On \(L^p\) estimates for square roots of second order elliptic operators on \(\mathbb R^n\) (Q1432099)

Consider \(A\) an \(n\times n\) matrix of \(L^\infty({\mathbb R}^n)\) functions satisfying the accretivity condition \[ \lambda | \xi |^2 \leq \Re\, A\xi \cdot\overline{\xi}, \; | A \xi\cdot \overline{\zeta} | \leq \Lambda | \xi| \, |\zeta|, \quad \xi,\zeta\in {\mathbb C} \] for some \(\lambda\), \(\Lambda\) such that \(0<\lambda\leq \Lambda <\infty\). Then \[ L f=-\text{div } (A\nabla f) \] can be seen as a maximal accretive operator in \(L^2({\mathbb R}^n)\), and the square root \(L^{1/2}\) is defined in the sense of maximal accretive operators. The Kato conjecture, that \[ \| L^{1/2} f\|_{L^2({\mathbb R}^n)}\sim \| \nabla f \|_{L^2({\mathbb R}^n)}, \quad n\geq 1 \] has a positive answer (see \textit{R. R~Coifman}, \textit{A.~McIntosh} and \textit{Y.~Meyer} [Ann. Math. (2) 116, 361--387 (1982; Zbl 0497.42012)] for \(n=1\), \textit{S.~Hofmann} and \textit{A.~McIntosh} [Publ. Mat., Barc. 2002, Spec. Vol., 143--160 (2002; Zbl 1020.47031)] for \(n=2\), and \textit{P.~Auscher}, \textit{S.~ Hofmann}, \textit{M.~Lacey, Michael}, \textit{A.~McIntosh} and \textit{Ph.~Tchamitchian} [Ann. Math. (2) 156, No. 2, 633--654 (2002; Zbl 1128.35316)] for general \(n\)). In the paper under review the norms of \(L^{1/2} f\) and \(\nabla f \) in \(L^p\) are compared, for some \(p\neq 2\), as a part of the programme initialized by the present author and \textit{Ph.~Tchamitchian} in [Square root problem for divergence operators and related topics (Astérisque 249, Société Mathématique de France, Paris) (1998; Zbl 0909.35001)]. In fact one proves that \[ \| L^{1/2} f\|_{L^p({\mathbb R}^n)}\sim \| \nabla f \|_{L^p({\mathbb R}^n)} \] whenever \[ \sup \big(1, \frac{2n}{n+4} -\epsilon\big) <p< \frac{2n}{n-2} +\epsilon. \] The method involves proving weak estimates by using properties of the semigroup \(\text{e}^{-tL}\) and relies on a Calderón-Zygmund decomposition for locally integrable functions with \(L^p\) gradients. The method is also generalized to high-order operators.