INTERACTING GALAXIES IN THE A901/902 SUPERCLUSTER WITH STAGES

Before outlining the methods (Sections 3.2 and 3.4) that we use to identify mergers, we briefly discuss the types of mergers that are of interest to our study, their morphological signatures, and their typical post-merger products.

According to simulations, galaxy mergers of mass ratio M₁/M₂ > 1/10 tend to have a significant impact on galaxy evolution. They include major mergers, which are defined to be those with a mass ratio of 1/4 <M₁/M₂⩽ 1/1, as well as minor mergers with 1/10 <M₁/M₂⩽ 1/4. Simulations show that major mergers of stellar systems typically destroy the outer disks, transforming them via violent relaxation into systems with a steep central surface brightness profile, associated with a high Sérsic index n (typically n > 2.5; Hopkins et al. 2009). These remnants include ellipticals and spheroids with a de Vaucouleurs-type stellar profile (n = 4) (e.g., Negroponte & White 1983; Barnes & Hernquist 1991; Mihos & Hernquist 1996; Struck 1997; Naab & Burkert 2001). In gas-rich major mergers, a disky component can form or survive inside the resulting spheroidal component (Hopkins et al. 2009), producing a remnant whose overall profile is less steep than a de Vaucouleurs one, with n < 4. Irrespective of the details of the remnants, the above simulations suggest that ongoing/recent major mergers are associated with arcs, shells, ripples, tidal tails, large tidal debris, extremely asymmetric light distributions, double nuclei inside a common body, and tidal bridges/envelopes of light linking systems of similar mass.

Conversely, minor mergers involving a spiral galaxy and a smaller satellite of mass ratio 1/10 <M₁/M₂⩽ 1/4, will not destroy the outer disk of the larger companion (e.g., Hernquist & Mihos 1995; Smith 1997; Jogee et al. 1999). Typically, the smaller companion sinks via dynamical friction, may excite bars, spirals, and other non-axisymmetric perturbations in the disk of the larger galaxy, and leads to tidal heating, arcs, shells, ripples, tidal tails, tidal debris, warps, offset rings, and highly asymmetric light distributions (e.g., Quinn et al. 1993; Hernquist & Mihos 1995; Mihos et al. 1995; Quinn et al. 1993; Smith 1997; Jogee et al. 1999; Jogee 2006 and references therein).

In this paper, we identify merging systems using two independent methods, which make use in different ways of the aforementioned morphological signatures seen in simulations. The first is a physically motivated classification system (Section 3.2), which is similar to that defined in Jogee et al. (2008, 2009), and is based on visual morphologies, stellar masses, and spectrophotometric redshifts. The second method (Section 4.2) uses the CAS merger criterion (A > 0.35 and A > S), which is based on quantitative asymmetry (A), and clumpiness (S) parameters derived using the CAS code (Conselice et al. 2000).

We visually classify the ACS F606W images of galaxies in the sample of 893 bright (R_Vega⩽ 24) intermediate-mass (M_* ⩾ 10⁹ M_☉) galaxies. This sample is complete for M_* ⩾ 10⁹ M_☉ (see Section 2). In Section 4.1, we will discuss whether further mass cuts should be applied in order to ensure the completeness of major and minor mergers, but for now we consider all the systems in the sample of 893 galaxies. A small fraction (below 1%) of the sample could not be classified due to image defects, low signal to noise, and a highly compact appearance. As illustrated in Figure 2, the remaining galaxies are classified into five main visual classes: Mergers of Types 1, 2a, and 2b, and Non-interacting systems of type Symmetric and Irr-1.

Zoom In Zoom Out Reset image size

The visual class "Mergers" is assigned to systems with evidence of a recent or ongoing merger of mass ratio >1/10. The mergers are subdivided into 3 groups called Type 1, Type 2a, and Type 2b, because different criteria are used to identify these three types of mergers and different techniques are used to separate them into major and minor mergers. Non-interacting systems are subdivided into the visual classes of Symmetric and Irr-1. The results of the visual classification are illustrated in Tables 1–2, and Figures 3–5. Below are details of the 5 main visual classes.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

(A) Mergers of Type 1. Mergers of Type 1 are systems, which appear as a single distorted remnant in the ACS F606W image (with effective PSF ∼01 or ∼280 pc at z = 0.165), and which host morphological distortions similar to those produced in the aforementioned simulations of mergers of mass ratio >1/10. Thus, mergers of Type 1 likely represent the very advanced phases of such a merger, at a point where the two progenitor galaxies have coalesced into a single merger remnant. Among the sample of M_* ⩾ 10⁹ M_☉ systems, we find 13 mergers of Type 1. Their properties are shown in Table 2 and Figure 3. They show morphological distortions such as tidal tails (Figure 3, Cases 10, 11), shells, ripples, warps, asymmetric tidal debris and distortions (e.g., Cases 5, 6, 7, 9), or a "train-wreck" morphology (Cases 7, 10, 11, 12).

For mergers of Type 1, a single redshift and stellar mass are available. Thus, the evidence for a merger of mass ratio >1/10 does not come from a measured stellar mass ratio M₁/M₂, but instead is inferred from the presence of the aforementioned morphological distortions, which are seen in simulations of mergers of mass ratio >1/10.

Ideally, one would like to further separate the mergers of Type 1 into major and minor mergers, as defined in Section 3.1. However, without individual stellar masses M₁ and M₂ for the progenitors, and with only morphological distortions as a guide, this is not possible in every case since the morphological disturbances induced depend not only on the mass ratio of the progenitors, but also depend to some extent on the orbital geometry (prograde or retrograde), the gas mass fraction, and structural parameters (e.g., Mihos & Hernquist 1996; Struck 1997; Naab & Burkert 2001; Mihos et al. 1995; di Matteo et al. 2007). Therefore, we separate the Type 1 mergers into three groups: likely major mergers, likely minor mergers, and ambiguous cases of "major or minor" mergers as follows:

• 1.  
  The class of likely major mergers includes systems, which host fairly unique tell-tale morphological distortions characteristic of a major merger, such as a train-wreck morphology (e.g., Cases 7, 10, 11, 12 in Figure 3), or two nuclei of similar luminosities.
• 2.  
  A system is classified as a likely minor merger if the outer disk has survived a recent merger. We base this on the results of simulations of mergers involving spiral galaxies where the outer disk of the spiral survives a minor merger, but not a major merger (Section 3.1). This is not true in every case: major mergers of extremely gas-rich disks with low SF efficiency can lead to a remnant with an extended stellar disk (Robertson et al. 2004), but such mergers are unlikely to be relevant for our study, which focuses on intermediate-mass (M_* ⩾ 10⁹ M_☉) systems at redshifts well below 1. An additional secondary criterion for classifying a system as a likely minor merger is that the light from morphological distortions, tidal debris, or accreted system in the surviving disk is a small fraction (between 1/4 and 1/10) of the total luminosity of the disk. Examples include Cases 8, 9, 13 in Figure 3 and Table 2.
• 3.  
  The class of ambiguous "major or minor" merger is assigned to systems hosting distortions, which could be due to both a major and a minor merger. Examples are Cases 1–6 in Figure 3 and Table 2.

