Validation: interaction timescales
Source: src/morrigan/interaction_timescales.py. Tests: tests/test_interaction_timescales.py.
| Test | Pin | Authority |
|---|---|---|
test_tau_vis_pins_its_closed_form |
1.74771e11 s for two Earths at 0.95/1.05 au, e = 0.05; exact rep_e^4 scaling | Kimura et al. (2025), eqs. 8-10, 23 |
test_tau_vis_pins_its_closed_form (unequal-mass case) |
6.41894e9 s for 1 and 10 Earth masses at e = 0.001 | Kimura et al. (2025), eq. 6 |
test_tau_col_pins_and_returns_a_time_not_a_rate |
1.97470e12 s for the equal-mass pair; 2.43544e11 s for 1 and 10 Earth masses at e = 0.001 | Kimura et al. (2025), eqs. 6, 11 |
test_interaction_wrapper_pins_the_analytic_stability_boundary |
Stability flips inside the bracket [4.95, 5.10] around the analytic root x_crit = 5.0332 for e = 2h | Kimura et al. (2025), eq. 28 |
Eq. 6 is resolved only by the unequal-mass pins. Eq. 23 takes the larger of sum(e_cross) and sum(e), and for the equal-mass reference pair those agree to within one bit, so the crossing eccentricities do not set rep_e there. Both timescales carry the shared denominator, so each is pinned separately: the swapped-denominator variant lands 4.8 % away for the viscous timescale and 2.4 % away for the collision timescale.
Guards: the factor-3 variant and dropped-2pi variant of the viscous timescale, the forgotten rate inversion of the collision timescale (twenty-five orders of magnitude), and the wrong-constant variants of the stability criterion, whose roots (4.62 and 5.77) fall outside the bracket. The Petit et al. (2020) crossing time is ordered across 4, 6, and 8 mutual Hill radii, with its overlap (0.0) and stable (1e20) sentinels asserted exactly.