Confirmation and physical characterization of the new bulge globular cluster Patchick 99 from the VVV and Gaia surveys

Globular clusters (GCs) are important tools to understand the formation and evolution of the Milky Way (MW). The known MW sample is still incomplete, so the discovery of new GC candidates and the confirmation of their nature are crucial for the census of the MW GC system. Our goal is to confirm the physical nature of two GC candidates: Patchick99 and TBJ3, located towards the Galactic bulge. We use public data in the near-IR from the VVV, VVVX and 2MASS along the with deep optical data from the Gaia DR2, in order to estimate their main physical parameters: reddening, extinction, distance, luminosity, mean cluster proper motions (PMs), size, metallicity and age. We investigate both candidates at different wavelengths. We use near-IR and optical CMDs in order to analyse Patchick99. We decontaminate CMDs following a statistical procedure and PM-selection. Reddening and extinction are derived by adopting reddening maps. Metallicity and age are evaluated by fitting stellar isochrones. Reddening and extinction are E(J-Ks)=0.12+/-0.02 mag, AKs=0.09+/-0.01 mag from the VVV data, whereas E(BP-RP)=0.21+/-0.03 mag, AG=0.68+/-0.08 mag from Gaia DR2. We estimate a distance d=6.4+/-0.2 kpc in near-IR and D=7.0+/-0.2 kpc in optical. We derive its metallicity and age fitting PARSEC isochrones, finding [Fe/H]=-0.2+/-0.2 dex and t=10+/-2 Gyr. The mean PMs for Patchick99 are pmRA=-298+/-1.74 mas/yr and pmDEC=-5.49+/-2.02 mas/yr. We confirm that it is a low-luminosity GC, with MKs=-7.0+/-0.6 mag. The radius estimation is performed building the radial density profile, finding r~10'. We recognise 7 RR Lyrae star members within 8.2 arcmin from its centre, confirming the distance found by other methods. We found that TBJ3 shows mid-IR emissions that are not present in GCs. We discard TBJ3 as GC candidate and we focus on Patchick99. We conclude that it is an old metal-rich GC, situated in the Galactic bulge.


Introduction
Globular Clusters (GCs) represent the most ancient stellar systems in the Milky Way (MW). For this reason, they are excellent tracers of the formation and chemodynamical evolution of the Galaxy, going from the outer halo to the innermost Galactic regions. In addition, they survived different dynamical processes, like dynamical friction, disk and bulge shocking, evaporation that also led some ancient GCs to destruction (Weinberg 1994). Today, we distinguish different kinds of GCs, depending on their metallicities, ages and orbits. The metal-poor GCs could be the oldest population (Barbuy et al. 2006(Barbuy et al. , 2009). More precisely, in the Galactic bulge, they may constitute the first stellar generation whose origin could coincide with the formation of the bulge itself or before the actual configuration of the bulge/bar component, examples are NGC 6522, NGC 6626 and HP 1 (Kerber et al. 2018(Kerber et al. , 2019. On the other hand, the metal-rich GCs may be the second generation in the MW (Muratov & Gnedin 2010). Furthermore, recent studies have demonstrated that some other luminous GCs may be actually Article number, page 1 of 13 arXiv:2103.03592v1 [astro-ph.GA] 5 Mar 2021 A&A proofs: manuscript no. newGCs_v28 fossil relics of accreted dwarf galaxies, such as Terzan 5 (Ferraro et al. 2009), Liller 1 (Ferraro et al. 2020), Terzan 9 (Ernandes et al. 2019) and also Omega Centauri (Lee et al. 1999;Sollima et al. 2008). Further, with the advent of Gaia Mission and deep surveys, it is now possible to associate individual GCs with specific merger events (e.g., Vasiliev 2019; Myeong et al. 2019;Massari et al. 2019).
