A morphological study of galaxies in ZwCl0024+1652, a galaxy cluster at redshift z $\sim$ 0.4

The well-known cluster of galaxies ZwCl0024+1652 at z $\sim$ 0.4, lacks an in-depth morphological classification of its central region. While previous studies provide a visual classification of a patched area, we used the public code called galaxy Support Vector Machine (galSVM) and HST/ACS data as well as WFP2 master catalogue to automatically classify all cluster members up to 1 Mpc. galSVM analyses galaxy morphologies through Support Vector Machine (SVM). From the 231 cluster galaxies, we classified 97 as early-types (ET) and 83 as late-types (LT). The remaining 51 stayed unclassified (or undecided, UD). By cross-matching our results with the existing visual classification, we found an agreement of 81%. In addition to previous Zwcl0024 morphological classifications, 121 of our galaxies were classified for the first time in this work. In addition, we tested the location of classified galaxies on the standard morphological diagrams, colour-colour and colour-magnitude diagrams. Out of all cluster members, $\sim$20% are emission line galaxies (ELG), taking into account previous GLACE results. We have verified that the ET fraction is slightly higher near the cluster core and decreases with the clustercentric distance, while the opposite trend has been observed for LT galaxies. We found higher fraction of ET (54%) than LT (46%) throughout the analysed central region, as expected. In addition, we analysed the correlation between the five morphological parameters (Abraham concentration, Bershady-Concelice concentration, Asymmetry, Gini and M20 moment of light) and clustercentric distance, without finding a clear trend. Finally, as a result of our work, the morphological catalogue of 231 galaxies containing all the measured parameters and the final classification is available in the electronic form of this paper.


