19.1. The role of mergers in galaxy evolution¶
From almost as soon as galactic nebulae were first observed and long before they were known to be external systems similar to our own Galaxy, it was clear that galaxies are often found near each other on the sky. Herschel (1811) discussed double and higher-multiple nebulae in his catalog of a few thousand nebulae (a precursor to the NGC catalog still used today) and argued that they must be the result of the splitting up of a single nebula. Once galaxies were recognized to be external stellar systems similar to the Milky Way, double galaxies and their implications for galaxy evolution became of greater interest. Increased observational capabilities allowed deeper images of double galaxies to be obtained, revealing bridges (e.g., Keenan 1935) and tails (e.g., Zwicky 1956). This early observational work culminated in the Arp (1966) Atlas of Peculiar Galaxies, a set of 338 images of galaxies with peculiar morphologies, many of which are now believed to result from galactic interactions and mergers.
While some argued against this (e.g., Vorontsov-Velyaminov 1962), the obvious explanation for the observed bridges and tails was that they were caused by tidal forces operating during a close encounter (e.g., Holmberg 1937; Zwicky 1956). However, clearly demonstrating this proved difficult, because galaxy interactions are not amenable to analytic studies and running large \(N\)-body simulations was not yet possible. Most of the early simulations used the restricted three-body approximation: interacting galaxies were represented as central point masses surrounded by rings of test particles—self-gravity within the galaxies was, thus, ignored—and the motion of the test particles in the gravitational field of the two point masses then traced the reaction to the overall gravitational perturbation from the other galaxy. Such simulations demonstrated that fly-bys can create spiral structure (Pfleiderer & Siedentopf 1961), bridges between galaxies (Yabushita 1971), and tails (Clutton-Brock 1972a; Wright 1972). The most influential paper of this era was that of Toomre & Toomre (1972), who provided a lucid explanation of the dynamics of the interactions that gives rise to tails and bridges and who were able to build models of four specific interaction pairs using the restricted three-body approximation. Their famous model for the interaction between M51 and its satellite NGC 5195 is shown in Figure 19.2 alongside the iconic Hubble observation of this system (already shown in Figure 1.1).
|
|
Figure 19.2: Restricted three-body approximation model for the NGC 5195/M51 interaction (Toomre & Toomre 1972; left) and the Hubble image of M51 (NASA/JPL-Caltech/Univ. of Arizona/DSS/SST; right).
While the lack of self-gravity in the simulations made it difficult for them to fully match the observed morphologies, observations of line-of-sight velocities in the tails and within the disturbed galaxies agreed with the predictions from the tidal-forces theory (e.g., Schweizer 1978; van der Hulst 1979). The tidal-force theory was, therefore, quickly accepted as the explanation for tails and bridges and these days tails and tidal tails are used synonymously. In addition to tails and bridges, galaxy interactions also give rise to rings and cartwheel galaxies (Lynds & Toomre 1976; Theys & Spiegel 1977; Toomre 1978), shells in elliptical galaxies (Quinn 1984), and polar rings (Athanassoula & Bosma 1985).
While tails and bridges provide convincing evidence for galaxy interactions, whether the involved galaxies actually end up merging was less obvious. Producing the observed tails and bridges required galaxies to approach each other on near-parabolic orbits and without a source of orbital dissipation, interacting galaxies would simply pass each other by and return to being isolated objects. The early simulations discussed so far did not include dark-matter halos, but incontrovertible evidence for their existence was accruing at the same time (see Chapter 8). Extended dark-matter halos around galaxies both significantly increase the cross section for interactions between galaxies—helping to explain the larger observed merger rate—and they provide a source of friction: dynamical friction. Dynamical friction was first discussed by Chandrasekhar (1943a) in the context of a star orbiting within a collection of other stars: Chandrasekhar (1943a) showed that some of the orbital energy of an orbiting star is transferred to random energy of the orbits of the collection of stars it orbits in and this gives rise to a frictional force. This force is proportional to the square of the mass of the star in question and is, thus, larger for more massive stars. As a purely dynamical gravitational effect, dynamical friction also occurs for other massive objects orbiting within a background of lower-mass objects (e.g., globular clusters in the centers of galaxies; Tremaine et al. 1975). If galaxies are surrounded by massive dark-matter halos, then dynamical friction will cause the orbits of interacting galaxies to decay (Ostriker & Tremaine 1975; Tremaine 1976a). Thus, dynamical friction is the missing ingredient that causes the slow, parabolic orbits that give rise to tidal tails and bridges to decay and that eventually leads to mergers. We discuss dynamical friction in more detail in Section 19.4.1.