(B) Mergers of Types 2a and 2b. Mergers of Types 2a and 2b are systems, which appear in ACS images as a very close (separation d< 10 kpc) overlapping pair of two galaxies, and whose properties suggest they are in the very late phases of a merger of mass ratio above 1/10. The difference between mergers of Type 2a and 2b is that the latter are resolved into two separate galaxies by the ground-based COMBO-17 data of resolution ∼15 (corresponding to ∼4.3 kpc at z ∼ 0.165), while the former are not. Thus, mergers of Type 2a only have a single redshift and a single stellar mass for the pair, while mergers of Type 2b have stellar masses (M₁, M₂) and spectrophotometric redshifts (z₁, z₂) for both pair members.

We focus only on very close pairs (with d< 10 kpc), and ignore widely separated pairs for several reasons. First, in this work, we are interested in systems with evidence for a recent or ongoing merger of mass ratio >1/10. The very close pairs represent the very late phases of such a merger, such that gravitationally bound members can coalesce into a remnant on a short timescale, comparable to the visibility timescale over which the morphological distortions in mergers of Type 1 persist. Second, we expect the morphological distortions from the ongoing merger to be more prominent in close pairs than widely separated ones. These morphological signatures in turn make it easier to separate real pairs from projection pairs. Third, as shown by our Monte Carlo simulations (see below), the statistical contamination from projection pairs increases significantly for distant pairs.

When identifying potential mergers of Type 2b, we consider very close pairs where one pair member has M_* ⩾ 10⁹ M_☉, and its companion satisfies the following two conditions. First, the companion has a mass such that the mass ratio is in the range 1/10 <M₁/M₂⩽ 1/1. Note that this condition allows for pairs where the companion can be of higher, as well as lower mass (see Table 2). Second, the companion has similar spectrophotometric redshift, such that the absolute value of (z₁ − z₂) is less than or equal to the 1σ redshift error σ_z. The latter criterion helps to remove some of the false projection pairs. However, there is still residual contamination by false pairs as σ_z is fairly large (see Section 2; σ_z∼ 0.028 at R_Vega = 23.0).

We identified 5 and 18 potential candidate mergers of Types 2a and 2b, respectively, in our final sample of 886 systems. They are shown in Figure 3 and Table 2. We use the term "potential mergers of Types 2a and 2b" when describing the 23 very close (d< 10 kpc) galaxy pairs listed in Table 2 (Cases 14–36), because these pairs could be a mix of real pairs and false projection pairs caused by chance line-of-sight superposition. In order to minimize the contamination from false projection pairs, we only include in our final analysis those pairs where at least one galaxy shows morphological distortions indicative of a galaxy–galaxy interaction. Such pairs are most likely to be real pairs, as opposed to projection pairs, and we will use these systems to set a firm lower limit to the merger fraction in Section 4.1. We find 7 such pairs (Cases 16–18, 20, 24, 34, 35), which are marked with a star in Column 1 of Table 2. This suggests that the remaining 16/23 or up to 70% of the Types 2a and 2b could potentially be pairs in projection, due to the dense cluster environment.

As a second step to gauge the level of contamination from projection pairs, we performed a Monte Carlo simulation. We generated a random galaxy distribution which has an identical redshift distribution to that of our sample of intermediate-mass (M_* ⩾ 10⁹ M_☉) galaxies, over the STAGES 05 × 05 field. We ran 100 realizations of this Monte Carlo simulation. For each realization, we took the ratio (N_obs/N_mc) of observed to simulated Monte Carlo pairs at different separations (d) out to 140 kpc. The same criteria outlined in Section 3.2 to identify the pairs in the observations are applied to the simulated galaxies. Namely, we apply the condition that the absolute value of the difference in spectrophotometric redshifts of the galaxies in a pair must be less than or equal to the 1σ redshift error σ_z, in order for the galaxies to be counted as a pair. Figure 4 shows the ratio (N_obs/N_mc) plotted as a function of d, with the four panels showing the mean, median, minimum, and maximum values of (N_obs/N_mc). Our results are qualitatively similar to those found by Kartaltepe et al. (2007). At large separations of d > 50 kpc, (N_obs/N_mc) is below 1, and false projection pairs can dominate the statistics. At small separations of d < 20 kpc, (N_obs/N_mc) is well above 1, suggesting that random chance superposition cannot fully account for all the observed pairs. Thus, by picking pairs with a separation of d < 20 kpc, one can reduce the contamination from false pairs. For d < 20 kpc, the mean value of the contamination level (defined as the fraction of observed pairs that are false) ranges from 16% to 55% and has a median value of 36%. The maximum contamination level is a factor of ∼1.3 lower than the factor of 70%, which we derived in the preceding paragraph, by assuming that all pairs without morphological distortions are projection pairs. One possible explanation for the difference is that the afore-described Monte Carlo simulations do not take into account the large-scale structure and possible overdensity of galaxies associated with the central parts of the supercluster.

We next attempt to separate the mergers of Type 2a and 2b into major and minor mergers. For mergers of Type 2b, where stellar masses (M₁, M₂) are available for each pair member, we can use the stellar mass ratio M₁/M₂ (see Table 2). By definition, systems with 1/10 <M₁/M₂⩽ 1/4 and 1/4 <M₁/M₂⩽ 1 are classified, respectively, as minor and major mergers. As shown in Table 2, potential major mergers of Type 2b are Cases 20–22, 24–26, 28, 29, 32, and 34, while Cases 19, 23, 27, 30, 31, 33, 35, and 36 are potential minor mergers of Type 2b.

Mergers of Type 2a, however, only have a single redshift and a single stellar mass, and hence the stellar mass ratio of the pair is not directly measurable. Instead, we use the stellar light ratio L₁/L₂, measured from the ACS images, as an approximate proxy (see Table 2). Examples of likely major mergers of Type 2a are Cases 16 and 18 in Figure 3, while Cases 14, 15, and 17 are examples of likely minor mergers of Type 2a.

(C) Non-interacting systems of Type Symmetric and Irr-1. The remaining systems, which show no indication of a recent merger, according to the criteria outlined in Sections A and B above, are classified as non-interacting systems. They are subdivided into two groups called "Non-interacting Irr-1" and "Non-interacting Symmetric." "Non-interacting Irr-1" systems exhibit internally triggered asymmetries, typically on scales of a few hundred parsecs. These asymmetries are generally due to stochastic SF and/or low ratios of rotational to random velocities. They can impart a clumpy morphology to a galaxy and are different from externally driven distortions, such as tails and ripples, which are often correlated on scales of several kpc. The class of "Non-Interacting Symmetric" consists of galaxies, which are relatively undistorted. Note that this class includes the undistorted galaxies, which are part of these Type 2a and 2b pairs identified as projection pairs in Section B. Examples of "Non-interacting Irr-1" and "Non-interacting Symmetric" galaxies are shown in Figure 5.

