RADIO LOUD AGNs ARE MERGERS

The first sample we consider is composed by the radio galaxies with $z\gt 1$ from the 3CR catalog (Spinrad et al. 1985). These are all powerful radio galaxies belonging to the Fanaroff–Riley class II (FR II; Fanaroff & Riley 1974), i.e., those in which the brightest components of the radio structure lie at the edges of the radio source, in contrast to the FR Is, in which the peak of the radio emission is located at the core.

The 3CR is a flux limited sample selected at low radio frequencies (${{S}_{178}}\gt 9$ Jy at 178 MHz). Since the radio emission at such frequencies is dominated by the radio lobes, the selection is independent of the AGN orientation. The original catalog includes both Type 1 (QSOs) and Type 2 objects (radio galaxies) and it is one of the best studied samples of radio-loud AGNs, being perfectly suitable to test AGN unification scenarios. The high redshift objects included in the 3CR are all firmly established radio-loud AGNs with powerful relativistic jets emanating from the central SMBH. Even at the lowest luminosities, the active nucleus is strongly radio-loud (see e.g., Chiaberge et al. 2005; Sikora et al. 2007), assuming the canonical threshold $R={{F}_{5\,{\rm GHz}}}/{{F}_{B}}\gt 10$ (Kellermann et al. 1989), where ${{F}_{5\,{\rm GHz}}}$ and F_B are the fluxes in the radio band at 5 GHz and in the optical B band, respectively. The $z\gt 1$ sub-sample of the 3CR catalog includes 58 objects. The highest redshift object is the radio galaxy 3C257 at z = 2.47.

The radio power of these AGNs is ${{L}_{151}}\sim {{10}^{35}}$ erg ${{{\rm s}}^{-1}}$ Hz⁻¹ or slightly higher, which corresponds to a bolometric luminosity ${{L}_{{\rm bol}}}\sim {{10}^{45-46}}$ erg s⁻¹ assuming standard bolometric corrections. This result is also supported by the estimates of the X-ray luminosity for some of these objects, which is typically in the range ${{L}_{2-10}}\sim {{10}^{44-45}}$ erg s⁻¹. Salvati et al. (2008) observed a sample of the most luminous and most distant radio galaxies and QSOs from the 3CR catalog with XMM. They found that the intrinsic X-ray luminosity is in the range $6\times {{10}^{44}}\lt {{L}_{2-10\,{\rm keV}}}\lt 2\times {{10}^{46}}$ erg s⁻¹, the Type 1 QSOs being a factor of ∼6 or higher brighter than the (Type 2) radio galaxies. These authors interpreted such a discrepancy as a result of the presence of a beamed component in the Type 1 QSOs. Wilkes et al. (2013) found similar results based on Chandra data. E. Torresi et al. (2015, private communication) analyzed both Chandra and XMM archival data for the 3CR radio galaxies observed with HST, and found intrinsic X-ray luminosities as low as ${{L}_{2-10\,{\rm keV}}}\sim 2\ \;\times \ \;{{10}^{44}}$ erg s⁻¹. However, the absorbing column density is poorly constrained, and it is possible that at least some of the Type 2 radio galaxies are in fact Compton-thick. In that case, the derived X-ray luminosity sets a lower limit to the intrinsic AGN luminosity in that band.

Twelve 3CR radio galaxies were observed with WFC3-IR and the F140W filter (in addition to WFC3-UVIS and F606W, not used in this work) as part of program SNAP13023 (PI M. Chiaberge, B. Hilbert et al. 2015, in preparation). The wide-band filter used for the HST observations (F140W) includes emission lines at the redshift considered here (mainly Hα, Hβ, and/or [O iii]5007, depending on the redshift of the source). However, the only object that is significantly contaminated by line emission is 3C 230 (see Steinbring 2011). The HST image of this specific object only shows the narrow-line region and thus we cannot observe the stellar component of the host. Therefore we exclude this source from our sample. ⁶ HST snapshot surveys of complete samples are well suitable for statistical studies, since the observations are scheduled by randomly picking objects from the original target list to fill gaps in the HST schedule. The observed sample spans the entire range of redshift of the original list, i.e., from z = 1.0 to 2.47. The comparison samples described in the following are tailored to match the properties of these 3CR galaxies, in either bolometric power or redshift range (or both).

In Table 1 we report the data for the 11 sources belonging to this sample. The 24 μm luminosities taken from Podigachoski et al. (2015) are used in the following to check that the relevant comparison sample is correctly matched to the Hz3C.

Table 1.  High-z 3CR Radio Galaxies

Name	R.A. (2000.0)	Decl. (2000.0)	Redshift	log ${{P}_{1.4\,{\rm GHz}}}$	${{L}_{24\,\mu m}}$
	hh:mm:ss.ss	dd:mm:ss.ss	z	log[erg s−1 Hz−1]	log[erg s−1 Hz−1]
3C 210	08:58:10.0	+27:50:52	1.169	35.13	45.65
3C 255	11:19:25.2	−03:02:52	1.355	35.29	<44.54
3C 257	11:23:09.2	+05:30:19	2.474	35.94	45.94
3C 297	14:17:24.0	−04:00:48	1.406	35.33	44.84
3C 300.1	14:28:31.3	−01:24:08	1.159	35.37	⋯
3C 305.1	14:47:09.5	+76:56:22	1.132	35.10	45.36
3C 322	15:35:01.2	+55:36:53	1.168	35.20	44.90
3C 324	15:49:48.9	+21:25:38	1.206	35.34	45.48
3C 326.1	15:56:10.1	+20:04:20	1.825	35.75	45.64
3C 356	17:24:19.0	+50:57:40	1.079	34.96a	45.52
3C 454.1	22:50:32.9	+71:29:19	1.841	35.60	45.67

Note. ${{P}_{\,1.4}}$ from Condon et al. (1998) except for item ^a from Laing & Peacock (1980). The 24 μm luminosities are derived from the fluxes published in Podigachoski et al. (2015).

