As galaxies orbit in the potential well of a galaxy cluster, stars are tidally stripped from their outer regions, mixing over time to form a diffuse "intracluster light" (ICL). First proposed by Zwicky (1951), the ICL has proved very difficult to study - at its brightest, it is only ~ 1% of the brightness of the night sky. Previous attempts to study the ICL have resulted in some heroic detections (Oemler 1973; Thuan & Kormendy 1977; Bernstein et al. 1995; Gregg & West 1998; Gonzalez et al. 2000), verified by observations of intracluster stars and planetary nebulae in Virgo (Feldmeier, Ciardullo, & Jacoby 1998; Ferguson, Tanvir, & von Hippel 1998; Arnaboldi et al. 2002). While the ICL is typically thought of as arising from the stripping of starlight due to the cluster potential, in fact the role of interactions between cluster galaxies in feeding the ICL is quite strong. Galaxy interactions significantly enhance the rate at which material is stripped; as illustrated in Figure 4, the strong, local tidal field of a close encounter can strip material from deep within a galaxy's potential well, after which the cluster tidal field can liberate the material completely. Interactions, particularly those in infalling groups, act to "prime the pump" for the creation of the ICL.
The properties of the ICL in clusters, particularly the fractional luminosity, radial light profile, and presence of substructure, may hold important clues about the accretion history and dynamical evolution of galaxy clusters. Material stripped from galaxies falling in the cluster potential is left on orbits that trace the orbital path of the accreted galaxy, creating long, low-surface brightness tidal arcs (e.g., Moore et al. 1996), which have been observed in a few nearby clusters (Trentham & Mobasher 1998; Calcáneo-Roldán et al. 2000). However, these arcs will only survive as discrete structures if the potential is quiet; substructure will dynamically heat these arcs, and the accretion of significant mass (i.e., a cluster merger event) may well destroy these structures. If much of the ICL is formed early in a cluster's dynamical history, before the cluster has been fully assembled, the bulk of the ICL will be morphologically smooth and well mixed by the present day, with a few faint tidal arcs showing the effects of late accretion. In contrast, if the ICL formed largely after cluster virialization, from the stripping of "quietly infalling" galaxies, the ICL should consist of an ensemble of kinematically distinct tidal debris arcs. Clusters that are dynamically younger should also possess an ICL with significant kinematic and morphological substructure.
Early theoretical studies of the formation of the ICL suggested that it might account for anywhere from 10% to 70% of the total cluster luminosity (Richstone 1976; Merritt 1983, 1984; Miller 1983; Richstone & Malumuth 1983; Malumuth & Richstone 1984). These studies were based largely on analytic estimates of tidal stripping, or on simulations of individual galaxies orbiting in a smooth cluster potential well. Such estimates miss the effects of interactions with individual galaxies (e.g., Moore et al. 1996), intermediate-scale substructure (Gnedin 2003), and priming due to interactions in the infalling group environment. As a result, these models underpredict the total amount of ICL as well as the heating of tidal streams in the ICL. Now, however, cosmological simulations can be used to study cluster collapse and tidal stripping at much higher resolution and with a cosmologically motivated cluster accretion history (e.g., Moore et al. 1998; Dubinski, Murali, & Ouyed 2001).
One example of the modeling of ICL is shown in Figure 5. These images are derived from the N-body simulations of Dubinski (1998), who simulated the collapse of a log(M / M) = 14.0 cluster in a standard cold dark matter Universe. Starting from a cosmological dark matter simulation, the 100 most massive halos are identified at a redshift of z = 2.2 and replaced with composite disk/bulge/halo galaxies, whereafter the simulation is continued to z = 0 (see Dubinski 1998 for more details). To quantify the diffuse light in these cluster models, we assign luminosity to the stellar particles based on a mass-to-light ratio of 1. The top panels show the cluster at two different times. On the left, the cluster is shown early in the collapse, at z = 2, where it consists of two main groups coming together. The right panels show the cluster at z = 0, when the cluster has virialized and formed a massive cD galaxy at the center. In each case, the lowest visible contour is at a surface brightness of µV 30 mag arcsec-2. The bottom panels show the effects of adding observational noise typical of our ICL imaging data (discussed below) and illustrate the difficulties in detecting this diffuse light.
Figure 5. Visualizations of the cluster simulations of Dubinski (1998). Top panels show the distribution of luminous starlight, with the faintest contours corresponding to a surface brightness of µV 30 mag arcsec-2. Bottom panels show the effect of adding noise characteristic of current observational limits. Left panels show the cluster early in collapse (at z = 2), while right panels show the virialized cluster at z = 0.
