A projection-based approximation method for solving differential and integral equations: one selects a finite-dimensional trial space and enforces that the residual (the PDE or operator applied to the trial solution minus the data) is orthogonal to a chosen test space, yielding a finite system for the expansion coefficients.

Principle

Approximate the infinite-dimensional problem by restricting solutions to a finite subspace and require the residual to be orthogonal to the test space (often the same as the trial space); this projects the problem onto a finite-dimensional subproblem, preserving variational structure when present.

Demonstration

Finite element method: choose piecewise-polynomial basis functions on a mesh as trial/test space; enforce weak form of −(p u')' = f so that ∫(p u'_h v') dx = ∫ f v dx for all test functions v in the space; solve resulting linear system for coefficients of u_h.

Misapplication

Choosing incompatible trial/test spaces (leading to instability), ignoring boundary conditions in the weak form, or using Galerkin with symmetric test/trial spaces on convection-dominated problems where upwinding or Petrov–Galerkin variants are needed; expecting spectral convergence without sufficient regularity or appropriate basis.

Consequence

Yields systematic, convergent discretizations under appropriate approximation and stability conditions (Céa-type estimates); it unifies many methods (FEM, spectral Galerkin, variational approximations) and preserves conservation/variational properties when the weak form is respected.

Reversal

Collocation or least-squares methods enforce the residual at points or minimize a norm of the residual rather than imposing orthogonality; Petrov–Galerkin generalizes Galerkin by using different test spaces to enhance stability.

Boundary

Applies to problems admitting a weak/variational formulation or operator view where projection makes sense; not universally optimal (e.g., some non-self-adjoint, convection-dominated, or non-smooth problems require modifications), and success depends on basis choice and mesh/refinement strategy.

Semantic Tension

Tension between Galerkin (orthogonality-based projection) and other approximation philosophies (pointwise collocation, minimization of residual norm, or direct discretization). There is also confusion between Galerkin as a recipe and specific instances (FEM, spectral methods).

Synthesis

The Galerkin method approximates infinite-dimensional operator problems by projecting onto finite-dimensional subspaces and enforcing residual orthogonality; with proper space choice it yields stable, convergent discretizations that respect variational structure, while variants (Petrov–Galerkin, stabilized schemes) address instabilities and non-self-adjointness.