Analytic systematics in next generation of effective-one-body gravitational waveform models for future observations

The success of analytic waveform modeling within the effective-one-body (EOB) approach relies on the precise understanding of the physical importance of each technical element included in the model. The urgency of constructing progressively more sophisticated and complete waveform models (e.g. including spin precession and eccentricity) partly defocused the research from a careful comprehension of each building block (e.g. Hamiltonian, radiation reaction, ringdown attachment). Here we go back to the spirit of the first EOB works. We focus first on nonspinning, quasicircular, black hole binaries and analyze systematically the mutual synergy between numerical relativity (NR) informed functions and the high post -Newtonian corrections (up to 5PN) to the EOB potentials. Our main finding is that it is essential to correctly control the noncircular part of the dynamics during the late plunge up to merger. When this happens, either using NR-informed nonquasicircular corrections to the waveform (and flux) or high-PN corrections in the radial EOB potentials oD; Q thorn , it is easy to obtain EOB/NR unfaithfulness similar to 10-4 with the noise of either Advanced LIGO or 3G detectors. We then improve the TEOBResumS-GIOTTO waveform model (dubbed TEOBResumSv4.3.2) for quasicircular, spin-aligned black hole binaries. We obtain maximal EOB/NR unfaithfulness F over bar max EOBNR similar to 10-3 (with Advanced LIGO noise and in the total mass range 10-200M circle dot) for the dominant l = m = 2 mode all over the 534 spin-aligned configurations available through the Simulating eXtreme Spacetime catalog. The model performance, also including higher modes, is then explored using the NR surrogates NRHybSur3dq8 and NRHybSur2dq15, to validate TEOBResumSv4.3.2 up to mass ratio m1/m2 = 15. We find that, over the set of configurations considered, more than 98% of the total-mass -maximized unfaithfulness lie below the 3% threshold when comparing to the surrogate models.