Let $s \in \mathbb{R}$ such that $0<s<1$. Consider the fractional Laplacian $(-\Delta)^s$ in the real line defined via Fourier series as follows: if $f:[-\pi,\pi] \subset \mathbb{R} \longrightarrow \mathbb{C}$ is a periodic function and is written as $$ f(x)=\sum_{n \in \mathbb{Z}} f_n e^{inx} $$ then $$ (-\Delta)^{s/2}f(x)=\sum_{n \in \mathbb{Z}} |n|^{s} f_n e^{inx}. $$

Question. If we define $g: \mathbb{R} \longrightarrow \mathbb{R}$ by $$ g(x):= |f(x)|,\; \forall \; x \in \mathbb{R} $$ then is true that $$ |(-\Delta)^{s/2}g(x)| \leq |(-\Delta)^{s/2}f(x)|? \tag{1} $$

I didn't make any progress as I couldn't get any relation between the Fourier coefficients of $f$ and $g$ (it would be ideal if we had $g_n=|f_n|$ for each $n \in \mathbb{Z}$). There is some relation? The inequality in $(1)$ makes sense?


If true, this would follow from the integral expression for the fractional Laplacian: $$(-\Delta)^{s/2} f(x) = \int_{-\pi}^\pi (f(x) - f(y)) \nu(x - y) dy$$ for an appropriate kernel $\nu$. But, unfortunately, the claimed inequality is false: if, for example, $f$ is a non-zero odd function, then $$(-\Delta)^{s/2} f(0) = 0$$ (by symmetry), while $$(-\Delta)^{s/2} |f|(0) < 0$$ (by the integral expression given above, or by a version of the maximum principle).

On the positive side, we have the following inequality ("Markov property") for the corresponding quadratic (Dirichlet) forms: $$ \langle |f|, (-\Delta)^{s/2} |f| \rangle \leqslant \langle f, (-\Delta)^{s/2} f \rangle . $$

  • $\begingroup$ Nice answer! The $\langle \cdot, \cdot \rangle$ is the $L^2$-inner product? Do you have some reference of the 'Markov Property'? $\endgroup$
    – Guilherme
    Sep 22 at 11:44
  • 1
    $\begingroup$ Well, the Markov property follows directly from $$\langle f, (-\Delta)^{s/2} f\rangle = \frac12 \int_{-\pi}^\pi \int_{-\pi}^\pi (f(y) - f(x)) \nu(y - x) dy dx.$$ And this is essentially one of the properties of a bilinear form required to call it a Dirichlet form. And yes, $\langle\cdot,\cdot\rangle$ is the $L^2$ inner product. $\endgroup$ Sep 22 at 11:52

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.