Frozen 150 nm SiN Fourier-to-pixel atom-hole campaign

This page preserves the terminal checkpoint selected by the user after the trajectory visibly saturated. It is not overwritten by the successor 200 nm campaign.

Executed configuration

Quantity	Value
Material stack	150 nm Si3N4 slab in SiO2
Atom	z-oriented, 780.24 nm
Immutable opening	1.0 µm-diameter through-slab air hole
Trainable region	9.6 µm-diameter extruded 2-D mask
Grid	25 nm; slab resolved by 6 cells
Symmetry	x/y/z, 8× reduced solve
Parameters	40,000 Fourier coefficients → 40,000 pixels
Terminal update	1283
Fourier-to-pixel handoff	update 991; executed-density RMSE 0.0
Initial / best LDOS	0.537764 / 0.682342
Best literal-binary LDOS	0.682342

The fixed-frequency matched-vacuum LDOS improved by 26.9%. The result remains below vacuum and no trustworthy resonant pole emerged. Forty-seven independent pulsed audits were attempted; none passed all field/energy agreement and window-stability gates. The largest fitted candidate Q was 42.53. At the final checkpoint, field- and energy-decay fits gave Q=38.61 and 172.65 respectively, so the conservative Q was not accepted and shifted optimization correctly remained disabled.

Geometry and material stack

The literal initial and terminal masks and the executed 150 nm xz stack.

Complete optimization and Q-surveillance history

Fourier and pixel stages share one update axis. Candidate pole fits remain visibly separate from trusted fits; no candidate crossed the fail-closed gate.

Terminal spatial spectrum

At the terminal pixel stage this is the diagnostic Fourier transform of the pixel controls, not a claim that Fourier coefficients remain active controls.

Terminal electric field

Paired atom-frequency xy and xz phasors from the same executed checkpoint.

The full resumable artifact is retained locally as benchmarks/artifacts/invdes_fourier_atom_hole_ldos_sin150_isotropic_v3.npz.