10.4.3 Processing steps

The selection of the variable AGN sample was performed by comparing several properties of the candidates with those of the Gaia-CRF3 sample. A series of consecutive filters were applied, with the goal of removing the maximum number of contaminants from the sample and at the same time losing the minimum number of Gaia-CRF3 objects. The selection procedure included filtering on the:

• •

  structure_function_index, i.e. the slope in the $\log \mathrm{SF}$ versus $\log \tau$ space, where $\tau$ is the time lag;

• •

  non_qso_variability versus qso_variability metrics, which were derived from the adaptation of the Butler and Bloom (2011) statistics to the Gaia data;

• •

  distribution of the ruwe parameter versus abbe_mag_g_fov value for $G$ FoV transits;

• •

  distribution of the $G-G_{\mathrm{RP}}$ versus $G_{\mathrm{BP}}-G$ colour indices;

• •

  parallax and pm for normal distributions, as expected from objects at cosmological distances;

• •

  maximum density of sources within 100${}^{\mathrm{\prime \prime }}$to reduce AGN misclassifications in crowded stellar fields;

• •

  upper limit to the scan angle correlation with photometric variations (suggesting artificial variability);

• •

  minimum general variability detection probability to focus on objects with clear variability.

Details on the computation of the parameters, their distributions, and the applied constraints are presented in Carnerero et al. (2023).

At the end of the above procedure, the sample consists of $\sim$869 000 sources. Further $\sim$3000 variable AGNs were added, which did not pass the above selection procedure because of peculiar properties (e.g. blazars). Therefore, the final variable AGN sample includes about 872 000 sources, $\sim$722 000 of which are already present in the Gaia-CRF3 sample. This left approximately 150 000 variable AGN candidates that are not in the Gaia-CRF3 sample and therefore require further investigation.

Table 10.1 reports the 0.01, 0.25, 0.5, 0.75, and 0.99 percentiles of the four parameters in the vari_agn table, calculated on the published variable AGN sample. An additional parameter, vari_agn_membership_score, is published in qso_candidates table; see Chapter 12), the computation of which is based on the above four parameters plus the abbe_mag_g_fov value.