The Role of Active Galactic Nuclei in the Quenching of Massive Galaxies in the SQuIGG$\vec{L}$E Survey

Ever since the realization that supermassive black holes (BHs) are ubiquitous at the centers of massive galaxies, we have invoked energy input from active galactic nuclei (AGN) as a mechanism for heating or removing the interstellar medium of galaxies and suppressing star formation (e.g., Silk & Rees 1998; Springel et al. 2005). While there is strong evidence that radio emission from central galaxies keeps the intracluster medium from cooling in the cluster environment (e.g., Fabian 2012), we do not know whether AGN play an active role in shutting off star formation in massive elliptical galaxies (e.g., Heckman & Best 2014). It is hard to define an experiment that can directly implicate the AGN, given that star formation necessarily lasts at least an order of magnitude longer in time than the typical mass doubling time of a BH (the Salpeter time of 3 × 10⁷ yr).

Large AGN surveys (predominantly identified using X-rays) have shown clearly that accretion depends on both the mass of the galaxy (and thus presumably the BH) and the star formation rate (e.g., Silverman et al. 2009; Aird et al. 2012, 2018, 2019; Chen et al. 2013; Yang et al. 2017, 2018). In cases where detailed spectroscopic information is available, there is some evidence that AGN tend to occur a few hundred million years after a burst of star formation (e.g., Kauffmann et al. 2003; Davies et al. 2007; Wild et al. 2010), although optical spectroscopy may be biased against finding accretion activity in the most luminous starbursts (e.g., Trump et al. 2015; Jones et al. 2016).

In this work, we identify galaxies for which a burst of star formation has ended recently, so-called post-starburst (PSB) galaxies. We are interested in massive galaxies whose primary epoch of star formation has recently ended. Since massive galaxies ended their star formation early (e.g., Thomas et al. 2005), observations at significant lookback times are required to find massive PSBs. Such galaxies are identifiable, although quite rare, at the modest redshift of $0.6\lt z\lt 1$ (e.g., Tremonti et al. 2007; Pattarakijwanich et al. 2016; Wild et al. 2016). In this work, we ask whether a sample of massive and recently quenched PSBs show any sign of enhanced nuclear activity compared with quiescent galaxies of similar mass. We adopt a standard concordance cosmology with ${H}_{0}=70$ km s⁻¹ Mpc⁻¹, ${{\rm{\Omega }}}_{m}=0.3$, ${{\rm{\Omega }}}_{{\rm{\Lambda }}}=0.7$.

In order to understand the detailed processes driving galaxy quenching, we have launched the Studying of QUenching in Intermediate-z Galaxies: Gas, angu$\vec{L}$ar momentum, and Evolution (SQuIGG$\vec{L}$E) Survey. In K. A. Suess et al. (2020, in preparation), we present a large new sample of 1318 PSB galaxies that are selected from the Sloan Digital Sky Survey Data Release 14 (SDSS DR14; Abolfathi et al. 2018). The sample ranges in redshift from z = 0.5–0.94 (median z = 0.68). The galaxies are relatively bright and massive, with i = 17.9–20.5 mag (median i = 19.5 mag), with nearly all having $10.5\lt \mathrm{log}{M}_{* }/{M}_{\odot }\lt 11.5$ (median M_* = 10¹¹ ${M}_{\odot }$).

The sample was selected from the SDSS DR14 spectroscopic catalog following Kriek et al. (2010). In brief, we calculate the flux in synthetic rest-frame U-, B-, and V-band filters for all SDSS spectra with z > 0.5. We require that the spectra have a signal-to-noise ratio (S/N) ≥ 6 in both the U and B filters and then use the color cuts U−B > 0.975 and −0.25 < B−V < 0.45 to isolate galaxies with the unique A-star-like spectral shape indicative of a recently quenched stellar population.

