Projects

Nova explosions are a crossroads between stellar astrophysics, nuclear physics and chemical synthesis, serving as laboratories to investigate different physical processes such as mass transfer in binaries or explosive nucleosynthesis. Novae are very luminous events that occur in interacting binary systems in which a white dwarf accretes on its surface material accreted from its companion. RS Oph is a symbiotic binary system, whose outbursts have a recurrence of about 15 years. However, despite being a well-studied system, there are still missing pieces to fully understand it, such as the orbit of its two components.  The aim of this work is to determine radial velocities of the two components from very high resolution spectra in order to parameterise the orbit of the system and to constrain the masses of the two components.

This project involves using publicly accessible archival images from DECam on the Blanco 4m telescope in Chile to find satellite streaks, measure their brightness, and (when possible) identify the source of each streak. We expect many of the streaks to originate from Starlink satellites. The student will therefore contribute significantly to a new paper on how the Starlink satellite population has changed over time from 2019-2025. The streaks that aren't Starlink satellites are interesting too, and all measured brightnesses for objects in the public satellite catalog will be uploaded to the Satellite Constellation Observation Repository (SCORE). The project will help quantify the impacts of satellite constellations on astronomy, which is necessary to effectively mitigate those impacts. The student will be expected to apply to join SatHub (it is free with no prerequisites, see cps.iau.org), and use and potentially contribute to SatHub software tools including SatChecker (satchecker.readthedocs.io) and SCORE (score.cps.iau.org).
In the last five years, Earth has entered a new era in orbital space. SpaceX alone has launched nearly 8,000 satellites since 2019, and multiple companies have proposed a total exceeding 1,000,000 satellites in the near future. Satellites in low-Earth orbit move fast across the sky and reflect sunlight, which can cause bright lines or streaks in telescope images. Sky observers worldwide now find themselves sharing increasingly crowded space across the electromagnetic spectrum. To address this challenge, the International Astronomical Union Centre for the Protection of the Dark and Quiet Sky from Satellite Constellation Interference (IAU CPS), and in particular SatHub, brings together experts and concerned parties to quantify impacts to astronomy and develop strategies to mitigate those impacts. More information is available at cps.iau.org.

In this internship, you will explore the intersection of cosmology and machine learning. Your main focus will be on developing a more precise model of the Point Spread Function (PSF). The PSF describes how a point source, such as a star, appears in an image after its light passes through the Earth's atmosphere and the telescope optics. Essentially, it shows the final "blurred" image of a point source, acting as a fingerprint of the entire imaging system. PSF is a critical factor when measuring cosmic shear—the subtle distortion of galaxy shapes due to gravitational effects, which is one way to probe the history of expansion of the Universe and the growth of structure.
Currently, PSF models are built using data from a single image/Charge-coupled device (CCD). However, modern wide field telescope are paved of multiple CCDs which means that modeling PSF on a single CCD often misses important information available across the entire Field of View (FoV) of a telescope. Your work will tackle this limitation by using deep learning techniques—specifically, a type of neural network called a Variational Autoencoder (VAE). A VAE is a tool that learns the underlying patterns in data, allowing it to capture complex structures like the PSF. While prior experience in linear algebra, machine learning, Python, or Linux is beneficial, it is not required.
What You'll Do:
- Learn about PSF, cosmic shear, cosmology, and deep learning.
- Work with precursor datasets (for example, data from the Subaru telescope) as a proof of concept.
- Learn and apply fundamental concepts of linear algebra and machine learning in the context of astrophysical data.
- Develop, train, and evaluate a VAE model to achieve a more accurate and global PSF representation.
- Gain experience in Python programming, Linux environment, and deep learning library such as PyTorch.

Young exoplanets (< 1 Gyr) offer us snapshots of early planetary evolution that can help to understand the link between the planet's formation and the observed population of old planets. Therefore, it is crucial to keep track of their ephemerides to make informed decisions on follow-up efforts such as radial velocity campaigns to measure planetary masses or transmission spectroscopy to study their atmospheres. K2-233 is a young star (~360 Myr) that was observed by the K2 NASA mission back in 2017. These data unveiled the presence of three transiting planets with periods of 2.5, 7.1, and 24.4 days. Despite the importance of this system, the planetary transits have not been observed since 2017. Seven years later, the NASA Transit Exoplanet Satellite Survey (TESS) will re-observe the star for 30 days in its Sector 91 (April 2025).  The objective of this project is to analyse the new TESS data and provide improved planetary orbital parameters and radii. The student will learn the transit method to detect exoplanets, detrending techniques to separate stellar and planetary signals, and parameter estimation using Markov chain Monte Carlo. 

