Skip to content

Campaign: atom_fishbone_coupler/anthracene_cavity_q25k_v1

Assembled 2026-09-05 14:54 UTC.

Engine: atom_fishbone_coupler · path: engines/atom_fishbone_coupler/campaigns/anthracene_cavity_q25k_v1/

README

Anthracene cavity Q≥25k campaign v1

Status: concluded at the direct-Q observability gate; success criterion not met.

The inverse-design topology became nearly binary and both differentiable surrogates improved, but 16 ringdowns across four continuation stages produced zero certified resonances. Free-form boundary optimization was therefore not run. No Q≥25,000 result is claimed.

Frozen physical contract

  • target wavelength: 780.24 nm;
  • 300 nm patterned Si3N4 (n=2.0) with air as the topology low-index material;
  • SiO2 substrate below (n=1.45);
  • uniform 100 nm anthracene film (n=1.8) from z=0.15 to 0.25 µm;
  • Ey point dipole at (0, 0, 0.20) µm, inside anthracene;
  • PVA above anthracene (n=1.5);
  • no protected vacuum aperture;
  • 10 × 3 µm design region with 700 nm symmetric feedthrough guides; and
  • x/y field symmetry (1, -1, 0).

Final accounting

  • Optimization ledger: 9.7850988610 of 12 FC spent, 2.2149011390 FC unused, no outstanding reservation.
  • Final-evaluation ledger: not created; the 2.5 FC remained locked because no post-boundary frozen geometry exists.
  • Final retained topology: beta=64, emitter-field contrast objective 1.288020492e10 → 1.447100974e10, gray fraction 0.065667.
  • Final contour handoff: threshold 0.50, 84 contours, geometry digest 3361071c5cfef3f24983b40c4b3fe7d50baeabae5b56ecb7ce1be42a9b392b6d.
  • Final direct-Q result: zero certified modes; Q and wavelength are null.
  • Final field check: max |Ey|=742.9734; RMS after 2 ps 0.10962599; RMS after 10 ps 0.10962617, demonstrating a stationary numerical floor.

Why the campaign stopped

The boundary optimizer requires a trackable pole. Without one, QNM gradients, finite differences, and line searches would all optimize numerical-fit noise. All Q tasks completed normally on a 35 nm fixed grid with 6,800,640 full-domain cells before symmetry, so this is not a RAM blockade. The topology objectives rewarded selective scattering and prompt local fields, not demonstrable stored energy.

Retained evidence

  • campaign_outcome.json: concise machine-readable disposition.
  • optimization_credit_ledger.json: aggregate paid-task accounting.
  • topology_beta*/: topology checkpoints, promoted density, exact contour handoffs, and server task provenance.
  • threshold_q_scan*/: raw Tidy3D SimulationData, metadata, per-candidate multi-probe analysis, and batch summaries.
  • Public report and figures: https://www.christianmlange.com/cavitygrad/experiments/2026-08-11_anthracene_cavity_q25k/

The next campaign should establish a known resonant positive control in the exact stack, optimize a stored-energy/loss objective with a coupling constraint, and require an observable direct-Q pole before any boundary step.