MNRAS 446, 3526–3544 (2015) doi:10.1093/mnras/stu2241

HIGHz: a survey of the most H I-massive galaxies at z ∼ 0.2

Barbara Catinella‹ and Luca CorteseCentre for Astrophysics & Supercomputing, Swinburne University of Technology, Hawthorn, VIC 3122, Australia

Accepted 2014 October 12. Received 2014 October 9; in original form 2014 July 27

ABSTRACTWe present the results of the HIGHz Arecibo survey, which measured the H I content of39 galaxies at redshift z > 0.16 selected from the Sloan Digital Sky Survey. These are allactively star-forming, disc-dominated systems in relatively isolated environments, with stellarand H I masses larger than 1010 M� and redshifts 0.17 ≤ z ≤ 0.25. Our sample includesnot only the highest-redshift detections of H I emission from individual galaxies to date, butalso some of the most H I-massive systems known. Despite being exceptionally large, the H I

reservoirs of these galaxies are consistent with what is expected from their ultraviolet andoptical properties. This, and the fact that the galaxies lie on the baryonic Tully–Fisher relation,suggests that HIGHz systems are rare, scaled-up versions of local disc galaxies. We show thatthe most H I-massive galaxies discovered in the Arecibo Legacy Fast ALFA survey are thelocal analogues of HIGHz, and discuss the possible connection between our sample and theturbulent, gas-rich discs identified at z ∼ 1. The HIGHz sample provides a first glimpse intothe properties of the massive, H I-rich galaxies that will be detected at higher redshifts by thenext generation H I surveys with the Square Kilometre Array and its pathfinders.

Key words: galaxies: evolution – galaxies: fundamental parameters – galaxies: kinematicsand dynamics – radio lines: galaxies.

1 IN T RO D U C T I O N

In the past decade or so, studies of the cold atomic (H I) and molec-ular (H2) hydrogen content of local galaxies have progressed fromobservations of relatively small samples to large surveys, includingfrom several hundreds to a few tens of thousands objects (Barneset al. 2001; Giovanelli et al. 2005; Catinella et al. 2010; Saintongeet al. 2011). This significant boost in number statistics, as wellas in data quality, has made it possible to characterize the cold gasproperties of galaxies from a statistical point of view, and to identifythe most important scaling relations connecting gas and other galaxyproperties in the local Universe (Catinella et al. 2010; Cortese et al.2011; Saintonge et al. 2011; Huang et al. 2012; Boselli et al. 2014).

One of the main challenges for H I and H2 astronomy is now toextend these investigations to cosmological distances, probing thevariation of the cold gas content of galaxies with the age of theUniverse. This is particularly important, as the remarkable decreasein the cosmic star formation rate (SFR) density from z ∼ 1 to 0(e.g. Lilly et al. 1996; Madau et al. 1996; Bell et al. 2005) must bea direct consequence of a change in the gas cycle of galaxies.

Thanks to the upgrade of the IRAM Plateau de Bure interferome-ter, observations of molecular hydrogen (as traced by the CO lines)have recently been able to reach redshifts that were unimaginablea decade ago, gradually revealing a population of exceptionally

� E-mail: bcatinella@swin.edu.au

gas-rich (MH2 /M�∼ 1), turbulent disc galaxies at z = 1–2 (Daddiet al. 2010; Genzel et al. 2011; Tacconi et al. 2013). Althoughonly upcoming surveys with the Atacama Large Millimeter Arraywill determine whether or not these are representative of the star-forming galaxies at those redshifts, or just the gas-rich tail of thedistribution, these pioneering observations confirm that the physicalconditions of the interstellar medium have changed significantly inthe last ∼8 billion years, and highlight the importance of the verygas-rich regime for galaxy evolution studies.

Unfortunately, H I astronomy lags behind in this respect. Due tosensitivity of current instruments, as well as man-made radio in-terference, H I observations are still struggling to detect the weak21 cm emission beyond z ∼0.16 (Verheijen et al. 2007; Catinellaet al. 2008; Jaffe et al. 2012; Fernandez et al. 2013). Thus, spectralstacking of optically selected samples that are undetected in H I

surveys is currently the most powerful technique to measure theaverage H I content of galaxies at higher redshifts. Although not asubstitute for H I detections, stacking has provided estimates of thecosmic H I density up to a redshift z ∼ 0.4 (Lah et al. 2009). Onlythe next generation, deep H I surveys with the Square KilometreArray (SKA, Carilli & Rawlings 2004) and its pathfinders, ASKAP(Johnston et al. 2008) and MeerKAT (Booth et al. 2009), will be ableto detect H I emission at these and higher redshifts. As theoreticalmodels predict a different evolution in the properties of the atomicand molecular gas phases (e.g. Obreschkow & Rawlings 2009;Lagos et al. 2011), quantifying how the H I reservoirs of galaxiesvary with redshift is of primary importance.

C© 2014 The AuthorsPublished by Oxford University Press on behalf of the Royal Astronomical Society

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3527

Despite current constraints, carefully planned observations withexisting radio telescopes can already peek into the z = 0.2 H I Uni-verse, providing us with a glimpse of the gas-rich galaxy populationthat will be detected by the SKA pathfinders, and whose physicalconditions might resemble those of higher redshift systems. Be-cause of its exquisite sensitivity, the Arecibo radio telescope hasa critical advantage for the detection of weak H I signals. Witha collecting area that is equivalent to one-tenth of the full SKA,Arecibo can already detect H I emission at z ∼ 0.2 with integrationsof less than 10 h per object, as opposed to the few hundred hoursnecessary with current interferometers. Naturally, the downside ofsingle-dish observations is their lack of spatial resolution, hencedetecting H I emission at these redshifts is practically restricted tocarefully selected galaxies in low-density environments. However,this drawback is partly compensated by the fact that gas-rich galax-ies (which are the most likely to be detected) are preferentially foundin isolation (Papastergis et al. 2013). Indeed, the feasibility of theseobservations was demonstrated by our pilot survey (Catinella et al.2008).

In this paper, we present the completed HIGHz survey, whichmeasured the H I content of 39 optically selected galaxies in the0.17 < z < 0.25 redshift interval. We describe our entire observingcampaign and investigate the properties of the detected galaxies.In addition to the highest redshift detection of H I emission from agalaxy to date (z = 0.25), this sample includes some of the mostH I massive galaxies currently known. We discuss the relevance ofHIGHz in the context of other surveys of exceptionally gas-richgalaxies, both in the local and in the higher redshift Universe.

All the distance-dependent quantities in this work are computedassuming � = 0.3, � = 0.7 and H0 = 70 km s−1 Mpc−1. ABmagnitudes are used throughout the paper.

2 SA M P L E SE L E C T I O N A N D A R E C I B OO B S E RVAT I O N S

As mentioned in Catinella et al. (2008), where we presented ourinitial results, this programme started as a pilot survey to detectH I emission from disc galaxies at z > 0.05, i.e. beyond the red-shift of past Arecibo surveys. We soon realized that the frequencyinterval corresponding to 0.11 < z < 0.16 was inaccessible becauseof Radio Frequency Interference (RFI), therefore we concentratedour efforts on the z > 0.16 targets. By then, the first data release(Abazajian et al. 2003) of the Sloan Digital Sky Survey (SDSS; Yorket al. 2000) became available, which provided the ideal data base tosearch for galaxies with potentially large H I content, although thespectroscopic coverage of the sky area accessible to Arecibo wasvery limited.

The targets for H I spectroscopy were selected from the mostrecent SDSS spectroscopic data release available at the time of the

observations (see Table 1) and according to the following criteria:(a) objects spectroscopically classified as galaxies; (b) observablefrom Arecibo during night-time (to minimize the impact of RFI andsolar standing waves on our data); (c) redshift 0.16 < z < 0.27. Thisinterval corresponds to frequencies from 1120 to 1220 MHz and isset by a filter; (d) inclination i ≥ 45◦ (computed from the axis ratio inr band as in Catinella et al. 2010), for use in disc scaling relations. Inthe most recent run, we relaxed this condition to include inclinationsi ≥ 30◦; (e) presence of Hα emission in the SDSS fibre, with linewidth between 100 and 700 km s−1 and equivalent width between5 and 50 Å (to avoid extreme SFRs, usually associated with mergersystems and/or starburst galaxies); (f) exponential disc profile, basedon the likelihood of exponential versus DeVaucouleurs fit to ther-band profile (likelihood(exp)/likelihood(DeV)>106).

After compiling the candidate list for each run, we discardedgalaxies with redshifts corresponding to the frequency of knownRFI, and/or in the vicinity of NVSS (NRAO VLA Sky Survey,Condon et al. 1998) continuum sources that would cause ripples inthe baselines. We then carefully inspected the SDSS image of eachremaining galaxy, and excluded those with interacting or peculiarappearance, and/or with H I emission possibly contaminated by thatof nearby objects. Specifically, we discarded targets with galaxiesof similar size or luminosity lying within a 4 arcmin radius (twicethe Arecibo beam). This was a crucial step in order to minimize thelikelihood that our H I measurements are contaminated by compan-ions with potentially comparable H I content, as the half power fullwidth of the beam at the frequencies of our observations, ∼4 arcmin,subtends ∼800 kpc at z = 0.2. From the final list, we gave priorityto the galaxies that looked more promising, i.e. those with largestapparent size and/or presence of spiral arms. Hence, although theparent sample out of which the targets were extracted is volume-limited and well defined, we deliberately picked our galaxies oneby one, in order to maximize our chances of detecting H I emission.Not surprisingly, we obtained a sample that is strongly biased to-wards H I-massive objects. We targeted 49 galaxies with effectiveon-source integration times ranging between 52 and 260 min, anddetected 39 objects. The 10 galaxies that were not detected haveredshifts between 0.23 and 0.26 (except one with z = 0.17). Anumber of other potential targets were observed for a short time andabandoned because of various issues (too close to RFI, continuumsources, bad baselines).