Download table as:  ASCIITypeset image

In Figure 1 we show the location of the 3CR sample in the radio power versus redshift plane, with respect to other samples used throughout the paper. Note that, at any redshift, 3CR objects are always the most powerful radio sources.

Zoom In Zoom Out Reset image size

The sample of high redshift low-luminosity radio galaxies (HzLLRGs) is derived from the Extended Chandra Deep Field South (ECDFS Lehmer et al. 2005). We use the Bonzini et al. (2012) catalog and AGN type classification to derive this sample. The criteria are as follows.

• (i)  
  The spectroscopic, if available, or photometric redshift of the source must be in the range $1\lt z\lt 2.5$. The redshift range is chosen to roughly match the range spanned by the 3CR sample described above.
• (ii)  
  The total radio power at 1.4 GHz must be greater than ${{P}_{1.4\,{\rm GHz}}}\gt {{10}^{30}}$ erg ${{{\rm s}}^{-1}}$ Hz⁻¹ and below the fiducial FRI/FRII separation ${{P}_{1.4\,{\rm GHz}}}\lt 4\times {{10}^{32}}$ erg ${{{\rm s}}^{-1}}$ Hz⁻¹ (Fanaroff & Riley 1974).
• (iii)  
  The source must be classified as a radio-loud AGN by Bonzini et al. (2013), i.e., it must show radio emission in excess of that produced by starbursts. This is measured by using the ${{q}_{24{\rm obs}}}$ parameter, defined as ${{q}_{24{\rm obs}}}={\rm log} ({{S}_{24{\rm obs}}}/{{S}_{r}})$, where ${{S}_{24{\rm obs}}}$ is the observed flux density at 24 μm and ${{S}_{r}}$ is the observed flux density at 1.4 GHz. Bonzini et al. (2013) define an object as radio-loud if ${{q}_{24{\rm obs}}}$ is 2σ off of the starburst locus, defined using the M82 template (see their Figure 2).
• (iv)  
  We also impose the additional constraint that its X-ray luminosity must not exceed ${{L}_{2-10\,{\rm keV}}}={{10}^{44}}$ erg s⁻¹. This is to ensure that the objects are consistent with typical low power radio galaxies at both low and high redshifts (e.g., Balmaverde et al. 2006; Tundo et al. 2012). The upper limit to the X-ray luminosity is also consistent with the results of Terashima & Wilson (2003). These authors showed that the radio-loud–radio-quiet dichotomy can be redefined using the X-ray emission instead of the flux in the optical band. This is particularly useful for Type 2 AGNs such as the ones considered in this work, since the optical emission form the accretion disk is heavily obscured in these object. According to such a scheme, powerful radio-loud AGNs are present for ${{R}_{{\rm X}}}\gtrsim -3.5$, where R_X is defined as the logarithm of the ratio between the radio luminosity ($\nu {{L}_{\nu }}$) at 1.4 GHz and the X-ray luminosity ${{L}_{2-10}}$. At low radio powers, it is easy to confuse a powerful radio-quiet AGN that shows some radio emission with a radio-loud object that is intrinsically weak at all wavelengths. Note that our selection criteria for the radio and X-ray luminosities correspond to ${{R}_{{\rm X}}}\gt -2.3$, which is a rather conservative value. We prefer to follow a conservative approach because for low-luminosity AGNs the transition between radio-quiet and radio-loud most likely occurs at higher values than for high power objects, similarly to the classical radio-loudness parameter R derived using the radio-to-optical luminosity ratio (Terashima & Wilson 2003; Chiaberge et al. 2005; Sikora et al. 2007).
• (v)  
  The source must be observed with HST WFC3-IR. Most of the images are taken from the CANDELS survey (Grogin et al. 2011; Koekemoer et al. 2011), or from surrounding fields observed as part of program GO-12866 (see Table 2). The ECDF sample includes six HzLLRGs.

Table 2.  High-z Low Luminosity Radio Galaxies

ID	R.A. (2000.0)	Decl. (2000.0)	Redshift	HST Survey or Prog. ID	log ${{P}_{1.4\,{\rm GHz}}}$	log ${{L}_{2-10}}$	q24
	hh:mm:ss.ss	dd:mm:ss.ss	z		log[erg s−1 Hz−1]	log[erg s−1]
ECDFS
65	03:31:23.30	−27:49:05.80	1.10 a	12866	31.85	<42.56	−1.54
127	03:31:34.13	−27:55:44.40	1.06	12866	30.51	<42.69	−0.09
215	03:31:50.74	−27:53:52.15	1.77	12866	31.13	<42.69	−0.46
338	03:32:10.79	−27:46:27.80	1.61 b	CANDELS	31.29	42.43	−0.98
410	03:32:19.30	−27:52:19.38	1.10 b	CANDELS	30.37	<42.27	−0.23
412	03:32:19.51	−27:52:17.69	1.06	CANDELS	30.93	<42.24	−0.51
UDS
48	02 18 18.38	−05 15 45.2	1.56	CANDELS	32.27	<43.93	0.09
124	02 17 04.77	−05 15 18.1	1.28 b	CANDELS	31.54	<43.73	−0.50

Note. For the ECDFS sources the ID corresponds to the source ID in the Bonzini et al. (2012) catalog. For the UDS galaxies, the ID is the Simpson et al. (2012) source number. Redshifts are photometric redshifts from Bonzini et al. (2012), except where stated otherwise. The UDS sources are undetected in the X-rays.

