pith. sign in

arxiv: 2512.20699 · v2 · submitted 2025-12-23 · 🌀 gr-qc · astro-ph.HE

Comparing next-generation detector configurations for high-redshift gravitational wave sources with neural posterior estimation

Filippo Santoliquido , Jacopo Tissino , Ulyana Dupletsa , Marica Branchesi , Jan Harms This is my paper

Pith reviewed 2026-05-16 19:58 UTC · model grok-4.3

classification 🌀 gr-qc astro-ph.HE
keywords gravitational wavesEinstein Telescopeneural posterior estimationhigh-redshift binary black holesdetector networkssky localizationmultimodal posteriorsCosmic Explorer
0
0 comments X

The pith

Two misaligned L-shaped Einstein Telescope detectors reduce sky and volume localization ambiguities for high-redshift black hole mergers compared to the triangular configuration, even as distance estimates become more complex.

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

This paper compares network layouts for next-generation gravitational-wave detectors by applying neural posterior estimation to simulated signals from massive, short-duration binary black holes at high redshift. It finds that a pair of misaligned L-shaped detectors produces luminosity-distance posteriors that can mix with sky position, yielding somewhat broader distance uncertainties than the triangular Einstein Telescope design. At the same time, the misaligned layout generates far fewer sky-position multimodalities than the eight expected from the triangle, which improves overall three-dimensional localization. Adding a Cosmic Explorer detector to either network further suppresses position degeneracies while the relative advantage of the misaligned pair persists.

Core claim

Validation against standard Bayesian sampling shows that neural posterior estimation accurately recovers complex, disconnected posterior structures for all tested networks. For the two-misaligned-L-shaped configuration the distance posteriors become multimodal and degenerate with sky location, producing less precise distance estimates than the triangular layout. However, the number of sky-location modes drops substantially below the eight modes characteristic of the triangle, which yields improved sky and volume localization. Inclusion of Cosmic Explorer in the network reduces these degeneracies further and preserves the localization benefit of the misaligned pair.

What carries the argument

Neural posterior estimation implemented with normalizing flows and importance sampling, which rapidly infers full posterior distributions, including multimodal and degenerate structures, for gravitational-wave signals observed by different detector-network geometries.

Load-bearing premise

The simulated population of short, massive, high-redshift binary black holes with detector-frame chirp mass above 100 solar masses faithfully represents the signals that will actually be detected, and the neural posterior estimation method reproduces the true posterior structure without systematic bias.

What would settle it

Detection of a real high-redshift massive binary black hole merger by an actual Einstein Telescope network followed by standard sampling that shows a sky-multimodality count near eight for the triangular layout and near zero for the misaligned L-shaped layout.

Figures

Figures reproduced from arXiv: 2512.20699 by Filippo Santoliquido, Jacopo Tissino, Jan Harms, Marica Branchesi, Ulyana Dupletsa.

