Bounded commuting cochain projections in finite element exterior calculus: unconditional, locally $L^2$-stable constructions in arbitrary dimension

The 2006 Acta Numerica survey that founded finite element exterior calculus (FEEC) built its stability theory for the Hodge Laplacian on smoothed commuting projections (extension, mollification, canonical projection, correction), while acknowledging that these required quasi-uniform meshes, that more general triangulations seem to need position-dependent smoothing, and that the construction was 'very possibly not the last word'. Quasi-uniformity was removed by globally defined smoothed projections (Christiansen-Winther 2008); locality was achieved in the graph norm by Falk-Winther (2014); Arnold-Guzmán (2021) obtained local $L^2$-bounded commuting projections in arbitrary dimension, but with the bound conditional on an unproved assumption of uniform regular-potential estimates on star patches; Ern-Guzmán-Potu-Vohralík (2025) established local $L^2$-stability unconditionally, but only in three dimensions and in vector-calculus form (its authors assert, without writing it out, that the results extend to arbitrary dimension in the FEEC framework), with trace-preserving refinements following in 2026. The surviving open core, formulated here, is an unconditional construction that is simultaneously commuting, patch-local, uniformly $L^2$-bounded on arbitrary shape-regular mesh families, boundary-condition preserving, and valid for both FEEC polynomial families, all form degrees, and arbitrary space dimension - or a rigorous obstruction showing that such a construction cannot exist.

Background

Finite element exterior calculus (FEEC), developed by Arnold, Falk, and Winther in the 2006 Acta Numerica survey (see also their 2010 Bulletin of the AMS survey and Arnold's 2018 book), is a framework for designing and analyzing finite element discretizations of PDEs whose well-posedness is governed by a differential complex. The model problem is the Hodge Laplacian on differential $k$-forms: a differential $k$-form is an antisymmetric tensor field that can be integrated over oriented $k$-dimensional surfaces, the language unifying the scalar Laplacian ($k=0$), curl-curl Maxwell problems ($k=1$), and mixed $H(\mathrm{div})$ formulations ($k=n-1$). The $L^2$ de Rham complex $L^2\Lambda^0(\Omega) \xrightarrow{d} L^2\Lambda^1(\Omega) \xrightarrow{d} \cdots \xrightarrow{d} L^2\Lambda^n(\Omega) \to 0$, whose arrows are the exterior derivative $d$ (gradient, curl, and divergence in vector-calculus disguise), encodes the compatibility structure of these problems. FEEC discretizes this complex by subcomplexes of piecewise polynomial differential forms on a simplicial mesh $\mathcal{T}_h$: two nested families of spaces, the full polynomial spaces $\mathcal{P}_r\Lambda^k(\mathcal{T}_h)$ and the trimmed (first-kind) spaces $\mathcal{P}_r^-\Lambda^k(\mathcal{T}_h)$, whose low-order instances are the Whitney forms, the Raviart-Thomas elements, and the Nédélec elements - the classical low-degree conforming element families for the Sobolev spaces $H(\mathrm{curl})$ and $H(\mathrm{div})$ of square-integrable vector fields with square-integrable curl or divergence, which govern Maxwell and mixed Poisson formulations.

The central analytic device is the commuting (cochain) projection: a uniformly bounded projection $\pi_h$ from the continuous complex onto the discrete subcomplex satisfying $d\pi_h = \pi_h d$. Its existence yields stability and quasi-optimal error estimates for mixed Galerkin formulations of the Hodge Laplacian; $L^2$-boundedness in particular upgrades error estimates to best-approximation quality and underwrites convergence of eigenvalue approximations (Theorems 3.11 and 3.19 of the 2010 Bulletin survey). The canonical interpolation operators defined by the finite element degrees of freedom do commute with $d$, but they act only on forms with substantially more regularity than elliptic theory supplies, so they cannot play this role. The 2006 survey therefore constructed smoothed projections: extend the form $\omega$ to a neighborhood of the domain, mollify at a scale proportional to the mesh size $h$, apply the canonical projection, and correct so the result lands in the finite element space. This requires the mesh family to be quasi-uniform (all element diameters comparable, condition (5.3) of the survey) - the survey notes this 'is too restrictive for many applications' and that more general triangulations seem to demand a position-dependent ($x$-dependent) smoothing radius - and the resulting operator is global rather than local. The survey states explicitly that the construction of bounded cochain projections is an active research subject and that its approach is 'very possibly not the last word' in the matter.

