Resampling to accelerate cross-correlation searches for continuous gravitational waves from binary systems

TL;DR

This paper introduces a resampling method that accelerates cross-correlation searches for continuous gravitational waves from binary systems, enabling longer integrations and improved sensitivity in less time.

Contribution

The paper presents a resampling technique that is approximately 20 times faster than previous methods, allowing more efficient and sensitive searches for gravitational waves from binary systems.

Findings

Resampling speeds up cross-correlation searches by 20x in key frequency bands.

Speed-up enables longer integration times, increasing sensitivity by up to 51%.

Method scales well with longer data sets, promising significant improvements in future searches.

Abstract

Continuous-wave (CW) gravitational waves (GWs) call for computationally-intensive methods. Low signal-to-noise ratio signals need templated searches with long coherent integration times and thus fine parameter-space resolution. Longer integration increases sensitivity. Low-mass x-ray binaries (LMXBs) such as Scorpius X-1 (Sco X-1) may emit accretion-driven CWs at strains reachable by current ground-based observatories. Binary orbital parameters induce phase modulation. This paper describes how resampling corrects binary and detector motion, yielding source-frame time series used for cross-correlation. Compared to the previous, detector-frame, templated cross-correlation method, used for Sco X-1 on data from the first Advanced LIGO observing run (O1), resampling is about 20x faster in the costliest, most-sensitive frequency bands. Speed-up factors depend on integration time and search setup. The speed could be reinvested into longer integration with a forecast sensitivity gain, 20 to 125 Hz median, of approximately 51%, or from 20 to 250 Hz, 11%, given the same per-band cost and setup. This paper's timing model enables future setup optimization. Resampling scales well with longer integration, and at 10x unoptimized cost could reach respectively 2.83x and 2.75x median sensitivities, limited by spin-wandering. Then an O1 search could yield a marginalized-polarization upper limit reaching torque-balance at 100 Hz. Frequencies from 40 to 140 Hz might be probed in equal observing time with 2x improved detectors.