The VANDELS ESO public spectroscopic survey
R. J. McLure, L. Pentericci, A. Cimatti, J. S. Dunlop, D. Elbaz, A. Fontana, K. Nandra, R. Amorin, M. Bolzonella, A. Bongiorno, A. C. Carnall, M. Castellano, M. Cirasuolo, O. Cucciati, F. Cullen, S. De Barros, S. L. Finkelstein, F. Fontanot, P. Franzetti, M. Fumana, A. Gargiulo, B. Garilli, L. Guaita, W. G. Hartley, A. Iovino, M. J. Jarvis, S. Juneau, W. Karman, D. Maccagni, F. Marchi, E. Marmol-Queralto, E. Pompei, L. Pozzetti, M. Scodeggio, V. Sommariva, M. Talia, O. Almaini, I. Balestra, S. Bardelli, E. F. Bell, N. Bourne, R. A. A. Bowler, M. Brusa, F. Buitrago, K. I. Caputi, P. Cassata, S. Charlot, A. Citro, G. Cresci, S. Cristiani, E. Curtis-Lake, M. Dickinson, G. G. Fazio, H. C. Ferguson, F. Fiore, M. Franco, J. P. U. Fynbo, A. Galametz, A. Georgakakis, M. Giavalisco, A. Grazian, N. P. Hathi, I. Jung, S. Kim, A. M. Koekemoer, Y. Khusanova, O. Le Fevre, J. M. Lotz, F. Mannucci, D. T. Maltby, K. Matsuoka, D. J. McLeod, H. Mendez-Hernandez, J. Mendez-Abreu, M. Mignoli, M. Moresco, A. Mortlock, M. Nonino, M. Pannella, C. Papovich, P. Popesso, D. P. Rosario, M. Salvato, P. Santini, D. Schaerer, C. Schreiber, D. P. Stark, L. A. M. Tasca, R. Thomas, T. Treu, E. Vanzella, V. Wild, C. C. Williams, G. Zamorani, E. Zucca
MMNRAS , 1–19 (2018) Preprint 15 May 2018 Compiled using MNRAS L A TEX style file v3.0
The VANDELS ESO public spectroscopic survey
R. J. McLure , L. Pentericci , A. Cimatti , , J. S. Dunlop , D. Elbaz , A. Fontana ,K. Nandra , R. Amorin , , M. Bolzonella , A. Bongiorno , A. C. Carnall ,M. Castellano , M. Cirasuolo , O. Cucciati , F. Cullen , S. De Barros ,S. L. Finkelstein , F. Fontanot , P. Franzetti , M. Fumana , A. Gargiulo ,B. Garilli , L. Guaita , , W. G. Hartley , A. Iovino , M. J. Jarvis , S. Juneau ,W. Karman , D. Maccagni , F. Marchi , E. M´armol-Queralt´o , E. Pompei ,L. Pozzetti , M. Scodeggio , V. Sommariva , M. Talia , , O. Almaini , I. Balestra ,S. Bardelli , E. F. Bell , N. Bourne , R. A. A. Bowler , M. Brusa , F. Buitrago , ,K. I. Caputi , P. Cassata , S. Charlot , A. Citro , G. Cresci , S. Cristiani ,E. Curtis-Lake , M. Dickinson , G. G. Fazio , H. C. Ferguson , F. Fiore ,M. Franco , J. P. U. Fynbo , A. Galametz , A. Georgakakis , M. Giavalisco ,A. Grazian , N. P. Hathi , I. Jung , S. Kim , A. M. Koekemoer , Y. Khusanova ,O. Le F`evre , J. M. Lotz , F. Mannucci , D. T. Maltby , K. Matsuoka ,D. J. McLeod , H. Mendez-Hernandez , J. Mendez-Abreu , , M. Mignoli ,M. Moresco , , A. Mortlock , M. Nonino , M. Pannella , C. Papovich , P. Popesso ,D. P. Rosario , M. Salvato , , P. Santini , D. Schaerer , C. Schreiber , D. P. Stark ,L. A. M. Tasca , R. Thomas , T. Treu , E. Vanzella , V. Wild , C. C. Williams ,G. Zamorani , E. Zucca Affiliations are listed at the end of the paper
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
VANDELS is a uniquely-deep spectroscopic survey of high-redshift galaxies with theVIMOS spectrograph on ESO’s Very Large Telescope (VLT). The survey has obtainedultra-deep optical ( . < λ < . µ m) spectroscopy of (cid:39) . ≤ z ≤ . , over a total area of (cid:39) . deg centred on the CANDELSUDS and CDFS fields. Based on accurate photometric redshift pre-selection, 85% ofthe galaxies targeted by VANDELS were selected to be at z ≥ . Exploiting the redsensitivity of the refurbished VIMOS spectrograph, the fundamental aim of the surveyis to provide the high signal-to-noise ratio spectra necessary to measure key physicalproperties such as stellar population ages, masses, metallicities and outflow velocitiesfrom detailed absorption-line studies. Using integration times calculated to produce anapproximately constant signal-to-noise ratio ( < t int < hours), the VANDELSsurvey targeted: a) bright star-forming galaxies at . ≤ z ≤ . , b) massive quiescentgalaxies at . ≤ z ≤ . , c) fainter star-forming galaxies at . ≤ z ≤ . and d) X-ray/
Spitzer- selected active galactic nuclei and
Herschel -detected galaxies. By targetingtwo extragalactic survey fields with superb multi-wavelength imaging data, VANDELSwill produce a unique legacy data set for exploring the physics underpinning high-redshift galaxy evolution. In this paper we provide an overview of the VANDELSsurvey designed to support the science exploitation of the first ESO public data release,focusing on the scientific motivation, survey design and target selection.
Key words: surveys – galaxies: high-redshift – galaxies: evolution – galaxies: starformation © a r X i v : . [ a s t r o - ph . GA ] M a y R. J. McLure et al.
Understanding the formation and evolution of galaxies re-mains the key goal of extra-galactic astronomy. However, de-lineating the evolution of galaxies, from the collapse of thefirst gas clouds at early times to the assembly of the com-plex structure we observe in the local Universe, continuesto present an immense observational (e.g. Madau & Dickin-son 2014) and theoretical challenge (e.g. Somerville & Dav´e2015; Knebe et al. 2015).From an observational perspective, the last fifteen yearshave been a period of unprecedented progress in our under-standing of the basic demographics of high-redshift galax-ies. As a direct consequence of the availability of deep,multi-wavelength, survey fields, we now have a good work-ing knowledge of how the galaxy luminosity function (e.g.McLure et al. 2013b; Bowler et al. 2015; Finkelstein 2016;Mortlock et al. 2017), stellar mass function (e.g. Muzzinet al. 2013; Tomczak et al. 2014; Davidzon et al. 2017) andglobal star-formation rate density (SFRD) evolve with red-shift (e.g. Magnelli et al. 2013; Novak et al. 2017). Indeed,Madau & Dickinson (2014) recently demonstrated the con-sistency (within a factor of ∼ ) between the integral ofcurrent SFRD determinations and direct estimates of theevolution of stellar-mass density.As a consequence, we can now be confident that the lowSFRD we observe locally is approximately the same as it waswhen the Universe was less than 1 Gyr old (i.e. z (cid:39) ), andthat in the intervening period the Universe was forming starsup to ≥ times more rapidly. However, despite this, it isstill perfectly plausible to argue that the peak in cosmic star-formation occurred anywhere in the redshift interval . < z < . , an uncertainty of 2.5 Gyr. Moreover, the results ofthe latest generation of semi-analytic and hydro-dynamicalgalaxy simulations (e.g. Genel et al. 2014; Henriques et al.2015; Somerville & Dav´e 2015) demonstrate that, from atheoretical perspective, even reproducing the evolution ofthe cosmic SFRD can still be problematic.Over the last decade it has become established thatthe majority of cosmic star formation is produced by galax-ies lying on the so-called ‘main sequence’ of star formation(Noeske et al. 2007; Elbaz et al. 2007; Daddi et al. 2007),a roughly linear relationship between star-formation rate(SFR) and stellar mass, the normalisation of which increaseswith look-back time. Furthermore, the evolving normalisa-tion of the main sequence over the last 10 Gyr is now rel-atively well determined, with the average SFR at a givenstellar mass increasing by a factor of (cid:39) between the lo-cal Universe and redshift z (cid:39) (e.g. Whitaker et al. 2014;Speagle et al. 2014; Johnston et al. 2015). However, at higherredshifts the evolution of the main sequence is still uncertain,despite a clear theoretical prediction that it should mirrorthe increase in halo gas accretion rates (i.e. ∝ ( + z ) . ; Dekelet al. 2009). Depending on their assumptions regarding star-formation histories, metallicity, dust and nebular emission,different studies find that the increase in average SFR be-tween z = and z = at a given stellar mass is anything froma factor of (cid:39) (e.g. Gonz´alez et al. 2014; M´armol-Queralt´oet al. 2016) to a factor of (cid:39) (e.g. de Barros et al. 2014);see Stark (2016) for a recent review.Although the decline of the global SFRD at z ≤ is nowwell characterised observationally, the relative importance of the different physical drivers responsible for the quenchingof star formation remains unclear. With varying degrees ofhard evidence and speculation, feedback from active galac-tic nuclei (AGN), stellar winds, merging and environmen-tal/mass driven quenching have all been widely discussedin the recent literature (e.g. Fabian 2012; Conselice 2014;Peng et al. 2015). At some level, quenching must be con-nected to the interplay between gas outflow, the inflow of‘pristine’ gas and morphological transformation. However,to date, the precise roles played by the different underlyingphysical mechanisms still remain uncertain, as does the po-tential redshift evolution of the quenching process. Indeed,recent evidence based on deep optical and near-IR spec-troscopy strongly suggests that the physical properties ofstar-forming galaxies at z = − are significantly differentfrom their low-redshift counterparts in terms of metallic-ity, α − enhancement and ionization parameter (e.g. Cullenet al. 2014; Shapley et al. 2015; Steidel et al. 2016; Cullenet al. 2016; Strom et al. 2017). Moreover, recent results atsub-mm and mm-wavelengths with Herschel and ALMA in-dicate that the dust properties of star-forming galaxies athigh redshift may also be significantly different (e.g. Capaket al. 2015; Bouwens et al. 2016; Reddy et al. 2018), althoughthe current picture is far from clear (e.g. Dunlop et al. 2017;Bourne et al. 2017; McLure et al. 2017; Koprowski et al.2018; Bowler et al. 2018).In summary, it now appears that progress in our un-derstanding of galaxy evolution at high redshift is often lesslimited by poor statistics than by the systematic uncertain-ties in our measurements of the crucial physical parameters,caused by the insidious and interrelated degeneracies be-tween age, dust attenuation and metallicity. It is also clearthat substantive progress in addressing these uncertaintieswill rely on combining the best available multi-wavelengthimaging with deep spectroscopy (e.g. Kurk et al. 2013).Within this context, a series of spectroscopic campaignswith VLT+VIMOS, such as the VIMOS Very Deep Survey(VVDS; Le F`evre et al. 2005), the COSMOS spectroscopicsurvey (zCOSMOS; Lilly et al. 2007) and the VIMOS UltraDeep Survey (VUDS; Le F`evre et al. 2015), have played akey role in improving our understanding of galaxy evolution,primarily through providing large numbers of spectroscopicredshifts over wide fields. The VANDELS survey is designedto complement and extend the work of these previous cam-paigns by focusing on ultra-long exposures of a relativelysmall number of galaxies, pre-selected to lie at high redshiftusing the best available photometric redshift information.The VANDELS survey is a major new ESO Public Spec-troscopic Survey using the VIMOS spectrograph on the VLTto obtain ultra-deep, medium resolution, red-optical spectraof (cid:39) high-redshift galaxies. The survey was allocated914 hours of VIMOS integration time and, between Au-gust 2015 and February 2018, each target galaxy received20–80 hours of on-source integration, obtained via repeatedobservations of the UDS and CDFS multi-wavelength sur-vey fields. The fundamental science goal of VANDELS isto move beyond redshift acquisition and obtain a spec-troscopic data set deep enough to study the astrophysicsof high-redshift galaxy evolution. The VANDELS spectro-scopic targets were all pre-selected using high-quality pho-tometric redshifts, with the vast majority ( (cid:39) ) drawnfrom three main categories. Firstly, VANDELS targeted
