Globular Cluster Kinematics in Twelve Early-type Galaxies

The SLUGGS Survey: Kinematics for over 2500 Globular Clusters in Twelve Early-type Galaxies

Vincenzo Pota, Duncan A. Forbes, Aaron J. Romanowsky, Jean P. Brodie,

Lee R. Spitler, Jay Strader, Caroline Foster, Jacob A. Arnold, Andrew Benson,

Christina Blom, Jonathan R. Hargis, Katherine L. Rhode, Christopher Usher

Centre for Astrophysics & Supercomputing, Swinburne University, Hawthorn VIC 3122, Australia
Department of Physics and Astronomy, San José State University, San Jose, CA 95192, USA
University of California Observatories, 1156 High Street, Santa Cruz, CA 95064, USA
Department of Physics and Astronomy, Michigan State University, East Lansing, Michigan 48824, USA
European Southern Observatory, Alonso de Cordova 3107, Vitacura, Santiago, Chile
The Observatories of the Carnegie Institution for Science, 813 Santa Barbara Street, Pasadena, CA 91101, USA
Indiana University, 727 East 3rd Street, Swain West 319, Blomington, IN 47405, USA
Email: vpota@astro.swin.edu.au, dforbes@astro.swin.edu.au
Released 2012 Xxxxx XX
Abstract

We present a spectro-photometric survey of extragalactic globular clusters (GCs) around twelve early-type galaxies, nine of which have not been published previously. Combining space–based and multi–colour wide field ground–based imaging, with spectra from the Keck DEIMOS instrument, we obtain an average of GC radial velocities per galaxy, with a high velocity precision of per GC. After studying the photometric properties of the GC systems, such as their spatial and colour distributions, we focus on the kinematics of metal-poor (blue) and metal-rich (red) GC subpopulations to an average distance of  effective radii from the galaxy centre.

Our results show that for some systems the bimodality in GC colour is also present in GC kinematics. The kinematics of the red GC subpopulations are strongly coupled with the host galaxy stellar kinematics. The blue GC subpopulations are more dominated by random motions, especially in the outer regions, and decoupled from the red GCs. Peculiar GC kinematic profiles are seen in some galaxies: the blue GCs in NGC  rotate along the galaxy minor axis, whereas the GC system of the lenticular galaxy NGC  appears to be strongly rotation supported in the outer region.

We supplement our galaxy sample with data from the literature and carry out a number of tests to study the kinematic differences between the two GC subpopulations. We confirm that the GC kinematics are coupled with the host galaxy properties and find that the velocity kurtosis and the slope of their velocity dispersion profiles is different between the two GC subpopulations in more massive galaxies.

keywords:
galaxies:star clusters – galaxies:evolution – galaxies: kinematics and dynamics – GCs
pagerange: The SLUGGS Survey: Kinematics for over 2500 Globular Clusters in Twelve Early-type GalaxiesLABEL:lastpagepubyear: 2012

1 Introduction

The reconstruction of the evolutionary history of galaxies in the local universe requires a comprehensive knowledge of their chemo-dynamic properties at all scales. Along these lines, the ATLAS project (Cappellari et al., 2011) has carried out a rich survey of galaxies employing an integral field unit (IFU) technique to map out the chemo-dynamics of galaxies within effective radius (). However, such a radius encloses % of the total galaxy mass (baryonic dark) and hence it may not be representative of the overall galaxy (e.g., Proctor et al., 2009). In contrast, beyond this radius (), the halo of the galaxy contains precious dynamical signatures of early merging or early collapse events that might have eventually built up the galaxy (Hopkins et al., 2009; Hopkins et al., 2010; Oser et al., 2010; Hoffman et al., 2010). Some recent examples of the dynamical studies of the outskirts of galaxies beyond the Local Group has mainly involved a few deep long-slit observations (e.g., Coccato et al. 2010), the pioneering technique of Proctor et al. (2009) and Norris et al. (2008) that probed the two-dimensional stellar kinematics to , and extragalactic planetary nebulae (PNe, Coccato et al. 2009) that can probe the galaxy potential up to .

Often labeled the “fossil record” of galaxy formation, GCs have been used to study galaxy haloes. GC systems have been found in galaxies of all morphological type and they extend beyond the detectable galaxy diffuse light, probing galactocentric distances where the are only few other gravitational constrains (Rhode & Zepf, 2001, 2004; Dirsch et al., 2003; Tamura et al., 2006; Forbes et al., 2011). Moreover, their old ages ( Gyr, Kissler-Patig et al., 1998; Cohen et al., 1998; Beasley et al., 2000; Forbes et al., 2001; Schroder et al., 2002; Brodie & Larsen, 2002; Beasley et al., 2004; Strader et al., 2005) suggest they have survived violent merging events, preserving the chemo-dynamical record of their parent galaxies.

Moreover, the well-studied dichotomy observed in the colour distribution of most GC systems (Zepf & Ashman, 1993; Ostrov et al., 1993; Whitmore et al., 1995; Gebhardt & Kissler-Patig, 1999; Larsen et al., 2001; Kundu & Whitmore, 2001; Peng et al., 2006; Sinnott et al., 2010) is thought to stem from different formation mechanisms that shaped the underlying host galaxy (Ashman & Zepf, 1992; Forbes et al., 1997; Côté et al., 1998). Although the reality of the GC metallicity bimodality has been recently called into question as the result of a strongly non-linear color-metallicity relation (Yoon et al., 2006; Yoon et al., 2011; Blakeslee et al., 2010), observations have shown that physical dissimilarities exist among these two subpopulations (Côté, 1999; Brodie & Strader, 2006; Peng et al., 2006; Chies-Santos et al., 2011; Forbes et al., 2011). For instance, the two GC subpopulations are found to have different physical sizes (Kundu & Whitmore, 2001; Jordán et al., 2005; Masters et al., 2010) and diverse spatial distributions around the host galaxy, with the metal-rich (red) GCs more centrally concentrated than the metal-poor (blue) GCs (Geisler et al., 1996; Ashman & Zepf, 1998; Brodie & Strader, 2006; Bassino et al., 2006; Faifer et al., 2011; Strader et al., 2011; Forbes et al., 2012).

Current spectroscopic studies of GC systems have shown that the kinematics (e.g. rotation directions and rotation amplitudes) of the two GC subpopulations are somewhat diverse. The kinematics of the red GCs is usually akin to that of the host galaxy stars (Schuberth et al., 2010; Strader et al., 2011), perhaps due to a similar formation history (Shapiro et al., 2010). Whereas, the velocity dispersion of the blue GCs is typically larger than that of the red GCs (e.g., Lee et al., 2008). Also intriguing is the fact that rotation has been detected for both the blue and the red GCs, regardless the mass or morphology of the host galaxy (Foster et al., 2011; Arnold et al., 2011).

Interpreting this variety of GC kinematics in the context of galaxy and GC formation has been limited due to the low number of galaxies with large GC radial velocity datasets. To date, this set includes only a dozen GC systems, most of which are nearby very massive ellipticals (see Lee et al. 2010, for a summary). On the other side, numerical simulations on this front have mainly focused on the origin of GC metallicity and colour bimodality (e.g., Weil & Pudritz, 2001; Kravtsov & Gnedin, 2005; Yoon et al., 2006; Yoon et al., 2011), rather than on GC kinematic properties (e.g., Bekki et al., 2005, 2008; Prieto & Gnedin, 2008).

This scenario has left open several questions regarding the kinematics of GC systems: does the colour bimodality also imply kinematic bimodality? If so, do the kinematical differences between the blue and red GC subpopulations found in the most massive ellipticals also hold for galaxies over the whole early-type sequence of the Hubble diagram? Furthermore, do the blue and red GC subpopulations rotate faster in the outer regions, as predicted in a formation in a disk-disk merging scenario (Bekki et al., 2005)? Can GC kinematics contribute to our understanding of the formation of lenticular galaxies (e.g., Barr et al., 2007)?