However, it seems that the frequency of the MW's GCs is lower than the Andromeda's GC system (van den Bergh 2010) and the MW's open clusters. Indeed, the total number of Galactic GCs (Harris et al. 2013) is about half those of our neighbour Andromeda (Barmby & Huchra 2001;Huxor et al. 2014), and the number of Galactic open clusters is vastly higher than that of GCs (Bonatto & Bica 2010). There are a number of issues that make their detection a challenging task, certainly we are still missing a number of them due to high stellar extinction and stellar crowding and very low-luminosities of some of them. Nevertheless, in the last decades, the number of star cluster candidates reported in the literature has increased significantly. Indeed, recent near-infrared (IR) surveys, like the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) and the VISTA Variables in the Via Láctea Survey (VVV; Minniti et al. 2010), and mid-IR survey, like Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) have allowed us to search for new GC candidates. In combination with near/mid-IR surveys, optical surveys are fundamental to finding new GCs, like Gaia Mission DR 2 (Gaia Collaboration et al. 2018), which provides optical photometry and proper motions (PMs). Thus, new GC candidates are being discovered, in particular in the Galactic bulge, using both the optical and near-IR images in order to help complete the census of GCs in the MW. Examples of recently discovered GCs toward to innermost Galactic regions are VVV-CL001 , VVV-CL002 and VVV-CL003 (Moni Bidin et al. 2011), Camargo 1102, 1103, 1104, 1105 and 11061 (Camargo 2018), Camargo 1107, 1108and 1109(Camargo & Minniti 2019, and several tens of candidates in Minniti et al. (2017); Minniti et al. (2017a,b) and Palma et al. (2019). Even so, the nature of some new GC candidates has yet to be confirmed and main parameters have still to be estimated.
In this paper, we focus on two recently discovered and not yet well characterised GC candidates, Patchick99 and TBJ 3 (a.k.a. TJ 3). We are studying for GC candidates in order to investigate the GC luminosity function in the Galactic bulge, especially at very faint luminosities. For that purpose, Patchick 99 and TBJ 3 were selected from Bica et al. (2018), who compiled a list of 10,000 good star cluster candidates in the MW, as they had not been studied before. We use a combination of optical and near-IR catalogues, aiming on having a more complete information about them. Our main goals are to confirm their true nature as GC and estimate their main parameters for the first time, such as reddening, distance modulus, heliocentric distance, size, total luminosity, mean cluster proper motions, metallicity and age.
In Section 2, the observational data are briefly described. In Section 3 we focus our attention to TBJ 3. Section 4 presents the methods used to estimate the physical parameters and the resulting values for Patchick 99. In Section 5, a summary and conclusions are drawn.

Observational Data
The effects of dust become increasingly prominent towards the centre of our Galaxy. Thus, we have to overcome problems like high differential reddening and extinction (e.g., Alonso-García et al. 2017), which introduce systematic errors in the determination of distances. Additionally, uncertainties in the extinction values affect the estimation of ages when isochrones are used. Also, the stellar density is very high in the inner regions and crowding does not always allow to resolve individual stars. Further uncertainties are due to the foreground and background field contamination that may alter the measurement of the total luminosity. For these reasons, we use combinations of optical and near-IR data in order to substantially reduce uncertainties in the measurements and obtain robust results. We analyse especially red clump (RC) and red giant branch (RGB) stars, since they are very bright in infrared and they allow us to estimate accurate values of reddening, extinction, distance and size for our GC candidates.
Below a short description of the observational material is provided.

Infrared Surveys: the VVV, VVVX and 2MASS
We use deep near-IR data from the VVV and VVVX Survey (Minniti et al. 2010;Minniti 2018), acquired with the VISTA In-fraRed CAMera (VIRCAM) at the 4.1m wide-field Visible and Infrared Survey Telescope for Astronomy (VISTA; Emerson & Sutherland 2010) at ESO Paranal Observatory. VVV and VVVX data are reduced at the Cambridge Astronomical Survey Unit (CASU; Irwin et al. 2004) and further processing and archiving is performed with the VISTA Data Flow System (VDFS; Cross et al. 2012) by the Wide-Field Astronomy Unit and made available at the VISTA Science Archive 1 . We use preliminary data from VIRAC (Smith et al. 2018) version 2, described in detail in Smith et al. (in preparation).In summary, VIRAC v2 is based on a psf fitting reduction of VVV and VVVX images using DoPhot (Schechter et al. 1993;Alonso-García et al. 2012. Their astrometry are calibrated to the Gaia DR 2 (Gaia Collaboration et al. 2018) astrometric reference system, and their photometry are calibrated against 2MASS using a globally optimised model of frame-by-frame zero points plus an illumination correction. The VVV and, in particular, the VVVX Surveys are sampling the Galactic bulge and the Southern Galactic Plane, using the J (1.25 µm), H (1.64 µm)), and K s (2.14 µm)) near-IR passbands, aiming to better disentangle and characterise the stellar populations located at these low latitude regions. The VISTA Telescope tile field of view is 1.501 deg 2 , hence 196 tiles are needed to map the bulge area and 152 tiles for the disk. Adding some X and Y overlap between tiles for a smooth match, the area of our unit tile covered twice is 1.458 deg 2 . Specifically, Patchick 99 is located in the VVV tile b251, whereas TBJ 3 is situated in the VVVX tile b462.
