Céa's Lemma in Finite Element Analysis

This lecture note explores Céa's lemma—a cornerstone result in finite element analysis that provides a quasi-optimal error estimate for finite element approximations of variational problems. We review the variational formulation, state the lemma with the necessary assumptions, outline its proof, and discuss its implications.

1. Introduction

Finite element methods (FEM) are widely used for numerically solving partial differential equations (PDEs). A central question in FEM is:

How close is the finite element solution to the true solution of the PDE?

Céa's lemma provides an answer by bounding the error between the true solution and its finite element approximation in terms of the best possible approximation error from the finite-dimensional space.

2. Variational Formulation

Consider a Hilbert space $V$ and a bilinear form $a(\cdot,\cdot)$ defined on $V \times V$. We are given a bounded linear functional $L$ on $V$. The continuous variational problem is to find $u \in V$ such that

The finite element method approximates $u$ by seeking a solution $u_h$ in a finite-dimensional subspace $V_h \subset V$ satisfying

3. Assumptions on the Bilinear Form

For the analysis, the bilinear form $a(\cdot,\cdot)$ must satisfy two key properties:

1. Continuity: There exists a constant $C_b > 0$ such that

   $a(u, v) \le C_b \|u\|_V \|v\|_V \quad \forall u,v \in V.$

2. Coercivity (Ellipticity): There exists a constant $C_s > 0$ such that

   $a(v,v) \ge C_s \|v\|_V^2 \quad \forall v \in V.$

These conditions ensure the well-posedness of the continuous problem via the Lax-Milgram theorem.

4. The Continuous and Discrete Problems

• Continuous Problem: Find $u \in V$ such that

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

• Finite Element Problem: Given the finite-dimensional subspace $V_h \subset V$, find $u_h \in V_h$ such that

  $a(u_h, v_h) = L(v_h) \quad \forall v_h \in V_h.$

The goal is to estimate the error $\|u - u_h\|_V$ between the true solution and its approximation.

5. Statement of Céa's Lemma

Céa's lemma provides the following error bound:

This inequality means that the finite element solution $u_h$ is quasi-optimal; it is nearly as good as the best approximation to $u$ from the space $V_h$, up to the constant factor $\frac{C_b}{C_s}$.

6. Outline of the Proof

The proof of Céa's lemma involves several key steps:

1. Galerkin Orthogonality:
   The finite element solution $u_h$ satisfies the property

   $a(u - u_h, v_h) = 0 \quad \forall v_h \in V_h.$

   This orthogonality is central to the error analysis.

2. Error Decomposition:
   For any $v_h \in V_h$, decompose the error as

   $u - u_h = (u - v_h) + (v_h - u_h).$

3. Application of Continuity and Coercivity:
   Using coercivity, we have

   $C_s \|u - u_h\|_V^2 \le a(u - u_h, u - u_h).$

   Then, by Galerkin orthogonality and the continuity of $a(\cdot,\cdot)$,

   $a(u - u_h, u - u_h) = a(u - u_h, u - v_h) \le C_b \|u - u_h\|_V \|u - v_h\|_V.$

   Dividing both sides by $C_s \|u - u_h\|_V$ (assuming $u \neq u_h$) yields

   $\|u - u_h\|_V \le \frac{C_b}{C_s} \|u - v_h\|_V.$

   Since this inequality holds for every $v_h \in V_h$, taking the infimum over $V_h$ leads to

   $\|u - u_h\|_V \le \frac{C_b}{C_s} \inf_{v_h \in V_h} \|u - v_h\|_V.$

7. Implications of Céa's Lemma

• Quasi-Optimality:
  The lemma shows that the finite element solution $u_h$ is nearly as good as the best approximation available in $V_h$. If the finite element space $V_h$ is chosen such that

  $\inf_{v_h \in V_h} \|u - v_h\|_V$

  is small, then the error $\|u - u_h\|_V$ will also be small.

• Error Analysis and Convergence:
  Combining Céa's lemma with approximation theory (which provides estimates for the best approximation error) enables the derivation of convergence rates for the finite element method.

• Design of Finite Element Spaces:
  Since the overall error is bounded by the best approximation error in $V_h$, much effort in FEM research is dedicated to constructing finite element spaces with excellent approximation properties.

8. Conclusion

Céa's lemma is a fundamental tool in finite element error analysis. It ensures that the error $\|u - u_h\|_V$ between the true solution and the finite element approximation is bounded by the best approximation error achievable within the finite-dimensional space $V_h$, scaled by the factor $\frac{C_b}{C_s}$. This quasi-optimal error estimate is crucial for both the theoretical understanding and practical implementation of finite element methods.