The outer halo globular cluster system of M31 - I. The final PAndAS catalogue
A. P. Huxor, A. D. Mackey, A. M. N. Ferguson, M. J. Irwin, N. F. Martin, N. R. Tanvir, J. Veljanoski, A. McConnachie, C. K. Fishlock, R. Ibata, G. F. Lewis
aa r X i v : . [ a s t r o - ph . GA ] M a y Mon. Not. R. Astron. Soc. , 1–26 (2014) Printed 21 September 2018 (MN L A TEX style file v2.2)
The outer halo globular cluster system of M31 – I. Thefinal PAndAS catalogue
A. P. Huxor , A. D. Mackey , A. M. N. Ferguson , M. J. Irwin , N. F. Martin ,N. R. Tanvir , J. Veljanoski , A. McConnachie , C. K. Fishlock , R. Ibata ,G. F. Lewis Astronomisches Rechen-Institut, Zentrum f¨ur Astronomie der Universit¨at Heidelberg, M¨onchhofstraße 12 - 14,69120 Heidelberg, Germany. Research School of Astronomy & Astrophysics, Australian National University, Mt. Stromlo Observatory, Cotter Road,Weston Creek, ACT 2611, Australia SUPA, Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK Institute of Astronomy, Madingley Road, Cambridge, CB3 0HA Observatoire de Strasbourg, 11, rue de l’Universit´e, F-67000, Strasbourg, France Department of Physics and Astronomy, University of Leicester, University Road, Leicester LE1 7RH NRC Herzberg Institute of Astrophysics, 5071 West Saanich Road, Victoria, British Columbia V9E 2E7, Canada Institute of Astronomy, School of Physics, A29, University of Sydney, NSW 2006, Australia
ABSTRACT
We report the discovery of 59 globular clusters (GCs) and two candidate GCs in asearch of the halo of M31, primarily via visual inspection of CHFT/MegaCam imageryfrom the Pan-Andromeda Archaeological Survey (PAndAS). The superior quality ofthese data also allow us to check the classification of remote objects in the RevisedBologna Catalogue (RBC), plus a subset of GC candidates drawn from SDSS imaging.We identify three additional new GCs from the RBC, and confirm the GC nature of11 SDSS objects (8 of which appear independently in our remote halo catalogue); theremaining 188 candidates across both lists are either foreground stars or backgroundgalaxies. Our new catalogue represents the first uniform census of GCs across the M31halo – we find clusters to the limit of the PAndAS survey area at projected radii of upto R proj ∼
150 kpc. Tests using artificial clusters reveal that detection incompletenesscuts in at luminosities below M V = − .
0; our 50% completeness limit is M V ≈ − . R proj = 25 kpc, and any new GCs within this radius. With these data weupdate results from Huxor et al. (2011), investigating the luminosity function (LF),colours and effective radii of M31 GCs with a particular focus on the remote halo. Wefind that the GCLF is clearly bimodal in the outer halo ( R proj >
30 kpc), with thesecondary peak at M V ∼ − .
5. We argue that the GCs in this peak have most likelybeen accreted along with their host dwarf galaxies. Notwithstanding, we also find, asin previous surveys, a substantial number of GCs with above-average luminosity inthe outer M31 halo – a population with no clear counterpart in the Milky Way.
Key words: galaxies: individual (M31) – galaxies: halos – galaxies: star clusters –galaxies: evolution
Globular cluster (GC) systems are thought to trace bothmajor star-formation episodes and accretion events. Assuch they have proven to be valuable tools for thestudy of their host galaxies (Georgiev, Goudfrooij, & Puzia2012) – from the seminal Milky Way (MW) work ofSearle & Zinn (1978) to recent studies of more distant galaxies (Forte, Vega, & Faifer 2012; Forbes et al. 2011;Chies-Santos et al. 2011).The GC system of M31 has naturally been thefocus of particular interest, providing (as a massive spi-ral galaxy) an excellent comparison to our own Milky c (cid:13) Huxor et al.
Way. Moreover, the proximity of M31 (at ∼
780 kpc) allows for detailed investigation of its GC populations,which have been extensively studied (e.g. Crampton et al.1985; Battistini et al. 1987; Elson & Walterbos 1988;Huchra, Brodie, & Kent 1991; Barmby et al. 2000;Perrett et al. 2002; Fan et al. 2008; Galleti et al. 2009;Caldwell et al. 2009; Fan, de Grijs, & Zhou 2010;Caldwell et al. 2011). Most of these studies have dealtwith the regions comparatively close to the centre of M31,typically within 20 −
25 kpc in projection. This is becausethe relative proximity of M31 also poses a problem in thatthe full extent of its stellar halo subtends a substantialangle on the sky ( ∼ > ◦ in diameter) which is difficult tosearch uniformly for GCs, especially those with low lumi-nosities and/or surface brightnesses. The Pan-AndromedaArchaeological Survey (PAndAS; McConnachie et al. 2009)almost completely obviates these issues: its imaging spansa very wide area, typically reaching a projected distance R proj ∼
150 kpc from M31 – and is yet sufficiently deep toallow the identification of even faint GCs.With high quality wide-field imaging such as that ob-tained for PAndAS, M31 halo GCs are much more easilylocated than those in more central regions where the back-ground and crowding due to the M31 disk hinders reliableidentification of star clusters in ground-based data. HaloGCs also offer the opportunity to study regions with verylong dynamical time-scales that can preserve evidence ofpast events. If formed in-situ , remote halo GCs will havebeen much less affected by tidal forces than those towardsthe centre; if accreted along with dwarf satellite galaxies,their properties may reflect the nature of the original hosts.This paper continues and extends earlier investigationsof the GC population of M31 by our group. In particu-lar, it provides the final catalogue of halo GCs from PAn-dAS, greatly extending our previous surveys and results –specifically those of Huxor et al. (2008) (hereafter, Hux08)and Huxor et al. (2011) (hereafter, Hux11). In Hux08 wepresented 40 new GCs from a precursor survey to PAn-dAS conducted using the Wide-Field Camera (WFC) on theIsaac Newton Telescope (INT) along with some early imag-ing from MegaCam on the Canada-France-Hawaii Telescope(CFHT), and updated the classifications of many entries inthe Revised Bologna Catalog (RBC) – the most completecatalogue of M31 GCs, and widely used by the community .Hux11 explored the “ensemble” properties of the updatedM31 GC sample from Hux08. In the present paper we ex-ploit the full, final PAndAS data, searching for new GCs,investigating candidate GCs from the RBC, and updatingmany of the results from Hux11 with a particular focus onthe properties of the GCs in the halo.In addition to M31, the PAndAS data (and its precedingINT/WFC survey) also extend to M33, and our work onthe GCs in this galaxy is published elsewhere (Huxor et al.2009; Cockcroft et al. 2011). We have also used PAndAS Throughout this paper we use the distance to M31 fromMcConnachie et al. (2005); see also Conn et al. (2012). Although this is still some distance short of the likely virialradius of M31. Note that at that time we worked with Version 3.0 of the RBC;for the present work we refer to Version 5 from August 2012 imaging to discover new GCs in the M31 dwarf elliptical(dE) satellites NGC 147 and NGC 185 (three GCs and oneGC respectively), as described in Veljanoski et al. (2013b).Although, strictly speaking, these clusters reside within thehalo of M31, we do not include them in the present paperas they possess clearly identified (and intact) host galaxies.The GCs listed in our previous catalogue (Hux08)provided targets for follow-up observations and analy-sis, both by our own group and by others. In partic-ular, our
Hubble Space Telescope ( HST ) observations ofmany of the halo GCs led to a number of studies oftheir colour-magnitude diagrams (CMDs) and structuralproperties (Mackey et al. 2007; Perina et al. 2009, 2011;Tanvir et al. 2012; Federici et al. 2012; Perina et al. 2012;Wang & Ma 2012). Many of those GCs were also observedspectroscopically with ground-based facilities – for example,Alves-Brito et al. (2009) observed several at high resolutionwith the Keck Telescope. Similarly, Ma (2012) used opticaland 2MASS photometry of many of our GCs to estimatetheir ages, masses and metallicities.The present paper is the first of a series of worksin which we use our catalogue to shed new light on theouter regions of the M31 halo. In an accompanying paper(Veljanoski et al. 2014) we investigate the kinematics of theremote GC system, while in two forthcoming works we willexplore the relationship between the GCs and the underlyingfield halo, and the resolved properties of the GCs through
HST imaging (Mackey et al. in prep).This paper proceeds as follows: in § §
3. In addition to discovering new GCs, we alsoused the same imaging data to clean previous samples ofpublished M31 GCs and GC candidates, and the results ofthis undertaking are given in §
4. The photometry of our newclusters, and all other GCs with a galactocentric distance ofgreater than 25 kpc, is described and tabulated in §
5. Nextwe assess the completeness of our sample, critical to properexploitation of the catalogue, in §
6. Finally, in §
7, we analysethe ensemble photometric properties of the M31 outer haloGC system, using our enlarged and improved catalogue.
The images and photometric catalogue employed in thisstudy were taken from the now-completed PAndAS surveyof M31, conducted using the CFHT on Mauna Kea, Hawaii.Details of this survey and its precursors can be found in anumber of previous works (e.g., McConnachie et al. 2009;Martin et al. 2006; Ibata et al. 2007; McConnachie et al.2008; Ibata et al. 2014), but we briefly summarise the keypoints here. The PAndAS imaging was undertaken withthe MegaPrime/MegaCam camera mounted on the CFHT,which comprises 36 CCDs (each 2048 × × .Three dithered 450s sub-exposures in each of the MegaCam g and i filters typically reach g ≈ i ≈ σ detection limit) once reduced andcombined. c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Crucial to our identification of GCs is the excellent PAn-dAS image quality. Many early exposures with relativelypoor seeing were re-observed towards the end of the surveyprogram, resulting in a mean seeing of 0 . ′′
67 in the g -bandand 0 . ′′
60 in the i -band, with an rms scatter between framesof 0 . ′′
10 in both cases.After initial reduction of the data at the CFHT, fur-ther image processing, calibration, and photometric mea-surements were conducted at the Cambridge Astronomi-cal Survey Unit (CASU). The CASU pipeline created finalstacked g - and i -band images at each pointing, and a mergedcatalogue providing photometric data and star/galaxy clas-sification for all detected sources, both stellar and non-stellar. The complete contiguous survey footprint, compris-ing 406 individual pointings, reaches to a projected distanceof ∼
150 kpc from M31 in most directions, thus encompass-ing almost the entire halo. This region is joined to a smallerarea around M33, extending to ∼
50 kpc from the centre ofthat galaxy.
We adopted a multi-strand search strategy based on ourexperience from Hux08, in which we found both classical“compact” M31 GCs, and also the more diffuse “extended”clusters. Our methodology is summarised in Figure 1.GC candidates were selected from the PAndAS pho-tometric catalogue based on their magnitude and colour.Known GCs (both compact and extended) inhabit a broadrange of absolute magnitudes and colours ( − . < M V < − .
5, and 0 . < ( V − I ) < .
7) – limits which we convertedto apparent MegaCam g and i -magnitudes by using the in-verse of the transformation equations (1 to 4) described inSection 5, below, and assuming an M31 distance modulus of24 .
47 and a typical foreground extinction E ( B − V ) ∼ . ∼ . ′′ − . ′′
7. Diffuse clus-ters, however, tend not to appear in the catalogue as a singlesource and can therefore easily be missed with this approach– we adopted additional search techniques for these objects(see below). Note that the CASU pipeline is not optimisedfor non-stellar source photometry. Hence, although the mag-nitudes and colours are sufficiently accurate to identify likelycompact GC candidates (especially given our very generousranges for both), we subsequently undertook our own be-spoke photometry of each GC we discovered (see section 5).We visually inspected a g -band image of every candi-date object and its local surrounding area, using a FITSimage viewer to overlay (and so highlight) the positions ofthe GC candidates with graphic markers. This ensured thatadequate attention was drawn to both the less luminous andthe more compact candidates. At the distance of M31, and Our previous experience revealed that the g -band is both moreeffective and more efficient for identifying GCs than the i -band.This is largely due to the greatly reduced prominence of the maincontaminants – background elliptical galaxies and foregrounddwarf stars – in the blue. with the high quality of the MegaCam images, GCs gener-ally take the form of a core that is slightly broader than thestellar point-spread function (PSF), surrounded by resolvedred giant branch (RGB) stars. This results in an easily dis-tinguished local “halo” of such objects in well populated-clusters, and/or a broadened core with an irregular appear-ance for less luminous examples. In almost all cases we foundit straightforward to unambiguously classify a GC candidateas a cluster or not. However, the search efficiency was low –in the vast majority of instances the candidates turned outto be distant background galaxies.Extended diffuse clusters (Huxor et al. 2005) are prob-lematic because they are typically semi- or completely re-solved across their full spatial extent in the MegaCam imag-ing and thus are not flagged in the PAndAS photometriccatalogue by the presence of a single unresolved source. Inmost cases, however, such objects are also not sufficientlywell populated or sufficiently uncrowded to appear as co-located groups of similar stars that could be detected bymeans of automated algorithms. Our previous experience(Hux08) showed that the most efficient and least biasedway to detect such objects is by simple visual inspectionof the full survey area. Although labour intensive, this in-spection, conducted by APH, led to the discovery of manyclusters ( ∼ Our search covered the PAndAS footprint to its largest ra-dial extent, which ranges from ∼ ◦ ( ∼
27 kpc), and an inclination of 77.5 ◦ .Within this region variable crowding makes it difficult toconduct a uniform search for GCs, particularly affecting thediscovery of low-luminosity compact GCs, and all extendedGCs. Taking the above into account, we expect the com-pleteness limits derived in Section 6 to be applicable outsidea projected galactocentric radius of ∼
25 kpc (although wedid locate a handful of GCs within this radius). c (cid:13) , 1–26 Huxor et al.
