4
$\begingroup$

What is the proof of the local well-posedness of the quadratic nonlinear Schrödinger equation $\mathrm{i} \,\partial_t u + \Delta u \pm \left|u\right| u = 0$ on the 1D torus in $H^s$ for $s > 1$ (a good reference would suffice)?

$H^s(\mathbb{T})$ is an algebra, but $\left|u\right| u$ is not of the form $u^2$, $\overline{u}u$ or $\overline{u}^2$ and so the LWP doest not immediately follow from the Banach contraction mapping principle.

The LWP should hold according to Tao's webpage (even for $s > 0$). However, the above problem is not covered by Theorem I in [Bo1993] (reference as on Tao's webpage) and in fact Remark (ii) after Proposition 5.73 states that uniqueness of the solutions is unclear.

$\endgroup$
2
  • $\begingroup$ Thanks you for your comment! The way I understand [Bo1993] is as follows: Propostion 5.73 indeed proves LWP in the Besov-type norm (as in (5.76)) but for $\alpha \geq 2$ (defined in (4.1)), which corresponds to at least a cubic nonlinearity. For $\alpha \geq 1$ (which covers quadratic nonlinearities), Bourgain suggests in Remark (ii) to use the usual $X^{s, 1}$ space but mentions that the uniqueness is lost. My question is how to prove the LWP and includes which norm to use. $\endgroup$ Jan 28, 2019 at 11:29
  • $\begingroup$ Ah, I wasn't reading carefully enough when I took a glance just then. That makes a lot more sense. It may be worthwhile to edit your question to include this explanation specifically. $\endgroup$ Jan 28, 2019 at 15:28

1 Answer 1

0
$\begingroup$

On a 1D torus, $H^s$, $s>1$, is embedded into $L^\infty$. A simple proof can be based on the semigroup $e^{i t\Delta}$, i.e., find a solution $u$ satisfying $$ u(t)=e^{i t\Delta}u_0 + i \int_0^t e^{i (t-s)\Delta} |u(s)|u(s) \, d s ,\quad t\in(0,T] , $$ where $u_0\in H^s$ is a given initial value. Given $v\in C([0,T];H^s)$ one can define $u=Mv$ as the solution of $$ u(t)=e^{i t\Delta}u_0 + i \int_0^t e^{i (t-s)\Delta} |v(s)|v(s) \, d s ,\quad t\in(0,T] . $$ Then one can show that the map $M$ is a contraction on the metric space $X$ if $T$ is small, where $$ X=\{v\in C([0,T];H^s): \|v-u_0\|_{H^s}\le 1 \} . $$

In fact, if $u_1=Mv_1$ and $u_2=Mv_2$, with $v_1,v_2\in X$, then $$ u_1(t)-u_2(t) =i \int_0^t e^{i (t-s)\Delta} (|v_1(s)|v_1(s)-|v_2(s)|v_2(s)) \, d s ,\quad t\in(0,T] . $$ Therefore \begin{align} \|u_1-u_2\|_{C([0,T];H^s)} &\le CT \sup_{0\le s\le t\le T} \|e^{i (t-s)\Delta} (|v_1(s)|v_1(s)-|v_2(s)|v_2(s))\|_{H^s} \\ &\le CT \sup_{0\le s\le t\le T} \||v_1(s)|v_1(s)-|v_2(s)|v_2(s) \|_{H^s} \\ &\le CT \|v_1-v_2\|_{C([0,T];H^s)} . \end{align} Therefore, if $T$ is sufficiently small then the map $M$ is a contraction. The key step is that for $s\in[0,2]$ and $\|v_1\|_{H^s}+\|v_2\|_{H^s}\le \|u_0\|_{H^s}+1$ there holds (since $|v_1(s)|v_1(s)$ is a quadratic term) $$ \||v_1(s)|v_1(s)-|v_2(s)|v_2(s) \|_{H^s} \le C \|v_1(s)-v_2(s)\|_{H^s} . $$ However, it may not be true for $s>2$.

$\endgroup$
1
  • $\begingroup$ Thanks! Could you elaborate more on how to obtain the last estimate above? $\endgroup$ Jan 11, 2019 at 14:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.