Local wellposedness of the 2d Anderson-Gross-Pitaevskii equation
S. Mackowiak · 2025
Read Local wellposedness of the 2d Anderson-Gross-Pitaevskii equation by S. Mackowiak (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-17063.txt.
Introduction
Samaël Mackowiak proves that a rough version of the Gross-Pitaevskii equation, a model of quantum fluids and condensates, behaves well in two dimensions, at least for short times. The roughness comes from the potential term, which can be as wild as spatial white noise. With such noise, the equation needs renormalization before it even makes sense, and standard methods fail.
Mackowiak first builds the associated Schrödinger operator from its quadratic form, then studies how smooth the functions in its domain are. Those facts feed a paracontrolled approach, a technique for taming rough equations, which yields Strichartz estimates. A contraction argument then gives local wellposedness: solutions exist, are unique, and depend sensibly on their starting data. The result widens the class of noisy quantum models on firm mathematical ground.
Catalog record
| CREATOR | S. Mackowiak |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 127 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-17063
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Local wellposedness of the 2d Anderson-Gross-Pitaevskii equation S. Mackowiak Abstract In this paper, the local wellposedness of a general Gross-Pitaevskii equation with rough potential is proven in dimension 2. The class of rough potentials we are considering is large enough to contain the spatial white noise and thus a renormalization procedure may be needed. We first construct the associated Schrödinger operator from its quadratic form. Then, the regularity of elements of its domain is explored. This allows to use a paracontrolled approach in order to obtain Strichartz estimates, which are used to prove the local wellposedness by a contraction argument. 1 Introduction In this paper, we are interested in solving the renormalized Gross-Pitaevskii equation with a real spatial white noise potential \xi in two dimensions of space {~}\begin{cases}\mathrm{i}\partial_{t}u+Hu+\xi u+\lambda\left|u\right|^{2\gamma}u=0\\ u(0)=u_{0},\end{cases} (1.1) where u=u(t,x) is a complex-valued space-time random field defined on I\times\mathbb{R}^{2}_{x} for some time interval I , H=\Delta-|x|^{2} is the Hermite operator, \lambda\in\mathbb{R} , and \gamma>0 . The study of such equation is motivated by the modelling of Bose-Einstein condensation. In fact, the deterministic equation, which is a non-linear Schrödinger equation with confining potential, is an extensively studied dispersive equation known as the Gross-Pitaevskii equation. It appears in the study of Bose-Einstein condensates as a mean-field limit to the coupled linear many-body Schrödinger equation with confining potential (see [ 10 ] ). The Gross-Pitaevskii equation with random potential can be seen as a model for Bose-Einstein condensates in presence of inhomogeneities (see for example [ 15 , 17 ] and references therein). Hence, the white noise potential can be seen as a toy model of random potential with null correlation length. In [ 25 ] the author hinted this equation may not admit solutions, due to the low regularity of the spatial white noise, even for smooth and localised initial data. Instead, the author solved the equation {~}\begin{cases}\mathrm{i}\partial_{t}u+:H+\xi:u+\lambda\left|u\right|^{2}u=0\\ u(0)=u_{0},\end{cases} (1.2) where :H+\xi: is the Wick renormalization of the Hermite operator with spatial white noise potential. The renormalized operator is defined through an exponential transform inspired by the one introduced in [ 21 ] to study the continuous parabolic Anderson model, and later used in [ 12 , 36 , 37 ] to solve the non-linear Anderson-Schrödinger equation on the torus \mathbb{T}^{2} and in [ 11 , 13 ] on the full space \mathbb{R}^{2} . Let Y=(-H)^{-1}\xi , it is almost surely of regularity 1- , so one can define \rho=\mathrm{e}^{Y} and search for a formula for (H+\xi)u when u is of the form u=\rho v . A formal computation gives (H+\xi)u=\rho\left(Hv+2\nabla Y\cdot\nabla v+xY\cdot xv+|\nabla Y|^{2}v\right). Unfortunately, \nabla Y is of regularity 0- , thus the product |\nabla Y|^{2} barely fails to exist. The author proved in [ 25 ] that the wick product :|\nabla Y|^{2}: can be defined as the almost sure limit, in suitable space of distributions of regularity 0- , of |\nabla Y_{N}|^{2}-\mathbb{E}[|\nabla Y_{N}|^{2}], where Y_{N} is a suitable regularisation of Y_{N} with only low modes. Thus, the renormalized operator can be formally defined as :H+\xi:u=\rho\left(Hv+2\nabla Y\cdot\nabla v+xY\cdot xv+:|\nabla Y|^{2}:v\right). This shows that knowing a trajectory of \xi is not sufficient to define the operator :H+\xi: . It is a usual fact in the study of Anderson operators that one needs more data in order