2 Globular Clusters

Globular clusters are gravitationally bound associations of 104 – 106 stars, distinct both from their smaller cousins, open clusters, and the larger, dark matter dominated dwarf galaxies that populate the low-mass end of the cosmological web of structure. Globular clusters are normally associated with a host galaxy and most galaxies, including the Milky Way, are surrounded and penetrated by a globular cluster system. A good estimate of the number of globular clusters in the Milky Way is the frequently updated catalogue by Harris [193Jump To The Next Citation Point], which has 157 entries as of 2010. Although fairly complete, a few new clusters have been discovered in recent years at low Galactic latitudes [234, 271] and there may be more hidden behind the galactic disc and bulge. The distribution of known globular clusters in the Galaxy is given in Figure 1View Image. Other galaxies contain many more globular clusters and the giant elliptical M87 alone may have over 10 000 [194]. The richness of the globular cluster system of a galaxy can be classified by the number of globular clusters associated with the galaxy normalized to its luminosity. One widely used measure of this is the specific frequency, SN = NGC × 100.4(MV +15) where NGC is the number of globular clusters and MV is the V-band magnitude of the galaxy [195]. SN can vary significantly between different galaxy types. For instance S ∼ 1.3 N for the local spiral galaxy M31 while S ∼ 14.4 N for M87. On the whole SN seems to be higher in massive elliptical galaxies than in spiral galaxies. For more information on extragalactic globular cluster systems see the review by Brodie & Strader [61Jump To The Next Citation Point].
View Image

Figure 1: Globular cluster distribution about the galaxy. Positions are from Harris [193Jump To The Next Citation Point] and are plotted as black circles on top of the COBE FIRAS 2.2 micron map of the Galaxy using a Mollweide projection. Image reproduced from Brian Chaboyer’s website [69].

Milky way globular clusters are old, having typical ages of 13 Gyr and an age spread of less than 5 Gyr [67]. This is on the order of the age of the Galaxy itself, thus Galactic globular clusters are thought to be left over from its formation. By contrast other galaxies such as the small and large Magellanic clouds (SMC and LMC) have intermediate age globular clusters (< 3 Gyr old, e.g., [346, 337]) and in some galaxy mergers, such as the Antennae, massive star-forming regions that may become globular clusters are observed [117]. Taken together, this implies that globular clusters of all ages are relatively common objects in the universe.

2.1 Stellar populations in globular clusters

Most of the detailed information on stellar populations in globular clusters comes from those in the Milky Way since only they are close enough for stars to be individually resolved. The stars in individual Galactic globular clusters all tend to have the same iron content [174Jump To The Next Citation Point] so globular clusters are thought to be internally chemically homogeneous. The colour-magnitude diagram (CMDs) for most Galactic globular clusters (e.g., M80, Figure 2View Image) also indicate a single stellar population with a distinct main-sequence, main-sequence turn-off, horizontal and giant branch. The single main sequence turn-off in particular indicates that all stars in the cluster have the same age. This leads to a so-called “simple stellar population” model for globular clusters where all stars have the same composition and age and differ only by their masses, which are set by the initial mass function (IMF). This simple picture has been challenged in recent years as observations have shown systematic star-to-star light element variations in globular clusters [174, 376]. Specific effects include different populations in s-process abundances (e.g., [321Jump To The Next Citation Point, 338Jump To The Next Citation Point]), anti-correlations between Na and O (e.g., [320Jump To The Next Citation Point, 321Jump To The Next Citation Point]), variations in CNO elements (e.g., [321Jump To The Next Citation Point, 338]) and even differences in iron abundance (e.g., [320, 321]). The best way of explaining these anomalies so far has been to use self-enrichment models where a single globular cluster experiences several bursts of star formation, each enriched by pollution from the previous generation [81Jump To The Next Citation Point]. How multiple populations affect the CMD of a globular cluster is shown in Figure 3View Image. The importance of these scenarios for relativistic binaries has not yet been explored but if the first and second generation have different IMFs this could affect the number of compact remnants. For this review, we will focus mainly on the case of a simple stellar population but we will discuss details of the multi-generation case further in Section 2.3.

