STAR-FORMING GALAXY EVOLUTION IN NEARBY RICH CLUSTERS

Cluster A2029 was originally observed at 24 μm with the Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004) in 2004 February with a total integration time of 80 s pixel⁻¹. Additional observations in 2005 January to look for faint intracluster emission extended the exposure time to ∼340 s pixel⁻¹ (Bai et al. 2007). While we have 24 μm observations over a larger area, for this study, we focused on the region with the deepest coverage, which covers a smaller area: ∼30' × ∼50'. The images were processed using the MIPS Data Analysis Tool (ver. 2.9; Gordon et al. 2005) as described by Bai et al. (2007).

Sources were initially selected using the IRAF routine DAOPHOT (Stetson 1987) and fluxes from point-spread function (PSF) fitting. Errors were estimated for each source by averaging the sigma-clipped flux of the nearest 30 randomly placed apertures. Sources detected at 3σ or above were considered real sources, while anything below 3σ significance was rejected. All possible sources were visually inspected so as not to include false detections such as high-signal image artifacts.

The DAOPHOT routine occasionally missed faint sources in the field, so we also used SExtractor to identify and extract IR sources. This routine first subtracts the background and filters with a Gaussian function we set to 4'' FWHM. The program also used a deblending routine to separate individual sources and remove any artifacts that may masquerade as sources. We used a 1σ detection threshold to identify and extract all possible sources, though only those above 3σ were retained as true detections. The SExtractor routine was able to identify more low-flux sources than DAOPHOT, though after filtering out active galactic nuclei (AGNs) and other sources (discussed later), few new galaxies were added. Given the better accuracy of this method, we use the SExtractor-selected galaxies in the rest of our analysis. The differences between these two methods are minor, however, and our overall results are independent of the photometry approach.

As a result, the 3σ detection limit was ∼135 μJy, or L₂₄ ∼ 2.6 × 10⁴¹ erg s⁻¹ and SFR ∼ 0.03 M_☉ yr⁻¹ according to the calibration of Calzetti et al. (2010) at the redshift of our cluster (z ∼ 0.08).

2.1.1. IR Source Matching and Errors

To confirm cluster members spectroscopically, we targeted objects down to r ∼ 20 in the inner 60' of the MIPS field with the Hectospec fiber spectrograph on the MMT at Mt. Hopkins (Fabricant et al. 2005). We used the 270 gpm grating, covering a wavelength range of 3650–9200 Å at a resolution of 6 Å, allowing us to observe the Hα emission line in galaxies out to z ∼ 0.4, far beyond the cluster redshift. The fibers subtend 15 on the sky, possibly necessitating aperture corrections for line fluxes. However, our use of continuum photometry for estimating fluxes implicitly includes a first-order aperture correction. The region was observed using two different configurations with three 10 minute exposures each for the bright galaxies and five different configurations with three 15 minute exposures each for the faint ones. The spectra were reduced using the HSRED³ IDL routine. We visually confirmed the redshift of each source using best-fitting templates, resulting in confident redshifts for 1362 objects.

Because 24 μm emission from AGNs can masquerade as star formation, we need to identify possible active galaxies in our sample. HSRED identifies AGNs and QSOs using the following emission line ratios:

$$\begin{eqnarray*} &&\mathrm{log\left(\frac{[{\rm O\,\scriptsize{III}}]}{H\beta }\right) > 0.7 - 1.2 \left(log\left(\frac{[{\rm N\,\scriptsize{II}}]}{H\alpha }\right) + 0.4 \right)} \end{eqnarray*}$$

(see the HSRED documentation for links and details). We reviewed the spectra by eye and compared the emission lines with classic BPT diagrams (Baldwin et al. 1981; Kewley et al. 2001; Kauffmann et al. 2003) to ensure we were only removing AGNs that were dominating the spectra. A few additional galaxies with exceptionally broad emission lines (but not identified as AGNs by HSRED) were also classified as AGNs and removed.

We selected far-UV (FUV) sources in the GALEX database⁴ from the A2029 pointed observation GI3-103 (PI: Hicks) described in further detail in Hicks et al. (2010). The GALEX 1.2 deg field of view, centered on the BCG, includes our entire Hectospec coverage area so that none of our spectroscopically confirmed cluster members lie near the edge of the GALEX image. This 1517 s observation is ∼80% complete down to ∼23.0 mag. Because this completeness limit is for 5σ or brighter sources, our SFR_FUV detection limit (0.043 M_☉ yr⁻¹) is a conservative estimation.⁵ We then matched the UV sources with our optical catalog using a 4'' radius.