The H I data were collected under four Arecibo programmes andduring several observing runs, which took place between Fall 2003and Spring 2011. Table 1 lists the Arecibo programme identifiers,the dates of the observations, the targeted redshift interval, the SDSSdata release used for target selection, and the amount of telescopetime allocated. The A1803 programme was split into a Fall and aSpring portion (we indicate the former as A1803f), and was partlydevoted to lower redshift observations. The A2428 programme wasaccepted in 2008 but scheduled on the telescope three years later.

Table 1. Arecibo observing runs.

Project ID Dates Redshift SDSS selection Allocation

A1803f 2003 Oct 8–18; Nov 8,10,22,23; Dec 6 0.09–0.18 DR1a 62 h (18 h @ z > 0.16)A1803 2004 Mar 8–10,24–28; Apr 11-16,25 0.09–0.25 DR1a,DR2b 121 h (110 h @ z > 0.16)A2008 2005 Apr 19–26 0.16–0.25 DR3c 72 hA2270 2007 Mar 8–16 0.16–0.25 DR5d 80 hA2428 2011 May 3–15 0.16–0.32 DR7e 76 h

Notes:aAbazajian et al. (2003); bAbazajian et al. (2004); cAbazajian et al. (2005); dAdelman-McCarthy et al. (2007);eAbazajian et al. (2009).

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3528 B. Catinella and L. Cortese

Including all the overheads, the total time spent on the z > 0.16targets was 356 h.

The observations were all carried out on site and in standardposition-switching mode: each observation consisted of an on/offsource pair, each integrated for 4 min (5 min in the last two observingruns), followed by the firing of a calibration noise diode. We used theL-band wide receiver, which operates in the frequency range 1120–1730 MHz, with a 1120–1220 MHz filter and a 750 MHz narrow-band (60 MHz) front-end filter to limit the impact of RFI on ourobservations. The interim correlator was used as a backend, and thespectra were recorded every second with nine-level sampling. Twocorrelator boards, each configured for 12.5 MHz bandwidth, onepolarization, and 2048 channels per spectrum (yielding a velocityresolution of 1.8 km s−1 at 1200 MHz before smoothing) werecentred at or near the frequency corresponding to the SDSS redshiftof the target. The Doppler correction for the motion of the Earthwas applied during offline processing. We note that, apart from thedifferent redshift interval, the observing setup and data reductionpipeline are identical to those used for the GALEX Arecibo SDSSSurvey (GASS; Catinella et al. 2010).

3 DATA R E D U C T I O N

The data reduction was performed in the IDL environment usingour own routines, which are based on the standard Arecibo dataprocessing library developed by Phil Perillat. In summary, the datareduction of each polarization and on-off pair includes Hanningsmoothing, bandpass subtraction, RFI excision, and flux calibra-tion. Once the data are flux calibrated, a total spectrum is obtainedfor each of the two orthogonal linear polarizations by combininggood quality records (those without serious RFI or standing waves).Each pair is weighted by a factor of 1/rms2, where rms is the rootmean square noise measured in the signal-free portion of the spec-trum. The two polarizations are separately inspected and averaged,yielding the final spectrum.

After boxcar smoothing and baseline subtraction, the H I-line pro-files are ready for the measurement of redshift, rotational velocityand integrated H I line flux. Recessional and rotational velocities aremeasured at the 50 per cent peak level from linear fits to the edgesof the H I profile. Our measurement technique is explained in moredetail, e.g. in Catinella, Haynes & Giovanelli (2007, Section 2.2).

Below we describe in more detail two very important steps of ourdata reduction, RFI excision and flux calibration.

3.1 RFI excision

The presence of RFI in our observations, which target a frequencyinterval that is well below the radioastronomy protected band(1400–1427 MHz), was the main challenge for our survey.

RFI can be generated both internally (by electrical equipmentsuch as digital correlators and computers) and externally (e.g. bybroadcasting radio and television stations, mobile phones, airportradars, and telecommunication satellites, but also by natural sourcessuch as lightning). Single-dish radio telescopes are especially vul-nerable to this problem because all incoming RFI, entering by scat-tering or reflection, enters the system coherently (Fridman & Baan2001). Due to the variety of its possible sources, RFI spans a widerange of characteristics, both in the time and frequency domains; itcan also be strongly polarized or completely unpolarized.

A few examples of the most common types of RFI that affectedour observations are illustrated in Fig. 1. Each panel is the result ofa processed 4 min integration for the first correlator board (which

Figure 1. Examples of RFI affecting our data (from the A1803 run). Eachpanel shows the time-frequency representation of a processed on-off pair forthe first correlator board. Spectra are recorded every second, thus the verticalscale can be read as time in seconds; the full 12.5 MHz bandwidth is shownon the horizontal axis. The top two panels illustrate examples of RFI thatcan be easily excised (see also Fig. 2). The RFI in (c) cannot be removed,but most of the frequency bandpass is still useful for the observations; (d)shows an observation that is completely compromised.

shows one polarization only; the second board is almost identicalunless the interference is strongly polarized); each row is a singlespectrum, recorded every second. We always refer to frequencychannels along the x-axis and records along the y-axis. In (a) theRFI appears for a very short time (a few seconds at most, aroundrecord 200), affecting all the bandpass; in (b) it is localized inthe frequency domain (between 1184 and 1186 MHz, and between1190 and 1192 MHz), but is present in several records. In bothcases, the interference can be easily removed during data reductionby excising the affected records. These examples demonstrate theimportance of a fast dump rate for spectrum recording, in order tominimize the amount of data that must be discarded because of RFIcontamination. Panel (c) shows strong RFI signals at 1169 MHzand 1174 MHz; in this case the RFI cannot be simply removed byexcising bad records, but the observation is not be compromised aslong as the interference does not sit on top of the galaxy emission.Indeed, this example shows how critical an accurate knowledge ofthe galaxy redshifts was for this project, in order to anticipate atwhich frequencies the expected H I signal (which starts appearingonly after many pairs have been co-added) would lie. Lastly, thepresence of strong RFI outside the 12.5 MHz bandpass in (d) makesthe data unusable. Our first observations immediately showed thatthe 1220–1230 MHz and 1270–1280 MHz frequency intervals wereinaccessible due to type (d) RFI.

Accurate planning of our observations was essential in order tominimize the impact of RFI on our data. As mentioned in Section 2,in addition to avoiding targets whose H I emission would lie inthe proximity of known RFI of the type shown in Fig. 1(c), we

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3529

used front-end bandpass filters to restrict as much as possible thefrequency interval ‘seen’ by the correlator (to avoid RFI of the typeof Fig. 1d).

Ideally, it would be convenient to rely on automatic algorithmsto identify and remove RFI, but defining properties that distinguishRFI from astronomical signals is not a trivial task, especially in thepresence of weak interference. Such algorithms are generally basedon thresholding techniques, by which a portion of the data in the timeand/or frequency domain is discarded when its mean (or some otherstatistical indicator) exceeds a certain value. A thorough testing isusually necessary to determine how to set the thresholds. Contraryto automatic algorithms, the human eye can easily identify RFIfrom the inspection of time-frequency representations such as thoseshown in Fig. 1 and thus, although impractical, manual excisionachieves much better results. Because reliable RFI excision wasextremely important for our survey, we resorted to an admittedlytime-consuming, hybrid approach.

RFI excision is applied to the data before flux calibration,and its output is an RFI mask, which records the pixels in thetime-frequency domain marked for rejection for each polarizationand pair. The process includes the following steps.

(1) Processing each polarization of the on-off pair. Because weprocess the pairs first, RFI appears as an emission or absorptionfeature in the spectra depending on whether it is present/stronger inthe on or in the off scan, respectively (see Fig. 2). This also impliesthat we discard data when RFI is present in at least one of the on/offscans.

(2) For each frequency channel, an rms noise spectrum acrossthe time direction is computed, and an iterative linear fit performed,rejecting any points that deviate by 3σ or more from the fit. Fittingof the same channel is repeated until no more points are rejected, or

an iteration limit is reached. Optionally, the data can be smoothedover Nsm frequency channels before searching for RFI. For our dataset, we found that the best solution was to smooth by 0.3 MHz, andwe did this by default.

(3) The rejected points are flagged, and time-frequency imagesfor the two polarizations with the RFI mask overlaid (along withhistograms showing the noise in each record) are displayed. This isillustrated in the bottom-left panel of Fig. 2 for one polarization ofone of our observations.

(4) At this point we inspect how well the automatic excisionworked, and either run the program again or move to the next step.It is often necessary to run the program a few times with a differentsigma threshold for pixel rejection and/or changing the number ofchannels Nsm for spectral smoothing in order to obtain the bestresult. RFI of the type seen in Fig. 1(a) usually requires to maskentire records, as many of the pixels in the affected records deviateby less than 3σ from the average noise. This is done by choosing athreshold r for record rejection (records with r per cent or more badpixels are flagged).

(5) If necessary, we perform manual excision by selecting regionsof the time-frequency image that should be rejected. The selectedregion is added to the RFI mask generated in point 4. Manual exci-sion is usually needed for weak RFI that is not adequately maskedby the automatic algorithm, and for unusual RFI (e.g. drifting infrequency).

(6) Lastly, the spectrum obtained after applying the RFI maskis shown (Fig. 2, bottom right). Here we can go back to (1)if needed, reject the pair if the data are unusable, or proceedto the flux calibration (the pair is reprocessed using the finalRFI mask), keeping track of the number of records used at eachfrequency.

Figure 2. RFI excision. Top: time-frequency representation of an observation of AGC 212887, showing strong RFI near 1185 and 1191 MHz (left), andcorresponding spectrum (right). The RFI is in the off-source observation, thus it appears as absorption-like features in the spectrum. Bottom: the RFI mask isshown on top of the data (white pixels) on the left; the corresponding spectrum is RFI-free (right), except for a very narrow spike at 1192.1 MHz that is notidentified by the RFI-masking algorithm.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3530 B. Catinella and L. Cortese

This algorithm is designed to identify strong RFI features inthe time domain and spectrally broad RFI which is time-variable.RFI that is present in all records at a given frequency, such asthat in Fig. 1(c), is not masked. Fig. 2 shows an example of theapplication of our automatic RFI excision algorithm to our data,and the resulting H I spectrum before and after RFI masking (right-hand panels).

