pith. machine review for the scientific record. sign in

arxiv: 2604.23625 · v1 · submitted 2026-04-26 · 🌀 gr-qc · astro-ph.IM· physics.comp-ph

Recognition: unknown

Physics informed operator learning of parameter dependent spectra

Haohao Gu , Sensen He , Hanlin Song , Bo Liang , Zhenwei Lyu , Xiaoguang Hu , Minghui Du , Peng Xu
show 1 more author
Bo-Qiang Ma
Authors on Pith no claims yet

Pith reviewed 2026-05-08 05:34 UTC · model grok-4.3

classification 🌀 gr-qc astro-ph.IMphysics.comp-ph
keywords quasinormal modesKerr black holesphysics-informed neural networksoperator learningblack hole spectroscopyspectral problemsgravitational wavesringdown modeling
0
0 comments X

The pith

One trained network learns the full spin-dependent quasinormal mode spectrum of Kerr black holes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents DeepOPiraKAN, a physics-informed operator learning network that trains a single model to map continuous parameters directly to the full spectrum of a differential operator. Applied to quasinormal modes of Kerr black holes, this replaces repeated solutions of the eigenvalue problem at separate spin values with one forward pass after training. The network resolves angular modes (2,0) and (2,1) with overtones up to n=7 across the entire spin range, matching Leaver's method to relative errors of order 10^{-6} for the fundamental mode and 10^{-4} for higher overtones. A sympathetic reader cares because these spectra govern the ringdown phase of gravitational wave signals, and efficient parameter sweeps are needed for data analysis with current and future detectors. The results position the architecture as a general tool for any spectral problem whose eigenvalues vary continuously with physical parameters.

Core claim

DeepOPiraKAN combines operator learning with physics-informed loss terms to learn the parameter-to-spectrum mapping for Kerr quasinormal modes. A single trained network resolves modes with (ℓ,m) in {(2,0),(2,1)} and overtones up to n=7 over the full spin range, achieving relative errors of O(10^{-6}) for the fundamental mode that rise gradually to O(10^{-4}) for higher overtones when benchmarked against Leaver's method.

What carries the argument

The DeepOPiraKAN neural architecture, which performs operator learning on the parameter-dependent spectral operator while enforcing the differential equation and boundary conditions through the loss function.

If this is right

  • Quasinormal mode frequencies and damping times become available for any spin value after a single training run.
  • Black hole spectroscopy and ringdown waveform modeling gain practical speed for parameter scans needed by observatories.
  • The same framework applies to other parameter-dependent spectral problems without repeated numerical solves at isolated points.
  • Accuracy holds across the full spin domain for the tested angular modes and overtones.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The trained model could speed up real-time parameter estimation from observed gravitational-wave ringdown signals.
  • Extension to higher overtones, different angular indices, or non-Kerr spacetimes would test the generality of the operator-learning step.
  • Embedding the network inside waveform templates might reduce the cost of generating large template banks for matched filtering.
  • The approach could transfer to eigenvalue problems in other wave equations that depend continuously on system parameters.

Load-bearing premise

The neural architecture and physics-informed training procedure can faithfully represent the continuous and sometimes sensitive dependence of the quasinormal mode frequencies on black hole spin without losing accuracy for higher overtones.

What would settle it

A direct comparison at an intermediate spin value where the network prediction for a low-overtone mode deviates from Leaver's method by more than the stated error tolerance.

Figures

Figures reproduced from arXiv: 2604.23625 by Bo Liang, Bo-Qiang Ma, Hanlin Song, Haohao Gu, Minghui Du, Peng Xu, Sensen He, Xiaoguang Hu, Zhenwei Lyu.

Figure 1
Figure 1. Figure 1: FIG. 1. The Physics Informed Deep Operator Network with Residual Adaptive KAN ( view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Relative accuracy of the first eight QNMs for the view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Solution accuracy for the real part of the first eight QNMs in the view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Solution accuracy for the imaginary part of the first eight QNMs in the view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Mean PDE loss view at source ↗
read the original abstract

