Radio galaxies in the 2SLAQ Luminous Red Galaxy Survey: I. The evolution of low-power radio galaxies to z~0.7

Abstract

We have combined optical data from the 2dF-SDSS Luminous Red Galaxy and QSO (2SLAQ) redshift survey with radio measurements from the 1.4 GHz VLA FIRST and NVSS surveys to identify a volume-limited sample of 391 radio galaxies at redshift 0.4<z<0.7. By determining an accurate radio luminosity function for early-type galaxies in this redshift range, we can investigate the cosmic evolution of the radio-galaxy population over a wide range in radio luminosity. The low-power radio galaxies in our LRG sample (those with 1.4 GHz radio luminosities in the range 10^{24} to 10^{25} W/Hz, corresponding to FR I radio galaxies in the local universe) undergo significant cosmic evolution over the redshift range 0<z<0.7, consistent with pure luminosity evolution of the form (1+z)^k where k=2.0+/-0.3. Our results appear to rule out (at the 6-7 sigma level) models in which low-power radio galaxies undergo no cosmic evolution. The most powerful radio galaxies in our sample (with radio luminosities above 10^{26} W/Hz) may undergo more rapid evolution over the same redshift range. The evolution seen in the low-power radio-galaxy population implies that the total energy input into massive early-type galaxies from AGN heating increases with redshift, and was roughly 50% higher at z~0.55 (the median redshift of the 2SLAQ LRG sample) than in the local universe. Comment: 18 pages, 15 figures, one 10-page data table in landscape format. Replaced with final version accepted for publication in MNRAS

Full text

Page 1

arXiv:astro-ph/0612019v2 11 Sep 2007
Mon. Not. R. Astron. Soc. 000, 1–?? (2006)Printed 5 February 2008(MN LATEX style file v1.4)
Radio galaxies in the 2SLAQ Luminous Red Galaxy
Survey: I. The evolution of low–power radio galaxies to
z ∼ 0.7
Elaine M. Sadler1, Russell D. Cannon2, Tom Mauch1, Paul J. Hancock1,
David A. Wake3, Nic Ross3, Scott M. Croom1, Michael J. Drinkwater4,
Alastair C. Edge3, Daniel Eisenstein5, Andrew M. Hopkins1, Helen M.
Johnston1, Robert Nichol6, Kevin A. Pimbblet4, Roberto De Propris7, Isaac
G. Roseboom4, Donald P. Schneider8, Tom Shanks3
1School of Physics, University of Sydney, NSW 2006, Australia
2Anglo-Australian Observatory, P.O. Box 296, Epping, NSW 2121, Australia
3Department of Physics, University of Durham, South Road, Durham DH1 3LE
4Department of Physics, University of Queensland, Brisbane, QLD 4072, Australia
5Steward Observatory, 933 N. Cherry Ave, Tucson, AZ 85721, USA
6Institute of Cosmology and Gravitation, University of Portsmouth, Portsmouth, PO1 2EG
7Cerro Tololo Inter-American Observatory, Casilla 63-D, La Serena, Chile
8Department of Astronomy and Astrophysics, Pennsylvania State University, 525 Davey Laboratory, University Park, PA 16802, USA
5 February 2008
ABSTRACT
We have combined optical data from the 2dF-SDSS Luminous Red Galaxy and QSO
(2SLAQ) redshift survey with radio measurements from the 1.4GHz VLA FIRST and
NVSS surveys to identify a volume–limited sample of 391 radio galaxies at redshift
0.4 < z < 0.7. By determining an accurate radio luminosity function for luminous
early–type galaxies in this redshift range, we can investigate the cosmic evolution of
the radio–galaxy population over a wide range in radio luminosity.
The low–power radio galaxies in our LRG sample (those with 1.4GHz radio lumi-
nosities in the range 1024to 1025WHz−1, corresponding to FRI radio galaxies in the
local universe) undergo significant cosmic evolution over the redshift range 0 < z < 0.7,
consistent with pure luminosity evolution of the form (1+z)k, where k = 2.0 ± 0.3.
Our results appear to rule out (at the 6–7σ level) models in which low–power radio
galaxies undergo no cosmic evolution. The most powerful radio galaxies in our sample
(with radio luminosities above 1026WHz−1) may undergo more rapid evolution over
the same redshift range.
The evolution seen in the low–power radio-galaxy population implies that the to-
tal energy input into massive early–type galaxies from AGN heating increases with
redshift, and was at least 50% higher at z ∼ 0.55 (the median redshift of the 2SLAQ
LRG sample) than in the local universe.
Key words: galaxies: radio continuum — galaxies: luminosity function — galaxies:
active — AGN: evolution
1INTRODUCTION
The strong cosmic evolution of the most powerful radio
galaxies was deduced more than forty years ago from radio
source counts (Longair 1966), which imply that the space
density of powerful radio galaxies at redshift z ∼ 2 was
roughly a thousand times higher than in the local universe
(Doroshkevich, Longair & Zeldovich 1970, Dunlop & Pea-
cock 1990).
Far less is known about the cosmic evolution of the
lower–power radio galaxies which comprise the overwhelm-
ing majority of the local radio AGN population. Relatively
few low–power radio galaxies have been observed at redshifts
c ? 2006 RAS

