0

Asssume that the following integral exists. Does there exists a non-trivial bounded measurable function $f(x)$ such that for all $y>0$, $$\int_0^\infty f(x) e^{-xy} dx=0?$$

If the the function $f(x)$ satisfies conditions(like being of exponential order)of Laplace transform, then $f(x)=0$ by the uniqueness of Laplace transforms.

stephan
  • 543
  • 4
    $f$ must be zero a.s. by the uniqueness of the Laplace transform (no need to assume that $f$ is of the exponential type). That is, if $f$ has a Laplace transform and is $0$, then $f$ just be $0$ a.s. Here is a posting about the invertibility of Laplace transform under several conditions. – Mittens Aug 21 '24 at 15:52
  • Got it. Thanks. – stephan Aug 21 '24 at 16:55
  • Hi @Mittens, I have a question about the rigorousness of your argument (not saying you are wrong, so don't get me wrong :) ). LT is defined for a complex variable and is valid in the region of convergence. Here, the question only talks about $y \gt 0$. If the LT is of the form where it is non-zero for left half of complex plane and zero in the right half of complex plane (specific interest is positive axis of $\sigma$ in $s = \sigma +i t$, the complex variable), it may invert to a non-zero $f(x)$. That is, unless we prove that such LT can never exist (in which case, argument becomes rigorous) – Srini Aug 23 '24 at 23:41
  • @Srini: What the OP has is the Laplace transform of a function $f$. I think some minimal assumptions in $f$ regarding local integrability for this to exist. Once can also look at this in the sense of distributions but I rather stay within the real of integration theory. Under the minimal conditions of local integrability you can use methods of complex analysis to show that the inverse Laplace transform of $f$ is $0$. See my link https://math.stackexchange.com/a/4707969/121671. – Mittens Aug 23 '24 at 23:51
  • What OP has is not LT strictly, no?. LT in traditional assumes $f(x) = 0$ for $x \lt 0$. In fact some texts explicitly use $f(x) \theta(x)$ (heaviside). If $f(x) = \theta(-x)$, OP's integral would vanish. Sorry I am confused. – Srini Aug 24 '24 at 00:21
  • Only the integration limit is over $(0,\infty)$. It says nothing about what $f(x)$ is for $x \lt 0$. Anyway, I have posted my comment as an answer now. Waiting for a barrage of downvotes :( – Srini Aug 24 '24 at 00:55

2 Answers2

1

Saying that $\int_0^{\infty}e^{-xy}f(x)dx=0$ for all $y>0$ postulates that $\int_0^{\infty}e^{-xy}|f(x)|dx<\infty$ for all $y>0$. With the notation $f=f^+-f^-$ (with $f^+=\max (f,0)$ and $f^-=(-f)^+$) we have $$L_+(y)=\int_0^{\infty}e^{-xy}f^+(x)dx=\int_0^{\infty}e^{-xy}f^-(x)dx=L_-(y)$$ One can prove without great effort that $L_{\pm}$ are extendable to two unique analytic functions on $H=\{z; \Re z>0\}$. Since they coincide on the real line, from the principle of isolated zeros they coincide on $H$. In particular $L_+(y_0+it)=L_-(y_0+it)$ for all real $t$, which means that the two Fourier transforms of the bounded positive measures $e^{-xy_0}f^{\pm}(x)dx$ coincide and from the Paul Levy theorem the measures themselves coincide. Therefore $f^+=f^-$ almost everywhere, meaning $f=0$ almost everywhere.

0

Since there is no condition on $f(x)$ for $x \lt 0$ in your question, for $f(x) = g(x) \theta(-x)$, where $\theta$ is Heaviside function and $g(x)$ is to be constructed to satisfy boundedness, non-trviality and measurablility, your integral

$$\int\limits_{0}^{\infty} f(x) e^{-xy} dx = \int\limits_{0}^{\infty} g(x) \theta(-x) e^{-xy} dx = 0$$ for $y \gt 0$.

Clearly $f(x)$ is not identically $0$ over the real axis.

However, if you were to edit your question to add that $f(x) = 0$ for $x \lt 0$ as an additional condition, comment from Mittens is valid. Laplace Transform is analytic in the region of convergence. If we can prove that no analytic function exists such that it is identically zero on the positive x-axis but non-zero elsewhere in the complex plane, it would rigorously prove the invertibility of LT would ensure that $f(x) = 0$.

Srini
  • 1,405
  • :It is irrelevant what is on (−∞,0), the integration is over (0,∞) and the inverse Laplace transform is on (0,∞). – Mittens Aug 24 '24 at 01:34