Figure 1
Figure 1. Figure 1: Top panel: Injections as function of sample effi￾ciency (ϵ, x-axis) and optimal SNR (ρ, y-axis) for each detec￾tor configuration (color-coded). Filled markers indicate in￾jections with sample efficiencies > 1%. Percentages in paren￾theses denote the fraction of sources with sample efficiency above 1%, while the horizontal colored lines indicate the me￾dian ρ for sources exceeding this threshold. Bottom pan… view at source ↗
Figure 2
Figure 2. Figure 2: also shows that Dingo-IS accurately cap￾tures the complex structure of multimodal posteriors. These multimodalities include both the eight-fold sky degeneracy arising from the triangular geometry of ET, extensively discussed in F. Santoliquido et al. (2025), [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Cumulative distributions of events as a function of the information gain (top-left panel, see also Section 2.3), the relative variations in luminosity distance (top-right panel, ∆DL/Dinj L ), sky (bottom-left panel, ∆Ω90%, see also [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Percentage of sky maps (y-axis) with ∆Ω90% ≤ 100 deg2 exhibiting one to eight or more discon￾nected modes (color-coded), for all detector configurations (x-axis). See Section 3.3 for details. observable with XG detectors (M. Branchesi et al. 2023; A. Abac et al. 2025), it targets a regime with strong potential for breakthrough discoveries. We assess the performance of seven configurations of XG detectors u… view at source ↗
Figure 5
Figure 5. Figure 5: Amplitude spectral densities (ASDs) adopted in this work for different XG detectors. See Section 2.1 and Appendix A for details. A. DETECTOR AND NETWORK CONFIGURATIONS We provide further details on the detector and network configurations considered in this study. We limit the frequency range from fmin = 6 Hz to fmax = 256 Hz, using 8-second time segments, which results in frequency bins of size df = 0.125 … view at source ↗
Figure 6
Figure 6. Figure 6: The log loss as a function of training epochs shown for all detector configurations considered in this study, with the solid line representing the training set and the dashed line representing the test set. See Appendix C for further details. et al. 2015; S. Talts et al. 2020; S. R. Green et al. 2020), by verifying whether the fraction of injected parameters recovered within a given credible interval follo… view at source ↗
Figure 7
Figure 7. Figure 7: Probability–probability (P–P) plots showing the confidence interval (p) on the x-axis versus the difference between the observed fraction of events within that interval and the interval itself (CDF(p) − p) on the y-axis, for each GW parameter. Posteriors are obtained with Dingo (without importance sampling) from 1000 injected binary black hole (BBH) signals sampled from the prior ( [PITH_FULL_IMAGE:figure… view at source ↗
Figure 8
Figure 8. Figure 8: Marginalized one- and two-dimensional posterior distributions for all parameters for the event shown in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Marginalized one- and two-dimensional posterior distributions for all parameters of the same event shown in [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Percentage of luminosity distance posteriors (y-axis) that are multimodal (dark blue) or unimodal (light blue) for all detector configurations (x-axis). See Appendix F for details. and bottom panel of [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Marginalized one- and two-dimensional posterior distributions for all parameters, with the exception that this time we show the source-frame chirp mass (Ms) and redshift (z). The results for the low-mass event (orange) are shown alongside those for the high-mass event (blue). Vertical and horizontal lines indicate the true injected values and contours represent the 68% and 95% credible regions. Top-right … view at source ↗
Figure 12
Figure 12. Figure 12: Cumulative distributions of events as a function of the optimal SNR (ρ, Equation D8), the relative variations in detector-frame chirp mass (∆Md/Minj d ), mass ratio (q), first (∆χ1) and second (∆χ2) aligned spins, and inclination angle (∆θjn) for all considered XG detector configurations. See Appendix H for details. Bosi, M., Bellomo, N., & Raccanelli, A. 2023, JCAP, 11, 086, doi: 10.1088/1475-7516/2023/1… view at source ↗
read the original abstract

The coming decade will be crucial for determining the final design and configuration of a global network of next-generation (XG) gravitational-wave detectors, including the Einstein Telescope (ET) and Cosmic Explorer (CE). In this study, and for the first time, we assessed the performance of various network configurations using neural posterior estimation (NPE) implemented in Dingo-IS-a method based on normalizing flows and importance sampling that enables fast and accurate inference. We focused on a specific science case involving short-duration, massive and high-redshift binary black hole mergers with detector-frame chirp masses $> 100~\mathrm{M}_\odot$. These systems encompass early-Universe stellar and primordial black holes, as well as intermediate-mass black hole binaries, for which XG observatories are expected to deliver major discoveries. Validation against standard Bayesian inference demonstrates that NPE robustly reproduces complex and disconnected posterior structures across all network configurations. For a network of two misaligned L-shaped ET detectors (2L MisA), the posterior distributions on luminosity distance can become multimodal and degenerate with the sky position, leading to less precise distance estimates compared to the triangular ET configuration. However, the number of sky-location multimodalities is substantially lower than the eight expected with the triangular ET, resulting in improved sky and volume localization. Adding CE to the network further reduces sky-position degeneracies, and the better performance of the 2L MisA configuration over the triangle remains evident.

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 / 2 minor

Summary. The manuscript evaluates next-generation gravitational-wave detector networks (primarily Einstein Telescope configurations with and without Cosmic Explorer) for parameter estimation of short-duration, high-redshift massive binary black hole mergers (detector-frame chirp mass >100 M_⊙) using neural posterior estimation via Dingo-IS. It validates NPE against standard Bayesian inference for reproducing complex multimodal posteriors and reports that a two-misaligned-L-shaped ET network (2L MisA) yields fewer sky-location modes than the triangular ET (reducing from eight expected modes), improving sky and volume localization despite increased distance multimodality; adding CE further mitigates degeneracies.

