Relaxation fixed-time numerical QA (2026-10-08)

Baseline main: 330d21eaa397a7baa06ef002c60a01c8e4b64b9d.

The Bloch relaxation explorer uses an adaptive time window tmax=clamp(4 max(T1,T2),1,32) microseconds, which supports readable exponential curves but inhibits absolute visual comparison of timescales across input settings. Retain this zoomed plot, and add normalized absolute-time observables at t=1 μs: M_z(t)=1-exp(-t/T1) and M_xy(t)=exp(-t/T2), with 1/T2=1/(2T1)+1/Tphi. These refer to the toy model’s zero-initial-population-recovery and unit-initial-transverse-coherence assumptions.

Numerical sweep: T1=[0.2,0.5,1,2,4,8] μs, Tphi=[0.2,0.5,1,3,10,20] μs. All 36 combinations gave finite values with normalized observables in [0,1] and T2<=2T1. At defaults T1=2 μs, Tphi=3 μs: T2=1.7143 μs, recovered longitudinal amplitude at 1 μs 39.3469%, remaining transverse amplitude at 1 μs 55.8035%; HTML fallback rounds to 39.3% and 55.8%.

The inventory remains 73 controls and is not yet exhaustively evaluated; this is an incremental 36-case quantitative test and source-level review, not a pixel-level browser inspection.