interTwin Use Case

A Digital Twin to simulate 'noise' in the Virgo Gravitational Wave interferometer

Creating a Digital Twin to realistically simulate 'noise' in the detector and study how it reacts to external disturbances. Noise reduction and subtraction will significantly increase Virgo capability for producing early and reliable alerts to other observatories.

Challenge

The discovery of gravitational waves (GW) was one of the main scientific results in recent years and was awarded the Nobel Prize in 2017. After that first detection, many other events have then been observed by the Advanced Virgo and Advanced LIGO interferometers.

Besides getting ready for their next observation runs, the gravitational-wave community is designing a next-generation observatory, the Einstein Telescope.

But the sensitivity of GW interferometers is limited by noise. Noise reduction and subtraction is one of the most important and challenging activities in GW research. The Digital Twin (DT) of an interferometer is meant to realistically simulate the noise in the detector, in order to study how it reacts to external disturbances and, in the perspective of the Einstein Telescope, to be able to detect noise “glitches” in quasi-real time, which is currently not possible. This will allow the low-latency search pipelines to veto or de-noise the signal, sending out more reliable triggers to observatories for multi-messenger astronomy.

Solution

The detection of gravitational waves is based on the measurement of the deformation of the interferometer arms, also known as strain.

The status of the various interferometer subsystems as well as the environmental conditions (wind, temperature, seismic motions) are recorded by several sensors, and stored in the so-called auxiliary channels. Transient noise, such as seismic activity, is recorded in both the auxiliary channels and the strain.

With the help of generative models such as GANs (Generative Adversarial Networks) we can generate simulated data in the strain channel starting from the auxiliary channels.

Proposed workflow

The realistic simulation of noise in the strain channel will allow INFN to build the Digital Twin (DT) of the interferometer.

The generated strain signal may be compared with the measured strain, and be used to veto events with a glitch. In a more advanced phase of the project, we can use the same generated signal for denoising rather than vetoing, i.e. the transient noise may be subtracted from the measured strain to allow for GW detection, even during a glitch in the interferometer (for example, due to on-going seismic activity).

The DT will be used directly in low-latency searches, in order to produce a first prototype of a low-latency noise analyser for veto generation (that can be used to stop processing for data segments that contain a large glitch, for example) and online de-noising of the signal.

In this way, we will significantly increase Virgo capability for producing early and reliable alerts to other observatories.

Background

What are gravitational waves?

In General Relativity, gravity can be interpreted as a consequence of the shape of spacetime: the presence of a massive body deforms the spacetime around it. Light travels in straight lines in flat spacetime, but close to a massive object light will also follow the curvature of the spacetime around the massive object. Gravitational waves are ripples in spacetime emitted in a wide range of frequencies and amplitudes by different kinds of astrophysical systems. When accelerated motion of mass and/or energy takes place without spherical or cylindrical symmetry, a GW is generated and propagates at the speed of light. Two black holes or neutron stars orbiting quickly around each other, or rotating neutron stars, may emit gravitational waves powerful enough to be detected.

On Earth, the passage of a gravitational wave will alternatively stretch and squeeze the distance between two massive objects. For example, the gravitational wave generated by the merging of two neutron stars in a galaxy close to ours will stretch the distance Earth-Sun (150 million of km) by the size of an atom. Such tiny distance variations may be measured by interferometers such as Virgo, located at Cascina, near Pisa in Italy and Ligo, located in two sites, the LIGO Hanford Observatory in eastern Washington and the LIGO Livingston Observatory, in Louisiana.

Binary Neutron Star merger simulation (Credits: EGO and The Virgo Collaboration)

Background

Multi Messenger Astronomy

The observation of GWs paved the way to Multi Messenger Astronomy (MMA), i.e. the observation of astronomical events by combining information by different “messengers”: photons, gravitational waves, neutrinos and/or cosmic rays. Events suitable to the observation in both the electromagnetic and gravitational domain, for example the merger of neutron stars, are very rare and often the localization of the source in the sky is poorly constrained. This, together with the short duration of the gamma ray burst, make this type of observation particularly complex.

“Low-latency searches” are meant to alert observatories early enough for them to point their telescopes and/or to decide whether to keep data in the relevant time interval around the event. In the current observation run (O4), that Virgo is foreseen to join in late 2023, the first alerts will be sent out in less than few tens of seconds after the possible GW signal is detected in the interferometer. Updates with improved sky localisation will follow a few minutes later. A challenge for next and future low-latency GW searches will be the detection of binary neutron stars signals early enough to generate the alert before the actual merger phase.