How can young massive clusters reach their present-day sizes?

Context. The classic question of how young massive star clusters attain the shapes and sizes, as we find them today, is still a difficult one. Both observational and computational studies of star-forming massive molecular gas clouds suggest that massive cluster formation is primarily triggered along the small-scale (≲0.3 pc) filamentary substructures within the clouds.
Aims: The present study investigates the possible ways in which a filament-like, compact, massive star cluster (effective radius 0.1-0.3 pc) can expand more than 10 times, still remaining massive enough (≳ 10⁴M_⊙) to become the young massive star cluster that we observe today.
Methods: To this end, model massive clusters (initially 10⁴-10⁵M_⊙) are evolved using Sverre Aarseth's state-of-the-art N-body code NBODY7. Apart from the accurate calculation of two-body relaxation of the constituent stars, these evolutionary models take into account stellar-evolutionary mass loss and dynamical energy injection due to massive, tight primordial binaries and stellar-remnant black holes and neutron stars. These calculations also include a solar-neighbourhood-like external tidal field. All the computed clusters expand with time, and their sizes (effective radii) are compared with those observed for young massive clusters (≲ 100 Myr) in the Milky Way and other nearby galaxies.
Results: In this study, it is found that beginning from the above compact sizes, a star cluster cannot expand on its own, I.e., due to two-body relaxation, stellar mass loss, and dynamical heating by primordial binaries and compact stars up to the observed sizes of young massive clusters; star clusters always remain much more compact than the observed ones.
Conclusions: This calls for additional mechanisms that boost the expansion of a massive cluster after its assembly. Using further N-body calculations, it is shown that a substantial residual gas expulsion with ≈ 30% star formation efficiency can indeed swell the newborn embedded cluster adequately. The limitations of the present calculations and their consequences are discussed.