© ESO 2020

1. Introduction

The study of local complexes of galaxies – galaxy clusters and groups– is crucial for attaining an understanding of the history of formation and evolution of the Universe through its building blocks. Local galaxy systems mark the endpoint of the evolution of galaxies after billion years of interactions, of varying intensities, with their companions (e.g., Mo et al. 2010).

A detailed study of the two extreme structures in terms of stellar density offers precious information on the history of formation and interactions of a galaxy: faint extended stellar features in the outskirts of galaxies, characterized by low star density and very long dynamical mixing timescales (Johnston et al. 2008), along with compact stellar systems, which are intrinsically bright, have typically old ages and have orbits that can trace recent and ancient accretion events (Brodie & Strader 2006). The stratification of dense star clusters and low surface brightness features can aid in probing a galaxy environment on different timescales from the earliest epoch of formation to the most recent merging events (e.g., West et al. 2004; Bournaud & Bournaud 2011).

In the last decade, also thanks to the advent of efficient large-format imaging cameras, a number of observational programs have carried out intensive surveys dedicated to covering large sections of nearby galaxy systems, superseding, in terms of both limiting magnitude and spatial resolution, any previous optical or near-IR study (e.g., Ferrarese et al. 2012; Iodice et al. 2016), thus providing a rich variety of data ideal for investigating compact stellar systems and faint stellar structures in different galaxy environments (de Jong et al. 2013; Muñoz et al. 2014; Durrell et al. 2014; Iodice et al. 2019; Venhola et al. 2019; Wittmann et al. 2019)

Within this framework, the Fornax Deep Survey (FDS) has carried out observations of the Fornax galaxy cluster centered on NGC 1399 out to one virial radius and further extended observations in the direction of the Fornax A sub-cluster in the South-West with its brightest member, NGC 1316, with a list of scientific topics: diffuse light and intracluster medium (Iodice et al. 2016), galaxy scaling relations (Iodice et al. 2019; Venhola et al. 2017, 2019; Raj et al. 2019), extragalactic star clusters and, more generally, compact stellar systems (D’Abrusco et al. 2016; Cantiello et al. 2018a), etc. In addition, the survey also contributes to research programs dealing with the study of the background galaxy population (e.g., identification of lensed systems and of their physical properties) and spectroscopic programs – for globular clusters (Pota et al. 2018), planetary nebulae (Spiniello et al. 2018), IFU study of galaxies in the cluster (Mentz et al. 2016).

The aim of this paper is to present the photometric and morphometric catalog of all point-like and slightly extended sources of the survey, along with a description of the methodology used to characterize the sources. As key topics of the survey, we present a preliminary study of compact stellar systems, including globular clusters (GCs) and ultra compact dwarf galaxies (UCDs).

Extragalactic, unresolved GCs are possibly the simplest class of astrophysical objects beyond stars. To a first approximation, GCs host a simple (that is single age and single metallicity) stellar population. In spite of the results on multiple populations in globular clusters (e.g., Piotto et al. 2007; Carretta et al. 2009; Bastian & Lardo 2018), it is doubtless that GCs host a stellar population that is much simpler than galaxies, in terms of the metallicity and age distributions, because their simpler star-formation history makes it possible to constrain the properties of these systems at a higher level of precision with regard to more complex and massive stellar systems.

The intrinsic simplicity of GCs, and of similar compact stellar systems, together with the old ages and the high luminosity, make these astronomical sources powerful and robust tracers of a galaxy and its environment, suitable to study a galaxy and its relevant structures out to cosmological distances (Alamo-Martínez et al. 2013; Janssens et al. 2017; Vanzella et al. 2017). The rich set of observables of stellar clusters makes them useful fossil records of the history of the evolution of their host galaxy and indicators of some of its physical property (distance, merging history, mass, metallicity, etc.). Here we focus on preliminary projected distribution maps of GCs and UCDs, and postpone further analysis of these sources to a forthcoming paper (Cantiello et al., in prep.).

In the following sections, we assume a distance modulus of (m − M)=31.51 ± 0.03 (ran.) ± 0.15 (sys.) mag for the Fornax galaxy cluster, corresponding to d = 20.0 ± 0.3 (ran.) ± 1.4 (sys.) Mpc (Blakeslee et al. 2009).

The paper is organized as follows. In Sect. 2, we describe the data, the procedures for source identification, calibration, and characterization, and we present the final FDS catalog of compact and slightly extended sources, as well as background galaxies. Section 3 is dedicated to a pilot application of the catalogs aimed at deriving 2D distributions of compact sources in the area. In Sect. 4, we report on the application to a science case for background sources. A brief summary of our conclusions is presented in Sect. 5.

2. Data and data analysis

2.1. Observations and data reduction

The observations used in this work are part of the now-completed FDS survey. The FDS consists of a combination of guaranteed time observations from the Fornax Cluster Ultra-deep Survey (FOCUS, P.I. R. Peletier) and the VST Early-type GAlaxy Survey (VEGAS, P.I. E. Iodice). The surveys were both performed with the ESO VLT Survey Telescope (VST), which is a 2.6 m diameter optical survey telescope located at Cerro Paranal, Chile (Schipani et al. 2010). The imaging is in the u, g,r and i-bands using the 1 × 1 square degree field of view camera OmegaCAM (Kuijken 2011).

The main body of the FDS dataset is centered on NGC 1399, the second brightest galaxy of the Fornax galaxy cluster in optical bands and the brightest galaxy of the main cluster. It consists of 21 VST fields with a complete ugri coverage. Further five fields in the gri bands extend in the south-west direction of the cluster, the Fornax A sub-cluster which covers the regions of the brightest cluster galaxy, the peculiar elliptical NGC 1316. For sake of clarity, in the following, we refer to the 21 FDS fields with ugri as FDS survey, and to the entire sample of 26 fields with gri coverage as FDS-extended, or FDSex. The FDS and FDSex areas are shown in Fig. 1; some of the known objects available from the literature and from previous FDS works are marked in the left panel of the figure.

Fig. 1. Left panel: FDS footprint of the area covered by ugri photometry (green solid line), and by only gri (dashed green line). Other sources from catalogs available in the literature are also shown, as labeled. Bright galaxies from the Fornax Cluster Catalog (Ferguson 1989) are subdivided into two categories: likely members brighter than BT = 17 mag and with 17 ≤ BT (mag) ≤ 18.5 (filled gray circles and blue triangles, respectively; from Ferguson 1989, Table II). Dwarf galaxies from FDS by Venhola et al. (2018), in the magnitude range 18.5 ≤ mg (mag) ≤ 21, are indicated with red crosses. The positions of the two brightest galaxies, NGC 1316 and NGC 1399, are also shown with a filled cyan triangle and a magenta square, respectively. Orange filled or empty five-pointed stars mark those stars with mV ≤ 7/≤9 mag, respectively. Right panel: FDS and FDSex area. Green lines mark the edges of the survey, green bullets show the edges of single pointings; the ID of the field is also indicated.

The data, data acquisition, and reduction procedures have been presented in a number of papers of the FDS series (Iodice et al. 2016, 2017a,b, 2019; Venhola et al. 2017, 2018, 2019). A full description of the observations and the pipeline used for data reduction (AstroWISE; McFarland et al. 2013) steps are given in the cited papers, and in Peletier et al. (in prep.). In the following, we describe two critical differences with respect to previous works, specifically related to the focus on compact stellar systems in the present work.

2.2. Multi-band image stacks

The FDS standard reduction pipeline produced imaging data for many different scientific cases, with a general focus on extended galaxies in the cluster (e.g., Spavone et al. 2017). In Cols. (2–5) of Table 1, we report the median full width at half maximum (FWHM) of the point spread function (PSF) in arcseconds for each FDS VST field and for each available band; the FWHM distributions are also shown in the histograms of Fig. 2. The large FWHM variation, up to ∼50% for different fields observed in the same passband, may represent a limitation to the effectiveness of the FDS dataset for the science cases related to compact objects (foreground MW stars, background galaxies, GCs host in Fornax, etc.). The typical FWHM scatter of the exposures combined to obtain the single FDS fields stacks is in u/g/r/i-band, respectively¹.

Fig. 2. Histograms of the median PSF FWHM of the FDS fields in the four available passbands, plus the multi-band a-stacks. The vertical dashed line shows the median of the ensemble.

To improve the detection and characterization of compact sources, we combined in a single coadded image all single VST exposures in g, r and i bands with a median FWHM lower than a fixed upper limit, u-band exposures were ignored because of the lower signal-to-noise and worse FWHM. After various experiments, we fixed the FWHM limit to : if a lower FWHM cut is adopted, the final resolution of the stack improves, at the expenses of a worse detection limit and larger field-to-field mean FWHM variability; a higher FWHM cut, instead, would make ineffective the use of multi-band stacks compared to single bands images. Hence, the cut is adopted as the trade off between needs of better resolution and uniformity of the master detection frame. The combined image was processed as the single band images, except for the photometric calibration which is not derived. In the following, we refer to the coadd of gri exposures with FWHM cut as a-stack, and use the subscript a to identify the quantities derived from it. With this procedure, a new frame with narrower and more stable FWHM compared with ugri bands is obtained, and used as master detection frame. This improved both the uniformity of detections over the different FDS fields, and the determination of the morphological properties of the sources, allowing more accurate characterization of compact and point-like objects. These a-stacks will not be used to define absolute quantities (like calibrated magnitudes), but only for relative ones (like the CI_n, see below), thus the wavelength dependence of the PSF and source morphology will not be an issue.