View Image

Figure 2: Colour–magnitude diagram for M80. Image reproduced from the catalogue of 52 globular clusters (see [413]). The entire catalogue is available at the Padova Globular Cluster Group website [166].
View Image

Figure 3: CMD for NGC 288 showing evidence of two populations, and models incorporating different metallicities, helium fraction, and age. The best fit model has an age difference of Δ Age = 1.5 Gyr. Image reproduced by permission from Roh et al. [412], copyright by IOP.

The IMF is thought to be universal [34] and is usually taken to be a power-law of the form

-dN- ∝ M −αi , (1 ) dM
where M is the mass of a star, N the number of stars, dN ∕dM is the number of stars in an infinitesimal mass range and αi can take different values for different mass ranges. For values above ∼ 1M ⊙, αi is usually assumed to have a single value, the so-called Salpeter slope, of ∼ 2.35 [416]. There is much more debate about the value of αi in low-mass regime. One solution is to treat the IMF as a broken power-law with a break around 1M ⊙, such as that proposed by Kroupa & Weidner [279, 280]:
α0 = 0.3 ± 0.7, 0.01 ≤ M ∕M ⊙ < 0.08 α = 1.3 ± 0.5, 0.08 ≤ M ∕M < 0.50 1 ⊙ (2 ) α2 = 2.3 ± 0.3, 0.50 ≤ M ∕M ⊙ < 1.00 α3 = 2.3 ± 0.7, 1.00 ≤ M ∕M ⊙.
Another possibility, introduced by Chabrier [70], uses a log-normal distribution of masses below 1 M ⊙ and a Salpeter slope above. In both cases the power-law strongly favours low masses so stars massive enough to form neutron stars (NSs) and black holes (BHs) will be rare. It is worth noting that both of these IMFs were derived in the context of clustered star formation and young open clusters in the solar neighbourhood. It is often assumed that these IMFs hold for globular clusters as well but, because nearby globular clusters are old, most of the stars above ∼ 0.8M ⊙ have already moved off the main-sequence. Those just above 0.8 M ⊙ will be on the red giant branch and are readily visible in most optical images of globular clusters, such as the image of M80 shown in Figure 4View Image. Those that are more massive have evolved past the giant and horizontal branches and become low-luminosity compact remnants. The fact that the original high-mass population is no longer visible produces an intrinsic uncertainty in our knowledge of the high-mass end of the IMF of globular cluster stars. It is possible that the characteristic mass for star formation is higher in dense, optically thick regions and this would lead to an IMF more biased towards high-mass stars [348]. This in turn would increase the number of massive stellar remnants and could have an effect on the compact binary population.
View Image

Figure 4: Hubble Space Telescope photograph of the dense globular cluster M80 (NGC 6093).

If we assume that the consequences of multiple stellar generations in globular clusers are negligible, then the picture of Galactic globular cluster stellar populations that emerges from this analysis is of a simple, nearly co-eval, chemically homogeneous set of luminous, low-mass population II stars combined with a low-luminosity population of high-mass stellar remnants. It is interactions with members of this remnant population that will be of particular interest for producing relativistic binaries.

2.2 The structure of globular clusters

Globular clusters are classically modelled as spherical N-body systems. This approximation is relatively good given that the mean ellipticity, e, of Galactic globular clusters is ∼ 0.08 (where e = 1 − (b∕a), a is the semi-major axis and b is the semi-minor axis) and that e < 0.3 for all Galactic globular clusters [193Jump To The Next Citation Point]. Globular clusters have a core-halo structure where the core is highly concentrated, reaching densities of up to 106M ∕pc3 ⊙, and strongly self-gravitating. The surrounding halo is of much lower density and is less strongly self-gravitating. The structure of a globular cluster can be classified using three basic radii: the core radius (rc), the half-mass radius (rh), and the tidal radius (rt). One definition of the core radius relates it to the central velocity and density through the equation