To differentiate cluster members from the field, we used the method of den Hartog & Katgert (1996), who select members based on each galaxy's relative line-of-sight velocity (V − V0, where V0 is the mean cluster velocity) as compared with its projected distance from the center of the cluster. Den Hartog & Katgert (1996) then use the maximum line-of-sight velocity at a given distance from the cluster center to define which galaxies are cluster members and which are interlopers. Figure 1 shows this method for A2029. The maximum line-of-sight velocity is plotted as a pair of dashed lines. Only galaxies within these lines are considered cluster members. To ensure we were as complete as possible, we retrieved all extended sources near the cluster redshift within our 24 μm field of view from the NASA/IPAC Extragalactic Database⁶ (NED) and the Sloan Digital Sky Survey⁷ (SDSS). We removed duplicate sources and included the remaining galaxies for identifying cluster members.

Zoom In Zoom Out Reset image size

After removing all non-cluster galaxies and limiting the field sample to galaxies with z ⩽ 0.2, we have a total of 585 cluster galaxies and 239 field galaxies with confirmed redshifts. Of those, 444 cluster members and 148 field galaxies were covered by Hectospec and deep MIPS observations. However, for this study, we did not include any of the NED or SDSS sources in our final sample because it is difficult to identify and correct for selection biases of these galaxies. All cluster members are listed in Table 1; this includes galaxies not explicitly used in our analysis.

For all non-AGN cluster galaxies not detected in Hα but detected in at least one other star formation indicator, 14 are only detected at 24 μm, 16 are only detected in the FUV, and 24 are detected in both. We examined these galaxies more in-depth to identify whether they are truly star forming or if they are merely quiescent sources.

Figure 2 shows the typical optical spectra (rest frame) of these 54 FUV- and/or IR-detected galaxies, with the Hα wavelength identified by a vertical dashed line. All of the galaxies' spectra (including the ones not shown here) are reminiscent of early-type galaxies. Other than the sky lines, the only obvious features are hydrogen and Ca H and K absorption lines; a few of the spectra shown also have [N ii] λ6583 emission lines. They do not look like star-forming galaxies at all.

Zoom In Zoom Out Reset image size

However, the Hectospec fibers subtend 15 on the sky, which means for the larger galaxies, we may only be probing the central bulges and missing star formation in the outer regions of the galaxies. If we look at optical images of this population from SDSS (Figure 3), we see that nearly all have early-type morphologies (E/S0). Some could be edge-on spiral galaxies, but given their red colors and the lack of any indications of star formation in their spectra, they seem to be completely passive. Only a handful show blue disks or spiral arms indicative of ongoing star formation; we classified these as true star-forming galaxies, while the rest were identified as being passive/quiescent.

Zoom In Zoom Out Reset image size

Forty-one of the original 54 galaxies only detected in the IR and/or FUV appear to be quiescent; if they truly are not forming stars, then what is the source of this emission? We previously covered some of the possibilities, but we would like to understand the source of the IR or FUV emission to be sure we are not excluding galaxies that may be forming stars at very low levels.

First of all, due to the large size of the 24 μm PSF, it is possible that the IR source is offset from the galaxies enough to cast doubt in terms of the 24 μm matching. While we estimated the matching criteria errors previously, the IR emission could still be from a nearby source. We investigated this possibility for the passive early-type galaxies, and most agreed in position almost exactly with their optical source. The few galaxies whose IR emission was visually identified as possibly coming from a companion object were assumed not to be detected at 24 μm.

Now that we are certain the 24 μm emission originates with the early-type galaxies, we can attempt to identify the source. Temi et al. (2009) use K_S-band luminosity compared with L₂₄ to identify both passive and star-forming early-type galaxies in the local universe. Figure 4 is our re-creation of the first plot of their Figure 1. We matched our cluster galaxies with those from the 2MASS extended and point-source catalogs and plotted them as black squares (passive galaxies) and gray squares (all other cluster members). The dashed line is our 24 μm detection limit at the redshift of the cluster. Temi et al. (2009) split their sample into blue (U − V < 1.1; blue circles) and red (U − V > 1.1; red circles) early types. Green circles indicate galaxies without available U and/or V colors, and green triangles indicate 24 μm upper limits. Temi et al. (2009) found a strong relation between $L_{K{_S}}$ and L₂₄ for red early-type galaxies with no apparent star formation (i.e., 24 μm emission is from old stars and not star formation or AGNs; solid black line). Our own passive early types mostly fall on or very close to the relation, indicating that the IR emission in these galaxies is dominated by the old stellar population rather than AGNs or ongoing star formation. If we convert our SDSS data to UBVRI colors (using KCORRECT; Blanton & Roweis 2007), all but two of our galaxies would easily fall in the Temi "red" galaxy sample; indeed, most are much higher than the Temi et al. (2009) U − V limit. A few of our passive galaxies lie significantly above the truly passive relation, but like the rest of the sample, they appear to be normal early-type galaxies with no emission lines (though some have weak [N ii] λ6583). The two early-type galaxies with the highest offset from the passive relation (also the least massive) do have blue colors (U − V < 1.1), but this could be due to AGNs. Since the nature of the 24 μm emission in these galaxies remains unknown, they remain in the "passive early-type" list. Even if they were truly star-forming galaxies, however, they would not affect our results.

