Simulating the formation of gravitational-wave sources from star clusters and isolated binaries
Main Author: BANERJEE, Sambaran (HISKP/AIfA)
Contact e-mail: sambaran@uni-bonn.de
Abstract
The ongoing detection of gravitational waves (GW) from merging double-compact binaries by the LIGO-Virgo-KAGRA (LVK) interferometer network has generated widespread interest on the origins of GW sources. A widely explored formation mechanism is dynamical encounters involving stellar-origin black holes (BH) inside dense, relaxation-driven, gravitationally bound stellar systems, such as globular clusters and young massive clusters. I will present data and results from an ongoing suite of direct N-body simulations, i.e., star-by-star-resolved simulations of evolution of model star clusters. These simulations, performed with the state-of-the-art N-body code NBODY7, incorporate gravitational interactions among the cluster members without any force-softening or reduction of the star-by-star resolution. That way, the dynamical interactions and relaxation of the cluster are treated to the fullest physics details. Additionally, NBODY7 incorporates evolution of the stars and binaries and post-Newtonian effects. Among many interesting outcomes, GW mergers of binary black holes occur in these simulations naturally, and consistently with the properties of the observed GW events. These computations comprise a key part of my ongoing DFG-funded project. Marvin, particularly its CUDA-enabled A40 GPU cluster, greatly facilitates such simulations by allowing to engage, typically, 4 GPUs and 64 CPU-threads per run. About 5-7 ∼ 105M⊙ star clusters can be evolved for at least a few billion years of age over a month. I am also engaged in studying evolution of massive stellar binaries - another debated formation channel of GW events. To obtain reliable statistics, many millions of binaries need to be evolved. The challenge here is evolving such huge binary populations, and mining the resulting data. Marvin’s CPU servers greatly facilitate these by allowing for parallel computing over a large number of threads (typically 64 threads per ∼ 106 -binary-set; 6-8 sets simultaneously). Some ∼ 107 binaries can be evolved and analysed within only 6-8 hours.
Banerjee, S., MNRAS, 500, 3002–3026 (2021), arXiv:2004.07382
Banerjee, S. and Olejak, A., arXiv:2411.15112 (revision after review in A&A)
Figure 1: Overall evolution of a low-metallicity, initially 105M⊙ model star cluster for 10 billion years. The blue lines show the time evolution of the Lagrangian radii (from lower to upper 1, 2, 5, 10, 20, 30, 40, 50, 67.5, 75, and 90 percent) of the luminous cluster members. The black and grey lines are the half-mass radii of the cluster’s BH and neutron star (NS) subpopulations, respectively. The BHs (about 70 of them at formation), being by far the heaviest members, segregate promptly to the cluster’s central region and remain concentrated therein. Close, energetic dynamical interactions in this dense ‘BH-core’ lead to the formation of binary BHs and their GW mergers. Such interactions also cause the sustained expansion of the cluster, termed as ‘BH-heating’, and prevent NSs from segregating to the cluster’s center. The noise in the curves arise from fluctuations due to the finite member numbers, and granularity of the cluster’s gravitational field. This particular computation, among others, has been performed on Marvin
Figure 2: Effective spin, χeff, versus mass ratio, q, of BH-BH GW mergers, as obtained from a modelled isolated evolution of 2×106 massive binaries at four metallicities, Z. The figure demonstrates a characteristic anti-correlation between χeff and q, particularly for the mass-ratio-reversed BHBHs where the second-formed BH is more massive than the first-formed one. Interestingly, the to-date-observed GW events by the LVK do suggest the existence of such a χeff−q anti-correlation