Figure 1.
A schematic summary of our multi-strand GC search strategy.
Following the procedure described in Section 2, we discov-ered 58 previously unknown GCs and two additional GCcandidates . All but one of these came from the indepen-dent visual inspection of (i) candidates, and (ii) the fullsurvey area. The exception, PAndAS-31, was discovered viathe automated search for M31 dwarf spheroidal galaxies (seeMartin et al. 2013). Note that, as described in that paper,the automated search also uncovered a small subset of theobjects discovered independently in our search by eye.The identity and location of each of these new ob-jects are listed in Table 1, and g-band thumbnail imagesare shown in Figure 2. The thumbnail images clearly revealthe unambiguous classification of each catalogued object asa GC, highlighting the quality of the data and the reasonwhy our search turned up so few candidates with indetermi-nate classification. Of the two such candidates in our sample,Cand-01 is a faint object set against a relatively dense stellarbackground that hinders discrimination between its identityas a cluster or a distant galaxy, while Cand-02 is an extendedobject (cluster or distant galaxy) largely cut-off at the edgeof an image. These are listed in Table 1 but not included inour subsequent analysis. We aim to obtain follow-up obser-vations of these objects to clarify their status.Note that with the 40 GCs presented in Hux08, we havefound a total of ∼
100 new GCs in the outer halo of M31.For completeness we note that a number of the clus-ters listed in Table 1 have formed the basis of previ-ous studies – specifically those by Mackey et al. (2010a)and Veljanoski et al. (2013a) who investigated the ensem-ble properties of M31 halo GCs including subsamples fromthe present list, and the works by Mackey et al. (2013a,b) We note that we subsequently also discovered one additionalGC (PA-59), as detailed in Section 4.1. who studied specific objects (PA-48, and PA-7 and PA-8,respectively).Our new catalogue represents the first detailed censusof GCs across the full M31 halo, greatly extending the workof Hux08. The vast majority of our discoveries (53 of 59)lie in the outskirts of M31, at projected radii R proj > R proj = 80 kpc, ofwhich 11 sit outside R proj = 100 kpc. Indeed we effectivelyfind GCs out to distances commensurate with the edge ofthe PAndAS footprint, confirming previous suggestions thatthe M31 cluster system is very extended (e.g., Mackey et al.2010b) and suggesting that additional GCs may be foundat even larger radii (see also di Tullio Zinn & Zinn 2013).Combined with previous discoveries, mostly from Hux08,we now know of 91 M31 GCs lying outside R proj >
25 kpc,which includes 12 at distances larger than R proj = 100 kpc.These observations stand in stark contrast to the haloGC population in the Milky Way, in which there are only ≈
13 objects known at Galactocentric radii larger than 30 −
35 kpc (corresponding to an average projected radius of ∼
25 kpc for random viewing angles), and in which the mostdistant known member sits at a Galactocentric distance of ≈
120 kpc (corresponding to an average projected distanceof ∼
95 kpc for random viewing angles). While the disparityin the number of GCs in the Milky Way and M31 within R proj ≈
25 kpc is roughly 3:1 in favour of M31, our newcatalogue reveals that outside this radius it is more like 7:1in favour of M31. We explore the differences between thesetwo GC systems in more detail in Section 7.The photometric properties (luminosities, colours andsizes) of our new GC sample are derived below in Section5. However, the excellent quality of PAndAS imaging alsoallowed us to examine and resolve the identity of many can-didate clusters previously identified in the literature, and wefirst turn our attention to these. c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Figure 2.
MegaCam g -band thumbnail images of our new M31 halo GCs. Each image is 1 ′ × ′ in size, with north to the top and eastto the left. PA-59 is shown in Figure 4.c (cid:13) , 1–26 Huxor et al.
Table 1.
Locations of newly-discovered PAndAS GCs.ID Position (J2000.0) R proj RA Dec (kpc)PAndAS-01 23 57 12.03 +43 33 08.28 118.92PAndAS-02 23 57 55.69 +41 46 49.25 114.74PAndAS-03 00 03 56.41 +40 53 19.20 100.00PAndAS-04 00 04 42.93 +47 21 42.47 124.62PAndAS-05 00 05 24.15 +43 55 35.70 100.60PAndAS-06 00 06 11.95 +41 41 20.97 93.66PAndAS-07 00 10 51.35 +39 35 58.55 85.95PAndAS-08 00 12 52.45 +38 17 47.86 88.26PAndAS-09 00 12 54.66 +45 05 55.86 90.82PAndAS-10 00 13 38.66 +45 11 11.13 90.00PAndAS-11 00 14 55.63 +44 37 16.35 83.23PAndAS-12 00 17 40.08 +43 18 39.02 69.21PAndAS-13 00 17 42.72 +43 04 31.83 67.98PAndAS-14 00 20 33.88 +36 39 34.46 86.20PAndAS-15 00 22 44.07 +41 56 14.16 51.90PAndAS-16 00 24 59.92 +39 42 13.11 50.81PAndAS-17 00 26 52.20 +38 44 58.11 53.93PAndAS-18 00 28 23.26 +39 55 04.86 41.55PAndAS-19 00 30 12.22 +39 50 59.27 37.87PAndAS-20 00 31 23.74 +41 59 20.12 30.59PAndAS-21 00 31 27.52 +39 32 21.84 37.68PAndAS-22 00 32 08.36 +40 37 31.62 28.73PAndAS-23 00 33 14.13 +39 35 15.93 33.74PAndAS-24 00 33 50.57 +38 38 28.04 42.81PAndAS-25 00 34 06.15 +43 15 06.65 34.79PAndAS-26 00 34 45.08 +38 26 38.05 43.92PAndAS-27 00 35 13.53 +45 10 37.85 56.58PAndAS-28 00 35 56.43 +40 48 44.98 18.60PAndAS-29 00 36 09.08 +40 08 09.85 23.04PAndAS-30 00 38 29.01 +37 58 39.21 46.35PAndAS-31 00 39 59.79 +43 03 19.67 25.38PAndAS-32 00 40 41.20 +40 00 54.95 17.94PAndAS-33 00 40 57.35 +38 38 10.24 36.28PAndAS-34 00 41 18.04 +42 46 16.51 20.85PAndAS-35 00 43 09.36 +40 36 38.23 9.07PAndAS-36 00 44 45.57 +43 26 34.79 30.14PAndAS-37 00 48 26.53 +37 55 42.14 48.06PAndAS-38 00 49 45.67 +47 54 33.12 92.33PAndAS-39 00 50 36.22 +42 31 49.29 26.40PAndAS-40 00 50 43.80 +40 03 30.20 26.51PAndAS-41 00 53 39.58 +42 35 14.98 33.09PAndAS-42 00 56 38.04 +39 40 25.93 42.18PAndAS-43 00 56 38.80 +42 27 17.77 38.92PAndAS-44 00 57 55.89 +41 42 57.01 39.35PAndAS-45 00 58 37.96 +41 57 11.48 41.66PAndAS-46 00 58 56.40 +42 27 38.29 44.31PAndAS-47 00 59 04.78 +42 22 35.06 44.26PAndAS-48 00 59 28.26 +31 29 10.64 141.34PAndAS-49 01 00 50.07 +42 18 13.25 48.21PAndAS-50 01 01 50.66 +48 18 19.22 106.68PAndAS-51 01 02 06.61 +42 48 06.64 53.42PAndAS-52 01 12 47.01 +42 25 24.87 78.05PAndAS-53 01 17 58.41 +39 14 53.20 95.88PAndAS-54 01 18 00.14 +39 16 59.93 95.79PAndAS-55 01 19 20.41 +46 03 11.52 111.50PAndAS-56 01 23 03.53 +41 55 11.02 103.34PAndAS-57 01 27 47.51 +40 40 47.20 116.41PAndAS-58 01 29 02.16 +40 47 08.66 119.42PAndAS-59 00 36 29.53 +40 38 16.83 18.28PAndAS-Cand-01 00 44 58.35 +40 21 37.92 13.70PAndAS-Cand-02 01 07 53.88 +48 22 41.79 114.60
There are two primary sources of candidate M31 halo clus-ters – the Revised Bologna Catalogue (RBC; Galleti et al.2004), and a recent search for M31 GCs in the Sloan DigitalSky Survey (SDSS) by di Tullio Zinn & Zinn (2013).
The RBC is the main repository for information on the M31GC system. It contains lists of confirmed GCs (classes 1and 8 in the catalogue) and candidate GCs (classes 2 and3), as well as a few H ii regions (class 5), and compilations ofobjects once suspected to be GCs but subsequently revealedas background galaxies (class 4) or foreground stars (class 6).The identities of these contaminants are retained in the RBCin order to avoid mis-classification in future GC surveys.We inspected 1 . ′ × . ′ PAndAS thumbnails of all ob-jects listed in Version 5 of the RBC (released in August2012) as having projected galactocentric radii larger than R proj ∼
15 kpc. Inside this radius the strong and variablebackground due to the M31 disk means that even with ourhigh quality PAndAS imaging it is frequently impossible toestablish a reliable target classification . Overall there are523 objects with R proj >
15 kpc in the RBC V5, of which497 have PAndAS imaging. The missing 26 entries typicallyfall into small gaps in the coverage resulting from the inter-row CCD spaces on the MegaCam array (the spaces betweenindividual CCDs on a given row were covered by the PAn-dAS dither pattern) or imperfect tiling of the PAndAS mo-saic, although a couple sit outside the survey footprint with R proj ∼ >
150 kpc.To avoid, as far as possible, prior knowledge introducingbias into our classifications, we employed a blind inspectionmethodology. One of us (ADM) generated thumbnails for alltargets, randomised the order, and supplied the images only,with no supplementary information, to APH for classifica-tion. Once the inspection process was complete, the resultswere returned to ADM for analysis.The original RBC classifications for the 497 objects weinspected broke down as follows: 72 GCs, 141 GC candi-dates, 166 background galaxies, 116 foreground stars, and 2H ii regions. We confirmed that all 282 of the contaminantobjects (galaxies and stars) were correctly identified as such.Of more interest are the GCs and GC candidates, and wewere able to greatly improve classifications for these targets.We found that two of the candidates were in fact gen-uine GCs. These objects are listed in Table 2 and their g -band thumbnails displayed in Figure 3: SK213C andSK255B. Both are located within the 25 kpc inner limitof our main survey, which was why they were not identifiedas part of that search. We also uncovered one particularlyinteresting GC candidate – SH06, which consists of a com-pact luminous source surrounded by nebulosity that is quiteevident in the g -band imaging (see Figure 3) but virtuallyinvisible in the i -band. This is suggestive of a massive youngstar cluster still embedded in gas, sitting at ∼
15 kpc from Note that this radius is smaller than the inner radius of ouruniform survey ( R proj = 25 kpc) as here we are not trying todiscover new GCs, but rather establish classifications for objectsfor which we already have positions.c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Table 2.
Updated globular cluster classifications in the Revised Bologna Catalogue V5.Name in Position (J2000.0) R proj Previous NewRBC V5 RA Dec (kpc) Classification a Classification a Promoted GCs + H ii PAndAS-59 00 36 29 .
35 +40 38 16 . . − .
05 +40 21 58 . .
19 2 5 (1?)B270D 00 45 49 .
22 +41 01 49 . .
57 2 1SK213C 00 46 58 .
77 +42 17 45 . .
71 2 1SK255B 00 49 03 .
02 +41 54 57 . .
39 2 1
Demoted GCs
SK002A 00 36 34 .
99 +41 01 08 . .
20 1 6SK004A 00 38 01 .
35 +42 04 06 . .
25 1 6BA11 00 48 45 .
59 +42 23 37 . .
71 1 4 + 6 a Classes: 1 , , ii region; 6 = star. b Classified as a candidate in previous versions of the RBC – see text.
Figure 3.
PAndAS g -band thumbnails for the three new globular clusters uncovered in the RBC V5, plus SH06 (see text). Each thumbnailis 1 ′ × ′ in size, with north to the top and east to the left. Figure 4.
PAndAS g -band thumbnail for our serendipitous dis-covery PAndAS-59 (circled). The object at the centre of the fieldis SK014B, correctly classified in the RBC as a star. The thumb-nail is 1 . ′ × . ′ , with north to the top and east to the left. the M31 centre. However, because we cannot be absolutelycertain that there is a cluster present, we conservatively clas-sify this object as a H ii region in Table 2.While inspecting the object SK014B, which is correctly classified in the RBC V5 as a star, we noticed a small clus-ter nearby which does not appear anywhere in the RBC. Wetherefore believe this to be a new discovery, which we namePAndAS-59. We include PA-59 in Table 2 and display thediscovery thumbnail in Figure 4. That we identified this ob-ject serendipitously in our sample of small thumbnail imagessuggests that a full search of the inner M31 halo, between ∼ −
25 kpc, may be quite fruitful – although we reiter-ate the caveat that the increased crowding would adverselyaffect the completeness of any such survey.To our sample of new RBC GCs we also add B270D.At R proj = 8 .
57 kpc, this object sits well inside both ourinner PAndAS search radius, and our inner RBC inspectionradius. In V5 of the RBC it is classified as a candidate object;we uncovered its cluster status by chance. It is listed in Table2 and displayed in Figure 3.In addition to confirming several new globular clustersamong RBC candidates, we also found three “confirmed”RBC GCs which were misclassified stars or galaxies. Theseobjects – SK002A, SK004A, and BA11 – are listed in Table2 and their thumbnails shown in Figure 5.All but one of the remaining 138 objects originally listedas GC candidates in the RBC V5 are either foreground stars(25 objects) or background galaxies (112 objects). Theseare listed in Table 3, and a few representative examples aredisplayed in Figure 6. For the last candidate, BH01, we couldnot find any object at the listed coordinates. This is likelybecause BH01 is a very faint object identified from
HST c (cid:13) , 1–26 Huxor et al.