INTRODUCTION
A consolidated observational fact is the outstanding difference in the properties of galaxies located in the cores (or regions of high local galaxy density) and in the external parts (or low density ones) of low-and intermediate-redshift clusters: the former regions are dominated by red, massive and passive early-type galaxies (ET galaxies, comprising elliptical and S0), while a substantial increase of the fraction of late-type galaxies (LT, comprising spiral and irregular objects) is observed in the latter. This was early identified by Zwicky (1942), and quantified by Dressler (1980) in the so-called morphology-density relation linking the increasing fraction of ET galaxies with local galaxy density. Similarly, a decrease of the fraction of star forming (SF) galaxies is observed with increasing local galaxy density (the SF-density relation, see for instance Pintos-Castro et al. 2013, and references therein). Moreover, these relations evolve with cosmic time, as was realized by Butcher & Oemler (1978), who found that cluster galaxy populations evolve as redshift changes in such a way that rich clusters at higher redshift (z > 0.2) are populated with a higher fraction of blue galaxies than low redshift clusters. This is the so-called Butcher-Oemler (BO) effect. Likewise, an increase of the cluster SF and active galactic nuclei (AGN) activity is observed (see for instance Haines et al. 2009;Martini et al. 2013).
The morphology-density relation seems to hold from nearby clusters up to redshifts as high as z ∼ 1.5 (e.g. Dressler et al. 1997;Postman et al. 2005;Holden et al. 2007;Mei et al. 2012;Nantais et al. 2013). Likewise, star formation takes place in low density regions where LTs dominate while high density regions are dominated by quiescent ET galaxies since z ∼ 1.5 to the local universe (e.g. Postman & Geller 1984;Kauffmann et al. 2004;Cooper et al. 2012;Wetzel et al. 2012;Woo et al. 2013). At higher redshift, there is some controversial evidence of the existence of a reversal of the SF-density relation: some authors, as Tran et al. (2010) find an (even dramatic) increase of the fraction of SF galaxies from low-to high-density regions in clusters at z ∼ 1.6, while other authors (e.g. Ziparo et al. 2014) do not find a clear evidence of such type of reversal when studying clusters at the same redshift. Quadri et al. (2012), using mass-selected samples from the UKIDSS Ultra-Deep Survey, conclude that galaxies with quenched SF tend to reside in dense environments out to at least z ∼ 1.8.
The structural and morphological properties of a galaxy are important tracers of its evolutionary stage. Thus, the correlation of the morphology (and/or SF activity) of the clusters' galaxies with the local density provides valuable information on the stage of infall at which galaxies experience the bulk of their transformations. To this end, it is important to perform wide-area surveys (to study the densitydependent effects) in clusters that span a range of redshifts (to assess the evolution with cosmic time).
The morphological taxonomy of galaxies can be backdated to Reynolds (1920). Visual inspection is the traditional method, and even now a very common way to perform morphological classification of galaxies (e.g. Lintott et al. 2008;Nair & Abraham 2010;Fasano et al. 2012;Kocevski et al. 2012;Kartaltepe et al. 2012Kartaltepe et al. , 2015Buitrago et al. 2013;Kuminski & Shamir 2016;Willett et al. 2013;Simmons et al. 2017; Willett et al. 2017). One of the drawbacks of the visual classification method is the subjectivity, that can be alleviated by performing multiple instances of the classification of each object carried out by different persons; an outstanding example is Galaxy Zoo project (Lintott et al. 2008(Lintott et al. , 2011. In the framework of this "citizen science" initiative, nearly one million galaxies from the Sloan Digital Sky Survey (SDSS) were classified by ∼ 10 5 participants who performed more than 4 × 10 7 classifications. Needless to say, when dealing with large number of sources, the visual classification method can be really time-consuming. It works well for closer and well resolved objects for accurate estimation. For such objects it agrees with the results of modern classification methods.
But with a currently overgrowing observational astronomical data, the above method is probably not the most appropriate or even unfeasible for high-redshift galaxies. Modern classification techniques include galaxy fitting algorithms, which can give reliable results for a large number of galaxies in a relatively shorter period and with minimal human resources. To deal with fast growing and big astronomical data, machine learning techniques employing Convolutional Neural Networks (CNN) are widely under use recently for morphological classification of galaxies (e.g. Banerji et al. 2010;Kuminski et al. 2014;Dieleman, Willett & Dambre 2015;Huertas-Company et al. 2015;Aniyan & Thorat 2017;Domínguez et al. 2018;Lukic et al. 2018). Modern galaxy classification methods can either be parametric or non-parametric.
Parametric methods use some parameters of the galaxies to classify them by fitting (one or two dimensional) mathematical models to their images assuming some predefined parametric model. In this approach Sérsic profile (Sérsic 1963) and a two-component profile (bulge + disk decomposition) are the commonly used models. The classification is obtained by fitting a two component profile as described in detail by Simard et al. (2002) and Peng et al. (2002). More recently, Simard et al. (2011) has performed a classification of 1.12 million galaxies using a bulge + disk decomposition approach with SDSS data release seven (Abazajian et al. 2009). In addition to this, a structural and morphological catalogue of 45 million sources have been presented by Tarsitano et al. (2018) with the Dark Energy Survey (The DES Collaboration 2016) data of the first year observation employing both a single Sérsic parametric fits and non-parametric methods. Parametric method in general is useful in that it gives a complete set of parameters describing the quantitative morphology. Since a large number of parameters need to be fitted, the results may be degenerated as shown in Huertas-Company et al. (2007). Degeneracy occurs as a result of correlation between parameters, the results of the local minima in the parameter space of the chi-square minimization, or by numerical divergence in the process of fitting (Peng et al. 2002(Peng et al. , 2010. The peculiar characteristic of the parametric method in general is the assumption that a galaxy is described well by a simple analytic model whereas this does not always work for well resolved as well as irregular and merging/interacting galaxies. On the other hand, the non-parametric approach does not assume any specific analytic model and is performed on the basis of measuring a set of well-chosen observables. The effects of seeing, being one of the major challenges in galaxy fitting, are not included in non-parametric mea-surements unlike the parametric ones where the assumed mathematical model is convolved with the PSF. The nonparametric method was introduced for the first time by Abraham et al. (1994Abraham et al. ( , 1996 with the definition of two observables: the Abraham concentration index and asymmetry. A third quantity, namely smoothness, was introduced by Conselice et al. ( , 2003. The classification has been further enhanced with additional observables: the GINI coefficient (Abraham et al. 2003); M20 moment of light (Lotz et al. 2004) and Conselice-Bershady concentration Bershady et al. 2000). These six parameters, together with ellipticity are described in more details in subsection 3.3. Non-parametric methods are in advantage when classifying large sample of galaxies at higher redshifts, when lower resolution data are available (e.g. Scarlata et al. 2007;Tasca et al. 2009;Pović et al. 2009Pović et al. , 2013Pović et al. , 2015Pintos-Castro et al. 2016). Furthermore, no analytic predefined profile is required in this approach.
In this paper, we apply a non-parametric classification method to a well-known intermediate redshift cluster, namely ZwCl0024+1652 at z = 0.395. This cluster has been extensively studied by several groups (e.g. Morrison et al. 1997;Broadhurst et al. 2000;Kneib et al. 2003;Treu et al. 2003;Moran et al. 2007;Geach et al. 2009;Natarajan et al. 2009;Sánchez-Portal et al. 2015). In particular, it has been observed by our team in the framework of the GaLAxy Cluster Evolution Survey (GLACE; Sánchez-Portal et al. 2015) in the Hα and [Nii] emission lines to trace the SF and AGN activity in a wide range of environments (see Sect. 5 below). Although this cluster has been deeply studied in some aspects, a visual morphological classifications has been performed for a limited number (214) of member galaxies within a clustercentric distance extending to 5 Mpc (Moran et al. 2007). The purpose of this work is to improve the knowledge about the morphological properties of the member galaxies by providing a reliable classification of the sources up to 1 Mpc of clustercentric radius. We use publicly available Hubble Space Telescope (HST) Advance Camera for Survey (ACS) data in F775W filter.
Only 66 galaxies have been classified by Moran et al. (2007) within this radius ( 31 % of the total sample). Second important objective of this work is to compare visual classification of Moran et al. (2007) with our non-parametric method -this may provide us with handy-tool for future works at higher redshifts.
In our present work, we use a non-parametric method called galSVM introduced by Huertas-Company et al. (2008). galSVM fits a number of parameters simultaneously and assigns probabilities for each galaxy to be classified. Then based on the probabilities, the galaxies are classified into two broad morphological classes, namely early-type (ET) and late-type (LT). For more details about this classification, we refer the reader to Section 3.5.
The paper is organized as follows: Section 2 describes the data, along with a brief description of the generated source catalogue. In Section 3 the galSVM code and its application to our sample are described. The analysis on the results from our classification are further developed in Section 4. A detailed discussion is presented in Section 5. Finally Section 6 presents brief conclusions of this work.

HST/ACS data
We used the public HST reduced scientific image of ZwCl0024+1652 1 from the observation made on 16 November 2004 with the ACS Wide Field Camera (WFC) using the F775W filter. The ACS/WFC has a pixel scale of 0.05 arcsec/pixel and field of view of 202 × 202 arcsec 2 . The cluster is centred at RA = 6.64433 deg and DEC = 17.16211 deg, and the used image covers the central part of cluster of ∼ 1 Mpc. The image data is shown in Fig. 1 with all the sources labeled.

