2020 Flatiron Institute Center for Computational Astrophysics Pre-Doctoral Program

Important Dates
  • Application Deadline
  • Applicant notification
  • Fellowship start date(s)
  • or
Contact Info
  • Please send inquiries
    about the program to ccainfo@flatironinstitute.org

Purpose

The Center for Computational Astrophysics (CCA) at the Flatiron Institute is a vibrant research center in the heart of New York City with the mission of creating new computational frameworks that allow scientists to analyze big astronomical datasets and to understand complex, multi-scale physics in a cosmological context.

The CCA Pre-Doctoral Program will enable graduate student researchers from institutions around the world to participate in the CCA mission by collaborating with CCA scientists for a period of 5 months on site. With this opportunity, the selected group of researchers will be able to participate in the many events at the CCA and interact with CCA scientists working on a variety of topics in computational astrophysics (including both numerical simulations and sophisticated analyses of observational data), thereby deepening and broadening their skill sets.

Application

CCA Pre-Doctoral Program participants will be employed for up to 5 months at the CCA as Research Analysts. More information about this paid position is available on the application page, which can be accessed by clicking ‘Apply Now’.

Research Analysts will collaborate with one or more CCA scientists on a project of mutual interest. Potential applicants can find a list of projects proposed by possible CCA mentors by going to the mentors tab above. Alternatively, applicants may propose a project related to the interests of one or more of the CCA mentors listed. Before applying, applicants must contact one or more potential mentors to discuss the project of interest in detail and specify the selected mentor(s) in the research proposal.

Deadline

Applications for the Research Analyst position should be submitted here by June 1. Applicants will be notified about the status of their applications by June 15.

Supporting material for the application includes the following:

  • CV and publication list
  • Description of previous research experience (not to exceed two pages)
  • Research proposal of not more than 2 pages outlining planned work at Flatiron
  • Two (2) letters of recommendation submitted confidentially by the letter writers to ccapredoc@flatironinstitute.org. One letter must be from the applicant’s PhD supervisor and must explicitly approve the applicant’s possible participation in the Pre-Doctoral Program

When submitting your application, please read and follow the guidelines from the AAS, which are available here: https://jobregister.aas.org/postdoc-application-guidelines

Read More
Important Dates
  • Application Deadline
  • Applicant notification
  • Fellowship start date(s)
  • or
Contact Info
  • Please send inquiries
    about the program to ccainfo@flatironinstitute.org

Position Description 

TITLE: Research Analyst REPORTS TO: Associate Research Scientists and/or Group Leaders
DEPARTMENT: Flatiron Institute FLSA STATUS: Exempt
LOCATION: NY office (162)

DATE: March 1, 2020

EMPLOYMENT CLASSIFICATION: Fixed-term

ESSENTIAL FUNCTIONS/RESPONSIBILITIES

The Center for Computational Astrophysics (CCA) at the Flatiron Institute seeks temporary full-time Research Analysts as part of its Pre-Doctoral Program. The aim of this program is to provide graduate students from institutions worldwide the opportunity to be employed at the CCA for 5 months for the purpose of working on a research project with one or more CCA staff mentors. The program is open to individuals who are currently pursuing a PhD in a relevant field.

Before applying, candidates for this position must contact one or more potential mentors. For further details about the program and a full list of mentors, please visit: https://www.simonsfoundation.org/grant/flatiron-institute-center-for-computational-astrophysics-pre-doctoral-program/

MINIMUM QUALIFICATIONS

Education

• Currently enrolled in a PhD program in a field relevant to the proposed research

Experience

• 1-2 years of advanced course work in a relevant field
• Demonstrated understanding of basic research skills

Related Skills & Other Requirements

• Knowledge of software engineering practices for working in groups, including software development life cycles, coding standards, code review and version control systems (e.g., Git)
• Expertise in algorithms and data structures and/or in computational methods
• Technical and scientific curiosity
• Professional communication skills

REQUIRED APPLICATION MATERIALS

• CV and publication list
• Description of previous research experience (not to exceed two pages)
• Research proposal of not more than 2 pages outlining planned work at Flatiron
• Two (2) letters of recommendation submitted confidentially by the letter writers to ccapredoc@flatironinstitute.org. One letter must be from the applicant’s PhD supervisor and must explicitly approve the applicant’s possible participation in the Pre-Doctoral Program

When submitting your application, please read and follow the guidelines from the AAS, which are available here: https://jobregister.aas.org/postdoc-application-guidelines

Deadline
• All applications must be submitted no later than June 1, 2020.

