I am trying to prove that for two integrable functions $f,g: \mathbb{R}^n \rightarrow \mathbb{R}$ the function $y \mapsto f(x-y)g(y)$ is integrable for almost every $x$. By using the holder inequality I reduced this to showing that if a function is integrable then also its square is integrable but after browsing a bit I found this so I guess this leads nowhere. Any hints are welcomed.
Asked
Active
Viewed 1,934 times
6
-
This is called the convolution. It is very important. Try to derive some of the other properties found at https://en.wikipedia.org/wiki/Convolution – Euler....IS_ALIVE Dec 28 '15 at 03:00
-
The function $x\mapsto\sqrt x$ on the interval $(0,1)$ is integrable but its square is not. ${}\qquad{}$ – Michael Hardy Dec 28 '15 at 03:44
1 Answers
8
Hint: consider $\int\int |f(x-y)g(y)| \, dy \, dx$, and use Tonelli's theorem to reverse the order of integration. If you can show this integral is finite, then $\int |f(x-y)g(y)| \, dy$ is finite for almost every $x$.
kccu
- 20,808
- 1
- 22
- 41
-
Tonelli says $$\int\int |f(x-y)g(y)| , dy , dx = \int\int |f(x-y)g(y)| , dx , dy.$$ The inside integral is plainly finite if $f$ is integrable and $g(y)$ is finite. But how does that imply the outside integral is finite? ${}\qquad{}$ – Michael Hardy Dec 28 '15 at 06:42
-
2Once you've switched the order of integration, $|g(y)|$ is just a constant in the inner integral, so you may bring it outside the (inner) integral (but still inside the outer integral)... – kccu Dec 28 '15 at 13:43
-
But you still have the factor $y\mapsto \int |f(x-y)|,dx$, so the integrability of $g$ is not enough at that point. ${}\qquad{}$ – Michael Hardy Dec 28 '15 at 17:00
-
2You also need to use the translation-invariance of the Lebesgue integral and the integrability of $f$. – kccu Dec 28 '15 at 20:47
-
aha! I guess I was distracted by other aspects of the problem. ${}\qquad{}$ – Michael Hardy Dec 28 '15 at 22:27