Spectral problems governed by differential operators underpin a wide range of physical systems, yet remain computationally challenging because their spectra depend sensitively on continuous parameters and often demand repeated evaluations across parameter space. Here we present $\texttt{DeepOPiraKAN}$, an open source physics informed neural network architecture for spectral analysis. By combining operator learning with enhanced optimization stability, it captures the underlying parameter-to-spectrum mapping in a single model, avoiding repeated spectral solutions at isolated points in parameter space. As a representative and stringent benchmark, we apply this framework to the computation of quasinormal modes of Kerr black holes. A single trained network accurately resolves modes with $(\ell,m)\in \{(2,0),(2,1)\}$ and overtones up to $n=7$ across the full spin range, achieving relative errors of $\mathcal{O}(10^{-6})$ for the fundamental mode and gradually increasing to $\mathcal{O}(10^{-4})$ for higher overtones, benchmarked against the Leaver's method. This level of accuracy is already significant for black hole spectroscopy and practical ringdown modelling for current and future observatories. More broadly, these results highlight the potential of $\texttt{DeepOPiraKAN}$ as a general and scalable framework for parameter dependent spectral problems across complex physical systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper introduces DeepOPiraKAN, an open-source physics-informed neural network architecture for operator learning of parameter-dependent spectra. As a benchmark application, it applies the framework to Kerr black hole quasinormal modes, claiming that a single trained network accurately resolves modes for (ℓ,m)∈{(2,0),(2,1)} and overtones up to n=7 across the full spin range a∈[0,1), with relative errors of O(10^{-6}) for the fundamental mode increasing to O(10^{-4}) for higher overtones when benchmarked against Leaver's method.

Significance. If the accuracy claims hold under rigorous validation, the work offers a potentially useful surrogate model for efficient computation of QNM spectra over continuous parameter ranges, which could benefit black hole spectroscopy and ringdown modeling for gravitational-wave observations. The combination of operator learning with physics-informed losses addresses a genuine computational challenge in spectral problems, and the open-source release supports reproducibility.

major comments (2)
  1. Abstract: The central claim of uniform O(10^{-4}) relative accuracy for n=7 overtones across a∈[0,1) is load-bearing. Kerr QNMs for high overtones exhibit increasingly steep dependence on a near extremality, so the training protocol must resolve this; without explicit confirmation that collocation points and loss weighting are sufficient near a=1, local errors could exceed the reported bound even if average errors are small.
  2. Methods/Results sections: The physics-informed loss enforcing the Teukolsky radial equation and outgoing boundary conditions must be specified in detail (including any adaptive weighting for stiff higher-n modes) to verify that the network faithfully captures the continuous parameter dependence without introducing spurious solutions or degradation for n=7.
minor comments (3)
  1. Introduction: Expand the acronym DeepOPiraKAN on first use and clarify the role of KANs versus standard PINNs in the operator-learning component.
  2. Results: Include error maps or tables showing maximum relative error versus a for each overtone n, particularly highlighting behavior near a=1, to make the accuracy claims more transparent.
  3. Notation: Ensure consistent use of symbols for the complex frequency ω(ℓ,m,n,a) and relative error definition throughout.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback and positive assessment of the potential utility of DeepOPiraKAN. We have revised the manuscript to provide the requested details on training and loss formulation, strengthening the validation of accuracy claims. Our point-by-point responses follow.

read point-by-point responses

  1. Referee: Abstract: The central claim of uniform O(10^{-4}) relative accuracy for n=7 overtones across a∈[0,1) is load-bearing. Kerr QNMs for high overtones exhibit increasingly steep dependence on a near extremality, so the training protocol must resolve this; without explicit confirmation that collocation points and loss weighting are sufficient near a=1, local errors could exceed the reported bound even if average errors are small.

    Authors: We agree that near-extremal accuracy for high overtones is a stringent test. The training protocol in the original manuscript already incorporated a non-uniform distribution of collocation points with increased density for a > 0.95, combined with the physics-informed loss that directly penalizes deviations from the Teukolsky equation. To make this explicit and address the concern about local errors, we have added a new figure in the revised Results section plotting relative error versus a for the n=7, (ℓ,m)=(2,1) mode, confirming that the error remains bounded by O(10^{-4}) across the full range, including a → 1^-. We have also updated the Methods section with the precise sampling strategy. revision: yes

  2. Referee: Methods/Results sections: The physics-informed loss enforcing the Teukolsky radial equation and outgoing boundary conditions must be specified in detail (including any adaptive weighting for stiff higher-n modes) to verify that the network faithfully captures the continuous parameter dependence without introducing spurious solutions or degradation for n=7.