Page 2

2
Sadler et al.
beyond z ∼ 0.3, partly because classical flux–limited radio
surveys like the Cambridge 3CR (Laing, Riley & Longair
1983) and Molonglo MRC (Large et al. 1981) sample only a
narrow range in radio luminosity at any given redshift (see
e.g. Blundell et al. 2002); but the observed distribution of
radio source counts implies that low–power radio sources
cannot evolve as rapidly as the most powerful sources do
(Longair 1966).
This finding led to the development of two classes of
model for the cosmic evolution of radio–loud active galax-
ies: single-population models, in which the rate of evolution
varies with radio power (Dunlop & Peacock 1990), and dual–
population models, in which the radio–source population is
assumed to be made up of a low–luminosity non–evolving
component and a high–luminosity rapidly–evolving compo-
nent (Jackson & Wall 1999). In the dual–population mod-
els, the two populations have been variously been identified
with FRI and FRII radio galaxies⋆(Wall 1980; later ex-
panded by Jackson & Wall 1999 to include BL Lac objects
and flat–spectrum QSOs as the beamed counterparts of FRI
and FRII objects respectively); or with objects having weak
and strong optical emission lines, independent of their radio
morphology (Willott et al. 2001).
Several recent studies imply that low–power radio
galaxies (i.e. those with 1.4GHz radio luminosities near or
below the FRI/FRII divide at ∼ 1026WHz−1) undergo lit-
tle or no cosmic evolution. In the model of Jackson & Wall
(1999), which is consistent with both radio source counts
and the observed redshift distribution of 3CR sources, there
is strong cosmic evolution of FRII radio galaxies but no evo-
lution of the FRI population. Clewley & Jarvis (2004) also
found no increase in the comoving density of radio sources
with luminosities below P325 MHz ∼ 1025WHz−1over the
redshift range 0 < z < 0.8.
In contrast, Brown, Webster & Boyle (2001) found sig-
nificant luminosity evolution (of the form (1 + z)k, where
3 < k < 5) for a sample of low–power radio galaxies over
the redshift range 0 < z < 0.5. Snellen & Best (2001) found
two distant FRI radio galaxies in the small area of sky cov-
ered by the Hubble Deep Field; from which they argue that
FRI radio galaxies must be significantly more abundant at
z > 1 than in the nearby universe. Willott et al. (2001) also
suggest that the comoving space density of low–luminosity
radio galaxies rises by about 1 dex between z ∼ 0 and ∼ 1.
The tight relation between the mass of luminous early–
type†galaxies and the mass of their central supermassive
⋆Fanaroff & Riley (1974) divided radio galaxies into two classes
based on the observed morphology of their radio emission and
showed that this morphology correlates with radio luminosity,
with the less luminous (FRI) objects having a jet-like appearance
and the more luminous (FRII) objects having edge-brightened ra-
dio hotspots. Later work (Bicknell 1995) showed that the classifi-
cation appears to have a physical basis, since FRII radio galaxies
have jets which remain relativistic over scales of tens to hundreds
of kiloparsec, whereas FRI radio jets rapidly decelerate to sub-
relativistic velocities. As a result, models of radio-galaxy evolution
often treat FRI and FRII radio galaxies as separate populations
which may evolve in different ways.
