NRTidalv2 Pv2 inherited remnant-spin receipt — 27 September 2026

Result: The source-complete in-plane remnant-spin map passed its independent scalar gate in Torch and JAX. Public IMRPhenomPv2_NRTidalv2 dispatch remains unqualified.

The frozen specification applies Dietrich et al. (2019), Sec. III.4 to the Pv2 technical note, Eqs. (29)–(30), (46), (51). The scalar reference independently evaluates the printed effective in-plane-spin and remnant-spin equations with mpmath. The Torch and JAX waveform drivers now call explicit helpers for the same remnant map; the historical signed retrograde behavior is unchanged. The tested paper domain has positive aligned final spin, so it makes no claim about that signed extension.

Three BNS-scale configurations cover equal and unequal masses, both bodies with nonzero in-plane spin, and either body dominating \(\chi_p\). For each, the native \(\chi_p\) and remnant-spin value agree with the scalar result to the frozen 2e-15 absolute bound in CPU FP64. Neither the aligned final-spin fit nor the QNM frequencies are treated as independently proved by this gate.

Validation ran on len, scratch cwd /home/xangma/tmp/diffgw-nrtidalv2-remnant-20260927, using the current available environment /home/xangma/repos/diffgw-worktrees/xas-whole-waveform-gate/.venv/bin/python (Torch 2.14.0+cpu, JAX 0.11.2, pytest 9.1.1, mpmath 1.3.0). Command:

PYTHONPATH=src /home/xangma/repos/diffgw-worktrees/xas-whole-waveform-gate/.venv/bin/python -m pytest -q tests/test_nrtidalv2_pv2_remnant_spin_math_reference.py tests/test_phenompv2_waveform_clean.py tests/test_nrtidalv2_pv2_aligned_carrier_math_reference.py

30 passed, 1 skipped in 11.66s; log pytest.log in the scratch cwd. After preserving the JAX driver's original multiplication order in the extracted helper, the same command passed again (30 passed, 1 skipped in 10.49s). The skipped existing Pv2 parity-harness test requires unavailable PyCBC; pytest -q -rs tests/test_phenompv2_waveform_clean.py confirmed its reason (18 passed, 1 skipped). No LAL source, fitted residual, phase alignment, or waveform retuning entered this gate.

The next public-dispatch boundary is specific. The inherited Pv2 technical note sets the longitudinal spin input for the NNLO angles to the larger body's aligned spin in Eq. (32), while the native Torch/JAX drivers use a mass-weighted mapping previously selected for BBH binary compatibility. The note's Eq. (26) prints 3.373 - 19η/8 - η²/24 for the 2PN orbital coefficient, while Pratten et al. (2020), Eq. (27) prints 27/8 - 19η/8 + η²/24; the native code has the mixed 27/8 - 19η/8 - η²/24. Therefore an exact paper angle gate against the current native drivers cannot pass without choosing and naming a convention. Dietrich Sec. III.4 directs insertion before rotation, but does not say whether the Pv2 remnant-frequency spline time shift is sampled from the BBH or finite-size carrier, nor does its phase decomposition set the BNS reference-frequency phase origin. Those choices need a separately frozen contract before exposing a complete public BNS dispatch.