Research Projects

Gamma-ray bursts (GRBs) are the most luminous and distant transients in the universe, offering crucial insights into relativistic astrophysics, compact object mergers, stellar collapse, and extreme radiation processes. Despite significant observational advances, key questions remain unresolved — including the nature of the central engine (black hole vs. magnetar), dominant radiation mechanisms, and viewing angles. This proposal addresses these challenges through two complementary computational approaches: (1) Magneto-hydrodynamic (MHD) simulations of relativistic jets with polarimetric radiation modeling, and (2) Deep learning–based analysis of GRB light curves as multivariate time series. Polarimetric observations are highly sensitive to jet structure, magnetic field geometry, and emission mechanisms. Instruments like AstroSat’s CZTI have demonstrated that GRB polarimetry is feasible, yet interpretation requires robust simulation frameworks. Currently, there is a critical gap in modeling time-resolved spectro-polarimetric signatures grounded in first-principles radiation-hydrodynamic simulations. Globally, such modeling is in its early stages, and in India, there has been no significant development on this front. A vast database of GRB light curves from missions like Fermi, Swift, and BATSE remains underutilised for extracting characteristic features, periodicities, or repetitions, as manual analysis is impractical. Applying machine learning and deep learning to these multivariate time-series datasets offers a powerful top-down approach to uncover hidden patterns and insights into GRB physics — an area that remains significantly under-explored both globally and in India. The scientific objectives of this project are: - To simulate the dynamical evolution of relativistic GRB jets using MHD models coupled with radiative transfer, producing synthetic time-resolved spectro-polarimetric light curves. - To explore how jet structure, magnetic field configuration, and viewing geometry influence polarisation. - To apply data-driven deep learning methods to uncover hidden patterns in GRB light curves. The hypotheses being tested include: - Whether time-evolving polarisation can distinguish between synchrotron and photospheric emission mechanisms. - Whether QPOs in GRB light curves indicate magnetar-driven engines or oscillatory processes in the jet. - Whether gravitational lensing events in GRBs can be identified via time-series similarity detection, offering probes of cosmological lensing. - Whether variabilities observed in the light curves decipher details of central engines and emission site dynamics. The methodology involves: Performing relativistic MHD simulations using codes like PLUTO to capture jet dynamics under varying magnetisation and jet geometries. Developing polarisation-enabled radiative transfer post-processing pipelines to generate synthetic polarisation light curves. Applying deep learning models—including unsupervised clustering, time-series similarity learning, and neural Fourier/Wavelet decomposition—to the vast GRB datasets from Fermi, Swift and BATSE. These will be used to detect gravitational lensing, characterise variability timescales, and identify QPOs. If successful, this will deliver India’s first simulation-based time-resolved spectro-polarimetric predictions for GRBs and among the few globally. It will aid interpretation of data from missions like AstroSat-CZTI, LEAP, POLAR-2, and COSI, advancing understanding of jet composition, magnetic fields, and emission in relativistic jets. The deep learning frameworks developed will extend beyond GRBs to FRBs, AGNs, X-ray binaries, and even non-astronomical domains like geophysical signals (e.g., earthquake precursors) and biomedical time-series (e.g., ECG, EEG). The project seeks HPC augmentation with compute nodes, high-memory workstations, storage, and AI-skilled manpower. All models will be shared publicly via GitHub to enable reuse and collaboration.