    Authors: We appreciate the request for greater specificity. The original manuscript outlined the composite loss in Section 3, but we acknowledge it omitted explicit weighting details. In the revised version we have expanded the Methods section to provide the full mathematical form of the loss terms (interior residual of the Teukolsky radial operator plus outgoing boundary conditions at the horizon and infinity). We now also describe the adaptive weighting procedure, in which per-mode weights are dynamically scaled by the running residual norm to mitigate stiffness for n ≥ 5. Additional cross-validation against Leaver’s method at 200 randomly sampled points (including n=7) shows no spurious modes and confirms the reported error bounds. revision: yes

Circularity Check

0 steps flagged

No significant circularity; accuracy claims benchmarked against independent solver

full rationale

The paper introduces DeepOPiraKAN as a physics-informed operator-learning network that maps black-hole spin to quasinormal-mode frequencies by enforcing the underlying differential operator via a loss function. The reported relative errors (O(10^{-6}) to O(10^{-4})) are obtained by direct numerical comparison to Leaver's method, an established, independent spectral solver. No equation in the abstract or described workflow defines the target spectrum in terms of the network's own fitted weights, nor does any load-bearing step rely on a self-citation chain that would render the result tautological. The central claim therefore remains an externally falsifiable numerical surrogate rather than a self-referential re-expression of its training inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central accuracy claim rests on the assumption that a physics-informed loss can enforce the underlying differential operator across the parameter space and that the network generalizes without post-hoc tuning.

free parameters (1)
  • Neural network weights and hyperparameters
    All trainable parameters of the DeepOPiraKAN model are fitted during training to match the target spectra.
axioms (1)
  • domain assumption The physics-informed loss function is sufficient to embed the differential operator governing the spectrum
    Invoked when the network is trained to respect the governing equations rather than purely data-driven fitting.

pith-pipeline@v0.9.0 · 5556 in / 1266 out tokens · 67041 ms · 2026-05-08T05:34:21.254582+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references · 50 canonical work pages · 5 internal anchors

  1. [1]

    Hern´ andez, J

    V. Hern´ andez, J. E. Rom´ an, V. Vidal, Slepc: Scalable library for eigenvalue problem computations, in: Inter- national Conference on High Performance Computing for Computational Science, Springer, 2002, pp. 377–391

    2002

  2. [2]

    J. E. Roman, C. Campos, E. Romero, A. Tom´ as, Slepc users manual, Departamento di Sistemas Informaticos y Computaci´ on, Universitat Politecnica de Valencia, TR DSIC-II/24/02, Rev 3 (2016)

    2016

  3. [3]

    K. D. Kokkotas, B. G. Schmidt, Quasinormal modes of stars and black holes, Living Rev. Rel. 2 (1999) 2.arXiv: gr-qc/9909058,doi:10.12942/lrr-1999-2

    work page doi:10.12942/lrr-1999-2 1999

  4. [4]

    R. A. Konoplya, A. Zhidenko, Quasinormal modes of 10 0.0 0.1 0.2 0.3 0.4 0.5 10−5 10−2 ϵrel n=7 0.0 0.1 0.2 0.3 0.4 0.5 10−4 10−1 ϵrel n=6 0.0 0.1 0.2 0.3 0.4 0.5 10−3 ϵrel n=5 0.0 0.1 0.2 0.3 0.4 0.5 10−4 10−1 ϵrel n=4 0.0 0.1 0.2 0.3 0.4 0.5 10−4 ϵrel n=3 0.0 0.1 0.2 0.3 0.4 0.5 10−4 ϵrel n=2 0.0 0.1 0.2 0.3 0.4 0.5 10−5 ϵrel n=1 0.0 0.1 0.2 0.3 0.4 0.5...

    work page doi:10.1103/revmodphys.83.793 2000

  5. [5]

    Quasinormal modes of black holes and black branes

    E. Berti, V. Cardoso, A. O. Starinets, Quasinormal modes of black holes and black branes, Class. Quant. Grav. 26 (2009) 163001.arXiv:0905.2975,doi:10. 1088/0264-9381/26/16/163001

    work page internal anchor Pith review arXiv 2009

  6. [6]

    doi:10.1103/PhysRevD.89.104059, https://arxiv.org/abs/1404.7149 arXiv:1404.7149 [gr-qc]

    E. Barausse, V. Cardoso, P. Pani, Can environmental effects spoil precision gravitational-wave astrophysics?, Phys. Rev. D 89 (10) (2014) 104059.arXiv:1404.7149, doi:10.1103/PhysRevD.89.104059

    work page doi:10.1103/physrevd.89.104059 2014

  7. [7]