In addition, we use 2MASS data (Skrutskie et al. 2006;Cutri et al. 2003). It is a near-IR Survey of the sky in J (1.25 µm), H (1.65 µm) and K s (2.17 µm) passbands. We adopt it specifically to extend the RGB VVV colour-magnitude diagrams (CMDs), as these stars are saturated for K s < 11 mag in VVV images. We downloaded the 2MASS data from the VizieR Online Data Catalogue 2 , and we merge it with the VVV catalogue, including all stars located within 10 from the targets centres. However, since the magnitude scale is different in the two photometric systems, we have transformed the 2MASS photometry into the VVV magnitude scale 3 .

Optical Surveys: Gaia DR 2
The Gaia astrometric mission maps the MW at optical wavelengths, with the aim of revealing motions, composition, formation and evolution of the Galaxy. It provides very accurate position, PMs and radial velocity measurements for sources both in our Galaxy and throughout the Local Group. We use Gaia DR 2 data (Gaia Collaboration et al. 2018) for both stellar cluster candidates, downloaded from the VizieR Online Data Catalogue. Gaia DR 2 data have been prepared by the Gaia Data Processing and Analysis Consortium (DPAC) and contain G-band magnitudes for over 1.6 × 10 9 sources with G < 21 mag and broadband colours G BP (330-680 nm) and G RP covering (630-1050 nm) for 1.4 × 10 9 sources. PM components in equatorial coordinates are available for 1.3 × 10 9 sources, with an accuracy of 0.06 mas yr −1 , 0.2 mas yr −1 , and 1.2 mas yr −1 , for sources with G < 15 mag, G ∼ 17 mag, and G ∼ 20 mag, respectively. We use the parallax (plx) values to make the first cut in such a way to exclude clear nearby foreground stars (those with large errors might scatter to plx < 0.5 mas). In our analysis, we did not make any Gaia photometric or colour cuts.
From Figure 1, it is clear that there is an object at these coordinates. Analysing it at different wavelengths and by looking especially at the WISE satellite images, we noted that TBJ 3 shows diffuse mid-IR emission, which is not present in GCs. This feature, as well as the oblate and very elliptical shape, lead us to discard TBJ 3 as a possible GC candidate, and consider this to be a background galaxy.

Confirmation of Patchick 99 as a new GC
Patchick 99 is located in the Galactic bulge at Equatorial coordinates R.A. = 18h15m47s and DEC = −29d48m46s (J2000) and Galactic coordinates l = 2 • .4884 and b = −6 • .1452 ( Figure  2). We found this cluster during our VVV search Minniti et al. 2018;Minniti 2018), but later realized that an object at that position had been reported before as part of the deep sky hunters (DHS) survey of the DSS and 2MASS images (Kronberger et al. 2006(Kronberger et al. , 2012, so their discovery and nomenclature take precedence. This cluster, which is named after the researcher who discovered it, was then listed in the compilation of Bica et al. (2018), but has not been studied in detail, and we therefore decided to investigate its nature and measure its physical parameters with the VVV data.

