Weakly modeled search for compact binary coalescences in the Einstein Telescope

A. Macquet 1, T. Dal Canton 1, T. Regimbau 2 · 2025

Read Weakly modeled search for compact binary coalescences in the Einstein Telescope by A. Macquet 1, T. Dal Canton 1, T. Regimbau 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-08377.txt.

Introduction

Adrian Macquet, Tito Dal Canton, and Tania Regimbau test a cheap way to find merging black holes and neutron stars in mock data, in a 2025 study. The data mimic the Einstein Telescope, a planned space-wave ear far more keen than those now in use. With so many finds due, the cost of the search itself could cap what teams can do.

Their search does not match signals to a set wave shape, so it runs cheap but sees less. It still caught 38 percent of black hole mergers in the mock run. It caught most heavy pairs and most near neutron star pairs, and it sized the pairs well. Loud signals broke the standard noise trick, so the team used a null stream, a blend with no true signal, to set the noise floor.

Catalog record

CREATORA. Macquet 1, T. Dal Canton 1, T. Regimbau 2
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE12 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-08377

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Weakly modeled search for compact binary coalescences in the Einstein Telescope A. Macquet 1, T. Dal Canton 1, T. Regimbau 2 Abstract We search for gravitational-wave (GW) signals from compact binary coalescences (CBC) in the 2024 2024 mock data challenge of the Einstein Telescope (ET) with a detection algorithm that does not rely on the waveform of the signal searched. With the increased sensitivity of ET compared to current GW detectors, a very high rate of detectable sources is expected in the data, and the computational cost of the searches may become a limiting factor. This is why we explore the behavior of a weakly modeled search algorithm, which is intrinsically less sensitive than optimal search methods based on matched filtering techniques, but computationally much cheaper. This search recovers a significant fraction of CBC signals present in the data: 38 % 38\% of the total number of binary black hole mergers, including 89 % 89\% of the systems with a total mass above 100 100 M ⊙ , as well as the majority of binary neutron star (BNS) mergers closer than 850 850 Mpc ( z = 0.17 z=0.17 ). It is also able to estimate the chirp mass of the recovered BNS with an average precision of 1.3 % 1.3\% . We also find that the usual method for estimating the background in transient GW searches, that consists in time-shifting the data from one detector with respect to the others, is impacted by the presence of loud CBC signals in the data, so we use the null stream instead as a signal-free channel to estimate the background of this search. 1 Introduction and motivation Expected in the mid- 2030 s, the Einstein Telescope (ET) [ 1 ] will be a ground-based gravitational wave (GW) detector which aims to increase the sensitivity by an order of magnitude compared to Advanced LIGO and Advanced Virgo in a frequency range between a few Hz to few kHz. Its current planed design consists of three nested interferometers forming an equilateral triangle with 10 km sides. Each detector will consist of two dual-recycled Fabry-Perot-Michelson interferometers tuned to be sensitive to high and low frequencies respectively, and the whole infrastructure will be built underground to limit the impact of seismic noise. At its design sensitivity, ET should be able to detect most of the binary black holes mergers (BBH) with total mass in the range 10^{1}-10^{3} M ⊙ up to a redshift z\sim 20 , and binary neutron stars mergers (BNS) up to z\sim 2 , allowing to probe the distribution of compact binary coalescences (CBC) at large redshift, among many other scientific objectives (a comprehensive description of the scientific targets of ET can be found in [ 2 ] ). In order to reach these objectives, current data analysis techniques used for LIGO/Virgo data must be adapted to deal with a significantly increased signal regime: the expected rate of detectable CBC in ET lies in the hundreds to thousands per day, compared to \sim 1 per day in Advanced LIGO at its design sensitivity. In addition, the extended sensitivity at low frequency means that CBC signals will be present for a longer time in the frequency band of the detectors, up to several hours or days in the case of BNS. Because CBC signals are well modeled and depend on a relatively small number of parameters ( 16 in the most general case), they can be searched nearly optimally in the data using matched filtering techniques, that consist in cross-correlating templates of the signals searched with the data. Conversely, when a signal is not well enough modeled to be searched with matched filtering techniques, one must rely on weakly modeled search algorithms that make no or few assumptions on the exact waveform of the signal searched. Such methods typically rely on looking for patterns of excess power in some time-frequency representation of the data. By construction, modeled searches are more sensitive to CBC signals than weakly modeled ones, but they are also much more computationally expensive because a large number of templates must be generated and cross-correlated with the data to cover the parameter space of CBC. In ET, that cost will further increase because of the large number of signals in the data, but also because of their extended duration, which will require much longer templates that are expensive to handle, up to a point that may become prohibitive. On the other hand, the improved sensitivity means that a large number of signals will have a high signal to noise ratio (SNR) and may therefore be detectable by weakly modeled searches. Here we study the ability of a weakly modeled search algorithm to recover CBC signals in ET. We run PySTAMPAS , a detection algorithm designed to search for long-duration ( \sim 1-1000 s) transient GW signals [ 3 ] , on the 2024 ET mock data challenge (MDC). This work is described in more details in [ 4 ] . 2 Dataset and methodology 2.1 ET mock data challenge The ET collaboration regularly issues MDCs in order to provide common datasets to test and demonstrate data analysis methods, and to precise the scientific objectives of the detector. The MDC analyzed in this work simulates 31 days of ET data for the 3 nested detectors forming the triangle. The data consist of Gaussian noise following the design sensitivity of ET, on top of which are injected \sim 70000 CBC signals following realistic astrophysical distributions compatible with the latest LVK observations [ 5 ] . Their distribution in total mass and redshift is represented in Fig. 1 . 2.2 Data analysis method PySTAMPAS is a data analysis pipeline originally designed to search for long-duration ( \sim 1-1000 s) transient GW signals in LIGO/Virgo data with minimal assumptions on the morphology of the signal. First, time-frequency maps (TF-maps) of the data are built by computing the short-time Fourier Transform of short segments of duration 0.5 , 1 , 2 , and 4 s, which are then combined into a multi-resolution TF-map. Each map spans a duration of 512 s and a frequency range 4-2000 Hz. Candidate GW signals ( triggers ) are then identified by running a pattern recognition algorithm on the individual TF-maps. At this step, loose constraints may be imposed on the time-frequency morphology of the signal searched (hence the search being described as weakly modeled). Here we explore two complementary strategies. First, we run a seed-based clustering algorithm that identifies bright pixels in the TF-map and groups them by proximity. This algorithms allows to reconstruct any signal morphology (provided it is contiguous in the time-frequency plane), but requires high SNR pixels. For long-lived BNS signals that may last up to several hours, the energy is spread over a large number of pixels in the TF-map, so individual pixels may not have a sufficiently high SNR to be picked up by the seed-based algorithm. Therefore we also use a seedless clustering algorithm, whose principle is to fit pre-determined time-frequency templates on each TF-map. Here we fit Newtonian chirp templates, characterized by their chirp mass ( 100 values between 1 and 5 M ⊙ ) and coalescence time (every 0.5 s). As a by-product,

Read the complete text