With regards smoothness:

Theorem: Let $\mu$ be a finite positive measure on $(\mathbb{R}^n,\mathscr{B}(\mathbb{R}^n))$. If $\partial^\alpha\widehat{\mu}(0)$ exits and is finite for all $|\alpha|=2m$ then, $\widehat{\mu}\in\mathcal{C}^{2m}(\mathbb{R}^n)$; furthermore, for all $\alpha\in\mathbb{Z}^n_+$ with $|\alpha|=2m$, $\int |x^\alpha|\,\mu(dx)<\infty$ and $\partial^\alpha\widehat{\mu}(t)=i^{|\alpha|}\int x^\alpha e^{ix\cdot t}\,\mu(dx)$.

Theorem: Suppose that $\mu$ is a complex measure on $(\mathbb{R}^n,\mathscr{B}(\mathbb{R}^n)$. If $$ \int_{\mathbb{R}^n}|x_j|^m|\mu|(dx)<\infty, $$ then the partial derivative $\partial^m_j\widehat{\mu}$ exists, is uniformly continuous, and $$\begin{align} \partial^k_j \widehat{\mu}(t)=i^k\int_{\mathbb{R}^n} x^k_j e^{i x\cdot t}\mu(dx),\quad 0\leq k\leq m. \end{align}\tag{1}\label{deriv-m-charac} $$ Moreover, if $|x|^m=\Big(\sum^n_{j=1}x^2_j\Big)^{\frac{m}{2}}\in\mathcal{L}_1(|\mu|)$, then $\widehat{\mu}\in\mathcal{C}^m(\mathbb{R}^n)$, and $$\begin{align} \widehat{\mu}(t)= \sum_{0\leq|\alpha|\leq m}\frac{i^{|\alpha|}}{\alpha!}t^\alpha \int x^\alpha\,\mu(dx) + o(|t|^m)\tag{2}\label{otaylor} \end{align}$$

Theorem: (Bochner--Herglotz) $\varphi:\mathbb{R}^d\rightarrow\mathbb{C}$ is the characteristic function of a finite nonnegative measure $\mu$ in $\mathscr{B}(\mathbb{R}^d)$ iff $\varphi$ is a bounded positive definite continuous function.