An Absence of Radio-Loud Active Galactic Nuclei in Geometrically Flat Quiescent Galaxies: Implications for Maintenance-Mode Feedback Models

TL;DR

This study finds that radio-loud AGN are predominantly present in geometrically round, quiescent galaxies, challenging the universality of maintenance-mode feedback models in preventing star formation across different galaxy morphologies.

Contribution

It reveals a strong correlation between galaxy shape and radio-loud AGN presence, questioning the assumption that such AGN feedback universally suppresses star formation in all massive galaxies.

Findings

Radio-loud AGN are more common in round galaxies than in flat, disk-like ones.

The fraction of radio-loud AGN increases with stellar velocity dispersion and mass.

Most disk-like quiescent galaxies lack radio-loud AGN, despite being massive.

Abstract

Maintenance-mode feedback from low-accretion rate AGN, manifesting itself observationally through radio-loudness, is invoked in all cosmological galaxy formation models as a mechanism that prevents excessive star-formation in massive galaxies (M$_*$ $\gtrsim$ 3$\times$10$^{10}$ M$_{\odot}$). We demonstrate that at a fixed mass the incidence of radio-loud AGN (L $>$ 10$^{23}$ WHz$^{- 1}$) identified in the FIRST and NVSS radio surveys among a large sample of quiescent (non-star forming) galaxies selected from the SDSS is much higher in geometrically round galaxies than in geometrically flat, disk-like galaxies. As found previously, the RL AGN fraction increases steeply with stellar velocity dispersion $\sigma_*$ and stellar mass, but even at a fixed velocity dispersion of 200-250 kms$^{-1}$ this fraction increases from 0.3% for flat galaxies (projected axis ratio of q $<$ 0.4) to 5% for round galaxies (q $>$ 0.8). We rule out that this strong trend is due to projection effects in the measured velocity dispersion. The large fraction of radio-loud AGN in massive, round galaxies is consistent with the hypothesis that such AGN deposit energy into their hot gaseous halos, preventing cooling and star-formation. However, the absence of such AGN in disk-like quiescent galaxies -- most of which are not satellites in massive clusters, raises important questions: is maintenance-mode feedback a generally valid explanation for quiescence; and, if so, how does that feedback avoid manifesting at least occasionally as a radio-loud galaxy?

Figures

Full text

††thanks: 1Max-Planck Institut für Astronomie, Königstuhl 17, D-69117, Heidelberg, Germany††thanks: 2Sterrenkundig Observatorium, Department of Physics and Astronomy, Ghent University, Belgium††thanks: 3Leiden Observatory, Leiden University, P.O.Box 9513, NL-2300 AA Leiden, The Netherlands††thanks: 5University of Pittsburgh, Department of Physics and Astronomy, 100 Allen Hall, 3941 OHara St, Pittsburgh PA 15260, USA††thanks: 4Department of Astronomy, University of Michigan, 1085 S. University Ave, Ann Arbor, MI 48109, USA††thanks: 6Academia Sinica Institute of Astronomy and Astrophysics, No.1 Section 4 Roosevelt Rd., 11F of Astro-Math Building, Taipei 10617, Taiwan††thanks: 7Leiden Observatory, Leiden University, P.O. Box 9513, NL-2300 RA Leiden, The Netherlands††thanks: 8European Southern Observatory, Karl-Schwarzschild-Straße 2, 85748 Garching bei Munchen, Germany

1 Introduction

To explain the mismatch between the theoretical dark matter halo mass function and the observed galaxy stellar mass function, semi-analytical and hydrodynamical simulations of galaxy formation and evolution have implemented a set of feedback mechanisms to reduce the efficiency of star-formation; they rely on supernova feedback in low halo mass systems, and super-massive black-hole (SMBH) feedback in systems with high mass halos (Croton et al., 2006; Bower et al., 2006; Fabian, 2012; Vogelsberger et al., 2014; Pillepich et al., 2017). The most common picture states that the SMBH feedback generally manifests through radio-loudness in active galactic nuclei (AGN) in the form of the jets (“radio-mode”) or gas outflows (“quasar-mode”) as a consequence of a low- or high-accretion rate (respectively) of the gas onto the SMBH. This feedback picture states that powerful jets transfer their energy to the surrounding gas, heating it up and therefore preventing it from cooling and forming new stars. “Radio-mode” feedback has thus been introduced as the leading mechanism in explaining the mass growth deficiency in the most massive early type galaxies with SMBH.