^a Photometric redshift from Cardamone et al. (2010). ^b Spectroscopic redshift.

Download table as:  ASCIITypeset image

We also use the same criteria to select HzLLRG in the CANDELS UDS field in the Subaru/XMM Newton Deep field survey (Ueda et al. 2008). The radio data are from Simpson et al. (2012), the 24 μm Spitzer IR data are taken from the Spitzer Public Legacy Survey of the UKIDSS Ultra Deep Survey (SpUDS, P.I. J.S. Dunlop) using the IRSA database. Only two galaxies that satisfy our selection criteria lie in the region covered by the HST WFC3 observations.

The properties of the eight galaxies included in the HzLLRG sample are reported in Table 2. In Figure 1 we show the radio power versus redshift distribution of these AGNs, compared with the other radio-loud samples. Note that the two radio-loud samples at $z\gt 1$ used in this paper (i.e., the 3CR and the HzLLRGs) are separated by about 4 dex in radio power, on average.

The sample of low-power Type 2 AGNs (LPTy2AGN) is selected from the 4Msec CDFS catalog and source classification (Xue et al. 2011). We include objects that are classified as AGNs, that do not show strong broad emission lines in the optical spectrum, and whose redshift is $1.0\lt z\lt 2.5$. We also use the additional constraint that the intrinsic (de-absorbed) X-ray luminosity (integrated between 2 and 10 keV, in the rest frame of the source) must be ${{L}_{2-10\,{\rm keV}}}\lt 2\times {{10}^{42}}$ erg s⁻¹, in order to be consistent with the properties of the corresponding radio-loud sample described in Section 2.2. The ${{L}_{2-10\,{\rm keV}}}$ luminosity is derived from the 0.5–8 keV luminosity listed in the catalog, converted to the 2–10 keV band using a photon index Γ = 1.8. Furthermore, for the objects that are detected in the radio, we require that the X-ray radio-loudness parameter R_X defined in Section 2.2 is $\lt -3.5$. The sample includes 26 objects. These are very low-power AGNs that are similar to Seyfert 2 galaxies in the local Universe. The properties of these AGNs are reported in Tables 3. The HST/WFC3-IR images for this sample are taken from the CANDELS survey (Grogin et al. 2011; Koekemoer et al. 2011).

Table 3.  Radio Quiet Low Power Ty2 AGNs at $1\lt z\lt 2.5$

ID	R.A.	Decl.	Redshift	log ${{L}_{2-10}}$
	hh:mm:ss.s	dd:mm:ss.s	z	log[erg s−1]
Low Power Type 2 AGNs (CDFS)
125	03:32:06.77	−27:49:14.10	1.050	42.04
147	03:32:09.22	−27:51:43.50	1.352	41.99
216	03:32:15.26	−27:44:38.60	1.109	41.70
225	03:32:15.91	−27:48:02.20	1.520	41.90
226	03:32:16.04	−27:48:59.90	1.413	41.91
247	03:32:17.84	−27:52:10.80	1.760	42.15
317	03:32:23.16	−27:45:55.00	1.224	42.25
318	03:32:23.17	−27:44:41.60	1.571	42.09
321	03:32:23.61	−27:46:01.40	1.033	41.98
376	03:32:27.04	−27:53:18.60	1.103	41.88
389	03:32:28.62	−27:45:57.20	1.626	41.97
394	03:32:28.85	−27:47:56.00	1.383	41.89
416	03:32:29.94	−27:52:52.80	1.017	41.76
419	03:32:30.05	−27:50:26.80	1.005	41.48
428	03:32:31.11	−27:49:40.00	1.508	41.79
455	03:32:33.06	−27:48:07.80	1.188	41.75
462	03:32:33.67	−27:47:51.20	1.388	41.84
463	03:32:33.84	−27:46:00.60	1.903	42.09
491	03:32:35.80	−27:47:35.10	1.223	42.13
493	03:32:35.98	−27:48:50.70	1.309	42.19
504	03:32:36.35	−27:49:33.40	1.508	41.81
536	03:32:39.07	−27:53:14.80	1.380	41.98
541	03:32:39.42	−27:53:12.70	1.381	42.00
545	03:32:39.65	−27:47:09.60	1.317	41.88
558	03:32:41.01	−27:51:53.40	1.476	42.19
579	03:32:43.45	−27:49:02.20	1.603	41.97

Note. The ID corresponds to the source ID in the Xue et al. (2011) catalog. The redshifts (${{z}_{{\rm adopt}}}$ in the Xue et al. catalog) are spectroscopic, if available, or photometric. The intrinsic X-ray luminosities (converted to the 2–10 keV band as explained in the text) are also from Xue et al. (2011).