3.2 Flux calibration

Standard position-switched observations of a source with Areciboare followed by two measurements of a calibration diode, whichis turned on for 10 s and then off for another 10 s, while the tele-scope points to the blank sky (the off galaxy position). The noisediode has a known temperature as a function of frequency for eachpolarization, hence processing its on/off observations allows oneto determine the ‘system temperature’, and convert correlator unitsinto Kelvin degrees. The conversion to spectral flux density unitsis achieved by applying the gain curve, which provides the pointsource gain of the telescope in K Jy−1 for the specific receiver as afunction of frequency and zenith angle of the observation. As dis-cussed in van Zee et al. (1997) and Springob et al. (2005) and mostrecently in Haynes et al. (2011, see their section 5.2), H I line fluxdensities derived from targeted single-dish observations in moderndata sets are typically accurate to not better than 15 per cent.

A technical problem made all the calibration scans taken between2004 March 26 and 2005 May unusable, thus we had to resort toa non-standard flux calibration for part of the A1803 and all theA2008 data. Because of a bug in the telescope control software(fixed in 2005 June), selecting the 750 MHz narrow-band filter alsosilently triggered the ‘winking’ calibration mode, which is used forpulsar observations. This caused our noise diode to be switched onand off with a frequency of 25 Hz (i.e. every 40 ms), which had twoeffects: (a) render our calibration scans useless, and (b) inject extranoise in our observations, as the diode was on half of the time. Thiseffectively increased the system temperature Tsys by ∼15 per cent(estimated from 0.5 × Tdiode/Tsys ∼ 1.15, where Tdiode and Tsys areapproximately 10 and 33 K for our observations, respectively).

In order to calibrate these data, Phil Perillat at Arecibo kindly pro-vided us with System Equivalent Flux Density (SEFD) curves forour receiver, which are obtained from a fit to the system temperatureas a function of gain of the telescope. These fits give the point sourceSEFD of the telescope in Jy, as a function of frequency and zenithangle of the observation. We processed all the good calibrationscans taken in 2003 and 2004, compared the diode calibration withthe SEFD one, and found the two to be consistent within less than10 per cent. The SEFD curves are based on archival measurementsof the system temperature taken under standard conditions, i.e.Tsys ∼ 33 K at the frequencies of HIGHz. Hence, in order to ac-count for the increased system temperature caused by the winkingcalibration, we multiplied the conversion factor obtained from theSEFD curves by 1.15. Conservatively, we consider the flux calibra-tion of the galaxies affected by the winking calibration to have anadditional uncertainty of 15 per cent over the standard calibration,or a total uncertainty of ∼21 per cent. This problem affected partlyor completely 20 out of 39 galaxies, however the larger uncertaintyin those H I fluxes does not affect our conclusions at all.

To our knowledge, these are the longest observations of individualgalaxies done with Arecibo. As can be seen in Fig. 3, the measuredrms noise of our observations computed from the unsmoothed H I

spectra (i.e. at a velocity resolution of 1.8 km s−1 at 1200 MHz)decreases as the square root of the integration time, as expected.

Figure 3. The rms noise is plotted versus on-source integration time, Tint,for individual galaxies (grey lines); red circles are running averages (com-puted when at least three data points were available). The rms decreases as√

Tint as expected (dotted line).

4 DATA PR E S E N TAT I O N

4.1 H I catalogue

The measured H I parameters for the 39 detected galaxies are listedin Table 2, ordered by increasing right ascension.

Col. (1): identification code in the Arecibo General Catalog(AGC, maintained by M.P. Haynes and R. Giovanelli at CornellUniversity).

Col. (2): SDSS identifier.Col. (3): on-source integration time of the Arecibo observation,

Ton, in minutes. This number refers to on scans that were actuallycombined, and does not account for losses due to RFI excision(typically of order of a couple of per cent). The fraction of usabledata for our sample, i.e. the ratio between Ton and the total on-sourcetime actually spent on each galaxy, varied between 50 and 100 percent, with an average of 84 per cent. Pairs were discarded becauseof RFI that could not be excised or bad baselines.

Col. (4): velocity resolution of the final, smoothed spectrum inkm s−1.

Col. (5): redshift, z, measured from the H I spectrum. The erroron the corresponding heliocentric velocity, cz, is half the error onthe width, tabulated in the following column.

Col. (6): observed velocity width of the source line profile in kms−1, W50, measured at the 50 per cent level of each peak. The erroron the width is the sum in quadrature of the statistical and systematicuncertainties in km s−1 (the latter depend on the subjective choiceof the H I signal boundaries and are usually negligible, see Catinellaet al. 2010 for details).

Col. (7): velocity width corrected for instrumental broadeningand cosmological redshift only, W50

c, in km s−1 (see Catinella et al.2012b). No inclination or turbulent motion corrections are applied.

Col. (8): observed, integrated H I-line flux density in Jy km s−1, F≡ ∫

Sdv, measured on the smoothed and baseline-subtracted spec-trum. The reported uncertainty is the sum in quadrature of the sta-tistical and systematic errors (see Catinella et al. 2010 for details).

Col. (9): rms noise of the observation in mJy, measured on thesignal- and RFI-free portion of the smoothed spectrum.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3531

Tabl

e2.

HIpr

oper

ties

ofH

Ide

tect

ions

.

Ton

�v

W50

W50

cF

rms

log

MH

I

AG

CSD

SSID

(min

)(k

ms−

1)

z(k

ms−

1)

(km

s−1)

(Jy

kms−

1)

(mJy

)S/

N(M

�)lo

gM

HI/

M�

QPr

ojec

tID

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

1017

50J0

0361

0.70

+14

2246

.415

628

0.19

087

254

3±

2144

40.

19±

0.03

0.12

8.2

10.5

1−

0.47

1a1

803f

1017

10J0

0384

0.25

+15

3912

.988

270.

176

535

598

±4

497

0.16

±0.

050.

164.

510

.37

−0.

722

1803

f12

2040

J020

849.

45+

1304

42.5

5627

0.16

821

033

2±

2327

30.

34±

0.04

0.17

15.0

10.6

50.

281

a180

3f18

1518

J082

522.

13+

3259

53.6

112

270.

171

022

515

±23

428

0.26

±0.

050.

178.

010

.55

−0.

361

a180

318

1593

J084

800.

91+

0618

37.2

176

290.

220

272

256

±9

198

0.13

±0.

040.

186.

010

.47

−0.

841

a200

8,a2

270

1815

59J0

8573

3.51

+03

4456

.813

228

0.18

909

454

6±

1344

70.

25±

0.05

0.17

7.4

10.6

3−

0.41

1a1

803

1983

10J0

9031

5.37

+07

3332

.423

029

0.21

751

748

1±

338

30.

15±

0.03

0.12

7.2

10.5

3−

0.56

1a2

270

1918

26J0

9195

4.42

+08

5344

.011

628

0.18

990

577

0±

2163

60.

42±

0.06

0.17

8.7

10.8

5−

0.34

1a2

008

1917

28J0

9195

7.00

+01

3851

.696

270.

176

425

219

±6

174

0.18

±0.

030.

189.

310

.42

−0.

471

a180

319

1787

J094

921.

53+

0442

39.7

104

280.

190

145

601

±5

493

0.26

±0.

060.

186.

210

.64

−0.

331

a180

319

1838

J095

923.

26+

0651

09.1

7238

0.18

999

853

8±

1343

60.

19±

0.06

0.16

5.0

10.5

1−

0.52

3a2

008

2021

25J1

0284

8.86

+04

2331

.911

228

0.20

701

048

0±

1738

60.

10±

0.04

0.16

3.4

10.2

9−

0.81

3a1

803

2087

50J1

0513

8.96

+06

1751

.790

410.

223

601

466

±79

364

0.15

±0.

060.

174.

310

.56

−0.

692

a242

820

8751

J105

230.

22+

0823

24.3

110

280.

187

710

492

±23

403

0.19

±0.

040.

147.

610

.51

−0.

421

a242

821

2966

J110

730.

34+

0723

50.2

124

360.

206

446

494

±16

395

0.13

±0.

040.

144.

410

.41

−0.

692

a200

821

2967

J111

353.

47+

0949

59.7

150

340.

247

755

414

±5

318

0.15

±0.

040.

165.

410

.63

0.01

3a2

270

2129

41J1

1164

5.15

+05

4210

.018

029

0.22

362

538

3±

3930

10.

18±

0.04

0.17

7.0

10.6

3−

0.36

1a1

803

2128

87J1

1463

4.33

+05

3442

.017

628

0.19

573

277

2±

2363

40.

29±

0.04

0.11

9.0

10.7

2−

0.25

1a1

803

2128

71J1

2005

2.83

+03

3754

.910

028

0.19

020

874

2±

761

20.

30±

0.06

0.17

6.2

10.7

1−

0.56

1a1

803

2243

21J1

2094

8.14

+10

0822

.580

280.

187

520

417

±14

339

0.26

±0.

050.

198.

910

.64

−0.

291

a200

822

8389

J122

036.

32+

0819

04.2

7529

0.22

176

037

7±

1029

70.

17±

0.05

0.18

6.3

10.6

0−

0.44

1a2

428

2391

10J1

3043

4.51

+27

1334

.975

290.

220

002

533

±3

425

0.30

±0.

050.

188.

310

.84

−0.

261

a242

823

2041

J132

325.

27+

0437

07.6

203

270.

186

225

591

±29

487

0.19

±0.

040.

127.

510

.50

−0.

551

a180

3,a2

270

2320

46J1

3332

7.34

+04

2244

.814

431

0.18

836

081

8±

567

50.

24±

0.05

0.14

5.1

10.6

0−

0.38

2a1

803,

a200

823

2055

J134

211.

36+

0532

11.5

9228

0.20

298

746

4±

1037

40.

26±

0.05

0.18

8.4

10.7

0−

0.12

1a2

008

2321

27J1

3594

9.19

+05

2731

.814

037

0.22

806

860

4±

1747

70.

