The SAMI Galaxy Survey: Stellar Population Gradients of Central Galaxies

The SAMI data consist of three-dimensional data cubes: two spatial dimensions and a spectral dimension.

The wavelength coverage is from 3750 to 5750 Å in the blue arm and 6300 to 7400 Å in the red arm, with a spectral resolution of R = 1812 (2.65 Å FWHM) and R = 4263 (1.61 Å FWHM), respectively (van de Sande et al. 2017), so that two data cubes are produced for each galaxy target. Only the blue arm is used to determine stellar population parameters, since there are only a few absorption features in the red wavelength range useful to this end.

The SAMI data reduction is fully described in Allen et al. (2015) and Sharp et al. (2015), and we give a brief summary here.

A dedicated SAMI Python package (that incorporates the 2DFDR package¹² ; Croom et al. 2004; Sharp & Birchall 2010) is used to perform bias subtraction and flat-field normalization, remove cosmic rays, and calibrate the absolute flux to reduce the raw SAMI observations (Allen et al. 2015). Each galaxy field was observed in a set of approximately seven 30 minute exposures that are aligned together by fitting the galaxy position within each hexabundle with a two-dimensional Gaussian and fitting a simple empirical model describing the telescope offset and atmospheric refraction to the centroids. The exposures are then combined to produce a spectral cube with regular 05 spaxels, with a median seeing of 21. More details for Data Release 1 and Data Release 2 reduction can be found in Green et al. (2018) and Scott et al. (2018), respectively.

The GAMA group sample includes all galaxy associations with two or more members, so to ensure the robustness of our group sample, we first select all SAMI observed galaxies that belong to GAMA halos with at least five members to ensure that they have a robustly estimated velocity dispersion (Robotham et al. 2011). Moreover, only galaxies classified as cluster members by Owers et al. (2017) are taken as cluster galaxies. This gives us an initial sample of 574 galaxies (out of a potential 2502 observed by GAMA) in 205 groups in the GAMA regions and 861 galaxies in eight clusters.

Stellar population measurements are compromised for 100 galaxies because nearby galaxies or stars affect their observation (van de Sande et al. 2017). In addition, we also exclude three more galaxies whose g − i colors suggest contamination from other objects in the field of view and eight galaxies whose annuli do not reflect the shape of their isophotes and/or the central bin is not well aligned with the peak of their flux map (since they could lead to a biased gradient).

In this paper, we are particularly interested in whether the evolution of central and satellite galaxies has an environmental dependence. We focus on the passive central galaxies to ensure a like-to-like comparison in determining how their central and satellite status affects their stellar population gradients. We use the SAMI spectroscopic classifications presented in Owers et al. (2019) to select a homogeneous sample. The galaxies are classified as star-forming, passive, or Hδ-strong using the absorption and emission line properties of each SAMI spectrum. We select 843 passive galaxies; of these, 98 are central and 745 are satellite galaxies (Figure 1).

Zoom In Zoom Out Reset image size

To ensure the stellar population measurements we use are reliable, we exclude all galaxies with masses less than ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })=9.5$, owing to the low signal-to-noise ratio (S/N) and completeness of these galaxies (we note that selecting galaxies with ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\gt 10$ does not change the conclusions we draw). This leaves 819 galaxies (98 central and 721 satellite galaxies). This excludes 40 nonpassive central galaxies with ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\gt 9.5$. The median halo mass of the nonpassive central galaxies is M₂₀₀ = 10^13.3 M_⊙.

We also make a quality cut in S/N and velocity dispersion. For each galaxy, we only use the annular measurements that have S/N ≥ 10 and velocity dispersions (σ) between 75 and 400 km s⁻¹. The [α/Fe] measurements were also restricted to annuli with S/N ≥ 20 owing to the greater uncertainties associated with measuring this parameter.

We select galaxies that have at least one annular measurement between 0.1 and 1 R_e and at least three in total, all meeting the S/N and σ criteria, in order to probe similar spatial regions for our sample of galaxies. This is due to some of the satellite galaxies having effective radii <11–12, such that the first aperture probes radii greater than 1 R_e. This radial criterion is the primary source of exclusion for our sample.

These selection criteria result in a final sample of 533 galaxies with metallicity and age gradients (96 central galaxies and 437 satellite galaxies) and 332 galaxies with [α/Fe] gradients (79 centrals and 253 satellites).

