Constraining the parameters of GW150914 & GW170104 with numerical relativity surrogates

Gravitational-wave detectors have begun to observe coalescences of heavy black holes at a consistent pace for the past few years. Accurate models of gravitational waveforms are essential for unbiased and precise estimation of the source parameters, such as masses and spins of component black holes. Recently developed surrogate models based on high-accuracy numerical relativity simulations provide ideal models for constraining physical parameters describing these heavy black hole merger events. In this paper we demonstrate the viability of these surrogate models as reliable parameter estimation tools, and show that within a fully Bayesian framework surrogates can help us extract more information from gravitational wave observations than traditional models. We demonstrate this by analyzing a set of synthetic signals and showing the improvement that the use of numerical relativity surrogates bring to our parameter estimates. We then consider the case of two of the earliest binary black holes detected by the LIGO observatories, GW150914 and GW170104, and reanalyze their data with a generically precessing numerical-relativity-based surrogate model. For these systems we find that overall results are quantitatively consistent with inferences performed with conventional models, except that our refined analysis estimates the sources of both GW150914 and GW170104 to be $10-20\%$ further away than previously estimated and constrain their orientation to be closer to either face-on or face-off configurations more strongly than in the past. Additionally, for GW150914 we constrain the effective spin parameter to be closer to zero. This work is a first step toward eliminating the approximations used in semi-analytic waveform models from GW parameter estimation.