We have been carrying out a project named SLUGGS111http://sluggs.swin.edu.au/ to investigate the GC systems in external galaxies (Brodie et al. 2012, in preparation). SLUGGS is the SAGES Legacy Unifying Globulars and Galaxies Survey, where SAGES is the Study of the Astrophysics of Globular Clusters in Extragalactic Systems. This survey exploits the combination of Subaru/Suprime-Cam wide-field imaging with spectra from the Keck/DEIMOS multi-object spectrograph. The results released so far have shown that the wide-field imaging can give clues about assembly history of the host galaxy (Blom et al., 2011; Forbes et al., 2011). If combined with the high velocity resolution of DEIMOS, this dataset can unravel, at the same time, the dynamics and the metallicity of the field stars (Foster et al., 2009; Proctor et al., 2009) and of the GCs (Foster et al., 2011; Arnold et al., 2011) deep into the galaxy halo, as well as a giving compelling view of galaxy dynamics (Strader et al., 2011; Romanowsky et al., 2009; Romanowsky et al., 2011).

In this work, we aim to study the GC kinematics for an unprecedented large sample of early-type galaxies (ETGs) with high-quality data. We investigate the global kinematics of blue and red subpopulations to study how they, and the underlying galaxy itself, formed. We also supplement our galaxy sample with literature data, and we compare the properties of this large sample with the existing numerical predictions.

The plan of this paper is as follows. From Section 1 to Section 5 we describe the reduction and the analysis of both the photometric and the spectroscopic data. In Section 5, we briefly discuss the significant findings for each galaxy. In Section 6, we give an overview on the current state of the GC formation models in order to compare their predictions to our generic results discussed in Section 7. In Section 8 we supplement our galaxy sample with data from the literature. In Section 9 we analyse the enlarged GC system dataset (our data plus literature) and discuss the results in Section 10. The summary of the paper is given in Section 11.

2 The sample

In this paper we discuss a subset sample from our survey. This includes nine new galaxies, in addition to other three galaxies already published: NGC  (Foster et al., 2011), NGC  (Arnold et al., 2011) and NGC  (Strader et al., 2011). The analysis and the specific results for these three galaxies have been extensively discussed in the respective papers. Therefore, their overall results will be discussed together with the other nine starting from Section 7.

The physical characteristics of the twelve galaxies are listed in Table 1, with their optical images shown in Figure 1. This galaxy sample extends the study of extragalactic GC systems into a new regime, because it is representative of a wide range of luminosity, morphological type (from lenticulars to giant ellipticals) and environment (from field to clusters), with a velocity resolution three times better than typical previous studies. This improvement is shown in Figure 2 in which we compare the intrinsic properties of our dataset with previous GC studies (that have employed various instruments including VLT/FLAMES, VLT/FORS2, Keck/LRIS or Gemini/GMOS).

Galaxy ID Hubble (
Type [km s] [arcsec] [mag] [Mpc] [mag] [mag] [degree]
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
NGC  E
NGC  SA
NGC  E
NGC  E
NGC  E
NGC  E
NGC  E
NGC  E
NGC  S
NGC  S
NGC  E
NGC  E
Table 1: General properties of our galaxy sample. The galaxy name (1) and Hubble Type (2) are from the NED database. The galaxy systemic velocity (3) and ellipticity corrected effective radius (4) are from Cappellari et al. (2011), otherwise from NED and the RC3 catalogue (de Vaucouleurs, 1991) if not in Cappellari et al. (2011). The distance modulus (5) and the respective distance in Megaparsec (6) are from Tonry et al. (2001) with a correction as advocated by Mei et al. (2007), respectively. If the galaxy is in the ACS Virgo Survey, we use the distances from Mei et al. (2007). We assume that NGC  and NGC  lie at the same distance, computed as the average of the respective Tonry et al. (2001) distances. The  band absolute magnitude (7) is from 2MASS apparent magnitude at the distances given in column 6 and corrected for the foreground Galactic extinction given in column 8 (NED database). The photometric position angle (9) and axis ratio (10) are from 2MASS (Skrutskie et al., 2006). The last three galaxies have been analysed in separate papers (see text).
Figure 1: Digitalized Sky Survey (DSS) images of our galaxy sample. Red circles and black ellipses represent the spectroscopically confirmed GCs found in this work and the  schematic isophotes corrected for the galaxy ellipticity respectively. The black line on the bottom-right spans arcmin in length. North up and the East on the left. The elliptical galaxy south-west of NGC  is NGC  whose GC system is shown separately in this figure. The last three galaxies have been analysed in separate papers.

3 Photometric observations and data analysis

3.1 Subaru data

Figure 2: Accuracy of GC radial velocity measurements. Open circles and open boxes represent our GC data and literature data, respectively. Literature data will be discussed in Section 8. Top panel. The median velocity uncertainty per GC system are shown as a function of absolute magnitude in the K band. Symbol sizes are , where is the number of spectroscopically confirmed GCs per galaxy. Bottom panel. Log is shown as a function of the absolute magnitude in the K band, where symbol sizes are . Our data extend the range of galaxy luminosity probed with three times better velocity accuracy.

Multi-band photometric observations were carried out using the Subaru/Suprime-Cam instrument with a field of view of  arcmin and a pixel scale of arcsec (Miyazaki et al., 2002). The galaxy sample was observed in the period between 2005 and 2010 with a standard Sloan Digital Sky Survey (SDSS) filter set. NGC  and NGC  were both downloaded from the SMOKA archive (Baba et al., 2002) and were observed with a and filter set, respectively. For NGC  and NGC  we present existing Suprime-Cam photometry published in Spitler et al. (2012) (see also Romanowsky et al. 2009) and Blom et al. (2011), respectively. For these two galaxies both the imaging reduction and the catalogue extraction were performed with the methodology described in this paper. NGC  is the only galaxy for which no Subaru imaging is available, and therefore we use WIYN/Minimosaic imaging in filters presented in Hargis et al. (2011) and we refer to this paper for a description of the data reduction.

In Table 2 the imaging observations are summarized. The overall seeing conditions were mainly sub-arcsec. For NGC  ( band), NGC  ( band) and NGC  ( band) the data suffer from cloudy conditions and poor ( arcsec) seeing.

Subaru raw images were processed using the SDFRED data pipeline (Ouchi et al., 2004) that yields standard flat field corrected images for each of the three filters. Photometric point source catalogues were extracted using standard IRAF/Daophot aperture photometry routines. We summarise here the main steps of the data reduction and we refer to Section 3 of Blom et al. (2011) for a detailed description of the method.

We obtain a raw list of object positions by running IRAF/Daofind on galaxy subtracted images in order to optimise the finding algorithm. The extraction threshold was typically set between and times the background depending on the filter and on the seeing conditions. The galaxy light was modelled with IRAF/Ellipse set to allow the position angle and ellipticity to vary. Next, we perform aperture photometry using IRAF/Phot on the preselected objects for a certain number of circular apertures from up to pixels (equivalent to arcsec to arcsec for the Suprime-Cam pixel scale). The extraction radius was chosen in order to maximise the signal of the source and minimise the sky contribution. The extracted magnitude was corrected for the computed aperture correction using IRAF/Mkapfile. Photometric zeropoints were estimated by boot-strapping the Suprime-Cam photometry to the Sloan Digital Sky Survey (SDSS) DR7 photometric system (Abazajian et al., 2009) using the brightest objects in common between the two datasets (typically with ). If not in SDSS. the zeropoints were calibrated using the flux from standard stars observed over the same night. Finally, we use the reddening given in Table 1 and the conversion table of Schlegel et al. (1998) to derive the Galactic extinction correction in our photometric bands. Hereafter, all magnitudes and colours are extinction corrected.

3.2 HST data

We use Hubble Space Telescope (HST) archive images from the Advanced Camera for Surveys (ACS) and Wide Field Planetary Camera 2 (WFPC2) to improve the quality of the photometric selection in the central regions of our galaxies. For most of them, we exploit existing photometric GC catalogues and we refer to the following authors for a detailed description of the data reduction and analysis: Spitler et al. (2008) for NGC , Forbes et al. (2006) for NGC  and NGC , Forbes et al. (1996) for NGC , Chomiuk et al. (2008) for NGC  and Blom et al. (2011) for NGC .

We obtained, from the Hubble Legacy Archive, new HST/ACS imaging for NGC , NGC  and NGC , respectively. The ACS camera has a pixel scale of arcsec and a field of view of arcmin.

The NGC  imaging consists of one pointing in F475W ( Sloan ) and F850LP ( Sloan ) filters and it was observed as part of the HST project ID 10554. NGC  (ID 9353) was imaged in F435W, F555W, F814W filters, equivalent to a configuration, respectively. Finally, the NGC  (ID 10835) data consists of four pointings in F475W and F850LP filters that probe the galaxy up to arcmin from the centre (Usher et al. 2012, in preparation).

