Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I need to show that $\int_{-\infty}^{\infty}|x-y|^{-p}f(y)\text{d}y<\infty$ for a.e. $x$ when $f\in L^1(\mathbb{R})$ and $p\in (0,1)$. Is the convolution still finite a.e. if instead of $|\cdot|^{-p}$ we use an arbitrary function in $L^1_{loc}(\mathbb{R})$?

share|cite|improve this question
up vote 1 down vote accepted

Hint: Show that $g(x) = \int_{x-1}^{x+1} |x - y|^{-p}f(y)\,dy$ is finite $a.e.$ by showing that $g(x)$ is in $L^1$. Showing the rest of $\int_{-\infty}^{\infty} |x - y|^{-p}f(y)\,dy$ is finite for all $x$ can be done by directly bounding the integrand.

share|cite|improve this answer
I think this is an elegant method. – timur Aug 16 '12 at 15:38

Supposing that $f$ is continuous at $0$, the reason for the integral $\int_{-\infty}^{\infty}|y|^{-p}f(y)\text{d}y$ being finite is that $|y|^{-p}$ is integrable near $y=0$, and that $f$ is globally integrable. So the proof must somehow separate the integral into two integrals.

By the Lebesgue theorem, $$ F(x) = \int_{-\infty}^xf(t)\mathrm{d}t, $$ is differentiable almost everywhere, with $F'(x)=f(x)$. We pick such an $x$, and without loss of generality, let us assume $x=0$. Also, let $\delta>0$. Then $$ \int_{-\infty}^\infty|y|^{-p}f(y)\mathrm{d}y = \int_{-\delta}^\delta|y|^{-p}f(y)\mathrm{d}y + \int_{|x|>\delta}|y|^{-p}f(y)\mathrm{d}y. $$ The second integral in the right hand side can be estimated easily as $$ \int_{|x|>\delta}|y|^{-p}f(y)\mathrm{d}y \leq \delta^{-p}\int_{|x|>\delta}|f(y)|\mathrm{d}y \leq \delta^{-p}\|f\|_{L^1}. $$ For the first integral, we have $$ \int_{-\delta}^\delta\frac{F'(y)}{|y|^{p}}\mathrm{d}y = \left.\frac{F(y)}{|y|^{p}}\right|_{-\delta}^{\delta} + p\int_{-\delta}^\delta\frac{yF(y)}{|y|^{p+2}}\mathrm{d}y. \qquad\qquad(*) $$ The first term in the right hand side is obviously finite. Note that by definition $$ F(y)=F(0)+yF'(0)+o(|y|), $$ and that we can take $F(0)=0$ since the left hand side of ($*$) depends only on $F'$. Taking this into account, we infer $$ \int_{-\delta}^\delta\frac{yF(y)}{|y|^{p+2}}\mathrm{d}y = \int_{-\delta}^\delta\frac{y^2F'(0)+o(|y|^2)}{|y|^{p+2}}\mathrm{d}y =O(\delta^{1-p}), $$ which completes the demonstration.

The answer to the second question is no. You can take $f$ to be a Gaussian, and convolve it with a fast growing function.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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