Testing the universality of the IMF with Bayesian statistics: young clusters

TL;DR

This study uses Bayesian statistics to analyze eight young stellar clusters, finding significant variations in IMF parameters that challenge the notion of a universal initial mass function across different environments.

Contribution

It introduces a Bayesian framework to compare IMF parameters across clusters, demonstrating non-universality with multiple likelihood functions and accounting for observational uncertainties.

Findings

IMF parameters vary significantly between clusters.

Most probable IMF parameters differ from the Galactic field IMF.

Variations are robust even after completeness and uncertainty corrections.

Abstract

The universality of the stellar initial mass function (IMF) is tested using Bayesian statistics with a sample of eight young Galactic stellar clusters (IC 348, ONC, NGC 2024, NGC 6611, NGC 2264, $\rho$ Ophiuchi, Chameleon I, and Taurus). We infer the posterior probability distribution function (pPDF) of the IMF parameters when the likelihood function is described by a tapered power law function, a lognormal distribution at low masses coupled to a power law at higher masses, and a multi-component power law function. The inter-cluster comparison of the pPDFs of the IMF parameters for each likelihood function shows that these distributions do not overlap within the $1\sigma$ uncertainty level. Furthermore, the most probable values of the IMF parameters for most of the clusters deviate substantially from their values for the Galactic field stellar IMF. We also quantify the effects of taking into account the completeness correction as well as the uncertainties on the measured masses. The inclusion of the former affects the inferred pPDFs of the slope of the IMF at the low mass end while considering the latter affects the pPDFs of the slope of the IMF in the intermediate- to high mass regime. As variations are observed in all of the IMF parameters at once and for each of the considered likelihood functions, even for completeness corrected samples, we argue that the observed variations are real and significant, at least for the sample of eight clusters considered in this work. The results presented here clearly show that the IMF is not universal.