Extreme mass ratio inspirals (EMRIs), the inspirals of compact objects into supermassive black holes, are important gravitational wave sources for the Laser Interferometer Space Antenna (LISA). We study the performance of various post-Newtonian (PN) template families relative to the waveforms that are high-precision numerical solutions of the Teukolsky equation in the context of EMRI parameter estimation with LISA. Expressions for the time-domain waveforms TaylorT1, TaylorT2, TaylorT3, TaylorT4 and TaylorEt are derived up to 22 PN order, i.e. $\mathcal{O}(v^{44})$ ($v$ is the characteristic velocity of the binary) beyond the Newtonian term, for a test particle in a circular orbit around a Schwarzschild black hole. The phase difference between the above 22 PN waveform families and numerical waveforms are evaluated during two-year inspirals for two prototypical EMRI systems with mass ratios ${10}^{-4}$ and ${10}^{-5}$. We find that the dephases (in radians) for TaylorT1 and TaylorT2, respectively, are about ${10}^{-9}$ (${10}^{-2}$) and ${10}^{-9}$ (${10}^{-3}$) for mass ratio ${10}^{-4}$ (${10}^{-5}$). This suggests that using 22 PN TaylorT1 or TaylorT2 waveforms for parameter estimation of EMRIs will result in accuracies comparable to numerical waveform accuracy for most of the LISA parameter space. On the other hand, from the dephase results, we find that TaylorT3, TaylorT4 and TaylorEt fare relatively poorly as one approaches the last stable orbit. This implies that, as for comparable mass binaries using the 3.5 PN phase of waveforms, the 22 PN TaylorT3 and TaylorEt approximants do not perform well enough for the EMRIs. The reason underlying the poor performance of TaylorT3, TaylorT4 and TaylorEt relative to TaylorT1 and TaylorT2 is finally examined.

• Received 2 May 2013

DOI:https://doi.org/10.1103/PhysRevD.88.024038