Bayesian surface photometry analysis for early-type galaxies

Abstract

We explore the application of Bayesian image analysis to infer the properties of an SDSS early-type galaxy sample including AGN. We use GALPHAT (Yoon et al. 2010) with a Bayes-factor model comparison to photometrically infer an AGN population and verify this using spectroscopic signatures. Our combined posterior sample for the SDSS sample reveals distinct low and high concentration modes after the point-source flux is modeled. This suggests that ETG parameters are intrinsically bimodal. The bimodal signature was weak when analyzed by GALFIT (Peng et al. 2002, 2010). This led us to create several ensembles of synthetic images to investigate the bias of inferred structural parameters and compare with GALFIT. GALPHAT inferences are less biased, especially for high-concentration profiles: GALPHAT S\'ersic index $n$, $r_{e}$ and MAG deviate from the true values by $6\%$, $7.6\%$ and $-0.03 \,\mathrm{mag}$, respectively, while GALFIT deviates by $15\%$, $22\%$ and $-0.09$\, mag, respectively. In addition, we explore the reliability for the photometric detection of AGN using Bayes factors. For our SDSS sample with $r_{e}\ge 7.92\,$arcsec, we correctly identify central point sources with $\mathrm{Mag_{PS}}-\mathrm{Mag_{Sersic}}\le 5$ for $n\le6$ and $\mathrm{Mag_{PS}}-\mathrm{Mag_{Sersic}}\le 3$ for $n>6$. The magnitude range increases and classification error decreases with increasing resolution, suggesting that this approach will excel for upcoming high-resolution surveys. Future work will extend this to models that test hypotheses of galaxy evolution through the cosmic time.