On the Kaup-Broer-Kupershmidt systems (Q6966442)

The authors start from the so-called abcd Boussinesq systems for surface water waves: \(\eta _{t}+\nabla \overset{.}{v}+\epsilon \nabla \cdot (\eta v)+\mu \lbrack a\nabla \cdot \Delta v-b\Delta \eta _{t}]=0\), \(v_{t}+\nabla \eta +\epsilon \frac{1}{2}\nabla \left\vert v\right\vert ^{2}+\mu \lbrack c\nabla \Delta \eta -d\Delta v_{t}]=0\), where \(\eta =\eta (x,t)\), \(x\in \mathbb{R}^{d}\), \(d=1,2\), \(t\in \mathbb{R}\), is the wave elevation, \(v=v(x,t) \) a measure of the horizontal velocity, \(\mu =\frac{h^{2}}{\lambda ^{2}}\) and \(\epsilon =\frac{\alpha }{h}\) the small shallowness and nonlinearity parameters, \(\alpha \) being a typical amplitude of the wave, \(h\) a typical depth and \(\lambda \) a typical horizontal wavelength. Considering the 1D case and assuming \(\epsilon \sim \mu \ll 1\), \(a=\pm 1\), \(b=c=d=0\), they obtain the Kaup-Broer-Kupershmidt system: \(\eta _{t}+v_{x}+(\eta v)_{x}+\alpha v_{xxx}=0\), \(v_{t}+\eta _{x}+vv_{x}=0\), with \(\alpha \pm 1\), that they transform as: \(\zeta _{t}+(\zeta v)_{x}+\alpha v_{xxx}=0\), \( v_{t}+\zeta _{x}+vv_{x}=0\), with \(\zeta =1+\eta \). The authors recall that this system is linearly ill-posed if \(\alpha =+1\) and linearly well-posed if \(\alpha =-1\). They recall the expression of the Hamiltonian, the expressions of traveling wave solutions and of stationary solutions. They recall the results obtained when applying the theory of integrable systems in the present context, leading to a hierarchy of Kaup-Broer-Kupershmidt Boussinesq systems in both cases. The main part of the paper presents a numerical approach to the Kaup-Broer-Kupershmidt system, mainly in the case \(\alpha =-1\). The authors recall that the energy \(E=\frac{1}{2}\int_{\mathbb{R}}(\eta ^{2}+(1+\eta )v^{2}+v_{x}^{2})dx\) is preserved. They apply Fourier transform to the Kaup-Broer-Kupershmidt system, then the discrete Fourier transform, which can be computed by the fast Fourier transform. They obtain a \(2N\) -dimensional system of equations of the form \(u_{t}=\mathcal{L}u+\mathcal{N}u \), where \(\mathcal{L}\) is a linear diagonal operator proportional to \(\pm ik \sqrt{1+k^{2}}\), and \(\mathcal{N}u\) a nonlinear term in \(\pm u\). To solve this system, they use exponential time differencing schemes based on equidistant time steps \(h\) and integrating this equation between the time steps \(t_{n}\) and \(t_{n+1}\), \(n=1,2,\ldots \) with an exponential integrator with respect to \(t\), leading to \(u(t_{n+1}=e^{\mathcal{L}h}u(t_{n})+ \int_{0}^{h}e^{\mathcal{L(}h-\tau )}\mathcal{N}(u(t_{n}+\tau ),t_{n}+\tau )d\tau \). The authors present results of numerical simulations in the case of an initial soliton, of an initial perturbed soliton, of localized initial data, and for dispersive shock waves.