WFP2 supercatalogue data
To extract redshift information and to identify cluster members, we used the public ZwCl0024+1652 master catalogue 2 described in Treu et al. (2003) and Moran et al. (2005). The catalogue consists of 73,318 sources, with photometric and/or spectroscopically confirmed redshifts available, and covering the area of 0.5 × 0.5 deg 2 up to the clustercentric distance of about 5 Mpc. All observations were carried out with the Canada-France-Hawaii Telescope (CFHT) and its CFH12K wide field camera, and/or the HST Wide Field and Planetary Camera (WFP2), as described in Treu et al. (2003). Beside redshifts, this catalogue includes the visual morphological classification of sources brighter than I = 22.5 . We cross-matched this catalogue with our SExtractor catalogue (3515 sources) using a maximum radius of 2 arcsec. This radius was selected after testing different ones from 1 to 5 arcsec and finding it to be the best compromise between being the counterparts and having multiple matches. We obtained a total of 255 counterparts (hereafter cluster sample) with available redshifts. In total, 126 and 129 sources have spectroscopic and photometric redshift measurements. The redshift distribution of the members is given in Fig. 2, including spectroscopic and photometric measurements.

MORPHOLOGICAL CLASSIFICATION
In this section we describe the morphological classification of the ZwCl0024+1652 cluster galaxies in detail. We first go briefly through the methodology used, obtained results, and final classification. The centre of the cluster is indicated by a large green dot. The larger red and blue crosses indicate galaxies classified as ET and LT, respectively. While the smaller cyan crosses show those galaxies for which the probabilities are not measured and the magenta crosses show galaxies with measured probabilities but undecided morphologies (see Sec. 3).

Methodology
In this work we use galSVM (Huertas-Company et al. 2008) to classify galaxies morphologically in the ZwCl0024+1652 cluster. The galSVM is a public code that uses a free library libSVM (Chang & Lin 2011) and works in IDL environment. It has been successfully tested previously, at different redshifts, and on both field and cluster galaxies (e.g., Huertas-Company et al. 2009, 2010Pović et al. 2012Pović et al. , 2013Pović et al. , 2015Pintos-Castro et al. 2016).
For source detection, flux extraction, and measurement of the morphological parameters we need for the morphologi-cal classification (e.g., ellipticity), we run SExtractor (Bertin & Arnouts 1996). We extracted 3515 possible sources, including cluster members and field galaxies. galSVM uses local sample with known visual morphologies (see Sec. 3.2) and use it to learn how these may be seen at redshfits and magnitudes distributions of our real sample. It consists of several steps. First, it simulates local galaxies, by placing them to the redshift and magnitude distributions characteristic of the real sample. Secondly, it drops simulated local galaxies into the background that corresponds to the real sample image. Third, it measures different morphological parameters (see Sec. 3.3), first of the simulated local sample, and then of the real sample of galaxies that we want to classify. Finally, it compares morphological parameters of the training simulated local galaxies with their known visual classification, and determines conditions inside the multiple-parameters space that are then applied to the real sample to be classified. The final classification is based on a number of Montecarlo (MC) simulations, where each simulation gives a probability that the galaxy is early-type (ET). The average probability (P avg) and measured error give the final classification that the galaxy is ET. The probability that galaxy is late-type (LT) will then be 1 − P avg. For more details regarding galSVM see Huertas-Company et al. (2008). For applying galSVM, to include morphology determination for more fainter galaxies we took a magnitude limit of F775W ≤ 26.

Training sample of local galaxies
We used a catalogue of 3000 visually classified local galaxies, with known redshifts and magnitudes. The sample was selected randomly from the Nair & Abraham (2010)  range of 0.01 -0.1, and most of galaxies are bright with r band magnitude between 13 and 17. The magnitude and redshift distributions of the training sample versus that of the real data are shown in the two plots of Fig. 3. The detailed description of the training sample can be found in Pović et al. (2013). The training sample of 3000 local galaxies was selected as a good compromise between the computing time and accuracy in classification (both being highly sensitive to the training sample size). In addition, equal number of ET and LT galaxies were taken into account to obtain more precise morphology, and the selected sample can be considered as representative of the whole data with respect to general galaxy properties, as shown in Pović et al. (2013).

Measured morphological parameters
We used the following six parameters simultaneously to run galSVM: ellipticity (obtained by SExtractor), asymmetry, Abraham concentration index, GINI coefficient, M20 moment of light, and Conselice-Bershady concentration index. The last five were measured using galSVM and are briefly described as follows.
(ii) Abraham concentration index (CABR), is defined as the ratio of fluxes of the inner isophote at 30 % to that of the outer isophote at 90 % (Abraham et al. 1994(Abraham et al. , 1996. (iii) GINI coefficient (GINI), is a statistical term derived from the Lorentz curve specifying the overall distribution function of the pixel values of the galaxy (Abraham et al. 2003;Lotz et al. 2004).
(iv) M20 Moment of light, describes the second order normalized moment of the 20 % brightest pixels of the particular galaxy (Abraham et al. 2003;Lotz et al. 2004).
(v) Bershady-Conselice concentration index (CCON), measures a light ratio within a circular inner aperture (radii comprising of 20 % of the total flux) to the outer aperture (radii containing 80 % of the total flux) of the galaxy Bershady et al. 2000;Conselice et al. 2003).
In all measurements, the total flux is defined as the amount of flux contained within 1.5 times the Petrosian radius, where the Petrosian radius was measured with SExtractor. The centre of the galaxy is defined by minimizing ASYM index. More details on all parameters can be found in Huertas-Company et al. (2008) and Pović et al. (2013).

