Flatiron Institute Center for Computational Astrophysics Pre-Doctoral Program

Julianne Dalcanton (Director)

Most of my research focus is on using nearby galaxies as laboratories for constraining astrophysics, with particular interests in stars, stellar populations, the interstellar medium, and their interactions. I’m also always interested in good data discovery or calibration puzzles.

Adrian Price-Whelan

Research Interests: Dark matter, galactic dynamics, Milky Way structure and evolution, data analysis and statistical inference, binary stars, methods, community software

Searching for Stellar Streams

Stellar streams form from disrupting stellar systems, like dwarf galaxies and globular clusters. Streams from star clusters are dynamically “cold” and are therefore very sensitive to perturbations from passing massive objects like dark matter subhalos, which make them extremely useful objects to find within the Milky Way. We will develop a new method to search for streams using data from the Gaia mission.

Exploring the distribution and evolution of neutron-capture elements in the Milky Way – Adrian Price Whelan (ARS) with FRFs Carrie Filion and Aida Behmard

The abundances of elements present in the atmospheres of stars can be inferred from stellar spectra. Elements are made through a number of different nucleosynthetic channels, and by examining high-resolution spectra of many stars, we can learn about these channels and Galactic chemical evolution as a function of location and time. The slow- and rapid neutron-capture process elements are particularly interesting as their origins and evolution are not as well-constrained as other channels.The most recent data release from the Galactic Archaeology with HERMES (GALAH) survey provides an unparalleled opportunity to investigate neutron-capture elements in the Milky Way thanks to its unique combination of spectral resolution, optical wavelength coverage, and survey size. We will use this dataset to constrain the distribution of a number of neutron-capture elements as a function of Galactic location in each of the structural components of the Milky Way. By comparing neutron-capture element distributions to those of elements resulting from other nucleosynthetic channels, we will gain insight into fundamental processes underlying Galactic chemical evolution. If there is time, we will derive stellar ages and incorporate a temporal dimension into our analysis. This project may also involve data-driven models of stellar spectra to re-derive or improve elemental abundance measurements from GALAH and other large spectroscopic surveys.

MARVELous DREAMS: Leveraging AI to explore baryonic subgrid physics in dwarf galaxies: Akaxia Cruz (FRF) in collaboration with Paco Villaescusa-Navarro (RS)

While cold dark matter agrees well with observational data on >~ Mpc scales, non-linear galactic scales have often challenged the LCDM paradigm. In particular, shallow dark matter cores are observed in dwarf galaxies, in tension with CDM-only simulations. The addition of galaxy formation (baryonic) physics in which star formation and subsequent feedback rearrange central dark matter densities has been proposed as a solution to create dark matter cores. However, current state-of-the-art galaxy formation simulations are unable to simultaneously resolve cosmological scales associated with dark matter halo growth and merging, and scales where stars form. Thus simulators often employ subgrid prescriptions for star formation and feedback, which are tuned with ‘subgrid parameters’. An alternative solution is dark matter self-interactions which collisionally thermalize dark matter particles and alter central halo densities. Since alternative dark matter and baryonic physics are both capable of producing dark matter cores, their signature may be degenerate in certain regions of baryonic subgrid and dark matter micro-physics parameter space. The goal of the DREAMS project is to use large simulation suites and machine learning to disentangle these regions.

In this project a set of 1024, Mhalo ~ 10^10 Msol at z = 0 dwarf galaxy simulations will be run with varied ICs, and N star formation and black hole parameters in the Charm N-body GrAvity solver (ChaNGa) code. The pre-doc will apply recent deep-learning techniques to examine this suite of simulations to help identify the best observables to use to further constrain the baryonic subgrid parameter space in a CDM cosmology. In particular the suite combined with AI techniques will be used to determine how scaling relationships like M200-M*, M*-MBH, size-mass, as well as core size and formation times change in the N-dimensional subgrid parameter space.

Explaining the polarization observed by the Event Horizon Telescope: Alisa Galishnikova (FRF) and Lorenzo Sironi (RS)

Current theoretical models of accretion flows around low-luminosity supermassive black holes struggle to predict the polarization properties observed by the Event Horizon Telescope. Specifically, without artificially adjusting the temperature of the synchrotron-emitting electrons, radiative transfer of General Relativistic Magnetohydrodynamic numerical models typically predict polarization fractions that are about 5 times higher than what has been observed. Additional turbulence in the accretion flow, which is not captured by current numerical models, may be responsible for this discrepancy. This project will investigate whether extensions of existing theoretical and numerical models can better explain the observed polarization properties and explore their other observational implications.

