Numerical study of the Amick-Schonbek system in 2D
Extracted source text · arXiv 2501.11483
\documentclass{amsart}
\usepackage{amssymb}
\usepackage[dvips]{epsfig}
\usepackage{graphicx}
\usepackage{color}
% \usepackage{showkeys}
\newcommand{\abs}[1]{\lvert#1\rvert}
\newcommand{\Abs}[1]{\lVert#1\rVert}
\newcommand{\nm}{\nabla_{X,z}^{\mu_2}}
\newcommand{\dm}{\Delta_{X,z}^{\mu_2}}
\newcommand{\cst}{\mbox{\textnormal{Cst }}}
\newcommand{\dsp}{\displaystyle}
\newcommand{\dt}{\partial_t}
\newcommand{\dx}{\partial_x}
\newcommand{\dz}{\partial_z}
\newcommand{\R}{{\mathbb R}}
\newcommand{\N}{{\mathbb N}}
\newcommand{\Z}{{\mathbb Z}}
\newcommand{\eps}{\varepsilon}
\newcommand{\uz}{\underline{\zeta}}
\newcommand{\uP}{\underline{\Phi}_2}
\newcommand{\tX}{{\mathcal X}}
\newcommand{\tY}{{\mathcal Y}}
\newcommand{\cG}{{\mathcal G}[\zeta]}
\newcommand{\cV}{{\mathcal V}[\zeta]}
\newcommand{\cZ}{{\mathcal Z}[\zeta]}
\newcommand{\tU}{\widetilde{U}}
\newcommand{\tzeta}{\widetilde{\zeta}}
\newcommand{\tbv}{\widetilde{\bf v}}
\newcommand{\tC}{\widetilde{C}}
\newcommand{\tm}{{\bf T}_{\mu}}
\newcommand{\tmd}{{\bf T}_{\mu_2}}
\newcommand{\D}{\vert D\vert}
\newcommand{\bv}{{\mathbf v}}
\newcommand{\bu}{{\mathbf u}}
\newcommand{\bw}{{\mathbf w}}
\newcommand{\mfR}{{\mathfrak R}[\zeta]\bv}
\newcommand{\mfRw}{{\mathfrak R}[\zeta]{\bf w}}
\newcommand{\mfS}{{\mathfrak S}[\zeta]\bv}
\newcommand{\mS}{{\mathfrak S}}
\newcommand{\curl}{\mbox{\textnormal{curl }}}
\newcommand{\Id}{\mbox{\textnormal{Id}}}
\newcommand{\init}{_{\vert_{t=0}}}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{xca}[theorem]{Exercise}
\newtheorem*{merci}{Acknowledgements}
\theoremstyle{remark}
\newtheorem{remark}{Remark}[section]
\def\R{{\mathbb R}}
\numberwithin{equation}{section}
\begin{document}
\title[Amick-Schonbek system in 2D]{Numerical study of the
Amick-Schonbek system in 2D}
\author[C. Klein]{Christian Klein}
\address{Institut de Math\'ematiques de Bourgogne, UMR 5584;\\
Institut Universitaire de France \\
Universit\'e de Bourgogne Europe, 9 avenue Alain Savary, 21078 Dijon
Cedex, France} %\\
\email{Christian.Klein@u-bourgogne.fr}
\author[J.-C. Saut]{Jean-Claude Saut}
\address{Laboratoire de Math\' ematiques, UMR 8628,\\
Universit\' e Paris-Saclay et CNRS\\ 91405 Orsay, France}
\email{jean-claude.saut@universite-paris-saclay.fr}
\date{\today}
\maketitle
\begin{center}
\emph{This paper is dedicated to the memory of Vladimir E. Zakharov}
\end{center}
\begin{abstract}
A numerical study of the 2D Amick-Schonbek Boussinesq system is
presented. Numerical evidence is given for the transverse stability
of the 1D solitary waves that are line solitary waves of the 2D
equations. It is shown that initial data not satisfying the
non-cavitation condition can lead to the formation of a gradient
catastrophe in finite time. The numerical propagation of localised
smooth initial data does not lead to the formation of stable
structures localised in both spatial directions. For initial data
satisfying the non-cavitation condition, smooth solutions appear to
exist for all times.
\end{abstract}
\section{Introduction}
This paper presents numerical simulations of a special case of
two-dimensional Boussinesq systems for the propagation of long, weakly nonlinear waves in the so-called Boussinesq regime. Those systems degenerate to the Korteweg-de Vries equation when restricted to one-dimensional, one way propagating waves.
They model water waves on a flat bottom propagating in
both directions in the aforementioned regime (see \cite{BCL, BCS1,
BCS2}). The rigorous derivation is obtained by expanding the
Dirichlet to Neumann operator occurring in the Zakharov-Craig-Sulem
formulation of the water waves system \cite{Za2,CSS2} with respect to the small
parameter $\epsilon$ (see below).
\subsection{Background}
More precisely we will consider a particular case of the so-called
abcd Boussinesq systems for surface water waves, see \cite{BCL,
BCS1,BCS2}. Note that Boussinesq \cite{Bou2} was the first to derive
particular Boussinesq systems, but not in the class of those studied here.
We refer to \cite{Da,Da2,Da3} for details and the history of
hydrodynamics in the nineteenth century. The systems can be put into
the form
\begin{equation}
\label{abcd2}
\left\lbrace
\begin{array}{l}
\eta_t+\nabla \cdot {\bf v}+\epsilon \nabla\cdot(\eta {\bf v})+\mu\lbrack a \nabla\cdot \Delta{\bf v}-b\Delta \eta_t\rbrack=0 \\
{\bf v}_t+\nabla \eta+\epsilon \frac{1}{2}\nabla |{\bf v}|^2+\mu\lbrack c\nabla \Delta \eta-d\Delta {\bf v}_t\rbrack=0.
\end{array}\right.
\end{equation}
Here $\eta= \eta(x,t), x\in \R^d, d=1,2, t\in \R$ is the elevation of the wave, ${\bf v}={\bf v}(x,t)$ is a measure of the horizontal velocity, $\mu$ and $\epsilon$ are the small parameters (shallowness and nonlinearity parameters respectively) defined as
$$\mu=\frac{h^2}{\lambda^2}, \quad \epsilon= \frac{\alpha}{h}$$
where $\alpha$ is a typical amplitude of the wave, $h$ a typical depth and $\lambda$ a typical horizontal wavelength.
In the Boussinesq regime, $\epsilon$ and $\mu$ are supposed to be
small and of same order, $\epsilon\sim\mu\ll1$. For simplicity we
will put $\epsilon=\mu,$ writing \eqref{abcd2} as
\begin{equation}
\label{abcd}
\left\lbrace
\begin{array}{l}
\eta_t+\nabla \cdot {\bf v}+\epsilon \lbrack\nabla\cdot(\eta {\bf v})+a \nabla\cdot \Delta{\bf v}-b\Delta \eta_t\rbrack=0 \\
{\bf v}_t+\nabla \eta+\epsilon\lbrack \frac{1}{2}\nabla |{\bf
v}|^2+c\nabla \Delta \eta-d\Delta {\bf v }_t\rbrack=0.
\end{array}\right.
\end{equation}
Particular cases are formally derived in \cite{Broer, Ding, Per}, see also \cite{Ding}.
The coefficients (a, b, c, d) are restricted by the condition
$$a+b+c+d=\frac{1}{3}-\tau,$$
where $\tau\geq 0$ is the surface tension coefficient.
% When restricted to one-dimensional, unidirectional motions,
% \eqref{abcd} leads to the Korteweg-de Vries (KdV) equation, see \cite{La1}
%
% $$u_t+u_x+\epsilon\left(\frac{1}{3}-\tau\right)u_{xxx}+\epsilon uu_x=0.$$
All (well-posed) members of the abcd class provide the same
approximation of the water wave system in the Boussinesq regime, with
an error of order $O(\epsilon^2t)$ (see \cite{BCL}), and their
dispersion relations are similar in the long wave regime.
Nevertheless they possess rather different mathematical properties
as nonlinear dispersive systems due to the different behavior of the
dispersion relation at high frequencies. In fact the order of their dispersive part can vary from $+3$ to $-1$, \cite{BCS1, BCS2}. We refer to \cite{Mel} for a careful analysis of linear dispersive estimates for the abcd systems.
Two particular one-dimensional cases of the abcd
systems have remarkable properties and were studied numerically in
\cite{KSAS,KSKBK}. We focus here on the two-dimensional version of one of them
This system
is a particular case of a system derived by Peregrine in \cite{Per},
and it is often referred to as the Peregrine or classical Boussinesq
system, but Schonbek and Amick were the first to recognize its
remarkable mathematical properties. A variant with slowly varying
bottom is derived in \cite{TW,W}. This particular case corresponds to $a=b=c=0$,
$d=1$ and thus
\begin{equation}
\label{AS}
\left\lbrace
\begin{array}{l}
\eta_t+v_x+\epsilon(\eta v)_x=0 \\
v_t+\eta_x+\epsilon(vv_x-v_{xxt})=0.
\end{array}\right.
\end{equation}
which can be seen as the BBM regularization of a (linearly ill-posed)
system obtained from the Zakharov-Craig-Sulem formulation of
the water wave system after expanding the Dirichlet-to-Neumann operator at first order in $\epsilon.$
% \begin{equation}
% \label{IP}
% \left\lbrace
% \begin{array}{l}
% \eta_t+v_x+\epsilon(\eta v)_x=0 \\
% v_t+\eta_x+\epsilon(vv_x+\eta_{xxx})=0,
% \end{array}\right.
