The SAMI Galaxy Survey: A Range in S0 Properties Indicating Multiple Formation Pathways

The observational data products used in this work were created by the SAMI Galaxy Survey (Bryant et al., 2015). The Sydney-AAO Multi-object Integral field spectrograph (SAMI; Croom et al., 2012) is mounted at the prime focus of the 3.9m the Anglo-Australian Telescope, which provides a 1 degree diameter field of view. SAMI uses 13 fused fibre bundles (Hexabundles; Bland-Hawthorn et al., 2011; Bryant et al., 2014) with a high (75 percent) fill factor. Each bundle contains 61 fibres of 1.6 arcsec diameter resulting in each hexabundle having a diameter of 15 arcsec. The hexabundles, as well as 26 sky fibres, are plugged into pre-drilled plates using magnetic connectors. SAMI fibres are fed to the double-beam AAOmega spectrograph (Sharp et al., 2006). For the SAMI Galaxy Survey its 570V grating was used with the blue arm (3700-5700A), giving a resolution of R=1730 (${\rm\sigma=74km/s}$), and the R1000 grating with the red arm (6250-7350A) giving a resolution of R=4500 (${\rm\sigma=29km/s}$) (van de Sande et al., 2017). At least six pointings on each galaxy are weighted and combined to produce data cubes with a pixel scale of $0.5\times 0.5$ arcseconds (Allen et al., 2015; Sharp et al., 2015). Here we use data products released in Data Release 2 (Scott et al., 2018).

The SAMI survey was designed to cover a wide range of environments to allow for studies of galaxy properties as a function of local environment. To achieve this, the survey targeted two regions. The first coincides with the Galaxy and Mass Assembly (GAMA) survey (Driver et al., 2011), which covers 286 square degrees and has a limiting $r-$band magnitude of 19.8. Since GAMA includes limited coverage of high density cluster regions, SAMI targeted additional galaxies within eight large clusters in order to achieve a more complete coverage of environments (Owers et al., 2017). In both regions, galaxies were selected using the same volume-limited mass limits. Table 1 displays the number of S0s in each region of the SAMI survey (see below for details on classification).

The high completeness of the GAMA survey enabled a complete group catalogue to be constructed for galaxies in this region (Robotham et al., 2011). We used this catalogue to identify the host environment of each SAMI galaxy, depending on the mass of their host group halo. We define host group masses in the range $10^{11}<{\rm M}_{\odot}<10^{13}$ as low-mass groups and those with ${\rm M}_{\odot}>10^{13}$ as high-mass groups. Galaxies not associated with a group are identified as field galaxies, and those located in the eight cluster regions outside of the GAMA field are identified as cluster galaxies.

High resolution images of all SAMI galaxies within the GAMA region were taken using the Hyper-Suprime Cam on the Subaru Telescope in Hawaii (Aihara et al., 2018). These images were taken in five colour bands - $g,i,r,Y$ and $z$. The depth of the imaging extends to $i$-band magnitudes of 28, and the images have a resolution of 0.168 arcseconds per pixel. The high depth and resolution of these images allows for the identification of both prominent and very faint structural features such as disks, shells and tidal arms within and around the galaxy, providing clues to their evolutionary history.

For galaxies in the GAMA region, the Sérsic index, ellipticity and effective radius are derived from a single component fit of SDSS imagery using the code GALFIT3 (Peng, et al., 2010; Driver et al., 2011). In the cluster regions, these physical parameters are derived from r-band images from the SDSS and VLT Survey telescope/ATLAS surveys using the code PROFIT (Robotham, et al., 2017; Owers, et al., 2019). The stellar masses in both regions were derived from mass/light ratios based on $g-i$ colours (Bryant et al., 2015).