galSVM applied to ZwCl0024+1652
To measure the morphologies of ZwCl0024+1652 cluster members, we run galSVM on the HST/ACS F775W image described in Sec. 2.1, and used the SExtractor catalogue of 255 sources with all needed input parameters and redshifts available (see Sec. 2.1 and 2.2). We went through all Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University. galSVM steps described in Sec. 3.1, using the 3000 SDSS local galaxies as a training sample (see Sec. 3.2). We measured all parameters described in Sec. 3.3 of both training and real samples. For final classification we run 15 MC simulations, where in each simulation we used 2000 different randomly selected local galaxies (out of 3000) with the same number of ETs and LTs. The number of MC simulations was selected as the best compromise between the computational time and accuracy of results (see Pović et al. 2013).
Taking into account previous results obtained in Pović et al. (2013), dividing a sample into different magnitude ranges can increase the accuracy of morphological classification by optimizing the galSVM code for fainter galaxies. Therefore in this work we run galSVM three times, using the following ranges.
For each range we provided the corresponding magnitude and redshift distributions of cluster members for simulating during the classification process. These distributions are shown in Fig. 4 for both training sample after being simulated and the real sample to be classified. For the final classification we follow the findings of Pović et al. (2013), and considered the results from the first magnitude bin, from the second, but only for those sources not present in the first one (for 22.0 < MAG AUTO ≤ 24.0), and finally from the third bin, but only for those sources not present in the previous two (for 24.0 < MAG AUTO ≤ 26.0).

Final classification
In all galSVM runs (i.e. for each magnitude bin) we obtain a final average probability from 15 MC simulations (for more details about the training sample and the running setup see Sections 3.2 and 3.4). Finally, we obtained PROBA AVG 4 with corresponding uncertainty values for 231 galaxies out of 255. For the remaining 24 galaxies, PROBA AVG was not measured either because one or more parameters have values 4 PROBA AVG is average probability measured by galSVM  , or they simply were not measured by galSVM. Of these, 50 % are located on image borders, while most of the remaining sources are merging/interacting systems or some are edge on galaxies. Only 3 galaxies (out of 24) have close companions, but the sample is not statistically significant for doing any additional studies. Tables 1 and 2 give the median values and Q1 to Q3 ranges 5 of average probability [Q1 -Q3] 6 and its error in three magnitude bins. It can be seen that most of the brightest galaxies (F775W ≤ 22.0) are characterized by larger probability (median value > 0.7) to be ETs, while for fainter galaxies (2nd and 3rd bin) probability to be ET is lower than 0.5. As we could expect, the error values increase for fainter bins (see Table 2). For the final classification we took into account the measured errors and considered a galaxy to be ET if PROBA FINAL = PROBA AVG ± PROBA ERR > 0.6 (or 0.7 in the last magnitude bin), and to be LT if PROBA FINAL < 0.4 (or 0.35 in the last magnitude bin), where PROBA ERR is uncertainty in measuring probability and PROBA FINAL is the final probability after error correction. For those galaxies with 0.4< PROBA FINAL < 0.6 (or between 0.35 and 0.7 in the last magnitude bin) we are not able to classify them morphologically, and they will remain inside the 'undecided class' (UD). To define the classification boundaries, we used previous works of Pović et al. (2012Pović et al. ( , 2013 and Pintos-Castro et al. (2016). Fig. 5 shows the PROBA FINAL distributions in the three magnitude ranges, while the final classification is summarized in Table 3. As can be seen, out of a total of 231 galaxies with measured final probabilities, we have 97 (42%), 83 (36%), and 51 (22%) galaxies classified as ET, LT, and UD, respectively. Figure 6 shows the PROBA FINAL of the whole clas-sified sample. We marked in Fig. 1 all the classified sources with red and blue crosses respectively for ET and LT galaxies respectively. Of the classified sources ET (LT) galaxies 59 (41) have spectroscopically confirmed and 38 (42) have photometric redshifts. Few bright galaxies (e.g. close to the cluster centre) remained unclassified, mainly due to the difficulty that galSVM had with classifying galaxies being in rich environments and with close companions. Moreover, it has also been deduced that only 21.6 % of the UD galaxies have spectroscopically confirmed redshifts while for the classified ones (ET or LT), 62.2 % have spectroscopic redshifts.

Comparisons with visual morphological classification
The visual morphological classification of 214 galaxies with spectroscopically confirmed redshifts in ZwCl0024+1652 was carried out previously by Moran et al. (2007), covering the clustercentric distance of 5 Mpc. In this section we compare our non-parametric classification of 231 galaxies, within  2), with the visual one. Within the region of our data (∼1 Mpc radius) we determined that there are 123 sources with spectroscopically confirmed redshifts having visual morphology as in Moran et al. (2007). While in our sample catalogue with measured probabilities (231 sources) we have 111 sources with spectroscopic redshifts and 120 sources with photometric redshifts. We cross-matched the two catalogues using the radius of 2 arcsec, and found a total 66 counterparts. The reason of a small number of counterparts is mainly because of the fact that Moran et al. (2007) visual classification was done only for galaxies with confirmed spectroscopic redshifts and only considers the best resolved galaxies. The I-band magnitude limit of galaxies in Moran et al. (2007) is 22.3, with 201 (95 %) of galaxies being brighter than I = 22, whereas our magnitude limit in F775W-band is 26. The I band magnitude distribution comparison of both works is given in Fig. 7. Out of 66 counterparts, 50 and 16 galaxies were classified visually by Moran et al. (2007) as ET and LT respectively. When compared with our results, 53 galaxies, or 81 %, match the visual classification, of these 41 being classified as ET, and 12 as LT. Of the remaining 13 galaxies, 7 have visual classification available, but were classified as UD in our work, while for the other 6 galaxies ET/LT classification is in disagreement between the two works. Visually checking these galaxies, we found that 3 of them are edge on (S0 in Moran et al. (2007) while LT in our work possibly Sa that both could be possible). The other three galaxies were classified as ET in in our work whereas 2 of them are Sa + b and one is Sc + d in Moran et al. (2007); our classification being right for one while one is observed to be an interacting system and the remaining one is peculiar galaxy. Finally, after these comparisons we conclude that 81 % of our classification is in a good agreement with the visual classification. Moreover, in this work we provide a reliable classification of additional 121 galaxies within 1 Mpc of clustercentric distance, being classified for the first time.

