A linear or nonlinear operator that transfers state or field data from a fine computational grid or model to a coarser representation by aggregation, averaging, sampling, or projection, used in multilevel methods, coupling, and analysis.

Principle

A restriction operator should map fine-scale information to coarse variables while preserving chosen invariants (mass, momentum, energy) or statistical moments as required by the application; it formalizes the coarse-graining step between hierarchical discretizations.

Demonstration

In a multigrid solver for Poisson's equation, the restriction operator computes coarse-grid residuals by volume-weighted averaging of fine-grid residuals; in finite-volume schemes, cell-average restriction maps fine-cell values into parent-cell averages conserving integral quantities.

Misapplication

Using pointwise sampling without accounting for volume weighting or conservation, which causes aliasing and loss of integral invariants, or applying the same restriction across heterogeneous meshes without adjusting for differing cell geometries.

Consequence

A properly designed restriction enables consistent multilevel solvers, conservative coupling between resolutions, and meaningful error estimators; poor restriction degrades convergence, introduces bias, and can violate conservation laws at the coarse scale.

Reversal

The inverse concept is the prolongation or interpolation (prolongation operator) that maps coarse representations to finer grids, often requiring reconstruction assumptions and not being unique in the absence of additional constraints.

Boundary

Applies to mappings between discretized representations (grids, finite elements, degrees of freedom) and excludes lossy data compression formats or observation down-sampling that introduce measurement noise rather than formal discretization projection.

Semantic Tension

Related to projection and filtering operators: restriction can act as a projection (L2, mass-weighted) or a filter; tension arises in choosing between simple sampling, conservative averaging, or orthogonal projection depending on whether conservation or error-minimization is prioritized.

Synthesis

A restriction operator is the formal mapping from fine to coarse discretizations realized by averaging, sampling, or projection that aims to preserve key invariants and support multilevel algorithms; its design balances conservation, accuracy, and computational convenience.