# Beta-first atom-gap inverse design

<div class="status-banner"><strong>STOPPED NO POLE</strong> · campaign <code>beta_first_v1</code> · updated 2026-08-20T19:30:56+00:00</div>

This is the literal live checkpoint for the simplified first stage: an Ez atom in a hard 1 µm vacuum gap, a 2-D mask extruded through 500 nm silicon, and a fixed 400 nm feedthrough waveguide that continues through both x-PMLs. The three mirror planes reduce the solve to one octant. The physical target is **βwg ≥ 90% total**, which symmetry divides equally into 45% left and 45% right.

| Quantity | Current value |
|---|---:|
| Binary coherent-retention updates | 60 / 60 |
| Guided branching ratio βwg | 74.78% total (37.39% per port) |
| Total normalized LDOS | 1.126 |
| Guided normalized LDOS | 0.9173 |
| Atom-frequency coherent retention | 0.1039 |
| Late-window atom-frequency coherence | 0.4766 |
| Gray material fraction | 0.00% |
| Projection stage | 7 / 7 |
| Campaign phase | coherent pole acquisition |
| Latest audited/ringdown Q | 135.1 |
| Latest pole detuning | 3.489% |
| Latest pole-fit residual | 15.500% |
| Purcell factor | pending |
| Guided Purcell βwg FP | pending |

![Literal geometry and routing objective](../_static/generated/scratch_atom_beta_geometry.png)

![Optimization history](../_static/generated/scratch_atom_beta_progress.png)

![Current electric field](../_static/generated/scratch_atom_beta_field.png)


## What is being optimized now

The fixed-frequency LDOS stage produced a broadband router rather than a cavity, so this replacement uses a pulsed solve and coherently demodulates only the atomic frequency. It rewards late atom-field amplitude and retention while constraining exact modal-overlap beta. Every evaluated mask is binary and isolated single-cell solid/void speckle is removed in the physical forward pass. Fitted Q is diagnostic only; a separate 4 ps normalized-LDOS audit must prove a clean Q ≥ 100 pole before shifting. A stricter foundry erosion/dilation audit remains a final acceptance gate rather than being imposed abruptly on this seed.

The state machine is finite: beta binarization → broadband pole audit → binary guided-LDOS acquisition → coherent pole acquisition if the fixed-frequency stage remains broadband → shifted guided-Purcell optimization → independent final validation. A clean pole is required before shifting, and no stage may reintroduce gray material after binarization.