We present the discovery of a wide retrograde moving group in the disk plane of the Milky Way using action-angle coordinates derived from the Gaia DR3 catalog. The structure is identified from a sample of its members that are currently almost at the pericenter of their orbit and are passing through the solar neighborhood. The motions of the stars in this group are highly correlated, indicating that the system is probably not phase mixed. With a width of at least 1.5 kpc and with a probable intrinsic spread in metallicity, this structure is most likely the wide remnant of a tidal stream of a disrupted ancient dwarf galaxy (age ∼12 Gyr, 〈[Fe/H]〉 ∼ −1.74). The structure presents many similarities (e.g., in energy, angular momentum, metallicity, and eccentricity) with the Sequoia merging event. However, it possesses extremely low vertical action J z , which makes it unique even among Sequoia dynamical groups. As the low J z may be attributable to dynamical friction, we speculate that these stars may be the remnants of the dense core of the Sequoia progenitor.

We report on the discovery in the Gaia DR3 astrometric and spectroscopic catalog of a new polar stream that is found as an overdensity in action space. This structure is unique as it has an extremely large apocenter distance, reaching beyond 100 kpc, and yet is detected as a coherent moving structure in the solar neighborhood with a width of ∼4 kpc. A subsample of these stars that was fortuitously observed by LAMOST has a mean spectroscopic metallicity of 〈 [ Fe / H ] 〉 = − 1.60 − 0.16 + 0.15 dex and possesses a resolved metallicity dispersion of σ ( [ Fe / H ] ) = 0.32 − 0.06 + 0.17 dex. The physical width of the stream, the metallicity dispersion, and the vertical action spread indicate that the progenitor was a dwarf galaxy. The existence of such a coherent and highly radial structure at their pericenters in the vicinity of the Sun suggests that many other dwarf galaxy fragments may be lurking in the outer halo.

Symbolic Regression is the study of algorithms that automate the search for analytic expressions that fit data. While recent advances in deep learning have generated renewed interest in such approaches, efforts have not been focused on physics, where we have important additional constraints due to the units associated with our data. Here we present Φ-SO, a Physical Symbolic Optimization framework for recovering analytical symbolic expressions from physics data using deep reinforcement learning techniques by learning units constraints. Our system is built, from the ground up, to propose solutions where the physical units are consistent by construction. This is useful not only in eliminating physically impossible solutions, but because it restricts enormously the freedom of the equation generator, thus vastly improving performance. The algorithm can be used to fit noiseless data, which can be useful for instance when attempting to derive an analytical property of a physical model, and it can also be used to obtain analytical approximations to noisy data. We showcase our machinery on a panel of examples from astrophysics.