Submillimetre surveys: The prospects for Herschel
aa r X i v : . [ a s t r o - ph . C O ] J un Mon. Not. R. Astron. Soc. , 1–6 (0000) Printed 1 November 2018 (MN L A TEX style file v2.2)
Submillimetre surveys: The prospects for Herschel
Chris Pearson , , ⋆ † , Sophia A. Khan , Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire OX11 0QX, UK Department of Physics, University of Lethbridge, 4401 University Drive, Lethbridge, Alberta T1J 1B1, Canada Department of Physics & Astronomy, The Open University, U.K. Harvard-Smithsonian Center for Astrophysics, 60 Garden Street MS-66 Cambridge, MA 02138, USA Shanghai Key Lab for Astrophysics, Shanghai Normal University, Shanghai 200234, China
Accepted 22nd June 2009.Received ;in original form 2009 May
ABSTRACT
Using the observed submillimetre source counts, from 250-1200 microns (including themost recent 250, 350 and 500 micron counts from BLAST), we present a model capableof reproducing these results, which is used as a basis to make predictions for upcom-ing surveys with the SPIRE instrument aboard the Herschel Space Observatory. Themodel successfully fits both the integral and differential source counts of submillime-tre galaxies in all wavebands, predicting that while ultra-luminous infrared galaxiesdominate at the brightest flux densities, the bulk of the infrared background is dueto the less luminous infrared galaxy population. The model also predicts confusionlimits and contributions to the cosmic infrared background that are consistent withthe BLAST results. Applying this to SPIRE gives predicted source confusion limitsof 19.4, 20.5 and 16.1mJy in the 250, 350 and 500 micron bands respectively. Thismeans the SPIRE surveys should achieve sensitivities 1.5 times deeper than BLAST,revealing a fainter population of infrared-luminous galaxies, and detecting approxi-mately 2600, 1300, and 700 sources per square degree in the SPIRE bands (with onein three sources expected to be a high redshift ultra-luminous source at 500 microns).The model number redshift distributions predict a bimodal distribution of local quies-cent galaxies and a high redshift peak corresponding to strongly evolving star-forminggalaxies. It suggests the very deepest surveys with Herschel-SPIRE ought to samplethe source population responsible for the bulk of the infrared background.
Key words:
Cosmology: source counts – Galaxies: surveys, evolution.
Over two decades ago, a large number of galaxies thatemit the bulk of their luminosity in the restframe far-IRwere detected in the
IRAS
All-Sky Survey (typically at z < . L ⊙ < L IR (8-1000 µ m) < L ⊙ and ULIRGs L IR > L ⊙ ) are powered by a com-bination of star formation and active galactic nucleus(Soifer et al. 1987), but only recently have they beenshown to be an important population in the early Uni-verse. This is in part due to the achievements of sub-millimetre continuum observations using ground-based fa-cilities: pioneering surveys at 850 µ m with SCUBA onthe JCMT begat the discovery of submillimetre galax- ⋆ E-mail: Chris Pearson ([email protected]) † ∼ cpp/research/ ies (SMGs; e.g., Smail, Ivison & Blain 1997, Barger et al.1998, Hughes et al. 1998, Eales et al. 1999) which weresubsequently constrained to be mainly distant star-forminggalaxies (e.g., Chapman et al. 2003). These characteristicswere shared with SMGs found in other submillimetre bandse.g., 1100 & 1200 µ m (Laurent et al. 2005, Bertoldi et al.2000) and 350 µ m (Khan et al. 2007). Larger surveys (e.g.the SCUBA SHADES survey, Mortier et al. 2005) have con-firmed these sources are strongly evolving (Coppin et al.2006). However, the discovery of SMGs still poses chal-lenges to semi-analytical hierarchical models of galaxy for-mation (e.g. Guiderdoni et al. 1998, Balland et al. 2003),and questions remain over their role in the formation of ellip-tical galaxies and supermassive black holes Magorrian et al.