†Throughout this paper, we use the term “early–type galaxy”
to refer to both luminous, passively–evolving distant galaxies
(LRGs) and giant E/S0 galaxies in the local universe.
black holes (Magorrian et al. 1998) implies that the evolu-
tion of galaxies and their central black holes are intimately
related. There is also increasing evidence that radio jets can
regulate and prevent star formation in luminous early–type
galaxies by heating the interstellar gas and stopping the on-
set of cooling flows (Binney & Tabor 1995; Rawlings & Jarvis
2004; Birzan et al. 2004; Springel, Di Matteo & Hernquist
2005). Since most radio galaxies have radio luminosities well
below the FRI/FRII break for most of their lifetimes, im-
proving our understanding of the cosmic evolution of these
lower–power radio galaxies is an essential first step in under-
standing their effects on the star–formation history of mas-
sive galaxies. This is particularly important in the context
of recent semi–analytic models of galaxy evolution which in-
corporate AGN heating (e.g. Croton et al. 2006; Bower et
al. 2006), as we discuss in §6.4 of this paper.
Previous studies in this area have relied strongly on the
analysis of radio source–counts and/or on the use of pho-
tometric redshifts to derive distances. However, the most
direct and accurate way to measure the cosmic evolution
of low–power radio galaxies is to compare the radio lumi-
nosity function (RLF) observed at different redshifts. This
requires a large volume–limited (rather than flux–limited)
sample of radio sources which is reasonably complete over
a wide range in radio power. Such samples can now be as-
sembled by combining a large optical redshift survey with a
sensitive, large–area radio continuum survey; this technique
has been used to measure accurate RLFs for galaxies in the
local universe (Sadler et al. 2002; Best et al. 2005a,b; Mauch
& Sadler 2006). In this paper, we extend the same technique
to higher redshift (0.4 < z < 0.8) by using optical spectra
from the 2SLAQ LRG survey (Cannon et al. 2006) and radio
data from the VLA FIRST (Becker et al. 1995) and NVSS
(Condon et al. 1998) surveys.
Throughout this paper, we use H0 = 70 km s−1Mpc−1,
Ωm = 0.3 and ΩΛ = 0.7.
2THE 2SLAQ LRG SURVEY
During 2003–5, the 2SLAQ (2dF–SDSS LRG And QSO) sur-
vey used the 2dF multi–object spectrograph on the Anglo–
Australian Telescope to obtain optical spectra of over 11,000
luminous red galaxies (LRGs) with i-band magnitude <19.8
and redshifts in the range 0.4 < z < 0.8. Full details of the
source selection, sample properties and spectroscopic obser-
vations are discussed by Cannon et al. (2006), and only a
brief outline is given here.
The total co-moving volume sampled by the 2SLAQ
LRG survey out to the median redshift of z=0.55 is just
over 108Mpc3, i.e. a larger volume than that sampled by the
2dF Galaxy Redshift Survey (2dFGRS; Colless et al. 2001),
which covered a larger area of sky to a shallower redshift
limit (up to z ∼ 0.3 for the most luminous galaxies).
2.1Colour selection of 2SLAQ LRGs
The 2SLAQ LRG sample was selected using ugriz (Fukugita
et al. 1996) photometry from the Sloan Digital Sky Survey
(SDSS; York et al. 2000) in two narrow strips along the
celestial equator covering a total area of about 180deg2.
c ? 2006 RAS, MNRAS 000, 1–??