Significance. If the central claims hold, the work is significant for guiding the final design of the ET and global XG network, as it quantifies configuration-specific trade-offs in handling degeneracies for a high-priority science case involving early-universe and intermediate-mass black holes. The use of NPE to enable rapid, accurate inference on disconnected posteriors across multiple networks is a methodological strength that supports efficient design studies.

major comments (2)
  1. [Validation section (abstract and §4)] Abstract and validation section: The claim that NPE 'robustly reproduces complex and disconnected posterior structures across all network configurations' is central to the 2L MisA vs. triangular ET comparison, yet it is unclear whether the validation set against standard Bayesian inference included the specific high-mass, high-redshift signals that produce the distance-sky degeneracies unique to the misaligned L-shaped pair. Without explicit confirmation that these cases were tested (including quantitative metrics such as mode recovery rates or KL divergence on multimodal distance posteriors), the reported reduction in sky multimodalities and localization improvement cannot be fully assessed.
  2. [§5 (results on 2L MisA)] Results on localization performance (abstract and §5): The statement that 2L MisA produces 'substantially lower' sky-location multimodalities than the eight expected for triangular ET, yielding improved volume localization, is load-bearing for the headline conclusion. This would be strengthened by reporting the actual distribution of mode counts (or fraction of events with reduced modes) and quantitative localization volumes (e.g., 90% credible volume statistics) across the simulated population rather than qualitative comparison alone.
minor comments (2)
  1. [Methods] The methods section should explicitly state the redshift and mass distribution assumptions for the simulated high-z BBH population to allow readers to evaluate how representative the results are of expected XG detections.
  2. [Figure captions] Figure captions describing posterior examples for different networks could include the number of events shown and the specific degeneracy features highlighted to improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback. We address each major comment below and have updated the manuscript with additional details and quantitative results to strengthen the validation and localization claims.

read point-by-point responses

  1. Referee: Abstract and validation section: The claim that NPE 'robustly reproduces complex and disconnected posterior structures across all network configurations' is central to the 2L MisA vs. triangular ET comparison, yet it is unclear whether the validation set against standard Bayesian inference included the specific high-mass, high-redshift signals that produce the distance-sky degeneracies unique to the misaligned L-shaped pair. Without explicit confirmation that these cases were tested (including quantitative metrics such as mode recovery rates or KL divergence on multimodal distance posteriors), the reported reduction in sky multimodalities and localization improvement cannot be fully assessed.

    Authors: The validation set was drawn from the identical population of high-mass, high-redshift signals (detector-frame chirp mass >100 M_⊙) used for the main results, explicitly including events that exhibit the distance-sky degeneracies characteristic of the 2L MisA configuration. We have revised §4 to state this explicitly and added quantitative metrics: mode recovery rates (fraction of events where all posterior modes are recovered within 1σ of the true values) and average KL divergence between NPE and standard Bayesian posteriors for the multimodal distance marginals, confirming comparable performance across all networks. revision: yes

  2. Referee: Results on localization performance (abstract and §5): The statement that 2L MisA produces 'substantially lower' sky-location multimodalities than the eight expected for triangular ET, yielding improved volume localization, is load-bearing for the headline conclusion. This would be strengthened by reporting the actual distribution of mode counts (or fraction of events with reduced modes) and quantitative localization volumes (e.g., 90% credible volume statistics) across the simulated population rather than qualitative comparison alone.

    Authors: We agree that quantitative reporting strengthens the result. The revised §5 now includes: (i) histograms of the number of sky-location modes recovered for each configuration, (ii) the fraction of events for which 2L MisA reduces the mode count relative to the triangular ET (typically from 8 to 2–4), and (iii) median and 90% credible intervals on the 90% sky-area and 3D volume localization across the full simulated population, confirming the improvement in volume localization despite the distance multimodality. revision: yes

Circularity Check

0 steps flagged

No circularity: forward simulation and NPE inference remain independent of reported metrics

full rationale

The paper generates simulated high-redshift BBH signals with detector-frame chirp mass >100 M⊙, runs NPE (Dingo-IS) to obtain posteriors for each network configuration, and compares quantities such as number of sky-location modes and volume localization. Validation is performed by direct comparison to standard Bayesian inference on the same simulated data, with no parameter fitted from the target metrics and then re-used as a prediction. No self-citation supplies a uniqueness theorem or ansatz that forces the central claims; the reported advantage of 2L MisA (fewer modes than the triangular ET's eight) follows from the explicit posterior samples rather than any definitional reduction or renaming of a known result.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claims rest on standard assumptions of gravitational-wave signal modeling and simulation; no new free parameters, axioms, or invented entities are introduced in the abstract.

pith-pipeline@v0.9.0 · 5575 in / 1171 out tokens · 24907 ms · 2026-05-16T19:58:19.954563+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Robust parameter inference for Taiji via time-frequency contrastive learning and normalizing flows

    gr-qc 2026-04 unverdicted novelty 6.0

    A glitch-robust amortized inference framework combining normalizing flows, time-frequency multimodal fusion, and contrastive learning outperforms MCMC for Taiji massive black hole binary parameter estimation under noi...

