Chapter Four · failure evidence
What Electromagnetic & Optical Simulation got wrong, from 35 dissertations
The records document various failures and computational bottlenecks encountered when modeling optical and electromagnetic systems across different regimes. Researchers frequently struggled with prohibitive computational costs in full-wave solvers, inaccurate wave approximations, and discrepancies between idealized simulations and fabricated devices. These records come from PhD theses at 13 institutions, 2021 to 2026. Each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Full-wave and fine-scale simulations become computationally intractable for large optical domains
Full 3D finite-difference time-domain methods and fine ray-tracing algorithms suffered from intractable runtimes and excessive memory requirements when scaled to large structures or long waveguides. Consequently, researchers faced computational bottlenecks or were forced to abandon direct 3D simulation in favor of approximate propagation alternatives.
Tried and failed
full-wave electromagnetic simulation applied to millimeter-scale optical metasurfaces. Outcome: infeasible cost. Reason: computational memory and time requirements scale intractably for large-area nanophotonic structures
Considered and rejected
Considered and rejected: Rejected full 3D simulation of planar waveguide polarizability extraction in favor of the TEM waveguide method due to large computational domain and lengthy runtimes (hours to days).
Advanced Metamaterials for Beamforming and Physical Layer Processing · DukeSpace
Considered and rejected
Considered and rejected: Simulating entire coupled resonator optical waveguides (CROWs) or large rings (R >= 20 um) directly in 3D FDTD, rejected due to prohibitive computational runtime (hours per simulation).
Statistical Modeling of the Effects of Process Variations on Silicon Photonics · MIT
Considered and rejected
Considered and rejected: Rejected full 3D FDTD simulations for long multimode waveguides due to prohibitive computational costs, using the split-step Fourier method instead.
Programmable Interactions between Optical Fields and Atom-like Systems in Integrated Circuits · MIT
Tried and failed
Monte Carlo ray tracing applied to 3D optical wave propagation. Outcome: infeasible cost. Reason: Prohibitively large sample sizes were required to resolve strong oscillations and reach acceptable accuracy in 3D.
A wave-kinetic numerical method for the propagation of optical waves · Virginia Tech
Tried and failed
Uniform Theory of Diffraction modeling applied to large-scale optical scattering simulation. Outcome: too slow. Reason: Excessive computational complexity in evaluating transition functions for large-scale structures compared to simpler asymptotic diffraction approximations.
Metrology of high aspect ratio nanoelectromechanical systems (NEMS) · Imperial
Tried and failed
exact Fermat ray tracing with circular reflections applied to optical path computation. Outcome: infeasible cost. Reason: chaotic scattering from curved reflective surfaces made exact computation computationally intractable
INFORMATION-DRIVEN MULTI-VIEW PATH PLANNING FOR UNDERWATER TARGET RECOGNITION · Cornell
Considered and rejected
Considered and rejected: Rejected FDTD for long waveguide simulation due to prohibitive computational inefficiency in paraxial propagation.
Development of hybrid optical simulation concepts for polymer waveguides · Leibniz Universität Hannover Repository
Simplified spatial and scattering approximations fail in complex waveguide and media environments
Dimensional reductions, diffusion assumptions, and Born scattering approximations produced large errors by ignoring boundary interactions, non-uniform scatterers, and vertical mode profiles. In certain waveguide formulations, these simplifying assumptions also led to unphysical reflection coefficients or violations of energy conservation.
