Published using Google Docs
MKI Zoom-SMURF2021
Updated automatically every 5 minutes

Zoom-Summer MKI Undergraduate Research Forum (SMURF)

When: August 25th, 8:30 AM-11:15 PM (Eastern Daylight Time),

Where: Zoom: https://mit.zoom.us/j/93619739436

Welcome and Introduction -- Dheeraj Pasham and George Ricker

------------------------------ TALK SESSION #1 ---------------------------------

                               8:30-9:45 AM: 8 mins presentation + 2 mins for Q&A

-----------------------------------------------------------------------------

  1. Lyman-alpha transmission in the THESAN cosmic reionization simulations

Clara Xu (Supervisor: Mark Vogelsberger)

The Epoch of Reionization is a time period when radiation from the first stars and galaxies ionized the neutral hydrogen throughout the universe. Transmission statistics of the Lyman-alpha spectral line of hydrogen are sensitive to the neutral gas distribution and can thus provide valuable constraints on reionization models. To do so, we utilize the new Thesan suite of cosmological radiation-hydrodynamic simulations to accurately model the properties of both individual galaxies and the intergalactic medium. The high-resolution Thesan-1 simulation is accompanied by several medium-resolution Thesan-2 runs with different ionizing escape fractions, dark matter models, and numerical convergence. This project investigates the impact of those differences on Lyman-alpha transmission including dependence on redshift, frequency from line center, galaxy brightness, and anisotropic covering fractions.

  1. Escape Fractions in THESAN simulation

Yuan-Chen Yeh (Supervisor: Aaron Smith)

The escape fractions determine whether galaxies are able to reionize the Universe in the epoch of reionization. In this project, we use a Monte-Carlo radiative transfer solver to explore the properties of galaxies that are responsible for the reionization.

  1. Planning optical telescope follow-up observations of fast radio bursts

Obinna Modilim (Supervisor: Kaitlyn Shin)

Over the course of this summer UROP, I worked toward the goal of creating a short list of notable FRBs to observe with the Magellan telescope in the hope of collecting more data on these bursts and potentially identifying host galaxies for them. To that end, I have looked into the measurable quantities of FRBs and decided which qualities could provide the most interesting follow-ups. After the desired parameters were picked, various Python scripts were used to filter the L4 database of FRBs until a modest list of about 25 candidates was obtained. Other candidates that were particularly interesting, like certain repeater FRBs, were also added to the list. Next, the candidates on the list were checked to see which would be visible during the first quarter of 2022, which is when the next opening for Magellan telescope time is available. With these steps complete, I have paved the way for writing a complete and justified telescope proposal for winter 2022.

  1. A New Application of Malmquist Bias: Caustic Areas, Time Delays, and Biases in the Measurement of the Hubble Constant

Derek Baldwin (Supervisor: Paul L. Schechter)

Quadruply lensed quasars are visible only when the source quasar lies within the diamond caustic of the lensing galaxy. This condition creates a Malmquist-like selection effect in the population of observed quadruply lensed quasars, biasing the inferred caustic areas low. The inferred time delays change as the square root of the caustic area, meaning model time delays are also biased low. Because the Hubble constant is inversely proportional to the time delay, this means that current measurements of H0 using quadruply-lensed quasars are biased low.

  1. Quadruple image configurations from asymptotically circular gravitational lenses

Chirag Falor (Supervisor: Paul L. Schechter)

The simplest models of quadruply lensed quasars span a three-dimensional subspace, discounting for translation, rotation, or rescaling. We consider the quadruple image configurations of “asymptotically circular” gravitational lenses, reducing the number of parameters to 2. As the "tangential" caustic — a kind of diamond-shaped bullseye — becomes vanishingly small, we find that keeping the position of source constant with respect to the caustic enables us to get stable configurations. We calculate the relative magnifications of images in this limiting case. Then, we reintroduce a non-trivial quadrupole in the form of an external shear and show how the circular configurations naturally give rise to the sheared ones. Finally, we show that these configurations aren’t just limited to gravitationally lensed quasars but also emerge in occultations of stars by planets in our own solar system.

