Star clusters with primordial binaries – I. Dynamical evolution of isolated models

In order to interpret the results of complex realistic star cluster simulations, which rely on many simplifying approximations and assumptions, it is essential to study the behaviour of even more idealized models, which can highlight the essential physical effects and are amenable to more exact methods. With this aim, we present the results of N -body calculations of the evolution of equal-mass models, starting with primordial binary fractions of 0–100 per cent, with values of N ranging from 256 to 16 384. This allows us to extrapolate the main features of the evolution to systems comparable in particle number with globular clusters.
In this range, we find that the steady-state 'deuterium main sequence' is characterized by a ratio of the core radius to half-mass radius that follows qualitatively the analytical estimate by Vesperini & Chernoff, although the N dependence is steeper than expected. Interestingly, for an initial binary fraction f greater than 10 per cent, the binary heating in the core during the post-collapse phase almost saturates (becoming nearly independent of f ), and so little variation in the structural properties is observed. Thus, although we observe a significantly lower binary abundance in the core with respect to the Fokker–Planck simulations by Gao et al., this is of little dynamical consequence.
At variance with the study of Gao et al., we see no sign of gravothermal oscillations before 150 half-mass relaxation times. At later times, however, oscillations become prominent. We demonstrate the gravothermal nature of these oscillations.     

Effects of external tidal field on the evolution of the outer regions of multi-mass star clusters

We present N -body simulations (including an initial mass function) of globular clusters in the Galaxy in order to study effects of the tidal field systematically on the properties of the outer parts of globular clusters. Using nbody6 , which correctly takes into account the two-body relaxation, we investigate the development of tidal tails of globular clusters in the Galactic tidal field. For simplicity, we have employed only the spherical components (bulge and halo) of the Galaxy, and ignored the effects of stellar evolution which could have been important in the very early phase of the cluster evolution. The total number of stars in our simulations is about 20 000, which is much smaller than the realistic number of stars. All simulations had been done for several orbital periods in order to understand the development of the tidal tails. In our scaled-down models, the relaxation time is sufficiently short to show the mass segregation effect, but we did not go far enough to see the core collapse, and the fraction of stars lost from the cluster at the end of the simulations is only ∼10 per cent. The radial distribution of extra-tidal stars can be described by a power law with a slope around −3 in surface density. The directions of tidal tails are determined by the orbits and locations of the clusters. We find that the length of tidal tails increases towards the apogalacticon and decreases towards the perigalacticon. This is an anti-correlation with the strength of the tidal field, caused by the fact that the time-scale for the stars to respond to the potential is similar to the orbital time-scale of the cluster. The escape of stars in the tidal tails towards the pericentre could be another reason for the decrease of the length of tidal tails. We find that the rotational angular velocity of tidally induced clusters shows quite different behaviour from that of initially rotating clusters.     

The evolution of binary fractions in globular clusters

We study the evolution of binary stars in globular clusters using a new Monte Carlo approach combining a population synthesis code ( startrack ) and a simple treatment of dynamical interactions in the dense cluster core using a new tool for computing three- and four-body interactions ( fewbody ). We find that the combination of stellar evolution and dynamical interactions (binary–single and binary–binary) leads to a rapid depletion of the binary population in the cluster core. The maximum binary fraction today in the core of a typical dense cluster such as 47 Tuc, assuming an initial binary fraction of 100 per cent, is only ∼ 5–10 per cent. We show that this is in good agreement with recent Hubble Space Telescope observations of close binaries in the core of 47 Tuc, provided that a realistic distribution of binary periods is used to interpret the results. Our findings also have important consequences for the dynamical modelling of globular clusters, suggesting that 'realistic models' should incorporate much larger initial binary fractions than has usually been the case in the past.     

2D Fokker–Planck models of rotating clusters

Globular clusters rotate significantly, and with the increasing amount of detailed morphological and kinematical data obtained in recent years on galactic globular clusters many interesting features show up. We show how our theoretical evolutionary models of rotating clusters can be used to obtain fits, which at least properly model the overall rotation and its implied kinematics in full 2D detail (dispersions, rotation velocities). Our simplified equal mass axisymmetric rotating model provides detailed two-dimensional kinematical and morphological data for star clusters. The degree of rotation is not dominant in energy, but also non-negligible for the phase-space distribution function, shape and kinematics of clusters. Therefore, the models are well applicable for galactic globular clusters. Since previously published papers on that matter by us made it difficult to do detailed comparisons with observations, we provide a much more comprehensive and easy-to-use set of data here, which uses as entries dynamical age and flattening of observed cluster and then offers a limited range of applicable models in full detail. The method, data structure and some exemplary comparison with observations are presented. Future work will improve modelling and data base to take a central black hole, a mass spectrum and stellar evolution into account.     

