Strategies for Obtaining Robust Spectral Energy Distribution Fitting Parameters for Galaxies at z ∼ 1 and z ∼ 2 in the Absence of Infrared Data
Chandler Osborne, Samir Salim
Abstract
Abstract Robust estimation of star formation rates (SFRs) at higher redshifts ( z ≳ 1) using UV–optical–near-infrared (NIR) photometry is contingent on the ability of spectral energy distribution (SED) fitting to constrain the dust attenuation, stellar metallicity, and star formation history (SFH) simultaneously. IR-derived dust luminosities can help break the degeneracy between these parameters, but IR data are often not available. Here, we explore strategies for SED fitting at z ≳ 1 in the absence of IR data using a sample of log M * > 10.2 star-forming galaxies from the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey (CANDELS) for which 24 μ m data are available. We adopt the total IR luminosity ( L TIR ) obtained from 24 μ m as the “ground truth,” which allows us to assess how well it can be recovered (as L dust ) from UV–optical–NIR SED fitting. We test a variety of dust attenuation models, stellar population synthesis models, metallicity assumptions, and SFHs separately to identify which assumptions maximize the agreement (correlation and linearity) between L TIR and L dust . We find that a flexible dust attenuation law performs best. For stellar populations, we find that Bruzual & Charlot models are favored over those of Eldridge et al. Fixing the stellar metallicity at solar value is preferred to other fixed values or leaving it as a free parameter. For SFHs, we find that minimizing the variability in the recent (<100 Myr) SFH improves the agreement with L TIR . Finally, we provide a catalog of galaxy parameters (including M * and SFR) for CANDELS galaxies with <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mi>log</mml:mi> <mml:msub> <mml:mrow> <mml:mi>M</mml:mi> </mml:mrow> <mml:mrow> <mml:mo>*</mml:mo> </mml:mrow> </mml:msub> <mml:mo>></mml:mo> <mml:mn>8</mml:mn> </mml:math> and 0.7 < z < 1.3, obtained using the models we found to be the most robust.