Table 3.
Updated classifications for globular cluster candidates listed in the Revised Bologna Catalogue V5.Name in Position (J2000.0) R proj New Name in Position (J2000.0) R proj NewRBC V5 RA Dec (kpc) Class a RBC V5 RA Dec (kpc) Class a SK001C 00 33 13 .
080 +40 05 26 .
00 29 .
47 4 SK019B 00 37 33 .
470 +40 05 28 .
70 20 .
96 4SK002C 00 33 15 .
820 +40 00 24 .
70 30 .
03 4 SK020B 00 37 35 .
680 +40 35 14 .
40 16 .
22 4SK001B 00 33 23 .
070 +40 04 40 .
70 29 .
21 4 SK048C 00 37 37 .
200 +40 05 39 .
60 20 .
83 4SK002B 00 33 32 .
200 +39 51 32 .
80 30 .
70 4 SK049C 00 37 37 .
270 +41 54 04 .
40 15 .
67 4SK004C 00 33 34 .
960 +40 08 16 .
30 28 .
32 4 SK050C 00 37 41 .
270 +40 04 42 .
90 20 .
89 4SK003B 00 33 37 .
030 +39 40 59 .
00 32 .
14 4 SK051C 00 37 41 .
790 +40 05 18 .
00 20 .
77 4SK005C 00 33 38 .
040 +39 35 35 .
70 32 .
96 4 SK022B 00 37 54 .
310 +40 17 26 .
70 18 .
31 4SK006C 00 33 46 .
110 +39 48 36 .
60 30 .
67 4 SK053C 00 38 00 .
760 +42 02 56 .
90 16 .
10 4SK007C 00 33 54 .
630 +39 34 36 .
60 32 .
61 4 SK054C 00 38 06 .
100 +40 24 30 .
20 16 .
80 6B133D 00 34 10 .
994 +39 50 50 .
27 29 .
52 4 SK058C 00 38 48 .
400 +40 03 01 .
20 19 .
53 4BH01 b
00 34 11 .
480 +39 23 59 .
10 33 . − SK066C 00 39 15 .
190 +42 22 50 .
70 17 .
59 6SK009C 00 34 12 .
200 +40 06 29 .
70 27 .
22 4 B186D 00 40 02 .
258 +39 23 12 .
11 26 .
68 4SK010C 00 34 26 .
850 +39 54 05 .
60 28 .
51 4 SK073C 00 40 04 .
300 +40 14 10 .
70 15 .
72 4B411 00 34 30 .
808 +41 33 44 .
09 21 .
46 4 B188D 00 40 14 .
038 +39 41 30 .
82 22 .
52 4SK004B 00 34 34 .
200 +40 02 49 .
40 26 .
98 4 B191D 00 40 17 .
893 +42 25 23 .
98 16 .
95 4SK011C 00 34 51 .
160 +39 55 33 .
10 27 .
50 4 SK090C 00 40 53 .
060 +40 00 43 .
30 17 .
84 4B412 00 34 55 .
281 +41 32 26 .
49 20 .
38 4 B460 00 41 54 .
817 +39 35 25 .
51 23 .
05 4SK012C 00 35 08 .
810 +40 07 32 .
60 25 .
13 4 SK104C 00 42 03 .
040 +40 03 48 .
80 16 .
58 4SK013C 00 35 09 .
240 +40 05 39 .
80 25 .
38 4 SK110C 00 42 33 .
090 +40 04 53 .
60 16 .
24 4B413 00 35 13 .
001 +41 29 07 .
81 19 .
51 4 B225D 00 43 13 .
440 +40 01 14 .
58 17 .
11 6BA22 00 35 13 .
608 +39 45 37 .
16 28 .
40 4 B233D 00 43 41 .
311 +39 36 45 .
96 22 .
78 4SK014C 00 35 14 .
860 +39 41 40 .
00 29 .
03 4 SK136C 00 44 04 .
430 +40 05 19 .
60 16 .
50 6SK015C 00 35 20 .
470 +39 35 04 .
10 30 .
02 4 SK160C 00 44 54 .
430 +40 06 44 .
10 16 .
78 4SK016C 00 35 22 .
000 +41 49 47 .
40 20 .
35 4 SK205B 00 45 33 .
250 +40 17 08 .
40 15 .
29 4SK017C 00 35 28 .
440 +39 32 25 .
10 30 .
27 6 SK193C 00 45 49 .
970 +40 05 09 .
10 18 .
05 4SK018C 00 35 29 .
320 +41 42 33 .
30 19 .
51 4 SK196C 00 45 51 .
580 +40 04 43 .
80 18 .
17 4B134D 00 35 30 .
298 +40 44 24 .
84 20 .
01 4 SK214B 00 45 54 .
060 +39 56 46 .
80 19 .
86 6SK006B 00 35 34 .
240 +41 11 53 .
00 18 .
45 4 SK197C 00 45 57 .
630 +40 17 09 .
50 15 .
82 4SK007B 00 35 45 .
260 +39 39 21 .
30 28 .
57 4 SK200C 00 46 06 .
090 +40 22 26 .
00 15 .
01 4SK020C 00 35 49 .
740 +41 50 02 .
40 19 .
28 4 SK221B 00 46 19 .
240 +40 23 42 .
00 15 .
12 6SK021C 00 35 50 .
830 +39 36 00 .
80 29 .
02 4 B281D 00 46 22 .
279 +40 18 08 .
00 16 .
22 6SK022C 00 35 51 .
760 +40 54 11 .
60 18 .
40 4 SK204C 00 46 22 .
920 +40 20 42 .
20 15 .
76 4SK023C 00 35 53 .
100 +41 51 23 .
70 19 .
27 4 SK223B 00 46 32 .
880 +40 06 36 .
50 18 .
67 6SK024C 00 35 53 .
830 +41 43 42 .
60 18 .
60 4 B488 c
00 46 34 .
287 +42 11 42 .
78 15 .
99 5SK025C 00 35 54 .
220 +41 46 53 .
80 18 .
84 4 B489 00 46 36 .
386 +40 00 26 .
86 19 .
95 4SK008B 00 35 58 .
150 +39 37 35 .
50 28 .
54 4 SH21 00 46 37 .
308 +39 23 57 .
85 27 .
49 6SK009B 00 36 00 .
230 +40 56 19 .
20 17 .
92 4 B291D 00 46 41 .
270 +40 03 02 .
00 19 .
55 6SK010B 00 36 01 .
700 +39 48 50 .
20 26 .
45 4 B293D 00 46 48 .
097 +40 02 21 .
72 19 .
84 6SK011B 00 36 02 .
020 +41 14 43 .
40 17 .
23 4 B390 00 46 51 .
632 +40 23 46 .
90 16 .
00 6SK026C 00 36 05 .
610 +39 58 04 .
90 24 .
77 4 SK231B 00 47 14 .
110 +40 22 23 .
20 16 .
89 6SK029C 00 36 22 .
260 +39 52 04 .
50 25 .
30 4 BA28 00 47 14 .
220 +42 21 42 .
20 18 .
82 4B139D 00 36 24 .
679 +39 45 07 .
43 26 .
47 6 SK232B 00 47 14 .
430 +40 25 38 .
80 16 .
36 4SK030C 00 36 27 .
360 +41 35 14 .
00 16 .
67 4 DAO93 00 47 46 .
178 +42 44 55 .
88 23 .
92 4SK031C 00 36 31 .
430 +42 06 24 .
60 19 .
56 4 DAO94 00 47 54 .
399 +42 44 01 .
58 23 .
94 4SK013B 00 36 31 .
700 +41 11 41 .
30 15 .
99 4 BA10 00 47 56 .
286 +42 28 43 .
73 21 .
18 4SK032C 00 36 33 .
360 +41 30 03 .
10 16 .
16 4 SK222C 00 47 59 .
480 +41 54 13 .
00 15 .
98 6B142D 00 36 33 .
831 +41 09 07 .
96 15 .
95 4 SK238B 00 48 01 .
660 +41 49 56 .
90 15 .
56 6B144D 00 36 36 .
647 +41 37 03 .
65 16 .
40 6 SK223C 00 48 04 .
610 +40 08 27 .
00 20 .
72 4SK033C 00 36 37 .
890 +42 14 46 .
20 20 .
51 4 SH24 00 48 15 .
545 +42 25 17 .
12 21 .
11 6SK034C 00 36 43 .
420 +39 34 56 .
20 27 .
87 4 SK240B 00 48 24 .
140 +40 06 43 .
10 21 .
58 4SK035C 00 36 46 .
660 +41 26 23 .
90 15 .
47 4 SK243B 00 48 27 .
190 +42 02 43 .
50 18 .
04 4SK036C 00 36 47 .
200 +40 04 09 .
50 22 .
52 4 SK225C 00 48 31 .
340 +42 01 05 .
50 17 .
97 4SK037C 00 36 49 .
190 +39 39 43 .
70 26 .
82 4 SK249B 00 48 32 .
960 +42 02 45 .
00 18 .
24 4SK039C 00 37 00 .
980 +39 33 27 .
60 27 .
74 6 SK252B 00 48 41 .
130 +41 31 54 .
60 15 .
66 6SK040C 00 37 03 .
400 +41 33 22 .
10 15 .
08 6 SK227C 00 48 44 .
050 +42 15 48 .
80 20 .
45 4SK041C 00 37 05 .
080 +40 01 06 .
30 22 .
52 4 B504 00 48 45 .
168 +40 08 45 .
94 21 .
87 4SK042C 00 37 06 .
280 +41 44 48 .
50 15 .
82 4 SK228C 00 48 46 .
490 +41 46 45 .
80 16 .
94 4SK043C 00 37 09 .
100 +39 49 10 .
40 24 .
56 6 DAO99 00 48 48 .
314 +42 32 45 .
20 23 .
29 4SK046C 00 37 28 .
930 +41 55 01 .
90 16 .
09 4 B334D 00 48 54 .
848 +39 35 56 .
07 27 .
92 4SK017B 00 37 30 .
440 +40 36 43 .
80 16 .
22 4 SK256B 00 49 05 .
380 +41 57 38 .
30 18 .
78 4SK018B 00 37 30 .
820 +40 18 23 .
90 18 .
86 4 SK229C 00 49 11 .
040 +41 57 53 .
10 19 .
01 4c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Table 3.
Continued.Name in Position (J2000.0) R proj New Name in Position (J2000.0) R proj NewRBC V5 RA Dec (kpc) Class a RBC V5 RA Dec (kpc) Class a SK257B 00 49 15 .
210 +41 01 29 .
40 17 .
09 6 B346D 00 50 03 .
750 +40 37 39 .
28 20 .
84 4B338D 00 49 15 .
765 +40 46 23 .
47 18 .
14 4 SK258B 00 50 17 .
460 +42 06 42 .
60 22 .
45 4B339D 00 49 17 .
493 +40 45 06 .
58 18 .
32 4 B348D 00 50 19 .
219 +40 58 02 .
78 19 .
95 4DAO104 00 49 21 .
347 +42 16 16 .
60 21 .
73 4 B511 00 50 43 .
418 +40 11 13 .
39 25 .
43 4SK232C 00 49 25 .
630 +42 06 06 .
70 20 .
52 4 B512 00 50 46 .
324 +39 53 19 .
91 28 .
13 4B340D 00 49 29 .
174 +41 04 32 .
10 17 .
56 4 B513 00 50 47 .
806 +41 25 46 .
27 20 .
79 4B506 00 49 34 .
905 +40 00 28 .
94 24 .
74 4 G355 00 51 33 .
740 +39 57 35 .
81 29 .
06 4SK233C 00 49 35 .
650 +42 11 42 .
80 21 .
58 4 SH25 00 52 04 .
054 +41 35 05 .
85 24 .
29 4B345D d
00 49 52 .
554 +40 53 10 .
10 19 .
12 6 a All objects were originally classified as cluster candidates (class 2 or 3) except for B488 (class 5). b No object is visible at the coordinates specified for BH01. This is a very faint candidate from archival
HST imaging. c We retain the classification of B488 as a H ii region but note there may also be a young cluster at this location. d B345D appears to be a star superposed on a galaxy, so could also be classed as 4.
Figure 5.
PAndAS g -band thumbnails for objects mis-classified as globular clusters in the RBC V5. Each thumbnail is 1 ′ × ′ in size,with north to the top and east to the left. SK002A is a star; SK004A is two barely-separated stars; and BA11 is a star superimposed ona background galaxy. Figure 6.
PAndAS g -band thumbnails for representative examples of objects classified as globular cluster candidates in the RBC V5,that are galaxies (upper row) or stars (lower row, left three panels). In addition we include on the lower row images for BH01, for whichno object is visible at the listed coordinates, and B488, which is a H ii region that may also contain a dispersed young cluster. Eachthumbnail is 1 ′ × ′ in size, with north to the top and east to the left.c (cid:13) , 1–26 Huxor et al.
Table 4.
