Sie sind auf Seite 1von 4

Dynamical Eects of Intermediate Mass Black Hole in Globular Clusters

Chenruo Qi
Department of Physics and Astronomy, University of British Columbia 6224 Agricultural Road, Vancouver, British Columbia, Canada, V6T 1Z1 (Dated: November 12, 2011) The black-hole-mass velocity-dispersion relation observed in galaxies suggests the existence of intermediate mass black holes (IMBHs), with masses between 100 M and 10,000 M , at the cores of globular clusters. Black holes of this mass range could bridge the gap between stellar mass black holes and super massive black holes. The velocity dispersion data of several globular cluster cores suggest the presence of IMBH. However, further studies show their existence is not mandatory. We will investigate the dynamical eects of a free-oating IMBH on the evolution of its host globular cluster. A series of simulations with dierent initial IMBH masses are planned to run on a machine with two NVIDIA Tesla GPUs using NBODY6, a direct N-body simulation program. If we are successful, we will be able to determine a more accurate relationship between the mass of IMBH and the properties of the host globular cluster.

I.

MOTIVATION

For a long time, only the stellar mass black hole, which is the product of stellar evolution, and the super-massive black hole, which resides at the center of a galaxy, were known to astronomers. It was suspected that a third class, the intermediate mass black hole (hereafter IMBH) with masses ranging from 100 M to 10,000 M , exists in the cores of globular clusters if the mass-velocity dispersion relation between the mass of the black hole and the stellar velocity dispersion at galactic center applies to globular clusters as well. M15 was the rst globular cluster to show signs of an intermediate mass black hole at its core. The high velocity dispersion measurements implied a 4,000 M IMBH [1]. Later studies with alternative models suggested that the central black hole could be replaced by a concentration of neutron stars and white dwarfs as a result of mass segregation [2]. G1, a globular cluster orbiting the Andromeda galaxy, was also found to harbour potentially a 20,000 M black hole [3]. Further X-ray and radio observations provided more support to the idea [4, 5]. However, an evolutionary model without a central IMBH was found in excellent agreement with the observational data as well [6]. Furthermore, the proposed 40,000 M black hole at the center of Omega Centauri was challenged by a detailed study with rened kinematical center location using proper motion data from HST [7]. The existence of intermediate mass black holes at the centers of globular clusters is still under debate. Since they are buried deep inside the cluster cores, the high concentration of stars crates observational diculties. Despite the lack of hard observational evidence, the formation of IMBH in clusters remains an active research topic in astronomy. There are a variety of IMBH formation theories. Physical collisions of stars are possible in very dense clusters. Massive stars tend to move towards the core due to the

johnqi67@interchange.ubc.ca

loss of kinetic energy during interactions with less massive stars. If massive stars reach the core before they undergo supernova explosions, they could merge and form a super massive star, which will later collapse and form an IMBH [8]. This mechanism is call runaway collision. Similarly, IMBH can form through repeated merging of stellar mass black hole binaries [9]. Gravitational waves carry energy away and eventually lead to the merger of the binary. If there are more than one such binary in a cluster, an IMBH could form at the center. Finally, a recent study done by Vesperini et al. suggests a seed black hole can grow into an IMBH by accreting a small fraction of the gas that is used to form the second generation stars in the cluster [10]. The presence of an IMBH has signicant impacts on the structure, kinematics, and dynamical evolution of the host cluster. There are a few mechanisms to inject energy into a clusters core, such as velocity kicks of white dwarfs and mass loss by stellar winds [11, 12]. Binary burning is the best studied mechanism. When a third star interacts with a binary, the binary hardens by decreasing its orbital separation while kinetic energy is added to the third star. However, most globular clusters have yet to reach the binary burning phase [13]. A central IMBH could provide the energy needed to prevent the core from undergoing core collapse. Furthermore, intermediate mass black hole is likely to form a binary with another stellar mass black hole or neutron star at the center of a cluster. The gravitational radiation emitted by the binary would be a rich source for future ground and space-based detectors. In this project, we are interested in investigating the eects of an IMBH on its host cluster through a series of direct N-body simulations. This study could potentially allow us to establish a relationship between the mass of the central IMBH and the physical properties of its host globular cluster. As a result, it is possible to put constraints on the mass of the IMBH given the morphology of the cluster. Also, we will be able to gain some insight into the growth of IMBHs with dierent masses in globular clusters.