The HST imaging was reduced and analysed using a custom built pipeline to find point-like sources and measure their magnitudes and half light radii. For details on the methods used by the pipeline including point spread function determination, we refer to Strader et al. (2006) and to Spitler et al. (2006). The extracted magnitudes and sizes for GCs in NGC  and NGC  were compared with those published by Chies-Santos et al. (2011) for objects in the NGC  pointing and in the two NGC  pointings. Both magnitudes and sizes show good agreement without any evidence of statistically significant offset from the published data.

Figure 3: Colour-colour diagram of objects around NGC . Grey and black points represent all the sources detected in the Suprime-Cam field-of view and all the photometrically selected GCs with and , respectively. Spectroscopically confirmed sources are shown as red points (GCs), green boxes (Galactic stars) and orange triangles (galaxies) respectively.

3.3 Photometric GC selection

Once the point-source catalogues have been constructed, they are photometrically selected to avoid contamination, such as unresolved galaxies and Galactic stars. As far as our new Subaru data are concerned, this process involves the following steps: 1. selection in a colour-colour space; 2. selection in a colour-magnitude space; 3. a visual check. The GC selection in the ground-based imaging for NGC  was carried out in Hargis et al. (2011) and we refer to this paper for a detailed description of their selection method. The steps listed above are subjected to variations in the procedures for the DEIMOS mask design.

Galaxy ID Obs date Filters Exp. time Seeing
[sec] [arcsec]
NGC  2008 Nov.
NGC  2008 Nov.
NGC  2008 Nov.
NGC  2005 Mar.
NGC  2008 Nov.
NGC  2002 Feb.
NGC  2008 Apr.
NGC  2009 Apr.
NGC  2009 Oct.
Table 2: Summary of the ground-based imaging observations. The galaxy name, observation date, filters employed, together with the respective exposure time and seeing are listed. All the observations were performed using Subaru/Suprime-Cam, except for NGC  observed with WIYN/MiniMo (Hargis et al., 2011).

GCs are known to populate a specific area of colour-colour diagrams (e.g., Rhode & Zepf, 2001; Faifer et al., 2011; Chies-Santos et al., 2011). With a filter set, this is enclosed within , , where these boundaries run diagonally to the colour axes, as shown in Figure 3 for the galaxy NGC . To take into account the dependence of the colour boundaries on the quality of the data, we flag as GC candidates all the objects deviating by less than 2 from these boundaries (Spitler et al., 2008).

Next, we apply a cut on the band magnitude ( band for NGC ). Given the ongoing debate regarding the uncertain separation between GCs and ultra compact dwarfs (UCDs, Mieske et al., 2006; Brodie et al., 2011), we decided to set the upper brightness magnitude at (), one magnitude brighter than the integrated magnitude of Cen, the brightest GC in the Milky Way. Nevertheless, in some cases we relax this criterion in order to include spectroscopically confirmed GCs that have magnitudes brighter than the set threshold. The separation between NGC  and NGC  objects will be discussed in Section 5.

Next, we calculate the radius at which the number of GC candidates per unit area flattens out (see §3.5 for the method), that is an estimate of the radius at which the contribution of the contaminants (Glactic stars and high redshift galaxies) becomes dominant. Therefore, we count out of the GC selection all the objects outside this background radius. We have tested that the effect of the contaminants on the GC colour distribution is minimal mag and that it does not affect considerably the GC colour bimodality (see also, Arnold et al., 2011).

Finally, we perform a visual check to make sure that no outliers, such as extended sources or image artefacts, contaminate the final GC catalogue. Most of the outliers turned out to be close to the galaxy (within arcmin) where the galaxy light contamination and the crowded field makes the Subaru photometry unreliable. Within this radius, the contribution of the HST imaging becomes crucial.

As far as the HST/ACS imaging is concerned, it is worth noting that the diffraction limited quality of HST imaging has the advantage of making extragalactic GCs partially resolved for all of our galaxies. Therefore, the GC selection in our space-based imaging is also based on a size selection in addition to magnitude and colour-colour criteria (if available). In NGC  and NGC , for which only and imaging are available, we flag as GC candidates all the objects with colour and sizes pc. Such a choice is motivated by the clear drop off in the density of the objects outside the adopted colour cut, as observed in Blom et al. (2011). For NGC , we adopt the same size-cut as above and select objects within and and an upper magnitude of . In Figure 4, the colour-magnitude diagrams of the GC candidates and of the spectroscopically confirmed GCs are shown.

Figure 4: GC system colour-magnitude diagrams. Photometrically selected GCs and spectroscopically confirmed GCs are shown as small and large black points respectively. Upper and bottom dashed lines represent the Cen magnitude () and the GC turnover magnitude ( mag) at the distances given in Table 1.

3.4 GC colour bimodality

To probe bimodality, we use a Kaye’s Mixture Model algorithm (KMM, Ashman et al. 1994) that returns colour peaks, variance and number of objects in the detected subpopulations. KMM was run on all the distributions of the Subaru GC candidates brighter than the turnover magnitude . In this paper we adopt , derived using the transformation from Faifer et al. (2011) to calibrate the band TOM given by Kundu & Whitmore (2001) into our photometric system.

For NGC , NGC  and NGC , we study the , and colour distributions respectively. We perform a bisector fit (Feigelson & Babu, 1992) to the bright GCs in common between the HST and Subaru images, in order to convert HST magnitudes of the spectroscopically confirmed GCs into the respective Subaru photometric system. For NGC , given the wide spatial coverage and better quality of its HST data, we transform Subaru magnitudes of the confirmed GCs into the HST photometric system. Results of the KMM analysis are presented in Table 3. We find that eight out of nine galaxies show significant GC colour bimodality. GCs are divided into blue and red according to the local minimum of the best fit Gaussians in Figure 5.

Galaxy ID p-value colour
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
Table 3: KMM results. For each galaxy, and represent the mean of the blue and red peak respectively, whereas and are the Gaussian for each peak. The last column gives the p-value, that is the confidence level with which the hypothesis of an unimodal colour distribution can be rejected (larger confidence for smaller p-values). The last column gives the colour distribution used in the analysis. NGC  is the only galaxy without significant bimodality detected.
Figure 5: Colour histograms of our galaxy sample. GC candidates and spectroscopically confirmed GCs are shown as empty and grey filled histograms respectively. For NGC , we only show GCs with . Solid lines show the sum of the KMM fits of the two subpopulations. No significant bimodality was detected in NGC . The dearth of red GCs in NGC  is due to the fact that we have excluded all the GC candidates within arcmin from the centre. The magnitude of these objects turned out to be corrupted by the underlying diffuse stellar light.

3.5 GC spatial distribution

We construct projected surface density profiles for the GC subpopulations of our galaxy sample. GC candidates brighter than the turnover magnitude were binned in circular annuli and corrected for areal incompleteness. The Subaru dataset was supplemented with HST photometry, in order to probe the GC surface density in the inner regions. An important caveat to bear in mind is that we do not correct the Subaru dataset for magnitude incompleteness. This effect becomes important especially for objects fainter than in the innermost regions ( arcmin) where the detection of sources is inhibited by the surface brightness contribution of the host galaxy (Faifer et al., 2011). Conversely, the completeness of the HST photometry in band is typically above per cent at the turnover magnitude even in the innermost regions, respectively (Larsen et al., 2001; Jordán et al., 2007). Therefore, we select all the GCs brighter than the turnover magnitude (see Figure 4) and we use circular radial bins in common between HST and Subaru to correct and adjust the Subaru GC surface density to that of the HST data. Although the matching between HST and Subaru data points is arbitrary, this approach preserves the relative slopes of the spatial distribution of the two subpopulations and allows us to verify whether or not blue and red GC subpopulations show different spatial distributions around the host galaxy as found from previous authors (e.g., Bassino et al., 2006).

Figure 6: Surface density profiles of the GCs brighter than the turnover magnitude. Blue and red GC subpopulations are shown as blue and red filled points (if from Subaru imaging) and blue and red open circles (if from HST imaging) respectively. Sérsic fits or alternatively power-law fits to blue and red GC subpopulations are shown as blue and red lines respectively. For NGC , the surface density and the fit to all GCs are shown as black points and black line respectively. Both the data points and the fits are background subtracted according to eq. 1. Also shown as dashed lines are the scaled and arbitrarily offset stellar surface brightness profiles obtained from Subaru imaging as described in §3.2 . For completeness, we also show the surface density profiles from Blom et al. (2011) for NGC  and from Hargis et al. (2011) for NGC , respectively.

