Simulation showing how symmetric flux variations can introduce a spurious RV shift for an asymmetric line profile. All the units are arbitrary. Left column: initial line profile (black curve) compared to the profile modified by the active regions (red curve). Middle column: RV derivation according to the linearisation of the Doppler shift (Bouchy et al. 2001). The RV is the weighted average of the local RVs. The uncertainties are at first order inversely proportional to the derivative of the flux. The zero RV (dashed line) is compared to the measured RV (blue line). Right column: RV derived using the barycentre of the CCF. First row: a 0.1 Doppler shift on a symmetric line profile, the Bouchy and CCF method extract the correct RV shift. Second row: change in width of a line induced by Zeeman broadening, pressure broadening, or other effects (Brandt & Solanki 1990). The extracted RV shift is null because each modification of the left wing is compensated for by the same modification of the right wing. Third row: depth modification induced by temperature-sensitive lines. Again, no RV shift is measured. Fourth–sixth rows: same as the three first rows, but a blend was inserted on the right side of the line. We note that in this case, the symmetric variations introduce a spurious RV shift. This RV shift would be inverted if the blend were inserted on the left wings instead of the right wing.