nx_stability_rh.nx

nx_stability_rh.nx -- the MOISTURE axis of stability, and the machinery to SEPARATE it from the temperature axis. ===== WHY THIS EXISTS ============================================ nx_stability recovers an activation energy from ICH's declaration that 6 months at 40 C / 75% RH supports 24 months at 25 C / 60% RH. Those two conditions differ in temperature AND in humidity, so what comes out is a LUMPED effective parameter carrying the moisture contribution inside it. nx_stability says so honestly but cannot do anything about it, because two conditions differing in two variables are underdetermined -- one equation, two unknowns, and no amount of arithmetic fixes that. This file adds the second axis and, with it, the ability to separate them when the experiment actually supports separation. ===== THE MODEL ================================================== The moisture-modified Arrhenius used by accelerated-stability practice: ln k = ln A - Ea/(RT) + B*RH so between two conditions, in base ten (this suite's convention): log10(k2/k1) = E10*(1/T1 - 1/T2) + B10*(RH2 - RH1) where E10 = Ea/(R ln10) and B10 = B/ln10. Two unknowns, so two independent contrasts -- three conditions -- determine both. ===== THE REFUSAL IS THE FEATURE ================================= Three conditions are NECESSARY but not SUFFICIENT: they must not be collinear in (1/T, RH) space. Three points on a line give a singular system, and a solver that "handles" that by picking one of the infinitely many solutions would emit a confident, specific, meaningless Ea. So the determinant is computed and a zero REFUSES. srh_ich_pair_alone_is_separable() returns 0 as a FUNCTION, so this file mechanically demonstrates the limitation nx_stability could only describe.