Morphological parameters
The distributions of different measured morphological parameters of 180 ET and LT classified cluster members are given in Fig. 8. In addition to the histograms, Table 4 summarizes the median values of each parameter and [Q1-Q3] range characteristic of cluster members classified as ET or LT. As can be seen from both Fig. 8 and Table 4, all parameters follow the expected trends of ET and LT galaxies, with concentration indices such as CABR, CCON and GINI being characterised with higher values in case of ETs, while ASYM, M20, and ELLIP are showing higher values for LTs. If we compare our results with those obtained by Pović et al. (2013), using the same methodology and data of the ALHAMBRA survey (Moles et al. 2008) in F613W band, ZwCl0024+1652 galaxies classified as ET seem to be slightly more concentrated (in terms of all concentration indices), and characterised with lower asymmetries in the case of both ET and LT.

Morphological diagnostic diagrams
In this section, we tested some of the commonly used morphological diagnostic diagrams by comparing the measured morphological parameters. Fig. 9 shows six different diagrams and relations between CABR and ASYM, GINI, and CCON (left plots, from top to bottom, respectively), and M20 and CCON, GINI, and CABR (right plots, from top to bottom, respectively). These relations have been used in many previous works, showing a separation between ET and LT galaxies (e.g., Abraham et al. 1994Abraham et al. , 1996  It can be seen in all plots that ZwCl0024+1652 cluster members classified as ET and LT are occupying different areas on diagrams, as expected. ETs are located again in the regions characterised with higher concentrations (larger values of CABR, GINI, and CCON and lower of M20), in comparison to LTs. ASYM parameter is much more delicate in separating sources, as has been commented previously (Pović et al. 2015), and as can be seen in Fig. 9 (top left plot). However, it can be efficient in selecting interacting systems, showing larger values as can be seen in the same plot for sources with ASYM > 0.5. The relationships depicted are all in agreement with recent works (e.g. Castellón et al. 2014;Parekh et al. 2015;Pintos-Castro et al. 2016). We took into account studies of Tarsitano et al. (2018) and their Figure  11. We reproduced Gini vs. M20 diagram color coded with CABR, CCON, ASYM, and ELLIP finding the results in line with our Fig. 9.

Colour-colour and colour-magnitude relations
In this section we tested the colour-colour and colourmagnitude diagrams for ZwCl0024+1652 cluster members classified as ETs and LTs. These diagrams have been tested at both lower and higher redshifts, and it is very well known that the distribution of galaxies on them is bimodal, with ETs being mainly located in the red sequence and LTs in the blue cloud (e.g., Bell et al. 2003;Cassata et al. 2007;Melbourne et al. 2007;Pović et al. 2013;Schawinski et al. 2014, etc.). In Fig. 10 we represented the relation between the R -K versus B -R rest-frame colours (left plot), and between the B -R rest-frame colour and absolute magnitude in the B band (right plot). We also represent the histograms of all parameters used in the 2d plots, and their distributions for both ET and LT galaxies. In the two plots we can see the area with a higher density of ET sources, and that in general brighter and redder regions have higher fractions of ETs, as

Morphology vs. clustercentric distance
The distance between member galaxy and the centre of the cluster is calculated using the spherical law of cosines as: cos(D s ) = sin(δ c )×sin(δ g )+cos(δ c )×cos(δ g )×cos(|α c −α g |), (1) where (α c , δ c ) are right ascension and declination of the cluster centre in radians, while (α g , δ g ) are galaxy coordinates.
To measure the clustercentric distance in Mpc, we used the following: where D cl = 1500M pc and is the distance to ZwCl0024+1652. Fig. 11 shows the distribution of clustercentric distance of 180 cluster members classified as ETs and LTs, while Table 5 provides the basic statistics (median and [Q1 -Q3] ranges) of both morphological types.
We also analysed the relation between the galaxy brightness, in terms of F775W magnitude and surface Figure 9. Standard morphological diagnostic diagrams showing the relation between CABR and ASYM, GINI, and CCON (left plots, from top to bottom, respectively), and M20 and CCON, GINI, and CABR (right plots, from top to bottom, respectively). In all the plots red solid and blue open triangles stand for ET and LT galaxies, respectively. brightness (MUMEAN), and clustercentric distance, as shown in Fig. 12. For the two morphological types, Table  5 gives again the main statistics regarding the brightness.
Finally, we analysed the relation between the clustercentric distance (R) and morphological parameters measured in previous section. Fig. 14 shows for the first time for ZwCl0024+1652 how the six morphological parameters vary with respect to the clustercentric distance in the case of cluster members classified as ET and LT. We also selected those sources classified as LTs in this work, and that taking into account previous studies of Parekh et al. (2015) and visual inspection seem to be mergers. We discussed all plots and statistics in Sec. 5.

