Machine learning for gravitational-wave detection: surrogate Wiener filtering for the prediction and optimized cancellation of Newtonian noise at Virgo

TL;DR

This paper introduces a machine-learning based surrogate Wiener filter to optimize seismometer array configurations for effective Newtonian noise cancellation in gravitational-wave detectors, enhancing sensitivity with fewer sensors.

Contribution

It presents a novel machine-learning approach to design optimal seismometer arrays for Newtonian noise cancellation in gravitational-wave detectors.

Findings

Efficient noise cancellation achievable with few sensors in optimal configurations.

The method is applicable to current and future gravitational-wave detectors.

Surrogate models improve the design of noise-cancellation systems.

Abstract

The cancellation of noise from terrestrial gravity fluctuations, also known as Newtonian noise (NN), in gravitational-wave detectors is a formidable challenge. Gravity fluctuations result from density perturbations associated with environmental fields, e.g., seismic and acoustic fields, which are characterized by complex spatial correlations. Measurements of these fields necessarily provide incomplete information, and the question is how to make optimal use of available information for the design of a noise-cancellation system. In this paper, we present a machine-learning approach to calculate a surrogate model of a Wiener filter. The model is used to calculate optimal configurations of seismometer arrays for a varying number of sensors, which is the missing keystone for the design of NN cancellation systems. The optimization results indicate that efficient noise cancellation can be achieved even for complex seismic fields with relatively few seismometers provided that they are deployed in optimal configurations. In the form presented here, the optimization method can be applied to all current and future gravitational-wave detectors located at the surface and with minor modifications also to future underground detectors.