(1998) and the energy budget between star-formation andaccretion in the Universe.In this work we present a galaxy evolution model thatsuccessfully reproduces the source counts from 250-1200 µ m, c (cid:13) C. Pearson, S. Khan including both the large area SCUBA surveys and the lat-est results from the BLAST telescope (Pascale et al. 2008).In Section 2 we describe the model and present fits to thegalaxy counts in Section 3. The launch of SPIRE on-boardthe Herschel Space Observatory offers an opportunity to ex-amine an SMG population that overlaps with ground-basedobservations and IR-luminous galaxies selected at mid – far-IR wavelengths (e.g., with IRAS, AKARI, Spitzer). SPIREwill perform surveys at 250, 350, 500 µ m and in Section 4we discuss the prospects for upcoming surveys with Her-schel. Throughout this work a concordance cosmology ofH o = 72 kms − Mpc − , Ω = 0 . , Λ = 0 . To model the submillimetre source counts we use a far-IR backward evolution framework following the models ofPearson (2005), Pearson et al. (2007). These models werepreviously successfully used to reproduce the combinedmid-infrared source counts from
ISO & Spitzer at 15 µ m& 24 µ m. These models have now been updated to pro-duce source counts from 1-1000 µ m and will be reportedin detail in Pearson (2009). Although submillimetre lu-minosity functions are available (e.g. Serjeant & Harrison2005), to model the counts we retain the 60 µ m luminos-ity function derived from the IRAS
Point Source Cata-logue (Saunders et al. 2000) since it is defined around thepeak of the dust emission and contains a large ensemble ofsources segregated by population class. The source countsare fit to the wavelength where the luminosity function isdefined, λ LF , which sets the baseline normalization of allparameters. To predict the counts at other wavelengths,the luminosity function is shifted to the observation wave-length, λ obs , using the ratio L ( λ obs ) /L ( λ LF ), obtained viamodel template spectra, no other priori is assumed. Spec-tral templates are drawn from four source populations, com-prising normal quiescent galaxies and three star-forminggroups consisting of, with increasing luminosity, starburstgalaxies ( L IR < L ⊙ ) , LIRGs, and ULIRGs (mod-elled on the archetype Arp220). An additional AGN com-ponent (based on the emission from a dust torus) is alsoincluded within the model framework of Pearson (2009),however it is found that AGN do not contribute signifi-cantly to the source counts in the submillimetre and al-though included, their contribution is not considered in thiswork. The normal galaxy spectral templates are from the li-braries described in Efstathiou & Rowan-Robinson (2003)which exhibit cold far-IR/submillimetre colours, with spec-tra peaking between 100-200 µ m. The adopted starburst,LIRG & ULIRG spectral templates are taken from thespectral models of Efstathiou et al. (2000), which providegood fits to the IRAS , ISO and
Spitzer galaxy populations( Rowan-Robinson et al. 2004, Rowan-Robinson et al.2005). Note that all templates are independent of the ob-served data sets being fitted.Follow-up SCUBA imaging of local IRAS-selectedgalaxies has implied colder far-IR-submillimetre colours inSMGs than those derived from galaxy spectra based purelyon
IRAS colours (Dunne et al. 2000, Vlahakis et al. 2005).The colours of our model templates agree with this, as theyfollow the trend of the local galaxy colours extremely well -3-2.5-2-1.5-1-0.50-0.5 0 0.5 1
SLUGSSEDS
Normal GalaxyStarburst GalaxyLIRGULIRG l g ( S / S ) lg(S100/S60) DISKSBULIRG
Figure 1.
