UVIT-HST-Gaia-VISTA study of KRON 3 in the Small Magellanic Cloud: A cluster with an extended red clump in UV
P. K. Nayak, A. Subramaniam, S. Subramanian, S. Sahu, C. Mondal, Maria-Rosa L. Cioni, Cameron P. M. Bell, A. Bandyopadhyay, Chul Chung
MMNRAS , 1–18 (0000) Preprint 5 February 2021 Compiled using MNRAS L A TEX style file v3.0
UVIT-
HST - Gaia -VISTA study of KRON 3 in the Small MagellanicCloud: A cluster with an extended red clump in UV
P. K. Nayak, , ★ A. Subramaniam, S. Subramanian, S. Sahu, C. Mondal, Maria-Rosa L. Cioni, Cameron P. M. Bell, A. Bandyopadhyay and Chul Chung Indian Institute of Astrophysics, 2nd Block, Koramangala, Bangalore, 560034, India Tata Institute of Fundamental Research, Homi Bhabha Road, Navy Nagar, Colaba, Mumbai 400005, India Leibniz Institute for Astrophysics Potsdam (AIP), An der Sternwarte 16, D-14482 Potsdam, Germany Aryabhatta Research Institute of Observational Sciences, Uttarakhand 263002, India Department of Astronomy & Center for Galaxy Evolution Research, Yonsei University, Seoul 03722, Republic of Korea
Accepted 2021 February 3. Received 2021 January 18; in original form 2019 December 22
ABSTRACT
We have demonstrated the advantage of combining multi-wavelength observations, from theultraviolet (UV) to near-infrared, to study Kron 3, a massive star cluster in the Small MagellanicCloud. We have estimated the radius of the cluster Kron 3 to be 2. (cid:48) − optical) colour-magnitude diagram (CMD).We found that extension of the RC is an intrinsic property of the cluster and it is not due tocontamination of field stars or differential reddening across the field. We studied the spectralenergy distribution of the RC stars, and estimated a small range in temperature ∼ ∼
60 - 90 L (cid:12) and radius ∼ (cid:12) supporting their RC nature. The rangeof UV magnitudes amongst the RC stars ( ∼ 𝑖𝑛𝑖 =0.23 to 0.28), anda small variation in age (6.5-7.5 Gyr) and metallicity ([Fe/H]= − − Key words: (galaxies:) Magellanic Clouds, galaxies: star clusters,
Star clusters were known to be the best examples of coeval systemsand thought to follow simple stellar population models. Recent Hub-ble Space Telescope (HST) observations of Galactic globular clus-ters (GCs) and massive star clusters in the Magellanic Clouds (MCs),however, have changed our understanding of cluster formation. It isfound that intermediate age (mostly younger than 2 Gyr) massive( ∼ a few times 10 M (cid:12) ) clusters in the MCs show an extended mainsequence turn-off (eMSTO), which can not be explained throughphotometric errors or binarity present in the clusters (Mackey et al.2008; Milone et al. 2008, 2009). Intrinsic age spreads and the ef-fects of stellar rotation are suggested to explain the presence ofsuch eMSTOs. Recent studies of Galactic open clusters (OCs) us-ing the Gaia Data Release (DR) 2 (Bastian et al. 2018; Cordoniet al. 2018; Marino et al. 2018; Lim et al. 2019; Sun et al. 2019;Li et al. 2019; Gossage et al. 2019) also show extended featuresin both main sequence (MS) and MSTO in the colour-magnitudediagrams (CMDs) and it is suggested that the distribution of the ★ E-mail: [email protected] projected rotational velocity of stars play a major role in the cre-ation of these extended features. The Galactic GCs (older than 10Gyr) and relatively older (more than 2Gyr old) massive clusters inthe MCs show multiple evolutionary sequence in the MS, sub-giantand red giant branch (SGB & RGB) (Piotto et al. 2007; Andersonet al. 2009; Piotto 2009). Multi-modal populations in age and/orenrichment in chemical abundance (i.e. spread in He, C, N, O, Naetc.) have been proposed to explain the observed split in differentevolutionary sequences. It is well known that there is a paucity ofclusters in the age range 4-9 Gyr in the Large Magellanic Cloud(LMC) (van den Bergh 1991) as well as in our Galaxy, whereas,the Small Magellanic Cloud (SMC) hosts a relatively large numberof rich clusters in the above age range. Niederhofer et al. (2017a,b)studied four clusters in the SMC in the above age ranges, which aresimilarly massive but significantly younger ( ∼ ∼ © a r X i v : . [ a s t r o - ph . GA ] F e b P. K. Nayak et al. of Kron 3 will add to the analysis of clusters of this type as well asprovide additional constraints to help us better understand multiplestellar population in massive star clusters.Gascoigne (1966) presented the first CMD of Kron 3 and re-ported it to be an intermediate-age ( ∼ 𝑉 =21 mag). Comparing itsCMD with theoretical stellar evolutionary models, Hodge (1982)reported a metallicity and age of Kron 3 of [Fe/H] = − ± ± 𝑅 =23 mag), using observations from the CTIO 4m tele-scope and found that Kron 3 has an age range of 5-8 Gyr. Theauthors also mentioned that the uncertainty in the distance modu-lus (DM) prevented a more precise estimate of the cluster age. Theauthors estimated the radius of the cluster to be ∼ (cid:48) .4 (42 pc) usingthe distribution of star counts around the cluster center. By fittingisochrones (Vandenberg & Bell 1985) to the CMDs, Alcaino et al.(1996) showed that the age of Kron 3 ranges from 8-10 Gyr with aDM value of 18.75 mag. They suggested that estimates of a higherDM could be the reason for getting younger ages in previous stud-ies. They questioned the previous estimation of radius by Rich et al.(1984). Alcaino et al. (1996) showed that stars located in a field2. (cid:48) ∼ (cid:48) or 105 pc) than previously estimated ( ∼ (cid:48) (cid:48) 𝐵 ) and F555W ( 𝑉 )bands, Mighell et al. (1998) estimated the age of the cluster as4.7( ± 𝑍 =0.001. Hence, there is a range ofages (1-10 Gyr) and radii ( ∼ − − 𝑍 ) ofthe cluster is 0.001 ([Fe/H]= − Gaia and the Visual and Infrared Survey Telescopefor Astronomy (VISTA) observations (for the outer regions of thecluster). The superior resolution of UVIT [ ∼ (cid:48)(cid:48)
5, three times betterthan that from the Galaxy Evolution Explorer (GALEX)] in thenear-UV and the large field-of-view (28 (cid:48) diameter) are the mainadvantages of this study.The remaining sections of this paper are arranged as follows. InSect. 2, we discuss the NUV data used for this study and use theseto determine a new estimate of the cluster radius. In Sect. 3, wepresent UV-Optical CMDs using UVIT and HST that demonstratethe extended nature of the RC in the NUV. In the Sect. 4, wediscuss the effect of photometric zero points variations in the UV-optical CMDs. In Sect. 5, we present UV-Optical CMDs using UVITand HST and have discussed the effect of field star contaminationin these CMDs. Sect. 6 provides the analysis to find the possiblereasons behind the extended RC in NUV. We discuss the results ofthis paper in Sect. 7 and summarise with our conclusions in Sect. 8.
The observations of Kron 3 were carried out with the UVIT tele-scope in far-UV (FUV) and near-UV (NUV) bands as part of aGuaranteed-Time proposal (G08) on 25th March 2018. Kron 3 wasobserved in one FUV (F148W : 125-175 nm) and one NUV (N242W: 203-281 nm) filter. The total exposure time was 7194 seconds. Theobservations were completed in multiple orbits. We applied correc-tions for spacecraft drift, flat-field and distortion, using the softwareCCDLAB (Postma & Leahy 2017) and created images for each or-bit. Then, the orbit-wise images were co-aligned and combined togenerate science ready images. The science ready images were cre-ated for an area of 4K ×
4K in size with a scale of 0. (cid:48)(cid:48) (cid:48)(cid:48)
24 and 1. (cid:48)(cid:48)
03, respectively. To detectthe sources, we used threshold as six times the average backgroundcount. Due to crowding in the central regions of the cluster, we haveperformed PSF photometry to the detected stars. First, a model PSFwas generated using isolated stars for both the science ready FUVand NUV images. Then the model PSF has been fitted to all thedetected stars to select the stars with good PSF values with theirPSF magnitude. We applied aperture corrections and saturationcorrections to PSF magnitudes and calculated the final magnitudesof the detected stars in the corresponding bands by adding zeropoint magnitudes. The values of zero point magnitudes have beentaken from Tandon et al. (2020).Figure 1 shows that there are only a few FUV detections in thecentral parts of Kron 3. Therefore, we used NUV data to estimatethe cluster radius. We counted the number of stars present in bins of
MNRAS000
MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure 1.
Colour composite 6 (cid:48) × (cid:48) image of the Kron 3 cluster. Yellowand blue colours correspond to stars detected in NUV and FUV band,respectively. The green circle indicates 2 (cid:48) radius around the cluster. Figure 2.
