The mean star formation rates of unobscured QSOs: searching for evidence of suppressed or enhanced star formation

Abstract

We investigate the mean star formation rates (SFRs) in the host galaxies of ∼3000 optically selected QSOs from the SDSS survey within the Herschel-ATLAS fields, and a radio-luminous sub-sample, covering the redshift range of z =0.2-2.5. Using WISE & Herschel photometry (12 -500µm) we construct composite SEDs in bins of redshift and AGN luminosity. We perform SED fitting to measure the mean infrared luminosity due to star formation, removing the contamination from AGN emission. We find that the mean SFRs show a weak positive trend with increasing AGN luminosity. However, we demonstrate that the observed trend could be due to an increase in black hole (BH) mass (and a consequent increase of inferred stellar mass) with increasing AGN luminosity. We compare to a sample of X-ray selected AGN and find that the two populations have consistent mean SFRs when matched in AGN luminosity and redshift. On the basis of the available virial BH masses, and the evolving BH mass to stellar mass relationship, we find that the mean SFRs of our QSO sample are consistent with those of main sequence star-forming galaxies. Similarly the radio-luminous QSOs have mean SFRs that are consistent with both the overall QSO sample and with star-forming galaxies on the main sequence. In conclusion, on average QSOs reside on the main sequence of star-forming galaxies, and the observed positive trend between the mean SFRs and AGN luminosity can be attributed to BH mass and redshift dependencies.

Fundamental differences in the radio properties of red and blue quasars: insight from the LOFAR Two-metre Sky Survey (LoTSS)

“…The distribution based on the IR luminosity function (IRLF), shown in orange, takes its overall shape from the IRLFs in Magnelli et al (2013), with redshift-independent low and high luminosity slopes, and a low-luminosity cutoff (L low ). Both µ IR,SF and L low are monotonically related to the average IR luminosity L IR,SF , shown by the solid vertical line, which is empirically determined for QSOs from the Herschel-based study of Stanley et al (2017). Both distributions shown here share the same L IR,SF , and their particular parameters are representative of cQSOs at z ∼ 1.5.…”

“…For a given L IR , L low is monotonically related to the average IR luminosity of the modified IRLF LIR,SF (see Figure 12 for a visualisation). Stanley et al (2017) have empirically determined LIR,SF for SDSS QSOs using Herschel and WISE stacked photometry. We use their relations to set L low as a function of redshift and the median bolometric luminosities of given QSO subpopulation.…”

“…Our approach to connect the FIR distributions to the observed QSO population is as follows. Through an analysis of composite IR SEDs of SDSS QSOs from Herschel and WISE data, Stanley et al (2017) measured the dependence of the mean IR luminosity attributable to SF as a function of redshift and AGN bolometric luminosity. In addition to a strong evolution with redshift, they reported a weak but definite trend in the average IR luminosities LIR,SF of QSOs with bolometric luminosity, which is governed primarily through a secondary correlation with SMBH mass (Rosario et al 2013;Stanley et al 2017).…”

“…Using L6µm and a 6 µm bolometric correction = 8 (Richards et al 2006), we constrain the median L6µm of the cQSOs (i.e., typical SDSS QSOs) and read off characteristic values for LIR,SF as a function of redshift directly from Figure 8 of Stanley et al (2017). Since we are reliant on their measurements, we perform our own analysis in bins of redshift that parallel those used in that earlier work: 0.5 < z < 0.8, 0.8 < z < 1.0, 1.0 < z < 1.5, 1.5 < z < 2.0, 2.0 < z < 2.5.…”

“…However, the simple application of an additional 0.2-0.3 dex to the nominal values of LIR,SF in these redshift intervals is sufficient to reproduce the observed L144 distributions. Such small differences are within the scope of the systematic uncertainties inherent to the SED fitting approach of Stanley et al (2017). We proceed with our analysis using the parameters of our nominal simulation, but note that such uncertainties on the empirical IR luminosity distributions -their averages and scatter -can quantitatively influence the comparisons of simulated and observed L144 distributions.…”