Hardy's uncertainty principle, convexity and Schrödinger evolutions (Q952509)

The authors study unique continuation properties for linear and nonlinear Schrödinger equations \[ \begin{aligned} i\partial_t u+\Delta u=V(x,t)u&\quad \text{in}\;{\mathbb R}^n\times[0,1],\\ i\partial_t u+\Delta u=F(u,\overline{u})&\quad \text{in}\;{\mathbb R}^n\times[0,1]. \end{aligned} \] For the linear equation it is assumed that potential \(V\) is bounded and either \(V(x,t)=V_1(x)+V_2(x,t)\) with \(V_1\) real-valued and \[ \sup_{[0,1]} \Big\| e^{| x| ^2\over (\alpha t+\beta(1-t))^2 }V_2(t)\Big\| _{L^\infty({\mathbb R}^n)} < \infty, \quad\text{or}\quad \lim_{R\to\infty}\| V\| _{L^1([0,1],L^\infty({\mathbb R}^n\backslash B_R))}=0. \] It is shown that if solution \(u\in C([0,1], L^2({\mathbb R}^n))\) to the linear Schrödinger equation is such that \(e^{| x| ^2/\beta^2}u(0)\in L^2({\mathbb R}^n)\), \(e^{| x| ^2/\alpha^2}u(1)\in L^2({\mathbb R}^n)\), \(\alpha>0, \beta>0, \alpha\beta<2\), then \(u\equiv 0\). For the nonlinear Schrödinger equation consider two strong solutions \(u_1,u_2\in C([0,1],H^k({\mathbb R}^n))\), \(k\in{\mathbb Z}^+\), \(k>n/2\), with \(F:{\mathbb C}^2\to{\mathbb C}\), \(F\in C^k\) and \(F(0)=\partial_uF(0)=\partial_{\overline{u}}F(0)=0\). If there exist \(\alpha>0, \beta>0, \alpha\beta<2\) such that \(e^{| x| ^2/\beta^2}(u_1(0)-u_2(0))\in L^2({\mathbb R}^n)\), \(e^{| x| ^2/\alpha^2}(u_1(1)-u_2(1)) \in L^2({\mathbb R}^n)\), then \(u_1\equiv u_2\). The proof is based on the logarithmic convexity of certain quantities, which measure the quadratic exponential decay at infinity and within two characteristic hyperplanes of solutions of Schrödinger evolutions. Corresponding results for heat evolutions are also obtained.