22±

0.05

0.14

6.1

10.7

4−

0.50

1a2

270

2420

91J1

4052

2.72

+05

2814

.615

628

0.19

530

250

2±

440

80.

25±

0.04

0.13

10.3

10.6

5−

0.38

1a1

803

2420

73J1

4085

6.04

+05

1616

.314

028

0.19

062

542

2±

234

30.

15±

0.03

0.13

7.4

10.4

2−

0.37

1a2

270

2495

58J1

4215

1.16

+10

0623

.685

280.

193

841

383

±6

309

0.15

±0.

030.

147.

210

.42

−0.

871

a242

824

2147

J142

735.

69+

0334

34.2

176

300.

245

370

499

±11

389

0.28

±0.

050.

188.

010

.90

−0.

361

a180

324

9559

J144

518.

88+

0250

12.3

9528

0.19

035

533

1±

7326

60.

18±

0.04

0.18

7.3

10.4

7−

0.70

1a2

428

2495

60J1

4482

3.96

+12

5551

.916

435

0.19

541

577

1±

963

00.

19±

0.04

0.12

5.0

10.5

3−

0.55

3a2

428

2525

80J1

5133

7.28

+04

1921

.184

270.

174

961

478

±8

396

0.32

±0.

060.

218.

410

.66

−0.

121

a180

325

2654

J153

240.

56+

3234

28.8

9232

0.20

113

679

2±

464

60.

26±

0.06

0.16

5.0

10.6

9−

0.35

2a2

008

2526

41J1

5340

0.67

+00

1133

.211

628

0.20

357

459

1±

8647

90.

31±

0.09

0.29

4.8

10.7

8−

0.24

3a1

803

2620

26J1

6043

3.51

+31

0900

.011

241

0.22

316

864

2±

250

90.

13±

0.05

0.14

3.2

10.4

9−

0.85

3a2

008,

a227

026

2029

J160

938.

00+

3129

58.5

136

290.

218

648

642

±56

515

0.34

±0.

050.

159.

010

.89

−0.

411

a200

826

2033

J162

515.

41+

2805

30.3

260

280.

187

113

643

±43

530

0.24

±0.

030.

119.

410

.59

−0.

491

a227

026

2015

J165

940.

12+

3443

07.8

9628

0.19

833

150

4±

1640

90.

42±

0.05

0.19

12.0

10.8

90.

061

a180

3

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3532 B. Catinella and L. Cortese

Figure 4. Distributions of (a) redshifts, (b) rotational velocities, and (c) H I masses for this sample (solid histograms). Hatched histograms do not include themarginal detections.

Col. (10): signal-to-noise ratio of the H I spectrum, S/N, estimatedfollowing ALFALFA and GASS (see e.g. Catinella et al. 2010).

Col. (11): base-10 logarithm of the H I mass, MH I, in solar units,computed via:

MH I

M�= 2.356 × 105

1 + z

[dL(z)

Mpc

]2 ( ∫S dv

Jy km s−1

)(1)

where dL(z) is the luminosity distance to the galaxy at redshift z asmeasured from the H I spectrum.

Col. (12): base-10 logarithm of the H I mass fraction, MH I/M�.Col. (13): quality flag, Q. Code 1 indicates detections with a S/N

ratio of 6.0 or higher. Code 2 is assigned to lower S/N, but still se-cure detections, and code 3 to marginal detections. The separationbetween codes 2 and 3 is not simply by S/N, but takes into accountH I profile and baseline quality – code 3 detections have more un-certain H I parameters, hence are shown with different symbols inour plots. In all cases, the H I redshift is consistent with the SDSSmeasurement.

Col. (14): Arecibo project identifier (see Table 1).For the 10 galaxies that were not detected (see section 2), we did

not obtain stringent constraints on the H I masses – the upper limitsare comparable to or higher than the H I masses of the detections.This is because, after the initial 1 or 2 h of on-source integration,we completed the observations of the target only if there was a hintof galaxy H I emission at the expected frequency.

SDSS postage stamp images and H I spectra of the HIGHzdetections can be found in the Appendix. Fig. 4 presents the distri-butions of measured redshifts, rotational velocities, and H I massesfor both high-quality (hatched) and marginal (solid histograms) de-tections. The HIGHz galaxies have H I masses that vary between1.9 and 7.9 × 1010 M�, at the top end of the H I mass function(H IMF; Martin et al. 2010). Because the H IMF drops rapidly abovelog MH I,∗/M� = 9.96, these systems are rare in the local Universe(for instance, galaxies with MH I≥5 × 1010 M� are over a factor of100 less common than MH I,∗ ones).

4.2 SDSS and GALEX data

This section summarizes the quantities derived from optical and UVdata used in this paper. All the optical parameters listed below were

obtained from Structured Query Language queries to the SDSS DR7data base server,1 unless otherwise noted.

The NUV magnitudes for our sample were obtained from theGALEX Unique Source Catalogs2 (Seibert et al. 2012). The mea-sured NUV−r colours are corrected for Galactic extinction follow-ing Wyder et al. (2007), from which we obtained ANUV − Ar =1.9807Ar (where the extinction Ar is available from the SDSS database and reported in Table 3 below). We did not apply internal dustattenuation corrections.

Table 3 lists the relevant SDSS and UV quantities for the galaxiespublished in this work, ordered by increasing right ascension:

Cols. (1) and (2): AGC and SDSS identifiers.Col. (3): SDSS redshift, zSDSS. The average uncertainty of SDSS

redshifts for this sample is 0.000 15.Col. (4): base-10 logarithm of the stellar mass, M�, in solar units.

Stellar masses are from the MPA/JHU SDSS DR7 catalogue,3

and are derived from SDSS photometry using the methodologydescribed in Salim et al. (2007, a Chabrier 2003 initial mass func-tion is assumed). In one case, we replaced the SDSS stellar massestimate with a different one (AGC 232041, see below).

Cols. (5) and (6): radii containing 50 and 90 per cent of thePetrosian flux in r band, R50 and R90, respectively, in arcsec.

Col. (7): radius containing 90 per cent of the Petrosian flux in rband, R90, in kpc.

Col. (8): base-10 logarithm of the stellar mass surface density,μ�, in M� kpc−2. This quantity is defined as μ� = M�/(2πR2

50),with R50 in kpc units.

Col. (9): Galactic extinction in r band, extr, in magnitudes, fromSDSS.

Col. (10): r-band model magnitude from SDSS, r, corrected forGalactic extinction.

Col. (11): minor-to-major axial ratio from the exponential fit in rband, (b/a)r, from SDSS.

Col. (12): inclination to the line-of-sight in degrees, computedfrom (b/a)r in the previous column assuming an intrinsic axial ratioq0 = 0.20 as in Catinella et al. (2012a).

1http://cas.sdss.org/dr7/en/tools/search/sql.asp2http://archive.stsci.edu/prepds/gcat/3http://www.mpa-garching.mpg.de/SDSS/DR7/; we used the improved stel-lar masses from http://home.strw.leidenuniv.nl/ jarle/SDSS/.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3533

Tabl

e3.

SDSS

and

UV

para

met

ers.

log

M�

R50

R90

R90

log

μ�

ext r

rin

clSF

RN

UV

−rA

GC

SDSS

IDz S

DSS

(M�)

(arc

sec)

(arc

sec)

(kpc

)(M

�kp

c−2)

(mag

)(m

ag)

(b/a

) r(d

eg)

(M�

yr−1

)(m

ag)

GA

LE

X(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)(1

0)(1

1)(1

2)(1

3)(1

4)(1

5)

1017

50J0

0361

0.70

+14

2246

.40.

1909

10.9

82.

425.

1723

.38.

110.

1817

.52

0.46

365

6.4

3.11

MIS

1017

10J0

0384

0.25

+15

3912

.90.

1769

11.0

92.

847.

0729

.38.

150.

1617

.59

0.39

070

2.6

3.34

MIS

1220

40J0

2084

9.45

+13

0442

.50.

1682

10.3

71.

724.

4817

.67.

910.

2517

.63

0.36

372

8.0

1.91

MIS

1815

18J0

8252

2.13

+32

5953

.60.

1709

10.9

12.

505.

3721

.48.

110.

1217

.49

0.42

967

10.4

2.43

MIS

1815

93J0

8480

0.91

+06

1837

.20.

2201

11.3

11.

773.

9721

.08.

570.

1417

.51

0.60

355

35.4

––

1815

59J0

8573

3.51

+03

4456

.80.

1894

11.0

42.

595.

7425

.78.

120.

1317

.50

0.51

661

13.7

2.66

MIS

1983

10J0

9031

5.37

+07

3332

.40.

2174

11.0

91.

944.

4123

.08.

280.

1518

.32

0.26

380

27.7

3.09

MIS

1918

26J0

9195

4.42

+08

5344

.00.

1899

11.1

93.

167.

0331

.58.

090.

1417

.43

0.28

079

13.1

2.76

AIS

1917

28J0

9195

7.00

+01

3851

.60.

1763

10.8

92.

034.

6119

.08.

250.

0817

.24

0.67

949

14.7

1.80

MIS

1917

87J0

9492

1.53

+04

4239

.70.

1903

10.9

72.

625.

5725

.08.

030.

1117

.66

0.47

564

7.5

2.66

MIS

1918

38J0

9592

3.26

+06

5109

.10.

1902

11.0

31.

834.

3419

.58.

400.

1117

.69

0.45

565

11.1

2.81

AIS

2021

25J1

0284

8.86

+04

2331

.90.

2069

11.1

02.

845.

5227

.28.

010.

0917

.74

0.34

174

7.9

2.95

MIS

2087

50J1

0513

8.96

+06

1751

.70.

2239

11.2

53.

527.

0237

.97.

900.

0917

.31

0.81

137

25.1

2.67

MIS

2087

51J1

0523

0.22

+08

2324

.30.

1880

10.9

32.

475.

0122

.28.

060.

0717

.57

0.46

165

4.7

2.38

MIS

2129

66J1

1073

0.34

+07

2350

.20.

2066

11.1

01.

363.

2115

.88.

650.

0917

.78

0.60

754

16.4

3.05

AIS

2129

67J1

1135

3.47

+09

4959

.70.