MNRAS , 1–19 (2018) he VANDELS spectroscopic survey bright ( i AB ≤ ) star-forming galaxies in the redshift range . ≤ z ≤ . (median z = . ). For these galaxies, thesignal-to-noise ratio (SNR) and wavelength coverage of theVANDELS spectra are designed to allow stellar metallic-ity and gas outflow information to be extracted for indi-vidual objects. Secondly, to study the descendants of high-redshift star-forming galaxies, VANDELS targeted a com-plementary sample of massive ( H AB ≤ . ) passive galaxiesat . ≤ z ≤ . (median z = . ). Again, in combination withdeep multi-wavelength photometry and 3D-HST grism spec-troscopy (Brammer et al. 2012), the high SNR spectra pro-vided by VANDELS are designed to provide age/metallicityinformation and star-formation history constraints for indi-vidual objects. Thirdly, VANDELS extended to fainter mag-nitudes and higher redshifts by targeting a large statisti-cal sample of faint star-forming galaxies ( ≤ H AB ≤ , i AB ≤ . ) in the redshift range ≤ z ≤ (median z = . ).Throughout the rest of the paper we will refer to the galax-ies in this sample as Lyman-break galaxies (LBGs), althoughthey were not selected via traditional colour-colour criteria(see Section 4). The final (cid:39) of VANDELS spectroscopicslits were allocated to AGN candidates or Herschel- detectedgalaxies with i AB ≤ . and z ≥ . (median z = . ).In this paper we provide an overview of the VAN-DELS survey to support the science exploitation of the firstdata release (DR1) via the ESO Science Archive Facility( archive.eso.org ). The structure of the paper is as fol-lows. In Section 2 we provide a brief review of the sciencecases that provided the principal motivation for VANDELS,along with the multiple legacy science cases which could befacilitated by the data. In Section 3 we describe the rea-soning behind the choice of survey fields. In Section 4 wedescribe the target selection process, including the gener-ation of photometric catalogues and the determination ofrobust photometric redshifts. In Section 5 we describe thebasic observing strategy before providing brief details of thedata reduction and spectroscopic redshift measurement pro-cedures in Section 6. In Section 7 we describe the contentsof the first data release, before reviewing the success of theVANDELS target selection process using the on-sky DR1data in Section 8. A full description of DR1, including a de-tailed discussion of the observing strategy, data reductionand spectroscopic redshift measurements is provided in acompanion data release paper (Pentericci et al. 2018). InSection 9 we provide a summary and an overview of thecontent and timeline for subsequent data releases. Through-out the paper we refer to total magnitudes quoted in theAB system (Oke & Gunn 1983). We assume the followingcosmology: Ω M = . , Ω Λ = . and H = km s − Mpc − ,and adopt a Chabrier (2003) initial mass function (IMF) forcalculating stellar masses and star-formation rates. The primary motivation behind the VANDELS survey wasto provide spectra of high-redshift galaxies with sufficientlyhigh SNR to allow absorption line studies both on individ-ual objects and via stacking. Armed with spectra of sufficientquality it should be possible, in combination with excellentmulti-wavelength photometry, to provide significantly im-proved constraints on key physical parameters such as stellar mass, star-formation rate, metallicity and dust attenuation.As a result, it is clear that the data set provided by VAN-DELS will have a potentially significant impact on manydifferent areas of high-redshift galaxy evolution science. Inthis section we provide a concise overview of the key sci-ence goals that motivated the original VANDELS surveyproposal, before briefly reviewing the legacy science case.
Tracing the evolution of metallicity is a powerful methodof constraining high-redshift galaxy evolution, due to its di-rect link to past star formation and sensitivity to interac-tion (i.e. gas inflow/outflow) with the inter-galactic medium(e.g. Mannucci et al. 2010). Moreover, accurate knowledge ofmetallicity is essential for deriving accurate star-formationrates and breaking the degeneracy between age and dust at-tenuation (e.g. Rogers et al. 2014). Consequently, it is clearthat extracting constraints on the metallicity and dust at-tenuation of high-redshift galaxies from VANDELS spectrais important to investigations of the build-up of the stellar mass-metallicity relation, accurately quantifying the peakin cosmic star-formation history (e.g. Castellano et al. 2014;Dunlop et al. 2017), and resolving the current uncertaintiesregarding the evolution of sSFR at z ≥ (e.g. Stark 2016).Recent studies using stacked spectra of relatively smallsamples (e.g. Steidel et al. 2016) have shown that is possibleto derive accurate stellar metallicities from the rest-frameUV spectra of galaxies at z ≥ , given a sufficiently highSNR. In addition, Steidel et al. (2016) also demonstratedthat rest-frame UV spectra can potentially be used to quan-tify the impact of binary stars in stellar population synthesismodels (e.g. Stanway et al. 2016; Eldridge & Stanway 2016)by fitting to the He ii emission line at ˚A.The high SNR and accurate flux calibration of the VAN-DELS spectra facilitates the measurement of stellar metallic-ities using photospheric UV absorption lines ( − ˚A),whose equivalent width is sensitive to metallicity and inde-pendent of other stellar parameters (e.g. Sommariva et al.2012; Rix et al. 2004). Moreover, within the context of dustattenuation, the VANDELS data set also has the potentialto differentiate between competing dust reddening laws (e.g.Cullen et al. 2017; McLure et al. 2017), and to constrain thestrength of the 2175˚A bump.The final VANDELS data set will provide individualand stacked measurements of stellar metallicity based on > ∼ spectroscopically-confirmed star-forming galaxies inthe redshift range . ≤ z ≤ . . These measurements can becompared with the gas-phase metallicities currently beingderived for z (cid:39) . galaxies by the MOSDEF (Shapley et al.2015) and KBSS-MOSFIRE (Strom et al. 2017) surveys andforthcoming observations with the James Webb Space Tele-scope ( JWST ). Along with stellar-metallicity measurements, a key sciencegoal for VANDELS is to investigate the role of stellar andAGN feedback in quenching star formation at high redshiftvia studies of outflowing interstellar gas. Over recent yearsit has become established that high-velocity outflows are
MNRAS000
MNRAS000 , 1–19 (2018)
R. J. McLure et al. likely to be ubiquitous for star forming galaxies at z > (e.g. Weiner et al. 2009), with mass outflow rates comparableto the rates of star formation (e.g. Bradshaw et al. 2013),and that very compact starbursts can produce outflows withvelocities > km s − , yielding winds that were previouslyonly thought possible from AGN activity (Diamond-Stanicet al. 2012). It seems likely that such outflows are playinga major role in the termination of star formation at highredshift and the build-up of the mass-metallicity relation.The individual and stacked spectra of star-forminggalaxies delivered by VANDELS will provide accurate mea-surements of outflowing ISM velocities from high and low-ionization UV interstellar absorption features (e.g. Shapleyet al. 2003), allowing the outflow rate to be investigated as afunction of stellar mass, SFR and galaxy morphology. Thisoffers the prospect of improving our understanding of the im-pact of galactic outflows on star-formation at z ≥ , directlytesting models of the evolving gas reservoir (e.g. Dayal et al.2013) and addressing the origins of the Fundamental Mass-Metallicity Relation (Mannucci et al. 2010). Finally, compar-ing the outflow velocities of star-forming galaxies with andwithout hidden AGN (e.g. Talia et al. 2017) will allow therole of AGN feedback in quenching star formation and thebuild-up of the red sequence to be investigated (e.g. Cimattiet al. 2013). A key sub-component of the VANDELS survey was obtain-ing deep spectroscopy of > massive, passive galaxies at . ≤ z ≤ . . This population holds the key to understand-ing the quenching mechanisms responsible for producing thestrong colour bi-modality observed at z < , together withthe significant evolution in the number density, morphologyand size of passive galaxies observed between z = and thepresent day (e.g. Bruce et al. 2012; McLure et al. 2013a;Tomczak et al. 2014; van der Wel et al. 2014). The physicalparameters which will be delivered by the VANDELS spec-tra offer the prospect of connecting these quenched galax-ies with their star-forming progenitors at z ≥ in a self-consistent way.For the majority of the passive sub-sample, theVANDELS spectra provide a combination of crucialrest-frame UV absorption-line information (e.g. MgUV,2640˚A/2900˚A breaks) and Balmer-break measurements.Combined with the unrivalled photometric data availablein the UDS and CDFS fields, it will be possible to breakage/dust/metallicity degeneracies and deliver accurate stel-lar mass, dynamical mass, star-formation rate, metallicityand age measurements via full spectrophotometric SED fit-ting (e.g. McLure et al. 2013a; Chevallard & Charlot 2016;Carnall et al. 2017). Although the science cases outlined above provided the pri-mary motivation, as an ESO public spectroscopy survey,the greatest strength of VANDELS is arguably its long-term legacy value to the astronomical community. In gen-eral terms, by providing high SNR continuum spectroscopyof galaxies which traditionally only have Ly α redshifts at best, VANDELS is guaranteed to open up new parame-ter space for investigating the physical properties of high-redshift galaxies.More specifically, the VANDELS spectra provide theopportunity to accurately determine the fraction of Ly α emitters amongst the general Lyman-break galaxy popula-tion in the redshift range . < z < . , thereby provid-ing an improved baseline measurement for studies withinthe reionization epoch (e.g. Curtis-Lake et al. 2012; Penter-icci et al. 2014; De Barros et al. 2017). In addition, VAN-DELS will also provide large samples of spectroscopically-confirmed galaxies at z (cid:39) with which to identify and studyLyman continuum emitters (e.g. Vanzella et al. 2016; de Bar-ros et al. 2016; Shapley et al. 2016; Marchi et al. 2017).Moreover, combining the VANDELS spectra with near-IRspectroscopy offers the prospect of directly comparing stel-lar and gas-phase metallicities out to z (cid:39) . , and constrain-ing the possible star-formation timescales via quantifyingthe level of α − enhancement (e.g. Steidel et al. 2016) as afunction of stellar mass and star-formation rate. We alsonote that additional science will be facilitated by the sam-ples of rarer Herschel- detected galaxies and AGN targetedby VANDELS. For these systems, the deep VANDELS spec-troscopy will make it possible to assess their physical con-ditions (e.g. metallicities, ionizing fluxes and outflow signa-tures) and compare them with those of less active systemsat the same redshifts.In terms of future follow-up observations, there is an ex-cellent synergy between VANDELS and the expected launchdate of the
JWST in 2020. The opportunity to combineultra-deep optical spectroscopy with the unparalleled near-IR spectroscopic capabilities of NIRSpec will make VAN-DELS sources an obvious choice for follow-up spectroscopywith
JWST . For high multi-plex follow-up observations,there is also an excellent overlap between the footprint ofthe VANDELS survey within the UDS and CDFS fields andthe the field of view of ESO’s forthcoming Multi Object Op-tical and Near-infrared Spectrograph (MOONS) for the VLT(Cirasuolo et al. 2014).Finally, it is also worth noting that the declinations ofthe UDS and CDFS fields make them ideal for sub-mm andmm follow-up observations with ALMA. One of the key sci-entific questions that VANDELS will help to address is theevolution of star formation and metallicity in galaxies at z ≥ . However, in order to derive a complete picture itwill be necessary to obtain dust mass and star-formationrate measurements at long wavelengths, which can now beprovided by short, targeted, continuum observations withALMA. The VANDELS survey targets two fields, the UKIDSS UltraDeep Survey (UDS: 02:17:38, − − MNRAS , 1–19 (2018) he VANDELS spectroscopic survey N E
Figure 1.
