A STRONG DICHOTOMY IN S0 DISK PROFILES BETWEEN THE VIRGO CLUSTER AND THE FIELD

One of the most interesting unresolved questions in the study of galaxy morphology is the origin of S0 galaxies. These galaxies are common in the denser parts of clusters (Dressler 1980), which suggests a cluster-based formation mechanism. The discovery that clusters evolve from being spiral-dominated at z ≳ 0.5 to being S0-dominated at lower redshifts (e.g., Dressler et al. 1997; Postman et al. 2005; Poggianti et al. 2008) is evidence that clusters may in fact be turning spiral galaxies into S0s. Indeed, transformation mechanisms such as ram-pressure stripping (Gunn & Gott 1972) require high orbital velocities and a relatively dense, hot intergalactic medium (IGM), as found in clusters.

However, S0 galaxies are also found in low-density environments such as small groups and even the local field (e.g., van den Bergh 2009; Wilman & Erwin 2011; Calvi et al. 2011), which suggests one of two things: either cluster-specific mechanisms such as ram-pressure stripping cannot create S0s, or there are actually multiple channels for forming S0s, some operating in clusters and others in lower-density regions. Moran et al. (2007) found evidence suggesting that both ram-pressure stripping and strangulation/starvation (Larson et al. 1980) may be operating in massive clusters, while Wilman & Erwin (2011) argue that a significant fraction of local field and group S0s probably became S0s as central galaxies within their dark-matter halos, which implies mechanisms other than interactions with the IGM (e.g., minor mergers; see Bekki 1998; Bournaud et al. 2005). If the multiple-channel scenario is true, then there may be differences in S0 galaxies as a function of environment, differences which could be used to test S0 formation models.

One galaxy feature that might show environmental influences quite clearly is the structure of the outer disk, which should be more vulnerable to effects such as gas stripping, tidal perturbations, and late-time accretion. Recent works (Erwin et al. 2005, 2008; Pohlen & Trujillo 2006; Gutiérrez et al. 2011) have shown that the surface-brightness profiles of outer disks—traditionally supposed to be purely exponential, or else exponential with sharp outer truncations—fall into three broad classes: single-exponential (Freeman [1970] Type I), truncated (Freeman Type II), or antitruncated (Type III; Erwin et al. 2005); see Figure 1 for examples. All three profiles have been observed in spiral galaxies out to z ∼ 1 (Pérez 2004; Trujillo & Pohlen 2005; Azzollini et al. 2008; Bakos et al. 2011). The analysis of Gutiérrez et al. (2011) indicated that local S0 disks are evenly divided between the three main classes. That study, however, was based on a sample of galaxies including both the Virgo Cluster and the local field. Clearly, it would be of interest to see whether there are differences in the outer disks of S0 galaxies as a function of environment. This Letter explores that possibility by comparing the outer-disk profiles of S0s in the Virgo Cluster with those of local field S0s.

Zoom In Zoom Out Reset image size

2.1. Virgo and Field Samples

Details of the process for extracting azimuthally averaged surface-brightness profiles of outer disks are presented in Erwin et al. (2008). Here, we briefly summarize the basic approach and discuss new automated techniques used to analyze SDSS images of the 28 galaxies not previously classified by Erwin et al. (2008) or Gutiérrez et al. (2011). Images for each galaxy were retrieved from the SDSS archive; if the galaxy was near the top or bottom of the image, the adjacent field was retrieved and the fields merged to form a larger image (adjacent fields in the vertical direction are from a single drift-scan observing run).

Careful sky subtraction is essential, as is careful masking of bright stars and other galaxies near the target galaxy. We started by running SExtractor (Bertin & Arnouts 1996) on the r-band images. The resulting catalog was processed into an SAOimage DS9 region file (converting stars to circles and galaxies to ellipses, with sizes being multiples of the SExtractor isophotal area).⁸ These regions were then displayed on the image and edited (adding additional masking if necessary and removing regions corresponding to the target galaxy). A mask image was then generated from the region file, for use in conjunction with the ellipse fitting (below). The mask file was also used for sky subtraction: median pixel values were determined for 100–150 10 × 10 pixel regions, with locations chosen randomly such that they fell outside masked regions and were not within 2.5 R₂₅ of the galaxy center (R₂₅ = half of the D₂₅ diameter). The final sky value was the mean of these median measurements, with uncertainty σ_sky computed by bootstrap resampling. We used σ_sky to compute a limiting surface brightness μ_crit = 4.94σ_sky; this is the level at which a 1σ_sky error in the sky subtraction would produce a 0.2 mag arcsec⁻² shift in the surface-brightness profile (see Pohlen & Trujillo 2006; Erwin et al. 2008). Median-smoothed images were inspected to identify any residual large-scale gradients (typically vertical, due to changes in sky brightness during the SDSS drift scan); if present, these were removed with the iraf imsurfit task.

