Constraining the Black Hole Mass Spectrum with LISA Observations II: Direct comparison of detailed models

TL;DR

This paper assesses how well LISA can differentiate between various models of massive black hole populations by analyzing simulated gravitational wave data and considering parameter estimation errors.

Contribution

It introduces a direct method to incorporate LISA's parameter estimation errors into model comparison, enhancing understanding of LISA's potential to constrain black hole formation scenarios.

Findings

LISA can distinguish between different MBH population models with high confidence.

Parameter estimation errors significantly affect the ability to differentiate models.

Using multiple parameters improves model discrimination.

Abstract

A number of scenarios have been proposed for the origin of the supermassive black holes (SMBHs) that are found in the centres of most galaxies. Many such scenarios predict a high-redshift population of massive black holes (MBHs), with masses in the range 100 to 100000 times that of the Sun. When the Laser Interferometer Space Antenna (LISA) is finally operational, it is likely that it will detect on the order of 100 of these MBH binaries as they merge. The differences between proposed population models produce appreciable effects in the portion of the population which is detectable by LISA, so it is likely that the LISA observations will allow us to place constraints on them. However, gravitational wave detectors such as LISA will not be able to detect all such mergers nor assign precise black hole parameters to the merger, due to weak gravitational wave signal strengths. This paper explores LISA's ability to distinguish between several MBH population models. In this way, we go beyond predicting a LISA observed population and consider the extent to which LISA observations could inform astrophysical modellers. The errors in LISA parameter estimation are applied with a direct method which generates random sample parameters for each source in a population realisation. We consider how the distinguishability varies depending on the choice of source parameters (1 or 2 parameters chosen from masses, redshift or spins) used to characterise the model distributions, with confidence levels determined by 1 and 2-dimensional tests based on the Kolmogorov-Smirnov test.