Reanalysis of the binary neutron star merger GW170817 using numerical-relativity calibrated waveform models

We reanalyze gravitational waves from a binary-neutron-star merger GW170817 using a numerical-relativity (NR) calibrated waveform model, the TF2+_KyotoTidal model. By imposing a uniform prior on the binary tidal deformability $\tilde{\Lambda}$ the symmetric $90\%$ credible interval of $\tilde{\Lambda}$ is estimated to be $481^{+436}_{-359}$ ($402^{+465}_{-279}$) for the case of $f_\mathrm{max}=1000~\mathrm{Hz}$ ($2048~\mathrm{Hz}$), where $f_\mathrm{max}$ is the maximum frequency in the analysis. We also reanalyze the event with other waveform models: two post-Newtonian waveform models (TF2$\_$PNTidal and TF2+$\_$PNTidal), the TF2+$\_$NRTidal model that is another NR calibrated waveform model used in the LIGO-Virgo analysis, and its upgrade, the TF2+$\_$NRTidalv2 model. While estimates of parameters other than $\tilde{\Lambda}$ are broadly consistent among different waveform models, our results indicate that there is a difference in estimates of $\tilde{\Lambda}$ among three NR calibrated waveform models. The difference in the peak values of posterior probability density functions of $\tilde{\Lambda}$ between the NR calibrated waveform models: the TF2+$\_$KyotoTidal and TF2+$\_$NRTidalv2 models for $f_\mathrm{max}=1000~\mathrm{Hz}$ is about 40 and is much smaller than the width of $90\%$ credible interval, which is about 700. The systematic error for the NR calibrated waveform models will be significant to measure $\tilde{\Lambda}$ in the case of GW170817-like signal for the planned third generation detectors's sensitivities.