A projection-based approximation technique for differential or integral equations that seeks approximate solutions in a finite-dimensional subspace by imposing that the residual be orthogonal to that subspace (or a test subspace).

Principle

Replace an infinite-dimensional variational problem by a finite-dimensional one: choose trial and test spaces, project the residual to zero on the test space, and obtain solvable algebraic systems whose solutions approximate the true solution under consistency and stability conditions.

Demonstration

To approximate the Poisson problem −Δu=f on a bounded domain, pick a finite basis of H^1_0(Ω) (e.g. piecewise linear hat functions) and require ∫Ω ∇u_h·∇v_h = ∫Ω f v_h for every basis test function v_h; assembling yields a sparse linear system for the coefficients of u_h.

Misapplication

Using non-conforming trial/test spaces, ignoring compatibility (inf-sup) conditions for mixed problems, or failing to control interpolation/stability leads to non-convergent schemes or spurious modes.

Consequence

Under appropriate approximation properties and stability (Galerkin orthogonality, Céa's lemma), the method yields convergent approximations with quantifiable error bounds and a practical route to numerical solution via matrix assembly and linear solvers.

Reversal

Choosing test spaces different from trial spaces yields Petrov–Galerkin schemes with different stability characteristics; collocation or least-squares methods abandon orthogonality in favor of pointwise or least-squares residual control.

Boundary

Applies to variational formulations on Hilbert or Banach spaces where suitable finite-dimensional subspaces exist; excludes schemes without projection structure (pure collocation without variational underpinning) or problems lacking a well-posed weak form.

Semantic Tension

Often contrasted with collocation, least-squares or spectral methods: Galerkin enforces weak orthogonality and inherits functional-analytic stability, while other approaches prioritize pointwise matching or global basis properties.

Synthesis

The Galerkin method projects an infinite-dimensional variational problem onto finite-dimensional subspaces, producing algebraic systems whose solutions approximate the true solution when consistency and stability (approximation and inf-sup) requirements are met.