Colour-colour distributions of the model normal, star-burst and U/LIRG templates compared with the local submil-limetre &
IRAS far-infrared colours from Dunne et al. (2000),Vlahakis et al. (2005), with locus of normal (DISK), starburst(SB) and the ULIRG ARP 220. The markers along the SED trackscorrespond to redshift steps of δ z=0.1. in Figure 1. Although deeper SCUBA surveys are expectedto principally select LIRG/ULIRGs (Blain et al. 2002), thislocal sample also comprises lower luminosity starburst andcooler normal galaxies (also predicted to contribute at higherredshifts Efstathiou & Rowan-Robinson 2003).The star-forming populations follow the burst evolution-ary scenario of Pearson (2005), Pearson (2009), modelled byan exponential function to z ∼ ∼ The model fits to the observed source counts at 250, 350,500, 850 and 1100 µ m are shown in Figure 2 with the totalmodel source counts shown alongside the respective contri-butions of the assumed galactic populations (normal andstarburst galaxies, LIRGs and ULIRGs).Figure 2 panels a, c & d show the model fits to the dif-ferential counts (normalised to a Euclidean universe) fromthe Balloon-borne Large-Aperture Submillimeter Telescope(BLAST, Pascale et al. 2008) survey in the GOODS field(Devlin et al. 2009), for the 250, 350 & 500 µ m bands re-spectively. In all the BLAST bands, it is predicted thatthe brightest counts ( > < µ m over the BLAST survey area of 8.7 deg , com-pared with the three sources found in the brightest bin ofthe source list of Devlin et al. (2009). The steep depar-ture from Euclidean counts is caused by the ULIRGs but c (cid:13) , 1–6 ubmillimetre surveys: The prospects for Herschel (a) 250 microns BLAST
All ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG d N / d S S . { m Jy . s r - } S/mJy (b) 350 microns SHARC-IISCUBAAll ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG N u m be r / s r S/mJy (c) 350 microns BLASTSHARC-II
All ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG d N / d S S . { m Jy . s r - } S/mJy (d) 500 microns BLAST
All ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG d N / d S S . { m Jy . s r - } S/mJy (e) 850 microns SCUBA-allSCUBA-SHADES
All ComponentsNormalStarburstLIRGULIRG N u m be r / s r S/mJy (f) 850 microns SCUBA-SHADES
All ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG d N / d S S . { m Jy . s r - } S/mJy (g) 1100 microns MAMBOBolocam
All ComponentsNormalStarburstLIRGULIRGupperboundlowerbound N u m be r / s r S/mJy (h) 1100 microns AzTEC
All ComponentsNormal GalaxiesStarburst GalaxiesLIRGULIRG d N / d S S . { m Jy . s r - } S/mJy
Figure 2.
Model fits to the observed Submillimetre counts from 250 – 1200 µ m. The total contribution and individual componentscorresponding to the normal, starburst, LIRG and ULIRG populations are overplotted. (a,c,d) Model fits to the observed BLASTcounts at 250, 350, 500 µ m from Devlin et al. (2009). Also shown is the SHARC-II 350 µ m observation from Khan et al. (2007). Sourcecounts are differential normalized to a Euclidean universe. (b) The Observed 350 µ m integral source counts from the SHARC-II surveyof the Bootes field by Khan et al. (2007); also plotted are the normalized 450 µ m SCUBA source counts from Smail et al. (2002). (e) Observed 850 µ m integral source counts from the various surveys by SCUBA (Smail et al. (1997), Hughes et al. (1998), Eales et al.(1999), Barger et al. (1998), Blain et al. (1999), Smail et al. (2002), Cowie et al. (2002), Scott et al. (2002), Knudsen et al. (2006)); andthe largest SCUBA survey, SHADES, covering ∼ (f) The SHADES differential counts normalized to aEuclidean universe. (g)
Observed 1100 µ m integral source counts are from the BOLOCAM instrument Black shaded area from Laurentet al. (2005) and the MAMBO survey of Greve et al. (2004) normalized from 1200 µ m. (g) Differential source counts normalized to aEuclidean universe for the AzTEC observations of Perera et al. (2008). at the peak of the differential source counts the less lumi-nous LIRGs are the dominant population. At 250 µ m, themodel slightly over-predicts the source counts at >
60 mJy,but this is within the BLAST error bars. There is a sharprise in the counts at the 200 mJy level, and a turn-over be-tween 100 and 20 mJy (although the BLAST counts in thisregion may be less reliable as the instrument is confusion-limited), with the model predicting a second turn-over atfainter flux densities ( <
