Skip to content

NRTidal v1 nonspinning tidal-phase qualification — 27 September 2026

Outcome: 33/33 passed, no skips, expected failures or deselection, for the bounded phase specification. This does not qualify the complete waveform or its physical accuracy.

The run used len, Linux 6.8.0-138-generic x86_64, in /home/xangma/tmp/diffgw-nrtidalv1-phase-20260927, copied from isolated checkout 518e49a33c6da8cdae860959c81ddc985cd3f512 with this slice's working files. The command was:

DIFFGW_REQUIRE_LAL_BINARY=1 OMP_NUM_THREADS=1 \
JAX_ENABLE_X64=1 JAX_PLATFORMS=cpu \
PYTHONPATH=/home/xangma/tmp/diffgw-nrtidalv1-phase-20260927/src:/home/xangma/tmp/diffgw-nrtidalv1-phase-20260927/tests \
/home/xangma/.cache/diffgw-taylorf2-hm-venv/bin/python -m pytest -q -s \
  tests/test_nrtidalv1_phase_math_reference.py \
  tests/test_nrtidalv1_phase_lal_binary.py

The complete log is qualification.log in that directory. Worker PID 1485390 exited after 17.78 s. Python was 3.12.3; NumPy 2.5.3, SymPy 1.14.0, mpmath 1.3.0, Torch 2.14.0+cpu, JAX 0.11.2, and LALSuite 7.26.15. The installed public LALSimulation extension was _lalsimulation.cpython-312-x86_64-linux-gnu.so, SHA-256 e816879d23e652e8b3928496d6fd261de0832b5c3781c50e33a458042495b42e. No LAL source was inspected.

The scalar reference derives the effective tidal coupling from Dietrich et al. (2017) Eq. (2) and transcribes its Eq. (7) Padé phase. It symbolically checks the mass-weighted deformability and exact 1PN coefficient; native Torch/JAX unit-complex BNS/BBH quotients match the explicitly rounded-carrier scalar values at five frequencies between 30 and 800 Hz. Selected frequency derivatives at 300 Hz meet the specified \(10^{-8}\) relative and \(10^{-10}\) rad/Hz absolute limits.

The predeclared binary comparison used the four main systems plus two holdouts stated in the specification, all at 30–800 Hz in 5 Hz steps. It compared unaligned unit-complex IMRPhenomD_NRTidal / IMRPhenomD quotients, requiring relative complex \(L_2\le10^{-6}\) and maximum phase residual \(\le10^{-5}\) rad. The largest observed LAL-to-rounded-scalar relative complex \(L_2\) was 2.09995954e-15; the largest Torch/JAX-to-LAL value was 2.66248986e-15. The largest respective pointwise phase residuals were 9.01201102e-15 and 1.2543367e-14 rad. These results show the native implementations, independent scalar rounded variant, and named binary agree on the declared phase quotient.

Source-to-carrier precision boundary. Eq. (7) defines \(d_1=n_1-3115/1248=-19.923993589\ldots\), whereas the native carriers and named binary use the rounded -19.924. This changes the paper-literal phase by up to 1.6497434e-5 rad for the equal-mass case and 1.9064569e-5 rad for the unequal-mass case over the same grid. The corresponding paper-literal versus rounded unit-complex relative \(L_2\) values are 5.7321178e-6 and 6.6094838e-6. The binary match therefore validates a three-decimal realization of the published fit, not exact implementation of its analytic 1PN constraint.

Amplitude, merger/taper, spinning binaries, other modes, FP32 and accelerator behavior, inherited IMRPhenomD physics, and comparisons with independent neutron-star simulations remain outside this receipt.