FALL 2020 Nonlinear Analysis Seminar Series

Associate Professor Dmitriy Stolyarov, St. Petersburg State University and St. Petersburg Department of Steklov Mathematical Institute

Title: Hardy--Littlewood--Sobolev inequality for $p=1$

Abstract:

The classical Sobolev embedding theorem says that the inequality $\|f\|_{L_q(\mathbb{R}^d)}\lesssim \|\nabla f\|_{L_p(\mathbb{R}^d)},\quad f \in C_0^\infty(\mathbb{R}^d)$ holds true provided $\frac{1}{p} - \frac{1}{q} = \frac{1}{d}$ and $1 \leq p < d$. The original Sobolev's proof was based on the Hardy--Littlewood--Sobolev (HLS) inequality $\|\mathrm{I}_\alpha [g]\|_{L_q(\mathbb{R}^d)} \lesssim \|g\|_{L_p(\mathbb{R}^d)},\quad \frac{1}{p} - \frac{1}{q} = \frac{\alpha}{d},\ 1 < p < q < \infty,$ here $\mathrm{I}_{\alpha}$ is the Riesz potential of order $\alpha$, i.e. a Fourier multiplier with the symbol $|\cdot|^{-\alpha}$. It is easy to see by plugging $g = \delta_0$ (the Dirac's delta) in the role of $g$ that the HLS inequality is false at the endpoint $p=1$. However, the Sobolev embedding is true in this case, as it was proved by Gagliardo and Nirenberg. The folklore principle, supported by the results of Bourgain--Brezis, Van Schaftingen, and many others, says that the HLS inequality becomes valid when we somehow separate the function $g$ from the set of delta-measures. We will discuss this effect in more details and state new results in this direction.

Please click here to join Meeting ID: 994 5746 3755