In addition to including dark-matter halos, the larger cosmological context that galaxy evolution occurs in also turned out to favor mergers. Simulations of merging galaxies in a cosmological setting demonstrated that the bottom-up hierarchical clustering inherent in CDM models causes a preponderance of close encounters with low angular momentum, that is, close, nearly head-on collisions (Aarseth & Fall 1980). The importance of mergers is therefore a crucial ingredient in our CDM cosmogony: the hierarchical clustering caused by the abundant small-scale structure in the primordial CDM power spectrum causes close encounters in which the extended dark-matter halos surrounding galaxies collide and transform orbital energy into random motions of the dark-matter particles, leading to an eventual merger of the two systems. Galaxies, occupying only a small volume at the center of their dark-matter halos, survive most of the merger relatively unscathed, until they too merge during the final stage, forming a new, transformed centrally-concentrated galaxy at the bottom of the potential well of the merged halos.
That galaxies with tidal tails can eventually merge is clear from objects that are demonstrably in the late stages of a merger that have two tidal tails emanating from the main body of the merging galaxy. NGC 2623, displayed in Figure 19.3, provides a beautiful example of this.
Figure 19.3: NGC 2623: two recently-merged spiral galaxies (credit: ESA/Hubble & NASA).
With mergers likely being quite common—recall the estimate of 5% among NGC galaxies from above—it was quickly realized that they could be an important process in galaxy evolution. Toomre (1977) pointed out that the estimated number of mergers in the NGC catalog agrees remarkably well with the number of large elliptical galaxies and hypothesized that elliptical galaxies could result from the mergers of disk galaxies. Investigating whether such mergers can quantitatively give rise to the observed structure of elliptical galaxies was difficult, because the \(N\)-body simulations that are required to properly incorporate the effects of dynamical friction and of self-gravity were only in their infancy. Simulations of merging spherical galaxies using hundreds of particles with the direct-summation technique from Chapter 12.4.1.1 (White 1978; White 1979), and using thousands to tens-of-thousands of particles with potential expansion methods (van Albada & van Gorkom 1977; Villumsen 1982; cf. Chapter 12.3) and grid-based Fourier methods (Miller 1978; Miller & Smith 1980), demonstrated that dynamical friction can indeed merge interacting galaxies. Similar simulations of mergers of disk galaxies embedded in dark-matter halos using a few hundred particles showed that they merge and form elliptical-like galaxies (e.g., Gerhard 1981; Negroponte & White 1983), but the numerical fidelity of these simulations was low, because of the large relaxation effects inherent in low-\(N\) simulations (see Chapter 5.1). While grid-based or expansion-based \(N\)-body methods could handle larger \(N\), their reliance on a fixed geometry is not well suited to following the highly dynamical process of a galaxy merger. Direct summation, on the other hand, can follow complex dynamical systems easily, but the \(\mathcal{O}(N^2)\) cost of force evaluations limited them to very low particle numbers. The introduction of the tree-based gravity algorithm by Barnes & Hut (1986) discussed in Chapter 12.4.1.2 was a crucial development in simulating mergers, because its tree-grid-based approach allows it to follow evolving, complex geometries with ease (cf. Figure 12.17) while its hiearchical-tree gravity solver lets it use large \(N\). Using the Barnes & Hut (1986) algorithm, Barnes (1988) demonstrated that mergers of disk galaxies can give rise to the observed tidal tails and bridges, that their dark-matter halos efficiently absorb orbital energy and angular momentum through dynamical friction, and that the merger happening because of the induced orbital decay of the interacting galaxies produces remnants with radial profiles, shapes, and kinematics similar to that of massive ellipticals.
Because disk galaxies generally contain a significant reservoir of gas, hydrodynamics, gas cooling, and star formation must play an important role in most galaxy mergers. But these processes were not considered in the first generations of merger simulations. Assessing the role of gas-dynamical processes in galaxy mergers required the development of simulation codes that can solve for gas hydrodynamics in the gravitational field of merging galaxies. Like for the \(N\)-body gravity calculations, the use of fixed grids to solve the fluid equations—common in other fields—is sub-optimal for following the complex, evolving geometry during a galaxy merger. Instead, smoothed particle hydrodynamics (SPH; Lucy 1977; Gingold & Monaghan 1977), which follows the gas using Lagrangian particles for which the fluid equations are solved using a kernel-based approach, can be used. The SPH particles naturally follow the evolving geometry of the galaxies, especially efficiently when SPH is combined with a tree-based gravity code (Hernquist & Katz 1989). SPH was the dominant method for simulating gas dynamics in galaxies until the advent of moving-grid codes that combine the advantages of SPH and grid-based codes (e.g., Springel 2010).