Download table as:  ASCIITypeset image

As for the low-power AGNs described above, the sample of high-power Type 2 AGNs (HPTy2AGN) is drawn from the CDFS. When selecting this sample we want to match as close as possible both the redshift range and the bolometric luminosity of the powerful 3CR radio galaxy sample.

The sample of HPTy2AGN includes objects classified as AGN with redshift $1\lt z\lt 2.5$ and intrinsic X-ray luminosity ${{L}_{2-10\,{\rm keV}}}\gt {{10}^{44}}$ erg s⁻¹. The lower limit in X-ray power is chosen to match the properties of the 3CR sample described in Section 2.1. Two of the objects (166 and 577) show relatively broad Mg ii and C iii] emission lines in the optical spectrum (Szokoly et al. 2004). However, the HST images clearly show the host galaxy and there is no evidence for the presence of any strong unresolved nuclear source, as it would be expected for a powerful Type 1 QSO. Therefore, for the purpose of this work, we consider those two objects as Type 2, regardless of the presence of broad lines.

As for the LPTy2AGN, for the objects that are detected in the radio, we require that the X-ray radio-loudness parameter R_X defined in Section 2.2 is $\lt -3.5$. This sample includes 9 AGNs (see Table 4). The HST/WFC3-IR images for this sample are taken from the CANDELS survey (Grogin et al. 2011; Koekemoer et al. 2011).

Table 4.  Radio Quiet High Power Ty2 AGNs at $1\lt z\lt 2.5$

ID	R.A.	Decl.	Redshift	log ${{L}_{2-10}}$	log ${{L}_{24\mu {\rm m}}}$
	hh.mm.ss.s	dd.mm.ss.s	z	log[erg s−1]	log[erg s−1]
166	03:32:10.93	−27:44:15.20	1.605	44.21	44.81
243	03:32:17.19	−27:52:21.00	1.097	44.05	44.39
257	03:32:18.35	−27:50:55.61	1.536	44.07	44.30
278	03:32:20.07	−27:44:47.51	1.897	44.06	45.04
351	03:32:25.70	−27:43:06.00	2.291	44.31	45.46
518	03:32:37.77	−27:52:12.61	1.603	44.23	45.38
577	03:32:43.24	−27:49:14.50	1.920	44.12	44.70
713	03:33:05.90	−27:46:50.70	2.202	44.02	⋯
720	03:33:07.64	−27:51:27.30	1.609	44.39	⋯

Note. The ID corresponds to the source ID in the Xue et al. (2011) catalog. The redshifts (${{z}_{{\rm adopt}}}$ in the Xue et al. catalog) are spectroscopic, if available, or photometric. The intrinsic X-ray luminosities (converted to the 2–10 keV band as explained in the text) are also from Xue et al. (2011). The 24 μm luminosities are from Cardamone et al. (2008).

Download table as:  ASCIITypeset image

Since the X-ray luminosity might be poorly constrained because of the uncertainty on the amount of nuclear obscuration, we should also make sure that this sample and the Hz3Cs have similar mid-IR luminosities. We performed a K–S test to check whether the distributions of 24 μm luminosity of the Hz3C and HPTy2AGN are different. ⁷ The resulting p-value is p = 0.1, therefore we cannot reject the null hypothesis that the two distributions are drawn from the same population.

This sample is derived from the 3D-HST survey of the GOODS-SOUTH field (Giavalisco et al. 2004; Brammer et al. 2012; Skelton et al. 2014). We select galaxies with spectroscopic redshift $1\lt z\lt 2.5$ and with magnitude in the F140W filter 19 $\lt \;{{m}_{{\rm F}140{\rm W}}}\lt$ 22. The magnitude range is chosen to match the magnitudes of the 3CR radio galaxies from Section 2.1 (B. Hilbert et al. 2015, in preparation). The full sample includes 145 galaxies. We limit the sample to the 50 objects included in the southern half area of the field observed with WFC3-IR and the F140W filter as part of the 3D-HST survey. Five of these objects are AGNs, therefore we are left with a sample of 45 galaxies. These are sufficient to provide us a statistically sound sample to be compared with the active galaxy samples described above. We refer to this sample as the bright sample of non-active galaxies. Note that we could in principle derive a larger sample by using the full area covered by the 3D-HST image of the GOODS-S field. However, this would not significantly improve the statistics for a sample that is already the largest we consider.

In the same area of the sky, we also select a comparable sample of fainter galaxies ($22\lt {{m}_{{{K}_{s}}}}\lt 24$) from the same HST survey. This sample spans the same magnitude range as the bulk of our radio-quiet AGN population. In the following, we will refer to this sample as the faint galaxies sample.

Even if we are selecting objects in the very same redshift range, the near-IR (rest-frame optical) magnitude might not be a good tracer of the stellar mass because of the possible presence of obscuration in some objects. We do not believe that this might significantly affect our results. However, in order to perform a sanity check, we also select a sample of non-active galaxies in the full GOODS-S area of 3D-HST matched to the stellar mass estimates of the Hz3C galaxies. At $1\lt z\lt 2.5$ these are typically between $1\times {{10}^{11}}$ and $5\times {{10}^{11}}$ ${{M}_{\odot }}$ (Seymour et al. 2007), although a small number of lower mass objects are also present. Although stellar mass measurements based on SED fitting heavily rely on models, it is important to check that our non-active galaxy samples are correctly representing the population of objects we need for comparison. If we restrict our range of masses between the above values, and after excluding those that are AGNs, we obtain 23 objects. About half of them are in common with the bright sample. While we do not consider this as our main control sample of non-active galaxies, results for this sample are briefly discussed in Section 4, for the sake of completeness.