Anna Wright (FRF)

Research Interests: Interested in using dwarf galaxies and other members of the low surface brightness universe to understand the incredible diversity that we observe across the galaxy population. The shallow gravitational potential wells of these galaxies make them excellent probes of the processes – both internal and external – that govern galaxy evolution and the fact that we typically don’t tune our simulations to them means that they can serve as unique tests of how we implement small-scale physics in simulations. I use high resolution cosmological simulations run with Gasoline/ChaNGa or Enzo to study dwarfs in all environments and I’m also interested in using synthetic data to make apples-to-apples comparisons with observations.

The fate of ram pressure stripped gas from resolved dwarf galaxies: Anna Wright (FRF) and Stephanie Tonnesen (ARS)

When low mass or “dwarf” galaxies fall into the gravitational potentials of larger galaxies like our own Milky Way, they often lose their gas through ram pressure stripping as they interact with the circumgalactic medium (CGM) of their host. However, most modern simulations struggle to resolve the diffuse gas that makes up the bulk of both galaxies’ CGM, obscuring the fate of this stripped gas. With this project, we will use cosmological simulations in which extra resolution elements have been added to the CGM to study how the gas from the dwarf galaxy – both ISM and CGM – evolves during and after ram pressure stripping: does it seed the formation of cold gas in the host’s CGM? Does it ultimately accrete onto the disk? How long does it remain coherent after being stripped? The student will analyze the outputs from multiple simulations to answer these questions.

How to make a polar ring: Anna Wright (FRF) and Stephanie Tonnesen (ARS)

Observations have revealed that a small fraction of galaxies have a secondary disk that is kinematically distinct from the central disk. The origins of these “polar ring” galaxies are unknown, but are generally thought to be connected to interactions with other galaxies and/or accretion from misaligned filaments in the cosmic web. The student will use a cosmological zoom-in simulation of a galaxy that forms a polar ring at low redshift to explore the contributions of ram pressure-stripped gas from dwarf galaxies, cold accretion, and other processes to the formation of the galaxy’s polar ring and its ignition of star formation.

Molecular hydrogen searches in archival JWST data at z>6: Blakesley Burkhart

We have tenatively detected molecular H2 in JWST stacked JADES galaxies. This project will deep dive into the archival spectra to determine which individual galaxies are most promising candidates to detect H2 lines individually.

The Life and Times of Molecular clouds in Starforge Simulations: Blakesley Burkhart

We will study the rates of molecular dissociation/formation in star forming clouds with the Starforge framework.

Elena Pinetti

Research Interests: Dark matter remains one of the most compelling mysteries in modern physics, accounting for approximately 85% of the universe’s total matter. So far, its existence has been inferred solely through gravitational effects. However, several theoretical models suggest that dark matter could also interact non-gravitationally, potentially producing a wide variety of astrophysical signals. Detecting such signals would be groundbreaking, providing the first evidence of dark matter interacting through a force beyond gravity. In particular, dark matter annihilation or decay could generate various astrophysical messengers, including photons, neutrinos, and charged cosmic rays. Each of these messengers requires distinct detection techniques and provides unique insights. Since different messengers are linked to various dark matter candidates, a multi-messenger approach offers a powerful framework for studying dark matter in the universe. The two proposed projects aim to 1) enhance our understanding of the dark matter distribution within cosmic filaments using cosmological simulations, and 2) identify a potential dark matter signature in MUSE telescope data.

Francisco Villaescusa-Navarro (RS)

Research Interests: I am interested in combining massive datasets from numerical simulations with machine learning to learn about cosmology, astrophysics, and dark matter. I am particularly interested in agentic AI and geometric deep learning.

Exploring the Dark Matter Distribution within Cosmic Filaments: Elena Pinetti (FRF) with Ken Van Tilburg (ARS), Francisco Villaescusa-Navarro (RS), Mariangela Lisanti (ARS)

In our universe, matter is arranged in a vast, web-like structure known as the cosmic web, consisting of filaments, galaxy clusters, and cosmic voids. Filaments are dense, elongated structures that connect galaxy clusters. One of the main challenges in studying cosmic filaments is their faintness, which makes direct observation difficult. As a result, many aspects of their properties and evolutionary dynamics remain unknown. Understanding the universe’s structure requires an accurate description of the dark matter content within these filaments. This project aims to investigate various filament properties across different dark matter models using cosmological simulations, with a particular focus on the behavior of small-scale structures within filaments.

