Shell-type Tidal Features Are More Frequently Detected in Slowly Rotating Early-type Galaxies than Stream- and Tail-type Features

To enhance our understanding of the impact of galaxy mergers on the kinematics of early-type galaxies (ETGs), we examine differences in specific stellar angular momentum within the half-light radius ( λRe ) among ETGs with different types of tidal features and those without such features. This is accomplished by categorizing tidal features, which serve as direct evidence of recent mergers, into shells, streams, and tails, through deep images from the DESI Legacy Survey, and by using MaNGA data for the analysis of the kinematics of 1244 ETGs at z < 0.055. We find that ETGs with tidal features typically have reduced λRe values that are lower by 0.12 dex than ETGs without tidal features. ETGs with shells contribute most to the reduction in λRe . Consequently, nearly half of ETGs with shells are classified as slow rotators, a fraction that is more than twice as high as that of ETGs with tails or streams, and over three times higher than that of ETGs without tidal features. These trends generally remain valid even when ETGs are divided into several mass bins. Our findings support the idea that radial mergers, which are more effective at reducing λRe than circular mergers, are more closely associated with the formation of shells rather than streams or tails. The detection of shells in slightly more massive ETGs compared to streams and tails may be attributed to the fact that massive satellite galaxies are more likely to be accreted through radial orbits, due to the nature of dynamical friction.

1. INTRODUCTION Early-type galaxies (ETGs) are in the mature stage of the cosmic evolution of galaxies, predominantly composed of old stellar populations.The low star formation rate of ETGs is a consequence of the depletion of cold gas in these systems.As a result, ETGs generally exhibit red colors (g − r ≳ 0.7) in the optical bands (Gallazzi et al. 2006;Graves et al. 2009;Schawinski et al. 2014;Lacerna et al. 2020).ETGs tend to possess smoother and simpler structures compared to late-type galaxies, which often exhibit spiral arms with intricate features and blobs of highly active star-forming regions (Nair & Abraham 2010).The stars in ETGs are distributed with a higher light concentration toward the center compared to those in late-type galaxies (Park & Choi 2005;Choi et al. 2010).Accordingly, the modeling of ETG surface brightness profiles typically employs centrally concentrated light profiles, such as the de Vaucouleurs profile yyoon@kasi.re.kr or the Sérsic profiles with high Sérsic indices of n ≳ 3 (Blanton & Moustakas 2009;Huertas-Company et al. 2013).
Beyond uncovering these photometrically derived ETG properties, a comprehensive understanding of the kinematic properties of ETGs has been attained through large surveys based on integral field unit (IFU) spectroscopy (Bacon et al. 2001;Bershady et al. 2010;Cappellari et al. 2011;Sánchez et al. 2012;Ma et al. 2014;Yoon et al. 2021;Sánchez et al. 2016).IFU surveys have provided new insights into ETGs.For instance, studies utilizing IFU survey data have revealed that a large portion of ETGs, including those with round and nondisky shapes, possess stellar components that exhibit significant rotation similar to disks in late-type galaxies (Cappellari 2016;Graham et al. 2018).Thus, a multitude of studies utilized a new classification system that categorizes ETGs into either slow or fast rotators based on resolved kinematics (Emsellem et al. 2007(Emsellem et al. , 2011;;Jesseit et al. 2009;Khochfar et al. 2011;Cappellari 2016;Graham et al. 2018).Indeed, this is increasingly being regarded as a more physically meaningful classification, replacing the conventional categorization into ellipticals and lenticulars, which strongly depends on line-of-sight inclinations (Emsellem et al. 2007;Cappellari et al. 2011;Cappellari 2016).
In the standard Λ cold dark matter universe, galaxy mergers play a crucial role in the formation and evolution of ETGs (Baugh et al. 1996;Christlein & Zabludoff 2004;De Lucia et al. 2006;De Lucia & Blaizot 2007;Wilman et al. 2013;Yoon et al. 2017), shaping the fundamental properties of ETGs.This is particularly significant for massive ETGs with M star > 10 10.5 M ⊙ , for which the contribution of ex-situ sources to mass assembly is considerable (Dubois et al. 2016;Davison et al. 2020).For example, galaxy mergers have the potential to generate red and quiescent remnants in the end that do not actively produce young stellar populations (Springel et al. 2005;Hopkins et al. 2008;Brennan et al. 2015), owing to the rapid depletion of available cold gas through intense star formation during the merger processes (Hernquist 1989;Mihos & Hernquist 1996;Springel et al. 2005).Strong feedback effects from active galactic nuclei (AGNs) triggered by the merger processes may also have the capability to quench the merger remnants (Hopkins et al. 2005(Hopkins et al. , 2008;;Springel et al. 2005).In addition, the concentrated steep light profiles, commonly observed in typical ETGs, can stem from mergers (Barnes 1988;Naab & Trujillo 2006;Hilz et al. 2013).
Likewise, galaxy mergers are expected to impact the stellar kinematics of ETGs, given that more massive ETGs, which more predominantly form and grow through mergers (De Lucia et al. 2006;De Lucia & Blaizot 2007;Yoon et al. 2017), typically have lower specific stellar angular momentum (Emsellem et al. 2007;Graham et al. 2018).Numerical simulations have been employed in previous studies to elucidate the influence of galaxy mergers on the stellar kinematics of merger remnants.For instance, the mass ratios and merger orbits of merger progenitors play pivotal roles in determining the kinematics of merger remnants, according to Jesseit et al. (2009), Bois et al. (2011) and Martin et al. (2018).The simulation of Choi & Yi (2017) demonstrated that galaxy mergers tend to statistically reduce rotation speeds, especially in massive galaxies.They also observed that frequent minor mergers exert significant cumulative effects on the kinematics of merger remnants.Similarly, Lagos et al. (2018a) and Schulze et al. (2018) also discovered a reduction in stellar rotations as a result of galaxy mergers in their simulations.In addition, the simulation of Choi et al. (2018) suggested that galaxy mergers are the dominant driver of the spin reduction for central ETGs in dense regions.
