Role of environment on nuclear activity

TL;DR

This study investigates how the environment influences nuclear activity in galaxies, finding that AGN presence correlates mainly with stellar mass and morphology, and is only weakly affected by local galaxy density.

Contribution

It provides a comparative analysis of AGN fractions in dense clusters versus voids, highlighting the minimal impact of environment on nuclear activity in the local universe.

Findings

AGN galaxies are generally old with low star formation in both environments.

The AGN fraction increases faster with stellar mass in clusters than in voids.

Nuclear activity correlates more strongly with stellar mass and morphology than with environment.

Abstract

Motivated by the apparently conflicting results reported in the literature on the effect of environment on nuclear activity, we have carried out a new analysis by comparing the fraction of galaxies hosting active galactic nuclei (AGNs) in the most overdense regions (rich galaxy clusters) and the most underdense ones (voids) in the local universe. Exploiting the classical BPT diagnostics, we have extracted volume limited samples of star forming and AGN galaxies. We find that, at variance with star-forming galaxies, AGN galaxies have similar distributions of specific star formation rates and of galactic ages (as indicated by the Dn4000 parameter) both in clusters and in voids. In both environments galaxies hosting AGNs are generally old, with low star formation activity. The AGN fraction increases faster with stellar mass in clusters than in voids, especially above 10^10.2 M(sun). Our results indicate that, in the local universe, the nuclear activity correlates with stellar mass and galaxy morphology and is weakly, if at all, affected by the local galaxy density.

Tables

Table 1. Table 1 KS probabilities, P 𝑃 P , that distributions of sSFRs or of \text ​ D n \text subscript 𝐷 𝑛 \text{D}_{n} 4000 for different sample pairs of SFGs and AGN galaxies are drawn from the same parent population. P = 0.0 𝑃 0.0 P=0.0 means that the probability is < 0.001 absent 0.001 <0.001 .

			LSM			HSM
			sSFR	$\text{text}{D}_n$4000		sSFR	$\text{text}{D}_n$4000
	Void-Cluster		0.0	0.0		0.0	0.0
SFG	Void-Control		0.355	0.916		0.089	0.052
	Cluster-Control		0.0	0.0		0.016	0.008
	Void-Cluster		0.002	0.0		0.002	0.147
AGN	Void-Control		0.014	0.041		0.147	0.052
	Cluster-Control		0.0	0.0		0.0	0.0

Full text

††slugcomment: Accepted in Astrophysical Journal(ApJ)††affiliationtext: amirnezamamiri@gmail.com

1. Introduction

It has long been known that, in the nearby universe, galaxy properties such as star formation rate (SFR), morphology and stellar mass are strongly correlated with the surrounding galaxy density (Oemler1974; Dressler1980; Hashimoto1998; Kauffmann2004; Baldry2006). Massive early-type galaxies in passive evolution are preferentially found in high density environments while the specific SFR increases towards lower density regions. This suggests that the environment has a substantial role in driving galaxy evolution, although the details of the physical processes involved are still poorly understood.

We focus here on the role of environment on nuclear activity. This is relevant to constrain processes that trigger and control it and to investigate its relationship to star formation. Several analyses have been carried out considering various definitions of the local density. It is generally agreed that the fraction of galaxies hosting active galactic nuclei (AGN) depends on stellar mass Pimbblet et al 2013 and references therein, but the dependence on environment at fixed stellar mass is debated.

Carter2001 claimed that the AGN fraction, $F_{\rm AGN}$, is insensitive to the environment. A similar conclusion was reached by Miller2003 who observed that while the fraction of star-forming galaxies decreases with density, the fraction of galaxies with an AGN remains constant from the dense cores of galaxy clusters to the low density field. Sabater2015 argued that the effect of the local density is minimal at fixed stellar mass and central star formation activity. This is in agreement with the analysis of Yang2018 who found that the sample-averaged accretion rate onto super-massive black holes does not show any significant dependence on overdensity or cosmic-web environment once the stellar mass of host galaxies is controlled, and with the finding by Karhunen2014 that quasar environments are not significantly different from those of normal galaxies with similar luminosities.