In summary after performing the visual classification of the 893 bright (R_Vega⩽ 24) intermediate mass (M_* ⩾ 10⁹) galaxies, we find the following results. We identify 13 merger remnants of Type 1 (Cases 1–13 in Table 2 and Figure 3), and 7 close pairs of Types 2a and 2b (Cases 16–18, 20, 24, 34, 35 in Table 2 and Figure 3) where at least one galaxy in the close pair shows morphological distortions indicative of a galaxy–galaxy interaction. Such pairs are most likely to be real pairs, as opposed to projection pairs. We discount those very close pairs where none of the members are morphologically distorted. Our final sample includes 20 mergers and 866 non-interacting galaxies (123 Irr-1 + 743 Symmetric; see Table 1).

Several tests were performed on the visual classes described in A, B, and C in Section 3.2. The visual classifications were calibrated by having three classifiers (AH, IM, SJ) train on a set of ∼100 nearby field galaxies. The full supercluster sample was then classified by AH and IM, with all ambiguous cases and interacting galaxies discussed by all three classifiers for final determination. Random checks on sub-samples of galaxies were performed by SJ. We denote as f_VC the fraction of intermediate-mass (M_* ⩾ 10⁹ M_☉) systems with different visual classes (e.g., mergers of Types 1, 2a, and 2b, and non-interacting systems of type Symmetric and Irr-1). In order to gauge the subjectivity inherent in any visual classification scheme, we estimate the maximum difference in the value of f_VC across the classifiers (δf_VC). The maximum percentage difference P defined as (100 × δf_VC/f_VC) is 4%, 19%, and 16% respectively for the visual classes of Non-interacting Symmetric, Non-interacting Irr-1, and Mergers. We conservatively adopt a value of 20% for P on the mean merger fraction, as a measure of the inherent subjectivity in the visual classification. When citing the final error bar on f_VC, we combine in quadrature the term δf_VC, which represents the subjectivity across classifiers, and the binomial error term $(\sqrt{f_{\rm VC}(1 - f_{\rm VC})/N})$, where N is the sample size of 886 systems. The raw values for f_VC are shown in Table 1.

It is also interesting to explore any systematic effects affecting the mergers we identified visually. The criteria we used, based on morphological distortions, might beg the question of whether these mergers are biased toward a gas-rich population. The extent to which a galaxy develops morphological distortions in an interaction depends on whether the encounter is prograde or retrograde, the interaction phase, the gas content, and the mass ratio (M₁/M₂). Di Matteo et al. (2007) found that the most influential parameters are the mass ratio and orbital geometry, and not the gas content alone. The visibility timescale (t_vis) on which the morphological signatures persist also depend on the orbital geometry, gas content, and the mass ratio. We do not have a direct measurement of the gas content of galaxies in A901/902, so we attempt an indirect first-order test in order to verify whether our methodology preferentially picks gas-rich mergers. We visually classify sample galaxies into three broad classes: "highly smooth," "highly clumpy," and "other." If a system shows a high degree of clumpiness (i.e., very patchy regions that are likely associated with significant amounts of gas, dust, and SF), it is classified as "highly clumpy." Conversely, systems with a very smooth non-clumpy appearance are classified as "highly smooth."

Out of the 20 mergers and 866 non-interacting galaxies, we find that 6/20 (30%) and 210/866 (24%) are classified as "highly clumpy," respectively. The comparable fraction of "highly clumpy" systems among non-interacting systems and mergers suggests that our method for identifying mergers does not show a strong bias toward preferentially selecting mergers, which are highly clumpy, gas-rich, and star forming. It is also interesting to recall here that our visual classification scheme takes special care to distinguish between Non-interacting Irr-1 and mergers (Section 3.2). The Non-interacting Irr-1 galaxies were defined as those where internally triggered small-scale asymmetries, often due to SF, are present. Thus, as expected, if we separate our non-interacting galaxies into Irr-1 and Symmetric, we find that a large fraction of the Non-interacting Irr-1 (111/123 or 90%) are "highly clumpy," compared to 99/713 (14%) of the Non-interacting Symmetric systems.

Using the CAS code (Conselice et al. 2000; Conselice 2003), quantitative structural parameters measuring the concentration (C), asymmetry (A), and clumpiness (S) were derived for the supercluster galaxies. CAS was run on the F606W images and the segmentation maps produced during the original source extraction (Caldwell et al. 2008) were used to mask neighboring galaxies.

The CAS concentration index C (Bershady et al. 2000) is proportional to the logarithm of the ratio of the 80%–20% curve of growth radii within 1.5 times the Petrosian inverted radius at r(η = 0.2):

$$\begin{equation} C = 5 \times {\rm log}\left(\!\frac{r_{80\%}}{r_{20\%}}\!\right). \end{equation} \tag{ 3 }$$

The CAS code derives the asymmetry index A (Conselice 2003) by rotating a galaxy image by 180°, subtracting the rotated image from the original image, summing the absolute intensities of the residuals, and normalizing the sum to the original galaxy flux. CAS improves the initial input center with the IRAF task "imcenter" and then performs a further refinement within a 3 × 3 grid, picking the center that minimizes A.

The clumpiness parameter S is a quantitative measure of the high spatial frequency patchiness of a galaxy. It is defined as the summation of the difference of the galaxy's original flux and the flux of an image with the high frequency structures muted by smoothing over a filter on order of the clumpiness, taking into account the background flux over that same area. This measurement is then divided by the summation of the original flux of the galaxy to obtain the clumpiness parameter S.

Conselice et al. (2000) argue that the CAS merger criterion (A > S and A > 0.35) at λ_rest > 4500 Å can be used to capture galaxies which have strong asymmetries indicative of major mergers. However, calibrations of the asymmetry value with N-body simulations (Conselice 2006; Lotz et al. 2008) have shown that for galaxies involved in major mergers of mass ratios 1:1–1:3, the asymmetries vary as a function of time during the merger. The A value reaches a peak near the midpoint of the merger and falls off both before and after this time to less than 0.35. The criterion A > 0.35 is fulfilled only ∼1/3 of the time for major mergers in these simulations, while minor mergers of mass ratios below 1:5 have A values significantly lower than 0.35.

Furthermore, in intermediate-mass (M_* ⩾ 10⁹ M_☉) field galaxies at z∼ 0.24–0.80, recent work (Jogee et al. 2008, 2009; Miller et al. 2008) has shown that the CAS criterion typically only picks up 50%–70% of galaxies visually typed as being disturbed or interacting. These results caution against using CAS without complementary visual classifications in field galaxies.

The effectiveness of the CAS criterion has not been explored in cluster environments where the dominant galaxy population is more gas-poor than in the field. In Section 4.2, we present the merger fraction (f_CAS) from CAS and perform one of the first systematic comparisons to date between CAS-based and visual classification results in clusters.

Next we discuss how to define and estimate the merger fraction. Our goal is to estimate the fraction of f_merge of systems with stellar mass above an appropriately chosen mass cut M_cut, which are involved in mergers of mass ratio ⩾1/10. The merger fraction f_merge is computed as (N_merge/N_tot), where N_merge is the number of major and minor mergers involving galaxies with M_* ⩾ M_cut, and N_tot is the total number of galaxies with M_* ⩾ M_cut.