Tried and failed
Analytical diffusion approximation with tissue homogenization applied to turbid media optical decorrelation modeling. Reason: Homogenization and diffusion assumptions produce up to 10x error in electric field decorrelation decay times
Computational optical imaging platform for volumetric large-field studying of cerebral hemodynamics · UT Austin
Tried and failed
dimensionally reduced 2.5D electromagnetic wave simulation applied to waveguides with boundary roughness perturbations. Reason: boundary perturbations violate vertical mode profile reduction assumptions, causing high error
Adjoint Methods and Inverse Modeling for Process Variation Analysis in Silicon Photonics · MIT
Tried and failed
first Born approximation for wave scattering applied to optical wave propagation modeling. Reason: fails to conserve energy around the beam axis for low refractive indices or plane waves
A wave-kinetic numerical method for the propagation of optical waves · Virginia Tech
Tried and failed
factored free-space scattering approximations applied to waveguide attenuation with heterogeneous scatterers. Outcome: did not generalise. Reason: neglects waveguide boundary interactions and non-uniform scatterer spatial distributions causing substantial prediction error
The effect of attenuation from fish on long-range active and passive acoustic sensing in the ocean · Woods Hole
Considered and rejected
Considered and rejected: Rejected standard travelling-wave formulations for MI waveguides because they predict non-physical reflection coefficients >1 for lossy and out-of-band passive junctions
Near field communication and localisation with slow waves · Imperial
Considered and rejected
Considered and rejected: Direct analytical modeling of DMA radiated fields rejected due to element coupling and waveguide interactions, requiring empirical near-field scans instead
Coherence in Dynamic Metasurface Aperture Microwave Imaging Systems · DukeSpace
Idealized electromagnetic models fail to match fabricated devices due to physical non-idealities
Theoretical and numerical simulations significantly overestimated modulation depth and transmission because they relied on pristine CAD geometries and nominal material parameters. Real physical implementations suffered from nanoscale fabrication defects, substrate gating inefficiencies, and impedance mismatches that broke theoretical model assumptions.
Tried and failed
idealized geometric electromagnetic simulations applied to fabricated nanophotonic transmission spectra. Outcome: did not generalise. Reason: nanoscale fabrication imperfections caused real optical behavior to diverge significantly from idealized CAD geometries
Nonlinear optics in nanoantennas · Imperial
Tried and failed
transfer matrix modeling of optical reflectance applied to suspended layered thin film heterostructures. Reason: uncertainties in dielectric functions at high bias voltages caused discrepancies between predicted and measured reflection
Quantum Optics with Excitons in Atomically Thin Semiconductors · Harvard
Tried and failed
physics-calibrated time-symmetric control pulses applied to waveguide state transfer efficiency. Outcome: worse than baseline. Reason: signal line distortions and impedance mismatches broke theoretical symmetry assumptions
Quantum Networking using Waveguide Quantum Electrodynamics · MIT
Lost to a baseline
Experimental bP waveguide modulation depth (8 dB/mm) fell short of theoretical/simulation predictions (>40 to 61 dB/mm).
Engineering Novel Semiconductors for Advanced Infrared Sensors · ResearchWorks
Lost to a baseline
Experimental integrated black phosphorus waveguide modulator achieved 8 dB/mm, falling short of the 40 dB/mm predicted by numerical simulations due to non-ideal planarization and substrate gate inefficiency.
Electromechanically control the optoelectronic properties of 2D semiconductors · ResearchWorks
Modal and energy transfer models break down in complex coupling regimes
Coupled mode theories and standard transfer approximations failed to capture complex effects like plasmon hybridization, multi-path parasitic pulses, and ultra-strong coupling corrections. Additionally, dipole and variational formulations mischaracterized spectral dissipation and either over-predicted or under-predicted energy transfer efficiency.
Tried and failed
dipole-dipole and planar FRET energy transfer models applied to deformable polymer nanoparticle optical sensors. Reason: Dipole-dipole models over-predicted energy transfer efficiency during matrix deswelling, while planar multi-fluorophore models systematically under-predicted it.
Considered and rejected
Considered and rejected: Rejected using a single gold nanorod geometry in DDA simulations because a T-shaped ligament model better captured localized surface plasmon hybridization and medium dielectric sensitivity.
POLYMER INFILTRATED NANOPOROUS GOLD: KINETICS AND OPTICAL PROPERTIES OF NOVEL POLYMER NANOCOMPOSITES · Penn
Lost to a baseline
Variational FHIP optical conductivity spectra produce unbroadened multi-phonon Franck-Condon peaks, failing to match the smearing/dissipation captured by Diagrammatic Monte Carlo (DiagMC).
Path integral methods for polarons in real materials · Imperial
Tried and failed
classical coupled mode theory applied to ultrastrongly coupled optical resonators. Reason: rotating wave approximation fails in the ultra-strong coupling regime without frequency-dependent off-diagonal corrections
Tried and failed
Iterative phase retrieval and gradient optimization applied to multi-mode waveguide pulse reconstruction. Outcome: did not generalise. Reason: Higher-order mode reconstruction could not account for parasitic multi-path side pulses
Terahertz ultrafast optics for bandstructure engineering · UT Austin
Ray tracing and geometric optics omit essential phase, interference, and tensor coupling effects
Geometric optical methods and simplified sensitivity models failed when simulating wave environments that depend heavily on phase accumulation, interference, or cross-field coupling. Neglecting these wave and local field factors led to distorted signal escape, erroneous phase tracking, and overly optimistic reach predictions.