% \end{equation}
It turns out that the hyperbolic structure of the underlying
Saint-Venant (Shallow Water) system can be used for solving the
Cauchy problem of the one-dimensional Amick-Schonbek system. In
particular the entropy of the Saint-Venant system helps to derive
an a priori bound for the solutions of the Amick-Schonbek
system, leading to the global well-posedness of the Cauchy
problem for arbitrary large initial data satisfying the physical
non-cavitation condition, see \cite{Am, Sch} and \cite{MTZ}, this
last paper presenting the sharpest results, in particular the
existence of global weak solutions. Note that the large time
behavior of these solutions is not known. We refer to \cite{KSAS}
for numerical simulations leading to relevant conjectures.
% Such a connection with the Saint-Venant system is not known in the two-dimensional case.
There have been many papers dealing with the numerical study of the one-dimensional
Amick-Schonbek system, see for instance \cite{An-Dou, An-Dou2, KSAS}. The
case of a non-flat bottom is considered in \cite{GAD} and the
periodic case in \cite{ADM3}. Solitary waves were constructed in
\cite{chen}, where their interaction was also studied.
The present paper is a continuation of a previous work \cite{KSAS} that considered the 1D case.
We focus here on the two-dimensional case for which the existence (and asymptotic behavior) of global solutions is still an open issue both for small and large initial data and thus for which numerical simulations are crucial to present relevant conjectures.
%As aforementioned the Amick-Schonbek is the only member of the
%(one-dimensional) abcd systems for which the global well-posedness
%of the Cauchy problem with arbitrary large initial data is known.
%The asymptotics in large time of these solutions is not established, and
%the simulations in this paper will give some insight into this issue.
%We will consider in particular the possible soliton decomposition.
%To this end we study the 1D case with a Fourier spectral method
%similar to the one for the Serre-Green-Naghdi equation in \cite{DK} and
%get, see also \cite{chen,An-Dou}, the following\\
%\textbf{Conjecture I:}\\
%The solitary waves of the Amick-Schonbeck system are asymptotically
% stable. The long time behavior of solutions to this system for
%Schwartz class initial data is given by solitary waves plus
%radiation.
% In addition we study
% the zero dispersion limit including the formation of dispersive
% shock waves (DSWs), zones of rapid modulated oscillations, and the behavior of solutions when the
% initial data $\eta_0$ do not satisfy the non-cavitation
% condition. We get\\
% \textbf{Conjecture II:}\\
%Solutions to initial data not satisfying the non-cavitation
% condition $1+\eta(x,0)>0$ can blow up in finite time.
% we numerically construct solitary waves. Perturbations
%of these solitary waves are studied in section 4. In section 5 we
%address the long time behavior of hump-like initial data from the
%Schwartz class. In section 6 we consider initial data
%not satisfying the non-cavitation condition. Dispersive shock waves
%are studied in section 7. We add some concluding remarks in section
%8.
The two-dimensional Amick-Schonbek system writes (we have kept the small parameter $\epsilon$):
\begin{equation}
\label{2D}
\left\lbrace
\begin{array}{l}
\eta_t+\nabla \cdot {\bf v}+\epsilon \lbrack\nabla\cdot(\eta {\bf v})=0 \\
{\bf v}_t+\nabla \eta+\epsilon\lbrack \frac{1}{2}\nabla |{\bf v}|^2-d\Delta {\bf v }_t\rbrack=0.
\end{array}\right.
\end{equation}
We are not aware of a connection between \eqref{2D} and the
hyperbolic structure the underlying two-dimensional Saint-Venant
system similar to the aforementioned results of the 1D case leading
for instance to the global existence of weak solutions. The local
well-posedness of the Cauchy problem is relatively standard, (see
\cite{BCS2} in the 1D case but the extension to the 2D case is
straightforward). The existence on long times of order $O(1/\epsilon)$ of solutions to the Cauchy problem for \eqref{2D} that is needed for the full rigorous justification of the system, see \cite{BCL}, was established in \cite{SWX}, Theorem 4.4, under the non-cavitation condition $1+\eta_0\geq H_0>0$ and in \cite{Bu} without this condition.
\begin{remark}
The long time existence for a two-dimensional Amick-Schonbek system in presence of a non-trivial bathymetry is proven in \cite{Meso}.
\end{remark}
On the other hand, as aforementioned, the {\it global} behavior of solutions to the 2D Amick-Schonbek system is unknown.
In particular the possibility of finite-time blow-up for solutions with initial data satisfying the non-cavitation condition is an open question. The aim of the present paper is to present numerical simulations leading to relevant conjectures,
on the qualitative properties of solutions of the Cauchy problem.
\subsection{Main results}
In this paper we first study for the 1D solitary waves the stability
which was also considered in the 1D setting in \cite{KSAS}. Obviously
these 1D solutions are infinitely extended (in $y$) solutions to the
2D system (\ref{2D}), so-called line solitary waves. We consider
numerically 2D perturbations of these line solitary waves and present
evidence for the
following
\begin{conjecture}
The line solitary waves of the Amick-Schonbek system are
asymptotically stable.
\end{conjecture}
Initial data not satisfying the non-cavitation condition were studied
in 1D numerically in \cite{KSAS}. We expand on these results and
consider initial data violating this condition in one point in the 2D
case. This leads to the following
\begin{conjecture}
Initial data violating the non-cavitation condition in one point
in 1D lead to a gradient catastrophe in finite time.
\end{conjecture}
Similar results appear to hold in 2D, but we do not have access to
sufficient numerical resolution in this case to present as convincing
evidence as in the 1D case.
We study the time evolution of 2D localised initial data and show
that it forms an annular structure which then leads to a similar
dynamical behavior as initial data that come close to violating the
non-cavitation condition. In particular we do not find lump solutions to this system as known
from the Kadomtsev-Petviashvili (KP) I equation. It appears that solutions
to the 2D system show a similar behavior as solutions to the
KP II equation. The Boussinesq equation appears to have a
\emph{defocusing} effect. We get the following
\begin{conjecture}
Localised smooth initial data satisfying the non-cavitation
condition do not lead to stable structures localised
in both spatial dimension. The solutions exist for all times and are
simply dispersed.
\end{conjecture}
Furthermore we study the
appearance of zones of rapid modulated oscillations in the vicinity
of shocks of the solutions to the same initial data for the
corresponding Saint-Venant system, so called \emph{dispersive shock
waves} (DSWs).
This paper is organized as follows. In section 2 we discuss the
transverse stability of line solitary waves. Section 3 is
devoted to the behavior of solutions corresponding to initial
data violating the non-cavitation condition. In section 4 the
time evolution of 2D localised smooth initial data is studied.
Section 5 is focussed on dispersive shock waves arising in the
zero dispersion limit. We add some concluding remarks in section
6.
%\section{Solitary waves of the Amick-Schonbek system}
%In this section, we will numerically construct localised traveling
%wave solutions of the system (\ref{AS}), i.e., solutions of the form
% $\eta=Q(x-Ct)$, $v = V(x-Ct)$ where $C$ is a constant real velocity.
% We follow the presentation in \cite{chen}, where the solitary waves
%were also studied numerically, see also \cite{An-Dou}.
%This ansatz implies for (\ref{AS}) after one integration taking care of the
%vanishing conditions at infinity,
%\begin{align}
% -CQ+V+QV & =0,
%\nonumber\\
%-CV+Q+\frac{1}{2}V^{2}+cV'' & =0,
%\label{ASsol}
%\end{align}
%where the prime denotes derivative with respect to the argument. Thus
%we get
%\begin{align}
%Q & =\frac{V}{C-V},
%\nonumber\\
%\label{ASV}
%\end{align}
%Thus solitary waves of the Amick-Schonbek system are given by a
%single ODE for $V$, the second equation in (\ref{ASV}). In order to
%get solutions that are exponentially decreasing as $e^{-\alpha|x|}$
%with $\alpha>0$
%for $|x|\to\infty$, one must have $\alpha = \sqrt{1-1/C^{2}}$, i.e.,
%$|C|>1$. Furthermore if $V_{C}$ is
%a solution to (\ref{ASV}), so is $-V_{-C}$, i.e., a change of sign
%of $C$ implies a change of sign of $V$. Therefore we will concentrate
%in the following on the case $C>1$.
%The second equation in (\ref{ASV})
%can be integrated once more to give
%\begin{equation}
%\frac{C}{2} ((V')^{2}-V^{2})+\frac{1}{6}V^{3}-V-C\ln(1-V/c)=0
%\label{ASVint},
%\end{equation}
%which implies that $V$ can be given in terms of quadratures. Due to
%the appearance of the logarithm, it does not seem possible to give
%$V$ in terms of elementary functions. Note that $Q$ gets very large
%once the maximum of $V$ is close to $C$.