It is important to determine the minimum stellar masses of the major and minor mergers involving galaxies with M_* ⩾ M_cut and thereby assess for what value of M_cut we can trace such mergers. Major merger pairs (defined as having mass ratio 1/4 <M₁/M₂⩽ 1) of mass ratio 1:1 to 1:3 will have minimum stellar masses ranging from 2 × M_cut to 4 × M_cut, in cases where galaxies of mass M_* ⩾ M_cut merge with systems at least as massive as themselves. However, if galaxies of mass M_* ⩾ M_cut merge with lower mass systems, the minimum mass of 1:1 to 1:3 mergers will range from 2 × M_cut to (4 × M_cut/3). Taken together, the above constraints imply that we are complete for major mergers involving galaxies with M_* ⩾ M_cut, as long as we can trace 1:3 major mergers with a minimum stellar mass of (4 × M_cut/3).

Repeating the exercise for minor mergers (defined as having 1/10 <M₁/M₂⩽ 1/4), it follows that the minimum stellar mass for 1:4 to 1:9 mergers ranges from 5 × M_cut to 9 × M_cut or from (5 × M_cut/4) to (10 × M_cut/9), depending on whether galaxies of mass M_* ⩾ M_cut merge with systems of higher mass or lower mass. Thus, we are complete for minor mergers involving galaxies with M_* ⩾ M_cut, as long as we can trace 1:9 minor mergers with a minimum stellar mass of (10 × M_cut/9).

For what value of M_cut are these conditions satisfied by the mergers of Types 1, 2a, and 2b, which we visually identified in the final sample (Section 3.2)? Since our final sample is complete for M_* ⩾ 10⁹ M_☉, it follows that we can identify all mergers of Types 1 and 2a with M_* ⩾ 10⁹ M_☉ from this sample. If we impose this value to the afore-defined criteria of (10 × M_cut/9) and (4 × M_cut/3), it implies that a mass cut M_cut ⩾ 0.9 × 10⁹ M_☉ would allow us to be complete for major and minor mergers of Types 1 and 2a. However, the situation is different for the mergers of Type 2b (close pairs resolved into two galaxies by COMBO-17) because the individual galaxies making up the pair are complete only for M_* ⩾ 10⁹M_☉. As a result, we can only completely trace 1:3 major mergers of Type 2b with total mass ⩾4.0 × 10⁹ M_☉. If we impose this value to the afore-defined criterion of (4 × M_cut/3), it implies that a mass cut of M_cut ⩾ 3.0 × 10⁹ M_☉ is needed to ensure completeness for major mergers of Type 2b. Similarly, we can only completely trace 1:9 minor mergers of Type 2b with total mass ⩾9.0 × 10⁹ M_☉, which implies that a mass cut M_cut ⩾ 9.0 × 10⁹ M_☉ is needed to completely trace all minor mergers of Type 2b.

Using a mass cut M_cut ⩾ 9.0 × 10⁹ M_☉ for computing the merger fraction f_merge would allow us to be complete for major and minor mergers of Types 1, 2a, and 2b. However, it leads to very small number statistics and is not viable. We therefore explore the value of f_merge for the less severe cut of M_cut ⩾ 3.0 × 10⁹ M_☉, as well as for M_cut ⩾ 10⁹ M_☉. The cut of M_cut ⩾ 3.0 × 10⁹ M_☉ ensures that we are complete for major and minor mergers of Types 1, 2a, but leaves us incomplete for mergers of Type 2b. It is encouraging that both cuts yield similar values for the merger fraction f_merge, as illustrated in Table 3: the merger fraction is 0.023 ± 0.007 and 0.021 ± 0.007, respectively, for M_cut ⩾ 10⁹ M_☉ and M_cut ⩾ 3.0 × 10⁹ M_☉. Note that in computing the merger fraction, we only include the 20 distorted mergers listed in Table 2, and avoid the potential projection pairs of Types 2a and 2b without signs of morphological distortions.¹⁹

While mergers may have played an important role in the evolution of cluster galaxies at earlier times (e.g., z > 2), hierarchical models (e.g., Gottloeber et al. 2001; Khochfar & Burkert 2001) predict that the merger fraction in dense clusters falls more steeply at z < 1 than the field merger fraction. As a result, at z < 0.3, the merger fraction for intermediate-mass cluster galaxies is predicted to be quite low (typically below 5%). The low merger fraction among intermediate-mass (M = 10⁹ to a few ×10¹⁰ M_☉) galaxies in the A901/902 clusters is consistent with the latter prediction.

How are the mergers distributed among major mergers, minor mergers, and ambiguous cases that could be either major or minor mergers? The results are shown in Columns 8–10 of Table 3, based on the classification listed in Column 8 of Table 2. The estimated fractions of likely major mergers, likely minor mergers, and ambiguous cases are 0.01 ± 0.004% (9/886), 0.006 ± 0.003% (5/886), and 0.007 ± 0.003% (6/886), respectively for M_cut ⩾ 10⁹ M_☉. For M_cut ⩾ 3.0 × 10⁹ M_☉, the corresponding fractions are 0.013± 0.005% (8/609), 0.008 ± 0.004% (5/609), and 0.0% (0/609), respectively.

In the rest of this paper, we continue to work with a mass cut of M_cut ⩾ 10⁹ M_☉, and the 20 distorted mergers (Table 2) applicable for this mass cut. However, where relevant, we cite many of our results for both of the mass cuts (M_cut ⩾ 10⁹ M_☉ and M_cut ⩾ 3.0 × 10⁹ M_☉) so that we can gauge the potential effect of incompleteness in tracing major and minor mergers.

The results of running CAS on the final classified sample of 886 intermediate-mass (M_* ⩾ 10⁹ M_☉) systems are shown in Table 4 and Figures 6 and 7. In Figure 6, the 20 mergers of Type 1, 2a, and 2b are plotted in different symbols. For Type 2b merger pairs, we plot the highest value of CAS A found between the two galaxies in each system. Using the CAS merger criterion (A > 0.35 and A > S) to identify mergers yields a merger fraction (f_CAS) in this sample of 18/886 or 0.02 ± 0.006. For the more conservative sample with a mass cut M_cut ⩾ 3 × 10⁹ M_☉, f_CAS is 7/609 or 0.011 ± 0.005.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

When citing the error on f_CAS, we take the largest of either the Poisson error or the systematic error (σ_CAS) in f_CAS due to the systematic errors in CAS A and S. The systematic error, σ_CAS, is calculated by taking the upper and lower bounds of f_CAS based on (A± error in A) and (S± error in S). Specifically, these limits on f_CAS are found by using the criteria of ((A± error in A) < (S± error in S)) and ((A± error in A) > 0.35).

