Finite element exterior calculus: from Hodge theory to numerical stability

Douglas N. Arnold; Richard S. Falk; Ragnar Winther

Source-linked AI summary

Finite element exterior calculus: from Hodge theory to numerical stability

Douglas N. Arnold, Richard S. Falk, Ragnar Winther

arXiv:0906.4325v3Jun 23, 2009math.NAmath.DGmath.GT

TL;DR

The paper addresses stable numerical discretization of PDEs linked to differential complexes, where de Rham cohomology and Hodge theory govern the continuous problem. It develops a Hilbert-complex framework and constructs de Rham finite element subcomplexes with bounded cochain projections, yielding well-posed formulations and stable discretization results.

Problem
PDEs related to differential complexes require discretizations that preserve the de Rham cohomology and Hodge-theoretic structures relevant to continuous well-posedness.
Method
The paper develops an abstract Hilbert-complex framework and constructs finite element differential-form subcomplexes with bounded cochain projections.
Results
The mixed formulation of the abstract Hodge Laplacian is well-posed, with a unique solution for every f ∈ W^k.
Takeaways & Limitations
The framework connects continuous Hodge-theoretic structure to stable finite element discretizations and also applies to elasticity equations.
Takeaways & Limitations
On non-simply connected domains such as an annulus, the boundary value problem is not generally well-posed and must be solved modulo harmonic vector fields, with an additional uniqueness constraint.

Abstract

from arXiv · show

This article reports on the confluence of two streams of research, one emanating from the fields of numerical analysis and scientific computation, the other from topology and geometry. In it we consider the numerical discretization of partial differential equations that are related to differential complexes so that de Rham cohomology and Hodge theory are key tools for the continuous problem. After a brief introduction to finite element methods, the discretization methods we consider, we develop an abstract Hilbert space framework for analyzing stability and convergence. In this framework, the differential complex is represented by a complex of Hilbert spaces and stability is obtained by transferring Hodge theoretic structures from the continuous level to the discrete. We show stable discretization is achieved if the finite element spaces satisfy two hypotheses: they form a subcomplex and there exists a bounded cochain projection from the full complex to the subcomplex. Next, we consider the most canonical example of the abstract theory, in which the Hilbert complex is the de Rham complex of a domain in Euclidean space. We use the Koszul complex to construct two families of finite element differential forms, show that these can be arranged in subcomplexes of the de Rham complex in numerous ways, and for each construct a bounded cochain projection. The abstract theory therefore applies to give the stability and convergence of finite element approximations of the Hodge Laplacian. Other applications are considered as well, especially to the equations of elasticity. Background material is included to make the presentation self-contained for a variety of readers.

1. Introduction

The paper develops finite element exterior calculus to obtain stable discretizations for PDEs connected to differential complexes, combining finite element methods with de Rham cohomology and Hodge theory. It introduces an abstract Hilbert-space framework and applies it to de Rham and elasticity complexes.

Motivation: Finite element convergence requires both consistency and stability, with stability meaning uniform well-posedness of the discrete problem.Consistency alone does not ensure stable discrete operators.
Contribution: The paper develops finite element exterior calculus to preserve de Rham cohomology and Hodge-theoretic structures in discrete finite element spaces.The target problems are PDEs related to differential complexes, including linear elliptic equations relevant to broader mathematical models.
Abstract framework: An abstract Hilbert-complex framework analyzes stability and convergence using mixed formulations of Hodge-Laplacian problems.The framework represents differential operators as maps between Hilbert spaces and captures structures needed for harmonic forms and cohomology.
Stability conditions: Stable discretization is obtained when finite element spaces form a subcomplex and admit a bounded cochain projection from the full complex.These hypotheses transfer the relevant continuous structures to the discrete setting.
De Rham application: The de Rham application constructs finite element differential-form spaces through the Koszul complex and arranges them into multiple subcomplexes.For each polynomial degree r, the spaces provide 2^n−1 ways to form de Rham subcomplexes in n dimensions.
Applications: The framework unifies finite element methods across fluid mechanics, solid mechanics, electromagnetics, and other applications, with implementations in scientific and commercial software.The paper also obtains stable mixed finite elements with polynomial shape functions for elasticity.