As shown in Table 1, the a-stacks have a median FWHM smaller by ∼15% and with an rms scatter a factor of ∼2.5 lower than the median and rms of the FWHM for the best passband, namely the r-band. In Fig. 3 we show a 1′ × 1′ thumbnail of the same FDS region in g, r, and i-band and the a-stack image centered on background spiral galaxy in the field FDS#5 (FCCB 1532, Ferguson 1989). In general, the depth of the coadded multiband a-stack does not change much compared with the best band of the field, because the reduced number of exposures used is compensated by the better S/N due to the higher spatial resolution. The spatial resolution, however, is in all cases enhanced, as shown in the FWHM_a column in Table 1.

Fig. 3. From left to right: g, r, i-band and a-stack of a background spiral galaxy in the field FDS#5 (FCCB 1532, Ferguson 1989). Rightmost panel: derived from the combination of the sub-exposures of the first three panels, selecting only the ones with lowest atmospheric turbulence (see text).

2.3. Photometry and photometric calibration

Catalogs were derived for each single FDS pointing; the identification of fields with available data is shown in the right panel of Fig. 1. To increase the contrast of faint sources close to the cores of extended galaxies, before running the procedures to obtain the photometry and the morphometry (like FWHM, elongation, flux radius, etc.; see Sect. 2.4 below) we modeled and subtracted all Fornax members brighter than B_T ∼ 18 mag. The fit of the isophotes is performed using the IRAF STSDAS task ELLIPSE, which is based on an algorithm by Jedrzejewski (1987).

To obtain the photometry of sources in FDS frames, we used a combination of procedures, based on SExtractor (Bertin & Arnouts 1996) and DAOphot (Stetson 1987) runs, and codes developed by the first author. We adopt AB mag photometric system, as in previous FDS works. The galaxy-subtracted frames used in this stage are already calibrated as described in the previous works of the FDS series (see below).

First, we used SExtractor to obtain the mean properties of each frame, like the FWHM; the reference morphometry for each source is obtained from the a-stacks, though we also derived the morphometric properties for all available passbands. Then, DAOphot is run on the a-stacks, and fed to our procedure to identify bright, non-saturated and isolated stars needed to obtain a variable PSF model over the single pointing. Typically, with this procedure we selected ∼200 candidate PSFs per single FDS field, that were visually inspected in all bands to remove candidates contaminated by faint companions, bright halos of galaxies or saturated stars, or other instrumental artifacts. Using this iterative process, we ended up with a typical list of 50 to 100 point-like sources to model the PSF with DAOphot for each filter and field. The list of PSFs was then fed to DAOphot for PSF modeling, adopting the variable PSF option. The first complete DAOphot run was on the a-stack. The output table for this run was used to (i) identify sources to define a master detection catalog, (ii) obtain the DAOphot sharpness parameter that would then be used as additional parameter for selecting good candidate compact sources.

The master detection catalog was then given as input to run DAOphot on each available filter and for all fields: ugri for the FDS area, gri for FDSex. We also run SExtractor on the full set of images, to obtain the aperture magnitude within 8-pixel diameter (MAG_APER) and the automated aperture magnitude derived from Kron (1980) for first moment algorithms (MAG_AUTO), with the respective photometric errors². For the aperture magnitudes, after some tests we adopted the eight-pixel diameter: larger diameters implied larger statistical errors on derived magnitudes (because of the noisier background and higher contamination from neighboring sources), smaller diameters suffered from larger systematic errors (because larger aperture corrections are needed). Both MAG_APER(8) and MAG_AUTO are stored in our final catalogs. It is, in particular, MAG_AUTO that provides a good choice for the magnitude of non-compact background objects.

The photometric calibration is carried out in two steps. The first is the same described in Venhola et al. (2018) and uses standard star fields observed each night and comparing their OmegaCAM magnitudes with the final data from the Sloan Digital Sky Survey Data III (Alam et al. 2015).

With such calibration, and after applying the field and pass-band dependent aperture corrections, the photometry of the same sources in different adjacent FDS pointings shows a spatially variable offset, with a median upper limit of ∼0.1 mag. This might be a consequence of the different (mean) photometric conditions for neighboring FDS fields during the FDS observing runs which span a time interval of ∼5 years.

As a second step of the photometric calibration, to improve the photometric uniformity and consistency over the FDS (and FDSex) area, and to derive the spatially and filter dependent aperture correction map, we compared our VST photometry of bright non-saturated point-like sources to the APASS photometry³ and obtained the two-dimensional map that best matches the two datasets. The map is derived for each field separately, using a support vector machine (SVM) supervised learning method, with a radial basis function (RBF) kernel (Pedregosa et al. 2011). Only isolated unsaturated stars, brighter than a given magnitude cut (19/17/17/16.5 mag in u/g/r/i band, respectively), are used in the regression algorithm.

The correction maps are derived from 200 to 300 stars per FDS field, the final median rms_{VST − APASS} between VST and APASS photometry over the full set of re-calibrated frames is reported in Table 2. Figure 4 shows an example of the correction maps derived for the field FDS#19. Each correction map is then applied to its specific field and passband, to correct the photometry of all sources detected in the specific FDS pointing.

Fig. 4. Example of the two-dimensional photometric correction maps for refining the photometry of FDS fields. The maps also include the aperture correction term. Field FDS#19 is shown: u, g, r, and i-band correction maps are plotted from upper to lower panels, respectively. For each passband, the surface correction map is shown with the same color coding and for different viewpoints in each of the three panels.

Because APASS lacks u coverage, for such passband we adopted a slightly different re-calibration strategy. After the preliminary calibration described above, the B-band magnitudes of stars from APASS were transformed to u-band using Lupton (2005) transformation equations available from the SDSS web pages⁴. In particular: u = B_APASS + 0.8116 ⋅ (u − g)_fit − 0.1313, where the (u − g)_fit color index is derived from the APASS (g − i) and (g − r) indices, using a second degree polynomial fit derived from SDSS data over different sky regions⁵. From this stage on, by using the u-band magnitudes of stars in APASS derived as a function of the B, g, r, and i photometry, we may proceed to derive and apply the u-band correction maps as in gri bands.

To further verify the validity of the calibration obtained with the strategy delineated above, especially for the more elaborate u-band, we matched and compared our photometry to the SkyMapper (SM) data (Wolf et al. 2018; Onken et al. 2019). The SDSS photometric systems of APASS and SM are not equivalent, the u and g bands, in particular, show differences of up to 0.5 mag in the two systems (Wolf et al. 2018). However, within the color interval |g − i| ≤ 1 mag, the SM to SDSS difference for uri-bands is ≲0.1 mag, while it is a factor of ∼4 larger in g-band (Wolf et al. 2018, see their Fig. 17 and Sects. 2.2, 5.4). Hence, as a further consistency check, we compare our VST re-calibrated photometry to SM data, within the color interval |g − i| ≤ 1 mag.

Over the entire FDS area covered with ugri observations, we found ∼46 500 sources in common with SM. After identifying bright and isolated stars, and with the given prescriptions on (g − i) color selection, the final sample contains ∼4600 objects (∼220 per FDS field).

Table 2 reports the median magnitude offsets between the FDS and SM photometry for the matched sources, together with the rms_MAD. With the only not unexpected exception of the g band we find good agreement between the u, r and i photometry, with magnitude offsets better than 0.02 mag in r and i bands and of ∼0.05 mag in u; the rms_MAD is ∼0.03 in gri and about twice larger in u-band.

For an independent check of the g-band photometry, we used the data from the HST/ACS Fornax Cluster Survey (ACSFCS; Jordán et al. 2007, 2015). In Fig. 5, we report a comparison of our and ACSFCS g-band magnitudes. We matched the ∼6.300 GC candidates from the ACSFCS with the FDSex gri catalog, to avoid the worse completeness limit of the u-band in the ugri catalogs. Adopting a matching radius of , a total of 3750 sources are found in common to both catalogs. The completeness of the matching is ∼90% or higher at bright magnitudes (m_g ≤ 23), decreases to ∼80% for m_g ≤ 24, and is lower than ∼70% for m_g ≤ 25. Hence, the completeness of the gri catalog drops quickly below m_g ∼ 24.5 (mag), which corresponds to ∼0.5 mag fainter than the turn over magnitude (TOM) of the GC luminosity function (GCLF) for galaxies in Fornax (Villegas et al. 2010).

Fig. 5. Left panel: g-band magnitudes from FDS compared with magnitudes of GC candidates from the ACSFCS. Blue symbols show the full matched set, red symbols identify compact sources in our reference catalog (see text). Middle panel: as left panel, but running averages are shown, with bin size of 100/50 objects for the blue/red symbols, respectively. Right panel: as left panel, but versus (g − i) color.

The left panel of Fig. 5 shows the VST to ACSFCS g-band magnitude difference versus m_g (blue dots in the figure). From the matched catalog, we selected a reference GC sample (see next section), marked as red dots in the figure. The running mean difference for both the full matched sample and the reference sample are shown in the middle panel, adopting window bin size 100/50 for the full/best sample, respectively. Finally, the right panel of the diagram shows the same quantities as in the left one, but versus the (g − i) color. In all cases shown, the difference is consistent with zero – Δg(FDS − ACSFCS)= − 0.03 ± 0.12 for the full sample of 3750 matched sources; Δg(FDS − ACSFCS)= − 0.01 ± 0.07 for the 1455 sources in the reference catalog – with no evidence of significant residual trends.