In the early stages of cluster collapse, material is being stripped out of galaxies and into the growing ICL component. This material has a significant degree of spatial structure in the form of thin streams and more diffuse plumes, much of it at observationally detectable surface brightnesses. At later times this material has become well mixed in the virialized cluster, forming a much smoother distribution of ICL and substructure that is visible only at much fainter surface brightnesses, well below current levels of detectability. Along these lines, the degree of ICL substructure may act as a tracer of the dynamical age of galaxy clusters.
Indeed, galaxy clusters do show a range of ICL properties. We (Feldmeier et al. 2002, 2003) have recently begun a deep imaging survey of galaxy clusters, aimed at linking their morphological properties to the structure of their ICL. As the detection of ICL is crucially dependent on reducing systematic effects in the flat fielding, we have taken significant steps to alleviate these issues, including imaging in the Washington M filter to reduce contamination from variable night sky lines, flat fielding from a composite of many night sky flats taken at similar telescope orientations, and aggressive masking of bright stars and background sources (see Feldmeier et al. 2002 for complete details). With this data, we achieve a signal-to-noise ratio of 5 at µV = 26.5 mag arcsec-2 and a signal-to-noise ratio of 1 at µV = 28.3 mag arcsec-2. We have targeted two types of galaxy clusters thus far: cD-dominated Bautz-Morgan class I clusters (Feldmeier et al. 2002) and irregular Bautz-Morgan class III clusters (Feldmeier et al. 2003).
Figure 6 shows results for the cD clusters Abell 1413 and MKW 7. Similar to the clusters studied by Gonzalez et al. (2000; see also Gonzalez, Zabludoff, & Zaritsky 2003), the cD galaxies are well fit by a r1/4 law over a large range in radius, with only a slight luminosity excess in the outskirts of each cluster. In each case, we search for ICL substructure by using the STSDAS ellipse package to subtract a smooth fit to the cD galaxy extended envelope. In the case of Abell 1413, we see little evidence for any substructure in the ICL; the small-scale arcs we observe are likely to be due to gravitational lensing. MKW 7 shows a broad plume extending from the cD galaxy to a nearby bright elliptical, but little else in the way of substructure.
Figure 6. Deep imaging of cD clusters from Feldmeier et al. (2002). Top panels show Abell 1413; bottom panels show MKW 7. The left panels show the full 10' × 10' view of the cluster, while the right panels show a close up of the clusters once a smooth elliptical fit to the cD cluster envelope has been removed. The oval shows the radius inside which the model has been subtracted.
In contrast, we see evidence for more widespread ICL substructure in our Bautz-Morgan type III clusters. Figure 7 shows our image of Abell 1914, binned to a resolution of 3'' after all stars and galaxies have been masked. Here we see a variety of features: a fan-like plume projecting from the southern clump of galaxies, another diffuse plume extending from the galaxy group to the east of the cluster, and a narrow stream extending to the northeast from the cluster center. The amount of substructure seen here is consistent with an unrelaxed cluster experiencing a merger, similar to the features seen in the unrelaxed phase of the model cluster shown in Figure 5. We see similar plumes in other type III clusters, suggesting that the ICL in these types of clusters does reflect a cluster that is dynamically less evolved than the cD-dominated clusters of Feldmeier et al. (2002).
While these studies point toward significant substructure in the ICL of galaxy clusters, imaging surveys continue to be hampered by systematic effects. With so much of the ICL substructure present only at surface brightnesses fainter than µV > 28 mag arcsec-2, issues of flat fielding, scattered light, and sky variability become severe. An interesting alternative is to use the significant numbers of intracluster planetary nebulae now being found in emission-line surveys of nearby galaxy clusters (Feldmeier et al. 1998; Arnaboldi et al. 2002). These studies have very different detection biases than deep surface photometry and have the potential to probe the ICL down to much lower surface densities. Planetary nebulae offer an added bonus: as emission-line objects, follow-up spectroscopy can determine the kinematics of the ICL, giving yet another view of the degree to which the ICL is dynamically relaxed (Dubinski et al. 2001; Willman 2003). An interesting analogy can be made between the search for kinematic substructure due to tidal stripping in galaxy clusters and the search for kinematic substructure due to tidally destroyed satellites in the Milky Way's halo (e.g., Morrison et al. 2002). In both cases, kinematic substructure can be used to trace the dynamical accretion history of the system. With the advent of multi-object spectrographs on 8-m class telescopes, new and exciting opportunities now exist for studying this substructure in the diffuse starlight of galaxy clusters.
Acknowledgements. I would like to thank all my collaborators for their contributions to this work, and John Feldmeier in particular for providing Figures 5 - 7. John Dubinski graciously provided the cluster simulations from which Figure 5 was made. I also thank the conference organizers for putting together such a wonderful scientific program. This research has been supported in part by an NSF Career Award and a Research Corporation Cottrell Scholarship.