At first sight, the CAS-based merger fractions f_CAS for the two mass cuts (0.020 ± 0.006 and 0.011 ± 0.005 for M_cut ⩾ 10⁹ M_☉ and M_cut ⩾ 3 × 10⁹ M_☉, respectively) are not widely different from the merger fraction f_merge based on visual classification (0.023 ± 0.007 and 0.021 ± 0.007 for M_cut ⩾ 10⁹ M_☉ and M_cut ⩾ 3 × 10⁹ M_☉, respectively; Section 4.1). However, the comparison of f_CAS and f_merge does not tell the whole story because the nature of the systems picked by the two methods can be quite different. The visual classes of the 18 systems, which satisfy the CAS criterion and are considered as mergers in the CAS system, are shown in Table 4. It turns out 7/18 (39 ± 14%) of these "CAS mergers" are visually classified as non-interacting systems. The results are illustrated in Table 4 and Figure 6. Figure 7 shows examples of these "contaminants": they tend to be dusty highly inclined systems and systems with low level asymmetries that seem to be caused by SF.

It is also useful to ask what fraction of the 20 systems visually identified as mergers satisfy the CAS criterion. We find that for M_cut ⩾ 10⁹ M_☉, the CAS criterion only captures 11 of the 20 (55 ± 16%) of the visually classified mergers. These results are illustrated in Figure 6. Figure 7 shows examples of merging galaxies missed by CAS. The missed mergers include galaxies with fainter outer tidal features, double nuclei where CAS puts the center between the nuclei, and pairs of very close connected galaxies.

It is also interesting to ask what percentage of the different types of mergers (Types 1, 2a, and 2b) does the CAS merger criterion recover. Figure 6 also shows the mergers divided up by merger Types 1, 2a and 2b. Of the 13 mergers of Type 1, CAS recovers 9/13 or 69% ± 18%. Of the three mergers of Type 2a, CAS recovers 1/3 or 33 ± 28%. Of the four mergers of Type 2b, CAS recovers 1/4 or 25% ± 22%.

Since the CAS criterion is widely used to pick major mergers, it is also interesting to explore how well this criterion picks up the systems that we visually classified as major mergers (Table 2). We find that the CAS merger criterion picks up 6/9 (67%± 20%) of the systems classified as major mergers. It is also interesting to note that the CAS criterion recovers 2/5 (40% ± 23%) and 4/6 (67% ± 23%) of the systems classified respectively as minor mergers and ambiguous mergers.

In order to define different regions of the A901a, A901b, and A902 clusters, we computed the projected number density n (Figure 8) for intermediate-mass (M_* ⩾ 10⁹ M_☉) galaxies as a function of clustocentric radius R by assuming a spherical distribution. Note here that each galaxy is assigned to the cluster closest to it, and R is measured from the center of that cluster.

Zoom In Zoom Out Reset image size

We consider the cluster core to be at R⩽ 0.25 Mpc, as this is the region where the projected number density n rises very steeply (Figure 8). The cluster virial radii are taken to be ∼1.2 Mpc, based on estimates from the DM maps derived from gravitational lensing by Heymans et al. (2008). Throughout this paper, we refer to the region at 0.25 Mpc <R⩽ 1.2 Mpc, between the cluster core and the cluster virial radius, as the "outer region of the cluster." The region outside the virial radius (1.2 Mpc <R⩽ 2.0 Mpc) is referred to as the "outskirt region of the cluster." The core, outer region, and outskirt region are labeled in Figure 8.

As shown in Table 5, the A901 clusters have central galaxy number densities (n) in the core region of 1000–1600 galaxies Mpc⁻³. These are lower than that of the rich Coma cluster and higher than that of the Virgo cluster.

Figure 9 shows the spatial distribution of the mergers visually identified among the sample of intermediate-mass (M_* ⩾ 10⁹ M_☉) galaxies in the A901a, A901b, and A902 clusters. We again consider here only the 20 distorted mergers listed in Table 2, for the reasons discussed in Section 4.1. We find that all the 20 mergers lie outside the cluster cores. We estimate that the number density (n_merge) of mergers is 0, 2.09, and 0.11 galaxies Mpc⁻³, respectively, in the core, outer region, and outskirt of the cluster (Table 6). When estimating n_merge in the three regions, we divided the number of mergers by the volume of the core region, and the volume of a spherical shell in the outer region and outskirt of the clusters.

Zoom In Zoom Out Reset image size

For mergers and non-interacting galaxies, we compute the minimum distance to the nearest cluster center (A901a, A901b, and A902), and the local values of various environmental parameters, such as the local galaxy surface density Σ₁₀, the local DM mass surface density κ, and the relative local ICM density. Following Wolf et al. (2005) and Gilmour et al. (2007), we define and compute the local galaxy surface density (Σ₁₀) as the number of galaxies per (Mpc/h)⁻² that are in an annulus bracketed by two circles whose radii are equal to the average distances to the 9th and 10th nearest neighbor. The DM mass surface density (κ) is computed locally using the coordinates of each galaxy on the weak lensing map provided by Heymans et al. (2008). The relative local ICM counts are measured at each galaxy position using the X-ray map from Grey et al. (2010, in preparation) and Gilmour et al. (2007). Figure 10 shows these local environmental parameters as a function of the minimum distance to the cluster center for mergers and non-interacting galaxies. Since the mergers lie outside the cluster core (0.25 Mpc <R⩽ 2 Mpc), they are associated with low values of κ and intermediate values of Σ₁₀, and ICM density (Figure 10).

Zoom In Zoom Out Reset image size

The timescale for collisions and close encounters is given by

$$\begin{equation} t_{\rm coll} = \frac{1}{n \sigma _{\rm gal} A}, \end{equation} \tag{ 4 }$$

where n is the galaxy number density, σ_gal is the galaxy velocity dispersion, A is the collisional cross section (πf(2r_gal)²), r_gal is a typical galaxy radius, and f is the gravitational focusing factor (Binney & Tremaine, 1987). Since the average velocity dispersion in clusters is high, the value for f will be low and we will assume it to be on the order of unity. Thus,

$$\begin{eqnarray} t_{\rm coll}&\sim& 800 \times \left(\!\frac{n}{10^{3} \rm \, {\rm Mpc}^{-3}}\!\right)^{-1}\! \times \left(\frac{\sigma }{10^{3}\,{\rm km\,s}^{-1}}\right)^{-1} \nonumber\\ &&\times \left(\!\frac{{r_{\rm gal}}}{10 \, {{\rm kpc}}}\!\right)^{-2}\ {\rm Myr}. \end{eqnarray} \tag{ 5 }$$

We compute n using the same method as in Section 4.3 for the core, and use spherical shells to compute n in the outer region and cluster outskirt for our sample of R_Vega⩽ 24, intermediate-mass (M_* ⩾ 10⁹ M_☉) galaxies. The local galaxy velocity dispersion profiles for A901a/b, A902, and South West Group (SWG) from kinematic modeling using ∼420 2dF redshifts (M. E. Gray et al. 2010, in preparation) are shown in Figure 11, and the average velocity dispersions σ_gal are shown in Table 5. The central galaxy velocity dispersion within the cores (R < 0.25 Mpc) of A901a,b and A902 typically range from 700 to 1000 km s⁻¹. Beyond the cluster core, in the outer region (0.25 Mpc <R⩽ 1.2 Mpc), the small number statistics leads to large error bars on the galaxy velocity dispersion, making it not viable to determine whether it remains high, drops, or rises. To estimate timescales, we take the average σ_gal for A901/902 to be ∼800 km s⁻¹ in three regions of the cluster. The timescales (t_coll) for close encounters are shown in Table 6 and range from 0.765 to 13.5 to 135 Gyr from the core (R⩽ 0.25 Mpc), to the outer region (0.25 Mpc <R⩽ 1.2 Mpc), to the outskirt region (1.2 Mpc <R⩽ 2.0 Mpc) of the cluster.

