Specialized Noise Elimination in Astronomical Data using Deep Learning

193524-Thumbnail Image.png
Description
Astronomy has a data de-noising problem. The quantity of data produced by astronomical instruments is immense, and a wide variety of noise is present in this data including artifacts. Many types of this noise are not easily filtered using traditional

Astronomy has a data de-noising problem. The quantity of data produced by astronomical instruments is immense, and a wide variety of noise is present in this data including artifacts. Many types of this noise are not easily filtered using traditional handwritten algorithms. Deep learning techniques present a potential solution to the identification and filtering of these more difficult types of noise. In this thesis, deep learning approaches to two astronomical data de-noising steps are attempted and evaluated. Pre-existing simulation tools are utilized to generate a high-quality training dataset for deep neural network models. These models are then tested on real-world data. One set of models masks diffraction spikes from bright stars in James Webb Space Telescope data. A second set of models identifies and masks regions of the sky that would interfere with sky surface brightness measurements. The results obtained indicate that many such astronomical data de-noising and analysis problems can use this approach of simulating a high-quality training dataset and then utilizing a deep learning model trained on that dataset.
Date Created
2024
Agent

Modeling SKYSURF Completeness Data for Comparison to the Hubble Space Telescope Exposure Time Calculator

Description
Using Wide Field Camera 3 (WFC3) data from the Hubble Space Telescope (HST) archival project "SKYSURF", we model completeness with respect to the exposure time and background of an image. This is accomplished by adding simulated objects with varying magnitudes

Using Wide Field Camera 3 (WFC3) data from the Hubble Space Telescope (HST) archival project "SKYSURF", we model completeness with respect to the exposure time and background of an image. This is accomplished by adding simulated objects with varying magnitudes and sizes into these HST images, and determining the matching rate for each set of parameters. The fifty percent completeness results then can be compared to the Exposure Time Calculator (ETC), in order to assess the differences between it and our analysis of the archive data. We find that for larger objects and exposures the ETC predicts higher completeness magnitudes, while for smaller objects, the ETC predicts lower magnitudes.
Date Created
2024-05
Agent

Impact of Low-Frequency Antenna Characteristics on Observations of Cosmic Dawn with EDGES and FARSIDE

172012-Thumbnail Image.png
Description
Standard cosmological models predict that the first astrophysical sources formed from a Universe filled with neutral hydrogen (HI) around one hundred million years after the Big Bang. The transition into Cosmic Dawn (CD) that seeded all the structures seen today

Standard cosmological models predict that the first astrophysical sources formed from a Universe filled with neutral hydrogen (HI) around one hundred million years after the Big Bang. The transition into Cosmic Dawn (CD) that seeded all the structures seen today can only be probed directly by the 21-cm line of neutral hydrogen. Redshifted by the Hubble expansion, HI signal during CD is expected to be visible in radio frequencies. Precisely characterized and carefully calibrated low-frequency instruments are necessary to measure the predicted ~10-200 mK brightness temperature of this cosmological signal against foregrounds. This dissertation focuses on improving the existing instrumental and analysis techniques for the Experiment to Detect the Global EoR Signature (EDGES) and building capabilities for future space-based 21-cm instruments, including the Farside Array for Radio Science Investigations of the Dark ages and Exoplanets (FARSIDE) concept.Frequency-dependent antenna beams of 21-cm instruments limit the removal of bright galactic foreground emission (~10^3 - 10^4K) from observations. Using three electromagnetic simulation packages, I modeled the EDGES low-band antenna, including the ground plane and soil, and quantified its variations as a function of frequency. I compared simulated observations to sky data and obtained absolute agreement within 4% and qualitatively similar spectral structures. I used the new open-source edges-analysis pipeline to carry out rigorous fits of the absorption feature on the same low-band data and lab calibration measurements as (Bowman et. al. 2018). Using a Bayesian framework, I tested a few calibration choices and found posteriors of the best-fit 21-cm model parameters well within the 1σ values reported in B18. To test for the ``global'' nature of the reported cosmic absorption feature, I performed a time-dependent analysis. Initial results from this analysis successfully retrieved physical estimates for the foregrounds and estimates of the cosmic signal consistent with previous findings. The array layout of FARSIDE, a NASA probe-class concept to place a radio interferometer on the lunar farside, is a four-arm spiral configuration consisting of 128 dual-polarized antennas with a spatial offset between the phase centers of its orthogonal polarizations. I modeled the impact of direction-dependent beams and phase offsets on simulated observations of all four Stokes parameter images of a model and quantified its effects on the two primary science cases: 21-cm cosmology and exoplanet studies.
Date Created
2022
Agent