4πG ----ρcr2c = v2c , (3 ) 3
where v2c is the mean-squared central velocity and ρc is the central density [199Jump To The Next Citation Point]. For an isothermal model this corresponds roughly to the radius at which the density drops to about one third, and thus the surface density to one half, of its central value. Observationally this corresponds to the radius at which the surface brightness drops to half of its central value. For multi-mass systems it is less clear how to arrive at an appropriate theoretical definition of the core radius and more empirical measures, such as the density-weighted average of the distance of each star from the density center are often used [68]. The half-mass radius is simply the radius thatcontains half of the mass of the system. The corresponding observational value is the half-light radius, which contains half the light of the system (the two radii do not necessarily agree). The tidal radius is the radius at which the gravitational field of the host galaxy becomes more important than the self gravity of the star cluster. A classical estimate of this for a circular orbit is given by Spitzer [443Jump To The Next Citation Point] as:
r3t = -MGC---R3GAL , (4 ) 2MGAL
where MGC is the mass of the globular cluster, MGAL the mass of the galaxy and RGAL the galactocentric radius of the circular orbit. In a time-dependent galactic field the escape process becomes significantly more complicated (see [119Jump To The Next Citation Point] for a more complete theory of tidal escape in a time-varyingexternal potential). For a given cluster, cluster orbit, and galaxy model the tidal radius of the cluster can, in principle, be clearly defined by comparing the effect of the galactic versus globular cluster gravitational field on a test mass. Observationally tidal radii can be difficult to determine due to the low stellar density of globular cluster halos and an imperfect knowledge of the gravitational potential of the host galaxy. Median values for rc, rh, and rt in the Galaxy are ∼ 1 pc, ∼ 3 pc and ∼ 35 pc respectively [199Jump To The Next Citation Point].

There are also two important timescales that characterize globular cluster evolution: the crossing time (t cr) and the relaxation time (t rlx). The crossing time is simply the time required for a star traveling at a typical velocity to cross some characteristic cluster radius. Thus, tcr ∼ R∕v where, for example, rc or rh might be typical radii of interest and v could be the velocity dispersion – normally taken to be the root-mean-square velocity and observed to be ∼ 10 km s–1 in Galactic globular clusters [193Jump To The Next Citation Point, 199Jump To The Next Citation Point]. tcr is also, roughly speaking, the orbital timescale for the cluster. For typical values of r h and v, t cr for Galactic globular clusters is on the order of 0.1 – 1 Myr but is longer at the tidal radius and much shorter in the core.

The relaxation time describes how long it takes for orbits to be significantly altered by stellar encounters. In particular trlx is often defined as the time necessary for the velocity of a star to change by an order of itself [57Jump To The Next Citation Point]. This can be thought of as the time necessary for a cluster to lose the memory of its initial conditions or, more exactly, the time necessary for stellar encounters to transform an arbitrary velocity distribution to a Maxwellian [443Jump To The Next Citation Point]. The relaxation time is related to the number and strength of encounters and, thus, to the number density and energy of a typical star in the cluster. It can be shown that the mean relaxation time in a globular cluster is [57Jump To The Next Citation Point, 443Jump To The Next Citation Point]

