How accurately can obscured galaxy luminosities be measured using spectral energy distribution fitting of near- through far-infrared observations?

TL;DR

The paper investigates how well infrared luminosities of obscured, infrared-luminous galaxies can be recovered from SED fitting under varying data designs. By simulating observations for a well-characterized local ULIRG sample and fitting with SMART using CYGNUS AGN templates, it quantifies biases and uncertainties in $L_{Tot}$, $L_{Sb}$, $L_{AGN}$, and $L_H$ as functions of wavelength coverage, S/N, flux calibration, angular resolution, and redshift. Key findings show that dense, broad wavelength sampling across the SED peak minimizes biases (with $L_{Sb}$ best shaped by far-IR data and $L_{AGN}$ by near-/mid-IR data), while host luminosities are consistently biased low; angular resolution and PAH-dominated vs AGN-dominated systems further modulate these biases. The results underscore the transformative potential of a far-IR instrument with dense coverage (20–300 μm) for studying infrared-luminous galaxies and provide quantitative bias/risk estimates for JWST and future facilities.

Abstract

Infrared-luminous galaxies are important sites of stellar and black hole mass assembly at most redshifts. Their luminosities are often estimated by fitting spectral energy distribution (SED) models to near- to far-infrared data, but the dependence of these estimates on the data used is not well-understood. Here, using observations simulated from a well-studied local sample, we compare the effects of wavelength coverage, signal-to-noise (S/N), flux calibration, angular resolution, and redshift on the recovery of starburst, AGN, and host luminosities. We show that the most important factors are wavelength coverage that spans the peak in a SED, with dense wavelength sampling. Such observations recover starburst and AGN infrared luminosities with systematic bias below $20\%$. Starburst luminosities are best recovered with far-infrared observations while AGN luminosities are best recovered with near- and mid-infrared observations, though the recovery of both are enhanced with near/mid-infrared, and far-infrared observations, respectively. Host luminosities are best recovered with near/far-infrared observations, but are usually biased low, by $\gtrsim20\%$. The recovery of starburst and AGN luminosity is enhanced by observing at high angular resolution. Starburst-dominated systems show more biased recovery of luminosities than do AGN-dominated systems. As redshift increases, far-infrared observations become more capable, and mid-infrared observations less capable, at recovering luminosities. Our results highlight the transformative power of a far-infrared instrument with dense wavelength coverage from tens to hundreds of microns for studying infrared-luminous galaxies. We tabulate estimates of systematic bias and random error for use with JWST and other observatories.

Figures (16)

• Figure 1: Comparison of telescope and astrophysical backgrounds in a $10\arcsec\times10\arcsec$ aperture, with Mrk 231 at $z=0,1,2,3,4,5$ (§\ref{['sec:obsgen']}). With $B(T)$ as the Planck function, the telescope backgrounds follow $\epsilon_{tel} B(T_{tel})$ with $\epsilon_{tel}=0.04$. Zodiacal light is modelled as $2\times10^{-7} B(T_{zod})$ with $T_{zod}=240\,$K leinert02. Milky Way emission is modelled via $40 ( (\lambda)/(\lambda_{0}) )^{-\beta} 2\times10^{-7} B(T_{ISM})$ (with $T_{ism} = 17.53\,$K, $\lambda_{0} = 23.05\,\mu$m, and $\beta = 1.55$) for wavelengths longward of $70\,\mu$m, and $5\times10^{-23} ((\lambda)/(\lambda_{0}))^{0.4}$ at wavelengths shortward of $70\,\mu$m finkbein99. For the cosmic infrared background, we adopt $1.6\times10^{-5}\left(\frac{\nu}{\nu_{0}}\right)^{(0.64)} B(T_{IRB})$. The CMB is modelled as $B(T_{CMB})$, with $T_{CMB} = 2.718$ K mather94.
• Figure 2: An example of an input SED, simulated data, and recovered fit (§\ref{['sec:obsrec']}). The left panel shows the original observed data and best-fit SED for Mrk 231 efs22. Also shown are the simulated Opt and Fs data (Table \ref{['tbl:instparams']}) from this best-fit "truth" SED. The right panel repeats the "truth" SED from the left panel, along with the SED obtained from fitting to the simulated Opt and Fs data. In this case the recovered starburst SED is well matched to the "truth" starburst SED, but the AGN and host components diverge significantly (§\ref{['sec:resnear']}).
• Figure 3: The fractional probability of recovering an observed-to-true luminosity ratio, $\alpha$ (Equation \ref{['eq:alphadef']}) at $z\sim0.1$, using observations with the Ns, Ms, NMp, Fs, and Fp instruments individually (§\ref{['sec:resnear']}). The top left panel shows the recovery of total infrared luminosity ($L_{Tot}$), the top right panel shows the recovery of starburst luminosity ($L_{Sb}$), the bottom left panel shows the recovery of AGN luminosity ($L_{AGN}$), and the bottom right panel shows the recovery of host galaxy luminosity ($L_{H}$).
• Figure 4: The effect of varying the S/N of the simulated observations on the recovery of total, starburst, AGN, and host luminosity (§\ref{['sec:ressnr']}). The lines have been slightly offset in the $x$-axis for clarity.
• Figure 5: The effect of a flux calibration error on the recovery of total, starburst, AGN, and host luminosity (§\ref{['sec:resfcalib']}). The lines have been slightly offset in the $x$-axis for clarity. Error bars have also been omitted for clarity but are comparable in size to those for the S/N=10 observations in Figure \ref{['fig:localsingle_snr_comp']}.