After sky subtraction, the galaxy disk orientation was determined by fitting ellipses to the r-band isophotes, using the iraf task ellipse. Ellipses corresponding to the outer disk (i.e., outside any bars and prominent rings) were used to determine the overall galaxy orientation, under the assumption that the outer disk is approximately circular. We then re-ran ellipse with ellipse shape and orientation fixed to that of the outer disk. This generates surface-brightness profiles corresponding to circular averaging if the galaxy were face-on. (This approximation fails at small radii if there is a prominent, rounder bulge, but this does not affect the outer profile.) A key advantage of this approach is that it allows measuring the profile to much fainter levels than is possible when the program must simultaneously fit variable ellipses to the isophotes; typically, we can trace the profile to at least twice R₂₅.

Finally, the surface-brightness profiles were classified into the three basic types (Figure 1), with additional subtyping as described in Pohlen & Trujillo (2006) and Erwin et al. (2008). Classifications and parameters for galaxies not in Erwin et al. (2008) or Gutiérrez et al. (2011) are presented in Table 1.

Since the influence of environmental effects is expected to be stronger in the outer disk, it makes sense to concentrate on differences in profiles at larger radii. Erwin et al. (2008) and Pohlen & Trujillo (2006) noted that a small fraction of Type II profiles had breaks at or interior to the bar radius ("Type II.i" profiles). It seems likely that these profiles are not produced by the same mechanisms that produce truncations at much larger radii; in fact, they appear to be produced spontaneously in isolated disks due to bar formation (P. Erwin et al. 2012, in preparation). Thus, in our statistical analyses we do not consider II.i profiles to be true "truncations," and group them with the Type I or III profiles (depending on the profile shape outside the bar). Because the distinction between II.i and II.o profiles is only possible for barred galaxies, we also look at the distribution of profile types for barred galaxies separately. For the S0 galaxies not previously studied in Erwin et al. (2008) or Gutiérrez et al. (2011), we carefully analyzed red (and, if available, near-IR) images for the presence of bars and measured any such bars that were found; these measurements will be presented elsewhere. The absolute numbers of Type II.i profiles are quite small: two in the Virgo Cluster S0s and none in the field.

The distribution of profile types is summarized in Figure 2, both for the complete sample (left panel) and for just the barred galaxies (right panel). Two features are immediately apparent. First, there is a complete absence of truncations in Virgo Cluster S0s, in contrast to the relatively high frequency of truncations in field S0s: 28.0^+6.7_{− 5.9}% of all field S0s (36.7^+9.1_{− 8.2}% of barred field S0s). For the DR7-only subsample (see Section 2.1), the frequencies in the field are 29.5^+7.3_{− 6.4}% for all field S0s and 38.5^+9.8_{− 8.9}% for the barred S0s. Second, Type I profiles are significantly more common in the Virgo Cluster (45.8^+10.1_{− 9.5}%) than in the field (26.0^+6.6_{− 5.7}%, or 23.3^+8.5_{− 6.8}% for just the barred galaxies; 22.7^+6.9_{− 5.7}% and 19.2^+8.8_{− 6.5}%, respectively, for the DR7-only subsample). Type III profiles, on the other hand, are equally common in both environments.

Zoom In Zoom Out Reset image size

The difference in profile frequencies seems strong; is it statistically significant? Rather than focus on, e.g., just the difference in truncation frequency (which amounts to selecting out the strongest deviation after the fact), we ask whether the distribution of all three profile types is similar for Virgo and field galaxies. For this, we can use the extended version of Fisher's exact test. For the complete Virgo and Field samples, the 3 × 2 Fisher's exact test⁹ gives P = 0.00623 for the null hypothesis that both field and Virgo distributions come from the same parent population, and P = 0.00366 when only the barred galaxies are considered. Restricted to the DR7-only subsample, the significance becomes even stronger: P = 0.00318 for all S0s and P = 0.00264 for barred S0s. (When the smaller number of unbarred galaxies is considered, the differences are not significant.) So, the difference in outer-disk profiles between the Virgo Cluster and the local field is both large and statistically significant.