In contrast to simulation studies, conducting direct observational studies on the effect of galaxy mergers on the properties of ETGs is relatively challenging.This is due to the fundamental limitation that we can only observe a snapshot of the universe, rather than the contin-uous flow of cosmic time.Nevertheless, we can mitigate this limitation by using tidal features.Tidal features are stellar debris generated by galaxy mergers (Toomre & Toomre 1972;Quinn 1984;Barnes 1988;Hernquist & Spergel 1992;Feldmann et al. 2008), which are generally fainter than the main bodies of galaxies, necessitating deep images for their detection and examination.Hence, tidal features such as tidal tails, streams, and shells serve as the most direct observational indicators of recent mergers, offering a means to examine the impact of mergers on the photometric (Schweizer & Seitzer 1992;Tal et al. 2009;Schawinski et al. 2010;Kaviraj et al. 2011;Hong et al. 2015;Yoon & Lim 2020) and kinematic (Krajnović et al. 2011;Duc et al. 2015;Oh et al. 2016;Yoon et al. 2022;Bílek et al. 2023) properties of ETGs.
For example, it has been observed that ETGs displaying blue optical colors are more prone to possess tidal features or morphological disturbances (Schweizer & Seitzer 1992;Tal et al. 2009;Schawinski et al. 2010;Kaviraj et al. 2011), suggesting a potential association between young stellar populations in ETGs and recent merger events.In a study by Hong et al. (2015), it was found that nearly half of luminous AGN hosts, primarily ETGs, exhibit tidal features, which is in contrast to the lower fraction observed in typical ETGs.This indicates that luminous AGNs in ETGs are likely to be activated by recent merger events.Combining tidal features identified in deep images with IFU spectroscopic data, Yoon et al. (2023) uncovered in detail that ETGs that have undergone recent mergers have different stellar population profiles compared to their counterparts that have not experienced recent mergers.Recently, Yoon et al. (2022) studied the impact of galaxy mergers on stellar kinematics of ETGs, using 167 ETGs in the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA; Bundy et al. 2015;Drory et al. 2015;Yan et al. 2016;Wake et al. 2017) IFU data that are in the Stripe 82 region of the Sloan Digital Sky Survey (SDSS).The main discovery of Yoon et al. (2022) is that galaxy mergers normally reduce the stellar angular momentum of ETGs.
Different types of tidal features (e.g., shells, tails, and streams) store information about recent mergers with different properties.For example, shells can be generated by mergers with radial orbits, while streams can be produced by mergers with circular orbits (Quinn 1984;Dupraz & Combes 1986;Johnston et al. 2008;Hendel & Johnston 2015;Pop et al. 2018;Karademir et al. 2019).Arm-and loop-shaped tidal tails may be formed by the dynamically cold material in the disks of merging galaxies (Schombert et al. 1990;Feldmann et al. 2008;Duc et al. 2015).Therefore, splitting tidal features into different types and exploring how these types are associated with the properties of galaxies can provide further insight into processes of galaxy mergers and evolution.
In this study, we extend the study of Yoon et al. (2022) by increasing the number of ETGs to over a thousand using the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Survey (Dey et al. 2019), which has a comparable surface brightness limit with that of the Stripe 82 coadded images but covers a far larger survey area.Using this substantially larger sample, we focus on examining how the stellar angular momentum of ETGs is different from each other, depending on the presence of the different types of tidal features.By doing so, we broaden our understanding of the impact of galaxy mergers on the stellar kinematics of ETGs.

SAMPLE AND ANALYSIS 2.1. SDSS-IV MaNGA
The MaNGA IFU spectroscopic survey is the fourth generation of the SDSS project (Blanton et al. 2017).This project collected observational data using the Astrophysical Research Consortium (ARC) 2.5m telescope.The MaNGA project employed 17 hexagonal shape fiber-bundled IFUs with sizes ranging from 12 ′′ -32 ′′ , depending on the number of fibers.These 17 IFUs are distributed across the 3 • field of view of the telescope focal plane.The spectrograph utilized in the MaNGA survey is identical to the one used in the Baryon Oscillation Spectroscopic Survey (Smee et al. 2013).This spectrograph covers a wavelength range of 3600-10300 Å , offering a midrange spectral resolution of R ∼ 2000.The target selection criteria of the MaNGA survey is based on i-band absolute magnitude and redshift (and nearultraviolet−i color for a small portion of targets).This selection process results in ∼ 10, 000 galaxies, evenly distributed across the color-magnitude space, with a uniform spectroscopic coverage up to 1.5 or 2.5 half-light radius along the major axis (R e ).Further details regarding the selection of target galaxies can be found in Wake et al. (2017).
Here, we utilize the MaNGA data of the final release version (Data Release 17).Approximately 1% of the MaNGA data were obtained from repeated observations of the same galaxy.In these cases, we select one observation from the duplicates, prioritizing the data with the larger IFU size.If the IFU sizes of the duplicate observations are identical, we choose the data with the highest blue channel signal-to-noise ratio (S/N).