THE SIMONS FOUNDATION’S DIVERSITY COMMITMENT

Many of the greatest ideas and discoveries come from a diverse mix of minds, backgrounds and experiences, and we are committed to cultivating an inclusive work environment. The Simons Foundation provides equal opportunities to all employees and applicants for employment without regard to race, religion, color, age, sex, national origin, sexual orientation, gender identity, genetic disposition, neurodiversity, disability, veteran status, or any other protected category under federal, state and local law.

Important Dates
  • Application Deadline
  • Applicant notification
  • Fellowship start date(s)
  • or
Contact Info
  • Please send inquiries
    about the program to ccainfo@flatironinstitute.org

Chiara Mingarelli

Supermassive black holes (SMBHs) in the million to 10 billion solar mass range form in galaxy mergers and live in the centers of galaxies with large and poorly constrained concentrations of gas and stars. When SMBHs merge, they create low frequency gravitational waves: ripples in the fabric of spacetime However, there are currently no observations of merging SMBHs— it is in fact possible that they stall at their final parsec (3.3 light years) of separation and never merge, called the final parsec problem. The only way to detect SMBHBs is by timing radio pulsars, which are excellent clocks. Their radio waves sweep across the Earth, like cosmic lighthouses, and GWs induce a delay or advance in the pulse arrival time. Thus, an array of precisely timed pulsars forms a galactic-scale GW detector, called a pulsar timing array. The cosmic merger history of SMBHs should generate a GW background (GWB), which will be detected in the next 5 years, providing a new and exciting avenue to explore galactic dynamics, SMBHB evolution, and fundamental physics not accessible by any other means.

Here I propose that we make the first attempt to assess the relative contributions of galaxy merger rates, SMBH mass estimates, and solutions to the final parsec problem, to the amplitude of the GWB, based on real galaxies instead of simulations. Importantly, the methods we develop here will provide insights into all aspects of low frequency GW astronomy: from the amplitude of the GWB to the likeliest SMBHB host galaxies — not possible with simulated populations. When the GWB is detected, we will in turn have the necessary tools to constrain the underlying astrophysics and demographics of the SMBHB population, addressing long-standing questions in galaxy evolution and fundamental physics.

Chris Hayward

I would be interested in mentoring a student on a project connected to any of my areas of research, so please email me if you have an ongoing project or project idea related to one or more of the following topics:

– Radiative transfer, dust and predicting observables from simulations
– Stellar feedback, turbulence and outflows
– Galactic magnetic fields
– Black hole accretion and feedback (with Daniel Anglés-Alcázar)
– Circumgalactic medium (CGM) and the physics of multiphase gas (with Drummond Fielding)
– Infrared/submillimeter-selected galaxies

Dan Foreman-Mackey

A systematic study of spectroscopic binaries observed by TESS

There are large catalogs of spectroscopic binary star systems detected using the radial velocity method but, in many cases, the precision with which we characterize these systems is limited because the orbits are not well sampled and there are covariances between the physical and orbital parameters. Many of these systems have also been observed by NASA’s TESS mission and in some cases eclipses and phase curves can be detected in the light curves. A joint analysis of these systems observed with both radial velocities and photometry yields extremely precise characterization of these stellar systems. In this project, we will perform a systematic analysis of all of these datasets and produce a catalog of benchmark stars that can be used to study the population of multiple star systems and place constraints on stellar evolutionary theory.

Detecting newborn planets with a neural network

(Co-mentors: Gabriella Contardo & Trevor David)

The properties of exoplanets are believed to evolve considerably over time, but evidence for this evolution has remained elusive. Young planets, for example, may be much larger than their older counterparts due to ongoing contraction and/or mass-loss. However, fewer than ten transiting exoplanets have been found around young stars (<100 Myr). One reason for the lack of evidence to date might be related to detection bias: conventional planet search methods are susceptible to under- or over-fitting the high-amplitude, rapid variability associated with young stars, and thus struggle to detect transits. A novel approach which makes no effort to model the underlying stellar variability may prove more efficient at detecting planets in these extreme cases. We will use machine learning methods to search for transiting exoplanets in the light curves of thousands of young stars from NASA’s TESS mission. The project has the potential to substantially increase the sample of young exoplanets and establish trends in the properties of those planets.

