The Hörmander multiplier theorem. I: The linear case revisited (Q1746514)

Let \(f\) be a Schwartz function on \(\mathbb{R}^n\). We consider the linear multiplier operator \(T_{m}\) associated with a bounded function \(m\) defined by \[ T_{m}(f)(x) = \int_{\mathbb{R}^n} \widehat{f}(\xi) \, m(\xi) e^{2\pi i x \cdot \xi}\, d\xi. \] For \(s>0\) let \(L^r_s\) space be the standard Sobolev space of all functions \(f\) on \(\mathbb{R}^n\) with norm \[ \|f\|_{L^r_s}:= \|(1-\triangle)^{s/2}f\|_{L^r} < \infty. \] Let \(\Psi\) be a Schwartz function, whose Fourier transform is supported on \(\{\xi: 1/2< |\xi| < 2\}\) and which satisfies \(\sum_{j}\widehat{\Psi}(2^{-j}\xi)=1\) for all \(\xi \neq 0\). Hörmander's extension of Mikhlin's theorem says that if \(1< r \leq 2\) and \(s>n/r\), a bounded function \(m\) satisfies \[ \sup_{k \in \mathbb{Z}} \| \widehat{\Psi}m(2^k \cdot) \|_{L^r_s} < \infty,\tag{1} \] then \(T_{m}\) is bounded on \(L^p(\mathbb{R}^n)\) for all \(1< p< \infty\) and of weak type \((1,1)\). It is known that \textit{A. P. Calderon} and \textit{A. Torchinsky} [Adv. Math. 24, 101--171 (1977; Zbl 0355.46021)] showed the \(L^p\) boundedness of operators \(T_{m}\) for \[ {\big|}\frac{1}{p}-\frac{1}{2}{\big|} < \frac{s}{n}\tag{2} \] and \({\big|}\frac{1}{p}-\frac{1}{2}{\big|} \leq \frac{1}{2}\). In this paper the authors consider the necessity conditions for the \(L^p\) boundedness of operators \(T_{m}\) that satisfy the hypotheses (1) of Hörmander multiplier theorem. More precisely, they prove the following theorem: Fix \(1 < r < \infty\) and \(0 < s \leq n/2\) such that \(rs > n\). Suppose that (1) holds. Then \(T_{m}\) maps \(L^p(\mathbb{R}^n)\) to \(L^p(\mathbb{R}^n)\) for all \(1< p< \infty\) such that \(|\frac{1}{p}-\frac{1}{2}|< \frac{s}{n}\). Moreover, if \(T_{m}\) maps \(L^p(\mathbb{R}^n)\) to \(L^p(\mathbb{R}^n)\) for all \(m\) such that (1) holds, then we must necessarily have \(|\frac{1}{p}-\frac{1}{2}| \leq \frac{s}{n}\). \noindent In order to prove it, they provide certain counterexamples related to the optimality of the hypotheses \(rs > n\) and \(|\frac{1}{p}-\frac{1}{2}|< \frac{s}{n}\). As for the endpoint case \(|\frac{1}{p}-\frac{1}{2}|=\frac{s}{n}\), they show \(L^p \rightarrow L^{p,2}\) estimates by replacing the Sobolev norm as in the condition (1) by some Besov norm when \(1< p < 2\).