Zoom In Zoom Out Reset image size

Though the IR emission from most of the passive A2029 galaxies seems to be from dust around old stars, weak AGNs could also be contributing. We removed the obvious AGNs from our sample already, but we could have missed low-luminosity AGNs. Inspection of our optical spectra reveals what could be faint [O iii] λ5007 emission lines in some of our IR early types, but this is hardly conclusive. Most of these galaxies are also detected in the FUV, which could be coming from AGNs (e.g., Agüeros et al. 2005), though, as mentioned previously, it could also be coming from old stars or recently quenched star formation (e.g., Greggio & Renzini 1990; Cortese et al. 2005; Yi et al. 2005; Donas et al. 2007; Atlee et al. 2009; Choi et al. 2009).

Instead, we can turn to X-ray observations of the cluster. Deep Chandra and XMM-Newton images from the central region of the cluster cover most of the 41 IR/FUV early-type galaxies we previously identified as not forming stars. Four are clearly detected in X-rays, implying that AGNs are likely responsible for the IR/FUV emission in at least some cases. It is also possible that some have low-luminosity AGNs that are too faint in X-rays to be detected in these observations. Additionally, the galaxies closest to the cluster center may not be detected due to their proximity to the bright X-ray gas of the cluster itself.

In conclusion, there are a variety of non-star-forming mechanisms that can produce low levels of IR and/or FUV emission in early-type galaxies. Generically, it appears that extremely low or nonexistent levels of star formation allow other IR and FUV sources that are usually hidden to dominate. Independent of the actual mechanism for the 24 μm and FUV emission in these passive early-type galaxies, it is highly unlikely to be from ongoing star formation. In total, 41 of the 153 galaxies detected using at least one of the three star formation indicators do not appear to be actively forming stars. This is 27% of the detected galaxies—not an insignificant fraction! This means that for SFR ≲ 1 M_☉ yr⁻¹, one must be careful when using only a single indicator (UV or IR) for identifying star-forming galaxies.

Now that we have identified the galaxies in the cluster currently forming stars, we can estimate their SFRs and study how the dense environment of A2029 has affected these galaxies over time. For all sources detected at 24 μm, we calculated the monochromatic 24 μm luminosity (L₂₄) with the general relation between flux and luminosity (νF_ν) using the luminosity distance. Hα luminosities (L_Hα) were calculated using the Hα equivalent width along with the continuum magnitude, which was estimated by linearly interpolating between the SDSS r- and i-band magnitudes to the Hα rest-frame wavelength. The linear behavior of the SEDs of these galaxies at r, i, and z (and sometimes g) makes linear interpolation appropriate in this case. For the few sources without SDSS observations, we simply use the r-band magnitude, which is not significantly different from the interpolated magnitude for most sources. Additionally, because our field galaxy sample only contains objects out to z ∼ 0.2, this method can be used for all field galaxies as well since Hα does not redshift beyond the i band until z ∼ 0.4. The equivalent widths were calculated by linearly fitting the continuum within 200 Å of the emission line and fitting a Gaussian to the Hα line on top of the continuum fit. All fits were visually checked; those where the continuum and/or Hα line were fit poorly by our automated routine were then fit manually (using the same method) so as to make the measurements as accurate as possible. The errors of these manually fit lines were modified to reflect the higher uncertainty in the fits.

Hα and emission by warm dust at 24 μm are both powered predominantly by very young stars in normal star-forming galaxies—see Calzetti et al. (2010, and references therein). We use their equations to convert 24 μm and/or Hα luminosities to total SFRs. As required by Calzetti et al. (2010), for galaxies with both 24 μm and Hα measurements, we used unobscured (uncorrected) L_Hα. For sources with only Hα detections and no 24 μm, we had two options: for a lower limit on the SFR (i.e., the unobscured component), we used the same conversion but without the IR contribution; as an upper limit, we added in our 24 μm detection limit for the IR component. Although the very low SFRs of the Hα-only detected sources suggest that the obscuration in these systems is insignificant, if we include the 24 μm upper limits, only about half of the SFRs of these galaxies remain reasonably unaffected; the other half change more significantly, indicating there could be a considerable obscured component for some sources. Despite this, we chose to display the non-corrected Hα SFRs in our figures unless stated otherwise; using the upper limits does not change our conclusions.

For estimating the Hα detection limit, however, we need to take into account the host galaxy's luminosity. Bright galaxies tend to have lower specific SFRs (SFR/M_*) than faint ones, so as a galaxy's luminosity increases, the stellar continuum increases in brightness faster than the Hα (Boselli et al. 2001; Brinchmann et al. 2004). This means that our Hα detection limit will change depending on the luminosity of the galaxy. Therefore, we estimated the detection limit using L_Hα of the brightest galaxies in the r band and converting them to SFRs using the methods already discussed. All but a few of the Hα-detected galaxies are above our limit (∼0.07 M_☉ yr⁻¹), and ∼ 90% of all galaxies in the cluster are fainter in the r band than the galaxies used to calculate the detection limit, giving us an estimate of the completeness limit of our Hα-detected galaxies as well.