%It is possible to numerically compute the integral with respect to $V$, but
%since we need the solution in the form $V(x)$, this has to be done
%for all interesting values of $x$ in order to be numerically able to
%invert the function $x(V)$. Therefore we use here the same approach
%as in \cite{KS15}. Since the solitary waves are rapidly decreasing,
%we can treat them as periodic functions that are smooth on a
%sufficiently large torus within the finite numerical precision (we
%work with double precision where the smallest difference between
%rational numbers is of the order of $10^{-16}$). Thus we consider
%$x\in L[-\pi,\pi]$ with $L\gg1$ and use the standard discretisation of
%the discrete Fourier transform (DFT) $x_{n} = -\pi L +nh$, $n=1,2,\ldots,N$
%with $h = 2\pi L/N$. The DFT can be conveniently treated with a fast
%Fourier transform (FFT). We denote the DFT of a discretised function
%$V$ by
%$\hat{V}$. Thus we approximate the second ODE in (\ref{ASV})
%\begin{equation}
%F:=-\hat{V}+\frac{1}{C(1+k^{2})}\left(\frac{1}{2}\widehat{V^{2}}+\widehat{\frac{V}{C-V}}\right)=0
%\label{ASF},
%\end{equation}
%which means we have to find the zero of an $N$-dimensional function,
%where $N$ is the number of DFT modes. This zero is identified
%iteratively with a Newton method where the action of the inverse of
%the Jacobian is computed as in \cite{KS15} with the Krylov subspace
%technique GMRES \cite{GMRES}.
%\section{Numerical approach}
\section{Transverse stability of line solitary waves}
The solitary waves constructed numerically for the 1D AS system, see
for instance \cite{KSAS} and references therein, are $y$-independent
traveling wave solutions to the 2D AS system (\ref{2D}). These
solitary waves infinitely extended in $y$-direction form so-called
line solitary waves. In this section we will study the transverse
stability of these waves, i.e., their stability under
small perturbations with a non-trivial $y$-dependence. We will
present numerical evidence for the transverse stability of the line
solitary waves.
\subsection{Numerical approach}
The numerical approach to be applied in this paper is essentially a
2D version of the 1D code in \cite{KSAS}. For the spatial coordinates
we use a discrete Fourier transform which is conveniently computed
with a 2D fast Fourier transform (FFT). This means we approximate a
setting in $\mathbb{R}^{2}$ by a setting in $\mathbb{T}^{2}$ where
the torus is chosen to be sufficiently large in both directions that
the effects of periodicity are not dominant. In other words the
period $2\pi L_{x}$, $L_{x}>0$ is large enough that the reentering
radiation is small compared to the effects we want to study. On the
other hand the period $2\pi L_{y}$ in $y$-direction is small meaning
that we essentially consider line solitary waves on $\mathbb{T}_{y}$,
but we expect that this will not affect the result (for the cases we
studied, no effect of the choice of $L_{y}$ was noted).
In other words we use the standard discretisation $x_{n}=-\pi
L_{x}+n2\pi L_{x}/N_{x}$, $n=1,\ldots,N_{x}$ and $y_{m}=-\pi
L_{y}+n2\pi L_{y}/N_{y}$, $m=1,\ldots,N_{y}$. For the time
integration of the resulting ODE system for (\ref{2D}) we use as in
\cite{KSAS} the classical explicit 4th order Runge-Kutta method. The
code is tested for the example of the line solitary waves that are
propagated to essentially the precision with which they were
constructed in \cite{KSAS} if the initial data are unperturbed.
\subsection{Perturbed line solitary wave with $c=2$}
The 1D solitary wave solutions $Q_{c}(x)$ exist for $|c|>1$, the amplitude of
the waves tending to the trivial solution for $|c|\to1$. We consider here only
positive values of $c$, i.e., waves traveling to the right. First we
consider the case $c=2$ with $N_{x}=2^{12}$ Fourier modes for $x\in
10[-\pi,\pi]$ and $N_{y}=2^{7}$ Fourier modes for $y\in 3[-\pi,\pi]$
with $N_{t}=10^{4}$ time steps for $t\leq20$. The first perturbation we
will look at is of the form
\begin{equation}
\eta(x,y,0) = Q_{2}(x) \pm 0.3\exp(-x^{2}-y^{2}),\quad
v_{x}(x,y,0) = V_{2}(x),\quad v_{y}(x,y,0)=0.
\label{c2gauss}.
\end{equation}
This corresponds to perturbations of the line solitary wave of the
order of $5\%$. Note that we need some finite perturbations to
numerically see an effect of the perturbation in finite time.
The solution $\eta$ for the initial data with the $+$ sign can be
seen in Fig.~\ref{figc2gauss} for $t=0$ and $t=20$. It appears
that the final state corresponds to a line solitary wave
with slightly larger velocity $c>2$ plus radiation.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_03gauss_t0eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_03gauss_t20eta.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2gauss}) for the $+$ sign, on the left $\eta$ for $t=0$, on the right for
$t=20$.}
\label{figc2gauss}
\end{figure}
This interpretation is confirmed by the velocities $v_{x}$ and $v_{y}$
which are shown for the same initial
data for $t=20$ in Fig.~\ref{figc2gaussv}.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_03gauss_t20vx.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_03gauss_t20vy.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2gauss}) for $t=20$, on the left $v_{x}$, on the right
$v_{y}$. }
\label{figc2gaussv}
\end{figure}
The $L^{\infty}$ norm of the solution $\eta$ in this case on the left
of Fig.~\ref{figc2gaussmax} shows oscillations around what appears to
be a constant final state, a line solitary wave. The above figures are very similar for the $-$sign
in (\ref{c2gauss}). Therefore we only show the $L^{\infty}$ norm of
$\eta$ in this case on the right of Fig.~\ref{figc2gaussmax}. The
final state is reached more slowly, but it appears to be again a line
solitary wave, this time with slightly smaller velocity than the
unperturbed case with $c=2$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_03gaussetamax.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_m03gaussmax.eps}
\caption{$L^{\infty}$ norm of the solution $\eta$ to the AS system (\ref{2D}) for the initial data
(\ref{c2gauss}), on the left for the $+$ sign, on the right for the
$-$ sign.}
\label{figc2gaussmax}
\end{figure}
If we perturb the velocity $v_{x}$ by a small Gaussian, i.e., if we
consider initial data of the form
\begin{equation}
\eta(x,y,0) = Q_{2}(x) ,\quad
v_{x}(x,y,0) = V_{2}(x)+ 0.1\exp(-x^{2}-y^{2}),\quad
v_{y}(x,y,0)=0,
\label{c2gaussvx}
\end{equation}
we get the solution shown in Fig.~\ref{figc2gaussvx}. Again the final
state appears to be a line solitary wave plus radiation.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_vx01gauss_t0vx.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_vx01gauss_t20vx.eps}\\
\includegraphics[width=0.49\textwidth]{ASsol_c2_vx01gauss_t20eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_vx01gauss_t20vy.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2gaussvx}), in the upper row on the left $v_{x}$ for $t=0$, on the right for
$t=20$, in the lower row on the left $\eta$ and on the right
$v_{y}$, both for t=20.}
\label{figc2gaussvx}
\end{figure}
The $L^{\infty}$ norm of the solution $\eta$ in
Fig.~\ref{figc2gaussvx} can be seen on the left of
Fig.~\ref{figc2max}. It appears to oscillate around a final state
corresponding to a line solitary wave.
\begin{figure}[htb!]
\includegraphics[width=0.32\textwidth]{ASsol_c2_vx01gaussmax.eps}
\includegraphics[width=0.32\textwidth]{ASsol_c2_vy01gaussmax.eps}
\includegraphics[width=0.32\textwidth]{ASsol_c2_cosetamax.eps}
\caption{$L^{\infty}$ norms of solutions $\eta$ to the AS system
(\ref{2D}) for various initial data, on the left for the solution
shown in Fig.~\ref{figc2gaussvx}, in the middle for the solution
shown in Fig.~\ref{figc2gaussvy}, on the right for the solution
shown in Fig.~\ref{figc2cosv}. }
\label{figc2max}
\end{figure}
If we perturb $v_{y}$ with a small Gaussian, i.e., if we
consider initial data of the form
\begin{equation}
\eta(x,y,0) = Q_{2}(x) ,\quad
v_{x}(x,y,0) = V_{2}(x),\quad v_{y}(x,y,0)=0.1\exp(-x^{2}-y^{2}),
\label{c2gaussvy}
\end{equation}
we get the solution shown in Fig.~\ref{figc2gaussvy}. Once more the
final state of the solution seems to be a line solitary wave plus
radiation. The $L^{\infty}$ norm of the solution $\eta$ in this case
can be seen in the middle of Fig.~\ref{figc2max}. It confirms the
interpretation of the final state being a line solitary wave plus
radiation.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_vy01gauss_t0vy.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_vy01gauss_t20vy.eps}\\
\includegraphics[width=0.49\textwidth]{ASsol_c2_vy01gauss_t20eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_vy01gauss_t20vx.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2cos}), in the upper row on the left $v_{y}$ for $t=0$, on the right for
$t=20$, in the lower row on the left $\eta$ and on the right
$v_{x}$, both for t=20.}
\label{figc2gaussvy}
\end{figure}
A different kind of perturbation is to deform the line solitary wave
periodically. Concretely we consider initial data of the form
\begin{equation}
\eta(x,y,0) = Q_{2}(x+0.4\cos(y)),\quad
v_{x}(x,y,0) = V_{2}(x),\quad v_{y}(x,y,0)=0.
