Optimization efforts

This section is organized by scientific effort, not by timestamped run directory. Each guide states the objective, current trust level, decisive evidence, and the dashboards that belong to it.

Effort	Role	Current state
Device23 full-boundary Q	Current focus	Physical seed improved; corrected surrogate ascent reached Q 2286.52 and stopped at a boundary-transfer floor.
Atom-accessible fishbone	Target-device concept	Q≈2283 diagnostic with a 0.5–1.0 µm open corridor; atom-referenced mode volume remains too large.
Design23 parameterized BFGS	Historical benchmark	Strong evidence that long surrogate ascent can transfer to FDTD despite absolute disagreement.
Quan–Lončar analytical optimizer	Mathematical lineage	Original open-cavity series engine and autonomous optimization experiments.
General 1-D cavity pole	Active campaign	Scatterer-generic outgoing pole; reproduce three papers, ≥2× Q, Tidy3D ≤15 credits.
Inverse-design prototypes	Target-device pipeline	Tidy3D topology creation, binarization, and boundary-handoff experiments for Si₃N₄/anthracene cavities.

The run hub defaults to the small set of runs needed to interpret present progress. Enable Show full archive for invalidated, superseded, empty, and secondary runs.

Shared optimization contract

Every promoted geometry should carry four distinct records:

1. the exact geometry and material stack;
2. the tracked field and complex pole;
3. the local objective/gradient evidence;
4. independent physical verification under a fixed protocol.

Analytical, FEM-surrogate, and FDTD quantities are labeled separately. A clean surrogate ascent is useful, but it is not relabeled as physical improvement until an independent solver measures it.