Delayed Convergence of S-Parameter Sensitivities to Design Parameters in FDTD Simulations

Abstract

In the finite-difference time-domain (FDTD) method, the sensitivities of S-parameters to design parameters come from discrete Fourier transforms of responses ending at a simulation time T. We show that, when a design parameter moves a resonance, the sensitivity converges later than the S-parameter. For a single mode with field decay time τ, the truncation error of the sensitivity is (1+T/τ)e^−T/τ, against e^−T/τ for the resonant part of the S-parameter, so a 1% sensitivity error needs T ≥ 6.6τ instead of 4.6τ. Automatic differentiation, the discrete adjoint and finite differences on the same run carry this error, so their agreement cannot reveal it. In three solvers, with no fitted parameter, the estimate predicts the simulated error within 4% at cavity resonances with Q of 1000-7000; at lower Q the error is larger. On a cavity whose slowest mode is weakly coupled, common end criteria leave |S₁₁| within 0.5 dB of a 1.3 dB dip but its sensitivity 10-72% in error. A matrix-pencil extrapolation inside the differentiable computation recovers the sensitivity on a cavity with field probes from a run 66 times shorter. In a single-pole filter loop with a symmetric objective, truncation leaves the maximum of the objective in place, and the optimizer reaches the resonance of a full run despite the wrong gradient. With two coupled poles, the truncated objective moves the optimum: on runs of 1τ it over-couples by Δk ≈ 1.24/Q, and an equal-ripple objective still shifts it by about 0.2α̅ at 3τ, but not at 4.6τ.

Sensitivities of a simulated Arctic stratocumulus to aerosol and microphysical parameters
Hannah Carolin Frostenberg, Jessie Creamean, Erik Schurch Thomson +12
Atmospheric Chemistry and Physics4 Nov 2026Peer-reviewedCC BY 4.0Full textGeneral chemistry Atmospheric science & meteorologyCopernicusCrossrefDOAJ
Optimal Hamiltonian Parameter Estimation in the Presence of Nuisance Parameters
Zhiyao Hu, Haidong Yuan, Liang Jiang +1
arXiv11 Sep 2026PreprintarXiv licenceFull textGeneral mathematics Quantum physics & quantum informationarXiv
Accelerating Quantum Simulations of Materials Through Parameter and Ansatz Transfer Strategies
Saurabh Shivpuje, Vinit Singh, Manas Sajjan +1
arXiv27 Sep 2026PreprintarXiv licenceFull textPhysical chemistry Condensed matterarXiv
Hierarchical Parameter Estimation for Intrinsic Parameters of Extreme-mass-ratio Inspirals Using Deep Learning
Ping Shen, Wen-Biao Han, Minghui Du
arXiv14 Oct 2026PreprintarXiv licenceFull textGravitation & general relativityarXiv
Evaluation of upper-scale mechanical parameters of particulate aluminum-based nanocomposites from molecular simulations
Bartosz Bucholc, Marcin Maździarz, Szymon Nosewicz
arXiv26 Oct 2026PreprintCC BY 4.0Full textComputational physics Condensed matterarXiv

Cite

BibTeX RIS

BibTeX

@misc{kim2026delayed,
  title = {{Delayed Convergence of S-Parameter Sensitivities to Design Parameters in FDTD Simulations}},
  author = {Byungkwan Kim},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.28496},
  eprint = {2610.28496},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.28496},
}

RIS

TY  - UNPB
TI  - Delayed Convergence of S-Parameter Sensitivities to Design Parameters in FDTD Simulations
AU  - Byungkwan Kim
PY  - 2026
DA  - 2026/10/20
PB  - arXiv
DO  - 10.48550/arxiv.2610.28496
UR  - https://arxiv.org/abs/2610.28496
AB  - In the finite-difference time-domain (FDTD) method, the sensitivities of S-parameters to design parameters come from discrete Fourier transforms of responses ending at a simulation time T. We show that, when a design parameter moves a resonance, the sensitivity converges later than the S-parameter. For a single mode with field decay time τ, the truncation error of the sensitivity is (1+T/τ)e^−T/τ, against e^−T/τ for the resonant part of the S-parameter, so a 1% sensitivity error needs T ≥ 6.6τ instead of 4.6τ. Automatic differentiation, the discrete adjoint and finite differences on the same run carry this error, so their agreement cannot reveal it. In three solvers, with no fitted parameter, the estimate predicts the simulated error within 4% at cavity resonances with Q of 1000-7000; at lower Q the error is larger. On a cavity whose slowest mode is weakly coupled, common end criteria leave |S_11| within 0.5 dB of a 1.3 dB dip but its sensitivity 10-72% in error. A matrix-pencil extrapolation inside the differentiable computation recovers the sensitivity on a cavity with field probes from a run 66 times shorter. In a single-pole filter loop with a symmetric objective, truncation leaves the maximum of the objective in place, and the optimizer reaches the resonance of a full run despite the wrong gradient. With two coupled poles, the truncated objective moves the optimum: on runs of 1τ it over-couples by Δk ≈ 1.24/Q, and an equal-ripple objective still shifts it by about 0.2α̅ at 3τ, but not at 4.6τ.
ER  -

Abstract

Related

Cite