Layout of the eight VANDELS pointings, four in UDS and four in CDFS. In each figure the VIMOS quadrants of a givenpointing are shown as a different colour, overlaid on a greyscale image showing the
HST H − band imaging provided by the CANDELSsurvey (Grogin et al. 2011; Koekemoer et al. 2011) in the central regions and ground-based H − band imaging from the UKIDSS UDS(Almaini et al., in preparation) and VISTA VIDEO (Jarvis et al. 2013) surveys covering the wider fields. The total area covered by theeight VIMOS pointings is (cid:39) . square degrees. The spectroscopic slits are all placed E-W on the sky, as recommended to minimise slitlosses during long VIMOS integrations on fields at these declinations (S´anchez-Janssen et al. 2014). Both the UDS and CDFS offer deep optical-nearIR
HST imaging provided by the CANDELS survey (Grogin et al.2011; Koekemoer et al. 2011) with the CDFS also offeringdeep
HST /ACS optical imaging from the original GOODSsurvey (Giavalisco et al. 2004) and ultra-deep X-ray imaging(Luo et al. 2017). Moreover, both fields feature the deep-est available
Spitzer
IRAC imaging on these angular scalesfrom the S-CANDELS survey (Ashby et al. 2015) and deepWFC3/IR grism spectroscopy from the public 3D-HST pro-gramme (Brammer et al. 2012). When combined with thedeepest available Y + K imaging from the HUGS survey(Fontana et al. 2014), it is clear that the UDS and CDFSare excellent legacy fields for studying the high-redshift Uni-verse.Given that a single pointing of the VIMOS spectrographcovers an area larger than the HST imaging in any of thefive CANDELS fields, another important consideration whenchoosing which fields to target with VANDELS was the qual- ity of the ancillary data over a wider area. The importanceof the wider-field ancillary data can be seen from Fig. 1,which shows the layout of the eight VIMOS pointings tar-geted by the VANDELS survey in UDS and CDFS. It canbe seen that, although the VIMOS pointings are arranged toensure that all of the deep WFC3/IR imaging is covered, ap-proximately 50% of the full VANDELS survey footprint liesoutside the central areas of the UDS and CDFS fields thatare covered by
HST imaging. Crucially, in both the UDS andCDFS, these wider-field regions are covered by high-quality,publicly-available, optical-nearIR imaging data from a widevariety of different ground-based telescopes (see Table 1).
The ideal situation when selecting targets for a spectroscopicsurvey is to utilise a single photometric catalogue that pro-vides consistent photometry with uniform wavelength cov-
MNRAS000
MNRAS000 , 1–19 (2018)
R. J. McLure et al.
Table 1.
Details of the imaging data included in the new photo-metric catalogues generated for the wide-field areas of the CDFSand UDS fields. Column 1 lists the field, column 2 lists the fil-ters, column three lists the median 5 σ depths measured withina (cid:48)(cid:48) − diameter aperture, column 4 lists the telescopes on whichthe imaging was obtained and column 5 lists the paper where thedata are presented. For the two filters tagged with a † in column2, the σ depth refers to the depth measured after the HST imag-ing was convolved to match the 1.0 (cid:48)(cid:48)
FWHM spatial resolution ofthe ground-based imaging in CDFS. The filters listed as ‘IA’ incolumn 2 are medium-band filters and NB921 is a narrow-bandfilter. The two z − band filters listed for the UDS field (z (cid:48) and z (cid:48) )refer to imaging obtained with the Suprime-Cam z (cid:48) − filter beforeand after the CCD detectors were upgraded. The references listedin column 5 correspond to: (1) Almaini et al., in preparation, (2)Furusawa et al. (2008), (3) Furusawa et al. (2016), (4) Sobralet al. (2012), (5) Jarvis et al. (2013), (6) Nonino et al. (2009), (7)Cardamone et al. (2010), (8) Rix et al. (2004), (9) Hsieh et al.(2012).Field Filter Depth( σ ) Telescope ReferenceUDS U 27.0 CFHT 1B 27.8 Subaru 2V 27.4 Subaru 2R 27.2 Subaru 2i (cid:48) (cid:48) (cid:48) † HST † HST
8Y 24.5 VISTA 5J 24.7 CFHT 9H 23.8 VISTA 5K 24.1 CFHT 9 erage over the full survey area. Unfortunately, this was notpossible when performing target selection for the VANDELSsurvey for two fundamental reasons. Firstly, given that VAN-DELS targeted two separate survey fields, covered by differ-ent sets of imaging data, it is clear that target selection hadto be performed using a minimum of two independent pho-tometric catalogues.Secondly, as described above, the footprint of the VAN-DELS survey within the UDS and CDFS fields covers boththe central areas with deep
HST imaging and the wider-field areas covered primarily by ground-based imaging (seeFig. 1). As a result, the VANDELS survey area is effec- tively divided into four regions: UDS-HST, UDS-GROUND,CDFS-HST and CDFS-GROUND, each of which requireda separate photometric catalogue. Consequently, the firststage in the target selection process was the adoption orproduction of robust photometric catalogues for each of thefour regions.
Within the two regions covered by the WFC3/IR imag-ing provided by the CANDELS survey (UDS-HST andCDFS-HST), we adopted the H − band selected photomet-ric catalogues produced by the CANDELS team (Galametzet al. 2013; Guo et al. 2013). Both catalogues pro-vide PSF-homogenised photometry for the available ACSand WFC3/IR imaging, in addition to spatial-resolutionmatched photometry from Spitzer
IRAC and key ground-based imaging data sets derived using the tfit softwarepackage (Laidler et al. 2007). We refer the reader toGalametz et al. (2013) and Guo et al. (2013) for full detailsof the production of these photometric catalogues for theCANDELS UDS and CANDELS CDFS fields, respectively.Within the wider-field areas there were no publiclyavailable, near-IR selected, photometric catalogues whichmet our target selection requirements. As a result, newmulti-wavelength photometric catalogues were generated us-ing the publicly available imaging. The imaging in both theUDS and CDFS fields was initially accurately registered andplaced on the same pixel scale and photometric zero-point.The imaging in the CDFS field had seeing which variedwithin the range . − . (cid:48)(cid:48) FWHM. As a result, it was nec-essary to PSF-homogenise the images to a common spatialresolution of . (cid:48)(cid:48) FWHM using Gaussian convolution ker-nels. The imaging in the UDS field had a much narrowerrange of seeing ( . ± . (cid:48)(cid:48) FWHM), meaning that PSF-homogenisation was not necessary.Following this initial processing, the photometric cat-alogues were generated with sextractor v2.8.6 (Bertin& Arnouts 1996) in dual-image mode, using the H − bandimages as the detection images. Object photometry wasmeasured within (cid:48)(cid:48) − diameter circular apertures, with ac-curate errors calculated on an object-by-object basis usingthe aperture-to-aperture variance between local blank-skyapertures (see Mortlock et al. 2017 for full details).In Table 1 we provide details of the imaging data in-corporated within the new photometric catalogues for theUDS-GROUND and CDFS-GROUND regions. All of thedepths listed in Table 1 refer to the data that were pub-licly available and included in the target selection cataloguesin summer 2015. We note that, since that date, many ofthe near-IR data sets have increased in depth significantly,particularly within the extended CDFS field. Therefore, toaccompany the final data release of the VANDELS survey,we are committed to publicly releasing updated photomet-ric catalogues, including deeper data where available, alongwith photometric redshifts and stellar-population parame-ters derived via SED fitting. A key element of the VANDELS survey strategy was the useof robust photometric-redshift pre-selection. For this pro-
MNRAS , 1–19 (2018) he VANDELS spectroscopic survey z spec z pho t Field: UDS-GROUNDoutlier rate = 1.4% σ dz = . z spec z pho t Field: CDFS-GROUNDoutlier rate = 2.4% σ dz = . z spec z pho t Field: CANDELS UDS+CDFSoutlier rate = 2.2% σ dz = 0.023bias = 0.007 Figure 2.