to define the operator (see for example [ 27 , 3 , 1 , 22 ] ). Following this idea, we construct a formal renormalization of H+\xi given an enhanced noise \Xi=(Y,Z) with Y=(-H)^{-1}\xi and Z playing the role of |\nabla Y|^{2} , as :H+\xi:u=\rho\left(Hv+2\nabla Y\cdot\nabla v+xY\cdot xv+Zv\right). This formula does not make sense directly, but as usual, the bilinear form of the operator can be defined as \langle-:H+\xi:u_{1},u_{2}\rangle=\re\int_{\mathbb{R}^{2}}\left(\nabla v_{1}\cdot\overline{\nabla v_{2}}+|x|^{2}(1-Y)v_{1}\overline{v_{2}}\right)\rho^{2}\mathrm{d}x-\langle Z,\overline{v_{1}}v_{2}\rho^{2}\rangle, where \rho=\mathrm{e}^{Y} , u_{i}=\rho v_{i} and \langle\cdot,\cdot\rangle stand for a duality bracket between suitable spaces. We then show that this bilinear form is associated to a self-adjoint lower semi-bounded operator of compact resolvent on \mathrm{L}^{2}(\mathbb{R}^{2}) which we take as a definition of -:H+\xi: . Up to a \Xi -dependent shift \delta_{\Xi} , one can define a positive self-adjoint operator -A=-:H+\xi:+\delta_{\Xi} . This makes -A a bit easier to study than :H+\xi: . The construction of -A and its elementary properties are the object of section 3 . A similar construction for :\Delta+\xi: on \mathbb{T}^{2} and \mathbb{T}^{3} has been done in [ 28 ] . In this paper, we follow a purely deterministic and analytic approach of "formal renormalization" which allows us to treat in the same way cases where a renormalization is needed and cases where |\nabla Y|^{2} exists as a \mathrm{L}^{q} function (for q>4 ), as for the radial white noise or any deterministic potential in a Sobolev space \mathrm{H}^{-1,2q} . Unfortunately, the domain of -A is not stable under multiplication, hence the standard fixed point fails to apply directly to {~}\begin{cases}\mathrm{i}\partial_{t}u+Au+\lambda\left|u\right|^{2\gamma}u=0\\ u(0)=u_{0}.\end{cases} (1.3) In this situation, it is well-known that one may take advantage of dispersive properties of the associated Schrödinger propagator (\mathrm{e}^{itA})_{t\in\mathbb{R}} in order to obtain local solutions. Given an admissible pair (p,r) , that is a pair (p,r)\in[2,+\infty)^{2} satisfying \frac{1}{p}+\frac{1}{r}=\frac{1}{2} , then one expects (\mathrm{e}^{itA})_{t\in\mathbb{R}} to verify \left|\mathrm{e}^{itA}u\right|_{\mathrm{L}^{p}_{(-T,T)}\mathrm{L}^{r}_{x}}\leqslant C|u|_{\mathcal{W}^{\sigma,2}} (1.4) for some T>0 and \sigma\geqslant 0 , where \mathcal{W}^{\sigma,2} is a Sobolev space associated to -H . This idea goes back to Strichartz, when studying in [ 32 ] the local smoothing property of wave equations. Such inequalities have been obtained for the free Schrödinger propagator (\mathrm{e}^{it\Delta})_{t\in\mathbb{R}} on \mathrm{L}^{2}(\mathbb{R}^{2}) , using the dispersive estimate |\mathrm{e}^{it\Delta}u|_{\mathrm{L}^{\infty}_{x}}\lesssim|t|^{-1}|u|_{\mathrm{L}^{1}_{x}}, and in that case T>0 is arbitrary and \sigma=0 , thus \mathcal{W}^{\sigma,2} is just \mathrm{L}^{2}(\mathbb{R}^{2}) . In the presence of a smooth confining potential V , there are two different cases depending on the growth at infinity of V . On the one hand, if V is subquadratic at infinity, that is \partial^{\alpha}V is bounded when |\alpha|\geqslant 2 , it was shown in [ 16 ] that (\mathrm{e}^{it(\Delta-V)})_{t\in\mathbb{R}} verifies the same kind of dispersive estimate on a short time. This implies that it verifies an estimate of the type equation 1.4 with T small enough and \sigma=0 . On the other hand, if V is superquatratic, then it is proven in [ 40 ] that (\mathrm{e}^{it(\Delta-V)})_{t\in\mathbb{R}} does not verify a dispersive estimate. Nevertheless, they obtain Strichartz inequalities with a loss of derivative on short time, that is equation 1.4 with T small enough and some \sigma>0 . On a compact manifold, Strichartz estimates with losses for (\mathrm{e}^{it\Delta})_{t\in\mathbb{R}} have been obtained in [ 7 ] , where they were proven to be stable under \mathrm{H}^{1} perturbations. The same method was then used in [ 29 , 41 ] to obtain Strichartz inequalities with losses for (\mathrm{e}^{it(\Delta+\xi)})_{t\in\mathbb{R}} on 2d compact manifolds. In this paper, we follow this idea of lower order perturbation of Strichartz inequalities for (\mathrm{e}^{itH})_{t\in\mathbb{R}} to obtain Strichartz inequalities for (\mathrm{e}^{itA})_{t\in\mathbb{R}} . However, A cannot be seen directly as a perturbation of H . First remark that, formally, \rho^{-1}A(\rho v)=Hv+2\nabla Y\cdot\nabla v+xY\cdot xv+Zv-\delta_{\Xi}v. Thus, \tilde{A}=\rho^{-1}A\rho is a better candidate to be a lower order perturbation of H . We prove in section 3 that this operator is well-defined on \mathrm{D}(-\tilde{A})=\rho^{-1}\mathrm{D}(-A)\subset\mathcal{W}^{2-,2} . Then, a paracontrolled approach of \tilde{A} allows to get rid of the worst regularity terms in the expression of \tilde{A} . The idea of paracontrolled calculus to study singular stochastic