skip to main content

gaia data release 3 documentation

14.2 Internal comparisons

14.2.8 Astrophysical Parameters

The main characteristics and properties of the astrophysical_parameters and astrophysical_parameters_supp tables are described in Creevey et al. (2023), Fouesneau et al. (2023), Delchambre et al. (2023) and Andrae et al. (2023), and caveats are shown in Babusiaux et al. 2023.

Statistics on the astrophysical_parameters and astrophysical_parameters_supp tables are shown in Section 17.1.2 and no major issues are identified.

There are some sources that do not pass the test that checks that the lower percentile is lower than the median value and this, in turn, lower than the upper percentile. The fails are:

  • 273 823 sources failing: lum_flame_lower<lum_flame<lum_flame_upper

  • 114 sources failing: mgfe_gspspec_lower<mgfe_gspspec<mgfe_gspspec_upper

We have checked the distributions for 855 604 sources in the 10×10 field centred on the galactic centre having absorption in temperature in both Gaia DR2 and DR3 (a_g_val and teff_val for Gaia DR2 and ag_gspphot and teff_gspphot for Gaia DR3) and having the same source_id. The strong correlation reported for Gaia DR2 is no longer present and the sources now fill a much larger area of parameter space, as shown in the top panels of Figure 14.32. The bottom panels show that the normalised differences (DR3-DR2) have a strong non-Gaussian component, especially for temperatures.

Figure 14.32: Comparison between absorption and effective temperature in Gaia DR2 and DR3 (top left and right, respectively). Bottom: Normalised differences (DR3-DR2) for absorption (left) and effective temperatures (right).

The maps of median values for a_g_val (Gaia DR2) and ag_gspphot (Gaia DR3) are shown in the top panels of Figure 14.33 and show basically the same features in Gaia DR2 and DR3, though the range of values is now larger. Similarly, in the bottom panels, the median maps of teff_val (Gaia DR2) and teff_gspphot (Gaia DR3) show the strong correlation present in Gaia DR2 and a much smoother behaviour in Gaia DR3. We still notice, though, that the hottest median temperatures are seen in areas with the smallest absorption.

Figure 14.33: Comparison between absorption and effective temperature in Gaia DR2 and DR3. Maps of median values of a_g_val (top left) and teff_val (bottom left) for Gaia DR2 and ag_gspphot (top right) and teff_gspphot (bottom right) for Gaia DR3.

Outlier analysis tables

Figure 14.34: Colour-magnitude diagrams highlighting the classes found by the outlier analysis. Left: highlighting objects labelled as ‘STAR_M’. Right: highlighting objects labelled as ‘QSO’.

The Outlier Analysis tables (oa_neuron_information, oa_neuron_xp_spectra) have been successfully checked for internal inconsistencies. Some issues were found:

  • There are sources (200k / 56M) with missing BP/RP photometry that have non-NULL neuron_oa_id.

  • About 8% (4M / 56M) of the sources with non-NULL neuron_oa_id have GSP-Phot entries. The derived GSP-Phot parameters are not always consistent with the outlier analysis class_label. For example, some of the objects classified as white dwarfs (‘WD’) by the outlier analysis that also have GSP-Phot effective temperatures are cooler than expected (25% of them have Teff<5200 K). Another example is that 10% of the objects classified as ‘STARS_M’ (expected to lie in the Teff range between 2500 K and 3500 K) with GSP-Phot entries are hotter than 4200 K (see left panel of Figure 14.34).

  • More than 3000 objects classified as ‘QSO’ have brighter G magnitudes than the optically brightest quasar, 3C 273 (G=12.8 mag), clearly indicating a mislabelling (see right panel of Figure 14.34). Similar problems exist with other subclasses.

  • The distribution of g_mag_max peaks at 22 for G, 24.5 for BP, and 22.5 for RP. The faintest source included in the OA has a G magnitude of 25.2. Spectra of these sources can be extremely problematic.

  • The bprp_colour_std_dev takes values between 0.1 and 1 mag. This means that for some neurons, the colour dispersion among the sources with supposedly similar spectra is quite high. Looking not at the standard deviation but at the range of GBP-GRP colours of the sources in nodes, these ranges are extremely high even for the best quality categories.

Total Galactic Extinction map tables

The tables total_galactic_extinction_map_opt and total_galactic_extinction_map are consistent with each other, and with the astrophysical_parameters table. The overall appearance of the maps is as expected: large-scale dust filaments are clearly visible where we expect them to be - and in the approximate expected relative intensity. Most of the missing HEALPix pixels are either close to the Galactic plane (as expected) or towards high latitude (for higher HEALPix levels) due to insufficient tracer statistics. In the populated part of the sky, the num_tracers_used to determine the total Galactic extinction can be quite small (=3) in a considerable number (more than 300 000) of HEALPix pixels.

The Magellanic Clouds (including the SMC) are showing up in the extinction maps. The extent of extinction (especially towards the LMC) seems to be overestimated - for a quantitative comparison with other maps over the full sky, see Section 14.4.