Skip to content

Source

engines/quan_loncar_v1/package/quan_loncar_optimizer/engine/quan_loncar_series_phase20/README.md · assembled 2026-07-29 15:57 UTC.


Quan--Loncar series, Phase 20

Phase 20 replaces brute-force treatment of the slow Maxwell longitudinal tail. For the current through-thickness-constant waveguide basis it evaluates the q_z integral of P_L analytically, evaluates only rapidly convergent one-dimensional q_y blocks numerically, and obtains the slow P_yy block exactly from

P_xx + P_yy + P_zz = I.

The remaining two-dimensional grid sees only P_T/(k^2-k_c^2), whose high-momentum tail is compact and much faster. A finite-grid algebra test proves that the decomposition reproduces the legacy combined Maxwell symbol before the infinite longitudinal completion is made.

run_phase20_regularized_pole.py then tracks the actual N=15 pole through transverse cutoffs 4, 6, and 8, doubles the residual one-dimensional longitudinal cutoff, and raises the compact transverse panel order.