Jiang, Zhang and Zheng (Commun. Comput. Phys. 13:559-582, 2013) developed a parallel hp-adaptive Nédélec edge-element method with residual-based a posteriori error estimates for time-harmonic Maxwell's equations on conforming unstructured tetrahedral meshes, and closed their study by acknowledging the limitations that kept such methods from being fully automatic, robust, and scalable: efficiency constants of the estimator that degrade with the polynomial-degree distribution, the absence of automatic anisotropic hp refinement for edge singularities, direct-solver memory limits for variable-p hp systems (roughly 2×1062\times10^{6} unknowns with MUMPS), fragile numerical quadrature on highly graded meshes, and the untreated lossless resonant regime Im⁡α≡0\operatorname{Im}\alpha\equiv0. The polynomial-degree-robustness gap has since been closed by reconstruction-based equilibrated estimators (Chaumont-Frelet, Ern, and Vohralík, 2021-2026), while the surviving open core — automatic anisotropic hp refinement with guaranteed convergence for edge singularities on unstructured tetrahedra, scalable preconditioning robust in the local polynomial order and in anisotropy, robust quadrature on graded singular meshes, and rigorous hp adaptive analysis at resonances — remains open.

Background

Time-harmonic (frequency-domain) Maxwell equations describe electromagnetic fields oscillating at a fixed frequency. On a polyhedral domain with perfectly conducting boundary they reduce, for the electric field EE, to a second-order boundary value problem of the form ∇×(μ−1∇×E)+αE=f\nabla\times(\mu^{-1}\nabla\times E)+\alpha E=f with n×E=0n\times E=0 on the boundary, where the complex coefficient α\alpha collects frequency and material data. Absorption, Im⁡α>0\operatorname{Im}\alpha>0, makes the problem coercive and uniquely solvable; in the lossless case Im⁡α≡0\operatorname{Im}\alpha\equiv0 uniqueness fails at the interior Maxwell eigenfrequencies, the resonances. The natural variational space for EE is H(curl)\boldsymbol H(\mathrm{curl}), the space of square-integrable vector fields with square-integrable curl. Nédélec edge elements are the finite elements conforming in this space; their degrees of freedom are tied to mesh edges and enforce tangential field continuity, which is the correct physical continuity of the electric field. In hp finite element methods the local mesh size hh and the local polynomial degree pp both vary over the mesh. For solutions that are piecewise analytic with singularities at corners and edges, hp refinement can converge exponentially in the number of unknowns, in contrast to the algebraic rates of pure h-refinement — the main attraction for electromagnetic cavity and eddy-current problems on polyhedral domains. On such domains Maxwell fields typically behave like rλr^{\lambda} times analytic functions near vertices, and near re-entrant edges they vary slowly along the edge but strongly in the two transverse directions. Optimal exponential hp convergence then requires anisotropic (needle) elements — stretched along the edge, small and/or low-degree in the transverse directions. A priori anisotropic edge-element approximation theory and mesh constructions go back to Nicaise (2001) and to the 3D hp framework of Demkowicz, Kurtz, Pardo, Paszyński, Rachowicz, and Zdunek (2008), but they require knowing the singular directions in advance. Adaptive methods replace a priori knowledge by computable a posteriori error indicators: the residual-based estimator used in the source paper bounds the true error from above (reliability) and, up to constants, from below (efficiency); its efficiency constants, however, depend on the polynomial-degree distribution Ph\mathcal{P}_h and may therefore degrade as degrees grow. An estimator is called polynomial-degree-robust (p-robust) when both bounds hold with constants independent of p. Finally, MUMPS — a multifrontal massively parallel sparse direct solver based on LU factorization — was used in the source to solve the linear systems by direct factorization; its memory and time costs limited the implementation to on the order of 2×1062\times10^{6} unknowns, and scalable alternatives are iterative solvers accelerated by preconditioners, notably auxiliary-space multigrid methods that reduce H(curl)\boldsymbol H(\mathrm{curl}) problems to nodal (scalar) multigrid solves. The source paper — Jiang, Zhang, and Zheng (2013) — implemented a parallel hp-adaptive Nédélec method with residual-based estimation on conforming unstructured tetrahedral meshes, demonstrated its superiority over h-adaptivity on problems with point and edge singularities and on an engineering benchmark of the TEAM Problem 21 family (electromagnetic device benchmarks posed by the international Testing Electromagnetic Analysis Methods workshops), and closed with five acknowledged limitations that define the problems considered here.

