Optimizing Parameter Space of Numerical Relativity Simulations for Eccentric Binary Black Hole Systems through Neural Network

Gravitational waveforms carry critical information about the properties of black holes, such as their masses, spins, and type of orbit. Numerical relativity simulations help model these waveforms, providing accurate predictions of what we expect to observe from binary black hole mergers. This work addresses the need to optimally populate the parameter space of binary black hole mergers and improve our understanding of how eccentricity affects the emitted gravitational waves. We present an efficient approach for calculating mismatches between gravitational waveforms from aligned spin eccentric binary black hole systems using a neural network built with TensorFlow. Our neural network focuses primarily on eccentric systems since observations from LIGO display signs of eccentricity in black hole mergers, and future detectors are expected to detect such systems more frequently. Understanding how eccentricity influences gravitational waveforms will allow for more precise parameter estimation and improve the ability to measure eccentricity in observed systems. The network is trained on gravitational waveforms obtained from numerical simulations that we carried out as well as public waveforms from the MAYA catalog. This waveform data incorporates varying mass, spin, and eccentricity. We demonstrate that the machine learning-based framework introduced provides an efficient and reliable means to compute mismatches for binary black hole systems, improving conventional computational techniques.