Embedded Files
Machine Learning & Artificial Intelligence
RealtorAI
RealtorAI
An AI-powered Agent that answers questions about Singapore HDB resale prices using LLMs with RAG (Retrieval-Augmented Generation).
ML on MRI
ML on MRI
Machine Learning based model for predicting gender in ADHD subjects using deep learning models, PCA, undersampling, and feature engineering.
Gravitational Lens Modeling
The lens model of galaxy-scale single-source-plane-lens AGEL0142 modeled using GLEE (left) and the lens model of galaxy-scale double-source-plane-lens AGEL1507 modeled using LENSTRONOMY (right). References: Sahu et al. (2024), Sahu et al. (2025), Github link to code
Cosmography with Gravitational Lenses
Double source plane gravitational lenses, with two background sources at widely separated redshifts, are a powerful probe of cosmology. The constraints from these lenses are orthogonal to, and therefore highly complementary with, existing standard probes such as Planck’s Cosmic Microwave Background (CMB) observations. Lens modeling of double source plane lenses provides an independent measurement of the distance ratio, known as β, between the sources, the deflector, and the observer. This distance ratio is sensitive to the matter density parameter and the dark energy equation of state parameter. If spectroscopic redshifts of the sources and deflectors are known, the independent measurement of β can be used to constrain cosmology. While constraints from a couple of DSPLs used for cosmography so far are already promising—improving CMB constraints for wCDM model by 30%—a larger sample of DSPLs with confirmed spectroscopic redshifts will be a powerful tool for testing cosmological models independently, addressing tensions in the concordance ΛCDM model, and investigating the nature of dark energy. Reference: Sahu et al. (2025)
How Do Galaxies Evolve with Time?
This project aims to study the evolution of total mass profile properties of galaxies with cosmic time and test the two-phase model of galaxy formation and evolution. The gravitational lens modeling can help obtain the total (baryonic plus dark matter) mass density profile of the foreground deflector galaxy. This can be used to understand the dark matter profile in the lens galaxies. Further, based on the correlations of density, mass, size, and dark matter fraction with the cosmic time (redshift) for a statistical sample of a particular class of galaxies, one can investigate the mass assembly history of that type of galaxy. In Sahu et al. (2024), I modeled a pilot sample of seven lenses from the Astro3d Galaxy Evolution with Lensing (AGEL) survey using state-of-the-art lens modeling software GLEE (Suyu and Halkola et al. 2010,2012). We updated the high redshift (z>0.5) end of the total mass density profile slope (γ) versus redshift (z) relation (AGEL data marked with blue stars). This paper compares the γ—z relation from advanced simulations, dynamical observations, and all lensing observations and investigates possible reasons behind conflicts between studies and how to resolve them. Reference: Sahu et al. (2024)
Black Hole--Galaxy Correlations
The correlation between black hole mass and host galaxy properties is dependent on galaxy morphology, which is shaped by the evolutionary track followed by a galaxy. These figures show the relation of black hole mass with host bulges mass (left) and bulge size (right). Early-type galaxies with a disk (ES, S0), early-type galaxies without a stellar disk (E), and late-type galaxies (LTGs, S) define different relations. The same morphological dependence is found between black hole mass--galaxy mass and black hole mass--density relations. These relations are obtained using the largest sample of galaxies with dynamically measured black hole masses, state-of-the-art galaxy modeling and multi-component decompositions, and statistical regressions.
References: Sahu et al. (2019a), Sahu et al. (2020), Sahu et. al. (2022), Graham & Sahu (2023a), Graham & Sahu (2023b)
A break in the M--σ relations
Core-Sérsic and Sérsic galaxies define two different relations between galaxy luminosity (absolute magnitude) and the central stellar velocity dispersion (left) and between black hole mass and central stellar velocity dispersion. The core-Sérsic galaxies are massive merger-driven galaxies, whereas Sérsic galaxies are gas-abundant and evolve through accretion and gas-rich mergers. The left plot is based on the V-band data from Lauer et al.(2007), and the right plot uses the exhaustive sample of dynamically measured black hole masses. Reference: Sahu et. al. (2019b)
Modeling Galaxy Image & Multi-component Decomposition
Modeling Galaxy Image
2D isophotal model of a galaxy, obtained using in-house software, capturing all the photometric and structural galaxy properties, e.g., surface brightness (μ), ellipticity (𝜖), position angle, and isophotal irregularities quantified by higher-order Fourier harmonic coefficients (An, Bn, up to n=12th order). The luminosity of a galaxy is modeled by fitting quasi-elliptical isophotes at each radius along the semimajor axis (Rmaj). References: Sahu et al. (2019a)
Multi-component Decomposition
After obtaining the isophote fit at along galactic radii, each isophote is uniformly sampled across the whole azimuthal range, using a natural angular coordinate for ellipses known as the “Eccentric Anomaly” (ψ) that provides the average intensity and associated parameters of the isophotes as a function of semi-major axis radii (shown in left figures).
Total galaxy light is then disassembled into its components as depicted in the (isophotal-averaged) galaxy surface brightness profile along its major-axis and geometric-mean axis. This analysis enables us to measure accurate stellar fluxes (which are converted to stellar mass using appropriate stellar mass-to-light ratio) and structural properties of galaxy components and the whole galaxy, in addition to the detailed galaxy morphology. Comprehensive image analysis and the final decomposition profiles for other galaxies are presented in Sahu et al. (2019a)
Page updated
Google Sites
Report abuse