Statistical and PM decontamination procedures
Contamination by Galactic bulge and disk stars is relevant, so we have made some selections in order to decontaminate the CMDs. As mentioned in Section 2.2, we consider only stars that have plx < 0.5 mas, in order to exclude foreground stars. Moreover, Figure 3 shows the VVV near-IR CMDs of a 3 radius region centred on Patchick 99 compared with the CMD of a nearby background region, and with the statistically decontaminated CMD. This last CMD was constructed following the procedure described by Piatti & Bica (2012) as applied by Palma et al. (2016Palma et al. ( , 2019 and Minniti et al. (2017). We notice clearly the presence of the cluster RGB and RC that remain in the decontaminated diagram, suggesting that a distance of 3 from the centre is an excellent compromise for our selection. However, there are some blue stars outside of the cluster RGB that probably belong to the foreground MW disk population that persist in this CMD. Therefore, we decided to use the PMs as an independent method to decontaminate the CMD for this cluster, since stars with similar PMs ensure belonging to the same cluster. In fact, we compared the vector PM (VPM) diagram for our cluster, including stars in the 3 radius region and for the field stars, located between 4 < r < 5 arcmin away from the cluster centre ( Figure 4). Visually, the main difference between the two PM distributions is that in the case of our cluster selection, this one shows two over-densities, suggesting a bimodal distribution, unlike the field stars that show a shallower homogeneous distribution when we move away from the centre of the cluster, indicating an unimodal distribution. In order to confirm our statement we have carried out the Gaussian mixture modelling (GMM) analysis (Muratov  & Gnedin 2010), which helps to give us the statistical significance of an unimodal versus a bimodal distribution. We have performed the GMM analysis with the heteroscedastic statistic for each sample. In summary, we perform the GMM analysis for our three subsamples: the field stars located between 4 < r < 5 arcmin away from the Patchick 99 centre, the position selection including all stars situated within 3 from the cluster centre, and our PM selection. Firstly, we do this for field stars and position selection in order to obtain the statistical significance for the two Gaussian distributions shown in Figure 4. Subsequently, we run the GMM algorithm to distinguish the two Gaussian peaks shown in Figure 5. The statistical parameters are shown in Tables 2 and 3. For a mixture of two normal distributions the means and standard deviations along with the mixing parameter (D, separation of the means relative to their widths) are usually used, as a total of five parameters. Ashman et al. (1994) noted that D > 2 is required for a clean separation between the modes, while if the GMM method detects two modes, but they are not separated enough (D < 2), then such a split is not meaningful. Applying the GMM analysis to our samples we noted that they may show a bimodal distribution, but with a low significance level. Doing a deep analysis of the GMM resulting parameters, we can confirm that the distribution of field stars located away from the cluster centre is unimodal since the D < 2 and the standard deviation of the second mode is of σ 2 ≈ 14, which is not physical. About the 3 radius region sample, the analysis becomes more complicated. We believe that the two peaks have a very small separation and the method is not able to distinguish both properly, indeed comparing the parameters of the first Gaussian with the second one we can appreciate that the first dominates over the second. On the other hand, the GMM method on the PM selection indicates a bimodal distribution, since D > 2. This may be misleading to interpret since we should have expected an unimodal distribution for the stars in a cluster. However, we have to highlight that any sample will be contaminated by foreground stars that will have different velocity (and PM) distribution. Thus, having two peaks in the sample of PM selection is not unexpected. In any case, this does not represent a problem since the second mode is very narrow and is not dominant over the first, meaning that our final sample has got a very small contamination.
Successively, we calculate the mean cluster PMs (adopting the σ-clipping method) measured from Gaia DR 2, yielding µ α = (−2.98 ± 1.74) mas yr −1 and µ δ = (−5.49 ± 2.02) mas yr −1 , which are in good agreement with the PM values found by the GMM method. After that, we selected as cluster members only stars within 3 mas yr −1 , following the procedure of Garro et al. (2020), indicated by the black circle drawn in Figure 4. Also, the size of the circle is set taking into consideration the typical size of GC members in the VPM diagram for other well studied MW bulge GCs. Additionally, in Figure 5 we show the PM distributions in a wide range (−20 < µ α < 15 mas yr −1 and −20 < µ δ < 10 mas yr −1 ), considering only stars situated within 3 from the Patchick 99 centre (blue histogram), attempting to minimise field stars. The resulting star sample after the PM cut is shown with the red histogram. Each histogram is constructed considering the sum of all the combined values per unit bin, selecting a fixed bin size of 0.8 mas yr −1 (value that allowed us to visually distinguish the two peaks of the Gaussian distributions). We can appreciate double peaked distributions for both blue histograms, with a very little separation (as expected from the GMM method), suggesting that we have at least two overlapping populations: the bulge field stars and the star cluster. Indeed, fitting two Gaussian distributions we assign a narrower and more peaked distribution for Patchick 99 (green curve) and a wider distribution for Bulge field stars (grey curve). The total distribution (black curve) is obtained by coadding the two Gaussians, in order to fit the entire sample. Figure 6 displays the PM decontaminated CMDs for VVV and Gaia data. We can recognise that the main GC features, like RGB and RC, remain but, in comparison with the statistical decontamination (Figure 3), the foreground disk population is now better removed.