The UV (and far-IR) emission of a galaxy is often powered by somewhat older (and later) stars than the emission in Hα and at 24 μm. The evolution of cluster galaxies may be fast enough to modify the calibration of the SFR in the UV (e.g., Boselli et al. 2009). However, these effects are not well calibrated, so we calculated UV SFRs based on field galaxy behavior. We corrected the FUV (1350–1780 Å) luminosity for obscuration effects using L₂₄ (for sources with both UV and IR detections) from Zhu et al. (2008). Then we used the method of Kennicutt et al. (2009) to convert this FUV luminosity to an SFR. For late-type galaxies only detected in the FUV (not 24 μm or Hα), as with Hα-only detected objects, we use the same conversion to SFR but without the 24 μm component, resulting in a lower limit on the SFR (i.e., the unobscured component). If we compare our unobscured SFR_Hα with the unobscured SFR_FUV (for objects with detections in both), the scatter is larger, though the SFR_Hα tends to be higher than the SFR_FUV. As with the Hα measurements, if we correct the UV-only SFRs with our 24 μm detection limit (resulting in an upper limit on the SFR), our conclusions do not change.

Stellar masses were estimated using the KCORRECT (ver. 4.2) routine SDSS_KCORRECT, which calculates mass using template and model fits with SDSS photometric data (see Blanton & Roweis 2007 for details). KCORRECT uses a different IMF and different cosmological values than we use (Kroupa 2001), so we corrected the original stellar mass output to the IMF and cosmology adopted in this paper.

We now look at the distributions of star-forming galaxies in the cluster. The locations of these galaxies can be an indicator of star formation suppression with respect to density. Indeed, other galaxy properties can vary with respect to cluster-centric distance, such as H i gas content (e.g., Cayatte et al. 1990; Solanes et al. 2001; Gavazzi et al. 2005, 2006), SFR (e.g., Lewis et al. 2002; Gómez et al. 2003; Bai et al. 2009; Edwards & Fadda 2011), and morphology (e.g., Melnick & Sargent 1977; Dressler 1980; Whitmore et al. 1993; Goto et al. 2004).

Figure 5 is a histogram of the projected radius (in Mpc) of all galaxies in A2029 (black solid line), all star-forming galaxies regardless of detection method (blue hashed bins), and the non-star-forming IR/FUV early-type galaxies (red hashed bins). The star-forming galaxies appear to be somewhat evenly distributed, while the passive galaxies are more concentrated closer to the center of the cluster.

Zoom In Zoom Out Reset image size

These trends could be biased by projection effects, however. To test this, we fit the overall galaxy distribution with a Navarro, Frenk, and White profile (NFW; Navarro et al. 1996) and created fake clusters with ∼500 galaxies randomly arranged in three-dimensional space according to the distribution. We then took the same fraction of star-forming galaxies as in A2029 (∼29%), placed them throughout the cluster, and created histograms of their projected radius. We ran this simple simulation hundreds of times and averaged the results. First, we distributed a star-forming galaxy population uniformly throughout the cluster, resulting in these galaxies having a similar (NFW) profile to the overall cluster population. Then, we distributed the same population uniformly beyond 1 Mpc from the center. This resulted in a relatively constant projected radial distribution, indicating that the ∼1 Mpc radius central region of the cluster is mostly devoid of star-forming galaxies. The lack of star-forming galaxies in the dense core of the cluster, as well as the increasing fraction of star-forming galaxies with increasing projected radius, suggests that the suppression mechanism becomes more efficient as the local density increases. Others have found similar results (e.g., Bai et al. 2009; Mahajan et al. 2010). The slight increase in the IR/FUV early-type galaxies toward the center of the cluster is likely due to the morphology–density relation (and, similarly, the mass–density relation).

Luminosity functions have been used frequently to study the overall distribution of galaxies emitting light at different wavelengths and luminosities. However, we are studying SFRs of galaxies detected using a variety of methods, making individual luminosity functions untenable. Instead, we have created SFR functions combining all of our detection methods. Our SFR functions are corrected for spectroscopic completeness, which was calculated by taking the number of sources with spectroscopic redshifts in a given r-band bin and dividing it by the total number of photometrically identified extended sources in each bin. The reciprocal of these values is the correction factor. We interpolated between these values given each source's r-band magnitude. The 24 μm detections are already 80% complete at the IR detection limit, and the vast majority of our 24 μm detections are far above this limit, so we do not correct for the incompleteness of IR source detections.

