On The Spectral Energy Distributions Of Dust-obscured Starburst-AGN Composite Galaxies

Since violent formation of stars (i.e. Starburst) and growth of SMBHs (i.e.active galactic nucleus, AGN) are coupled and ongoing together, dust-obscuredStarburst-AGN composite galaxies represent important phases in the formationand evolution of galaxies, and are ideal laboratories for studying starburst-AGNconnections. The spectral energy distributions (SEDs) are encoded with infor-mation about star formations, stellar populations, interstellar mediums (ISM)and AGN in these interesting objects. However, it is currently still very chal-lenging to efciently extract the basic physical properties of these galaxies fromthe analysis of their SEDs.In this thesis, we frstly described the feld of the analysis of SEDs (or ft-ting), including the basic concept of SED, the construction of observed multi-wavelength SED of galaxies, and the methods for the analysis of SEDs. Then,we present our work on this feld in detail.The luminosity function (LF) of AGN is an important observable quantityfor our understanding of AGN activities and their evolution. In practice, the LFsof AGN are measured independently from diferent wavelength bands, all of whichare sufered by diferent limitations. To obtain an accurate determination of theLF of AGN, multi-wavelength observations need to be combined self-consistently.Given these, we present a detailed comparison between the210keV hard X-ray and infrared (IR) LF of AGN, by employing a simple but well tested modelfor the SEDs of AGN. We fnd that the IRLFs predicted from HXLFs tend tounderestimate the number of the most IR-luminous AGN. This is independent ofthe choices of HXLF, and even more obvious for the HXLFs recently measured.We show that the discrepancy between the independently obtained HXLFs andIRLFs of AGN can be largely resolved when the anticorrelation between the UVto X-ray slope αoxand UV luminosity LUVis appropriately considered in themodel for the SEDs of AGN. We also discuss other possible explanations for thediscrepancy, such as the missing population of Compton-thick AGN and possible contribution of star-formation in the host to the mid-IR. Meanwhile, we fndthat the HXLFs and IRLFs of AGN can be more consistent with each other ifthe obscuration mechanisms of Quasars and Seyferts are assumed to be diferent.This is consistent with the widely accepted idea that the mechanisms of the twotypes of AGN are fundamentally diferent.The dust-obscured Starburst-AGN composite galaxies are more complicatedthan AGN, and the analysis of their multi-wavelength SEDs is more challenging.Given these, we have built BayeSED, a general purpose tool for doing Bayesiananalysis of SEDs by using whatever pre-existing model SED libraries or theirlinear combinations. The artifcial neural networks (ANNs), principal compo-nent analysis (PCA) and multimodal nested sampling (MultiNest) techniquesare employed to allow a highly efcient sampling of posterior distribution andthe calculation of Bayesian evidence. As a demonstration, we apply this toolto a sample of hyperluminous infrared galaxies (HLIRGs). The Bayesian evi-dences obtained for a pure Starburst, a pure AGN, and a linear combination ofStarburst+AGN models show that the Starburst+AGN model have the highestevidence for all galaxies in this sample. The Bayesian evidences for the threemodels and the estimated contributions of starburst and AGN to infrared lumi-nosity show that HLIRGs can be classifed into two groups: one dominated bystarburst and the other dominated by AGN. Other parameters and correspondinguncertainties about starburst and AGN are also estimated by using the modelwith the highest Bayesian evidence. We found that the starburst region of theHLIRGs dominated by starburst tends to be more compact and has a higherfraction of OB star than that of HLIRGs dominated by AGN. Meanwhile, theAGN torus of the HLIRGs dominated by AGN tend to be more dusty than thatof HLIRGs dominated by starburst. Overall, we believe that BayeSED could bea reliable and efcient tool for exploring the nature of complex systems such asdust-obscured Starburst-AGN composite systems from decoding their SEDs.