Titanium, Iron, Neutrinos, and Asymmetric Explosions: An Exploration of Supernovae and Their Remnants

168500-Thumbnail Image.png
Description
The lives of high-mass stars end with core-collapse supernovae, which distribute energy and chemical elements into the interstellar medium. This process is integral to the Galactic ecosystem, since stars and planets will form from the enriched interstellar medium. Since most

The lives of high-mass stars end with core-collapse supernovae, which distribute energy and chemical elements into the interstellar medium. This process is integral to the Galactic ecosystem, since stars and planets will form from the enriched interstellar medium. Since most supernovae are detected at intergalactic distances, opportunities to examine them in detail are rare. Computer simulations and observations of supernova remnants are frequently employed to study these events and their influence on the universe. I explore the topic of supernovae using a multi-pronged approach, beginning with an examination of the core-collapse supernova engine. The radioisotopes 44Ti and 56Ni, produced in the innermost ejecta, provide a probe of this central engine. Using a three-dimensional supernova simulation with nucleosynthesis post-processing, I examine the production of these isotopes and their thermodynamic histories. Since production of 44Ti is especially sensitive to the explosion conditions, insights can be gained by comparing the model with 44Ti observations from supernova remnant Cassiopeia A. Next, I consider supernova remnants as potential sources of high-energy neutrinos within the Milky Way galaxy. The developing field of neutrino astronomy has yet to identify the origins of the diffuse neutrino flux first detected by the IceCube Neutrino Observatory in 2013. In principle, high-energy Galactic sources like supernova remnants could contribute measurably to this flux. I also consider Galactic open clusters, environments which are rich in supernovae and other energetic phenomena. Statistical analysis finds no evidence of causal association between these objects and the IceCube neutrino events. I conclude with a series of asymmetric three-dimensional supernova models, presented as a comparative analysis of how supernova morphology affects nucleosynthetic yields. Both real supernovae and simulations frequently exhibit aspherical morphologies, but the detailed thermodynamic consequences and the ultimate effects on yields are poorly understood. The simulations include symmetric and bipolar explosion geometries for both 15- and 20-solar-mass progenitor stars. Across the spectrum of models, I show how small changes in the peak temperatures and densities experienced by ejecta can influence the production of notable isotopes such as 44Ti.
Date Created
2021
Agent

Properties of the Faint (uJy) Radio Source Population

164759-Thumbnail Image.png
Description
I present a multi-spectral analysis of the faint, uJy, radio source population in the James Webb Space Telescope North Ecliptic Pole Time Domain Field. Very Long Baseline Array pointings at the 127 brightest of ~2500 radio galaxies identified with the

I present a multi-spectral analysis of the faint, uJy, radio source population in the James Webb Space Telescope North Ecliptic Pole Time Domain Field. Very Long Baseline Array pointings at the 127 brightest of ~2500 radio galaxies identified with the Very Large Array indicate active galactic nucleus contamination of approximately 9.45%. My estimates of 4.8 GHz brightness of this radio source population indicate an upper bound on this contamination of 10.6%. This is well within acceptable limits, in population studies, for the use of the radio-FIR relation in the JWST NEP TDF. This improves the utility of the field to the community by reducing the need for expensive FIR observations. I have also developed an extensive catalog of magnitudes and other data in visible bands of this population. My analysis in these bands does not give any conclusive criteria for distinguishing between AGN and SFGs. The strongest trends I do identify appear to be due to reddening by interstellar dust. Future follow-up will focus on characterizing individual sources in further depth.
Date Created
2022-05
Agent

SED Analysis of 43 Spectroscopically Confirmed Galaxies at z ~ 6 to Constrain Possible Relationships between UV Slope, Model Dust Attenuation, and Escape Fraction

148332-Thumbnail Image.png
Description

The reionization of the Universe is thought to have completed by redshift z~5.5. To probe this era, galaxy observations in the Subaru Deep Field (SDF) have identified more than 100 galaxies at z~6, many spectroscopically confirmed through follow-up observations. Using