The HST/WFC3-IR images are taken as part of the 3DHST survey with the F140W filter.

The goal of this work is to identify merging systems in samples of objects that were observed as part of different surveys or observing programs. All of the images were taken using the WFC3-IR camera. Its extremely high sensitivity and spatial resolution allows us to detect low surface brightness features that characterize recent merger events at z ∼ 1–2 even in short (1 orbit or less) observations. However, for the sake of clarity, we report in Table 6 the 5σ surface brightness limits for the different surveys we use. To derive the limits we used the WFC3-IR Exposure Time Calculator (ETC). We assume a 2 × 2 pixel extraction area and a spectrum of an elliptical galaxy with a UV upturn to perform the conversion between F160W (used for the CANDELS observations) and F140W magnitudes. Although the CANDELS deep images we used for the radio-quiet samples and for the HzLLRGs are deeper than all other data, we checked that the merger classification is not different if the shallower 3D-HST images are considered. The images show the very same features, irrespective of the redshift of the source. The faintest tidal structures are less prominent in the shallower images, but that does not significantly affect our classification. The short exposure times of the 3C snapshot survey are also sufficient to detect the faint features we are interested in, in the range of redshift of our sample, as shown by the fact that that sample has the greater observed merger fraction.

Table 6.  Surface Brightness Limits

Survey or HST Prog. ID	Sample	HST Camera/Filter	5σ Surface Brightness Limit	Sensitivity
			${{\mu }_{{\rm F}140{\rm W}}}$ (AB mag arcsec−2)	AB mag
SNAP-13023	Hz3C	WFC3/F140W	23.9	25.8
CANDELS wide	RQ and HzLLRG	WFC3/F160W	24.1	26.5
CANDELS deep	RQ and HzLLRG	WFC3/F160W	25.6	27.2
3D-HST	non-active	WFC3/F140W	24.1	26.2
GO-12866	HzLLRG	WFC3/F160W	25.0	26.7
Low-z Samples
GO-9045	3CRR,6C,7C,TOOT	WFPC2/F785LP	20.7	24.9
SNAP-10173	3CR ($z\lt 0.3$)	NICMOS/NIC2/F160W	22.5	24.3

Note. The 5σ limits are estimated using the WFC3-IR ETC and assuming a 2 × 2 pixel extraction region. In column 4 we report the limit surface brightness for each data set converted to the WFC3-IR F140W filter to allow for an easy comparison between the different surveys. An elliptical galaxy spectrum with UV upturn redshifted to the appropriate redshift for each sample was used for the conversion between the two WFC3 filters. RQ in the samples column refers to both the LP and HPTy2AGN samples. For the low redshift samples (GO9045 and SNAP10173) we used synphot to convert magnitudes to the WFC3 filter system. In column 5 we report the 5σ image sensitivities calculated for point sources in each of the bands used for the observations.

Download table as:  ASCIITypeset image

One important goal of this work is to establish whether the merger fraction for radio-loud AGN depends on either redshift or luminosity. In order to test if that is the case, we used additional lower redshift samples, for comparison with the $z\gt 1$ data.

McLure et al. (2004) observed a sample of 41 intermediate (z ∼ 0.5) radio galaxies with HST and WFPC2 (GO 9045, P.I. Willott). The sample is taken from four different complete catalogs, spanning about 4 dex in radio power. The four catalogs are all complete, flux limited and radio selected at low frequencies. The sample includes objects from the 3CRR, 6CE, 7CRS, and the TexOx-1000 (TOOT sample, hereinafter) (McLure et al. 2004, and references therein). The redshift range spanned by the objects observed with WFPC2 is between z = 0.40 and 0.59. We will refer to these galaxies as the Willott sample throughout the paper.

For these objects, we retrieved the HST data from the MAST and we reduced them using Astrodrizzle (Fruchter et al. 2012). We then classify the objects between mergers and non-mergers, since no classification was provided in the original paper by McLure et al. (2004), and for consistency with the other samples. Note that the filter used for these observations (F785LP) provides images at ∼5300 Å in the rest frame. This is similar to the rest frame wavelength of the WFC3-IR observations for the $z\gt 1$ samples described above. In Table 8 we report the merger fractions for this sample. The merger fractions are statistically compatible with those for the radio-loud samples at higher redshifts.

Table 8.  Estimated Bayesian Probabilities For the $z\lt 1$ Radio-loud AGN Samples

Sample	Size	Merger Fraction	95% Credible Interval
Willott Samples
3CRR	13	74%	0.53–0.94
6CE	7	68%	0.38–0.93
7CRS	9	65%	0.37–0.89
TOOT	12	65%	0.39–0.87
Full Sample	41	70%	0.56–0.83
Low Redshift Radio galaxies
$z\lt 0.3$ 3CR	101	88%	0.81–0.94

Download table as:  ASCIITypeset image

At even lower redshifts, we use the observations of the 3CR sample with $z\lt 0.3$ taken with HST-NICMOS (program SNAP10173) and the F160W filter (Madrid et al. 2006). Although these observations are in the rest-frame IR, they are more suitable for our purposes than the optical data of the same sample taken with WFPC2 (de Koff et al. 1996; Martel et al. 1999) because the NICMOS images are significantly deeper. For this sample, we consider the results published by Floyd et al. (2008). These authors classified all of the objects in mergers, pre-mergers, tidal-tails, major, and minor companions. The field of view was covered by NIC2 is 192 × 192, which corresponds to 50 × 50 kpc² at the median redshift of the sample. They found that 89 out of 101 objects fall into at least one of these categories. This corresponds to a merger fraction of 0.88, with a Bayesian 95% credible interval [0.81–0.94]. If we exclude objects with only minor companions (i.e., $\gtrsim 1$ mag fainter than the radio galaxy host, as defined by Floyd at al., 2008), the fraction is reduced to 0.76 with a Bayesian 95% credible interval [0.68–0.84].