N trlx ∝ -----tcr. (5 ) lnN
Taking a typical value of 5 N = 10 and tcr as before, typical values of trlx are 0.1 – 1 Gyr. Thus, with ages typically greater than 10 Gyr, Galactic globular clusters are expected to be dynamically relaxed objects. In reality the value of trlx varies significantly within a globular cluster due to the highly inhomogeneous density distribution of the core-halo structure. It is possible for the core of a globular cluster to be fully relaxed while the halo remains un-relaxed after 13 Gyr. By making various approximations for tcr it is possible to relate trlx to local cluster properties. For instance the criterion of Meylan and Heggie [326Jump To The Next Citation Point, 288]:
2 3∕2 0.065⟨v-⟩---- trlx = ρ⟨m ⟩ln(γN ) (6 )
relates trlx to the local mass density (ρ), the mass-weighted mean squared velocity (2 ⟨v ⟩), and the average mass (⟨m⟩). The criterion by Spitzer [443Jump To The Next Citation Point]:
0.138N 1∕2r3∕2 trlx = ----1∕2--1∕2------- (7 ) ⟨m ⟩ G ln(γN )
relates trlx to a characteristic radius (r), the average mass, and the number of stars in the system. In practise rh is normally used for the characteristic radius in the Spitzer criterion and trlx is renamed the half-mass relaxation time, trh. The factor ln γN that appears in both definitions is called the Coulomb logarithm and describes the relative effectiveness of small and large angle collisions. It is calculated as an integral over the impact parameters for two-body scattering encounters, b:
∫ bmax b (b ) ln(γN ) ≡ ---= ln -max- , (8 ) bmin db bmin
where γ is a constant of order unity. The exact value of γ is a matter of some debate and must be determined empirically. Values in the literature range from 0.02 – 0.4 depending on the mass distribution of the system [156, 157]. Both definitions of trlx are used extensively in stellar dynamics.

On timescales shorter than a relaxation time interactions between individual stars do not govern the overall evolution of a stellar system and the granularity of the gravitational potential can be ignored. On these timescales the background structure of the cluster can be modelled using a static distribution function, f, that describes the probability of finding a star at a particular location in a 6-D position-velocity phase space. Formally, f depends on position, velocity, mass, and time so we have f (⃗x, ⃗v,m, t). For times less than trlx, however, the evolution of f is described by the collisionless Boltzmann equation:

∂f- ⃗ ⃗ ∂f- ∂t + ⃗v ⋅∇f − ∇ ϕ ⋅ ∂⃗v = 0 (9 )
and the explicit time dependence can be removed: ∂f ∕∂t = 0. The gravitational potential, ϕ, is given by Poisson’s equation:
⃗∇2 ϕ = 4πG ρ (10 )
and can be calculated at any position by integrating the distribution function over mass and velocity:
∫ ⃗∇2 ϕ = 4πG f (⃗x,⃗v,m )d3⃗v dm . (11 )
Solutions to Eq. 9View Equation are often described in terms of the relative energy per unit mass, ℰ ≡ Ψ − v2∕2, rather than in terms of the phase-space variables. Here, Ψ = − ϕ + ϕ0 is the relative potential and ϕ0 is defined such that no star has an energy less that zero (f > 0 for ℰ > 0 and f = 0 for ℰ < 0). A simple class of solutions to Eq. 9View Equation are Plummer models [378]:
7∕2 f(ℰ) = F ℰ , (12 )
the stellar-dynamical equivalent of an n = 5 ploytrope [199Jump To The Next Citation Point]. Another class of models that admit anisotropy and a distribution in angular momentum, L, are known as King–Mitchie models [264, 327]. Their basic distribution function is:
2 [ ] f (ℰ,L) = ρ (2π σ2)−3∕2exp( −-L--) e ℰ∕σ2 − 1 , ℰ > 0 , (13 ) 1 2r2aσ2
where σ is the velocity dispersion, ra the anisotropy radius where the velocity distribution changes from nearly isotropic to nearly radial, and ρ1 is a constant related to the density. Although not as well theoretically supported as the single-mass case, King–Mitchie models have been extended to include a spectrum of stellar masses [187] and even external gravitational field [200]. Multi-mass King models in particular are often fit to observed globular cluster cluster surface brightness profiles in order to determine their masses. A good example of the construction of a multi-mass King–Mitchie model is found in the appendix of Miocchi [339].

