Projection methods preserving Lyapunov functions (Q981671): Difference between revisions

The authors propose explicit Runge-Kutta methods for the numerical solution of initial value problems for autonomous ordinary differential equations which have a known Lyapunov function. The integrators preserve this geometric quantity by proceeding as follows: First, one step of the Runge-Kutta method is performed. Subsequently, an approximation to the Lyapunov function is computed based on the dense output of the Runge-Kutta method and Gaussian quadrature. Finally, the solution is projected to the manifold characterized by the property that the exact Lyapunov function coincides with its numerical approximation previously computed. The theoretical analysis gives the well-posedness of the process and proves the convergence order. Numerical experiments illustrate the theory and demonstrate the favorable performance of the new methods.