Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Physics and Astronomy

Committee Chair/Advisor

Dr. Marco Ajello

Committee Member

Dr. Dieter Hartmann

Committee Member

Dr. Mark Leising

Committee Member

Dr. Hugo Sanabria


Active galactic nuclei (AGN) are supermassive black holes (SMBHs) in the center of galaxies that accrete surrounding gas and emit across the entire electromagnetic spectrum. They are the most energetic persistent emitters in the Universe, capable of outshining their host galaxies despite their emission originating from a region smaller than our Solar System. AGN were some of the first sources discovered that helped teach us that there were galaxies outside of our own, and they proved the existence of black holes. Moreover, AGN can give us valuable insights into other branches of astrophysics. For example, they can be used to study the center of SMBHs and learn more about black-hole physics; test aspects of General Relativity; analyze the intergalactic medium from the early Universe; and constrain crucial cosmological parameters and the nature of dark energy. AGN are an invaluable astronomic tool and thus it is crucial to understand how they operate and how they have evolved over time.

One of the best ways to study AGN through cosmic time is via the cosmic X-ray background (CXB), i.e., the diffuse X-ray emission from 1 to ∼200-300 keV. For example, the CXB can be used to constrain the X-ray luminosity function (XLF) and thus population synthesis models going back through time. While almost all of the low-energy (≤ 2 keV) CXB has been resolved into point sources, there is a significant portion (15-20%) of the peak of the CXB (∼30 keV) that remains unaccounted for. It is believed this peak is generated by a largely undetected population of heavily obscured AGN with line-of-sight column density > 10^24 cm^−2 . Recent population synthesis models predict these so-called Compton-thick (CT-) AGN represent between 20% and 50% of all AGN. However, the current observed fraction of CT-AGN is between 5 and 10%. Therefore, new methods are needed to detect this important subclass of AGN.

In this thesis, I discuss two novel methods designed to solve this problem by identifying CT-AGN in the local Universe (z ≤ 0.1). First, we selected known Seyfert 2 (Sy2) galaxies or sources identified as galaxies with low redshifts and no counterparts from ROSAT, and obtained short 10 ks observations from Chandra. This Chandra data was jointly fit with spectra from Swift-BAT using physically motivated models. Of the nine sources I analyzed, two were determined to be CT-AGN candidates. Then, we obtained simultaneous deep NuSTAR and XMM-Newton observations of a total of four CT-AGN candidates. We found one source to be unobscured and highly variable in flux, two sources to be Compton-thin (10^23 < N_H < 10^24 cm^−2 ), and one source to be a confirmed CT-AGN.

Our second method utilized a multiple linear regression machine learning algorithm to accurately predict line-of-sight column density values based on mid-infrared (MIR), soft, and hard X-ray data. Using WISE colors, Swift-BAT count rates, soft X-ray hardness ratios, and an MIR-soft X-ray flux ratio, we were able to accurately classify 75% of our test sample as either unobscured (< 10^22 cm^−2 ), obscured (10^22 < N_H < 10^23 cm^−2 ), Compton-thin, or CT-AGN. This represents a dramatic improvement over previously published methods, which yielded 42% and 30% accurate classifications. In particular, our method was highly accurate (78%) at identifying unobscured AGN when compared to these previous methods (8% and 11%). This method will be used to help identify the missing CT-AGN and resolve the remaining portions of the CXB.

This thesis is organized as follows: Section 1 discusses the history and basics of studying AGN in X-rays. Section 2 details the Chandra-BAT analysis of nine AGN, while Section 3 details the NuSTAR-XMM analysis of four CT-AGN candidates. In Section 4, I discuss the machine learning algorithm and compare the results to previous methods. Finally, Appendix A details a paper on identifying X-ray candidates of γ-ray detected sources, and uses a machine learning algorithm to identify the subclass of the likely blazars.

Author ORCID Identifier




To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.