Characterizing NGC 6383: A study of fundamental properties, pre-main sequence stars, mass segregation, and age using Gaia DR3 and 2MASS.
Date
2026
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
Presentamos una caracterización bayesiana integral del cúmulo abierto galáctico joven NGC 6383, combinando astrometría y fotometría de Gaia DR3 con datos infrarrojos cercanos de 2MASS a través de COSMIC (Characterisation Of Star clusters using Machine learning Inference and Clustering), un pipeline de código abierto que integra agrupamiento basado en densidad (HDBSCAN) con modelamiento jerárquico bayesiano mediante PyMC y el muestreador No-U-Turn (NUTS). A partir de una muestra de trabajo de 15 276 fuentes de Gaia DR3 dentro de 40 arcmin del centro del cúmulo, el pipeline identifica 254 miembros probables con un umbral de pseudoprobabilidad ˜pi > 0.6, con un análisis de sensibilidad al umbral que demuestra que la población recuperada varía entre 161 y 321 miembros en cuatro umbrales de probabilidad. Un modelo jerárquico basado en paralajes entrega una distancia d = 1,110 ± 0,060 kpc, confirmada independientemente por el módulo de distancia isocronal μDM = 10,3±0,26 mag. Los movimientos propios medios son μα∗ = 2,540±0,010 mas año−1 y μδ = −1,710±0,009 mas año−1, con dispersiones intrínsecas de ∼ 0,15 mas año−1. El ajuste isocronal bayesiano, realizado con ASteCA utilizando isócronas MIST y muestreo NUTS, produce una edad modal τ = 3,53+1,40 −1,00 Ma, metalicidad Z = 0,024±0,008 y extinción visual AV = 1,24±0,26 mag. Las clasicaciones de estrellas pre-secuencia principal de Sagitta y el parámetro Q libre de enrojecimiento confirman independientemente un cúmulo muy joven con un episodio extendido de formación estelar que abarca ∼ 1–6 Ma y una fracción elevada de objetos estelares jóvenes Yfrac = 0,28 (IC 95% [0,22; 0,34]). El análisis estructural revela un núcleo compacto (Rc = 1,95 ± 0,19 arcmin) inmerso en una envolvente extendida (Rt = 40,4 ± 14,3 arcmin), con concentración C = 1,316 y 39 potenciales estrellas en escape más allá del radio de Hill que comparten la firma cinemática del cúmulo. El tiempo de relajación de media masa trh = 18,4 ± 2,1 Ma supera la edad del cúmulo por un factor ∼ 5 (τ/trh ≈ 0,19), situando a NGC 6383 firmemente en el régimen dinámicamente joven. La segregación de masa entre estrellas binarias y de alta masa persiste cuando la muestra se restringe a estrellas cuyas escalas temporales de fricción dinámica exceden la edad del cúmulo (tseg > τ ; m < 8,29 Mf), constituyendo evidencia de segregación de masa primordial o cuasi-primordial consistente con predicciones de simulaciones de N-cuerpos de fusión jerárquica de sub-grupos. La binaria O7V+O7V HD 159176, históricamente central en los debates sobre la edad de NGC 6383, se demuestra como no-miembro: su movimiento propio se desvía 6,6σ y su paralaje ∼ 60σ respecto a las distribuciones posteriores del cúmulo. Este resultado resuelve una ambigüedad de décadas en la que la membresía asumida de este sistema infaba las estimaciones publicadas de edad a 6–10 Ma. Se identifican cinco contribuciones metodológicas: (i) una arquitectura híbrida de membresía HDBSCAN+pseudoprobabilidad que combina descubrimiento no paramétrico con filtrado iterativo de robustez; (ii) el reemplazo del muestreador por deecto de ASteCA por NUTS, proporcionando distribuciones posteriores exactas y diagnósticos de convergencia geométricos; (iii) una separación formal y doctrinal entre la razón de persistencia de HDBSCAN y las probabilidades de membresía posteriores calibradas; (iv) un análisis de sensibilidad al umbral que cuantifica cómo los resúmenes físicos dependen del corte de membresía; y (v) una disciplina de estatus de afirmaciones que gobierna toda inferencia física, requiriendo condicionamiento explícito sobre el modelo generativo, la especificación de priors y el estado de convergencia. El código fuente de COSMIC, el catálogo de membresía y las muestras posteriores completas se publican abiertamente.