Morphological classes
It was pointed out that the evolution of ET proportion is affected by redshift in addition to density and clustercentric distance (Smith et al. 2005;Postman et al. 2005;Simard et al. 2009). The BO effect (Butcher & Oemler 1984) verified with works for different redshift ranges, have been indifferently shown that LT proportion increases with redshift (e.g. Fairley et al. 2002 for z ∼ 0.2 -0.5, de Lucia et al. 2007 for z ∼ 0.4 -0.8, Barrena et al. 2012 for z ∼ 0.2 -0.5 and Castellón et al. 2014 for z ∼ 0.17 -0.6). These studies show that the proportion of LTs at z ∼ 0.4 accounts about ∼35 % -40 %. In our current work the fraction of LT galaxies is ∼36 %, which is in agreement with previous results. Moreover, according to Parekh et al. (2015) it was determined that galaxies are classified into most relaxed, relaxed and non-relaxed ones based on values of GINI coefficient where the non relaxed (peculiar/most disturbed) galaxies being characterized by GINI < 0.4 criterion. It is also described for the most disturbed galaxies that GINI value is small because bright pixels are not compact while equally distributed in the given aperture radius. Accordingly eight non-relaxed galaxies (peculiar) were identified from LT class closer to the cluster core leaving the spiral population near the core to be very small.
For the overall galaxy population, Moran et al. (2007) determined that for 123 matching galaxies within the 1Mpc region, 65.6 % were ET while 34.1 % being LT. In the same work the morphology was determined for MS0451-0305 galaxy cluster at z ∼ 0.5 with 52 % and 48 % being ET and LT, respectively. The galaxy population in our work follows nearly the same trend as what has been on board for cluster studies confirming ET population is greater than the LT population (see Postman et al. 2005& Moran et al. 2007). As shown in Sec 3.5, out of the 231 galaxies we have 42 % and 36 % galaxies classified as ET and LT, respectively.

Morphology versus ELGs
Using GLACE survey data, Sánchez-Portal et al. (2015) has presented a catalogue of 174 unique emission line galaxies (ELGs) in our cluster within 4 Mpc clustercentric distance. Accordingly ∼ 37 % of the ELGs (64 galaxies) were shown to be AGN (broad line AGN (BLAGN) and narrow line AGN (NLAGN)) whereas ∼ 63 % being star forming (SF) galaxies (110 in number). Out of the 174 ELGs, 79 galaxies (52 SFs (∼ 66 %) and 27 AGNs (∼ 34%)) were within the clustercentric distance of 1 Mpc (region of our concern). Matching the GLACE result with ours we found 43 (∼ 54.4 %) counterparts. Here 26 ELGs had no match in our catalogue may be because Sánchez-Portal et al. (2015) was working only on ELGs but this is not the case of our work. Out of the matching 43 sources, 26 (∼ 60.5 %) are SF while the remaining 17 (∼ 39.5 %) are AGN. Morphologically comparing the matching ELGs; 11 galaxies (∼ 26 %) correspond to ET, 28 galaxies (∼ 65 %) belong to LT and the remaining 4 galaxies (∼ 9 %) correspond to UD class in our results. More specifically 18 SF galaxies are in LT class where 5 SF galaxies fall in ET while the remaining 3 SF galaxies belong to UD class. Similarly for AGN; LT contributes 10, ET contributes 6 while UD contributes only 1 AGN. This confirms that ELGs (SF as well as AGN) are mostly galaxies rather than ET galaxies.

Morphological fractions
In Fig. 13 we compared the morphological fraction with both F775W magnitude and clustercentric distance. To compare it with magnitude (left plot of Fig. 13), morphological fraction is computed for each 1 magnitude bin in such a way that a particular class fraction is the ratio of the number of a given class galaxies to the total number of galaxies of all classes within the same bin. This for instance, is given for ET fraction in a given bin as: where number of ETs ≡ ETs with magnitudes within the range of the bin; total number of galaxies ≡ number of all galaxies (ET+LT+UD) with magnitudes in the same magnitude range.
Once it is computed for all the bins it is plotted against the centre of the bin. It can be seen that the fraction of ET galaxies decreases as a function of increasing magnitude while that of LT galaxies increases up to F775W ∼ 22.5, remaining nearly constant for fainter magnitudes. The median F775W value for ET is determined to be 21.28 and that for LT is 22.16 while that for UD galaxies is 23.20. Hence we can see that the brightest galaxies are most likely to be resolved and classified into ET/LT whereas fainter galaxies could not easily be resolved, significant number of these galaxies are unlikely to be classified then left as UD. For magnitudes where F775W > 24.5, the number of sources are very small that the statistics is very poor to draw a conclusion.
According to Fasano et al. (2012), the fraction of ET galaxies is high near the centre of a nearby cluster while decreasing as a function of clustercentric distance. The fraction of LT galaxies on the other hand being smaller closer to the core while increasing as a function of clustercentric distance (see also Zwicky 1942;Dressler 1980;Whitmore et al. 1993;Pintos-Castro et al. 2016). Here to compare with clustercentric distance (right panel of Fig. 13) we compute the morphological fraction in each 0.2 Mpc bin for 0 to 0.6 Mpc, 0.1 Mpc bin for 0.6 to 0.7 Mpc and 0.3 Mpc bin for 0.7 Mpc to 1 Mpc in the same way as in eq. 3. Once it is computed for all the bins it is plotted against the centre of the bin. Results of our current work (see Fig. 11 and the right plot of Fig. 13) confirm that closer to the core, the ET population fraction is higher than the LT fraction, but ET fraction is observed decreasing and LT fraction increasing until the clustercentric distance of ∼ 0.3 Mpc. Whereas beyond ∼ 0.3 Mpc fractions of both populations continue nearly flat in parallel up to a clustercentric distance; R ∼ 1 Mpc. For clustercentric distances where R > 0.7 Mpc, the number of sources are very small that the statistics is very poor to conclude. We can see in general on Fig. 13 throughout the entire region that the fraction of ET galaxies is consistently higher than the LT fraction and more fraction of galaxies is classified into . ET/LT near the core (lower UD fraction) than in far distances (where higher UD fraction) from the centre. Hence, our results are in a good agreement with previous results.
Moreover, out of the total number of 231 galaxies with in the cluster; 111 have spectroscopically confirmed redshifts while 120 have photometric redshifts. Different trends are observed in morphological fractions throughout the clustercentric distance. It can easily be seen that for galaxy population with spectroscopic redshifts, ET fraction is greater than the LT fraction throughout the region. While for galaxies with photometric redshifts, the LT fraction dominates throughout over the ET faraction.