The distribution of g − i color with stellar mass for the galaxies selected is illustrated in Figure 1. Figures 2 and 3 illustrate the completeness of the selected sample. As shown, our sample is representative of the parent sample in stellar mass and halo mass, with 96% completeness for passive galaxies with ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\gt 10.5$.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Stellar masses are estimated from g and i magnitudes using an empirical proxy developed from GAMA photometry (Taylor et al. 2011; Bryant et al. 2015). For cluster galaxies, stellar masses are derived using the same approach (Owers et al. 2017).

The g and i magnitudes are taken from Sloan Digital Sky Survey (SDSS; York et al. 2000) images for GAMA galaxies and VLT Survey Telescope (VST) ATLAS (Shanks et al. 2015) and SDSS DR9 (Ahn et al. 2012) observations for cluster galaxies. The VST/ATLAS data were reprocessed as described in Owers et al. (2017).

Circularized effective radii, R_e, were determined by fitting a Sérsic profile to r-band images. For group galaxies, these radii are taken from the GAMA Sérsic catalog v0.7 (Kelvin et al. 2012), whereas for cluster galaxies, the R_e, expressed as the radius of the major axis, are taken from Owers et al. (2019) and converted to a circularized radius using the axial ratio for each galaxy. For the galaxies that are in common, cluster galaxies' R_e are consistent when measured from VST/ATLAS and SDSS/GAMA (Owers et al. 2019).

We use the GAMA Galaxy Group Catalogue (G3C; Robotham et al. 2011) to define the galaxy groups in the GAMA regions. In this catalog, galaxies are grouped using an adaptive friends-of-friends algorithm, taking advantage of the high spectroscopic completeness of the GAMA survey (∼98.5%; Liske et al. 2015). Following the recommendation of Robotham et al. (2011), we use the "iterative center" as the center of our halo. This center is found through an iterative procedure: at each step, the r_AB-band luminosity centroid of the halo is found and the most distant galaxies are rejected. When only two galaxies remain, the brightest is taken as the central galaxy.

For the GAMA groups, we use the halo mass, M₂₀₀, the mass contained within R₂₀₀ (the radius at which the density is 200 times the critical density; White 2001), calculated from the GAMA group velocity dispersion and calibrated using halos from simulated mock catalogs (see Robotham et al. 2011). The cluster halo masses are taken from Owers et al. (2017), who calculated them from a caustic analysis of the galaxy velocities and their distance from the cluster center.

It is important to note that, following Owers et al. (2017), we apply a scaling factor of 1.25 to the cluster halo masses to be consistent with the GAMA halo masses. We also scale the GAMA halo masses (H₀ = 100 km s⁻¹ Mpc⁻¹) to be consistent with the cosmology used here.

We define the halo central galaxy as the most massive galaxy within 0.25 R₂₀₀ (e.g., Oliva-Altamirano et al. 2017). To identify the central galaxies for the group sample, we check that the galaxy identified as "iterative central" (Robotham et al. 2011) is also the most massive galaxy. This is true for 188/205 groups in our sample. Seventeen groups have different galaxies selected as the iterative central and most massive. For these 17 groups, we find the most massive galaxy within a radius of 0.25 R₂₀₀. For most of the groups (12/17), the most massive galaxy within 0.25 R₂₀₀ is the iterative central galaxy. For the five groups where another galaxy is found to be the most massive within 0.25 R₂₀₀, we select that galaxy as the central galaxy.

A similar procedure is carried out to select the central galaxy in the clusters in order to ensure consistency between the samples. We identify the galaxy that sits closest to the center of the cluster (cluster centroids are taken from Owers et al. 2017), as well as the most massive galaxy in the cluster. For three out of eight clusters, these galaxies are the same. For the other five clusters, we find the most massive galaxy within a radius of 0.25 R₂₀₀. For two clusters, this galaxy is also the central one, whereas for three out of five clusters (A168, A2399, and A4038), the most massive galaxy within 0.25 R₂₀₀ is not the galaxy closest to the center. This is consistent with the dynamical state of these clusters as discussed in Brough et al. (2017) and Owers et al. (2017). We therefore select the most massive galaxy within 0.25 R₂₀₀ as the central galaxy for these clusters.