Top: photometric redshifts derived by the VANDELSteam compared to robust spectroscopic redshifts in the wide-arearegion of the UDS (red data-points are catastrophic outliers with | dz | > . ). Middle: equivalent plot for the wide-area region of theCDFS. Bottom: photometric redshift versus spectroscopic redshiftfor those objects in the top two panels for which photometricredshifts derived by the CANDELS survey team were available(see text for more details). The catastrophic outlier fraction, σ dz and bias are displayed in the top-left corner of each panel. cess to be successful it was of paramount importance to ei-ther adopt or derive photometric redshifts of equal qualitywithin all four of the VANDELS regions. For the two re-gions covered by deep HST near-IR imaging (UDS-HST andCDFS-HST), we adopted the photometric redshifts madepublicly available by the CANDELS survey team (Santiniet al. 2015). As discussed in Dahlen et al. (2013), thesephotometric redshifts are derived by optimally combiningthe independent estimates produced by a variety of differ-ent photometric-redshift codes.For the wider-area regions outside of the CANDELSWFC3/IR imaging footprint, new photometric redshiftswere generated within the VANDELS team, based on thenew UDS-GROUND and CDFS-GROUND photometric cat-alogues. These photometric redshifts were derived by takingthe median value of z phot for each galaxy, based on a total offourteen different photometric redshift estimates derived bydifferent members of the VANDELS team. The fourteen dif-ferent photometric redshift estimates were produced usinga variety of different publicly-available codes (e.g. Arnoutset al. 1999; Bolzonella et al. 2000; Ilbert et al. 2006; Bram-mer et al. 2008; Feldmann et al. 2006) and in-house software(e.g. Fontana et al. 2000; McLure et al. 2011), using a widevariety of different SED templates, star-formation histories,metallicities and emission-line prescriptions.In order to optimise their respective photometric-redshift codes, each member of the VANDELS team takingpart in the photometric-redshift exercise was initially allo-cated a spectroscopic training set for the UDS-GROUNDand CDFS-GROUND regions. Each training set consistedof approximately one thousand high-quality spectroscopicredshifts, and were used by each team member to optimisethe performance of their code. The second step in the pro-cess was to allocate spectroscopic validation sets to eachmember of the photometric-redshift team. The spectroscopicvalidation sets were identical in size and quality to the cor-responding training sets, the only difference being that thespectroscopic redshifts were not disclosed to the team mem-bers. The accuracy of the results on these blind validationsets was used to ensure that each set of photometric-redshiftestimates was adding useful information to the overall result.For the UDS-GROUND region the robust spectroscopic red-shifts used for training and validation purposes were drawnfrom the VIPERS survey (Guzzo et al. 2014a), the 3D-HST survey (Momcheva et al. 2016) and the UDSz survey(Almaini et al., in preparation). For the CDFS-GROUNDregion the robust spectroscopic redshifts were drawn fromthe large number of spectroscopic redshift campaigns previ-ously undertaken within the field (e.g. Le F`evre et al. 2005;Mignoli et al. 2005; Vanzella et al. 2008; Balestra et al. 2010;Cooper et al. 2012; Le F`evre et al. 2013; Momcheva et al.2016).To quantify the quality of the photometric redshift es-timates we calculate three statistics. To quantify any sys-tematic off-set between the photometric and spectroscopicredshifts we calculate the bias, which we define as the medianvalue of dz = ( z spec − z phot ) / ( + z spec ) . Secondly, to quan-tify the accuracy of the photometric redshifts, we calculate σ dz using the robust median absolute deviation (MAD) esti-mator. Finally, we also calculate the fraction of catastrophicoutliers, where an object is considered to be a catastrophicoutlier if | dz | > . . Based on the spectroscopic validation MNRAS000
Top: photometric redshifts derived by the VANDELSteam compared to robust spectroscopic redshifts in the wide-arearegion of the UDS (red data-points are catastrophic outliers with | dz | > . ). Middle: equivalent plot for the wide-area region of theCDFS. Bottom: photometric redshift versus spectroscopic redshiftfor those objects in the top two panels for which photometricredshifts derived by the CANDELS survey team were available(see text for more details). The catastrophic outlier fraction, σ dz and bias are displayed in the top-left corner of each panel. cess to be successful it was of paramount importance to ei-ther adopt or derive photometric redshifts of equal qualitywithin all four of the VANDELS regions. For the two re-gions covered by deep HST near-IR imaging (UDS-HST andCDFS-HST), we adopted the photometric redshifts madepublicly available by the CANDELS survey team (Santiniet al. 2015). As discussed in Dahlen et al. (2013), thesephotometric redshifts are derived by optimally combiningthe independent estimates produced by a variety of differ-ent photometric-redshift codes.For the wider-area regions outside of the CANDELSWFC3/IR imaging footprint, new photometric redshiftswere generated within the VANDELS team, based on thenew UDS-GROUND and CDFS-GROUND photometric cat-alogues. These photometric redshifts were derived by takingthe median value of z phot for each galaxy, based on a total offourteen different photometric redshift estimates derived bydifferent members of the VANDELS team. The fourteen dif-ferent photometric redshift estimates were produced usinga variety of different publicly-available codes (e.g. Arnoutset al. 1999; Bolzonella et al. 2000; Ilbert et al. 2006; Bram-mer et al. 2008; Feldmann et al. 2006) and in-house software(e.g. Fontana et al. 2000; McLure et al. 2011), using a widevariety of different SED templates, star-formation histories,metallicities and emission-line prescriptions.In order to optimise their respective photometric-redshift codes, each member of the VANDELS team takingpart in the photometric-redshift exercise was initially allo-cated a spectroscopic training set for the UDS-GROUNDand CDFS-GROUND regions. Each training set consistedof approximately one thousand high-quality spectroscopicredshifts, and were used by each team member to optimisethe performance of their code. The second step in the pro-cess was to allocate spectroscopic validation sets to eachmember of the photometric-redshift team. The spectroscopicvalidation sets were identical in size and quality to the cor-responding training sets, the only difference being that thespectroscopic redshifts were not disclosed to the team mem-bers. The accuracy of the results on these blind validationsets was used to ensure that each set of photometric-redshiftestimates was adding useful information to the overall result.For the UDS-GROUND region the robust spectroscopic red-shifts used for training and validation purposes were drawnfrom the VIPERS survey (Guzzo et al. 2014a), the 3D-HST survey (Momcheva et al. 2016) and the UDSz survey(Almaini et al., in preparation). For the CDFS-GROUNDregion the robust spectroscopic redshifts were drawn fromthe large number of spectroscopic redshift campaigns previ-ously undertaken within the field (e.g. Le F`evre et al. 2005;Mignoli et al. 2005; Vanzella et al. 2008; Balestra et al. 2010;Cooper et al. 2012; Le F`evre et al. 2013; Momcheva et al.2016).To quantify the quality of the photometric redshift es-timates we calculate three statistics. To quantify any sys-tematic off-set between the photometric and spectroscopicredshifts we calculate the bias, which we define as the medianvalue of dz = ( z spec − z phot ) / ( + z spec ) . Secondly, to quan-tify the accuracy of the photometric redshifts, we calculate σ dz using the robust median absolute deviation (MAD) esti-mator. Finally, we also calculate the fraction of catastrophicoutliers, where an object is considered to be a catastrophicoutlier if | dz | > . . Based on the spectroscopic validation MNRAS000 , 1–19 (2018)
R. J. McLure et al. sets, the fourteen individual photometric-redshift runs pro-duced bias values in the range . − . , values of σ dz inthe range . − . and catastrophic outlier rates between2% and 16%. The equivalent statistics for the adopted me-dian combined z phot results are bias= 0.008, σ dz = . anda catastrophic outlier rate of 1.9%. Compared to the best-performing individual photometric redshift run, the processof median combination has produced a improvementin both σ dz and the catastrophic outlier fraction, with thesame level of bias. In Fig. 2 we show the accuracy of thefinal photometric redshifts adopted for the wider-area UDS-GROUND and CDFS-GROUND regions, based on the spec-troscopic validation sets.Within the final spectroscopic validation sets used todefine the accuracy of the VANDELS photometric redshifts,44% of the galaxies also had photometric redshifts deter-mined by the CANDELS team. As a result, it was possi-ble to perform a useful comparison of the quality of ournew photometric redshifts, based on the photometric datalisted in Table 1, and the photometric redshifts derivedby the CANDELS survey team based on a combination ofdeep HST imaging, ground-based imaging and
Spitzer
IRACimaging. For the objects in common, the VANDELS pho-tometric redshifts have a catastrophic outlier rate of 2.0%and σ dz = . , virtually identical to the statistics for thefull validation sets. The equivalent statistics for the CAN-DELS photometric redshifts are an outlier rate of 2.2% and σ dz = . (see bottom panel of Fig. 2). The results of thiscomparison suggest that the VANDELS photometric red-shifts are slightly more accurate that the photometric red-shifts derived by the CANDELS survey team.In summary, we are confident that by combining theresults of the CANDELS and VANDELS teams we were ableto produce a final set of photometric redshifts of consistentquality over all four of the VANDELS regions, irrespectiveof the availability of deep HST imaging data.
In order to produce the cleanest selection catalogue pos-sible, it was necessary to remove potential stellar sources.Due to the high angular resolution provided by
HST , thiswas a straightforward process for the photometric cata-logues within the UDS-HST and CDFS-HST regions. Allsources originating from the Galametz et al. (2013) and Guoet al. (2013) catalogues were excluded if they had a sex-tractor (Bertin & Arnouts 1996) stellaricity parameterof CLASS STAR ≥ . . Following the application of thiscriteria to remove stellar sources, it was confirmed that theUDS-HST and CDFS-HST photometric catalogues no longerdisplayed a stellar locus in a variety of different colour-colourdiagrams.For the two ground-based photometric catalogues, allsources consistent with the stellar locus on the BzK dia-gram (Daddi et al. 2004) were excluded. In addition, all re-maining sources had their SED fitted with a range of stellartemplates drawn from the SpeX archive . All sources whichproduced an improved SED fit with a stellar template andwere consistent with being a point source at ground-based http://pono.ucsd.edu/ adam/browndwarfs/spexprism/ resolution were excluded. It should be noted that < ofthe objects in the two ground-based photometric catalogueswere excluded as being potentially stellar. Morevoer, it isnoteworthy that of the excluded objects had z phot < and would therefore not even have entered the VANDELSparent sample (see Section 4.5). At this stage, a final run of SED fitting was carried outin order to derive star-formation rates, stellar masses andrest-frame photometry. This SED fitting was performed us-ing Bruzual & Charlot (2003) templates with solar metallic-ity and no nebular emission. Exponentially-declining star-formation histories were employed, with τ in the range . ≤ τ ≤ Gyr, and ages were constrained to lie be-tween 50 Myr and the age of the Universe at the redshift ofinterest. Dust attenuation was described using the Calzettiet al. (2000) starburst attenuation law, with A V in the range . ≤ A V ≤ . , and IGM absorption was accounted for us-ing the Madau (1995) prescription. These parameters wereadopted following the results of Wuyts et al. (2011), whoshowed that this parameter set does a reasonable job ofrecovering the total star-formation rate of main-sequencegalaxies, provided that they are not heavily obscured. Wealso note that this SED parameter set is very similar to thatadopted by the 3D-HST survey team (Momcheva et al. 2016)and delivers stellar-mass estimates in good agreement withthose derived for the CANDELS CDFS and UDS photo-metric catalogues by Santini et al. (2015). During the SED-fitting process the redshift was fixed at the median value de-rived from the multiple photometric-redshift runs describedin Section 4.2.Further cleaning of the sample was carried out basedon the results of the SED fitting. For each of the four pho-tometric catalogues, plots of the SED fits for the objectscomprising the worst 10% of fits (i.e. highest χ ), were vi-sually examined. Objects that were revealed by this processto have unreliable or discrepant photometry were excludedfrom the sample ( (cid:39) of objects). Armed with catalogues providing robust photometry, photo-metric redshifts and physical properties, it was then possibleto select the parent sample of potential spectroscopic targets.The vast majority (i.e. (cid:39) %) of the potential targets weredrawn from three main target categories: • Bright star-forming galaxies in the range . ≤ z ≤ . • Lyman-break galaxies in the range . ≤ z ≤ . • Passive galaxies in the range . ≤ z ≤ . while the remaining (cid:39) % of potential targets were eitherknown or candidate AGN ( (cid:39) %), or Herschel- detectedgalaxies ( (cid:39) %). This sub-sample consists of bright star-forming galaxieswithin the redshift range . ≤ z ≤ . with i ≤ . The MNRAS , 1–19 (2018) he VANDELS spectroscopic survey redshift range is designed to ensure that the UV absorp-tion features necessary for investigating stellar metallicitylie within the . < λ < . µ m wavelength coverage of theVANDELS spectra. The magnitude constraint is designed toensure that the final VANDELS spectra have sufficient SNRto allow absorption-line studies on individual objects. In or-der to be classified as actively star-forming, each member ofthis sub-sample was required to satisfy: sSFR > − ,where sSFR is the specific star-formation rate (SFR/ M ∗ )derived from the SED fitting described in Section 4.4. Inreality, 99% of this sub-sample satisfy the criteria: sSFR > − , ensuring that they are fully consistent with beinglocated on the main sequence of star formation (see Fig. 3). This sub-sample consists of fainter star-forming galaxieswithin the redshift range . ≤ z ≤ . . The vast major-ity (95%) of the galaxies in this sub-sample lie in the red-shift interval . ≤ z ≤ . and in the HST regions have ≤ H ≤ ∧ i ≤ . . In the wider-field regions theseobjects have i ≤ . . The remainder of the sub-sampleconsists of galaxies selected to have redshifts in the range . ≤ z ≤ . and, in the HST regions, to have ≤ H ≤ and z (cid:48) ≤ . (UDS-HST) or z ≤ . (CDFS-HST).In the wider-field regions these objects have z (cid:48) ≤ . and z ≤ . in the UDS-GROUND and CDFS-GROUND re-gions, respectively. The change in selection criteria for the z ≥ . targets was mandatory, due to the impact of IGMabsorption on i − band photometry at these redshifts. Onceagain, the formal requirement for these galaxies to be classi-fied as star-forming was that sSFR > − . However, inreality, 99% of the galaxies in this sub-sample have sSFR > − and provide a good sampling of the main sequenceof star formation (see Fig. 3). This sub-sample consists of
UV J − selected (Williams et al.2009; Whitaker et al. 2011) passive galaxies in the red-shift interval . ≤ z ≤ . with H ≤ . ∧ i ≤ . The H − band magnitude constraint for this sub-sample is de-signed to impose an effective lower stellar-mass limit of log ( M ∗ / M (cid:12) ) ≥ . As with the bright star-forming galaxysub-sample, the i − band magnitude constraint is designed toensure that the final individual spectra are deep enoughto allow detailed absorption-line studies. The UV J selec-tion was performed using the rest-frame photometry derivedfrom the SED fitting described in Section 4.4. Galaxies whichsatisfied all of the following criteria were identified as pas-sive: U − V > . ( V − J ) + . , U − V > . , V − J < . . (1)We note here that although these galaxies are classified aspassive, it is not the case that they are necessarily expectedto exhibit no on-going star-formation. Based on the resultsof the SED fitting, 94% of the UV J − selected passive galaxiesdo have estimated values of sSFR < − , clearly sepa-rating them from main-sequence galaxies. However, of the UV J − selected passive galaxies have sSFR > − ,placing them in a location on the SFR − M ∗ diagram consis-tent with the low-SFR tail of the main sequence. This is notunexpected, given that UV J selection is inevitably vulnera-ble to contamination by dusty star-forming galaxies at somelevel.