Next, we fit the resulting (HST Subaru) GC density profile with a Sérsic law (Sérsic, 1963) similar to that commonly carried out for galaxy surface brightness profiles, but in practice we fit a variation of the Sérsic profile from Graham et al. (2005) in order to quantify the background level:

(1)

where , is the Sérsic index,  is the effective radius of the GC system, is the surface density at that radius and is the background contamination level. In case a Sérsic fit was not feasible due to small numbers we fit a power-law, i.e. that has been also used for similar analyses (Spitler et al., 2008). In Figure 6 the background subtracted GC surface density profiles for our galaxies are shown. As found in other galaxies, the red GCs are more centrally concentrated than the blue GCs (e.g., Geisler et al., 1996; Bassino et al., 2006; Faifer et al., 2011; Strader et al., 2011).

We also compare the GC surface density to the galaxy surface brightness for each galaxy, obtained from the Subaru images using IRAF/Ellipse as described in §3.2. The surface brightness profiles were corrected for the local galaxy ellipticity and then shifted by an arbitrary constant for comparison purposes (different for each galaxy). Figure 6 shows that, qualitatively, the slope of the red GC subpopulation agrees with that of the surface brightness of the host galaxy.

4 Spectroscopic observations and data analysis

Spectroscopic observations were performed with the DEep Imaging Multi-Object Spectrograph (DEIMOS, Faber et al. 2003) mounted on the 10 m Keck-II telescope. Galaxies were targeted in the period between 2006 and 2011, employing a different number of masks for each galaxy depending on the richness of the GC system. Objects selected for spectroscopic follower were chosen according to their likelihood of being GCs, i.e. giving priority to the objects selected with methods as Section 3.3. The spectroscopic observations are summarized in Table 4.

The large collecting area of the 10-meter Keck primary mirror combined with the arcmin of DEIMOS is the ideal combination to investigate the outskirts of galaxies where GCs are expected to be one of the best tracers of the total galaxy potential. For all the galaxies, DEIMOS was set up with the 1200 l/m grating centred on Å together with arcsec wide slits, allowing coverage of the region between Å with a resolution of  Å. Raw spectra were reduced using the DEIMOS/spec2d reduction pipeline provided online, that produces calibrated and sky subtracted spectra for each slit (Newman et al., 2012; Cooper et al., 2012).

We estimate the radial velocity of the GCs by measuring the Doppler shift of Calcium Triplet (CaT) absorption lines that characterises the infrared part of their spectra at 8498 Å, 8542 Å, 8662 Å, respectively. We measure radial velocities with IRAF/Fxcor that performs a cross-correlation between the Fourier transformed science spectrum and template Galactic star Fourier transformed spectra. The template spectra were observed with the same DEIMOS setup used for scientific spectra and they cover a wide range of spectral type, luminosity and metallicity (from F to M type). Fxcor was configured to have all the science and template spectra in the same wavelength range from to Å with the same DEIMOS spectral resolution. The radial velocity for each object was estimated as the mean of the radial velocity resulting from the correlation with each template star. The respective errors were evaluated by adding in quadrature the default error given by fxcor as described in Tonry & Davis (1979) to the standard deviation among the stellar templates, which is an estimate of the systematics.

Galaxy ID Masks Exp. time N N N
[hr]
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC  0
NGC 
NGC 
Table 4: Summary of the spectroscopic observations for our twelve galaxies. The table lists the galaxy name and the total number of the DEIMOS masks used. To take into account the different seeing conditions over different nights, we show the effective exposure time weighted by the mean seeing conditions during the observation time as done in Coccato et al. (2009). Also shown are the total number of spectroscopically confirmed GCs (including marginal GCs), Galactic stars and background galaxies respectively. The datasets for NGC  and NGC  also include GCs from external datasets as described in Arnold et al. (2011) and Strader et al. (2011) respectively.

4.1 Kinematic selection criteria

Figure 7: GC radial velocity distributions with galactocentric radius. Confirmed GCs, marginal GCs and Galactic stars are shown as filled points, open points and green boxes respectively. GCs are colour coded according to their subpopulation (blue or red) membership. Galactocentric radii have been translated into effective radii on the top axis. Dotted curves and dot-dashed lines are the clipping envelope as defined by the friend-less algorithm (see text) and the galaxy systemic velocities from Table 1, respectively. Also shown as thick dotted lines are the systemic velocities of the galaxies that might contaminate the GC system of the target galaxy. GCs have a mean velocity similar to the galaxy systemic velocity.

Once the final radial velocities of the GC candidates are obtained, we redshift-correct the spectra and perform a visual check to verify that the CaT lines are real and that they lie at the expected rest wavelength. Our selection criteria require the presence of at least two “visible” CaT lines (typically the brightest  Å and  Å lines) and of the H absorption line at  Å, if probed in the redshift corrected spectra. The visual analysis is performed by at least two members of the team and it produces a final spectroscopic consensus catalogue. Spectra that show a radial velocity consistent with a GC, but for which it was not possible to reach a consensus, were flagged as “marginal” and are not included in the kinematic analysis. These objects have usually low signal-to-noise spectra or suffer from bad sky-line subtraction making the line identification subjective.

Besides background galaxies that can be usually spotted by the emission lines in their spectra, the main outliers of our spectroscopic selection are Galactic stars because they show a stellar spectrum with radial velocity  km s. In most cases GCs and Galactic stars are well-separated in velocity, but for galaxies with Galactic stars might introduce a low velocity tail in the observed GC candidate velocity distribution. We decided to use a friendless algorithm introduced in Merrett et al. (2003) that flags objects deviating by more than from the velocity distribution of their nearest neighbours. We use and (depending on the galaxy) to exclude both stars and possible outliers that lie outside the envelope from the kinematic analysis.

In Figure 7 we show the distribution of all the spectroscopically confirmed GCs and Galactic stars in a radius-velocity phase space. In this plot and hereafter, galactocentric distances are expressed as a equivalent radius, that is defined as:

(2)

where is the axis ratio defined as the ratio of minor over the major axis of the galaxy (see Table 1), and are Cartesian coordinates of an object in the galaxy rest frame, with the origin at the galaxy centre, and with and aligned along the photometric major axis and minor axis respectively. The final spectro-photometric catalogues of our galaxy sample are available on-line and they include spectroscopically confirmed GCs, “marginal” GCs, Galactic stars and galaxies.

Figure 8: Comparison of our GC radial velocities with previous studies. For each galaxy the root-mean-square of the velocity difference between the two samples is also shown. Literature datasets come from Brodie et al. (2005) (black points) and Larsen et al. (2003) (open circles) for NGC ; Chomiuk et al. (2008) for NGC ; Puzia et al. (2004) for NGC . The one-to-one line is shown as a dashed line.

4.2 Repeated GC measurements

We searched through the literature for GCs observed in our surveyed galaxies, finding existing datasets for three galaxies. In summary, we have re-observed: 1. 6 GCs of the 26 GCs confirmed by Puzia et al. (2004) in NGC  using VLT/FORS2; 2. 7 GCs of the 13 GCs found by Chomiuk et al. (2008) in NGC  using Keck/LRIS ; 3. 24 GCs of the 33 GCs resulting from the combination of the Brodie et al. (2005) and Larsen et al. (2003) catalogues of NGC , both observed with Keck/LRIS.

Figure 9: Repeated GC radial velocity measurements. In the top panel we show the radial velocity of a GC against the radial velocity of the same GC observed in a different night . Data points are colour coded according their host galaxy membership listed on the right. The dotted line is a one-to-one line. In the bottom panel we show the difference as a function of the band magnitude of the object. The dotted line shows the constant to guide the eye. Radial velocities from different nights are in good agreement with each other, with an overall root-mean-square (rms) of . No significant trend is seen as a function of GC magnitude.

In Figure 8 we compare our DEIMOS dataset with literature studies. There is a rough agreement between external datasets and our DEIMOS radial velocities, although the root-mean-square of the velocity difference between the two datasets is larger than what was found with DEIMOS repeated measurements.

We supplement our GC catalogues with external GC radial velocities corrected for the mean offset between the literature and our DEIMOS measurements. This offset is for both NGC  and NGC . In NGC , the offset between DEIMOS radial velocities and Brodie et al. (2005) is effectively zero, and we only correct the dataset of Larsen et al. (2003) by .