    Franchini, S

    N. Franchini, S. H. V¨ olkel, Testing General Relativity with Black Hole Quasi-normal Modes, 2024.arXiv: 2305.01696,doi:10.1007/978-981-97-2871-8_9

    work page doi:10.1007/978-981-97-2871-8_9 2024

  8. [8]

    Black hole spectroscopy: from theory to experiment

    J. Abedi, et al., Black hole spectroscopy: from theory to experiment (5 2025).arXiv:2505.23895

    work page internal anchor Pith review arXiv 2025

  9. [9]

    B. P. Abbott, et al., Observation of Gravitational Waves from a Binary Black Hole Merger, Phys. Rev. Lett. 116 (6) (2016) 061102.arXiv:1602.03837,doi:10.1103/ PhysRevLett.116.061102

    work page internal anchor Pith review arXiv 2016

  10. [10]

    Laser Interferometer Space Antenna

    P. Amaro-Seoane, et al., Laser Interferometer Space An- tenna, e-prints (2 2017).arXiv:1702.00786

    work page internal anchor Pith review arXiv 2017

  11. [11]

    Combinatorial optimization enhanced by shallow quantum circuits with 104 superconduct- ing qubits

    W.-R. Hu, Y.-L. Wu, The Taiji Program in Space for gravitational wave physics and the nature of gravity, Natl. Sci. Rev. 4 (5) (2017) 685–686.doi:10.1093/nsr/ nwx116

    work page doi:10.1093/nsr/ 2017

  12. [12]

    Du, et al., Towards realistic detection pipelines of Taiji: New challenges in data analysis and high-fidelity simulations of space-based gravitational wave antenna, Sci

    M. Du, et al., Towards realistic detection pipelines of Taiji: New challenges in data analysis and high-fidelity simulations of space-based gravitational wave antenna, Sci. China Phys. Mech. Astron. 69 (4) (2026) 249501. arXiv:2505.16500,doi:10.1007/s11433-025-2870-8

    work page doi:10.1007/s11433-025-2870-8 2026

  13. [13]

    Quasinormal modes of black holes and black branes

    J. Luo, et al., TianQin: a space-borne gravitational wave detector, Class. Quant. Grav. 33 (3) (2016) 035010.arXiv:1512.02076,doi:10.1088/0264-9381/ 33/3/035010

    work page doi:10.1088/0264-9381/ 2016

  14. [14]

    Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations

    M. Raissi, P. Perdikaris, G. E. Karniadakis, Physics- informed neural networks: A deep learning framework for solving forward and inverse problems involving non- linear partial differential equations, J. Comput. Phys. 378 (2019) 686–707.arXiv:1711.10561,doi:10.1016/ j.jcp.2018.10.045

    work page Pith review arXiv 2019

  15. [15]

    Anitescu, B

    C. Anitescu, B. ˙Ismail Ate¸ s, T. Rabczuk, Physics- informed neural networks: Theory and applications, in: Machine Learning in Modeling and Simulation: Methods and Applications, Springer, 2023, pp. 179–218

    2023

  16. [16]

    DOI 10.1140/epjc/ s10052-022-10942-5

    D. Tseneklidou, A. Torres-Forne, P. Cerda-Duran, To- wards asteroseismology of neutron stars with physics- informed neural networks, Eur. Phys. J. C 85 (10) (2025) 1218.arXiv:2504.12183,doi:10.1140/epjc/ s10052-025-14942-z

    work page doi:10.1140/epjc/ 2025

  17. [17]

    A. S. Cornell, A. Ncube, G. Harmsen, Using physics- informed neural networks to compute quasinormal modes, Phys. Rev. D 106 (12) (2022) 124047.arXiv: 2205.08284,doi:10.1103/PhysRevD.106.124047

    work page doi:10.1103/physrevd.106.124047 2022

  18. [18]

    Patel, A

    N. Patel, A. Aykutalp, P. Laguna, Calculating Quasi- Normal Modes of Schwarzschild Black Holes with Physics Informed Neural Networks (1 2024).arXiv:2401.01440

    work page arXiv 2024

  19. [19]

    R. Luna, J. Calder´ on Bustillo, J. J. S. Mart´ ınez, A. Torres-Forn´ e, J. A. Font, Solving the Teukolsky equa- tion with physics-informed neural networks, Phys. Rev. D 107 (6) (2023) 064025.arXiv:2212.06103,doi:10. 1103/PhysRevD.107.064025

    work page arXiv 2023

  20. [20]

