Computational cost for detecting inspiraling binaries using a network of laser interferometric detectors

TL;DR

This paper extends a coherent network data-analysis method to post-Newtonian waveforms for inspiraling binaries, estimating the computational costs for different detector networks and comparing with coincidence detection strategies.

Contribution

It introduces a method for analyzing post-Newtonian waveforms in a network of detectors and estimates the associated computational costs, including realistic detector configurations.

Findings

Computational costs range from a few Gflops for single detectors to tens of Tflops for three-detector networks.

The number of templates increases with the inclusion of post-Newtonian parameters.

Coherent detection strategies have different computational demands compared to coincidence methods.

Abstract

We extend a coherent network data-analysis strategy developed earlier for detecting Newtonian waveforms to the case of post-Newtonian (PN) waveforms. Since the PN waveform depends on the individual masses of the inspiraling binary, the parameter-space dimension increases by one from that of the Newtonian case. We obtain the number of templates and estimate the computational costs for PN waveforms: For a lower mass limit of a solar mass, for LIGO-I noise, and with 3% maximum mismatch, the online computational speed requirement for single detector is a few Gflops; for a two-detector network it is hundreds of Gflops and for a three-detector network it is tens of Tflops. Apart from idealistic networks, we obtain results for realistic networks comprising of LIGO and VIRGO. Finally, we compare costs incurred in a coincidence detection strategy with those incurred in the coherent strategy detailed above.