Three refinements define the modern form of the question. First, locality, in the sense of Clément-type interpolation: the projected form on an element $T$ should depend only on the input restricted to a bounded patch of elements around $T$, which is what makes such projections usable in local a posteriori and $hp$ analysis. Second, uniform local $L^2$-stability on shape-regular families: writing $S(T)$ for the patch of $T$, the bound $\|\pi_h\omega\|_{L^2\Lambda^k(T)} \le C\|\omega\|_{L^2\Lambda^k(S(T))}$ with $C$ independent of the mesh (no quasi-uniformity; shape regularity means a bounded ratio of element diameter to inscribed-ball diameter) and ideally of the polynomial degree. Third, unconditional proofs: the arbitrary-dimension construction of Arnold-Guzmán (2021) achieves locality and $L^2$-boundedness but derives its crucial weight-function bounds from regularized Poincaré (right-inverse) estimates of Costabel-McIntosh type on extended star patches $\mathrm{es}(\sigma)$ - the union of mesh simplices meeting a given simplex $\sigma$ - whose constants are unproved in general, because such patches need not be star-shaped with respect to a ball (their Assumption 2.4). Because this bound is the only conditional ingredient of the 2021 argument, proving the uniform regular-potential bounds would render properties 1-3 unconditional in arbitrary dimension; the 2021 paper, however, does not address boundary-condition preservation (property 4), and its extension to the full polynomial family $\mathcal{P}_r\Lambda^k$ is asserted rather than carried out, so closing the bound alone would not settle the full problem. The 2025 construction of Ern-Guzmán-Potu-Vohralík closes this gap, but only in three dimensions and in vector-calculus language (Lagrange, Nédélec, Raviart-Thomas, and discontinuous elements). What a complete theory still lacks is the combination: arbitrary space dimension, both FEEC polynomial families, all form degrees, and unconditional local $L^2$-stability with boundary-condition preservation.

Problem Statement

Construct commuting projection operators $\pi_h^k : L^2\Lambda^k(\Omega) \to V_h^k$ onto the FEEC finite element spaces - for both choices $V_h^k = \mathcal{P}_r\Lambda^k(\mathcal{T}_h)$ (polynomial degree $r \ge 0$) and $V_h^k = \mathcal{P}_r^-\Lambda^k(\mathcal{T}_h)$ (polynomial degree $r \ge 1$, the trimmed space $\mathcal{P}_0^-\Lambda^k$ being trivial for $k < n$), for every form degree $0 \le k \le n$ - where $\mathcal{T}_h$ runs over arbitrary shape-regular simplicial mesh families of a bounded Lipschitz domain $\Omega \subset \mathbb{R}^n$, in arbitrary space dimension $n$, whose extended stars $\mathrm{es}(\sigma)$ - the union of the mesh simplices meeting a given mesh simplex $\sigma$ - are contractible, the standing assumption under which all the constructions cited here operate and which is needed for exactness of the local polynomial de Rham sequences, such that all of the following are proved unconditionally, i.e. without unproved assumptions:

1. (Commutation) For $0 \le k \le n-1$, $d^k \pi_h^k u = \pi_h^{k+1} d^k u$ for every $u$ in the graph space $H\Lambda^k(\Omega) = \{u \in L^2\Lambda^k(\Omega) : d^k u \in L^2\Lambda^{k+1}(\Omega)\}$, so the projections form a cochain map onto the discrete de Rham subcomplex. (The identity is imposed on $H\Lambda^k(\Omega)$, as in the cited literature, because $d^k$ is unbounded on $L^2\Lambda^k(\Omega)$; the projections themselves are defined on all of $L^2\Lambda^k(\Omega)$.)
2. (Locality) For each element $T \in \mathcal{T}_h$, the restriction $(\pi_h^k u)|_T$ depends only on $u$ restricted to a bounded patch of mesh elements around $T$ (Clément-type locality, with the patch bound independent of the particular mesh in the family).
3. (Uniform local $L^2$-stability) For every element $T$ with associated patch $S(T)$, $\|\pi_h^k u\|_{L^2\Lambda^k(T)} \le C\, \|u\|_{L^2\Lambda^k(S(T))},$ with $C$ depending only on the shape-regularity constant, $n$, $k$, and (unless polynomial-degree robustness is additionally claimed) $r$ - in particular with no quasi-uniformity assumption on the mesh family.
4. (Boundary conditions) For inputs on which the homogeneous essential boundary conditions are defined, $\pi_h^k$ preserves them on any prescribed union of boundary faces.
5. (Uniform coverage) The same construction applies across the whole de Rham complex in one framework, as the 2006 survey's smoothed projections do.

