ISTEDDAS: a new direct N-body code on GPU to study merging compact-object binaries in star clusters

On February 11th, 2016, the LIGO and Virgo scientific collaborations announced the first direct detection of gravitational waves (GWs), a signal caught by the LIGO interferometers on September 14th, 2015, and produced by the coalescence of two stellar-mass black holes. The discovery represented the beginning of an entirely new way to investigate the Universe. The latest gravitational-wave catalog by LIGO, Virgo, and KAGRA brings the total number of gravitational-wave events to 90, and the count is expected to significantly increase in the next years when additional ground-based and space-born interferometers will be operational. From the theoretical point of view, we have only fuzzy ideas on where the detected events came from, and the answers to most of the five Ws and How for the astrophysics of compact binary coalescences are still unknown. However, two main formation channels have been proposed so far for the formation of merging compact objects (neutron stars - NSs, and black holes BHs). In the isolated binary channel, two progenitor stars are bound since their formation, evolve, and then turn into (merging) compact objects at the end of their life, without experiencing any kind of external perturbation. This scenario is driven by single and binary stellar evolution processes, and it is sometimes referred to as the “field” scenario because it assumes that binaries are born in low-density environments, i.e., that they evolve in isolation. In contrast, in the dynamical channel, two compact objects get very close to each other after one (or more) gravitational interactions with other stars or compact objects. This evolutionary scenario is quite common in dense stellar environments (e.g., star clusters), and it is driven mainly by stellar dynamics. In reality, the two formation pathways might have a strong interplay. In star clusters, the orbital parameters of binaries might be perturbed by many passing-by objects. One of the main issues is that most stars form in dense stellar environments, and numerical simulations of merging compact-object binaries in such crowded stellar systems are very challenging. However, to investigate the origins and the properties of merging compact objects we need a powerful N-body code, which can handle, at the same time, the large spatial and time-evolutionary scales of star clusters (∼pc and ∼Gyr), and the small scales typical of tight binaries (∼AU and ∼days). Therefore, in this thesis, I discuss the innovative algorithms behind isteddas, a new direct N-body code I developed in C++ from scratch that natively supports CUDA to run on Graphics Processing Unit (GPU) supercomputers. I coupled isteddas with the few-body code tsunami, which numerically integrates the orbits of tight systems (e.g., binaries or three-body encounters) with very high accuracy, and also with the population-synthesis code sevn, which includes up-to-date prescriptions for the evolution of both single and binary stars. In this Thesis, I will explain the complex machinery behind isteddas. In particular, the second, third, and fourth chapters are overviews of isteddas, tsunami, and sevn, respectively. In those chapters, I will go through the implementation details of the codes and I will explain how they are interfaced with each other. In the fourth chapter, I will show some results that validate the first version of the isteddas code.