Figure inventory¶
All current figures are regenerated by analysis/build_figures.py and audited
through results/figure-manifest.json.
| Figure | Role | Source record |
|---|---|---|
| Method overview | Main method/data-flow schematic | derivation guides |
| Quan geometry | Matched-axis frozen endpoints | quan_baseline.json, quan_optimized_phase58.json |
| Geometry profiles | Per-hole aspect ratio and 2r_y/W, both devices | results/geometry-metrics.json |
| Design23 geometry | Matched-axis exact step 0/105 endpoints | retained 128-entry geometry vectors |
| Design23 trajectory panels | Q, V_eff, λ vs step; protocol change and excluded fit marked | fdtd_series.json, surrogate-fidelity.json |
| Surrogate fidelity | Analytical vs FDTD Q; bias drift along one trajectory | results/surrogate-fidelity.json |
| Ringdown adequacy | Window vs fitted lifetime; endpoint adequacy bars | results/ringdown-adequacy.json |
| Design23 stack | Layered z cross-section | cases/design23/case.yaml |
| Historical endpoint bars | Two-device orientation | run summaries |
| Design23 fields | XY-midplane ( | E |
Single-device paper (manuscript-quan/)¶
Regenerated by analysis/build_quan_figures.py, audited through
results/quan-figure-manifest.json.
| Figure | Role | Source record |
|---|---|---|
| Loop schematic | The propose/dispose optimization loop | results/quan-campaign.json |
| Geometry | Seed vs step-56 design on matched axes, plus per-hole change | seed JSON, campaign state.json |
| Q vs iteration | Analytical Q every step, FDTD checkpoints with accept/reject, overlap gate, trust radius | results/quan-campaign.json |
| Field profiles | Tidy3D ( | E |
| Ringdown bound | Windows per lifetime along the campaign; the two observables against each other | results/quan-campaign.json |
Still required¶
- matched FDTD mesh/runtime/PML/padding panels, and matched spectra and ringdowns from them;
- new Quan endpoint fields (none were retained historically) and matched Design23 fields;
- timing/memory scaling and an operator-size table.
Analytical convergence and gradient checks are now reported as SI tables (S14) rather than as figures, since the underlying records are small tables and a plot would add nothing.
Every new plot needs matched axes where applicable, material legend, caption, source artifact, field quantity/component, wavelength, slice or integration direction, normalization, and color-scale provenance.