A divergence-conforming DG-mixed finite element method for the stationary Boussinesq problem (Q2204884)

New numerical methods are developed for the natural convection problem with the aim to approximate the temperature directly. The authors propose and analyze a fully \(H(\operatorname{div})\) divergence-conforming finite element method for the numerical simulation of the Boussinesq problem. For the fluid flow equation, the standard velocity-pressure formulation is considered. The gradient of the temperature is introduced as a further unknown and a dual-mixed formulation is employed for the heat equation. The resulting unknowns are then given by the velocity, the pressure, the temperature, and its gradient. The fluid equation is discretized with the divergence-conforming Galerkin scheme, whereas the heat equation is discretized using \(H(\operatorname{div})\) conforming elements for the gradient of the temperature and discontinuous elements for the temperature. In this way, the gradient of the temperature is approximated directly avoiding numerical differentiation of the temperature field. It was proven that the method is optimal convergent without including any stabilization parameter. In addition, it preserves the divergence-free velocity constraint exactly. The analysis of the continuous and discrete problems is carried out by means of a fixed-point strategy under a sufficiently small data assumption. Furthermore, optimal error estimates are derived in the mesh size for smooth solutions. Numerical results demonstrate the performance of the primal-mixed finite element method and confirming the theoretical rates of convergence.