Jacob Lange

University of Texas Austin, March 29, 12:00 pm

On the Importance of Expanding and Improving Numerical Relativity Simulations

Since the original breakthrough of the numerical relativity (NR) evolution codes, multiple research groups have been able to solve Einstein’s equations numerically on supercomputers to simulate merging binary black hole gravitational wave sources. While these NR simulations can be directly compared to the data using special configurations parameter estimation codes, these waveforms are mostly used to calibrate and verify the accuracy semi-analytical gravitational wave models. These models are in turn used for production level parameter estimation analyses of gravitational wave detections from the LIGO-Virgo-KAGRA Collaboration. Because of this, it is essential not only to expand our current NR simulation grid into more exotic parts of parameter space (i.e. more unequal mass ratios, more extreme precessing spins, inclusion of eccentricity) but also to improve the accuracy of our current simulations as we prepare for next generation detectors. For the former, I present a novel algorithm that suggests new NR simulation placement based off of the interpolated likelihood as well as the error to that interpolated likelihood when directly comparing the NR waveforms to a real event. This allows the placement of new simulations to be in relevant parts of parameter space (high likelihood) as well as in sparse parts of parameter space for the existing grid (high error in interpolated likelihood). For the latter, I present parameter estimation results investigating the effects of waveform systematics due to NR resolution. With our current expected signal-to-noise ratios (SNR) for signals from our current ground-based detectors, the resolutions used to generate the current catalogs of NR simulations are considered large enough to be indistinguishable from an infinite resolution simulation (i.e. statistical errors dominate). As our current and next generation detectors increase our sensitivity, the systematic errors due to our finite resolution becomes more important. Following up from a previous mismatch study by Ferguson et. al. comparing different resolutions for a given simulation using different detectors, I present preliminary work that quantifies the impact of resolution error by injecting these different resolution simulations at different SNRs based off the criteria investigated in Ferguson et. al.