\label{c2cos}
\end{equation}
The initial condition for $\eta$ can be seen on the left of
Fig.~\ref{figc2cos}. The deformed solitary wave emits some radiation
and appears to reach another solitary wave as a final state as
suggested by the figure on the right of Fig.~\ref{figc2cos} for
$t=20$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_cos_t0eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_cos_t20eta.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2cos}), on the left $\eta$ for $t=0$, on the right for
$t=20$.}
\label{figc2cos}
\end{figure}
This is confirmed by the velocities at the final time $t=20$ shown in
Fig.~\ref{figc2cosv}. The $L^{\infty}$ norm of $\eta$ in the same
figure appears to approach rapidly the final state as can be seen on
the right of Fig.~\ref{figc2max}.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c2_cos_t20vx.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c2_cos_t20vy.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c2cos}) for $t=100$, on the left $v_{x}$, on the right
$v_{y}$. }
\label{figc2cosv}
\end{figure}
\subsection{Perturbed line solitary wave with $c=1.1$}
As an example of solitary waves with velocities close to the
threshold $c=1$, i.e., small slow solitary waves, we consider the
case $c=1.1$. For the numerical experiments we use $N_{x}=2^{14}$ Fourier modes for $x\in
40[-\pi,\pi]$ and $N_{y}=2^{7}$ Fourier modes for $y\in 3[-\pi,\pi]$
with $N_{t}=2*10^{4}$ time steps for $t<100$. We consider again
localised perturbations with initial data of the form ($\kappa>0$,
$\alpha>0$)
\begin{equation}
\eta(x,y,0) = Q_{1.1}(x) + 0.01\exp(-x^{2}-\alpha y^{2}),\quad
v_{x}(x,y,0) = V_{1.1}(x),\quad v_{y}(x,y,0)=0.
\label{c11gauss}
\end{equation}
This means the perturbation has a maximum of the order of $5\%$ of the
wave crest. The initial data for $\eta$ and the solution for $t=20$
are shown in Fig.~\ref{figc11gauss}. It can be seen that the final
state appears to be a line solitary wave of slightly different
velocity as in the case for $c=2$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c11_001gauss_t0eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c11_001gauss_t100eta.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c11gauss}), on the left $\eta$ for $t=0$, on the right for
$t=100$.}
\label{figc11gauss}
\end{figure}
The plots for the velocities $v_{x}$ and $v_{y}$ are presented in
Fig.~\ref{figc11gaussv}. Once more the final state appears to be
a line solitary wave plus radiation.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c11_001gauss_t100vx.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c11_001gauss_t100vy.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c11gauss}) for $t=100$, on the left $v_{x}$, on the right
$v_{y}$. }
\label{figc11gaussv}
\end{figure}
This interpretation is confirmed by the $L^{\infty}$ norm of $\eta$
shown on the left of Fig.~\ref{figc11gaussmax}. It appears to settle
on a constant level for larger $t$ indicating that the final state is
a line solitary wave of slightly higher velocity than the unperturbed
wave with $c=1.1$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c11_001gaussmax.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c11_04cosmax.eps}
\caption{$L^{\infty}$ norm of the solution to the AS system (\ref{2D}) for the initial data
(\ref{c11gauss}) on the left, on the right for the initial data
(\ref{c11cos}).}
\label{figc11gaussmax}
\end{figure}
A periodic deformation of the line solitary wave similar to
(\ref{c2cos}),
\begin{equation}
\eta(x,y,0) = Q_{1.1}(x+0.4\cos(y)),\quad
v_{x}(x,y,0) = V_{1.1}(x),\quad v_{y}(x,y,0)=0,
\label{c11cos}
\end{equation}
gives similar results as in the case $c=2$ as can be seen in
Fig.~\ref{figc11cos}. The final state is once more a line solitary
wave plus radiation. This is confirmed by the $L^{\infty}$ norm of
$\eta$ in Fig.~\ref{figc11gaussmax} on the right.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{ASsol_c11_04cos_t0eta.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c11_04cos_t100eta.eps}\\
\includegraphics[width=0.49\textwidth]{ASsol_c11_04cos_t100vx.eps}
\includegraphics[width=0.49\textwidth]{ASsol_c11_04cos_t100vy.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{c11cos}), in the upper row on the left $\eta$ for $t=0$, on the right for
$t=100$, in the lower row on the left $v_{x}$ and on the right
$v_{x}$, both for $t=100$.}
\label{figc11cos}
\end{figure}
\section{Non-cavitation condition}
In this section we consider initial data which violate the
non-cavitation condition or come close to doing so. We first recall
and expand on the case in 1D discussed in \cite{KSAS}. Then we
discuss similar examples in 2D.
\subsection{Non-cavitation condition in 1D}
In \cite{KSAS}, we have considered initial data of the form
$$\eta(x,0)=-\exp(-x^{2}),\quad v(x,0)=0.$$
We applied the approach by Sulem, Sulem and Frisch in
\cite{SSF} to numerically identify singularities. It uses the fact
that an essential singularity in the complex plane of the form $u\sim
(z-z_{j})^{\mu_{j}}$, $\mu_{j}\notin \mathbb{Z}$,
with $z_{j}=\alpha_{j}-i\delta_{j}$ in the lower half
plane ($\delta_{j}\geq 0$) implies for $k\to\infty$
the following asymptotic behavior of the Fourier
transform (see e.g.~\cite{asymbook}, here denoted in the same way as the DFT),
\begin{equation}
\hat{u}\sim
\sqrt{2\pi}\mu_{j}^{\mu_{j}+\frac{1}{2}}e^{-\mu_{j}}\frac{(-i)^{\mu_{j}+1}}{k^{\mu_{j}+1}} e^{-ik\alpha_{j}-k\delta_{j}}
\label{fourierasym}.
\end{equation}
For a single such singularity with positive $\delta_{j}$, the modulus of the Fourier
transform decreases exponentially for large $|k|$ until
$\delta_{j}=0$, when
this modulus has an algebraic dependence
on $|k|$ for large $|k|$. The same behavior is expected in the discrete version of the
Fourier transform, the DFT. In a dynamic situation, both $\delta_{j}$
and $\mu_{j}$ are expected to depend on $t$, and the time that
$\delta_{j}$ vanished corresponds to the singularity hitting the real
axis, i.e., to the solution forming a singularity. This will be
identified by fitting the DFT coefficients to the (\ref{fourierasym})
via some linear regression.
Note that this fitting is much less reliable for the factor $\mu_{j}$
than for the factor $\delta_{j}$ since this is an algebraic correction to
an exponential term. In the 1D case \cite{KSAS}, we used $2^{18}$
DFT modes, but in 2D below we will be limited on the used computers to values
of $2^{12}$ to $2^{13}$. Note that if there are several singularities
of the form (\ref{fourierasym}), $j=1,\ldots,N_{s}$, the asymptotic
expression (\ref{fourierasym}) will involve a summation over $j$, but
this will not be considered further (therefore we will drop the index
$j$ in the following).
For the Gaussian example, the code in \cite{KSAS} broke for $t\sim
4.681$ since the fitted value of $\delta$ vanished. We show the
solution at the final recorded time in a close-up in
Fig.~\ref{AS1dcloseup}. It can be seen that though the $L^{\infty}$
norm of $\eta$ gets large, a possible cusp near the origin appears to
be the singular feature. The function $v$ on the right of the same
figure appears to stay regular though the fitted value of $\delta$
for $v$ vanishes at the same time.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS1dmgaussetat468zoom.eps}
\includegraphics[width=0.49\textwidth]{AS1dmgaussvt468zoom.eps}
\caption{Solution to the AS system in 1D for the initial data
$\eta(x,0)=-\exp(-x^{2}),\quad v(x,0)=0$ for $t=4.681$ in a
close-up, on the left $\eta$, on the right $v$.}
\label{AS1dcloseup}
\end{figure}
The found value for $\mu$ for the function $\eta$ is negative ($\sim
-0.85$) which would indicate an $L^{\infty}$ blow-up, whereas the
corresponding value for $v$ is positive ($\sim1/3$) which would
indicate the formation of a cusp. However as mentioned, these
values have to be taken with a grain of salt since they are
numerically problematic as an algebraic correction to an exponential
decrease. To get a better insight into the behavior of the solution,
we therefore consider various norms in Fig.~\ref{AS1dnorms}. Whereas
the $L^{\infty}$ norm of $\eta$ appears to grow, this is not the case
for various $L^{p}$ norms, for instance the $L^{4}$ norm which
appears to be bounded. Though numerically one cannot get close enough
to a potential singularity, this is clearly an indication that there
is no $L^{\infty}$ blow-up to be observed in this example. The
$L^{2}$ norms of the gradients in the lower row of the figure
indicate, however, the blow-up of the gradient of $\eta$, but not of
$v$. Thus it appears that there is a cusp formation in $\eta$ in
finite time whereas $v$ stays regular at this time (of course we
cannot rule out that higher Sobelev norms explode, it is just that
$v$ is more regular at this time than $\eta$).
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS1dmgaussetamax.eps}
\includegraphics[width=0.49\textwidth]{AS1dmgausseta4.eps}\\
\includegraphics[width=0.49\textwidth]{AS1dmgaussetaH1.eps}
\includegraphics[width=0.49\textwidth]{AS1dmgaussvH1.eps}
\caption{Solution to the AS system in 1D for the initial data
$\eta(x,0)=-\exp(-x^{2}),\quad v(x,0)=0$, in the upper row on the
left the $L^{\infty}$ norm, on the right the $L^{4}$ norm, in the
lower row the $L^{2}$ norm of the gradient of $\eta$ on the left
(normalized to 1 for $t=0$ as the $L^{4}$ norm) and
of $v$ on the right.}
\label{AS1dnorms}
\end{figure}
\subsection{Non-cavitation condition in 2D}
Concretely we
look at initial data of the form (here $\kappa<0$, $\alpha>0$)
\begin{equation}
\eta(x,y,0) = \kappa \exp(-(x^{2}+\alpha y^{2}),\quad v_{x}(x,y,0) =
v_{y}(x,y,0) = 0.