Following these procedures, we have obtained the decontaminated near-IR (K s versus J − K s ) and optical (G versus BP − RP) CMDs for our cluster, as shown in the right panels of Figures 6 and in Figure 7. We use those CMDs to derive the cluster's main parameters. Both diagrams clearly show the RC at K s = (12.5 ± 0.04) mag and G = (15.5 ± 0.05) mag, suggesting that P99 is a metal-rich GC. It is crucial to use RC stars in this kind of studies, because these stars are very good distance indicators.
Using our estimates of the reddening and distance we were able to fit PARSEC 4 isochrones (Bressan et al. 2012;Marigo et al. 2017) to the CMDs of Patchick 99. This allows us to have estimation of the metallicity and the age of the GC, finding [Fe/H] = (−0.2 ± 0.2) dex and t = (10 ± 2.0) Gyr, respectively. In order to obtain a more robust result, we have also derived the metallicity from the slope of the RGB (α = 0.11±0.01) using the A&A proofs: manuscript no. newGCs_v28  Figure 7 shows a good fit of the isochrones with the above mentioned values of metallicity and age in both our optical and near-infrared CMDs. We determined the errors in age and in metallicity by changing the isochrones at a fixed metallicity and a fixed age, respectively, until they do not fit the cluster sequence simultaneously in the near-IR and optical CMDs. Therefore, it is clear that although age is one of the fundamental parameters to characterise GCs, its correct determination is still one of the most complicated tasks to perform, especially in environments with high stellar density and with differential reddening, such as the Galactic bulge. This is mainly due to difficulties to reach the main sequence turn-off (MS-TO) point. We can appreciate that problem in Figure 8, which displays the VVV and 2MASS PM decontaminated  CMDs. We include 2MASS cluster members with K s < 11 mag, as they are more sensitive to the metallicity variations (e.g. , McQuinn et al. 2019). We can see that variations in metallicity have an especially strong effect along the RGB and RC, and vice-versa, we can notice evident changes in age in the MS-TO region. For this reason our age should be taken as a crude estimate, since we are unable to measure it precisely.
All of the previously derived parameters, including position, metallicity and kinematics confirm that Patchick 99 is a bulge GCs, but if we consider its age, it could be even younger than typical metal-rich GCs present in the Galactic bulge, which have ages between 12 and 13 Gyr (e.g., Carretta et al. 2000. However, many studies have demonstrated that GCs show a relatively wide range of ages in the bulge (Zoccali et al. 2003;Clarkson et al. 2011;Valenti et al. 2013;Bernard et al. 2018). Figure 9 shows the Gaia+VVV optical-near-IR colourcolour diagram (CCD) for the whole catalogue without selection cuts, and for only cluster members of Patchick 99 with K s < 15 mag. We can clearly notice that the colour distribution reveals a tighter sequence for star members (circles, right panel) than for the field stars (squares, left panel). In particular, we can appreciate that there is a little differential extinction, in the right panel of Figure 9. Based on the high resolution reddening map by Surot et al. 2020, the mean reddening is E(J − K s ) = 0.05 mag in this field, which differs from our reddening value by ∆(E(J − K s )) = 0.07 mag at most. Hence, we believe that our estimation is in good agreement with the adopted map, discarding this as a significant problem.