Confirmed globular clusters in the dTZZ13 catalogue.Name in PAndAS Position (J2000.0) R proj dTZZ13 Name RA Dec (kpc)SDSS1 −
00 36 01 . . −
00 39 13 . . . . −
00 42 27 . . . . . . . . . . . . . . . . WFPC2 imaging (Barmby & Huchra 2001). The thumbnailfor this target is also displayed in Figure 6.Finally, of the two H ii regions listed in our sample, wefound no compelling reason to alter the classification of one(DAO88), while at the coordinates of the second, B488, wefound a dispersed sample of luminous blue stars and a smallamount of nebulosity. This is consistent with its classifica-tion as a H ii region; however we note that there may possiblyalso be a young cluster at this location. A thumbnail for thisobject is shown in Figure 6.In summary, we inspected PAndAS thumbnails for 497objects listed at R proj >
15 kpc in the RBC V5. Of these, 141were originally classified as GC candidates; we were able toreclassify these as genuine GCs (2 objects) plus a H ii regionwith a possible embedded young massive cluster, foregroundstars (25 objects), and background galaxies (112 objects),while in one case no object was visible. Of the 72 targetsoriginally listed as definite GCs, we confirmed 69 but foundthat three were either foreground stars or background galax-ies. We did not change the classification of the 2 H ii regionsin the RBC list, and we confirmed the identity of the 282 ob-jects originally listed as contaminants. Finally, we added twomore new GCs (B270D and PAndAS-59) located by chanceas discussed above. During the preparation of this paper, di Tullio Zinn & Zinn(2013, hereafter dTZZ13) released a catalogue of M31 GCsand GC candidates derived from SDSS imaging. The areacovered by SDSS overlaps substantially with the PAndASfootprint, allowing us to check the identity of many of theobjects in the dTZZ13 catalogue – although a number alsolie well beyond the edge of the PAndAS coverage.The dTZZ13 catalogue consists of two primary lists.The first contains 18 objects classified as high confidenceGCs, while the second contains 75 lower confidence candi-date GCs. We located 17 of the high confidence targets inour PAndAS imaging, along with 42 of the candidates, andassessed these in the same manner as for objects in the RBC.The remaining dTZZ13 targets are at large radii from M31,150 ∼ < R proj ∼ <
230 kpc, and thus do not lie within the PAn-dAS footprint.We found that ten of the 17 high confidence objects
Table 5.
Non-clusters in the dTZZ13 catalogue.Name in Position (J2000.0) R proj ClassdTZZ13 RA Dec (kpc)SDSS2 00 38 26 . .
25 4SDSS5 00 41 47 . .
83 4SDSS7 00 47 41 . .
72 4SDSS10 00 55 28 . .
06 4SDSS13 01 16 41 . .
35 4SDSS14 01 22 20 . .
73 4SDSS18 23 49 09 . .
73 4C2 00 08 19 . .
23 4C3 00 08 34 . .
14 4C14 00 39 32 . .
00 4C15 00 40 09 . .
55 4C16 00 40 14 . .
13 4C17 00 40 31 . .
52 4C18 00 41 38 . .
96 4C20 00 42 09 . .
90 4C22 00 43 03 . .
71 4C23 00 43 32 . .
73 4C24 00 43 44 . .
93 4C26 00 44 01 . .
48 4C27 00 45 40 . .
22 4C30 00 48 25 . .
77 4C31 00 49 33 . .
39 4C32 00 49 37 . .
45 4C33 00 50 22 . .
12 4C34 00 51 12 . .
88 4C36 00 51 32 . .
37 4C37 00 51 47 . .
68 4C39 00 52 34 . .
33 4C40 00 54 06 . .
23 4C41 01 00 12 . .
83 4C45 01 05 43 . .
59 4C47 01 06 14 . .
64 4C48 01 06 40 . .
44 4C50 01 08 33 . .
81 4C51 01 09 40 . .
46 4C52 01 10 50 . .
72 4C55 01 14 29 . .
46 4C56 01 17 36 . .
10 4C58 01 19 43 . .
57 4C59 01 22 56 . .
27 4C60 01 26 10 . .
66 4C61 01 27 37 . .
50 4C63 01 28 38 . .
18 4C65 01 31 17 . .
68 4C66 01 32 45 . .
74 4C67 01 33 59 . .
41 4C68 01 34 06 . .
88 4C69 01 34 39 . .
82 4 that we inspected are indeed GCs, the remaining seven be-ing either stars or distant galaxies. Classifications for theseobjects are listed in Table 4. All but three of the ten GCs ap-pear independently in our PAndAS catalogue, as indicatedin the Table. The three outstanding clusters are at relativelysmall projected radii, R proj ∼ <
20 kpc, and thus fall withinthe inner limiting radius of our uniform search area. Thisadds further weight to the suggestion from our RBC workabove that a thorough search for GCs in the inner M31 halo c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Figure 7.
PAndAS g -band thumbnails for the three confirmed non-PAndAS globular clusters in the SDSS catalogue ofdi Tullio Zinn & Zinn (2013). Each thumbnail is 1 ′ × ′ in size, with north to the top and east to the left. may be fruitful. We show g -band thumbnails of the threenew SDSS clusters in Figure 7.Of the 42 candidate objects inspected, we only con-firmed one as a genuine GC. This is C62, which we alsolist in Table 4 and which also appears in our PAndAS cat-alogue. All of the other candidate objects turned out to bebackground galaxies; we list these in Table 5.Based on a simple extrapolation of our results, it is mod-erately likely ( ∼ R proj ∼ ∼ . R proj >
150 kpc, it would be veryworthwhile tracking these down.
We performed aperture photometry on each of our 59 newly-discovered GCs using the phot task in iraf . We also pho-tometered our two GC candidates, the additional 6 newly-confirmed GCs from the RBC and dTZZ13 listed in Tables2 and 4, SH06, and, to ensure a complete uniform sampleof measurements for the outer M31 system, all other knownGCs lying at R proj >
25 kpc (38 objects, predominantlyfrom Hux08). Our results may be found in Table 6.For each target, we used phot to measure the flux inconcentric apertures of increasing radius, and constructed acurve-of-growth. We employed the centroiding algorithm inthe phot task to accurately determine the cluster centres .This worked very well except on the most diffuse objects inour sample, which are fully resolved in the PAndAS imag-ing. For such targets we determined the centroid by eye, andverified that our photometric measurements were robust tochanges of a few pixels ( ∼ . ′′ ) in any direction about thispoint. For each GC we combined the central coordinates de-termined (independently) from the g - and i -band images in Note that our calculated coordinates for PA-31, and the hand-ful of other GCs detected by Martin et al. (2013), are somewhatdifferent than those listed in that paper. This is because Mar-tin et al. report the coordinates of the spatial grid point in theircalculation corresponding to the local probability maximum. Thecoordinates determined here are more accurate. a straight average, and these are the positions reported inTable 1. In all cases the difference in coordinates from the g - and i -band images was less than 0 . ′′ , and in most casesless than 0 . ′′ . To estimate the background flux for a givenGC we used the median level in an annulus of width 10 ′′ sitting outside the selected maximum photometry aperture.In practice the precise position of this background annu-lus was necessarily determined iteratively together with themaximum aperture itself.For an isolated cluster with little or no foreground orbackground contamination, we would define the maximumaperture r max to sit at a point where the increase in cumula-tive flux with radius (i.e., the curve-of-growth) is flat – thusensuring the inclusion of essentially all cluster light in themeasurement. Figure 8 shows an example for the GC PA-27.However, only ∼
60% of our systems conform to this ideal.A few objects are badly impacted by their proximity to theedge of a CCD (e.g., B517), a very bright star (H13, PA-9)or galaxy (HEC11), or, for more centrally located clusters( R proj <
25 kpc), the presence of moderately dense M31field populations (e.g., PA-32, PA-35). Such cases cannoteasily be corrected and thus for this type of object we wereforced to artificially constrain r max to a point on the curve-of-growth where the gradient is not necessarily flat, leadingto an under-estimate in the flux. Note that wherever possiblein this situation we kept the background annulus substan-tially outside the enforced limiting radius for photometry soas to avoid any cluster contribution to the estimated back-ground level – although in such cases this would never bethe dominant source of uncertainty in any event. On a fewrare occasions (e.g., PA-51, PA-55) a cluster fell so close toa CCD edge that it was only (partially) visible in one pass-band. In this situation useful photometry is not possible.The most common non-ideal scenario we encounteredfell between the two extremes of a completely isolated clus-ter and an object severely impacted by a chip edge or anexcessively bright local contaminant. Typically, a given tar-get might have an unrestricted maximum aperture, but a few( ∼ <
5) sources lying within this aperture that were obviouslyeither background galaxies or foreground stars of sufficientbrightness to noticeably impact the measured flux. In gen-eral we found it straightforward to mask these objects suchthat the affected pixels were not used in the flux calculation.We also note that a few clusters (e.g., G1, PA-53) are suf-ficiently bright so as to be mildly saturated at their centresin the PAndAS imaging. While this affects the shape of the c (cid:13) , 1–26 Huxor et al.
Figure 8.
Example of our photometric measurements for PAndAS-27. The left panel shows the g -band image of the cluster, with themaximum aperture ( r max = 14 . ′′ ) marked in red, and the colour aperture (3 . ′′ ) marked in green. North is to the top of the page, andeast to the left; the moderately bright foreground star within r max to the SSE of the cluster was masked during the procedure. The PSFFWHM is slightly broader than the PAndAS g -band median at 0 . ′′ . The central panel shows the curve-of-growth for PA-27; note thatthis has clearly levelled out by the time the maximum aperture is reached. The measured half-light radius is r h = 1 . ′′ ∼ . curve-of-growth at small radii, in no case was the saturationsevere enough to alter the total flux measured within r max .Given the variety of different circumstances seen acrossour sample, we assigned a flag to each object to indicatethe quality of the photometric measurement. These sit on ascale of A to D, with the following meanings: • A. An ideal isolated cluster, with an unrestricted max-imum aperture and limited or no masking of contaminantsources necessary. • B. Minor issues, such as the necessity for moderatemasking of contaminants, or a slightly restricted maxi-mum aperture due to a CCD edge, nearby bright star, ornon-trivial field background – but not sufficient to under-estimate the flux by more than ∼ . − . • C. Major issues and potentially significant unreliability,due to, for example, scattered light from a very nearby brightstar or galaxy, a strongly limited maximum aperture, or acontaminant coincident with the cluster centre, the maskingof which interfered substantially with the cluster flux. • D. Fatal problems, such as the majority of the clus-ter falling off the edge of a CCD, or the centre falling pre-cisely coincident with a bright contaminant that could notbe masked. Useful photometry is not possible for objects inthis category.The quality flags are included in Table 6 along with notesindicating the specific issues, if any, arising for each particu-lar GC (for example, whether r max was truncated, and if sowhy). For any analysis utilising our photometric measure-ments, only objects in categories A and B should be used.Photometry for objects flagged with a C is useful only fordetermining indicative properties such as whether a clusteris “bright” or “faint”, or “compact” or “diffuse”.Because all the GCs for which we derived photometryare either brand new, or sit at large galactocentric distances,there is minimal overlap between our sample and the set ofobjects possessing high precision luminosity measurements in the literature. We found eight compact category ‘A’ or‘B’ clusters in our sample for which luminosities were mea-sured from HST imaging by Tanvir et al. (2012) – H1, H4,H5, H10, H23, H24, H27, and B514. Because the
HST imag-ing is in different filters than our PAndAS data, we comparethe integrated absolute V -band luminosities, M V , calculatedfrom the total g - and i -band magnitudes as detailed in Sec-tion 5.4 below. The mean offset in M V between our mea-surements and those from Tanvir et al. is +0 .
09 mag, andthe dispersion about this value is 0 .
13 mag. Our luminosi-ties are typically a little fainter than the
HST measurements,which is not surprising as resolved photometry allows clus-ter members to be isolated even at radii well beyond ouradopted r max values.We located four additional compact category ‘A’ or ‘B’clusters in our sample that have previous luminosity mea-surements from HST imaging calculated by Barmby et al.(2007) – G1, G2, G339, and G353. When added to the Tan-vir et al. clusters, the mean offset in M V between our mea-surements and those from the literature drops to +0 .
01 mag,but the dispersion rises somewhat to 0 .
18 mag.Finally, there are two very diffuse clusters in our samplethat were measured by Tanvir et al. – HEC7 and HEC12.For these two objects we find M V to be more substantiallyunder-estimated, by 0 .
46 and 0 .
52 mag respectively. It is notclear why our M V estimates are ∼ . HST values – most likely this reflects an inherent system-atic limitation in integrating the extremely faint diffuse lightcomponent of these objects on medium-deep ground-basedimaging.
In addition to determining the GC luminosities, we alsoused the curves-of-growth to obtain an empirical measureof their structures – as parametrised by the half-light radius r h , which is the projected radius of an aperture encircling c (cid:13)000
In addition to determining the GC luminosities, we alsoused the curves-of-growth to obtain an empirical measureof their structures – as parametrised by the half-light radius r h , which is the projected radius of an aperture encircling c (cid:13)000 , 1–26 he final PAndAS catalogue of M31 outer halo GCs half a cluster’s flux. We report r h for each GC in Table6; this quantity for a given target is the straight average ofthe (independent) measurements from the g - and i -band im-ages . The quoted sizes are not meant to represent extremelyprecise measurements of the cluster structures – performingsuch work on distant objects such as these using ground-based imaging is challenging and complex, and beyond thescope of the present paper. Rather, our estimates of r h areintended to provide a quantitative indication of whether aGC is compact or diffuse, or somewhere in between. It hasbeen known for some time that the halo of M31 hosts nu-merous unusually extended clusters (see e.g., Huxor et al.2005, 2011), and it is thus of interest to be able to examine,even if just at an indicative level, the distribution of GCsizes across the complete sample.We first note that our method of estimating r h is robustonly if r max falls on the flat part of the curve-of-growth. Ifnot, then both the total luminosity and the half-light ra-dius will be under-estimated. As described above, clustersfor which r max was truncated are flagged in the table; quotedsizes for these objects should be treated with caution.An additional, and arguably more important factor toconsider is the effect of the seeing profile on our size mea-surements. Compact GCs in both the Milky Way and M31have r h ∼ − ∼ . − . ′′ at theM31 distance ( µ = 24 . . ′′ in g and 0 . ′′ in i . Thusobservations of r h for very compact clusters in our samplelargely reflect the seeing profile of the PAndAS image inwhich the object falls, rather than the intrinsic propertiesof the GC. In principle this problem may be corrected bycareful deconvolution of the local image point-spread func-tion (PSF) and the radial brightness profile of the cluster;this problem will be addressed by an upcoming analysis of asubstantial new HST dataset (Mackey et al. 2014, in prep.).For now, we used artificial GC images generated to assessthe completeness of our PAndAS catalogue (see Section 6)to explore the impact of the PSF on our size measurements.We constructed two representative samples from the fullsuite of 4760 artificial GCs. As we describe in detail in Sec-tion 6, these objects were generated by first specifying astructure and luminosity, and then constructing a realisticimage assuming the median PAndAS seeing. Both of oursamples contained GCs spanning the full range of input half-light radii r h ∼ −
35 pc, but for one ensemble the clusterluminosities fell within the range M V = − . ± . M V = − . ± .