Unveiling Dark Matter Signatures with the MUSE telescope: Elena Pinetti (FRF) with Ken Van Tilburg (ARS), Francisco Villaescusa-Navarro (RS), Mariangela Lisanti (ARS)

Numerous dark matter models predict that dark matter particles could decay into photons, producing a distinctive line feature. If the mass of these dark matter candidates is on the order of the electronvolt, optical telescopes are well-suited to search for the corresponding emission line. In particular, the Multi-Unit Spectroscopic Explorer (MUSE), a cutting-edge instrument at the Very Large Telescope (Chile), combines a wide field of view with high spatial resolution and a large spectral range, making it ideal for dark matter searches. This project aims to use available MUSE data to search for a dark matter signal.

Reconstructing the Initial Conditions of the Multimodal Cosmos: Francisco Villaescusa-Navarro (RS) and Adrian Bayer (FRF)

In this era of Big Data, we have access to exceptionally high-resolution images of the Universe from a variety of telescopes, offering an unprecedented opportunity to understand our place in the Cosmos. The most promising way to maximize what we learn from these observations is by reconstructing the initial conditions of the Universe with cutting-edge computational techniques.

In this project, we will develop this approach for the vast datasets provided by surveys such as DESI, Euclid, and the Simons Observatory by optimizing high-dimensional probabilistic inference, designing efficient machine learning architectures, and integrating multi-modal methods to combine data from galaxy surveys, weak lensing, and the CMB into a unified framework. In particular, we will explore Monte Carlo methods, diffusion models, and/or Schrödinger Bridges to enable more precise and efficient reconstructions.

We will apply these techniques to state-of-the-art cosmological simulations of the multimodal cosmos, such as HalfDome and Backlight. There are also opportunities to extend this work to study the baryonic Universe or explore alternative dark matter models, leveraging simulation suites such as CAMELS and DREAMS. By combining Bayesian methodologies with modern AI-driven inference, we will push the boundaries of our ability to reconstruct the early Universe from real data, and constrain cosmological and astrophysical parameters like never before.

Numerical simulations and AI: Francisco Villaescusa-Navarro (RS) and Adrian Bayer (FRF)

At CCA, we have a very large collection of state-of-the-art simulations such as Quijote, CAMELS, DREAMS, Backlight. Students interested in:
– analyzing the simulations
– contributing to their development
– training machine learning models with them
– building AI agents to automate scientific discovery
are encouraged to apply. For more information about these simulations and their scientific applications see:

Quijote: https://quijote-simulations.readthedocs.io
CAMELS: https://camels.readthedocs.io
DREAMS: https://dreams-project.readthedocs.io

Laura Sommovigo (FRF)

Research Interests: high-redshift, dust, ISM, galaxy evolution, the interface between models and observations.

Investigating the Nature of Little Red Dots: Laura Sommovigo (FRF) and Rachel Somerville (SRS)

In its first year, JWST has shattered records by detecting the earliest galaxies and supermassive black holes (SMBHs) at z>>5. However, the bulk of the high-redshift AGN population likely remains hidden in dust-obscured, low-luminosity systems. A compelling candidate population for these missing AGN is represented by the so-called “”little red dots”” — faint, compact, and red sources whose nature remains uncertain. These objects could either be quiescent galaxies, heavily dust-obscured starbursts, or elusive AGNs that have so far evaded UV-based detection methods.

We aim to investigate the fraction of these sources that could harbor obscured AGN by leveraging a semi-analytical framework that models the interplay between galaxy-scale dust attenuation and AGN torus obscuration. By varying both galaxy and torus parameters, we will generate synthetic observations to compare against JWST data. This will allow us to test whether the little red dots are consistent with an AGN origin, assess how dust-related biases impact our census of the high-redshift AGN population, and ultimately place constraints on the physical mechanisms driving SMBH growth in the early Universe.

Matteo Cantiello (RS)

Research Interests: Stellar Physics, Stellar Evolution, Binary Stars, MESA, Fluid Dynamics, Numerical Simulations, Magnetohydrodynamics, Planets, Active Galactic Nuclei, Black Holes, Gravitational Waves.

Modeling Stellar and Planetary Ingestions with MESA: Matteo Cantiello (RS)

Planets and stars can be engulfed when, e.g., their host or companion star ascends the giant branch. Using a 1D code (MESA) it is possible to account for the energy deposited during the spiral in and determine the evolution of the primary star. We will use 1D modeling (MESA) to track the evolution of the primary, and 3D hydro simulations to study details of the energy and angular momentum deposition processes. This project aims at creating a grid of light-curves to be used as templates for observations with VRO/LSST.