    A. S. Cornell, S. R. Herbst, A. M. Ncube, H. Noshad, Solving the Regge-Wheeler and Teukolsky equations: supervised versus unsupervised physics-informed neural networks (2 2024).arXiv:2402.11343

    work page arXiv 2024

  21. [21]

    A. M. Pombo, L. Pizzuti, Teukolsky by Design: A Hybrid Spectral-PINN solver for Kerr Quasinormal Modes (11 2025).arXiv:2511.15796

    work page arXiv 2025

  22. [22]

    R. Luna, D. D. Doneva, J. A. Font, J.-H. Lien, S. S. Yazadjiev, Quasinormal modes in modified gravity us- ing physics-informed neural networks, Phys. Rev. D 109 (12) (2024) 124064.arXiv:2404.11583,doi:10. 1103/PhysRevD.109.124064. [23]https://github.com/PaddlePaddle/PaddleScience/ tree/develop/examples/qnm(2025)

    work page arXiv 2024

  23. [23]

    L. Lu, P. Jin, G. Pang, Z. Zhang, G. E. Karni- adakis, Learning nonlinear operators via deeponet based on the universal approximation theorem of operators, Nature Machine Intelligence 3 (3) (2021). doi:10.1038/s42256-021-00302-5. URLhttp://dx.doi.org/10.1038/ s42256-021-00302-5

    work page doi:10.1038/s42256-021-00302-5 2021

  24. [24]

    P. Jin, S. Meng, L. Lu, Mionet: Learning multiple-input operators via tensor product (2022).arXiv:2202.06137. URLhttps://arxiv.org/abs/2202.06137

    work page arXiv 2022

  25. [25]

    S. Wang, H. Wang, P. Perdikaris, On the eigenvector bias of fourier feature networks: From regression to solv- ing multi-scale pdes with physics-informed neural net- works, Computer Methods in Applied Mechanics and En- gineering 384 (2021) 113938.doi:10.1016/j.cma.2021. 113938. URLhttp://dx.doi.org/10.1016/j.cma.2021.113938

    work page doi:10.1016/j.cma.2021 2021

  26. [26]

    S. Wang, A. K. Bhartari, B. Li, P. Perdikaris, Gradi- ent alignment in physics-informed neural networks: A second-order optimization perspective (2025).arXiv: 2502.00604. URLhttps://arxiv.org/abs/2502.00604

    work page arXiv 2025

  27. [27]

    A. Daw, J. Bu, S. Wang, P. Perdikaris, A. Karpatne, Mitigating propagation failures in physics-informed neu- ral networks using retain-resample-release (R3) sampling, in: A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, J. Scarlett (Eds.), Proceedings of the 40th International Conference on Machine Learning, Vol. 202 of Proceedings of Machine Le...

    2023

  28. [28]

    Gopakumar, S

    V. Gopakumar, S. Pamela, D. Samaddar, Loss land- scape engineering via data regulation on pinns, Machine Learning with Applications 12 (2023) 100464. doi:https://doi.org/10.1016/j.mlwa.2023.100464. URLhttps://www.sciencedirect.com/science/ article/pii/S2666827023000178

    work page doi:10.1016/j.mlwa.2023.100464 2023

  29. [29]

    Rathore, W

    P. Rathore, W. Lei, Z. Frangella, L. Lu, M. Udell, Chal- lenges in training pinns: A loss landscape perspective (2024).arXiv:2402.01868. URLhttps://arxiv.org/abs/2402.01868

    work page arXiv 2024

  30. [30]

    Z. Liu, Y. Wang, S. Vaidya, F. Ruehle, J. Halverson, M. Soljaˇ ci´ c, T. Y. Hou, M. Tegmark, Kan: Kolmogorov- arnold networks (2025).arXiv:2404.19756. URLhttps://arxiv.org/abs/2404.19756

    work page internal anchor Pith review arXiv 2025

  31. [31]

    Y. Wang, J. Sun, J. Bai, C. Anitescu, M. S. Eshaghi, X. Zhuang, T. Rabczuk, Y. Liu, Kolmogorov–arnold- informed neural network: A physics-informed deep learn- ing framework for solving forward and inverse problems based on kolmogorov–arnold networks, Computer Meth- ods in Applied Mechanics and Engineering 433 (2025) 12 117518.doi:10.1016/j.cma.2024.117518...