  1. Machine Learning based Stellar Variability Classification using TESS lightcurves

Manan Agarwal (Supervisor: Michael Fausnaugh)

The Transiting Exoplanet Survey Satellite (TESS) produces high-quality lightcurves with a baseline of at least 27 days and cadence of 30 minutes for sources across the whole sky. This provides an unprecedented opportunity to classify and study stellar variability in great detail. We built a neural network based multi-class classifier to automatically classify sources which are flagged as variable by GAIA. The combination of GAIA determined photometry (magnitudes in G, BP and RP) and empirical stellar parameters (effective temperature, luminosity and radius), along with the photometric data from TESS FFIs, is used as input to the neural network model. We classify sources into six broad categories: RR Lyrae stars, Cepheids, Delta Scuties, Eclipsing variables, rotation modulated stars and long period variables. The classifier has an accuracy of 99% at identifying long period variables and an accuracy of around 93% for RR Lyraes and Cepheids. We classify more than 200,000 sources and find new candidate variable sources which are not present in SIMBAD. By applying anomaly detection algorithms and manual inspection of lightcurves, we also find some rare astronomical objects with strange looking lightcurves.

  1. Using Machine Learning for transient detection

Muhammad Abdullah (Supervisor: Rahul Jayaraman)

The group I work with does a blind search for transients in each TESS observing sector. The current workflow requires human review of over two thousand light curves every month. Over the summer, we designed a pipeline that uses a combination of machine learning and statistical techniques to classify and prioritize candidates for human review. We have currently achieved a reduction of ~90%.  

----------------------------- Break: 9:45-10:00 AM ------------------------------

------------------------------ POSTER SESSION ---------------------------------

                                  10-10:10 AM: 3 mins presentation + 2 min for Q&A

-----------------------------------------------------------------------------

1. TESS: Classifying Variable Stars with Decision Trees

Ally Hong (Supervisor: Michael Fausnaugh)

Using TESS data, I have been building a variable star classifier. This summer, I have focused on training decision trees through analyzing branches and cluster groups inherent in the data – as well as experimenting with different input data sets to study machine-made decisions while maximizing accuracy. This poster pitch will showcase interesting tree plots, explore methods of analysis, and discuss steps moving forward.

2. Characterizing Variability in the Sun to Better Understand Removal of Stellar Activity Signals That Interfere With Exoplanet Detection

Archana Mohandas (Supervisor: Gabor Furesz)

Over 4000 planets orbiting other stars (exoplanets) have been discovered in the past two decades, and these exoplanets are helping to inform scientists about how planets form and evolve, and to place our own solar system and Earth in context. One of the main hurdles in detecting Earth-analogs is that the stars around which they orbit produce activity signals that are much larger than the signal from the planet. A promising approach to better characterizing and removing these stellar activity signals is to start by studying our own Sun, where we can obtain higher cadence and higher signal-to-noise data than is possible for any other star.

Using newly obtained spectra of the sun taken with a set of USB spectrographs at Lowell Observatory in AZ, we developed methods for removing windows of time in the data impacted by various types of clouds in order to create a ‘clean’ data set that can be used to study how different regions of the solar spectrum are impacted by the presence of active regions (e.g. sun spots) on the surface of the sun. This analysis will serve as an important step towards motivating future, space-based, stellar activity satellites by determining whether or not ground-based data can provide the same scientific impact as satellite missions despite the interference of the Earth’s atmosphere.

3. Exploring the discrepancy of hot Jupiter occurrence rates between Kepler and radial velocity data

Maya Beleznay (Supervisor: Michelle Kunimoto)

This project utilizes data from the MIT Transiting Exoplanet Survey Satellite (TESS) to contribute to the study of exoplanet occurrence rates, which could provide valuable information for planet evolution theories. We search for hot Jupiters around FGK dwarf stars (which range in temperature from 3500 to 7500 degrees Kelvin), in order to enable a comparison with planets around FGK stars observed by the Kepler mission. The results of this study could contribute to understanding the large discrepancy in the occurrence rate of hot Jupiters around FGK dwarf stars found by Kepler and radial velocity surveys. We also aim to search for possible correlation between stellar metallicity, mass and effective temperature and hot Jupiter occurrence rates by selecting for a stellar group that has metallicity data recorded in the TESS stellar catalogue.

------------------------------ TALK SESSION #2 --------------------------------

                           10:10 AM-11:15 PM: 8 mins presentation + 2 mins for Q&A

-----------------------------------------------------------------------------

1. Discovery of a candidate Quasi-Periodic Eruption from a systematic XMM-Newton search

Joheen Chakraborty (Supervisor: Erin Kara)

X-ray Quasi-Periodic Eruptions (QPEs) are a recently discovered phenomenon associated with supermassive black holes at the centers of galaxies. They are high-amplitude soft X-ray flares that recur on timescales of hours, but what causes these bursts remains unknown. In the two years since their original discovery, four known QPE-hosting galaxies have been found, with varying properties and levels of activity. We have conducted a blind algorithm-assisted search of the XMM-Newton Source Catalog and found a candidate fifth QPE source, XMMSL1 J024916.6-041244, a star-forming galaxy at z=0.0186 hosting an AGN. In this talk, I will cover (I) the strategy we used to systematically search through XMM archival data; (II) the properties of the host galaxy and its QPE flares; and (III) the relative behaviors and properties of the QPE sample to date, now 5 members large.

