Galerkin Approximation Method

• Galerkin Approximation Method is a technique that constructs finite-dimensional approximations for infinite-dimensional variational problems, ensuring rigorous convergence and error control.
• It systematically projects problems onto nested subspaces in Hilbert spaces, making it applicable to both linear and nonlinear partial differential equations.
• The method underpins finite element and spectral techniques through algorithmic steps like stiffness matrix assembly and error estimates via Céa’s lemma.

The Galerkin approximation method is a foundational strategy for constructing finite-dimensional approximations to infinite-dimensional variational problems arising in partial differential and integral equations. Originating in Hilbert space analysis, the method systematically projects the original problem onto a sequence of subspaces, generating computable linear (or nonlinear) systems whose solutions converge to that of the original problem. Rigorous convergence results, optimality estimates, computational implementations, and extensions to nonlinear problems are key ingredients of its modern formalism (Venkatraman, 2011).

1. Variational Formulation in Hilbert Spaces

The essential setting is a real Hilbert space $V$ equipped with norm $\|\cdot\|_V$ and inner product $(\cdot,\cdot)_V$. One is given:

• A bilinear form $a: V \times V \to \mathbb{R}$, assumed bounded ($|a(u,v)| \leq M\|u\|_V\|v\|_V$ for some $M>0$) and coercive, or $V$-elliptic ($a(v,v)\geq \alpha\|v\|_V^2$ for some $\alpha>0$).
• A continuous linear functional $L\in V'$.

The model variational problem is: find $u\in V$ so that

$a(u,v) = L(v)\quad\forall v\in V.$

The Lax–Milgram lemma applies to guarantee unique solvability provided $a(\cdot,\cdot)$ is bounded and coercive.

2. Galerkin Discretization and Algebraic System

One constructs a sequence of nested finite-dimensional subspaces $V_1\subset V_2\subset\cdots\subset V$ with $\dim V_n=n$ and $\bigcup_n V_n$ dense in $V$. For fixed $n$, a basis $\{\varphi_1,\dots,\varphi_n\}$ spans $V_n$. The Galerkin approximation $u_n\in V_n$ is defined by

$a(u_n,v) = L(v)\quad\forall v\in V_n.$

Expressing $u_n = \sum_{j=1}^n u_j\varphi_j$, and testing against basis vectors, gives the linear system of equations

$$a\left(\sum_j u_j\varphi_j,\varphi_i\right) = L(\varphi_i) \quad i=1,\dots,n,$$

or equivalently in matrix form, $A u = b$ with $$A_{ij} = a(\varphi_j,\varphi_i),\; b_i = L(\varphi_i)$$.

3. Convergence Analysis and Error Estimates

Céa’s Lemma

With $u$ the exact solution and $u_n$ the Galerkin approximation,

$$\|u-u_n\|_V \leq \frac{M}{\alpha} \inf_{v_n\in V_n}\|u-v_n\|_V.$$

This asserts that the Galerkin error is controlled up to the extremal constants of $a$ by the best-approximation error in $V_n$.

Density and Convergence Rate

If $\bigcup_n V_n$ is dense in $V$, then $\inf_{v_n\in V_n}\|u-v_n\|_V\to0$ as $n\to\infty$, implying $\|u-u_n\|_V\to0$. Suppose further that $u$ is sufficiently smooth so that best approximation in $V_n$ yields $\inf_{v_n\in V_n}\|u-v_n\|_V = O(h^s)$ for mesh size $h\sim n^{-1/d}$; then

$\|u-u_n\|_V = O(h^s).$

4. Algorithmic Implementation and Computational Issues

A typical Galerkin computational pipeline includes:

1. Stiffness Matrix Assembly: Compute $$A_{ij}=\int_\Omega \nabla\varphi_j\cdot\nabla\varphi_i$$ (or more general $a(\varphi_j,\varphi_i)$) via numerical quadrature.
2. Load Vector Assembly: Compute $b_i=\int_\Omega f\,\varphi_i$.
3. Linear Solve: Solve $Au=b$ using direct or iterative linear solvers.

Methodological insights:

• Polynomial bases $x^i(1-x)$ tend to yield ill-conditioned systems for large $n$; sine or orthogonal trigonometric bases produce diagonal or well-conditioned matrices.
• Quadrature rules should be selected to match the degree and smoothness of basis functions and data.
• Numeric evaluation of the condition number (e.g., cond(A,2) in MATLAB) is essential to monitor stability and solve accuracy.

5. Nonlinear Problem Extensions via Galerkin

For nonlinear PDEs, such as the semilinear Dirichlet problem

$$-\Delta u - \lambda u + u^3 = f \quad \text{in } \Omega, \;\; u=0 \text{ on } \partial\Omega,$$

the weak formulation seeks $u\in H^1_0(\Omega)$ so that

$$\int_\Omega \nabla u\cdot\nabla v - \lambda u v + u^3 v = \int_\Omega f v\quad\forall v\in H^1_0(\Omega).$$

The Galerkin scheme projects onto finite-dimensional orthonormal subspaces $W_m$ of $H^1_0(\Omega)$, yielding a finite system $F(\xi)=0$ for the coefficients in the basis expansion. Existence is established using fixed-point arguments and compactness; as $m\to\infty, u_m$ converges weakly to a solution $u$ of the original nonlinear equation.

6. Context and Significance in Numerical Analysis

The Galerkin approximation method provides a systematic variational framework for discretizing PDEs and integral equations. Its major contributions are:

• Generality: Applicable to broad classes of linear and nonlinear problems.
• Rigorous Error Control: Céa’s lemma and density arguments establish convergence rates aligned with best-approximation theory.
• Algorithmic Foundations: It forms the basis for finite element, spectral, and various modern computational methods.
• Nonlinear and Existence Theory: Extends to nonlinear systems with guaranteed existence under coercivity and boundedness.

The Hilbert-space Galerkin method synthesizes functional analytic rigor, computational tractability, and extensibility to nonlinear variational problems (Venkatraman, 2011).