\label{NCini}
\end{equation}
Note that we cannot characterize potential singularities in the
solutions due to a lack of resolution on the available computers.
First we study the case $\kappa=-0.9$, i.e., initial data satisfying
the non-cavitation condition, but close to violating it.
We use $N_{x}=N_{y}=2^{12}$ DFT modes for $x\in 5[-\pi,\pi]$, $y\in
5[-\pi,\pi]$ and $N_{t}=10^{4}$ time steps for the considered time
interval. The solution
stays smooth for all $t\leq10$, there is as expected no indication of
the appearance of a singularity. To illustrate the interesting
dynamics of the solution, we first show it on the $x$-axis in
Fig.~\ref{figm09gaussaxis} (recall the radial symmetry of the
solution). For the function $\eta$ shown on the left of the figure,
there is a closing of the initial gap in the form of two peaks coming
from the edges and eventually colliding to a central peak of high
elevation, that then leads to another valley before recombining. The
velocity $v_{x}$ on the right of the same figure shows a similar
agitated behavior.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dm09gaussetaaxiswater.eps}
\includegraphics[width=0.49\textwidth]{AS2dm09gaussvxaxiswater.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-0.9$ and $\alpha=1$ on the $x$-axis, on
the left $\eta$, on the right $v_{x}$.}
\label{figm09gaussaxis}
\end{figure}
% We show the $L^{\infty}$ norm and
% the minimum of $\eta$ in dependence of time in
% Fig.~\ref{figm09gaussinf}. It can be seen that the maximum of the
% solution first decreases before reaching a strong peak before
% decreasing again. The minimum on the other hand increases before
% decreasing once more to a small value similar to the initial minimum.
% \begin{figure}[htb!]
% \includegraphics[width=0.49\textwidth]{AS2dm09gaussmax.eps}
% \includegraphics[width=0.49\textwidth]{AS2dm09gaussmin.eps} \caption{Solution to the AS system (\ref{2D}) for the initial data
% (\ref{c11cos}), in the upper row on the left $\eta$ for $t=0$, on the right for
% $t=100$, in the lower row on the left $v_{x}$ and on the right
% $v_{x}$, both for t=100.}
% \label{figm09gaussinf}
% \end{figure}
To illustrate the 2D situation, we show the solution $\eta$ for
several values of time in Fig.~\ref{figm09gausseta}. It is clearly
visible that the peak for $t=4$ is almost cusp-like, but then the
solution forms an annular structure and a pronounced minimum appears
inside the former peak.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t25eta.eps}
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t4eta.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t51eta.eps}
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t10eta.eps}
\caption{Solution $\eta$ to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-0.9$ and $\alpha=1$, in the upper row on the left for
$t=2.5$, on the right for
$t=4$, in the lower row on the left for $t=5.1$ and on the right
for $t=10$.}
\label{figm09gausseta}
\end{figure}
The corresponding plots for $v_{x}$ can be seen in
Fig.~\ref{figm09gaussvx}. The solution appears to be always smooth
though strong gradients appear. Note that the solution is always well
resolved both in space and in time. The plots for $v_{y}$ are very
similar, just rotated by 90 degrees, and are therefore not shown.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t25vx.eps}
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t4vx.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t51vx.eps}
\includegraphics[width=0.49\textwidth]{AS2dm09gauss_t10vx.eps}
\caption{Solution $v_{x}$ to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-0.9$ and $\alpha=1$, in the upper row on the left for
$t=2.5$, on the right for
$t=4$, in the lower row on the left for $t=5.1$ and on the right
for $t=10$.}
\label{figm09gaussvx}
\end{figure}
The situation is different if the non-cavitation condition is not
satisfied for the initial data, for instance for initial data of the
form (\ref{NCini}) with $\kappa=-1$, $\alpha=1$ where there is one point with
$\eta(x,y,0)=0$. We consider $N_{x}=N_{y}=2^{12}$ DFT modes for $x\in
3[\pi,\pi]$ and $x\in
3[\pi,\pi]$.
Nonetheless we find in the case $\kappa=-1$, $\alpha=1$ that the code breaks for
$t=4.0857$ since the fitting of the DFT coefficients on the $k_{x}$
axis leads to a vanishing $\delta$ in (\ref{fourierasym})
indicating that a singularity
appears on the $x$-axis. The fitted value for $\mu$ is $0.604$, i.e.,
compatible with a cusp, but as we mentioned we do not have the needed
numerical resolution to make a precise statement. We can only report
that there is an indication of a singularity which needs to be
checked on more powerful computers. Note that the same
values of the fitted parameters are obtained on the $y$-axis. This
would indicate the formation of a cusp for $\eta$ of the form
\begin{equation}
\eta\sim (x^{2}+y^{2})^{\mu/2}
\label{cusp}
\end{equation}
at the critical time.
Note further that the fitted values of the DFT
coefficients for $v_{x}$ and $v_{y}$ appear to indicate that they are
still regular at the critical time for $\eta$.
We show the solutions at the final recorded time in
Fig.~\ref{figmgauss}. The behavior of $\eta$ near the origin is
complicated, a strong peak with a caldera.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dmgaussetac.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussvxc.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-\alpha=-1$, on the left $\eta$ for
$t=4.0857$, on the right $v_{x}$ at the same time.}
\label{figmgauss}
\end{figure}
% The tracing of the $L^{\infty}$ norm of the solution and the minimal
% value is shown in Fig.~\ref{figmgaussmax}. The minimum stays very
% close to $-1$ for a long time until a peak of $\eta$ emerges
% that will eventually lead to the cusp shown in Fig.~\ref{figmgauss}
% on the left.
% \begin{figure}[htb!]
% \includegraphics[width=0.49\textwidth]{AS2dmgaussmax.eps}
% \includegraphics[width=0.49\textwidth]{AS2dmgaussmin.eps}
% \caption{Solution $\eta$ to the AS system (\ref{2D}) for the initial data
% (\ref{NCini}) with $\kappa=1$, on the left the $L^{\infty}$ norm of
% $\eta$, on the right its minimum.}
% \label{figmgaussmax}
% \end{figure}
It appears best to visualize the behavior of the solution on the
$x$-axis which is done in Fig.~\ref{figmgaussaxis}. It can be seen
that the initial depression for $\eta$ is closed by two
peaked maxima that are approaching each other, but that a gap remains
in between these two
peaks. The velocity $v_{x}$ is seen in the lower row of the same
figure. It seems to develop a cusp near the strongly peaked minimum
of $\eta$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dmgaussetaxwater.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussetacaxis.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dmgaussvxaxiswater.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussvxcaxix.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-\alpha=-1$ on the $x$-axis, in the upper row on
the left in dependence of time, on the right at the critical time
in a close-up,
in the lower row on the left $v_{x}$ and on the right
a close-up at the critical time.}
\label{figmgaussaxis}
\end{figure}
As in the 1D case, we explore the possibly singular behavior via
certain norms of the solution. In Fig.~\ref{figmgaussnorms}, the
$L^{\infty}$ norm of $\eta$ appears to grow as can be seen on the
left of the upper row. However as in the 1D case, a possible
singularity only appears as a cusp at the minimum of the solution.
This is confirmed by the $L^{2}$ norm of $\eta$ (higher $L^{p}$ norms are very
similar) on the right. The $L^{2}$ norm of the $x$-derivative of
$\eta$ can be seen in the lower row. It appears to grow, but we
clearly do not have the needed resolution to reach the potential cusp
singularity. The corresponding norm for $v_{x}$ does not indicate any
singularity formation at the considered times.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dmgaussetainf.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgausseta2.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dmgaussetaH1.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussvxH1.eps}
\caption{Various norms of the solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-\alpha=-1$, in the upper row
the $L^{\infty}$ and the $L^{2}$ norm of $\eta$ (normalized to 1 at
$t=0$), in the lower row
the $L^{2}$ norms of the $x$-derivative of $\eta$ (normalized to 1 at
$t=0$) and of $v_{x}$.}
\label{figmgaussnorms}
\end{figure}
For completeness we consider also a situation without radial
symmetry, initial data of the form (\ref{NCini}) with $\kappa=-1$ and
$\alpha=0.5$. We use the same computational domain as above, but
$N_{x}=2^{13}$, $N_{y}=2^{11}$ DFT modes. The code breaks for
$t=4.0857$ since the fitted $\delta$ on the $k_{x}$-axis vanishes.
We show the
solution at the final time in Fig.~\ref{figmgaussy2}.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dmgaussy2etac.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussy2etaxc.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dmgaussy2vxc.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussy2vxxc.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-1$ and $\alpha$=0.5. In the upper row on
the left the solution $\eta$ at the critical time, on the right a
close-up of the solution on the $x$-axis. In the lower row the
solution $v_{x}$ at the critical time on the left and on the right a
close-up on the $x$-axis.}
\label{figmgaussy2}
\end{figure}
The corresponding fitted value for $\mu\sim-.07$ is even negative,
but as in the 1D case this is not reliable. Therefore we show in
Fig.~\ref{figmgaussy2norms} some norms of the solution.
The $L^{\infty}$ norm of $\eta$ grows as
in the radially symmetric case, but the $L^{2}$ norm stays finite.