Deconvolution of IFU Data
Observations from ground-based telescopes are inevitably influenced by instrument-induced aberrations and atmospheric conditions, causing the seeing effect.The effect of seeing is particularly pronounced in data obtained through fiber-based IFUs, owing to the large physical gaps between the sampling elements (IFU fibers).Reducing this effect enables us to derive more reliable spatially resolved kinematics information, especially in the central regions of galaxies.In order to alleviate the impact of seeing, we apply the Lucy-Richardson (LR) deconvolution algorithm (Richardson 1972;Lucy 1974) to the MaNGA IFU data as in Chung et al. (2021) (the same algorithm was also used in Yoon et al. 2021 andYoon et al. 2022).The LR deconvolution algorithm is an iterative process designed to restore an original image that has undergone convolution by a point spread function (PSF).This algorithm is characterized by its minimal parameter requirements, making it well suited for the deconvolution of very large datasets.Comprehensive information regarding the algorithm and its application to MaNGA data cubes can be found in Chung et al. (2021).Thus, in this paper, we provide a concise description of the application of the LR algorithm to the IFU data.
The LR algorithm can be expressed using a simple equation, where u n represents the nth estimate of the maximum likelihood solution, while d stands for the original PSFconvolved image (hence u 0 = d).The parameter p denotes a two-dimensional (2D) PSF, and the symbol ⊗ denotes 2D convolution.
The LR algorithm is applied to the MaNGA cube data by deconvolving the 2D image slice at each wavelength bin individually.The MaNGA data cubes provide PSF full width half-maximum (FWHM) values for the g, r, i, and z bands.We conduct a linear fitting on these FWHM values at the respective wavelengths of the four bands.The FWHM of the PSF at each wavelength bin in data cubes is derived from this linear model through interpolation. 1 In the deconvolution process for a specific wavelength, the 2D Gaussian function with the interpolated FWHM value is applied. 2 The number of iterations (N iter ) in the LR algorithm is set to 20, which is determined to be optimal through the tests in Chung et al. (2021).Going beyond N iter = 20 does not lead to a significant enhancement in deconvolution quality; instead, it introduces additional artifacts in the image with amplified noise.
The test for the applications of the LR algorithm to simulated IFU data in Chung et al. (2021) demonstrates that the deconvolution process enables us to effectively recover the true stellar kinematics of galaxies.For example, the test indicates that the luminosity-weighted stellar angular momentum can be restored with an underestimation of less than ∼ 5% in the majority of cases where the deconvolution process is used.By contrast, without applying deconvolution, the level of underestimation can reach 20%-30%.

Extracting Stellar Kinematics
We use the method outlined in Yoon et al. (2021) to derive stellar kinematics from the MaNGA data, such as line-of-sight velocities and velocity dispersions.The code used here for extracting stellar kinematics is the Penalized Pixel-Fitting (pPXF) method, which is designed to use the maximum penalized likelihood formalism for full spectrum fitting on galaxy spectra.We utilize MILES single stellar population models (Sánchez-Blázquez et al. 2006;Vazdekis et al. 2010;Falcón-Barroso et al. 2011) with the Padova+00 isochrone (Girardi et al. 2000) and the initial mass function of Chabrier (2003) for input model templates in pPXF.The model templates encompass six metallicity values spanning −1.71 ≤ [M/H] ≤ 0.22 3 and 10 age values ranging from 0.07-12.59Gyr, totaling 60 model templates.Following the approach of Belfiore et al. (2019), we incorporate an eighth-order additive Legendre polynomial into the fit to improve the quality of the derived stellar kinematics (see Emsellem et al. 2004).
As in the method of Cappellari (2017), the model templates are convolved with a Gaussian function to align with the resolution of the MaNGA spectra.Additionally, the spectra are shifted to the rest frame before extracting stellar kinematics.The fitting process is carried out after masking the pixels around known emission lines and bad pixels (such as those with low coverage depth, dead fibers, or contamination from foreground stars, etc.) flagged in the MaNGA data reduction pipeline (Law et al. 2016).The fitting range is confined to the wavelength range of 3700-7400 Å to match the wavelength coverage of the model templates (3540-7410 Å).In the third and fourth rows of Figures 1-5, we present examples of the 2D line-of-sight stellar velocity and velocity dispersion maps for our ETG sample.
We calculate the dimensionless spin parameter luminosity-weighted specific stellar angular momentum, 3 As mentioned in Yoon et al. (2022), the use of alternative MILES models derived from the BaSTI isochrone (Hidalgo et al. 2018), which covers up to a higher metallicity of [M/H] = 0.4, yields similar kinematics for the most massive ETGs with log(Mstar/M ⊙ ) > 11.2, whose metallicities are expected to be high.
λ R following the method in Emsellem et al. (2007), in which F i represents the flux of the ith bin, while R i denotes the circular radial distance from the center to the ith bin.Here, we use the F i of r-band images, which is deconvolved in the same manner as with IFU data.As for V i and σ i , these refer to the line-of-sight velocity and velocity dispersion of the ith bin, respectively.The summation is performed over N pixels located within the photometric ellipse.The spin parameter λ R is scaled by V 2 i + σ 2 i , which serves as a proxy for mass.The parameter λ R approaches unity in the cases where the system is rotation dominated, and converges to zero when the system is predominantly pressure supported (dominated by random motions of stars).For the computation of λ R , we consider only spaxels with a median S/N ≥ 5. We do not use the spaxels with σ < 40 km s −1 in the calculation, as such low σ values can be unreliable due to the instrumental resolution limit (Penny et al. 2016;Lee et al. 2018). 4We further remove spurious spaxels with |V | ≥ 500 km s −1 in the calculation of λ R .