The large-scale environment is one of the main external drivers of galaxy evolution, while the active galactic nucleus (AGN) is a key internal driver. Both play a fundamental role in regulating the star formation and chemical evolution of galaxies. Metallicity is a crucial indicator of these processes, as it reflects the chemical enrichment and dynamics that have shaped a galaxy. Previous studies have linked metallicity to gas accretion, stellar winds, AGN, supernova explosions, stellar mass, star formation rate, and gas fraction. In turn, many of these factors depend on the cosmic environment.
The mass-metallicity relation (MZR) shows a positive correlation between stellar mass and metallicity. However, for a given mass, the dispersion in metallicity suggests the influence of other processes.
This project will investigate how the environment affects the MZR in AGN galaxies. Emission lines will be measured in optical spectra in a sample of AGNs located in cosmic filaments, analysing the evolution of metallicity with radial distance to the filament taking into account the stellar mass. Our goal is to understand when and where chemical enrichment occurs in the large-scale structure of the Universe.

​This project aims to identify key spectral signatures for the chemical characterisation of ices present in different astrophysical objects, using public spectra from the James Webb Space Telescope (JWST). The observational data will be compared with laboratory spectra to select the most suitable candidates for describing the behaviour of the spectral signals. Minimal knowledge of spectroscopy and preferably experience with Python is required.

In this project, we will explore the potential of neural networks, specifically with a Deep Learning (DL) algorithm, to estimate cosmological parameters from a series of numerical simulations based on the standard ΛCDM model of cosmology. To train the algorithm, different hydrodynamical configurations will be taken from the CAMELS project, which include both the effects of gravity, baryons, and different astrophysical processes (e.g., supernova feedback and AGN). Our main goal is to optimize the DL algorithm to find 6 free parameters used to build these simulations, focusing on certain fields with physical properties. We have multiple cosmological models with 1000 realizations (or LH cubes), each of them with astrophysical and cosmological variations, which change the large-scale matter distribution, and therefore generate different patterns that will be captured by the algorithm during its training. In addition, the project will allow us to test how the physical models expressed in the simulations change with redshift and contribute to shaping our current universe. Currently, PSF models are built using data from a single image/Charge-coupled device (CCD). However, modern wide field telescopes are paved with multiple CCDs, which means that modeling PSF on a single CCD often misses important information available across the entire Field of View (FoV) of a telescope. Your work will tackle this limitation by using deep learning techniques—specifically, a type of neural network called a Variational Autoencoder (VAE). A VAE is a tool that learns the underlying patterns in data, allowing it to capture complex structures like the PSF. While prior experience in linear algebra, machine learning, Python, or Linux is beneficial, it is not required.
What You'll Do:
- Learn about PSF, cosmic shear, cosmology, and deep learning.
- Work with precursor datasets (for example, data from the Subaru telescope) as a proof of concept.
- Learn and apply fundamental concepts of linear algebra and machine learning in the context of astrophysical data.
- Develop, train, and evaluate a VAE model to achieve a more accurate and global PSF representation.
- Gain experience in Python programming, Linux environment, and deep learning libraries such as PyTorch.

Disk galaxies are the main star-forming galaxies in the present Universe, and today we can already simulate them with a good level of complexity. However, it is not fully understood what physical ingredients regulate the rate at which galaxies form stars, as there are many factors that can influence: the level of galaxy rotation, the amount of total mass, the amount of baryonic mass, the amount of gas mass in the disk, possible interactions with other galaxies, the shape of the clouds, the metallicity of the galaxies, the parameters of the dark matter halo, the relative orientation between dark matter and baryonic matter, and so on. In this project we want to simulate a large variety of galaxies, sweeping all these parameters, and search for possible dependencies of the star formation rate on these parameters. To do this we will simulate galaxies, both isolated, in idealized situations, and galaxies coming from cosmological simulations, but increasing the resolution. The project will allow us to learn how to initialize, run, visualize and analyze numerical simulations made with supercomputers in Mexico to which we have access.