2011MNRAS.416..972D

There is currently no accepted theoretical framework for the formation of the most massive stars, and the manner in which protostars continue to accrete and grow in mass beyond ∼10 M_☉ is still a controversial topic. In this study we use several prescriptions of stellar accretion and a description of the Galactic gas distribution to simulate the luminosities and spatial distribution of massive protostellar population of the Galaxy. We then compare the observables of each simulation to the results of the Red MSX Source (RMS) survey, a recently compiled data base of massive young stellar objects (YSO). We find that the observations are best matched by accretion rates which increase as the protostar grows in mass, such as those predicted by the turbulent core and competitive accretion (i.e. Bondi–Hoyle) models. These `accelerating accretion' models provide very good qualitative and quantitative fits to the data, though we are unable to distinguish between these two models on our simulations alone. We rule out models with accretion rates which are constant with time, and those which are initially very high and which fall away with time, as these produce results which are quantitatively and/or qualitatively incompatible with the observations. To simultaneously match the low- and high-luminosity YSO distribution we require the inclusion of a `swollen-star' pre-main-sequence phase, the length of which is well-described by the Kelvin–Helmholz time-scale. Our results suggest that the lifetime of the YSO phase is ∼10⁵ yr, whereas the compact H II region phase lasts between ∼2 and 4 {x} 10⁵ yr depending on the final mass of the star. Finally, the absolute numbers of YSOs are best matched by a globally averaged star formation rate for the Galaxy of 1.5–2 M_☉.