2482

10.6

21.

503.

4721

.07.

900.

0618

.40

0.68

348

12.7

2.02

AIS

2129

41J1

1164

5.15

+05

4210

.00.

2239

10.9

92.

734.

9126

.57.

860.

2117

.53

0.58

956

11.9

1.85

AIS

2128

87J1

1463

4.33

+05

3442

.00.

1961

10.9

72.

034.

4820

.88.

220.

0517

.78

0.36

072

9.7

––

2128

71J1

2005

2.83

+03

3754

.90.

1904

11.2

72.

936.

7530

.48.

230.

0717

.23

0.42

767

24.7

––

2243

21J1

2094

8.14

+10

0822

.50.

1879

10.9

32.

936.

4028

.47.

900.

0717

.27

0.69

847

11.5

2.58

AIS

2283

89J1

2203

6.32

+08

1904

.20.

2218

11.0

42.

425.

3628

.68.

020.

0617

.70

0.63

652

4.2

1.99

MIS

2391

10J1

3043

4.51

+27

1334

.90.

2201

11.1

04.

488.

4144

.57.

550.

0317

.33

0.63

052

4.9

2.28

MIS

2320

41J1

3232

5.27

+04

3707

.60.

1861

11.0

5�1.

984.

5920

.19.

690.

0817

.53

0.51

161

20.7

––

2320

46J1

3332

7.34

+04

2244

.80.

1888

10.9

82.

435.

6825

.38.

110.

0817

.60

0.38

071

6.0

2.78

AIS

2320

55J1

3421

1.36

+05

3211

.50.

2031

10.8

22.

365.

4526

.37.

900.

0817

.75

0.54

059

9.1

2.31

AIS

2321

27J1

3594

9.19

+05

2731

.80.

2281

11.2

42.

365.

0928

.08.

210.

0817

.70

0.60

754

17.9

3.06

MIS

2420

91J1

4052

2.72

+05

2814

.60.

1954

11.0

33.

126.

4229

.77.

910.

0717

.23

0.45

865

17.5

2.63

MIS

2420

73J1

4085

6.04

+05

1616

.30.

1906

10.7

92.

114.

6120

.88.

030.

0717

.66

0.57

657

12.7

2.37

MIS

2495

58J1

4215

1.16

+10

0623

.60.

1938

11.2

93.

336.

5430

.08.

130.

0817

.00

0.79

039

16.8

2.62

MIS

2421

47J1

4273

5.69

+03

3434

.20.

2455

11.2

62.

495.

8735

.18.

110.

0917

.54

0.62

953

26.6

2.49

MIS

2495

59J1

4451

8.88

+02

5012

.30.

1906

11.1

72.

626.

2128

.08.

230.

1117

.28

0.77

340

15.6

3.01

MIS

2495

60J1

4482

3.96

+12

5551

.90.

1958

11.0

83.

086.

3329

.47.

970.

0717

.41

0.71

745

9.8

2.70

MIS

2525

80J1

5133

7.28

+04

1921

.10.

1754

10.7

83.

607.

9432

.67.

640.

1317

.57

0.44

166

6.1

2.38

MIS

2526

54J1

5324

0.56

+32

3428

.80.

2001

11.0

42.

445.

3325

.38.

120.

0617

.66

0.36

572

16.3

––

2526

41J1

5340

0.67

+00

1133

.20.

2036

11.0

22.

244.

8423

.58.

150.

2117

.49

0.54

759

6.2

2.07

MIS

2620

26J1

6043

3.51

+31

0900

.00.

2228

11.3

43.

186.

8236

.68.

070.

0817

.16

0.50

762

26.3

2.75

AIS

2620

29J1

6093

8.00

+31

2958

.50.

2188

11.3

02.

876.

4433

.88.

150.

0917

.20

0.60

854

20.5

2.84

AIS

2620

33J1

6251

5.41

+28

0530

.30.

1870

11.0

83.

266.

8230

.17.

970.

1917

.66

0.24

782

5.1

––

2620

15J1

6594

0.12

+34

4307

.80.

1982

10.8

32.

695.

5926

.37.

830.

0617

.54

0.59

155

7.7

2.15

MIS

Not

e:�SD

SSva

lue

was

12.3

7(s

eete

xt).

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3534 B. Catinella and L. Cortese

Col. (13): total SFR in M� yr−1, from the MPA/JHU SDSSDR7 catalogue. These SFRs are based on the technique discussedin Brinchmann et al. (2004).

Col. (14): NUV−r observed colour, corrected for Galactic ex-tinction.

Col. (15): source of the UV photometry: GALEX All-sky ImagingSurvey (AIS) or Medium Imaging Survey (MIS; see Martin et al.2005).

All our targets have SDSS MPA/JHU stellar masses in the in-terval log M�/M�=[10.3, 11.4] except for AGC 232041, whichhas an unreasonably large value of log M�/M�=12.37. Hence wecomputed new stellar mass estimates from the i-band luminosity andthe g − i colour following Zibetti, Charlot & Rix (2009), and ap-plying k-corrections based on analytical approximations by Chilin-garian, Melchior & Zolotukhin (2010). For the other galaxies in oursample, there is an excellent correlation between these estimatesand the MPA/JHU values, with an offset of 0.21 dex (Zibetti’s val-ues are systematically larger). Hence, we replaced the stellar massof AGC 232041 with the new estimate, log M�/M�=11.05 (aftersubtracting the systematic offset), which falls well within the stellarmass interval of the other galaxies in our sample. This galaxy isidentified by a different symbol in our plots.

5 R ESULTS

By selection, the HIGHz sample includes unusually H I-richgalaxies (see Fig. 4). Are these the tip of the iceberg of the z ∼0.2 disc galaxy population, or just peculiar objects? Are their hugegas reservoirs due to recent accretion, or are these systems simplyscaled-up versions of local gas-rich discs?

The properties of the HIGHz systems are best understood whencompared with those of a representative sample of local galaxieswith similar stellar masses, for which homogeneous H I measure-ments are available. Such a reference sample is uniquely providedby GASS, which measured the H I content of ∼800 galaxies selectedonly by stellar mass (10 < log (M�/M�) < 11.5) and redshift (0.025< z < 0.05). Not only GASS and HIGHz span the same stellar massinterval, but the H I observations were also taken and processed inthe same way. We use here the final data release of GASS (Catinellaet al. 2013) and, for comparisons that do not involve the gas con-tent, the full parent sample, which is the super-set of 12 006 galaxiesmeeting GASS selection criteria out of which the targets for Areciboobservations were extracted.

We plot the HIGHz galaxies on the main gas fraction scalingrelations identified by GASS in Fig. 5. Clockwise from the top left,we show how the gas mass fraction MH I/M� depends on stellar mass,stellar mass surface density, observed NUV−r colour (a proxy forspecific SFR, or SFR per unit of stellar mass) and concentrationindex R90/R50 (a proxy for bulge-to-total ratio) for both samples.HIGHz H I detections are shown as dark blue circles, while marginaldetections are presented in light blue; the magenta circle is AGC232041, for which we computed our own estimate of stellar mass(see Section 4.2). GASS detections and non-detections are indicatedby grey dots and green triangles, respectively. For consistency withHIGHz, we recomputed stellar mass surface densities for GASSusing r-band Petrosian radii (as opposed to z band, as in the GASSdata release papers).

The gas fraction versus stellar mass plot confirms that the HIGHzsample is extremely H I-rich: these galaxies have MH I/M� ratios thatare a factor of ∼10 higher than the average value of GASS at thesame stellar masses. These large gas fractions are comparable to

those of Malin 1 (log MH I/M�=10.82, log M�/M�=10.88, Lelli,Fraternali & Sancisi 2010; Huang et al. 2014) and HIZOA J0836-43(log MH I/M�=10.88, log M�/M�=10.64, Cluver et al. 2010), twoof the most H I-massive galaxies known.

Interestingly, however, the large difference between HIGHz andGASS samples completely disappears if we consider the other scal-ing relations shown in Fig. 5. At fixed stellar mass surface densityand NUV−r colour, HIGHz galaxies lie exactly on the trends fol-lowed by local massive galaxies. This automatically implies that, atfixed stellar mass, HIGHz galaxies should be outliers in the colourstellar mass diagram (and in the μ�–M� plot, see next section). In-deed, as shown in the left-hand panel of Fig. 6, HIGHz galaxies lieoutside the region that includes most of the GASS parent sample,indicated by the grey contours. Specifically, the z ∼ 0.2 galaxies areunusually blue for their stellar masses, and similar systems are rarein the local Universe. As NUV−r is an excellent proxy for specificSFR, this implies that HIGHz galaxies formed the bulk of their starsat later times than the typical massive galaxy at z = 0 (see also thenext section). We note that the peculiarity of our galaxies would notappear in an optical colour magnitude diagram, as all our targets lieon the optical red sequence. This is not surprising because, at suchhigh stellar masses, optical colours do not properly trace currentstar formation activity (e.g. Cortese 2012).

The results presented in Fig. 5 support the idea that the H I contentof galaxies is mainly driven by their colour and stellar mass surfacedensity, with the trend with stellar mass being just a secondary, notphysically driven relation (e.g. Catinella et al. 2010; Fabello et al.2011). This is reinforced in the right-hand panel of Fig. 6, whichshows the gas fraction plane defined by GASS. This is a relationbetween the gas fraction measured by our H I observations and thatpredicted from the NUV−r colours and stellar surface densities ofthe galaxies. We use here the gas fraction prediction calibrated onGASS galaxies with NUV−r≤4.5 mag (see Catinella et al. 2013),4

which is clearly the most appropriate for HIGHz. Our z ∼ 0.2galaxies follow the same relation identified by GASS, implyingthat their gas content is exactly what is expected based on theircolours and stellar mass surface densities.