We created an SFR function by plotting all cluster galaxies in terms of their detection method (Figure 6). Galaxies only detected at 24 μm, Hα, or FUV are indicated by orange open triangles, blue open stars, or purple open squares, with detection limits shown by the orange dashed line, blue dot-dashed, and purple dotted line, respectively. The green open circles indicate sources detected at both Hα and 24 μm, and the red down-pointing triangles are galaxies detected with both FUV and 24 μm. The black filled circles are all sources, regardless of detection method, with Poisson errors. (Lines connecting the data points are used only to help guide the eye.)

Zoom In Zoom Out Reset image size

Because we do not have sufficient data at the faint end to constrain the shape of our SFR function, we use the faint-end slope of previous studies to compare with A2029. The dashed gray line in Figure 6 is the multi-cluster composite Schechter function from Bai et al. (2009), normalized to A2029 using a χ² minimization routine. We then fit a Schechter function to our cluster using the same faint-end slope but allowing SFR* to vary (solid gray line). This results in log(SFR*/(M_☉ yr⁻¹)) ∼ 0.34 ± 0.04, a lower value than the composite, which is ∼0.47$^{+0.13}_{-0.12}$, converted to SFR using our method and evolved to our cluster redshift using Le Floc'h et al. (2005). However, given the error bars and the overall fits to the data, both Schechter functions are comparable and reasonably good representations of A2029.

The Bai et al. (2006, 2009) studies only use 24 μm detections for their luminosity functions, though. To directly compare their Coma and A3266 composite luminosity function with our cluster, we plotted a total IR luminosity (L_TIR) function for A2029, shown in Figure 7. We estimated L_TIR using Rieke et al. (2009), which was shown in Bai et al. (2009) to be comparable to their own template-fitting method for estimating IR luminosity and SFRs. The black filled circles are for all 24 μm detected cluster members, while the blue triangles only include IR sources identified as actively forming stars. The vertical dashed black line indicates the 24 μm detection limit; we do not use the data point below this limit in any fits. The dashed and solid red lines show the best-fit Schechter function from the Bai et al. (2009) composite luminosity function normalized to all 24 μm emitting A2029 galaxies and only 24 μm emitting galaxies forming stars, respectively. The composite Schechter function is a reasonable fit in both cases. Even if Coma and A3266 also have a population of non-star-forming 24 μm galaxies that were included in the luminosity functions of Bai et al. (2009), the IR luminosities of the galaxies are low enough that they would be below our completeness limit for A2029. In other words, accidentally including non-star-forming 24 μm galaxies in Coma and A3266 does not change the similarities between their luminosity function and that of A2029. All three clusters can be fitted by the same luminosity function, found by Bai et al. (2009) to be very similar to the field luminosity function as well.

Zoom In Zoom Out Reset image size

Our SFR function has a turnover at log(SFR) ≲ −0.8, which we found to be due to a selection effect. For our spectroscopy, we only targeted galaxies with r ⩽ 20 mag, which will automatically filter out low-mass galaxies. This effect can be seen in Figure 8, where we compare the stellar masses of the galaxies with their total SFRs (IR, Hα, and/or FUV). Symbols connected by lines indicate galaxies detected at Hα or FUV (but not 24 μm), where the lower value is for Hα or FUV SFRs only and the upper values include our 24 μm detection limit. A2029 galaxies are shown as red circles, and field galaxies are shown as blue stars. Galaxies with no detected star formation indicator are plotted as downward arrows, indicating that these points are upper limits (based on our 24 μm detection limit). The dashed line is our approximate mass-dependent Hα detection limit for the cluster. If we look at all of the quiescent galaxies, we seem to be complete in mass only down to log(M_*) ∼ 9.7. Some of the galaxies are lost due to the r-band magnitude limit, but many could be due to an inability to identify redshifts. It is difficult to determine the redshifts of quiescent galaxies because of their lack of emission lines; this is even harder for faint galaxies. Star-forming galaxies, on the other hand, have bright emission lines, making it much easier to measure their redshifts (as can be seen in the figure). We estimate that for r ∼ 19.5 mag star-forming galaxies, we are complete in mass down to at least log(M_*) ∼ 9.0, if not lower. However, the greater difficulty in measuring redshifts accounts for the higher mass completeness limit for quiescent galaxies. Given what we know of our completeness in the star formation indicators, we estimate that our SFR function is complete down to log(SFR) ∼ −1.0.

Zoom In Zoom Out Reset image size

Let us look more closely at Figure 8. Recall that red circles are galaxies in the cluster and blue stars are field galaxies (note that our field sample is incomplete below log(SFR) ∼ −0.07). Open symbols connected by lines indicate galaxies with FUV or Hα detections but no 24 μm detections. The lower and upper limits in SFR for these sources are indicated by the unobscured SFR (FUV or Hα only) and maximum SFR (FUV or Hα corrected with the 24 μm detection limit), respectively. Overall, the star-forming galaxies in the cluster and field are comparable where the field is complete. The quenching process in this cluster appears to be primarily responsible for a larger proportion of quiescent galaxies, not a change in the mass–SFR relation for the active galaxies.