Page 3

Radio–galaxy evolution to z = 0.73
Figure 1. Photometric selection criteria for the 2SLAQ LRG
sample. Small dots show the full ∼15,000-galaxy spectroscopic
sample and large filled circles the z > 0.4 radio–detected LRGs
discussed in §3. The regions occupied by Samples 8 and 9 (as
defined by Cannon et al. 2006) are also labelled, along with the
colour–cut lines cparand dperpused to select them. A small num-
ber of 2SLAQ galaxies belong to other photometric samples, and
fall outside the area covered by Samples 8 and 9.
The colours used for selection were the SDSS extinction-
corrected modelMag colours as described by Stoughton et
al. (2002), and the colour selection criteria were similar to
those used by Eisenstein et al. (2001) to select SDSS LRGs,
modified as described by Cannon et al. (2006) to select tar-
get objects in the redshift range 0.4 < z < 0.8. A (dered-
dened) cutoff magnitude of i=19.8 was applied along with
the cuts in g − r and r − i colour.
Figure 1 shows the colour cuts used to select target
2SLAQ LRGs for the spectroscopic survey. As discussed in
detail by Cannon et al. (2006), 0.4 < z < 0.7 LRGs whose
light is dominated by an old, passively–evolving stellar pop-
ulation are expected to lie along a vertical track in this di-
agram, with g − r ≃ 1.7. In this redshift range, the r − i
colour of early–type galaxies becomes rapidly redder with
increasing z as the 4000˚ A break moves through the band,
whereas the g − r colour remains almost constant.
The dashed line sloping downward from left to right in
Figure 1, defined by setting a constant value of 1.6 for the
quantity
cpar = 0.7(g − r) + 1.2(r − i − 0.18),
is used to separate LRGs from star–forming galaxies, which
lie to the left of this line at 0.4 < z < 0.7. The lines sloping
upwards from left to right in Figure 1, defined by constant
values of
dperp = (r − i) − [(g − r)/8.0]
select early–type galaxies at increasingly high redshift for
larger values of dperp (Eisenstein et al. 2001; Cannon et al.
2006)
Two main samples of LRGs were observed. The pri-
mary sample (Sample 8) consists of objects with z > 0.45 at
a surface density such that most targets can be accessed in
a single pass. This sample has over 90% completeness both
spatially and in terms of redshift reliability. A secondary
sample (Sample 9) contains mostly lower-redshift objects
with z ∼ 0.45; it is photometrically homogeneous and has
high redshift completeness but very variable spatial cover-
age.
The photometric selection technique worked very suc-
cessfully; over 90% of the objects observed were LRGs in the
target redshift range. As discussed by Wake et al. (2006),
these galaxies lie well above the ‘knee’ in the optical lumi-
nosity function and have r–band luminosities in the range
2–15L∗.
2.2 2dF spectroscopy
All the 2SLAQ spectra were obtained with the 2dF fibre
spectrograph (Lewis et al. 2002), and the observing and re-
duction techniques are described in detail by Cannon et al.
(2006). The wavelength coverage of the 2dF spectra was
typically 5000–7250˚ A, as shown in Figure 2. In the redshift
range targeted by the 2SLAQ survey, the main features seen
in the 2dF spectra are the CaII H and K absorption lines
and 4000˚ A break.
For 2SLAQ galaxies with redshifts above z ∼ 0.45, the
Hβ line is shifted out of the 2dF wavelength range and the
only observable strong emission line is [OII] 3727˚ A, which
is seen in the spectra of just over 25% of the radio–detected
2SLAQ LRGs. Higher–excitation emission lines of [NeV]
3426˚ A and [NeIII] 3870˚ A are also detected in a few 2SLAQ
LRGs (like J092203.20−004443.5 in Figure 2), as discussed
later in §4.3. The great majority (> 70%) of radio–detected
2SLAQ LRGs show absorption–line spectra like the galaxy
in Figure 2(a), and only a handful have an e+A spectrum
like that shown in 2(d).
3RADIO SOURCE IDENTIFICATIONS
We adopted a two–step approach similar to that described
by Best et al. (2005a) to identify radio sources associated
with the 2SLAQ LRGs, using radio–source catalogues from
both the VLA FIRST survey (Becker et al. 1995) and the
NVSS (Condon et al. 1998)
The FIRST and NVSS radio surveys have complemen-
tary properties. NVSS, with a 45arcsec beam, accurately
samples the total flux density of extended radio sources.
FIRST, with a 5arcsec beam, has higher spatial resolution
but at the expense of resolving out extended radio emission
on scales larger than a few arcseconds (and hence under-
estimating the total flux density of galaxies with extended
radio components). We therefore use the FIRST positions to
identify 2SLAQ radio galaxies and the NVSS flux densities
to calculate the radio luminosity function.
3.1 Background
The surface density of bright galaxies (B<19.5mag.) is
low enough that reliable radio identifications can usually
be made from the NVSS survey alone (see e.g. Sadler et
c ? 2006 RAS, MNRAS 000, 1–??