The parameter λ R exhibits little variation over a broad range of viewing angles, except when the angle is nearly face-on (Emsellem et al. 2007;Jesseit et al. 2009;Bois et al. 2011).This stability arises from the concurrent decrease of ⟨V ⟩ and ⟨σ⟩ as the inclination decreases, resulting in only minor variations in their ratio (Jesseit et al. 2009).Consequently, λ R serves as a robust observational indicator for the intrinsic angular momentum in the majority of galaxies.
For this reason, λ R has been widely used in many previous studies examining the stellar angular momentum of galaxies (Jesseit et al. 2009;Emsellem et al. 2011;Fogarty et al. 2015;Cappellari 2016;Oh et al. 2016;Choi & Yi 2017;Graham et al. 2018).These works have used λ R within R e (hereafter, λ Re ) for statistical analyses of galaxy stellar angular momentum, and we also use λ Re throughout this study.Here, the parameters R e and ellipticities (ε) of galaxies are derived using the elliptical Petrosian flux in the r band. 5The stellar masses (M star ) of the galaxies used in this study are likewise computed from the elliptical Petrosian fluxes.These parameters are sourced from the NASA-Sloan Atlas catalog, which is the base catalog for selecting target galaxies in the MaNGA project (Wake et al. 2017).

ETG Sample
We limit our sample to galaxies within the redshift range of z < 0.055.We exclude galaxies at higher redshifts, due to their small angular sizes and the cosmological surface brightness dimming effects (see Equation 6in Yoon & Park 2020), which render the detection of tidal features challenging.As in Yoon et al. (2022), we use galaxies with M star ≥ 10 9.65 M ⊙ .Note that tidal features are rarely detected at M star < 10 9.65 M ⊙ (Yoon et al. 2022).The number of MaNGA galaxies after implementing the redshift and stellar mass criteria is 5745.
For the selection of ETGs among MaNGA galaxies, we use a value-added catalog (MaNGA Visual Morphologies from SDSS and DESI images; Vázquez-Mata et al. 2022), which contains morphology classification information based on visual inspection of SDSS and DESI images, for all galaxies in MaNGA Data Release 17.The morphology classifications of this catalog are in good agreement with the visual classifications of Nair & Abraham ( 2010) and the machine-learning-based results of Domínguez Sánchez et al. (2022), showing median scatters in T type of 1.2 and 1.48, respectively (Vázquez-Mata et al. 2022).We double-check 1838 ETGs (T type≤ 0 in the catalog) through visual inspection of SDSS and DESI color images and exclude 26 galaxies, as they have morphologies more similar to late-type galaxies.
In order to use reliable λ Re values in our analysis, we do not include IFU data where over 25% of spaxels within λ Re are excluded, due to the conditions described in Section 2.3, such as median S/N < 5, σ < 40 km s −1 , |V | ≥ 500 km s −1 , or contamination from foreground stars, etc.We additionally rule out IFU data where the total number of spaxels within R e is fewer than 45.The number of remaining ETGs is 1312, after excluding such IFU data.Finally, 68 ETGs are excluded due to poor image quality resulting from their proximity to bright sources (Section 2.5).As a result, the final sample consists of 1244 ETGs.

Detection and Classification of Tidal Features
For the detection of tidal features, we use images of the DESI Legacy Survey Data Release 10 ( Dey et al. 2019).The DESI Legacy Survey is a combination of three widearea surveys, which are the Dark Energy Camera Legacy Survey, the Beijing-Arizona Sky Survey, and the Mayall z-band Legacy Survey.These surveys cover a total area of ∼ 14, 000 deg 2 .The median surface brightness limit (1σ of the background noise over a 1 ′′ × 1 ′′ region) of g-and r-band DESI images is ∼ 27 mag arcsec −2 .This limit is similar to that of deep coadded r-band images of the Stripe 82 region of SDSS (Yoon & Lim 2020;Yoon et al. 2022Yoon et al. , 2023)), which have been widely used to identify low-surface brightness tidal features around galaxies (e.g., Kaviraj 2010;Schawinski et al. 2010;Hong et al. 2015).
We perform a visual inspection of the g-band and rband images, and composite color images of the g, r, and z bands to identify tidal features.If necessary, especially when examining the inner regions of galaxies, we also inspect the residual images of the g and r bands, in which the 2D light models of galaxies6 are subtracted from the original images.During the visual inspection of the images, we fine-tune the scale of the pixel values and enhance signals by smoothing images using Gaussian kernels of various sizes for the better identification of diffuse and faint tidal features.In this process, we find 68 galaxies that are located too close to bright stars and large galaxies.We exclude them, as mentioned in Section 2.4 since it is difficult to identify tidal features in the images with the high background levels caused by nearby very bright sources.
We classify the detected tidal features into three types, which are tails, streams, and shells.These are commonly used categories in previous studies (Duc et al. 2015;Mancillas et al. 2019;Bílek et al. 2020Bílek et al. , 2023;;Sola et al. 2022).Tidal tails are thick, radially elongated structures that are visibly connected to the host galaxy.These elongated stellar features, potentially formed during major mergers, are morphologically similar to streams but possess a higher thickness, sometimes reaching the size of the host galaxy itself.However, there is no obvious visual difference between tails and streams in some cases (Bílek et al. 2020;Sola et al. 2022).Tidal streams are thin, elongated structures that usually resemble narrow filaments, likely tracing minor mergers.In some cases, they are connected to smaller companion galaxies.Shells are concentric, arc-shaped features with sharp edges.The alignment of arc features can either follow a common axis or appear to be randomly spread around the host galaxy.As shells extend to larger radii, they become more diffused.The exact conditions for their formation remain a topic of debate, with radial mergers generally being favored (Quinn 1984;Dupraz & Combes 1986;Johnston et al. 2008;Hendel & Johnston 2015;Pop et al. 2018;Karademir et al. 2019).If a galaxy has more than two types of tidal features, we assign the galaxy to all corresponding groups.For example, ETGs both with shell and stream features are classified as ETGs with shell-type tidal features, and at the same time, as ETGs with streams.