2
II. A. THEORY

Source of Intra-cluster Gas

The gas content in a cluster has been constantly replenished in dierent ways. When the hydrogen supply in the core of a main-sequence star is depleted, the core contracts and heats up. Once the core reaches sucient temperature ( 3 108 K ), helium burning is initiated. After helium burning is completed at the core, the star now has a carbon core enclosed in a helium burning shell. This is the early asymptotic giant branch (E-AGB) stage. When most of the helium in the shell is exhausted, the leftover helium layer is too thin to raise the material above it, thus the temperature increases rapidly. Eventually the accumulated energy is released in the form of a thermal pulse. This late AGB stage is named thermally pulsing AGB (TP-AGB). The pulsation expands the atmosphere so that the temperature is low enough at large distances for molecules to form. Radiation pressure pushes these molecules out, and their collisions with the gas further increase the mass lose rate [14]. This process is expected to produce most of the gas in a cluster. Stars do not suer from mass lose in the TP-AGB stage alone. The stellar wind is constantly ejecting surface materials throughout the life of a star. The mass lose rate of main-sequence stars has a strong dependence on the stellar mass. Wolf-Rayet stars (> 20M ) typically lose 1 105 M per year [15], while the mass lose rate of midsized stars are much less (approximately 1 1014 M annually for the Sun). Streams of ejected material from dierent stars collide with each other and lose angular momentum. Eventually, they fall towards the center of the cluster and feed the growth of the black hole.

FIG. 1. The accretion of two particles, A and B, with initial velocity v, onto the black hole located at BH. C is the location where the collision of the two particles occurs. is the distance from the black hole within which particles do not have sucient radial velocity to escape after collision at point C. Credit: [16]

The accretion rate in this scenario is then


2 2 BH = 2 v = 4G M , M 3 v

(2)

where is the density of the gas cloud and M is the mass of the black hole. Further analysis with the addition of an accretion column in front of the black hole shows the accretion rate is only half of the value in equation (2) [17]. However, this accretion model does not include the eects of uid dynamics. In Bondis study on spherically symmetric accretion onto a point mass, the uid eects were examined in detail [18]. The corresponding accretion rate has a similar form,
2 2 BH = 2G M M 2 2 (cs + v )3/2

B.

Accretion Rate

(3)

Imagine a black hole moving through a cloud of gas, the cloud moves from right to left in the black holes frame as shown in gure 1. Let us now consider the path of particle A with initial velocity v . If the particle comes too close to the black hole, its speed and direction of motion are completely changed. However, if the particle is far away, its velocity remains more or less the same. The characteristic distance that separates the two cases is called the impact parameter. Now, if we consider both particles, A and B, it is clear that a collision will happen at point C in gure 1. The collision destroys the angular momentum of the particles about the black hole, and they will be accreted if the remaining radial velocity is insucient for them to escape. Therefore, we can nd a radius, , within which all particles will be accreted. 2GMBH = v2 (1)

where is the undisturbed density of the gas, and cs is the sound speed. The gas density is related to the change in mass of the clusters gas content by the continuity equation, gas = 4r2 cs , M c (4)

where rc is the core radius of the cluster. Most of the stars in the cluster are located within this radius. cs is the sound speed of the gas. We can solve for , which is = gas M . 2c 4rc s (5)

From equation (3) and (5), we have


2 2 gas M BH = 1 G M M . 2 2 3/2 2 rc cs (c2 s + v )

(6)

3
TABLE I. Cluster node specications and simulation time CUDA JMD CPU 2 X Intel Xeon E5504 AMD Opteron 844 GPU 2 X Nvidia Tesla C1060 Not used RAM 6 X 2GB 2GB Time(h:m:s) 05:12:00 88:24:40

tailed treatments of binaries, close encounters, and stellar evolution. The program has most of the functions necessary for the simulations. Its GPU package enables us to take advantage of the GPU node on JMD. There are a few modications need to be made. 1. Implementation of black hole accretion rate 2. Calculation of mass loss due to stellar wind 3. Position and velocity output for kinematic analysis.
FIG. 2. Advances in clock rate and core count of CPUs (circles) and GPUs (diamonds) in the past decade. Each data point is averaged over two consecutive years. Credit: [19]

4. Potential energy corrections due to black hole accretion. Globular clusters are subject to the gravitational pull of the galaxy they are orbiting, so stars in the outskirts of a cluster are easily lost. The tidal eld of the host galaxy also has impact on the morphology of the cluster, as many globular clusters in the Milky Way and Andromeda galaxy are elliptical [20]. In our simulations, the clusters are treated as isolated. Without the complication of external elds, we can better relate the mass of the IMBH and the clusters morphology. In addition to the isolated cluster assumption, we will not include primordial binaries. Since primordial binaries have signicant inuence on the time scale of core collapse. They will blur the eects of IMBH on the clusters evolution. The two reasonable simplications should help lowering the computational cost of the project. All cluster models assume an initial density prole given by a King model with Wo = 7.0, and the initial mass distribution is modelled by Kroupa mass function with a lower mass limit of 0.1M and an upper limit of 30M . The metallicity of the clusters will be set to 0.001 since globular clusters contains mostly (50%-80%) stars with little metal content. In order to fairly distribute the computing resources among all users, every job running on the JMD cluster is killed after a certain amount of time. This means the simulations will not run continuously from start to nish. Therefore, a simulation strategy is developed. NBODY6 allows the user to set a CPU time, which is the amount of time the program is allowed to run. When the CPU time is exceeded, the current state of the simulation is saved to a le so that it could be continued at a later time by using the restart facility. We decided to run the simulations in alternating order so that we could start data processing and analysis for all simulations before they are nished.

III.