We test the reliability of our spectroscopic measurements by comparing the radial velocity of two or three GCs observed over different nights. Overall, we have collected 118 repeated GC radial velocities and these are shown in Figure 9. One-third of the total repeated sample is from a NGC  mask that was observed twice with a one night offset. We find that repeated GC measurements are in good agreement in all the galaxies. The root-mean-square (rms) of the velocity difference between two observations is .

4.3 Kinematic analysis

We study the kinematic properties of our GC systems using a maximum likelihood approach. We summarize here the main points of this analysis and refer to Foster et al. (2011) for the details. We divide the data in radial bins and then we fit the amplitude of the rotation (), the velocity dispersion () and the kinematic position angle () simultaneously. The bin size varies from galaxy to galaxy (wider for larger datasets) and it was set to have roughly the same number of objects per bin. For the radial bin we minimise the function:

(3)

where:

(4)

In eq. 4, , and are the position angle, recession velocity and uncertainty on the recession velocity for the GC, respectively. The axis ratio of GC systems is challenging to derive directly because of low number statistics, but it has been constrained in galaxies with large photometric datasets (Kissler-Patig et al. 1997, Strader et al. 2011, Blom et al. 2011). Therefore, the kinematic axis ratio of the GC system was assumed to be equal to the respective photometric axis ratio of the galaxy light (see Table 1). The galaxy systemic velocity was fixed to the values given in Table 1 because the GC mean velocity is in good agreement with the galaxy systemic velocity itself. Uncertainties on the fits to the GC kinematics are obtained using a bootstrapping method similar to that used by Côté et al. (2001). We obtain 1000 “mock” GC kinematic samples for each galaxy, by sampling with replacement from our measured distribution. Think kind of kinematic modelling tends to overestimate the rotation amplitude when the kinematic position angle is a free parameter. The correction for this bias is described in Appendix A.

By way of example, the best fit to eq. 4 for the GCs in NGC  and its kinematic dataset are shown in Figure 10 and Table 5, respectively.

The kinematic method described above was used to investigate the kinematic properties of our surveyed galaxies in two different ways:

  • First, the kinematic properties of each GC system were calculated as a function of the galactocentric radius. This approach allows us to compare the kinematics of a given GC system to different kinematic probes (e.g., long-slit spectroscopy and/or PNe kinematics). We generally use between and GCs per bin depending on the galaxy. These results are shown in Figure 11.

  • Secondly, the confirmed GCs were sorted by their colour to study the GC kinematics as a function of colour. Here we use moving colour bins of equal width (with usually GCs per bin) to investigate the effect of each GC on the final fit. This analysis is independent of the photometric dividing colour and it has the advantage of testing whether or not there is a transition in the kinematics between the two subpopulations. These results are shown in Figure 13.

To better appreciate the global kinematic properties of our GC systems, we also construct a 2D smoothed velocity field for our elongated early-type galaxies (i.e., galaxies with : NGC , NGC , NGC , NGC ). These galaxies are all part of the ATLAS sample (Krajnović et al., 2011) and are indeed good candidates to compare their moderate–fast stellar rotation in the innermost regions with the kinematics of their GC system. At every position on the sky we compute the local radial velocity via interpolation of the weighted average radial velocities of the nearest neighbours, similar to the technique used in Coccato et al. (2009):

(5)

with the weights being the reciprocal of the square distance between two neighbour GCs. The 2D velocity field was smoothed using a Gaussian filter kernel of variable width for each galaxy on a regularly spaced grid. The kernel size was arbitrarily set between and kpc depending the sampling of the GC system and on the size of the galaxy. For the sake of comparison, we analysed the SAURON data-cubes (Emsellem et al., 2007) of the same four early-type ellipticals in order to reproduce their 2D stellar velocity fields using our technique. These results are shown in Figure 15.

Figure 10: Rotation of NGC  GC system with position angle in different radial bins. From the top to the bottom, panels show the rotation for all, blue and red GC subpopulations respectively. Whereas from the left to the right, panels show the rotation in the radial bin reported on the top left for each panel. The photometric major axis and the systemic velocity of the galaxy are both from Table 1 and represented with dashed vertical and dashed horizontal lines respectively. The best fit to eq. 4 and envelope are shown as filled and dotted lines respectively.

We also compute the root-mean-square velocity:

(6)

This quantity coincides with the velocity dispersion in eq. 4 if the contribution of the rotation is negligible. The uncertainty on the is estimated with the formulae provided by Danese et al. (1980).

ID RA Dec V V
[Degree] [Degree] [ km s] [ km s] [mag] [mag] [mag] [mag] [mag] [mag]
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
NGC_GC1 8 3
NGC_GC2 10
NGC_GC3 13
NGC_GC4 17
NGC_GC5 14 1
NGC_GC6 15
NGC_GC7 18
NGC_GC8 5
NGC_GC9 6
NGC_GC10 7
NGC_star1 1
NGC_star2
NGC_gal1
NGC_gal2
Table 5: Spectro-photometric catalogue of objects around NGC . The horizontal line divides the sample into spectroscopically confirmed GCs, spectroscopically confirmed Galactic stars and background galaxies. The first column gives the object ID, composed of the galaxy name and the object identification. Columns 2 and 3 give position in right ascension and declination (J), respectively. Columns 4 and 5 are the observed heliocentric radial velocities and uncertainties respectively. The remaining columns are the Subaru photometry in gri and respective uncertainties. The full table for all the surveyed galaxies is available in the online version.

5 Notes on individual galaxies

In this section we briefly discuss the kinematic results for our new nine GC systems that are presented in Figure 11, Figure 13 and Figure 15.

5.1 Ngc 821

NGC  is an isolated E6 galaxy (de Vaucouleurs, 1991) with photometric and kinematic signatures of an edge-on stellar disk (Emsellem et al. 2007; Proctor et al. 2009). The dark matter content of this galaxy has been debated in the literature. Romanowsky et al. (2003), and recently Teodorescu et al. (2010) found that the velocity dispersion of the PNe decreases with radius. In contrast, Weijmans et al. (2009), Forestell & Gebhardt (2010) and Proctor et al. (2009) found a flat velocity dispersion for the stellar component within 100 arcsec. A photometric study of the GC system of NGC  was carried out by Spitler et al. (2008) using wide-field WIYN and HST observations. They were able to detect significant colour bimodality only after combining ground-based and space-based observations.

Our Subaru observations suffer from moderate band seeing ( arcsec). However, we detect significant GC colour bimodality, although the blue and the red peaks are not clearly visible. We confirm that the surface density profile of the GC system extends up to arcmin, as suggested from the WIYN imaging.

The combination of the poor band seeing, and the presence of the th magnitude star arcmin from the galaxy centre, resulted in a low return rate of spectroscopically confirmed GCs. In total, we obtained radial velocities for GCs over DEIMOS masks. We adopt a colour split at . This value was used to analyse the kinematics of blue and red GC subpopulations separately.

We detect significant rotation only for the blue GCs and for a small group of red GCs. The blue GCs are found to rotate at along PA deg, consistent with the photometric minor axis and in agreement with that found by Coccato et al. (2009) using PNe. Such a peculiarity is clearly visible in Figure 13. We will discuss this feature and its implications in more detail later. Interestingly, we note that the direction of the GC and PNe rotation coincides with an elongated jet/outflow structure detected in the X-ray (Pellegrini et al., 2007). With the adopted colour split, the kinematic position angle of the red GCs is generally unconstrained, except the outer red GCs that counter rotate with respect to the host galaxy stars. The velocity dispersion of both the red and the blue GCs declines with radius with a slope similar to that of the PNe and host galaxy stars.

Galaxy ID PA PA PA
[ km s] [ km s] [ km s] [ km s] [deg] [deg] [deg]
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
NGC 
Table 6: GC kinematic results for our galaxy sample. Column (2) shows the systemic velocity of the host galaxy obtained from the weighted average of the GC radial velocities of the azimuthal-complete innermost region. The remaining columns are the overall root-mean-square velocity , the rotational dominance parameter and kinematic position angle PA for all, blue and red GC subpopulations respectively.