Radial distribution of stellar density at 0. (cid:48) ∼ (cid:48) . (cid:48) ∼ (cid:48)
0. For theremainder of this study, we will adopt a cluster radius of 2 (cid:48) .0. Figure3 shows the photometric error of stars within the cluster radius as afunction of PSF magnitudes for NUV (black points) and FUV (red
Figure 3.
Photometric errors of stars within the cluster radius (2.0 (cid:48) ) as afunction of magnitude in both NUV and FUV bands. points) bands. The figure suggests that the photometric errors are ≤ ∼ Artificial-star (AS) tests provide a crucial role to determine thecompleteness level of the data used in this study. We have addedartificial stars to original image with a spatial density distribution ofstars similar to the distribution of original stars. For each iteration,the number of inserted artificial stars were 20% of detected starskeeping the magnitude same for all stars and the positions of the starswere generated randomly. The detected NUV stars have a magnituderange 22.0 to 25.0 magnitude and we have divided the range in 9bins to perform the AS test for the corresponding magnitude. Torecover added stars from the image, we followed similar proceduresas used for the real stars. We have considered a star to be recoveredwhen it has a magnitude difference less than 0.75 mag and spatialdifference less than 0.5 pixel. The completeness is measured as aratio of recovered stars to the added stars for a fixed magnitude.Figure 4 shows the completeness of NUV data as a function ofcluster radius for different NUV magnitudes. We found that the datareached ∼
50% of its completeness level at 24.5 mag at the densestor core region ( < (cid:48) ) of the cluster. Only 54 stars were detected in the FUV band compared to 1623stars in the NUV within a cluster radius of 2 (cid:48) .0, making it hard toconstruct a FUV − NUV CMD to study the properties of the cluster.On the other hand, we can combine NUV data with data obtainedat optical and near-infrared (near-IR) wavelegths. This will giveus a broader colour range for the various evolutionary sequencesof the cluster, particularly near the MSTO region which is usedto estimate the fundamental parameters of the cluster (e.g. age).We used the publicly available optical photometric (DAOPHOT)data in F555W ( ∼ 𝑉 ) and F814W ( ∼ 𝐼 ) filters obtained from theHubble Legacy Archive (HLA), that covers the central region of MNRAS , 1–18 (0000)
P. K. Nayak et al.
Figure 4.
Completeness as a function of radius and V magnitudes. the cluster. We considered stars brighter than 22.5 mag in V whilecross-matching with the NUV data because stars fainter than thisare not detected in the NUV image due to the sensitivity limit ofUVIT. We have compared the isochrone with the general sensitivityof UVIT to put the limit in F555W. Not considering stars fainterthan 22.5 magnitude in V also helped to reduce the crowding in thecentral part of the cluster, making it easy to cross-match with theNUV data. We cross matched the NUV detections with HLA databy considering a maximum separation of 1. (cid:48)(cid:48)
0. In our analysis weexcluded those common detections where we found multiple HLAdetections within 1. (cid:48)(cid:48) (cid:48)(cid:48)
0. Using the above mentionedcriteria, we found 749 detections in common between both data sets.In the top-left panel of Figure 5, we plotted 𝑉 vs ( 𝑉 − 𝐼 ) CMDof the full HLA data of Kron 3. We noticed a large spread in theMS and MSTO region. We also over-plotted a Padova isochrone(Marigo & Girardi (2007); Marigo et al. (2008) ; cyan line) of age7 Gyr and metallicity Z=0.001, after correcting for reddening using 𝐸 ( 𝑉 − 𝐼 ) = 0.033 mag and a DM of 18.8 mag. We took the age,reddening, DM and metallicity values from Glatt et al. (2008). Togenerate the isochrone, we have used the Flexible Stellar PopulationSynthesis (FSPS) model (Conroy et al. 2009; Conroy & Gunn 2010)and convolved the Padova models with the effective area curves ofdifferent filters to obtain the model magnitudes. The aforementionedreddening and DM are used throughout this study to fit the CMDs.In the top right panel we highlighted cross-matched stars and inparticular, the RC stars. We plotted also an isochrone. The bottomright panel shows 𝑉 vs ( 𝑁𝑈𝑉 − 𝑉 ) CMD of only cross matched stars(black points) and RC stars, which are highlighted in red. We fittedan isochrone after correcting for reddening, extinction and DM. Wecan see that the RC is no longer a clump in the UV-optical CMD,instead it appears stretched along the colour axis. The extension ofRC stars towards both bluer and redder colours indicates that the RCstars span a range of NUV magnitudes. Furthermore, the SGB, RGBand a part of MS stars are not fitted well by a single isochrone, whilethey are fitted well in the optical CMD. The spread in the MSTOregion noticed in the 𝑉 vs ( 𝑉 − 𝐼 ) CMD becomes even larger in theV vs ( 𝑁𝑈𝑉 − 𝑉 ) CMD. We find that a few SGB, RGB and MS starsget brighter in NUV and appear bluer in ( 𝑁𝑈𝑉 − 𝑉 ). In the bottom-left panel we plotted HLA-NUV cross-matched stars, highlightingRC stars, in the NUV vs ( 𝑁𝑈𝑉 − 𝑉 ) CMD. This plot clearly shows that RC stars extend more than two magnitudes in both colour andmagnitude. We also notice that UVIT reaches its detection limit nearthe MSTO region, and the SGB as well as part of the RGB becomefainter than the MSTO in the NUV band. Photometric errors are alsorelatively large near the detection limit. The bottom of the RGB evenexceeds the detection limit of UVIT. Therefore, the stars near thebottom of the RGB in the NUV-optical CMDs not fitted by theisochrone could be due to a poor detection in NUV. Henceforth, wefocused our study on the evolved stellar population, mainly RC starswhich are detected in NUV with less uncertainties. The extensionof the RC shall not be due to photometric errors.The observed extension in the RC could be an effect of uncer-tainties in the photometric zero points. The variation can occur dueto differential reddening in the cluster region and/or to photomet-ric inaccuracies. Poor sky determination or not accounting for thespatial variation of the PSF to generate the PSF model causes inac-curacies in photometry (Milone et al. 2012a). Therefore, we need tocheck whether differential reddening or PSF variations across thefield can cause the observed extension in RC. Typically, it is assumed that the foreground dust distribution isuniform across a given cluster. In some cases, however, it has beenfound to be patchy, resulting in reddening variations across the clus-ter and leading to an artificial broadening in the stellar sequences onthe CMDs (Milone et al. 2012a). Another source of non-intrinsicbroadening can manifest due to PSF variations across the CCD,leading to slight shifts in the photometric zero point (see Andersonet al.2008 for details). To unambiguously determine the presenceof multiple populations within a given cluster, one must first ac-count the main sources of non-intrinsic scatter/broadening withinthe photometry. In this Section, we investigate the effects of bothdifferential reddening across the cluster as well as PSF variationsin the photometric zero points to assess if they have a significantimpact on the inferred spread of the RC.