Lastly, also the rotational velocities of HIGHz galaxies areconsistent with those expected from nearby systems. This is shownin Fig. 7, which reproduces the baryonic Tully–Fisher (TF) relationfor the subset of GASS galaxies with inclinations larger than 40◦

and R90/R50≤2.8 from Catinella et al. (2012a). Our sample meetsthese two criteria by selection (see Fig. 5 and Table 3), except fortwo galaxies with inclinations slightly below the cut (37◦ and 39◦)that we show anyway. Overall, the z ∼ 0.2 sample lies on the bary-onic TF relation, except for three outliers on the low velocity side(from top to bottom, these are AGC 181593, 191728, and 122040)and a marginal H I detection (AGC 249560) on the high-velocityside. Note however that there are similar outliers at lower baryonicmass also in GASS.

To summarize, the comparison with GASS shows that the HIGHzgalaxies have unusually high gas content (in terms of both H I massand H I gas fraction) and blue NUV−r colours for their stellarmass, but are not outliers in the other scaling relations, including

4The calibration of the gas fraction plane was obtained with z-band stellarsurface densities, hence the coefficients shown in Fig. 6 are strictly forz-band μ�, but we plotted r-band μ� instead. Because the difference betweenz-band and r-band calibrations is barely noticeable, we prefer not to providea new version of the gas fraction plane in this paper, and simply note thatthis small difference does not affect our conclusions at all.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3535

Figure 5. Gas fraction scaling relations. The H I mass fraction of the sample is plotted here as a function of stellar mass, stellar mass surface density,concentration index, and observed NUV−r colour (large blue symbols; light blue circles are marginal detections, and the magenta circle indicates AGC 232041– see text). For comparison, we also show the H I detections (grey dots) and the non-detections plotted at their upper limits (green upside-down triangles) fromthe GASS sample. The dashed line in the top-left panel indicates the H I detection limit of GASS.

the gas fraction plane. Their low values of stellar mass surfacedensity and concentration index are typical of disc-dominatedsystems, and star formation appears to proceed as expected, despitetheir huge H I reservoirs, with specific SFRs that are on averageone order of magnitude larger than those of GASS H I detections.In other words, all observational evidence suggests that the HIGHzgalaxies are rare, scaled-up versions of disc galaxies in the localUniverse, with no clear signs of peculiarity or recent interaction.Hence, although these objects are not spatially resolved in H I, itseems reasonable to assume that their H I is distributed in a disc aswell, rotating at the expected velocity given the baryonic mass of thesystem.

Incidentally, the results presented above should also dispel anydoubts that our H I observations might be significantly contaminatedby beam confusion. This is because, if we had detected the inte-grated H I emission of multiple galaxies within the beam instead, theHIGHz systems would be outliers in the gas fraction plane and/orthe baryonic TF relation. This suggests that the effect of beamconfusion, if present, is well within our observational uncertainties,as expected given our careful target selection (see Section 2).

6 D I SCUSSI ON

Having established how HIGHz galaxies compare to the typicalmassive systems in the local Universe, we turn our attention to howthese might fit into our general picture of galaxy evolution. On oneside, we ask whether H I reservoirs as large as the ones we observedare still present at z ∼0, despite being rare, or have completelydisappeared. On the other side, it is natural to wonder if HIGHzgalaxies might be the progeny of the very gas-rich turbulent discsobserved at z ∼1 (Tacconi et al. 2013). To answer these questions,we compare HIGHz to some of the most gas-rich samples (withstellar masses above 1010 M�) currently known between z ∼0and ∼1.

6.1 Are there local analogues of the HIGHz galaxies?

As the most extreme example of massive, H I-rich galaxies at z

= 0, we use the HIghMass (Huang et al. 2014) sample, a set of34 exceptionally H I-rich local galaxies identified by ALFALFA.HIghMass galaxies were selected from the 40 per cent ALFALFA

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3536 B. Catinella and L. Cortese

Figure 6. Left: NUV−r colour–stellar mass diagram. Contours and grey-scales show the distribution of the GASS parent sample for comparison; blue symbolsindicate HIGHz galaxies as in Fig. 5. Galaxies in the HIGHz sample are unusually blue for their stellar masses. Right: GASS gas fraction plane, showing therelation between measured (y-axis) and predicted (x-axis) H I mass fractions; the dashed line indicates the 1:1 relation, and the symbols are as in Fig. 5. TheHIGHz galaxies are not outliers in this relation.

Figure 7. Baryonic TF relation. Grey dots are the ‘Tully-Fisher subset’ ofinclined, disc-dominated GASS galaxies, reproduced from fig. 2 of Catinellaet al. (2012a); the dotted line indicates the inverse fit to the GASS data points,which is in excellent agreement with the relation from McGaugh et al. (2000,red solid line). The HIGHz sample is shown with the same symbols used inFig. 5.

data release (Haynes et al. 2011) to have both large H I mass(above 1010 M�) and large H I gas fraction for their stellar mass(i.e. MH I/M� more than 1 σ above the average, see their fig. 1).Other samples of massive local galaxies claimed to be ‘unusuallyH I-rich’ (i.e. Bluedisks, Wang et al. 2013, and the H I ‘Monsters’,Lee et al. 2014) are, in fact, significantly less extreme than HIgh-Mass in terms of gas fractions, and therefore are not shown in ourplots. For consistency with our sample, we extracted stellar massesfor HIghMass from the MPA/JHU SDSS DR7 catalogue, and keptonly galaxies with log (M�/M�) ≥ 10. Stellar mass surface densi-ties of HIghMass galaxies were computed using r-band petrosianradii and axis ratios obtained via elliptical aperture photometry ofSDSS images (kindly provided by S. Huang, see section 2 in Huanget al. 2014).

In Fig. 8, we plot both HIGHz (blue) and HIghMass (red) onsome of the main scaling relations followed by local galaxies. Itis clear that HIghMass and HIGHz galaxies have similarly highH I gas fractions for their stellar masses (top-left panel), implyingthat z = 0 analogues of our galaxies do indeed exist. The fact thatsome of the z ∼ 0.2 galaxies have more extreme gas fractions isnot unexpected, as these were selected from a much larger volume(hence they are rarer) than the ALFALFA galaxies (which havez < 0.06). Moreover, both samples lie preferentially on the upperenvelope of the main sequence of star-forming galaxies (dashed linein the bottom-right panel, from equation 12 of Salim et al. 2007,which applies to galaxies with NUV−r<4 mag), confirming thattheir large H I reservoir is actively feeding the formation of newstars at a significant rate, as already hinted by Fig. 6.

Intriguingly, the only clear difference between HIGHz and HIgh-Mass is in their stellar distribution (bottom-left panel): HIGHz sys-tems have stellar surface densities that are ∼0.6 dex smaller thanHIghMass, implying that their optical radii are significantly largerthan what observed in local, massive discs.5 This is purely a selec-tion effect since, as mentioned in Section 2, we intentionally pickedthe galaxies with the largest apparent sizes from the SDSS imagesas targets for Arecibo observations. Therefore, we ended up with asample of very uncommon galaxies, even rarer than HIghMass.

As discussed by Huang et al. 2014, the low μ� values of HIghMasscompared to the rest of ALFALFA points out to their large halospin parameters. It is a well-known prediction of galaxy formationmodels that stellar discs formed in dark matter haloes with higherangular momentum content are more extended and have highergas fractions (e.g. Mo, Mao & White 1998; Boissier & Prantzos2000). Hence, HIGHz galaxies are most likely the tail of the highspin parameter distribution identified by HIghMass. Indeed, thegas fractions of HIGHz galaxies are consistent with the values

5We checked the reliability of our μ� measurements by taking into accountthe effect of seeing on the stellar radii (the median seeing at r band is1.43 arcsec, according to the SDSS DR7 web site), and the conclusionremains the same, as most points in the plot would move upward by lessthan the symbol size.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3537

Figure 8. Comparison with literature samples. Top left: H I gas fraction versus stellar mass, as in Fig. 5; HIGHz and HIghMass galaxies with stellar masseslarger than 1010 M� are shown as blue circles and red diamonds, respectively. Top right: total gas fraction versus stellar mass (see text for details on how thetotal gas fractions are computed). Stars and green squares indicate PHIBSS and DYNAMO galaxies, respectively; dark green squares connected by lines arethe four DYNAMO systems with available CO measurements (the leftmost one is an upper limit, as shown by the downward arrow). Bottom left: the stellarsurface density is plotted as a function of stellar mass for the same data sets; grey dots show the full GASS parent sample. Bottom right: SFR as a function ofstellar mass, same symbols as bottom-left panel. A dashed line indicates the star-forming main sequence from Salim et al. (2007, equation 12).

predicted for spin parameters in the range ∼0.07–0.09 (Boissier &Prantzos 2000). This confirms that the H I content, SFR and stellardistribution of our galaxies are what expected from their mass andangular momentum, and that no recent accretion and/or interaction(sometimes invoked to explain unusually gas-rich galaxies, e.g.Cluver et al. 2010) are required.

Lastly, HIGHz (and HIghMass, as discussed in Huang et al. 2014)systems are not analogues of giant low surface brightness galaxieslike Malin 1 (Bothun et al. 1987). Indeed, the low SFR of Malin1 (∼0.1 M� yr−1, Impey & Bothun 1989, but see also Lelli et al.2010) indicates that its huge H I reservoir (log MH I/M� = 10.83,Pickering et al. 1997; Lelli et al. 2010) is largely inert, whereasHIGHz galaxies are actively star forming, with an average SFR of13.5 M� yr−1 (see Table 1). This is demonstrated by the fact that theaverage depletion time (MH I/SFR) of HIGHz systems is ∼3 Gyr, inline with the typical value observed in normal star-forming galaxies(Schiminovich et al. 2010; Boselli et al. 2014).

6.2 How do HIGHz galaxies compare with turbulent gas-richdiscs at z ∼1?

As a representative sample of the high gas fraction population of z

∼1 discs, we consider PHIBSS, the IRAM Plateau de Bure HIgh-z Blue Sequence CO(3–2) Survey (Tacconi et al. 2013). PHIBSSmeasured the molecular gas content of 52 star-forming galaxies intwo redshift slices, at z ∼ 1.2 and 2.2; most of these systems arerotationally supported turbulent discs. Here, we restrict the PHIBSSsample to the 38 galaxies in the lower redshift interval (1.00 < z <

1.53), which all have CO detections.We also include in this comparison DYNAMO (DYnamics of

Newly-Assembled Massive Objects, Green et al. 2014), a surveyof local star-forming galaxies selected by H α emission from theSDSS to be potential analogues of the PHIBSS systems. Althoughfor DYNAMO there are no direct H I observations, and moleculargas measurements are available for only four objects (Fisher et al.2014), given their possible connection to high-redshift galaxies it is

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3538 B. Catinella and L. Cortese

very interesting to see how they relate to our HIGHz sample. As forHIghMass, we extract stellar masses and stellar surface densitiesfor DYNAMO from SDSS, consistently with our sample.