Could we be missing truncations at larger radii in the Virgo S0s and thus mis-classifying them as Type I? This seems quite unlikely. Figure 3 shows that break radii for field S0s occur at R ∼ 2–12 kpc, with surface brightnesses at the break of μ_{brk, R} < 24 mag arcsec⁻²; the mean μ_{brk, R} is 22.7. Also plotted are limits on any possible breaks for Type I profiles. The latter are all >26 mag arcsec⁻² much fainter than any observed breaks.

Zoom In Zoom Out Reset image size

We can summarize the difference between outer-disk profiles in the Virgo Cluster and the field thus: Type III profiles are (roughly) equally common in Virgo and in the field; in Virgo, the remaining profiles are all Type I, while in the field they are half Type I and half Type II. (We note that Maltby et al. 2011 found no significant environmental differences for spiral disk-profile frequencies in and around a multiple-cluster system at z ∼ 0.17; however, their classification scheme used a narrow surface-brightness range and is thus rather different from ours.¹⁰) If we focus on the difference between Type I and II profiles and assume that it is indeed directly due to environmental effects, we can posit two possibilities: either something in the cluster (or proto-cluster) environment transforms Type II profiles into Type I, or else something prevents a Type I to Type II transition which is common in the field.

The most popular models for disk truncations combine two elements. The first is a radial dropoff in efficient star formation, either because the gas density falls below some critical star formation threshold (Kennicutt 1989; Schaye 2004; Elmegreen & Hunter 2006) or because the gas density profile itself has a sharp break, possibly due to accretion-induced warping (Roškar et al. 2008a; Sánchez-Blázquez et al. 2009; Martínez-Serrano et al. 2009). The second element is the outward scattering of stars from the inner disk, such as that produced by transient spiral arms (Sellwood & Binney 2002; Roškar et al. 2008a, 2008b; Sánchez-Blázquez et al. 2009; Martínez-Serrano et al. 2009). The main problem is that these models tend to approach disk truncation as a universal phenomenon and do not, in general, explain how and why truncations might not occur.

If we assume that a significant fraction of present-day cluster S0s originally had truncations, then we can look for something that could erase truncations. One possibility is "harassment" (e.g., Moore et al. 1996, 1999), where a galaxy's motion through the (evolving) cluster potential leads to repeated tidal shocks. The "high-surface-brightness" model galaxy in Moore et al. (1999, their Figure 7) had its initially single-exponential profile transformed into a mild antitruncation by this process; this suggests that a Type II profile could "flatten out" into a Type I profile (but see Gnedin 2003). More detailed simulations involving disks with initially truncated profiles are needed to see if this is a viable mechanism. (Of course, this cannot be the only way to form Type I profiles, since they are also found in the field.)

Alternatively, we can consider cluster-based mechanisms that might prevent Type II profiles from forming in the first place. If, as suggested by Erwin et al. (2008) and P. Erwin et al. (2012, in preparation), truncations in S0s and early-type spirals are predominantly related to the same sort of bar–outer Lindblad resonance (OLR) interactions that produce outer rings, then one possible scenario might be the following. The long dynamical times in outer disks suggest—and simulations agree—that large-scale changes there such as outer-ring formation require several Gyr (see references in Buta & Combes 1996). Since bar–OLR interactions are strengthened by the presence of significant gas in the outer disk (gas, being dynamically cooler than the stellar disk, more readily absorbs angular momentum from the bar), the removal of gas should weaken OLR effects. One possible signature of this might be a tendency of gas-deficient spirals to be less likely to have Type II profiles and outer rings.¹¹ If S0s in the field were able to retain gas in their outer disk for longer periods, they would be more likely to show the effects of bar–OLR interactions and develop Type II profiles. S0 galaxies in Virgo, on the other hand, could have lost their gas—particularly in the outer disk—earlier due to, e.g., a combination of ram-pressure stripping and strangulation.

We thank Dave Wilman for helpful comments, Michael Pohlen for the initial version of the catalog-parsing code, and the referee for suggesting the possible relevance of H i depletion. P.E. was supported by DFG Priority Program 1177.