To check for differential reddening, we adopted the method byMilone et al. (2012a), in which they used MS stars as a tool toestimate the effect of reddening. As we have only a few detectionsin the FUV to make a UV CMD and the UVIT reaches its detec-tion limit near the turn-off of Kron 3, we instead used HST datato calculate the reddening levels across the cluster and convertedthese to the corresponding values in the NUV band to de-redden theNUV-optical CMD.We briefly summarise the method below, but refer the readerto Sec. 3.1 of Milone et al. (2012a) for more details. First we definean arbitrary point near the MSTO (o) and then with respect to thatpoint rotate the CMD counter-clockwise by an angle theta=arctan[ 𝐴 𝑉 /( 𝐴 𝑉 − 𝐴 𝐼 ) ] so that the abscissa became parallel to the redden-ing vector. 𝐴 𝑉 and 𝐴 𝐼 are nothing but the absorption coefficientsin the V and I bands corresponding to the average reddening of thecluster. The reason for rotating the CMD is that it is much moreintuitive to determine a reddening difference on the horizontal axisrather than along the oblique reddening vector. Now “abscissa” and“ordinate” will be indicated as the abscissa and ordinate of the ro-tated reference frame, respectively. The left panel of Figure 6 showsthe highlighted (black points) region of MS, used as reference MSstars to determine the effect of differential reddening. The red arrow MNRAS000
To check for differential reddening, we adopted the method byMilone et al. (2012a), in which they used MS stars as a tool toestimate the effect of reddening. As we have only a few detectionsin the FUV to make a UV CMD and the UVIT reaches its detec-tion limit near the turn-off of Kron 3, we instead used HST datato calculate the reddening levels across the cluster and convertedthese to the corresponding values in the NUV band to de-redden theNUV-optical CMD.We briefly summarise the method below, but refer the readerto Sec. 3.1 of Milone et al. (2012a) for more details. First we definean arbitrary point near the MSTO (o) and then with respect to thatpoint rotate the CMD counter-clockwise by an angle theta=arctan[ 𝐴 𝑉 /( 𝐴 𝑉 − 𝐴 𝐼 ) ] so that the abscissa became parallel to the redden-ing vector. 𝐴 𝑉 and 𝐴 𝐼 are nothing but the absorption coefficientsin the V and I bands corresponding to the average reddening of thecluster. The reason for rotating the CMD is that it is much moreintuitive to determine a reddening difference on the horizontal axisrather than along the oblique reddening vector. Now “abscissa” and“ordinate” will be indicated as the abscissa and ordinate of the ro-tated reference frame, respectively. The left panel of Figure 6 showsthe highlighted (black points) region of MS, used as reference MSstars to determine the effect of differential reddening. The red arrow MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure 5. (top-left) CMDs of HLA stars within the central 2’ of Kron 3. Red line indicated the fiducial line of MS stars. (top-right) CMDs of stars cross-matchedbetween HLA and NUV data (black) over-plotted onto the full HLA sample (grey) with RC stars marked in red. The cross-matched stars are shown in the NUVvs (NUV-V) CMD (bottom-left) and V vs (NUV-V) CMD (bottom-right) where the distribution of RC stars appears stretched. In all panels an isochrone (cyan)corresponding to an age of 7 Gyr and a metallicity Z=0.001 is overlaid onto the CMDs after correcting for extinction and DM. indicates the reddening vector, also defined as direction of abscissaaxis of the rotated CMD. The middle panel shows the rotated CMDwith respect to the point "o".We have used only MS stars to generate red fiducial line. First,we binned the MS in 0.2 mag along ordinate and calculated medianabscissa associated to median ordinate for each bins. The fiducialline is then created by fitting these points with a cubic spline. Theuse of median points help us to minimize the contamination effectfrom binary stars, field stars and stars with poor photometry. Thered line in the middle panel of Figure 6 represents spline fit to themedian points (blue) of each bin. Then we calculated the distance toeach star from the fiducial line along the reddening axis ( 𝛿 abscissa).The right panel of Figure 6 shows the plot of ordinate vs 𝛿 abscissa.We choose these black points as references to calculate the differ- ential reddening suffered by each stars in the ordinate vs 𝛿 abscissadiagram.The final step is explained in the Figure 7. The idea is if thecluster does not effected by differential reddening then median valueof stars from any small part of the cluster region in the 𝛿 ( 𝑎𝑏𝑠𝑐𝑖𝑠𝑠𝑎 ) vs ordinate diagram will coincide with the red fiducial line. On theother hand, if the cluster has differential reddening the median valuewill show an offset to the red fiducial line. Here we describe theprocedure to calculate differential reddening suffered by a targetstar. We choose a small region around that target star in the spatialdiagram and over plotted on the ordinate vs abscissa diagram. Then,we calculate the median value of 𝛿 abscissa of these stars with re-spect to the fiducial line, which gives us the differential reddeningassociated to that star. The top left panel of Figure 7 shows the MNRAS , 1–18 (0000)
P. K. Nayak et al.
Figure 6. (left) CMD of HLA stars within the central 2’ of Kron 3. The red arrow indicates the reddening vector, also defined as abscissa axis and parallel tothe arrow is defined as ordinate axis. (middle panel) distribution of stars in ordinate vs abscissa reference frame. Red line is the MS fiducial line in referenceframe. Black points are reference stars used to determine differential reddening. (right panel) Distribution of stars with respect to fiducial line. spatial distribution of reference stars from the ordinate vs 𝛿 abscissaplot and the blue points indicates nearby 49 stars around the targetstar (red point) of interest, which is further zoomed in the bottomleft panel. The top right panel shows distribution of 49 stars (blue)in ordinate vs 𝛿 abscissa diagram, over plotted on the distribution ofall the reference stars. The blue line indicated the median value of 𝛿 abscissa of that 49 stars which is slightly shifted from the median 𝛿 abscissa (red line) of reference stars. The difference (0.025) pro-vides us differential reddening suffered by the target star. The bottomright panel shows the histogram of 𝛿 abscissa of that 49 stars. Thisstep is repeated for all the reference stars to get the distribution ofdifferential reddening suffered by all the reference stars across thefield.After we subtracted 𝛿 abscissa to the abscissa of each star inthe rotated CMD, we obtain an improved CMD. We have furtherused that CMD to to derive a more accurate selection of the sampleof MS reference stars and derive a more precise fiducial line. Wehave iterated the procedure and after three iterations the procedurehas converged. The improved abscissa and ordinate values are thenconverted to 𝑉 and 𝐼 magnitudes. Now the comparison betweenthe original magnitudes with the corrected ones will provide usvariation in 𝐸 ( 𝑉 − 𝐼 ) for each star. Figure 8 shows the reddeningmap (variation in 𝛿𝐸 ( 𝑉 − 𝐼 ) ) of the cluster. The maximum variationin differential reddening suffered by the stars is ∼ ± As demonstrated in Sec. 4.1, Kron 3 exhibits very low levels ofintrinsic and differential reddening. In this section, we check forspatial variation of the photometric zero point due to small, un-modelable PSF variations. As for the case of differential reddening,we adopted the method described by Milone et al. (2012a), whichsuggests that the most evident manifestation is a shift in the color ofthe cluster sequence as a function of the location in the field (An-derson et al.2008). However, we have only one filter (NUV) with asufficient number of sources to construct a UV CMD. So, we areunable to carry out the method for NUV CCD. Whereas, in recentcalibration paper by Tandon et al. (2020), it is shown that for theSMC calibration field that there is no variation in PSF (FWHM of2.63 sub-pixels or 1.08 (cid:48)(cid:48) ) with 7.5 arcmin around the center of the UVIT CCD. Kron 3 is located within 3 arcmin of the center and assuch we expect there to be no variation in the PSF.The adopted method is applied to the HST data and is verysimilar to as described in section 4.1. The only difference is thatCMD has not been rotated with respect to the reddening vector, sothat we made the correction along the colour axis. All other steps aresame. The maximum colour variation is around 0.01 mag for HSTdata, which corresponds to 0.03 mag in
𝑁𝑈𝑉 − 𝑉 colour. So, theexpected spread due to PSF variation is also significantly smallerthan the observed spread seen in the RC stars in the UV.We found that photometric zero point variations (differentialreddening and/or PSF variation) cannot explain the extended RC inthe UV-optical CMD of Kron 3. Now we have to examine whetherfield contamination could account for the observed spread. In orderto ensure whether these bright stars are cluster members or not,it is important to remove field stars from the CMD. As the HSTdata covers only the central part of the cluster, it is not possible toremove field stars in optical or UV-optical CMDs. Hence, we used Gaia
DR2 to isolate field stars in the cluster CMD, as described inthe following section.
GAIA
DATA
We used the
Gaia
DR2 catalog and considered detections within 15 (cid:48) radius around the cluster center to study the field star distributionand its effect on the cluster properties.
Gaia
DR2 sources have largeastrometric errors in crowded fields such as in the central regionsof dense clusters, which are measured by astrometric_excess_noiseparameters (Lindegren et al. 2012, 2018). A zero value for this pa-rameter indicates a reliable astrometric solution, while high valuescorrelate with unreliable astrometric solutions. We selected starswith astrometric_excess_noise not exceeding a median value of 1.3mas at G=19 mag; faint sources in crowded regions correspond tohigher values. A large fraction ( ∼ ∼ (cid:48) Gaia
DR2 is not useful to study the core of this cluster.We have also used
Gaia
DR2 to determine the radius ofthe cluster. We calculated the radial density of stars (number ofstars/area) with a bin width of 0. (cid:48) ∼ (cid:48)
0. Therefore,
MNRAS000
MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure 7. (Top left) HST/ACS field of view of Kron 3 reference stars to calculate differential reddening. Blue points indicate nearby 49 stars around target star.(top right) ordinate vs 𝛿 abscissa distribution of reference stars and 49 nearby stars. Red and blue lines are median 𝛿 abscissa of reference stars and 49 nearbystars. (bottom left) zoomed version of top left plot. (bottom right) 𝛿 abscissa histogram of 49 stars. Median 𝛿 abscissa is also over plotted. the radius of the cluster based on optical Gaia DR2 is similar to thevalue obtained from UVIT-NUV data.In the top-left panel of Figure 11 we plotted G vs ( 𝐺 BP − 𝐺 RP )CMD for the cluster region (stars within a radius of 2. (cid:48) 𝐺 BP − 𝐺 RP ) CMDfor cross-matched Gaia stars (black points) on top of the CMD of
Gaia stars (grey points) in the cluster region; RC stars are marked inred. The isochrone was also plotted onto the CMD as in the previouspanel. RC stars extend over 0.4 mag in the colour axis in both figures.The bottom-left panel of Figure 11 shows the G vs (NUV − G) CMD of the cross-matched Gaia-UVIT stars (238) where RC stars spreadover two magnitudes in colour. The criteria used to cross-match theGaia-UVIT data is similar to that obtain HLA-UVIT cross-matcheddata. The bottom-right panel shows the NUV vs (NUV − G) CMDfor the same stars. Here, RC stars exhibit a similar spread in bothcolour and magnitude. Thus, a single isochrone is not able to fit thecolour extension of RC stars. To investigate whether the extensionis due to field star contamination, we statistically decontaminate thefield stars as follows.
