Black-hole ringdown with templates capturing spin precession: a critical re-analysis of GW190521

TL;DR

This paper develops and tests physics-informed ringdown templates that incorporate spin precession and applies them to GW190521 using a simulation-based (TSNPE) inference framework. By combining NR-calibrated QNM amplitude fits with a remnant-mass/spin surrogate, the authors quantify how precession alters ringdown-inferred binary parameters and mode excitations. They find modest, systematic shifts due to precession and no strong evidence for precession from the ringdown alone, underscoring the viability of precessing ringdown templates while highlighting limitations at current SNRs. The work establishes a foundation for spin-precession detection in ringdown-dominated gravitational-wave events and guides future improvements with higher-SNR data.

Abstract

The ringdown phase of a binary black-hole merger provides a clean probe of strong-field gravity, as it can be modeled with minimal assumptions. The quasi-normal-mode frequencies encode the mass and spin of the Kerr black-hole remnant, while the mode excitation depends on the progenitor binary. In this paper, we implement a recently developed amplitude model that captures spin precession in a simulation-based inference pipeline that specifically targets ringdown signals. We present a critical re-analysis of GW190521 -- a short-duration, merger-dominated event with conflicting interpretations. Spin-aligned and precessing analyses at two ringdown start times show that precession induces modest but systematic shifts in inferred parameters and subdominant mode amplitudes, although such ringdown-only analyses provide no strong evidence for precession. Our results demonstrate the feasibility of physics-informed precessing ringdown modelling, paving the way for the identification of spin precession in gravitational-wave events using solely their ringdown stages, where waveform systematics are expected to be substantially less prominent.

Figures (5)

• Figure 1: Joint posterior distributions for the binary parameters inferred with the spin-aligned (blue) and precessing (orange) ringdown models, using the GPR mean predictions for the remnant mass and spin and for the QNM amplitudes. The lower (upper) triangular panels correspond to a ringdown start time of 6 ms (12 ms) after the strain peak. The IMR posterior (green) from the LVK analysis 2020PhRvL.125j1102A2020ApJ...900L..13A is overlaid for comparison. All contours represent the 68% and 90% credible regions.
• Figure 2: Joint posterior distributions for the QNM amplitudes inferred from the spin-aligned (blue) and precessing (orange) ringdown models, using the GPR mean predictions for the remnant mass and spin and for the QNM amplitudes. We show the absolute amplitude of the dominant $(2,2,0)$ mode, $A_{220}$, together with the relative amplitudes $A_{330}/A_{220}$ and $A_{210}/A_{220}$. The lower (upper) triangular panels correspond to a ringdown start time of 6 ms (12 ms) after the strain peak. For the 6 ms case, we overlay the results of Ref. 2023PhRvL.131v1402C (pink), obtained using a $(2,2,0)+(3,3,0)$ Kerr model. Contours indicate the 68% and 90% credible regions.
• Figure 3: Residual errors on our remnant GPR fits for $\left(M_f,\chi_{fz},\chi_{f\perp}\right)$, respectively, computed via 10-fold cross validation. The dashed vertical lines delimit the regions comprising the 90% density of each distributions.
• Figure 4: Joint posterior distributions for the binary parameters inferred with the spin-aligned (blue) and precessing (orange) ringdown models, using randomly sampled GPR realizations of the remnant mass and spin and of the QNM amplitudes. The lower (upper) triangular panels correspond to a ringdown start time of 6 ms (12 ms) after the strain peak. The IMR posterior (green) from the LVK analysis 2020PhRvL.125j1102A2020ApJ...900L..13A is overlaid for comparison. All contours represent the 68% and 90% credible regions.
• Figure 5: Joint posterior distributions for the QNM amplitudes inferred from the spin-aligned (blue) and precessing (orange) ringdown models, using randomly sampled GPR realizations of the remnant mass and spin and of the QNM amplitudes. We show the absolute amplitude of the dominant $(2,2,0)$ mode, $A_{220}$, together with the relative amplitudes $A_{330}/A_{220}$ and $A_{210}/A_{220}$. The lower (upper) triangular panels correspond to a ringdown start time of 6 ms (12 ms) after the strain peak. For the 6 ms case, we overlay the results of Ref. 2023PhRvL.131v1402C (pink), obtained using a $(2,2,0)+(3,3,0)$ Kerr model. Contours indicate the 68% and 90% credible regions.