James Cho

My research interests mainly center on studying hydro and MHD dynamics using simulations. I am also interested in topological and algebraic techniques for analyzing multi-dimensional data and differential equations. I would be interested in working with a student on any of the following:

– Weakly and/or strongly ionized planetary/stellar atmospheric dynamics (including possible coupling to radiative transfer)
– Vortices or instability in protoplanetary/accretion disks
– Convection and mean flows in giant planet or stellar interiors
– Hamiltonian fluid dynamics (wave-mean flow interaction or analysis of quasi-linear equations)
– Topological characterization/analysis of data from simulations or observations.

John Forbes

Designing a high-performance neural network emulator for galaxy formation

(Possible co-mentors: Shirley Ho, Chris Hayward, Rachel Somerville, Dan Foreman-Mackey)

Emulators are a growing and increasingly powerful tool to speed up complex calculations with many different input parameters. Essentially, a neural network or some other regression method is trained on many cleverly chosen example runs of a simulation code to create a fast function that approximates the mapping between the input parameters and the outputs of interest from the simulation. This allows even modestly expensive simulations to be approximated and hence efficient exploration of and inference in parameter space. I am particularly interested in constructing an emulator for a radially resolved disk galaxy evolution model, perhaps employing a brand-new method of automated network architecture searches that may drastically reduce the necessary number of examples for training.

Constructing a galaxy catalog with covariances and selection functions

(Possible co-mentors: Rachel Somerville, Chris Hayward)

Our knowledge of galaxy properties comes from the interpretation of photometry and spectra. It has recently become possible to derive the joint posterior distribution of a wide variety of galaxy properties – stellar mass, star formation rate, redshift, metallicity, sizes, and even star formation histories – in a Bayesian framework. In collaboration with Joel Leja at Harvard, I am interested in using this code to produce a catalogue of probabilistic galaxy properties, where the properties of each galaxy have correlated uncertainties both among their properties and across all galaxies. This catalogue will be extremely useful in large hierarchical Bayesian inference problems to constrain the physics of galaxy evolution.

Kathryn Johnston & Adrian Price-Whelan

Stellar streams around the Milky Way provide a fossil record of the merger history of our Galaxy. By comparing the orbits of the many known streams to orbits of (undisrupted) satellite galaxy and globular cluster populations around our Galaxy, we should therefore be able to use these objects to measure the infall and destruction rate of satellites over cosmic time. A visiting graduate student could work with Adrian Price-Whelan and Kathryn Johnston on combining data from the Gaia mission with spectroscopic surveys (to obtain full phase-space information) for known streams and satellites, compare the orbital properties of these populations, and compare these populations with satellite and stream populations from cosmological simulations. This work would lay the groundwork for more detailed modeling of the merger rate and assembly history of the Milky Way.

Kung-Yi Su & Chris Hayward

We will study the magnetic field amplification and structure in massive galaxies and cluster ellipticals with halo masses ranging from 10^12-10^14 solar masses. We will start from the isolated galaxy simulations presented in Su et al. (2019, 2020) with different halo masses and AGN feedback toy models. This could be an extension of Su et al. 2018 (MNRAS 473, L111–L115) but with higher halo masses, hot halo gas, and different AGN feedback models. We want to do a more detailed analysis by analyzing the magnetic field power spectrum and field lines’ curvature regarding how much of the amplification is due to flux-freezing compression versus the turbulent dynamo and whether the picture can be altered by AGN feedback. We would also like to create Faraday rotation maps and compare them with observations. We would possibly also extend the study to some FIRE zoom-in simulations with halo masses of ~10^12 solar masses.

Matteo Cantiello & Adam Jermyn

More info at: https://www.stellarphysics.org/projects

Modeling Stellar Convection

Stellar modeling is limited by our understanding of turbulence. We are looking for a student interested in studying stellar convection, either near the star’s surface where it can cause measurable brightness fluctuations or in deeper regions where it plays a crucial role in distributing angular momentum and setting stellar rotation rates. These projects likely involve a combination of hydrodynamic simulations with the Athena or Dedalus software instruments, as well as semi-analytic work developing prescriptions for 1D stellar evolution instruments like MESA.

Stars in AGN Disks

Stars are likely formed in, or captured by, the disks of active galactic nuclei (AGN). The gas in these disks can accrete and block heat escaping from the star, and both effects can profoundly change the star’s evolution. This project could involve either modeling the accretion stream with radiation hydrodynamics software instruments like Athena, 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.

APOGEE Binaries

Binary stars represent a powerful probe of stellar evolution. Data from the APOGEE radial velocity survey have been used to identify a large population of binary main-sequence and red giant stars. This project would involve studying the chemical abundances, eccentricities, periods, and rotation periods of the APOGEE binaries, and leveraging semi-analytic models as well as 1D stellar evolutionary models to gain insight into how both single and binary stars evolve.