In contrast, Kauffmann2004 reported a strong anti-correlation of nuclear activity on local density at fixed stellar mass for powerful AGNs ($L({\rm[OIII]})>10^{7}\,L_{\odot}$), generally residing in massive galaxies with significant star formation activity. The strong dependence of the prevalence of nuclear activity with density and interactions was found after taking into account the effect of mass. According to Constantin2008, however, such a trend holds only for moderately bright ($M_{\rm r}\approx-20$), and moderately massive ($10<\log(M_{\star}/M_{\odot})<10.5$) galaxies; their data do not show any statistically significant excess of any type of AGN in the most massive void galaxies relative to those in relatively crowded regions (“walls”). Deng2012 found a different environmental dependence for low and high stellar mass galaxies. In the former $F_{\rm AGN}$ depends very little on local density but, in the latter, it decreases with increasing density. Lopes2017 confirmed these conclusions.

The opposite conclusion was reached by Manzer2014 who reported a significant increase in the AGN fraction in group environments compared with isolated galaxies. A larger fraction of AGN in denser environment was also found by Argudo2018 for quenched and red galaxies.

A related issue is the role of interactions/mergers in triggering nuclear activity. This problem has been addressed in different ways. Schawinski2010 looked for major morphological disturbances in a sample of spheroidal galaxies in the process of migrating from the blue cloud to the red sequence via an AGN phase. Other studies compared the AGN fraction in samples of close galaxy pairs with that in a control sample of galaxies with similar masses and redshifts but no nearby companion (Ellison2011; Ellison2015; saty_ellisonetal2014; ScottKaviraj). The general conclusion of these investigations is that interactions/mergers substantially enhance the specific SFR, but the effect on AGN activity depends on the AGN type. The AGN fraction was found to be substantially higher in close galaxy pairs (Ellison2011; Ellison2015; saty_ellisonetal2014), implying that mergers can trigger nuclear activity and enhance the accretion rate. However, the fraction of low excitation radio galaxies (LERGs) was found to be independent of projected separation, implying that LERGs are not fuelled by mergers (Ellison2015). ScottKaviraj reported a modest decrease of the Seyfert fraction in close pair galaxies compared to isolated galaxies, suggesting that either mergers may not trigger AGN activity at the close pair stage or may trigger a different AGN type. Alternatively, the lack of a clear link between merger features and AGN activity may be due to the large time delay between the merger-driven starburst and the peak of AGN activity, allowing the merger features to decay (Schawinski2010).

The apparently contradictory results on the effect of environment on nuclear activity have motivated our re-analysis. In this paper, we compare the extremes of the density distribution of large scale structures: the most overdense regions (galaxy clusters) and the most underdense ones (voids). In Sect. 2, we present the samples used and describe their properties. The results of our analysis are reported in Sect. 3 while in Sect. LABEL:sec:conclusions we summarize our main conclusions.

2. Sample selection and galaxy classification

We have selected a volume-limited sample of void galaxies brighter than $\text{M}_{r}=-18$ drawn from the void catalogue by Tavasoli2015. The sample covers the redshift range $\text{0.01}<{z}<{0.04}$, the maximum redshift being determined by the limiting absolute magnitude, and comprises 1159 galaxies.

Cluster galaxies in the same redshift and absolute magnitude range of void galaxies were extracted from the catalogue by Tempel14, considering only clusters with at least 15 spectroscopic members. Brightest cluster galaxies (BCGs) themselves were excluded in consideration of their different evolution compared to the other member galaxies (Lauer2014). The final sample contains 2870 cluster galaxies.

To separate star-forming galaxies (SFGs) from galaxies hosting AGNs (hereafter AGNs) we exploited the classical BPT diagnostics (Baldwin1981) based on emission line ratios. The emission line intensities for both void and cluster galaxies were taken from the Max-Planck-Institute for Astrophysics (MPA)-Johns Hopkins University (JHU) SDSS DR7 catalog (MPA/JHU; kauf03; Brinchmann2004) with $\text{S/N}>3$ for all objects. All the 1159 void galaxies and 2865 out of the 2870 cluster galaxies have emission lines and are included in the MPA/JHU catalog.

Figure 1 shows the distribution of our galaxies in the [OIII]5007/H$\beta$ versus the [NII]6584/H$\alpha$ diagram together with the boundary lines between SFGs and AGNs defined by Kewley01 and by kauf03. The demarcation line by Kewley01 is a theoretical upper limit on the location of SFGs in this diagram, obtained using a combination of photo-ionization and stellar population synthesis models. It yields a conservative AGN selection. kauf03 have revised the boundary line on the basis of the observational evidence that SFGs and AGNs are distributed along two separated sequences. It yields a conservative SFG selection.

In order to avoid ambiguous classifications we have adopted the more conservative selection for both SFGs and AGNs, excluding from further consideration galaxies located between the two lines. The excluded objects are a minor fraction: $\sim 10\%$ of void galaxies and $\sim 16\%$ of cluster galaxies. Thus this exclusion cannot significantly bias our results. This classification yielded 588 cluster AGNs, 118 void AGNs, 1077 cluster SFGs and 902 void SFGs.