Finally, we derive the total luminosity of our cluster, using the catalogue with only cluster members. Once the flux for each star is measured, we derive the absolute magnitude from the total flux, obtaining M K s = (−7.0 ± 0.6) mag, equivalent to M V = (−4.5 ± 0.8) (for a typical GC colour of (V − K s ) = (2.5 ± 0.5) mag), indicating that it is a very faint GC. This value is an underestimate of the total luminosity, since the faintest stars are missing. Regarding this, we compared Patchick 99 luminosity with that other known GCs similar to Patchick  Harris (1996) catalogue. Therefore, estimating the M V differences and assuming the similarities between the GCs, we derive that the total luminosity for Patchick 99 is M V = −5.2 mag, ∼ 2.2 less luminous of the MW GC luminosity function peak (M V = (−7.4 ± 0.2) mag from Harris 1991;Ashman & Zepf 1998) All these parameters are summarised in Table 1. Fig. 9. Gaia+VVV optical-near-IR CCDs for Patchick 99, as density plot. On the left: squares are the observed stars, using the entire catalogue without selection cuts; on the right: circles represent star members of Patchick 99, considering bright stars previously selected. At the top-left of the panel, the red arrow represents the reddening vector.

The derivation of the surface density profile
A detailed description of the structural parameters of Patchick 99 would need a careful treatment of completeness as well as relatively large numbers of member stars from which a radial profile can be derived. Traditionally, King profiles (King 1962) have been fitted to a significant fraction of MW old clusters, with however more and more evidence that other profiles might provide better fits, especially when extragalactic or massive GCs are included (e.g., Mackey & Gilmore (2003), de Boer et al. (2019a). Although rich clusters allow for model-dependent sizes and structural properties, our sample of member stars is not enough to adequately address a comparison among the different models available. Instead, a King function is assumed to be a good approximation for the radial distribution of the stellar density.
The correct determination of the cluster centre has a fundamental importance, since an error in its position can lead to altered results regarding the density profile. For this reason, we have used the K-Means algorithm in Python to determine the centre in a robust way from our catalogue of cluster member stars. In summary, this is a machine-learning algorithm that searches for subgroups (or literally "clusters") in a given parameter space among a sample of data points. We checked visually that the resulting centres were close to the expected density peaks and we found that the initial position (R.A. = 18h15m47s and DEC = −29d48m46s (J2000)) coincides almost-perfectly with the resulting values of the centre, with a small separation of ∆RA ≈ 2 and ∆DEC ≈ 5 . Starting from these new coordinates, we divided our sample of member stars into ten circular annuli, out to a radius of 1.5 , in steps of 0.15 . We do not need to apply geometric corrections as the resulting circular annuli are within the FOV of our catalogue. We performed the radial density profile of Patchick 99 (see Figure 10) dividing the total number of stars in each radial bin by the area of that ring in order to derive the density in each bin.
Poisson errors are also shown. We note that the error bars are relatively large, mainly due to the low-number statistics in each radial bin, especially in the innermost point (see Figure 10). No background has been subtracted since we are dealing with a list that is expected to be free from foreground stars and other contaminants.
We have overplotted a family of King profiles adopting different values of core radius (r c = 1.8 , 1.5 , 1.2 ), but keeping the tidal radius fixed at 10 . The choice to use that value of tidal radius is justified following de Boer et al. (2019b) (see their Figures 13 and 14), where a clear correlation between the tidal radius and galactocentric radius of a given GC is illustrated. As shown in Figure 10, the observational points are best represented by r c = 1.2 , which is equivalent to r c ∼ 1.1 pc. Also, we consider the tidal radius r t = 10 as the size of the cluster, which corresponds to a physical size of 9.3 pc.

The RR Lyrae stars in Patchick 99
RR Lyrae variable stars represent a very old-population in our Galaxy, found in the Galactic halo and bulge. In particular, according to Pietrukowicz et al. (2015) their spatial density ranges between 80 and 440 RRL/deg 2 in the bulge. They are also considered powerful tools to reveal the nature of a GC candidate because if detected in a stellar cluster, RR Lyrae stars guarantee that it is an old GC. Although these variables are used as excellent tracers of old and metal-poor populations, there are observational evidences that suggest the presence of RR Lyrae also in metal-rich GCs, such as NGC 6388 ([Fe/H] = −0.44 dex) and NGC 6441 ([Fe/H] = −0.46 dex) (Pritzl et al. 2002;Clementini et al. 2005), and NGC 6440 [Fe/H] = −0.36 dex), all of which are somewhat more metalrich than Patchick 99. Additionally, they are good reddening and distance indicators.