2.3 The dynamical evolution of globular clusters

Although static models can be used to describe the instantaneous structure of a globular cluster, there is no stable equilibrium for self-gravitating systems [57Jump To The Next Citation Point] and, therefore, their structure changes dramatically over time. Accessible descriptions of globular cluster evolution are given in Hut et al. [241Jump To The Next Citation Point], Meylan and Heggie [326] and Meylan [325] and have also been the subject of several texts (e.g., Spitzer [443Jump To The Next Citation Point] and Heggie and Hut [199Jump To The Next Citation Point]). Here we merely outline some of the more important aspects of globular cluster evolution.

The initial conditions of globular clusters are not well constrained since they seem to form only very early in the history of galaxy formation or in major mergers [150, 61Jump To The Next Citation Point] – both situations quite different from the environment of our Galaxy today. However, it is possible to make some general statements. Like all stars, the stars in globular clusters collapse out of molecular gas. Due to their small age spread and chemical homogeneity, it seems likely that all the stars in a globular cluster formed by the collapse and fragmentation of a single giant molecular cloud [61]. However, the exact details of how this collapse proceeds are unclear. In the classical picture a globular cluster forms in a single collapse event – all stars form rapidly at almost exactly the same time from a globally collapsing cloud of gas and, thus, have almost exactly the same age and chemical composition. Not all of the gas in the collapsing cloud is necessarily converted into stars and the fraction of gas that becomes stars is described through the star formation efficiency (SFE). An SFE of less than 100% implies that not all of the primordial gas forms stars and the resulting globular cluster will be less massive than its parent cloud. Star formation is then terminated and the left-over gas expelled by a combination of ionizing radiation from young stars and energy injection from supernovae. Although more applicable to open clusters, Dale et al. [85] provides some the the most recent results on the details of how this process works in young star clusters. If the star formation efficiency is low, then the amount of mass loss through gas expulsion can leave the cluster out of virial equilibrium and may lead to its immediate dissolution [171, 172Jump To The Next Citation Point, 484], a process called “infant mortality”. It is estimated that more than half of young star clusters in the local universe are destroyed in this manner [172] and even the surviving clusters will lose a large fraction of their stars. Even if the cluster survives the gas expulsion, the rapid change in potential will cause the energy of individual stars to change in a mass-independent manner [57Jump To The Next Citation Point]. This process, called violent relaxation, means that the positions, velocities, and masses of cluster stars will be initially uncorrelated.

As mentioned in Section 2.1, observations are now beginning to challenge some of the details of this simplistic picture of globular cluster formation. Rather than a single burst of star formation the observed abundance anomalies suggest a more drawn-out formation scenario where a first generation of stars forms, evolves and enrichs the cloud with ejecta processed through nuclear burning. Later generations form from the enriched gas and carry the chemical tracers of the pollution. A general description of how such a scenario might proceed can be found in Conroy and Spergel [81] however there is little consensus on the exact details. The initial cloud could suffer a global collapse, experience star formation, have the collapse halted by feedback and then re-collapse to form the second generation. Conversely, different regions of the cloud could collapse into sub-clumps at different times and enrich neighbouring regions with their ejecta. The globular cluster would then assemble from the merger of these sub-clumps. It is not clear what effect, if any, these different formation scenarios will have on the compact binary population of globular clusters. There is some suggestion that the early generations must have an IMF biased towards massive stars in order to explain the observed abundance anomalies (e.g., [91Jump To The Next Citation Point, 92]) and if so this would certainly have implications for the number of compact objects and compact binaries. However, the question is far from resolved. Due to the uncertainties in the multi-generation scenarios and in particular the fact that the details are not necessarily important for compact binary production, we will assume a single star formation event in globular clusters for the purposes of this review.