Page 4

4
Sadler et al.
Figure 2. Sample 2dF spectra of some 2SLAQ radio galaxies from Table 3. J144541.59−003409.6 at z = 0.5360 (a) is an absorption–line
galaxy at about the median redshift of the survey and is typical of the majority of 2SLAQ radio galaxies, which show no obvious optical
emission lines. J143000.12+001555.2 at z = 0.4343 (b) is a lower–redshift galaxy with a weak [OII] emission line. J225439.77−001501.2
(c) is an example of an emission–line galaxy with strong lines of [OII], [NeIII] and [NeV] and weaker Balmer emission lines of Hγ
and Hδ, while J093142.54−000306.1 (d) shows the strong Hδ absorption feature characteristic of post–starburst ‘e+A’ galaxies. The
effective spectral resolution is about 5˚ A.
al. 2002, Mauch & Sadler 2006). For fainter objects, how-
ever, more accurate radio positions are needed. Best et al.
(2005a) have developed a multi-stage method using infor-
mation from both FIRST and NVSS to produce a sample of
radio–source IDs with high completeness and reliability, and
have used this to identify the radio counterparts of galaxies
(14.5 < r < 17.8mag.) in the second data release of the
SDSS.
Experience shows that the measured radio luminosity
function of volume–limited galaxy samples such as the SDSS
and 2dFGRS is very robust against small changes in the way
the sample is selected. For example, the RLFs measured by
Best et al. (2005a) for local AGN and star–forming galaxies
agree well with those measured by Sadler et al. (2002) and
Mauch & Sadler (2006) even though slightly different radio
identification criteria were used in all three investigations.
In this study we use a similar approach to that of Best
et al. (2005a), though with some modifications as described
below. In particular, we check all our candidate FIRST radio
IDs visually on radio–optical overlay plots. This matches
the procedure used by Mauch & Sadler (2006) to identify a
large sample of radio galaxies from the 6dF Galaxy Survey
(Jones et al. 2004), which we use here as the local benchmark
to measure the redshift evolution of the radio luminosity
function.
We carried out the radio–source identification in three
stages:
• Early in the project, we performed a series of tests on
the 2003 version of the 2SLAQ LRG input photometric cat-
alogue. This allowed us to estimate both the reliability of
the final radio sample and the expected number of extended
double and multiple radio sources, as described in §3.2.1.
• Once the 2SLAQ spectroscopic observations were com-
plete, we cross-matched the final spectroscopic catalogue
with the FIRST survey. This allowed us to identify a sample
of 367 2SLAQ LRGs associated with FIRST radio sources
(see §3.3.1).
• We also cross-matched the 2SLAQ LRG spectroscopic
catalogue with the NVSS catalogue as discussed in §3.3.2.
This allowed us to identify six very extended radio sources
with angular sizes greater than 1arcmin, as well as 22 weak
radio sources which were not listed in the FIRST catalogue.
The NVSS catalogue provides accurate total flux densities
for most of the sources identified in the FIRST catalogue.
c ? 2006 RAS, MNRAS 000, 1–??