The reionization of the Universe is thought to have completed by redshift z~5.5. To probe this era, galaxy observations in the Subaru Deep Field (SDF) have identified more than 100 galaxies at z~6, many spectroscopically confirmed through follow-up observations. Using available optical/IR data, we model with CIGALE the spectral energy distributions (SEDs) of 43 SDF galaxies, including newly acquired data from the UKIRT WFCAM K-band for seven previously studied objects. In particular, modeling deep IR photometry is sensitive to the galaxy's Lyman continuum (LyC) escape fraction (fesc). We find the median implied fesc value as ~0.4+/-0.1 (mean error). Significant uncertainties in data and fitting result in a large range of fesc for individual objects, but analysis suggests that fesc is likely high enough for galaxies to finish reionization by z~6. More importantly, we find trends between the CIGALE UV slope b, fesc, and dust extinction E(B-V): for a given E(B-V), b appear steeper by ~0.4 than at z=0. Lower fesc values appear to be associated with bluer b and lower E(B-V), but only weakly. This suggests that LyC could have escaped through holes with sufficiently wide opening angles surrounding the ISM from outflows of supernovae and/or weak AGN to escape, but resulting in a large range of implied fesc values depending on the orientation of each galaxy. The current HST, Spitzer and ground-based photometric and model errors for the 43 galaxies are large, so IR spectroscopic observations with the James Webb Space Telescope are needed to better constrain this possibility.

Date Created
2021-05
Agent

Galaxy Evolution in the Local and the High-z Universe Through Optical+near-IR Spectroscopy

158521-Thumbnail Image.png
Description
A key open problem within galaxy evolution is to understand the evolution of galaxies towards quiescence. This work investigates the suppression of star-formation through shocks and turbulence at low-redshift, and at higher-redshifts, this work investigates the use of features within

A key open problem within galaxy evolution is to understand the evolution of galaxies towards quiescence. This work investigates the suppression of star-formation through shocks and turbulence at low-redshift, and at higher-redshifts, this work investigates the use of features within quiescent galaxy spectra to redshift estimation, and passive evolution of aging stellar populations to understand their star-formation histories.

At low-$z$, this work focuses on the analysis of optical integral field spectroscopy data of a nearby ($z\sim0.0145$) unusual merging system, called the Taffy system because of radio emission that stretches between the two galaxies. This system, although a recent major-merger of gas-rich spirals, exhibits an atypically low star-formation rate and infrared luminosity. Strong evidence of shock heating as a mechanism for these atypical properties is presented. This result (in conjunction with many others) from the nearby Universe provides evidence for shocks and turbulence, perhaps due to mergers, as an effective feedback mechanism for the suppression of star-formation.

At intermediate and higher-$z$, this work focuses on the analysis of Hubble Space Telescope (HST) Advanced Camera for Surveys (ACS) G800L grism spectroscopy and photometry of galaxies with a discernible 4000\AA\ break. The usefulness of 4000\AA/Balmer breaks as redshift indicators by comparing photometric, grism, and spectrophotometric redshifts (SPZs) to ground-based spectroscopic redshifts, is quantified. A spectral energy distribution (SED) fitting pipeline that is optimized for combined HST grism and photometric data, developed for this project, is presented. This pipeline is a template-fitting based routine which accounts for correlated data between neighboring points within grism spectra via the covariance matrix formalism, and also accounts for galaxy morphology along the dispersion direction. Evidence is provided showing that SPZs typically improve the accuracy of photometric redshifts by $\sim$17--60\%. For future space-based observatories like the Nancy Grace Roman Space Telescope (formerly the Wide Field InfraRed Survey Telescope, i.e., WFIRST) and Euclid, this work predicts $\sim$700--4400 galaxies\,degree$^{-2}$, within $1.6 \lesssim z \lesssim 3.4$, for galaxies with 4000\AA\ breaks and continuum-based redshifts accurate to $\lesssim$2\%.

This work also investigates the star-formation histories of massive galaxies ($\mathrm{M_s \geq 10^{10.5}\, M_\odot}$). This is done through the analysis of the strength of the Magnesium absorption feature, Mgb, at $\sim$5175\AA. This analysis is carried out on stacks of HST ACS G800L grism data, stacked for galaxies binned on a color vs stellar mass plane.
Date Created
2020
Agent

Attempting to Produce Helium White Dwarfs From Low-Metallicity Solar Mass Stars

132092-Thumbnail Image.png
Description
I examine the effects of metallicity on solar mass stellar evolution, trying to replicate a previous result in Windhorst et.al., 2018, in which a zer metallicity solar mass star did not reach the AGB, and thus may turn into a