To demonstrate the success of our selection, we look at two key stellar absorption indices (Figure 1). The equivalent width of Hδ_A combined with the ${D}_{n}4000$ index (the ratio in flux at 3850–3950 Å with that at 4000–4100 Å) are a powerful nonparametric way to track stellar age (e.g., Kauffmann et al. 2003). We show two tracks in Figure 1 based on two-component stellar population models with an old population plus a young burst with varying burst fraction. SQuIGG$\vec{L}$E galaxies had bursts of star formation that ended 100–500 Myr in the past. The actual quenching time is degenerate with the burst fraction, but the light-weighted ages tend to be ∼500 Myr for a D_n4000 = 1.25 (D. Setton et al. 2020, in preparation).

Zoom In Zoom Out Reset image size

Within SDSS, objects are targeted spectroscopically for many reasons, including radio emission. To avoid any preselection based on active galaxy attributes, we restrict attention to the 1207 galaxies that fall within the "CMASS" SDSS sample. This sample is targeted based solely on griz color and magnitude (Dawson et al. 2013). The inverse variance–weighted mean spectrum of the full sample, normalized at 4020 Å, is shown in Figure 2.

Zoom In Zoom Out Reset image size

SQuIGG$\vec{L}$E aims to study the gas fractions, angular momenta, stellar populations, and merger fractions of this PSB sample. With the Atacama Large Millimeter/submillimeter Array (ALMA), we have searched for CO 2–1 in 13 galaxies (Suess et al. 2017), and found that roughly half contain surprisingly large gas reservoirs, with gas fractions of ∼10%–20% (R. Bezanson et al. 2020, in preparation).

2.1. LEGA-C Comparisons

We select AGN using the ratio between [O iii]λ5007 and Hβ, requiring reliable measurements of the emission-line strengths of these two key lines. Due to the strong stellar absorption in Hβ, we jointly model the stellar continuum and emission lines ([Oii]$\,\lambda \lambda 3726,3729$, Hβ, and [O iii]$\,\lambda \lambda 4959,5007$) using the publicly available penalized pixel-fitting code pPXF (Cappellari & Emsellem 2004). Our stellar templates are the single stellar population models from Vazdekis et al. (2010) built from the MILES stellar library (Sánchez-Blázquez et al. 2006), and we include an order 12 additive polynomial to account for flux calibration uncertainties. We model the line-of-sight stellar velocity distribution as a Gaussian and fix the velocities and line widths of all emission lines to the same value.

We generate mock spectra to measure uncertainties on our flux measurements. We start with the best-fit pPXF model fit to the stars and gas, and for each of 100 artificial runs we draw the noise spectrum randomly from the variance array, and refit. The standard deviation in flux values from these runs is our adopted uncertainty, which is 20% higher than the statistical errors returned by pPXF. We select all galaxies with [O iii] line equivalent width (EW) > 5, which corresponds to a minimum S/N = 4 in the line, although all but 10 of the galaxies have S/N > 10. Hβ is most challenging to detect, given the strong absorption. In the case of nondetections of Hβ emission, we assign the 3σ upper limit as the flux for the purpose of calculating line ratios.

We select two complementary AGN samples. First, we look for high-ionization optical emission-line ratios indicative of a hard ionizing spectrum. Then, we use the Faint Images of the Radio Sky at Twenty-cm (FIRST; Becker et al. 1995) survey to find AGN through radio emission. We adopt the FIRST survey because it overlaps so well with the SDSS footprint, and it remains the most uniform survey available over this area with a beam size roughly matched to that of galaxies.

4.1. Optical Emission-line Selection

We search for high-ionization emission lines in the SDSS PSB spectra. Classic narrow-line region selection relies on two-dimensional emission-line ratio plots known as BPT diagrams (Baldwin et al. 1981). Unfortunately, at the redshifts of our sample, the SDSS spectra do not include Hα or [N ii] λλ6548, 6584, which play a central role in standard BPT diagrams. Instead, we are forced to rely on [O iii]/Hβ alone.