Megan Bedell

I am broadly interested in spectroscopy of Sun-like stars, stellar variability, and extreme precision radial velocity (EPRV) data analysis. Some general directions of student projects could include:
– application of the “wobble” RV extraction method to NIR spectra for improved tellurics mitigation
– data-driven approaches to spectral disentangling of SB2 systems
– modeling stellar spectral variability in EPRV data sets using data-driven methods
– uniform retrieval of planetary signals from EPRV survey data

Please email me to discuss further if you are interested in any of the above topics or have another idea for collaboration.

Melissa Ness

What remains in stellar spectra once you remove stellar parameters and age?

In Ness et al. (2019), we showed that the majority of information from the many abundance measurements for low-alpha disk stars is captured with two parameters, [Fe/H] and age. In this project, we will take a set of APOGEE stellar spectra and remove the variability at each wavelength caused by stellar parameters and age. The remaining variability in the spectra of the population is presumably a consequence of the intrinsic chemical element abundance spread. We will focus on the set of element absorption features in Feeney et al. (2019) and quantify the variance of these features not accounted for by Poisson noise. We will compare the results for the field disk stars to a population of open cluster stars – the stars of which presumably come from a single gas cloud. Using stellar models, we will calculate how any residual spectral variability maps to abundance variability, for 20 elements. This analysis will quantify signal to noise requirements for APOGEE spectra to measure any intrinsic abundance variation beyond overall ([Fe/H],age).

Natascha Manger & Phil Armitage

Turbulence properties in protoplanetary disks are still not fully understood, but current models and observations suggest that hydrodynamic instabilities can develop near the disk mid plane. Understanding which instabilities develop at different radii and how they interact with the disk magnetic field is crucial in finding a consistent model of early disk development and potential routes to planetesimal formation.
The student would work with Natascha Manger and Phil Armitage on projects related, but not limited to, simulations of VSI turbulence at different disk radii (optionally with non-ideal MHD), or dusty protoplanetary disks and the onset of planetesimal formation.

Robyn Sanderson

Understanding the orbit structure of cosmologically accreted stellar halos

The stellar halos of galaxies like the Milky Way are made up of stars accreted from many smaller galaxies, spread into tidal streams by the gravitational field of the host galaxy. Stars in the same tidal stream are expected to have similar orbital invariants (constants of motion) since they come from the same progenitor globular cluster or dwarf galaxy, which occupies a smaller total phase space volume than the total available in the host. In principle, stars can be assigned to a particular stream by association with the nearest invariant-space cluster, which would be a powerful way to determine the accretion history of our Galaxy. However, in a real galaxy some or all of these quantities will only be approximately invariant, both due to time-evolution in the potential (which could be non-adiabatic in some cases) and due to departures from perfect symmetry in the potential (which breaks the assumption of separable equations of motion underlying the choice of a coordinate system). Both of these types of symmetries are expected to be broken to different degrees at different locations in the gravitational potential, which can lead to distortion of the cluster corresponding to a single stream. Additionally, the extent of each structure in invariant space varies in proportion to the mass of the progenitor, while filamentary infall can lead several progenitors to create streams on similar orbits and hence partially overlap in invariant space. Several aspects of a realistic galactic potential and cosmological background can thus impact the certainty with which one can assign stars to a particular tidal stream.

The student will work with simulated stellar halos from the FIRE suite, as well as simulated Gaia surveys of these systems, to understand how these different effects influence the structure of [approximately] invariant space in cosmologically accreted stellar halos, especially compared to the effect of current measurement uncertainties in calculating orbital invariants. Possible topics of investigation include the effect on constraints on the mass distribution from action-space clustering, how to evaluate the certainty with which stars can be assigned to streams, and whether distortions in invariant space can point to specific discrepancies between a model and the actual galactic gravitational field.

Ruth Angus & Melissa Ness

Characterizing stars with TESS

TESS is a new NASA space mission which monitors the brightness variations (light curves) of hundreds-of-thousands of near-by stars. Stellar light curves reveal the rotation periods, masses and surface gravities of stars, which can be used to discover new stellar, planetary and Galactic astrophysics. Using TESS data, it is possible to measure or predict the rotation rates, masses, and perhaps even ages of stars using both classical techniques and machine learning. In this project, the student will apply new machine learning methods for analyzing TESS light curves and characterizing stars based on their time-variability. Time-permitting, they may use these stellar properties to study the relationships between planets and stars, or the evolution of the Galaxy itself.

