We use a novel cluster identification tool StarGO to explore the metal poor ([Fe/H] $<$ -1.5) outer stellar halo (d $>$ 15 kpc) of the Milky Way using data from Gaia, LAMOST and SDSS. Our method is built using an unsupervised learning algorithm, a self-organizing map, which trains a 2-D neural network to learn the topological structures of a data set from an n-D input space. Using a 4-D space of angular momentum and orbital energy, we identify three distinct groups corresponding to the Sagittarius, Orphan, and Cetus Streams. For the first time we are able to discover a northern counterpart to the Cetus stream. We test the robustness of this new detection using mock data and find that the significance is more than 5-sigma. We also find that the existing southern counterpart bifurcates into two clumps with different radial velocities. By exploiting the visualization power of StarGO, we attach MW globular clusters to the same trained neural network. The Sagittarius stream is found to have five related clusters, confirming recent literature studies, and the Cetus stream has one associated cluster, NGC 5824. This latter association has previously been postulated, but can only now be truly confirmed thanks to the high-precision Gaia proper motions and large numbers of stellar spectra from LAMOST. The large metallicity dispersion of the stream indicates that the progenitor cannot be a globular cluster. Given the mean metallicity of the stream, we propose that the stream is the result of a merger of a low-mass dwarf galaxy that hosted a large nuclear star cluster (NGC 5824).
تحميل البحث