The $L^{2}$ norm of the $x$-derivative of $\eta$ grows, but more
slowly than in 1D which once more indicates that we lack resolution to get
closer to the potential singularity. The $L^{2}$ norm of the
$x$-derivative of $v_{x}$ stays finite for the studied times.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dmgaussyetainf.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussyeta2.eps}\\
\includegraphics[width=0.49\textwidth]{AS2dmgaussyetax2.eps}
\includegraphics[width=0.49\textwidth]{AS2dmgaussyvxx2.eps}
\caption{Various norms of the solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=-1$ and $\alpha=0.5$, in the upper row
the $L^{\infty}$ and the $L^{2}$ norm of $\eta$ (normalized to 1 at
$t=0$), in the lower row
the $L^{2}$ norms of the $x$-derivative of $\eta$ (normalized to 1 at
$t=0$) and of $v_{x}$.}
\label{figmgaussy2norms}
\end{figure}
\section{Localised initial data}
In this section we study the time evolution of localized initial
data of the form (\ref{NCini}) with positive $\kappa$. As in the
study of the transverse stability of the line solitary waves, no
stable structures localized in two dimensions are observed.
First we look at initial data with a small positive elevation, say
$\kappa=1$ in (\ref{NCini}). The same numerical parameters as in the
previous section are applied. In Fig.~\ref{figgaussmax} we show the
global minimum of the solution $\eta$ and its $L^{\infty}$ norm in
dependence of time. It can be seen that an annular structure forms
near the original maximum, but that the observed minimum reaches only
values of the order of $-0.5$ before a new peak is formed. Thus one
observes a similar dynamics as in the previous section. There is no
indication of a stable structure emerging from these initial
conditions.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dgaussetamax.eps}
\includegraphics[width=0.49\textwidth]{AS2dgaussetamin.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=\alpha=1$, on the left the $L^{\infty}$ norm of $\eta$, on the
right its global minimum, both in dependence of time.}
\label{figgaussmax}
\end{figure}
This behavior of the solution can as before be best observed on the
$x-$axis since the solution will be once more radially symmetric, see
Fig.~\ref{figgaussaxis}. The initial hump for $\eta$ on the axis spreads
overall, but several oscillations between annular structures and
peaks can be observed. The corresponding plot for $v_{x}$ on the
right of the same figure shows a similar behavior starting with the
trivial initial condition.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dgaussetaxwater.eps}
\includegraphics[width=0.49\textwidth]{AS2dgaussvxxwater.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=\alpha=1$ on the $x$-axis in dependence of time, on
the left $\eta$, on the right $v_{x}$.}
\label{figgaussaxis}
\end{figure}
The situation becomes more agitated if initial data of larger norm
are considered, for instance $\kappa=10$ in (\ref{NCini}). The time
evolution of the potential on the $x-$axis can be seen in
Fig.~\ref{fig10gaussaxis}. The initial peak leads to an annular
structure which develops a hole that approaches the ground to the
order of $-0.993$, see the lower row of the same figure on the right.
It is at this point that the dynamics of the last
section for near cavitation initial data takes over which is why we
stop the time evolution shortly afterwards. As can be seen on the right of the lower row
of the figure for the $L^{\infty}$ norm, the near cavitation leads to
the formation of a peak as discussed in the previous section.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2d10gaussetaxwater.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gaussvxxwater.eps} \\
\includegraphics[width=0.49\textwidth]{AS2d10gaussmax.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gaussmin.eps}
\caption{Solution to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=10$, $\alpha=1$ on the $x$-axis in
dependence of time in the upper row, on
the left $\eta$, on the right $v_{x}$, in the lower row the
$L^{\infty}$ norm of the solution on the left and the global minimum
on the right.}
\label{fig10gaussaxis}
\end{figure}
To get a better understanding of the solution in the 2D setting,
i.e., ignoring the radial symmetry, we show the solution $\eta$
at several times in Fig.~\ref{fig10gausseta}.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2d10gausseta_t2.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gausseta_t4.eps} \\
\includegraphics[width=0.49\textwidth]{AS2d10gausseta_t692.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gausseta_t10.eps}
\caption{Solution $\eta$ to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=10$, $\alpha=1$ for several values of
time, in the upper row for $t=2$ and $t=4$, in the lower row for
$t=6.92$ (time close to the minimal value) and $t=8$.}
\label{fig10gausseta}
\end{figure}
The corresponding plots for $v_{x}$ can be seen in
Fig.~\ref{fig10gaussvx}.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2d10gaussvx_t2.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gaussvx_t4.eps} \\
\includegraphics[width=0.49\textwidth]{AS2d10gaussvx_t692.eps}
\includegraphics[width=0.49\textwidth]{AS2d10gaussvx_t10.eps}
\caption{Solution $v_{x}$ to the AS system (\ref{2D}) for the initial data
(\ref{NCini}) with $\kappa=10$, $\alpha=1$ for several values of
time, in the upper row for $t=2$ and $t=4$, in the lower row for
$t=6.92$ and $t=8$.}
\label{fig10gaussvx}
\end{figure}
\section{Dispersive shock waves}
It is well known that the solutions to nonlinear dispersive PDEs can
have zones of rapid modulated oscillations near shocks of solutions
for the same initial data to the corresponding dispersionless
equations. These zones are called \emph{dispersive shock waves}
(DSWs). A convenient way to study DSWs is to consider initial data
slowly varying on length scales of order $1/\varepsilon$ where
$\varepsilon\ll1$ for times of order $1/\varepsilon$. This can be
addressed by a rescaling of both time and spatial variable with
$\varepsilon$. In an abuse of notation (we keep the same symbols as
before in order to avoid cluttered notation) we get for (\ref{2D})
\begin{equation}
\label{2De}
\left\lbrace
\begin{array}{l}
\eta_t+\nabla \cdot {\bf v}+ \lbrack\nabla\cdot(\eta {\bf v})=0 \\
{\bf v}_t+\nabla \eta+\lbrack \frac{1}{2}\nabla |{\bf v}|^2-\varepsilon\Delta {\bf v }_t\rbrack=0.
\end{array}\right.
\end{equation}
The formal limit $\varepsilon\to0$ leads to a Saint-Venant system
expected to have shock solutions,
\begin{equation}
\label{2DSV}
\left\lbrace
\begin{array}{l}
\eta_t+\nabla \cdot {\bf v}+ \lbrack\nabla\cdot(\eta {\bf v})=0 \\
{\bf v}_t+\nabla \eta+\lbrack \frac{1}{2}\nabla |{\bf v}|^2\rbrack=0.
\end{array}\right.
\end{equation}
To illustrate the behavior of the solutions for small values of
$\varepsilon$, we consider once more Gaussian initial data for
$\varepsilon=10^{-2}$. The solution for $t=5$ is shown in
Fig.~\ref{figgausse2}. It can be seen that the time evolution is
similar to what was shown in the previous section, an initial peak
develops into some annular structure for $\eta$. But the small value
$\varepsilon$ leads to very strong gradients in an almost tubular
structure, and similarly for $v_{x}$.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5eta.eps}
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5vx.eps}
\caption{Solution to the AS system (\ref{2De}) for the initial data
(\ref{NCini}) with $\kappa=\alpha=1$ and $\varepsilon=10^{-2}$ for
$t=5$, on the left $\eta$, on the right $v_{x}$.}
\label{figgausse2}
\end{figure}
However there are small oscillation on what appears to be a steep
front in Fig.~\ref{figgausse2}. To illustrate this, we show the
potentials in Fig.~\ref{figgausse2} on the axis together with a
close-up of the oscillatory zone in Fig.~\ref{figgausse2zoom}. It
can be seen that there are modulated oscillations near the strong
gradients.
\begin{figure}[htb!]
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5etaaxis.eps}
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5etaaxiszoom.eps} \\
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5vxaxis.eps}
\includegraphics[width=0.49\textwidth]{AS2dgauss1e2_t5vxaxiszoom.eps}
\caption{Solution to the AS system (\ref{2De}) for the initial data
(\ref{NCini}) with $\kappa=\alpha=1$ and $\varepsilon=10^{-2}$ for
$t=5$ on the $x$-axis, in the upper row $\eta$, in the lower $v_{x}$,
both with a close-up of the oscillatory zone on the right.}
\label{figgausse2zoom}
\end{figure}
\section{Conclusion}
In this paper we have presented a numerical study of the 2D
Amick-Schonbek system (\ref{2D}). Numerical evidence for the
transverse stability of line solitary waves was presented as well as
for the absence of stable lump solitary waves, i.e., for solitary
waves localised in both spatial dimensions. It was shown that
solutions to the system can exhibit zones of rapid modulated
oscillations known as dispersive shock waves. It was argued that
initial data not satisfying the non-cavitation condition lead to the
formation of cusps in finite time. A precise characterization of
these cusp was not possible due to a local of numerical resolution.
This will be the subject of further research.
Similar numerical studies will be performed for the Boussinesq system
integrable in 1D and numerically studied in \cite{KSKBK}. The results
are to be compared to other equations appearing in the context of
water waves as the Serre-Green-Naghdi equations \cite{GN1,GN2,Ser}.
The goal is to expand the study of \cite{GK} to the questions
addressed in the present paper. This will be the subject of future
work.
\begin{merci}
The work of the first author was partially supported
by the ANR project ANR-17-EURE-0002 EIPHI.
Both authors were partially supported by the ANR project ISAAC-ANR-23-CE40-0015-01.
\end{merci}
\bibliographystyle{amsplain}
\begin{thebibliography}{99}
%\bibitem{ABM}\textsc{D.M. Ambrose, J.L. Bona and T. Milgram}, {\it Global solutions and ill-posedness for the Kaup system and related systems}, Indiana Math. J. {\bf 68} (4) (2019), 1173-1198.