Galaxies in the SAMI Survey were visually classified by a subset of survey team members people using Sloan Digital Sky Survey imaging (Cortese et al., 2016). Firstly, the galaxy was classed as an early or late type galaxy based on the absence or presence of a clear disk, spiral arm features and areas of star formation. In the second step, galaxies in the early-type category were classed as lenticular (S0) if a disk component was visible or elliptical (E) otherwise. Late-type galaxies were classed as either early spirals (Sa, bulge present) or late type spirals (Sp, no bulge visible). Classes were assigned only if 66 percent of the classifiers agreed on the type. Intermediate classes (e.g. E/S0) were assigned to cases where 66 percent agreed it was one or the other. For all other cases, the galaxy was classed as unknown. In our analysis, the Sa, Sa/Sp and Sp classes are grouped into a single category labelled spirals and E/S0, S0 and S0/Sa are grouped together as S0s.

In order to allow for direct comparisons of S0 kinematics with the Diaz et al. (2018) simulation, we use the SAMI kinematic maps to derive radial profiles of the inclination-corrected velocity, the velocity dispersion and their ratio for the stellar and gas components.

The method used to derive stellar kinematic maps for each galaxy is detailed in van de Sande et al. (2017). Briefly, both the red and blue arms are used for fitting the stellar kinematics, with the red arm convolved to match the spectral resolution in the blue arm. Elliptical annuli are used to derive the optimal stellar template for each region of the galaxy in order to mitigate strong radial gradients. The penalised pixel fitting code (pPXF Cappellari & Emsellem, 2004) is then used on each spaxel, producing two-moment and four-moment data products. Here we use the two-moment velocity and velocity dispersion, derived from Gaussian line of sight velocity dispersion fits to each spaxel. The fitting tool LZIFU (Ho et al., 2016) is used to derive the emission line physics for each spaxel (see Green et al., 2018 for full details). LZIFU simultaneously fits all the strong emission lines with both a single-component Gaussian fit and a multi-component fit; here we use the single component fit to investigate the emission line and gas kinematic profiles. The single component fit produces measurements of the emission line velocity and velocity dispersion, as well as the flux for the strong emission lines.

For the gas and stellar velocity maps, spaxels with uncertainties greater than $\pm 15$ km/s were removed from the analysis. For the velocity dispersion maps, following van de Sande et al. (2017) we removed spaxels where the uncertainty is greater than $10\%$ of the dispersion value plus 25 km/s.

To assess the degree of rotational support in the SAMI S0s, we employed the widely-used parameter $v/\sigma$, where v is the observed line-of-sight velocity and $\sigma$ is the velocity dispersion. Following recent work with IFU surveys (e.g. van de Sande et al., 2018; Cappellari et al., 2007) we calculate $v/\sigma$ using the following equation:

where the sum is over all spaxels within a de-projected radius of 1.5 ${\rm R_{e}}$ and $F_{i}$ is the flux of spaxel $i$. Values of $v/\sigma$ approaching 1 indicate a rotationally supported structure, while values significantly lower than 1 indicate a pressure-supported structure. The de-projected radius of each spaxel corresponds to the major axis of the ellipse on which it is located, which was determined using the $r$-band-derived axis ratio and position angle (Cortese et al., 2016).

The position angles of the gas and stellar components were found by rotating a 3-pixel-wide slit through 360 degrees and adopting the angle which resulted in the maximum velocity gradient.

Following (Cortese et al., 2016), to recover true radial velocity profiles from the line-of-sight velocity projections, we estimated the galaxy’s inclination from the measured axial ratio:

where $i$ is the inclination, $q_{0}$ is the assumed intrinsic ellipticity, and $a$ and $b$ are the major and minor axis respectively. We used a commonly-assumed intrinsic ellipticity for a disk galaxy of 0.2 (Tully & Pierce, 2000; Cortese et al., 2016). We then applied this inclination to the observed line-of-sight velocity $V_{k\rm{los}}$ of each spaxel $k$ as follows:

where $\theta$ is the azimuthal angle of the spaxel $k$ in the galaxy coordinate frame. We determined the direction of rotation of the stellar and gas components by rotating a slice through the 2D map and determining the angle which produced the greatest velocity gradient along that slice. We removed all but the spaxels within 1.5" of the line along the strongest gradient and binned the remaining spaxels by their distance from the galaxy centre. From this, we constructed the radial profile of the velocity, velocity dispersions and $v/\sigma$ using the median values in each bin.