DETAILS ON PROPOSED SIMULATION

Since around 2005, the clock rate of CPU ceased to increase. Instead, manufacturers included more processing cores to deliver higher performance [19]. Mainstream workstations typically have CPUs with 2-8 cores. However, GPUs normally have 30-60 times more cores operating at a lower clock rate. This means GPUs could handle parallel tasks must faster than CPUs. With proper libraries, GPU can even outperform the GRAPE6, a specialized hardware for force calculations in N-body simulations, under certain circumstances. Given the time constraint on the project, we will simulate the evolution of start clusters on a GPU node of Dr. Heyls JMD computing cluster. Based on a previous simulation with 16,000 stars. The performance of the GPU node on JMD is close to 17 times faster than that of the CPU node. The details are shown in Table 1. All simulations for this project are planned to begin with 65,000 stars. The CPU computation cost of dynamical simulations scales as N 3 . After a short calculation, the simulations will take approximate 15 months using CPUs. If the computing time ratio between GPU and CPU also holds for the 65K simulations, data will be produced in a reasonable amount of time. In this project, three clusters will be simulated to examine the inuences of intermediate mass black holes on globular clusters. The models contain free-oating IMBHs of masses 0.3%, 1.0%, and 3.0% of the total mass of the star cluster. Aarseths direct n-body simulation program, NBODY6, will be used. It is built on the feature-rich NBODY5, which is also developed by Sverre Aarseth in late 1970s, with more accurate fourthorder Hermite method for particle integrations and de-

4 After all simulations are done, we will inspect the morphology of the clusters by taking snapshots separated by 10 Myrs near the end of their evolution. Velocity dispersions and RMS velocities will be calculated to study their kinematics. Then, we can relate the properties of the clusters to the masses of the IMBHs they are hosting. In addition, the growth of the black holes could also be determined. This project will give us a more precise relation between globular clusters and the masses of the central IMBHs. lab, is needed for data processing and analysis.

V.

PLANNED SCHEDULE

The code modication of NBODY6 and subsequent testing will be completed at the beginning of January 2012. The simulations are expected to take 8 weeks or more, but there will be some data for the progress report in February. Data processing and analysis should nish in the rst half of March. The thesis write-up will begin around the same time and the rst draft will nish in the third week of March.

IV.

RESOURCES LIST VI. ACKNOWLEDGEMENTS

A computer with two Nvidia Tesla GPUs is required for the simulations. The cuda-matt node on the JMD computing cluster will be used to provide most of the computational needs for this project. In addition, a typical workstation, such as the ones in the physics computer

I would like to thank my supervisor Dr. Jeremy S. Heyl for all the helpful conversations about the BondiHoyle accretion theory in clusters, and the computational resources provided for this project.

[1] Gerssen, J., van der Marel, R. P., Gebhardt, K., Guhathakurta, P., Peterson, R. C., & Pryor, C. 2002, AJ, 124, 3270 [2] Baumgardt, H., Hut, P., Makino, J., McMillan, S., & Portegies Zwart, S. 2003b, ApJL, 582, L21 [3] Gebhardt, K., et al. 2003, ApJ, 583, 92 [4] Pooley, D., & Rappaport, S. 2006, ApJ, 644, L45 [5] Ulvestad, J. S., Greene, J., & Ho, L. 2007, ApJL, accepted [6] Baumgardt, H., Makino, J., Hut, P., McMillan, S., & Portegies Zwart, S. 2003c, ApJL, 589, L25 [7] Anderson, J., & van der Marel, R. P. 2010, ApJ, 710, 1032 [8] Portegies Zwart S. F., Makino J., McMillan S. L. W., Hut P., 1999, A&A, 348, 117 [9] Mouri, H., and Taniguchi, Y. 2002, ApJ, 566, L17

[10] Vesperini, E., McMillan, S. L. W., DErcole, A., & DAntona, F. 2010, ApJL, 713, L41 [11] Davis, D. S., Richer, H. B., King, I. R., Anderson, J., Coey, J., Fahlman, G. G., Hurley, J., & Kalirai, J. S. 2008, MNRAS, 383, L20 [12] Baumgardt, H., & Kroupa, P. 2007, MNRAS, 380, 1589 [13] Fregeau, J. M. 2008, ApJ, 673, L25 [14] Zijlstra, A. A. 2006, in IAU Symposium, Vol. 234, Planetary Nebulae in our Galaxy and Beyond, ed. M. J. Barlow & R. H. M endez, 5562 [15] Langer, N., 1989, A&A 220, 135 [16] Hoyle, F. & Lyttleton, R. A. 1939, Proc. Camb. Phil. Soc., 35, 405 [17] Bondi, H. & Hoyle, F. 1944, MNRAS, 104, 273 [18] Bondi, H. 1952, MNRAS, 112, 195. [19] Barsdell B. R., Barnes D. G., Fluke C. J., 2010, MNRAS, 408, 1936 [20] Han, C., & Ryden, B. S. 1994, ApJ, 433, 80

Das könnte Ihnen auch gefallen