Zoom In Zoom Out Reset image size

It may at first seem surprising that mergers do not preferentially lie in the cluster core, where the probability for close encounters is high due to the associated short value of the timescale t_coll for collisions and close encounters. However, if σ_gal is large, a close encounter is unlikely to lead to a merger or a large amount of tidal damage (i.e., to strong morphological and kinematical distortions). This is because a large galaxy velocity dispersion will likely cause the energy E of the reduced particle in a binary encounter to be positive, causing the orbit to become unbound (Binney & Tremaine, 1987).

All merging systems lie at a projected clustocentric radius of 0.25 Mpc <R⩽ 2 Mpc, between the core and cluster outskirts, although in this region the timescale for collisions and close encounters is quite large. We suggest two possible reasons why strong interactions and mergers might populate the outer region and cluster outskirt. The first possibility is that the velocity dispersion of galaxies falls between the core and outskirt, due to the intrinsic velocity dispersion profile of the clusters. Many clusters show a flat velocity dispersion profile from 1 Mpc outward, but several also show a declining profile (den Hartog & Katgert 1996, Carlberg et al. 1997). In the latter cases, the velocity dispersions can fall from 1500 to 100 km s⁻¹ from the cluster core to cluster outskirt. Since the local velocity dispersion profiles in the regions of interest in A901/902 (Figure 11) are based on a small number of galaxies and therefore large errors in σ_gal, we cannot determine if the velocity dispersion profile falls or remains flat. An additional difficulty is that the member galaxies of the neighboring clusters influence σ_gal at ∼0.5 Mpc for A901a/b and ∼1 Mpc for A902.

The second possibility is that the mergers are part of groups or field galaxies that are being accreted along filaments by the A901/902 clusters. This would be in line with the idea that clusters continuously grow by mergers and accretion of groups (e.g., Zabludoff & Franx 1993; Abraham et al. 1996; Balogh et al. 2000). Groups have lower galaxy velocity dispersion (typically below 300 km s⁻¹) than clusters, as shown in Table 6. Furthermore, simulations show that merger rates are highest as field and group galaxies accrete into cluster along cosmological filaments (see below; van Kampen & Katgert 1997; Figure 12).

Zoom In Zoom Out Reset image size

To further quantify the possibility of groups accreting into the A901/A902 clusters, we compare the data to model predictions from simulations of the STAGES A901/A902 supercluster (E. van Kampen et al. 2010, in preparation). The simulation aims to reproduce the environment as observed for A901/A902 by using constrained initial conditions that produce clusters with overall properties similar to those of A901, A902, and the neighboring A907 and A868. The simulation box is large enough to contain all these, and hence reproduce the overall large-scale density and velocity fields similar to what is observed. A phenomenological galaxy formation model is then used to simulate the galaxy population in the same box, allowing us to select mock galaxies above a certain luminosity or stellar mass cut, as done in the observed data set. A stellar mass cut of M_*⩾ 10⁹ M_☉ is applied in the simulations, corresponding to the mass cut in the observational sample. The solid lines in Figure 12 show the predicted number density (n) and fraction (f) of galaxy mergers with mass ratio M₁/M₂ > 1/10, as a function of environment, as characterized by the local overdensity (δ^G). The latter is calculated by smoothing the density of DM halos with a Gaussian of width 0.4 Mpc to take out the effect of individual galaxies. Typical values of δ^G are ∼10–100 for group overdensities, ∼200 at the cluster virial radius, and ≳1000 in the core of rich clusters.

Typically, in the simulations, as field and group galaxies fall into a cluster along filaments, the coherent bulk flow enhances the galaxy density and causes galaxies to have small relative velocities, thus leading to a high probability for mergers at typical group overdensities. Closer to the cluster core, galaxies show large random motions rather than bulk flow motion, leading to a large galaxy velocity dispersion, and a sharp drop in the probability of mergers.

In order to compare the data to simulation results, we use the number (N_merge) and fraction (f_merge) of mergers of mass ratio ⩾1/10, which we computed in Section 4.1. The solid curve in Figure 12 shows the merger fraction and number density predicted by the simulations as a function of overdensity. The three dashed lines in Figure 12 show the estimated range in the observational number density (n_merge) and fraction (f_merge) of mergers in the three different regions (the core, the outer region, and the outskirt) of the A901/902 clusters, as defined in Section 4.3. The points at which the dashed lines cross or approach the solid curve tell us the typical overdensities at which we expect to find such merger fraction or merger number densities in the simulations. It can be seen that the low merger density seen in the core region of the cluster corresponds to those expected at typical cluster core overdensities. On the other hand, the larger merger fraction we observe between the cluster core and outer region (0.25 Mpc <R⩽ 2 Mpc) is close to those seen in typical group overdensities. Our results are in agreement with the above scenario, whereby the accretion of group and/or field galaxies along filaments, where they have low relative velocities and enhanced overdensities, leads to enhanced merger rates. We note that similar conclusions are reached if we perform the comparisons between data and simulations using the more conservative mass cut of M_* ⩾ 3 × 10⁹ M_☉, discussed in Section 4.1.

However, there are a couple of caveats when comparing our data with simulations. One caveat is that we directly compare the projected values of the number density or fraction of mergers from two-dimensional observations, to the predicted number density of mergers from three-dimensional simulations. Projection effects can introduce uncertainties in our observational estimate. The magnitude of the uncertainties due to projection effects depends on the detailed environment and will be investigated in the full STAGES supercluster simulation data set (E. van Kampen et al. 2010, in preparation).

A further indirect evidence for group accretion stems from comparing semi-analytic galaxy catalogs to STAGES observations. R. Rhodes et al. (2010, in preparation) find an overabundance in galaxies in A902 compared to its mass. This could be explained by two or more galaxy groups in projection, consistent with the idea of group accretion.

We first recapitulate our results on the visually based merger fraction (f_merge) in the A901/902 clusters (Section 4.1). Among intermediate-mass (M_* ⩾ 10⁹ M_☉) systems, we find that the fraction f_merge of systems which show evidence of a recent or ongoing merger of mass ratio >1/10 is 0.023 ± 0.007 (Table 3). Most of these mergers have M_V ∼ −19 to −22 and L ⩽ L* (see Figure 1). We also estimated that the fraction of likely major mergers, likely minor mergers, and ambiguous cases to be 0.01% ± 0.004% (9/886), 0.006% ± 0.003% (5/886), and 0.007% ± 0.003% (6/886), respectively.

