GPU-accelerated massive black hole binary parameter estimation with LISA

The Laser Interferometer Space Antenna (LISA) is slated for launch in the early 2030s. A main target of the mission is massive black hole binaries that have an expected detection rate of _ 20 yr􀀀1. We present a parameter estimation analysis for a variety of massive black hole binaries. This analysis is performed with a graphics processing unit (GPU) implementation comprising the PhenomHM waveform with higher-order harmonic modes and aligned spins; a fast frequency-domain LISA detector response function; and a GPU-native likelihood computation. The computational performance achieved with the GPU is shown to be 500 times greater than with a similar CPU implementation, which allows us to analyze full noise-infused injections at a realistic Fourier bin width for the LISA mission in a tractable and e_cient amount of time. With these fast likelihood computations, we study the e_ect of adding aligned spins to an analysis with higher-order modes by testing di_erent con_gurations of spins in the injection, as well as the e_ect of varied and _xed spins during sampling. Within these tests, we examine three di_erent binaries with varying mass ratios, redshifts, sky locations, and detector-frame total masses ranging over three orders of magnitude. We discuss varied correlations between the total masses and mass ratios; unique spin posteriors for the larger mass binaries; and the constraints on parameters when _xing spins during sampling, allowing us to compare to previous analyses that did not include aligned spins