SIAM Journal on Scientific Computing · 2006 · 71 citations · 12 references
ComputingMathematicsComputational MechanicsFinite Element AnalysisFinite Element MethodsBoundary ConditionsBoundary-value ProblemsFinite Difference MethodsFinite Element DiscretizationFinite Element FormulationNumerical AlgorithmsDiscrete Maximum PrincipleDiscretizationFinite Difference DiscretizationFinite Difference Solution Of Partial Differential EquationsGalerkin DiscretizationNumerical DiscretizationNumerical Solutions Of Partial Differential EquationsParabolic Boundary-value Problems
In this paper, we analyze the connections between the different qualitative properties of numerical solutions of linear parabolic problems with Dirichlet‐type boundary condition. First we formulate the qualitative properties for the differential equations and shed light on their relations. Then we show how the well‐known discretization schemes can be written in the form of a one‐step iterative process. We give necessary and sufficient conditions of the main qualitative properties of these iterations. We apply the results to the finite difference and Galerkin finite element solutions of linear parabolic problems. In our main result we show that the nonnegativity preservation property is equivalent to the maximum‐minimum principle and they imply the maximum norm contractivity. In one, two, and three dimensions, we list sufficient a priori conditions that ensure the required qualitative properties. Finally, we demonstrate the above results on numerical examples.
12