One way to define the maximal Hilbert transform of a function, $f$, is by $$\mathcal{H}[f](x):=\sup_{\varepsilon>0} \left| \int_{|x-t|\geq\varepsilon} \frac{f(t)}{x-t} \, dt\right|, \quad x\in\mathbb{R},$$ and $\mathcal{H}$ has been shown to be bounded from $L_p(\mathbb{R})\to L_p(\mathbb{R})$, $1<p<\infty$.

If one only considers the Hilbert transform (i.e. the principal value of the integral over $\mathbb{R}$ without absolute value on the outside), then one can show that an iterated version of the Hilbert transform in higher dimensions is $L_p(\mathbb{R}^d)\to L_p(\mathbb{R}^d)$ bounded. I can think of two ways to formulate a similar maximal version. Restricting our attention to two dimensions, consider the first:

$$\mathcal{H}_1[f](x_1,x_2):= \sup_{\varepsilon>0} \left|\int_{|x_2-t_2|\geq\varepsilon} \int_{|x_1-t_1|\geq\varepsilon} \frac{f(t_1,t_2)}{(x_1-t_1)(x_2-t_2)} \, dt_1 \, dt_2\right|.$$

Or one could consider different parameters:


The question then is: are these bounded on $L_p(\mathbb{R}^2)$? I thought there would be a fairly simple way to use the univariate result, but so far such an argument has eluded me.

  • $\begingroup$ Sorry I wrote that kind of off the cuff, I edited the formula to relay what I intended. This does look close to the Riesz transform, but slightly different. $\endgroup$ – Keaton Hamm Feb 26 '15 at 5:28
  • $\begingroup$ I'm a little confused--I think in the one-variable definition you want the region of integration to be $|x-t|\geq \epsilon$. But then the corresponding regularization of the multivariate version doesn't work, since the region $\|x-t\|_2\geq \epsilon$ still contains points where the denominator of the integrand vanishes. Indeed, it seems one would have to remove a neighborhood of the union of the coordinate hyperplanes (translated by $x$), and of course any such region is unbounded. $\endgroup$ – Mike Jury Apr 10 '15 at 20:22
  • $\begingroup$ You are right, that is something I missed. If one modifies the definition to look at something like an iterated version of $\mathcal{H}$ then this I think is well defined, and possibly sufficient to consider. I will try to formulate it correctly and edit the question later. $\endgroup$ – Keaton Hamm Apr 10 '15 at 21:39

The maximal Beurling transform seems to be such an analog in two dimensions: P. Z. Agranovich and V. N. Logvinenko use it in https://zbmath.org/?q=py:1991+ai:agranovich.p-z to obtain the two-dimensional version of the one-dimensional result from https://zbmath.org/?q=an:03396282 , where the maximal Hilbert transform is applied.


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.