2.4. Morphometry

As already anticipated in Sect. 2.2, by morphometry we mean the measurement of all characteristics related to the shape of the source, our reference frames for the morphological characterization of sources are the multi-band a-stacks derived from gri exposures with the best seeing. We placed a particular emphasis on deriving quantities useful for distinguishing between point-like and extended sources and identified a number of useful features: FWHM, CLASS_STAR, flux radius, and elongation (major-to-minor axis ratio) derived with SExtractor, as well as the sharpness parameter derived from DAOphot.

For each source detected, we also measured the magnitude concentration index, described in Peng et al. (2011), defined as the difference in magnitude measured at two different radial apertures. Following various tests, we adopted as a reference the concentration index derived from the a-stacks aperture magnitudes at four and six pixels, namely: CI = mag_{4 pix} − mag_6 pix. For point-like sources, after applying the aperture correction to the PSF magnitudes of isolated stars at both radii, CI should be statistically consistent with zero. The concentration index is constant for point-like objects, while extended sources have variable CI larger than zero.

Because the a is not a real photometric band and because of the field-to-field variations for simplicity, we decided to normalize the CI index to 1, rather than to zero⁶. The normalization was derived as follows: for each field, we first estimated the CI from the magnitude difference within the two chosen apertures (so no aperture correction is applied), then derived the median CI of candidate point-like isolated and bright sources. Finally, the CI of the full sample was normalized to the median CI such that compact sources should, by construction, be characterized by normalized CI values, CI_n, of ∼1. Figure 6 shows the procedure described, for sources in the field FDS#13: as expected, compact sources (selected here using the morphological parameters from SExtractor) occupy a flat sequence of constant CI (left panel), normalized to one in the right panel of the figure.

Fig. 6. Normalization procedure for the CI, data for the field FDS#13 are shown. Left panel: concentration index CI = mag4 pix − mag6 pix versus the uncalibrated a-stack magnitude. For the sake of clarity, only the brightest magnitude range is shown. Blue dots refer to the full sample, red symbols to candidate compact sources used to derive the median CI factor for normalization. Right panel: same as left panel, but over a larger magnitude range and after normalization to the median CI of bright point-like sources (red dots). Point-like sources candidates are aligned along the sequence parallel to the x-axis, around CIn ∼ 1 (green dot-dashed line).

2.5. Final catalog and data quality

The DAOphot and SExtractor catalogs of sources in the FDS fields can then be combined in one single catalog (the same is done, independently, for the FDSex regions). The final catalog contains: (i) source identification adopting the IAU naming rules⁷ and position from the a-stacks; (ii) the calibrated AB magnitudes from PSF photometry derived with DAOphot in all available bands; (iii) the uncorrected aperture and Kron-like magnitudes from SExtractor; (iv) the morphometric parameters for a-stacks (FWHM, CLASS_STAR, flux radius, elongation and sharpness), as well as the latter for all other available bands. The FDS catalog provides data based on the 21 FDS field in the ugri-bands, and for the a-stacks; a second gri-bands catalog for the full FDSex area is also generated.

In the catalogs, we include the extinction correction term, assuming the Galactic extinction values from the Schlafly & Finkbeiner (2011) recalibration of the Schlegel et al. (1998) infrared based dust maps. Figure 7 shows a selection of extinction corrected color magnitude and color-color diagrams for the full sample of sources in the FDS catalog.

Fig. 7. Hess color–magnitude and color-color diagrams of the full sample with ugri photometry. Extinction corrected PSF magnitudes are used in all cases.

As an overall photometric quality assessment, we used the principal colors, described in Ivezić et al. (2004). Principal colors are linear combinations of the SDSS colors of stars. We adopted the coefficients and selection parameters given in Tables 1–3 of Ivezić et al. (2004). The colors are combined to obtain a new color perpendicular to the stellar locus. Assuming the position of the locus to be fixed, the value of the principal colors is then an internal measure of the absolute photometric calibration of the data. Table 1 provides the median and rms width of three principal colors, P2(s), P2(w) and P2(x) for each FDS field; the median P2 values over the full set of fields is < 0.02 with rms ≲ 0.03. The P₂(s) depends on the u-band photometry, and cannot be determined over the FDSex fields. The overall ⟨P₂⟩ and σ[P₂] values, and the values for each field, are consistent with the same value reported by Ivezić et al. (2004) for SDSS photometry.

Finally, we obtain the limiting magnitudes reported in Table 1 for all fields and bands, derived as 5σ magnitude integrated over the PSF, determined from the median S/N estimated as . The median g-band limiting magnitude is g_lim ∼ 25.4 ± 0.2 mag; we note that the faintest GCs matched with the ACSFCS reach m_g ∼ 25.6 mag, which increases to m_g ∼ 25.2 mag for the sources in the reference catalog.

All catalogs are available via a dedicated web-interface of the FDS team⁸, and are made available through the CDS. An extract of the data for the ∼1.7 million ugri matched sources in the FDS catalog is reported in Table 3 (an extract for the ∼3.1 million sources in the FDSex gri catalog is given in Table 4).

3. A preliminary map of GCs and UCD galaxies over the FDS area

One of the goals of the FDS survey is to map the distribution of GCs and UCDs in Fornax out to the virial radius. In the following sections and, in further detail, in a forthcoming dedicated paper (Cantiello et al., in prep.), we analyze and discuss the cluster-wide properties of these two classes of compact stellar systems, with more emphasis on GCs.

Unambiguously identifying GCs from purely optical photometry is unfeasible. In Cantiello et al. (2018b) we showed that also spectroscopic samples might be affected by non negligible contamination. Muñoz et al. (2014) demonstrated that optical data including the u band, combined with K-band near-IR data can dramatically reduce the contamination by fore and background sources.

Lacking a publicly available deep near-IR survey covering the FDS area, we proceeded as previously in an earlier work on GCs from the VEGAS and FDS surveys (Cantiello et al. 2015, 2018a; Cantiello 2016; D’Abrusco et al. 2016). Briefly, we identify a master catalog of GCs, and UCDs, and use the main properties of confirmed sources to constrain the mean loci of several photometric (magnitudes, colors, etc.) and morphometric (CI_n, galaxy/star classification, etc.) indicators. In the following section we discuss the procedures adopted for identifying the loci of GCs using several parameters.

3.1. GCs and UCDs Master Catalogs

We define a master catalog of GCs and UCDs, taking as reference spectroscopic and photometric studies from the literature, adopting M_g = −10.5 mag as GC/UCD separation criteria, corresponding to M_V ∼ −11 mag (∼10⁷ M_⊙), and to an apparent m_g = 21 magnitude at the adopted distance to Fornax (e.g., Mieske et al. 2004; Hilker et al. 2007). We collected photometric data from the previously mentioned ACSFCS survey (Jordán et al. 2007, 2015). The advantage of ACS with respect to other imagers is the very high resolution allowed by the space-based observations. At the distance of Fornax, GCs observed with the ACS camera appear as partially resolved sources, so their physical size can be estimated and used as a further parameter to reliably separate them from foreground stars and background galaxies. From the ACSFCS GC sample, we selected only GC candidates with a high probability p_GC of being a GC (p_GC ≥ 0.75, derived according to a maximum-likelihood estimate, Jordán et al. 2009).

The spectroscopic sample is a combination of Pota et al. (2018) and Schuberth et al. (2010) datasets. By matching the spectroscopic and photometric catalogs –cleaned up by the common sources– with our FDS ugri catalog, we obtained a list of ∼3.250 GCs. We completed our master catalogs of reference compact stellar systems with 68 bright sources in Fornax, confirmed UCD compiled from the available spectroscopic and photometric literature for this class of objects in Fornax. The GC and UCD master catalogs are given in the Tables 5 and 6.

The upper panels in Fig. 8 shows the same color-color diagrams as in Fig. 7 with a zoom over the color-color region of GCs and UCDs. The contour levels of sources from the master catalog are reported with thick dark-blue lines (we adopt linear spacing for contour levels). In the figure we also report the SPoT simple stellar population models (Brocato et al. 1999; Cantiello et al. 2003; Raimondo et al. 2005), for an age range of 4–14 Gyr and metallicity [Fe/H] = −1.3 to 0.4 dex. The consistency between the empirical loci of GCs and stellar population models for the typical age and metallicity ranges of GCs, provides further independent support to the reliability of the calibration approach adopted. In the (u − r)–(g − i) plane, the most metal-rich old stellar population models do not match with the observed GC distribution. One possible explanation is the combination of two effects: the small number of observed old GCs with such high metallicity (age ≥10 Gyr, [Fe/H] = 0.4, more than twice solar metallicity) and, consequently, the uncertainties of stellar population models is this regime.