We evaluate the robustness of the separation of tidal features by combining independent categorizations provided by two authors (W.B. and K.C.) and adjusting the initial classifications (established by Y.Y. with assistance from J.K. and H.C.) based on the majority rule.We find that 93.7% (238/254) of the tidal feature separations remain unchanged from the initial classifications when decisions are made using the majority rule among the three sources.Only 14 out of the 254 cases (5.5%) are modified from the initial classifications. 8For the remaining two cases, where all classifications from the three sources differ from each other, we follow the initial categorizations.The outcome of this process indicates that the categorization of tidal features into three types is quite robust, even though it relies on visual inspections that can be subjective The number of ETGs with the tidal features of tails, streams, and shells are 59, 177, and 72, respectively.The numbers of ETGs that have a purely single type of tidal features are 30, 128, and 45, for those with tails, streams, and shells, respectively.There are 27 ETGs that have both tails and streams, while 25 ETGs have both streams and shells.Five ETGs have tail-and shelltype tidal features.Three ETGs exhibit all three types of tidal features simultaneously.
Image depth can introduce potential biases in the detection of tidal features, thereby influencing our results.For example, the simulation of Mancillas et al. (2019) suggests that the detection of streams can be largely dependent on the surface brightness limit of images.In order to roughly figure out the impact of image depth on our results, we divide the 254 ETGs with tidal features into two groups: one group consists of ETGs with prominent tidal features that are clearly visible without the need for smoothed or residual images, or are 8 One ETG with a shell feature is reclassified as an ETG with a stream feature.Nine ETGs with streams are changed to seven ETGs with tails and two ETGs with tails+streams.One ETG with a shell+stream feature is modified to an ETG with a shell+tail feature.Three ETGs with tails+streams are reclassified as three ETGs with streams.
even visible in shallow images of SDSS (151 ETGs); the other group comprises the remaining ETGs with faint tidal features (103 ETGs).We find that the respective results obtained based on these two groups are consistent with the main findings derived from the full sample.For example, the median λ Re for ETGs with prominent tidal features is 0.20 ± 0.01.The median λ Re values for ETGs with prominent shells, tails, and streams are 0.13±0.02,0.21±0.03,and 0.22±0.02,respectively.The fraction of slow rotators among ETGs with prominent tidal features is 0.27 ± 0.04, while the slow rotator fractions for those with prominent shells, tails, and streams are 0.47 ± 0.07, 0.20 ± 0.06, and 0.20 ± 0.04, respectively.These quantities, including those from faint tidal features, are consistent within the margins of error with those derived from the full sample specified in Section 3.This implies that image depth is not a critical factor affecting our results within the depth range we can probe with DESI images.However, the use of much deeper images from future large-area surveys will enable us to more fully understand the influence of image depth.
Another factor that can produce potential biases is the projection angle.According to Mancillas et al. (2019), the detection of streams and tails does not depend on projection angles, while the detection of shells can be more substantially affected by the projection.However, if only ∼ 20% of shells are missed when restricted to a single projection, as found in the simulations of Pop et al. (2018), such dependency in shell detections is unlikely to greatly impact our main results.Moreover, if the detection of prominent tidal features with high surface brightness is less affected by projection angles, as shown by Martin et al. (2022), the fact that the respective results derived from both the samples of prominent and faint tidal features do not show significant discrepancies from the overall findings from the full sample implies that biases due to projection angles may not be substantial enough to fundamentally alter our results.

RESULTS
In this section, we first present the results when ETGs are divided into those with and without tidal features.Following that, we demonstrate our novel findings when ETGs with tidal features are further divided into three categories based on the types of tidal features.
The left panel of Figure 6 displays the distribution of ETGs with/without tidal features in the λ Re versus ε plane and its projected histogram for λ Re .The figure exhibits that ETGs with tidal features have lower λ Re than those without tidal features.The median λ Re for ETGs with tidal features is 0.19 ± 0.01, whereas for ETGs without tidal features, it is 0.31 ± 0.01.We perform a Kolmogorov-Smirnov (KS) test for the distributions of λ Re for the two ETG categories to assess the significance of the difference.The test yields that the probability (0 ≤ p ≤ 1) of the null hypothesis in which  the two distributions originate from the same distribution is p = 6.0 × 10 −11 , which indicates that the two λ Re distributions are significantly different from each other.The fraction of slow rotators (f slow ) for the ETGs with/without tidal features, derived from the criterion of λ Re < 0.08 + ε/4 and ε < 0.4 (under the solid black line in Figure 6; Cappellari 2016), is f slow = 0.27 ± 0.03 for ETGs with tidal features and f slow = 0.14 ± 0.01 for ETGs without tidal features.9This suggests that more than one-fourth of ETGs with tidal features are slow rotators, a fraction that is nearly twice as high as that of ETGs without tidal features.