Predictions for triple stars with and without a pulsar in star clusters

Though about 80 pulsar binaries have been detected in globular clusters so far, no pulsar has been found in a triple system in which all three objects are of comparable mass. Here, we present predictions for the abundance of such triple systems, and for the most likely characteristics of these systems. Our predictions are based on an extensive set of more than 500 direct simulations of star clusters with primordial binaries, and a number of additional runs containing primordial triples. Our simulations employ a number N _tot of equal-mass stars from   N _tot= 512  to  19 661  and a primordial binary fraction from 0 to 50 per cent. In addition, we validate our results against simulations with   N = 19 661  that include a mass spectrum with a turn-off mass at  0.8 M_⊙  , appropriate to describe the old stellar populations of Galactic globular clusters. Based on our simulations, we expect that typical triple abundances in the core of a dense cluster are two orders of magnitude lower than the binary abundances, which in itself already suggests that we do not have to wait too long for the first comparable-mass triple with a pulsar to be detected.     

Black holes and core expansion in massive star clusters

In this study we present the results from realistic N -body modelling of massive star clusters in the Magellanic Clouds. We have computed eight simulations with   N ∼ 10⁵  particles; six of these were evolved for at least a Hubble time. The aim of this modelling is to examine in detail the possibility of large-scale core expansion in massive star clusters, and search for a viable dynamical origin for the radius–age trend observed for such objects in the Magellanic Clouds. We identify two physical processes which can lead to significant and prolonged cluster core expansion – mass-loss due to rapid stellar evolution in a primordially mass-segregated cluster, and heating due to a retained population of stellar mass black holes, formed in the supernova explosions of the most massive cluster stars. These two processes operate over different time-scales and during different periods of a cluster's life. The former occurs only at early times and cannot drive core expansion for longer than a few hundred Myr, while the latter typically does not begin until several hundred Myr have passed, but can result in core expansion lasting for many Gyr. We investigate the behaviour of each of these expansion mechanisms under different circumstances – in clusters with varying degrees of primordial mass segregation, and in clusters with varying black hole retention fractions. In combination, the two processes can lead to a wide variety of evolutionary paths on the radius–age plane, which fully cover the observed cluster distribution and hence define a dynamical origin for the radius–age trend in the Magellanic Clouds. We discuss in some detail the implications of core expansion for various aspects of globular cluster research, as well as the possibility of observationally inferring the presence of a significant population of stellar mass black holes in a cluster.     

Basic N-body modelling of the evolution of globular clusters – I. Time scaling

We consider the use of N -body simulations for studying the evolution of rich star clusters (i.e. globular clusters).The dynamical processes included in this study are restricted to gravitational (point-mass) interactions, the steady tidal field of a galaxy, and instantaneous mass loss resulting from stellar evolution. With evolution driven by these mechanisms, it is known that clusters fall roughly into two broad classes: those that dissipate promptly in the tidal field, as a result of mass loss; and those that survive long enough for their evolution to become dominated by two-body relaxation.
The time-scales of the processes we consider scale in different ways with the number of stars in the simulation, and the main aim of the paper is to suggest how the scaling of a simulation should be done so that the results are representative of the evolution of a 'real' cluster. We investigate three different ways of scaling time. One of these is appropriate to the first type of cluster, i.e. those that dissipate rapidly; similarly, a second scaling is appropriate only to the second (relaxation-dominated) type. We also develop a hybrid scaling, which is a satisfactory compromise for both types of cluster. Finally we present evidence that the widely used Fokker–Planck method produces models that are in good agreement with N -body models of those clusters that are relaxation-dominated, at least for N -body models with several thousand particles, but that the Fokker–Planck models evolve too fast for clusters that dissipate promptly.     