Direct observational evidence for this picture in local galaxies comes from observations of cavities in the X-ray emitting gas caused by jets, connecting the presence of radio-loud (RL) AGN with the absence of star-formation (McNamara & Nulsen, 2007; Heckman & Best, 2014). Indirect evidence in the local universe has also provided additional support to this radio-mode feedback picture (Matthews et al., 1964; Kauffmann et al., 2003), including studies that show a strong, increasing trend in the RL fraction with M∗ and stellar velocity dispersion $\sigma_{*}$ (Best et al., 2005). According to the van den Bosch (2016) BH mass $-$ $\sigma_{*}$ relation (see also Gebhardt et al., 2003; Beifiori et al., 2012), this implies an increasing trend of the RL fraction with the BH mass, suggesting the importance of the BH mass as a parameter in setting the probability for a galaxy to become RL (see also Häring & Rix, 2004; Terrazas et al., 2017).

In this work we explore how the incidence rate of RL AGN varies with geometric shape (flat or disk-like vs. round). We test this by examining the correlation of the RL fraction with $\sigma_{*}$ and the projected axis ratio for a sample of galaxies drawn from the Sloan Digital Sky Survey (SDSS; York et al., 2000), NRAO VLA Sky Survey (NVSS; Condon et al., 1998) and Faint Images of the Radio Sky at Twenty centimeter (FIRST; Becker et al., 1995) survey, and we present evidence that round galaxies host RL AGN much more frequently than disk-like galaxies at any fixed $\sigma_{*}$ and M∗. According to the ’radio-mode’ feedback picture, and the association of the presence of RL AGN with quiescence, this raises a question to why are these galaxies not forming new stars in the absence of a visible heating source.

In Section 2 we give an overview of the data-set we use in this study, and in Section 3 we present or results and discuss the implications. In section 4 we give a brief summary of our main results. We use standard cosmology: $H_{0}$ = 70(km/s)/Mpc, $\Omega_{M}$ = 0.3, $\Omega_{\Lambda}$ = 0.7 .

2 Data and sample selection

2.1 SDSS

The SDSS survey (York et al., 2000) provides unique and high quality information on various physical properties of a large number of present-day galaxies. Physical properties of those galaxies are measured and derived using photometry and spectroscopy, and include $\sigma_{*}$111http://classic.sdss.org/dr7/algorithms/veldisp.html, star-formation rate (SFR), M∗, redshift and the projected axis ratio. The resolution of the spectroscopic data-set is $R$ = 1850 $-$ 2200, covering the wavelength range between 3800 Å to 9200 Å, with a minimum signal-to-noise S/N $\sim$ 14 Å*-1*.

In this study we make use of the SFRs and M∗ obtained from modelling galaxy spectral energy distributions (SED; Chang et al., 2015) based on SDSS Data Release 7 (DR7; Abazajian et al., 2009) and WISE (Cutri et al., 2013).

In order to create a volume-selected sample (303,785222 We removed 11058 galaxies (3.5%) that lack rest-frame u-, r-, and z-band photometry and a measurement of the axis ratio.) complete in M∗ we apply a redshift-dependent mass-cut: log(M∗) $>$ 10.6 + 2.28$\times$log($z$/0.1)) (Chang et al., 2015), consequently spanning a redshift range from 0.02 $<$ z $<$ 0.14. For sample classification between star-forming and quiescent galaxies, we apply a color-based criteria following Chang et al. (2015) ($u$ - $r$ $>$ 1.6 $\times$ ($r$ - $z$) + 1.1) for quiescent galaxies. Star-forming galaxies are considered to be all other galaxies that fall outside of this color-based criteria.

2.2 NVSS and FIRST

The NVSS (Condon et al., 1998) 1.4 GHz sky survey north of $\delta=-40^{o}$ has been conducted in the period between 1993 and 1996. The angular resolution of the observed data-set is 45$\arcsec$, with the noise level rms of 0.45 mJy beam*-1* and the point source flux density 50% completeness level at 2.5 mJy. For more details see Condon et al. (1998). The FIRST (Becker et al., 1995) 1.4 GHz survey of the northern Galactic cap has been conducted in the period between 1993 and 2004. The average rms of the images is 0.15 mJy, with an angular resolution of 5$\arcsec$ and the flux density limit of 1 mJy. Both surveys have been conducted using the Very Large Array (VLA) radio interferometer.