We use data from the SAMI v0.11 internal data release (reduced identically to DR2; Scott et al. 2018). This data release consists of 3071 galaxies, with repeat observations available for 277 galaxies.

Out of 3071 SAMI galaxies, 3019 have several annular measurements (up to four) of age, metallicity, and α-element abundance ratio (and their corresponding errors). These are made in elliptical apertures centered on the center of the cube, with constant spacing along the major axis. The position angle, PA, and ellipticity, , of the galaxy are determined using the find galaxy Python routine of Cappellari (2002) from the image generated by summing the cube along its wavelength axis. The spectra of each of the spaxels in each aperture are summed, and the resulting spectrum is analyzed in order to extract the stellar population measurements for each annular aperture.

Stellar population measurements were determined as described in Scott et al. (2017) using an approach based on Lick absorption line strengths. We summarize the method here and identify the differences Scott et al. (2018) took for aperture measurements compared to the total measures presented in Scott et al. (2017).

The Lick indices used for this analysis consist of indices defined by Worthey & Ottaviani (1997) and Trager et al. (1998): Balmer lines (Hδ_A, Hδ_F, Hγ_A, Hγ_F, and Hβ), iron-dominated lines (Fe4383, Fe4531, Fe5015, Fe5270, Fe5335, and Fe5406), molecular indices (CN1, CN2, Mg1, and Mg2), and Ca4227, G4300, Ca4455, C4668, and Mgb. All of these indices are in the SAMI blue wavelength range.

Before measuring the absorption line strength in the annular spectra, emission from star formation was removed by comparing the observed spectra with synthetic spectra unaffected by star formation. The best-fitting model was found using the penalized Pixel Fitting (pPXF) code of Cappellari & Emsellem (2004). This was done by fitting each spectrum with a set of 30 stellar template spectra chosen from the MILES library (Sánchez-Blázquez et al. 2006; Falcón-Barroso et al. 2011). The fit was performed for each input three times: the first run determines the noise in the spectra, the second run identifies bad pixels and emission-affected pixels, and the third fit replaces those pixels with the values of the best-fitting emission-free spectra (see Scott et al. 2017 for details). This replacement method is more robust than the simple subtraction of emission lines from the spectra, especially in cases of low continuum S/N or weak emission or in the wings of emission lines.

In order to measure the Lick indices, all annular spectra were then broadened (by convolving with a Gaussian) in order to match the Lick resolution of the relevant index, also taking into account the combined instrumental broadening, intrinsic broadening due to the galaxy's velocity dispersions, and additional broadening due to binning over many spaxels. For every index, a Monte Carlo procedure was used to estimate the uncertainties: noise was randomly added to the best-fitting spectrum, the indices were remeasured for 100 different realizations, and the standard deviation thus found was adopted as the error on each index (see Scott et al. 2017 for more details).

The indices measured were then converted into single stellar population (SSP) equivalent age, metallicity, and α-element abundance through comparison with stellar population models that predict Lick indices as a function of logarithmic age, metallicity, and [α/Fe]. For the SAMI galaxies, two different models were used: Schiavon (2007, hereafter S07) and Thomas et al. (2011, hereafter TMJ). These models were interpolated to a fine grid (with a resolution of 0.02 in [Z/H] and log age and 0.01 in [α/Fe]). A χ² minimization approach (first implemented by Proctor et al. 2004) was used to find the SSP that best reproduced the indices measured. As described in Scott et al. (2017), this also iteratively excluded indices if they were more than 1σ outside the range covered by the model. Moreover, the fit would not return a solution if the fitted indices did not include at least one Balmer index and at least one Fe index. A robust estimation of α-element abundance was not always possible due to the S/N of the spectra (S/N < 20). The absorption lines corresponding to the relevant indices were sometimes too shallow or not broad enough to be detected at a high enough S/N.

For each spectrum, the values of age, [Z/H], and [α/Fe] were taken from the minimum χ² fit, and their uncertainties were determined from a χ² distribution in the 3D space of the three parameters.