10 mJy). Due to the strong nega-tive K-corrections in the submillimetre (Franceschini et al.1991), the flux densities of distant galaxies are enhancedsuch that the luminosity function at lower luminosities issampled at fainter flux densities, with any break in thecounts being attributed to a change in the dominant popu-lation. The BLAST counts are derived from a P(D) analysisrather than source catalogues and provide a statistical con-straint on the slope of the source counts at faint fluxes whichare already source confused. Encouragingly at 350 µ m, thefaintest BLAST counts are consistent with the differentialcounts from the deeper (non-confused) 350 µ m survey usingSHARC II in the Bootes field by Khan et al. (2007). Fig-ure 2 b shows the 350 µ m integral source counts from samesurvey and the SCUBA 450 µ m counts (Smail et al. 2002;assuming an Arp 220 spectral template to transform thecounts to this band). The model fits these observations well,predicting breaks in the source counts at ∼
40 and ∼
10 mJy, and that the deeper SHARC II results are dominated byLIRGs. In the 500 µ m BLAST band, the model fit is excep-tionally good, from the steep rise from Euclidean values at S <
300 mJy, to the turn-over between 30-6 mJy. The modelpredicts turnovers in the counts at fainter flux densities of10, 8, 5 mJy in the 250, 350 & 500 µ m bands, all withinthe constraints imposed by equating the integrated surfacebrightness of the BLAST sources to the emission from theinfrared background derived from a power-law extrapola-tion and naive cut-off of sources estimated by Devlin et al.(2009) to be 7.0 ± ± ± µ m respectively.At longer submillimetre wavelengths, the model fits arecompared with the observed integral source counts from themyriad surveys carried out with SCUBA at 850 µ m (Figure2 e ). These observations span two orders of magnitude in fluxdensity and thus provide the best pre-Herschel constraintson the galaxy counts. The models provide a good fit to thecounts from the brightest flux densities down to 0.5 mJy –below the SCUBA-850 µ m confusion limit of 2mJy (from thelensed surveys of Smail, Ivison & Blain 1997, Smail et al.2002). At these levels, due to the strong negative K-corrections, we expect to be able to observe relatively moder-ate starburst galaxies. At brighter flux densities, ∼
10 mJy,ULIRGs are the dominant population (although a signifi-cant increasing contribution from normal galaxies cannot be c (cid:13) , 1–6 C. Pearson, S. Khan ruled out (Efstathiou & Rowan-Robinson 2003), but fromthe model they are predicted to dominate at > ∼
50 mJy).The largest 850 µ m survey to-date (the ∼ SHADESsurvey Mortier et al. 2005) detected 120 sources, effectivelydoubling the number of known SMGs. The SHADES dif-ferential source counts (Coppin et al. 2006) are shown inFigure 2 (f ) . The best-fitting model requires a break at ∼ ∼ g the integral source counts at millimetrewavelengths for the surveys with the BOLOCAM instru-ment at 1100 µ m (from the maximum likelihood analysisof Laurent et al. (2005)) and the MAMBO instrument at1200 µ m (normalising the counts to 1100 µ m). BOLOCAM,MAMBO & SCUBA have all surveyed the same area – theLockman Hole – and in essence, the counts suggest the mil-limetre observations are sampling the same brighter portion(S > ∼ µ m population, expectedto be dominated by ULIRGs or even HLIRGs (Hyper Lu-minous Infra-Red Galaxies, L IR > L ⊙ ). This is sim-ply due to the longer wavelengths sampling further downthe Rayleigh-Jeans slope and therefore preferentially select-ing the higher luminosity, high redshift objects. Finally, themodel fits to the differential counts from the recent AzTECobservations of Perera (2008) in the GOODS fields are pre-sented in Figure 2 h . There is a good fit to these observedcounts, with the higher luminosity sources providing themain population. The flattening seen in both the integraland differential source counts at fluxes of ∼ ∼ The Herschel Space Observatory (Pilbratt 2008), launchedon 14th May 2009, is ESA’s next generation infrared mission.The Spectral and Photometric Imaging Receiver (SPIRE)instrument is one of the focal plane instruments and is de-signed for photometry and spectroscopy between 200-550 µ m(Griffin et al. 2008). The three SPIRE bolometer arrays(PSW, PMW and PLW, respectively centered on 250, 350 &500 µ m, λ/ ∆ λ ∼
3, with 139, 88, and 43 pixels) allow simulta-neous observations over a FOV of 4 ′ × ′ in the three bands.In SPIRE’s large map scanning mode the 5 σ , 1 hour pointsource sensitivities are expected to be 3.7, 5.3 & 4.6 mJyfor the respective arrays. Despite being near identical to thethree arrays on BLAST, the 3.5m Herschel primary mirror(2m on BLAST) offers superior resolution.The ultimate sensitivity of any survey will be the con-fusion limit, defined as the threshold of fluctuations in thebackground sky brightness caused by (unresolved) pointsources below which sources cannot be discretely detected N u m be r / s qua r e deg r ee Redshift
Figure 3.