5.2 Ngc 1400

NGC  has been classified both as a face-on S (Jarrett et al., 2003) and as an E0 (e.g., da Costa et al. 1998). It is the second brightest galaxy in the Eridanus group after NGC . We have assumed that NGC  and NGC  lie at the same distance of Mpc. NGC  is characterised by a uniformly old stellar age up to (Spolaor et al., 2008) and by an unusually low systemic velocity (V) for its distance. Forbes et al. (2006) studied the GC system of this galaxy using Keck/LRIS in imaging mode, detecting significant bimodality.

The photometric GC selection was performed within arcmin from the galaxy centre to minimise the contamination from NGC  GC system, resulting in a clear bimodal distribution, with the colour separation occurring at . The GC surface density was corrected for the local NGC  contribution. This galaxy shows the steepest red surface density profile among our galaxies, similar to the slope of the galaxy surface brightness. In contrast, the profile of the blue GC subpopulation is more radially extended and also requires a power-law fit, because a Sèrsic function does not return a satisfactory solution.

We present here radial velocities for blue and red GCs respectively, for a total of spectroscopically confirmed GCs. Despite their small angular separation arcmin), NGC  and NGC  also have a large peculiar velocity difference (  km s), assuring a reliable separation of their spectroscopically confirmed GCs. The red GCs mimic the rotation of the stars in the inner regions, whereas the rotation of the blue GCs is consistent with zero with a marginal signature of counter-rotation in the inner radial bin. The velocity dispersion of the red GCs is in agreement with long–slit data in the region of overlap, with a slightly increasing trend towards the outer regions. Conversely, the velocity dispersion of the blue GCs decreased with the radius.

5.3 Ngc 1407

NGC  is a massive E galaxy at the centre of the dwarf galaxy dominated Eridanus A group (Brough et al., 2006). It shows moderate rotation along the photometric major axis and has a weak central AGN (Zhang et al., 2007). The stellar population analysis of Spolaor et al. (2008) found the galaxy to possess a uniformly old age within . A dynamical analysis of NGC  was given by Romanowsky et al. (2009). They used Suprime-Cam imaging and DEIMOS spectra for GCs and found a massive dark halo.

In this work, we use the photometric results presented in Romanowsky et al. (2009) in which the Suprime-Cam imaging was reduced and analysed with the same methodology described in this paper. The colour distribution shows clear bimodality with the colour separation occurring at . We supplement the spectroscopic sample of Romanowsky et al. (2009) with additional DEIMOS masks, that make NGC , with a total of spectroscopically confirmed GCs, the most populous spectro-photometric dataset in our galaxy sample.

We detect rotation along the photometric major axis for both GC subpopulations. The blue GCs rotate in the innermost radial bin along the photometric minor axis and between and  along the photometric major axis. Similarly, the rotation of the red GCs occurs along the photometric major axis directions, but in this case the rotation signal is generally larger and better constrained than that of the blue GCs. Both GC subpopulations appear to rotate in the outermost regions of the galaxy along the major axis of the galaxy. The velocity dispersion of the red GC subpopulation in the inner regions is consistent with stellar data from Spolaor et al. (2008) and Proctor et al. (2009) showing a decreasing profile up to 10 . Conversely, the velocity dispersion of the blue GCs increases with radius.

5.4 Ngc 2768

NGC  is classified as E in de Vaucouleurs (1991) and as S in Sandage & Bedke (1994). Spectroscopic studies of the very central regions (McDermid et al., 2006) have also suggested a young ( Gyr) stellar population associated with the disk, as supported by the recent supernova SN2000ds (Filippenko & Chornock, 2000). Kundu & Whitmore (2001) studied the HST/WFPC2 photometry for 113 GC candidates in this galaxy, finding a statistically significant probability of it having a bimodal colour distribution.

In this work, NGC  was imaged with filters, in good seeing conditions but with the lowest exposure time among our sample of galaxies. Using KMM we found the colour distribution to be bimodal, with the blue and red peaks at and respectively. Bimodality was also found in the HST/ACS West pointing used to design the DEIMOS masks in the central region. The colour peaks occur at (F435W  F814W)  and respectively.

We find 109 spectroscopically confirmed GCs over 5 DEIMOS masks observed in sub-arcsec seeing conditions. Using a colour split at and excluding the marginal GCs, we investigate the kinematics of the resulting blue and red GCs respectively.

We find significant rotation only for the red GCs, which rotate roughly along the photometric major axis in agreement with the host galaxy stars. The –D velocity field of the red GCs is akin to that of ATLAS in the inner regions (Figure 15). The blue GCs have a marginally higher velocity dispersion profile than the red GCs.

Figure 11: GC system kinematics compared to stellar data. For each galaxy (labeled on the top right of each panel) the plots represent the tern of parameters that minimise eq. 3 as a function of the radius. Data points are joined for clarity, with the error bars representing the per cent confidence intervals. The dashed horizontal lines in top and middle panels represent the photometric deg from Table 1 and , respectively. All the PNe data (green stars) are from Coccato et al. (2009). Integrated stellar light data are shown as small black points (if from long–slit) and large black points (if from Proctor et al. 2009). References for long–slit data are: Forestell & Gebhardt (2010) for NGC ; Proctor et al. (2009) for NGC  and NGC ; Fried & Illingworth (1994) for NGC ; Coccato et al. (2009) for NGC ; van der Marel & Franx (1993) for NGC ; Bender et al. (1994) for NGC ; Kronawitter et al. (2000) for NGC ; Simien & Prugniel (2000) for NGC . Black trangles in NGC  are SAURON-IFU data from Weijmans et al. (2009). For NGC , because of the low number statistics, we only show the total GC kinematic profile in black.
Figure 12: continued

5.5 Ngc 3377

Figure 13: GC kinematics as a function of colour. Kinematic position angle, rotation velocity and velocity dispersion are shown in the top, central and bottom panel respectively. Dotted lines represent per cent confidence intervals. Also shown are the photometric major axis (horizontal dotted lines) from Table 1 and the colour split between blue and red GC subpopulations (vertical dashed line) as derived from the KMM analysis.
Figure 14: continued

NGC  is a classical E6 elliptical in the Leo I group and the closest galaxy in our sample. The resolved stellar properties of NGC  have been studied by (Harris et al., 2007) that found no evidence for any young ( Gyr) stellar population. Coccato et al. (2009) studied PNe in this galaxy, detecting significant rotation within arcmin and a kinematic major axis twist of deg from the major towards the minor axis. Kundu & Whitmore (2001), Chies-Santos et al. (2011) and Cho et al. (2012) studied the GC system of this galaxy using HST imaging, finding the colour distribution to be likely bimodal.

NGC  was imaged with filters and gz filters for ground based and space based observations respectively. We found a high probability for the colour distribution to be bimodal, both in our Suprime-Cam and in the supplementary ACS archive imaging, as already found in Chies-Santos et al. (2011).

In this work, we present radial velocities for 126 GCs, observed in 4 DEIMOS masks. It is worth noting that we extend the kinematics of this galaxy to kpc, two times further than the PNe studies. We rule out two GCs that are likely to belong to the spiral galaxy NGC A (V 573  km s) that lies arcmin North-West from NGC . With a colour split at we study the kinematics of blue and red GCs, respectively.

We detect significant rotation for the red GC subpopulation along the photometric major axis in agreement with PNe and long–slit data (Coccato et al., 2009). As found for the PNe, we discover that the kinematic position angle of the red GC subpopulation twists with radius from deg to deg (see also Figure 10). The –D velocity field of the red GCs is consistent with ATLAS in the inner regions (Figure 15). The position angle of the blue GC subpopulation is unconstrained, implying an overall null rotation with radius. The velocity dispersion of the red GC subpopulation is very flat and it matches the stellar and the PNe. The blue GCs have an overall higher velocity dispersion profile than the red GC subpopulation.

5.6 Ngc 4278

NGC  is an elliptical (E) member of the Coma I cloud. It has been extensively studied at different wavelengths: in radio, Nagar et al. (2005) detected two sub-parsec jets; in X-rays, a long (months) time scale variability (Ho et al., 2001) and a dominant nuclear source (Younes et al., 2010) were detected. It has long been known for its massive ( M) H I disc extending beyond  (Gallagher et al., 1977) and used to infer the dark matter content at large radii (Bertola et al., 1993). The bimodality of the GC system of NGC  has been under debate. Both Forbes et al. (1996) and Kundu & Whitmore (2001) investigated the GC system of NGC  with HST/WFPC2, finding a marginal or “likely” probability for the distribution to be bimodal. Recently, Chies-Santos et al. (2011) combined HST/ACS and WHT/LRIS imaging, finding evidence for bimodality in space-based but not ground-based datasets.