Problem Statement

The problem, as posed in the concluding discussion of Jiang, Zhang, and Zheng (2013), is to complete the hp-adaptive conforming edge-element framework for time-harmonic Maxwell's equations on unstructured tetrahedral meshes so that it becomes fully automatic, provably reliable, and computationally scalable for singular and resonant three-dimensional problems. Of the five limitations the authors listed, one — efficiency of a posteriori estimation degrading with the polynomial-degree distribution Ph\mathcal{P}_h — has since been settled by reconstruction-based p-robust estimators and is now an available ingredient rather than a target. The surviving open core consists of four linked gaps, each stated for the source setting (conforming Nédélec hp discretizations on polyhedral domains): (1) Automatic anisotropic hp refinement for edge singularities: design an algorithm that, driven only by computed error indicators and without a priori knowledge of singular directions, performs anisotropic h- and p-refinement so that problems with edge singularities reach the near-exponentially convergent error decay that tailored a priori anisotropic (needle-element) refinements achieve, together with a reliability guarantee and a convergence or optimality analysis of the adaptive loop. (2) Scalable iterative solution of the variable-p hp H(curl)\boldsymbol H(\mathrm{curl}) systems thus produced: preconditioners or solver strategies whose convergence is demonstrably robust with respect to the local polynomial-degree distribution and to element anisotropy, and that scale in parallel beyond the roughly 2×1062\times10^{6}-unknown ceiling of the direct MUMPS-based approach. (3) Robust quadrature: numerical integration on highly graded, high-order hp meshes with singular (rλr^{\lambda}-type) integrands such that quadrature error no longer degrades the optimality of the adaptive solution, replacing the ad hoc regularization parameter used in the source. (4) Resonant regimes: a rigorous hp adaptive analysis — error estimation, convergence, and solver strategy — covering the lossless case Im⁡α≡0\operatorname{Im}\alpha\equiv0, including behavior near and at interior Maxwell eigenfrequencies where the boundary value problem loses uniqueness. These gaps jointly block a single objective: the solver (2) must handle the anisotropic variable-p matrices produced by (1), the quadrature (3) underpins the accuracy of both, and (4) extends the same framework to the parameter regime the source excluded. Completing the framework therefore requires addressing all four gaps (1)-(4) jointly: resolving only one of them, or any proper subset, does not resolve the task.