Comparing the stellar population parameters found with the two models, a good agreement is found for [Z/H] and [α/Fe] (although the latter presents a small offset between the two models). The poorest agreement is found comparing ages from the two models, with TMJ predicting older ages. The broad scatter is the result of the fact that the stellar populations are luminosity-weighted; therefore, the equivalent mean ages are more sensitive to the youngest populations. The TMJ model predicts older ages at a given Hβ because several galaxies have Balmer line indices that lie outside the TMJ model grid (Kuntschner et al. 2010); therefore, the age measurements at low metallicities found by this model are not reliable. After considering all of these factors, Scott et al. (2017) adopted the S07 model to derive SSP-equivalent ages, while the TMJ model was used to derive the [Z/H] and [α/Fe] values as a best compromise given the limitations, noting that neither model perfectly describes the full set of indices in the SAMI sample. We found that this approach produced good agreement with previous literature studies, whereas the TMJ ages showed an unphysical upturn to much older ages at low masses/metallicities. However, our work focuses on relative age differences, not absolutes. Considering that relative ages are more accurate, and that the galaxies in the sample presented here do not reach the low metallicities where age reliability is problematic, our results are robust against this issue.

A log–log linear fit is applied to the annular stellar population measurements of each galaxy using the Python package scipy.optimize.curve_fit (Virtanen et al. 2020) in order to derive the corresponding stellar population gradient. The fit takes into account the errors from the stellar population measurements and uses nonlinear least squares to fit a straight line to the data. For the age measurements, since the age uncertainties are asymmetric, we fit the gradient using the mean error on each point.

The fitted slope is taken as the gradient, and the uncertainty on the gradient is derived using a Monte Carlo procedure, where each annular value is randomly taken in their uncertainty window and the linear fit is applied again (with the errors taken into account) for 1000 different realizations. The standard deviation from the mean value of the slope is then taken as the uncertainty on the gradient derived.

We note that, as pointed out by Oyarzún et al. (2019), a nonlinear fit could be a better model when deriving the gradients, since the slope of the radial profile can vary going toward the outskirts of the galaxy. However, having only three to four radial bins for galaxies in our sample, a linear fit with logarithmic radius is the most robust solution. We also note that choosing to fit the gradients up to 1 R_e instead of to the last available radial bin does not change the conclusions we draw.

We do not take the inclination of the galaxies into account when deriving the gradients, consistent with previous studies (e.g., Greene et al. 2015; Oliva-Altamirano et al. 2015; Goddard et al. 2017; Zheng et al. 2017; Ferreras et al. 2019).

In order to consider the effects of seeing on our results, we compared the half-width at half-maximum (HWHM) of the point-spread function (PSF) with the effective radius and the first and last radial bins available, as shown in Figure 4. Some of the gradients may be strongly affected by the PSF width, since it is significant compared to the galaxy size (six galaxies out of 533 have R_e < PSF).

Zoom In Zoom Out Reset image size

To measure the effect of the PSF width on the derivation of the gradients, we first analyze the radial profiles of the 23 galaxies in our final sample that have multiple observations available. A small number of the repeat observations exhibit small-scale variations in the flux (aliasing; see Green et al. 2018 for an extended discussion) that can affect the measured indices and therefore the derived population parameters. In such cases, we identify the discrepant observation by eye, selecting the cube with the more physical spectral variations. For the remaining 35 galaxies with repeated observations, we measure the slope of the best fit for each observation, finding a mean change in metallicity gradient of Δ[Z/H] = −0.03 ± 0.05. Figure 5 shows that, within the uncertainties, the measurements are consistent for the majority of the galaxies and show no trend with seeing. However, a more detailed analysis that accounts for the PSF is required to extract robust gradients with meaningful error bars.

Zoom In Zoom Out Reset image size

Following Ferreras et al. (2019), we derive the gradients of each galaxy by fitting the observed population parameters in each radial bin to a linear function model convolved with the PSF of the observation. From this model, stellar population values are extracted in the same radial bins used for the observed values and compared with the original observations, applying a standard likelihood function. The best-fit parameters for the slope and the intercept of the linear model are then retrieved (with corresponding uncertainties) using a Markov Chain Monte Carlo sampler (EMCEE; Foreman-Mackey et al. 2013). More details on this process can be found in Ferreras et al. (2019).

The gradients derived following this method are in general steeper than those derived without taking into account the effect of the PSF width, due to the fact that the PSF tends to wash out the gradients. Figure 6 shows a radial metallicity profile of a typical galaxy with the gradient derived using both methods.

Zoom In Zoom Out Reset image size

Example stellar population maps and radial profiles are shown in Figure B1 in Appendix B for two example galaxies (one central and one satellite galaxy). Hereafter, we refer to the gradients derived after correcting for the effect of the PSF. However, considering the gradients derived without taking into account the effect of the PSF does not change the conclusions we draw.