MNRAS , 1–18 (0000)
P. K. Nayak et al.
Figure 8.
Differential reddening map of Kron 3.
Figure 9.
Differential reddening corrected CMDs of Kron 3.
Figure 10.
Radial density profile of Kron 3 in steps of 0. (cid:48)
Figure 11. (top-left) CMD of Gaia DR2 stars within the central 2’ of Kron3. (top-right) CMD of Gaia DR2 stars cross-matched with NUV data (black)over-plotted onto the full Gaia DR2 sample (grey) with red clump stars indi-cated in red. The same stars are plotted in the G vs (NUV-G) CMD (bottom-left) and in the NUV vs (NUV-G) CMD (bottom-right). The isochroneplotted in all panels is as in Figure 5.
Figure 12.
CMDs of stars within the central 2’ of Kron 3 cross-matchedbetween Gaia DR2 and NUV data for the full sample (left) and for onlycluster stars where field stars have been removed (right). CMDs for both Gvs ( 𝐺 BP − 𝐺 RP ) and NUV vs ( 𝐺 BP − 𝐺 RP ) are shown in top and bottomrows, respectively. The isochrone plotted in all panels is as in Fig. 5. To carry out a statistical process, we considered that field stars areuniformly distributed in the sky over a few arcmin around the cluster.An annular field region, with an inner radius of 3 (cid:48) .0 and an areaequal to that of the cluster region was chosen outside the cluster tocarry out the process. We first constructed G vs (NUV − G) CMD forthe cluster and field regions. The field stars within the cluster regionare then removed by considering each star in the field CMD and
MNRAS000
MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV finding its nearest counterpart in the cluster CMD. We considereda grid of [magnitude, colour] bins with different sizes, starting with[ Δ G , Δ (NUV − G)] = [0.02, 0.01] up to a maximum of [0.5, 0.25],where the units are in magnitude.In the top-left panel of Figure 12, we plotted the G vs (NUV − G)CMD for the cluster (black) and field region (cyan); RC stars aremarked in red. The top-right panel shows the CMD of the clusterafter subtracting field stars. We notice the presence of an extendedRC distribution even after the removal of field stars. The bottompanels show NUV vs (NUV-G) CMDs for the same stars as in thetop panels. This figure suggests that the extension of the RC is notdue to field contamination, rather it is an intrinsic property of thecluster. It is therefore essential to check all the intrinsic propertieswhich can contribute to make the RC stars brighter and hotter inUV-optical CMD.
As the RC stars are found to be members of the cluster and theobserved spread in RC is an intrinsic property of the cluster, then thepossible reasons for the extended RC could be (i) variable mass loss,(ii) presence of age and metallicity gradient among the stars, (iii)presence of multiple stellar population due to variation in the lightelemental abundances (CNO), (iv) variation in the initial heliumabundances within the cluster.Stars with higher mass loss rate can expose their inner hotterlayer by expelling the outer layer and get brighter in NUV. Starswith different metallicity values will appear in different location inthe CMD. The metal poor stars will appear brighter and bluer inthe CMD whereas the metal rich stars will appear relatively fainterand redder. So, the presence of metallicity gradient can cause theextended feature in the RC. This feature becomes more prominentin the UV-optical CMD as the UV region is more sensitive tometallicity. Study by Hollyhead et al. 2018 showed the presence ofvariation in C and N abundances in the RGB stars of this clusters. RCstars are not yet been reported for the presence of CNO variation.Presence of molecular bands (CN, NH, CH) in the NUV regioncan show the variation in NUV flux due to variation in the C-N-O abundances, which may produce the observed spread. Massiveclusters ( > 𝑀 (cid:12) ) in the Magellanic Clouds are found to host2G stars with enrichment in initial helium abundances (Chantereauet al. 2019). Here, we did not able to separate two distinct populationbut rather observed a spread in the CMD. It will be interestingto check if helium can be one of the reason behind the observedspread for this massive SMC cluster. Hence, there are more thanone possibilities to get a large spread in the RC. We have analyzedall the above mentioned possibilities to ensure we understand thereason for getting an extended RC in the cluster Kron 3. Tailo et al. 2019, 2020 showed that the mass loss during the red giantbranch(RGB) phase is an fundamental ingredients to constrain thehorizontal-branch (HB) morphology in the globular clusters (GCs).Therefore, we have also taken into account the effect of mass lossin the our study to check if mass loss can be one of the cause inthe extended features of RC. We have generated synthetic CMDsfor two different mass loss rate using BaSTI model for F555W andN242W filters and over plotted on observed NUV-optical CMDafter correcting for reddening and distance modulus. In Figure 13,blue and Cyan points shows two synthetic CMDs for mass loss rate
Figure 13.
NUU vs (NUV-V) CMD for HLA-UVIT cross-match data. BaSTIisochrones for different mass loss rate with fixed age has been over plottedon CMD.
To check whether the age and metallicity variation in Kron 3 canexplain the spread in the RC in the NUV, we used HLA-UVIT cross-matched data. In Figure 14 we plot V vs ( 𝑉 − 𝐼 ), V vs ( 𝑁𝑈𝑉 − 𝑉 )and NUV vs ( 𝑁𝑈𝑉 − 𝑉 ) CMDs highlighting RC stars. Then, weover-plotted isochrones on each CMD with ages ranging from 5to 8 Gyr, keeping the metallicity value constant at 𝑍 =0.001. Thisfigure suggests that although the age spread can explain the broadMS in the V vs ( 𝑉 − 𝐼 ) CMD, it is unable to fit the MS spread inboth the V vs ( 𝑁𝑈𝑉 − 𝑉 ) and NUV vs ( 𝑁𝑈𝑉 − 𝑉 ) CMDs. The (Vvs 𝑉 − 𝐼 ) CMD suggests that the cluster has a spread in age from6 to 8 Gyr. Despite the difference in age, the positions of the RCfrom the various isochrones are almost identical in both colour andmagnitude, and thus a simple age spread is unable to account forthe observed spread in the RC stars in the NUV.In Figure 15, we plot the same CMDs as shown in Fig. 14,however we now show isochrones of the same age, but with differentmetallicities ranging from Z=0.0002 to 0.002. The Figure showsthat isochrones of different metallicities are able to reproduce theobserved spread in the RC. We suggest that a spread in metallicityfrom Z=0.0002 to 0.002 is likely to be one of the reason behind theextended RC in the NUV.A similar spread in the NUV − optical colour has also beenobserved among RC stars in the Milky Way, which appears to bestrongly correlated with spectroscopic metallicities (Mohammedet al. 2019). These authors used a sample of 5175 RC stars, ob-served as part of the Sloan Digital Sky Survey Apache Point Obser-vatory Galactic Evolution Experiment (APOGEE) and combinedthese with GALEX All Sky Imaging Survey (GIAS; Martin et al.2005) and Gaia DR2 photometry, to construct NUV-optical CMDsof RC stars. They aimed to establish a relation between absolute(NUV − optical) colour and the spectroscopic metallicity in orderto photometrically determine the metallicity of any RC star, whosedistance is known. They noticed that a spread of 0.7 mag in absolute MNRAS , 1–18 (0000) P. K. Nayak et al.
Figure 14.
CMDs of the HLA-UVIT cross-matched sample within 2. (cid:48) 𝑉 − 𝐼 ), V vs ( 𝑁𝑈𝑉 − 𝑉 ) and NUV vs( 𝑁𝑈𝑉 − 𝑉 ), respectively. Red points denote the red clump stars. Isochrones for a range of ages (5-8 Gyr, marked with different colours) are over-plotted ontoeach CMD with the metallicity fixed at Z=0.001. Figure 15.
