Skip to content

Pre-localization of Massive Black Hole Binaries in the Millihertz Band

Home / Papers / Pre-localization of Massive Black Hole Binaries in the Millihertz Band

Pre-localization of Massive Black Hole Binaries in the Millihertz Band

Authors: Xue-Ting Zhang, Jonathan Gair, Chris Messenger, Natalia Korsakova, Yi-Ming Hu, Hong-Yu Chen Date: 2026-04-27 Paper ID: openalex:2604.24330

Summary

This paper addresses the challenge of providing low-latency sky localization for massive black hole binaries (MBHBs) in space-borne gravitational-wave detectors. The authors propose an amortized Bayesian inference pipeline that leverages learned time-series embeddings and neural spline flows (NSF) to achieve rapid parameter estimation. The method successfully provides accurate pre-merger localizations, enabling timely electromagnetic follow-up within a one-minute analysis window.

Key Contributions

  • Introduces a normalizing flow-based inference pipeline using neural spline flows for low-latency gravitational wave parameter estimation.
  • Achieves ~20 deg² pre-merger sky localization for massive black hole binaries approximately 15 minutes before coalescence.
  • Demonstrates inference speed and accuracy comparable to traditional parallel-tempered Markov chain Monte Carlo (PTMCMC) methods while operating within a one-minute per-event latency budget.

Open Questions & Future Work

Key Concepts

  • amortized-bayesian-inference-gw: A framework that uses learned neural embeddings coupled with normalizing flows to replace traditional MCMC sampling for real-time parameter estimation of gravitational wave signals.

Archivist Review

I approved the concept of ‘Amortized Bayesian Inference for GW Parameter Estimation’ as it represents a significant methodological shift in signal processing for real-time scientific inference. The open question on ‘source confusion’ was approved as a critical, non-trivial bottleneck for future multi-messenger observation architectures, whereas the original candidate was too broad and focused on general incremental precision improvements.

Approved Concepts

  • Amortized Bayesian Inference for GW Parameter Estimation: This approach shifts the computational burden from inference time to training time, which is the key enabler for real-time multi-messenger astronomy pipelines.

Approved Open Questions

  • MBHB Source Confusion Bottleneck: Addressing source confusion is a foundational requirement for the viability of future space-borne detectors like LISA.

Rejected Candidates

  • [open_question] Improving MBHB Inference Precision (improving-mbhb-inference-precision-and-robustness) - other: The candidate is partially conflated with general performance improvements; the proposed slug focuses more clearly on the specific, critical bottleneck of source confusion in multi-source environments.

Metadata & Links

url
https://arxiv.org/abs/2604.24330
paper_id
2604.24330
paper_source
openalex
domain
nlp
tags
architectures
datasets
concept_slugs
amortized-bayesian-inference-gw
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-04-30T06:04:52Z