3. We passed each arti-ficial cluster in these two samples through our photometrypipeline. Note that we only selected objects that would havebeen classified in category ‘A’ in terms of the quality of thephotometric measurement.The results of this process are presented in Figure 9.The marked error-bar for a given size bin corresponds to Although r h may, in principle, be intrinsically slightly variablebetween various passbands (if, for example, a GC is mass segre-gated), in practice our individual measurement errors of ∼ > Figure 9.
Measured versus input half-light radii for artificial clus-ters with luminosities M V ∼ − ∼ − . x -axis. Figure 10.
Half-light radii derived in this paper versus those de-rived from
HST imaging by Barmby et al. (2007) (open points)and Tanvir et al. (2012) (filled points). Circles represent compactGCs, while triangles are diffuse GCs. Note that the two trianglesshould sit at ∼
20 pc and ∼
30 pc, but have been plotted atsmaller radii to maintain clarity. The inclined dotted line repre-sents a straight linear fit to all points in Figure 9 for which theinput size was below 9 pc. The apparently deviant point fromBarmby et al. (2007) is G2, which is mildly saturated in our im-ages (pushing r h to a larger value). The point for G1 falls well offthe top of the plot as it is quite strongly saturated in the PAndASimages. the standard deviation in the measured GC sizes withinthat bin. It is clear that for GCs with input r h larger than ∼ −
10 pc, we recover a very reasonable estimate of the ob-ject’s size. This appears to be true irrespective of luminosity,although not surprisingly the scatter noticeably increases forlower-luminosity GCs compared to higher-luminosity GCs.Our tests indicate that typical uncertainties in the measuredvalues of r h are ∼ <
8% for M V ∼ − ∼ <
12% for M V ∼ − . ∼ r h .Below r h ∼ −
10 pc it is clear that the size mea-surements are significantly affected by the seeing profile, asexpected. This limit corresponds to roughly three times theFWHM of the PSF used when constructing the artificialcluster images. It is interesting to note that while the mea-sured r h values become increasingly different from the inputvalues when moving to smaller sizes, within the limitationsof the scatter the ordering is preserved. That is, a GC thatis intrinsically more compact than another will still be mea-sured as such by our photometry pipeline even when strongly c (cid:13) , 1–26 Huxor et al.
Table 6.
Photometric measurements for PAndAS globular clusters and selected others.Cluster E ( B − V ) r max g i ( g − i ) M V ( V − I ) r h Quality NotesName (arcsec) (pc) FlagPAndAS-1 0.099 14.8 17.55 16.74 0.62 -7.48 0.83 7.1 A ...
PAndAS-2 0.106 19.2 18.30 17.29 0.70 -6.82 0.90 25.7 A ...
PAndAS-3 0.087 13.9 20.89 19.90 0.65 -4.17 0.86 27.4 B cPAndAS-4 0.133 14.1 18.07 17.17 0.69 -7.09 0.89 4.7 A ...
PAndAS-5 0.078 11.8 19.94 19.08 0.79 -5.05 0.97 17.0 A cPAndAS-6 0.068 14.1 16.92 16.15 0.67 -8.02 0.87 4.4 A ...
PAndAS-7 0.088 11.8 20.18 18.77 0.70 -5.00 0.89 13.3 A cPAndAS-8 0.109 10.4 19.89 18.29 0.87 -5.40 1.03 9.5 A ...
PAndAS-9 0.090 5.2 18.23 17.51 0.62 -6.75 0.83 3.8 C b,r ∗ PAndAS-10 0.094 12.2 19.61 18.75 0.75 -5.43 0.93 15.2 A cPAndAS-11 0.088 12.6 18.29 17.41 0.67 -6.74 0.87 9.4 A ...
PAndAS-12 0.060 7.4 19.63 18.72 0.75 -5.33 0.94 13.7 B e,rPAndAS-13 0.063 7.4 18.43 17.66 0.65 -6.49 0.85 4.7 B m,b,rPAndAS-14 0.069 12.6 17.93 17.17 0.71 -7.01 0.90 10.9 A ...
PAndAS-15 0.069 3.7 19.92 19.08 0.74 -5.04 0.93 6.0 C b,e,r ∗ PAndAS-16 0.072 16.3 16.54 15.66 0.79 -8.44 0.97 5.5 A ...
PAndAS-17 0.067 14.1 16.87 15.77 1.00 -8.17 1.14 4.4 A ...
PAndAS-18 0.062 14.8 19.61 18.71 0.76 -5.35 0.94 23.0 A cPAndAS-19 0.055 8.1 20.17 19.39 0.72 -4.73 0.91 7.3 A ...
PAndAS-20 0.067 8.1 19.57 18.58 0.83 -5.43 1.00 7.4 A ...
PAndAS-21 0.054 14.1 17.84 17.06 0.67 -7.06 0.87 4.0 A ...
PAndAS-22 0.063 10.4 18.79 17.87 0.91 -6.18 1.06 7.3 A ...
PAndAS-23 0.054 6.7 19.98 18.89 1.04 -5.02 1.17 6.7 A ...
PAndAS-24 0.059 11.8 20.27 19.39 0.73 -4.68 0.91 16.9 A cPAndAS-25 0.064 4.8 19.78 18.80 0.88 -5.21 1.04 6.6 B b,rPAndAS-26 0.061 5.6 19.88 18.92 0.94 -5.10 1.09 7.4 B e,rPAndAS-27 0.075 14.8 17.31 16.41 0.75 -7.69 0.93 5.2 A ...
PAndAS-28 0.066 8.3 19.26 18.56 0.64 -5.65 0.85 12.7 B f,rPAndAS-29 0.058 5.2 20.58 19.75 0.72 -4.35 0.91 10.7 B c,f,rPAndAS-30 0.064 7.4 19.57 18.58 0.79 -5.42 0.96 10.9 A cPAndAS-31 0.073 9.3 20.62 19.60 0.82 -4.41 0.99 18.5 B c,mPAndAS-32 0.075 6.7 19.48 18.37 0.97 -5.58 1.11 8.0 B f,rPAndAS-33 0.059 18.5 19.56 18.67 0.68 -5.39 0.88 35.8 B c,bPAndAS-34 0.068 12.6 18.37 17.38 0.81 -6.64 0.98 10.4 A ...
PAndAS-35 0.086 5.2 19.91 18.59 1.09 -5.24 1.21 10.3 B c,f,rPAndAS-36 0.073 10.4 17.69 16.79 0.75 -7.30 0.94 5.8 A ...
PAndAS-37 0.057 9.6 17.66 16.56 1.00 -7.35 1.13 4.6 A ...
PAndAS-38 0.159 13.9 20.76 19.75 0.61 -4.50 0.83 24.4 B c,mPAndAS-39 0.086 9.6 18.85 17.90 0.88 -6.19 1.04 13.0 B c,f,rPAndAS-40 0.058 10.0 19.80 18.97 0.77 -5.13 0.95 10.0 A ...
PAndAS-41 0.096 − − − − − − −
D e,r ∗ PAndAS-42 0.060 15.5 18.54 17.04 0.89 -6.59 1.05 15.4 C c,bPAndAS-43 0.093 4.4 19.79 18.85 0.79 -5.27 0.97 6.0 B c,e,rPAndAS-44 0.062 9.6 17.18 16.48 0.61 -7.72 0.82 3.1 A ...
PAndAS-45 0.083 7.4 20.97 20.05 0.79 -4.06 0.96 8.9 B cPAndAS-46 0.072 16.3 16.27 15.52 0.64 -8.67 0.85 4.3 B sPAndAS-47 0.070 5.6 19.39 18.26 1.01 -5.66 1.14 3.8 A ...
PAndAS-48 0.066 13.7 20.21 19.41 0.59 -4.73 0.81 21.2 A cPAndAS-49 0.067 11.1 20.24 19.11 0.92 -4.81 1.07 16.4 A ...
PAndAS-50 0.163 14.8 18.93 17.78 0.95 -6.38 1.10 17.1 A ...
PAndAS-51 0.074 − − − − − − −
D e,xPAndAS-52 0.063 13.3 17.38 16.49 0.78 -7.58 0.96 6.5 B bPAndAS-53 0.053 12.6 15.79 15.07 0.64 -9.09 0.85 4.2 B s,b,rPAndAS-54 0.053 12.6 16.30 15.57 0.63 -8.58 0.84 5.1 C b,mPAndAS-55 0.070 − − − − − − −
D e,xPAndAS-56 0.050 14.1 17.27 16.45 0.70 -7.63 0.89 4.7 A ...
PAndAS-57 0.066 8.9 19.24 18.44 0.71 -5.70 0.91 10.3 A cPAndAS-58 0.062 11.5 18.82 17.83 0.88 -6.17 1.04 9.3 A ...
PAndAS-59 0.068 3.7 19.96 19.32 0.53 -4.93 0.76 5.6 B f,rPAndAS-Ca1 0.067 4.4 20.84 20.27 0.45 -4.03 0.70 8.1 B f,rPAndAS-Ca2 0.175 4.4 20.02 19.08 0.65 -5.26 0.86 10.4 C e,r ∗ c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Table 6.
Continued.Cluster E ( B − V ) r max g i ( g − i ) M V ( V − I ) r h Quality NotesName (arcsec) (pc) FlagG1 0.057 32.6 14.17 13.23 0.77 -10.79 0.95 8.7 B s,b,mG2 0.052 25.2 15.97 15.21 0.67 -8.92 0.87 5.0 B sG339 0.093 11.8 17.47 16.55 0.75 -7.58 0.94 5.3 A ...
G353 0.082 10.4 17.39 16.58 0.67 -7.60 0.87 4.8 A ...
B514 0.058 25.2 16.02 15.17 0.77 -8.91 0.95 6.6 A ...
B517 0.065 − − − − − − −
D e,r ∗ EXT8 0.068 15.5 15.79 14.58 0.56 -9.28 0.79 4.3 B sMGC1 0.086 37.4 15.60 14.15 0.71 -9.59 0.91 8.8 A ...
H1 0.070 18.5 16.25 15.45 0.68 -8.70 0.88 4.5 B sH2 0.059 18.5 17.43 16.60 0.73 -7.50 0.91 5.2 A ...
H3 0.069 8.9 18.48 17.53 0.85 -6.52 1.01 5.5 A ...
H4 0.073 17.0 17.17 16.28 0.74 -7.82 0.93 5.4 A ...
H5 0.075 17.0 16.51 15.76 0.64 -8.44 0.85 9.5 A ...
H7 0.057 17.0 17.76 16.92 0.73 -7.17 0.92 10.5 A ...
H8 0.062 11.1 19.33 18.18 0.87 -5.71 1.03 11.8 B fH9 0.055 − − − − − − −
D e,xH10 0.065 23.7 16.25 14.88 0.77 -8.86 0.95 5.8 A ...
H11 0.071 11.8 17.10 16.23 0.75 -7.88 0.93 4.1 A ...
H12 0.066 13.3 16.75 15.95 0.68 -8.19 0.88 4.0 A ...
H15 0.057 11.8 18.46 17.20 0.62 -6.60 0.83 10.3 A cH17 0.052 9.6 17.85 16.47 0.79 -7.23 0.96 3.3 A ...
H18 0.087 14.8 16.92 16.08 0.70 -8.09 0.90 4.1 A ...
H19 0.059 7.4 17.66 16.77 0.74 -7.29 0.92 4.9 A ...
H22 0.050 11.1 17.26 16.43 0.74 -7.65 0.93 4.2 B gH23 0.051 8.9 17.00 15.55 0.85 -8.09 1.01 3.6 B b,rH24 0.098 14.8 17.97 17.04 0.75 -7.10 0.94 8.9 A ...
H25 0.094 14.8 17.12 16.21 0.76 -7.93 0.94 5.8 A ...
H26 0.053 14.8 17.66 16.34 0.70 -7.40 0.90 5.6 A ...
H27 0.055 14.8 16.66 15.39 0.66 -8.39 0.86 4.9 A ...
HEC1 0.060 12.6 19.06 18.39 0.65 -5.82 0.86 15.7 A ...
HEC2 0.055 12.2 19.51 18.03 0.78 -5.60 0.96 12.4 B c,e,rHEC3 0.070 15.5 19.63 18.71 0.83 -5.36 1.00 17.6 A cHEC6 0.073 18.5 19.09 18.12 0.75 -5.92 0.94 26.7 A cHEC7 0.087 16.7 18.48 17.53 0.82 -6.57 0.99 19.5 A cHEC10 0.106 20.4 18.97 17.98 0.79 -6.14 0.97 22.5 A cHEC11 0.048 8.9 18.41 17.03 0.70 -6.65 0.90 14.6 B c,g,rHEC12 0.049 20.4 18.93 17.48 0.82 -6.16 0.99 29.9 A cHEC13 0.048 13.7 19.48 18.27 0.64 -5.54 0.85 20.7 B c,e,rB270D 0.087 6.7 17.77 16.86 0.76 -7.26 0.94 6.6 B f,rSK213C 0.121 3.3 19.43 18.43 0.81 -5.72 0.98 4.4 B f,rSK255B 0.070 8.1 18.01 17.00 0.89 -7.01 1.04 5.6 B b,f,rSH06 0.168 8.1 16.43 16.55 -0.49 -8.45 0.00 9.9 A ...