To quantify the mass–SFR distribution of cluster galaxies, we need to know the mass–SFR relation for the field and minimize any systematic differences between the field and our cluster. We adopted the relation of Brinchmann et al. (2004), who found a strong correlation between stellar mass and SFR for a large sample of field galaxies. Up to log(M_*/M_☉) ∼ 10.5, this relation is linear; the SFRs of higher-mass galaxies drop off and become more uncertain due to lower numbers of galaxies and the larger proportion of early types at high mass. We extrapolate the relation above log(M_*/M_☉) = 10.5 to represent only the late-type population (as is also done by Elbaz et al. 2007). To avoid this region, and because of our completeness limit, we only compare galaxies in the mass range 9.5 ≲ log(M_*/M_☉) ≲ 10.5. The mode of the Brinchmann et al. (2004) relation follows the distribution of star-forming galaxies, so we use the mode as the basis of our field comparison.⁸

To minimize any systematic differences between Brinchmann et al. (2004) and our study, we calculated the SFR and stellar mass in the same manner as A2029 for Hα and 24 μm detected galaxies from the Boötes field (Rujopakarn et al. 2010; not including AGNs) and combined it with our own field galaxies. We then fit a line to the mode of the Brinchmann et al. (2004) relation and normalized it to our combined field sample. Our result can be seen in Figure 9. Our field galaxies are blue stars, the Boötes sample galaxies are black triangles, and the galaxies used in the fitting (i.e., in the region of the plot where these samples are the most complete) are circled in red. This normalized Brinchmann et al. (2004) relation is shown by the black dashed line; we will refer to it as the overall "field relation" from now on. We plotted the Elbaz et al. (2007) field relation (for blue SDSS galaxies at z ∼ 0) as a green dotted line for comparison.

Zoom In Zoom Out Reset image size

Earlier, we compared A2029's SFR function with the Coma cluster; it would be prudent to compare both clusters' mass–SFR distributions to the field as well. While many groups have studied Coma using multiwavelength data that are now publicly available (e.g., Iglesias-Páramo et al. 2002; Gavazzi et al. 2003, 2006; Cortese et al. 2008; Hammer et al. 2012), we chose to utilize the recent results from Edwards & Fadda (2011) due to their observations being similar to those we have for A2029. Additionally, Edwards & Fadda (2011), unlike the previous studies, were able to disentangle the Hα line from the surrounding [N ii] lines, which is necessary for accurately estimating SFRs.

Figure 10 shows the same total SFR versus stellar mass as Figure 8, including Coma (blue filled triangles; Edwards & Fadda 2011) with A2029 (red filled circles). We included Coma galaxies with both 24 μm and Hα detections; only a few Coma galaxies with spectra are not detected at Hα (probably because only galaxies detected at 24 μm were targeted). We calculated total SFRs for Coma in the same way as for our cluster. A subset of the Coma galaxies was listed in SDSS, allowing us to compare our method for estimating stellar masses with theirs. Because the two techniques resulted in masses within a factor of two for most of the galaxies, we use the masses listed in Edwards & Fadda (2011) for all Coma galaxies. The horizontal dotted line is our 24 μm SFR detection limit. While the Coma 24 μm detection limit is much lower than A2029, the Hα detection limit should be comparable to ours.

Zoom In Zoom Out Reset image size

A2029 and Coma seem mostly to follow the field relation, though Coma appears to have a larger number of high-mass, low-star-forming galaxies. This is easier to see in Figure 11, where we show the SFR offset between the cluster galaxies and the field for 9.5 ≲ log(M_*/M_☉) ≲ 10.5. The solid black line is the distribution of field galaxies at these masses (Brinchmann et al. 2004), while the red dashed line and blue dotted line are the distributions of A2029 and Coma galaxies, respectively. A2029 follows the field very closely, while Coma has more low-SFR galaxies than either the field or A2029, as we noted earlier. In addition, due to the targeting of IR-bright galaxies for spectroscopy (Edwards & Fadda 2011), massive quiescent galaxies are missing from the Coma sample. The detection limit for both clusters (vertical dashed line) is log(M_*/M_☉) = 9.5. The incompleteness slowly increases for masses further below this limit. A completeness-corrected version of the figure is discussed later.

Zoom In Zoom Out Reset image size

One issue with the Coma SFRs is the Hα aperture corrections (Edwards & Fadda 2011). The size of the spectral fibers used for Coma tends to be smaller than the galaxies themselves, indicating that some Hα luminosity is lost. Edwards & Fadda (2011) estimate corrections for this lost light based on the galaxy continuum photometry and state that the median aperture correction is ∼2, though some can be as high as 10.

Because the star-forming regions are likely to be distributed differently from the optical continua, we test the effects of different aperture corrections. Figure 12 shows the field (black dashed line) and Coma (blue solid line) distributions with aperture correction factors of one (no correction), two, four, and eight. These plots show that even with the most extreme correction, the population of low-star-forming galaxies is still readily apparent. This population of star-forming galaxies is clearly real and not from losing Hα due to fiber size.

