Core Formation, Coherence and Collapse: A New Core Evolution Paradigm Revealed by Machine Learning

We study the formation, evolution and collapse of dense cores by tracking density structures in a magnetohydrodynamic (MHD) simulation. We identify cores using the dendrogram algorithm and utilize machine learning techniques, including principal component analysis (PCA) and the k-means clustering algorithm to analyze the full density and velocity dispersion profiles of these cores. We find that there exists an evolutionary sequence consisting of three distinct phases: i) the formation of turbulent density structures (Phase I), ii) the dissipation of turbulence and the formation of coherent cores (Phase II), and iii) the transition to protostellar cores through gravitational collapse (Phase III). In dynamically evolving molecular clouds, the existence of these three phases corresponds to the coexistence of three populations of cores with distinct physical properties. The prestellar and protostellar cores frequently analyzed in previous studies of observations and simulations belong to the last phase in this evolutionary picture. We derive typical lifetimes of 1.4$\pm$1.0$\times$10$^5$ yr, 3.3$\pm$1.4$\times$10$^5$ yr and 3.3$\pm$1.4$\times$10$^5$ yr, respectively for Phase I, II and III. We find that cores can form from both converging flows and filament fragmentation and that cores may form both inside and outside the filaments. We then compare our results to previous observations of coherent cores and provide suggestions for future observations to study cores belonging to the three phases.