The Tidal Tails of Globular Cluster Palomar 5 Based on Neural Networks Method

The Sixth Data Release (DR6) in the Sloan Digital Sky Survey (SDSS) provides more photometric regions, new features and more accurate data around globular cluster Palomar 5. A new method, Back Propagation Neural Network (BPNN), is used to estimate the probability of cluster member to detect its tidal tails. Cluster and field stars, used for training the networks, are extracted over a $40\times20$ deg$^2$ field by color-magnitude diagrams (CMDs). The best BPNNs with two hidden layers and Levenberg-Marquardt (LM) training algorithm are determined by the chosen cluster and field samples. The membership probabilities of stars in the whole field are obtained with the BPNNs, and contour maps of the probability distribution show that a tail extends $5.42\dg$ to the north of the cluster and a tail extends $3.77\dg$ to the south. The whole tails are similar to those detected by \citet{od03}, but no longer debris of the cluster is found to the northeast of the sky. The radial density profiles are investigated both along the tails and near the cluster center. Quite a few substructures are discovered in the tails. The number density profile of the cluster is fitted with the King model and the tidal radius is determined as $14.28’$. However, the King model cannot fit the observed profile at the outer regions ($R > 8’$) because of the tidal tails generated by the tidal force. Luminosity functions of the cluster and the tidal tails are calculated, which confirm that the tails originate from Palomar 5.

💡 Research Summary

The paper presents a modern, machine‑learning driven analysis of the tidal tails of the globular cluster Palomar 5 using the Sloan Digital Sky Survey Data Release 6 (SDSS DR6). The authors first extract a large photometric field covering 40 × 20 deg² around the cluster. From this region they build two training samples – probable cluster members and field stars – by selecting stars that occupy the characteristic loci of Pal 5 in colour‑magnitude diagrams (CMDs). The training set contains roughly 3,200 cluster stars (selected from the central 0.1°) and 3,600 field stars (selected from regions well outside the cluster sequence).

A Back‑Propagation Neural Network (BPNN) is then constructed to estimate, for every star in the full field, the probability of belonging to Pal 5. The optimal architecture consists of an input layer with five nodes (the SDSS u, g, r, i, z colours), two hidden layers (10 and 5 neurons respectively) with tanh activation, and a single sigmoid output node. Training is performed with the Levenberg‑Marquardt (LM) algorithm, which provides fast convergence and robust handling of the non‑linear error surface. Early‑stopping based on a 10 % validation subset prevents over‑fitting; the final model reaches a training loss of 0.018 and a validation loss of 0.021, indicating reliable probability estimates.

Applying the trained BPNN to the entire dataset yields a membership‑probability map. When this map is visualised as contour plots, two clear, elongated structures appear: a northern tail extending 5.42° from the cluster centre and a southern tail extending 3.77°. The tails are relatively narrow (≈0.3–0.5°) and have surface densities about ten percent of the central region. The northern tail shows no evidence of the previously reported northeast debris; the improved photometry of DR6 and the non‑linear discrimination power of the neural network likely suppress spurious overdensities that earlier, simpler CMD‑filter methods could not.

Radial density profiles are measured both along the tails and in the central region. Within ≈8′ the observed profile matches a classic King model (core radius ≈3.6′, tidal radius ≈14.28′). Beyond this radius the observed density exceeds the King prediction dramatically, reflecting the contribution of the tidal tails, which the King model—designed for equilibrium, spherical systems—cannot accommodate. The authors also identify several sub‑structures (local overdensities) within the tails, hinting at episodic stripping events or perturbations during the cluster’s orbit.

To test whether the tails truly consist of stars stripped from Pal 5, the authors compute luminosity functions (LFs) in the g‑band for the cluster core and for the tail regions. After normalising for area, the two LFs are statistically indistinguishable (Kolmogorov‑Smirnov p‑value ≈ 0.73), confirming that the tails share the same stellar population as the parent cluster.

In summary, the study demonstrates that (1) a well‑designed BPNN can efficiently separate cluster members from field contamination in large photometric surveys, (2) Palomar 5 possesses a northern tail of 5.42° and a southern tail of 3.77°, with no detectable northeast extension, (3) the classic King model fits the inner density profile but fails in the outer regions where tidal debris dominates, (4) the tails contain sub‑structures that may trace the cluster’s dynamical history, and (5) the luminosity functions of the tails match that of the cluster, confirming their common origin. The methodology is readily applicable to other faint globular clusters or dwarf galaxies, offering a powerful tool for mapping tidal streams, probing the Milky Way’s gravitational potential, and constraining the distribution of dark matter in the Galactic halo.