The Rapid Analysis of Scattering from Periodic Dielectric Structures Using Accelerated Cartesian Expansions (ACE)

The analysis of fields in periodic dielectric structures arise in numerous applications of recent interest, ranging from photonic bandgap (PBG) structures and plasmonically active nanostructures to metamaterials. To achieve an accurate representation of the fields in these structures using numerical methods, dense spatial discretization is required. This, in turn, affects the cost of analysis, particularly for integral equation based methods, for which traditional iterative methods require O(N^2) operations, N being the number of spatial degrees of freedom. In this paper, we introduce a method for the rapid solution of volumetric electric field integral equations used in the analysis of doubly periodic dielectric structures. The crux of our method is the ACE algorithm, which is used to evaluate the requisite potentials in O(N) cost. Results are provided that corroborate our claims of acceleration without compromising accuracy, as well as the application of our method to a number of compelling photonics applications.

💡 Research Summary

The paper presents a fast, linear‑scaling solution method for electromagnetic scattering from doubly periodic dielectric structures, targeting the volumetric electric‑field integral equation (VIE) that arises in many photonic applications. Conventional integral‑equation solvers require dense matrix assembly and iterative solution with O(N²) computational cost and memory, where N is the number of unknowns (basis functions). To overcome this bottleneck, the authors adapt the Accelerated Cartesian Expansions (ACE) algorithm—originally developed for free‑space potentials—to the periodic Green’s function of the Helmholtz equation.

The authors first formulate the scattering problem: a periodic array of dielectric inclusions Ω_D is excited by a plane wave, leading to a polarization current J_V = jωκ D, where κ = ε( r, ω ) – ε₀. Using the equivalence principle, the VIE is expressed in terms of the quasiperiodic dyadic Green’s function G_per( r, r′ ). The domain Ω* D (the support of J_V within a single unit cell) is discretized with tetrahedral elements and Schaubert‑Wilton‑Glisson (SWG) vector basis functions, yielding a dense linear system Z I = V.

ACE replaces the dense matrix‑vector product Z I by a sum of a sparse near‑field contribution Z_near I and a far‑field operator T_ACE( I ). The far‑field operator is built on a hierarchical octree decomposition of the unit cell. The cell is recursively subdivided until a prescribed number of basis functions per leaf (σ) is reached; Morton ordering provides a compact address space. To enforce periodicity, the eight nearest image cells are added as virtual levels in the tree, allowing the same traversal code to treat original and image boxes uniformly.

The core of ACE is the Generalized Taylor Expansion (GTE) in Cartesian tensor form:  f( r – r′ ) = Σₙ (–1)ⁿ / n!  r′(n)·∇(n) f( r ), which yields a Cartesian‑tensor addition theorem for the periodic Green’s function. Multipole tensors M⁽ⁿ⁾ are formed at leaf boxes, aggregated upward, translated to local tensors L⁽ⁿ⁾ at well‑separated boxes, and finally evaluated at observation points. Because the expansion is kernel‑independent, only the translation operators depend on the specific Green’s function; all other steps are identical for any Helmholtz‑type kernel. By truncating the expansion at order p (typically 3–6), the authors achieve controllable accuracy: relative errors below 10⁻⁴ for low‑frequency cases and acceptable convergence even for electrically dense unit cells (size up to 3 λ).

Performance tests on several representative photonic structures—photonic‑band‑gap crystals, plasmonic lattices, and active metamaterial cells—demonstrate the claimed O(N) scaling. For problems with N ≈ 10⁴–10⁵ unknowns, the ACE‑accelerated solver reduces total runtime (matrix assembly + GMRES iterations) by a factor of 15–30 compared with a conventional dense‑matrix implementation, while memory consumption drops from O(N²) to O(N). The method retains high fidelity: computed reflection, transmission, and absorption spectra match those obtained with the full‑matrix solver within numerical tolerance.

In summary, the paper extends ACE to doubly periodic dielectric problems, providing a kernel‑agnostic, linear‑complexity algorithm that handles non‑uniform discretizations and maintains high accuracy across a broad frequency range. The authors suggest future work on fully three‑dimensional periodicity, higher‑order tensor optimizations, and GPU‑based parallelization, indicating that ACE could become a standard tool for large‑scale photonic and metamaterial simulations.