2. Finite element discretizations

Finite element discretizations use Galerkin trial spaces and require analysis of consistency, stability, and approximation. For complex PDEs, mixed formulations and suitable discrete spaces address failures of straightforward methods and yield convergence when stability conditions hold.

Finite element methods: Finite element methods discretize PDEs by selecting finite-dimensional trial spaces, often piecewise polynomial spaces assembled from local shape functions and degrees of freedom.Locally supported bases make the coefficient matrix sparse and the resulting system efficient to solve.
Galerkin methods: Galerkin discretization replaces the continuous variational problem with a finite-dimensional algebraic system defined by the discrete space and bilinear and linear forms.The generalized formulation also permits discrete spaces that are not subspaces of the continuous space and additional approximations such as numerical integration.
Stability and convergence: Consistency measures whether the discrete problem approximates the continuous problem, while stability requires uniform well-posedness as the discretization parameter decreases.The stability constant is defined through the norm of the inverse discrete solution operator.
Stability and convergence: A consistent, stable generalized Galerkin method converges, with the discrete solution approaching the restricted continuous solution as h tends to zero.For standard Galerkin methods, the resulting approximation is quasioptimal up to a constant relative to the best approximation in the subspace.
Mixed formulations: Mixed finite element discretizations require compatible spaces because nonsingularity and stability depend on relationships between the discrete spaces.For the mixed Poisson formulation, the condition dim(div Σh) ≥ dim Vh is necessary for a particular nonsingularity implication.
Examples and limitations: Standard Galerkin methods can fail for vector Laplacian problems on reentrant-corner or multiply connected domains, whereas mixed methods can provide convergent or accurate approximations.On an annulus, the mixed solution is computed modulo harmonic fields and made unique by orthogonality to them; for the eigenvalue problem, Raviart–Thomas elements give a provably good approximation for any mesh.

3. Hilbert complexes and their approximation

The paper develops a Hilbert-complex framework that transfers Hodge-theoretic structure to finite element approximation and supports stability, convergence, and error analysis.

3. Hilbert complexes and their approximation: Hilbert complexes abstract essential features of the L2 de Rham complex for analyzing finite element exterior calculus.The framework targets approximation of geometrical quantities such as cohomology spaces and error estimates for the abstract Hodge Laplacian.
3. Hilbert complexes and their approximation: The framework is used for abstract Hodge Laplacian source and eigenvalue problems, with concrete de Rham subspaces treated subsequently.The later construction verifies the hypotheses required to apply the abstract results.
3. Hilbert complexes and their approximation: Hilbert complexes extend beyond the canonical L2 de Rham complex to variations for more general PDEs and boundary-value problems.The paper also discusses elasticity, which requires a different Hilbert complex containing a second-order differential operator.
3. Hilbert complexes and their approximation: The paper begins by recalling homological algebra, functional analysis, and notation needed for the framework.This background is intended to support the subsequent abstract development.

3.1. Basic definitions.

This section defines cochain and Hilbert complexes, their cohomological and Hodge-theoretic structures, and the analytic properties used to establish approximation results.

3.1. Basic definitions.: A cochain complex is a sequence of vector spaces and differentials whose consecutive composition is zero.It may equivalently be represented as a graded vector space with a degree +1 operator satisfying d ◦ d = 0.
3.1. Basic definitions.: Cocycles are null-space elements, coboundaries are range elements, and cohomology is the quotient of cocycles by coboundaries.Cochain maps preserve cocycles and coboundaries and therefore induce maps on cohomology.
3.1. Basic definitions.: A subcomplex consists of subspaces preserved by the differential, while a cochain projection maps the full complex onto it compatibly with the differential.The inclusion and projection induce maps on cohomology; under the stated condition, subcomplex cohomology has dimension at most that of the larger complex.
3.1. Basic definitions.: A Hilbert complex uses Hilbert spaces and closed, densely defined differentials satisfying the complex property; bounded complexes have bounded differentials.Its domain complex is a bounded Hilbert complex with the graph norm, and closedness or Fredholmness is preserved between the original and domain complexes.
3.1. Basic definitions.: Harmonic forms are cocycles orthogonal to coboundaries and represent reduced cohomology, while a closed complex admits a Hodge decomposition.The domain complex has the same null spaces, ranges, and harmonic forms as the original complex.
3.1. Basic definitions.: Closed range yields a Poincare inequality, and the condition that the range is closed is also necessary for that inequality.The Poincare inequality follows from bounded invertibility of the differential on the orthogonal complement of its null space.
3.1. Basic definitions.: The compactness property implies finite-dimensional harmonic spaces and yields the implication compactness property ⇒ Fredholm ⇒ closed.The compactness property is defined through compact inclusion of the joint domains of a differential and its adjoint.
3.1. Basic definitions.: The abstract Hodge Laplacian is L = dd∗ + d∗d, an unbounded operator associated with a Hilbert complex.The dual complex reverses the differential direction through densely defined adjoint operators.