In order to compare the gas content of all samples, which haveH I (HIGHz, HIghMass) or H2 (PHIBSS) measurements, or totalgas masses simply inferred from SFRs (DYNAMO), we estimatetotal gas masses as follows. For GASS, HIGHz, and HIghMass weassume a molecular-to-atomic hydrogen mass ratio MH2 /MH I=0.3,which is the average value measured for a representative subset ofGASS galaxies by COLD GASS (Saintonge et al. 2011). Hence,Mgas = 1.768 MH I for these samples, taking into account the heliumcontribution (a factor 1.36). It should be kept in mind that these arerough estimates, as the relation between MH2 and MH I in this stellarmass regime has a large scatter (0.41 dex, see fig. 8 in Saintongeet al. 2011). For DYNAMO, we use their published total gas contentestimates, which were inferred from the SFR densities using theKennicutt–Schmidt (KS; Kennicutt 1998) law, and correct them forthe Helium contribution as above. We assume the same MH2 /MH I

ratio obtained from GASS to compute Mgas for the four galaxieswith available CO data. For PHIBSS, we assume that the hydrogenis all in molecular phase, i.e. Mgas = MH2 (Tacconi et al. 2013; theirmeasurements already include the He contribution).

The results of our comparisons are again presented in Fig. 8.Keeping in mind all the assumptions in our calculations, thetop-right panel shows that the total gas content of the bulk ofz ∼ 1.2 galaxies is comparable to that of both HIGHz and HIgh-Mass, although the partition between atomic and molecular phasesis obviously very different. It is beyond the reach of current andupcoming radio facilities to detect H I emission from individualPHIBSS galaxies, but only the presence of large H I reservoirs thatare unaccounted for (and considered unlikely by current simula-tions, e.g. Lagos et al. 2014; Popping, Somerville & Trager 2014)can make the bulk of PHIBSS galaxies significantly different fromHIGHz and HIghMass.

Conversely, DYNAMO galaxies have less overlap with the otherthree samples considered here. Indeed, DYNAMO does not includegalaxies with stellar masses above ∼1011 M�, hence the parameterspace in common with HIGHz and PHIBSS is somewhat limited.Because DYNAMO gas fractions are only estimates, we also plotthe four galaxies with actual CO measurements (dark green squares,connected by lines to the corresponding KS estimates), which agreepretty well with the KS predictions. Thus, taking all the predictedgas masses at face value, about half of DYNAMO systems seem tohave gas fractions as extreme as those of the other samples.

Despite their similar total gas fractions, PHIBSS and HIGHzsamples are remarkably different when we look at their SFRs andstellar mass surface densities in the bottom panels of Fig. 8. First, asexpected, the SFRs (and hence star formation efficiency) of PHIBSSgalaxies are significantly higher (∼1 dex) than all the other samplesconsidered here. Second, the average stellar mass surface densitiesof PHIBSS systems (as well as those of HIghMass and DYNAMO)are entirely consistent with those of disc-dominated, local galaxieswith similar stellar masses, and only a couple of objects exhibitvalues of μ� comparable to those of HIGHz. As mentioned in theprevious section, the fact that HIGHz galaxies occupy a region ofparameter space in the μ�–M� diagram that is practically untouchedby the other samples is the result of a selection effect. As we arelooking at two families of galaxies separated by ∼6 Gyr of evolution,the differences in optical and SFR properties between HIGHz andPHIBSS samples are not surprising, and it is interesting to speculatewhether or not we can confidently exclude an evolutionary linkbetween the two.

If PHIBSS galaxies are representative of the massive populationat z ∼ 1, it would be extremely hard to explain why typical massivegalaxies at z ∼ 1 evolve into such a rare population of gas-richsystems as HIGHz. Conversely, if gas-rich z ∼ 1 systems are thetip of the iceberg of the high-redshift population of disc galaxies,as hinted by recent numerical simulations (e.g. Lagos et al. 2014;Popping et al. 2014), the main obstacle to reconcile the two popu-lations is just the difference in μ�, which would require significantdisc growth from z ∼ 1.2 to ∼0.2. Observations and semi-analyticalmodels (Dutton et al. 2011) suggest that, between z ∼ 1 and 0, amassive disc galaxy increases its stellar mass by a factor of ∼2.5and its radius by a factor of ∼2. This implies a decrease in μ�

of ∼0.2 dex. Although this shift is too small to reconcile HIGHzand PHIBSS galaxies, it is intriguing to note that, once HIGHz andHIghMass are treated as a single population of ‘local’ H I massivesystems, the differences with the population of gas-rich turbulentdiscs at high-redshift almost entirely disappear.

Unfortunately, until a proper characterization of the gas prop-erties of a representative population of disc galaxies at z ∼ 1 isobtained, it is impossible to establish whether gas-rich galaxies atvarious redshifts are really linked, and this discussion remains justan intriguing speculation. Nevertheless, our findings highlight theimportance of investigating the still vastly unexplored regime ofvery high gas content for our understanding of galaxy evolutionacross cosmic time.

7 SU M M A RY A N D C O N C L U S I O N S

In this paper, we presented HIGHz, a survey that measured the H I

content of 39 disc galaxies at redshift z ∼ 0.2 using the Areciboradio telescope. This sample includes the highest redshift detectionsof H I emission from individual galaxies published to date, whichare also among the most H I-massive systems known. By selection,HIGHz galaxies are disc-dominated systems in relatively isolatedfields, with stellar masses M�= 2–22 × 1010 M�, H I masses MH I

= 2–8 × 1010 M�, SFRs of 3–35 M� yr−1 and redshifts z =0.17–0.25.

We showed that the HIGHz galaxies have unusually large H I

gas fractions and blue NUV−r colours for their stellar masses.However, when we look at more physical relations, such as the gasfraction plane and the baryonic TF relation, HIGHz galaxies areindistinguishable from the average systems. In other words, theirgas content is exactly what is expected from their UV and opticalproperties, and there is nothing unusual in the way in which starformation proceeds in these galaxies, or in the relation betweentheir dynamical and baryonic masses. We concluded that HIGHzgalaxies are rare, scaled-up versions of disc galaxies in the localUniverse, and there is no need to invoke unusual episodes of gasaccretion to explain their large reservoirs.

When compared to HIghMass, which includes the most H I-rich local galaxies extracted from the H I-blind ALFALFA survey,HIGHz systems show striking similarities in their gas content andstar formation. The only significant difference is that, by selection,HIGHz galaxies have ∼0.6 dex lower stellar surface densities, sug-gesting higher values of their spin parameters. Therefore, HIghMassgalaxies appear to be the local counterparts of HIGHz, with the lat-ter mapping into the high end of the spin parameter distribution ofHIghMass.

It is more difficult to establish a connection with the gas-rich,turbulent discs identified by PHIBSS at z ∼ 1. It is intriguingthat the total gas content of HIGHz and PHIBSS seems to besimilar (although the phase is clearly different) but, unless the

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3539

PHIBSS discs are not representative of the general disc population atz ∼ 1, it is very unlikely that they could be the progenitors of sucha rare population at z ∼ 0.2.

In addition to probing the H I Universe beyond z = 0.2 for the firsttime with direct detections, the HIGHz survey provides importantinsights into the properties of the massive, H I-rich systems thatwill likely dominate the samples detected at higher redshift byfuture H I surveys with the SKA and its precursor telescopes. Thisis particularly relevant for deep H I surveys such as DINGO (DeepInvestigation of Neutral Gas Origins, Meyer 2009) and especiallyLADUMA (Looking At the Distant Universe with the MeerKATArray, Holwerda, Blyth & Baker 2012), which will open the higherredshift Universe to H I exploration.

AC K N OW L E D G E M E N T S

We wish to thank Shan Huang for kindly providing us with HIgh-Mass data in advance of publication. BC warmly thanks MarthaP. Haynes and Riccardo Giovanelli for their contributions to theinitial stages of the survey, and Phil Perillat, Ganesh Rajagopalanand the telescope operators at Arecibo for their help and assis-tance. We thank our referee, Barbel Koribalski, for useful com-ments that helped us improving the clarity of our manuscript.BC is the recipient of an Australian Research Council FutureFellowship (FT120100660). LC acknowledges support under theAustralian Research Council’s Discovery Projects funding scheme(DP130100664).

This research has made use of the NASA/IPAC ExtragalacticDatabase (NED) which is operated by the Jet Propulsion Laboratory,California Institute of Technology, under contract with the NationalAeronautics and Space Administration.

The Arecibo Observatory is operated by SRI International un-der a cooperative agreement with the National Science Foundation(AST-1100968), and in alliance with Ana G. Mendez-UniversidadMetropolitana, and the Universities Space Research Association.

GALEX (Galaxy Evolution Explorer) is a NASA Small Explorer,launched in 2003 April. We gratefully acknowledge NASA’s supportfor construction, operation, and science analysis for the GALEX mis-sion, developed in cooperation with the Centre National d’EtudesSpatiales (CNES) of France and the Korean Ministry of Scienceand Technology.

Funding for the SDSS and SDSS-II has been provided bythe Alfred P. Sloan Foundation, the Participating Institutions, theNational Science Foundation, the US Department of Energy, theNational Aeronautics and Space Administration, the JapaneseMonbukagakusho, the Max Planck Society, and the Higher Ed-ucation Funding Council for England. The SDSS Web Site ishttp://www.sdss.org/.

