Skip to content

Results dashboard

The top section shows the latest complete checkpoint from the active workstation optimization, including its step, stage, objective, LDOS, guided beta, tracked pole Q, and numerical-validity state. The browser checks for a newly published checkpoint every 30 seconds. Older tiny and smoke artifacts remain visible as explicitly labeled integration checks.

The visual section shows the current filtered/projected material density, an amplified density-change map relative to step zero, and a normalized physical \(|E_z|^2\) top view on the slab midplane over the full x-y design. The field map is aligned with the geometry, outlines the material boundary, and marks the centered dipole. Geometry updates every checkpoint; the field profile updates every ten completed steps to limit simulation overhead.

Each optimization-history chart uses the actual optimization step on its horizontal axis and names the plotted metric on its vertical axis. Markers show exact step/value pairs on hover or keyboard focus.

The planned geometry is an 8 µm by 2 µm nanobeam envelope. Its top and bottom 0.5 µm strips are fixed air, leaving a continuous 1 µm-wide trainable center.

The optimization maximizes peak log-LDOS as the single-pole \(Q/V\) proxy. Its guided-beta constraint is

\[ \beta = \frac{P_\mathrm{desired\ waveguide}}{P_\mathrm{total}} \ge 0.90, \]

which is equivalent to at least 9:1 desired-waveguide decay versus all other decay. “Other” includes free-space radiation and any undesired modes; beta is a constraint rather than the primary objective.

Loading published run data…