The left panel of Figure 7 presents the distribution of ETGs with/without tidal features in the λ Re versus log M star plane.The upper panels of Figure 8 illustrate the mean λ Re for ETGs with/without tidal features as a function of log M star and effective velocity dispersion within R e (hereafter σ e ).The parameter σ e is determined using Equation 3 in Graham et al. (2018).These figures show that λ Re of ETGs with tidal features are lower by ∼ 0.06 than that of ETGs without tidal features in all the M star and σ e bins.KS tests on the distributions of λ Re for the two ETG categories, corrected for the variation of λ Re as a function of M star and σ e ,10 yields p < 1.1 × 10 −8 , which demonstrates that the λ Re values of the two ETG categories are still significantly different from each other even after accounting for the relation between λ Re and mass.
The fraction f slow follows exactly the same trend as shown in the middle panels of Figure 8, so that in all the M star and σ e bins, f slow of ETGs with tidal features is higher by ∼ 0.08 than that of ETGs without tidal features.We note that the mean λ Re is the lowest; hence, f slow is the highest in the most massive bin of log(M star /M ⊙ ) ≥ 11.0, which is consistent with the results of previous studies (Emsellem et al. 2007;Graham et al. 2018).
The histograms presented in the lower panels of Figure 8 indicate a higher occurrence of tidal features in more massive ETGs.The median log(M star /M ⊙ ) values are 10.88 ± 0.03 and 10.54 ± 0.02 for ETGs with and without tidal features, respectively.A KS test performed on the distributions of log(M star /M ⊙ ) for the two ETG categories provides p = 1.8 × 10 −15 , indicating a highly significant difference between the two distributions.This result aligns with those reported in previous studies (Bílek et al. 2020(Bílek et al. , 2023;;Yoon & Lim 2020).
Our results and trends regarding λ Re when ETGs are separated into those with and without tidal features are almost identical to those presented in Yoon et al. (2022).Therefore, by using a sample of ETGs more than six times larger than that used in Yoon et al. (2022), we verify the principal findings of Yoon et al. (2022).Now, we present the results when ETGs with tidal features are subdivided into three categories based on the types of tidal features, such as tails, streams, and shells.The right panel of Figure 6 shows the distribution of ETGs with shells, tails, and streams, as well as those without tidal features, in the λ Re versus ε plane.The histograms of the λ Re distribution for each ETG category are shown on the right side of the panel.Figure 6 demonstrates that among ETGs with tidal features, those with shells exhibit a much lower λ Re than those with tails or streams.The median λ Re for ETGs with shells is 0.13 ± 0.02, while for those with tails and streams, the median λ Re values are 0.22 ± 0.02 and 0.22 ± 0.01, respectively.KS tests reveal that the distributions of λ Re are significantly different (with p ∼ 6×10 −4 ) between ETGs with shells and those with streams or tails, whereas there is no significant difference in the λ Re distributions (with p = 0.89) between ETGs with streams and those with tails.However, the λ Re values of ETGs with tails or streams are still significantly lower than those of ETGs without tidal features, given that p is less than 0.01.
The fractions f slow , calculated from the criterion of Cappellari (2016), are 0.46 ± 0.06, 0.20 ± 0.05, and 0.22 ± 0.03 for ETGs with shells, tails, and streams, respectively.11Thus, nearly half of ETGs with shells are slow rotators.This fraction is more than twice as high as that of ETGs with tails or streams, and more than three times higher than that of ETGs without tidal features.
The right panel of Figure 7 displays the distribution of ETGs with shells, tails, and streams, as well as those without tidal features, in the λ Re versus log M star plane.The upper and middle panels of Figure 9 show the mean λ Re and f slow , respectively, as a function of log M star and σ e for ETGs with shells, tails, streams, and those without tidal features.The lower panels of Figure 9 display histograms of log M star and σ e for each ETG category.
Figure 9 reveals that ETGs with streams and those with tails exhibit very similar λ Re and f slow in each bin, with the differences in λ Re and f slow being less than 0.008.They also display comparable trends as a function of mass.Moreover, they have similar mass distributions, with median log(M star /M ⊙ ) values of 10.8.Due to their similarity, ETGs with streams and those with tails are grouped together in the subsequent description.
Figures 7 and 9 demonstrate that λ Re of ETGs with streams + tails are lower by ∼ 0.05 than that of ETGs without tidal features in all the M star and σ e bins.Furthermore, ETGs with shells have λ Re lower by ∼ 0.07 than ETGs with streams + tails in all the M star and σ e bins.KS tests on the distributions of λ Re for ETGs with streams + tails and those without tidal feature, corrected for the variation of λ Re as a function of M star and σ e , give p < 5 × 10 −5 .The same tests on the distributions of λ Re for ETGs with shells and those with streams + tails yield p < 5 × 10 −3 .These KS tests suggest that the difference in λ Re between any two ETG groups remains significant even after accounting for the relation between λ Re and mass.
These results are reflected in the trend for f slow as displayed in the middle panels of Figure 9.In all the M star and σ e bins, f slow of ETGs with shells is higher by ∼ 0.2 than that of ETGs with streams + tails, while f slow of ETGs with streams + tails is higher by ∼ 0.04 than that of ETGs without tidal features in the low-mass range.
We note that ETGs with purely shell-type tidal features and no features of other types have slightly lower λ Re by ∼ 0.01 (resulting in a higher f slow by ∼ 0.04) in each mass bin than ETGs in which at least one of the tidal features is a shell.By contrast, ETGs that have purely stream-or tail-type tidal features without shells exhibit very slightly higher λ Re by ∼ 0.01 (hence a lower f slow by ∼ 0.02) in each mass bin than ETGs in which at least one of the tidal features is a stream or a tail. 12he histograms shown in the lower panels of Figure 9 show that shells are typically found in ETGs that are more massive by 0.2 dex than those with streams and tails.The median log(M star /M ⊙ ) values are 11.01 ± 0.05 and 10.82 ± 0.03 for ETGs with shells and those with streams + tails, respectively.A KS test conducted on the distributions of log(M star /M ⊙ ) for the two ETG categories yields p = 8 × 10 −4 , indicating a significant difference between the two distributions.