Zoom In Zoom Out Reset image size

To demonstrate that the correction should be no larger than a factor of two (the median aperture correction as reported by Edwards & Fadda), we carried out another test. The 24 μm fluxes represent the total star formation over the full galaxy (beam of 6'', extended emission captured to first-order through SExtractor), and we can compare them with the Hα measurements to see if there is a tendency for the latter to be systematically too small. For this test, we used the SINGS sample, for which integrated 24 μm and Hα fluxes are available (Kennicutt et al. 2003). We restricted the test to galaxies with SFRs between 0.1 and 1 M_☉ yr⁻¹ to have good statistics and comparable SFRs in the samples. We found that there was no significant difference between the SFR₂₄/SFR_Hα for the Coma galaxies (mean 10, median 4) and the SINGS galaxies (mean 8.6, median 6.8), suggesting that for most of the Coma Hα measurements, only small aperture corrections (if any) may be needed.

Despite that the other Coma studies previously mentioned did not separate Hα and [N ii], we can use them as an additional check on aperture effects. The available data from Edwards & Fadda (2011) include [N ii] λ6583; given our own spectra, we can safely assume that [N ii] λ6548 tends to be approximately 50% of the longer-wavelength [N ii] λ6583 line. We combined the fluxes of the three lines and compared the Edwards & Fadda (2011) values with the blended Hα/[N ii] fluxes from Iglesias-Páramo et al. (2002), GOLDMine (Gavazzi et al. 2003), and Gavazzi et al. (2006), for all overlapping Coma galaxies. (We focused on GOLDMine since it has the most complete sample and highest number of overlapping sources.) The blended GOLDMine line fluxes were an average of ∼2 times higher (median ∼2.5) than the combined Hα/[N ii] lines from Edwards & Fadda (2011). Since the dominant difference between the two studies should be due to aperture corrections, this supports the median aperture correction of 2 that Edwards & Fadda (2011) cite in their study (and that we have chosen to use here).

In the case of A2029, the projected fiber diameters are already about twice those for Coma, and our use of continuum photometry to convert equivalent widths to line fluxes implicitly makes a first-order aperture correction. We therefore make no additional adjustments for the size of the spectroscopic aperture.

4.4.1. Anemic Coma Galaxies

The analysis in the preceding section has identified a population of star-forming galaxies in Coma that fall significantly below the field mass–SFR relation. How are these galaxies different from the other star-forming galaxies?

We extracted the subset of 13 massive (10 ≲ log(M_*/M_☉) ≲ 10.5) galaxies that lie significantly below the field relation (offset ≲−0.7) for closer study. These galaxies are listed in Table 2. First of all, these galaxies are fairly evenly distributed in terms of projected area in the cluster, which, as we showed earlier with A2029, means most if not all of the galaxies are in the outskirts. Second, almost all are early-type galaxies, with over half being S0s, specifically. None show obvious signs of ongoing merger activity, though some have indications of previous interactions (e.g., rings). We classify this small population of high-mass, low-star-forming Coma galaxies as "anemic," a term defined by van den Bergh (1976) to identify spiral galaxies with abnormally low amounts of gas and star formation as compared with "normal" spirals. Van den Bergh identified the brightest Coma cluster spiral, NGC 4921, as anemic as well, and we include it in our table of anemic Coma galaxies.

Table 2. Anemic Coma Cluster Galaxies

Catalog ID	R.A.	Decl.	Morphology	log(M*)	Total SFR	AGN?
				(M☉)	(M☉ yr−1)
...	194.48608	27.037508	...	10.25	0.106	Y
...	194.14433	27.22756	S0	10.37	0.120	Ya
...	194.29483	27.404902	Sa	10.14	0.055	Y
...	194.65758	27.464002	S0	10.35	0.049	Ya
MCG+05-31-036	194.47652	27.490639	SBb	10.25	0.044	Y
...	194.17664	27.548255	S0	10.32	0.021	N
...	194.26768	27.730046	S0	10.02	0.041	N
MCG+05-31-007	193.85438	27.798033	Sp	10.38	0.072	N
IC 4042	195.17815	27.971266	SB0	10.50	0.070	Y
NGC 4907	195.20335	28.158341	SBb	10.49	0.060	Y
...	196.04228	28.247957	S0	10.08	0.032	N
MCG+05-31-110	195.67002	28.371308	S0/Sp	10.07	0.042	N
...	194.56276	28.521786	...	10.01	0.080	N
NGC 4921b	195.35896	27.885944	SB(rs)ab	11.18	0.181	Y

Notes. Description of columns: (1) catalog ID of the galaxy (if available); (2) and (3) SDSS coordinates; (4) galaxy morphological type as listed by the SIMBAD database; (5) stellar mass, in solar units, adapted from Edwards & Fadda (2011); (6) total SFR using 24 μm and Hα from Edwards & Fadda (2011), calculated using the same method as A2029 (assuming all 24 μm emission is from star formation); (7) flag for whether the galaxy is known to have an AGN or not (most identified via SIMBAD). ^aWe retrieved 20 cm data from Miller et al. (2009) and found the IR–radio q parameter (q = log(F₂₄/F_20 cm)) to be indicative of an AGN. ^bClassic anemic galaxy in Coma, not identified as anemic in our sample due to its high mass. Morphology from de Vaucouleurs et al. (1991).