Fig. 8. Upper panels: color–color Hess diagrams for the sample of sources with ugri photometry, over the color interval expected for GCs and UCDs. The dark-blue lines show the linear spaced contour levels of sources in the master GC catalog. Filled squares show the integrated colors from the SPoT stellar population synthesis code. White, light-gray, black, and dark-gray symbols indicate metallicity [Fe/H] = [−1.3, −0.7, 0.0, 0.4], respectively; symbols size scales with increasing model age, ranging between 4 and 14 Gyr, with 2 Gyr step. Same metallicity models are connected with dashed lines. Left color-color panel: we also draw with dotted lines the color intervals of GCs assuming ±3 − rmsMAD with respect to the median values in Table 7. Middle and lower panels: (g − i) and (u − r) color histograms, respectively, for the master spectroscopic (left panel, green histogram), photometric (middle, yellow), and combined (right, blue) GC catalogs. In the third histograms the data of UCDs are also shown (orange), expanded by a factor of five for sake of clarity. Only sources brighter than mg = 23.5 mag are considered.

The middle and lower panels of the figure also show the (g − i) and (u − r) color histograms for the photometric, spectroscopic and combined samples, for sources brighter than m_g = 23.5 mag. The asymmetric appearance of the color distribution is a consequence of the well-known color bimodality of GC systems in some filters (Ashman & Zepf 1992; Yoon et al. 2006; Blakeslee et al. 2010; Usher et al. 2012; Cantiello et al. 2014), here smoothed as the GC sample is a combination of GCs around ∼30 galaxies in Fornax, each one with different morphological types and magnitudes, hence with different properties in terms of GCs color peaks (Peng et al. 2006).

3.2. GCs and UCDs Selection by shape and photometric properties

At the assumed distance of Fornax, our best resolution for (e.g., field FDS#1 a-stack) corresponds to a physical size of ∼68 pc. Using specific analysis tools (e.g., Baolab, Larsen 1999), sources down to ∼FWHM/10, ∼7 pc for us, are marginally resolved, and can be analyzed and identified as slightly resolved sources. Typical GC half light radii of 2–4 pc are found in Fornax GCs from high-resolution ACS data (Jordán et al. 2009; Masters et al. 2010; Puzia et al. 2014). Using as reference the catalog of Fornax GC candidates by Jordán et al. (2015), ∼0.5% of the best sample (p_GC ≥ 0.75) has an half light radius r_h ≥ 7 pc estimated in both g and z bands. Hence, even at the best resolution, we can assume the largest fraction of GCs in our catalogs are indistinguishable from point like sources.

To identify compact stellar systems we adopted a procedure similar to our previous works (Cantiello et al. 2018a,b). We relied on several indicators of compactness derived from the multi-band a-stacks, as on such frames we have the lowest field-to-field variation, and, by construction, the best seeing over the entire FDS and FDSex areas. As in previous works, we combined the selection based on CI_n to other morphometric indicators from DAOphot and SExtractor (elongation, flux radius, FWHM, class star, sharpness). This refines and further cleans the final sample of compact sources by the possible outliers not identified by using the sole CI_n, or by any other single indicator.

A comparison of the CI_n distribution for the full ugri sample and for the GCs in the master catalog is shown in Fig. 9 (upper left panel). From the comparison with the reference sample (dark contour levels in the panel) we find that the GC locus extends over the CI_n ∼ 1 line, with a tail toward larger CI_n values at fainter m_g magnitudes. UCDs are also reported in the figure, with black filled dots, and show small but noticeable offsets with respect to the median properties of confirmed GCs, in particular for the size-dependent parameters (like flux radius and FWHM). Such an effect depends on the evidence that UCDs can have effective radii a factor of several times larger than GCs (Mieske et al. 2008; Misgeld & Hilker 2011), i.e. they appear resolved, or slightly resolved, in our multi-band best seeing image stacks.

Fig. 9. Hess diagrams of several morphometric and photometric indicators used to select GC candidates, overlaid with the contour levels of GCs in the master reference catalog, and UCDs (black circles).

In Fig. 9, we also show some of the other indicators used to identify GCs, together with UCDs and contour levels of the master catalog for the appropriate diagram. To define the best GC selection intervals for each indicator, we analyzed the master catalog using GCs brighter than m_g = 23.5, and derived the median and the rms_MAD for each indicator. The results are reported in Table 7. In the table we also show the median properties for the reference sample of 68 UCDs.

In addition to morphology, we refine the catalog of candidate compact sources by their photometric characteristics: the shape of the GCLF (or the magnitude interval for known UCDs), the color intervals and the errors on colors.

In our previous works, which have mostly been focused on NGC 1399, we adopted as bright magnitude cut to the GCLF the magnitude 3σ_GCLF above the turn-over of this bright cD galaxy at the photo-center of Fornax. The Fornax cluster, with an estimated total line of sight depth of ∼2 Mpc (Blakeslee et al. 2009), has member galaxies located at different physical distances. Adopting the ACSFCS results, the median g-band GCLF turn-over magnitude and σ_GCLF values are mag and σ_GCLF = 0.94 ± 0.11 mag (Jordán et al. 2007). A 3σ_GCLF cut above the median TOM corresponds to m_g ∼ 21.2 mag. For a rough estimate of the number of GCs lost with such bright cut level, we again take as reference the ACSFCS full list of GCs hosted by 43 Fornax galaxies (Jordán et al. 2015). The list contains 53 GCs brighter than m_g = 21.2 mag (∼0.8% of the sample⁹). Hence, in what follows we assume m_g = 21 mag as bright cut of the GCLF, which includes 99.5% of the likely GCs sample in the ACSFCS sample. The bright cut is needed for having a sample of GC candidates with lower stellar contamination, at the cost of an expected minimal impact on the GC population. We will in any case also analyze candidates within 19.0 ≤ m_g ≤ 21.0 mag, the magnitude interval corresponding to UCDs in Fornax (Mieske et al. 2012). These systems share many characteristics with GCs but, as mentioned above, have larger effective radii than GCs (see Fig. 9).

As a maximum color uncertainty, we chose Δ(g − i)_max = 0.15 and Δ(u − r)_max = 0.3, corresponding approximately to half of the separation between the blue and red peaks of the GCs color sub-populations host in typical bright galaxies (Cantiello et al. 2018a).

Thanks to the multiple color coverage, the selection of candidates can be improved using color-color criteria, rather than flat single-color ranges. The contour levels in the color-color diagrams of Fig. 8 reveal the relatively narrow color-color loci of GCs. A simple color-color selection box (e.g., black dotted lines in the upper left panel of the figure) would imply a trivial contamination from either stars or background objects. Instead, we proceed by inspecting in the color-color planes all sources satisfying the morpho-photometric parameters identified above. Finally, only the sources inside the color-color contours of the reference sample are identified as candidates and used for further analysis (see next section).

In summary, to identify the least contaminated and most complete possible GCs (and UCDs) catalog from our photometry, we adopted a three step strategy. First, we generated a master GCs (and UCDs) catalog using confirmed sources in the literature. From the GCs catalog we cut out all sources fainter than m_g = 23.5 mag, to better identify the morpho-photometric loci of GCs; the cut is adopted only for the reference catalog, for the GC identification and analysis on the FDS catalogs, we adapted a ∼1 mag fainter limiting magnitude to increase the sample of GC candidates (see below). Second, we used the control parameters shown in Fig. 9 and the properties of the master catalogs to define the best intervals for GCs and UCDs selection. These selection criteria are then independently applied to the FDS and FDSex catalogs. For some parameters, we adopted as confidence intervals the ranges from the master catalogs, using the median ± N × rms_MAD, with N = 4/2 for GCs/UCDs respectively (median and rms from Table 7); for the GCLF, colors, and color errors, we proceeded as described above. The complete list of parameters, together with the used ranges, is reported in Table 8. Third, the sample of compact sources after the previous steps was inspected in the color-color plane to further narrow down the contamination using the contour levels derived from the master catalog.

3.3. Surface distribution of compact sources over the FDS area

The analysis of the GCs over the FDS and FDSex area, together with the comparison with similar datasets, will be presented in more detail in a forthcoming paper. In the following, we show a preliminary determination of GCs and UCDs surface density maps as an example use of the FDS catalogs, based on the source selection strategies described in the previous section; in Sect. 4, we also show an example of use of the catalogs for the study of background galaxies.

3.3.1. Globular clusters and UCDs distribution maps

Using the identification scheme described above, we inspect the GC distribution maps over the FDS and FDSex areas using as reference the ugri and gri selections, respectively.

GC candidates are derived by cross-matching the color-color regions of pre-selected GC candidates (Table 8), with the color-color loci of GCs identified in the master sample. Candidates falling in the contour levels of higher GCs density in the two-color diagram have higher likelihood of being true GCs. However, the narrow color-color range also implies lower completeness. In what follows, then, we analyze the GC density maps for candidates over different color-color contour levels.

Figure 10 shows the two-dimensional projected distribution over the ∼21 sq. degree area of FDS. In the left panels of the figure we plot the color-color Hess diagrams of all sources identified with the selection criteria in Table 8, overplotting the contour levels of the GCs in the master sample. Even after all morpho-photometric cleaning of the sample (except for the color-color selection), a substantial fraction of selected candidates lies outside the expected GCs color-color region identified by the contour levels in the panel.