To check the stability of our classification we have re-classified our sources using the criterion by LaraLopez2010 which exploits a different set of emission lines (H$\alpha$, [NII] and [SII]). We got the same classification for $\simeq 97\%$ of void and $\simeq 82\%$ of cluster SFGs, and for $\simeq 83\%$ void and $\simeq 85\%$ cluster AGNs.

We note that while the adopted classification proved to be solid and free from contamination by objects of uncertain type, there is a price to pay for that: we are missing galaxies with significant contributions from both nuclear activity and star-formation, which are located in the intermediate region between pure SFGs and pure AGNs in the BPT diagram; however, as already mentioned above, they are a small fraction of the initial sample.

We are also missing strongly obscured AGNs which may fall in the SFG region of the diagram as well as dust-enshrouded SFGs which do not show emission lines. However only a small fraction of AGNs are heavily obscured ($7.6(+1.1,-2.1)\%$; Ricci2015), and the mean AGN fraction in the local universe is substantially lower than that of star-forming galaxies, as shown by the many references on that cited in Section 1. The fraction of dust-enshrouded SFGs is also very small (HwangGeller2013). We can thus safely assume that both source types are too rare to significantly affect our results.

As mentioned in Section 1, it is widely agreed that $F_{\rm AGN}$ strongly correlates with stellar mass, $M_{\star}$, so that the effect of environment must be investigated at fixed $M_{\star}$. To this end to compare the properties of SFGs and AGNs in clusters and in voids in a way as homogeneous as possible, AGN and SFG galaxies in both environments were randomly drawn from the parent samples in such a way to have the same number of objects in each stellar mass bin so that the stellar mass distribution is the same for both populations (Fig. 2). Each sample has been subdivided in two stellar mass ranges containing similar numbers of galaxies: $9.4\leq\log(M_{\star}/M_{\odot})<10.2$ (low stellar mass; LSM, 60 galaxies per subsample) and $10.2\leq\log(M_{\star}/M_{\odot})\leq 11.0$ (high stellar mass; HSM, 49 objects per subsample). Galaxy properties come from the MPA/JHU catalog which includes accurate aperture corrections (Brinchmann2004). For AGNs, SFRs could not be estimated from classical SFR diagnostics like the H$\alpha$ luminosity because it is affected by the AGN component. Therefore the $\text{D}_{n}4000$ values were used instead. In turn, the star-formation contribution affects the strengths of AGN lines, diluting their ratios. This however does not affect our analysis since we are using line ratios only to select AGNs and the use of the conservative Kewley01 limit ensures that our AGN sample is clean.

3. Results

The four upper panels of Fig. 3 contrast the distributions of specific SFRs (sSFRs, i.e. SFRs per unit stellar mass) in clusters and in voids for SFGs and AGNs subdivided into LSM and HSM galaxies. The distributions for field SFGs and AGNs in the same ranges of redshift and stellar mass (control samples) are also shown for comparison. The control samples where built randomly selecting from the SDSS DR7 catalog galaxies with the same criteria used for our SFG and AGN samples, excluding those in clusters and in voids. The selected galaxies were then classified as SFGs or AGNs in the same way as void and cluster galaxies. In this way we obtained SFG and AGN control samples five times larger than void and cluster samples, i.e. containing 545 galaxies each. In both stellar mass ranges the distribution of sSFRs of cluster SFGs galaxies is broad and extends to low values, consistent with the notion that the central regions of clusters are populated by red galaxies in essentially passive evolution, while star-forming late-type galaxies populate the cluster outskirts. The red, passive population is rare in voids, whose sSFR distribution is concentrated at relatively large values. The distribution of the control sample is intermediate between those in clusters and in voids (although somewhat closer to that in voids) for the HSM galaxies but much closer to that in voids for LSMs. A quantitative comparison is provided by the probabilities of the null hypothesis (same parent distribution for different SFG samples) for void/cluster, void/control sample and cluster/control sample SFGs given by the Kolmogorov-Smirnov (KS) test, listed in Table LABEL:tab:KS. The distributions of sSFRs of void and cluster AGNs are more similar, with a single peak at quite low sSFRs. However the distributions for both HSM and LSM galaxies in voids have a substantial tail extending to much higher sSFRs than in clusters. Again the distributions of control samples are more similar to those in voids, especially for HSM galaxies; see Table LABEL:tab:KS for a quantitative comparison with the KS test.