Morphology-density relation
An important point to be raised is morphology -density relation. As shown by Hoyle et al. (2012), there is a trend of an increase in the population of ET galaxies towards the cluster centre accompanied by a strong morphology -density relation. Previous studies have already described a high -intermediate -low density regions in a cluster (see Jee et al. 2005;Demarco et al. 2010). Analysing a cluster at z = 0.84, Nantais et al. (2013) determined that the cluster outskirts (intermediate to low density region) is characterised by higher LT while lower spiral with more peculiar (merging) galaxy population. Whereas a high density region (cluster core) with dominating ET population, few peculiar galaxies and almost devoid of spirals. It has been determined near the cores of clusters that the proportion of ET galaxies is ∼ 47 %, that is ∼ 2.8 times greater than the ET fraction in the field at the same intermediate redshift (see Delgado-Serrano et al. 2010& Nantais et al. 2013. This is a relationship that holds also for low redshift rich clusters as determined for galaxies with redshift of z ∼ 0.1 -0.2 by Fasano et al. (2000). In our case, we were working on the broad classes (ET and LT). While there is no clear classification into peculiar (merging) galaxies and these will be included in our classification either into LT or UD. In our work, the proportion of ET is recorded decreasing as going outwards (decreasing den-sity) from the cluster core at least up to ∼ 0.7 Mpc (see the right plot in Fig. 13). As mentioned previously, after the R = 0.7 Mpc the number of sources decrease significantly in all three morphological groups, which affects the measured fractions. Moreover, LT population decreases approaching to the cluster core in agreement with existing results.

Relevance of morphological parameters
In Parekh et al. (2015) while working on galaxy classification into relaxed versus dynamically disturbed system using the data of clusters at different redshifts from Chandra archive, they indicated GINI, M20 and Concentration as very promising parameters for identifying mergers. Accordingly, the criteria set for the most relaxed system is that GINI > 0.65, M20 < -2.0 and Concentration > 1.55. For the most dynamically disturbed (non relaxed) system GINI < 0.4, M20 > -1.4 and Concentration < 1 were set. Intermediate between the two extreme conditions is the mildly disturbed situation. They identified that GINI is the most useful parameter in determining substructure because it does not depend on the exact position of the centre. Our classification was done with six morphological parameters (subsection 3.3) to classify the galaxies into ET and LT. Adapting the criteria from Parekh et al. (2015) for our work, GINI < 0.4 gave us 12 galaxies classified into ET/LT (ET=1 (8.3 %), LT=11 (91.7 %)). These 11 LT galaxies are checked visually to be the most perturbed (non relaxed) galaxies, where M20 > -1.4 for 9 galaxies out of the 11 LTs (with ∼ 82 % agreement). On the other hand corresponding to GINI > 0.65; 58 galaxies were classified into ET/LT (ET=49 (85 %), LT=9 (15 %)) and are the most relaxed ones accordingly. Hence our result in this aspect is subject to 85 % agreement with previous works (amount of ETs). But in our work, M20 < -2.0 is too small to be used while it is better to take M20 cut off for the most relaxed galaxies to be -1.8 to establish an accuracy of at least 79 %. Moreover, since concentration parameters are defined as in subsection 3.3, from our results the cut off values of CABR < 0.2 and CCON < 7.0 can be used for the most perturbed galaxies with ∼ 91 % agreement while CABR > 0.45 and CCON > 7.5 can segregate about ∼ 94 % of the most relaxed galaxy population. Therefore with this cut off limits, CCON and CABR parameters could also be equally important parameters as GINI and M20 for morphological classification of galaxies.

Morphological parameters vs. clustercentric distance
In this work for the first time, we studied the properties of different morphological parameters in relation to the clustercentric distance (R). In Fig. 14 we showed how GINI, ELLIP, M20, CABR, ASYM and CCON change with R for ET and LT galaxies. In general, we do not find any clear trend in case of ASYM, CCON and ELLIP with R. In case of GINI, CABR and M20 a slight trend is observed of decreasing GINI and CABR showing and increasing M20 moment of light, suggesting that as going outwards from the cluster centre the light concentration decreases. However, much better statistics are needed to confirm this result. As shown in the right plot of Fig. 13 for ET class galaxies, the median and [Q1 -Q3] range of R being lower on average than LT, and this can be explained against each parameter. This is accompanied by higher values of GINI, CCON and CABR while lower values of ELLIP, M20 and values about zero for ASYM for ET galaxies. Whereas for LT galaxies, the median and [Q1 -Q3] range of R is slightly higher on average than ET galaxies. This can be seen from the plot describing morphological fraction in Fig. 13. It can also be seen that GINI value slightly decreases as a function of increasing R for LT galaxies. Similar trend is observed from the plot of CABR versus R but very slow decrease for both classes in this case. For other parameters the values almost remain stagnant with R.

