Gaussian mixture models of the total mass distribution of stellar black holes from LIGO-Virgo GWTC-2: Implications on the origin of GW190521

The recently discovered very massive stellar black holes (BHs), i.e., the binary BH (BBH) merger for the gravitational-wave (GW) event GW190521 by LIGO-Virgo O3a with the component masses of ${85}_{\ensuremath{-}14}^{+21}\text{ }\text{ }{M}_{\ensuremath{\bigodot}}$ and ${66}_{\ensuremath{-}18}^{+17}\text{ }\text{ }{M}_{\ensuremath{\bigodot}}$, have aroused much attention. Our statistical knowledge of the mass distributions of the BBHs observed in LIGO-Virgo GW detections is helpful to understanding the origin of GW190521. Hence, in order to determine whether GW190521 is consistent with the same population as the other LIGO-Virgo BBHs, we analyze the BBH total masses in the GWTC-2 catalog with the classical Gaussian mixture models (GMMs). Based on the model selection results in this analysis, we cannot confidently conclude that GW190521 is a new population of BBHs distinct from the other LIGO-Virgo BBHs due to the limitation of the current small sample size of GW-observed BBHs. Moreover, our results suggest that there is a clear preference for a two-component GMM for the sample of BBHs in GWTC-2 without GW190521, implying that two (or more) distinct underlying populations of these BBHs may exist because of the differing astrophysical origins. After a consideration for the detailed features of the above two-component GMM, we find statistical evidence that tends to support a mixture model for LIGO-Virgo BBH formation, which is in line with previous studies that also supported a multicomponent model (for example, one consisting of isolated binary evolution and dynamical channels).