    work page doi:10.1016/j.cma.2024.117518 2025

  32. [32]

    S. Wang, B. Li, Y. Chen, P. Perdikaris, Piratenets: Physics-informed deep learning with residual adaptive networks (2024).arXiv:2402.00326. URLhttps://arxiv.org/abs/2402.00326

    work page arXiv 2024

  33. [33]

    Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Ragha- van, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T

    M. Tancik, P. P. Srinivasan, B. Mildenhall, S. Fridovich- Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. T. Barron, R. Ng, Fourier features let networks learn high frequency functions in low dimensional domains (2020). arXiv:2006.10739. URLhttps://arxiv.org/abs/2006.10739

    work page arXiv 2020

  34. [34]

    S. Wang, Y. Teng, P. Perdikaris, Understanding and mit- igating gradient pathologies in physics-informed neural networks (2020).arXiv:2001.04536. URLhttps://arxiv.org/abs/2001.04536

    work page arXiv 2020

  35. [35]

    S. J. Anagnostopoulos, J. D. Toscano, N. Stergiopulos, G. E. Karniadakis, Residual-based attention and connec- tion to information bottleneck theory in pinns (2023). arXiv:2307.00379. URLhttps://arxiv.org/abs/2307.00379

    work page arXiv 2023

  36. [36]

    Kolmogorov-Arnold Fou rier networks,

    J. Zhang, Y. Fan, K. Cai, K. Wang, Kolmogorov-arnold fourier networks (2025).arXiv:2502.06018. URLhttps://arxiv.org/abs/2502.06018

    work page arXiv 2025

  37. [37]

    Farea, M

    A. Farea, M. S. Celebi, Learnable activation functions in physics-informed neural networks for solving partial differential equations (2025).arXiv:2411.15111. URLhttps://arxiv.org/abs/2411.15111

    work page arXiv 2025

  38. [38]

    S. Wang, H. Wang, P. Perdikaris, Learning the solution operator of parametric partial differential equations with physics-informed deeponets, Science Advances 7 (40) (2021) eabi8605.arXiv:https:// www.science.org/doi/pdf/10.1126/sciadv.abi8605, doi:10.1126/sciadv.abi8605. URLhttps://www.science.org/doi/abs/10.1126/ sciadv.abi8605

    work page doi:10.1126/sciadv.abi8605 2021

  39. [39]

    B. P. Abbott, et al., Tests of general relativity with GW150914, Phys. Rev. Lett. 116 (22) (2016) 221101, [Erratum: Phys.Rev.Lett. 121, 129902 (2018)].arXiv: 1602.03841,doi:10.1103/PhysRevLett.116.221101

    work page doi:10.1103/physrevlett.116.221101 2016

  40. [40]

    A. G. Abac, et al., GW231123: A Binary Black Hole Merger with Total Mass 190–265 M⊙, Astrophys. J. Lett. 993 (1) (2025) L25.arXiv:2507.08219,doi:10.3847/ 2041-8213/ae0c9c

    work page arXiv 2025

  41. [41]

    Black Hole Spectroscopy and Tests of General Relativity with GW250114 (9 2025).arXiv:2509.08099

    work page arXiv 2025

  42. [42]

    Reitze, et al., Cosmic Explorer: The U.S

    D. Reitze, et al., Cosmic Explorer: The U.S. Contri- bution to Gravitational-Wave Astronomy beyond LIGO, Bull. Am. Astron. Soc. 51 (7) (2019) 035.arXiv:1907. 04833

    2019

  43. [43]

    Punturo, et al., The Einstein Telescope: A third- generation gravitational wave observatory, Class

    M. Punturo, et al., The Einstein Telescope: A third- generation gravitational wave observatory, Class. Quant. Grav. 27 (2010) 194002.doi:10.1088/0264-9381/27/ 19/194002

    work page doi:10.1088/0264-9381/27/ 2010

  44. [46]

    S. A. Teukolsky, Rotating black holes - separable wave equations for gravitational and electromagnetic pertur- bations, Phys. Rev. Lett. 29 (1972) 1114–1118.doi: 10.1103/PhysRevLett.29.1114

    work page doi:10.1103/physrevlett.29.1114 1972

  45. [47]

    S. A. Teukolsky, Perturbations of a rotating black hole

  46. [48]