The candidate AGN all lie within the CDFS field and wereselected based on either a power-law SED shape in the mid-IR (Chang et al. 2017) or X-ray emission (Xue et al. 2011;Rangel et al. 2013; Hsu et al. 2014). Within the CDFS-HSTregion the candidate AGN were restricted to z ≥ . and i ≤ . , while in the CDFS-GROUND region they were re-stricted to z ≥ . and i ≤ . The Herschel- detected galax-ies all lie within the UDS-HST and CDFS-HST regions, have z ≥ . and i ≤ . , and are detected in at least one Her-schel band (c.f. Pannella et al. 2015). We note here that thephotometric redshifts derived for the AGN candidates arebased on SED fitting with the same set of galaxy templatesdiscussed in Section 4.2, and are therefore not expected tobe as accurate as the photometric redshifts derived for therest of the VANDELS sample.
Following the application of the selection criteria outlinedabove, a final visual check was performed on the entire sam-ple to ensure that no image artefacts had survived the se-lection procedure. The resulting parent sample of potentialVANDELS spectroscopic targets consisted of 9656 galaxies,split roughly equally between the UDS and CDFS fields. Thedistribution of the parent sample on the SFR − M ∗ plane isshown in Fig. 3, from which it can be seen that the adoptedselection criteria successfully isolated the main sequence ofstar formation and the high stellar-mass quenched popula-tion. Overall, the parent VANDELS sample spans 3.5 dex instellar mass and 4.5 dex in star-formation rate. Using the parent sample as input, extensive simulation workwas undertaken in order to maximise the number of slitswhich could be allocated across the eight VIMOS pointings.In addition to the total number of spectroscopic slits, theprimary goal of this experimentation was to maximise thenumber of slits allocated to bright star-forming galaxies andmassive passive galaxies, the two classes of targets with thelowest surface densities. Apart from the photometric redshiftand magnitude constraints outlined above, the only addi-tional constraint applied to the simulations was the desire toallocate the slits to objects requiring 20, 40 and 80 hours ofintegration in an approximately 1:2:1 ratio. Crucially, dur-ing the slit allocation process, no additional prioritisationwas applied based on source brightness, redshift or position.The overall result of the target selection process was afinal sample of 2106 galaxies being allocated to spectroscopicslits. The distribution of the spectroscopic slits between thetwo survey fields, the different target classifications and thedifferent amounts of required exposure time are detailed
MNRAS000
MNRAS000 , 1–19 (2018) R. J. McLure et al.
Figure 3.
The distribution of the VANDELS parent sample on the SFR − M ∗ plane. The blue-shaded 2D histogram shows the locationof the star-forming galaxies (including additional candidate AGN and Herschel sources) in the redshift interval . ≤ z ≤ . (medianredshift z = . ). The red-shaded histogram shows the location of the passive galaxy sub-sample in the redshift interval . ≤ z ≤ . (median redshift z = . ). The horizontal and vertical colour bars indicate the number of galaxies within each 2D bin. The blue and greendashed lines show determinations of the main sequence of star-formation at z = and z = . by Speagle et al. (2014) and Whitakeret al. (2014), respectively. It can be seen that the VANDELS galaxies successfully sample the main sequence of star-formation and thearea of parameter space occupied by massive, quenched galaxies. In total, the VANDELS spectroscopic sample spans 3.5 dex in stellarmass and 4.5 dex in star-formation rate. in Table 2. The final spectroscopic samples of bright star-forming galaxies and passive galaxies are random (approx-imately 1 in 4) sub-samples drawn from the correspondingtargets within the input parent spectroscopic sample. Like-wise, the final spectroscopic sample of Lyman-break galax-ies is a random (approximately 1 in 5) sub-sample of theLyman-break targets within the parent spectroscopic sam-ple. In Fig. 4 we compare the photometric-redshift distribu-tion of the final VANDELS sample to the spectroscopic red-shift distributions of comparable large-scale spectroscopicsurveys previously carried out using the VIMOS spectro-graph. As illustrated in Fig. 1, the VANDELS survey consists ofa total of eight VIMOS pointings, four overlapping point-ings in UDS and four overlapping pointings in CDFS. Inboth fields the pointing centres were chosen to provide bothcontiguous coverage and to fully sample the central areaswith deep
HST imaging. Fully covering the deep
HST imag-ing was essential in order to allow access to a high surface-density of faint z ≥ targets. The VANDELS observing strategy was designed to provideconsistently high SNR continuum detections for the brightstar-forming and passive galaxy sub-samples. For those ob-jects with i ≤ . , the final 1D spectra are designed to have MNRAS , 1–19 (2018) he VANDELS spectroscopic survey Table 2.
The distribution of the 2106 spectroscopic slits targetedwithin the VANDELS survey between the two survey fields, thedifferent target classifications and the different integration times.The first column lists the survey field. Column two lists the num-ber of slits allocated to bright star-forming galaxies (SFG), col-umn three lists the number of slits allocated to massive, passivegalaxies (PASS), column four lists the number of slits allocated tofainter star-forming galaxies (LBG) and the fifth column lists thenumber of slits allocated to AGN candidates or
Herschel- detectedgalaxies (AH). Note that all of the AGN candidates were selectedin the CDFS field due to the availability of ultra-deep X-ray data(Luo et al. 2017).The final three columns list the number of slitsallocated to objects which require 20, 40 and 80 hours of on-sourceintegration, respectively.FIELD SFG PASS LBG AH 20 40 80UDS 224 151 693 10 303 550 225CDFS 200 117 656 55 238 528 262TOTAL 424 268 1349 65 541 1078 487 a SNR in the range − per resolution element, withinthe wavelength range < λ < ˚A, based on 20 or40 hours of on-source integration (where one resolution ele-ment is 4 pixels, or 10.2˚A). For the faintest objects in thesesub-samples ( i (cid:39) ), the final spectra are designed to haveSNR (cid:39) , based on 80 hours of integration. For the fainter( H ≤ ∧ i ≤ . ) Lyman-break galaxies at z ≥ , the VAN-DELS observing strategy is designed to provide SNR ≥ inthe continuum, and a consistent Ly α emission-line detectionlimit of (cid:39) × − erg s − cm − ( σ , integrated over a lineprofile with FWHM=10˚A).In order to achieve the desired SNR, targets were allo-cated 20, 40 or 80 hours of on-source integration accordingto two different exposure time schemes. The bright star-forming and passive galaxies were allocated 20 hours of inte-gration time if i ≤ . , 40 hours in the range . < i ≤ . and 80 hours in the range . < i ≤ . (where i is the i − band magnitude measured in a 2 (cid:48)(cid:48) − diameter circu-lar aperture at ground-based resolution ). The LBGs, AGNcandidates and Herschel- detected galaxies were allocated 20,40 or 80 hours of integration time within the following threemagnitude ranges: . < i ≤ . , . < i ≤ . and . < i ≤ . . The highest-redshift LBG targetsat z ≥ . followed the same exposure time scheme as themain LBG sub-sample, except with the i − band magnitudesreplaced with z − band magnitudes. To accommodate the required range of exposure times, theVANDELS survey employed a nested slit allocation strategy.Each of the eight VIMOS pointings was observed using foursets of masks, with each set receiving 20 hours of on-sourceintegration time. Consequently, objects which required 80hours of integration were retained on all four masks, thoserequiring 40 hours were included on two masks and those the typical off-set between i and the total i − band magnitudesused throughout the rest of the paper is (cid:39) . mag. z . . . . . . F r e qu e n c y zCOSMOS DeepVVDS DeepVUDSVLRSVANDELS Figure 4.
A comparison of the redshift distributions of large-scalespectroscopic surveys carried out with the VIMOS spectrograph.The deep component of the zCOSMOS survey (Lilly et al. 2007) isshown in blue and the deep component of the VIMOS VLT DeepSurvey (VVDS) is shown in green (Le F`evre et al. 2013). The VI-MOS Ultra Deep Survey (VUDS) is shown in red (Le F`evre et al.2015) and the VLT LBG Redshift Survey (VLRS) is shown inorange (Bielby et al. 2013). The black histogram shows the pho-tometric redshift distribution of the final sample of 2106 galaxiestargeted by the VANDELS survey. requiring 20 hours only appeared on a single mask. As canbe seen from Table 2, approximately 75% of the galaxiestargeted by the VANDELS survey received 40+ hours ofon-source integration.
All of the VANDELS observations used the MRgrism+GG475 order sorting filter, 1 arcsec slit widths and aminimum slit length of 7 arcsec. This set-up provides wave-length coverage of 480 − R (cid:39) . All ofthe slits were oriented E-W on the sky, as recommendedfor minimising slit losses when pursuing long integrations ofthe UDS and CDFS fields from Paranal (S´anchez-Janssenet al. 2014). To ensure that the VIMOS slits were placedwith maximum accuracy, short R − band pre-images were ob-tained in service mode during P94, in order to properly ac-count for VIMOS focal plane distortions and allocate 1–2bright reference stars to each VIMOS mask.All observations were obtained using observing blocks(OBs) designed to deliver a total of one hour of on-sourceintegration time. Each OB consisted of three integrations of1200s, obtained in a three-point dither pattern, with off-setsof 0, − − . and+1.64 arcsec respectively. One arc frame and one flat-fieldframe were obtained for calibration purposes after the execu- MNRAS000
All of the VANDELS observations used the MRgrism+GG475 order sorting filter, 1 arcsec slit widths and aminimum slit length of 7 arcsec. This set-up provides wave-length coverage of 480 − R (cid:39) . All ofthe slits were oriented E-W on the sky, as recommendedfor minimising slit losses when pursuing long integrations ofthe UDS and CDFS fields from Paranal (S´anchez-Janssenet al. 2014). To ensure that the VIMOS slits were placedwith maximum accuracy, short R − band pre-images were ob-tained in service mode during P94, in order to properly ac-count for VIMOS focal plane distortions and allocate 1–2bright reference stars to each VIMOS mask.All observations were obtained using observing blocks(OBs) designed to deliver a total of one hour of on-sourceintegration time. Each OB consisted of three integrations of1200s, obtained in a three-point dither pattern, with off-setsof 0, − − . and+1.64 arcsec respectively. One arc frame and one flat-fieldframe were obtained for calibration purposes after the execu- MNRAS000 , 1–19 (2018) R. J. McLure et al. tion of two consecutive OBs. A spectrophotometric standardwas observed at least once every seven nights and at leastonce per observing run. Further details of the VANDELS ob-servations can be found in the data release paper (Pentericciet al. 2018).