Fig. 10. Surface density maps of GC candidates over the FDS area. Upper left panel: color-color Hess diagram for GC candidates selected using the parameters in Table 8. The contour lines refer to the master GC sample. All GC candidates in the color-color contour level shown with a thick magenta solid line (also evidenced with a gray shaded area) are used for the density maps in the middle and center panels. Upper middle panel: density map of the GC candidates within the shaded area highlighted in the left panel. The density is in number of candidates per square arcmin. East is left, north is up. The light green line shows the FDS footprint; filled green dots mark the limits of single pointings; five pointed stars mark stars with mV ≤ 7 mag; yellow squares show galaxies bcenterer than BT = 16 mag, with symbol size scaled to galaxy total magnitude; NGC 1399 is also marked with a magenta empty square. Upper center panel: as upper middle panel, except that all reference sources and lines are not plotted to highlight the GC structures in the area. Second to fourth row of panels: as upper row, but for the other narrower contour levels of the color–color diagram, as evidenced with the magenta contour in the first column of panels. From upper to lower panels: the number of GC candidates within the color–color region identified with magenta contour level is: 5.650, 3.650, 2.170, and 900, respectively.

The middle and center panels of Fig. 10 show the maps of GCs identified adding also the color-color contour level selection, that is, of all sources falling in the contour levels marked in the left panels of Fig. 10. Each row of panels in the figure refers to a different contour level, indicated by the thick magenta contour in the left panel. Again, the inner contours pinpoint regions with higher GCs density in the color-color diagram, thus the level of contamination from non-GCs decreases in the maps from the upper to lower panels in Fig. 10; vice-versa, because of the smaller color-color intervals, lower panels suffer due to higher incompleteness fractions. In particular, the lowermost panel is limited to a blue color-color region, hence mostly representative on the blue-GCs sub-population, also discussed below.

We calculate the smooth density maps using non-parametric kernel density estimates based on FFT convolution¹⁰. After various tests, we adopted a grid mesh size of ∼0.1′ spacing, smoothed with an Epanechnikov kernel, with kernel bandwidth¹¹ five times the grid size.

Although obvious differences appear between GC maps drawn from the diverse color-color contour levels, there are several recurrent patterns appearing at various levels of selection, that is at different levels of GC contamination and incompleteness. The recurrence of the sub-structures over various GC color-color contours supports the reality of the sub-structure itself. Some of these patterns were also discussed in our works (D’Abrusco et al. 2016; Cantiello et al. 2018a), over a smaller survey area and using partially different data and algorithms; yet, here we observe several new features, that are possible extensions to those described previously.

Central over-density. For sake of clarity, in Fig. 11, we plot the density map relative to the third contour plot (third row in Fig. 10). The peanut shaped distribution of GCs, elongated in the E-W direction of the cluster, with a marked peak on NGC 1399, was already found in our studies relying on data of the central FDS area, within 52.5 ≤ RA (deg) ≤ 56.5 and −37 ≤ Dec (deg) ≤  − 35 (a total of ∼7.5 sq. degrees).

Fig. 11. Single-panel view of the 2-D GC surface distribution. Iso-density contours and symbols are the same as in Fig. 10 (third contour level plots). Light blue arrows and labels indicate the GC overdensities discussed in the text. The blue dashed line shows the ∼10° tilt in the direction of NGC 1336 (“E” label in the figure).

In the new dataset, covering about four times the area previously inspected, we find a ∼10 deg tilt of the position angle for the broad distribution of inter-galactic GC candidates, tilting in the direction of NGC 1336 (the tilt direction is also indicated with a blue dashed line in Fig. 11). The length of the last isodensity contour is a = 2.6 ± 0.2 deg (or 920 ± 60 kpc), obtained combining the sizes from the four maps in Fig. 10. The width of the distribution is of b = 0.89 ± 0.03 deg (or 310 ± 10 kpc), implying an ellipticity ϵ = 1 − b/a ∼ 0.65, slightly larger than what was previously found on smaller scales (Kim et al. 2013; Cantiello et al. 2018a).

F & G features. In the distribution, aside from the obvious case of NGC 1399 and its fainter close companions, we observe several regions of marked over-density in correspondence with bcenter galaxies or pair of galaxies: NGC 1427, NGC 1374/1375, NGC 1351, all with B_T ≤ 12 mag, and of NGC 1336, which is ∼1.5 mag fainter than the others. The GCs peaks on these regions were already commented in our previous works. However, thanks to the larger area analyzed and the different detection strategy the new photometric sample reaches ∼1 mag deeper u-band, we now find that such structures are connected and extend to larger clustercentric radii. The F and G features described in D’Abrusco et al. (2016) (arrows in Fig. 11) extend ∼1.5 degrees (∼0.5 Mpc) South-West and North-East of the cluster core, respectively. These substructures do not cross any galaxy bcenterer than B_T = 16 mag, both overlap a handful of galaxies with 16 ≤ B_T ≤ 18 mag (absolute magnitude −15.5 ≤ M_B, tot (mag) ≤  − 13.5), and a dozen of fainter galaxies, down to B_T ∼ 20 mag (M_B, tot = −11.5 mag). The F extension, points toward a group of five galaxies with magnitudes B_T between 13.5 and 16 mag, dominated by ESO 358-050, where no GC structure or overdensity is noticeable in any of the GC color-contour selections.

The level of persistence of the F and G structures changes with the selection contours. To estimate the level of significance of both these overdensities we proceed as follows. First, taking as reference the third contour level in Fig. 10, we count the number of GC candidates in the F and G feature density contours (N_X, with X referred to the F or G region). Then, to define a background level, we move the same density contours around the FDS area, avoiding the central overdensity and the regions with galaxies bcenterer than B_T ∼ 16, and count the number of candidates in such regions. For each feature, we identified seven independent regions for background estimation over the survey area; then we used the median and rms_MAD of the GC number counts in the seven regions (N_X, back., rms_X, back.) to quantify the F and G overdensity ratio as follows:

By definition, Σ[(in − out)/err] quantifies the ratio between the difference of counts in and out the X feature, and the squared sum of the standard deviation of both counts, assuming a poissonian fluctuation for N_X (). We obtain Σ[(in − out)/err]∼4.2, for F and ∼4.4 for G, meaning that the GC candidates overdensity, with respect to the diffuse background GCs component, is at least factor of four larger than the estimated total expected counts fluctuation in both regions. A similar result, albeit for smaller regions, with a different (shallower) sources catalog and with independent algorithms, was found by D’Abrusco et al. (2016).

The F is more evident in the wider color-contours selections (upper two panels in Fig. 10), which also include the red GCs that are mostly expected to be closely bound to the galaxies; because of the wider selection intervals, this feature is also likely to have higher fore or back-ground contamination. The G structure, instead, appears more connected to the blue GC population (lowermost panel in the figure); the properties of such coherent structure extending over cluster scale, over an area devoid of bcenter galaxies and composed mostly of blue GCs –the GC sub-population typically found in the outer galactic regions– suggest its inter-galactic nature. We speculate that the G feature might be connected with NGC 1404, as a stream of blue GCs possibly leading or tailing from the galaxy; the galaxy has an overall z-band specific frequency S_N, z = 0.30 ± 0.00, and within one effective radius S_N, z, In = 0.12 ± 0.01 (Liu et al. 2019). The whole median of the ACSFCS sample is ⟨S_N, z⟩=0.82 ± 0.37, or 0.93 ± 0.26 if limited to the five bcenterest galaxies in the main Fornax cluster after excluding NGC 1399 and NGC 1404 itself¹²; for the S_N, z, In from the combined Fornax and Virgo cluster sample (Table 4 in Liu et al. 2019), and limited to galaxies bcenterer than M_z ∼ −20.7 mag, we obtain ⟨S_N, z, In⟩=0.32 ± 0.18. Hence, in all cases NGC 1404 is a noteworthy case of bcenter galaxy with a GC population consistently lower than average. Bekki et al. (2003) have a dynamical model for the GCs system of NGC 1404, explaining its low specific frequency as an effect of the tidal stripping of GCs by the gravitational field around cluster core, dominated by NGC 1399. The authors find that at given models input conditions (highly eccentric orbit, initial scale-length of the GCs system twice as large as the galaxy effective radius), NGC 1404 GCs population can be reduced through stripping to the presently observed value. One of the observable characteristics predicted by Bekki et al. is the formation of an elongated or flattened tidal stream of GCs.

Furthermore, the complex structure of the Fornax X-ray halo (Paolillo et al. 2002; Su et al. 2017) has been explained by Sheardown et al. (2018) using hydrodynamics simulations, by the orbital motion of NGC 1404 within the cluster, assuming that the galaxy is at its second or third passage through the cluster center.

NGC 1336. The new photometry confirms the peculiarity of NGC 1336 with respect to the rest of the cluster: we find its GCs overdensity (E feature in D’Abrusco et al. 2016) isolated with respect to the rest of the cluster-wide GCs system. The distinctiveness of NGC 1336 is also discussed by Liu et al. (2019), who find that it has the second highest GC specific frequency, after NGC 1399, and the largest 3D clustercentric distance in the ACSFCS sample. The relative isolation of the galaxy from the Fornax core, at ∼2 times of the cluster virial radius, also supported by the lack of GC streams toward the core, might strengthen the hypothesis by Liu et al. that it is an infalling central galaxy with a higher total mass-to-light ratio, resembling the behavior of the most massive ETGs. Its GC system has possibly experienced fewer external disruption processes, and the GCs may have a higher survival efficiency. The presence of two kinematically decoupled cores (Fahrion et al. 2019b), most probably evidencing a major merger that has altered the structure of NGC 1336 significantly, might further support such hypothesis.