The inclusion of gas dynamics in merger simulations does not substantially change the formation of tails, bridges, and other morphological merger signatures (Barnes & Hernquist 1996). But tidal torques operating during the merger causing gas to flow to the center of the merger remnant and the compression of gas and of the merging galaxies’ molecular-cloud system both lead to enhanced star formation during mergers that has observational consequences. Collisions between molecular clouds are an important driver of star formation and the rate of collisions increases dramatically as the luminous galactic components start to merge (e.g., Noguchi & Ishibashi 1986). Tidal torques can lead to the formation of a central bar that funnels gas to the center through torques resulting from the bar and through further-enhanced cloud-cloud collisions (Noguchi 1988; Barnes & Hernquist 1991; Mihos & Hernquist 1996; Barnes & Hernquist 1996). Thus, strong bursts of star formation should be common during galaxy mergers, especially in the central regions where it is also often accompanied by nuclear activity resulting from accretion onto the central supermassive black hole that is observed in the form of Seyfert galaxies and quasars. Stellar population synthesis modeling of the colors of merging galaxies indeed showed that star formation rates are high and occur in bursts (Larson & Tinsley 1978), in some cases leading to very high star formation rates (Joseph & Wright 1985). Observations of the molecular gas content of merging galaxies confirmed the presence of abundant molecular gas in the process of forming stars at a rate that is much enhanced compared to that in isolated galaxies (Young et al. 1986; Scoville et al. 1986). Essentially all of the galaxies with the highest observed star formation rates—ULIRGs, with infrared luminosities \(L_{\mathrm{FIR}} > 10^{12}\,L_{\mathrm{FIR},\odot}\)—are strongly interacting systems (Sanders et al. 1988; Melnick & Mirabel 1990). Bursts of star formation are clearly visible in the image of the Antennae galaxies and, to a lesser extent, in that of Stephan’s Quintet in Figure 19.1. Much larger, more recent \(N\)-body simulations of galaxy mergers with better treatments of gas dynamics, star formation, and the energetic feedback processes associated with massive stars, supernova explosions, and accretion onto the central black hole (see Section 19.3.3 below), find similar enhancements in star formation rate and black hole feeding, but emphasize the role that feedback has in shutting off the cold gas supply (Di Matteo et al. 2005; Springel et al. 2005). Thus, soon after a merger, star formation drops dramatically or ceases altogether and the central black hole accretes gas more slowly or becomes fully quiescent.
So far we have focused on so-called major mergers—mergers with mass ratios of 3:1 or less. Because of the strong gravitational forces at play, galaxies are significantly transformed during major mergers and, for example, disk galaxies rarely survive a major merger intact. But more numerous than major mergers are minor mergers—mergers with mass ratios of 3:1 or higher (however, the exact boundary between major and minor mergers varies between studies). Considering minor mergers down to 10:1, minor mergers are about three times as common as major mergers and this ratio is approximately constant with redshift (Lotz et al. 2011). Like major mergers, minor mergers drive starbursts (Mihos & Hernquist 1994) and nuclear activity (Hernquist & Mihos 1995), although to a lesser degree. Because the dynamical-friction force is proportional to the square of the mass of the merging companion, it is less effective at causing the orbital decay of lower-mass companion galaxies: as we will see in Section 19.4.1 below, very roughly the merger time scale is given by the product of the mass ratio and the dynamical time, so minor mergers take factors of a few to orders of magnitude longer to complete than major mergers. Minor mergers, especially at high mass ratios, are therefore often better described as accretion, because any addition of gas or stars from the merging companion occurs rather slowly. The dynamical effects of minor mergers scale as the mass ratio and can therefore be quite large for 10-ish:1 minor mergers, causing in particular vertical heating of galactic disks (e.g., Toth & Ostriker 1992; Quinn et al. 1993; Villalobos & Helmi 2008), although this can be damped by the presence of gas in the disk (Moster et al. 2010a). However, the effect of minor mergers on the stellar content of most galaxies is small, because of the dramatically lower stellar-mass–to–halo-mass ratio at low halo masses (see Equation 18.27 and Figure 18.16): a minor 10:1 merger only accretes \(\approx 1\%\) additional stellar mass for galaxies like the Milky Way or smaller (with halo masses \(\lesssim 10^{12}\,M_\odot\)). Only for massive ellipticals and groups/clusters (with halo masses \(\gtrsim 10^{12}\,M_\odot\)) do minor mergers bring in substantial fractions of stars, because the stellar mass is only weakly-dependent on halo mass at those halo masses (Equation 18.28). Minor mergers can therefore contribute significantly to, e.g., the size evolution of massive elliptical galaxies (Bournaud et al. 2007; Naab et al. 2009). Similarly, in groups and clusters of galaxies, mergers contribute significantly to the intracluster light—the light between individual galaxy members of a group or cluster that typically contains the same amount of mass for groups and an order of magnitude more for clusters than the central galaxy (Behroozi et al. 2013a).