At first glance, this single line ratio does not appear sufficient to classify a galaxy as an AGN. In standard BPT diagrams, there are many galaxies without nuclear activity but with high values of [O iii]/Hβ. However, in the absence of AGN activity, high [O iii]/Hβ is predominantly seen in low-mass, low-metallicity galaxies (e.g., Moustakas et al. 2006). At higher masses, star-formation-powered emission lines have lower [O iii]/Hβ ratios. This realization was formalized by Juneau et al. (2011) in what they call the Mass-Excitation diagram, which plots [O iii]/Hβ against stellar mass, and shows that local galaxies with ${M}_{* }\gt {10}^{10.8}$ ${M}_{\odot }$ and [O iii]/Hβ > 3 are all AGN. While strong line ratios are known to shift within the BPT diagram at higher redshift (e.g., Shapley et al. 2005), in our redshift range of interest $z\sim 0.7$, galaxies are observed to have interstellar medium (ISM) emission lines similar to local galaxies (e.g., Kewley et al. 2013). Furthermore, even at $z\sim 2$, higher-mass galaxies like those studied here do not appear significantly offset (e.g., Shapley et al. 2015).

We apply the simple joint criteria that (a) the [O iii] line EW > 5 Å and (b) [O iii]/Hβ > 3 (e.g., Ho 2008). This ensures adequate signal to noise in the line and corresponds to a flux limit of 10⁻¹⁶ erg s⁻¹ cm⁻². All but four of the galaxies with [O iii] above our EW cut qualify as AGN based on line ratios. These two cuts identify a total of 64 AGN, out of 1207 galaxies, for an AGN fraction of 5 ± 0.7% (see Table 1). As additional confirmation that we are selecting bona fide AGN, we note that the [O iii]/Hβ distribution is peaked at a ratio of 10, and that our qualitative results would not change if we selected the 50 objects with [O iii]/Hβ > 5.

Table 1.  AGN Fractions

Sample	Subset	NTot	NAGN	fAGN
Optical	All	1207	64	0.053 ± 0.007
Optical	Young	603	53	0.088 ± 0.012
Optical	Old	604	11	0.018 ± 0.005
Optical	Low Mass	602	26	0.043 ± 0.008
Optical	High Mass	605	38	0.063 ± 0.010
LEGA-C	Optical	803	10	0.012 ± 0.004
Radio	All	1207	51	0.042 ± 0.006
Radio	Young	603	26	0.043 ± 0.008
Radio	Old	603	38	0.063 ± 0.010
Radio	Low Mass	602	14	0.023 ± 0.006
Radio	High Mass	605	37	0.061 ± 0.010
LEGA-C	Radio	675	10	0.015 ± 0.005

Note. AGN fraction over galaxy parameters. We divide old versus young at D_n4000 = 1.25 and high versus low mass at M_* = 10¹¹ ${M}_{\odot }$.

Download table as:  ASCIITypeset image

In Figure 3, we plot the [O iii] luminosity distribution of the detected AGN. Our objects span only one order of magnitude in [O iii] luminosity. Sources with fluxes <10⁻¹⁶ erg s⁻¹ cm⁻² or ${L}_{[{\rm{O}}{\rm\small{III}}]}\lt {10}^{41}$ erg s⁻¹ at z = 0.7 are not included in our sample. We are likely less complete in AGN at lower stellar masses, given the correlation between L_{[O iii]} and M_*. We quantify this correlation and investigate the impact of this incompleteness in Section 5.1. Furthermore, the most luminous targets may be too blue to either be included in the CMASS sample or the PSB color selection, due to scattered light from the obscured quasar (e.g., Zakamska et al. 2006; Liu et al. 2009), but such objects are very rare (Reyes et al. 2008). For physical context, BHs of ${M}_{\mathrm{BH}}\approx {10}^{8}$ ${M}_{\odot }$ (for typical stellar masses of $\sim {10}^{11}$ ${M}_{\odot }$; Häring & Rix 2004) have a range in Eddington ratio of 1%–10%, assuming a bolometric correction of 10³ (Liu et al. 2009) for [O iii].

Zoom In Zoom Out Reset image size

Through stacking, we compare the average spectra of the [O iii]-selected AGN with the full inactive sample, as well as stacks of the younger and older galaxies divided at the median D_n4000 = 1.25. We can see that the AGN have a bluer stellar continuum than the full stack, matching that of the younger inactive stack (Section 5.1), but the line emission is much stronger in the AGN stack than in the young inactive stack. We also detect [Ne v]λ3426 Å, a line that is very hard to excite without the hard ionizing spectrum of an AGN (e.g., Goulding & Alexander 2009). In Figure 2(b), we show that the [Ne v] line is clearly detected in the stacked AGN spectrum, providing some confidence that a significant fraction of our optically selected AGN are powered by accreting BHs.

