Hierarchical accretion flow from the G351 infrared dark filament to its central cores

Context. Quantifying the accretion flow from large cloud scales down to individual protostars is a central ingredient to the understanding of (high-mass) star formation. Aims. We characterize and quantify this multi-scale flow for a prototypical high-mass star-forming region. Methods. In a multi-scale analysis from parsec to ∼50 au scales, we combined multiple single-dish and interferometric observations to study the gas flow from large-scale sizes of several parsec (Mopra) via intermediate-scale filamentary gas flows (ALMA-IMF) to the central cores (ALMA DIHCA and configuration 10 data). The highest-resolution multi-configuration ALMA dataset achieved a spatial resolution of 0.027″ × 0.022″ or 50 au. Results. This multi-scale study allows us to follow the gas from the environment of the high-mass star-forming region (∼2 pc) via intermediate-scale (∼0.25 pc) filamentary gas flows down to the innermost cores within the central few 1000 au. The intermediatescale filaments connect spatially and kinematically to the larger-scale cloud as well as the innermost cores. We estimate a filamentary mass inflow rate around 10−3 M⊙ yr−1, feeding into the central region that hosts at least a dozen mm cores. While the flow from the cloud via the filaments down to 104 au appears relatively ordered, within the central 104 au the kinematic structures become much more complicated and disordered. We speculate that this is caused by the interplay of the converging infalling gas with feedback processes from the forming central protostars. Conclusions. This multi-scale study characterises and quantifies the hierarchical gas flow from clouds down to the central protostars for a prototypical infrared dark cloud with several embedded cores at an unprecedented detail. While comparatively ordered gas flows are found over a broad range of scales, the innermost area exhibits more disordered structures, likely caused by the combination of inflow, outflow and cluster dynamical processes.