# 6.2.3 Auxiliary data

The input data of the spectroscopic pipeline include also some auxiliary data. The auxiliary data are:

The ground-based radial velocities of standard stars are used in wavelength calibration to fix the zero point (Section 6.3).

The pipeline uses the 2568 standard stars described in Soubiran et al. (2018), for which the radial velocity has been obtained with high accuracy. In addition, another 5729 standard stars have been selected from XHIP measurements, having uncertainties $e_{\rm RV}<1$  km s${}^{-1}$ and quality A and B, and having a match in the master list in Crifo et al. (2010). The list of 5729 standard stars is the result of the cleaning done using the daily pipeline results. The RVS daily observations and the dry run of the spectroscopic pipeline permitted to measure the radial velocities of most of these stars in various transits; they were excluded from the list if the radial velocity measured with the RVS differed by more than 3  km s${}^{-1}$ from the one provided. Table 6.3 lists the catalogues of standard stars used for calibration, together with the catalogues used for validation.

## Atmospheric parameters

The knowledge of the atmospheric parameters of the stars is essential for the pipeline. The parameters are used to select the appropriate synthetic spectrum to generate a template as much as possible similar to the RVS spectrum, which is then used to estimate, via Pearson correlation with the RVS spectrum (Section 6.4.7), the spectroscopic radial velocity of the star. The atmospheric parameter information is also used to select the stars having a spectral type required by the calibration module. For example, wavelength calibration needs spectra with deep lines, and the input spectra selection criterium uses $T_{\rm eff}$: 4500 $<$ $T_{\rm eff}$ $<$ 6500 K (it uses also magnitude).

In Gaia DR2 the spectroscopic processing took place before the atmospheric parameters processing (Chapter 8) and could not have access to the $T_{\rm eff}$ produced with Gaia data. For this reason, a list of auxiliary atmospheric parameters obtained using ground-based data are used in the pipeline. The auxiliary parameter list contains: $T_{\rm eff}$ information for 1.8 million bright stars, some are taken from the literature and described in Soubiran et al. (2014), and some have been estimated using the 2MASS photometric catalogue, $\log g$ for 633 050 stars and [Fe/H] for 650 505 stars, all taken from the literature.

Approximately 15 % of the stars treated in Gaia DR2 have the auxiliary ground-based atmospheric parameters associated. The treatment for the other stars is described in Section 6.4.4.

## Synthetic spectral libraries

The synthetic spectra are used to produce the templates (i.e. the synthetic spectrum is convolved with the RVS LSF-AL, and re-sampled, see Section 6.4.7.

A large set of synthetic spectra has been calculated in Sordo et al. (2011), for both the low-resolution spectrophotometry and the medium-resolution RVS data. The RVS spectra have been recalculated in 2016; in particular the MARCS library grid has been extended to the cool stars, several line lists have changed and the continuum opacity treatment has been improved.

The spectroscopic pipeline uses 5256 spectra selected among the following grids:

• MARCS spectra: $T_{\rm eff}$: 2500–8000 K, step 100 K in the range 2500–3900 and 250 K in the range 4000–8000; $\log g$: -0.5–+5.0, step 0.5; [Fe/H]: -5.0, -4.0, -3.0, -2.5, -2.0, -1.5, -1.0, -0.75, -0.50, -0.25, 0.0, +0.25, +0.5, +0.75, +1.0;

• A-type spectra: $T_{\rm eff}$: 8500–15 000 K, step 500 K; $\log g$: 0.5–5, step 0.5; [Fe/H]: -0.5–+0.25, step 0.25;

• OB-type spectra $T_{\rm eff}$: 15 000–55 000 K, step 1000 K in the range 15 000–30 000 K, step 2500 K in the range 30 000–50 000 K; $\log g$: highest value: 4.75, lowest: approximately linearly from 1.75 at 15 000 K to 4.0 at 55 000 K; [Fe/H]: -0.3, 0.0, +0.3.

For more information on the synthetic spectra used in the spectroscopic pipeline see Blomme et al. (2017).

## Restricted library of template spectra

To reduce the computation time in some of the software modules, a subset of the synthetic spectra dataset (Section 6.2.3) was used, consisting of the 28 models listed in Table 6.2. This selection is based on the expected number of stars of a given spectral type/luminosity class, determined from the Gaia Universe Model Snapshot (Robin et al. 2012, specifically their Tables 16 and 17).

The subset is used notably in the modules DetermineAP (Section 6.4.4) and TodCor (Section 6.4.7).