The complex KdV equation with or without dissipation (Q558613)

The authors are concerned with the complex KdV equation \[ u_t + 2u u_x + u_{xxx}=0 \] and with the complex KdV--Burgers equation \[ u_t + 2u u_x + u_{xxx} - \nu u_{xx}=0, \] where \(\nu\) is the diffusion or dissipation coefficient. Both equations are suplemented with the initial condition \[ u(x,0)=u_0(x), \] where \(u_0\) is a reasonably smooth, \(2\pi\)-periodic complex-valued function. Sections 2 and 3 are devoted to the complex KdV-equation, section 4 to the complex KdV-Burgers equation, and section 5 to numerical experiments with both equations. In section 2 the authors derive, for any complex-valued solution of the KdV equation, an explicit relationship between the regularity of the real part and that of the imaginary part. They show that the real part is bounded in \(L^2\) or in \(H^k\) (\(k\geq 1\)) if and only if the imaginary part is bounded in the same space. As a consequence, any solution of the complex KdV equation is regular for all time if its real part remains so. In section 3, solutions of the form \[ u(x,t)=\sum_{k=1}^\infty a_k(t)e^{ikx} \] are considered, where \(a_k(t)\) may be complex-valued. The complex KdV equation is then formally reduced into a system of infinite ordinary differential equations of \(a_k\)'s. These equations allow the authors to represent each \(a_k\) as a series in the temporal variable \(t\). With this idea, the solution is finally written in the following form: \[ u(x,t)=\sum_{k=1}^\infty \sum_{h=k}^{k^3} \alpha_{k,h} e^{iht} e^{ikx}. \] Iterative relations for the coefficients \(\alpha_{k,h}\) are derived. These relations, in particular, imply that \(\alpha_{k,h}=0\) for any \(k\geq 1\) and \(k^3-3k^2+3k<h<k^3\). In addition, an explicit formula for \(\alpha_{k,k}\) is found through these relations. As a consequence, it is proved that any solution of this type becomes unbounded in \(L^2([0,2\pi]\times [0,2\pi])\) in a finite time if \(| a_1(0)| \geq 6\), being \(a_1(0)\) the first coefficient in the initial datum \(u_0\). This result is sharp in the sense that there exists a series with \(| a_1(0)| <6\) which is regular for all time. Aiming at determining the \(L^2\)-convergence of the double series, the authors compute symbolically through a Maple procedure the expressions of the \(a_k\)'s in terms of the initial coefficients \(a_k(0)\)'s. Using these formulas, they compute \(\| u\| _{L^2([0,2\pi]\times [0,2\pi])}\) for a special class of initial data \(u_0\), namely \(a_k(0)=1/k^\beta\) with \(\beta > 1/2\). The numerical results indicate that \(u\) is bounded and its \(L^2\)-norm has an interesting connection to its initial \(L^2\)--norm when \(\beta=1\). It is also investigated the question of how many \(\alpha_{k,h}\) are nonzero for each fixed \(k\geq 1\). Section 4 is devoted to the complex KdV-Burgers equation. The point is to see how dissipation modifies the regularity of solutions to the complex KdV equation. It is found an explicit bound \(M_0\) for which \[ \| u(\cdot,t)\| _{ L^2}\leq M_0\quad \text{for}\;t<T^\ast, \] where \(T^\ast\) is represented in terms of the dissipation coefficient \(\nu\) and the initial datum \(u_0\). Either increasing \(\nu\) or decreasing \(\| u_0\| _{L^2}\) lengthens \(T^\ast\). The authors demonstrate that, if the \(L^2\)-norm were bounded for all time (it is not known if \(u\) remains bounded in \(L^2\) for \(t\geq T^\ast\)), then any \(H^k\)-norm with \(k\geq 1\) is also bounded for all time. This result implies that any possible finite-time singularity of the complex KdV equation must develop in the \(L^2\)--norm. Section 5 presents the results of systematic numerical simulations. Numerical experiments are carried out to test how the sizes of initial data and dissipation impact the regularity of the solution of the complex KdV equation. The fully discrete schemes amount to the Gauss-Legendre methods on the systems of ordinary differential equations that arise from the Galerkin semidiscretization. Adaptive mechanisms are used to adjust the temporal and local spatial grids to retain accuracy in the situation of large dependent variable. Subsection 5.2 reports numerical computations on the complex KdV equation. Attention is focused on how the regularity of its solutions changes with the magnitude of the initial data. In subsection 5.3, numerical solutions of the complex KdV-Burgers equation are computed. The authors place a special emphasis on the role of dissipation in the regularity. Solutions corresponding to a range of \(\nu\) are computed. The results indicate that for any given initial datum \(u_0\) there exists a critical value \(\nu^\ast\) such that the solution is smooth for all time if \(\nu > \nu^\ast\). Efforts are made to pin down the critical \(\nu^\ast\)'s.