Sarah Pearson & Robyn Sanderson

Galactic Bar Formation and Destruction in FIRE Simulations

The Milky Way Galaxy is a spiral galaxy and hosts a so-called Galactic “bar” at its center. The Galactic bar consists of billions of stars on orbits moving much faster than the rest of the stars in the Galaxy, many crossing quite near to the Galactic center. Together, the billions of stars form a coherent structure, which has persisted for at least several Gyrs. When we look out in the Universe, however, only half of spiral galaxies like our own Milky Way host galactic bars, and it is unknown why, when and how the bars form and get destroyed.

In this proposed project, we aim to investigate the signatures of bar formation and destruction in the FIRE simulations of Milky Way type galaxies. You will analyze the (already run) FIRE simulations and look for signatures in the distribution and motion of stars in the simulated galaxies as galactic bars form and dissolve. The hope is to predict how observers can look for these signatures in future observing programs and ultimately to answer why only some spirals host galactic bars.

Sasha Philippov

My main research area is relativistic plasma astrophysics. Interested students may work on subjects including (but not limited to) kinetic plasma simulations of pulsars, binary neutron star and black hole magnetospheres and astrophysical jets. These first-principles simulations are instrumental for understanding plasma production, particle acceleration and emission of nonthermal photons in the environments of compact objects. I’m also broadly interested in production of coherent radio emission in the universe, ranging from solar radio bursts to pulsar radio emission and the enigmatic fast radio bursts.

Shy Genel

The CAMELS project is a new suite of many thousands of cosmological simulations that cover a large high-dimensional space of both cosmological and baryonic sub-grid parameters, using both the fiducial IllustrisTNG and SIMBA models as bases. A range of research projects can be developed around this rich data set, focusing on machine learning applications as well as more traditional investigations of how different physical models affect galaxy formation, depending on the interests of the candidate. Discussion of additional ideas in numerical galaxy formation outside of the CAMELS context will also be welcome.

Simone Aiola & Sigurd Naess

Ground based surveys of the Cosmic Microwave Background (CMB) like the Atacama Cosmology Telescope (ACT) and Simons Observatory (SO) are able to map more than half of the sky with unprecedented resolution. Although located in one of the driest place on the planet, the noise from the atmosphere is still not negligible and can greatly reduce the containing power of the data –– being able to describe the noise properties of the multifrequency data is vital! Modern machine learning techniques should be able to help us describing the noise as function of basic parameters and inform the procedure to best compute reliable CMB maps.

With this project, the research will be exposed to machine learning techniques, map-making for CMB experiments, and map-space and Fourier-space analysis of maps.

Stephanie Tonnesen

I am generally interested in galaxy evolution, and on how galaxies quench. I run idealized simulations of individual galaxies, and study the galaxies and intergalactic gas in large-scale cosmological simulations. My interests are quite broad, so feel free to contact me with your own ideas for collaboration! Some projects that I would propose are:

What Causes Outside-In Star Formation in Cosmological Simulations?

Galaxies are generally found to have older bulges that may be embedded in young, star-forming disks. However, observations of low-mass galaxies have found that this age gradient may flatten and even invert (Gallart et al. 2008; Perez et al. 2013). In Tuttle & Tonnesen (2020) we found an observational sample of galaxies with star-forming bulges in red disks that reside at a range of masses. In this project, we would search large cosmological simulations, Simba and/or TNG, to find galaxies with outside-in star formation. Examining these galaxies will give us insight into what drives the distribution of star formation within galaxies.

Do Satellites Stir the CGM?

The CircumGalactic Medium (CGM) is an important reservoir of gas that has accreted into a galaxy halo, or been ejected from a galaxy disk. Because this gas is very difficult to observe, the distribution of gas density, temperature and velocity is poorly constrained. In this project, we (the student, Drummond Fielding, and myself) would examine z=0 Milky-Way mass galaxies in cosmological simulations, and determine the effect of satellites on the CGM. Some questions to consider are: Do satellites drive turbulence in the CGM? Induce cooling? Effect the metallicity?

Ram Pressure Stripping

Ram Pressure Stripping is a gas-removal mechanism that acts on satellite galaxies in which an interaction between the ISM and surrounding gas removes the ISM. Observations of these galaxies have led to several interesting questions: Will ram pressure enhance the feeding of a black hole? Does ram pressure induce star formation in the remaining gas disk? Under what circumstances does star formation occur in stripped gas? And many more! We would examine these questions using idealized “wind-tunnel” simulations of individual galaxies.