The reduction of the VANDELS data set is performed withthe fully-automated easylife pipeline, starting from theraw data and ending with the fully wavelength- and flux-calibrated one-dimensional spectra. The easylife pipeline(Garilli et al. 2012) is an updated version of the original vipgi system (Scodeggio et al. 2005). The original vipgi system was used to reduce all the spectra from the VVDS(Le F`evre et al. 2005; Garilli et al. 2008), zCosmos (Lillyet al. 2007) and VUDS surveys (Le F`evre et al. 2015), whilethe updated system easylife was used to reduce all of thespectra from the recently completed VIPERS survey (Guzzoet al. 2014b). A detailed description of the full data reduc-tion process can be found in Pentericci et al. (2018).In addition to the reduced spectra, it is a requirementof the ESO public survey agreement for VANDELS thatthe team provide spectroscopic redshift measurements foreach of the spectra released via the ESO data archive. Thespectroscopic redshift measurements were made by a dedi-cated group of VANDELS team members using the ez soft-ware package (Garilli et al. 2010). The core algorithm of ez is cross-correlation using galaxy templates that, for VAN-DELS spectra, were predominantly derived from previousVIMOS surveys. The redshift for each galaxy was indepen-dently measured by two team members, who were subse-quently required to reach agreement on the spectroscopicredshift measurement and the associated quality flag. As afinal check, the spectroscopic redshifts and associated qual-ity flags for all spectra released in DR1 were independentlychecked by the two Co-PIs.The quality of the spectroscopic redshift measurementswas quantified using the system originally employed by theVVDS team (Le F`evre et al. 2005), in which every galaxy isallocated a quality flag of 0, 1, 2, 3, 4 or 9. Galaxies for whichit was not possible to measure a spectroscopic redshift areallocated flag=0, while galaxies with spectroscopic redshiftmeasurements that are believed to be 50% or 75% reliableare allocated flag=1 and flag=2, respectively. The galaxieswith the most secure redshifts, based on multiple absorp-tion/emission features, are allocated flag=3 or 4, dependingon whether their redshift measurements are believed to be95% or 100% reliable. Galaxies which have redshift measure-ments based on a single emission line, in most cases Ly α , areallocated flag=9. The first public data release for the VANDELS survey(DR1) was made by the ESO Science Archive Facility( archive.eso.org ) on 29th September 2017, and featuresspectra obtained during the first VANDELS observing sea-son from August 2015 until February 2016; ESO run num- bers 194.A-2003(E-K). The data release includes fully flux-and wavelength-calibrated 1D spectra, plus wavelength cal-ibrated 2D spectra, for all the VANDELS targets that re-ceived their total scheduled integration time during seasonone. In addition, the data release also includes spectra forthose targets that had received 50% of their scheduled inte-gration time by the end of season one.In total, DR1 contains spectra for 879 galaxies, 415 fromthe CDFS pointings and 464 from the UDS pointings. In Fig.5 we show finding charts for the CDFS and UDS fields whichshow the locations of the full VANDELS target list in blue,with the locations of those VANDELS targets featured inDR1 in white. In addition to the reduced spectra, DR1 alsofeatures an associated catalogue which provides coordinates,optical+nearIR photometry, photometric redshifts, spectro-scopic redshifts and spectroscopic redshift quality flags foreach target. In Figs. 6 & 7, we show examples that illustratethe potential for using the DR1 data set to produce highSNR stacked spectra.
Based on the extensive testing described in Section 4.2, itwas determined that the typical accuracy of the photomet-ric redshifts adopted in the VANDELS target selection was σ dz (cid:39) . , with a catastrophic outlier rate of ≤ . How-ever, as is often the case, the samples of galaxies used tovalidate the photometric redshifts have i − band magnitudesthat are significantly brighter than those of the real VAN-DELS targets. Indeed, the median i − band magnitude of thegalaxies used to validate the photometric redshifts is twomagnitudes brighter than the median i − band magnitude ofthe DR1 galaxies. Consequently, it is clearly of interest touse the DR1 galaxies to review the accuracy of the selectionprocess based on real, on-sky, data.In the top panel of Fig. 8 we show a plot of z phot ver-sus z spec for the galaxies released in DR1 with spectroscopicredshift quality flags 3 and 4, which together comprise 55%of the full DR1 sample. For these galaxies σ dz = . witha catastrophic outlier rate of only 0.8%. The middle panelin Fig. 8 is the equivalent plot for those DR1 galaxies withspectroscopic redshift quality flags 1, 2 and 9, which have σ dz = . and a catastrophic outlier rate of 3.6%. Takentogether, the full DR1 sample (i.e. flags − ) has an accuracyof σ dz = . with a catastrophic outlier rate of . .It is worth noting that the fraction of catastrophic out-liers is actually significantly biased by the inclusion of arelatively small number of AGN candidates and Herschel- detected galaxies. If the statistics are restricted to the 97%of objects drawn from the three principal classificationsof VANDELS targets (see Section 4.5), the accuracy is σ dz = . and the catastrophic outlier rate is a remark-ably low 1.2% (flags − ). Given the relative faintness ofthe VANDELS targets, these figures provide a clear valida-tion of the accuracy and robustness of the target selectionprocedure described in Section 4. Moreover, the low num-ber of catastrophic outliers amongst those objects allocatedspectroscopic quality flags 1 and 2 suggests that the VAN-DELS quality flags are somewhat conservative. In reality,for many of the flag 1 and 2 objects we can be very confi-dent that the spectroscopic redshift lies within a relatively MNRAS , 1–19 (2018) he VANDELS spectroscopic survey Figure 5.
Finding charts showing the location of the VANDELS spectroscopic targets within the CDFS (top) and UDS (bottom) fields.The 415 targets in the CDFS and 464 targets in the UDS with spectra released in VANDELS DR1 are shown in white, with the remainingtargets shown in blue. The black dashed rectangles show the approximate location of the CANDELS near-IR
HST imaging (Grogin et al.2011; Koekemoer et al. 2011). The background images are ground-based H − band data from the VISTA VIDEO (Jarvis et al. 2013) andUKIDSS UDS (Almaini et al., in preparation) surveys.MNRAS000
HST imaging (Grogin et al.2011; Koekemoer et al. 2011). The background images are ground-based H − band data from the VISTA VIDEO (Jarvis et al. 2013) andUKIDSS UDS (Almaini et al., in preparation) surveys.MNRAS000 , 1–19 (2018) R. J. McLure et al. λ rest / ˚A f λ / − e r g s − c m − ˚ A − C III S i II O I C II O I V S i I V S i I V S i II C I V F e II A III N i II A III F e II F e II F e II L y α S i II S i II S i II H e II O III C III S i III λ rest / ˚A f λ / − e r g s − c m − ˚ A − C III S i II O I C II O I V S i I V S i I V S i II C I V F e II A III N i II A III F e II F e II F e II L y α S i II S i II S i II H e II O III C III S i III λ rest / ˚A f λ / − e r g s − c m − ˚ A − C III S i II O I C II O I V S i I V S i I V S i II C I V F e II A III N i II A III F e II F e II F e II L y α S i II S i II S i II H e II O III C III S i III
Figure 6.
Median-stacked spectra of Lyman-break galaxies from VANDELS DR1. The top panel shows a stack of 105 LBGs from DR1with robust redshifts in the range . ≤ z ≤ . (median redshift z = . ). The middle panel shows a stack of the 61/105 galaxies thatdisplay Ly α in emission. The bottom panel shows a stack of the 44/105 galaxies that display Ly α in absorption. In all three panels,common absorption (dotted lines) and emission (dot-dashed lines) features are highlighted. narrow range, but the spectral features simply do not allowcompeting redshift solutions to be reliably differentiated.In the bottom panel of Fig. 8 the redshift distributionof the galaxies released in DR1 is shown as the filled bluehistogram, based on their measured spectroscopic redshifts.The histogram indicated by the thin grey line shows the red-shift distribution of the VANDELS parent sample, based onthe input photometric redshifts. A comparison of the twoclearly indicates that the spectroscopic redshift distribution of the real VANDELS spectra is in very close agreement tothe distribution predicted by the photometric-redshift selec-tion procedure.The galaxies targeted by the VANDELS survey arefainter than those typically targeted by previous large spec-troscopic surveys of high-redshift galaxies. Consequently, itis clearly of interest to explore how the accuracy of the VAN-DELS photometric redshifts varies as a function of targetmagnitude. MNRAS , 1–19 (2018) he VANDELS spectroscopic survey λ rest / ˚A . . . . f λ / − e r g s − c m − ˚ A − [ O II ] M n II F e II M g II M g I M g I F e I H H C N C a II H C a II K H δ λ rest / ˚A . . . . f λ / − e r g s − c m − ˚ A − [ O II ] M n II F e II M g II M g I M g I F e I H H C N C a II H C a II K H δ λ rest / ˚A . . . . f λ / − e r g s − c m − ˚ A − [ O II ] M n II F e II M g II M g I M g I F e I H H C N C a II H C a II K H δ Figure 7.
Median-stacked spectra of passive galaxies from VANDELS DR1. The top panel shows a stack of 65 passive galaxies from DR1with robust redshifts in the range . ≤ z ≤ . (median redshift z = . ). The middle panel shows a stack of the 33/65 passive galaxiesthat display [O ii ] emission. The bottom panel shows a stack of the 32/65 passive galaxies without [O ii ] emission. Common absorption(dotted lines) and emission (dot-dashed lines) features are highlighted in each panel. All but three of the VANDELS galaxies released in DR1have i − band magnitudes in the range . ≤ i ≤ . .Consequently, Fig. 9 shows a comparison between spectro-scopic and photometric redshifts in three i − band magni-tude ranges: . < i ≤ . , . < i ≤ . and One passive galaxy has i = . and two further galaxies with i ≥ . were selected as z ≥ . LBGs based on their z − bandmagnitudes. . < i ≤ . , and includes all objects with spectro-scopic redshift quality flags − . The middle panel of Fig.9 is representative of the i − band magnitude of the typicalVANDELS source, whereas the top and bottom panels il-lustrate the photometric redshift accuracy at the bright andfaint ends of the target magnitude distribution, respectively.The relevant statistics quantifying the quality of the agree-ment between the spectroscopic and photometric redshiftsare displayed in the top-left corner of each panel of Fig. 9. MNRAS000
Median-stacked spectra of passive galaxies from VANDELS DR1. The top panel shows a stack of 65 passive galaxies from DR1with robust redshifts in the range . ≤ z ≤ . (median redshift z = . ). The middle panel shows a stack of the 33/65 passive galaxiesthat display [O ii ] emission. The bottom panel shows a stack of the 32/65 passive galaxies without [O ii ] emission. Common absorption(dotted lines) and emission (dot-dashed lines) features are highlighted in each panel. All but three of the VANDELS galaxies released in DR1have i − band magnitudes in the range . ≤ i ≤ . .Consequently, Fig. 9 shows a comparison between spectro-scopic and photometric redshifts in three i − band magni-tude ranges: . < i ≤ . , . < i ≤ . and One passive galaxy has i = . and two further galaxies with i ≥ . were selected as z ≥ . LBGs based on their z − bandmagnitudes. . < i ≤ . , and includes all objects with spectro-scopic redshift quality flags − . The middle panel of Fig.9 is representative of the i − band magnitude of the typicalVANDELS source, whereas the top and bottom panels il-lustrate the photometric redshift accuracy at the bright andfaint ends of the target magnitude distribution, respectively.The relevant statistics quantifying the quality of the agree-ment between the spectroscopic and photometric redshiftsare displayed in the top-left corner of each panel of Fig. 9. MNRAS000 , 1–19 (2018) R. J. McLure et al. z spec z pho t DR1 flags 3+4 (N=483)outlier rate = 0.8% σ dz = 0.026bias = 0.003 flag=3flag=4 z spec z pho t DR1 flags 1+2+9 (N=390)outlier rate = 3.6% σ dz = 0.036bias = 0.002 flag=1flag=2flag=9 z spec . . . . . F r e qu e n c y Figure 8.