The C feature. A further structure, labeled C in D’Abrusco et al., ranges from NGC 1380 North-West in the direction of the ringed barred spiral NGC 1350. The feature appears less coherently connected than the F and G in the maps of Fig. 10, and it crosses four galaxies with B_T ≤ 16, thus, it might be the result of the projected superposition of several adjacent GC systems, rather than an intra-cluster GCs structure.

Blue and red GCs, foreground stars. We also plot the map of blue and red GC candidates in Fig. 12, using the color-contours shown in the left panels. To improve the blue and red GCs separation, taking advantage of the availability of two colors, the separation between red and blue GC is taken from a linear fit to the (u − r)–(g − i) sequence of the master GC sample, then taking the blue/red separation from the dip in the distribution projected along this axis, a procedure we already used in Angora et al. (2019, see their Fig. 7, upper panel). The blue and red surface density maps show the property already anticipated above of red GCs being concentrated on galaxies, especially on bcenter ellipticals, and blue GCs covering a wider area, including the intra-cluster regions.

Fig. 12. Two-dimensional density maps of blue (upper panels) and red (lower panels) GC candidates. Symbols are the same as in Fig. 10.

For comparison with the previous maps, Fig. 13 shows the stellar density map, where stars are identified as the bcenter sources 16 ≤ m_g (mag) ≤ 20.5, with the same photo-morphometric properties of GCs (Table 8) except no color selection is applied. The stellar map shows both the lack of any obvious structure over the field, and the large contamination from MW stars: the map, limited to the bcenterest part of the field MW stellar population, is derived from ∼23.000 stars, versus the ∼5.600/900 GCs used for the GC maps in Fig. 10, and the ∼2.200/1200 blue/red GCs selected for the maps in Fig. 12.

Fig. 13. Surface density maps of bcenter stars. Symbols are the same as in Fig. 10.

UCD galaxies. A further map from the ugri catalog is shown in Fig. 14, with the UCD surface density distribution over the FDS area, derived using the selection parameters for UCDs, reported in Table 8, and the color contours of known UCDs from the reference sample (magenta solid lines in the figure). Unsurprisingly, the surface density maps show the concentration of UCDs rises around the central square degree area of NGC 1399. The map is mostly shown for completeness, as number of UCDs is known to be small, so even a small contamination can significantly alter the analysis. With our selection we identify 160 sources, which probably include a substantial fraction of contaminating stars, especially in the bcenterest magnitude bin (19 ≤ m_g ≤ 20), and bcenter GCs with morphological parameters consistent with the UCDs. Inspecting separately the maps of bcenter or faint UCDs candidates, adopting m_g = 20 mag as the separation limit, we observe that the map for the faint magnitude bin – m_g = 20–21 containing 105 candidates at the given selection criteria doesn’t change notably with respect to Fig. 14 and it shows an elongated density structure with a peak close to the cluster core, along with two secondary maxima at [RA, Dec] = [53.7, −37.6] and [52.3, −33.5]. The map of the bcenter component – m_g = 19–20, 55 candidates – does not show any noteworthy pattern, with sources appearing evenly distributed in the region, a behavior suggesting large contamination from MW stars in this magnitude range. The study of the UCD distribution over the area requires a dedicated analysis to characterize and identify all the selected UCD candidates, which is beyond the scopes of this study, and will be addressed in a forthcoming work, also using near-IR photometry (Saifollahi et al., in prep.).

Fig. 14. Surface density maps of UCD candidates. Symbols are the same as in Fig. 10.

In conclusion, it is worth highlighting that all the sub-structures described in this section are relatively insensitive to the main parameters chosen to identify GC or UCD candidates, and to the details of the algorithms used to derive the maps themselves, except minor details which leave unaltered the general presentation above.

3.3.2. GCs distribution maps over the FDSex area

The lack of u-band photometry over the FDSex area implies that any sample of compact sources selected in the area using the same procedures adopted in the previous section, yet based only on gri photometry, is more contaminated. In Fig. 15, we plot the contour levels of the master GC sample (blue lines and shaded area), and the contour levels of compact (green color, C.I_n ∼ 1) and extended (red colors, CI_n ≥ 1.3) sources, all bcenterer than m_g = 22.5 mag, using the ugri catalog. The bcenter magnitude cut is adopted to reduce the scatter due to increased photometric errors at fainter magnitudes. The diagrams show that the sequence of GCs/UCDs/stars in the (g − i)–(g − r) matches with the sequence of extended objects, while in the (g − i)–(u − r) diagram the degeneracy is less dramatic, maximizing the efficiency of the separation of compact extended sources.

Fig. 15. Color–color contour plots of the GC master catalog (blue contours and shaded area), of point-like sources (green) and of extended sources (red).

To obtain a rough estimate of the increase of contamination due to the lack of u-band photometry we proceed as follows. Using the ugri catalog in the FDS area, we adopt the GC selection scheme described in the previous section but use the (g − i)–(g − r) color combination, instead of the (g − i)–(u − r), for the selections on the color-color plane. By comparing the number of GC candidates identified using the (g − i)–(u − r) color–color, , with the number of candidates identified using (g − i)–(g − r) color-color, , we find . Therefore, this single change in the criteria for GC selections implies the number of sources identified as GC candidates is nearly doubled over the FDS area. Such increase is not spatially uniform: it is close to ∼80% in background regions, that is, far from bcenter galaxies and their host GC system, and drops to ∼15% around bcenter galaxies. This difference shows that GC selection in the central cluster area, where GCs have a high surface density, is already quite efficient with a 3-band combination. In contrast, the addition of the u-band makes a significant difference in GC selection in the outer parts of the cluster where the fractional background contamination, mostly due to MW stars, is higher.

In spite of the higher level of contamination, the FDSex gri-band catalog also includes the area of NGC 1316, Fornax A, the bcenterest galaxy in the cluster in optical bands, a peculiar giant elliptical, suggested being in its second stage of mass assembly (Iodice et al. 2017b). It is then of particular interest to show here, for the first time, the global properties of the GCs over such wide area. We should, however, be aware that NGC 1316 is known to contain relatively young GCs (∼2−3 Gyr, e.g. Gómez et al. 2001; Goudfrooij et al. 2001; Sesto et al. 2017), which are not part of our reference sample. Young GCs are in general bluer and bcenterer than equally massive old GCs; hence, we bear in mind that our selection is intrinsically biased toward old GCs.

Using the same procedures described in the previous sections, except that (g − r) is used instead of (u − r), we analyze the surface distribution maps over the 27 sq. degrees of the FDSex area. For sake of clarity, in Fig. 16 we only show the second color-color density contour, corresponding to the iso-density contour level of 15 GCs from the master catalog. In the panels of the figure, no obvious GC substructure appears bridging the core of the main Fornax cluster to the Fornax A sub-group. The two bcenterest galaxies, NGC 1399 and NGC 1316, are ∼3.6deg apart (∼1.3 Mpc in projection) and the density map of the ∼10.200 GCs selected does not reveal any hint of residual GC tails along the direction connecting the bcenter ellipticals, with the possible only exception of the East-West elongation of GCs around the cluster core still visible in the gri map, although with less details compared with the ugri maps.

Fig. 16. Same as in Fig. 10, except that over the FDSex area, and the (g − i)–(g − r) color-color diagram is used for GCs selection. The position of NGC 1316 is shown with light-blue empty triangle in the middle panel.

The higher level of contamination of the gri maps appears in some spurious features. Figure 16 shows a structure around the area of coordinates [RA = 52 deg, Dec = −34 deg], characterized by nearly the same geometric appearance of the FDS fields #14, #19 and #31. Such structure is completely unseen in the ugri maps which also cover the area; inspecting the three FDS fields we find slightly deeper limiting magnitudes and slightly poorer source compactness relative to the neighboring fields: combined, two effects generate larger number of detections with poorer morphologic characterization, hence a higher fraction of GCs contamination.

To have a less contaminated sample, we narrowed the sample of GC candidates by using a bcenterer magnitude cut, more stringent ranges on the various morphological parameters in Table 8, and narrower color-color regions. Using narrower selections, the spurious structure around the fields FDS#14/19/31 disappears. Nevertheless, no matter how much the GC sample is narrowed with more strict selections, no GC substructure emerges along the NGC 1316/NGC 1399 direction.

By counting the number of GCs candidates within a given radius centered on each of the two bcenter galaxies within the respective environments, we find that the number of GCs around NGC 1399 outnumbers NGC 1316 by a factor of 4–4.7 at galactocentric radii R_gal of ∼6′ and ∼24′, and by a factor of ∼3 out to R_gal ∼ 40′. Figure 17 shows the number ratio versus galactocentric distance for GCs candidates (black line in the figure), for galaxies bcenterer than a given limit (as labeled in the figure), and the flux ratio of the r-band integrated magnitudes of the two galaxies (from Iodice et al. 2016, 2017b).

Fig. 17. Number ratio of the total number of sources around NGC 1399 and NGC 1316 within a given galactocentric radius in the respective environment: . The number ratio for GCs is shown with a black solid line; number ratios for galaxies at a given bcenter magnitude cut are also shown, and labeled. The r-band flux ratio between the two galaxies within Rgal is shown with light-blue solid line and pentagons.