For these reasons, we have searched for these variable stars and we have detected 21 RR Lyrae stars within 12 arcmin from the Patchick 99 centre, matching the OGLE (Soszyński et al. 2014), the VVV (Majaess et al. 2018) and the Gaia (Clementini et al. 2019) catalogues. They are listed in Table 4, where we summarised: identification, location, period, photometry in different filters, heliocentric distance, distance from the centre of the GC, type for each source and their PMs.
Initially we constructed their phased light curves, in order to discriminate their pulsation mode. We identify 13 fundamental mode pulsators (ab-type RR Lyrae stars) with long periods (P 0.4 days) and large amplitudes, and 8 first-overtone pulsators (c-type RR Lyrae stars), with short period (P 0.4 days) and small amplitudes. The RRc and RRab pulsators also differ in the morphology of their light curves, since the RRab's have typically asymmetrical light curves, unlike RRc's. Figure 11 displays the Amplitude-Period diagram (so-called Bailey diagram, Bailey et al. 1919) in such a way to highlight that differences. Indeed, we consider the Bailey diagram a solid tool for verifying our estimations. Our main goal is to identify which RR Lyrae stars are real members of Patchick 99. Firstly, we measured their projected distance from the centre using the centroid of each variable stars. Based on these distances and considering the size of the cluster, we believe that seven RR Lyrae stars are members of Patchick 99, as shown in Figure 12 and evidenced by a star symbol in Table 4. In addition, we compare their projected distance with the resulting heliocentric distance (Figure 12), confirming again the membership. In detail, we used the period-luminosity (PL) relation in K s -band following Muraveva et al. (2015) (with RRc stars "fundamentalized" by adding 0.127 to the logarithm of the period) in order to derive their heliocentric distance. Once selected the RR Lyrae members, we estimate the mean distance, finding a value of D = 7.06 ± 0.48 kpc, consistent with the heliocentric distances obtained from the VVV and Gaia photometries (see Section 4), since these values differ by less than 2σ. On the other hand, we discard the others as members since they have heliocentric distances larger than those found using VVV near-IR photometry, or they are located outside our assumed tidal radius. However, we can restrict the RR Lyrae sample to the ones that have PMs matching the mean GC PM within the errors, we find that only 3 of them fulfil this requirement: OGLE-BLG-RRLYR-35312, 35476 and 35459, favouring a short distance of 6.3 ± 0.5 kpc in the mean. Additionally, Figure 12 does not show OGLE-BLG-RRLYR-35364 and OGLE-BLG-RRLYR-35381, since their distance values are very large and Fig. 12. Spatial distribution of RR Lyrae variable stars detected within 12 from the centre of Patchick 99. The magenta points represent the seven RR Lyrae stars considered cluster members, whereas we highlight the PM members with magenta squares. The distance errorbars are of 0.5 kpc for each point. The dotted line depicts the heliocentric distance at D = (6.4 ± 0.2) kpc derived from the VVV near-IR photometry.
probably they are located in the Galactic halo or in the Sagittarius dwarf galaxy. We also tested two other PL relations from Gaia Collaboration et al. (2017) and Navarrete et al. (2017), yielding slightly larger distance values. However, we use the comparison between the three PL relations, in order to assign a mean distance error of 0.5 kpc. Once absolute magnitudes are obtained, we also build-up the PL relation diagram (Figure 13). This allowed us to confirm again the membership, since we find a stronger relation when the least-square fit is used on the star cluster members (red line), rather than on the whole sample (blue line). We also show the VVV phased light curves for each RR Lyrae member (Figure 14). We note that the light curve profile for OGLE-BLG-RRLYR-35459 is slightly worse than the others, since this variable appears a bit blended. Moreover, we derive the expected spatial density from Pietrukowicz et al. (2015) and Navarro et al. (2020). Specifically, following Navarro et al. (2020), we are able to predict a mean density of 4 field RR Lyrae type-ab in the same field of Patchick 99. We found 21 variables in total, including 13 RRab. Hence, this value represents a clear excess over the background (∼ 4.5σ detection).