    Fundamental equations for gravitational electromag- netic and neutrino field perturbations, Astrophys. J. 185 (1973) 635–647.doi:10.1086/152444

    work page doi:10.1086/152444 1973

  47. [49]

    E. W. Leaver, An Analytic representation for the quasi normal modes of Kerr black holes, Proc. Roy. Soc. Lond. A 402 (1985) 285–298.doi:10.1098/rspa.1985.0119

    work page doi:10.1098/rspa.1985.0119 1985

  48. [50]

    C. V. Vishveshwara, Scattering of Gravitational Radia- tion by a Schwarzschild Black-hole, Nature 227 (1970) 936–938.doi:10.1038/227936a0

    work page doi:10.1038/227936a0 1970

  49. [51]

    Chandrasekhar, S

    S. Chandrasekhar, S. L. Detweiler, The quasi-normal modes of the Schwarzschild black hole, Proc. Roy. Soc. Lond. A 344 (1975) 441–452.doi:10.1098/rspa.1975. 0112

    work page doi:10.1098/rspa.1975 1975

  50. [52]

    B. F. Schutz, C. M. Will, BLACK HOLE NORMAL MODES: A SEMIANALYTIC APPROACH, Astrophys. J. Lett. 291 (1985) L33–L36.doi:10.1086/184453

    work page doi:10.1086/184453 1985

  51. [53]

    S. Iyer, C. M. Will, Black Hole Normal Modes: A WKB Approach. 1. Foundations and Application of a Higher Order WKB Analysis of Potential Barrier Scattering, Phys. Rev. D 35 (1987) 3621.doi:10.1103/PhysRevD. 35.3621

    work page doi:10.1103/physrevd 1987

  52. [54]

    New approach to the quasinormal modes of a black hole.Phys

    V. Ferrari, B. Mashhoon, New approach to the quasi- normal modes of a black hole, Phys. Rev. D 30 (1984) 295–304.doi:10.1103/PhysRevD.30.295

    work page doi:10.1103/physrevd.30.295 1984

  53. [55]

    H. Yang, D. A. Nichols, F. Zhang, A. Zimmerman, Z. Zhang, Y. Chen, Quasinormal-mode spectrum of Kerr black holes and its geometric interpretation, Phys. Rev. D 86 (2012) 104006.arXiv:1207.4253,doi:10.1103/ PhysRevD.86.104006

    work page arXiv 2012

  54. [56]

    O’Toole, R

    C. O’Toole, R. Macedo, T. Stratton, B. Wardell, Quasi- normalmodes: A black hole perturbation toolkit package, https://github.com/BlackHolePerturbationToolkit/ QuasiNormalModes, a package for computing quasi- normal modes of Schwarzschild and Kerr black holes (2025)

    2025

  55. [57]

    Inc,https://www.paddlepaddle.org.cn/en(2025)

    B. Inc,https://www.paddlepaddle.org.cn/en(2025)

    2025

  56. [58]

    Baibhav and E

    V. Baibhav, E. Berti, Multimode black hole spectroscopy, Phys. Rev. D 99 (2) (2019) 024005.arXiv:1809.03500, doi:10.1103/PhysRevD.99.024005

    work page doi:10.1103/physrevd.99.024005 2019

  57. [59]

    L. C. Stein, qnm: A Python package for calculating Kerr quasinormal modes, separation constants, and spherical- spheroidal mixing coefficients, J. Open Source Softw. 4 (42) (2019) 1683.arXiv:1908.10377,doi:10.21105/ joss.01683

    work page arXiv 2019

  58. [60]

    Carullo, Black hole spectroscopy: status report, Gen

    G. Carullo, Black hole spectroscopy: status report, Gen. Rel. Grav. 57 (5) (2025) 76.doi:10.1007/ s10714-025-03408-y

    2025

  59. [61]

    K. D. Kokkotas, B. F. Schutz, W-modes: A New family of normal modes of pulsating relativistic stars, Mon. Not. Roy. Astron. Soc. 225 (1992) 119

    1992

  60. [62]

    Benhar, E

    O. Benhar, E. Berti, V. Ferrari, The Imprint of the equation of state on the axial w modes of oscillat- ing neutron stars, Mon. Not. Roy. Astron. Soc. 310 (1999) 797–803.arXiv:gr-qc/9901037,doi:10.1046/ j.1365-8711.1999.02983.x

    work page arXiv 1999