In sum, ETGs with tidal features have a lower λ Re and hence a higher f slow than ETGs without tidal features.When tidal features are subdivided into shells and streams/tails, ETGs with shells, which exhibit a much lower λ Re than those with tails or streams, contribute the most to reducing λ Re (increasing f slow ), while the λ Re values of ETGs with tails or streams are still lower than those of ETGs without tidal features.These findings generally remain valid even when ETGs are divided into several mass bins.In addition, tidal features are more frequently detected in more massive ETGs, with shells typically found in slightly more massive ETGs compared to streams and tails.
4. DISCUSSION Our results imply that mergers that generate shelltype tidal features also lower λ Re more effectively than mergers that create tidal streams or tails.The simulations in previous studies (Moody et al. 2014;Lagos et al. 2018b;Li et al. 2018) show that radial infall of satellites (radial mergers with orbits of low-angular momentum) can reduce the stellar angular momentum (rotation support) of merger remnants more effectively than circular infall of satellites (mergers with orbits of high angular momentum).Another set of simulations (Quinn 1984;Dupraz & Combes 1986;Johnston et al. 2008;Hendel & Johnston 2015;Pop et al. 2018;Karademir et al. 2019) suggests that radial mergers are capable of producing shell-type tidal features, whereas satellite infall with circular orbits can make stream-like linear tidal features.Therefore, our results support the scenario that radial mergers, which are better at reducing λ Re compared to circular mergers, are more closely associated with the formation of shell-type tidal features rather than tidal streams or tails.
An alternative hypothesis that may explain our results is that the infall of galaxies into ETGs with already low λ Re tends to result in the formation of shell-type tidal features.However, this should be verified through the analysis of galaxy simulations.Another potential explanation for our findings is that ETGs with streams could be, on average, at earlier stages of merging (hence λ Re has not fully reduced yet) than those with shells because circular orbits have longer merger timescales than radial orbits.
The finding that tidal features are more frequently detected in more massive ETGs aligns well with the understanding that the formation of more massive ETGs involves a higher number of mergers (Dubois et al. 2016;Yoon et al. 2017;Davison et al. 2020).Our finding that shells are typically found in slightly more massive ETGs compared to streams and tails may be due to the nature of dynamical friction.Dynamical friction is stronger in the case when infall galaxies are more massive (major mergers).This is because massive satellites exert a greater influence on their surroundings and consequently experience stronger dynamical friction.Since dynamical friction can make the infall orbit of satellite galaxies more radial, massive satellites are more likely to be accreted through radial orbits in the later stages of mergers (Amorisco 2017;Pop et al. 2018).This could explain our observational result that shell-type tidal features are detected in slightly more massive ETGs, which is also found in the simulation of Pop et al. (2018) and in the observation of Bílek et al. (2023).We note that lowmass satellites require almost purely radial infall orbits at the initial accretion time in order to produce shells in the simulation of Pop et al. (2018).
As stated in Section 2.5 and illustrated in Figures 6 and 7, among the 254 ETGs with tidal features, 27 have both tails and streams, and 25 have both shells and streams.However, in contrast, only five ETGs exhibit both shells and tails, which is a notably smaller count.This can be explained if tails and shells predominantly stem from major or intermediate mergers, whereas streams, the most prevalent tidal features observed in 177 ETGs in our sample, typically arise from more frequent minor mergers.Indeed, many previous studies associate tails with major mergers and streams with minor mergers (Duc et al. 2015;Mancillas et al. 2019;Bílek et al. 2020Bílek et al. , 2023;;Sola et al. 2022), and our findings indicate that shells are more likely formed through the accretion of massive satellites.Therefore, the simultaneous occurrence of shells and tails is rare, likely due to the low possibility of two major or intermediate mergers occurring consecutively within the lifespan of tidal features.
In addition, if the mass ratio of galaxy mergers is the decisive factor for distinguishing between the formation of tails and streams, our finding that ETGs with tails and those with streams exhibit identical distributions in λ Re implies that the kinematics of merger Yoon et al. (2024) remnants may not be significantly determined solely by the mass ratio of the merging galaxies, highlighting the importance of other factors influencing their kinematics.
5. SUMMARY We investigate differences in stellar kinematics (λ Re ) among ETGs with several types of tidal features and those without tidal features.This is done by categorizing tidal features, which serve as direct evidence of recent mergers, into shells, streams, and tails.Through this study, we broaden our understanding of the impact of galaxy mergers on λ Re of ETGs.This study is an extension of that of Yoon et al. (2022), using a sample of ETGs that is more than six times larger, which enables us to divide tidal features into several categories.
We use MaNGA IFU data for the analysis of the stellar kinematics of ETGs.In order to reduce the seeing effect in stellar kinematics, we apply deconvolution to the MaNGA IFU data using the LR algorithm.The pPXF code, which conducts full spectrum fitting on galaxy spectra, is used to extract stellar velocities and velocity dispersions.We detect and categorize tidal features through a visual inspection of DESI Legacy Survey images, which provide sufficient depth for the study of tidal features.The final sample consists of 1244 ETGs with redshifts of z < 0.055 and stellar masses of M star ≥ 10 9.65 M ⊙ .The main results of this study are as follows.