We use the template-fitting code STARLIGHT (Cid Fernandes, et al., 2005) to determine stellar age profiles for SAMI galaxies. We binned pixels into radial elliptical annuli, with each annulus covering a radial extent of 2 pixels. For each bin, we determined the median flux at each wavelength of all spaxels. We used stellar population templates from the BC03 model (Bruzual & Charlot, 2003) for the fitting process. During the fitting process, STARLIGHT determines the mixture of stellar population templates which produces the best match to the observed spectrum. From this template mixture, we determined the final age of each radial bin (along with its metallicity) by finding the weighted average age of templates giving the best match. We estimated uncertainties in the derived stellar ages by using a Monte Carlo method, whereby random noise was added to the median spectrum in proportion to the standard error of the flux at each wavelength in the bin.

We first looked at the distribution of S0s across the different environments covered by SAMI. The fraction of all galaxies in the SAMI survey classified as S0s is shown in Figure 2 as a function of environment. Here the full galaxy sample is used (’All’ in Table 1). As the mass of the host group increases, there is a strong increase in the fraction of S0 galaxies, consistent with previous findings (e.g. Dressler, 1980). S0 galaxies account for around half of all galaxies located in the cluster regions, while within the field and low mass group environments, S0 galaxies still make up a non-negligible fraction of 0.16 and 0.24 respectively. This also demonstrates that our SAMI sample contains a significant fraction of S0s across all environments including the field and low mass groups.

We next looked to characterise the range of physical properties of S0s within the SAMI sample. A large range, or any bimodality present in the distributions, may indicate a wide range of conditions in the formation of S0s, or indeed the presence of multiple formation pathways. We focus here on the Sérsic index, effective radii and stellar $v/\sigma$, since these properties together physically characterise the structure of the S0s and are affected by both the initial transformation and subsequent evolution. Figure 3 presents the distribution of the final S0 sample (including both the GAMA and cluster regions) in the Sérsic vs ellipticity, Sérsic vs galaxy mass and effective radius vs stellar mass planes, with the points coloured by their stellar $v/\sigma$ values. The distributions of S0s over each of the parameters, and across each of the parameter spaces shown, appears to be continuous with no clear separation into bimodal populations.

It is however evident that there is a wide range in rotational support within the S0 population. In particular, the range in $v/\sigma$ over the different parameter spaces forms a strong gradient in each case. S0s with low rotational support ($v/\sigma$ below 0.5) generally feature lower ellipticities, higher Sérsic indices, slightly higher stellar masses and smaller effective radii. These features indicate a more compact system with a higher degree of random stellar motions. The more rotationally-supported S0s ($v/\sigma$ above 0.5), meanwhile, tend to feature larger ellipticities, lower Sérsic indices, and larger effective radii. These taken together indicate that they feature a more prominent rotationally supported disc component. While the correlation between $v/\sigma$ and ellipticity will be partly due to projection affects, the additional correlations between other structural parameters indicate that there is also an intrinsic shift in the structure of the S0s from high to low measured $v/\sigma$ values. The lower ellipticites in the pressure-supported population may therefore also indicate an overall shift to more spherical structures, supported by the larger Sérsic index and higher degree of random motion.

Figure 4 shows the average stellar $v/\sigma$ values for S0s in the field, low mass (up to $10^{13}M_{\odot}$, high mass (over $10^{13}M_{\odot}$ and cluster regions. The observed trend in the groups is marginal relative to the uncertainty range, with S0s in the field featuring the lowest average stellar $v/\sigma$, increasing towards the high-mass groups, and then falling back in the cluster region. This reversal in the clusters could be due to either an increase in harassment experienced by these galaxies or the merging of group and field S0s into the clusters; this is discussed in Section 5.5. Alternatively, this may be a bias resulting from different sample selections between the GAMA and Cluster regions.

Hyper Suprime Cam (HSC) images of S0s with stellar $v/\sigma$ above and below 0.5 are shown in Figure 5. These high-resolution images are currently only available for the GAMA region. The galaxies shown here were blindly selected from a list of randomised catalogue IDs. More rotationally supported S0s generally feature smooth, well-defined disks with hints of remnant spiral structure in some cases, suggesting a passive evolutionary history. Many S0s in the pressure-supported regime (for example 323854 and 508315) feature signs of interactions and mergers, including shells, tidal features and structures with orientations different to that of the rest of the galaxy.