In this work we present new wide field Suprime-Cam imaging (with the band in poor seeing conditions), combined with four and HST/ACS pointings downloaded from the Hubble Legacy Archive (see Section 3.2). We find statistically significant bimodality both in our ground-based and space-based observations, with the colour separation occurring at and , respectively.

We spectroscopically confirm GCs over four DEIMOS masks observed in good seeing conditions. Given the wide field coverage of the four HST pointings, we decided to use HST magnitudes to divide the kinematics of blue and red GC subpopulations, converting colours into if only Subaru photometry was available. As for other group members, we exclude from the kinematics analysis GCs that might be associated with other galaxies. We select and remove three GCs whose position and radial velocity are consistent with the galaxy NGC  that lies ( kpc) arcmin from NGC .

We find no clear evidence for red GC rotation, but there is a hint that the blue GCs rotate in the outer regions along a direction intermediate between the major and the minor axis. The velocity dispersion of the red and blue GCs seem to be in good agreement with long–slit data in the inner regions although there is no direct overlap between the two datasets. The blue GCs have an overall higher velocity dispersion at intermediate radii.

5.7 Ngc 4365

NGC  is a massive elliptical (E3) behind the Virgo cluster with signs of an ongoing merger (Mihos et al. 2012, in preparation). It has a kinematically distinct core (Krajnović et al., 2011) and an unusual stellar rotation along the photometric minor axis (Bender et al., 1994). Its GC system is mainly known for an odd GC colour distribution that seems to suggest the presence of three, instead of the classic two, subpopulations (Larsen et al., 2003; Larsen et al., 2005; Brodie et al., 2005; Blom et al., 2011).

We make use of the photometric results presented in Blom et al. (2011) based on HST/ACS and Suprime-Cam imaging. In this work, we will treat the GC system of NGC  as made up of two classic subpopulations and we refer to Blom et al. (2012, in preparation) for a detailed study of the kinematics of the three subpopulations. We use a colour split at that includes the “green” GCs as part of the red subpopulation. We represent here the surface density profile published in Blom et al. (2011) that shows that red GCs is more centrally concentrated than the blue GC subpopulation, with a trend similar to that of the galaxy stellar light (Figure 6).

We spectroscopically confirm GCs over 6 DEIMOS masks. We study the kinematics of blue and red GCs respectively. We also add 9 GCs from the combined dataset of Larsen et al. (2003) and Brodie et al. (2005) as described in §4.2. We find that the red GCs rotate within arcsec along the photometric minor axis, mimicking the kinematics of the stars. The rotation of the blue GCs is only significant at intermediate radii. The velocity dispersion profile is identical for the blue and the red GCs.

Figure 15: Two-dimensional smoothed velocity fields of our disky early-type galaxies. Black points and ellipses represent the locations of the spectroscopically confirmed GCs for a given population and the  isophotes corrected for the galaxy ellipticity respectively. The white boxes are the maximum red-shifted and minimum blue-shifted velocities in the galaxy rest frame in unit of  km s. The SAURON stellar velocity maps are also shown in the innermost region of the galaxies. The agreement between the two datasets (except for NGC ) is extremely good.

5.8 Ngc 5846

NGC  is the brightest member of a galaxy group. The dynamics of NGC  has been modelled within 1 (Cappellari et al. 2007; Kronawitter et al. 2000). Recently, Das et al. (2008) derived the mass of this galaxy up to 11 exploiting X-ray and PNe observations. An HST/WFPC2 analysis of the NGC  GC system was given in Forbes et al. (1997) and then revisited by Chies-Santos et al. (2006). In addition to the classic bimodality, they also discovered that the GC system is better aligned with the galaxy’s minor axis than its major axis and that this galaxy has an unusually low specific frequency compared to similar dominant ellipticals in groups or clusters.

In this work we present new Subaru photometry in filters, with the band in moderate seeing conditions ( arcsec). Although the red peak is not clearly visible (Figure 5), KMM returned a high probability for the colour distribution to be bimodal with a colour split at .

We spectroscopically confirm GCs over DEIMOS masks. We also note that some radial velocities might be associated with other bright members of the group as NGC A () and NGC  (). However the similar systemic velocities of these galaxies with NGC  () makes any attempt to distinguish their GC populations problematic. We also supplement our DEIMOS catalogue with 22 GCs from the dataset of Puzia et al. (2004) as described in §4.2. In summary, we study the kinematics of blue and red GCs.

We detect rotation only for the blue GCs, which rotate between and arcsec close to photometric major axis. This feature might be caused by the GCs of NGC A, which contaminates our sample within arcsec. The velocity dispersion of the red GCs is flat with radius and consistent with other studies, although the PNe seem to suggest a slightly decreasing slope. The blue GCs have a systematically higher velocity dispersion than the red GC subpopulation.

5.9 Ngc 7457

NGC  is an isolated S with a pseudo-bulge detected both photometrically (Tomita et al., 2000) and kinematically (Pinkney et al., 2003) that shows an unusually low central velocity dispersion for its luminosity. Emsellem et al. (2007) revealed that this galaxy has a small counter rotating core that might be the result of a merger. Chomiuk et al. (2008) gave an overview of the GC system of NGC  using HST observations and Keck/LRIS spectra for GCs. They find evidence for a third intermediate population of GCs sharing the same age ( Gyr) as the young nuclear (radius of 1 arcsec) stellar population discovered by Sil’chenko et al. (2002).

NGC  is the only galaxy in our sample not observed with Suprime-Cam. We have instead used the photometric results of Hargis et al. (2011) to design our DEIMOS masks, and we refer to their paper for a detailed description of the data reduction and analysis. They provide wide field WIYN/Minimosaic photometry of the GC system of this galaxy, finding that the colour distribution is not bimodal and showing the total population to be made up of GCs with a radial extent of kpc.

We find 21 spectroscopically confirmed GCs in two DEIMOS masks observed in average arcsec seeing conditions. We also obtained spectra for of the GCs already observed by Chomiuk et al. (2008), finding their LRIS and our DEIMOS radial velocities to be consistent within the errors, but with a mean difference of (see §4.2). Therefore, we add the remaining Chomiuk GCs corrected for the velocity offset to our GC sample. The final GC catalogue consists of 27 GC radial velocities.

Our confirmed GCs appear to be distributed within 2 arcmin  kpc from the galaxy centre, reflecting the expected poor extent of the underlying GC subpopulation. The mean velocity of the system () is in good agreement with both Chomiuk et al. (2008) and literature data. As this galaxy appears to be unimodal in colour, we fit the kinematics of all the GCs without any split in colour and we compare the results with the long-slit data of Simien & Prugniel (2000). We find that the GC system of NGC  is rapidly rotating along the photometric major axis, with a flat rotation curve at up to . In contrast, the velocity dispersion is low at all radii. The agreement with the ATLAS velocity map is also very good (Figure 15). We will discuss the implication of these results in more detail in §10.4.

6 GC formation models

Having determined the photometric and kinematic properties of our GC sample, in the following sections we will discuss these results as a whole. We look for common kinematic features that might retain key information about the formation history of the galaxies themselves and then we compare these results to theoretical models. In order to do so, it is first worth giving a brief summary of the main GC formation scenarios proposed in the literature, focusing on their kinematic predictions.

To date, the formation of GC bimodality has been investigated both in a cosmological context (Kravtsov & Gnedin, 2005; Diemand et al., 2005; Moore et al., 2006) and at smaller scales, with models fine tuned to reproduce the properties of specific galaxies (see Vesperini et al. 2003 for NGC  or Deason et al. 2011 for the Milky Way and M31). The challenge of producing simulations of GC bimodality formation is intertwined with our poor understanding of how GC systems formed in the first place (e.g., Beasley et al., 2002; Elmegreen et al., 2012). As a consequence, the three classic formation scenarios proposed in the literature and summarized below have few, if any, theoretical predictions directly comparable with the observations. In particular, there is a general dearth of GC kinematics predictions.

In the major merger scenario (Ashman & Zepf, 1992) two or more gas-rich disk galaxies with pre-existing blue and red GC subpopulations merge to form an elliptical galaxy. New red GCs may form from the star formation induced by the merger (Bekki et al., 2002).