Ulrich Steinwandel & Chris Hayward

Star formation models with variable star formation efficiency

Typical star formation models in galaxy-scale simulations are tuned to the Schmidt-Kennicutt relation, which is mainly constrained by observations of nearby galaxies. In this picture, the star formation rate density increases with the gas surface density as a power law with a slope of around 1.5. The local star formation rate can be written as the well-known Schmidt law, \rho_{\star} = \epsilon_{eff} \cdot \frac{\rho}{t_{ff}}, where epsilon is the star formation efficiency, rho is the local gas density and t_{ff} is the local free-fall time. Many groups adopt a star formation model that operates with a constant star formation efficiency. This is mostly constrained by the sub-grid physics that they include in their codes. Many astrophysical galaxy formation codes adopt some sub-grid treatment in which the galactic ISM remains unresolved on sub-kiloparsec scales and thus the detailed turbulent motion of the gas that sets the star formation efficiency remains unresolved.

Recently, we have developed more realistic (resolved) ISM models for galactic-scale simulations, in which we form single stars, follow their stellar evolution and explode the massive ones in core-collapse supernova events. In every individual core-collapse event, the momentum and hot mass that is needed to drive supersonic turbulence and outflows is generated in the resolved Sedov-Taylor phase, the resolved pressure-driven snow-plough phase and the resolved momentum-conserving snow-plough phase. We therefore develop a highly turbulent and pressurized galactic ISM on all scales. In such a simulation, it is reasonable to assume that the ISM is resolved at sufficiently high resolution to actually account for the local variation in the star formation efficiency as a function of the Mach number of the local fluid flow. We would thus implement the star formation recipe from Kretschmer & Teyssier (2020) into our version of the GADGET code, which was recently updated to utilize the GIZMO hydro solver used for the FIRE-2 simulations.

The effect of runaway stars on dwarf galaxy evolution

The orbit of stars can be altered by several phenomena, including gravitational interactions within stellar many-body systems, stellar collisions and nearby supernova explosions. These effects can result in the ejection of a star from its local gravitationally bound group. These stars are called runaway stars, as they move faster than the average speed in their local galactic environment. Modern galactic-scale simulations can account for the formation of single stars but typically neglect the effect of runaway stars. However, in a galactic context, they could be very important if it is possible that they travel a very large distance before they undergo a core-collapse supernova (if they are more massive than 8 solar masses). Thus, they can explode in areas of much lower density compared to the density of the molecular cloud in which they originally formed.

In Steinwandel et al. (2019c), we showed that supernova feedback is effective at driving a strong galactic wind. Statistically, roughly 20 per cent of stars end up as runaway stars and could alter the outflow properties of galaxies. The goal of this project is to implement a simple model for runaway stars. This can range from random kicks to star particles at their time of formation to a more sophisticated approach that accounts for the binary population within the IMF. We will run two simulations like those presented in Hu et al. (2017; see also Steinwandel et al. 2019c, Figure 20). One simulation will incorporate the effect of runaway stars, and the other will not. The final goal is to determine the ability of both models to drive galactic outflows and investigate if the effect of runaway stars is significant. If the effect is significant, the simulation results could be used to construct a sub-grid model that could be applied in future large-scale cosmological galaxy formation simulations.

Yan-Fei Jiang

(Co-mentors: Yin Li & Shirley Ho)

Studying AGN variability with machine learning techniques

Active Galactic Nuclei (AGN) shows variabilities over a wide range of time scales and amplitude. Both numerical simulations and statistical studies of observational data suggest that the AGN lightcurve variability is closely connected to the fundamental properties of the central black holes (such as black hole mass) and the properties of its accretion disks (for example accretion rate). The student will work with Yan-Fei Jiang, Yin Li and Shirley Ho to apply machine learning techniques to study AGN variability for the first time. Specifically, the student will first use a set of observed and simulated AGN lightcurves with known black hole mass and accretion rate to train machine learning models, and recover the already established properties of AGN lightcurves. Then we will apply the network to a large dataset to see how this method works compared with the traditional technique.

Cosmic ray-driven outflows

Cosmic Rays have been realized to play an important role to drive outflows in galaxies. The second project will apply the recently developed two moment approach for CR magnetohydrodynamics to study the formation of cold clouds with thermal instability and how the cold clouds will behave in the outflow driven by CRs.

Advancing Research in Basic Science and MathematicsSubscribe to Flatiron Institute announcements and other foundation updates