ETGs with tidal features typically have reduced
λ Re values that are lower by 0.12 (hence a higher fraction of slow rotators by 0.13) compared to ETGs without tidal features, showing a significant difference in λ Re distributions between the two ETG categories.
2. ETGs with shells, which have λ Re values lower by 0.1 than ETGs with tails or streams, contribute the most to reducing λ Re .As a result, nearly half of ETGs with shells are classified as slow rotators.This fraction is more than twice as high as that of ETGs with tails or streams, and over three times higher than that of ETGs without tidal features.
3. The λ Re values of ETGs with tails or streams are slightly lower than those without tidal features, while there is no significant difference in λ Re between ETGs with streams and those with tails.
4. These trends generally remain valid even when ETGs are segmented into several mass bins.
5. Our findings support the idea that radial mergers of low-angular momentum orbits, which are more effective at reducing λ Re than circular mergers, are more closely related to the formation of shell-type tidal features than streams or tails.6. Shells tend to be found in slightly more massive ETGs compared to streams and tails.This may be accounted for by the fact that massive satellite galaxies are more likely to be accreted through radial orbits, due to the nature of dynamical friction.
Increasing the size of the galaxy sample using deep and large survey images enhances our understanding of the origins of tidal features and the impact of galaxy mergers, as demonstrated in this study.We expect that much deeper images from future large-scale surveys will further advance our comprehensive understanding of galaxy mergers and tidal features.This research was supported by the Korea Astronomy and Space Science Institute under the R&D program (Project No. 2024-1-831-00), supervised by the Ministry of Science and ICT.

λFigure 1 .λλλλFigure 2 .λλλλFigure 3 .λλλFigure 4 .Figure 5 .
Figure 1.Examples of normal ETGs without tidal features.First row: color images from SDSS.The plate ID (e.g., 7978) and IFU design ID (e.g., 6104) are shown in the color image.The horizontal bar in the color image denotes the angular scale of the image.The green square denotes the window size of the 2D line-of-sight velocity and velocity dispersion maps shown in the third and fourth rows.The λR e and log(Mstar/M⊙) of each galaxy are shown in the bottom of each color image.Second row: r-band deep images of the DESI Legacy Survey.The angular scale of the deep image is the same as that of the color image.Third row: 2D line-of-sight velocity maps within 1.5Re.Fourth row: 2D line-of-sight velocity dispersion maps within 1.5Re.The black ellipses in the third and fourth rows indicate the areas within which λR e values are calculated.The color bars positioned above the panels in the third and fourth rows represent the color-coded velocity and velocity dispersion scales, respectively.The symbols ∆α and ∆δ denote relative R.A. and decl., respectively.
. The examples of shallow SDSS color images and deep DESI images for ETGs without tidal features are shown in the first and second rows of Figure 1.The examples for ETGs with tidal features are displayed in Figures 2-5 (Figure 2: ETGs with shells; Figure 3: ETGs with streams; Figure 4: ETGs with tails + streams; Figure 5: ETGs with shells + streams/tails).

Figure 6 .Figure 7 .
Figure6.Distribution of ETGs in the λR e versus ε plane.In the left panel, ETGs are divided into two categories based on the presence of tidal features.In the right panel, ETGs with tidal features are additionally subdivided into three categories based on the types of tidal features, as indicated in the legend.The right side of each panel exhibits a histogram of λR e for each ETG category, with the horizontal dashed lines indicating the median λR e for each respective ETG category.The solid black line and the black-dashed line represent the criteria for identifying slow rotators, as defined inCappellari (2016) andEmsellem et al.  (2011), respectively.The pink line on the λR e versus ε plot represents λR e values for the edge-on isotropic rotator fromBinney (2005) with various intrinsic ellipticities(Cappellari 2016).The light blue line denotes λR e values for edge-on galaxies following the anisotropy versus intrinsic flattening relation described in Equation11ofCappellari (2016).The gray-dotted lines indicate different versions of the light blue line at different inclination angles with a step size of 0.1 • .The gray-dashed line traces the path of galaxies with a constant intrinsic ellipticity as the inclination angle varies.

Figure 8 .
Figure 8. Upper panels: mean λR e for ETGs with/without tidal features as a function of log Mstar and σe.The error bars indicate the standard error of the mean.Middle panels: the fraction of slow rotators (f slow ) for ETGs with/without tidal features as a function of log Mstar and σe.Here, we use the criterion for defining slow rotators from Cappellari (2016) (the solid black line in Figure 6.)The error bars denote the standard error of the proportion.The gray vertical lines in the upper and middle panels indicate the boundaries of the bins used for calculating f slow and mean λR e .Lower panels: the histograms of log Mstar and σe for ETGs with and without tidal features.The vertical dashed lines represent the median values of log Mstar or σe for each ETG category.

Figure 9 .
Figure 9.This figure is identical to Figure 8, except that ETGs with tidal features are subdivided into those with shells, tails, and streams.The horizontal bars in the upper and middle panels indicate the ranges of the bins used for computing f slow and mean λR e .The gray vertical dashed lines in the bottom panels indicate the median values for ETGs without tidal features.
and Yoon et al. (2022), our classifications of tidal features in the Stripe 82 region are compared with those made by Kaviraj (2010), who classified ETGs with M r 7 < −20.5 at z < 0.05 in the Stripe 82 region into normal (relaxed) ETGs and ETGs with tidal features.The comparison reveals that over ∼ 90% of the classifications are in agreement with each other (see Yoon & Lim 2020 and Yoon et al. 2022 for more details).