Page 5

Radio–galaxy evolution to z = 0.75
Figure 3. Number of candidate FIRST radio detections of
2SLAQ LRGs (in successive 1-arcsec annuli) plotted against the
offset ∆ between the radio and optical positions. Filled circles
represent candidate radio IDs from the 2SLAQ photometric cat-
alogue. Open circles show matches found to positions randomly
offset from the 2SLAQ objects, and allow us to estimate the opti-
mal search radius for radio IDs. The dashed line shows the num-
ber of chance coincidences expected if the FIRST radio sources
are uniformly distributed on the sky with a surface density of
90deg−2.
3.2Tests using the 2SLAQ LRG photometric
catalogue
3.2.1Radio–source identification
Of the 70,582 galaxies listed in the 2003 2SLAQ LRG input
photometric catalogue, 60,290 (85%) lay within the region of
sky covered by the FIRST radio catalogue. We checked for
possible matches of these galaxies with FIRST radio sources,
using a 30arcsec maximum offset between the radio and op-
tical positions (because both our own tests and the study
by Best et al. (2005a) showed that only a handful of gen-
uine radio IDs are expected at larger separations). For the
redshift range covered by the 2SLAQ LRGs (0.4 < z < 0.8),
30arcsec corresponds to a projected linear distance from the
optical galaxy of 160–220 kpc. The few radio galaxies with
radio lobes more distant than this are best identified in the
lower-resolution NVSS images as discussed below.
A total of 2,782 of the LRGs in the input catalogue (4%)
had one or more FIRST radio sources within 30arcsec of the
optical position, and so were candidate radio galaxies. We
repeated the matching process using a set of ‘random’ po-
sitions offset by 10arcmin from the position of each galaxy
in the 2SLAQ input catalogue. This Monte Carlo test al-
lows estimation of the number of unrelated foreground or
background radio sources which are seen by chance. Figure
3 plots the number of sources seen in 2SLAQ and random
fields as a function of the offset between radio and optical po-
sitions. Note that the vertical axis is logarithmic, so the ex-
cess of sources at separations ≤ 3arcsec is very large. As ex-
pected, the number of matched sources approaches the value
expected by chance for offsets larger than about 20arcsec.
The population of FIRST radio sources within 30arcsec
of a 2SLAQ LRG will be a mixture of the following:
(a) Single radio sources which are genuinely associated with
a 2SLAQ galaxy.
(b) Components of double or triple radio sources which are
genuinely associated with a 2SLAQ galaxy.
(c) Unrelated foreground or background radio sources. The
likely numbers of such objects can be estimated from Monte
Carlo tests.
(d) Components of single, double or triple radio sources as-
sociated with neighbouring galaxies at the same redshift as
the 2SLAQ galaxy.
For each 2SLAQ LRG with a candidate radio source
within 30arcsec, we overlaid radio contours from the FIRST
survey onto greyscale optical images (taken mainly from the
SDSS DR3, with a small number from the SuperCOSMOS
images). These overlays were then inspected by at least two
team members, who flagged each candidate source as ‘ac-
cept’ or ‘reject’ based on the following guidelines:
(i) All sources less than 3.0arcsec from a 2SLAQ galaxy were
accepted as genuine IDs.
(ii) A single FIRST source separated by 3–10arcsec from
a 2SLAQ galaxy is accepted as an ID if (a) it is spatially
resolved and extended in the direction of the optical galaxy,
(b) the separation is no larger than the projected major axis
of the radio source, and (c) no other optical object closer to
the radio position is visible on the overlay images.
(iii) Two FIRST components of similar flux density are ac-
cepted as IDs if the optical galaxy lies within 5.0arcsec of
the radio centroid (unless another optical object is closer).
(iv) Where three or more FIRST components are present,
a decision on whether each is associated with the optical
galaxy is based on visual inspection alone.
3.2.2 The effects of clustering
Of the 2,871 FIRST matches with the 2SLAQ input pho-
tometric catalogue, 1602 (56%) were classified as genuine
associations with 1362 2SLAQ galaxies. This corresponds to
a radio detection rate of 2.3% for the 2003 2SLAQ input
catalogue. Of the accepted radio galaxies, 87% had a single
FIRST component, 10% were doubles and 3% were resolved
into three or more FIRST components. Our Monte Carlo
tests imply that the excess of ‘real’ over ‘random’ sources
in this sample should be roughly 1750±40. This is signifi-
cantly higher than the 1602 sources we accepted as genuine
associations with 2SLAQ galaxies, and at first glance might
suggest that we have failed to recognize up to 150 genuine
matches of FIRST radio sources with galaxies in the 2SLAQ
input catalogue.
However, as noted by Best et al. (2005a), Monte Carlo
tests will not give a reliable estimate of completeness if the
2SLAQ galaxies are strongly clustered. If the overall space
density of galaxies is higher in the vicinity of a 2SLAQ LRG,
then the probability of finding a radio source within 30arcsec
of the LRG will also be higher. For the 2003 2SLAQ in-
put catalogue, we have a statistical excess of 148±40 ra-
dio sources which are not identified with 2SLAQ galaxies
but also cannot be explained by chance associations of fore-
ground or background objects. If these excess radio sources
c ? 2006 RAS, MNRAS 000, 1–??