A structure preserving discretization for the Derrida-Lebowitz-Speer-Spohn equation based on diffusive transport (Q6926531)

The authors develop and analyze a novel spatial discretization for the fourth-order nonlinear Derrida-Lebowitz-Speer-Spohn (DLSS) equation on the circle,\N\[\N\partial_t \rho = -\partial_{xx}(\rho\,\partial_{xx} \log \rho),\N\]\Nalso known as the quantum drift-diffusion equation. Their approach is grounded in a new gradient flow formulation with respect to a \textit{diffusive transport metric}, which generalizes martingale transport. The resulting discrete dynamics preserves the essential structural features of the continuous model -- mass conservation, nonnegativity, dissipation of discrete Lyapunov functionals (entropy, Fisher information, and logarithmic entropy), and contractility in the discrete Hellinger distance. Furthermore, the discrete system admits dual gradient flow formulations in both the discrete Wasserstein and diffusive transport metrics.\N\NThe main analytical result proves convergence of the discrete solutions to weak solutions of the DLSS equation as the mesh size tends to zero, based on a discrete nonlinear functional inequality. The exposition combines a clear structural motivation with a rigorous convergence proof and detailed numerical implementation, including adaptive implicit time-stepping that maintains Lyapunov monotonicity.\N\NThis work contributes significantly to the theory of structure-preserving numerical schemes for higher-order nonlinear diffusion equations, extending the geometric and variational perspective of Wasserstein gradient flows to a novel metric setting.