Taking advantage of the high resolution FIRST data-set, Best & Heckman (2012) cross-match sources with galaxies in the SDSS DR7. Flux density measurements are subsequently performed on the NVSS data-set. The resulting cross-matched catalog, with a limiting NVSS flux density value of 5 mJy, is complete for radio-loud AGN down to L1.4GHz $\sim$ 1023 WHz*-1* at $z=0.1$. Since the sample is incomplete below L1.4GHz $<$ 2.5 $\times$ 1023 WHz*-1* at highest redshifts, we apply a volume-correction that reaches at most a factor 4 for L1.4GHz $\sim$ 1023 WHz*-1* at z = 0.14.

2.3 Selection of Radio-Loud AGN Hosts

To select a radio-loud sub-sample we consider galaxies with detected radio emission at 1.4 GHz by the NVSS and FIRST. From the total number of galaxies in the selected galaxy sample, the number of those detected with F1.4GHz $>$ 5 mJy is 4,571, and they are shown with red and blue symbols in Figure 1.

We use Condon (1992) luminosity relation ($\alpha$ = - 0.8), to convert the NVSS measured flux density values F1.4GHz into luminosities. Derived luminosities at 1.4 GHz are then converted to radio-based SFRs using the Bell (2003) radio$-$far-infrared calibration.

We follow Best et al. (2005) 1.4 GHz luminosity limit L1.4GHz $>$ 1023 WHz*-1* to select radio-loud galaxies. We subtract the contribution of star-formation as follows: L1.4GHz - (3 $\times$ L1.4GHz,SFR) $>$ 1023 , where L1.4GHz is the luminosity based on the 1.4 GHz flux density value, and L1.4GHz,SFR is the expected L1.4GHz luminosity based on the SFR (taken from Chang et al. (2015)). In order to select RL AGN with high confidence we multiply L1.4GHz,SF by a factor 3 to account for uncertainties in radio- and mid-IR based SFR estimates.

We keep the color-based classification from Chang et al. (2015) to separate the radio-detected sample between quiescent (red) and star-forming (blue symbols) galaxies. The selected radio-loud AGN sub-sample is indicated with the purple dashed line in Figure 1.

3 Which Parameters Correlate the Radio-Loud AGN Fraction?

In the following subsections we explore the RL fraction trend with the $\sigma_{*}$ (or equivalently BH mass; van den Bosch, 2016) for all, quiescent and star-forming galaxies. We also explore the aspect of radio-loudness having a preference to occur more frequently among galaxies of a specific shape.

3.1 Radio-Loud AGN Fraction Increases with $\sigma_{*}$

It has been shown in the local universe that the RL fraction of massive elliptical galaxies exhibits a strong increasing trend with $\sigma_{*}$ (Best et al., 2005). According to the BH mass$-\sigma_{*}$ relation (van den Bosch, 2016) this implies a strong dependence of the RL fraction on the BH mass. In Figure 2 (left panel) we show the fraction of quiescent galaxies in the selected sample (violet crosses), and the RL fraction among all (yellow circles), quiescent (red empty circles) and star-forming (blue empty circles) galaxies as a function of $\sigma_{*}$ and the corresponding BH mass (van den Bosch, 2016). It is evident that the RL fraction exhibits an increasing trend with $\sigma_{*}$, which reproduces the result from Best et al. (2005).

Evidently, the RL fraction is increasing among star-forming galaxies as well, and these galaxies appear to host RL AGN as frequently as quiescent galaxies at fixed $\sigma_{*}$ (i.e. Janssen et al., 2012). Since the star-forming galaxies host RL AGN at a comparable rate as the quiescent galaxies, then the radio-loudness is not a sufficient condition for a galaxy to become quiescent.

3.2 Geometry Plays a Role in Radio-Loud Fraction

For the remainder of this letter we only consider quiescent galaxies, as we are interested in constraining maintenance-mode feedback. We divide the sample into six evenly spaced bins of $\sigma_{*}$ and four in the projected axis ratio. In each bin we determine the fraction of RL galaxies. The right panel of Figure 2 presents the heat-map of the projected axis ratio distribution as a function of $\sigma_{*}$ for quiescent galaxies, color-coded by the fraction of RL galaxies. The top number inside of each bin corresponds to the number of RL galaxies, while the bottom number corresponds to the total number of galaxies. We observe that RL fraction among quiescent galaxies exhibits an increasing trend with both parameters – $\sigma_{*}$ and the projected axis ratio.

