The process of replacing continuous mathematical objects — functions, operators, domains — with discrete counterparts (grids, matrices, finite elements, etc.) to enable numerical computation and algorithmic solution.

Principle

Approximation and convergence: discretization constructs a finite-dimensional representation that approximates the continuous problem; analysis of consistency, stability and convergence ensures the discrete solutions approach the continuous solution as resolution is refined.

Demonstration

Finite-difference discretization of the one-dimensional Poisson equation on a uniform grid replaces second derivatives by central difference quotients, producing a sparse linear system Ax=b whose solution approximates the continuous solution at grid nodes.

Misapplication

Using an overly coarse grid or an inconsistent stencil that violates stability conditions (e.g., ignoring CFL conditions in time-dependent problems), leading to numerical dispersion, instability, or convergence to an incorrect limit.

Consequence

A sound discretization yields a computable algebraic problem whose solution converges to the true continuous solution under mesh refinement and whose error can be estimated and controlled for practical computation.

Reversal

Instead of discretizing continuous models, one can formulate discrete models as primary (e.g., agent-based or network models) and only seek continuum limits; reversal emphasizes deriving continuum behavior from discrete rules rather than approximating continuous equations numerically.

Boundary

Applies to problems where the continuous formulation is known and where numerical approximation is intended; excludes purely symbolic manipulation that yields exact closed-form solutions and contexts where discretization destroys essential structure (e.g., some geometric invariants) unless handled with structure-preserving schemes.

Semantic Tension

Discretization as numerical approximation versus modeling choice: as approximation it is a surrogate for a known continuous model; as modeling it can be the primary representation of phenomena, creating tension about interpretation of errors and validity.

Synthesis

Discretization turns continuous mathematical formulations into finite algebraic systems suitable for computation, guided by approximation properties and error analysis to ensure fidelity and convergence to the underlying continuous problem.