What Are Finite Element Methods?

Finite element methods are a family of numerical techniques for finding approximate solutions to partial differential equations by subdividing the problem domain into a mesh of small subregions called elements, approximating the unknown function within each element using polynomial basis functions, and assembling the element contributions into a global system of equations. The method converts boundary-value problems and initial-value problems that govern physical phenomena into linear or nonlinear algebraic systems that digital computers can solve. Its scope covers structural mechanics, heat transfer, fluid dynamics, acoustics, and electromagnetics, making it one of the most widely applied computational frameworks in science and engineering.

The mathematical foundations of the finite element method rest on the variational calculus and the Galerkin weighted-residual formulation, which recast differential equations as equivalent integral statements. Structural engineers at Boeing published the first systematic description of the method in 1956, and applied mathematicians working in the 1970s established its rigorous error and convergence theory, showing that the solution error decreases in a predictable manner as the mesh is refined. The method is closely related to, but distinct from, finite difference methods: where finite difference schemes approximate derivatives directly on a structured grid, finite element methods approximate the solution space itself on an unstructured mesh, which allows complex geometries to be handled more naturally.

Formulation and Basis Functions

The key idea in finite element methods is the choice of a finite-dimensional function space to approximate the exact solution. In the classical Galerkin approach, the unknown field is expressed as a linear combination of basis functions that are nonzero only over a small number of adjacent elements. Lagrange polynomials of degree one (linear elements) are the most common choice; higher-degree elements (quadratic, cubic) achieve greater accuracy per degree of freedom at the cost of larger element matrices. The residual of the governing equation, weighted by the same basis functions, is set to zero over the domain, yielding a system of equations whose unknowns are the values of the field variable at the mesh nodes. Detailed exposition of this process appears in Zienkiewicz and Taylor's standard reference text on finite element methods, which traces the method's development from structural to multiphysics applications.

Variants and Extensions

The classical conforming finite element method has spawned a family of variants designed to address its shortcomings in particular problem classes. Mixed finite element methods introduce additional unknown fields (such as stress alongside displacement) to achieve better performance for nearly incompressible materials or saddle-point problems. Discontinuous Galerkin methods relax the continuity requirements between elements, offering advantages for convection-dominated transport and high-order wave propagation. Extended finite element methods (X-FEM) enrich the standard basis with discontinuous functions to model cracks and interfaces without remeshing. The hp-FEM combines mesh refinement (h-refinement) with polynomial degree increase (p-refinement) to achieve exponential convergence for problems with smooth solutions. These developments are surveyed in NIST's Digital Library of Mathematical Functions and in the SIAM Journal on Numerical Analysis, which has published much of the foundational theory.

Error Analysis and Convergence

A key advantage of finite element methods over many alternative numerical schemes is the availability of rigorous a priori and a posteriori error estimates. A priori estimates bound the error in terms of the mesh size and the regularity of the exact solution, guiding the choice of element type and polynomial degree. A posteriori estimates, computed from the numerical solution itself, drive adaptive mesh refinement algorithms that automatically concentrate computational resources where the error is largest.

Applications

Finite element methods have applications in a range of fields, including:

• Computational electromagnetics for antenna, motor, and transformer design
• Crash and impact simulation in automotive and aerospace structures
• Geomechanical modeling for petroleum reservoir and tunnel engineering
• Biomedical device simulation and patient-specific surgical planning
• Semiconductor device modeling and MEMS design