The same as Fig. 14 but over plotting isochrones with a range of metallicities (Z=0.0002-0.002, marked with different colours) at a constant age of7 Gyr. ( 𝐺 BP − 𝐺 RP ) increased to over 4 mag in absolute NUV − G colour.They found a strong correlation between (NUV − G) and [Fe/H] witha standard deviation of about 0.16 mag. Using stellar evolutionarymodels from MESA Isochrones and Stellar Tracks (MIST) project,Mohammed et al. (2019) also showed that the relation betweenNUV − G colour and metallicity for RC stars closely matches theobserved trend and it does not depend much on the initial mass andage. Hence, this relation can be used to estimate photometric metal-licities where a spectroscopic metallicity measurement is missing.This study supports the suggestion that the extension of RC stars ofKron 3 in the UV-optical CMD, is probably due to the presence ofa metallicity range within the cluster.In Figure 16, we plot the distribution of RC stars against colour.The left-panel shows the histogram of ( 𝑉 − 𝐼 ) colour of RC starswith a bin size of 0.04 mag for full HLA data of Kron 3 andHLA-UVIT cross-matched data, denoted in blue and green colourrespectively. Both distributions peak at 𝑉 − 𝐼 ∼ 𝑉 ) ∼ . If there is a metallicity gradient among the RC stars, then the NUVflux of metal poor stars should be higher than that of the metalrich stars. It is therefore necessary to calculate the expected fluxfor RC stars in different bands from theoretical synthetic spectraand to determine whether there is excess flux in the NUV withinthe metallicity range mentioned above. We used the atlas9 modelspectra (Castelli et al. 1997) obtained from the Virtual ObservatorySpectral Energy Distribution (SED) Analyzer (VOSA; Bayo et al.2008) tool to estimate the flux in different bands. The model spectracover a large range of the following parameters: the metallicityranges from [Fe/H]= − 𝑒 𝑓 𝑓 ranges from 3500 to 50000 K. [Fe/H] and log(g) vary insteps of 0.5 whereas T 𝑒 𝑓 𝑓 varies in steps of 250 K. In the isochronetable we found that RC stars have a surface gravity value log(g) ∼ 𝑒 𝑓 𝑓 = 5100-5600 K for a metallicity range Z=0.0002-0.002 at an MNRAS000
The same as Fig. 14 but over plotting isochrones with a range of metallicities (Z=0.0002-0.002, marked with different colours) at a constant age of7 Gyr. ( 𝐺 BP − 𝐺 RP ) increased to over 4 mag in absolute NUV − G colour.They found a strong correlation between (NUV − G) and [Fe/H] witha standard deviation of about 0.16 mag. Using stellar evolutionarymodels from MESA Isochrones and Stellar Tracks (MIST) project,Mohammed et al. (2019) also showed that the relation betweenNUV − G colour and metallicity for RC stars closely matches theobserved trend and it does not depend much on the initial mass andage. Hence, this relation can be used to estimate photometric metal-licities where a spectroscopic metallicity measurement is missing.This study supports the suggestion that the extension of RC stars ofKron 3 in the UV-optical CMD, is probably due to the presence ofa metallicity range within the cluster.In Figure 16, we plot the distribution of RC stars against colour.The left-panel shows the histogram of ( 𝑉 − 𝐼 ) colour of RC starswith a bin size of 0.04 mag for full HLA data of Kron 3 andHLA-UVIT cross-matched data, denoted in blue and green colourrespectively. Both distributions peak at 𝑉 − 𝐼 ∼ 𝑉 ) ∼ . If there is a metallicity gradient among the RC stars, then the NUVflux of metal poor stars should be higher than that of the metalrich stars. It is therefore necessary to calculate the expected fluxfor RC stars in different bands from theoretical synthetic spectraand to determine whether there is excess flux in the NUV withinthe metallicity range mentioned above. We used the atlas9 modelspectra (Castelli et al. 1997) obtained from the Virtual ObservatorySpectral Energy Distribution (SED) Analyzer (VOSA; Bayo et al.2008) tool to estimate the flux in different bands. The model spectracover a large range of the following parameters: the metallicityranges from [Fe/H]= − 𝑒 𝑓 𝑓 ranges from 3500 to 50000 K. [Fe/H] and log(g) vary insteps of 0.5 whereas T 𝑒 𝑓 𝑓 varies in steps of 250 K. In the isochronetable we found that RC stars have a surface gravity value log(g) ∼ 𝑒 𝑓 𝑓 = 5100-5600 K for a metallicity range Z=0.0002-0.002 at an MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure 16.
Distributions of RC stars as a function of ( 𝑉 − 𝐼 ) colour (left) and ( 𝑁𝑈𝑉 − 𝑉 ) colour (right) for the HLA data (blue) and the HLA-UVITcross-matched data (green). Figure 17.
Model spectra for a metallicity from Z=0.0002 to 0.006 (markedin different colours) are presented in the upper panel, for T 𝑒 𝑓 𝑓 =5250 K andlog(g)=2.5. In the lower panel the plot shows response curves for pass bandsused in this study as indicated. age of 7 Gyr. In Figure 14 we plot model spectra covering a rangeof metallicities, assuming a fixed value for the effective temperature( 𝑇 eff = 𝑔 =2.5 dex). Figure 17 showsthat there is a significant difference in flux for different metallicitiesat wavelengths below 400 nm, which is not prominent at longerwavelengths. In the lower panel we showed normalised responsecurves for different filters used in this study, taken from SpanishVirtual Observatory (SVO) filter profile service. Note that the NUVfilter N242W falls within the region where the variation in flux isnoticeable. The zoomed-in version of this region is shown in Figure18, to highlight the flux variation in the above mentioned metallicityrange. It is apparent that the continuum flux reduces to almost halfwhen the metallicity changes from Z=0.0002 to 0.002 at 5250 K.A flux decrease of a factor of two, corresponds to a magnitudedifference of 0.75 mag. We also noticed that absorption spectrallines get deeper with increasing metallicity causing a reduction Figure 18.
Zoomed-in version of Fig. 17 to show the absorption lines andflux variation prominently below 400 nm for different metallicities. in flux. Figures 17 and 18 suggest that the UV region is moresensitive to metallicity than optical and near-IR regions. As mostof the absorption lines due to metals appear in the UV, the higherthe metallicity the greater the absorption by metals present at thesurface of stars which will make it fainter. As there are more metalliclines present in the UV region than in the optical or near-IR regions,the variation in flux due to metallicity is more prominent in theUV region. This could be one of the reasons for some of the RCstars getting brighter in the NUV due to a low metallicity, whichcontributes to extending the RC in the NUV-optical CMDs, but notin optical CMDs.To get a quantitative estimate of the expected flux in all passbands, we convolved the filter responses with the model spectra fordifferent temperatures and metallicities using the VOSA tool. Wechose a temperature range 5000-5500 K and a [Fe/H] from − − MNRAS , 1–18 (0000) P. K. Nayak et al.
Figure 19.
The same as Fig. 15 but for a range of metallicity Z=0.0002-0.002 over plotting also the expected colours and magnitudes for RC stars for differenttemperatures (5000, 5250 and 5500 denoted by orange, blue and green points, respectively) and metallicities (Z=0.0002, 0.0006, 0.002 and 0.006) aftercorrecting for DM and reddening. (R 𝑐 /D) to calculate the absolute flux, where R 𝑐 is the average ra-dius and D the distance (10 pc) of RC stars. The absolute flux wasconverted to absolute magnitude ( 𝑀 ) using the standard relation, 𝑀 = − ∗ /F ), where F ∗ is the flux of the star and F is thezero-point flux in a given filter. Zero-point flux values were are pro-vided as part of VOSA. In Figure 19 we overlay the model-generatedcolours and magnitudes of RC stars after accounting for DM andreddening. The green points indicate the magnitude obtained fordifferent metallicities at a temperature of 5500 K. The blue and or-ange points denote the same for temperatures of 5250 and 5000 K,respectively. We also overlay isochrones for a range of metallicity(the same values as discussed earlier when describing Fig. 15) butat a fixed age of 7 Gyr. The observed RC clump stars are markedin red. The left panel shows the optical CMD (V vs 𝑉 − 𝐼 ) usingHLA data. The figure also shows that model generated magnitudesare clumped together for a constant temperature, suggesting thatthe RC population is not stretched in either colour or magnitude,due to metallicity variations. On the other hand, model generatedmagnitudes appear stretched in ( 𝑁𝑈𝑉 − 𝑉 ) colour at a constant tem-perature and show only a small variation in magnitude. The middlepanel shows that the range of temperatures and metallicities candescribe the extension of the RC, which gets more prominent in theNUV vs ( 𝑁𝑈𝑉 − 𝑉 ) CMD (right panel). In this panel, the extensionsof the RC stars in both colour and magnitude are represented wellby the theoretical models. This analysis supports and strengthensthe suggestion that RC stars in Kron 3 have a range of metallicities. To verify that Kron 3 contains RC stars with a range of metallicities,we fit the optical/near-IR spectral energy distributions (SEDs) of asample of Kron 3 RC stars using VOSA. This tool performs multipleiterations to fit the observed flux distribution with the theoreticalmodel flux for different combinations of T 𝑒 𝑓 𝑓 , log(g), [Fe/H] and 𝑀 𝑑 values, and gives the best fitted parameters after performing a 𝜒 minimization. The scaling factor ( 𝑀 𝑑 ) is used to scale the modelflux to match the observed flux and is defined as ( 𝑅 𝑐 / 𝐷 ) where 𝑅 𝑐 is the radius of the star and D is its distance. We provided theextinction value of the cluster, A 𝑉 = 0.082 (Glatt et al. 2008) asan input parameter to the VOSA tool. In this study we determined reduced 𝜒 , which is defined as 𝜒 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 = 𝑁 − 𝑛 𝑁 ∑︁ 𝑘 = ( 𝐹 𝑜,𝑘 − 𝑀 𝑑 × 𝐹 𝑚,𝑘 ) 𝜎 𝑜,𝑘 (1)where N is the number of photometric data points, n is the num-ber of input free parameters, 𝐹 𝑜,𝑘 is the observed flux and 𝐹 𝑚,𝑘 is the model flux. To obtain a better estimate of parameters, it isnecessary to have many data points covering a wide range of wave-lengths to generate SEDs and compare them with theoretical mod-els. Therefore, we included near-IR data from the VISTA survey ofthe Magellanic Clouds system (VMC; Cioni et al. 2011).We used RC stars detected in all wavebands from UVIT (NUV),HLA, Gaia DR2 and VMC. We found 18 such stars and generatedtheir SEDs after converting magnitudes to fluxes. We fitted theseSEDs with the same atlas9 model spectra as discussed in Sec.6.2.2. The Gaia , UVIT and VMC fluxes were fitted well with thespectra. As the HST wavelengths are covered by the