We explore the consequences that active galactic nuclei (AGN) could have on our results in Appendix C, finding that there is no significant influence from AGN emission on our stellar population measurements.

In order to determine whether there is any dependence of the gradients on host halo mass, we split our central galaxy sample into three subsets of halo mass. Each subset has an equal number of galaxies (32 galaxies for metallicity and age gradients and 26 galaxies for [α/Fe] gradients). We then bin the satellite galaxies into the same halo mass subsets as the central galaxies.

Figure 10 shows the stellar population gradients of central and satellite galaxies as a function of stellar mass in the low halo mass (10.96 < M₂₀₀ < 13.28) and high halo mass (13.77 < M₂₀₀ < 15.29) subsets. Each subset has been grouped into bins by stellar mass (similar to the mass bins described at the beginning of Section 4) to better visualize any underlying trend with stellar mass.

Zoom In Zoom Out Reset image size

The metallicity gradients of central and satellite galaxies show no differences between those in low- and high-mass halos. In the stellar mass bins where we have values for both central and satellite galaxies, the metallicity gradients of central galaxies are consistent with those of satellite galaxies in structures of similar halo mass. The metallicity gradients of galaxies in low-mass halos (both satellite and central galaxies) become shallower with increasing stellar mass compared to the gradients of galaxies in denser environments, which do not show any trend with stellar mass. Using Kendall's correlation coefficient, we find a value of τ = 0.24 with a probability of correlation of 98.9%. This weak trend could be the driver of the weak trend with stellar mass that we see for the whole sample of galaxies.

Age gradients are generally positive for both central and satellite galaxies, with no trend with stellar mass regardless of the host halo mass.

Central and satellite galaxies show similar [α/Fe] gradients for both environments.

We observe shallow metallicity gradients in our sample (mean ${\rm{\Delta }}[{\rm{Z}}/{\rm{H}}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})$ = −0.25 ± 0.02 with a 1σ deviation of 0.24 for central galaxies and mean ${\rm{\Delta }}[{\rm{Z}}/{\rm{H}}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})$ = −0.30 ± 0.01 with a 1σ deviation of 0.24 for satellite galaxies). These are in agreement with previous results from Kuntschner et al. (2010) and Oliva-Altamirano et al. (2015).

Examining the relationship of metallicity gradients with stellar mass (Figure 7), we find a potential weak positive trend such that gradients are shallower (less negative) with increasing stellar mass. This is in agreement with Zheng et al. (2017), who found evidence of a weak positive trend of metallicity gradients with stellar mass, so that the gradients become less negative with increasing stellar mass, in a sample of 577 early-type MaNGA galaxies. Their trend is more evident when examining galaxies in low-mass halos (sheet and void environment).

However, we do not observe a transition to a shallower metallicity gradient at ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })=10.54$, as observed by Kuntschner et al. (2010) and predicted by Taylor & Kobayashi (2017) from chemodynamical simulations.

Our lack of significant correlation with stellar mass is in contrast to Li et al. (2018), who found a clear decrease of the metallicity gradients (i.e., more positive gradients) with increasing stellar mass, and Goddard et al. (2017), who found a weak negative correlation of gradients with stellar mass in a sample of MaNGA ETGs. However, we note that if we examine the gradients derived by Goddard et al. (2017) and Li et al. (2018) for galaxies in the same mass range as those in our sample ($9.5\lt {\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\lt 11.8$), they show no strong dependence with stellar mass, in agreement with our results.

Tortora & Napolitano (2012), in a sample of about 4500 SDSS ETGs, found that age and metallicity gradients generally do not depend on environmental density, in agreement with our results. However, they found a residual dependence of metallicity gradient with environment for central galaxies (shallower metallicity gradients in denser environments).

Ferreras et al. (2019) made independent stellar population measurements for 522 ETGs in the SAMI Galaxy Survey and found a marginal steepening of metallicity gradient with increasing stellar mass. They found the steepening to be stronger in the denser cluster environments, whereas the group environments appear to have shallower metallicity gradients. While we do not see a steepening in the metallicity gradients in denser environments, Ferreras et al.'s (2019) mean gradients are consistent with our derived gradients, as shown in Figure 7 (lower panel), and the relationships are consistent within their errors. The difference in the metallicity gradient–stellar mass relationship observed by Ferreras et al. (2019) could therefore be driven by the different selection of halo structures.