Next, we compare our merger fraction in the A901/902 clusters with the published merger fraction in other clusters and group galaxies out to z∼ 0.8, over the last 7 Gyr (Figure 13). These comparisons are difficult to make as different studies apply different luminosity or mass cuts. Furthermore, some studies consider only major mergers, while others consider all interacting galaxies, which are likely candidates for both major and minor interactions.

Zoom In Zoom Out Reset image size

The variation in galaxy populations sampled at different epochs must also be kept in mind when comparing results at lower redshift with those out to z∼ 0.8. Low redshift samples typically sample a small volume and therefore tend to host only a small number of the brightest and most massive systems. Conversely, higher redshift magnitude-limited samples will suffer from Malmquist bias and tend to under-represent the faint (L < L*) galaxies.

There have been several studies of galaxy mergers and interactions in intermediate redshift clusters (Lavery & Henry 1988; Lavery et al. 1992; Dressler et al. 1994; Oemler et al. 1997; Couch et al. 1998), but only few to date with high-resolution HST imaging. Couch et al. (1998) used WFPC/WFPC2 observations and spectroscopy of two clusters at z ∼ 0.3 and ∼0.4, focusing on galaxies with R ∼ 22.5 or L < L*. In their study, they classified ∼7/200 or ∼3.5% galaxies to be merging based on separations and visible distortions. The merger fractions of these two intermediate redshift clusters are plotted in Figure 13.

It is also interesting to note that the merging systems in these two intermediate redshift clusters tend to be blue and preferentially located in the outskirt of the clusters. In fact, in the Couch et al. (1998) sample, ∼60% of the merging galaxies among L < L* systems are blue. The mergers in these two intermediate redshift clusters are similar in many ways to the mergers in the A901/902 clusters: the latter mergers lie in between the cluster core and outskirt (0.25 Mpc <R⩽ 2 Mpc; Section 4.3) of the A901/902 clusters, and for the absolute magnitude range of M_V ∼ −19 to −22, 16/20, or 80% ± 18% of these mergers lie on the blue cloud (Section 4.6).

In a study by Dressler et al. (1994), based on WFPC images of a z ∼ 0.4 cluster, the fraction of bright (M_V < −18.5) galaxies identified as mergers was 10/135. Another WFPC study of four clusters at z = 0.37–0.41 by Oemler et al. (1997) placed a lower limit of 5% on the fraction of merging galaxies with M_V < −19 or L ⩽ L*.

A high-redshift (z = 0.83) cluster was studied by van Dokkum et al. (1999) using HST WFPC2 images and spectroscopic redshifts for 80 confirmed members. They found a high fraction (17%) of luminous (M_B ∼ −22; L ∼ 2L*) systems, which exhibit signatures of ongoing mergers. These mergers were found to be red, bulge-dominated, and located in the outskirts of the cluster. A later study by Tran et al. (2005) used a spectroscopic follow-up to confirm that 15.7% ± 3.6% of these galaxies were bound pairs with separations ≲30 hr⁻¹ kpc and relative velocities ≲300 km s⁻¹. These merging pairs are located outside the cluster centers, similar to the mergers in A901/902. However, the galaxy sample at high redshift is dominated by more luminous galaxies (M_B ∼ −22; L ∼ 2L*) than the A901/902 sample where most galaxies have M_V ∼ −19.5 to −22 (Figure 1). The number statistics for bright (∼2L*) mergers in A901/902 are not robust enough to allow a direct comparison with the merger fraction cited in van Dokkum et al. (1999).

Finally, we look at the merger fraction in groups. A study by McIntosh et al. (2008) finds that ∼1.5% of massive (M ⩾ 5 × 10¹⁰ M_☉) galaxies in SDSS groups (with M_halo > 2.5 × 10¹³ M_☉) are major mergers at z of 0.01–0.12. Most of the mergers involve two red sequence galaxies and they are located between 0.2 and 0.5 Mpc from the group center. A study by McGee et al. (2008) also finds an enhancement in asymmetric bulge-dominated galaxies in groups, consistent with a higher probability for merging in the group environment. Additional evidence for mergers in groups is shown in a study of a supergroup at z ∼ 0.37 by Tran et al. (2008), who report dry dissipationless mergers and signatures thereof in three of four brightest group galaxies.

A merger fraction of 6% was found by Zepf (1993) in Hickson compact groups at z < 0.05, among systems with luminosities L ⩽ L*. The merger fraction is significantly higher than that in SDSS groups. The increased merger fraction is expected since Hickson compact groups are different from loose groups: they have high number densities comparable to those in cluster cores, but low galaxy velocity dispersions. These two conditions provide an environment most favorable to strong tidal interactions and mergers, as argued in Section 4.4.

It is not straightforward to draw conclusions about the evolution of the merger fraction in cluster galaxies over z∼ 0.05–1.0 from the above studies because they sample different types of systems and different luminosity ranges. If we conservatively restrict ourselves to only studies of L ⩽ L* cluster galaxies, then we have four data points over z∼ 0.17–0.4, shown as solid filled circles in Figure 13. The mean value of the data points allows for evolution by a factor of ∼3.2, in the merger fraction of L ⩽ L* cluster galaxies over z ∼ 0.17–0.4. However, if one uses the full range of merger fractions allowed by the error bars on the data points, then we can admit a wider spectrum of scenarios, ranging from no evolution to evolution by a factor of ∼5 over z∼ 0.17–0.4. Having additional deep, large-volume, high-resolution studies, which are based on larger samples with smaller error bars would help to separate between these scenarios, and thereby test hierarchical models of galaxy evolution.

As mentioned in Section 4.1, hierarchical models (e.g., Gottloeber et al. 2001; Khochfar & Burkert 2001) predict that the merger fraction in dense clusters falls more steeply at z < 1 than the field merger fraction, such that at z < 0.3, the predicted merger fraction is quite low (typically below 5%) among intermediate-mass cluster galaxies. The low merger fractions among intermediate-mass (10⁹ to a few ×10¹⁰ M_☉) or intermediate luminosities (L < L*) galaxies in the A901/902 clusters and other low-redshift clusters are consistent with these predictions, but we cannot yet test the predicted rate of evolution of the merger fraction in clusters with redshfit.

We explore the properties of galaxies on the blue cloud and in the red sample among the final sample of 886 systems (20 mergers and 866 non-interacting galaxies; Section 3.2). The red sample was defined in Wolf et al. (2009) and contains both passively evolving spheroidal galaxies on the red sequence, as well as dusty red galaxies that lie between the red sequence and the blue cloud.

Figure 14 shows the rest-frame U − V color plotted against stellar masses for galaxies of different visual classes: Mergers, Non-interacting Irr-1, and Non-interacting Symmetric (Section 3.2). The visual classes of galaxies on the blue cloud and in the red sample are shown in Table 7. As described in Section 4.1, we only consider the 20 distorted mergers listed in Table 2, and avoid the potential projection pairs without signs of morphological distortions. The 20 mergers are divided into 13 mergers of Type 1, 3 mergers of Type 2a, and 4 mergers of Type 2b. For mergers of Type 2b, which are resolved into two galaxies with separate COMBO-17 colors, we plot the average U − V color of the galaxies in the pair.