The median for galaxies in the range of 6′−24′ is ∼2.0 with rms_MAD = 0.3. Assuming a nearly uniform contamination of the FDSex catalogs around the two regions, we estimate the overdensity of GCs around NGC 1399 compared to NGC 1316 () is a factor of ∼2 larger than the overdensity of galaxies in the magnitude range 11.5 ≤ B_T (mag) ≤ 16.5 (−20 ≤ M_B, tot (mag) ≤ −15). Hence, even accounting for the larger density of bcenter and faint galaxies of all morphological types, the population of GCs is considerably larger in the region of 6′ ≤ R_gal ≤ 24′ around NGC 1399 compared with NGC 1316, and mainly composed of blue GCs.

This overpopulation of GCs is likely associated with the intra-cluster GCs component; on the contrary, the relative GCs under-density around Fornax A, and the lack of any major accretion events of NGC 1316 that could have significantly increased the specific frequency of blue GCs, is possibly at the basis of the lack of any significant GC substructure. Furthermore, as expected from the known factor of ∼2 higher total magnitude of NGC 1316 compared to NGC 1399, the r-band flux ratio between the two ellipticals is ∼0.4 ± 0.1 (light–blue line in Fig. 17), a factor of 8−10 lower than the GCs count ratio.

Figure 17 also shows some other features : (a) the GCs and bcenter galaxies with B_T ≤ 11.5 mag and B_T ≤ 13.5 mag (M_B, tot = −20 and −18 mag, respectively) have at R_gal ≥ 30′, while for the fainter galaxy bin limits we find ; (b) the nearly flat GCs N_ratio within 9 ≤ R_gal (′) ≤ 20, which assumes a value of 4.66 ± 0.04. A more detailed analysis of such properties combined with the data in other galaxy clusters is in progress (Cantiello et al., in prep.).

4. FDS catalogs of background sources and related science

The depth and spatial resolution of the FDS images, together with ancillary data from other spectral ranges available in this field, provide the opportunity to study the stellar populations and structural properties of galaxies beyond the cluster, as well as to discover rare astrophysical objects, such as compact massive galaxies and strong gravitational lenses (e.g., Tortora et al. 2018; Petrillo et al. 2017). The FDS image quality is similar to the one of the KiDS survey (Kuijken et al. 2019), since the longer exposures in FDS images are balanced by a slightly poorer seeing (r-band FWHM of for FDS, vs. in KiDS). The limiting magnitudes in the two surveys are quite similar, but the FDS is deeper in the i-band.

Taking advantage of the FDS data, we aim at determining photometric redshifts, stellar masses, galaxy classifications, and structural parameters of thousands of background galaxies. The goal is to provide a complete characterization of the background galaxy population over the area meant for investigating the evolution of the structural and stellar properties of galaxies as a function of redshift and mass. The tools for deriving all required quantities are already available and well-tested among our team (e.g., La Barbera et al. 2008; Cavuoti et al. 2017; Roy et al. 2018).

As a first test on the background galaxy population in FDS, we run a code to find galaxy-galaxy lens candidates. The code uses a machine learning classification method based on convolutional neural networks (CNNs), and was already applied to the KiDS survey (de Jong et al. 2017; Petrillo et al. 2019). We performed the run of CNN with a network trained on a large sample of r-band (or combined g-r-i) KiDS images, to the equivalent FDS images. Although the network is not customized and trained on FDS images, KiDS and FDS are based on data from the same telescope and camera, and, as mentioned above, are comparable in both FWHM and depth. Therefore, this is a valid approach to search gravitational lenses in FDS. In fact, we have already discovered several gravitational lens candidates in the FDS fields; two of them with FDS catalogue ID and coordinates: FDSJ032720.32-365821.81 at [51.834682; −36.972725] and FDSJ034739.60-352516.23 at [56.91502, −35.421176], which are presented in Fig. 18 as an example of the potential of this approach.

Fig. 18. Two example lens candidates found in the FDS fields applying the CNN code; the image cutout have side. Left: FDSJ032720.32-365821.81. Center: FDSJ034739.60-352516.23.

5. Conclusions

In this paper, we present the photometric and morphometric catalog of compact and slightly extended sources in the Fornax galaxy cluster, derived with VST observations within the FDS survey over an area of ∼21 square degrees in ugri-bands, and in gri-bands for a total of ∼27 square degrees.

The ugri data of FDS cover the main body of Fornax, centered on NGC 1399, and extend out to ∼1 Mpc, the virial radius of the cluster. The gri coverage, FDSex, extends to the South-West region of the Fornax A sub-cluster with its bcenterest galaxy, NGC 1316.

Because of the large FWHM variation from field to field, to improve the uniformity of sources detection and their morphological characterization, we derived a master-detection frame by coadding all gri single exposures with ; starting from a median FWHM ranging from to with rms within for the various bands, adopting the multi-band stacking procedure we ended up with a master-detection frame with a median FWHM of , ∼15% improvement over the median FWHM of the highest resolution imaging (r-band), and a factor of ∼2.5 lower rms.

We calibrated the photometry using a two-step procedure, to reduce the effect of the independent calibration of the FDS fields, which generate a non-negligible photometric offset between neighboring fields. The first calibration step follows the standard calibration plan of VST frames. As a second step, we used the APASS photometry to derive a matrix to match the full FDS catalog to a unique reference. With this approach, the photometric offset between fields becomes negligible, and the re-calibrated photometry shows a general good match to existing literature data from SKyMapper, from the HST/ACSFCS survey and to predictions from stellar population synthesis models.

The catalogs are available through the project web pages, and will also be available on CDS. In the catalogs we provide the position, the photometry and the morphometry for 1.7 million sources with ugri detections, and for 3.1 million sources with gri data. As a preliminary use of the catalogs, we analysed the 2-D distribution of compact stellar systems in the area, with particular attention given to GCs.

With the FDS instrumental setup and at the distance of Fornax, GCs are by all means point-like sources, except for a possible fraction of ≲0.5% of the population. Hence, GCs can be identified by their compactness.

To obtain the least contaminated GCs sample, we selected a number of morpho-photometric features and analysed them over a reference catalog of confirmed GCs and UCDs in Fornax. Such catalog is build by cross-matching the FDS catalogs with available spectroscopic and photometric datasets of confirmed GC/UCD. The reference catalog is then used to define the GCs loci in the parameter space, for the chosen photometric and morphometric parameters.

The GCs maps over the FDS area confirm the results of previous studies, about the presence of a large inter-galactic GC population around the main body of the cluster, centered on NGC 1399, stretched along the East-West direction. Here we find a small tilt of the distribution in the direction of NGC 1336 by ∼10deg. The distribution appears to extend over ∼1 Mpc from side to side, highly flattened, with an ellipticity of ∼0.65. In addition to our previous results, we find that one of the features already discussed, which extends from the main cluster body to the South-West direction, might be a tail of relatively blue GCs from NGC 1404, a bcenter galaxy close to the cluster core and with a peculiarly poor GCs population.

Of the GCs features already commented in the past, we here highlight the case of NGC 1336, which we confirm to be relatively isolated from the cluster, and with a high specific frequency of GCs; this might support the hypothesis that it is an infalling massive galaxy, with a GCs system that possibly experienced only few disruption processes. We also inspected the blue/red GCs maps, and confirm the known property of blue GCs residing in the wider cluster area, and red GCs being more concentrated on massive galaxies.

Systems selected to fit the color-magnitude range of spectroscopically confirmed UCDs show a substantial overdensity in the central cluster. The 160 UCD candidates are about three times more than the currently known UCDs in Fornax and would require spectroscopic follow up to learn more about their nature.

We also derived the GCs maps over the FDSex area, which has the disadvantage of suffering for larger contamination because of the lack of u-band over the NGC 1316 area, but has the advantage of covering this bcenterest cluster galaxy. With the caveat that the gri catalogs do not allow the detailed analysis allowed over the FDS area, despite our attempts to obtain a cleaner GC candidates sample, we do not find significative GCs structures along the NGC 1399–NGC 1316 direction, which extends over a projected distance of ∼1.3 Mpc. This might be due to the lower efficiency of the GC identification. However, assuming similar contamination of the gri catalogs over the NGC 1399 and NGC 1316, we find that the GC population of the former outnumbers the second by a factor of ∼4, and by about a factor of ∼10 when normalized to galaxy luminosity, within a galactocentric range of 6′−24′, and remains a factor of ∼3 higher than NGC 1316 at larger galactocentric radii, out to ∼40′. Hence, the “contrast” of the GC populations towards NGC 1316 might be too low for the purpose of our study, – in spite of its luminosity twice larger than NGC 1399 – and might explain the difficulty in finding GCs sub-structures, which intrinsically need a large number of candidates over a given region to be identified. The rich intra-cluster GCs population around NGC 1399 does not seem to be matched by a similarly rich system around NGC 1316, the bcenterest galaxy of the Fornax A sub-cluster. In spite of this, the lack of obvious GC sub-structures between these two bcenter and massive galaxies might also be consequence of the NGC 1316 sub-cluster being in its first infalling phase and evolving autonomously, a result also supported by an independent analysis of FDS data for galaxy surface bcenterness profiles and intracluster light (Iodice et al. 2017b; Raj et al. 2019). A deeper analysis of the 2-D maps and other characteristics of the GCs over the FDS and FDSex area is in progress and will be presented in a dedicated paper.