Equipartition of energy dictates that the most massive stars should have the lowest kinetic energies [57Jump To The Next Citation Point, 443Jump To The Next Citation Point] but violent relaxation leaves the velocities and masses uncorrelated. Therefore, as soon as the residual gas is expelled from a young globular cluster, massive stars with large kinetic energies will start to transfer this energy to low-mass stars through stellar encounters. As the massive stars lose kinetic energy they will sink to the center of the cluster while the low-mass stars gain kinetic energy and move to the halo. This process is known as mass segregation and proceeds on a timescale tms ∝ mi ∕⟨m ⟩ × trlx [442Jump To The Next Citation Point, 479Jump To The Next Citation Point, 257Jump To The Next Citation Point]. Mass segregation can also occur in other situations, such as around super-massive black holes in galactic centres [136]. Due to mass-loss from stellar evolution, compact remnants rapidly become the most massive objects in globular clusters as the star population ages [419]. Thus they, along with binaries, are strongly affected by mass segregation and are particularly likely to be found in cluster cores [475]. Mass segregation continues until energy equipartition has been achieved. There are, however, initial conditions for which it is impossible to achieve energy equipartition [479, 257] and it is formally impossible to halt mass segregation, leading to a singularity in the core of the cluster. This phenomenon was first noted by Spitzer [442] and is thus called the “Spitzer instability”. In reality, the massive objects in such systems form a strongly interacting subsystem, dynamically decoupled from the rest of the cluster. Due to their high masses, black holes in star clusters are particularly likely to experience the Spitzer instability and this has been the starting point for several investigations of BH binaries in star clusters (e.g. [360Jump To The Next Citation Point]).

The longer-term evolution of star clusters is driven by two-body relaxation, where the orbits of stars are perturbed by encounters with their neighbours. The theory of two-body relaxation was first quantified by Chandrasekhar in 1942 [71]. Two-body relaxation becomes important on timescales longer than the local relaxation time. The evolution of a globular cluster over these timescales can still be described (at least formally) by the Boltzmann equation but with a collisional term added to the right-hand side. Eq. 9View Equation then takes on the form:

∂f- ⃗ ⃗ ∂f- ∂t + ⃗v ⋅ ∇f − ∇ϕ ⋅ ∂⃗v = Γ [f]. (14 )
Eq. 14View Equation is sometimes called the collisional Boltzmann equation and the term Γ [f ] describes the effect of two-body (and in principle higher-order) interactions on the distribution function. Practically speaking, it is not possible to evaluate Γ [f] analytically and various numerical approximations must be employed. Approaches include the Fokker–Planck method, where Γ [f] is approximated in the weak scattering limit by an expansion in powers of the phase-space parameters; the Monte Carlo method, where Γ [f] is approximated by a Monte Carlo selection of weak encounters over a time shorter than the relaxation time; the anisotropic gas model where the cluster is approximated as a self-gravitating gaseous sphere; or direct N-body integration, where rather than solving Eq. 14View Equation the orbits of each star in the cluster are explicitly integrated. Each method has its strengths and weaknesses and will be discussed further in Section 5.

Thermodynamically speaking, strongly self-gravitating systems have a negative heat capacity. This can be seen by relating the kinetic energy of the system to a dynamical temperature [57Jump To The Next Citation Point]:

1 3 -mv2ave = -kBT . (15 ) 2 2
This, in combination with the virial theorem, can be used to define a total internal energy for the cluster: 5
3- E = − 2N kBT , (16 )
where N is the number of bodies in the system. Finally, this can be used to calculate a heat capacity:
dE- 3- C = dT = − 2 N kB . (17 )
Since all the constants on the right hand side of Eq. 17View Equation are positive, C is always negative. A negative heat capacity means that heating a self-gravitating system actually causes it to lose energy. For a core-halo star cluster, the core is strongly self-gravitating while the halo is not, so the halo acts as a heat bath for the core. Any perturbation in which the core becomes dynamically hotter than the halo causes energy to flow into the halo. The negative heat capacity means that the core becomes even hotter, increasing the flow of energy to the halo in a runaway process. This causes the core to contract, formally to a singularity. The runaway process is called the gravothermal catastrophe and the consequent shrinking of the core is called core collapse. It affects all self-gravitating systems and was first noted in the context of star clusters by Antonov [18]. In equal-mass systems core collapse will occur after 12– 20 trh [443Jump To The Next Citation Point] but may be accelerated in systems with a spectrum of masses due to mass segregation. Core collapse not only appears in analytic models, but has also been found in a variety of numerical simulations such as the model shown in Figure 5View Image [253Jump To The Next Citation Point]. Furthermore, the Harris catalogue lists several Galactic globular clusters that from their surface brightness profiles are thought to have experienced a core collapse event [193Jump To The Next Citation Point].
View Image