We present a comprehensive Bayesian characterisation of the young Galactic open cluster NGC 6383, combining Gaia DR3 astrometry and photometry with 2MASS near-infrared data through COSMIC (Characterisation Of Star clusters using Machine learning Inference and Clustering), an open-source pipeline that integrates HDBSCAN density-based clustering with hierarchical Bayesian modelling via PyMC and the No-U-Turn Sampler (NUTS). From a working sample of 15 276 Gaia DR3 sources within 40 arcmin of the cluster centre, the pipeline identifies 254 probable members at a pseudoprobability threshold ˜pi > 0.6, with a threshold-sensitivity analysis demonstrating that the recovered population spans 161–321 members across four probability thresholds. A parallax-based hierarchical model yields a distance d = 1.110 ± 0.060 kpc, independently confirmed by the isochronal distance modulus μDM = 10.3 ± 0.26 mag. Mean proper motions are μα∗ = 2.540 ± 0.010 mas yr−1 and μδ = −1.710 ± 0.009 mas yr−1, with intrinsic dispersions of ∼ 0.15 mas yr−1. The Bayesian isochronal fit, performed with ASteCA using MIST isochrones and NUTS sampling, yields a modal age τ = 3.53+1.40 −1.00 Myr, metallicity Z = 0.024 ± 0.008, and visual extinction AV = 1.24 ± 0.26 mag. Sagitta pre-mainsequence classifications and the reddening-free Q-parameter independently confirm a very young cluster with an extended star-formation episode spanning ∼ 1–6 Myr and an elevated young stellar object fraction Yfrac = 0.28 (95% CI [0.22, 0.34]). The structural analysis reveals a compact core (Rc = 1.95±0.19 arcmin) embedded within an extended envelope (Rt = 40.4 ± 14.3 arcmin), with concentration C = 1.316 and 39 potential escapers beyond the Hill radius sharing the cluster’s kinematic signature. The half-mass relaxation time trh = 18.4 ± 2.1 Myr exceeds the cluster age by a factor of ∼ 5 (τ/trh ≈ 0.19), placing NGC 6383 firmly in the dynamically young regime. Mass segregation among binary and high-mass stars persists when the sample is restricted to stars whose dynamical friction timescales exceed the cluster age (tseg > τ ; m < 8.29 Mf), constituting evidence for primordial or nearprimordial mass segregation consistent with predictions from N-body simulations of hierarchical sub-clump mergers. The O7V+O7V binary HD 159176, historically central to age debates about NGC 6383, is demonstrated to be a non-member: its proper motion deviates by 6.6σ and its parallax by ∼ 60σ from the cluster posteriors. This result resolves a decades-long ambiguity in which assumed membership infated published age estimates to 6–10 Myr. Five methodological contributions are identified: (i) a hybrid HDBSCAN+pseudoprobability membership architecture that combines non-parametric discovery with iterative robustness filtering; (ii) the replacement of ASteCA’s default sampler with NUTS, providing exact posteriors and geometric convergence diagnostics; (iii) a formal doctrinal separation between the HDBSCAN persistence ratio and calibrated posterior membership probabilities; (iv) a threshold-sensitivity analysis quantifiying how physical summaries depend on the membership cut; and (v) a claim-status discipline that governs every physical inference, requiring explicit conditioning on the generative model, prior specification, and convergence status. The COSMIC source code, membership catalogue, and full posterior samples are publicly released. thods: data analysis; techniques: photometric; astrometry.
We present a comprehensive Bayesian characterisation of the young Galactic open cluster NGC 6383, combining Gaia DR3 astrometry and photometry with 2MASS near-infrared data through COSMIC (Characterisation Of Star clusters using Machine learning Inference and Clustering), an open-source pipeline that integrates HDBSCAN density-based clustering with hierarchical Bayesian modelling via PyMC and the No-U-Turn Sampler (NUTS). From a working sample of 15 276 Gaia DR3 sources within 40 arcmin of the cluster centre, the pipeline identifies 254 probable members at a pseudoprobability threshold ˜pi > 0.6, with a threshold-sensitivity analysis demonstrating that the recovered population spans 161–321 members across four probability thresholds. A parallax-based hierarchical model yields a distance d = 1.110 ± 0.060 kpc, independently confirmed by the isochronal distance modulus μDM = 10.3 ± 0.26 mag. Mean proper motions are μα∗ = 2.540 ± 0.010 mas yr−1 and μδ = −1.710 ± 0.009 mas yr−1, with intrinsic dispersions of ∼ 0.15 mas yr−1. The Bayesian isochronal fit, performed with ASteCA using MIST isochrones and NUTS sampling, yields a modal age τ = 3.53+1.40 −1.00 Myr, metallicity Z = 0.024 ± 0.008, and visual extinction AV = 1.24 ± 0.26 mag. Sagitta pre-mainsequence classifications and the reddening-free Q-parameter independently confirm a very young cluster with an extended star-formation episode spanning ∼ 1–6 Myr and an elevated young stellar object fraction Yfrac = 0.28 (95% CI [0.22, 0.34]). The structural analysis reveals a compact core (Rc = 1.95±0.19 arcmin) embedded within an extended envelope (Rt = 40.4 ± 14.3 arcmin), with concentration C = 1.316 and 39 potential escapers beyond the Hill radius sharing the cluster’s kinematic signature. The half-mass relaxation time trh = 18.4 ± 2.1 Myr exceeds the cluster age by a factor of ∼ 5 (τ/trh ≈ 0.19), placing NGC 6383 firmly in the dynamically young regime. Mass segregation among binary and high-mass stars persists when the sample is restricted to stars whose dynamical friction timescales exceed the cluster age (tseg > τ ; m < 8.29 Mf), constituting evidence for primordial or nearprimordial mass segregation consistent with predictions from N-body simulations of hierarchical sub-clump mergers. The O7V+O7V binary HD 159176, historically central to age debates about NGC 6383, is demonstrated to be a non-member: its proper motion deviates by 6.6σ and its parallax by ∼ 60σ from the cluster posteriors. This result resolves a decades-long ambiguity in which assumed membership infated published age estimates to 6–10 Myr. Five methodological contributions are identified: (i) a hybrid HDBSCAN+pseudoprobability membership architecture that combines non-parametric discovery with iterative robustness filtering; (ii) the replacement of ASteCA’s default sampler with NUTS, providing exact posteriors and geometric convergence diagnostics; (iii) a formal doctrinal separation between the HDBSCAN persistence ratio and calibrated posterior membership probabilities; (iv) a threshold-sensitivity analysis quantifiying how physical summaries depend on the membership cut; and (v) a claim-status discipline that governs every physical inference, requiring explicit conditioning on the generative model, prior specification, and convergence status. The COSMIC source code, membership catalogue, and full posterior samples are publicly released. thods: data analysis; techniques: photometric; astrometry.
Description
Tesis presentada para optar al grado de Magíster en Astronomía.
Keywords
Star clusters, Astrometry, Photometry