Gaia passbands, we excluded the HST fluxes in the fits. Initially we set allthe input parameters ([Fe/H], log(g) and temperature) as free valuesto obtain the best fit parameters. All the stars were fitted with asingle spectrum. We found that for most of the stars the parameterslie between − − ∼ (cid:12) and ∼ (cid:12) , respectively. Estimated values of effective temperature,luminosity and radius from SEDs are consistent with those expectedof RC stars (Gallenne et al. 2018; Wan et al. 2015). In Figure 20we show model SED fits for two of these stars. The observed andexpected model fluxes are denoted in blue and red, respectively, ontop of the best fit theoretical spectrum (grey). In the left panel theobserved fluxes of the RC star in different bands are fitted with amodel spectrum of [Fe/H], log(g) and T 𝑒 𝑓 𝑓 values of − 𝑒 𝑓 𝑓 as − − model flux)/observed flux) for all filters, which appear all close tozero. 𝜒 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 values for the fitted SEDs presented in Figure 20 MNRAS000
Gaia passbands, we excluded the HST fluxes in the fits. Initially we set allthe input parameters ([Fe/H], log(g) and temperature) as free valuesto obtain the best fit parameters. All the stars were fitted with asingle spectrum. We found that for most of the stars the parameterslie between − − ∼ (cid:12) and ∼ (cid:12) , respectively. Estimated values of effective temperature,luminosity and radius from SEDs are consistent with those expectedof RC stars (Gallenne et al. 2018; Wan et al. 2015). In Figure 20we show model SED fits for two of these stars. The observed andexpected model fluxes are denoted in blue and red, respectively, ontop of the best fit theoretical spectrum (grey). In the left panel theobserved fluxes of the RC star in different bands are fitted with amodel spectrum of [Fe/H], log(g) and T 𝑒 𝑓 𝑓 values of − 𝑒 𝑓 𝑓 as − − model flux)/observed flux) for all filters, which appear all close tozero. 𝜒 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 values for the fitted SEDs presented in Figure 20 MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV are found to be 18.1 and 10.6. The model fitted SEDs of the rest ofthe 16 stars are presented in the appendix in Figures A1 and A2.The values of model fitted parameters are noted in those figures.Theoretical stellar evolutionary models suggest that RC starshave a surface gravity of 2.5 dex and that this values is insensitiveto changes in both age and metallicity. Therefore, we tried to fit theSEDs keeping log(g) fixed at 2.5 and leaving [Fe/H] and T 𝑒 𝑓 𝑓 asfree parameters. We obtained that the best fit spectra for RC starsspan a smaller range of metallicities, from − 𝜒 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 valuesdid not show much variation with respect to the previous estimatesexcept for two stars. In Figure 21, we show the distribution of [Fe/H]derived using SED fitting for 18 stars. The blue histogram representsthe distribution of [Fe/H] when all the input parameters are leftfree to vary, whereas the green histogram shows the distribution of[Fe/H] when log(g)=2.5 and both [Fe/H] and T 𝑒 𝑓 𝑓 are left to vary.The distribution of [Fe/H] spans a range of values from − − − Almost all GCs shows split or broad red HB, that is mainly dueto the combined effects of variations in light elemental abundancesand helium. CN, CH and NH bands coincide with different photo-metric filters, and stars with different level of C and N exhibit fluxvariations in those filters, thereby leading to distinct populationssplit in colour and magnitude (Milone et al. 2020b, 2012b; Piottoet al. 2015). Lagioia et al. (2019) found helium Variation in fourSmall Magellanic Cloud Globular Clusters. Chantereau et al. (2019)found a spread in helium abundance in an intermediate age SMCcluster, Lindsay 1. Given that we have demonstrated the extendednature of RC stars in Kron 3, when viewed in the UV-opitcal CMD,we need to investigate whether variations in both the elemental andinitial helium abundances could produce such a spread.
To check the effect of CNO variation in Kron 3, we first gener-ated synthetic spectra for RC stars for a fixed temperature (5250K)and logg (2.5) but for different elemental abundances. The valuesof temperature and logg values are obtained from Padova theoreti-cal isochrone model. We have employed the one-dimensional localthermodynamic equilibrium LTE stellar atmospheric models (AT-LAS9; Castelli & Kurucz 2004) and the spectral synthesis codeTURBOSPECTRUM (Alvarez & Plez 1998) to construct syntheticspectra. Version 12 of the turbo-spectrum code for spectrum syn-thesis and abundance estimates was used for the analysis. We havegenerated the spectra by varying the abundances of C and N for aspectral range from 1500 to 10500 angstrom (for the adopted stel-lar model) in order to measure the variation in flux. Taking solarabundances as the reference, we have varied [N / Fe] between -1.0to +1.50 dex in steps of 0.50, whereas [C / Fe] was varied between0.0 and -1.0 dex following the typical abundance range for globularclusters of a given metallicity. After generating the spectra, we haveconvolved them with filters’ response curves using VOSA moduleto calculated the expected flux in each filter for different elemen-tal abundances. We converted the model flux to expected observedflux using scaling factor ( 𝑀 𝑑 = ( 𝑅 𝑐 / 𝐷 ) ). The equation for flux tomagnitude conversion has been used to convert the expected flux to magnitude unit. The zero point flux values for the corresponding fil-ter system. As the converted magnitudes are Vega systems, we usedthe Vega to AB conversion factors for corresponding filters fromMESA website. The converted AB magnitude then over plotted onobserved (NUV – V) vs NUV CMD after correcting for extinction.Figure 22 shows the UV-optical CMD of the cluster where RCstars are highlighted with red points. All the points for the range of[N / Fe] abundances (-1.0 to +1.50 dex) and for the range for [C / Fe]abundances (0.0 to -1.0 dex) merge together in a single blue pointin the Figure 22. The Figure suggest that C and N variations areunable to produce the observed spread. In the Figure 23, we haveplotted synthetic spectra for different N abundances in the spectralregion of F555W and N242W filter systems. The Figure clearlyshows that there is almost no variation in the flux for variation inN abundances. As shown in figure 32 of Milone et al. 2012b, Nvariation affects are seen more near ∼ ∼ Chung et al. (2013, 2017) showed that implications and prospectsfor the helium-enhanced populations in relation to the second-generation populations found in the Milky Way GCs using YonseiEvolutionary Population Synthesis (YEPS) model. The model pro-vides us Yonsei–Yale (Y ) stellar evolutionary tracks and BaSel 3.1flux libraries. To construct synthetic CMD for Helium-enhancedred HB stellar populations , we have used Y stellar libraries withenhanced initial helium abundances (Y 𝑖𝑛𝑖 ) (Lee et al.2015). Wechoose two values for Y 𝑖𝑛𝑖 as 0.23 and 0.28, at fixed Z value of0.001 ([Fe/H]=-1.5 dex) and age of 7 Gyr. We have also consid-ered 𝛼 -enhancement [ 𝛼 /Fe]=0.3, under the 𝛼 -elements mixture ofKim et al.(2002). We chose Salpeter’s IMF for the model (Salpeter1955). The left panel of Figure 24 shows that synthetic CMDs fortwo different Y 𝑖𝑛𝑖 values 0.23 (blue) and 0.28 (black), over plottedon the observed NUV-optical CMD where observed RC stars arehighlighted in red. The right hand panel shows the zoomed versionof RC or red HB population. The observation photometric errorshave been included to generate the synthetic CMD. We notice thatthe spread in Y 𝑖𝑛𝑖 can produce NUV bright stars and is able toexplain a large fraction of observed extension in the RC. Therefore,we can suggest that initial variation in helium is one of the reasonbehind the observed spread in RC population.Now, to check if a small spread in age and metallicity alongwith the variation in Y 𝑖𝑛𝑖 can accout for the full spread observedin RC, we generated synthetic CMD for HB stars for two differentcombinations of age and metallicity along with initial helium vari-ation of Y 𝑖𝑛𝑖 =0.23 & 0.28: (1) metal-poor ([Fe/H]=-1.5 dex) andslightly younger (6.5 Gyr); (2) metal-rich ([Fe/H]=-1.3 dex) andslightly older age (7.5 Gyr). In Figure 25, We show synthetic CMDsfor these combinations on observed RC stars, which clearly sug-gests that a spread in helium abundance along with a small spreadin age and metallicity able to produce observed extension in the redclump. So, we can suggests that the cluster hosts stars with a small MNRAS , 1–18 (0000) P. K. Nayak et al.