Since the observations are not homogeneous, these two lower redshift samples are only considered here for comparison. However, we do not base our conclusions on these samples only.

In this section we discuss the results we obtained for the different radio-loud samples. It is in fact particularly interesting to investigate whether the merger fraction in these objects may depend on, e.g., redshift, luminosity, or on the original criteria for each sample selection. However, while the statistical analysis clearly shows that radio-loud objects are almost all associated with mergers, one important caveat is that the samples are small. This is particularly true for the radio-loud samples (except for the low-z 3CR), but also for the high-z radio-quiet comparison sample. Unfortunately, this prevents us from drawing statistically firm conclusions on any possible trend between the different samples of radio galaxies. For example, we cannot determine whether the HzLLRG's are less likely to be associated with mergers than the Hz3C's even if the observed fraction is lower in the former than in the latter. Similarly, we cannot determine whether each of the samples at z ∼ 0.5 that compose the Willott sample behave differently, for example as a result of their different radio power.

However, it is interesting to consider radio-loud samples grouped by redshift bin. We can thus compare the merger fractions for the low-z 3CR, the whole Willott sample (TOOT+7CRS+6CE+3CRR), and the high-z objects (3CR+HzLLRG). In Section 4 we showed that we find no statistical evidence that that the observed merger fractions are different for any of those groups. Therefore, while we wish to stress that this does not imply that they are the same, we can conclude that the data show no evidence for a redshift evolution. By comparing the 95% Bayesian credible intervals, we can also conclude that the merger fractions do not differ by more than ∼20%. Since the samples are well separated in radio luminosity, it is straightforward to perform a similar analysis for samples grouped by radio luminosity, and conclude that the merger fraction in low- and high-power samples does not differ by more than 20%. This is a notable result, since these samples are separated by more than 4 dex in radio power, and span a wide range of redshift (see Figure 1).

In Figure 5 we show the Bayesian 95% credible intervals for each of the groups of radio-loud objects plotted against the redshift (left panel) and radio luminosity (right panel) range spanned by each group. In the left panel (merger fraction versus redshift) we also include the radio-quiet AGNs and the two samples of non-active galaxies. The radio-quiet AGNs and the non-active galaxies are not plotted in the right panel, since any trends with the radio power would be irrelevant. In fact, the origin of radio emission in radio-quiet AGNs is still debated, as it could be either thermal or non-thermal, and a possible contribution from starbursts could not be excluded at the lowest luminosities. Starbursts are instead the most likely origin for the radio emission in non-active galaxies. These objects would lie on the bottom-left of the figure in the right panel, but any correlation with the merger fraction would be meaningless. As it is clear from the figure (left panel), the only group of radio-loud AGNs that is still marginally compatible with the radio-quiet samples is the Willott sample. However, this might be due to the fact that the images were taken with WFPC2, which was significantly less sensitive than WFC3-IR. In Table 6 we report the limit magnitude estimated using the WFPC2 ETC, and converted to the F140W filter using synphot to allow an easier comparison with the WFC3 observations. The value ${{m}_{{\rm F}140{\rm w}}}=20.7$ mag arcsec⁻² was derived using an elliptical galaxy spectrum redshifted to z = 0.5. As a result of that, some images might not show faint surface brightness structures such as e.g., asymmetries or tidal tails, which might lead us to misclassify some of the objects as non-mergers. Deeper images with WFC3 or with ACS in the I band should be taken in order to achieve a more reliable estimate of the merger fraction in the Willott sample.

Zoom In Zoom Out Reset image size

The right panel of Figure 5 shows the merger fraction for each sample, against radio power at 151 MHz. ¹³ Here we plot the Willott samples separately, since they belong to different luminosity bins. The low-z 3CR span a large range in radio power. Most of these 3CR objects are confined to the lower luminosity bin, as it is clear from Figure 1. As noted above, except for the low-z 3CR for which the merger fraction is much better constrained, the number statistics is small and the error on the merger fraction is large. However the figure clearly shows that all of the samples are located in the upper part up the diagram, and that no trend with radio power is visible.

Finally, in Figure 6 we show the merger fraction against the radio-loudness parameter R_X. In order to represent the average value of R_X for each sample, we calculated average values for both the radio and the X-ray luminosities (see the caption for references). The data are very uncertain, since the information is extremely sparse. This especially holds for the X-ray data of the Hz3C and the Willott sample, while most of the HzLLRG only have upper limits in the X-ray band, so the radio-loudness of the HzLLRGs is represented as a lower limit. Note that ∼20% of the radio-quiet objects are detected in the radio band. The value of R_x for the radio-quiet samples is thus calculated using the average radio luminosity for the detected objects, and the points in the figure are shown as upper limits for the radio-loudness parameter. For all other samples, to be safe, we assume uncertainties up to ∼ ±0.5 dex in R_X. This is reasonable considering the uncertainties in the X-ray luminosity and the range of radio and X-ray luminosity spanned by the objects.

