Deep Learning Based Classification and Parameter Estimation of Gravitationally Lensed Gravitational Waves

Afroz, Saqlain (2025) Deep Learning Based Classification and Parameter Estimation of Gravitationally Lensed Gravitational Waves. Masters thesis, Indian Institute of Science Education and Research Kolkata.

Text (MS Dissertation of Saqlain Afroz (20MS230)) 20MS230_Thesis_file.pdf - Submitted Version Restricted to Repository staff only Download (3MB)

Abstract

Since centuries, we humans have tried to probe cosmos using telescopes in the visible light domain of electromagnetic spectrum, and in recent years, we have seen advancements in radio astronomy, infrared astronomy, X-ray astronomy, and gamma-ray astronomy, each delivering unique information about celestial objects and cosmic phenomena. Similar to EM waves, GWs can also provide significant information about some of the extreme phenomena in the Universe, as these waves are highly non-interacting with the intervening matter. But the problem with detection of gravitational waves are that they are very weak and there is also the presence of different kinds of noise in the signal. In the past few years scientific and technological advancements have made it possible to detect GWs, using LIGO-Virgo-Kagra detectors and we have detected ∼ 200 events as of the current ongoing O4 observing run [1]. As detectors achieve greater sensitivity, the signal-to-noise ratio increases and we can actually probe to larger distances in the Universe. This also increases the probability of the GWs getting gravitationally lensed, which is an effect where the trajectory of GWs seem to bend as it passes through the vicinity of a heavy object. The GWs that get gravitationally lensed, can get (de-)amplified, there can be some amplitude and phase modulations and there can be multiple copies of the same signal arriving at different time delays. The lensing of gravitational waves help us map the dark matter distribution across the Universe, can help us put constraints on alternate theories of gravity, and measure Hubble’s constant. This study addresses the methodology used to do accurate and fast classification of different kinds of GW signals, i.e. unlensed, precessing, eccentric and lensed and the critical need for parameter estimation of gravitationally lensed GWs to avoid biases introduced by neglecting lensing effects [2], and faster EM follow-up. For simplicity, we have taken the case of Point mass lens case, where we consider the GWs to be microlensed, and we need to work in the wave optics regime.

Actions (login required)

View Item