The Computer Journal · 1964 · 39 citations · 0 references
Natural SciencesMathematicsComputational PhysicsDifferential EquationsFluid DynamicsBoundary ConditionsBoundary-value ProblemsError EstimationFinite Difference MethodsNumerical Methods For Differential EquationsPartial Differential EquationsDiscretizationDiscretization ErrorDiscretization Of Partial Differential EquationsFinite Difference DiscretizationFinite Difference SchemesInitial Boundary ConditionsNumerical DiscretizationParabolic Partial Differential Equations
For some simple type of boundary conditions, errors introduced by the use of finite-difference representations in the region of a discontinuity, either in the initial conditions or between the initial conditions and the boundary conditions, may persist right through to the steady-state solution of a one-dimensional parabolic partial differential equations. These errors depend on the mesh length, δx, in the x direction, and, in the simplest cases, are equal to k(δx)p, where k is a constant depending on the initial distribution, and where the discontinuity occurs in the (p − 1)th derivative. Analytic expressions for these persistent discretization errors are given in simple cases, with a general initial distribution defined by its values at discrete points, δx apart. The error persist when the steady-state solution depends on the initial condition as well as the boundary conditions.