Suppose $f, g \in L^1(\mathbb{R})$. I want to show that their convolution is continuous.

I can show continuity if one of the functions were in $L^\infty(\mathbb{R})$.

I have tried to approximate one of the functions, say $f$ with $f_M = f\cdot \mathbb{I}\{ |f(x)| \leq M\}$. But I have problem bounding the residual, which would be of form:

$$ \int_\mathbb{R} |f(x-z) - f_M(x-z)|\cdot |g(z)| dz$$

Any help/hint would be greatly appreciated!


This is not true. The counterexample I know is:

$f(x) = g(x) = \begin{cases} x^{-3/4}&0<x< 1\cr 0& \text{otherwise} \end{cases}$

then $\lim_{x \to 0^+}f*f(x) = \infty$ and $\lim_{x \to 0^-}f*f(x)=0$.

I guess you need at least one of $f$ or $g$ to be $L^{\infty}.$

  • 1
    $\begingroup$ $x^{-3/4}\notin L^1.$ $\endgroup$ – zhw. Feb 22 at 0:56
  • $\begingroup$ So I'll just cut it at $1$. $\endgroup$ – Matematleta Feb 22 at 1:04
  • $\begingroup$ Thanks! Do you happen to have a simple way of showing that the $\lim_{x \to 0^+} f*f(x) = \infty$? $\endgroup$ – is it normal Feb 22 at 1:25
  • $\begingroup$ I think we can do this: Note that if $0<x<1,\ f*f(x)=\int^x_0f(x-y)f(y)dy$. Then, since $y<x$, we compute $\frac{1}{x-y}\cdot \frac{1}{y}=\frac{1}{x}\left(\frac{1}{x-y}+\frac{1}{y}\right)\ge \frac{1}{x}\cdot \frac{1}{y}$ and so we get $f*f(x)\ge \frac{1}{x^{3/4}}\int^x_0y^{-3/4}dy=\frac{4}{x^{3/4}}\cdot x^{1/4}=\frac{4}{x^{1/2}}\to \infty $ as $x\to 0^+$. $\endgroup$ – Matematleta Feb 22 at 1:56
  • $\begingroup$ That's perfect! I did an underestimate with rectangles and got $1/\sqrt{x}$ too. $\endgroup$ – is it normal Feb 22 at 2:07

$f*g$ is not a well defined function from $\mathbb R$ to itself, in general. For example, if $f(x)=g(x)=\frac 1 {\sqrt {|x|}}$for $0<|x| \leq1 $ and $0$ elsewhere then $(f*g)(0)=\infty$.

  • $\begingroup$ Thanks for the reply! I can see that convolution of two continuous functions is continuous but I'm not sure how to use the uniformity of limits. I don't think approximation by continuous function is uniform on $L^1(\mathbb{R})$? $\endgroup$ – is it normal Feb 22 at 0:18
  • $\begingroup$ The point is that if you convolve two $L^1$ functions then you still get an a.e. defined $L^1$ function. This follows directly from Young's inequality with $p=q=r=1$. $\endgroup$ – Shalop Feb 22 at 23:27

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.