\bibitem{Ada}\textsc{K. Adamy}, {\it Existence of solutions for a Boussinesq system on the half line and on a finite interval}, Discrete Cont. Dyn. Systems {\bf 29} (1) (2011), 25-49.
\bibitem{Am}\textsc{C.J. Amick}, {\it Regularity and uniqueness of solutions for the Boussinesq system of equations}, J. Differential Eq. {\bf 54} (1) (1984), 231-247.
\bibitem{An-Dou}\textsc{D. Antonopoulos and V. Dougalis}, {\it Numerical solution of the "classical" Boussinesq system}, Math. Comp. Simul. {\bf 82} (2012), 984-1007.
\bibitem{An-Dou2}\textsc{D. Antonopoulos and V. Dougalis}, {\it Error bestimates for Galerkin approximations of the "classical" Boussinesq system}, Math. Comp. {\bf 82} (2013), no. 282, 689-717.
\bibitem{ADM3}\textsc{D.C. Antonopoulos, V. A. Dougalis and D. E. Mitisotakis}, {\it Galerkin approximations of periodic solutions of Boussinesq systems}, Bull. Greek Math. Soc. {\bf 57} (2010), 13-30.
%\bibitem{Ang}\textsc{J. Angulo Pava}, {\it On the Cauchy problem for a Boussinesq-type system}, Adv. Diff. Eq. {\bf 4} (1999), 457-492.
%\bibitem{AFM}\textsc{L. Arlen, T. Fromcke and P. Nabelek}, {\it On the integrable Kaup-Broer system with strong capillarity}, preprint (2022).
\bibitem{BeLi}\textsc{H. Berestycki and P.-L. Lions}, {\it Nonlinear scalar field equation I}, Arch. Rat. Mech.Anal. {\bf 32} (1983), 313-346.
\bibitem{BMT}\textsc{L. Brudvik-Lindner, D. Mitsotakis and A. E. Tzavaras}, {\it Oscillatory and regularized shock waves for a dissipative
Peregrine–Boussinesq system}, IMA Journal of Applied Mathematics (2023)
https://doi.org/10.1093/imamat/hxad030
\bibitem{BCL} \textsc {J. L. Bona, T. Colin and D. Lannes}, {\it Long-wave approximation for water waves},
Arch. Ration. Mech. Anal. {\bf 178}, (2005), 373-410.
\bibitem{BCS1} \textsc {J. L. Bona, M. Chen and J.-C. Saut}, {\it Boussinesq equations and
other systems for small-amplitude long waves in nonlinear dispersive
media I : Derivation and the linear theory}, J. Nonlinear Sci. {\bf
12} (2002), 283-318.
\bibitem{BCS2} \textsc{J. L. Bona, M. Chen and J.-C. Saut},
{\it Boussinesq equations and other systems for small-amplitude long
waves in nonlinear dispersive media. II : The nonlinear theory},
Nonlinearity {\bf 17} (2004) 925-952.
\bibitem{Bou2}\textsc{J. Boussinesq}, {\it Th\' eorie des ondes et des remous qui se propagent le long d'un canal rectangulaire horizontal, en communicant au liquide contenu dans ce canal des vitesses sensiblement pareilles de la surface au fond}, J. Math. Pures Appl. {\bf 7} (2) (1872), 55-108.
\bibitem{Broer}\textsc{L.F.J. Broer}, {\it Approximate equations for long water waves}, Appl. Sci. Res. {\bf 31} (1975), 377-395.
\bibitem{Bu}\textsc{C. Burtea}, {\it New long time existence results for a class of Boussinesq-type systems}, J. Math. Pures Appl. {\bf 106} (2) (2016), 203-236.
%\bibitem{Bu2}\textsc{C. Burtea}, {\it Long time existence results for bore-type initial data for BBM-Boussinesq systems}, .J. Diff. Equations {\bf 261} (2016), 4825-4860.
%\bibitem{CDS}\textsc{R. Carles, R. Danchin and J.-C.Saut}, {\it Madelung, Gross-Pitaevskii and Korteweg}, Nonlinearity {\bf 25} (2012), 2843-2873.
\bibitem{asymbook}
{\sc G.~Carrier and M.~K.~C. Pearson}, {\em Functions of a {C}omplex
{V}ariable, {T}heory and {T}echnique}, Society for Industrial and Applied
Mathematics (SIAM), Philadelphia, PA, 2005.
\bibitem{chen}\textsc{Min Chen}, {\it Solitary-wave and multi-pulsed traveling-wave solutions of Boussinesq systems}, Appl. Anal. {\bf 75} (1-2) (2000), 213-240.
%%\bibitem{CIKP} \textsc{T. Congy, S.K. Ivanov, A.M. Kamchatnov and N. Pavloff}, {\it Evolution of initial discontinuities in the Riemann problem for the Kaup-Boussinesq equation with positive dispersion}, Chaos {\bf 27} (8) (2017), 083107.
%\bibitem{CS}\textsc{P. Constantin and J.-C. Saut}, {\it Local smoothing properties of dispersive equations}, J. Amer. Math. Soc. {\bf 1} (1988), 413-439.
\bibitem{CSS2} \textsc{W. Craig and C. Sulem}, {\it Numerical simulation of gravity waves}, J. Comput. Phys. {\bf 108} (1993),
73-83.
\bibitem{Da}\textsc{O. Darrigol}, {\it Worlds of flow. A hydrodynamics from the Bernoullis to Prandtl}, Oxford University Press 2005.
\bibitem{Da2}\textsc{O. Darrigol}, {\it The Spirited Horse, the Engineer, and the Mathematician : Water Waves in Nineteenth -Century Hydrodynamics}, Arch. Hist. Exact Sci. {\bf 58} (2003), 21-95.
\bibitem{Da3}\textsc{O. Darrigol}, {\it Joseph Boussinesq's legacy in fluid mechanics}, C.R. mechanics {\bf 345} (2017), 427-445.
\bibitem{Ding}\textsc{M.W. Dingemans},{\it Water wave propagation over uneven bottoms. Part 2- Non-linear Wave Propagation}, Advances Series on Ocean Engineering-Volume 13. World Scientific, (2000).
%0n of initial discontinuities in the Riemann problem for the Kaup-Boussinsq equation with positive dispersion}, Chaos {\bf 27} (8) (2017), 083107.
%\bibitem{Doi}\textsc{S.I. Do\"{i}}, {\it Remarks on the Cauchy problem for Schr\"{o}dinger-type equations}, Comm. Partial Diff. Eq. {\bf 21} (1996), 163-178.
%\bibitem{DDS}\textsc{V.A. Dougalis, A. Duran and L. Saridaki}, {\it On solitary-wave solutions of Boussinesq/Boussinesq systems for internal waves}, Physica D (2021).
%\bibitem{EGP}\textsc{G.A.El, R.H.J. Grimshaw and M.V. Pavlov}, {\it Integrable shallow-water equations and undular bores}, Studies in Appl. Math. {\bf 106} (2001), 157-186.
\bibitem{DK} V. Duch\^ene, C. Klein, Numerical study of the
Serre-Green-Naghdi equations and a fully dispersive
counterpart, Discr. \& Cont. Dyn. Syst. B, 27(10), 5905-593 (2021) doi:
10.3934/dcdsb.2021300
\bibitem{GK} S. Gavrilyuk, C. Klein, Numerical study of the
Serre-Green-Naghdi equations in 2D, Nonlinearity 37 045014
DOI 10.1088/1361-6544/ad2eb8
\bibitem{GN1}A. E. Green, N. Laws and P. M. Naghdi, On the theory of water waves, Proc. R. Soc. Lond.
A 338 (1974), 43–55.
\bibitem{GN2} A. E. Green and P. M. Naghdi, A derivation of equations for wave propagation in water of
variable depth, J. Fluid Mech. 78 (1976), 237–246.
\bibitem{Fok}\textsc{A.S. Fokas}, {\it A unified transform method for solving linear and certain nonlinear PDE's}, Proc. Roy. Soc. A {\bf 453} (1997), 1411-1443.
\bibitem{FP}\textsc{A.S. Fokas and B. Pelloni}, {\it Boundary value problems for Boussinesq type systems}, Analysis and Geometry {\bf 8} (2005), 59-96.
\bibitem{GAD}\textsc{G. Kounadis, D.C. Antonopoulos, V.A. Dougalis}, {\it Galerkin finite element methods for the numerical solution of two classical Boussinesq type systems over variable bottom}, Wave Motion {\bf 102} (2021), 102715.
%\bibitem{GSS}\textsc{M. Grillakis, J. Shatah and W.A. Strauss}, {\it Stability theory of solitary waves in the presence of symmetry. I}, J. Funct.Anal. {\bf 74} (1987), 160-197.
\bibitem{Gr}\textsc{M.D. Groves}, {\it Hamiltonian long-wave approximations for water waves in a uniform channel}, in Nonlinear Dispersive Wave Systems, Lokenath Debnath Ed. World Scientific (1992), 99-125.
%\bibitem{Hu}\textsc{Junqi Hu}, {\it Global well-posedness of the BCL system with viscosity}, Chin. App. Math. Ser. B {\bf 30} (2) (2009), 153-172.
%\bibitem {KKU}\textsc{A.M. Kamchatnov, R.A. Kraenkel and B.A. Umarov}, {\it Asymptotic soliton train solutions of Kaup-Bousinesq equations}, Wave Motion {\bf 38} (203),,355-365.