3.2. The abstract Hodge Laplacian and the mixed formulation.

The abstract Hodge Laplacian is formulated as a mixed problem whose well-posedness follows from closedness and the Poincare inequality, then extended to discrete subcomplexes.

3.2. The abstract Hodge Laplacian and the mixed formulation.: The mixed formulation combines the differential, codifferential, and harmonic component to represent the abstract Hodge Laplacian.Its variational equations are Euler–Lagrange equations for a quadratic functional, and the critical point is a saddle point.
3.2.1. Interpretation of the mixed formulation.: Harmonic forms are exactly the homogeneous solutions, and nonzero data must be orthogonal to the harmonic space for solvability.The harmonic component is the orthogonal projection of f and the orthogonality condition fixes a particular solution.
3.2. The abstract Hodge Laplacian and the mixed formulation.: The mixed formulation is computationally useful because an efficient finite element approximation of V^k ∩ V∗^k may not be available.The alternative formulation also incorporates nonuniqueness caused by harmonic forms.
3.2. The abstract Hodge Laplacian and the mixed formulation.: Closedness is crucial because the Poincare inequality makes the mixed formulation well-posed.The theorem guarantees a unique solution for every f, with a bound controlled by the Poincare constant.
3.2.1. Interpretation of the mixed formulation.: The solution operator K maps f to the unique solution orthogonal to harmonic forms, with σ = d∗Kf, u = Kf, and p = PHf.The operator commutes with d and d∗ in the stated sense.
3.2.1. Interpretation of the mixed formulation.: The mixed variables decompose f into exact, harmonic, and coexact components, reproducing the Hodge decomposition.The corresponding restricted B∗ and B problems provide uniquely determined solutions of Lu = f under their stated data conditions.
3.2.2. Well-posedness of the mixed formulation.: An inf-sup condition for the mixed bilinear form provides the abstract well-posedness estimate.Theorem 3.2 supplies γ > 0 depending only on the Poincare constant, and Theorem 3.1 follows from this bound.
3.2.2. Well-posedness of the mixed formulation.: The result extends to general dual data, for which the solution correspondence is an isomorphism onto the dual space.The argument is related to Brezzi’s mixed finite element theorem, with additional symmetry and positivity assumptions in the stated special-case reduction.

3.3. Approximation of Hilbert complexes.

The section characterizes finite-dimensional Hilbert subcomplexes through bounded cochain projections and derives their cohomological, harmonic, and stability properties.

Approximation assumptions: Finite-dimensional spaces must form a subcomplex and admit a bounded cochain projection from the full complex.These are the central structural assumptions used throughout the approximation theory.
Discrete Hodge structures: A bounded cochain projection provides quasioptimal approximation and transfers key Hodge-theoretic structures to the discrete complex.The resulting discrete harmonic forms faithfully approximate the continuous harmonic forms.
Cohomology: Under the stated assumptions, the induced map on cohomology is an isomorphism.The proof uses the Hodge decomposition and an approximability condition for harmonic forms.
Cohomology: For de Rham finite element discretizations, the cohomology isomorphism holds for all mesh sizes, not only sufficiently small meshes.In general applications, the corresponding estimate guarantees the result only for sufficiently small h.
Stability: A bounded cochain projection yields a discrete Poincaré inequality whose constant depends only on the continuous constant and projection norm.This provides uniform control across families of subcomplexes when those quantities are uniformly bounded.
Stability: Conversely, suitable discrete Poincaré and harmonic-space bounds imply existence of a bounded cochain projection.The projection norm can be bounded in terms of the two assumed constants.