2. Kinematic clustering of ultra-faint dwarf galaxies

Hillary Diane Andales (Supervisor: Anna Frebel)

The Milky Way stellar halo largely contains stars accreted from dwarf galaxies. The smallest and oldest of these dwarfs–the ultra-faint dwarf galaxies (UFDs)–are vital building blocks in the Galaxy’s formation. Since the kinematics of halo stars retain information about their origins, a cluster of halo stars in kinematic phase space may be the remnant of a progenitor UFD–but this correspondence is uncertain. Here, we examine how well clustering algorithms identify true UFD remnants. Using simulations of Milky Way-like halos from the Caterpillar suite, we compare six different clustering algorithms (HDBSCAN, Friends-of-Friends, K-Means, Mean-shift, Agglomerative Clustering, Gaussian mixtures) and find that HDBSCAN performs best at identifying true UFD remnants–albeit with a large scatter. On average, among HDBSCAN clusters within 5 kpc of the Sun, only one (15.5+34.5-15.5%) is a true remnant. For clusters within 50 kpc, around five (7.7+42.3-5.2%) are true remnants. The other algorithms achieve only 5.0% at best. We also find that all clustering algorithms perform best at identifying true remnants with higher kinetic energies, higher velocities, and more recent accretion times.

3. Comparison of sky-maps from multiple gravitational-wave event-finding methods

Pinchen Fan (Supervisor: Erik Katsavounidis)

Real-time searches for GW transients with LIGO-Virgo invoke multiple searches for the identification of such events, each resulting in a separate sky map that describe the error area of the candidate source. In this work, we study the similarities of sky maps from the different methods considered and compare them with ones obtained from offline analyses that use sophisticated stochastic samplers. Understanding the relatedness of such sky maps obtained in real time is important for the prompt follow-up of GW candidates in the electromagnetic spectrum and in neutrinos, as well as in helping discriminate noise events from signals. We will report on a preliminary analysis of these sky maps using a subset of the public alerts from LIGO-Virgo in O3.

4. Analysis of iDQ Glitch Detection Statistics During LIGO’s Standard Veto Flags

Nezir Alic (Supervisor: Erik Katsavounidis)

To optimize gravitational wave searches, it is necessary for LIGO and other interferometric gravitational-wave detectors to accurately distinguish between true astrophysical signals, and non-Gaussian noise, or “glitches.” This can be difficult at times due to the similarity between these two types of signals. However, algorithms exist that address this problem, including iDQ, which provides low-latency, probabilistic information such as the likelihood of a glitch. Statistical comparisons between the conclusions of iDQ and those of human-made veto definers are presented. Segments of active detector time from both LIGO detectors that are flagged by such veto definers as vetoes of various categories, as well as segments that are not flagged, are selected. The iDQ values of these times are then analyzed in order to establish whether the probability of a glitch, as determined by iDQ, differs in a statistically significant way during veto times. Preliminary results suggest that there may indeed be a particularly high p(glitch) for those times corresponding to veto flags, but this requires more review and verification; the project is not yet complete.

5. Algorithm and Interface Development for a Manual and Robotic Telescope

Cruz Soto and Josh Glass (Supervisor: Robert Simcoe)

WINTER telescope research team: in which Josh Glass has developed a GUI for manual use of the telescope and Cruz Soto has worked on software to find precise focus before shooting. They are also starting on scientific observations after wrapping up most of our software development, taking aim at deep sky objects, namely supernovae and transits.

6. Using Scintillation to Probe the Host Environments of Fast Radio Bursts

Eve Schoen (Supervisors: Calvin Leung/Kiyoshi Masui)

Fast Radio Bursts (FRBs) are short pulses of powerful radio waves. Not much is known for sure about their host environments, but through measuring FRB scintillation--the same phenomenon that causes stars to twinkle--we can learn about their host environments. By observing the timescales of variations in intensity of an FRB through autocorrelating the spectrum, we can probe the variations in electron densities of the host environment and the Milky Way. We measure the autocorrelation function (ACF) for several FRBs detected by CHIME/FRB and comment on the interpretation of the result.