%\bibitem{Ka}\textsc{D.J. Kaup}, {\it A higher-order water-wave equation and the method for solving it}, Progr. Theoret. Phys. {\bf 54} (1975), 396-408.
%\bibitem{KK} \textsc{C.E. Kenig and K. D. Koenig,}
%\newblock \emph{On the local well-posedness of the Benjamin-Ono and modified Benjamin-Ono equations,} \newblock Math. Res. Let., \textbf{10} (2003), 879--895.
%\bibitem{KPV}\textsc{C.E. Kenig, G. Ponce and L. Vega}, {\it Smoothing effects and local existence theory for the generalized nonlinear Schr\"{o}dinger equations},
%Invent. Math. {\bf134} (3) (1998), 489-545.
%\bibitem{KS}\textsc{N. Kita and J. Segata}, {\it Well-posedness for the Boussinesq-type system related to the water wave}, Funkcial. Ekvac. {\bf 47} (2004), 329-350.
%\bibitem{Kup}\textsc{B.A. Kuperschmidt}, {\it Mathematics of dispersive water waves}, Communications in Mathematical Physics {\bf 99} (1985), 51-73.
%\bibitem{KMPP}\textsc{C. Kwak, C. Munoz, F. Poblete and J.C. Pozo}, {\it The scattering problem for Hamiltonian ABCD Boussinesq systems in the energy space}, J. Math. Pures et Appl. {\bf 127} 121-159, (2019).
%\bibitem{KM}\textsc{C. Kwak and C. Munoz}, {\it Asymptotic dynamics for the small data weakly dispersive one-dimensional Hamiltonian abcd systems}, Trans. Am; Math. Soc. {\bf 373} (12) 1043-1107, (2019).
\bibitem{JGM}\textsc{C.M. Johnston, C.T. Gartman and D. Mantzavinos}, {\it The linearized classical Boussinesq system on the half-line}, Studies in Applied Math. {\bf 146} (2021), 635-657
\bibitem{KR} \textsc{C.~Klein and K.~Roidot}, \emph{Numerical study of shock formation in the dispersionless
Kadomtsev-Petviashvili equation and dispersive
regularizations}, Physica D, Vol. 265, 1--25, 10.1016/j.physd.2013.09.005 (2013).
\bibitem{KSAS} C. Klein, J.-C. Saut, Numerical study of the
Amick-Schonbek system, Stud Appl Math. 2024;1-22. https://doi.org/10.1111/sapm.12691
\bibitem{KSKBK}C. Klein, J.-C. Saut, On the Kaup-Broer-Kupershmidt
systems,
https://doi.org/10.48550/arXiv.2402.17576
\bibitem{KS15} \textsc{C.~Klein and J.-C.~Saut}, \emph{A numerical
approach to blow-up issues for dispersive perturbations of
Burgers' equation}, Physica D: Nonlinear Phenomena (2015),
46-65, 10.1016/j.physd.2014.12.004
\bibitem{KSbook} \textsc{C.~Klein and J.-C.~Saut}, Nonlinear dispersive equations -
Inverse Scattering and PDE methods,
Applied Mathematical Sciences 209 (Springer, 2022)
\bibitem{La1}\textsc{D. Lannes}, {\it Water waves : mathematical theory and asymptotics}, Mathematical Surveys and Monographs, vol 188 (2013), AMS, Providence.
% \bibitem{LS}\textsc{D. Lannes and J.-C.Saut}, {\it Weakly transverse Boussinesq systems and the KP approximation}, Nonlinearity {\bf 19} (2006), 2853-2875.
%\bibitem{LPS}\textsc{F. Linares, D. Pilod and J.-C. Saut}, {\it Well-posedness of strongly dispersive two-dimensional surface waves Boussinesq systems}, SIAM J. Math. Analysis, {\bf 44} (6) (2012), 4195-4221.
%\bibitem{MY}\textsc{V.B. Matveev and M.I. Yavor}, {\it Solutions presque p\' eriodiques et N-solitons de l'\' equation hydrodynamique non lin\' eaire de Kaup}, Annales de l'IHP, section A, {\bf 31} (1) (1979), 25-41.
\bibitem{Mel}\textsc{B. Melinand}, {\it Dispersion estimates for non-homogeneous radial phases : an application to weakly dispersive equations and water waves models}, J. Funct. Analysis {\bf 286} (2023), 110204.
%\bibitem{MSZ} \textsc{M. Ming, J.-C. Saut and P. Zhang}, {\it Long time existence of solutions to Boussinesq systems}, SIAM. J. Math. Anal. {\bf 44} (6) (2012), 4078--4100.
\bibitem{Meso}\textsc{B. Mesognon-Gireau}, {\it The Cauchy problem on large time for a Boussinesq-Peregrine equation with large topography variations}, Ann. I.H. Poincar\' e-AN {\bf 34} (2017), 89-118.
\bibitem{MTZ}\textsc{L. Molinet, R. Talhouk and I. Zaiter}, {\it The classical Boussinesq system revisited}, Nonlinearity {\bf 34} (2) (2021), 744-775.
%\bibitem{MV}\textsc{L. Molinet and S. Vento}, {\it Improvement of the energy method for strongly nonresonant dispersive equations and applications}, Analysis and PDE {\bf 8} (6) (2015), 1455-1495.
%\bibitem{Nab}\textsc{P.V. Nabelek}, {\it Private communication February 2022.}
%\bibitem{NAF}\textsc{P. Nabelek, A. Arlen and T. Fromcke}, {\it On periodic and finite genus solutions to the inytegrable Kaup-Btroer suystem for capillary waves}, Applied Numerical Mathematics (20234)
%\bibitem{NZ}\textsc{P.V. Nabelek and V.E. Zakharov}, {\it Solutions to the Kaup-Broer system and its (2+1) dimensional integrable generalization via a dressing method}, Physica D {\bf 409} (2020), 132478.
\bibitem{Per}\textsc{D.H. Peregrine}, {\it Long waves on a beach}, J. Fluid Mech.{\bf 27} (1967), 815-827.
%\bibitem{P} \textsc{G. Ponce}, {\it On the global well-posedness of the Benjamin-Ono equation}, Diff. Int. Equations {\bf 4} (3) (1991), 527-542.
\bibitem{GMRES}
\textsc{Y. Saad, M. Schultz}, {\it GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems}, SIAM J. Sci. Stat. Comput. {\bf 7} (3), 856--869 (1986).
\bibitem{SWX} \textsc{J. C. Saut, C. Wang and L. Xu}, The Cauchy problem on large time for surface waves Boussinesq systems II, \emph{SIAM Journal on Mathematical Analysis}, {\bf 49} (4) (2017), 2321--2386.
%\bibitem{SX} \textsc {J.-C. Saut and Li Xu}, {\it The Cauchy problem on large time for surface waves Boussinesq systems}, J. Math.Pures Appl. {\bf 97} (2012), 635--662.
%\bibitem{SX2} \textsc{J.-C. Saut and Li Xu}, {\it Well-posedness on large time for a modified full dispersion system of surface waves}, J. Math. Phys. {\bf 53}, 115606 (2012).
%\bibitem{SX3} \textsc{J.-C. Saut and Li Xu}, {\it Long time existence for a strongly dispersive Boussinesq system}, SIAM J. Math. Anal. {\bf 52} (3) (2020), 2803-2848.
%\bibitem{SX4} \textsc{J.-C. Saut and Li Xu}, {\it Long time existence for a two-dimensional strongly dispersive Boussinesq system}, Communications on PDE {\bf 46} (11) 2057-2087, (2021).
%\bibitem{T} \textsc{T. Tao}, {\it Global well-posedness of the Benjamin-Ono equation in $H^1(\R)$}, J. Hyperbolic Differ. Equ. {\bf 1} (2004), 27-49.
\bibitem{Sch}\textsc{M.E. Schonbek}, {\it Existence of solutions to the Boussinesq system of equations}, J. Differential Eq. {\bf 42} (1981), 325-352.
\bibitem{Ser}F. Serre, Contribution `a l’´etude des ´ecoulements permanents et variables dans les canaux,
La Houille Blanche 8 (1953), 374–388.\bibitem{SSF}
{\sc C.~Sulem, P.~Sulem, and H.~Frisch}, {\em Tracing complex singularities
with spectral methods}, J. Comp. Phys., 50 (1983), pp.~138--161.
%\bibitem{V}\textsc{C.Valls}, {\it Stability of some waves in the Boussinesq system}, Communications in Pure and Applied Analysis {\bf 5} (4), 929-939.
%\bibitem{ZW}\textsc{Junyi Zhu and Xueru Wang}, {\it Broer-Kaup system revisit : inelastic interaction and blow-up}, J. Math. Anal. Appl. {\bf 496}(2021), 124791.
\bibitem{TW}\textsc{M. H. Teng and T. Y. Wu}, {\it Nonlinear water waves in channels of arbitrary shape}, J. Fluid Mech. {\bf 242} (1992), 211-233.
\bibitem{W}\textsc{T. Y.Wu}, {\it A bidirectional long wave model}, Methods and Applications of Analysis. {\bf 1} (1) (1994), 108-117.
\bibitem{Za2} \textsc{V.E. Zakharov}, {\it Weakly Nonlinear Waves on Surface of Ideal Depth Fluids}, Amer. Mat. Soc. Transl. Ser. 2 {\bf 182}, 167-197.
\end{thebibliography}
\end{document}