Figure 6 a) shows the distribution of our S0 sample on the widely-used stellar $v/\sigma$ vs ellipticity plot for the GAMA and cluster region. For comparison, Figure 6 b) shows the same distribution for ellipticals in the GAMA region which meet the same kinematic selection criteria, with the slow rotators highlighted in orange. This illustrates that the distribution in kinematics seen among the S0s here encompasses a wider range than that covered by the slow and fast-rotating ellipticals. The cluster S0s appear to be shifted towards lower ellipticities relative to the GAMA region. As a check for whether this is caused by a systematic bias in ellipticity measurements between the two regions due to different imagery (SDSS images for the GAMA region and VST images for the cluster regions, see Section 2.3), we also looked at the distribution of ellipticity for spiral galaxies and found no such shift. In addition, where an overlap between SDSS and VST was present, parameters derived from the two sets of imagery were checked for consistency (Owers et al., 2017). This suggests that S0s in the cluster region have intrinsically lower ellipticities, which could be due to the increased harassment these galaxies experience in the denser environment.

As a comparison, Figure 6 b) also illustrates the stellar $v/\sigma$ vs ellipticity distribution of the elliptical population within GAMA. This demonstrates that the well-known separation of centrally slow and fast rotator ellipticals occurs outside the large range of rotational support seen amongst the S0s (Cappellari, 2016).

To gain further insight into the structure and history of the S0 sample, we next investigated the distribution of gas kinematics within the S0s and determine how these relate to the stellar component. Not all S0s in our sample have line emission strong enough to derive accurate gas kinematics (see Table 1), as would be expected given the typically low gas contents of this class of galaxy. This is particularly true for the cluster regions, where both the larger redshifts and the generally lower gas contents of galaxies in cluster environments would result in more S0s falling below the emission detection threshold. The distribution of the $H\alpha$ flux weighted by the stellar flux for both the GAMA and cluster region is shown in Figure 7. The distribution for S0s in the GAMA region (i.e. those in field and group environments) is broader with a tail towards higher $H\alpha$ flux, while the majority of cluster S0s contain very little detected gas emission. Because of this, we cannot derive accurate gas kinematics for many S0s in the cluster region.

For S0 galaxies meeting the additional selection criteria for inclusion in the analysis of gas kinematics, Figure 8 shows the distributions of the stellar and gas $v/\sigma$, as well as the difference in position angle between the stellar and gas components, again coloured by their stellar kinematics. S0s with unusual PAs were visually investigated to confirm that these differences are not due to poor kinematic maps; three examples are shown in Figure 8 b). Most evident here is that the S0s with highly rotationally-supported star kinematics also feature rotationally-supported gas kinematics, as well as very small differences in the position angles of the two components. This indicates that the stellar and gas components in these galaxies are co-rotating in a common structure. S0s with $v/\sigma$ values below 0.5, however, feature a large range of misalignments between the stellar and gas components. The six S0s with PAs between 50 and 150 degrees have lower $v/\sigma$ values relative to those near 0 or 180 degrees; such configurations are likely more stable in systems where the stellar component has a lower degree of angular momentum. Misalignments of S0 star and gas components in SAMI have already been noted (Bryant, et al., 2019); here we are demonstrating that this misalignment only occurs in S0s with low rotational support.

Figure 9 shows typical examples of kinematic profiles for S0s in the pressure-supported S0s (top row) and rotationally-supported S0s (bottom row). In the top row, the gas components are rotating significantly faster than the stellar components, while in the S0s from the rotationally-supported group, the rotational velocities of the gas and stellar components are similar. This further suggests that the star and gas components in the rotationally-supported S0s have evolved together, while the gas component in the pressure-supported group may have a separate origin to the stellar component. However, we note that S0s with low stellar $v/\sigma$ may also be more likely to feature mis-aligned gas kinematics simply due to a larger bulge to disk ratio.