Figure 20.
SEDs of RC stars are presented in UVIT, Gaia DR2 and VMC data. The blue and red points represent the observed and expected fluxes, respectivelyin different passbands.
Figure 21.
Distribution of metallicity ([Fe/H]) values for RC stars obtainedfrom the fit of their SEDs with theoretical models when metallicity, tem-perature and gravity are allowed to vary (blue) and when metallicity andtemperature vary while gravity is fixed at log(g)=2.5 (green). spread in age and metallicity along with a variation in initial heliumabundances.
Using both the UVIT and
𝐺𝑎𝑖𝑎 -DR2 data sets, it is found that thedistribution of stellar density around the centre of Kron 3 matcheswith the field star distribution at a distance of ∼ (cid:48) , suggesting theradius of Kron 3 to be ∼ (cid:48) ( ∼
35 pc). This estimate is slightly largerthan a previous estimate of 1. (cid:48)
Figure 22.
NUV vs (
𝑁𝑈𝑉 − 𝑉 ) CMD of the HLA-UVIT cross-matchedsample within 2. (cid:48) compared to an estimate (2. (cid:48)
4) by Rich et al. (1984). Our resultdoes not support the claim by Alcaino et al. (1996), where theauthors suggested a radius of 6 (cid:48) , as we do not see any variation inthe star count beyond 2 (cid:48) in both the optical and NUV.In general, the presence multiple stellar populations can beattributed to large variations in the elemental abundances and havebeen identified in Galactic globular clusters, which were oncethought to be chemically homogeneous. Similar variations haverecently been found in the only (cid:48) classical (cid:48) globular cluster NGC121 ( ∼
10 Gyr old) and a few intermediate-age ( ∼ 𝐹 𝑊 ,𝐹
𝑊 ,𝐹 𝑁 = (F336W − F438W)
MNRAS000
MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure 23.
Model generated synthetic spectra for different N abundances [N / Fe] = -1.0 to +1.50 dex (marked in different colours) are presented in the upperpanels, temperature 5250 K and log(g)=2.5. In the lower panel the plot shows response curves for N242W and F555W pass bands.
Figure 24.
NUV vs (
𝑁𝑈𝑉 − 𝑉 ) CMD of the HLA-UVIT cross-matched sample within 2. (cid:48) 𝑖𝑛𝑖 ) of 0.23 and 0.28, respectively, but for a fixed age (7 Gyr) and metallicity [Fe/H]= − Figure 25.
The same as Fig. 24 but for different age and metallicity combi-nations. − (F438W − F343N). If C and N abundance variations are presentwithin a given cluster, the use of the pseudo-colour-magnitude di-agram acts to separate the different RGB populations in a way thatstandard colour-magnitude diagrams are unable to do. The fractionof N enriched and C poor stars found in NGC 121, NGC416, NGC339 and Lindsay 1 are ∼
32, 45, 25 and 36 per cent, respectively(Niederhofer et al. 2017a,b). These fractions are found to be lowerthan the average value for Galactic globular clusters. Hollyhead et al.(2017, 2018) provided spectroscopic evidence of N-enriched starsin the RGB population of Lindsay 1 and Kron 3. All the clustersmentioned above are equally massive with a mass range of 1-2 × M (cid:12) and located in different regions of the SMC, hence mass andenvironmental effects are not key for the presence of chemical en-richment in these cluster (Martocchia et al. 2017). The morphologyof the RC in these clusters has not been studied so far; here wepresent the first detailed analysis of the RC in Kron 3, one of theyoungest clusters where elemental abundance variations identified.UVIT images in the NUV allow us to detect RC stars in Kron MNRAS , 1–18 (0000) P. K. Nayak et al. ∼ (cid:12) and ∼ (cid:12) , respectively, which areconsistent with those expected of RC stars (Gallenne et al. 2018;Wan et al. 2015).All three methods suggest that the RC stars span a wide rangein metallicity and Kron 3 is probably the first cluster where this hasbeen quantified. A spectroscopic abundance study of RC stars inKron 3 has not been performed so far. Dias et al. (2010) estimated themetallicity of Kron 3 as Z=0.0002, which falls within the metallicityrange found in this study. Our study supports that Kron 3 is arelatively young cluster of age 7 Gyr with respect to ancient GCshosting multiple stellar populations.Metallicity gradient among the stars suggest that variation inthe strength of absorption lines due to variation by Mg, C, Fe II,Iron peak elements etc in the NUV region causes the extended RC.Whereas, the variation in the light elemental abundance (He, C,N) can also cause variation in molecular absorption line as well asin continuum. Therefore, we have also examined whether effect ofvariation in elemental abundances in RC stars. We generate syntheticCMDs for C and N abundances for a fixed temperature (5250K),logg (2.5), metallicity (Z=0.001) and convolve with filter response curve to observed spread in UV-optical CMD. We found almost novariation in the CMD, suggest that elemental abundance variationis not the cause of observed spread in RC. As, previous study byHollyhead et al. 2018 suggested presence of C and N variations inthe RGB stars but we do not see any signature of that in RC starsusing NUV filter, so we also suggest that NUV filter might not besensitive to the elemental abundance variation.We also tried to check if spread in initial helium abundance(Y 𝑖𝑛𝑖 ) can mimic the observed extension in the RC, as previousstudy by Chantereau et al. 2019 showed the presence of variation ininitial helium abundances in the RC distribution of a very similar(having similar age, metallicity and reddening) SMC cluster Lindsay1. We used alpha enhanced ([ 𝛼 /Fe]) Y isochrone model to generatesynthetic CMD of RC stars for two different initial value of helium(0.23 and 0.28) keeping age and metallicity constant as average valuefor the cluster 7 Gyr and Z=0.001 (or [Fe/H]= − 𝑖𝑛𝑖 can mimic the observed spread in RC to a large extend. Wehave further found that combination of a small spread in metallicityby 0.2 dex, a small age spread of 6.5-7.5 Gyr along with variationin Y 𝑖𝑛𝑖 able to produce the observed spread very nicely.In optical CMD, RC stars have an almost constant luminositythroughout their lifetime but T 𝑒 𝑓 𝑓 can vary depending on theirmetallicities and initial masses (Girardi 2016; Choi et al. 2016). Fora constant stellar mass, the luminosity of an RC star varies with themetallicity. Stars with a lower metallicity will have higher luminos-ity and vice versa (Girardi 2016). In our study, we found that com-bination effect of variation in Y 𝑖𝑛𝑖 and a small spread in metallicitycan cause a large spread in NUV luminosity of RC stars which showa constant luminosity and very compact distribution in the opticalCMD. Mohammed et al. (2019) observed NUV bright RC stars inthe solar neighborhood and found that the absolute (NUV − G) colouris strongly correlated with spectroscopic metallicities, whereas theabsolute ( 𝐺 BP − 𝐺 RP ) colour has a weak correlation with metallicityand a large scatter. Bluer stars tend to be hotter and have a lowermetallicity and vice versa. This relation can be used to obtain pho-tometric metallicities from other stars in the same CMD space asRC candidates using only their UV-optical colour. They also foundthat metallicity vs absolute (NUV − G) colour in the MIST modelsis only weakly dependent on the initial mass and age during the RCphase of evolution. Though they have not taken care of the effectof initial helium variation among the stars. Our study of Kron 3suggests that the NUV − optical colour can be used as a tool to in-vestigate the presence of multiple stellar populations due to initialhelium variation and also for metallicity spread within clusters.After taking into account the effect helium variation, we ex-clude the possibility of a large spread in metallicity. We suggeststhat the observed spread in RC is an combined effect of photometricerror, variation in Y 𝑖𝑛𝑖 , variable mass loss, a small spread in metal-licity and age. We are not quantifying the value of initial heliumspread, age and metallicity spread as there will be an uncertaintiesin those values due to relatively large photometric error in fainterend of NUV magnitude and to degeneracy between metallicity andhelium spread. The recent studies by Marino et al. 2019; Miloneet al. 2020a,b found the presence of Type II GCs in the MilkyWay using universal chromosome map (ChM). All the GCs showsboth 1G and 2G population, suggesting the multiple populationsin them. The universal ChM helps to reveal the internal variationof elemental abundances or metallicity ( > MNRAS , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV stars, which is explained by both variation in metallicity ( ∼ We summarize the results of our study of the Kron 3 cluster in theSMC as follows.We present the analysis of UVIT-HST-
Gaia -VISTA data for theintermediate-age cluster Kron 3 in the SMC. For the first time, wereport the identification of NUV bright RC stars and the extensionof the RC in the CMD. This study thus demonstrates the powerof UVIT-HST-
Gaia -VISTA combination to study clusters in theMagellanic Clouds. We take advantage of the high spatial resolutionof the HST in the central region of the cluster and of the wide-areacoverage of Gaia in combination with the UVIT data. We find thatthe extended RC is an intrinsic property of the cluster and that it isnot due to the influence of field stars. We estimated the radius of thecluster as 2. (cid:48)
Gaia
DR2 data.We suggest that the cluster exhibits multiple stellar populationswith a small range in age (6.5-7.5 Gyr) and metallicity (0.2 dex)along with a variation in initial helium abundance Y 𝑖𝑛𝑖 =0.23 to0.28. We suggest that NUV filter can be used as a tool to investigatethe presence variation in initial helium abundance and metallicityspread within a cluster. A spectroscopic follow-up study of RC starsare suggested to check if Kron 3 is a probable candidate of Type IIGCs in the SMC.