Two partial results delimit the remaining gap: Arnold-Guzmán (2021) achieves properties 1-3 in arbitrary dimension but with the $L^2$ bound conditional on an unproved assumption of uniform regularized Poincaré (regular-potential) bounds on extended star patches; Ern-Guzmán-Potu-Vohralík (2025) establish the corresponding properties unconditionally but only in three space dimensions and in vector-calculus form - the authors state that their results could be proved in any dimension in the FEEC framework, but this extension is asserted rather than written, so the full differential-form construction in arbitrary dimension remains absent from the literature. The question is resolved either by such a construction or by a rigorous obstruction showing that the stated combination of properties cannot be achieved in the stated generality.

References

1. D. N. Arnold, R. S. Falk, and R. Winther (2006), 'Finite element exterior calculus, homological techniques, and applications', Acta Numerica 15, 1-155. DOI: 10.1017/S0962492906210018. https://doi.org/10.1017/S0962492906210018
2. D. N. Arnold, R. S. Falk, and R. Winther (2010), 'Finite element exterior calculus: from Hodge theory to numerical stability', Bulletin of the American Mathematical Society 47, 281-354. DOI: 10.1090/S0273-0979-10-01278-4. https://doi.org/10.1090/S0273-0979-10-01278-4
3. D. N. Arnold (2018), 'Finite Element Exterior Calculus', CBMS-NSF Regional Conference Series in Applied Mathematics 93, SIAM. DOI: 10.1137/1.9781611975543. https://doi.org/10.1137/1.9781611975543
4. S. H. Christiansen and R. Winther (2008), 'Smoothed projections in finite element exterior calculus', Mathematics of Computation 77, 813-829. DOI: 10.1090/S0025-5718-07-02081-9. https://doi.org/10.1090/S0025-5718-07-02081-9
5. R. S. Falk and R. Winther (2014), 'Local bounded cochain projections', Mathematics of Computation 83, 2631-2656. DOI: 10.1090/S0025-5718-2014-02827-5. https://doi.org/10.1090/S0025-5718-2014-02827-5 (arXiv:1211.5893)
6. R. S. Falk and R. Winther (2015), 'Double complexes and local cochain projections', Numerical Methods for Partial Differential Equations 31, 541-551. DOI: 10.1002/num.21922. https://doi.org/10.1002/num.21922
7. M. W. Licht (2019), 'Smoothed projections and mixed boundary conditions', Mathematics of Computation 88, 607-635. DOI: 10.1090/mcom/3330. https://doi.org/10.1090/mcom/3330 (arXiv:1710.06868)
8. A. Ern, T. Gudi, I. Smears, and M. Vohralík (2022), 'Equivalence of local- and global-best approximations, a simple stable local commuting projector, and optimal hp approximation estimates in $H(\mathrm{div})$', IMA Journal of Numerical Analysis 42, 1023-1049. DOI: 10.1093/imanum/draa103. https://doi.org/10.1093/imanum/draa103
9. T. Chaumont-Frelet and M. Vohralík (2024), 'A stable local commuting projector and optimal hp approximation estimates in $H(\mathrm{curl})$', Numerische Mathematik 156, 2293-2342. DOI: 10.1007/s00211-024-01431-w. https://doi.org/10.1007/s00211-024-01431-w
10. D. N. Arnold and J. Guzmán (2021), 'Local $L^2$-bounded commuting projections in FEEC', ESAIM: Mathematical Modelling and Numerical Analysis 55, 2169-2184. DOI: 10.1051/m2an/2021054. https://doi.org/10.1051/m2an/2021054 (arXiv:2104.00184)