Download table as:  ASCIITypeset image

Another interesting aspect of these galaxies is that over half were previously identified as hosting an AGN (shown in Figure 10 as larger blue triangles), along with NGC 4921. Indeed, we need to demonstrate that their IR outputs are dominated by star formation, not by AGNs. This possibility could be tested using the diagnostic diagrams proposed by Lacy et al. (2004) or Stern et al. (2005) since both approaches identify AGNs that dominate the mid-IR output of a galaxy by the filling in of the spectral dip near 5 μm by warm dust heated by the AGN. However, the IRAC measurements needed for this diagnostic are available for only two of the anemic galaxies.

Donley et al. (2012) show that the color selection methods are basically equivalent to selecting objects with power-law mid-IR spectra, so we will use an analog of the Lacy/Stern plots using WISE photometry instead. The resulting diagnostic diagram is shown in Figure 13. The dark green line shows the power-law locus, running from indices of −0.25 to −2.25. The green triangles are Seyfert galaxies, divided roughly equally between Type 1 and Type 2 (NGC 1275, 3227, 3516, 3783, 4051, and 4151; Mrk 1, 6, 9, 78, 79, 176, 198, 273, and 348). The red stars are a selection of low-luminosity star-forming galaxies (8 < log(L_TIR) < 10; Sanders et al. 2003), and the black square is the color of a low-luminosity star-forming galaxy based on the templates of Rieke et al. (2009). The Coma anemic galaxies are the blue circles. They fall exactly in the locus of the other star-forming galaxies, demonstrating that their mid-IR outputs are dominated by stellar power and that their SFRs are correctly deduced from IR luminosities. We also show the colors of a number of massive early-type galaxies in Coma, selected to be passive with no indications of star formation in their optical spectra (i.e., no Hα emission; e.g., Moss & Whittle 2005; Miller et al. 2009). Two of these galaxies fall in the star-forming zone, indicating that low levels of star formation may have been overlooked in the optical spectroscopy; the remaining 11 galaxies define a distinct locus that shows the behavior of the IR excess in quiescent early-type galaxies. Thus, the properties of the anemic galaxies in the mid-IR are inconsistent with those of AGN-dominated galaxies or of quiescent ones. They do indicate significant levels of star formation.

Zoom In Zoom Out Reset image size

It is difficult to determine the prevalence of these high-mass, low-star-forming galaxies in Coma, both because of the AGN contamination and because of the detection limit at log(M_*/M_☉) ∼ 9. We need to correct for any galaxies we may be missing below our detection limit. We do not have enough data to correct for the incompleteness for galaxies below log(M_*/M_☉) ∼ 9.5, so we focus on galaxies from 9.5 ≲ log(M_*/M_☉) ≲ 10.5. For each of the first two bins below offset ∼ −1, we calculate a completeness correction using the ratio of areas and number of sources above and below the detection limit in each bin. The results are shown in Figure 14. The solid black, dashed red, and dotted blue lines (the field, A2029, and Coma, respectively) are the same as in Figure 11 except for the completeness corrections for the clusters. Our new completeness limit, with the corrections, is shown as a vertical black dashed line. Even with completeness corrections, A2029 closely follows the field (though it does have a few low-star-forming galaxies), while the majority of Coma's star-forming galaxies are far below the field distribution.

Zoom In Zoom Out Reset image size

4.4.2. 24 μm Emitting Early-type Galaxies

We also discovered that Coma has non-star-forming early-type galaxies with 24 μm emission, as in A2029. These galaxies are shown in Figures 13 and 15. We collected a sample of 13 normal early-type galaxies in Coma, all of which have WISE and 24 μm coverage but do not have Hα emission. In Figure 13, the WISE colors of these galaxies (purple downward-pointing triangles) indicate that most are indeed passive; only a couple have possible residual star formation.

Zoom In Zoom Out Reset image size

All but two of these Coma galaxies are detected at 24 μm, so we added them to the Temi et al. (2009) plot used earlier to show that the IR emission from passive, early-type A2029 galaxies is mostly from the old stellar population (Figure 15). All points are the same as in the previous plot (Figure 4) except that the only A2029 galaxies we include are the passive early-type galaxies. The Coma galaxies, shown as yellow stars, occupy the same region of the plot as the passive A2029 galaxies, indicating that the IR emission from all of these galaxies is from the same source—old stars—rather than star formation.