The number redshift distribution for SPIRE bands atthe confusion limit of 19.4, 20.5 & 16.1mJy for the PSW 250,PMW 350 & PLW 500 µ m arrays. Redshift bin size is δ z = 0.1. in the telescope beam λ /D, where D is the telescope diam-eter. The confusion due to faint galaxies is more severe atlonger wavelengths and smaller apertures and is often char-acterized by the number of beams per source, with classicallimits of 20-40 beams per source often adopted ( Hogg 2001,Jeong et al. 2006). The confusion limits for Herschel-SPIREand BLAST can therefore be compared using the sourcecount model: for BLAST the 20 beams per source confusionlimit is predicted to be 33.7, 33.6 & 23.9mJy in the 250, 350& 500 µ m bands – agreeing very well with the estimates fromDevlin et al. (2009) of 33 ±
4, 30 ± ± µ m bands respectively,implying that equivalent surveys with Herschel offer a sen-sitivity improvement of 1.5 over BLAST. This advantage issignificant given the steep nature of the source counts inthis flux regime, with the model predicting a shift from theULIRG dominated counts to LIRG domination. Using P(D)analysis , the BLAST source counts are able to probe be-neath the conventional confusion limit, however P(D) anal-ysis is only able to constrain the slope of the source counts.The intrinsically deeper images afforded by Herschel shouldbe able to resolve and reliably sample this emerging popu-lation, which our model predicts will contribute the bulk ofthe Cosmic Infra-Red Background (CIRB).Using a similar P(D) analysis, SPIRE surveys couldalso be pushed well below the corresponding SPIRE con-fusion limits allowing an actual detection of the turn overin the differential counts (note that the BLAST survey canonly place upper limits). However, surveys with SPIRE tothese depths are not expected to resolve the sources re-sponsible for the break in the source counts (eluded to byKhan et al. 2007 at 350 µ m) although this population willbe accessible using ground-based facilities such as SHARC-II/CSO, SCUBA-2/JCMT Holland et al. (2006), CCATSebring et al. (2008). In the deepest SPIRE confusion-limited surveys, we expect 2600, 1300 and 700 sources persquare degree for the 250, 350 & 500 µ m bands respectively,of which ∼ c (cid:13) , 1–6 ubmillimetre surveys: The prospects for Herschel
250 microns350 microns500 microns P e r c en t age F r a c t i on o f C I R B S/mJy
Figure 4.
CIRB fraction as a function of flux for theSPIRE/BLAST bands. solid-lines are the integral and dashed-lines are the differential contribution. Also shown are the confu-sion limited sensitivities for SPIRE circles and BLAST squares . In Figure 3 we show the number redshift distribution forthe three SPIRE bands at a survey sensitivity correspond-ing to the SPIRE confusion limit. In all bands, but mostpredominantly in the short wavelength 250 µ m band, a bi-modal distribution is seen which can be interpreted as a lo-cal contribution from quiescent normal galaxies and a highredshift contribution from evolving starburst galaxies, withthe high redshift peak becoming more pronounced to longerwavelengths. The median redshift of the N-z distribution liesbetween 2 < z <
3, consistent with the redshift distributionof SCUBA-850 µ m sources (Chapman et al. 2003).Using the model to integrate to the faintest flux levels,the total contribution of faint sources to the cosmic infraredbackground in the SPIRE bands is estimated, giving inten-sities of 11.0, 6.0 & 2.4 nW/m /sr in the 250, 350 & 500 µ mbands. These values agree well with the COBE-FIRAS re-sults of 10.4 ± ± ± ∼
80, 85 & 90% of the total back-ground resides at z > µ m respectively. InFigure 4 the integral and differential percentage contribu-tions as a function of flux for the SPIRE and BLAST bandsare shown, alongside the corresponding 20-beam confusionlimits for both instruments. We expect Herschel to resolve ∼
30% (70%), 20% (60%) & 12% (45%) at the confusion(and optimal instrumental 1-hour integration) level in itsPSW-250 µ m, PMW-350 µ m & PLW-500 µ m arrays, with acorresponding peak in the background emission to occuringat flux densities of 10-25 mJy, 8-20 mJy & 5-10 mJy in therespective bands. This implies the very deepest surveys withHerschel-SPIRE should sample the dominant source of thebackground, which from our model is expected to be mostlyLIRGs (rather than more luminous ULIRGs, whose contri-bution peaks at slightly brighter fluxes in all wavebands). Atfaint flux densities ( < We thank Steve Willner for helpful comments and AndreasEfstathiou for providing his galaxy templates. We thank thereferee for constructive comments that improved this work.