This result illustrates that the important parameters in determining the probability for a galaxy to “switch on” as a RL AGN is not only its $\sigma_{*}$ but also the geometry of the galaxy.

We note that the $\sigma_{*}$ measurement for all but face-on galaxies is to a certain degree affected by contribution of the rotational velocity. Due to this effect, the actual $\sigma_{*}$ should be lower for inclined galaxies, further suggesting lower RL fractions. Assuming that the M∗ roughly corresponds to the virial mass, then the ratio between the galaxy M∗ and its inclination-independent semi-major axis (half-light) radius R, can serve as a proxy for the intrinsic $\sigma_{*}$ free of the rotational velocity effects, and independent on the inclination. Figure 3 shows the RL fraction trend with the projected axis ratio for quiescent galaxies, in narrow bins of the $\sigma_{*}$ (left hand), M∗ (middle) and the $\sigma_{*}$ proxy log(M∗/R) (right panel). These trends are representative of the sample as a whole. The increasing trend in all three panels is likely dominated by a dependence of the RL fraction on the intrinsic shape of the galaxy, rather then its inclination. We note that replacing the semi-major axis radius with the inclination-independent circularized radius in estimating the $\sigma_{*}$ proxy log(M∗/R) results in an even stronger trend than seen in the right panel of Figure 3. Therefore, correcting $\sigma_{*}$ measurement of inclined galaxies for the rotational velocity effects would not significantly alter the increasing trend we see, which implies the preference for RL AGN to occur in round galaxies.

3.3 Radio-Loud AGN Fraction of Intrinsically Round and Flat Galaxies

Chang et al. (2013) developed a two-population model to reproduce the observed axis ratio distribution of quiescent galaxies up to $z=2.5$, combining round (triaxial) and flat (oblate) populations of galaxies. They perform their modelling in three different mass bins, and to simplify the analysis we only use the fit parameters of the highest stellar mass bin (M∗ $>$ 10.8) 333This mass limit corresponds to a $\sigma_{*}$ limit of $\sim 120\rm{km~{}s}^{-1}$, but even below this limit the shape parameters are not very different and our analysis would not be affected. where the intrinsic axis ratio of the oblate component is b444b = intermediate axis of the triaxial ellipsoid = 0.29$\pm$0.07. The best-fit triaxility (T555T = [1 - $\beta^{2}$]/[1 - $\gamma^{2}$] = 0.64$\pm$0.06) and ellipticity (E666E = 1 - $\gamma$ = 0.41$\pm$0.02) parameter values, and their standard deviations ($\sigma_{T}$, $\sigma_{E}$ = 0.08$\pm$0.05, 0.19$\pm$0.02) are used to generate distributions of intrinsic intermediate-to-long axis ratio $\beta$ and short-to-long axis ratio $\gamma$. Combining these distributions with random viewing angles provide us with model triaxial and oblate intrinsic axis ratio distributions.

We split the sample of quiescent galaxies into six equally spaced $\sigma_{*}$ bins, and in each bin we find the best-fit linear combination of model triaxial and oblate shape distributions which reproduce the observed axis ratio distribution. This gives us the triaxial and oblate fraction of galaxies in the specific $\sigma_{*}$ bin. As an example, in Figure 4 we show a normalized histogram (gray bins) of the observed axis ratio (q) distribution, and a corresponding function describing the distribution obtained after binning, for quiescent (left panel) and for RL quiescent (right panel) galaxies in the range of $\sigma_{*}$ values from 150 - 200 km s*-1*. For illustration, we choose the $\sigma_{*}$ bin in which the total population is similarly populated by both triaxial and oblate galaxies according to the two-population model. Blue and red dashed curves represent best-fit oblate and triaxial fractions, respectively. We see that the triaxial fraction (100 $\pm$ ${}^{0}_{7}$%) of RL AGN in the observed quiescent population in this $\sigma_{*}$ bin completely dominates over the oblate fraction, implying a more frequent occurrence of RL AGN among round galaxies.

The best-fit model triaxial and oblate fraction for the observed axis ratio distribution of quiescent galaxies can be rescaled to fit the observed axis ratio distribution of RL quiescent galaxies. Multiplying these rescaling factors by the RL fraction gives the RL fraction among triaxial and oblate galaxies.

