I write the exact statement of the problem:

Show that for any $g \in L^1$ and $f ∈ L^p(\mathbb{R})$, p $\in (1, \infty)$, the integral for f ∗g converges absolutely almost everywhere and that $∥f ∗ g∥_p ≤ ∥g∥_1∥f∥_p$.

I've tried to show that the convolution is well defined for f,g $\in L^1$ in this way:


I've used Tonelli's theorem since the function is non-negative


Now I've thought that since $L^1 \subset L^p$ this is true also for f $\in L^p$ (can I do it? )

I've tried with some colleagues something with Hölder's inequality. I've chosen a q $\in$ (1,$\infty$) such that $\frac{1}{p}+\frac{1}{q}=1$

$$∥f ∗ g∥_p^p=\int_{\mathbb{R^d}}|\int_{\mathbb{R^d}}f(x-y)g(y)dy \space |^pdx=\int_{\mathbb{R^d}}\space |\int_{\mathbb{R^d}}f(x-y)g(y)^{\frac1p}g(y)^{\frac1q}dy\space |^pdx$$

$$\le\int_{\mathbb{R^d}}\space \int_{\mathbb{R^d}}|f(x-y)g(y)^{\frac1p}g(y)^{\frac1q} \space |^pdy\space dx=\int_{\mathbb{R^d}}\space \int_{\mathbb{R^d}}|f(x-y)^pg(y)g(y)^{\frac{p}{q}}|\space dy\space dx$$

$$\int_{\mathbb{R^d}}\space \int_{\mathbb{R^d}}|f(x-y)^pg(y)g(y)^{\frac{p}{q}}| \space dy\space dx=\int_{\mathbb{R^d}}\space \int_{\mathbb{R^d}}|f(x-y)^pg(y)^p| \space dy\space dx$$

$$= \int_{\mathbb{R^d}}\space \int_{\mathbb{R^d}}|f(x-y)^pg(y)^p| \space dx\space dy=∥g∥_1^p∥f∥_p^p \implies ∥f ∗ g∥_p ≤ ∥g∥_1∥f∥_p$$

This works is it not?

  • 1
    $\begingroup$ $L^1(\mathbb{R})$ isn't contained in $L^p(\mathbb{R})$. You need a bit of Hölder [there probably are other ways too] to reach the conclusion. $\endgroup$ Jan 16 '14 at 16:47
  • $\begingroup$ The problem with inequalities such as $L^p(B) \subset L^1(B)$ for $p>1$ is that the proof requires $B$ to be of finite measure. In that case, $p<q$ implies $L^q(B)\subset L^p(B)$ (by an adequate use of Hölder's inequality). Just write $|f(x)|=(|f(x)|^p)^{1/p} 1^{1/q}$ and apply Hölder's ineq. $\endgroup$ Jul 23 '14 at 9:01

Now I've thought that since $L^1 \subset L^p$ this is true also for f $\in L^p$ (can I do it? )

No because we don' t have the inclusion $\mathbb L^1(\mathbb R)\subset\mathbb L^p(\mathbb R)$ (consider $x^{1/p}\chi_{(0,1)})$.

However, we can use Jensen's inequality for the probability measure $\mu$ on the real line, given by $$\mu(B)=\frac 1{\lVert g\rVert_1}\int_B|g(x)|\mathrm dx.$$ We have $$\left(\int_{\mathbb R}|f(x-y)|\cdot |g(y)|\mathrm dy\right)^p =\left(\int_{\mathbb R}|f(x-y)|\mathrm d\mu(y)\right)^p\lVert g\rVert_1^p\leqslant\int_{\mathbb R}|f(x-y)|^p|\mathrm d\mu(y)\lVert g\rVert_1^p.$$ Now integrate over $x$ with respect to Lebesgue measure and use Fubini-Tonelli's theorem to conclude.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.