I examine the effects of metallicity on solar mass stellar evolution, trying to replicate a previous result in Windhorst et.al., 2018, in which a zer metallicity solar mass star did not reach the AGB, and thus may turn into a helium white dwarf. In trying to replicate this result, I used the M.E.S.A. stellar evolution code and was unable to reproduce this result. While M.E.S.A has undergone several updates since the previous result was obtained, more current evidence suggests that this may have been a one-time occurrence, as no helium white dwarfs were produced for low-metallicity models. Nonetheless, interesting results were obtained, including a lowest metallicity value for which CNO burning does not significantly contribute during the main sequence, 1 −10 Z , which produces noticeable effects on post main sequence evolution. All models are run with no rotation, one solar mass, and a series of MESA parameters kept constant, with the only exception being metallicity. Any metallicity value listed as Nd −10 is an absolute mass fraction, and Z is relative to solar metallicity, 2d*10 −2 .
Date Created
2019-12
Agent

The study of astronomical transients in the infrared

157735-Thumbnail Image.png
Description
Several key, open questions in astrophysics can be tackled by searching for and

mining large datasets for transient phenomena. The evolution of massive stars and

compact objects can be studied over cosmic time by identifying supernovae (SNe) and

gamma-ray bursts (GRBs) in other

Several key, open questions in astrophysics can be tackled by searching for and

mining large datasets for transient phenomena. The evolution of massive stars and

compact objects can be studied over cosmic time by identifying supernovae (SNe) and

gamma-ray bursts (GRBs) in other galaxies and determining their redshifts. Modeling

GRBs and their afterglows to probe the jets of GRBs can shed light on the emission

mechanism, rate, and energetics of these events.

In Chapter 1, I discuss the current state of astronomical transient study, including

sources of interest, instrumentation, and data reduction techniques, with a focus

on work in the infrared. In Chapter 2, I present original work published in the

Proceedings of the Astronomical Society of the Pacific, testing InGaAs infrared

detectors for astronomical use (Strausbaugh, Jackson, and Butler 2018); highlights of

this work include observing the exoplanet transit of HD189773B, and detecting the

nearby supernova SN2016adj with an InGaAs detector mounted on a small telescope

at ASU. In Chapter 3, I discuss my work on GRB jets published in the Astrophysical

Journal Letters, highlighting the interesting case of GRB 160625B (Strausbaugh et al.

2019), where I interpret a late-time bump in the GRB afterglow lightcurve as evidence

for a bright-edged jet. In Chapter 4, I present a look back at previous years of

RATIR (Re-ionization And Transient Infra-Red Camera) data, with an emphasis on

the efficiency of following up GRBs detected by the Fermi Space Telescope, before

some final remarks and brief discussion of future work in Chapter 5.
Date Created
2019
Agent

Correlating Galactic Magnetic Fields with Regions of Dense Star Formation using LOFAR and CALIFA

132911-Thumbnail Image.png
Description
I test the hypothesis that galactic magnetic fields originate from regions of dense
star formation (Dahlem et al. 2006) by comparing maps of 120-240 MHz synchrotron emission and hydrogen alpha (Hα) emission of the tidally-interacting, edge-on, barred spiral galaxy UGC 9665.

I test the hypothesis that galactic magnetic fields originate from regions of dense
star formation (Dahlem et al. 2006) by comparing maps of 120-240 MHz synchrotron emission and hydrogen alpha (Hα) emission of the tidally-interacting, edge-on, barred spiral galaxy UGC 9665. Synchrotron emission traces magnetic field strength to a rough first order, while Hα emission traces recent massive star formation. UGC 9665 was selected because it was included in the LOw Frequency ARray (LOFAR) TwoMetre Sky Survey (LoTSS; Shimwell et al. (2017)) as well as the Calar Alto Legacy Integral Field Area Survey (CALIFA; Sanchez et al. (2012)). I generated vertical intensity profiles at several distances along the disk from the galactic center for synchrotron emission and Hα in order to measure how the intensity of each falls off with distance from the midplane. In addition to correlating the vertical profiles to see if there is a relationship between star formation and magnetic field strength, I fit the LOFAR vertical profiles to characteristic Gaussian and exponential functions given by Dumke et al. (1995). Fitting these equations have been shown to be good indicators of the main mode of cosmic ray transport, whether it is advection (exponential fit) or diffusion (Gaussian fit) (Heesen et al. 2016). Cosmic rays originate from supernova,
and core collapse supernovae occur in star forming regions, which also produce
advective winds, so I test the correlation between star-forming regions and advective regions as predicted by the Heesen et al. (2016) method. Similar studies should be conducted on different galaxies in the future in order to further test these hypotheses and how well LOFAR and CALIFA complement each other, which will be made possible by the full release of the LOFAR Two-Metre Sky Survey (LoTSS) (Shimwell et al. 2017).
Date Created
2019-05
Agent