Zoom In Zoom Out Reset image size

This plot summarizes the main result of this work, i.e., that radio-quiet sources are systematically associated to smaller merger fractions (and they are located in the lower-left side of the figure), while radio-loud AGNs are unambiguously associated with mergers (and they are located in the upper-right side of the figure).

Our results basically agree with the findings of Ramos Almeida et al. (2012) and Ramos Almeida et al. (2013). Those studies focus on samples of relatively low-redshift ($z\lt 0.7$) sources with deep ground based observations. These authors found that the large majority (∼80%) of radio-loud objects show disturbed morphologies, and they also reside in dense environments. However, the parameter space covered by those papers is limited to luminous low-redshift objects, while our work spans a significantly larger range both in redshift and luminosity. The same group also studied a sample of 20 Type 2 quasars (Bessiere et al. 2012) that includes at least two RLQSOs, based on their location in the ${{L}_{5\,{\rm GHz}}}$ versus L_{[O iii]} plane (Xu et al. 1999). As expected in light of our results, both RLQSOs are associated with mergers. The merger fraction in the full sample is as high as 75%. If we only exclude those two sources that exceed their definition of radio-loudness, the observed merger fraction among the remaining radio-quiet objects is still quite large (72%). While this might appear in disagreement with our measured fraction for the RQAGNs, the uncertainty on the value found by these authors is large, as a result of the small number statistics (13 possible mergers out of 18 objects). Since the samples are small, we cannot reject the null hypothesis that a 72% merger fraction for that sample is different from the fraction measured in our sample of RQAGNs.

One of the major points here is that we showed that there is no evidence for a trend with luminosity, at least for the radio-loud samples. This is apparently at odds with the results of Treister et al. (2012), where only the most powerful samples appear to predominantly reside in merging systems. However, it is interesting to briefly discuss their work in the light of our results. Two samples show a significantly greater merger fraction than any other samples treated in that work. Those are the sample of dust-reddened Type 1 QSOs of Urrutia et al. (2008) and the Bahcall et al. (1997) sample of Type 1 QSOs. The former is most likely biased in favor of a high merger fraction, because of the nature of those obscured quasars. The latter, instead, is more interesting. Of the 20 QSOs observed with HST by Bahcall et al. (1997), 14 are radio-quiet and 6 are radio-loud. All of the 6 RLQSOs are apparently merging or show irregular features that might be explained with a merger, in agreement with our results. Thus if we only limit the analysis to the RQQSOs, the fraction of mergers is ∼50%, in agreement with our findings. Therefore, we argue that the high fraction of mergers in the full Bahcall sample (65%) could in principle be explained, in light of our results, by the fact that a significant fraction of the objects are radio-loud.

First of all, we should point out that most mergers do not generate an RLAGN. This is clear from the fact that a fraction of non-active galaxies at $z\gt 1$ are seen to be merging, but they show no signs of radio-loud activity. If there is an association between these two phenomena, it is not a univocal cause/effect relationship. Based on our results, the same also holds for radio-quiet AGNs since a fraction of those are associated with mergers. Thus, we conclude that mergers are unrelated to radio-quiet AGNs or, alternatively, only a fraction of those may be triggered by mergers. However, it is worth noting that our definition of merger includes objects that still have to merge as well as objects for which the signs of a past merger are somehow still visible. Therefore, the timescale during which we observe a merger is probably of the order of at least ∼1–2 Gyr (e.g, Di Matteo et al. 2005). The time scale for radio-loud activity is most likely significantly shorter ($\sim {{10}^{7}}-{{10}^{8}}$ years). Therefore, we cannot exclude that some of the non-active galaxies that we observe in a merger phase are turned-off radio-loud AGNs, or, alternatively, they still have to be turned-on. Summarizing, we believe that not all mergers may directly generate a radio-loud AGN. Below we discuss a few possible conditions for a merger to trigger RLAGN activity.

We argue that when certain conditions are met, mergers may trigger radio-loud nuclear activity. What we ultimately wish to know is what those conditions are. An important piece of information here is that the association between mergers and RLAGNs is robustly established at all redshifts. While both recent simulations and observations show that the galaxies merger rate increases with redshift (e.g., Conselice et al. 2003; Hopkins et al. 2006; Lotz et al. 2011; Rodriguez-Gomez et al. 2015), our results show that there is no evidence that the merger fraction for RLAGNs is higher at higher z. Of course, this should be confirmed through the analysis of larger samples that may better constrain the merger fractions and highlight any possible trends with redshift. But even if the uncertainties remain large with the present samples, the existence of such a tight relationship between the two phenomena means that somehow an RLAGN needs a merger (at all redshifts) in order to manifest itself. The predominance of major mergers among our radio-loud samples (at least at $z\gt 1$) that is apparent from visual inspection needs to be confirmed through a more quantitative analysis. However, this may imply that one of the conditions that must be met in order to trigger an RLAGN is that the merger needs to be between two galaxies (and thus between two BHs) of similar mass. It is well known that radio galaxies are ubiquitously associated with very massive hosts ($\gtrsim {{10}^{10-11}}\;{{M}_{\odot }}$, e.g., Best et al. 2005) and high-mass SMBH ($\gtrsim {{10}^{8}}$ ${{M}_{\odot }}$, Laor 2000; Dunlop et al. 2003; Chiaberge & Marconi 2011; Best & Heckman 2012). Therefore, we expect those major mergers to involve high-mass objects only.

