Simulated three-body gravitational encounters used for training an XGBoost classifier to predict binary formation

This dataset contains N-body simulations of three-body gravitational encounters, used to study binary formation and train machine learning models for binary formation prediction. Methodology:The simulations were performed using REBOUND IAS15 N-body integrator that solves the Newtonian equations of motion for three initially unbound bodies. Data Structure:The dataset is split into multiple CSV files, organized into two categories:- no_binary_dataset_XXX.csv: Simulations where no binary formed (bin == False)- binary_formed_dataset_XXX.csv: Simulations where a binary formed (bin == True) Each CSV file contains 5,000 rows of simulation data. The data files are in CSV format with the following column structure: id: Unique simulation identifier (string) sim_e_err: Energy conservation error sim_time: Total simulation time t_enc: Encounter time r_enc: Encounter radius m1, m2, m3: Masses of the three bodies xi1, yi1, zi1, xi2, yi2, zi2, xi3, yi3, zi3: Initial positions for bodies 1, 2, and 3 bx1, by1, bz1, bx2, by2, bz2, bx3, by3, bz3: Impact parameter vectors for bodies 1, 2, and 3 vxi1, vyi1, vzi1, vxi2, vyi2, vzi2, vxi3, vyi3, vzi3: Initial velocities for bodies 1, 2, and 3 theta1, phi1, theta2, phi2, theta3, phi3: Angular coordinates for bodies 1, 2, and 3 xf1, yf1, zf1, xf2, yf2, zf2, xf3, yf3, zf3: Final positions for bodies 1, 2, and 3 vxf1, vyf1, vzf1, vxf2, vyf2, vzf2, vxf3, vyf3, vzf3: Final velocities for bodies 1, 2, and 3 bin: Boolean (True/False) indicating whether a binary formed a, e, p: Binary orbital parameters (semi-major axis, eccentricity, period) - populated only if bin == True E: Total orbital energy of the binary hx, hy, hz: Specific angular momentum components of the binary inc: Inclination of the binary orbit bin_comp1, bin_comp2: Indices of the two bodies that formed the binary escapee: Index of the escaping body All simulations use G = 4 * pi ^ 2 in au, year, solar mass units. Usage Notes:This dataset was generated to provide training and testing data for machine learning models predicting binary formation in three-body encounters. The data is suitable for supervised learning tasks where the target variable is the 'bin' column. Associated Publication:Farhani Asl, Ahmad (2026). Machine learning prediction of binary formation in three-body gravitational encounters. A&A, in press. @misc{asl2026machinelearningpredictionbinary, title={Machine learning prediction of binary formation in three-body gravitational encounters}, author={Ahmad Farhani Asl}, year={2026}, eprint={2607.16776}, archivePrefix={arXiv}, primaryClass={astro-ph.GA}, url={https://arxiv.org/abs/2607.16776}, } Please cite this dataset with the paper above.