ACKNOWLEDGEMENTS
This research was supported by a DST-DAAD exchange grant(REMAP) between the Indian Institute of Astrophysics (IIA, Ban-galore) and the Leibniz Institute for Astrophysics Potsdam (AIP,Potsdam). This paper makes use of observations collected at theEuropean Organisation for Astronomical Research in the South-ern Hemisphere under ESO programme 179.B-2003. We thank theCASU and the WFAU in Edinburgh for providing calibrated dataproducts under the support of the Science and Technology FacilityCouncil (STFC) in the UK. M.-R.L. Cioni and Cameron P. M. Bellacknowledge support from the European Research Council (ERC)under the European Union’s Horizon 2020 research and innova-tion programme (grant agreement no. 682115). S. Subramanianacknowledges support from the Science and Engineering ResearchBoard (SERB), India through Ramanujan Fellowship. We thank thereferee for valuable suggestions to help in improving the manuscript.P. K. Nayak thank William Chantereau and Nate Bastian for valuablediscussion with them.
REFERENCES
Alcaino G., Liller W., Alvarado F., Kravtsov V., Ipatov A., Samus N.,Smirnov O., 1996, AJ, 112, 2004Alvarez R., Plez B., 1998, A&A, 330, 1109Anderson J., Piotto G., King I. R., Bedin L. R., Guhathakurta P., 2009, ApJ,697, L58Bastian N., Kamann S., Cabrera-Ziri I., Georgy C., Ekström S., CharbonnelC., de Juan Ovelar M., Usher C., 2018, MNRAS, 480, 3739 Bayo A., Rodrigo C., Barrado Y Navascués D., Solano E., Gutiérrez R.,Morales-Calderón M., Allard F., 2008, A&A, 492, 277Bica E., Bonatto C., Dutra C. M., Santos J. F. C., 2008, MNRAS, 389, 678Castelli F., Kurucz R. L., 2004, ArXiv Astrophysics e-prints,Castelli F., Gratton R. G., Kurucz R. L., 1997, A&A, 318, 841Chantereau W., Salaris M., Bastian N., Martocchia S., 2019, MNRAS, 484,5236Choi J., Dotter A., Conroy C., Cantiello M., Paxton B., Johnson B. D., 2016,ApJ, 823, 102Chung C., Yoon S.-J., Lee S.-Y., Lee Y.-W., 2013, ApJS, 204, 3Chung C., Yoon S.-J., Lee Y.-W., 2017, ApJ, 842, 91Cioni M.-R. L., et al., 2011, A&A, 527, A116Conroy C., Gunn J. E., 2010, The Astrophysical Journal, 712, 833Conroy C., Gunn J. E., White M., 2009, The Astrophysical Journal, 699,486Cordoni G., Milone A. P., Marino A. F., Di Criscienzo M., D’Antona F.,Dotter A., Lagioia E. P., Tailo M., 2018, ApJ, 869, 139Dias B., Coelho P., Barbuy B., Kerber L., Idiart T., 2010, A&A, 520, A85Gallenne A., et al., 2018, A&A, 616, A68Gascoigne S. C. B., 1966, MNRAS, 134, 59Gascoigne S. C. B., 1980, in Hesser J. E., ed., IAU Symposium Vol. 85, StarClusters. pp 305–313Girardi L., 2016, ARA&A, 54, 95Glatt K., et al., 2008, AJ, 136, 1703Gossage S., et al., 2019, arXiv e-prints, p. arXiv:1907.11251Hodge P. W., 1982, ApJ, 256, 447Hollyhead K., et al., 2017, MNRAS, 465, L39Hollyhead K., et al., 2018, MNRAS, 476, 114Lagioia E. P., Milone A. P., Marino A. F., Dotter A., 2019, ApJ, 871, 140Li C., Sun W., de Grijs R., Deng L., Wang K., Cordoni G., Milone A. P.,2019, ApJ, 876, 65Lim B., Rauw G., Nazé Y., Sung H., Hwang N., Park B.-G., 2019, NatureAstronomy, 3, 76Lindegren L., Lammers U., Hobbs D., O’Mullane W., Bastian U., HernándezJ., 2012, A&A, 538, A78Lindegren L., et al., 2018, A&A, 616, A2Mackey A. D., Broby Nielsen P., Ferguson A. M. N., Richardson J. C., 2008,ApJ, 681, L17Marigo P., Girardi L., 2007, A&A, 469, 239Marigo P., Girardi L., Bressan A., Groenewegen M. A. T., Silva L., GranatoG. L., 2008, A&A, 482, 883Marino A. F., Milone A. P., Casagrande L., Przybilla N., Balaguer-NúñezL., Di Criscienzo M., Serenelli A., Vilardell F., 2018, ApJ, 863, L33Marino A. F., et al., 2019, MNRAS, 487, 3815Martin D. C., et al., 2005, ApJ, 619, L1Martocchia S., et al., 2017, MNRAS, 468, 3150Mighell K. J., Sarajedini A., French R. S., 1998, AJ, 116, 2395Milone A. P., et al., 2008, ApJ, 673, 241Milone A. P., Bedin L. R., Piotto G., Anderson J., 2009, A&A, 497, 755Milone A. P., et al., 2012a, A&A, 540, A16Milone A. P., et al., 2012b, ApJ, 744, 58Milone A. P., et al., 2020a, MNRAS, 491, 515Milone A. P., et al., 2020b, MNRAS, 497, 3846Mohammed S., Schiminovich D., Hawkins K., Johnson B., Wang D., HoggD. W., 2019, ApJ, 872, 95Niederhofer F., et al., 2017a, MNRAS, 464, 94Niederhofer F., et al., 2017b, MNRAS, 465, 4159Piotto G., 2009, in Mamajek E. E., Soderblom D. R., Wyse R. F. G.,eds, IAU Symposium Vol. 258, The Ages of Stars. pp 233–244,doi:10.1017/S1743921309031883Piotto G., et al., 2007, ApJ, 661, L53Piotto G., et al., 2015, AJ, 149, 91Postma J. E., Leahy D., 2017, PASP, 129, 115002Rich R. M., Da Costa G. S., Mould J. R., 1984, ApJ, 286, 517Salpeter E. E., 1955, ApJ, 121, 161Stetson P. B., 1987, PASP, 99, 191Subramaniam A., et al., 2016, in Space Telescopes and InstrumentationMNRAS , 1–18 (0000) P. K. Nayak et al. arXiv:1608.01073 ),doi:10.1117/12.2235271Sun W., de Grijs R., Deng L., Albrow M. D., 2019, ApJ, 876, 113Tailo M., Milone A. P., Marino A. F., D’Antona F., Lagioia E. P., CordoniG., 2019, ApJ, 873, 123Tailo M., et al., 2020, MNRAS, 498, 5745Tandon S. N., et al., 2017a, Journal of Astrophysics and Astronomy, 38, 28Tandon S. N., et al., 2017b, AJ, 154, 128Tandon S. N., et al., 2020, AJ, 159, 158Vandenberg D. A., Bell R. A., 1985, ApJS, 58, 561Wan J.-C., Liu C., Deng L.-C., Cui W.-Y., Zhang Y., Hou Y.-H., Yang M.,Wu Y., 2015, Research in Astronomy and Astrophysics, 15, 1166van den Bergh S., 1991, ApJ, 369, 1
APPENDIX A: MODEL FITTED SEDS OF RC STARS
This paper has been typeset from a TEX/L A TEX file prepared by the author. MNRAS000
This paper has been typeset from a TEX/L A TEX file prepared by the author. MNRAS000 , 1–18 (0000) ron 3: A cluster with an extended red clump in the UV Figure A1.
SEDs of eight RC stars are presented in UVIT, Gaia DR2 and VMC data. The blue and red points represent the observed and expected fluxes,respectively in different passbands.MNRAS , 1–18 (0000) P. K. Nayak et al.
Figure A2.
The Figure represents the same as Fig. A1, but for another eight RC stars. MNRAS000