Figure 5: Lagrange radii indicating the evolution of a Plummer model globular cluster for an N-body simulation and a Monte Carlo simulation. The radii correspond to radii containing 0.35, 1, 3.5, 5, 7, 10, 14, 20, 30, 40, 50, 60, 70, and 80% of the total mass. Image reproduced by permission from Joshi et al. [253Jump To The Next Citation Point], copyright by IOP.

Core collapse can be halted, at least temporarily, by an energy source in the core. For stars (also self-gravitating systems) this energy source is nuclear burning. In star clusters tightly bound binaries perform a similar role. Stars in the core scatter off these binaries and gain kinetic energy at the expense of the orbital energy of the binary. This process reverses the temperature gradient and consequently the gravothermal instability which in turn causes the core to re-expand. The core will then cool again due to the expansion, the temperature gradient will again reverse and the process of core collapse will repeat. These repeated core expansions and contractions are know as gravothermal oscillations [56]. The heating from the binaries is the trigger for the process and, in analogy to nuclear burning in stars, is called “binary burning”. The binaries taking part in binary burning may be either primordial (binaries where the stars were born bound to each other) or dynamically formed by a variety of interactions that will be discussed in Section 4.3. It is worth noting that while tight binaries serve as energy sources for the cluster, loosely bound binaries with orbital velocities below the local velocity dispersion can actually act as energy sinks and may significantly hasten the onset of core collapse [131Jump To The Next Citation Point]. The importance of this effect on the evolution of star clusters remains largely unexplored.

Stars can escape from a star cluster if they gain a velocity greater than the cluster’s escape velocity, ⟨v2e⟩ = − 4U ∕M where U is the potential energy of the star cluster and M its total mass. Using the virial theorem it is possible to show that ⟨v2e⟩ = 4⟨v2⟩ where ⟨v2⟩ is the RMS velocity in the cluster [57Jump To The Next Citation Point]. There are two means through which a star can reach the escape velocity. The first is ejection where a single strong interaction, such as occurs during binary burning, gives the star a sufficient velocity impulse to exceed ve. This process is highly stochastic. The second is evaporation where a star reaches escape velocity due to a large number of weak encounters during the relaxation process. Relaxation tends to maintain a local Maxwellian in the velocity distribution and, since a Maxwellian distribution always has a fraction γ = 7.38 × 10− 3 stars with v > 2vRMS, there are always stars in the cluster with a velocity above the escape velocity. Thus, it is the fate of all star clusters to evaporate. The evaporation time can be estimated as:

t ≈ trlx = 136 t . (18 ) evap γ rlx
This is much longer than a Hubble time so few globular clusters are directly affected by evaporation. Evaporation can, however, be accelerated by the presence of a tidal field. In this case stars whose orbits extend beyond r t are stripped from the cluster and lost. Tidal dynamics are more complicated than a simple radial cutoff would imply and detailed prescriptions taking into account orbital energy and angular momentum as well as a time-varying field for star clusters in elliptical orbits are necessary to capture all of the important processes [457, 119]. It seems likely that the ultimate fate of most globular clusters is destruction due to tidal effects [167, 152]. Since most compact objects are likely to live deep in the core of globular clusters where ejection will normally be due to violent interactions, the details of tidal stripping are unlikely to be critical for the treatment of relativistic binaries in star clusters.

  Go to previous page Scroll to top Go to next page