The top panel shows a comparison between the in-put photometric redshifts and measured spectroscopic redshiftsfor DR1 galaxies with redshift quality flags 3 and 4. The middlepanel is the equivalent plot for DR1 galaxies with redshift qualityflags 1, 2 and 9. Those galaxies falling outside the dashed lines arecatastrophic outliers with | dz | > . . In both panels, candidateAGN and Herschel- detected galaxies are plotted as open sym-bols. The bottom panel shows a comparison of the spectroscopicredshift distribution of the DR1 galaxies (solid blue histogram)and the photometric redshift distribution of the full VANDELSparent sample (open histogram). z spec z pho t . < i ≤ . ( i med = . ) outlier rate = 1.5% σ dz = 0.025bias = 0.002 z spec z pho t . < i ≤ . ( i med = . ) outlier rate = 1.5% σ dz = 0.031bias = 0.002 z spec z pho t . < i ≤ . ( i med = . ) outlier rate = 3.7% σ dz = 0.036bias = 0.010 Figure 9.
The top panel shows a comparison between the inputphotometric redshifts and measured spectroscopic redshifts forDR1 galaxies in the magnitude range . < i ≤ . . The mid-dle and bottom panels show the equivalent plots for DR1 galaxiesin the magnitude ranges . < i ≤ . and . < i ≤ . ,respectively. All three panels include all DR1 galaxies with spec-troscopic redshift quality flags in the range − .MNRAS , 1–19 (2018) he VANDELS spectroscopic survey It is clear from Fig. 9 that in terms of bias and catas-trophic outlier rate, the VANDELS photometric redshiftsperform very well within the two brighter magnitude bins.Over the full magnitude range there is a gradual decrease inthe photometric redshift accuracy, with σ dz dropping from0.025 to 0.036. However, given the factor of (cid:39) drop inbrightness between the top and bottom panels, the decreasein accuracy is not particularly dramatic. In contrast, it isclear from the bottom panel of Fig. 9 that the photometricredshifts for the faintest VANDELS targets with i > . ( (cid:39) of the DR1 objects) do show a notable increase inboth the fraction of catastrophic outliers and the bias.Overall, the quality of the VANDELS photometric red-shifts is in-line with expectations based on the spectroscopicredshift validation data (see Section 4.2). For all DR1 ob-jects with spectroscopic quality flags − , an accuracy of σ dz = . and a catastrophic outlier rate of 2.1% comparesfavourably with the results from the spectroscopic validationsets ( σ dz = . and . catastrophic outliers), despitethe i − band magnitudes of the VANDELS galaxies being twomagnitudes fainter than the validation objects, on average.Interestingly, compared to the DR1 data, the overall system-atic bias of the photometric redshifts is only . ± . .This is actually better than the expectation from the spec-troscopic validation data ( . ± . ), albeit only at the (cid:39) . σ level. In this paper we have provided an overview of the VAN-DELS spectroscopic survey, focusing on the scientific moti-vation, survey design and target selection. The original moti-vation for the VANDELS survey was to move beyond simpleredshift determination and to provide the high SNR spec-tra necessary to study the physical properties of the high-redshift galaxy population. The spectra released in DR1demonstrate that the original goals of the survey are withinreach, and that the VIMOS spectrograph can be used to in-tegrate for 20–80 hours without the final SNR being domi-nated by systematic effects. Combined with the unparalleledancillary data available within the CDFS and UDS surveyfields, it is clear that the VANDELS survey has the potentialto become a key legacy data set for studying the evolutionof high-redshift galaxies for many years to come.The observations for the VANDELS survey were fullycompleted in February 2018. The second ESO public datarelease is currently scheduled for June 2018 and will featureall of the spectra completed, or 50% completed, by the endof the second VANDELS observing season in February 2017.The third ESO public data release is scheduled for June 2019and will consist of the entire VANDELS spectroscopic dataset. A final data release is currently scheduled for June 2020and will formally mark the end of the project. It is cur-rently intended that the final data release will feature are-reduction of the entire spectroscopic data set, incorpo-rating improvements in the data reduction process whichhave been implemented over the course of the survey. Inaddition, the VANDELS team is committed to release twofinal catalogues to enhance the legacy value of the survey.The first catalogue will contain physical properties for each target (i.e. stellar masses, star-formation rates, dust atten-uation and rest-frame colours) based on SED fitting of thefinal data set. The second catalogue will provide measure-ments of the fluxes and equivalent widths of significant emis-sion/absorption features identified in the VANDELS spec-tra, along with their corresponding uncertainties.
ACKNOWLEDGEMENTS
Based on data products from observations made with ESOTelescopes at the La Silla Paranal Observatory under pro-gramme ID 194.A-2003(E-K). We thank the ESO staff fortheir continuous support for the VANDELS survey, par-ticularly the Paranal staff, who helped us to conduct theobservations, and the ESO user support group in Garch-ing. RJM, AM, EMQ and DJM acknowledge funding fromthe European Research Council, via the award of an ERCConsolidator Grant (P.I. R. McLure). AC acknowledges thegrants PRIN-MIUR 2015 and ASI n.I/023/12/0. RA ac-knowledges support from the ERC Advanced Grant 695671“QUENCH”. F.B. acknowledges the support by Funda¸c˜aopara a Ciˆencia e a Tecnologia (FCT) via the postdoctoralfellowship SFRH/BPD/103958/2014 and through the re-search grant UID/FIS/04434/2013. PC acknowledges sup-port from CONICYT through the project FONDECYT reg-ular 1150216.
REFERENCES
Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F.,Fontana A., Giallongo E., 1999, MNRAS, 310, 540Ashby M. L. N., et al., 2015, ApJS, 218, 33Balestra I., et al., 2010, A&A, 512, A12Bertin E., Arnouts S., 1996, A&AS, 117, 393Bielby R., et al., 2013, MNRAS, 430, 425Bolzonella M., Miralles J.-M., Pell´o R., 2000, A&A, 363, 476Bourne N., et al., 2017, MNRAS, 467, 1360Bouwens R. J., et al., 2016, ApJ, 833, 72Bowler R. A. A., et al., 2015, MNRAS, 452, 1817Bowler R. A. A., Bourne N., Dunlop J. S., McLure R. M., McLeodD. J., 2018, preprint, ( arXiv:1802.05720 )Bradshaw E. J., et al., 2013, MNRAS, 433, 194Brammer G. B., van Dokkum P. G., Coppi P., 2008, ApJ, 686,1503Brammer G. B., et al., 2012, ApJS, 200, 13Bruce V. A., et al., 2012, MNRAS, 427, 1666Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Calzetti D., Armus L., Bohlin R. C., Kinney A. L., Koornneef J.,Storchi-Bergmann T., 2000, ApJ, 533, 682Capak P. L., et al., 2015, Nature, 522, 455Cardamone C. N., et al., 2010, ApJS, 189, 270Carnall A. C., McLure R. J., Dunlop J. S., Dav´e R., 2017,preprint, ( arXiv:1712.04452 )Castellano M., et al., 2014, A&A, 566, A19Chabrier G., 2003, PASP, 115, 763Chang Y.-Y., et al., 2017, ApJS, 233, 19Chevallard J., Charlot S., 2016, MNRAS, 462, 1415Cimatti A., et al., 2013, ApJ, 779, L13Cirasuolo M., et al., 2014, in Ground-based and Air-borne Instrumentation for Astronomy V. p. 91470N,doi:10.1117/12.2056012Conselice C. J., 2014, ARA&A, 52, 291Cooper M. C., et al., 2012, MNRAS, 425, 2116MNRAS000
Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F.,Fontana A., Giallongo E., 1999, MNRAS, 310, 540Ashby M. L. N., et al., 2015, ApJS, 218, 33Balestra I., et al., 2010, A&A, 512, A12Bertin E., Arnouts S., 1996, A&AS, 117, 393Bielby R., et al., 2013, MNRAS, 430, 425Bolzonella M., Miralles J.-M., Pell´o R., 2000, A&A, 363, 476Bourne N., et al., 2017, MNRAS, 467, 1360Bouwens R. J., et al., 2016, ApJ, 833, 72Bowler R. A. A., et al., 2015, MNRAS, 452, 1817Bowler R. A. A., Bourne N., Dunlop J. S., McLure R. M., McLeodD. J., 2018, preprint, ( arXiv:1802.05720 )Bradshaw E. J., et al., 2013, MNRAS, 433, 194Brammer G. B., van Dokkum P. G., Coppi P., 2008, ApJ, 686,1503Brammer G. B., et al., 2012, ApJS, 200, 13Bruce V. A., et al., 2012, MNRAS, 427, 1666Bruzual G., Charlot S., 2003, MNRAS, 344, 1000Calzetti D., Armus L., Bohlin R. C., Kinney A. L., Koornneef J.,Storchi-Bergmann T., 2000, ApJ, 533, 682Capak P. L., et al., 2015, Nature, 522, 455Cardamone C. N., et al., 2010, ApJS, 189, 270Carnall A. C., McLure R. J., Dunlop J. S., Dav´e R., 2017,preprint, ( arXiv:1712.04452 )Castellano M., et al., 2014, A&A, 566, A19Chabrier G., 2003, PASP, 115, 763Chang Y.-Y., et al., 2017, ApJS, 233, 19Chevallard J., Charlot S., 2016, MNRAS, 462, 1415Cimatti A., et al., 2013, ApJ, 779, L13Cirasuolo M., et al., 2014, in Ground-based and Air-borne Instrumentation for Astronomy V. p. 91470N,doi:10.1117/12.2056012Conselice C. J., 2014, ARA&A, 52, 291Cooper M. C., et al., 2012, MNRAS, 425, 2116MNRAS000 , 1–19 (2018) R. J. McLure et al.