SDSS1 0.068 8.1 18.33 17.22 0.95 -6.71 1.10 12.0 B mSDSS3 0.066 6.7 19.06 17.95 1.03 -5.98 1.16 7.2 A ...
SDSS6 0.079 − − − − − − −
D e,xNotes: b=nearby bright star, poorly masked or not maskable; c=centroided by eye; e=affected by CCD edge;f=high field star density; g=nearby bright galaxy, poorly masked or not maskable; m=masking required formany contaminanting sources; r=restricted maximum aperture; r ∗ =severely restricted maximum aperture;s=saturated at centre; x=missing in one or both filters. affected by the PSF. We must bear in mind that the seeingprofile does vary between PAndAS images, unlike for ourartificial clusters; however, as previously reported the rmsscatter about the mean seeing values is small ( ∼ . ′′ ).We return briefly to the sample of 12 compact GCs and2 diffuse GCs for which high precision photometry and struc-tural measurements exist in the literature. Figure 10 shows acomparison between our r h measurements and those derivedfrom HST imaging by Barmby et al. (2007) and Tanvir et al. (2012). This strongly resembles Figure 9 – for cluster sizesbelow ∼ −
10 pc, our estimated r h values are clearly toolarge; however the correct ordering is preserved. Indeed ourmeasurements appear to behave exactly as predicted by theartificial cluster tests – the inclined dotted line represents astraight linear fit to all points in Figure 9 with input sizebelow r h = 9 pc, and this provides an excellent descrip-tion of how strongly our measured quantities deviate fromthose obtained via HST . As a final note, we see that for the c (cid:13) , 1–26 Huxor et al. two diffuse clusters our r h measurements match those from HST to better than ∼ Comparing our g - and i -band flux measurements for a givenGC allows us to derive the integrated colour of that object.In principle, we could calculate ( g − i ) directly from the totalluminosities measured within r max . However, previous workhas found that employing a smaller aperture can lead to amore robust result (e.g., Huxor et al. 2009; Veljanoski et al.2013b). This is perhaps not too surprising, as the largerthe aperture for colour measurement, the more sensitive theresult is to (i) the presence of unidentified contaminants,and (ii) the accuracy of the estimated background level.For the present work, there are some subtleties associ-ated with determining an optimal colour aperture. First, thismethodology is predicated on the absence of intrinsic colourgradients within the target GCs. This appears reasonable– high resolution studies from HST have not revealed anysuch gradients (e.g., Tanvir et al. 2012). Second, our cata-logue spans a very large range of cluster sizes, so it doesnot make sense to simply apply a uniform colour apertureacross the entire sample. A small aperture that might workwell for a compact GC could lead to a very misleading resultfor a diffuse GC, as it would be extremely sensitive to thepresence, or absence, of a handful of bright stars at the cen-tre of such an object. Much better is to define an aperturethat samples the same region in every cluster – for example, r h ; however this introduces a third issue which is that, aswe have already seen, compact GCs are strongly affected bythe seeing profile. In the context of a colour measurement,it is important to recognise that any difference in the PSFwidth between the g - and i -band images leads to an artificialcolour gradient at the centre of the object due to the differ-ential redistribution of flux. If the colour aperture is set tobe too small, any such gradient will result in an erroneousmeasurement.Fortunately, we already know from Figure 9 the radiusat which the effect of the seeing profile becomes negligible.Thus, we set the colour aperture to be equal to r h for allGCs down to a conservative limit of r h = 3 . ′′ ∼
13 pc;for any clusters with r h smaller than this, the colour aper-ture is set at 3 . ′′ . This lower limit matches the uniformcolour aperture used in comparable studies, such as that ofVeljanoski et al. (2013b) for GCs in the M31 dwarf ellipti-cal (dE) companions NGC 147 and 185 (although note thatthat sample did not span anything like the range in size asdoes the present sample). We conclude this section by summarising the measurementsreported in Table 6. For each GC we list the foregroundcolour excess E ( B − V ) as derived from the Schlegel et al.(1998) maps, and the maximum photometry aperture r max .Next, we list the total integrated g - and i -band AB magni-tudes within r max , along with the ( g − i ) colour determinedfrom a more central aperture as described above. The colour as reported has been dereddened using the appropriate co-efficients from the study of Schlafly et al. (2011): g = g − . E ( B − V ) i = i − . E ( B − V ) . (1)An important subtlety is that in mid-2007 theCFHT/MegaCam i -band filter was broken, and subse-quently replaced with a new filter possessing a slightlydifferent transmission profile. As a result, instrumental i -band magnitudes for GCs falling in images taken priorto June 2007 are calibrated to a slightly different systemthan for GCs taken after this date. To ensure a consistentset of measurements across the entire sample, we usethe relationship from Ibata et al. (2014) to transform thephotometry for GCs imaged with the old i -band filter ontothe system of the new filter. Since all our objects have( g − i old ) < . i new = i old + 0 .
031 ( g − i old ) − . . (2)To facilitate comparison with GCs in the inner parts of M31,as well as in systems belonging to other galaxies, we alsolist in Table 6 our photometry transformed to the standardJohnson-Cousins system using the relations from Hux08 (seealso Veljanoski et al. 2013b). We first convert from AB mag-nitudes to Vega magnitudes: g = g + 0 . i = i old − . , (3)and then transform to V and I : V = g − .
42 ( g − i ) + 0 .
04 ( g − i ) + 0 . I = i − .
08 ( g − i ) + 0 . . (4)Note that, as explicitly denoted in Equation 3, these rela-tions are valid only for photometry in the old i -band system.Thus we transform all our measurements into this systemusing the inverse of Equation 2 prior to implementing theabove procedure. In Table 6 we list the absolute V -bandmagnitude M V and the dereddened ( V − I ) colour. We cal-culate these from V and I assuming a distance modulus µ = 24 .
47, the relevant E ( B − V ), and, as before, the ap-propriate coefficients from Schlafly et al. (2011): V = V − . E ( B − V ) I = I − . E ( B − V ) . (5) A thorough, quantitative assessment of detection complete-ness is critical to the utility of our globular cluster catalogue.We have identified two major sources of incompleteness as-sociated with the PAndAS data and our search technique,and we quantify each of these below.
Although PAndAS does an excellent job of achieving uni-form imaging of the M31 halo to R proj ≈ −
150 kpcin all directions, there are myriad small gaps in its spatialcoverage. These arise from two sources: (i) spaces betweenthe first and second, and third and fourth rows of CCDs on c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs the MegaCam focal plane, which were typically not filled inby the small telescope dithers employed during the observa-tions; and (ii) imperfect tiling of the PAndAS mosaic.In general this spatial incompleteness is of no conse-quence for the primary goal of the PAndAS survey – study-ing the properties of the resolved M31 field halo and thelarge-scale substructures and overdensities that are foundwithin it. Individual satellite dwarf galaxies of M31 are alsotypically large enough to span the missing regions. GCs, onthe other hand, are sufficiently small on the sky that theycan easily fall into a gap in the coverage and not be detected.We are aware of at least one such case of this occurring –the object H9 from Hux08, which sits at R proj = 56 kpc andwas originally discovered in INT/WFC observations, doesnot appear in any PAndAS image because it sits squarely inan inter-chip gap.Fortunately, this kind of incompleteness is straightfor-ward to quantify. It affects all GCs equally, irrespective oftheir morphology or luminosity; all that is required is tocalculate the fraction of missing coverage as a function ofprojected galactocentric radius. To do this, for every PAn-dAS image we used the WCS information in the header todetermine the coordinates of the four corners of each of the36 CCDs and hence the equations defining their edges inRA and Dec. Note that we treated g - and i -band imagesindependently in the list, as there are often small spatialoffsets between images in the two filters and we only neededcoverage from one filter to identify a GC.Next, we constructed circular annuli about the M31 cen-tre and filled each one with points generated at random po-sitions so as to achieve uniform coverage of the area withinthe annulus. Each annulus was of width 0 . . ◦ on the sky. We generated enough pointsper annulus to achieve a minimum density of 100 arcmin − – sufficient to properly sample the inter-row gaps betweenCCDs ( ∼ ′′ wide). Using the complete list of CCD edgeequations, we tested each point to see whether it fell withinthe area covered by any given chip, and hence within theimaged region of the PAndAS footprint. The fractional cov-erage for a particular annulus was then simply the ratio ofimaged to total points generated inside that annulus.Figure 11 shows our results. In the central regions ofM31, within R proj ≈
10 kpc, there are sufficient overlappingimages that the spatial coverage is complete. Beyond this,the coverage falls to ∼
96% all the way out to R proj = 105kpc, where the irregular edge of the PAndAS footprint grad-ually begins to affect the completeness. There is a shallowdecline to ∼
80% coverage at R proj = 130 kpc, and thenbeyond this a rapid drop to ∼
20% coverage at 150 kpc.How does this affect our GC sample? Includingpreviously-catalogued objects, we know of 82 GCs with pro-jected radii in the range 25 R proj <
105 kpc; however only81 of these appear in PAndAS imaging (recall that H9 fallsin an inter-chip gap). The 96% spatial coverage over thisradial range leads us to expect 84 . R proj <
130 kpc we know of 8 GCs,all of which appear in the PAndAS imaging. A crude inte-
Figure 11.
Fractional spatial coverage of the PAndAS surveyimaging as a function of projected radius from the M31 centre. gration of the completeness function suggests we are miss-ing ∼ ∼ < R proj <
150 kpc, we have found only 1 cluster inthe PAndAS imaging (Mackey et al. 2013a); there is proba-bly ∼ < ≈ − R proj <
150 kpc due to the incompletespatial coverage of the PAndAS imaging.
The completeness of our catalogue is also affected by ourability to identify objects as GCs. That is, it is certain thatwe miss some clusters due to them being too small, faint,compact, or diffuse (or some combination of these) to recog-nise as GCs. There is also the possibility of human error toconsider – missing objects due to, say, a lapse in attentionwhile searching images.All indications suggest that human error is a negligiblefactor for our search. As a first pass, one of us (ADM) in-spected ∼
30% of the images previously searched by APH,including a number with no GCs as well as some of thosemore heavily populated with GCs. In all cases the con-sistency of the results was excellent, suggesting that ourmethodology, at least in uncrowded regions of low back-ground, is robust. In addition to this, we recovered all knownGCs in our primary search area ( R proj >
25 kpc) – from boththe RBC (Galleti et al. 2004) and the previous INT survey(Hux08) – with no omissions (barring H9). Finally, the auto-mated search for dwarf spheroidal satellites of M31 devisedby NFM (Martin et al. 2013) recovered just one missed GCacross our entire survey area. This object (PA-31) is dif-fuse and very faint, falling near our 50% completeness limit(see below) – so its original omission is not surprising. Thesearch algorithm is sensitive only to objects possessing asufficient number of resolved but relatively uncrowded stars.This describes just a relatively small fraction of our final GCcatalogue, but includes bright objects such as MGC1 (e.g.,Martin et al. 2006; Mackey et al. 2010b), as well as fainterextended clusters like PA-31. That it did not return a signif-icant number of missed systems is another indication thathuman error has not introduced appreciable incompletenessinto our catalogue.To quantify how our ability to identify GCs in PAndASimaging is affected by cluster luminosity and structure, we c (cid:13) , 1–26 Huxor et al. used a sample of artificial GCs. Ideally, these would be addedinto a wide variety of the PAndAS images themselves andthen “discovered” (or not) via a search methodology identi-cal to that which we originally employed. This would havethe added benefit of facilitating a more precise quantificationof any incompleteness arising due to human error, as wellas that due to the presence of very bright foreground stars(which we assume to be negligible due to the small numberof such objects). Unfortunately, however, this technique isnot practical. To achieve barely viable statistics requires aminimum sample approaching ∼ ∼
50 times the area of the PAndAS footprint. Even dis-tributing the artificial GCs with an unrealistic factor of tenhigher density would still require a search of several timesthe PAndAS footprint.To circumvent this issue we employed a simpler tech-nique. We constructed small thumbnails with our artificialGCs at the centre, one per thumbnail, and then inspectedeach of these with the aim of determining, as objectivelyas possible, which would have been identified as a GC andwhich not. This methodology facilitated both our main aimof quantifying the faint limit of our survey, and our sec-ondary aim of exploring how strongly this varies with clusterstructure.Under the assumption that incompleteness due to hu-man error is negligible, there ought to be no difference be-tween results derived from our simple inspection techniqueand those derived via a full search for artificial GCs. This as-sumption is a good approximation for luminous and/or com-pact GCs – objects which (i) were targeted by our inspectionof colour-magnitude selected candidates, and (ii) were typ-ically prominent and thus easily-located in the blind visualsearch. However, the approximation may break down subtlyfor objects of very low surface brightness because we knew, apriori , that each artificial thumbnail hosted an object at itscentre. This is in contrast to the real situation where it wasnecessary to first find these objects in the blind search .Ultimately this mild systematic bias could mean that ourderived faint completeness limits are too generous by a fewtenths of a magnitude – and indeed, as we point out later,we may observe weak evidence for such an effect.Our artificial clusters were generated across a binnedgrid in luminosity and concentration , extending betweenthe limits − M V − . c .
