Discrete Fourier restriction associated with Schrödinger equations (Q2256074)

Let \(p>1\), \(p'=p/(p-1)\) and \(f\in L^p(\mathbb T^{d+1})\), \(d\in\mathbb N\). Its Fourier coefficients may be written in the form \(\hat{f}({\mathbf n},m)\), \({\mathbf n}=(n_1,\dots,n_d)\in\mathbb Z^d\), \(m\in\mathbb Z\). Further \(|{\mathbf n}|=\left(\sum^d_{i=1}|n_i|^2\right)^{1/2}\), \(\|\cdot\|_p\) is a norm in \(L^p(\mathbb T^{d+1})\). The following theorem gives a new proof and sharpening of J. Bourgain's result. Theorem 1.1. For any \(\sigma>0\), \(d\in\mathbb N\) and \(p>4(d+2)/d\) there exists a constant \(C>0\) independent of \(N\) such that for all \(f\in L^p(\mathbb T^{d+1})\) and \(N\in\mathbb N\) \[ \sum_{{\mathbf n}\in\mathbb Z^d}e^{-\sigma|{\mathbf n}|^2/N^2}|\hat{f}({\mathbf n},|{\mathbf n}|^2)|^2\leq CN^{d-2(d+2)/d}\|f\|^2_{p'}. \] The following theorem is applied to estimates of multilinear maximal functions. Theorem 1.7. Let \(N_1,\dots, N_d\in\mathbb N\) and \(S_{N_1,\dots, N_d}(\mathbf{x},t)=\sum_{{\mathbf n}\in S(N_1,\dots, N_d)} e^{2\pi i{\mathbf n}{\mathbf x}}e^{2\pi i|{\mathbf n}|^2t}\), where \(S(N_1,\dots, N_d)=\{{\mathbf n}\in\mathbb Z^d:|n_i|\leq N_i, 1\leq i\leq d\}\). For any \(\varepsilon>0\) there exist a constant \(C\) not depending on \(N_1,\dots, N_d\) such that \[ \|S_{N_1,\dots,N_d}\|_{2(d+2)/d}\leq C(N_1\dots N_d)^{d/(2(d+2))}\max(N_1,\dots,N_d)^{d/(d+2)+\varepsilon}. \]