Minimal smoothness conditions for bilinear Fourier multipliers (Q1955672)

For \(m\in L^\infty(\mathbb R^{2n})\), the bilinear Fourier multiplier operator \(T_m\) is defined by \[ T_m(f_1,f_2)(x)=\frac{1}{(2\pi)^{2n}}\int_{\mathbb R^{2n}} e^{ix\cdot (\xi_1+\xi_2)}m(\xi)\hat f_1(\xi_1) \hat f(\xi_2)\,d\xi_1 d\xi_2. \] For \(s_1,s_2\in \mathbb R\) and \(F\in \mathcal S(\mathbb R^{2n})\), the product type Sobolev norm \(\|F\|_{W^{(s_1,s_2)}(\mathbb R^{2n})}\) is defined by \[ \|F\|_{W^{(s_1,s_2)}(\mathbb R^{2n})} =\left(\int_{\mathbb R^{2n}}\langle \xi_1\rangle^{2s_1} \langle \xi_2\rangle^{2s_2}|\hat F(\xi_1,\xi_2)|^2\,d\xi_1 d\xi_2\right)^{1/2}. \] Take a function \(\Psi\in \mathcal S(\mathbb R^{2n})\) that satisfies \(\text{supp}\Psi\subset \{\xi\in\mathbb R^{2n}; 1/2\leq|\xi|\leq2\}\), \(\sum_{k\in\mathbb Z}\Psi(\xi/2^k)=1\) for \(\xi\in\mathbb R^{2n}\setminus \{0\}\), and, for \(m\in L^\infty(\mathbb R^{2n})\) and \(j\in \mathbb Z\), define \(m_j(\xi)=m(2^j\xi_1,2^j\xi_2)\Psi(\xi_1,\xi_2)\). The authors' main result is the following: Let \(0<p_1,p_2,p\leq \infty\) and \(1/p_1+1/p_2=1/p\). Then \(\|T_m\|_{H^{p_1}(\mathbb R^n)\times H^{p_2}(\mathbb R^n)\to L^p(\mathbb R^n)} \lesssim \sup_{j\in\mathbb Z}\|m_j\|_{W^{(s_1,s_2)}}(\mathbb R^{2n})\), if \[ s_1>\max\Bigl\{\frac{n}{2},\frac{n}{p_1}-\frac{n}{2}\Bigr\}, s_2>\max\Bigl\{\frac{n}{2},\frac{n}{p_2}-\frac{n}{2}\Bigr\}, \text{ and } s_1+s_2>\frac{n}{p_1}\frac{n}{p_2}-\frac{n}{2}, \] and only if \[ s_1\geq\max\Bigl\{\frac{n}{2},\frac{n}{p_1}-\frac{n}{2}\Bigr\}, s_2\geq\max\Bigl\{\frac{n}{2},\frac{n}{p_2}-\frac{n}{2}\Bigr\}, \text{ and } s_1+s_2\geq\frac{n}{p_1}\frac{n}{p_2}-\frac{n}{2}, \] where \(H^{p_1}(\mathbb R^n)\times H^{p_2}(\mathbb R^n)\to L^p(\mathbb R^n)\) is replaced by \(L^\infty\times L^\infty\to \text{{BMO}}\) if \(p_1=p_2=p=\infty\). This gives precise informations on differentiability assumptions, relating to the preceding works on bilinear Fourier multipliers by Coifman-Meyer and Grafakos-Torres.