3.4. Stability and convergence of the mixed method.

The mixed finite element method is stable and convergent when its spaces form suitable Hilbert subcomplexes with uniformly bounded cochain projections.

Discrete mixed method: Galerkin discretizations of the Hodge Laplacian use finite-dimensional subspaces of the domain complex and require a subcomplex structure.A bounded cochain projection is the second main assumption.
Stability: Uniformly bounded cochain projections give an inf-sup stability constant depending only on the Poincaré constant and projection norms.The discrete problem therefore has a stability bound uniform in the mesh parameter under uniform projection bounds.
Convergence: The stable mixed discretization admits an error estimate controlled by approximation errors and the harmonic projection term.The estimate compares continuous and discrete variables through the bilinear form and projection operators.
Implementation: Discrete harmonic forms can be computed as the null space of a matrix.For finite element de Rham sequences, more direct computation may also be possible.
Convergence: When the spaces approximate the domain spaces, the discrete variables converge to the continuous solution as h → 0.This establishes convergence of the Galerkin method for the Hodge Laplacian.

3.5. Improved error estimates.

Additional compactness and W-boundedness assumptions yield improved Hodge-Laplacian error estimates, including optimal-order convergence under suitable regularity.

Rates: The basic V-norm estimate for the three variables is O(h^r).Approximation theory identifies this as the best possible rate for ∥u − uh∥_V in that setting.
Assumptions: Improved estimates require compactness of the Hilbert complex and cochain projections uniformly bounded in the W norm.The earlier stability theory requires only V-bounded projections.
Approximation rates: For piecewise polynomial spaces, the approximation quantities satisfy η = O(h), δ = O(h^min(2,r+1)), and µ = O(h^(r+1)).Here r is the largest degree of complete polynomials in the space.
Improved estimates: The improved estimate bounds ∥u − uh∥ using best-approximation errors for u, du, σ, dσ, p, and P_Bu.The terms are weighted by η, δ, and µ according to the regularity and decomposition components.
Optimal convergence: For convex domains with sufficiently smooth solutions, all components converge at the optimal order allowed by the polynomial degree.Elliptic regularity gives η = O(h), δ = O(h^2), and µ = O(h^2) in the stated example.
Whitney forms: For Whitney forms, the improved result gives an O(h) estimate instead of the earlier O(h|log h|) estimate.The passage attributes η and µ both order O(h) in this case.

3.6. The eigenvalue problem.

The eigenvalue analysis converts operator convergence into convergence of discrete Hodge-Laplacian eigenvalues and eigenspaces, while preserving the same subcomplex framework.

Convergence concept: Eigenvalue convergence is defined to approximate eigenvalues with multiplicity and eigenspaces while excluding spurious discrete eigenvalues and eigenvectors.The definition uses a gap between exact and discrete eigenspaces.
Operator criterion: Operator-norm convergence of K_hP_h to K is sufficient and necessary for the stated eigenvalue-approximation convergence.This links the abstract operator theory directly to spectral convergence.
Main theorem: Families of subcomplexes with uniformly W-bounded cochain projections yield convergence of the discrete eigenvalue problems.The result applies under the same structural framework used for source-problem stability and convergence, with W-boundedness required here.
Rates: For a simple eigenvalue with polynomial degree r, the eigenvalue error is O(h^(2r)), twice the source-problem rate.The stated rate assumes a convex domain and sufficient smoothness of the solution w = Ku.
Rates: For Whitney forms, the eigenvalue error is O(h^2), improving on the O(h|log h|) estimate.This is the corresponding example given after the general simple-eigenvalue theorem.
Problem decomposition: The full Hodge-Laplacian eigenproblem decomposes into B and B* problems whose eigenvalues correspond to those of the full problem.The associated eigenvectors are related through d*u_B = σ.

4. Exterior calculus and the de Rham complex

The de Rham complex supplies the differential-form setting for connecting exterior calculus, Hilbert-complex theory, and Hodge-theoretic analysis of PDEs on bounded Lipschitz domains.