We find mildly positive age gradients (i.e., younger centers) with mean ${\rm{\Delta }}{\rm{log(Age/Gyr)}}/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=0.13\pm 0.03$ for central galaxies and mean ${\rm{\Delta }}{\rm{log(Age/Gyr)}}/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\,=0.17\pm 0.01$ for satellite galaxies with 1σ deviations of 0.33 and 0.29, respectively.

Evidence of flat or slightly positive gradients in age has been previously observed by numerous studies, such as Sánchez-Blázquez et al. (2007), Brough et al. (2007), Kuntschner et al. (2010), Loubser & Sánchez-Blázquez (2012), Greene et al. (2015), and Goddard et al. (2017).

We find no correlation of age gradients with stellar mass, in agreement with Zheng et al. (2017).

Ferreras et al. (2019) found that there is a slight steepening of the age gradients toward high-mass galaxies (with a slope of −0.07 ± 0.04); in particular, galaxies in groups appear to have steeper age gradients (i.e., more negative) than galaxies in clusters, as shown by the red and gray lines in Figure 8 (lower panel). When examining our sample as a function of halo mass, we do not see any trend of age gradient with stellar mass or any dependence on halo mass (in agreement with Tortora & Napolitano 2012). While our relationships are not consistent with Ferreras et al. (2019), their mean gradients are consistent with our derived gradients, as shown in Figure 8. The difference in relationship is likely driven by the different selection of low- and high-mass halos for the analysis (Ferreras et al. 2019 divided galaxies into group versus cluster galaxies).

We find flat and shallow [α/Fe] gradients with a mean value of ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=0.01\pm 0.03$ for central galaxies and${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=0.08\pm 0.01$ for satellite galaxies with standard deviations of 0.23 and 0.18, respectively. These results are in good agreement with Kuntschner et al. (2010), who found ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})$ between 0.01 and 0.07.

We find a weak negative correlation of [α/Fe] gradients with stellar mass, such that the gradients are more negative with increasing masses (Figure 9). Greene et al. (2015), with a sample of 50 MASSIVE galaxies and 50 lower-mass galaxies, found similar [α/Fe] gradients, but their correlation is opposite to that found here: galaxies in their low-mass bin ($10.1\,\gtrsim {\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\lesssim 11.2$) have ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\,=-0.09\pm 0.06$, and galaxies in their higher-mass bin $({{\rm{log}}}_{10}({M}_{\ast }/{M}_{\odot })\gt 16)$ have ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\,=0.1\pm 0.08$. However, the gradients are consistent within our 1σ deviation.

Greene et al. (2019) found that [α/Fe] gradients for very high-mass galaxies (11.6 $\lt {\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\lt$ 12.2) are consistent with zero, with a median value of ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=-0.03\pm 0.02$, in agreement with our median value for central galaxies, ${\rm{\Delta }}\mathrm{log}[\alpha /\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\,=0.01\pm 0.03$.

Ferreras et al. (2019) found evidence of [α/Fe] gradients varying from positive to more negative with increasing mass for cluster galaxies (Figure 9), in agreement with what we see for central galaxies in high-mass halos. However, we do not see more negative [α/Fe] gradients for satellite galaxies in high-mass halos.

Similarly shallow positive [α/Fe] gradients are also found in simulations. Hirschmann et al. (2015) derived mildly positive age gradients (${\rm{\Delta }}{\rm{log(Age/Gyr)}}/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\approx 0.04$). Taylor & Kobayashi (2017) found small positive [α/Fe] gradients (${\rm{\Delta }}\mathrm{log}[{\rm{O}}/\mathrm{Fe}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})\approx 0.1\mbox{--}0.2$).

In this paper, we find negative metallicity gradients, positive age gradients, and flat [α/Fe] gradients. The metallicity gradients show evidence of a potential weak trend with stellar mass, so that gradients become less negative with increasing stellar mass, while [α/Fe] gradients become more negative with increasing mass. When examining the galaxies as a function of halo mass, we find evidence of shallower (less negative) metallicity gradients with increasing stellar mass for galaxies in low-mass halos but no significant difference between the gradients of central and satellite galaxies.