Finally, the magenta points in Figure 7 define the position of the seven RR Lyrae stars in the VVV CMD. Their near-IR magnitudes are on average much fainter than the RC (∼ 1 − 1.5 mag). However, both theory and observations suggest that the RR Lyrae indeed should be fainter than the RC in the near-IR CMDs and also indicate that more metal-rich RR Lyrae are fainter than metal-poor ones. For example, Cassisi et al. (2004) explored the predicted behavior of the pulsators as a function of the horizontal branch morphology and over the metallicity range Z=0.0001 to 0.006, finding that the RR Lyrae luminosity decreases with increasing metallicity. Further, Marconi et al. (2018) constructed new sets of helium-enhanced (Y = 0.30, Y = 0.40) nonlinear, time-dependent convective hydrodynamical models of RR Lyrae stars covering a broad range in metal abundances (Z = 0.0001 -0.02), finding that an increase in helium content from the canonical value (Y = 0.245) to Y = 0.30 -0.40 causes a simultaneous increase in stellar luminosity and in Fig. 13. M K s − log(P) diagram for all RR Lyrae sample (black points) and RR Lyrae star members (magenta points, while the squares represent the PM variable members). Blue and red lines represent the leastsquares fit for all RR Lyrae sample and cluster members, respectively. pulsation period (see their Figure 1 and 2). Also, from the observational point of view, we have compared the location of our RR Lyrae with those confirmed PM members of the cluster NGC 6441 from Alonso- García et al. (in preparation). The RR Lyrae stars in Patchick 99 fall in the expected region of the near-IR CMD for this metal-rich cluster. In any case, spectroscopic metallicities are needed to confirm if this is the most metal-rich GC known with RR Lyrae.

Summary and Conclusions
We have investigated the GC candidates Patchick 99 and TBJ 3 in order to understand their real nature. They are analysed in optical and near-IR wavelengths with the goal of determining their physical parameters. The visual examination of TBJ 3 at various wavelengths confirms that it is not a GC but from its mid-IR emission we conclude that TBJ 3 is a background galaxy. Therefore, we have focused our work on Patchick 99. We have measured its reddening, extinction, distance modulus and heliocentric distance, size, proper motion, total luminosity, metallicity and age. We summarised all those values in Table 1. Even though the field contamination is severe, we used two different methods (statistical and PM decontamination) recovering the same results, with the GC RGB and RC clearly present. We conclude that Patchick 99 is a metal-rich GC, located in the Galactic bulge at a heliocentric distance of D = 6.6 ± 0.6 kpc, obtained as an average distance of three-independent methods (Gaia DR 2 and VVV photometries, and three RR Lyrae star members). We define it as a genuine GC because both CCD and CMDs yield a population with an age of 10 Gyr. These results are confirmed by the presence of seven RR Lyrae stars, located at d < 12 from the cluster centre and at the same distance of the cluster from the Sun. However, analysing their PM we can confirm that three of them are surely member of our cluster: OGLE-BLG-RRLYR-35312, 35476 and 35459, since their PMs match the mean cluster PM within the errors. Patchick 99 may be the most metal-rich globular cluster with RR Lyrae, to be confirmed by spectroscopic observations. As mentioned previously, even if position, metallicity and kinematics confirm that it is a bulge E.R. Garro et al.: Confirmation of the new bulge globular cluster Patchick 99 from the VVV and Gaia surveys  −1.96 ± 0.58 4.08 ± 2.8 −2.54 ± 0.55 10.71 ± 3.16 0.07 ± 0.06 PM selection −3.35 ± 0.13 1.42 ± 0.05 −0.73 ± 0.14 0.53 ± 0.08 2.46 ± 0.09 Table 3. GMM analysis for each sub-sample (field stars, position and PM selection) applied to PM in DEC. The subscripts 1 and 2 indicate the first and second mode of the Gaussian distribution.