Exterior calculus: Differential forms are alternating multilinear forms on tangent spaces, with wedge products, exterior derivatives, pullbacks, traces, and integration providing the basic calculus.The exterior derivative satisfies d(dω)=0, while pullbacks respect exterior products and derivatives.
Exterior calculus: Stokes’s theorem relates the integral of an exterior derivative to the boundary trace and yields integration-by-parts formulas for differential forms.The trace operator extends to Sobolev spaces, with boundary terms interpreted through appropriate pairings on Lipschitz boundaries.
Exterior calculus: The Hodge star maps k-forms to (n−k)-forms and transfers properties of the exterior derivative to the coderivative.On Riemannian manifolds, the Hodge star is an isometry at every point.
The de Rham complex: On bounded Euclidean domains with piecewise smooth Lipschitz boundaries, the L2 spaces and exterior derivative form a Hilbert complex.The exterior derivative is realized as a closed densely defined operator with domain HΛk.
The de Rham complex: The adjoint of the exterior derivative is the coderivative with the stated boundary-condition domain, and the de Rham complex satisfies compactness, Hodge decomposition, and Poincaré inequalities.These properties allow the abstract Hilbert-complex theory to apply to the de Rham setting.
The Hodge Laplacian: For k = 3, the associated problem is the Dirichlet problem for Poisson’s equation, which has a unique solution because there are no non-zero harmonic forms.The passage identifies the problem as having a unique solution.

5. Finite element approximation of the de Rham complex

Finite element exterior calculus constructs finite-dimensional differential-form spaces that preserve the de Rham complex structure and support stable approximation through bounded cochain projections.

Finite element subcomplexes: The finite element spaces must form a subcomplex, admit uniformly bounded cochain projections, and provide good approximation properties.These conditions allow mixed Galerkin error estimates to be applied to the finite element solution.
Polynomial differential forms: Polynomial differential-form spaces P_rΛ^k and P−_rΛ^k are used to build many polynomial subcomplexes of the de Rham complex.The two families are analyzed using the Koszul differential and Koszul complex.
The Koszul complex: The Koszul differential contracts with a radial vector field, increases polynomial degree, and decreases form degree, opposite to the exterior derivative.The identity (dκ + κd)ω = (r + k)ω underlies the exactness results.
Exact sequences: The homogeneous polynomial de Rham and Koszul complexes are exact except for the lowest-degree constant cohomology represented by R.The Koszul complex has vanishing cohomology in the stated non-exceptional degrees.
Polynomial decompositions: The Koszul identity also yields a direct-sum decomposition of homogeneous polynomial differential forms into κ- and d-generated components.For r + k > 0, the decomposition is established as a direct sum.
Polynomial subcomplexes: For each polynomial degree, the construction produces 2^n−1 complexes ordered by subcomplex inclusion, with all distinct when r ≥ n.Some complexes coincide for smaller degrees because later spaces vanish.

5.2. Degrees of freedom and finite element differential forms.

Finite element differential-form spaces are assembled from polynomial shape functions and geometrically indexed degrees of freedom, then organized into de Rham subcomplexes with canonical cochain projections.

Finite element construction: On a triangulation, the P_rΛ^k and P−_rΛ^k families are assembled from elementwise polynomial shape functions and degrees of freedom attached to subsimplices.The resulting spaces are subspaces of HΛ^k(Ω).
Degrees of freedom: Degrees of freedom use integrals of traced k-forms wedged with forms on subsimplices, and their single-valuedness determines interelement continuity.The geometric decomposition of the dual space, rather than a particular basis, determines the assembled space.
Degrees of freedom: The assembled degrees of freedom enforce exactly the continuity required for the finite element spaces to belong to HΛ^k(Ω).This property is established in the cited theorem on the assembly process.
De Rham subcomplexes: The two finite element families can be combined into 2^n−1 de Rham subcomplexes for each polynomial degree and mesh.These complexes are linearly ordered by inclusion and are distinct for r ≥ n.
Classical finite elements: The spaces include higher-order Whitney, Sullivan–Whitney, Lagrange, Raviart–Thomas, Brezzi–Douglas–Marini, and Nédélec finite elements under standard identifications.The identifications depend on form degree, dimension, and polynomial degree.
Canonical projection: The canonical projection defined by the degrees of freedom is a cochain map because it commutes with the exterior derivative, with commutation verified using Stokes’s theorem.The projection maps the smooth de Rham complex to the constructed finite element complexes.

