Campaign: `general_1d_cavity/paper_reproduction_v1`¶

Assembled 2026-09-05 14:55 UTC.

Engine: general_1d_cavity · path: engines/general_1d_cavity/campaigns/paper_reproduction_v1/

campaign.yaml¶

schema_version: 1
campaign_id: paper_reproduction_v1
engine: general_1d_cavity
objective: >
  Reproduce Quan 2011, Sawfish 2022, and Fryett 2017; raise model Q by
  at least 2x; confirm each gain with matched Tidy3D.
tidy3d_flexcredit_cap: 15.0
fdtd_submit_from_build: false
families:
  - quan2011
  - sawfish
  - fryett2017
status: active

README¶

Paper-reproduction campaign¶

Goal. Represent, optimize, and independently confirm three published 1-D photonic-crystal cavities with the general outgoing-pole engine.

Budget. 15 Tidy3D flexcredits. Ledger: tidy3d_credit_ledger.json (0 spent at campaign start).

Done when STATUS.md records, for each family below, a near-paper wavelength, ≥2× model \(Q\), and ≥2× matched FDTD \(Q\).

Families¶

Family	Paper	First model	Seed target
Quan 2011	arXiv:1108.2675	Si 700×220 nm, air, elliptical holes	~1500 nm, dielectric \(E_y\)
Sawfish 2022	arXiv:2210.04702	diamond, sidewall teeth	~619 nm SnV
Fryett 2017	arXiv:1709.02032	SiN in \(n=1.5\) cladding, circular holes	~737–780 nm

Safety¶

Do not submit Tidy3D unless authorize_submission succeeds.
Analytical Q and FDTD Q are distinct observables.
Cheap P0/y1 numbers inflate absolute Q; they certify machinery only.
Production Quan must recover the preserved Phase-57/58 lineage before any new FDTD spend on that family.

Current cheap-basis evidence¶

See runs/cheap_poles.json. Six-pair Quan P0/y1: model \(Q\) 70 556 → 367 350 in five implicit steps, wavelength held at 1875.89 nm. That is not a publication claim.

Production Quan (15-pair P2-z1/y3-z1) recovers the packaged preflight pole at \(Q=7722.35\), \(\lambda=1521.77\) nm. Four implicit steps raise model \(Q\) to 16835 (\(2.180\times\)), \(\lambda=1521.41\) nm. Evidence: runs/quan/q_ascent.json. Not an FDTD claim.

Fryett: the cheap 6-pair P0/y1 pole sits at \(Q\approx 129\), \(\lambda\approx 946\) nm. A 10-pair P1/y1 circular reduction on the paper 450×330 nm beam (\(a=233\) nm) gives an outgoing pole at \(Q=907\), \(\lambda=776.7\) nm. Two implicit radius steps raise model \(Q\) to 6184 (6.82×) with \(\lambda\) held to 0.09 nm. Evidence: runs/fryett/. Not an FDTD claim.

Sawfish: the isolated 6-pair \(\cos^6\) P1-x pole sits at \(Q=268.5\), \(\lambda=619.71\) nm. One 2 nm implicit (period, neck) step raises model \(Q\) to 1831 (6.82×) with \(\lambda=619.99\) nm. Evidence: runs/sawfish/cosine_n6_a200_P1_qsteps/. Absolute cheap-basis \(Q\) is not a production claim. Not an FDTD claim.

Campaign: general_1d_cavity/paper_reproduction_v1¶