11. M. Costabel and A. McIntosh (2010), 'On Bogovskiĭ and regularized Poincaré integral operators for de Rham complexes on Lipschitz domains', Mathematische Zeitschrift 265, 297-320. DOI: 10.1007/s00209-009-0517-8. https://doi.org/10.1007/s00209-009-0517-8
12. A. Ern, J. Guzmán, P. Potu, and M. Vohralík (2025), 'Local $L^2$-bounded commuting projections using discrete local problems on Alfeld splits', preprint, arXiv:2502.03601. https://arxiv.org/abs/2502.03601
13. A. Ern, J. Guzmán, and P. Potu (2026), 'Bounded, Commuting, Discrete-trace Preserving Projections', preprint, arXiv:2604.28103. https://arxiv.org/abs/2604.28103
14. A. Ern, J. Guzmán, P. Potu, and M. Vohralík (2024), 'Discrete Poincaré inequalities: a review on proofs, equivalent formulations, and behavior of constants', preprint (accepted, IMA Journal of Numerical Analysis), arXiv:2412.11796. https://arxiv.org/abs/2412.11796
15. N. Nigam and J. Phillips (2012), 'High-order conforming finite elements on pyramids', IMA Journal of Numerical Analysis 32, 448-483. DOI: 10.1093/imanum/drr015. https://doi.org/10.1093/imanum/drr015
16. F. Fuentes, B. Keith, L. Demkowicz, and S. Nagaraj (2015), 'Orientation embedded high order shape functions for the exact sequence elements of all shapes', preprint, arXiv:1504.03025. https://arxiv.org/abs/1504.03025
17. B. Cockburn and G. Fu (2017), 'A systematic construction of finite element commuting exact sequences', SIAM Journal on Numerical Analysis 55, 1650-1688. DOI: 10.1137/16M1073352. https://doi.org/10.1137/16M1073352 (arXiv:1605.00132)
18. M. Ainsworth and G. Fu (2017), 'A lowest-order composite finite element exact sequence on pyramids', Computer Methods in Applied Mechanics and Engineering 324, 110-127. DOI: 10.1016/j.cma.2017.05.030. https://doi.org/10.1016/j.cma.2017.05.030 (arXiv:1705.00064)
19. F. Bonaldi, D. A. Di Pietro, J. Droniou, and K. Hu (2025), 'An exterior calculus framework for polytopal methods', Journal of the European Mathematical Society. DOI: 10.4171/jems/1602. https://doi.org/10.4171/jems/1602
20. J. Hu and S. Zhang (2016), 'Finite element approximations of symmetric tensors on simplicial grids in $\mathbb{R}^n$: the lower order case', Mathematical Models and Methods in Applied Sciences 26, 1649-1669. DOI: 10.1142/S0218202516500408. https://doi.org/10.1142/S0218202516500408
21. S. H. Christiansen, J. Gopalakrishnan, J. Guzmán, and K. Hu (2023), 'A discrete elasticity complex on three-dimensional Alfeld splits', Numerische Mathematik 156, 159-204. DOI: 10.1007/s00211-023-01381-9. https://doi.org/10.1007/s00211-023-01381-9
22. P. Lederer and R. Stenberg (2024), 'Analysis of weakly symmetric mixed finite elements for elasticity', Mathematics of Computation 93, 523-550. DOI: 10.1090/mcom/3865. https://doi.org/10.1090/mcom/3865
23. P. D. Brubeck and P. E. Farrell (2022), 'A scalable and robust vertex-star relaxation for high-order FEM', SIAM Journal on Scientific Computing 44, A2991-A3017. DOI: 10.1137/21m1444187. https://doi.org/10.1137/21m1444187
24. R. Dabetić (2026), 'Multigrid Preconditioning for FEEC using Mass-Lumping and Transforming Smoothers', preprint, arXiv:2606.17934. https://arxiv.org/abs/2606.17934