Stars in AGN Disks: Matteo Cantiello (RS)

Stars are likely formed in, or captured by, the disks of active galactic nuclei (AGN). The disk conditions profoundly change the star’s evolution, with AGN stars accreting large amounts of mass and becoming massive / very massive. This project could involve either improving the 1D treatment of the accretion stream using existing 3D results from radiation hydrodynamics simulations, modeling the long-term stellar evolution in the MESA software instrument, or studying the interplay of stellar dynamics, AGN disk models, and evolution, tying together output from a variety of tools with semi-analytic models. Another possible project could involve calculating the rate of visible explosive transients in AGN disks from a population of massive AGN stars. This predictions could be useful for e.g. VRO/LSST, LIGO/VIRGO.

Megan Bedell (RS)

Research Interests: Exoplanet detection, stellar spectroscopy, stellar abundances, extreme precision radial velocity, spectroscopic data analysis, tellurics, stellar variability.

Improved Diagnostics of Stellar Activity Cycles: Megan Bedell (RS) with Ryan Rubenzahl (FRF)

Using Sun-as-a-star spectra, develop and test data-driven indicators for stellar activity levels, then apply these metrics to long-running RV survey archives to better characterize the prevalence of Solar-like activity cycles.

(Super)granulation Across Spectral Types: Megan Bedell (RS)with Michael Palumbo (FRF)

Building off of high-resolution solar data and scaling relations, create a model to predict spectral line shape changes due to convective motions in stars across the FGK spectral type range, and test it against standard star observations.

Mike Grudić (Software ARS)

Research Interests: Star formation, interstellar medium, stellar feedback, numerical techniques.

Sub-grid star formation recipes revisited with STARFORGE: Mike Grudić (Software ARS)

Hydrodynamical galaxy formation simulations do not resolve the gas flows that form individual stars, but do require some numerical rule for converting gas into newborn stars. The current state-of-the-art prescriptions are based upon simplistic assumptions that do not hold in real star-forming molecular clouds, and may predict orders-of-magnitude incorrect star formation rates.

It is now possible to revisit this issue with the latest generation of numerical star formation simulations. The STARFORGE simulations account for all key processes responsible for regulating the rate of star formation: turbulence, gravity, thermodynamics, and all forms of stellar feedback. The goal of this project will be to 1. assess how well (or poorly) the existing prescriptions predict the rate of star formation in the STARFORGE simulations, and 2. use the simulation data to calibrate a next-generation sub-grid star formation model for use by the galaxy simulations community.

Evolution of feedback-driven bubbles in a turbulent magnetized medium: Mike Grudić (Software ARS)

Massive stars affect their natal clouds through feedback in the form of jets, radiation, winds, and supernova explosions. Exact, well-understood solutions exist for spherical, unmagnetized clouds, but real molecular clouds are neither.

This project will be an analysis of an existing suite of high-resolution, numerical radiation-MHD simulations of stellar feedback in a turbulent, magnetized GMC, run with the GIZMO code with the STARFORGE feedback and ISM physics packages. The key goal will be to compare the behaviour of the 3D clouds with corresponding 1D solutions.

New simulations may be run by the student if more parameter coverage is needed and the student is interested in developing this skillset.

Ore Gottlieb (FRF)

Research Interests: Numerical simulations, Magnetohydrodynamics, General-relativity, Multi-messenger Astrophysics, Black Holes, Neutron Stars, Black Hole–Neutron Star Mergers, Neutron Star–Neutron Star Mergers, Supernovae, Gamma-ray Bursts, Fast Blue Optical Transients, Accretion Disks, Active Galactic Nuclei, Gravitational Waves, and Particle Emission.

Connecting stellar death and jet formation: simulating the dynamo effect in collapsing stars: Ore Gottlieb (FRF)

The collapse of some rapidly rotating massive stars may initiate the formation of long gamma-ray bursts (LGRBs) – the most luminous explosions in the Universe. LGRBs are generated by relativistic magnetized jets emerging from the black hole formed in the core of the stellar progenitor. The development of these Poynting-flux dominated jets necessitates the presence of strong poloidal magnetic fields threading the black hole. However, the differential rotation in the star generates magnetic fields of a toroidal geometry. This underscores the significance of a dynamo effect that converts the toroidal field into a poloidal one for jet formation. This project, utilizing advanced 3D general-relativistic magnetohydrodynamic simulations, aims to numerically investigate the dynamo effect in collapsing massive stars and decipher how jets are born. This will establish a self-consistent connection between the demise of massive stars and the formation of jets for the first time.