This scenario has been simulated by Bekki et al. (2005) (hereafter B05). In this simulation the pre-existing blue and red GCs are assumed to have a Milky Way-like spatial distribution, but they are both pressure-supported. This is a reasonable assumption for the blue but not for the red GCs of the Milky Way that are known to be rotation-supported (Côté, 1999; Deason et al., 2011). B predicted that a merger with even mass-to-mass ratio produces strong rotation for both GC subpopualtions in the outer regions of the remnant. This result is independent of the orbital configuration of the merger. The velocity dispersion profile is predicted to decrease with radius, but it would be flatter in case of multiple mergers. Also, the blue GCs are expected to show a larger central velocity dispersion than red GC subpopulation. Finally, the ratio of the maximum rotational velocity to the central velocity dispersion of the GC systems ranges from 0.2 to 1.0 within 6 for both the blue and red GC subpopulations, but most GC systems, viewed from various angles, have . Bekki & Peng (2006) carried out a simulation similar to that of B (with the rotation of the disk in the spiral progenitors included), fine tuned to study the dynamics of the planetary nebulae (PNe) in elliptical galaxies. They found that the effect of the residual spin disk (additional to the initial orbital angular momentum) on the final PNe kinematics enhances the rotation at all radii, making the merger remnant rotation-supported.

In the multiphase dissipational collapse scenario (Forbes et al., 1997), GCs are the result of an early two phase collapse, and hence two main star formation episodes, that the galaxy undergoes. The blue GCs form during the first star formation episode in a metal-poor cloud, whereas red GC subpopulation form in a second phase after the gas in the galaxy is self-enriched. In this scenario, a fraction of blue GCs might also come from accreted satellite galaxies, similar to the most recent two-phase galaxy formation model (Oser et al., 2010). No numerical simulations have been performed so far to test this model. Since the red GCs are coeval with the galaxy stellar component, Forbes et al. (1997) infer that they should share the same kinematic properties as the galaxy stars, and hence also the same spatial distribution. No significant rotation is expected for the blue GC subpopulation.

In the dissipationless accretion scenario (Côté et al., 1998), the red GCs form from the monolithic collapse of a giant protogalactic cloud, whereas the blue GCs are accreted from low mass galaxies due to mergers or tidal stripping. To date, simulations based on similar scenarios have been performed (e.g., Oser et al., 2010) but they do not include GC particles, and they have not made kinematic predictions. In addition to the observation of infalling co-moving groups, Côté et al. (1998) also infer that we should expect to observe blue GCs with radially biased orbits without overall rotation.

Bekki et al. (2008) (hereafter B08) performed a high resolution N-body cosmological simulation combined with semi-analytic models. They found that almost half of the simulated galaxies show clear bimodality in their GC metallicity distribution. The majority of GCs form in low-mass galaxies at redshift greater than 3 with the blue GCs being slightly older ( Gyr) than the red GC subpopulation. B08 also made predictions for the overall kinematics of each GC system, whereas current spectroscopic observations only sample a smaller fraction of it. They find that the velocity dispersion of blue and red GC subpopulations increases with the total luminosity of the host galaxy and that the ratio of the velocity dispersion of the blue to red GC subpopulation is for a wide range of luminosity. Finally, they also predicted that the GC systems of most galaxies are mainly pressure supported with (. This is in contrast with the wider range of found by B.

Figure 16: GC rotation velocities and velocity dispersions compared to the galaxy stars/PNe. Rotation velocities (left panel) and velocity dispersions (right panel) are extracted within 3. The exact extraction radius is usually set to 2 but this threshold is arbitrarily adjusted by in case the GCs do not overlap with the stellar data. The blue and the red GC subpopulations are shown in the bottom and upper panel respectively. Dashed lines represent a one-to-one relationship. Blue GC data for NGC  and NGC  are not shown due to limited radial range. NGC  is not shown. The red GC subpopulation shows better agreement with stars than the blue GCs.

7 Global results from our GC sample

In this section we summarize the results for the GC systems presented in this paper, including the three galaxies previously published by us (i.e., NGC , NGC , NGC ). Section 8 will incorporate other literature data, and tackle topics not covered here.

7.1 Spatial Distribution

Figure 17: Velocity dispersion profiles. For each galaxy, we show the ratio of the velocity dispersion of blue to red GC subpopulations normalised to  (vertical dashed line) as a function of the effective radius. Different colours correspond to different galaxies. The horizontal and vertical lines show and , respectively. NGC  and NGC  are not shown.

Significant colour bimodality was detected in all the GC systems, except for NGC . All GC system formation scenarios, as well as numerical simulations, predict that this bimodality should be reflected in different spatial distributions of the two subpopulations around the parent galaxy. A glance at Figure 6 reveals that this is also the case for our galaxy sample. Generally speaking, the red GCs are more centrally concentrated than the blue GCs, as already found in other studies (e.g., Schuberth et al., 2010; Faifer et al., 2011; Strader et al., 2011).

The advent of wide field imaging has shown that the radial distribution of the red GC subpopulation matches that of the host galaxy light, suggesting that they might have shared a similar formation history (Bassino et al., 2006). Here we have exploited the arcmin field of view of our Suprime-Cam imaging to compare the GC surface density to the surface brightness of the respective host galaxy (shifted by an arbitrary constant). Qualitatively speaking, the galaxy starlight has a similar slope to that of the red GC subpopulation. The discrepancies between the surface brightness and the GC spatial distribution in innermost regions (e.g., see NGC  and NGC  in Figure 6) is due to the core-like distribution of the GCs, that makes the GC density profile flatter. This feature is interpreted as the effect of the GC disruption, stronger in the central regions (Baumgardt, 1998; Baumgardt & Makino, 2003).

7.2 Rotation

We find a large variety of GC rotation profiles. Both the blue and red GCs show some degree of rotation, but there does not seem to be a clear correlation between the rotation patterns and the property of the host galaxy. A glance at Figure 11 reveals that the red GCs rotate more consistently with the photometric major axis than the blue GC subpopulation does. For our galaxy sample the photometric position angle coincides with the photometric major axis of the galaxy stars (except for NGC ), which means that the rotation velocity of the red GC subpopulation is similar to that of the galaxy stars. This is also true for the overall GC system of NGC .

To quantify this phenomenon, in Figure 16 we compare the major axis rotation velocity of the blue and the red GC subpopulations within 3 to the rotation velocity of the stars and PNe (if available) at the same galactocentric distance. The exact extraction radius was set to 2, but in some cases this was relaxed by to maximise the overlap between the GCs and the stars. The rotation velocity of the host galaxy stars was extracted along the photometric major axis for all the galaxies, except for NGC  in which the bulk of the stellar rotation is occurring along the photometric minor axis. The red GC rotation for our galaxies is similar to those of the host galaxy stars, at least for galaxies with a conspicuous amount of stellar rotation. On the other hand, the star-GC system coupling is also evident for only blue GC subpopulations, such as in NGC , NGC  and NGC .

The rotation of the blue GCs is more puzzling than the red GCs. Overall the rotation of the blue GCs is lower than the red GCs and often consistent with zero, yet we also detect minor axis rotation in NGC , also seen in PNe, in contrast with integrated starlight.

We detect GC rotation at large galactocentric radii for some galaxies. This feature is only significant in NGC , NGC  and only marginal in NGC . The outer rotation occurs in the same direction for both the blue and the red GC subpopulations and it usually coincides with the photometric major axis.

Figure 18: Summary of GC system kinematics as a function of the host galaxy properties. Literature data and our sample are shown as red and black points respectively. From the top to the bottom: root-mean-square velocity for the overall GC system (), ratio of the of the blue to the red GC subpopulation (), ratio of the rotation velocity to the velocity dispersion for the overall GC system . From the left to the right: X-ray luminosity (log ) from O’Sullivan et al. (2001) at the distance given in Table 1 and Table 7, central stellar velocity dispersion () from Prugniel & Simien (1996) and total K-band magnitude (M) from 2MASS. The black and red arrows indicate the position of NGC  at and of NGC  at , respectively. The horizontal dashed line marks to guide the eye. The grey shaded shaded region represents the theoretical predictions from the cosmological simulation of Bekki et al. (2008).

7.3 Velocity Dispersion

Galaxy ID Hubble N Ref.
Type [mag] [Mpc] [ km s] [ km s] [ km s]
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
NGC  SA0 P04
NGC  E1 S
NGC  cD M
NGC  E1 P; B
NGC  E6 P04
NGC  E4