Skip to content

Author decisions

This log records choices that materially constrain the paper and repository.

Date Decision Consequence
2026-08-17 One method, two device families. Quan–Lončar and Design23 share one main narrative and one SI.
2026-08-17 Preprint-first publication sequence. Build a complete auditable preprint before journal reformatting.
2026-08-17 LaTeX is authoritative. Website prose is a workbench; generated PDFs are canonical.
2026-08-17 Historical Design23 trajectory is primary only after exact matched revalidation. Step 0 and 105 are frozen; separate seed and repaired-run results remain auditing controls.
2026-08-17 Author site is private/authenticated. No route or publish action until shared-infrastructure approval.
2026-08-17 Public unauthenticated drafting route supersedes the private-site decision. Publish the working site at /gme-1d/ with noindex; a future site-wide login is separate shared infrastructure.
2026-08-17 Initial release surface is site plus polished PDFs. Reviewer/code/data packages are deferred until the paper is written and evidence audited.
2026-08-17 Preserve raw data by checksum outside Git. 69 HDF5 artifacts remain local; compact evidence and SHA-256 registry enter Git.
2026-08-18 Quantitative claims are the retained historical endpoints, honestly qualified. Complete submission-quality manuscript written now; one clean substitution point (manuscript/values.tex) replaces panel numbers later without touching prose.
2026-08-18 The Design23 dielectric→air mode-character change is a secondary observation. Reported as characterized geometric and spectral restructuring (aspect ratio, 2r_y/W, blueshift, V_eff). No band-structure work, no HDF5 re-analysis, no band-edge-swap assertion.
2026-08-18 Venue: physics.optics primary, cross-list physics.comp-ph and quant-ph. Two-column article[10pt], ~11-page main text; SI single-column.
2026-08-18 The SI is a self-contained rewrite, not a stitched derivation archive. S1–S13 written in one notation from scratch; derivations/generated/ becomes an audited provenance table in S19.
2026-08-18 The headline result is direction transfer without magnitude transfer. New claim result.direction_not_magnitude_transfer; §7 of the main text; per-step sign agreement (19/31, p=0.28) explicitly not claimed.
2026-08-18 No number is typed into the LaTeX. Three evidence generators plus two exporters; tests/test_evidence.py pins their outputs; a record change without a test change fails the build.
2026-08-18 The predeclared Quan panel axis is amended, not reinterpreted. At the optimized endpoint the declared 60/90 ps gives N = 0.63/0.95 lifetimes; promotion requires extending the axis until N ≳ 3. Recorded in §4 and SI S15.
2026-08-18 A second, narrower draft covers the Quan–Lončar campaign alone. manuscript-quan/ is an independent manuscript built from results/quan-campaign.json. It uses the 56-step indefinite campaign, not the 3-step Phase-58 endpoint quoted in paper A, so the two papers' numbers differ by construction and must not be cross-quoted.
2026-08-18 Paper B states its own certified bound rather than its endpoint. The FDTD checkpoints used a fixed 64 ps window while the photon lifetime grew; adequacy falls below one lifetime at step 9. The largest gain measured over a full lifetime is 2.12×, and that is the number the paper certifies.

Add a row whenever a choice changes claims, evidence, scope, authorship, release boundaries, or the final simulation protocol.