CONCLUSIONS
In this work as part of a complete morphological study of the cluster ZwCl0024+1652 at an intermediate redshift z ∼ 0.4, we presented a broad classification of member galaxies with available redshifts within the clustercentric distance of 1 Mpc using the HST/ACS image. We have classified galaxies up to the I -band magnitude of 26. By running galSVM code on a sample of 255 galaxies, 6 morphological parameters were measured and classification was provided for 231 galaxies. Of these, 111 have spectroscopic and 120 photometric redshift measurements. From our classification and analysis we have drawn the following conclusions: • Out of all the 231 galaxies 97 (∼ 42 %) were classified as ET, 83 (∼ 36 %) as LT and 51 (∼ 22 %) stayed unclassified. If we take the well classified galaxies (180 in number); 97 (∼ 54 %) were classified as ET whereas 83 (∼ 46 %) fall into an LT class.
• Comparing with the visual classification results in Moran et al. (2007) we have classified 53 galaxies matching with their previous visual morphologies, 6 galaxies classified to different classes and 121 new sources which didn't have any reported morphological classification within ∼ 1 Mpc) radius are newly classified in our work. Therefore this work work gives the most complete and largest morphological catalogue available up to now for our galaxy cluster.
• Moreover, our comparison with the existing visual classification of Moran et al. (2007) is in a good agreement of 81 %. Hence, applying galSVM for morphological classification can be taken as a reliable technique to be used even for a large sample.
• We have tested that ET and LT galaxies follow the expected distributions for different standard morphological diagrams, colour -colour and colour -magnitude diagrams.
• The ET morphological fraction is higher near the cluster core decreasing outwards with LT fraction being lower at core increasing outwards. Throughout the region of 1 Mpc radius, the fraction of ET galaxies is consistently greater than the LT fraction for our cluster in the region of our concern (R out to 1 Mpc). Hence, the ET/LT fraction in the cluster is in agreement with previous studies.
• Morphological fractions in our galaxy cluster at z ∼ 0.4 evolves with magnitude in such a way that ET fraction dominates in the brightest magnitude limit decreasing towards the fainter end while the LT fraction increases as magnitude goes fainter.
• We compared our results with Sánchez-Portal et al. (2015) and found 43 ELG counterparts. As a result out of these counterparts, 11 galaxies (∼ 26 %) correspond to ET while 28 galaxies (∼ 65 %) were found to belong to LT. with the remaining 4 galaxies (∼ 9%) stayed unclassified in our work. Moreover with the star forming ELGs; 18 SF galaxies are LT, 5 SF galaxies fall in ET and the remaining 3 SF galaxies belong to UD class. Similarly for AGN; LT contributes 10, ET contributes 6 while UD contributes only 1 AGN. Hence, in general we deduce that ELGs are more of LT in morphology than ET.
• We have analysed the morphological parameters as a function of clustercentric distance out to 1 Mpc for the first time. In general we do not find any clear trend, however better statistics would be valuable in future studies to revise the change of galaxy light concentration with R.
This work contributes significantly in the area of studies related to evolution of galaxies in clusters involving morphological classification, and provides the most complete morphological catalogue of ZwCl0024+1652. In our future studies within the GLACE survey, we are planning to compare morphological properties with metallicities, star formation, and AGN contribution using the tunable filters data. Finally, a complete morphological catalogue that resulted from our work can be accessed with an electronic version of this paper. The first seven rows and the descriptions for all the columns are presented as an appendix in this paper. Figure 14. From top to bottom, and from left to right: Relation between the GINI, ellipticity, M20 moment of light, CABR concentration index, asymmetry, and CCON concentration index and distance from the cluster centre. In all plots red solid triangles stand for ET, and blue open triangles for LT galaxies. The disturbed (merging) galaxies selected from LT based on GINI < 0.4 criteria are indicated with green dots (solid circles). Median values of each parameter with clustercentric distance are shown with the red solid and blue dashed lines of ET and LT galaxies, respectively. phase of the paper. Moreover ZBA acknowledges Kotebe Metropolitan University for granting a study leave and giving material supports. MP and SBT acknowledge financial support from the Ethiopian Space Science and Technology Institute (ESSTI) under the Ethiopian Ministry of Innovation and Technology (MoIT). MP also acknowledges support from the Spanish MINECO under projects AYA2013-42227-P and AYA2016-76682-C3-1-P, and from the State Agency for Research of the Spanish MCIU through the "Center of Excellence Severo Ochoa" award for the Instituto de Astrofísica de Andalucía (SEV-2017-0709). In this work, we made use of Virtual Observatory Tool for OPerations on Catalogues And Tables (TOPCAT) and IRAF. IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. We also used ACS/HST data based on observations made with the NASA/ESA HST, and obtained from the Hubble Legacy Archive, which is a collaboration between the Space Telescope Science Institute (STScI/NASA), the Space Telescope European Coordinating Facility (ST-ECF/ESA) and the Canadian Astronomy Data centre (CADC/NRC/CSA).This work was supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under the grants AYA2014 -58861 -C3 -2 -P, AYA2014 -58861 -C3 -3 -P, AYA2017 -88007 -C3 -1 -P and AYA2017 -88007 -C3 -2 -P.