Connecting black hole natal properties to black hole populations: Ore Gottlieb (FRF)

The collapse of massive stars is the primary formation channel for black holes in the universe. Some of these stars rotate rapidly during collapse – so-called collapsars – which produces strong outflows that can expel the stellar envelope and limit the mass accreted onto the black hole. Concurrently, the outflows extract angular momentum from the black hole, reducing its spin. This project employs state-of-the-art 3D general-relativistic magnetohydrodynamic simulations of collapsars to investigate how the interplay between black hole spin and mass, mediated by this feedback mechanism, shapes the population of black holes formed via this channel. Ultimately, it aims to provide a predictive model for probing the origins of black holes through gravitational-wave data.

Constraining the formation pathways of black holes in the early universe: Rachel Somerville (SRS)

One of the big puzzles in galaxy formation is how supermassive black holes formed and how they were able to grow so quickly in the early universe. The James Webb Space Telescope is continuously discovering more and more of these early black holes and probing further back in lookback time. We are interested in using insights from detailed small-scale numerical simulations to understand the physics of how black hole seeds formed, how black holes accrete, and how they merge. We then use these insights to develop recipes that can be included in much more computationally efficient semi-analytic models, in order to predict observable properties of the SMBH population over a broad range of cosmic time. Projects may involve analyzing numerical simulations, developing, running, or analyzing semi-analytic models, and/or creating and analyzing synthetic observations. There are also opportunities to incorporate machine learning and/or simulation based inference techniques.

Sebastian Wagner-Carena (FRF)

Research Interests: I work at the intersection of machine learning and astrophysics research. My interests include dark matter physics, large-scale structure, simulation-based inference, contrastive learning, Bayesian machine learning, strong lensing, and differentiable simulators. I’m open to discussing potential projects relating to any of the above interests.

Astrophysics with Probabilistic Contrastive Learning Sebastian Wagner-Carena (FRF) and Francisco Villaescusa-Navarro (RS)

In astrophysics, we work with diverse data modalities to study celestial objects. A galaxy, for instance, can be observed using a spectroscopic fiber, a targeted space-based telescope, or as part of a large ground-based survey. Combining these heterogeneous observations to jointly constrain the properties of astrophysical objects remains a significant challenge. This project aims to leverage modern contrastive learning techniques to develop a data-driven, multi-modal model for astrophysical data, enabling probabilistic integration of diverse observations. A successful model would facilitate rare object searches and improve photometric redshift (photo-z) estimation.

Hydrodynamics in the Latent Space: Sebastian Wagner-Carena (FRF) and Francisco Villaescusa-Navarro (RS)

This project would focus on building a shared latent space encoding for different hydrodynamic models (i.e. TNG, SIMBA, etc.). The goal of the project would be to build a physically informative mapping in a data-driven fashion that would allow us to 1) express the similarity of competing hydrodynamics prescriptions, 2) directly infer which hydrodynamics prescriptions best align with our universe, and 3) steer our future simulations towards using our observations.

Thomas Pfeil (FRF)

Research Interests: My research is focused on the combination of different methods for simulations of protoplanetary disks to better understand the planet formation process. In particular, I am interested in modeling the evolution of dust particles and their connection to turbulence.

For my research, I’m conducting multifluid hydrodynamic simulations with PLUTO and Athena++ for which I am developing subgrid methods to model dust evolution on top of the hydrodynamic calculations.

Potential projects will be co-mentored by CCA Research Scientist Dr. Yan-Fei Jiang, who is an expert in radiation hydrodynamic modeling and the main developer of the radiation module for the Athena++ GRMHD code.

Yan-Fei Jiang (RS)

Research Interests: I am generally interested in multi-dimensional radiation (magneto-)hydrodynamic simulations of various astrophysical simulations. Examples include spectrum, structures and dynamics of accretion disks around compact objects, proto-planetary disks, 3D structures of (binary) stars, as well as radiation driven outflows in different contexts.

Interaction of Dust and Radiation in Protoplanetary Disk Simulations: Thomas Pfeil (FRF) and Yan-Fei Jiang (RS)

Dust particles are not only the building blocks of planets, but also determine the thermal structure of protoplanetary disks.

Combining methods for radiation-hydrodynamic modeling and dust evolution is thus necessary to understand the evolution of planet-forming disks.

We welcome project ideas related to hydrodynamic and radiation-hydrodynamic simulations with dust in protoplanetary disks.