5.3. Computational bases.

The computational realization uses dual-space decompositions and local bases, with explicit barycentric formulas available in important low-order cases and reference-element precomputation for higher orders.

Dual bases: The finite element assembly process provides bases for the dual spaces, indexed by subsimplices and their associated degrees of freedom.The dual-space construction supports local finite element bases.
Whitney bases: Whitney forms provide an especially simple dual basis, with one dual basis function for each k-dimensional subsimplex.For degree one, the associated function is given by integrating the traced form over the subsimplex.
Local bases: Dual-basis functions associated with a subsimplex produce finite element basis functions supported only on simplices containing that subsimplex.This gives a local basis suitable for computation.
Higher-order bases: For higher-degree spaces, the dual basis is computed by inverting a d × d matrix on one reference simplex and transferring it by affine transformation.The matrix dimension d equals the dimension of the shape-function space.
Explicit bases: Explicit barycentric-coordinate bases analogous to Bernstein bases are available for all spaces in the P and P− families, although the computational section is not essential to the rest of the paper.The tables display representative cases.
Three-dimensional examples: In three dimensions, Tables 5.1 and 5.2 summarize bases for P−_rΛ^1 and P_rΛ^2 at polynomial degrees r = 1, 2, and 3.The table discussion illustrates basis-function counts and barycentric-coordinate formulas for edges and faces.

5.4. Approximation properties.

The finite element spaces provide optimal-order approximation of differential forms, while canonical interpolation operators alone lack the boundedness and cochain properties needed by the abstract stability theory.

Approximation properties: The spaces PrΛk(Th) and P−rΛk(Th) provide optimal-order approximation of differential k-forms as the mesh size decreases.Optimal order is defined relative to polynomial degree and the smoothness of the approximated form.
Interpolation limitations: Canonical projections are not bounded on L2 or HΛk because they use traces onto lower-dimensional simplices.The Clément interpolant fixes boundedness but is neither a projection nor commuting with the exterior derivative.
Clément construction: The Clément interpolant is constructed by applying local L2 projections on element patches to determine the finite element degrees of freedom.For each degree of freedom, the relevant local polynomial projection is taken over the union of elements containing the associated simplex.
Clément interpolation: For shape-regular triangulations, the Clément interpolant is uniformly bounded in L2 independently of the element and mesh size.Its bounds rely on shape regularity, with constants possibly depending on polynomial degree and dimension.
Clément interpolation: The Clément interpolant preserves local polynomials and yields O(h^(r+1)) L2 approximation for sufficiently smooth k-forms.This rate is optimal because the spaces contain degree-r polynomials but not degree-(r+1) polynomials.

5.5. Bounded cochain projections.

The paper modifies canonical interpolation through smoothing and correction to construct uniformly bounded cochain projections that preserve projection and commuting properties while retaining approximation accuracy.

Construction: The smoothed projection construction combines canonical interpolation with a smoothing operator to recover bounded cochain projections.The construction builds on earlier work by Schöberl and Christiansen and is extended from quasi-uniform to general shape-regular meshes.
Construction: Smoothing by averaged pullbacks commutes with the exterior derivative, and a commuting extension operator produces the required smoothed operator.The smoothing maps L2 forms to smooth or continuous forms depending on the mesh assumptions.
Correction: The intermediate operators Q^ε_h are uniformly bounded in L2 and commute with the exterior derivative, but need correction to become projections.They converge to the identity in L2 operator norm as ε tends to zero, uniformly in h.
Result: For sufficiently small fixed ε, the corrected operators are uniformly bounded projections commuting with d.The resulting operators provide the stability structure required by the abstract Hilbert-complex theory.
Result: The corrected projections converge to the identity in L2 as h tends to zero.This convergence supplies the approximation statement for the finite element spaces.

5.6. Approximation of the de Rham complex and the Hodge Laplacian.

The constructed finite element spaces form discrete de Rham complexes with bounded cochain projections, allowing the abstract theory to establish cohomological consistency and Hodge Laplacian convergence.