Understanding the details of this issue is central to our future work on the subject, but it is beyond the scope of this paper. However, we can build on our results and speculate on a possible scenario. One of the effects of mergers is to lower the specific angular momentum of the gas in the galaxy, and thus to drive the gas toward the center (e.g., Hernquist 1989; Hopkins & Quataert 2011). While this may naturally happen in gas-rich mergers, as those observed in our high-z 3CR sample (e.g., Barthel et al. 2012; Tadhunter et al. 2014; Podigachoski et al. 2015), the hosts of low-redshift radio galaxies (in particular those of low radio powers) are often relatively gas–poor systems. Therefore, this effect may play a role at higher redshifts, but it is very unlikely to be the ultimate cause of radio-loud activity in general. Furthermore, tidal effects happen in all mergers, thus merger events should affect both RQ and RL AGNs at the same level, contrary to our results.

Another effect of mergers is to alter the spin and the mass of the central BH. That can be achieved in two ways, i.e., either via accretion, or via BH–BH merger (e.g., Volonteri et al. 2013). The former implies that a large amount of gas is driven toward the central region of the galaxy, for a significant amount of time. If the accretion of matter is coherent, i.e., if the flow of matter occurs at a fixed angular momentum axis, that ultimate leads to spinning-up the BH. On the other hand, if accretion results from multiple merger events that drive the matter toward the BH from different directions, the BH is spinned-down. In any case, as already pointed out above, the expected amount of gas in major mergers (at least at low redshifts) is probably too small to alter the BH spin significantly in these objects.

In the case of BH–BH merger, the two BH coalesce and the resulting object has increased mass and, in most cases, a lower spin per unit mass. But there are scenarios in which those events lead to the opposite result. For example, it has been shown that a single major BH–BH merger, where the ratio between the masses of the two involved BHs approaches unity, may generate a spinning BH, even if the two merging BH are initially not spinning (Hughes & Blandford 2003; Baker et al. 2006; Li et al. 2010; Giacomazzo et al. 2012; Schnittman 2013, for a recent review). However, current simulations are unable to reproduce BH with dimensionless spin parameter greater than ∼0.94 as a result of BH–BH mergers alone (Hemberger et al. 2013).

According to the so-called spin paradigm (Blandford et al. 1990, p. 97; Wilson & Colbert 1995), radio-loud AGNs are associated with rapidly spinning BHs, while BHs in radio-quiet AGNs are expected to spin less rapidly. Those major mergers that result in rapidly spinning BHs may provide the link between our observations and the physics behind the RLAGNs phenomenon as a whole. Clearly, the objects that we are seeing in a pre-merger phase, or the very few ones that are associated with a minor merger, do not fit the above scheme. In those cases, we must assume that another major merger event happened in the recent past. This does not seem unreasonable, since all of these objects lie in over-dense environments, but it should be proven by finding the signatures of that previous merger.

In a framework in which BHs are spun up by major BH–BH mergers, we expect a range of resulting spin values. Therefore, it is possible that different radio morphologies (and radio powers) are associated with different BH spin levels. For example, only the BHs that spin more rapidly might be able to produce the most relativistic jets, in a framework in which the jet is powered by energy extracted from the rotating BH (e.g., Blandford & Znajek 1977; McKinney et al. 2012; Ghisellini et al. 2014). Such a scenario has been explored for X-ray binaries jets, so far leading to contrasting results (e.g., Fender et al. 2010; Narayan & McClintock 2012; Russell et al. 2013; Gardner & Done 2014). It is interesting to note that Fanidakis et al. (2011) were able to reproduce the RL and RQ AGN populations in the context of galaxy evolution. Their model is based on a scheme that includes a bimodal BH spin distribution mainly caused by the the combination of accretion of matter onto the BH and by the different type of mergers that galaxies of different stellar mass undergo.

As pointed out in the previous section, BHs could also be spun-up by merger-triggered accretion. Such a mechanism could in principle account for our observations of mergers at $z\gt 1$, but only if those are gas-rich mergers. However, it would not completely explain the origin of radio-loud activity in low redshift, gas-poor galaxies (see also the discussion in the next section).

A scenario in which BH–BH mergers are directly implied in triggering radio-loud AGN activity seems to be supported by the observed properties of low-redshift RLAGN hosts. Capetti & Balmaverde (2006) and de Ruiter et al. (2005) showed that all radio-loud AGNs in their samples are hosted by core-galaxies, i.e., galaxies that show a flat radial brightness profile in the central regions. While this analysis is clearly limited to very low redshifts (z ≲ 0.1), where the core radius can be resolved in HST images, and to objects that are not affected by significant amounts of dust in the central kiloparsecs, it very clearly shows that there is a strong connection between the presence of a core profile and RL activity. Interestingly, one of the most likely explanations for the presence of core profiles is related to a major BH–BH mergers, in which the binary BH formed during the merger ejects stars from the central regions before the two BHs coalesce (e.g., Graham 2004; Merritt 2006). While the direct physical connection between such a phenomenon and radio-loud activity is still a matter of debate (see e.g., Chiaberge & Marconi 2011), it provides one more piece of evidence that major mergers and radio-loud AGNs are strictly connected.