$$ iu_t + \Delta u = \Phi \cdot \nabla u$$

with initial data $u_0 \in H^s(\mathbb{R}^d)$ and $\Phi\colon \mathbb{R}^d \rightarrow \mathbb{R}^d$ sufficiently regular. Suppose I want to use Strichartz:

$$\|e^{it\Delta} u\|_{L^p W^{s,q}} \lesssim \|u_0\|_{H^s} \\ \|\int_0^t e^{i(t-s)\Delta} (\Phi \cdot\nabla u)(s) ds \|_{L^p W^{s,q}}\lesssim \|\Phi \cdot\nabla u\|_{L^{\tilde{p}'} W^{s,\tilde{q}'}}$$

where $(p,q),(\tilde{p},\tilde{q})$ admissible and $s\geq 0$ sufficiently large. Normally I want to close the argument in $L^pW^{s,q} \cap L^{\infty}H^1$ and use Banach's fixed point theorem. However I encounter a loss of derivative when trying to estimate

$$ \|\Phi \cdot\nabla u\|_{L^{\tilde{p}'} W^{s,\tilde{q}'}} $$

because of $\nabla u$ and so I always have to put $u$ in a space with $s+1$ derivatives. Is there a way to circumvent this and be able to use Strichartz estimates?

  • 1
    $\begingroup$ You may find some of the results and techniques in arxiv.org/pdf/math/0509314.pdf to be helpful for your problem. See also arxiv.org/pdf/1309.3291.pdf for how one can construct local solutions to Schrodinger equations with derivatives (but using local smoothing estimates in place of Strichartz estimates). $\endgroup$
    – Terry Tao
    Commented Oct 1, 2022 at 2:56
  • $\begingroup$ Thanks for the references. Indeed I know Stefanovs paper (a bit) but one has to go rather deep into Besov territory, right? I was just wondering if there is a „quick and easy“ way to get LWP for sufficiently smooth initial data. $\endgroup$ Commented Oct 1, 2022 at 6:46

1 Answer 1


It is not clear to me how you plan to close your argument unless you assume that the coefficient $\Phi$ is sufficiently small. I think it is not easy to get rid of first order terms by a perturbative approach without smallness.

In some cases, large first order terms can be handled by including them in the operator. E.g., Strichartz estimates are available for some equations of the form $iu_{t}+Au=0$, where $A$ is a selfadjoint operator like $A=-\Delta+i \nabla \cdot(a u)+ia \cdot \nabla u$, provided $a(x)$ is smooth enogh and decays at infinity (but it can be large). However, your equation does not seem to fit into this form.

Let me also mention that there exists another class of estimates, the so called smoothing estimates, which ensure a gain of one full derivative w.r.to the forcing term, and may be possibly useful in your case. Note that smoothing estimates are weaker than Strichartz, indeed they only allow to control a weighted $L^{2}$ norm of the solution.

  • $\begingroup$ I know about some of the "magnetic Strichartz estimates" in the literature (e.g. Stefanov as mentioned by Terry Tao). The motivation for my question is if there is any "brute force" (in the sense of "not optimal") way to obtain LWP without having to resort to magnetic Strichartz estimates. $\endgroup$ Commented Oct 1, 2022 at 10:48
  • $\begingroup$ One brute force attack might be to consider the full space-time operator $iu_t+\Delta u=Lu$, which can be inverted as an operator on suitable Strichartz spaces. Then your problem would be to invert $L+V(x)$, and for this problem there exist several techniques, notably Fredholm theory $\endgroup$ Commented Oct 1, 2022 at 19:34

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.