Finally, we examine whether there is any difference in the stellar age profiles of S0s at the extreme ends of the stellar $v/\sigma$ distribution. Figure 10 shows the age profiles of pressure-supported and rotationally-supported S0s, along with the average profile for each. For this comparison, we use the limits $v/\sigma$ > 0.75 for the rotationally-supported group (blue) and $v/\sigma$ < 0.25 for the pressure supported group (red), resulting in 19 and 20 S0s respectively.

The average age profiles have central ages consistent with each other within the uncertainties. The rotationally-dominated S0s do show signs of a rapid drop in age towards the outer regions, suggesting that the more prominent disk structure in these galaxies consists of younger stars. However this difference relative to the pressure-supported S0s is not significant within this sample.

A large amount of work has recently been done on the separation of early type galaxies (ellipticals and S0s) into slow and fast rotators. Slow rotators are defined to be those galaxies which fall within the region below $\lambda_{\rm R_{e}}=0.08+e/4$ (Cappellari, 2016). Recent work has suggested that slow and fast rotators may have different evolutionary pathways, with the slow rotators resulting from merger activity and the fast rotators following a more passive evolution (Penoyre, et al., 2017). However, the majority of early-type galaxies which fall within the slow-rotator class are ellipticals; for example in the work of Emsellem et al. (2007), while 10 out of 25 ellipticals are classed as slow rotators, only two out of 22 S0s are classed as slow rotators while the remaining were classed as fast rotators. Here we found the same behaviour for our S0 sample; only 12 out of 219 S0s lie within the slow-rotator region. This highlights that the spread in kinematics of S0s we find here lies in a different region of the ellipticity vs $\lambda_{\rm R_{e}}$ parameter space. Therefore we conclude that the spread in $v/\sigma$ we observe is not caused by subpopulations corresponding to the slow and fast rotators, but is caused by a range of processes intrinsic within the S0 class. However, it should be noted that kinematics at larger radii considered here can differ from the kinematics derived from smaller radii, as demonstrated in Foster, et al. (2016).

The more pressure-supported population of S0s (those with $v/\sigma$ less than 0.5) feature a pressure-supported stellar component, high Sérsic indices, lower ellipticites, and are more prominent in lower density environments. Many of these S0s show signs of disruption in their optical imagery, such as shells and tidal streams. Their gas content is lower than that of the S0s with a higher degree of rotational support, and the position angle of the gas rotation relative to the stellar component covers a large range from aligned to anti-aligned, suggesting that the gas component has a different origin to the stellar component. This is in contrast to spiral galaxies, for which the vast majority feature aligned gas and stellar components (Bryant, et al., 2019).

S0s forming via minor mergers or tidal interactions are expected to feature a higher degree of pressure support (Rizzo, Fraternali & Iorio, 2018), consistent with what is seen in these S0s. Comparing the pressure-supported S0s with spiral galaxies in the SAMI survey (Figure 12) shows that the pressure-supported S0s feature higher Sérsic indices and higher stellar masses than the SAMI spiral population. If they descended directly from spirals, their progenitors would have been a subpopulation of spirals with high masses and a dominating bulge. However, their $v/\sigma$ values are lower than spirals of similar mass, making them difficult to explain with a passive faded-spiral pathway.

All these features could indicate that these S0s have formed via a disruptive process such as a minor merger or tidal interaction, rather than via the fading spiral pathway. However, it cannot be ruled out that these are older systems which have experienced a greater amount of harassment, ram pressure stripping or significant disruption since their initial formation.

Alternatively, as seen in Cortese, et al. (2019), passive galaxies may also feature lower v/$\sigma$ values due to a slow-down in angular momentum build-up once they enter a denser environment. S0s with lower v/$\sigma$ values seen in the larger group environments may therefore also reflect a passive evolution rather than a merger event. Comparisons with cosmological simulations may help to determine whether an altered passive evolution or a more disruptive merger event is resulting in S0s with a greater degree of pressure support in these environments.

The comparison of our observational results to the fiducial model in Diaz et al. (2018) are consistent with the claim that at least some S0s may have resulted from the cE+disk merger pathway. The Diaz et al. (2018) simulations predicted that the cE+disk merger would lead to an S0 with low stellar $v/\sigma$, a faster-rotating gas component and a negative age gradient. The final S0 in the fiducial model has a mass of around $10^{10.5}{\rm M}_{\odot}$.