Tried and failed
spectroscopic modeling neglecting thin-film interference effects applied to nonlinear optical response of thin interfaces. Reason: optical interference and local field factors distort signal escape across polarization geometries
Development and application of electronic sum-frequency generation to molecular solid interfaces · UT Austin
Lost to a baseline
The published LZ collaboration S2-only toy sensitivity model gave an overly optimistic reach compared to the thesis's more realistic simulation accounting for electron trains and optical propagation
Tried and failed
geometric ray tracing in turbulent media applied to wave scattering across plasma filaments. Reason: neglects leading-order cross-field coupling term linear in off-diagonal dielectric tensor elements
Impact of wave-filament scattering in the scrape-off layer during lower hybrid current drive · MIT
Tried and failed
ray-tracing software for relative phase difference simulation applied to optical interferometry and wavefront analysis. Reason: optical path difference phase accumulation was calculated incorrectly, requiring the term to be manually zeroed
Lifetime of beam driven wakes at the FACET-II facility · UT Austin
Left open by the authors
Problems the authors named and did not get to.
Left open
Perform FDTD simulations of the optical probe system to model phase accumulation and resolve Code V optical path difference discrepancies. Blocker: None
Lifetime of beam driven wakes at the FACET-II facility · UT Austin
Left open
Refine the fiber Bragg grating spectral analysis model by implementing anisotropic dielectric perturbation tensors and modal deformation calculations. Blocker: None
Investigation of Bragg Gratings in Few-Mode Fibers with a Femtosecond Laser Point-by-Point Technique · Virginia Tech
Left open
Simulate vertical coupling in stacked nanodisks separated by dielectric films to model an anapole nanotower using FDTD and multipole decomposition. Blocker: None
Strong coupling in nanoplasmonic systems, and anapoles · Imperial
Left open
Simulate coupled Maxwell-heat equations to determine transverse mode instability power thresholds in scaled short optical fibers. Blocker: Requires access to the thesis author's specialized parallel hp-adaptive DPG finite element codebase.
A scalable hp-adaptive finite element software with applications in fiber optics · UT Austin
Left open
Implement time-domain adjoint topology optimization using FDTD to design photonic crystal waveguides for pulse shaping, pulse delay, and particle acceleration. Blocker: None
Topology Optimization for Dispersive (3D) Meta‑Optics via the Time‑Domain Adjoint Method using FDTD and High‑Performance Computing · Leibniz Universität Hannover Repository
Left open
Model the full diffraction orders and reflectance bandwidth for 2D self-assembled periodic nanocolumns using standard optical simulation tools. Blocker: None
Magnetically responsive dynamic nanostructures · UT Austin
Left open
Extend the CCPR-ADE time-domain adjoint dissipation framework and analytical gradient evaluation to nonlinear optical media in FDTD. Blocker: None
Topology Optimization for Dispersive (3D) Meta‑Optics via the Time‑Domain Adjoint Method using FDTD and High‑Performance Computing · Leibniz Universität Hannover Repository
Left open
Extend the HEDA optical extinction model to 2D or 3D anisotropic plasmonic nanocrystal geometries. Blocker: Generalizing Mie theory/Maxwell-Garnett core-shell approximation to multi-dimensional anisotropic geometries requires developing or integrating numerical electrodynamics solvers without explicit formulation provided.
Nonuniform charge carrier distribution within plasmonic semiconductor nanocrystals · UT Austin
Left open
Develop a deep learning FDTR model taking optical parameters, sample layering geometry, and phase lag data simultaneously as inputs to predict thermal properties. Blocker: None
Investigation of memristor devices and materials through multi-scale modeling and metrology · Georgia Tech
Left open
Develop theoretical formalism and numerical simulation for multi-tonal Floquet-Fresnel optical reflectivity accounting for finite mode lifetimes. Blocker: None
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.