References

  1. Xue Jiang, Linbo Zhang, Weiying Zheng, Adaptive hp-Finite Element Computations for Time-Harmonic Maxwell's Equations, Communications in Computational Physics 13(2):559-582, 2013. DOI 10.4208/cicp.231111.090312a, https://doi.org/10.4208/cicp.231111.090312a
  2. Théophile Chaumont-Frelet, Asymptotically constant-free and polynomial-degree-robust a posteriori error estimates for time-harmonic Maxwell's equations, Numerische Mathematik 158(3):951-983, 2026. DOI 10.1007/s00211-025-01519-x (preprint arXiv:2402.17309), https://doi.org/10.1007/s00211-025-01519-x
  3. Théophile Chaumont-Frelet, Martin Vohralík, p-Robust Equilibrated Flux Reconstruction in H(curl) Based on Local Minimizations: Application to A Posteriori Analysis of the Curl-Curl Problem, SIAM Journal on Numerical Analysis 61(4):1783-1818, 2023. DOI 10.1137/21M141909X, https://doi.org/10.1137/21M141909X
  4. Théophile Chaumont-Frelet, Alexandre Ern, Martin Vohralík, Polynomial-degree-robust H(curl)-stability of discrete minimization in a tetrahedron, Comptes Rendus. Mathématique 358(9-10):1101-1110, 2021. DOI 10.5802/crmath.133, https://doi.org/10.5802/crmath.133
  5. Théophile Chaumont-Frelet, Patrick Vega, Frequency-Explicit A Posteriori Error Estimates for Finite Element Discretizations of Maxwell's Equations, SIAM Journal on Numerical Analysis 60(4):1774-1798, 2022. DOI 10.1137/21M1421805, https://doi.org/10.1137/21M1421805
  6. Ankit Chakraborty, Stefan Henneking, Leszek Demkowicz, An anisotropic hp-adaptation framework for ultraweak discontinuous Petrov-Galerkin formulations, Computers & Mathematics with Applications 167:315-327, 2024. DOI 10.1016/j.camwa.2024.05.025 (preprint arXiv:2309.00726), https://doi.org/10.1016/j.camwa.2024.05.025
  7. Vít Dolejší, Georg May, Anisotropic hp-Mesh Adaptation Methods: Theory, Implementation and Applications, Nečas Center Series, Springer, 2022. DOI 10.1007/978-3-031-04279-9, https://doi.org/10.1007/978-3-031-04279-9
  8. J. Kurtz, L. Demkowicz, A fully automatic hp-adaptivity for elliptic PDEs in three dimensions, Computer Methods in Applied Mechanics and Engineering 196(37-40):3534-3545, 2007. DOI 10.1016/j.cma.2006.10.053, https://doi.org/10.1016/j.cma.2006.10.053
  9. Leszek Demkowicz, Jason Kurtz, David Pardo, Maciej Paszyński, Waldemar Rachowicz, Adam Zdunek, Computing with hp-Adaptive Finite Elements, Volume 2: Frontiers: Three-Dimensional Elliptic and Maxwell Problems with Applications, Chapman & Hall/CRC, 2008. DOI 10.1201/9781420011692, https://doi.org/10.1201/9781420011692
  10. Serge Nicaise, Edge Elements on Anisotropic Meshes and Approximation of the Maxwell Equations, SIAM Journal on Numerical Analysis 39(3):784-816, 2001. DOI 10.1137/S003614290036988X, https://doi.org/10.1137/S003614290036988X
  11. Ralf Hiptmair, Jinchao Xu, Nodal Auxiliary Space Preconditioning in H(curl) and H(div) Spaces, SIAM Journal on Numerical Analysis 45(6):2483-2509, 2007. DOI 10.1137/060660588, https://doi.org/10.1137/060660588
  12. Tzanio V. Kolev, Panayot S. Vassilevski, Parallel Auxiliary Space AMG for H(curl) Problems, Journal of Computational Mathematics 27(5):604-623, 2009. DOI 10.4208/jcm.2009.27.5.013, https://doi.org/10.4208/jcm.2009.27.5.013
  13. A. Nicolopoulos, M. Campos-Pinto, B. Després, A stable formulation of resonant Maxwell's equations in cold plasma, Journal of Computational and Applied Mathematics 362:185-204, 2019. DOI 10.1016/j.cam.2019.05.007, https://doi.org/10.1016/j.cam.2019.05.007
  14. Yingying Xie, Ming Tang, A posteriori error estimate for a WG method for indefinite time-harmonic Maxwell equations, Journal of Computational and Applied Mathematics 481:117281, 2026. DOI 10.1016/j.cam.2025.117281, https://doi.org/10.1016/j.cam.2025.117281
  15. Pablo D. Brubeck, Patrick E. Farrell, Robert C. Kirby, Charles Parker, Fast solvers for the high-order FEM simplicial de Rham complex, Mathematics of Computation, 2026 (online first). DOI 10.1090/mcom/4252 (extended preprint arXiv:2506.17406), https://doi.org/10.1090/mcom/4252