Zoom In Zoom Out Reset image size

Out of the 886 visually classified systems in our sample, we find that 310/886 or 35% ± 7% lie on the blue cloud. Out of the 20 visually classified mergers with distortions, 16/20 or 80% ± 18% lie on the blue cloud (Table 7). Conversely, out of 866 non-interacting galaxies, 294/866 or 34% ± 7% are on the blue cloud. Thus, the fraction of mergers, which lie on the blue cloud (f_merg-BC) is over two times larger than the fraction of non-interacting galaxies (f_non-int-BC), which lie on the blue cloud. This implies that mergers and interacting galaxies are preferentially blue, compared to non-interacting galaxies.

The observed higher fraction of blue galaxies among mergers compared to non-interacting galaxies may be caused by several factors. It may be due to enhanced levels of unobscured SF in mergers (see Section 4.7), translating to bluer colors on average. It may also in part be the result of the mergers being part of accreted field and group galaxies (Section 4.4), which are bluer than the average cluster galaxy. It is also relevant to ask whether we are overestimating the fraction of blue galaxies among interacting systems due to the visibility timescale t_vis of morphological distortions induced by interactions being longer in bluer galaxies than redder ones. While this is possible, it is non-trivial to correct for this effect because no direct unique relation exists between t_vis and color. As discussed in Section 4.1, t_vis depends on the mass ratio of an interaction as well as the gas content. A higher gas content may result in a longer t_vis for certain gas distributions, but it can lead to either redder colors (e.g., enhanced level of dusty SF) or bluer colors (enhanced level of unobscured SF), compared to interacting systems.

This work uses SFRs based on UV data from COMBO-17 (Wolf et al. 2004) and Spitzer 24 μm imaging (Bell et al. 2007). The unobscured SFR_UV is derived using the 2800 Å rest-frame luminosity (L_UV = 1.5νl_ν,2800) as described in Bell et al. (2005, 2007). The UV spectrum is dominated by continuum emission from massive stars and provides a good estimate of the integrated SFR from the younger stellar population in the wavelength range of 1216–3000 Å. The SFR_IR is derived using the 24 μm flux to construct the integrated IR luminosity (L_IR) over 8–1000 μm following the methods of Papovich & Bell (2002). The total SFR is derived using identical assumptions of Kennicutt (1998) from PEGASE assuming a 100 Myr old stellar population with constant SFR and a Chabrier (2003) IMF:

$$\begin{equation} {\rm SFR}_{\rm UV + IR} = 9.8 \times 10^{-11}(L_{\rm IR} + 2.2L_{\rm UV}). \end{equation} \tag{ 6 }$$

The factor of 2.2 on the UV luminosity accounts for light being emitted longward of 3000 Å and shortward of 1216 Å by young stars. The total SFR accounts for both the dust-reprocessed (IR) and unobscured (UV) SFs.

We work with our final sample of 886 classifiable bright massive (M_* ⩾ 10⁹ M_☉) systems, and only consider here the 20 distorted mergers in Table 2. All of these systems have UV-based SFR from COMBO-17 observations. Of this sample, ∼11% (94/886) were not observed with Spitzer, ∼23% (206/886) were observed and detected at 24 μm, while the rest had no detection at the ∼4σ depth of 83 μJy.

The UV-based SFR (SFR_UV) versus stellar mass is plotted in Figure 15 for all 886 systems. The SFR_UV ranges from ∼0.01 to 14 M_☉ yr⁻¹. The UV-based SFR represents a lower limit to the total SFR for galaxies on the blue cloud and most star-forming galaxies on the red sample. However, for some old red galaxies, the SFR_UV may overestimate the true SFR as the UV light from such systems may not trace massive OB stars, but rather low- to intermediate-mass stars.

Zoom In Zoom Out Reset image size

Figure 16 shows the UV+IR-based SFR (SFR_UV+IR), which ranges from ∼0.2 to 9 M_☉ yr⁻¹. For the 206 galaxies that were observed and detected at 24 μm, the implied UV+IR-based SFR is plotted as stars in the lower panel of Figure 16. For the 586 galaxies that are observed but undetected with Spitzer, we use the 24 μm detection limit as an upper limit on the 24 μm flux. The corresponding upper limit on the UV+IR-based SFR is plotted as inverted triangles in the lower panel of Figure 16, and included in the calculation of the average UV+IR-based SFR, plotted in the middle panel.

Zoom In Zoom Out Reset image size

In a cluster environment, the competition between processes that enhance the SFR and those that depress the SFR, ultimately determine the average SFR of cluster galaxies. The first class of processes are strong close interactions: tidal and mergers (e.g., Toomre & Toomre 1972), and harassment (e.g., Moore et al. 1996), which refers to the cumulative effect of weak interactions. The second class of processes include ram pressure stripping of cold gas out of the galaxy by the hot ICM (e.g., Gunn & Gott 1972), and strangulation (e.g., Larson et al. 1980; Balogh et al. 2000).

For the few mergers (orange line) present, the average SFR_UV is typically enhanced by an average modest factor of ∼2 compared to the both Non-interacting Symmetric and Irr-1 galaxies (purple, green, and black lines). Similarly, the UV+IR-based SFR (SFR_UV+IR; and Figure 16) of merging galaxies is typically enhanced by only an average factor of ∼1.5 compared to the Non-interacting Symmetric galaxies (purple line) and to all Non-interacting galaxies (i.e., Symmetric + Irr-1; black line). We note that a similar modest enhancement in the average SFR, by a factor of 1.5–2 is also found in mergers in the field over z∼ 0.24–0.80 by Jogee et al. (2008, 2009). This modest enhancement is consistent with the theoretical predictions of di Matteo et al. (2007; see their Figure 10), based on a recent statistical study of several hundred simulated galaxy collisions. Modest SFR enhancements are also seen in galaxy pair studies in the field (Barton et al. 2000, 2003; Lin et al. 2004; Ellison et al. 2008) and in mixed environments (Robaina et al. 2009; Alonso et al. 2004).

While mergers in the A901/902 clusters enhance the SFR of individual galaxies, it is clear that they do not contribute much to the total SFR of the cluster. We compute the SFR density of mergers to non-interacting galaxies in the same volume of A901/902 by taking the ratio of the total SFR in each class. If we include only the 20 distorted mergers in Table 2, we find that the contribution of mergers to the SFR density of the clusters to be 10%. Alternatively, if we include all of the 36 visually classified mergers in Table 2, before accounting for false projection pairs, we find the contribution of mergers to the SFR density to be 15%. Thus, we find that mergers contribute only a small fraction (between 10% and 15%) of the total SFR density of the A901/902 clusters compared to non-interacting galaxies. The small contribution of mergers to the total cluster SFR density is likely due to the low number of mergers in A901/902, and the fact that these mergers only cause a modest SFR enhancement.