10.6.2 Properties of the input data

This SOS work package analyzed several million systems that were classified as ellipsoidal variables by the union of multiple classifiers after applying different probability cuts. For each source, the Variability Characterization work package (Section 10.2.3) uses a cleaned $G$-band light curve of Gaia (Section 10.2.3), and calculates a Least-Squares frequencygram to derive the frequency of the highest peak $f_{\mathrm{char}}$. The analysis is performed on the cleaned light curve, folded with twice the Characterization period: $P=2P_{\mathrm{char}}=2/f_{\mathrm{char}}$, assuming the pipeline detects the predominant ellipsoidal second-harmonic periodicity, which is half the orbital period. The period error is estimated by $P_{\mathrm{err}}=P(f_{\mathrm{char},\mathrm{err}}/f_{\mathrm{char}})$.

The first selection criteria:

1. 1.

   $P>0.25$ day,

2. 2.

   number of $G$-band FoV transits $>25$,

3. 3.

   $P/P_{\mathrm{err}}>10$,

4. 4.

   $G$-band time range,

ensured a minimal sampling for period search, with results constrained according to the relative period error and the expected range of values.