  2. Not too close! Evaluating the impact of the baseline on the localization of binary black holes by next-generation gravitational-wave detectors

    gr-qc 2026-04 conditional novelty 4.0

    Baselines of 8-11 ms light travel time for two CE detectors provide a reasonable compromise for BBH sky localization, with third detectors eliminating multimodality for most or all events.

Reference graph

Works this paper leans on

6 extracted references · 6 canonical work pages · cited by 2 Pith papers · 3 internal anchors

  1. [1]

    The Science of the Einstein Telescope

    Abac, A., et al. 2025, arXiv. https://arxiv.org/abs/2503.12263 Abac, A. G., et al. 2025, Phys. Rev. Lett., 135, 111403, doi: 10.1103/kw5g-d732 Aghanim, N., et al. 2020, Astron. Astrophys., 641, A6, doi: 10.1051/0004-6361/201833910 All´ en´ e, C., et al. 2025, arXiv. https://arxiv.org/abs/2501.04598 Antonini, F., Gieles, M., & Gualandris, A. 2019, Mon. Not...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/kw5g-d732 2025

  2. [2]

    Fast Fisher Matrices and Lazy Likelihoods

    aligned spins, and inclination angle (∆θjn) for all considered XG detector configurations. See Appendix H for details. Bosi, M., Bellomo, N., & Raccanelli, A. 2023, JCAP, 11, 086, doi: 10.1088/1475-7516/2023/11/086 Branchesi, M., et al. 2023, JCAP, 07, 068, doi: 10.1088/1475-7516/2023/07/068 Buchner, J. 2022, Research Notes of the American Astronomical So...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/1475-7516/2023/11/086 2023

  3. [3]

    http://www.jstor.org/stable/2241144 Hartigan, P. M. 1985, Journal of the Royal Statistical Society. Series C (Applied Statistics), 34,

    work page arXiv 1985

  4. [4]

    2016, Mon

    http://www.jstor.org/stable/2347485 Hartwig, T., Volonteri, M., Bromm, V., et al. 2016, Mon. Not. Roy. Astron. Soc., 460, L74, doi: 10.1093/mnrasl/slw074 Hawking, S. W. 1974, Nature, 248, 30, doi: 10.1038/248030a0 Heinzel, J., & Vitale, S. 2025, arXiv. https://arxiv.org/abs/2509.07221 Hu, Q., & Veitch, J. 2025, Phys. Rev. D, 112, 084039, doi: 10.1103/dj7k...

    work page doi:10.1093/mnrasl/slw074 2016

  5. [5]

    2024, Phys

    https://d3qi0qp55mx5f5.cloudfront.net/stat/docs/tech- rpts/tr348.pdf Kritos, K., Strokov, V., Baibhav, V., & Berti, E. 2024, Phys. Rev. D, 110, 043023, doi: 10.1103/PhysRevD.110.043023 Kullback, S., & Leibler, R. A. 1951, The Annals of Mathematical Statistics, 22, 79 , doi: 10.1214/aoms/1177729694 Lanchares, D., Freitas, O. G., Gonz´ alez-Nuevo, J., & Fon...

    work page doi:10.1103/physrevd.110.043023 2024

  6. [6]

    Cosmic Explorer: The U.S. Contribution to Gravitational-Wave Astronomy beyond LIGO

    https://arxiv.org/abs/1907.04833 Rizzuto, F. P., Naab, T., Rantala, A., & et al. 2023, Mon. Not. R. Astron. Soc., 521, 2930, doi: 10.1093/mnras/stad734 22F. Santoliquido et al. Romano, J. D., & Cornish, N. J. 2017, Living Rev. Rel., 20, 2, doi: 10.1007/s41114-017-0004-1 Romero-Shaw, I. M., et al. 2020, Mon. Not. Roy. Astron. Soc., 499, 3295, doi: 10.1093/...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stad734 1907