While it is clear that mergers happen and significantly affect the galaxies involved in the merger, how much of an effect mergers have in the grand scheme of galaxy formation and evolution requires a statistical determination of the rate of mergers as a function of cosmic time. This is difficult to obtain, because it is hard to unambiguously detect in-progress mergers and merger remnants in large, statistical galaxy surveys and because one has to estimate how long any given merger will take to complete to turn the fraction of observed mergers into a merger rate. Moreover, the merger rate depends on a wide range of variables such as galaxy mass and environment. Due to these issues, all observational determinations of the merger rate are somewhat inaccurate and different studies over the years have come to significantly different conclusions regarding the merger rate. A simple, early estimate of the merger rate that we have already discussed was given by Toomre (1977), who estimated that \(\approx 5\%\) of galaxies in the NGC catalog have been affected by mergers. Early statistical determinations of the merger rate confirmed this rate (Patton et al. 2000) and found that the merger rate increases with redshift approximately as \(\propto (1+z)^3\) up to \(z=1\) (Le Fevre et al. 2000). This agrees with the expect rate of mergers of dark-matter halos in \(\Lambda\)CDM (e.g., Gottlober et al. 2001), but the merger rate and its redshift dependence also strongly depends on galaxy luminosity (Conselice et al. 2003), with mergers becoming increasingly common for massive galaxies at higher redshift. Subsequent, larger studies, however, found little evolution in the merger rate with redshift (Lotz et al. 2008), although this conclusion depends on how galaxies are matched across redshift: requiring galaxies at different redshifts to have the same comoving number density—a better proxy for following the same type of galaxy as it evolves than stellar mass or luminosity—finds strong \(\propto (1+z)^3\) evolution again (Lotz et al. 2011). Thus, for any given galaxy, the probability of a merger increases with redshift, although exactly by how much is still somewhat up in the air. Studies of close pairs at redshifts up to \(z=6\) demonstrate that the probability of mergers increases by two orders of magnitude between \(z < 1\) and \(z = 6\) (Duncan et al. 2019). This conclusion is in line with the increase in the fraction of peculiar galaxies with redshift that we discussed in Chapter 18.3 (see Figure 18.18).
However, the impact of all of these mergers appears to be limited. Combining merger rates with estimates of how much stellar-mass growth they cause, shows that they only contribute a small fraction of the overall stellar-mass growth determined from the evolution of the stellar mass function (e.g., Bundy et al. 2009; Moustakas et al. 2013). One powerful way of looking at this is using a variation of the abundance matching approach that we introduced in Chapter 18.2 where the intertwined dark-matter and galaxy growth over time is determined by simultaneously matching the stellar mass function, cosmic star formation history, and galaxy star formation rates at different redshifts for galaxies matched to simulated dark-matter halos (Behroozi et al. 2013a). This exercise shows, remarkably, that mergers have almost no effect on the overall stellar-mass growth of galaxies, which is instead dominated by in-situ star formation. That is, stars are overwhelmingly formed from gas within the galaxy that they reside in, not within another galaxy that merged with their present galaxy, at all redshifts. The exception to this are massive galaxies, groups, and clusters (with halo masses larger than \(10^{12}\,M_\odot\)) at low redshift, for which most of their present-day mass growth comes from mergers (for the reason discussed above).
Thus, while mergers are crucial to understanding many interesting classes of galaxies, such as peculiar galaxies, active galactic nuclei, and quasars, and while mergers are increasingly common in the early Universe, their net impact on the stellar-mass growth of galaxies is limited. The Milky Way provides an instructive example: mergers that do not destroy disks deposit the companion’s stars in the stellar halo rather than the disk, so the mass of the stellar halo provides an upper limit to the amount of stellar mass obtained through mergers (an upper limit, because some of the stars in the stellar halo may have been kicked out of the disk). As we saw in Chapter 1.2, the Milky Way’s total stellar mass is \(\approx 6 \times 10^{10}\,M_\odot\), while the mass of its stellar halo is \(\approx 1.5 \times 10^9\,M_\odot\). Thus, only \(\lesssim 2.5\%\) of the stars in the Milky Way were accreted during mergers. While the Milky Way has a particularly quiescent merger history compared to similar galaxies, the typical accreted ratio for Milky-Way-like galaxies is \(\lesssim 10\%\). While it may appear counterintuitive that it is precisely at higher redshifts that galaxy mergers have the least impact on stellar-mass growth while mergers are far more common than they are today, this is because the cosmic star-formation rate is also an order of magnitude higher than it is today (see Figure 18.26) , more than compensating for the increased merger rate in the contest between in-situ star formation and accretion from mergers.
.
.
.