Central energy equipartition in multimass models of globular clusters

In the construction of multimass King–Michie models of globular clusters, an approximated central energy equipartition between stars of different mass is usually imposed by scaling the velocity parameter of each mass class inversely with the stellar mass, as if the distribution function were isothermal. In this paper, this 'isothermal approximation' has been checked and its consequences on the model parameters studied by a comparison with models including central energy equipartition correctly. It is found that, under the isothermal approximation, the 'temperatures' of a pair of components can differ to a non-negligible amount for low concentration distributions. It is also found that, in general, this approximation leads to a significantly reduced mass segregation in comparison with that given under the exact energy equipartition at the centre. As a representative example, an isotropic three-component model fitting a given projected surface brightness and line-of-sight velocity dispersion profiles is discussed. In this example, the isothermal approximation gives a cluster envelope much more concentrated (central dimensionless potential   W = 3.3  ) than under the true equipartition  ( W = 5.9 × 10⁻²)  , as well as a higher mass function logarithmic slope. As a consequence, the inferred total mass (and then the global mass-to-light ratio) is a factor of 1.4 times lower than the correct value and the amount of mass in heavy dark remnants is 3.3 times smaller. Under energy equipartition, the fate of stars having a mass below a certain limit is to escape from the system. This limit is derived as a function of the mass and W of the component of giant and turn-off stars.     

The influence of gas expulsion and initial mass segregation on the stellar mass function of globular star clusters

Recently, De Marchi, Paresce & Pulone studied a sample of 20 globular clusters and found that all clusters with high concentrations have steep stellar mass functions while clusters with low concentration have comparatively shallow mass functions. No globular clusters were found with a flat mass function and high concentration. This seems curious since more concentrated star clusters are believed to be dynamically more evolved and should have lost more low-mass stars via evaporation, which would result in a shallower mass function in the low-mass part.
We show that this effect can be explained by residual-gas expulsion from initially mass segregated star clusters, and is enhanced further through unresolved binaries. If gas expulsion is the correct mechanism to produce the observed trend in the   c –α  -plane, then observation of these parameters would allow to constrain cluster starting conditions such as star formation efficiency and the time-scale of gas expulsion.     

Post-Newtonian N-body simulations

We report on the first fully consistent conventional cluster simulation which includes terms up to the third-order post-Newtonian approximation. Numerical problems for treating extremely energetic binaries orbiting a single massive object are circumvented by employing the special 'wheel-spoke' regularization method of Zare which has not been used in large- N simulations before. Idealized models containing   N = 1 × 10⁵  particles of mass  1 M_⊙  with a central black hole (BH) of  300 M_⊙  have been studied on GRAPE-type computers. An initial half-mass radius of   r _h≃ 0.1  pc is sufficiently small to yield examples of relativistic coalescence. This is achieved by significant binary shrinkage within a density cusp environment, followed by the generation of extremely high eccentricities which are induced by Kozai cycles and/or resonant relaxation. More realistic models with white dwarfs and 10 times larger half-mass radii also show evidence of general relativity effects before disruption. An experimentation with the post-Newtonian terms suggests that reducing the time-scales for activating the different orders progressively may be justified for obtaining qualitatively correct solutions without aiming for precise predictions of the final gravitational radiation wave form. The results obtained suggest that the standard loss-cone arguments underestimate the swallowing rate in globular clusters containing a central BH.     

The evolution of the binary population in globular clusters: a full analytical computation

I present a simplified analytical model that simulates the evolution of the binary population in a dynamically evolving globular cluster. A number of simulations have been run spanning a wide range in initial cluster and environmental conditions by taking into account the main mechanisms of formation and destruction of binary systems. Following this approach, I investigate the evolution of the fraction, the radial distribution, the distribution of mass ratios and periods of the binary population. According to these simulations, the fraction of surviving binaries appears to be dominated by the processes of binary ionization and evaporation. In particular, the frequency of binary systems changes by a factor of 1–5 depending on the initial conditions and on the assumed initial distribution of periods. The comparison with the existing estimates of binary fractions in Galactic globular clusters suggests that significant variations in the initial binary content could exist among the analysed globular cluster. This model has been also used to explain the observed discrepancy found between the most recent N -body and Monte Carlo simulations in the literature.     

Scaling of N-body calculations

We report results of collisional N -body simulations aimed at studying the N dependence of the dynamical evolution of star clusters. Our clusters consist of equal-mass stars and are in virial equilibrium. Clusters moving in external tidal fields and clusters limited by a cut-off radius are simulated. Our main focus is to study the dependence of the lifetimes of the clusters on the number of cluster stars and the chosen escape condition.
We find that star clusters in external tidal fields exhibit a scaling problem in the sense that their lifetimes do not scale with the relaxation time. Isolated clusters show a similar problem if stars are removed only after their distance to the cluster centre exceeds a certain cut-off radius. If stars are removed immediately after their energy exceeds the energy necessary for escape, the scaling problem disappears.
We show that some stars that gain the energy necessary for escape are scattered to lower energies before they can leave the cluster. As the efficiency of this process decreases with increasing particle number, it causes the lifetimes not to scale with the relaxation time. Analytic formulae are derived for the scaling of the lifetimes in the different cases.