Skip to content

IMRPhenomT dominant ringdown supplied-input gate, 2026-09-27

The component specification records the printed sign conflict between Estellés et al. THM Eqs. (25) and (27a). The tested branch uses the minus sign inferred from Eq. (27a) and the peak join in Eq. (31b). This is a conditional equation-level qualification, not a claim that the literal Eq. (25) or the calibrated input fits are verified.

On len CPU, a clean snapshot of integration commit 3ac8cc3 plus this slice was tested in /home/xangma/tmp/diffgw-phenomt-ringdown-20260927. Python /home/xangma/.cache/diffgw-taylorf2-hm-venv/bin/python was 3.12.3, with NumPy 2.5.3, PyTorch 2.14.0+cpu, JAX 0.11.2, mpmath 1.3.0, and pytest 9.1.1. The current installed CPU packages were used; no environment update was needed.

JAX_ENABLE_X64=1 JAX_PLATFORMS=cpu PYTHONPATH=src:tests /home/xangma/.cache/diffgw-taylorf2-hm-venv/bin/python -m pytest -q tests/test_phenomt_ringdown_frequency_math_reference.py --reference-device=cpu -ra
JAX_ENABLE_X64=1 JAX_PLATFORMS=cpu PYTHONPATH=src:tests /home/xangma/.cache/diffgw-taylorf2-hm-venv/bin/python -m pytest -q tests/test_phenomt_ringdown_frequency_math_reference.py tests/test_phenomt_reference.py::test_author_reference_high_spin_unequal_masses tests/test_phenomt_remnant.py::test_remnant_override_preserves_merger_and_ringdown_join --reference-device=cpu -ra

The isolated gate returned 7 passed in 11.12 s. The combined gate returned 11 passed in 16.83 s, with no skips. Three binary states (30,30,0,0), (35,25,0.7,-0.5), and (60,6,0.9,0.1) supplied the native peak/QNM/shape inputs. At six nonnegative times, both frontends' production carrier frequencies, phase increments, and analytic slopes agreed with the 80-digit scalar formulas at the specification's FP64 tolerances. The scalar identities also passed independent quadrature and finite-difference checks. The existing pinned author regression and remnant-override join checks passed in the combined run; the override equations themselves remain outside this component's source-backed gate.

No native waveform code, constants, defaults, or calibration were changed. The input fits, amplitudes, other modes, full-family assembly, and physical accuracy still need separate evidence.