Cullen F., Cirasuolo M., McLure R. J., Dunlop J. S., BowlerR. A. A., 2014, MNRAS, 440, 2300Cullen F., Cirasuolo M., Kewley L. J., McLure R. J., Dunlop J. S.,Bowler R. A. A., 2016, MNRAS, 460, 3002Cullen F., et al., 2017, preprint, ( arXiv:1712.01292 )Curtis-Lake E., et al., 2012, MNRAS, 422, 1425Daddi E., Cimatti A., Renzini A., Fontana A., Mignoli M.,Pozzetti L., Tozzi P., Zamorani G., 2004, ApJ, 617, 746Daddi E., et al., 2007, ApJ, 670, 156Dahlen T., et al., 2013, ApJ, 775, 93Davidzon I., et al., 2017, A&A, 605, A70Dayal P., Ferrara A., Dunlop J. S., 2013, MNRAS, 430, 2891De Barros S., et al., 2017, A&A, 608, A123Dekel A., et al., 2009, Nature, 457, 451Diamond-Stanic A. M., Moustakas J., Tremonti C. A., Coil A. L.,Hickox R. C., Robaina A. R., Rudnick G. H., Sell P. H., 2012,ApJ, 755, L26Dunlop J. S., et al., 2017, MNRAS, 466, 861Elbaz D., et al., 2007, A&A, 468, 33Eldridge J. J., Stanway E. R., 2016, MNRAS, 462, 3302Fabian A. C., 2012, ARA&A, 50, 455Feldmann R., et al., 2006, MNRAS, 372, 565Finkelstein S. L., 2016, Publ. Astron. Soc. Australia, 33, e037Fontana A., D’Odorico S., Poli F., Giallongo E., Arnouts S., Cris-tiani S., Moorwood A., Saracco P., 2000, AJ, 120, 2206Fontana A., et al., 2014, A&A, 570, A11Furusawa H., et al., 2008, ApJS, 176, 1Furusawa H., et al., 2016, ApJ, 822, 46Galametz A., et al., 2013, ApJS, 206, 10Garilli B., et al., 2008, A&A, 486, 683Garilli B., Fumana M., Franzetti P., Paioro L., Scodeggio M., LeF`evre O., Paltani S., Scaramella R., 2010, PASP, 122, 827Garilli B., Paioro L., Scodeggio M., Franzetti P., Fumana M.,Guzzo L., 2012, PASP, 124, 1232Genel S., et al., 2014, MNRAS, 445, 175Giavalisco M., et al., 2004, ApJ, 600, L93Gonz´alez V., Bouwens R., Illingworth G., Labb´e I., Oesch P.,Franx M., Magee D., 2014, ApJ, 781, 34Grogin N. A., et al., 2011, ApJS, 197, 35Guo Y., et al., 2013, ApJS, 207, 24Guzzo L., et al., 2014a, A&A, 566, A108Guzzo L., et al., 2014b, A&A, 566, A108Henriques B. M. B., White S. D. M., Thomas P. A., Angulo R.,Guo Q., Lemson G., Springel V., Overzier R., 2015, MNRAS,451, 2663Hsieh B.-C., Wang W.-H., Hsieh C.-C., Lin L., Yan H., Lim J.,Ho P. T. P., 2012, ApJS, 203, 23Hsu L.-T., et al., 2014, ApJ, 796, 60Ilbert O., et al., 2006, A&A, 457, 841Jarvis M. J., et al., 2013, MNRAS, 428, 1281Johnston R., Vaccari M., Jarvis M., Smith M., Giovannoli E.,H¨außler B., Prescott M., 2015, MNRAS, 453, 2540Knebe A., et al., 2015, MNRAS, 451, 4029Koekemoer A. M., et al., 2011, ApJS, 197, 36Koprowski M. P., et al., 2018, preprint, ( arXiv:1801.00791 )Kurk J., et al., 2013, A&A, 549, A63Laidler V. G., et al., 2007, PASP, 119, 1325Le F`evre O., et al., 2005, A&A, 439, 845Le F`evre O., et al., 2013, A&A, 559, A14Le F`evre O., et al., 2015, A&A, 576, A79Lilly S. J., et al., 2007, ApJS, 172, 70Luo B., et al., 2017, ApJS, 228, 2Madau P., 1995, ApJ, 441, 18Madau P., Dickinson M., 2014, ARA&A, 52, 415Magnelli B., et al., 2013, A&A, 553, A132Mannucci F., Cresci G., Maiolino R., Marconi A., Gnerucci A.,2010, MNRAS, 408, 2115Marchi F., et al., 2017, preprint, ( arXiv:1710.10184 ) M´armol-Queralt´o E., McLure R. J., Cullen F., Dunlop J. S.,Fontana A., McLeod D. J., 2016, MNRAS, 460, 3587McLure R. J., et al., 2011, MNRAS, 418, 2074McLure R. J., et al., 2013a, MNRAS, 428, 1088McLure R. J., et al., 2013b, MNRAS, 432, 2696McLure R. J., et al., 2017, preprint, ( arXiv:1709.06102 )Mignoli M., et al., 2005, A&A, 437, 883Momcheva I. G., et al., 2016, ApJS, 225, 27Mortlock A., McLure R. J., Bowler R. A. A., McLeod D. J.,M´armol-Queralt´o E., Parsa S., Dunlop J. S., Bruce V. A.,2017, MNRAS, 465, 672Muzzin A., et al., 2013, ApJ, 777, 18Noeske K. G., et al., 2007, ApJ, 660, L43Nonino M., et al., 2009, ApJS, 183, 244Novak M., et al., 2017, A&A, 602, A5Oke J. B., Gunn J. E., 1983, ApJ, 266, 713Pannella M., et al., 2015, ApJ, 807, 141Peng Y., Maiolino R., Cochrane R., 2015, Nature, 521, 192Pentericci L., et al., 2014, ApJ, 793, 113Pentericci L., et al., 2018, preprint, ( arXiv:1803.07373 )Rangel C., Nandra K., Laird E. S., Orange P., 2013, MNRAS,428, 3089Reddy N. A., et al., 2018, ApJ, 853, 56Rix H.-W., et al., 2004, ApJS, 152, 163Rogers A. B., et al., 2014, MNRAS, 440, 3714S´anchez-Janssen R., Mieske S., Selman F., Bristow P., Hammer-sley P., Hilker M., Rejkuba M., Wolff B., 2014, A&A, 566,A2Santini P., et al., 2015, ApJ, 801, 97Scodeggio M., et al., 2005, PASP, 117, 1284Shapley A. E., Steidel C. C., Pettini M., Adelberger K. L., 2003,ApJ, 588, 65Shapley A. E., et al., 2015, ApJ, 801, 88Shapley A. E., Steidel C. C., Strom A. L., Bogosavljevi´c M.,Reddy N. A., Siana B., Mostardi R. E., Rudie G. C., 2016,ApJ, 826, L24Sobral D., Best P. N., Matsuda Y., Smail I., Geach J. E., CirasuoloM., 2012, MNRAS, 420, 1926Somerville R. S., Dav´e R., 2015, ARA&A, 53, 51Sommariva V., Mannucci F., Cresci G., Maiolino R., Marconi A.,Nagao T., Baroni A., Grazian A., 2012, A&A, 539, A136Speagle J. S., Steinhardt C. L., Capak P. L., Silverman J. D.,2014, ApJS, 214, 15Stanway E. R., Eldridge J. J., Becker G. D., 2016, MNRAS, 456,485Stark D. P., 2016, ARA&A, 54, 761Steidel C. C., Strom A. L., Pettini M., Rudie G. C., Reddy N. A.,Trainor R. F., 2016, ApJ, 826, 159Strom A. L., Steidel C. C., Rudie G. C., Trainor R. F., PettiniM., Reddy N. A., 2017, ApJ, 836, 164Talia M., et al., 2017, MNRAS, 471, 4527Tomczak A. R., et al., 2014, ApJ, 783, 85Vanzella E., et al., 2008, A&A, 478, 83Vanzella E., et al., 2016, ApJ, 825, 41Weiner B. J., et al., 2009, ApJ, 692, 187Whitaker K. E., et al., 2011, ApJ, 735, 86Whitaker K. E., et al., 2014, ApJ, 795, 104Williams R. J., Quadri R. F., Franx M., van Dokkum P., Labb´eI., 2009, ApJ, 691, 1879Wuyts S., et al., 2011, ApJ, 738, 106Xue Y. Q., et al., 2011, ApJS, 195, 10de Barros S., Schaerer D., Stark D. P., 2014, A&A, 563, A81de Barros S., et al., 2016, A&A, 585, A51van der Wel A., et al., 2014, ApJ, 788, 28van der Wel A., et al., 2016, ApJS, 223, 29 Institute for Astronomy, University of Edinburgh, RoyalObservatory, Edinburgh, EH9 3HJ, UK
MNRAS , 1–19 (2018) he VANDELS spectroscopic survey INAF, Osservatorio Astronomico di Roma, Monteporzio,Italy Dipartimento di Fisica e Astronomia, Universit`a diBologna, Via Gobetti 93/2, I-40129, Bologna, Italy INAF - Osservatorio Astrofisico di Arcetri, Largo E. Fermi5, I-50157, Firenze, Italy Laboratoire AIM-Paris-Saclay, CEA/DRF/Irfu, CNRSFrance MPE, Giessenbachstrasse 1, D-85748 Garching, Germany Kavli Institute for Cosmology, University of Cambridge,Madingley Road, Cambridge CB3 0HA, UK Cavendish Laboratory, University of Cambridge, 19 J. J.Thomson Avenue, Cambridge CB3 0HE, UK INAF-Osservatorio di Astrofisica e Scienza dello Spazio diBologna, via Gobetti 93/3, I-40129, Bologna, Italy European Southern Observatory, Karl-Schwarzschild-Str.2, D-85748 Garching b. Munchen, Germany Observatoire de Gen`eve, Universit´e de Gen`eve, 51 Ch. desMaillettes, 1290, Versoix, Switzerland Department of Astronomy, The University of Texas atAustin, Austin, TX 78712, USA INAF - Astronomical Observatory of Trieste, via G.B.Tiepolo 11, I-34143 Trieste, Italy INAF-Istituto di Astrofisica Spaziale e Fisica Cosmica Mi-lano, via Bassini 15, 20133, Milano, Italy N´ucleo de Astronom´ıa, Facultad de Ingenier´ıa, Universi-dad Diego Portales, Av. Ej´ercito 441, Santiago, Chile Department of Physics and Astronomy, University CollegeLondon, Gower Street, London WC1E 6BT, UK INAF-Osservatorio Astronomico di Brera, via Brera 28,20122 Milano, Italy Astrophysics, The Denys Wilkinson Building, Universityof Oxford, Keble Road, Oxford OX1 3RH, UK Kapteyn Astronomical Institute, University of Groningen,Postbus 800, 9700 AV, Groningen, The Netherlands European Southern Observatory, Avenida Alonso de C´or-dova 3107, Vitacura, 19001 Casilla, Santiago de Chile, Chile School of Physics and Astronomy, University of Notting-ham, University Park, Nottingham NG7 2RD, UK University Observatory Munich, Scheinerstrasse 1, D-81679 Munich, Germany Department of Astronomy, University of Michigan, 311West Hall, 1085 South University Ave., Ann Arbor, MI48109-1107, USA Instituto de Astrof´ısica e Ciˆencias do Espa¸co, Universidadede Lisboa, OAL, Tapada da Ajuda, P-1349-018 Lisbon, Por-tugal Departamento de F´ısica, Faculdade de Ciˆencias, Univer-sidade de Lisboa, Edif´ıcio C8, Campo Grande, PT1749-016Lisbon, Portugal Instituto de Fisica y Astronomia, Facultad de Ciencias,Universidad de Valparaiso, 1111 Gran Bretana, Valparaiso,Chile Institute d’Astrophysique de Paris, CNRS, Universit´ePierre et Marie Curie, 98 bis Boulevard Arago, 75014, Paris,France National Optical Astronomy Observatory, 950 NorthCherry Ave, Tucson, AZ, 85719, USA Harvard-Smithsonian Center for Astrophysics, 60 GardenSt, Cambridge MA 20138, USA Space Telescope Science Institute, 3700 San Martin Drive,Baltimore, MD, 21218, USA Dark Cosmology Centre, Niels Bohr Institute, Universityof Copenhagen, Juliane Maries Vej 30, DK-2100 Copen-hagen, Denmark Imperial college, Kensington, London SW7 2AZ, UK Astronomy Department, University of Massachusetts, Amherst, MA01003, USA Pontificia Universidad Cat´olica de Chile, Instituto de As-trof´ısica Avda. Vicu˜na Mackenna 4860, Santiago, Chile Aix Marseille Universit´e, CNRS, LAM (Laboratoired’Astrophysique de Marseille) UMR 7326, 13388, Marseille,France Instituto de Astrof´ısica de Canarias, Calle V´ıa L´actea s/n,E-38205 La Laguna, Tenerife, Spain Departamento de Astrof´ısica, Universidad de La Laguna,E-38200 La Laguna, Tenerife, Spain Faculty of Physics, Ludwig-Maximilians Universit¨at,Scheinerstr. 1, 81679, Munich, Germany Department of Physics and Astronomy, Texas A&M Uni-versity, College Station, TX 77843-4242, USA Excellence Cluster, Boltzmann Strasse 2, D-85748 Garch-ing, Germany Department of Physics, Durham University, South Road,DH1 3LE Durham, UK Leiden Observatory, Leiden University, 2300 RA, Leiden,The Netherlands Steward Observatory, The University of Arizona, 933 NCherry Ave, Tucson, AZ, 85721, USA Department of Physics and Astronomy, PAB, 430 PortolaPlaza, Box 951547, Los Angeles, CA 90095-1547, USA School of Physics and Astronomy, University of St. An-drews, SUPA, North Haugh, KY16 9SS St. Andrews, UK
This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000