0
$\begingroup$

It is well-known that the Laplace transform, defined by $$\mathcal{L} \colon f(x) \in L^2(\mathbb{R}_+) \to \hat{f}(\xi) \in L^2(\mathbb{R}_+)$$ via $$\hat{f}(\xi) = \int_{\mathbb{R}_+} f(x)\,\mathrm{e}^{-\xi x} \mathrm{d}x,$$ is a linear and bounded operator with $\|\hat{f}(\xi)\|^2_{L^2(\mathbb{R}_+)} \leq \pi\,\|f(x)\|^2_{L^2(\mathbb{R}_+)}$ (actually, it can be shown that the $L^2$ operator norm of $\mathcal{L}$ is $\|\mathcal{L}\| = \sqrt{\pi}$). Now I am wondering if a reversed inequality can be shown, namely, can we find a generic positive constant $C > 0$ such that $$\|\hat{f}(\xi)\|^2_{L^2(\mathbb{R}_+)} \geq C\,\|f(x)\|^2_{L^2(\mathbb{R}_+)} \tag{1}\label{1}$$ for all $f \in L^2(\mathbb{R}_+) $? I dig harder into the literature and it seems that I can not find relevant materials.


Edit: I forgot to mention that I am only interested in the validity of \eqref{1} when $f$ is a real-valued function.

$\endgroup$
7
  • 2
    $\begingroup$ There should probably be a $dx$ in he integral, not $d\xi$. Anyway, the answer is no, consider for example $f_n(x) = e^{inx}\chi_{(0, 1)}(x)$. Then $||f_n||_{L^2} = 1$, but direct computation should give that $||\hat{f_n}||_{L^2}\to 0$. $\endgroup$ Commented Sep 30, 2023 at 2:02
  • 1
    $\begingroup$ The Laplace transform has absolutely continuous spectrum $[-\sqrt\pi,\sqrt\pi]$ of multiplicity 1, in particular its inverse is unbounded and no inequality as that you ask for holds. You can find more general results here: www2.math.upenn.edu/~cle/papers/laplce_rev2.pdf -cf p6 for the Laplace transform. It may also be possible to deduce that from Mackey's paper pnas.org/doi/pdf/10.1073/pnas.34.4.156 with his notion of (not) "strongly in $L^2$". You can find here conditions under which the Laplace transform is invertible: rose-hulman.edu/~bryan/invlap.pdf $\endgroup$
    – plm
    Commented Sep 30, 2023 at 2:06
  • $\begingroup$ @AlekseiKulikov Hello Aleksei, thank you for pointing out a typo in the OP, I forgot to mention that I am only interested in real-valued functions (see the edited version of my post). $\endgroup$
    – Fei Cao
    Commented Sep 30, 2023 at 2:18
  • $\begingroup$ Come on, just replace it with $sin(nx)\chi_{(0, 1)}(x)$ then $\endgroup$ Commented Sep 30, 2023 at 2:24
  • $\begingroup$ @plm Thank you for pointing out these illuminating references $\endgroup$
    – Fei Cao
    Commented Sep 30, 2023 at 2:26

1 Answer 1

2
$\begingroup$

A comment but I amn’t entitled. This is well trod territory. The one-sided Laplace transform as in your example is an equivalence between the $L^2$-space on the positive axis and the Hardy space on the right half plane, i.e., a suitable Hilbert space of analytic functions there which satisfy appropriate growth conditions.

A precise formulation can be found on p. 131 of the 1962 classic “Banach Spaces of Analytic Functions” by K. Hoffman, which is readily available online.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .