Galaxy And Mass Assembly: The 1.4GHz SFR indicator, SFR-M* relation and predictions for ASKAP-GAMA

0.98 MB·

by  L. J. M. Davies

MNRAS000,1–14(2016) Preprint24January2017 CompiledusingMNRASLATEXstylefilev3.0
Galaxy And Mass Assembly: The 1.4 GHz SFR indicator,
SFR-M relation and predictions for ASKAP-GAMA
∗
L. J. M. Davies1(cid:63), M. T. Huynh1, A. M. Hopkins2, N. Seymour3, S. P. Driver1,4,
A. G. R. Robotham1, I. K. Baldry5, J. Bland-Hawthorn6, N. Bourne7, M. N. Bremer8,
M. J. I. Brown9, S. Brough2, M. Cluver10, M. W. Grootes11, M. Jarvis12,13, J. Loveday14,
A. Moffet1, M. Owers2,15, S. Phillipps, E. Sadler6, L. Wang16,17, S. Wilkins14,
7
1 A. Wright1
0
2
1 ICRAR, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia
n 2 Australian Astronomical Observatory, P.O. Box 915, North Ryde, NSW 1670, Australia
a 3 ICRAR, Curtin University, Bentley, WA 6102, Australia
J 4 SUPA, School of Physics and Astronomy, University of St Andrews, North Haugh, St Andrews, Fife, KY16 9SS, UK

3 5 Astrophysics Research Institute, Liverpool John Moores University, IC2, Liverpool Science Park, 146 Brownlow Hill, Liverpool, L3 5RF, UK
2 6 Sydney Institute for Astronomy, School of Physics A28, University of Sydney, NSW 2006, Australia.
7 Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK

] 8 Astrophysics Group, School of Physics, University of Bristol, Tyndall Avenue, Bristol BS8 1TL
A 9 School of Physics and Astronomy, Monash University, Clayton, Victoria 3800, Australia
G 10 Department of Physics and Astronomy, University of the Western Cape, Robert Sobukwe Road, Bellville, 7535, South Africa
11 ESA/ESTEC, 2201 AZ Noordwijk, The Netherlands
.
h 12 Astrophysics, University of Oxford, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, UK
p 13 Department of Physics, University of the Western Cape, Bellville 7535, South Africa
- 14 Astronomy Centre, University of Sussex, Falmer, Brighton BN1 9QH, UK
o 15 Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia
r 16 SRON Netherlands Institute for Space Research, Landleven 12, 9747 AD, Groningen, The Netherlands
t
s 17 Kapteyn Astronomical Institute, University of Groningen, Postbus 800, 9700 AV Groningen, the Netherlands
a
[

1 AcceptedXXX.ReceivedYYY;inoriginalformZZZ
v
2
4
ABSTRACT
2
6
We present a robust calibration of the 1.4GHz radio continuum star formation
0
rate (SFR) using a combination of the Galaxy And Mass Assembly (GAMA) survey
.
1 and the Faint Images of the Radio Sky at Twenty-cm (FIRST) survey. We identify
0 individually detected 1.4GHz GAMA-FIRST sources and use a late-type, non-AGN,
7 volume-limited sample from GAMA to produce stellar mass-selected samples. The
1
latter are then combined to produce FIRST-stacked images. This extends the robust
:
v parametrisation of the 1.4GHz-SFR relation to faint luminosities. For both the indi-
i vidually detected galaxies and our stacked samples, we compare 1.4GHz luminosity
X
to SFRs derived from GAMA to determine a new 1.4GHz luminosity-to-SFR relation
r withwellconstrainedslopeandnormalisation.Forthefirsttime,weproducetheradio
a
SFR-M relation over 2decades in stellar mass, and find that our new calibration is
∗
robust,andproducesaSFR-M relationwhichisconsistentwithallotherGAMASFR
∗
methods. Finally, using our new 1.4GHz luminosity-to-SFR calibration we make pre-
dictionsforthenumberofstar-formingGAMAsourceswhicharelikelytobedetected
in the upcoming ASKAP surveys, EMU and DINGO.
Key words: galaxies: star-formation - galaxies: evolution - radiation mechanisms:
non-thermal - radio continuum: galaxies
1 INTRODUCTION
(cid:63) E-mail:[email protected] The rate at which galaxies are forming new stars (the star
formation rate, SFR) is critical to our understanding of the
(cid:13)c 2016TheAuthors
2 L. J. M. Davies et. al.
formation of stellar mass in galaxies and the global evo- ferent spectral slope to non-thermal emission (e.g. Condon
lution of baryonic matter in the Universe. However, accu- 1992),andinthisworkweassumethatthatthermalcontri-
rately measuring SFRs is problematic. This is largely due butionat1.4GHzisnegligible(asfoundforthemajorityof
tothefactthatcommonmethodsforderivingSFRarelim- local star-forming galaxies, Rabidoux et al. 2014).
ited by either dust obscuration (e.g. see Meurer, Heckman, Inordertorobustlyuse1.4GHzemissiontoprobedust-
&Calzetti1999)and/oraperturecorrectionstoaccountfor unbiased star-formation, we require the observed 1.4GHz
missing flux in fibre-based spectroscopy (for example using radio power to be well calibrated against reliable measures
theHαemissionlinetoderiveSFRs,e.g.Hopkinsetal.2013; ofstar-formationusingothertracers.Therearetwodifferent
Gunawardhana et al. 2013). approaches to perform such a calibration: (i) using detailed
Apotentiallymorerobustapproachistomeasureboth observations of well studied nearby galaxies in the high sig-
the UV and total IR emission simultaneously (UV+TIR or nal to noise regime, primarily with dedicated observations
full spectral energy distribution (SED)-derived SFRs e.g. (e.g.Kennicuttetal.2009;Kennicutt&Evans2012;Heesen
Bell et al. 2005; Papovich et al. 2007; Barro et al. 2011), et al. 2014), and (ii) the statistical approach of identifying
probingboththedustobscuredandunobscuredSFRs.This multiple faint sources in large area surveys (e.g. Bell 2003;
approach does not require obscuration corrections, as one Hopkinsetal.2003).Untilrecently,thelatterapproachhas
completely observes the full (direct and reprocessed) emis- relied on SFR tracers that require dust obscuration and/or
sion from young stars. The number of sources with robust aperturecorrectionstocalibratethe1.4GHzSFRindicator
UV+TIRmeasurementshowever,hashistoricallybeenvery and have been limited to relatively high radio luminosity
small,hamperingeffortstoanalyselargesamplesofgalaxies systems. With the new full SED and well calibrated SFR
using this method. measuresfromsurveyssuchastheGalaxyAndMassAssem-
Recently great strides have been made in improving bly (GAMA Driver et al. 2011; Liske et al. 2015) as in D16
techniquestoderiverobustSFRsforlargesamplesofgalax- however,wecanbegintoexplorethe1.4GHzSFRindicator
ies (see Davies et al. 2016, hereafter D16). Complex pre- withouttheneedforcomplexcorrectionsandtosignificantly
scriptions for the treatment of obscuration corrections in lowerradioluminosities.InthisworkweutilisetheUV+TIR
theUV,suchasusingradiativetransfer(RT)models(Tuffs and magphys-derived SFRs from D16 (these are described
et al. 2004; Wood et al. 2008; Popescu et al. 2011; Popescu brieflyinSection4)toproduceanew1.4GHzluminosityto
&Tuffs2013;Grootesetal.2013,2014,Grootesinprepand SFR calibration and use stacking techniques to extend the
D16), have dramatically improved our ability to reduce the 1.4GHz luminosity - SFR relation to faint luminosities.
scatter in UV derived SFRs to that of the intrinsic popu- Such calibrations will become extremely powerful with
lation. Furthermore, samples of UV+TIR detected sources the next generation of deep large area radio continuum
have increased dramatically with the extensive surveys of surveys from the Square Kilometre Array (SKA) and its
GALEX (Martin et al. 2005) and Herschel (Pilbratt et al. precursors, such as ASKAP-EMU (Norris et al. 2011)
2010), and improvements to SED modelling, such as mag- andMeerKAT-MIGHTEE(Jarvis2012).Inpreparationfor
phys (da Cunha et al. 2008) and cigale (e.g. Noll et al. these future studies, it is essential that we fully exploit ex-
2009),haveallowedustoprobestatisticallyrobustsamples isting datasets in order to explore SFRs derived from the
usingUV+TIRSFRs(forexampleseeD16andSmithetal. 1.4GHz radio emission. Here we use the current state of
2012). theartlargearearadiosurvey,FIRST,incombinationwith
Despitetheseimprovements,itisalsopossibletoavoid GAMA to investigate the 1.4GHz SFR indicator and make
sourcesoferrorinducedbyobscurationcorrectionsandaper- predictions for number of GAMA sources what will be de-
turecorrectionsbyusingameasureofstar-formationwhich tectablewithASKAP.Throughoutthispaperweuseastan-
isunaffectedbydustobscuration,integratedoverthewhole dard ΛCDM cosmology with H0=70kms−1Mpc−1, ΩΛ=0.7
galaxy and probing down to faint levels. The radio contin- and ΩM=0.3.
uum is ideally suited to this. It has long been known that
there is a tight correlation between FIR emission and rest-
frame 1.4GHz radio power (e.g. van der Kruit 1971; Helou,
Soifer,&Rowan-Robinson1985;Condon1992;Yun,Reddy, 2 DATA
& Condon 2001). This relation arises because emission at
2.1 GAMA
bothwavelengthsisconnectedwithongoingstarformation.
Emissionfromstarforminggalaxiesat1.4GHzisdominated TheextendedGAMAsurvey(GAMAII)covers286deg2 to
by synchrotron radiation arising from relativistic electrons a main survey limit of r < 19.8mag in three equatorial
AB
thought to be accelerated by supernovae shocks (e.g. Har- regions(G09,G12andG15)andtwosouthernregions(G02
wit&Pacini1975).Giventhatmassivestarsdominateboth andG23 surveylimitofi <19.2maginG23)(Liskeetal.
AB
thesupernovarateanddustheating,theFIR-radiocorrela- 2015). The limiting magnitude of GAMA was initially de-
tion arrises through the same underlying sources producing signed to probe all aspects of cosmic structures on 1kpc
the emission at both wavelengths. As the supernova rate is to 1Mpc scales spanning all environments and out to a
intimately linked to the birth of high mass stars and emis- redshift limit of z ∼0.4. The spectroscopic survey was un-
sion at these wavelengths is unencumbered by dust obscu- dertakenusingtheAAOmegafibre-fedspectrograph(Sharp
ration, the non-thermal radio luminosity provides a robust et al. 2006; Saunders et al. 2004) in conjunction with the
and dust-insensitive measure of the current star-formation Two-degree Field (2dF, Lewis et al. 2002) positioner on
on∼100Myrtimescales(e.g.seeCondon,Cotton,&Broder- the Anglo-Australian Telescope and obtained redshifts for
ick2002).Thermalradioemissionisalsostronglycorrelated ∼280,000 targets covering 0 < z (cid:46) 0.5 with a median red-
with star-formation (e.g. Galvin et al. 2016), but has a dif- shiftofz∼0.2,andhighlyuniformspatialcompleteness(see
MNRAS000,1–14(2016)
GAMA: 1.4GHz Radio SFR 3
FIRST half beam width, see similar crossmatching in e.g.
1210 Sadler et al. 2007) between the GAMA IIEq galaxies with
● robust redshifts and photometric measurements, and the
FIRST catalogues. Where multiple GAMA sources are
Stellar Mass, Mo 91011101010 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● GAMA mwmappaacposnrlrreeaaeueodtttsdniahlccssFAitushh)ekssccIeeeiRoweGddglsamyunShtAagbtToptoaMthsoarhltpigarogAaeabeaaxbhnrb-eliuctsafFeloliisflxeAsoaIntruRislmgGteiexnoclsSsfeNarTttaitdlanhihFcepwbmesetnItorahRihsasosomGiaieitrScnttmieAhpyTioGolnpoMcednAlfol.dioemAbmMemWmooetiupAettbavienwrslctote.tcaahrlexiGuhteieoGineggm.nnmAeihiTvtoedMlarh(en,ihtag<nebsAcwidh.ysh1it-ht2oF.Crhi%b4tecIahheehsRGtmtauionwSHtwlwfitgTeszseteessehionnirtosinauesuonaiSfrummels1cD.rraae9bc(mnSs9tteei)ndS1oosr-,
● ● GAMA−Spirals star-formation, we opt to exclude all sources which poten-
810 GAMA−FIRST tiallyhavesomefractionoftheirradioemissionarisingfrom
0 ● 0.1 0.2 0.3 0.4 an AGN and apply multiple cuts to produce a robust, but
bydesign,incompletesampleofstar-formingradiogalaxies.
Redshift
1.4GHz luminosities for the GAMA-FIRST sample are
calculated using the total integrated flux densities (FINT)
Figure 1. The redshift-M∗ distribution of GAMA IIEq galaxies fromtheFIRSTcatalogue,convertedtointrinsicluminosity
(contours),theGAMAvolumelimitedspiralssampleofGrootes
using the GAMA redshifts and k-corrected assuming a
et al. (2014) used in our stacking analysis (red circles), and the
finalstar-formingGAMA-FIRSTsample,excludingallpotential power law slope of Sν ∝ ν−0.7 (assuming emission from
AGNsources(goldsquares).Thecolouredshadedregionsdisplay optically thin synchrotron radiation). For completeness, we
thevolumestackedwithincarefullydesignedstellarmassbins. alsoperformouranalysisassumingaS ∝ν−0.6 andS ∝ν−0.8
ν ν
slope and find that it does not significantly change our
results. To remove potential AGN-like sources, we apply
Baldryetal.2010;Robothametal.2010;Driveretal.2011,
the following steps:
for summaries of GAMA observations).
Full details of the GAMA survey can be found in Hop-
•First,weexcludesourceswhichareidentifiedasAGN
kinsetal.(2013);Driveretal.(2011,2016)andLiskeetal.
using the BPT diagnostic (Baldwin, Phillips, & Terlevich
(2015). In this work we use the data obtained in the three
1981). We select all GAMA-FIRST galaxies which have
equatorial regions, which we refer to here as GAMA II .
Eq [OIII],Hβ,[NII]andHαlinesdetectedat>2σ.Thetopleft
Stellar masses for the GAMA II sample are derived from
Eq panel of Figure 2 displays the distribution of these sources
the ugriZYJHK photometry using a method similar to that
in the BPT diagram. We use the AGN-SF dividing line
outlined in Taylor et al. (2011) assuming a Chabrier IMF
of Kauffmann et al. (2003), to exclude sources which are
(Chabrier 2003). Figure 1 displays the stellar mass-redshift
identified as AGN via their optical emission line ratios (i.e.
distributionoftheGAMAII sample.Allphotometryused
Eq we remove all black points in Figure 2 from our sample).
in this work comes from the lambdar catalogue discussed
This removes 236 optically identified AGN.
in Wright et al. (2016) and spectral line analysis will be
detailed in Gordon et al (in prep).
• This process does not account for heavily obscured
(optically thick) AGN, which may not be identified via the
2.2 FIRST BPT method but can still show strong radio emission. In
order to remove such sources we apply the Wide-field In-
The Faint Images of the Radio Sky at Twenty-cm (FIRST)
frared Survey Explorer (WISE) colour selection of obscured
survey (Becker, White, & Helfand 1995) is a 1.4GHz con-
AGN in a similar manner to, for example, Stern et al.
tinuum survey in the Northern hemisphere and contains
(2012) and Mateos et al. (2013). Figure 2 highlights this.
∼90 sources deg−2 at the 1mJy survey threshold to an
The top right panel of Figure 2 shows the WISE colours
rmssensitivityof∼0.15mJybeam−1.Thesurveywasunder-
for all GAMA sources (contours) and our GAMA-FIRST
taken by the VLA in B configuration with a synthesized
matchedsample(gold).Hereweapplyaconservative(more
restoring beam of 5.4(cid:48)(cid:48) full width at half-maximum. We use
strict than previous works) selection of W1-W2<0.125;
the‘14Dec17’FIRSTcataloguewhichcontainsobservations
where W1 and W2 are the observed magnitudes in WISE-1
from1993to2011.Thiscatalogueconsistsof946,432sources
(3.4µm) and WISE-2 (4.6µm) bands respectively, taken
covering ∼10,500 deg2 (i.e.∼95deg−2). fromtheGAMAlambdarcatalogue(removing70sources).
• We also remove sources which have WISE colours
3 COMBINING GAMA AND FIRST consistent with passive galaxies (as their radio emission is
likely to arise from an AGN not SF), using the colours of
3.1 GAMA-FIRST Detected Sample
passive spirals outlined in Fraser-McKelvie et al. (2016),
To identify GAMA galaxies which have a detection in W3−W2>−0.5, where W3 are the observed magnitudes in
FIRST, we perform a 3(cid:48)(cid:48) cross match (comparable to the WISE-3(12µm).Thisremovesafurther1277sources.None
MNRAS000,1–14(2016)
4 L. J. M. Davies et. al.
2
AGN l l l 1.0
5
1. l
Log[OII/Hb] 00.51 lll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllll lll W1−W2 −0.50.00.5 AGN
lll l
ll l
SFG l Passive
0.5 −1.0 SFG
-
- 1.5 - 1 - 0.5 0 0.5
−3 −2 −1 0 1 2 3
Log[NII/Ha]
W3−W2
l
2 5 l lll
l ll
l
AGN l lllll ll lll l l
l l lll
log(S/S)1.4K -011 SFG S/SNVSSFIRST 12 llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllll l ll l l ll ll l
l
2 l l l
- l
- 1 0 1 2 0 5 10 15 20 25
log(SW4/S1.4) r−band Re, arcsec
Figure2.Identificationofstar-forminggalaxiesinGAMA-FIRSTsample,regionswheresourcesareexcludedinoursampleselectionare
shadedred.Topleft:BPTclassificationusedtoselectstar-forminggalaxiesfromtheGAMA-FIRSTcrossmatchedsample.Pointsdisplay
GAMA-FIRSTmatchedsourceswhicharedetectedat>2σinallemissionlinesrequiredfortheBPT.Thereddashedlinedisplaysthe
SF-AGNdividinglineofKauffmannetal.(2003).WeexcludeGAMA-FIRSTmatchedsourceswhichareidentifiedaspectroscopic-AGN
via the BPT diagram (black points). Gold points remain in our sample. Top Right: WISE colour selection of obscured AGN sources.
ContoursdisplaytheGAMAsample,whilegoldsquaresdisplaytheremainingGAMA-FIRSTsample,afterBPTrejectionofAGN.We
applyaconservativecutinW3−W2<0.125,redhorizontalline,toexcludeGAMA-FIRSTsourceswhichpotentiallycontainanobscured
AGNandalsoexcludesourcewithW3−W2>−0.5,redverticalline,assuchsystemshavecoloursconsistentwithpassivegalaxies(and
as such their radio emission is unlikely to arise from star-formation. Bottom left: W4/1.4GHz and 1.4GHz/K-band AGN selection of
Seymouretal.(2008).GoldpointsshowsourcesthatremaininourselectionafterboththeBPTandWISEselection.Bottomright:The
NVSS/FIRST1.4GHzfluxdensityratioasafunctionofr-bandeffectiveradius(Re).Goldpointsdisplaysourceswhicharestillinour
selection after all previous cuts, while black points display all sources which are detected in both FIRST and NVSS. Note that not all
oftheremainingsameareshowninthispanelasonlyasub-samplehaveNVSSdetections.WeexcludeallsourceswithRe >5(cid:48)(cid:48) which
potentiallyhaveresolvedoutfluxinFIRST.
MNRAS000,1–14(2016)
GAMA: 1.4GHz Radio SFR 5
of the sources removed here are identified as star-forming for our final sample as cyan points, and highlight that the
using the BPT diagnostic as they have do not have the addition of ‘resolved out’ flux would not strongly affect our
required BPT emission lines. derived relations.
• We then exclude any source which has a rest-frame • Last, we then visually inspect all remaining sources
1.4GHz luminosity of > 1023.5WHz−1, as such high lumi- for broad Hα line emission (potentially broad-line AGN)
nosities may be representative of an AGN (this luminosity and/or extended and two component radio emission (po-
would imply SFR>200M yr−1 using previous calibrations), tentially lobed radio galaxies), and exclude a further 4
(cid:12)
and also sources with exceedingly large 1.4GHz luminosity systems.
in comparison to their measured UV+TIR SFR, excluding
sources with log [L ]>log [SFR]+23 (displayed as the This leaves a final, highly robust, non-AGN star-
10 1.4 10
grey shaded region in Figure 3). This selection may remove forming GAMA-FIRST sample of just 144 galaxies. While
ultra luminous infrared galaxies (ULIRGS) which poten- this sample is small, we have made every possible effort
tially have all of their emission arising from star-formation. to exclude any sources of AGN contribution to the radio
These sources generally reside at higher redshifts than the emission.WepresentthefinalGAMA-FIRSTsampleasthe
GAMA sample however, and thus their potential removal gold squares in Figure 1. This sample largely consists of
will not affect our derived calibrations. This selection sourceswithhighradioluminositiesandstarformationrates
removes a further 170 sources. (as they are individually detected in the relatively shallow
FIRST data). To push to lower radio powers requires the
• In the bottom left panel of Figure 2 we exclude stacking of well-defined populations.
remaining sources which meet the radio-NIR/MIR AGN
selection of Seymour et al. (2008). We use a conserva-
3.2 GAMA-FIRST Stacking
tive selection to exclude as AGN the 115 sources with
log[S22µm/S1.4GHz]<0.5 (where S22µm is the lambdar WISE-4 To supplement the individually detected GAMA-FIRST
(22µm)flux).Thismayleadtotheremovaloflowmetallic- galaxies described above, we also perform a stacking anal-
ity dwarf galaxies, but this is unlikely to significantly affect ysis of stellar mass selected star-forming galaxies within a
our sample. volume limited sample from GAMA.
We use the low contamination and high complete-
This leaves 172 non-AGN star-forming galaxies in the ness, volume limited sample of spiral galaxies outlined in
GAMA-FIRST sample. Using the high resolution FIRST Grootes et al (submitted) and D16, and selected follow-
data however, also leads to the possibility of radio flux ing the method presented in Grootes et al. (2014) - here-
being‘resolvedout’forlargeangularsizesourceswithfaint after GAMA-SPIRALS. Briefly, the sample uses a non-
radio emission in their extremities (i.e. Jarvis et al. 2010). parametric, cell-based, morphological classification algo-
This could potentially lead to an underestimation of source rithm to identify spiral galaxies at 0 < z < 0.13. The mor-
fluxdensityandthusbiasanyderivedcalibrations.Inorder phological proxy parameters used in Grootes et al. are the
to investigate this we use the the NRAO VLA Sky Survey r-band effective radius, i-band luminosity and single-S´ersic
(NVSS, Condon et al. 1998), a 1.4GHz survey using the index(takenfromKelvinetal.2012),importantlyavoiding
VLA in the more compact D configuration. This compact observables which are themselves SFR indicators. We refer
configurationhaspoorerresolutionthanFIRSTbutgreater the reader to Grootes et al. (2014) and Grootes et al. (sub-
sensitivity to extended outlying structure. As such, NVSS mitted) for further details.
providesarobustmeasurementoftotal1.4GHzfluxdensity, TheredpointsinFigure1displaytheGAMA-SPIRALS
but is more likely to be affected by source confusion. To sample, which contains 6,366 sources. We then also exclude
estimate the fraction of flux that is potentially resolved galaxieswhichareidentifiedasAGNusingtheBPTdiagnos-
out, we match to the NVSS catalogue and find 54 sources tic,leaving6,149sources.Thisprocessmaystillretainheav-
in our remaining sample have NVSS detections. Figure 2 ilyobscuredAGN,whichareproblematictoremovefromthe
(bottom right) displays the NVSS to FIRST flux density sample prior to stacking. If included such sources could po-
ratioagainstr-bandeffectiveradiustakenfromGAMA.We tentiallycauseaslightoverestimationinthestacked1.4GHz
display both our robust SF sample (gold squares) and all measurements. We do not exclude galaxies that would be
other FIRST-NVSS matches from our initial 1991 sources identified as AGN using the WISE colour selection in Sec-
(black points). Clearly, NVSS measures a larger 1.4GHz tion 3.1 as we wish to keep our GAMA-SPIRALS sample
flux density than FIRST for many sources, but typically identicaltothatusedinD16,butnotethatonly21(<0.5%)
finding differences of less then a factor of two. Here we opt sources in our sample would meet the such a selection.
toexcludelargesourceswhicharemostlikelytobeaffected We split the resulting sample into six stellar mass bins
missingfluxinFIRST.Wedonotexcludesourcesbasedon from 9.25<log[M /M ]<11.25. We include four intermedi-
∗ (cid:12)
their NVSS to FIRST flux density ratio as not all sources ate mass bins of ∆log[M /M ]=0.25, boundedby two larger
∗ (cid:12)
have NVSS detections. ∆log[M /M ]=0.5 bins at the high and low mass end to in-
∗ (cid:12)
crease signal to noise in the resultant stacks where either
• In a similar manner to Hopkins et al. (2003) but sources are radio faint (the low mass end) or the number
with a more conservative cut, we exclude all sources from density of galaxies is low (the high mass end). Our stacked
our sample with r-band effective radius > 8(cid:48)(cid:48), removing a volumesaredisplayedasthecolouredshadedregionsinFig-
further 24 galaxies. In Figure 3, displaying our 1.4GHz ure 1. Stellar mass ranges, median redshifts and number
luminosity to SFR relation, we show NVSS measurements densities of the stacked samples can be found in Table 1.
MNRAS000,1–14(2016)
6 L. J. M. Davies et. al.
Table 1. Properties of the GAMA-FIRST stacked samples. Column 1: the stellar mass range over which our volume-limited sample
of spiral galaxies is stacked. Column 2: the median redshift of the stacked sample. Column 3: the number of sources in the stacked
sample.Column4:thenumberofindividuallydetected(peakfluxdensity>0.9mJy)sourcesinthestack.Column5:stackedfluxdensity
measurementforallsources.Column6:stackedfluxdensitymeasurementexcludingindividuallydetectedsources.Column7:Luminosity
measurement from flux density-measured stack, using stacked flux density and median redshift. Column 8: Luminosity measurement
from luminosity-measured stack, using individual source redshifts. Luminosity measurements are derived from the full stacks only (not
excludingdetections),buttheerrorrangeincorporatesthedifferencebetweenthefullstackandthestackwithdetectedsourcesremoved.
StellarMass Median # # S1.4 -full S1.4 -nodetect Lflux−measured Llum−measured
log[M∗/M(cid:12)] Redshift full detected µJy±rms µJy±rms ×1021 WHz−1 ×1021 WHz−1
(1) (2) (3) (4) (5) (6) (7) (8)
9.25-9.75 0.100 2261 7 24.2±5.6 24.2±5.5 0.60±0.14 0.39±0.07
9.75-10.00 0.107 706 4 46.3±9.6 43.5±9.7 1.33±0.38 0.85±0.20
10.00-10.25 0.106 565 1 76.3±11.1 76.3±11.2 2.14±0.31 1.64±0.22
10.25-10.50 0.106 456 3 94.1±12.0 93.4±12.0 2.63±0.34 2.11±0.22
10.50-10.75 0.108 213 6 91.8±17.2 86.0±17.5 2.67±0.67 2.37±0.64
10.75-11.25 0.111 126 5 100.0±23.6 97.0±23.7 2.98±0.70 2.45±0.48
We perform the stacking analysis using two different pixelsintheoffsetpositionusingauniqueredshiftfromthe
modesbothstackingtheFIRSTdatadirectly,notcatalogue GAMA-SPIRALSsample(thusreplicatingthesameredshift
measurements. In both modes we apply median stacking distributioninourrmsmeasurements).Wedonotdisplaylu-
to exclude outlying pixels without the need to apply arbi- minositymeasurementsusingthestackedsamplewithindi-
trary cutoffs to the distribution. Median stacking has been vidually detected sources removed, but highlight that these
found to work successfully when investigating faint sources only marginally differ from the full stacked sample (<5%).
in FIRST, for example White et al. (2007). Wealsoincludethedifferencebetweenthefullsampleanda
First,weproducestacksbymediancombiningthepixel sample excluding detected sources in our luminosity errors.
values of the FIRST data centred on the positions of the
GAMA-SPIRAL samples in each mass bin. We then mea-
sure the total integrated flux density at the central beam
of the median stack using the miriad maxfit function and Inordertoavoidincludingradioemissionfromsources
outside of the GAMA-SPIRALS sample, or repeat stacking
derivea1.4GHzluminosityusingthemedianredshiftofall
within the sample, we confirm that none of our GAMA-
sources in the mass bin and k-correcting assuming a power
lawslopeofα=−0.7(medianredshiftsaregiveninthesec- SPIRALS sample overlaps with another GAMA source
within 5(cid:48)(cid:48), and thus the FIRST beam size. As such, we do
ond column of Table 1). Hereafter, we will refer to this as
not have to exclude potentially confused sources. However,
the flux density-measured stack. This stacking process es-
this does not rule out contributions to the emission arising
sentiallyassumesthatthereisnoevolutionovertheredshift
fromsourcesbelowtheGAMAr-bandselectionlimit(these
rangeofoursampleandthatsourcesareevenlydistributed
sources are likely to be faint in radio emission) or high red-
over the redshift range probed.
shiftsourceswhichsitwithinthebeamoftheGAMAgalaxy.
Secondly,wedeterminetheindividualluminosityofthe
Given it is impossible to remove such sources (as there are
FIRST data at the position of each of the GAMA-SPIRAL
nodeeperspectroscopicobservationsintheGAMAregions)
samples.ForthisweextractaregionoftheFIRSTdatacen-
we cannot make assessments regarding their contribution
tred on the position of the GAMA-SPIRALS source, then
to the observed flux density. However, given that the re-
convert every pixel value into a luminosity at the source’s
sults in the following sections display consistency between
redshift(againassumingα=−0.7).Wethenmediancombine
our stacked samples and individually detected sources, it is
thepixelvaluesineachextractedregionandagainmeasure
unlikely that such faint galaxies strongly contribute to our
the total integrated luminosity at the central beam. Here-
derivedfluxdensities.Wealsodonotexcludesourceswhich
after,wewillrefertothisastheluminosity-measuredstack.
have r-band effective radius>8(cid:48)(cid:48) in our stacked samples, as
Thisstackedsampleusesalldistancemeasurementsforindi-
intheindividualdetections.Whilethesesourcesmaypoten-
vidual sources, and hence avoids the assumption of no evo-
tially have ‘resolved out’ flux, we wish to keep the stacked
lution and even distribution over the redshift range.
sample identical to that used in D10, and note that an r-
For each stellar mass range we also produce identical band effective radius<8(cid:48)(cid:48) cut would only remove 35 sources
stacked samples with the individually detected sources re- (∼1%) from our GAMA-SPIRALS sample.
moved. In Table 1 we display the median flux density stack
measurements for both the full stacks and the stacks with
individuallydetectedsourcesremoved.Wealsodisplaylumi-
nositymeasurementsforboththefluxdensity-measuredand Figure4displaysthefullstackedGAMA-SPIRALsam-
luminosity-measured stacks using the full sample. In order plesindifferentstellarmassbins.Allstackedvaluesincludea
toestimatermserrors,westackthesamenumberofsources multiplicationfactorof1.4toaccountfor“CLEAN”bias(see
as in each stellar mass bin, but at random offset positions White et al. 2007, for further details). We obtain a >4.25σ
in the FIRST data and measure the resultant rms. For the detectioninallbinsinourfluxdensity-measuredstacksand
luminosity-measuredstack,wecalculatetheluminosityofall >4.5σ detections in our luminosity-measured stacks .
MNRAS000,1–14(2016)
GAMA: 1.4GHz Radio SFR 7
l l
l
l
00 Free Fit: m=0.66 c=−14.02 00
1 Fixed Fit: m=1 c=−21.62 1 Free Fit: m=0.75 c=−15.96 l l l
Fixed Fit: m=1 c=−21.59
Stack Fit: m=0.63 c=−13.27
D16 SFR−UV+TIR, M/yro 101 HCBooospnekdliolni+ns+1+59023llllll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l l SFR−MAGPHYS, M/yro 101 Stack Fit: m=l0.l4l8l cl=−10lllll.l1ll2lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l l l
l
l UV+TIR l l ll MAGPHYS
l GAMA−FIRST detections l
GAMA−FIRST Luminosity Stack
GAMA−FIRST Flux Stack l
0.1 GAMA−NVSS detections 0.1
1021 1022 1023 1021 1022 1023
1.4GHz Luminosity, W Hz-1 1.4GHz Luminosity, W Hz-1
Figure 3. Correlation between 1.4GHz luminosity and SFR indicators from GAMA outlined in Davies et al. (2016): UV+Total IR
derived SFR (left) and magphys SED-derived SFR (right). Circles display the GAMA-FIRST detected sample. Open squares display
ourluminosity-stackedsamples,whilefilledtrianglesdisplayourfluxdensity-stackedsamples-whereerrorsaresmallerthantheplotted
points.Wefittherelationsusinghyperfitforbothafreeslopeandnormalisation(magentalines),afixedm=1relation(bluelines),a
freefittojustthefluxdensity-measuredstackeddatapoints(greenline).WeshowtheHopkinsetal.(2003),Condon(1992)andBoselli
etal.(2015)relationsasthedarkgreen,orangeandpurple-dashedlinesrespectively.Forcompleteness,wealsoshowthesmallnumber
of sources in our sample with NVSS detections (cyan diamonds) to highlight that including potential ‘resolved-out’ flux in our sample
wouldnotsignificantlychangeourderivedrelations.Greyshadedregiondisplayswheresourceswitherroneouslyhigh1.4GHzluminosity
areexcluded.Theexcludedpointsfalloffthisfigureandthegreyshadedregionisonlyintendedtoshowthatwearenotbiasingourfits
byexcludingobjectsinthisregion.
4 1.4GHZ LUMINOSITY-SFR RELATION band photometry outlined in Driver et al. (2016); using
GALEX-UV, SDSS-optical, VIKING-NIR, WISE-MIR and
Using both the individually detected GAMA-FIRST galax-
Herschel-ATLAS-FIR data. We follow a Bayesian process,
ies and our stacked samples, we investigate the 1.4GHz
with uniform/uninformative priors on the templates (i.e.
luminosity-SFR relation. D16 provides multiple SFR esti-
each template is assumed to be equally likely). For a par-
matesusing12differentmethodsforderivingSFRandpro-
ticulartemplate,thebestfit/maximumlikelihoodvalueand
duces consistent measurement of star-formation across all
theformaluncertaintyareanalytic(throughtheusualprop-
methods. Here we only compare to the full SED measures
agationofuncertainties).Theposteriorforthebest-fitvalue
of star-formation, UV+TIR (UV+TIR1 in D16) and mag-
template is given by marginalising over the full set of tem-
phys(daCunhaetal.2008).Giventherecalibrationprocess
plates. By effectively marginalising over template number
in D16, all other GAMA SFRmethods willproduce similar
asanuisanceparameter,wefullypropagatedtheerrors,in-
results to the UV+TIR measurement. We also expect the
cluding uncertainties due to template ambiguities.
1.4GHz SFRs to be most closely correlated with long du-
ration measures of star-formation, as they arise from SNe- magphys SFRs use the Bruzual & Charlot (2003) stel-
driven emission. We opt to use the full SED measurements larpopulationswithaChabrier(2003)IMFandassumesan
of star-formation over FIR emission only (as has previously angle-averaged attenuation model of Charlot & Fall (2000).
been used when calibrating 1.4GHz via the FIR-radio re- This is combined with an empirical NIR-FIR model ac-
lation), as the UV+TIR SFR estimation combines the SF counting for PAH features and near-IR continuum emis-
information derived in the FIR with that observed in the sion, emission from hot dust and emission from thermal
UV, and as such is likely to produce a more representa- dust in equilibrium. The code defines a model library over
tive measure of the total star-formation. We also include a wide range of star formation histories, metallicities, and
the magphys SFR asitgives analternative estimate of the dust masses and temperatures, and fits the photometry -
SF,essentiallyusinginformationfromtheUV+TIR,butde- forcing energy balance between the observed TIR emission
rived using a different fitting method. Both SFR measures and the obscured flux in the UV-optical. Physical proper-
used here assume a Chabrier IMF. ties (SFR, SFH, metallicity, dust mass, dust temperature)
Briefly,theUV+TIRSFRusestheBrownetal.(2014) for the galaxy are then estimated from the model fits, giv-
spectrophotometrically calibrated library of galaxy spectra ing various percentile ranges for each parameter. Here we
to derive UV and TIR luminosities, from the GAMA 21- usethemedianSFR parameter,whichprovidesanesti-
0.1Gyr
MNRAS000,1–14(2016)
8 L. J. M. Davies et. al.
0 0 0 0 0 0
2 30 2 30 2 60 2 15 2 30 2 60
oJy20 oJy20 oJy40 oJy10 oJy20 oJy40
10 Flux, micr100 10 Flux, micr100 10 Flux, micr200 10 Flux, micr 05 10 Flux, micr100 10 Flux, micr200
ec −10 ec −10 ec −20 ec −5 ec −10 ec −20
arcs 0 arcs 0 arcs 0 arcs 0 arcs 0 arcs 0
ec, ec, ec, ec, ec, ec,
D D D D D D
0 0 0 0 0 0
1 1 1 1 1 1
− − − − − −
0 0 0 0 0 0
2 2 2 2 2 2
− − − − − −
−20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20
RA, arcsec RA, arcsec RA, arcsec RA, arcsec RA, arcsec RA, arcsec
9.50<log[M*/M(cid:1)]<9.75 9.75<log[M*/M(cid:1)]<10.00 10.00<log[M*/M(cid:1)]<10.25 9.25<log[M*/M(cid:1)]<9.75 9.75<log[M*/M(cid:1)]<10.00 10.00<log[M*/M(cid:1)]<10.25
20 60 20 60 20 100 20 60 20 60 20 100
10 Flux, microJy24000 10 Flux, microJy24000 10 Flux, microJy500 10 Flux, microJy24000 10 Flux, microJy24000 10 Flux, microJy500
Dec, arcsec 0 −20 Dec, arcsec 0 −20 Dec, arcsec 0 −50 Dec, arcsec 0 −20 Dec, arcsec 0 −20 Dec, arcsec 0 −50
−10 −10 −10 −10 −10 −10
−20 −20 −20 −20 −20 −20
−20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20
RA, arcsec RA, arcsec RA, arcsec RA, arcsec RA, arcsec RA, arcsec
10.25<log[M /M]<10.50 10.50<log[M /M]<10.75 10.75<log[M /M]<11.00 10.25<log[M /M]<10.50 10.50<log[M /M]<10.75 10.75<log[M /M]<11.25
* (cid:1) * (cid:1) * (cid:1) * (cid:1) * (cid:1) * (cid:1)
Figure 4. Stacked 1.4GHz images of our volume limited late-type sample. The green ellipse shows the FIRST beam shape, centred
onthestackposition.Wealsoproducestacksexcludingindividuallydetectedsources,andthoseproducedwhenstackinginluminosity
spaceforeachsample,butforclaritywedonotshowthemhere;seetextfordetails.
matefortheSFRaveragedoverthelast0.1Gyrs.Errorson sources are not uniformly distributed over the redshift of
SFRmagphys are estimated from the 16th-84th percentile our stacked sample.
range of the SFR parameter, which encompasses both WeshowpreviouslypublishedrelationsoutlinedinHop-
0.1Gyr
measurement and fitting errors. kins et al. (2003) (from SDSS-FIRST), Boselli et al. (2015)
(from the Herschel Reference Survey, K-band selected sam-
ForfurtherdetailsoftheseSFRs,seethemoredetailed ple) and Condon (1992), as the dark green, purple-dashed
descriptions in D16. We do not use the favoured radiative andorangesolidlinesrespectively.TheHopkinsetal.(2003)
transfer-derivedSFRsofD16inthisworkaswedonothave lineisplottedasabrokenpowerlawtoaccountforthescal-
these SFRs for the full GAMA-FIRST sample. ing for non-thermal radio continuum emission from dwarf
galaxies,appliedintheirrelation.Allrelationsarescaledto
Figure3displaysthe1.4GHzLuminosity-SFRrelation a Chabrier IMF using the conversions outlined in Haarsma
for both the UV+TIR SFR and magphys SFRs from D16. et al. (2000), for Miller-Scalo to Salpeter, and Driver et al.
IndividuallydetectedsourcesfromtheGAMA-FIRSTsam- (2013), for Salpeter to Chabrier.
ple are displayed as circles while the flux density-measured We then fit the 1.4GHz luminosity-SFR relation lin-
andluminosity-measuredstacksaredisplayedasfilledtrian- early in a number of ways using the multi-dimensional
glesandopensquaresrespectively.Bothmethodsfordeter- MCMC fitting [r] package hyperfit1 (Robotham &
miningthestackedfluxesarefoundtobewithinthescatter Obreschkow 2015). Firstly, we fit the full distribution using
of the individually detected sources, suggesting that stack- a fixed, m=1, slope (blue line), these fits are almost iden-
ing method does not strongly affect our results. We do find tical for both the UV+TIR and magphys SFRs but have
thattheluminosity-measuredstacksproducesystematically an offset normalisation from the Hopkins et al. (2003) and
lower luminosity measurements, which potentially suggests
that in using a flux stack and median redshift over-predicts
to true luminosity. This may be due to the fact that the 1 http://hyperfit.icrar.org/
MNRAS000,1–14(2016)
GAMA: 1.4GHz Radio SFR 9
0 0.4 0 0.4
10 UV+TIR 10 MAGPHYS l l
l l
0.3 0.3 l
l l
R l R l l
SFR, M/yro 101 000..12ledshiftll l l ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l SFR, M/yro 101 000..12ledshiftll l ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l
l
l l
GAMA−FIRST Stack
l lllll GAMA−FIRST detections
Davies+16b SFR−M* l
Davies+16b Evolution
SDSS−Ha
0.1 l GAMAI+SDSS−Ha 0.1
109 1010 1011 l109 1010 1011
Stellar Mass, Mo Stellar Mass, Mo
Figure 5.The1.4GHzSFR-M∗ relationintheGAMAregions,derivedusingusingourfree-fitluminosity-to-SFRrelationforUV+TIR
(left) and MAGPHYS (right). Our new 1.4GHz-derived SFR-M∗ relation is consistent with the SFR-M∗ relation from Davies et al.
(2016)atlog[M∗/M(cid:12)]<10.5,buttunsoveratthehighmassend(theknownturnoverintheSFR-M∗relationathighstellarmass).Black
trianglesandopensquaresdisplayourflux-measuredandluminosity-measuredstackedsamplesrespectively,witherrorbarshowingthe
stackedsamplerangeinstellarmass.ColouredpointsshowtheGAMA-FIRSTmatchedsamplecolourcodedbyredshift.Wealsoshow
the SFR-M∗ fit from Davies et al. (2016) scaled to various redshifts, given the normalisation evolution taken from Eq 20 of D16 and
colourcodedonthesameredshiftscaleasthedatapoints.TheblacklinedisplaysthedirectSFR-M∗ fitfromD16toanidenticalsample
used in our stacking analysis here. Consequently, the stacked data points should be directly compared to the black line. We also show
theHα-derivedSFR-M∗ fitsfromSDSSatz=0(Elbazetal.2007)andGAMAI+SDSSatz<0.1(Lara-Lo´pezetal.2013)asthegreen
dashedanddottedlinesrespectively.ErrorsinmedianSFR(includingfittingerrors)aresmallerthanthesymbols.
Condon (1992) relations. Secondly we fit the distributions issub-linear.Thisisconsistentwithnon-calorimetricmodels
with a free slope and normalisation (magenta line), these of non-thermal emissioningalaxy disks (e.g. Niklas & Beck
fitshaveaslightlydifferentslopebetweentheUV+TIRand 1997; Bell 2003; Lacki, Thompson, & Quataert 2010; Irwin
magphys SFRs. Interestingly for both the UV+TIR and etal.2013;Basuetal.2015),wherecosmicrayelectronsdo
magphys SFRs this fit has a similar slope and normaliza- not lose all of their energy before escaping galaxies and not
tion to the lower 1.4GHz broken power-law component, for alloftheirenergyisradiatedassynchrotronradioemission.
dwarf galaxies, of the Bell (2003) and Hopkins et al. (2003) These models predict a SFR∝L 0.73−0.9 relation (consistent
1.4
relation (i.e. the dark green and magenta fits have a similar slopeofourmagphysfits).However,thesomewhatextreme
slopeat L <6.4×1021WHz−1).Lastly,wefitthedistri- non-calorimetric model are seemingly in conflict with the
1.4GHz
butions using just the flux density-measured stacks (green tightness of the far-IR-radio relation over a broad range of
line). physical properties of the host galaxy (e.g see diecussion in
All fits take the form of: Lacki, Thompson, & Quataert 2010). While linear, calori-
metric, fits (m = 1, blue lines) are not in strong conflict
log10[SFR(M(cid:12)yr−1)]=m×log10[L1.4GHz(WHz−1)]+C (1) withourdata(specificallyforthemagphysrelations),non-
with parameters, m and C, given in the figure. Given our calorimetric models for radio emission will require further
freefit(whicharethebestfittothefulldataset)wesuggest investigation in the MeerKAT/ASKAP/SKA era.
a new calibration to the 1.4GHz-SFR relation as:
5 THE 1.4GHZ SFR-M RELATION
log10[SFRUV+TIR]=0.66±0.02×log10[L1.4]−14.02±0.39 (2) ∗
Using the 1.4GHz luminosity-SFR calibration derived
above, it is possible to explore the 1.4GHz SFR-M rela-
∗
log [SFR ]=0.75±0.03×log [L ]−15.96±0.58 (3) tion (Figure 5). We display the flux-measured stacked data
10 MAGP 10 1.4 pointsassolidblacktriangles,luminosity-measuredstacked
Interestingly, we find best fit relations with sub-linear data points as open squares and the individually detected
slopes (i.e. m(cid:44)1). Given that thermal radio emission scales GAMA-FIRST sample are shown as circles colour coded
linearlywithSFR(fromfundamentaltheoryoftheemission by their redshift. We show the SFR-M relation fit for the
∗
processes),thismustmeanthatthenon-thermalcomponent GAMA-SPIRALSsampleusingtheradiativetransferSFRs
MNRAS000,1–14(2016)
10 L. J. M. Davies et. al.
from D16 as the black solid line, and the same fit at vari- lowdepthofcurrentradiocontinuumsurveyssuchasFIRST
ous redshifts (colour coded in the same manner as the data andNVSS,andthesmallareaofdeepradiocontinuumsur-
points)usingtheevolutionofthenormalisationoftheSFR- veys, such as VLA-COSMOS (Schinnerer et al. 2007) and
M relation using Eq 20 of D16. Green dashed and dotted ATLAS1.4GHz(Halesetal.2014),havelimitedthenumber
∗
lines show the Hα-derived SFR-M fits from SDSS at z=0 ofsourceswithdetectable1.4GHzcontinuumemissionwith
∗
(Elbaz et al. 2007) and GAMA I + SDSS at z<0.1 (Lara- which to derive SFRs. This is set to change dramatically
Lo´pez et al. 2013) respectively. These do not include the with the advent of new deep large area continuum surveys
turnover at high stellar masses as they are fit linearly. from the Square Kilometre Array (SKA) and its precursors
Wefindthattheslopeandnormalisationofthe1.4GHz such as ASKAP-EMU (Norris et al. 2011) and MeerKAT-
SFR-M relation from our stacked samples, using our new MIGHTEE (Jarvis 2012). One of the key scientific goals of
∗
calibration (black triangles), has the same slope to that de- the SKA is to measure the cosmic star-formation history
rivedinD16atlog[M /M ]<10.5(Figure5);theblacklinein usingtheradiocontinuumasadust-unbiasedtracerofstar-
∗ (cid:12)
thisfigureisthefitusingthesamesamplethatisstackedin formation(seeCiliegi&Bardelli2015;Jarvisetal.2015a,b).
thiswork,butwithslightnormalisationoffset(∼0.05dexfor A potential limiting factor in the use of the 1.4GHz
the flux-weighted stacks using UV+TIR). While the mag- SFR tracer in large area surveys however, is the lack of ro-
phys stacked data points are ∼0.3dex lower than the D16 bust spectroscopic redshifts, with which to derive 1.4GHz
relation,thisisroughlyconsistentwiththeoffsetinnormal- luminositiesfromobservedfluxdensitiesandaidinthesep-
isation between the UV+TIR and magphys SFR-M rela- aration of AGN/SF-like sources. EMU is likely to detect 70
∗
tions in Figure 8 of D16. milliongalaxies,ofwhichonlyasmallfactionwillhavespec-
The slope of the 1.4GHz SFR-M∗ relation flattens at troscopic redshifts, mostly at low-z (z < 0.25) from EMUs
log[M /M ]>10.5. This is expected given the well known sibling HI spectral line survey WALLABY (see Koribalski
∗ (cid:12)
turnoverintheSFR-M relationathighstellarmasses(see 2012) and the local galaxy redshift survey, Taipan. Beyond
∗
Whitaker et al. 2014; Johnston et al. 2015; Lee et al. 2015; theverylocalUniverse,EMUwillhavetoeitherrelyonpho-
Schreiber et al. 2015; Gavazzi et al. 2015; Tomczak et al. tometricredshifts,undertakeadditionalspectroscopicobser-
2016, and discussion in D16). The turnover observed here vations, or use redshifts from existing large area surveys.
is severe however, given that our stacked sample is based The GAMA survey and upcoming Wide Area VISTA
on purely spiral galaxies. We do highlight that there is a Survey (WAVES, Driver et al. 2016b), are ideally suited to
turn over observed in other SFR indicators using the same providingalargenumberofspectroscopicredshifts.GAMA
sample (see coloured circles in Figure 8 of Davies et al. contains redshifts for ∼280,000 galaxies in the EMU foot-
2016), but this is less extreme (although only measured to print at z < 0.4. In addition, GAMA provides an exten-
log[M∗/M(cid:12)]=10.5). Potentially we are simply observing the sive database of multi-wavelength observations and value
increasingcontributionofpassivebulges,intermsofspecific added catalogues of FIR luminosities, stellar masses, dust
SFR, in galaxies at the high mass end. Further studies into masses, metallicities, environmental metrics and most im-
the high mass turn over in comparison to galaxy morphol- portantly, multiple metrics of star formation with which to
ogy and components will be the subject of upcoming work compare to the observed EMU luminosities (see D16). The
(Davies et al, in prep). upcoming WAVES survey will add ∼ 2M galaxies to this
Primarily the GAMA-FIRST individually detected sampletoz<0.8,whichwillbeinvaluableinprovidingred-
galaxiesliewellabovetheSFR-M∗ relationattheirredshift, shifts, environmental metrics, and derived parameters for
suggesting they are star-bursting galaxies. This is unsur- EMU sources. The combination of GAMA/WAVES with
prising given that they are detected in the relatively shal- the ASKAP surveys (EMU, WALLABY and Deep Inves-
low FIRST data. The exceptions to this are the very local tigationsofNeutralGasOrigins,DINGO,Meyer2009)will
galaxies (red points), which are mostly consistent with the produce a formidable dataset with which to study galaxy
SFR-M∗ relation; very nearby sources can be detected by evolution over an extensive redshift baseline.
FIRST to lower SFRs. Using the 1.4GHz luminosity to SFR relations derived
It is also interesting to note that using the individu- in the previous section we make predictions for the number
ally detected GAMA-FIRST galaxies, one would not have ofGAMAstar-forminggalaxiesthatarelikelytobedetected
been able to define the 1.4GHz SFR-M∗ relation given the in upcoming deep radio continuum surveys using ASKAP.
small number of sources spread over a large redshift range. WetakethefullGAMAII SFRsderivedinD16fora
Eq
Thishighlightsthepowerinperformingoptically-motivated numberofdifferentSFRmethods,andusethe1.4GHzlumi-
source stacking of radio continuum data using surveys such nositytoSFRrelationtopredicttherest-frame1.4GHzlu-
as GAMA. The stacked data points allow us explore the minosityforallGAMAII sources.Assumingα=−0.7and
Eq
1.4GHzSFR-M∗ relationtolowerstellarmassesthanthose the GAMA redshift, we then convert each luminosity to a
probed by the individual detected sources and for the first predictedobservedfluxdensity.Wethenexcludeallsources
time, show that the slope and normalisation of the 1.4GHz whicharedetectedasanAGNusingtheBPTdiagnosticor
SFR-M∗ relationover2decadesinstellarmassisconsistent haveWISEcoloursconsistentwithanAGN(W1−W2>0.125
with previous estimates using other multiple SFR tracers. - as in the top two panels of Figure 2). We also exclude all
sourceswhichdonothavea>2σdetectionintheobservable
usedtodetermine thesource’sSFR;suchsourcesmay have
erroneousmeasurementsofstarformation.Formagphyswe
6 PREDICTIONS FOR GAMA-ASKAP
onlyconsidersourceswherethederivedSFRisgreaterthen
Despite the recent advancements in studying 1.4GHz emis- twice the error.
sion from galaxies in large area surveys, the relatively shal- Figure 6 displays the predicted distribution of 1.4GHz
MNRAS000,1–14(2016)