4.1.1. LEGA-C Comparison Sample

We select as LEGA-C AGN those galaxies with [O iii]/Hβ > 3 and an [O iii] flux above the SQuIGG$\vec{L}$E [O iii] flux limit of 10⁻¹⁶ erg s⁻¹. The resulting five galaxies are shown as colored squares in Figure 3 corresponding to an AGN fraction of 1.4 ± 0.3%. We also relax the [O iii] luminosity cut and show all galaxies with AGN-like line ratios in LEGA-C (gray squares). Unlike the SQuIGG$\vec{L}$E sample, LEGA-C does not contain many massive PSBs, and it has very few AGN above our flux limit. Both are almost certainly due to the small volume.

4.2. Radio Selection

To find the radio-detected AGN within the SQuIGG$\vec{L}$E sample, we search the FIRST archive for cross-matches within 5'' of the position of each SQuIGG$\vec{L}$E galaxy. In practice, all but two are matched within 3'' of the SDSS position. One of those is an extended jet source, leaving one possible mismatched source. We do not inspect the images for possible radio detections at larger radius (as in, for instance, Best et al. 2005). For ease of comparison with other radio AGN samples, we calculate our radio detection limit at 3 GHz rather than 1.4 GHz, assuming a standard radio spectral slope of α = −0.7. FIRST has a uniform flux limit of 1 mJy, which corresponds to a luminosity limit range L_3GHz = (0.6–2) × 10²⁴ W Hz⁻¹, and L_3GHz = 10²⁴ W Hz⁻¹ for a median distance of z = 0.7.

This selection yields 52 radio AGN, or a radio AGN fraction of 4 ± 0.6%. The distribution of radio luminosities are plotted in Figure 4. The star formation rate needed to explain the observed radio emission is many hundreds of solar masses per year (e.g., Bell 2003). Given that we observe upper limits on the star formation rates of tens of solar masses per year from [Oii] fluxes (Suess et al. 2017), we view that the radio emission we observe arises from star formation as highly unlikely.

Zoom In Zoom Out Reset image size

In contrast to the [O iii]-selected sample, we observe a very wide range of radio luminosities, spanning nearly three decades (Figure 4). Radio luminosity does not have a simple or well-understood relationship with the accretion rate (see references in Heckman & Best 2014), although radio emission contributes a larger fraction of the bolometric luminosity at lower AGN accretion rates (e.g., Ho 2008, and references therein). We detect [O iii] emission from only five of our radio galaxies. The majority appear to be "low-excitation" (Hine & Longair 1979) sources that, likely due to very low Eddington ratios, do not show radiative signatures of standard optically thick but geometrically thin accretion disks (e.g., Shakura & Sunyaev 1973; Best & Heckman 2012). Conversely, among the [O iii] sources, we detect few radio AGN due to our flux limit; [O iii]-selected AGN in SDSS with ${L}_{[{\rm{O}}{\rm\small{III}}]}\approx {10}^{41}$ erg s⁻¹ have radio luminosities of ${L}_{r}\approx {10}^{22.5}-{10}^{23}\,$ W Hz⁻¹ as probed by stacks (e.g., de Vries et al. 2007).

4.2.1. LEGA-C Comparison

Barišić et al. (2017) built a catalog of radio AGN in LEGA-C; we simply take the subset of these that obey our flux limit of ${f}_{1.4\mathrm{GHz}}=1$ mJy and fall within our stellar mass range. We find a radio AGN fraction of 1.5 ± 0.4%. In the case of radio AGN, the fraction in SQuIGG$\vec{L}$E is consistent within 2σ of the radio-loud fraction in a mass-matched LEGA-C sample.

Having identified a sample of recently quenched galaxies that host AGN, we now examine trends between the AGN and host galaxy properties. As shown in Figure 1, D_n4000 tracks stellar age; the median D_n4000= 1.25 corresponds to a light-weighted age of ∼500 Myr. In the following, binned quantities are constructed to have equal numbers of AGN per bin, and error bars are Poisson.