The gradients we find are consistent with those predicted for galaxies formed in a two-phase process (White 1980; Kobayashi 2004; Oser et al. 2010). In this scenario, galaxies first undergo a rapid formation phase during which "in situ" stars are formed within the galaxy. The more massive ETGs start forming stars at earlier epochs (de Lucia et al. 2006; Dekel et al. 2009). The resulting galaxy is compact, with a steep negative metallicity gradient (with gradients ranging from ${\rm{\Delta }}[{\rm{Z}}/{\rm{H}}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})$ = −0.35 to −1.0; Kobayashi 2004) and a positive age gradient (due to the presence of a younger stellar population in the center of the galaxy). More specifically, the metallicity and age gradients would naturally correlate with galaxy mass, as star formation lasts longer in the center of more massive systems that have deeper central potential wells (e.g., Kobayashi 2004; Pipino et al. 2010), so more massive galaxies would have younger and more metal-rich centers, more positive age gradients, and more negative metallicity gradients.

After the star formation is quenched, a second phase of slower evolution is dominated by dissipationless mergers. In particular, galaxy interactions will either steepen or flatten the metallicity gradients; major dissipationless mergers tend to cause gradients to become shallower as metal-rich stars from the center of the galaxy are redistributed further out (Kobayashi 2004; di Matteo et al. 2009; Rupke et al. 2010; Navarro-González et al. 2013; Taylor & Kobayashi 2017), and minor dissipationless mergers lead to steeper negative metallicity gradients due to metal-poor stars being accreted in the outer regions from lower-mass galaxies (Hilz et al. 2012, 2013; Hirschmann et al. 2015).

Central galaxies, due to their privileged position in the center of the potential well, are expected to experience a greater number of galaxy interactions compared to satellite galaxies. In this scenario, simulations predict flat or slightly positive age gradients due to old stars being added in the outer regions (at radii >2 R_e; Hirschmann et al. 2015), and [α/Fe] profiles are expected to be flat as a result of massive ETGs assembling via mergers with low-mass systems (Gu et al. 2018). These predictions are in agreement with our findings; the metallicity gradients we find (see Table 1) are shallower than those predicted by a monolithic collapse (ranging from ${\rm{\Delta }}[{\rm{Z}}/{\rm{H}}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=-0.5$ (Carlberg & Freedman 1985) to ${\rm{\Delta }}[{\rm{Z}}/{\rm{H}}]/{\rm{\Delta }}\mathrm{log}(r/{R}_{e})=-1.0$ (Kobayashi 2004)), hinting at a flattening of the gradients due to mergers. The age gradients are slightly positive (with central stars having younger ages than stars in the outskirts) and generally flat [α/Fe] gradients. However, since evidence of minor later accretions is expected at radii >1.5–2 R_e, and the majority of the radial profiles for our central galaxies only extend to a maximum of 2 R_e (as shown in Figure 11), our results point to a similar evolution path for the inner regions of both central and satellite galaxies.

Zoom In Zoom Out Reset image size

Previous works have seen clear evidence that satellite galaxies in these environments are found to be older than central galaxies of the same stellar mass in higher-density environments (Thomas et al. 2005; Pasquali et al. 2010; la Barbera et al. 2014; Scott et al. 2017). However, we do not see a difference in the gradients of central and satellite galaxies when taking into account the mass of their host halos.

The fact that our metallicity gradients for central and satellite galaxies are consistent within the standard deviation suggests that the inner regions of central galaxies (up to 2 R_e) form in a similar fashion to similarly massive satellite galaxies. Moreover, the gradients for galaxies in high-mass halos show no trend with stellar mass. This seems to point to a similar evolution for galaxies in denser environments, regardless of their current position in the halo.

The potential trend of shallower metallicity gradients with increasing stellar mass in low-mass halos could point to a greater number of interactions in the group environment. Galaxy interactions in this environment are more likely due to the lower relative velocity at which galaxies are moving. In high-mass halos, the frequency of galaxy mergers is low due to the fast orbital motions of the galaxies and the long timescales for dynamical friction (e.g., Park & Hwang 2009).

The consistent stellar population gradients ([Z/H], age, and [α/Fe]) presented here for both central and satellite galaxies suggest a similar formation and evolution history for the inner regions (R < 2 R_e) of central and satellite galaxies, with no dependence on their host halo mass.