A thin region near a domain boundary where discretization or numerical treatment produces steep gradients, spurious oscillations, or error amplification that resemble a physical boundary layer but originate from numerical approximation.

Principle

Discretization and boundary treatment can introduce local imbalance between resolved scales and imposed constraints, producing high local error gradients concentrated in a few mesh cells or time steps adjacent to the boundary.

Demonstration

In finite-difference solutions of advection-dominated transport with an incompatible inflow profile, a sharp layer of oscillatory or smeared error forms in the first few grid cells from the boundary — a numerical boundary layer.

Misapplication

Interpreting all steep boundary gradients as physical rather than diagnosing discretization-induced layers can lead to incorrect parameter tuning, excessive mesh refinement, or wrong physical conclusions.

Consequence

Recognizing numerical boundary layers permits targeted fixes (improved boundary discretization, upwinding, local refinement, or filtering) and avoids misattributing artifacts to model physics.

Reversal

A genuine physical boundary layer is produced by the model physics (viscosity, diffusion) and scales with physical parameters; identifying the reversal requires analysis of convergence with mesh/time refinement and parameter limits.

Boundary

Refers specifically to discretization-induced localized error phenomena near computational boundaries; excludes physical boundary layers and large-scale global discretization errors away from boundaries.

Semantic Tension

Tension exists between interpreting a boundary-localized feature as a numerical artifact versus a legitimate physical layer; distinguishing them often requires mesh refinement studies and analytic scaling arguments.

Synthesis

A numerical boundary layer is the boundary-localized pattern of steep errors or oscillations generated by numerical approximation or boundary treatment; diagnosing it separates numerical artifact from physical structure and guides corrective discretization strategies.