5.1. [O iii]-selected AGN

5.2. Radio-selected Sample

K. A. Suess et al. (2020, in preparation) have assembled a large sample of massive and rare PSB galaxies at z ∼ 0.5–0.9 using SDSS DR14 spectroscopy. We study the nuclear activity of the SQuIGG$\vec{L}$E sample using both optical narrow emission-line diagnostics and radio emission from FIRST. The radio AGN are likely dominated by systems that are accreting at very low Eddington fractions and are thus unobservable with optical diagnostics, while the optically selected AGN are accreting at a few percent of Eddington.

We replicate the previously observed trend between radio luminosity and stellar mass in radio AGN that has been reported previously (e.g., Best et al. 2005; Barišić et al. 2017), extending this finding to recently quenched galaxies. Radio AGN fractions are not a strong function of M_* or stellar age. AGN fractions are not a strong function of M_* for [O iii]-selected samples either, but the optical AGN fraction is a strong function of D_n4000. The youngest PSBs are 10 times more likely to harbor an optical AGN than the oldest PSBs. These younger galaxies, particularly those with D_n4000 < 1.3, are also most likely to harbor significant molecular gas fractions. What may this tell us about the relationship between BH and galaxy growth?

Temporal connections between merging, star formation, BH growth, and AGN-driven outflows have long been considered in the literature (e.g., Sanders et al. 1988; Canalizo & Stockton 2001; Springel et al. 2005). We note also that there is an ongoing debate about the relative roles of stellar mass and star formation rate on accretion rate distributions (e.g., Yang et al. 2018; Aird et al. 2019). Since we consider a very narrow range in stellar mass and redshift, we will focus only on the possible connections between star formation and accretion here. Because of the very short timescales for accretion compared with star formation, establishing an observational temporal link is challenging. Even if trends are observed, the connection could be due to either fueling or feedback.

A number of local studies of [O iii]-selected AGN have reported a boost in AGN fractions ≳600 Myr after a starburst (e.g., Kauffmann et al. 2003; Davies et al. 2007). These studies suggest that gas availability links AGN and star formation, perhaps with winds from massive stars providing the fuel for the AGN activity (Wild et al. 2010). Interestingly, Aird et al. (2019) find an increase (by 2–5 times) in the AGN fraction for X-ray-selected AGN in galaxies that live below the star-forming main sequence as compared with galaxies on the main sequence. These results all may point to a similar fueling mechanism. We are also seeing a dramatic increase in availability of molecular gas at the lowest D_n4000 in our sample. We thus postulate that the increase in AGN activity toward younger ages is a function of increased gas availability leading to fueling, rather than AGN feedback shutting down star formation in older systems. Of course, both could be true. Thus, we last consider whether we might link AGN activity to quenching in these systems.

Crudely speaking, theorists often conceptually divide the feedback roles of low and high accretion rate sources into a "maintenance" (or "radio") mode, and a radiatively efficient quenching mode (e.g., Croton et al. 2006). The former happens in a quasi-continuous fashion to keep hot gas envelopes from cooling (e.g., Fabian 2012). We see clear evidence for such activity in our galaxies, and if our radio observations were deeper we would undoubtedly uncover a very high radio active fraction. The latter is more elusive; despite some evidence for massive outflows in luminous AGN (e.g., Greene et al. 2012; Liu et al. 2013), we still do not have a clear verdict on the role of AGN in shutting down star formation in galaxies. In SQuIGG$\vec{L}$E, we will search for signs of outflow in integral-field unit (IFU) data (Hunt et al. 2018). So far, we have found one intriguing object showing extended ionized and molecular gas (J. Spilker 2020, in preparation). We will see if there are compelling signs of outflow in this or other targets to more cleanly disentangle the relationship between fueling and feedback in massive galaxies.

J.E.G., R.S.B., and D.N. gratefully acknowledge support from NSF grant AST-1907723. R.F. acknowledges financial support from the Swiss National Science Foundation (grant No. 157591). This work was performed in part at the Aspen Center for Physics, which is supported by National Science Foundation grant PHY-1607611.

Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III website is http://www.sdss3.org/.

SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofisica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group, University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University.