5. We generallyadopted bin sizes of 0 .
25 mag in M V and 0 . M V − c > . M V and c , in order to uniformly sample the bin. Thisresulted in a total ensemble of 4760 artificial GCs.We generated thumbnail images of these objects usingthe SimClust software (Deveikis et al. 2008). This pack-age generates a random realisation of a GC given its age, Recall that diffuse clusters typically did not appear in the listof colour-magnitude selected candidates. The concentration c = log( r t /r c ) where r t is the cluster tidalradius and r c the core radius, assuming a King (1962) modelfit to the radial surface density profile. A concentration c = 2 . metallicity, mass ( M cl ), and structural parameters ( r c and r t ), and then ”observes” this model to produce a realis-tic image. For simplicity we assumed a uniform age of 13Gyr and metallicity of [Fe / H] = − . M/L = 2, which is ap-propriate for the assumed age and metallicity. The structuralparameters were determined by first assigning to each GC arandom 3D galactocentric radius within 25 R gc
145 kpcto match the range of projected radii observed for our PAn-dAS GC sample. This then defined the tidal radius accordingto the usual relationship r t = R gc ( M cl /M g ) / where we as-sumed a galactic mass M g = 1 . × M ⊙ for M31, andthen our randomly generated concentration parameter de-termined r c . Given this set of input parameters, SimClust randomly selects stars from an appropriate Padova isochrone(Marigo et al. 2008) according to a set mass function, untilthe desired cluster mass is reached. We used the segmentedpower-law mass function of Kroupa (2001) for our GCs. Thestars are randomly distributed spatially according to a King(1962) model with appropriate r c and r t . SimClust converts all stellar positions and luminosi-ties for a given artificial GC to “observed” quantities ac-cording to a specified distance and foreground extinction –we used µ = 24 .
47, and the typical colour excess across thePAndAS footprint of E ( B − V ) = 0 . SkyMaker software pack-age (Bertin 2009), which generates the thumbnail images . Skymaker also requires a model PSF, which we generatedusing the iraf psf and seepsf tasks assuming a Gaussianprofile of FWHM ∼ . ′′ , corresponding to the mean g -bandstellar profile in PAndAS. We further specified the remain-der of the SkyMaker parameters to have values appropri-ate for CFHT/MegaCam and PAndAS. Finally, we also em-ployed the ability of
SkyMaker to randomly add field starsacross each thumbnail; we tweaked the relevant parametersto match, empirically, the range of field densities observedlocally about our PAndAS GCs. Figure 12 shows 1 ′ × ′ cen-tral cut-outs from a handful of representative artificial GCimages, across the luminosity-concentration plane.Once all the artificial GC images had been generated,the order was randomised and the full set supplied to APHfor inspection. To ensure a completely blind test, no ac-companying information on individual cluster properties wasprovided. Once the inspection was complete, the classifica-tions (a simple yes or no for each object) were returnedto ADM for analysis. Figures 13 and 14 show the results.Our survey is complete to at least M V = − c ∼ .
25; there is a gradual fall-offfor concentrations within ± .
75 of this value, and a greater By default
SimClust provides
UBV RIJHK magnitudes to
SkyMaker , which then produces images in these passbands. Wemade a small modification to the software in order to produce g -band magnitudes (and images), which we calculated accordingto the inverse of the transformation equations in Section 5.4.c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Figure 12.
Examples of g -band artificial cluster images across the luminosity-concentration plane. These are central 1 ′ × ′ cut-outsfrom our 1 . ′ × . ′ thumbnails. Each assumes a stellar FWHM of 0 . ′′ , typical for the vast majority of PAndAS g -band imaging. fall-off for very concentrated clusters with c >
2. Whereasour catalogue is >
95% complete down to M V ∼ − c <
2, it is only ∼
80% complete at M V = − c >
2. The main reason for this is that,except for the most diffuse examples, GCs in PAndAS arepredominantly recognisable as a group of resolved giant starssurrounding an unresolved, or partially resolved, core. Forthe most concentrated systems the resolved halo vanisheswith decreasing luminosity, leaving just a small central corethat is indistinguishable from a foreground star or compactbackground galaxy. This effect is clearly visible in Figure 12.Considering the sample as a whole, our 50% complete-ness level occurs at M V = − .
1. The effect of differing clus-ter structures is to move this level by a few tenths of amagnitude in either direction about the mean value. The50% completeness levels for GCs with 0 . c < . . c < . . c < . M V = − .
8, while for the most compact clusters with 2 . c < . M V = − .
6. Note that irrespective of structure, thereis essentially no chance of detecting GCs with M V ∼ > − . M V = − .
06 (PA-45). Be-tween − < M V < −
3, we expect to detect about 20% ofany clusters that are present (see the middle panel of Figure14). However, we found none – suggesting that (i) few suchGCs exist in the halo of M31, and/or (ii) the PAndAS imag-ing data are somewhat more demanding than the syntheticGC data used to estimate the completeness, and/or (iii) themild selection bias we described above for very low surfacebrightness clusters has pushed our faint-end completenesslimits too low by a few tenths of a magnitude.It is informative to consider our completeness limits interms of the half-light surface brightness Σ g,h (that is, themean g -band surface brightness within the cluster half-lightradius r h ). Here, the fall-off is very sharp – our mean 50%limit occurs at Σ g,h = 26 . − , and there is es- c (cid:13) , 1–26 Huxor et al.
Figure 13.
Detection completeness as a function of luminosityand concentration from our artificial cluster tests. sentially no chance of detecting clusters with Σ g,h > . − . Again, this ties in well with our detections.We have just one GC with Σ g,h ≈
27 mag arcsec − (PA-03), and only another four with Σ g,h ∼ >
26 mag arcsec − .As usual, we are assuming µ = 24 .
47 for M31, and a typicalforeground extinction of E ( B − V ) = 0 . V (i.e., Σ V,h ) would be ∼ .
35 mag arcsec − brighter. Notethat it is unnecessary to consider the effect of cluster struc-ture on these limits because the faint end of the function(Σ g,h >
26 mag arcsec − ) samples only diffuse ( r h >
10 pc)and relatively low luminosity ( M V > −
6) GCs.
In this section we explore the properties of the enlarged M31halo GC system, using the new clusters described above andexploiting our analysis of completeness which was not avail-able in any of our previous work (e.g., Hux11). As in Hux11we study the ensemble photometric properties of the M31GC system and compare them to those of the GC system ofthe Milky Way (MW). When taking photometry and struc-tural measurements from the present paper, we only includethose clusters which have a quality flag of either ‘A’ or ‘B’(see Table 6). We also exclude the two candidate GCs fromour analysis.In the analysis that follows, we supplement our cat-alogue of outer halo GCs with confirmed GCs from themost recent revision of the RBC (almost all of which areat R proj <
25 kpc). Since Hux11 there have been a num-ber of significant changes to the RBC – Hux11 used ver-sion 3.5, and this has now been updated to version 5.In particular, the latest version adds the photometry ofFan, de Grijs, & Zhou (2010) and Peacock et al. (2010), andthe spectroscopy of Caldwell et al. (2009).The sample of M31 GCs we take from the RBC is de-
Figure 14. Upper panel:
Detection completeness as a func-tion of cluster luminosity, collapsed into four concentration binsas marked.
Middle panel:
Detection completeness as a functionof cluster concentration, collapsed into four luminosity bins asmarked.
Lower panels:
Detection completeness across the en-tire sample of artificial clusters, as a function of luminosity (left)and g -band half-light surface brightness Σ g,h in magnitudes persquare arcsecond (right). fined in a manner comparable to that used in Hux11, andexploits a number of flags provided in the RBC that helpclassify the characteristics of the GCs. We only use thoseobjects for which the RBC flag ‘f’ is set to either 1 or 8(indicating confirmed compact and extended GCs respec-tively – the extended clusters all appear to be old metal-poor systems, so we treat them equally). We thus effectivelyexclude all objects in the RBC V5 that do not have imagingor spectroscopy confirming their status as GCs. M31 pos-sesses a population of younger clusters, predominantly setagainst the galactic disk, which we also exclude as thereare no comparable clusters in our MW sample. This wasachieved by ensuring the flag ‘yy’ is 0 – indicating clustersthat are “not young” according the data of Fusi Pecci et al.(2005), based on the ( B − V ) colour or the strength of theH β spectral index. Additional young clusters are excludedby removing objects for which the flag ‘ac’ is 1 or 2 (which c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs indicate an age estimate of less than 1 Gyr, or ∼ − E ( B − V ) values compared to version 3.5. In addition tothose of Fan, de Grijs, & Zhou (2010), which were availablefor Hux11, new values have been derived by Caldwell et al.(2011) from their spectroscopy. If any particular GC isin both Fan, de Grijs, & Zhou (2010) and Caldwell et al.(2011), we take the mean of the two E ( B − V ) values givenfor that cluster. If a GC within 25 kpc of the centre of M31has no colour excess from either of these sources, we apply anaverage value of E ( B − V ) derived from the medians of bothsamples. However, for those GCs with R proj >
25 kpc, we usethe E ( B − V ) values derived from the Schlegel et al. (1998)dust maps, as additional internal M31 reddening towardsthese objects is probably negligible. Note that we employthe updated reddening values from Schlafly et al. (2011) aslisted in Table 6.Data on the MW GC system have also been updatedsince Hux11 was written. We now use the current version(dated December 2010) of the McMaster catalogue (seeHarris 1996) , although there are no major changes sincethe previous catalogue. In the analysis that follows we fo-cus on the photometric properties of the GCs, derived fromtheir observed magnitudes and colours. With this in mind,we exclude from our plots the 28 (of 157 total) MW GCslisted in the McMaster catalogue as having E ( B − V ) > . A λ is not constant at high extinction. These objects are,in any case, almost all rather poorly studied.Analysis of the spatial layout of the M31 GCs, and theirrelationship to stellar substructures in the halo, will be ad-dressed in detail in an accompanying paper in this series(Mackey et al. 2014, in prep). The M31 globular cluster luminosity function (GCLF) isshown in Figure 15. The median value of M V for M31 is − .
6, compared to the − . − .
3, the same as in Hux11 despite thevarious minor updates to the McMaster catalogue.In Hux11 we suggested that the then available data in-dicated a secondary peak in the M31 GCLF at M V ∼ − R proj <
25 kpc) and outer halo ( R proj >
25 kpc) Figure 15.
Histogram of M V , showing the distribution for allM31 GCs, taking the additions and updates from this paper intoaccount (black solid line), and the M31 GCs with a projectedgalactocentric radius R proj >
30 kpc (green solid line) comparedto the MW (red line). The solid black and dashed red verticallines indicate the median values for the M31 and MW GC sys-tems respectively. The solid red regions show the GCs associatedwith the Sagittarius dwarf galaxy. The completeness limits forour PAndAS GC search are also shown (blue vertical lines).
GCs were considered separately, suggesting that this mightindeed be a real feature.The new data reveal a more complex situation. Withthe updated RBC, the second peak for the full M31 sample(black histogram) no longer appears. This is primarily dueto the reclassification of many of the Kim et al. (2007) GCsas stellar contaminants (Peacock et al. 2010), which reducesthe number of confirmed GCs in this magnitude range. How-ever, if we consider only the outer halo clusters (in this case R proj >
30 kpc), shown in green in Figure 15, a bimodal-ity in the GCLF is very clear with peaks at ∼ − . ∼ − .
5. The fainter secondary peak sits between our 100%and 50% completeness limits. Hence it is possible that ad-ditional GCs exist around this luminosity, that we have notdetected. These would further increase the prominence ofthe feature, and the location of the peak may shift slightly(probably towards slightly fainter magnitudes).It is natural to ask about the typical nature of the GCsresiding in the secondary peak. Mackey et al. (2010a) arguethat a substantial fraction (perhaps up to ∼ R proj >
30 kpc have been accreted into theM31 halo along with their parent dwarf galaxies. Hence, thefainter peak in the GCLF, which is prominent only for theouter halo system, might well be primarily driven by thepresence of this type of GC.This scenario finds additional support if we consider theMW GCs that are believed to be associated with the Sagit- c (cid:13) , 1–26 Huxor et al.
Figure 16.
Plot of M V against log ( r h ) for M31 GCs in theouter halo, with a projected distance greater than 30 kpc. Themore luminous clusters ( M V < − .
5) are relatively compact,whilst the fainter clusters span a broad range of effective radii.Recall that size measurements for any GC with r h ∼ < −
10 pcare upper limits. The dashed line shows the location of our 50%completeness limit. tarius dwarf galaxy (Law & Majewski 2010). Although thereare only perhaps eight such GCs (Arp 2, NGC 6715, NGC5634, Terzan 7, Terzan 8, NGC 5053, Pal 12 and Whiting 1),five of these have luminosities fainter than M V = −
6. Moregenerally, Mackey & van den Bergh (2005) found a similarfainter peak in GCLF of the“young halo” GCs of the MW,which they argued (from indirect evidence) are most likelyaccreted objects.The M31 halo GCs near the secondary peak in theGCLF differ in other ways. A plot of M V against r h forthe M31 GCs beyond 30 kpc (Figure 16) shows that clusterswith a luminosity near the fainter peak of the GCLF span avery broad range of half-light radii, while clusters near themore luminous peak are primarily compact.Our GCLF for the outer M31 halo suggests that there isa substantial population of luminous GCs outside R proj = 30kpc. A plot of absolute magnitude against projected radius(Fig. 17) makes this more explicit – our new PAndAS GCsearch has yielded many more luminous clusters in the outerhalo compared to Hux11. Note that the MW GCs are plottedwith an “average projected distance”, via the relationship R proj = R gc × ( π/ M V > −
4) in the MW, which we would likely not see in
Figure 17.