Discrete de Rham complexes: The piecewise polynomial spaces form 2^n−1 distinct discrete de Rham complexes.The spaces include appropriate choices of polynomial differential forms arranged into subcomplexes.
Cohomology: Bounded projections produce commuting diagrams from the de Rham complex to each discrete complex and induce cohomology isomorphisms for sufficiently small h.The projections converge to the identity in the HΛk norm as the mesh is refined.
Cohomology: Whitney forms are isomorphic to simplicial cochains, so their cohomology is simplicial cohomology and agrees with de Rham cohomology on sufficiently fine triangulations.The same cohomological identification extends to the other discrete complexes.
Hodge Laplacian: The Galerkin method approximates the Hodge Laplacian using finite element spaces from the discrete de Rham complexes.The subcomplex and bounded cochain projection properties combine with the abstract error estimates and approximation estimates.
Hodge Laplacian: Convergence rates are limited by data smoothness, elliptic regularity, and the degree of complete polynomials in the finite element shape functions.With sufficient elliptic regularity, the error achieves the optimal order allowed by the subspace.

6. Variations of the de Rham complex

Variations of the de Rham Hilbert complex extend the framework to variable coefficients and essential boundary conditions, while preserving Hodge-theoretic structure and well-posed finite element formulations.

Variable coefficients: Equivalent coefficient-weighted inner products allow the same finite element de Rham subcomplexes to handle variable-coefficient Hodge Laplacians.The coefficient operator is bounded, symmetric, and positive definite.
Variable coefficients: In three dimensions, choosing a1=ε and a2=µ^-1 yields the standard Maxwell eigenvalue problem with dielectric tensor ε and magnetic permeability µ.These coefficients may be scalar or matrix-valued and may vary across the domain.
Boundary conditions: Using the domain ˚HΛk for the exterior derivative produces a de Rham complex with essential boundary conditions.Both boundary conditions for the resulting Hodge Laplacian are imposed through membership in the solution spaces.
Boundary conditions: Compact inclusion yields Hodge decomposition, the Poincaré inequality, well-posed mixed Hodge Laplacian formulations, and related results for the boundary-condition complex.The analogous compactness result follows by replacing k with n−k and applying the Hodge star operator.
Poincaré duality: The Hodge star maps harmonic (n−k)-forms isomorphically onto harmonic k-forms with boundary conditions, establishing Poincaré duality.The dimension of the boundary-condition harmonic space equals the corresponding Betti number.
Three-dimensional applications: In R3, the de Rham differentials become grad, curl, and div, yielding scalar and vector Laplacian boundary-value problems.The formulation includes Dirichlet, curl, and div boundary conditions together with orthogonality conditions for uniqueness.

7. The elasticity complex

The elasticity complex provides a differential-complex framework for linear elasticity that differs from the de Rham complex by including a second-order differential operator. Finite element spaces, pullbacks, smoothing, and bounded cochain projections yield stable and convergent mixed discretizations under a star-shaped-domain assumption.

The elasticity complex: The elasticity complex applies differential-complex methods to linear elasticity and includes a differential operator of second order.It is related to de Rham-type constructions through the Bernstein–Bernstein–Gelfand resolution.
The elasticity equations: Linear elasticity combines constitutive and equilibrium equations for stress and displacement fields under material and boundary assumptions.The compliance tensor is bounded, symmetric, and uniformly positive definite, with clamped displacement prescribed on the boundary.
Mixed formulation: The mixed formulation characterizes stress and displacement through the Hellinger–Reissner functional and a weak problem in H(div; Ω; S) × L2(Ω; V).The stress field is symmetric and has square-integrable divergence, while displacement is square-integrable.
Discrete complex: The elasticity discretization is tied to a subcomplex whose operators are the rotated Hessian J and divergence, with affine pullbacks commuting with both.The finite element construction uses commuting diagrams and modified interpolation operators to obtain bounded cochain projections.
Stability and convergence: For star-shaped domains and sufficiently small fixed smoothing parameter ϵ, bounded cochain projections establish well-posed Hodge Laplacians and stable discretizations.Using the stress spaces V 1_h, the construction gives a stable, convergent mixed discretization of the elasticity equations.