Dataset Open Outputs from Gaia Net for Gaia XP and Gaia RVS spectra

This repository contains the tables from the paper "Gaia Net: Towards robust spectroscopic parameters of stars of all evolutionary stages" by Hydson et al. (2025, ApJ)

We present a new processing of XP spectra for 220 million stars released in Gaia DR3. The new data model is capable of handling objects with Teff between 2000 and 50,000 K, and with log g between 0 and 10, including objects of multitude of masses and evolutionary stages. This includes for the first time ever robust processing of spectroscopic parameters for pre-main sequence stars, with log g sensitivity towards their age. Through this analysis we examine the distribution of young low mass stars with ages of up to 20 Myr in the solar neighborhood, and we identify a new massive (>1000 stars) population, Ophion, which is found east of Sco Cen. This population appears to be fully disrupted, with negligible kinematic coherence. Nonetheless, due its young age it appears to still persist as a spacial overdensity. Through improved determination of ages of the nearby stars, it may be possible to better recover star forming history of the solar neighborhood outside of the moving groups.

One of the tables in the repository are the spectroscopic parameters derived from Gaia XP spectra, the other - from Gaia RVS spectra. 

Both tables have the same data structure, see below for an example.

source_id
ra
dec
logTeff
e_logTeff
logg
e_logg
FeH
e_FeH

4295806720
44.99615537864534
0.005615226341865997
3.740103
0.0057907104
4.497857
0.011642456
-0.8710861
0.09863281

38655544960
45.004978371745516
0.019879675701858644
3.6812372
0.0014762878
4.520626
0.023925781
-0.12406255
0.06124878

1275606125952
44.993270784169155
0.07633404499591856
3.718066
0.007709503
4.3352485
0.118774414
-0.30459395
0.119018555

Parquet table format has been integrated into standard software and libraries for table management, including TOPCAT, as well as Astropy tables and Pandas, e.g.,

from astropy.table import Tablet=Table.read('filename.parquet')

or

import pandas as pddf=pd.read_parquet('filename.parquet')

The tables are also duplicated in the AAS MRT format with truncated decimal precision for greater compression.