Its degree of rotational support lies among the pressure-supported S0s; the position of the model S0 on the ellipticity vs $v/\sigma$ diagram is shown as the black diamond in Figure 6. Figure 3 also displays the locations of the fiducial model within the Sérsic vs ellipticity, Sérsic vs galaxy mass and effective radius vs stellar mass planes relative to the observed S0 distributions. An example of an S0 galaxy lying near this point in ellipticity vs $v/\sigma$ space is shown in Figure 11. Consistent with the fiducial model, the stellar $v/\sigma$ remains below 0.5 out to 1.5 effective radii, while the gas component rotates significantly faster.

The kinematic predictions presented in Diaz et al. (2018) corresponded to only a single scenario. It is certainly possible that a range of masses and properties of the cE and the disk galaxy, along with their relative motions, would lead to S0s with a range of masses and kinematic properties. For example, a more massive disk with a faster relative velocity may lead to a more rotationally supported S0 with a relatively larger disk component and higher $v/\sigma$ values. Therefore, the fraction of the S0 population consistent with the model could be significantly higher than what can be determined at this stage.

The more rotationally-supported S0 galaxies are located more frequently in higher density groups. The majority of these S0s feature smooth, well-defined disks, some of which show hints of remnant spiral features. The gas in these S0s is also co-rotating with the stellar component in the majority of cases, which suggests that the gas has co-evolved with the stellar component.

A spiral galaxy which fades to an S0 would be expected to maintain a constant mass, since no significant amount of new mass enters the galaxy through a merger (Rizzo, Fraternali & Iorio, 2018). In addition, the lack of any significant disruptions would leave the concentration of the stellar component similar to the original spiral, and the galaxy would be expected to maintain its degree of rotational support. We would therefore expect a faded spiral to have similar mass, $v/\sigma$ and Sérsic indices to spiral galaxies. Comparing the rotationally-supported S0s to the mass-matched sample of spiral galaxies (Figure 12) shows that in both the GAMA and cluster regions, these S0s feature similar concentrations and kinematics to the spiral population, which lends support to the idea that these S0s form via the faded spiral pathway. The shift to higher Sérsic and lower $v/\sigma$ values from the GAMA to the cluster regions evident in Figure 12 is likely a reflection of the morphology-density relation, where the higher-density cluster environment contains a higher proportion of bulge-dominated systems.

The mass distribution of the full SAMI spiral population ranges from $10^{8}$ to $10^{11}M_{\odot}$; the lack of rotationally-supported S0s with lower masses may indicate that only high-mass spirals undergo this process. In addition, barring any major subsequent inflow of gas, S0s originating from faded spirals would retain the co-rotation of the stellar and gas component seen in spiral galaxies, which is what we also observe in the rotationally-supported S0 population.

Due to the higher relative velocities between galaxies in higher mass groups and clusters, the occurrence of mergers and slow gravitational interactions is expected to be less than that in lower-density environments (Barnes & Hernquist, 1992). Meanwhile, the increased density of the intra-cluster medium in higher-mass systems may lead to the increased occurrence of fading pathways such as ram pressure stripping. The increase in $v/\sigma$ observed in higher mass groups (see Figure 4) could therefore reflect a decreasing occurrence of the merger formation pathway and an increase in the prevalence of the fading-spiral transformation pathway.

Interestingly, the decrease in the average $v/\sigma$ seen in the cluster regions goes against the trend seen in the groups (Figure 4). This decrease could be the result of S0s which have formed within small groups merging into the clusters. Indeed, Just et al. (2010) showed that S0 formation is occurring more prominently in low-mass groups, and that the relative numbers of S0s seen in clusters can be explained by the subsequent merging of these low-mass groups into the clusters. Alternatively, additional harassment in the cluster environment, such as multiple flybys of other cluster members, may lead to further heating of the disk. Future work will look at the morphology-density relation of S0s within clusters, in particular their stellar $v/\sigma$ as a function of radius.