Finally, we offer an example use of the catalogs for analyzing background galaxies. Using machine-learning methods, which have already been tested on the KiDS survey with VST, we identified two lens candidates in the FDS area.

Acknowledgments

This research was made possible through the use of the AAVSO Photometric All-Sky Survey (APASS), funded by the Robert Martin Ayers Sciences Fund and NSF AST-1412587. This work is based on visitor mode observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under the following VST GTO programs: 094.B-0512(B), 094.B-0496(A), 096.B-0501(B), 096.B-0582(A). INAF authors acknowledge financial support for the VST project (P.I. P. Schipani). We acknowledge the use of data from the SKyMapper survey. The national facility capability for SkyMapper has been funded through ARC LIEF grant LE130100104 from the Australian Research Council, awarded to the University of Sydney, the Australian National University, Swinburne University of Technology, the University of Queensland, the University of Western Australia, the University of Melbourne, Curtin University of Technology, Monash University and the Australian Astronomical Observatory. SkyMapper is owned and operated by The Australian National University’s Research School of Astronomy and Astrophysics. The survey data were processed and provided by the SkyMapper Team at ANU. The SkyMapper node of the All-Sky Virtual Observatory (ASVO) is hosted at the National Computational Infrastructure (NCI). Development and support the SkyMapper node of the ASVO has been funded in part by Astronomy Australia Limited (AAL) and the Australian Government through the Commonwealth’s Education Investment Fund (EIF) and National Collaborative Research Infrastructure Strategy (NCRIS), particularly the National eResearch Collaboration Tools and Resources (NeCTAR) and the Australian National Data Service Projects (ANDS). MP acknowledges financial contribution from the agreement ASI-INAF n.2017-14-H.O. JFB acknowledges support through the RAVET project by the grant AYA2016-77237-C3-1- P from the Spanish Ministry of Science, Innovation and Universities (MCIU) and through the IAC project TRACES which is partially supported through the state budget and the regional budget of the Consejería de Economía, Industria, Comercio y Conocimiento of the Canary Islands Autonomous Community. GvdV acknowledges funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under grant agreement No 724857 (Consolidator Grant ArcheoDyn). CT acknowledges funding from the INAF PRIN-SKA 2017 program 1.05.01.88.04.

References

All Tables

All Figures

Fig. 1. Left panel: FDS footprint of the area covered by ugri photometry (green solid line), and by only gri (dashed green line). Other sources from catalogs available in the literature are also shown, as labeled. Bright galaxies from the Fornax Cluster Catalog (Ferguson 1989) are subdivided into two categories: likely members brighter than BT = 17 mag and with 17 ≤ BT (mag) ≤ 18.5 (filled gray circles and blue triangles, respectively; from Ferguson 1989, Table II). Dwarf galaxies from FDS by Venhola et al. (2018), in the magnitude range 18.5 ≤ mg (mag) ≤ 21, are indicated with red crosses. The positions of the two brightest galaxies, NGC 1316 and NGC 1399, are also shown with a filled cyan triangle and a magenta square, respectively. Orange filled or empty five-pointed stars mark those stars with mV ≤ 7/≤9 mag, respectively. Right panel: FDS and FDSex area. Green lines mark the edges of the survey, green bullets show the edges of single pointings; the ID of the field is also indicated.

In the text

	Fig. 2. Histograms of the median PSF FWHM of the FDS fields in the four available passbands, plus the multi-band a-stacks. The vertical dashed line shows the median of the ensemble.
In the text

	Fig. 3. From left to right: g, r, i-band and a-stack of a background spiral galaxy in the field FDS#5 (FCCB 1532, Ferguson 1989). Rightmost panel: derived from the combination of the sub-exposures of the first three panels, selecting only the ones with lowest atmospheric turbulence (see text).
In the text

Fig. 4. Example of the two-dimensional photometric correction maps for refining the photometry of FDS fields. The maps also include the aperture correction term. Field FDS#19 is shown: u, g, r, and i-band correction maps are plotted from upper to lower panels, respectively. For each passband, the surface correction map is shown with the same color coding and for different viewpoints in each of the three panels.

In the text

Fig. 5. Left panel: g-band magnitudes from FDS compared with magnitudes of GC candidates from the ACSFCS. Blue symbols show the full matched set, red symbols identify compact sources in our reference catalog (see text). Middle panel: as left panel, but running averages are shown, with bin size of 100/50 objects for the blue/red symbols, respectively. Right panel: as left panel, but versus (g − i) color.

In the text

Fig. 6. Normalization procedure for the CI, data for the field FDS#13 are shown. Left panel: concentration index CI = mag4 pix − mag6 pix versus the uncalibrated a-stack magnitude. For the sake of clarity, only the brightest magnitude range is shown. Blue dots refer to the full sample, red symbols to candidate compact sources used to derive the median CI factor for normalization. Right panel: same as left panel, but over a larger magnitude range and after normalization to the median CI of bright point-like sources (red dots). Point-like sources candidates are aligned along the sequence parallel to the x-axis, around CIn ∼ 1 (green dot-dashed line).

In the text

	Fig. 7. Hess color–magnitude and color-color diagrams of the full sample with ugri photometry. Extinction corrected PSF magnitudes are used in all cases.
In the text

Fig. 8. Upper panels: color–color Hess diagrams for the sample of sources with ugri photometry, over the color interval expected for GCs and UCDs. The dark-blue lines show the linear spaced contour levels of sources in the master GC catalog. Filled squares show the integrated colors from the SPoT stellar population synthesis code. White, light-gray, black, and dark-gray symbols indicate metallicity [Fe/H] = [−1.3, −0.7, 0.0, 0.4], respectively; symbols size scales with increasing model age, ranging between 4 and 14 Gyr, with 2 Gyr step. Same metallicity models are connected with dashed lines. Left color-color panel: we also draw with dotted lines the color intervals of GCs assuming ±3 − rmsMAD with respect to the median values in Table 7. Middle and lower panels: (g − i) and (u − r) color histograms, respectively, for the master spectroscopic (left panel, green histogram), photometric (middle, yellow), and combined (right, blue) GC catalogs. In the third histograms the data of UCDs are also shown (orange), expanded by a factor of five for sake of clarity. Only sources brighter than mg = 23.5 mag are considered.

In the text

	Fig. 9. Hess diagrams of several morphometric and photometric indicators used to select GC candidates, overlaid with the contour levels of GCs in the master reference catalog, and UCDs (black circles).
In the text

Fig. 10. Surface density maps of GC candidates over the FDS area. Upper left panel: color-color Hess diagram for GC candidates selected using the parameters in Table 8. The contour lines refer to the master GC sample. All GC candidates in the color-color contour level shown with a thick magenta solid line (also evidenced with a gray shaded area) are used for the density maps in the middle and center panels. Upper middle panel: density map of the GC candidates within the shaded area highlighted in the left panel. The density is in number of candidates per square arcmin. East is left, north is up. The light green line shows the FDS footprint; filled green dots mark the limits of single pointings; five pointed stars mark stars with mV ≤ 7 mag; yellow squares show galaxies bcenterer than BT = 16 mag, with symbol size scaled to galaxy total magnitude; NGC 1399 is also marked with a magenta empty square. Upper center panel: as upper middle panel, except that all reference sources and lines are not plotted to highlight the GC structures in the area. Second to fourth row of panels: as upper row, but for the other narrower contour levels of the color–color diagram, as evidenced with the magenta contour in the first column of panels. From upper to lower panels: the number of GC candidates within the color–color region identified with magenta contour level is: 5.650, 3.650, 2.170, and 900, respectively.

In the text

	Fig. 11. Single-panel view of the 2-D GC surface distribution. Iso-density contours and symbols are the same as in Fig. 10 (third contour level plots). Light blue arrows and labels indicate the GC overdensities discussed in the text. The blue dashed line shows the ∼10° tilt in the direction of NGC 1336 (“E” label in the figure).
In the text

	Fig. 12. Two-dimensional density maps of blue (upper panels) and red (lower panels) GC candidates. Symbols are the same as in Fig. 10.
In the text

	Fig. 13. Surface density maps of bcenter stars. Symbols are the same as in Fig. 10.
In the text

	Fig. 14. Surface density maps of UCD candidates. Symbols are the same as in Fig. 10.
In the text

	Fig. 15. Color–color contour plots of the GC master catalog (blue contours and shaded area), of point-like sources (green) and of extended sources (red).
In the text

	Fig. 16. Same as in Fig. 10, except that over the FDSex area, and the (g − i)–(g − r) color-color diagram is used for GCs selection. The position of NGC 1316 is shown with light-blue empty triangle in the middle panel.
In the text

Fig. 17. Number ratio of the total number of sources around NGC 1399 and NGC 1316 within a given galactocentric radius in the respective environment: . The number ratio for GCs is shown with a black solid line; number ratios for galaxies at a given bcenter magnitude cut are also shown, and labeled. The r-band flux ratio between the two galaxies within Rgal is shown with light-blue solid line and pentagons.

In the text

	Fig. 18. Two example lens candidates found in the FDS fields applying the CNN code; the image cutout have side. Left: FDSJ032720.32-365821.81. Center: FDSJ034739.60-352516.23.
In the text