2.5D LDOS cavity optimizerhoodlab · 780 nm

Checks before claims

Validation status

Tests span the analytic slab limit, radiation-boundary derivative, complete nonlinear objective gradient, cross-basis identity, vector-field identity, and withheld full-vector spectra.

Reduced-model numerical checks

Slab dispersion error
1.444e-04
17 vertical nodes
DtN derivative error
3.465e-10
Objective gradient error
2.696e-05
Protected-strip density
0.0
exact invariant

Centered-emitter symmetry reduction

Launch unknowns
1,701 → 275
even x/y/z sector · 6.185x smaller
Response error
2.092e-15
reduced versus full basis
Gradient error
2.050e-15
complete atom objective
Field correlation
1.000000000000
lifted to the common grid
Measured speedup
1.81x
small launch solve; fixed overhead included
Parity Q-gradient error
2.696e-05
complete nonlinear derivative

Even-sector gate: PASS. The scalar reduction is algebraically exact for the centered source and enforced symmetric topology. It does not yet reduce the full-vector staggered-Yee validation solve.

Cross-basis result

QuantityProduction basisRicher basis
Unknowns1,7013,025
Initial Q15.3532916.25395
Final Q33.1158132.80768
Gain2.1569×2.0184×
Initial/final field correlation0.983866 / 0.984896

Reduced-model gate: PASS. The same scalar mode improves on both bases. This establishes a basis-stable 2.5D candidate, not a calibrated full-vector Q.

Vector transfer and linewidth checks

3D branch identity

Selected-branch complex vector overlap: 0.997021. Energy-map overlap: 0.998693.

PASS The transferred topology is compared on the same reacquired vector branch.

Current linewidth step

Fit Q improves by 0.1406%; withheld Q improves by 0.1010%.

15/15 pass Controlled-grid only; refinement remains pending.

780 nm full-vector scale-up qualification

Readiness decision
PASS
5/5 top-level gates
Current diagnostic Q
not the optimized objective
Boundary LDOS change
0.00%
limiting-absorption selector
Atom-gradient error
3.164e-08
complete full-vector chain rule
Grid field overlap
0.999235
density-only transfer and reacquisition
Target-window Q
No pole
81 points · seven complex probes
Domain field overlap
0.999287
expanded PML/domain
Q reporting
Honest
no numeric Q assigned to analytic background
Readiness gateResult
complete exact production gradientPASS
fixed geometry grid and domain convergedPASS
restart resources and improving trial

Scope of this decision. Passing authorizes the next higher-resolution/larger-domain atom-coupling optimization. The refinement comparison changes the transverse y spacing while holding x and z spacing fixed, so this is a launch gate—not a full x/y/z device-convergence claim. No rigorous QNM mode volume, Purcell factor, useful-port beta, tweezer-scattering budget, material loss, or fabrication tolerance is established.