The SDSS is managed by the Astrophysical ResearchConsortium for the Participating Institutions. The Participating In-stitutions are the American Museum of Natural History, Astrophys-ical Institute Potsdam, University of Basel, University of Cam-bridge, Case Western Reserve University, University of Chicago,Drexel University, Fermilab, the Institute for Advanced Study, theJapan Participation Group, Johns Hopkins University, the Joint In-stitute for Nuclear Astrophysics, the Kavli Institute for ParticleAstrophysics and Cosmology, the Korean Scientist Group, theChinese Academy of Sciences (LAMOST), Los Alamos NationalLaboratory, the Max-Planck-Institute for Astronomy (MPIA), theMax-Planck-Institute for Astrophysics (MPA), New Mexico StateUniversity, Ohio State University, University of Pittsburgh, Univer-

sity of Portsmouth, Princeton University, the United States NavalObservatory, and the University of Washington.

R E F E R E N C E S

Abazajian K. et al., 2003, AJ, 126, 2081Abazajian K. et al., 2004, AJ, 128, 502Abazajian K. et al., 2005, AJ, 129, 1755Abazajian K. N. et al., 2009, ApJS, 182, 543Adelman-McCarthy J. K. et al., 2007, ApJS, 172, 634Barnes D. G. et al., 2001, MNRAS, 322, 486Bell E. F. et al., 2005, ApJ, 625, 23Boissier S., Prantzos N., 2000, MNRAS, 312, 398Booth R. S., de Blok W. J. G., Jonas J. L., Fanaroff B., 2009, preprint

(arXiv:0910.2935)Boselli A., Cortese L., Boquien M., Boissier S., Catinella B., Lagos C.,

Saintonge A., 2014, A&A, 564, A66Bothun G. D., Impey C. D., Malin D. F., Mould J. R., 1987, AJ, 94, 23Brinchmann J., Charlot S., White S. D. M., Tremonti C., Kauffmann G.,

Heckman T., Brinkmann J., 2004, MNRAS, 351, 1151Carilli C. L., Rawlings S., 2004, New. Astron. Rev., 48, 979Catinella B., Haynes M. P., Giovanelli R., 2007, AJ, 134, 334Catinella B., Haynes M. P., Giovanelli R., Gardner J. P., Connolly A. J.,

2008, ApJ, 685, L13Catinella B. et al., 2010, MNRAS, 403, 683Catinella B. et al., 2012a, MNRAS, 420, 1959Catinella B. et al., 2012b, A&A, 544, A65Catinella B. et al., 2013, MNRAS, 436, 34Chabrier G., 2003, PASP, 115, 763Chilingarian I. V., Melchior A.-L., Zolotukhin I. Y., 2010, MNRAS, 405,

1409Cluver M. E., Jarrett T. H., Kraan-Korteweg R. C., Koribalski B. S., Appleton

P. N., Melbourne J., Emonts B., Woudt P. A., 2010, ApJ, 725, 1550Condon J. J., Cotton W. D., Greisen E. W., Yin Q. F., Perley R. A., Taylor

G. B., Broderick J. J., 1998, AJ, 115, 1693Cortese L., 2012, A&A, 543, A132Cortese L., Catinella B., Boissier S., Boselli A., Heinis S., 2011, MNRAS,

415, 1797Daddi E. et al., 2010, ApJ, 713, 686Dutton A. A. et al., 2011, MNRAS, 410, 1660Fabello S., Catinella B., Giovanelli R., Kauffmann G., Haynes M. P., Heck-

man T. M., Schiminovich D., 2011, MNRAS, 411, 993Fernandez X. et al., 2013, ApJ, 770, L29Fisher D. B. et al., 2014, ApJ, 790, L30Fridman P. A., Baan W. A., 2001, A&A, 378, 327Genzel R. et al., 2011, ApJ, 733, 101Giovanelli R. et al., 2005, AJ, 130, 2598Green A. W. et al., 2014, MNRAS, 437, 1070Haynes M. P. et al., 2011, AJ, 142, 170Holwerda B. W., Blyth S.-L., Baker A. J., 2012, in Tuffs R. J., Popescu

C. C., eds, Proc. IAU Symp. 345, The Spectral Energy Distribution ofGalaxies. Cambridge Univ. Press, Cambridge, p. 496

Huang S., Haynes M. P., Giovanelli R., Brinchmann J., 2012, ApJ, 756, 113Huang S. et al., 2014, ApJ, 793, 40Impey C., Bothun G., 1989, ApJ, 341, 89Jaffe Y. L., Poggianti B. M., Verheijen M. A. W., Deshev B. Z., van Gorkom

J. H., 2012, ApJ, 756, L28Johnston S. et al., 2008, Exp. Astron., 22, 151Kennicutt R. C., Jr, 1998, ApJ, 498, 541Lagos C. D. P., Baugh C. M., Lacey C. G., Benson A. J., Kim H.-S., Power

C., 2011, MNRAS, 418, 1649Lagos C. D. P., Baugh C. M., Zwaan M. A., Lacey C. G., Gonzalez-Perez

V., Power C., Swinbank A. M., van Kampen E., 2014, MNRAS, 440,920

Lah P. et al., 2009, MNRAS, 399, 1447Lee C., Chung A., Yun M. S., Cybulski R., Narayanan G., Erickson N.,

2014, MNRAS, 441, 1363

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3540 B. Catinella and L. Cortese

Lelli F., Fraternali F., Sancisi R., 2010, A&A, 516, A11Lilly S. J., Le Fevre O., Hammer F., Crampton D., 1996, ApJ, 460, L1+McGaugh S. S., Schombert J. M., Bothun G. D., de Blok W. J. G., 2000,

ApJ, 533, L99Madau P., Ferguson H. C., Dickinson M. E., Giavalisco M., Steidel C. C.,

Fruchter A., 1996, MNRAS, 283, 1388Martin D. C. et al., 2005, ApJ, 619, L1Martin A. M., Papastergis E., Giovanelli R., Haynes M. P., Springob C. M.,

Stierwalt S., 2010, ApJ, 723, 1359Meyer M., 2009, in Heald G., Serra P., eds, Proc. Panoramic Radio Astron-

omy: Wide-field 1-2 GHz Research on Galaxy Evolution. Available at:http://pos.sissa.it/cgi-bin/reader/conf.cgi?confid=89

Mo H. J., Mao S., White S. D. M., 1998, MNRAS, 295, 319Obreschkow D., Rawlings S., 2009, ApJ, 696, L129Papastergis E., Giovanelli R., Haynes M. P., Rodrıguez-Puebla A., Jones

M. G., 2013, ApJ, 776, 43Pickering T. E., Impey C. D., van Gorkom J. H., Bothun G. D., 1997, AJ,

114, 1858Popping G., Somerville R. S., Trager S. C., 2014, MNRAS, 442, 2398Saintonge A. et al., 2011, MNRAS, 415, 32Salim S. et al., 2007, ApJS, 173, 267Schiminovich D. et al., 2010, MNRAS, 408, 919Seibert M. et al., 2012, in American Astronomical Society Meeting Ab-

stracts, Vol. 219, The Galaxy Evolution Explorer (GALEX) Source Cat-alogs. Am. Astron. Soc., Washington, DC, #340.01

Springob C. M., Haynes M. P., Giovanelli R., Kent B. R., 2005, ApJS, 160,149

Tacconi L. J. et al., 2013, ApJ, 768, 74van Zee L., Maddalena R. J., Haynes M. P., Hogg D. E., Roberts M. S.,

1997, AJ, 113, 1638Verheijen M., van Gorkom J. H., Szomoru A., Dwarakanath K. S., Poggianti

B. M., Schiminovich D., 2007, ApJ, 668, L9

Wang J. et al., 2013, MNRAS, 433, 270Wyder T. K. et al., 2007, ApJS, 173, 293York D. G. et al., 2000, AJ, 120, 1579Zibetti S., Charlot S., Rix H.-W., 2009, MNRAS, 400, 1181

A P P E N D I X A : SD S S IM AG E S A N D A R E C I B OH I SPECTRA

We present here SDSS postage stamp images and Arecibo H I-linespectra for the 39 HIGHz galaxies discussed in this work. Theseare organized as follows: Fig. A1 shows the 28 H I detections withquality flag Q = 1 in Table 2, and Figs A2 and A3 show five galax-ies with Q = 2 and 6 marginal detections with Q = 3, respectively.The objects in each of these figures are ordered by increasing AGCnumber (indicated on the top-right corner of the SDSS image). TheSDSS images show a 1 arcmin square field, i.e. only the centralpart of the region sampled by the Arecibo beam (the half powerfull width of the beam is ∼4 arcmin at the frequencies of theseobservations). The H I spectra are always displayed over a velocityinterval corresponding to the full 12.5 MHz bandwidth adopted forour observations. The H I-line profiles are calibrated, smoothed (to avelocity resolution between 27 and 41 km s−1, as listed in Table 2),and baseline subtracted. A red, dotted line indicates the heliocen-tric velocity corresponding to the optical redshift from SDSS. Theshaded area and two vertical dashes show the part of the profile thatwas integrated to measure the H I flux and the peaks used for widthmeasurement, respectively.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3541

Figure A1. SDSS postage stamp images (1 arcmin square) and H I-line profiles of the best quality detections (code 1) presented in this work, ordered byincreasing AGC number (indicated on each image). The H I spectra are calibrated, smoothed and baseline-subtracted. A dotted line and two dashes indicatethe heliocentric velocity corresponding to the SDSS redshift and the two peaks used for width measurement, respectively. The redshift measured from the H I

profile is indicated on the top right corner of the spectrum.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3542 B. Catinella and L. Cortese

Figure A1 – continued

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

HIGHz: H I-massive galaxies at z ∼ 0.2 3543

Figure A1 – continued

Figure A2. Same as Fig. A1 for lower signal-to-noise detections (code 2).

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

3544 B. Catinella and L. Cortese

Figure A3. Same as Fig. A1 for marginal detections (code 3).

This paper has been typeset from a TEX/LATEX file prepared by the author.

MNRAS 446, 3526–3544 (2015)

at Swinburne U

niversity of Technology on M

ay 11, 2016http://m

nras.oxfordjournals.org/D

ownloaded from