Plot of M V against projected galactocentric radius R proj , with the completeness limits of our PAndAS search againshown (blue). In the case of the MW GCs the actual distance(R gc ) in this, and subsequent plots, is converted to an “averageprojected distance” via the relationship R proj = R gc × ( π/ the PAndAS data – if they were present – as at the M31distance they lie well below our 50% completeness limit.However, in the MW, these faint GCs are found at moder-ately large (projected) galactocentric radii, suggesting thatdeeper imaging in the future may indeed reveal such objectsin the halo of M31. The distribution of ( V − I ) colours (Figure 18) shows al-most no difference to that found by Hux11. The median( V − I ) values for GCs in M31 and the MW are almostindistinguishable at 0.95 and 0.93, respectively.When viewed as a function of galactocentric radius (Fig-ure 19), the results are again similar to those reported byHux11. In that paper, we found a flat colour-radius relationfor GCs in the outer halo. The new data are consistent withthis, exhibiting only a marginal slope of − . ± . R proj >
25 kpc, while Colucci, Bernstein, & Cohen (2012)find a nearly constant metallicity for GCs with R proj > c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs Figure 18.
Histogram of ( V − I ) . The vertical lines show themedian values for the full sample (solid) and the MW (red). Figure 19.
Plot of ( V − I ) against R proj . The black line showsa linear fit to the M31 GCs with R proj >
30 kpc.
It is noticeable that although the ( V − I ) colours ofthe full M31 and MW GC systems are almost identical, forthe outer halo the MW GCs typically appear slightly redderthan the bulk of the M31 GCs at a comparable distance– although as there are very few MW GCs at these largegalactocentric radii, it is difficult to draw firm conclusions. Figure 20.
Plot of log ( r h ) against R proj . There is an appar-ently continuous range of half-light sizes at large galactocentricradii. One result that differs significantly from that seen by Hux11concerns the distribution of half-light radii for outer haloM31 GCs. In Hux11 we suggested this may take a bimodalform, with one peak corresponding to the typical sizes oftraditional compact GCs ( r h ∼ − r h ∼ >
15 pc. Wang & Ma (2013) also reporteda size bimodality for M31 GCs at R proj >
40 kpc, but theyused a small sample of clusters from Hux08 which were alsoincluded in our Hux11 analysis. We can now address thisquestion definitively with our larger halo GC sample.For comparison purposes we also assemble a set of (in-ner) M31 GCs with recent size measurements in the RBC.The largest RBC sample comes from the compilation ofPeacock et al. (2009, 2010), which we supplement with mea-surements from Barmby et al. (2007) when only the latter isavailable. Note that Peacock et al. (2009) provided a care-ful demonstration that their GC size measurements showedexcellent consistency with those derived from
HST imagingby Barmby et al. (2007).We remind the reader of the reliability of our deter-mination of cluster effective radius as described in Section5. For clusters smaller than r h ∼ −
10 pc, the size mea-surements are significantly affected by the seeing profile (seeFigure 9) and we thus over-estimate their values, typicallyby ≈ − HST imaging by Tanvir et al.(2012). We take these in preference where available, to tryand minimise the effects of this issue.In Figure 20 we show r h for M31 and MW GCs againstprojected distance from the centre of their host galaxy. Theapparent bimodality in cluster size at large galactocentricradii for M31 GCs, as noted in Hux11, has now vanished.There are many clusters with r h between 5 and 20 pc,in which range the size distribution appears rather evenlyspread at all galactocentric radii outside ∼
10 kpc. The orig- c (cid:13) , 1–26 Huxor et al. inal observation of bimodality may have been partly dueto the difficulty in measuring accurate GC sizes from ourINT/WFC data. We observe significant changes in the in-ferred sizes of some objects moving to the superior PAn-dAS data – for example, the clusters H7, H8 and H15 wentfrom our default value for “compact” GCs of 4.5 pc, usedin Hux11, to ∼
10 pc, while by contrast, some of the moreextended clusters have smaller sizes measured from the PAn-dAS data than those obtained by Hux11.The apparently even spread of cluster sizes larger than ∼ −
10 pc in the M31 halo strongly suggests that theextended clusters first identified by Huxor et al. (2005) aresimply objects selected from the upper tail of the GCsize distribution. This is consistent with their constituentstellar populations, which appear indistinguishable fromthose observed in typical metal-poor compact GCs (see e.g.,Mackey et al. 2006, 2007). It is also noticeable from Figure20 that the largest clusters observed in the remote MW haloare comparable to the sizes of many of the more extendedclusters in M31 – that is, there do appear to be a few counter-parts of the M31 extended clusters seen in the MW halo. Thelargest M31 clusters have greater r h than any GCs found inthe MW halo, but this is perhaps not surprising given themuch more numerous M31 halo GC population.Figure 21 shows a histogram of r h for M31 and MWGCs. Those for the full systems appear to share a very sim-ilar shape. However it is notable that the distribution of r h for M31 GCs with a galactocentric distance >
30 kpcis quite unlike that for the full M31 sample. Even takinginto account the tendency for our PAndAS measurementsto over-estimate the sizes of GCs with r h ∼ < −
10 pc,the distribution of half-light radii for clusters more than30 kpc from M31 would still be considerably flatter thanthat of the full sample. That is, the ratio of the number ofGCs with r h above 8 −
10 pc to those with r h below thislevel is substantially greater for M31 GCs with R proj > In this paper we present the final catalogue of M31 halo GCsfrom the PAndAS survey. Of these, 57 were identified by ourusual method of visually searching the new image data, andone further cluster was found by a code searching for faintdwarf galaxies. Our catalogue represents the first detailedand uniform census of GCs across nearly the full extent ofthe M31 halo. We find numerous clusters with very largeprojected galactocentric radii ( R proj ∼ >
100 kpc), reflectingthe huge spatial extent of the M31 GC system.We located a few additional GCs by revisiting outerhalo candidates listed in the RBC. We found that threesuch candidates are indeed GCs, while one is a H ii regionwith a possible embedded young cluster; and we also lo- Figure 21.
Histogram (logarithmic in N ) of r h . For large galac-tocentric radii (blue line), the distribution of half-light radii isconsiderably flatter than for the full M31 sample. cated one further new discovery that serendipitously fallsnear a star that was the source of the RBC entry. In addi-tion, we found that three “definite” outer halo GCs listedin the RBC are not clusters after all. Finally, we confirmthat ten of the 17 “high-confidence” SDSS clusters listed bydi Tullio Zinn & Zinn (2013) are indeed GCs, based on ourhigher-quality PAndAS imaging. However, only one of their42 “candidate” objects that we were able to examine wasfound to be a cluster.Experiments with artificial clusters suggest that our GCsurvey is complete down to a cluster luminosity of M V = − .
0, and has 50% completeness limit at roughly M V ≈− . ∼ − ≈
150 kpc of M31), due tosmall gaps in the survey coverage. We cannot rule out thatthere may also be many very faint clusters with M V ∼ > − R proj = 25kpc. The results of this process confirm most of the findingsfrom Hux11 with a much larger sample. The bimodality ofthe luminosity function constructed using M31 halo GCswith R proj >
30 kpc is perhaps the most notable feature.This bimodality is not seen in the LF constructed using morecentral clusters, and we suggest it may be a consequence ofthe dwarf galaxy accretion history of the outer M31 halo.The colours of the halo GCs show only a marginally sig-nificant shallow gradient with projected radius, while thedistribution of half-light radii for the M31 halo GCs revealsan apparently continuous spread of cluster sizes, rather thanthe bimodality suggested by previous studies that used muchsmaller samples and shallower imaging.Many of the new GCs described here have already been c (cid:13) , 1–26 he final PAndAS catalogue of M31 outer halo GCs followed up by the PAndAS collaboration. For example, alarge fraction of these objects is included in the studies ofVeljanoski et al. (2013a) and the companion paper to thepresent work by Veljanoski et al. (2014), where radial veloc-ities have been used to explore the kinematics of the M31outer halo GC system.Individual clusters have also proved of interest. InMackey et al. (2013b) we investigated two of the new PAn-dAS GCs (PA-7 and PA-8), which are almost certainly as-sociated with a prominent halo substructure known as theSouth-West Cloud (see Lewis et al. 2013; Bate et al. 2014).These objects appear to be at least 2 Gyr younger thanthe oldest MW GCs, and thus fit with the trends identi-fied by Perina et al. (2012), and show strong similarities tothe supposedly-accreted “young halo” clusters in the MW(Mackey & van den Bergh 2005).Our new clusters also provide a substantial number ofGCs which exhibit properties unlike those studied in theMW. Examples include the few very most extended clusters,and the luminous, compact clusters found in the far haloof M31. Some of the new GCs may be of major interest.For example, PA-48 has a structure and ellipticity that maybe more akin to a very faint dwarf galaxy than a typicalglobular cluster (see Mackey et al. 2013a). HST imaging reaches to below the horizontal branch atthe distance of M31 in a just a couple of orbits – althoughit is a challenge to go much deeper. Brown et al. (2004) re-quired a total of 3.5 days of exposure time to reach to 1.5mag below the old main sequence turn-off of the M31 glob-ular cluster SKHB 312. However, this situation will changewith the launch of
JWST , which should be able to reach themain sequence turn-off for M31 GCs with manageable expo-sure times, allowing us to investigate the GC system of M31in a manner comparable to our current understanding of theGalactic GC system. With low contaminating backgrounds,the GCs presented here will be ideal targets for such studies.
ACKNOWLEDGMENTS
We would like to thank the referee, Flavio Fusi Pecci, fora detailed and constructive report that helped improve thispaper. We also appreciate the careful reading by LucianaFederici and Silvia Galleti.ADM is grateful for support by an Australian ResearchFellowship (Grant DP1093431) from the Australian Re-search Council. AMNF, APH, and ADM acknowledge sup-port by a Marie Curie Excellence Grant from the EuropeanCommission under contract MCEXT-CT-2005-025869, dur-ing which this work was initiated.This work is based on observations obtained withMegaPrime/MegaCam, a joint project of CFHT andCEA/DAPNIA, at the Canada-France-Hawaii Telescope(CFHT) which is operated by the National Research Coun-cil (NRC) of Canada, the Institute National des Sciences del’Univers of the Centre National de la Recherche Scientifiqueof France, and the University of Hawaii.
REFERENCES
Alves-Brito A., Forbes D. A., Mendel J. T., Hau G. K. T.,Murphy M. T., 2009, MNRAS, 395, L34Barmby P., Huchra J. P., Brodie J. P., Forbes D. A.,Schroder L. L., Grillmair C. J., 2000, AJ, 119, 727Barmby P., Huchra J. P., 2001, AJ, 122, 2458Barmby P., McLaughlin D. E., Harris W. E., HarrisG. L. H., Forbes D. A., 2007, AJ, 133, 2764Bate N.F., et al., 2014, MNRAS, 437, 3362Battistini P., Bonoli F., Braccesi A., Federici L., Fusi PecciF., Marano B., Borngen, F., 1987, A&AS, 64, 447Bertin E., 2009, MmSAI, 80, 422Brown T. M., Ferguson H. C., Smith E., Kimble R. A.,Sweigart A. V., Renzini A., Rich R. M., VandenBergD. A., 2004, ApJ, 613, L125Caldwell N., Harding P., Morrison H., Rose J. A., SchiavonR., Kriessler J., 2009, AJ, 137, 94Caldwell N., Schiavon R., Morrison H., Rose J. A., HardingP., 2011, AJ, 141, 61Chies-Santos A. L., Larsen S. S., Kuntschner H., Anders P.,Wehner E. M., Strader J., Brodie J. P., Santos J. F. C.,2011, A&A, 525, A20Cockcroft R., et al., 2011, ApJ, 730, 112Colucci J., Bernstein R. A., Cohen J., 2012, Proceedingsof the XII International Symposium on Nuclei in the Cos-mos,Conn A. R., et al., 2012, ApJ, 758, 11Crampton D., Cowley A. P., Schade D., Chayer P., 1985,ApJ, 288, 494Da Costa G. S., Grebel E. K., Jerjen H., Rejkuba M., Sha-rina M. E., 2009, AJ, 137, 4361Deveikis V., Narbutis D., Stonkut˙e R., Bridˇzius A., Van-seviˇcius V., 2008, BaltA, 17, 351di Tullio Zinn G., Zinn R., 2013, AJ, 145, 50Dotter A., et al., 2010, ApJ, 708, 698Elson R. A., Walterbos R. A. M., 1988, ApJ, 333, 594Fan Z., Ma J., de Grijs R., Zhou X., 2008, MNRAS, 385,1973Fan Z., de Grijs R., Zhou X., 2010, ApJ, 725, 200Fan Z., Huang Y.-F., Li J.-Z., Zhou X., Ma J., Wu H.,Zhang T.-M., Zhao Y.-H., 2011, RAA, 11, 1298Federici L., Cacciari C., Bellazzini M., Fusi Pecci F., GalletiS., Perina S., 2012, A&A, 544, A155Forbes D. A., Spitler L. R., Strader J., Romanowsky A. J.,Brodie J. P., Foster C., 2011, MNRAS, 413, 2943Forte J. C., Vega E. I., Faifer F., 2012, MNRAS, 421, 635Fusi Pecci F., Bellazzini M., Buzzoni A., De Simone E.,Federici L., Galleti S., 2005, AJ, 130, 554Galleti S., Federici L., Bellazzini M., Fusi Pecci F., MacrinaS., 2004, A&A, 416, 917Galleti S., Bellazzini M., Buzzoni A., Federici L., Fusi PecciF., 2009, A&A, 508, 1285Georgiev I. Y., Goudfrooij P., Puzia T. H., 2012, MNRAS,420, 1317Harris W.E., 1996, AJ, 112, 1487Huchra J. P., Brodie J. P., Kent S. M., 1991, ApJ, 370, 495Huxor A.P., Tanvir N.R., Irwin M.J., Ibata R., Collett J.L.,Ferguson A.M.N., Bridges T., Lewis G.F., 2005, MNRAS,360, 1007Huxor A.P., Tanvir N.R., Ferguson A.M.N., Irwin M.J.,Ibata R.A., Bridges T., Lewis G.F., 2008, MNRAS, 385, c (cid:13) , 1–26 Huxor et al. c (cid:13)000