REFERENCES
Balland C., Devriendt J.E.G., Silk J., 2003, MNRAS, 343, 107Barger A.J.et al., 1998, Nature, 394, 248Bertoldi F., et al., 2000, A&A, 360, 92Blain A., Kneib J.P., Ivison R.J., Smail I., 1999, ApJ, 512, 87Blain A., Smail I., Ivison R.J., Kneib J.-P., Frayer D.T., 2002,Physics Reports, 369, 111Chapman S.C., Blain A.W., Ivison R.J., Smail I.R., 2003, Na-ture, 422, 695Coppin K. et al., 2006, MNRAS, 372, 162Cowie L.L.; Barger A.J., Kneib J.-P., 2002, ApJ, 603, 69Devlin M. et al., 2009, Nature, 458, 737Dunne L., Eales S., Edmunds M.G., Ivison R., Alexander P.,Clements D. L.,2000, MNRAS, 315, 115Eales S. et al., 1999, ApJ, 515, 518Efstathiou A., Rowan-Robinson M., Siebenmorgen R., 2000,MNRAS, 313, 734Efstathiou A., Rowan-Robinson M., 2003, MNRAS, 343, 322Fixsen, D.J., Dwek, E., Mather, J.C., Bennett, C.L., Shafer,RA., 1998, ApJ,. 508, 123Franceschini A., Toffolatti L., Mazzei P., Danese L., De ZottiG., 1991, AAS 89, 285Guiderdoni, B., Hivon, E., Bouchet, F.R., Maffei B., 1998, MN-RAS, 295, 877Greve T.R., Ivison R.J., Bertoldi F., Stevens J.A., Dunlop J.S.,Lutz D., Carilli, C.L., 2004, MNRAS, 354, 779Griffin M. et al., 2008, Proc. SPIE 7010, 701006Hogg D., 2001, AJ, 121, 1207Holland W. et al., 2006, Proc. SPIE 6275, 62751Hughes D. et al., 1998, Nature, 457, 616Jeong W-S., Pearson C.P., Lee, H.M., Pak, S-J.,Nakagawa T.,2006, MNRAS, 369, 281Khan S. et al., 2005, ApJ, 631, 9Khan S. et al., 2007, ApJ, 665, 973Knudsen K.K. et al., 2006, MNRAS, 368, 487Laurent G.T. et al., 2005, ApJ, 623, 742Magorrian J. et al., 1998, AJ 115, 2285Mobasher B. et al., 2009, ApJ, 690, 1074Mortier A.M.J. et al., 2005, MNRAS, 363, 563Pascale E. et al., 2008, ApJ, 681, 400Pearson C.P. 2005, MNRAS, 358, 1417Pearson, C.P. et al. 2007, Adv.Space Res. 40, 605Pearson C.P. 2009, in preparationPerera E.L. 2008, MNRAS 391, 1227Pilbratt G.L., 2008, Proc. SPIE 7010, 701002Rowan-Robinson M. et al, 2004, MNRAS, 351, 1290Rowan-Robinson M. et al, 2005, ApJ, 129, 1183Rowan-Robinson M. et al, 2009, ApJ, 394, 117Saunders W. et al. 2000, MNRAS, 317, 55Scott S.E. et al., 2002, MNRAS, 331, 838Sebring T.A., Radford S., Giovanelli R., Glenn J., Woody D.,2008, Proc. SPIE, 7012, 70121Serjeant S., Harrison D., 2005, MNRAS, 356, 192Smail I., Ivison R.J.,Blain A.W., 1997, ApJ, 490, L5c (cid:13) , 1–6
C. Pearson, S. Khan
Smail I., Ivison R.J., Blain A.W., Kneib J.-P., 2002, MNRAS,331, 495Soifer, B.T., Neugebauer G., Houck J.R., 1987, ARAA 25, 187Vlahakis C., Dunne L., Eales S., 2005, MNRAS, 364, 1253 c (cid:13)000