Figure 5 shows the RL fraction among triaxial (red filled squares) and oblate (red open squares) galaxies for quiescent population as a function of the $\sigma_{*}$. The overall triaxial fraction (black crosses) increases with the $\sigma_{*}$. The increase in RL fraction with $\sigma_{*}$ is entirely due to its increase among the triaxial population. In fact, the inferred RL fraction among oblate population is essentially zero. This analysis, by accounting for round (face-on) disk-like galaxies, further strengthens our result that RL AGN preferentially reside in intrinsically round galaxies. We note that we do not claim that the RL fraction among disk-like quiescent galaxies is, in fact, zero; the triaxial population is defined to consist of objects with a wide range in intrinsic $\beta$ and $\gamma$ values (see above), including a subset with $\beta$ $\sim$1 and small $\gamma$, that is, disk-like geometries.

3.4 Satellites, Centrals and Halo Mass

We now investigate whether the correlation between the shape and RL fraction is driven by an underlying connection between dark matter halo properties and RL AGN. This may be expected in case disk-like galaxies live in less massive halos than round galaxies, or if satellite galaxies, which are more often disk-like than centrals (van der Wel et al., 2010), are less likely to be RL AGN.

We cross-match our sample with the SDSS DR7 based ”petroC”777 http://gax.sjtu.edu.cn/data/Group.html group catalog created by Yang et al. (2007). For this halo catalog, centrals are chosen based on stellar mass, and halo masses are assigned to groups based on their total stellar mass. We focus on quiescent galaxies with $200<\sigma_{*}/(\rm{km}~{}\rm{s^{-1}})<230$, which shows a strong correlation between the shape and RL AGN incidence, and for which the satellite fraction is non-neglible ($\sim 20$%, van den Bosch et al., 2008). We find that RL fractions are similar: 2.3% for the satellites, and 2.8% for the centrals, and their trends with galaxy shape are consistent with each other. There is no substantial difference ($<$0.15 dex) between the halo masses of round central galaxies and flat central galaxies, and selecting galaxies by halo mass still contains a strong correlation between RL AGN fraction and galaxy shape.

We conclude that the strong dependence of RL AGN incidence on galaxy shape is intrinsic and cannot be explained by differences between RL AGN fractions in satellite and central galaxies, nor to differences in halo mass.

4 Discussion

In this paper we demonstrate that RL AGN occur more frequently in round than in flat quiescent galaxies at any M∗ or $\sigma_{*}$. The analysis of the projected axis ratio distribution of RL AGN hosted by quiescent galaxies shows that the increased RL fraction at higher $\sigma_{*}$ values is entirely driven by intrinsically round (triaxial) galaxies. The RL fraction reaches $\sim$20% for the most massive, round quiescent galaxies ($\sigma_{*}$ $\sim$300 km s*-1*), whereas geometrically flat (disk-like) quiescent galaxies have an extremely low RL AGN fraction ($<$ $\sim$10*-3*), even at high $\sigma_{*}$ ($>$200 km s*-1*).

This result is consistent with the common notion that luminous radio AGN reside in massive, elliptical galaxies. However, thus far, the correlation between the RL AGN occurrence in quiescent galaxies and a second global property, besides its $\sigma_{*}$ (BH mass), has not been explicitly demonstrated. The roundness of a galaxy likely reflects its merger history, and the connection between an active merger history and the occurrence of low-accretion rate radio AGN has been made by many authors (De Ruiter et al., 2005; Capetti & Balmaverde, 2006; Sikora et al., 2007; Fanidakis et al., 2010). The physical basis for this connection has been referred to as the “spin paradigm” (Blandford & Znajek, 1977; Wilson & Colbert, 1995; Hughes & Blandford, 2003; McKinney & Gammie, 2004) in which only BHs with large spin parameters can launch radio jets. In turn, a large spin parameter is the result of BH-BH mergers that reflect the merger history of the host galaxy.

Regardless of the physical explanation, the lack of RL AGN in disk-like quiescent galaxies raises important questions. In galaxy formation and evolution models, low-accretion rate AGN are universally assumed to play a crucial role in preventing excessive SF in massive galaxies, independent of their structure. Our findings suggest that, if this model is generally valid, the AGN feedback does not manifest itself at radio wavelengths for galaxies with disk-like structures. We conclude that there is compelling evidence for maintenance-mode feedback for intrinsically round galaxies, but that further study is needed to bolster the evidence for such feedback as a general explanation for the lack of star-formation activity in massive galaxies.

5 Acknowledgements

We thank the anonymous referee for providing useful feedback. Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England888http://www.sdss.org/. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement 683184). GvdV acknowledges funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under grant agreement No 724857 (Consolidator Grant ArcheoDyn)