Radial velocity variations in the young eruptive star EX Lupi

Context. EX Lup-type objects (EXors) are low-mass pre-main sequence objects characterized by optical and near-infrared outbursts attributed to highly enhanced accretion from the circumstellar disk onto the star. Aims: The trigger mechanism of EXor outbursts is still debated. One type of theory requires a close (sub)stellar companion that perturbs the inner part of the disk and triggers the onset of the enhanced accretion. Here, we study the radial velocity (RV) variations of EX Lup, the prototype of the EXor class, and test whether they can be related to a close companion. Methods: We conducted a five-year RV survey, collecting 54 observations with HARPS and FEROS. We analyzed the activity of EX Lup by checking the bisector, the equivalent width of the Ca 8662 A line, the asymmetry of the Ca II K line, the activity indicator SFEROS, the asymmetry of the cross-correlation function, the line depth ratio of the VI/FeI lines, and the TiO, CaH 2, CaH 3, CaOH, and Halpha indices. We complemented the RV measurements with a 14-day optical/infrared photometric monitoring to look for signatures of activity or varying accretion. Results: We found that the RV of EX Lup is periodic (P = 7.417 d), with stable period, semi-amplitude (2.2 km s-1), and phase over at least four years of observations. This period is not present in any of the above-mentioned activity indicators. However, the RVs of narrow metallic emission lines suggest the same period, but with an anti-correlating phase. The observed absorption line RVs can be fitted with a Keplerian solution around a 0.6 Ms central star with msini = (14.7 ± 0.7) MJup and eccentricity of e = 0.24. Alternatively, we attempted to model the observations with a cold or hot stellar spot as well. We found that in our simple model, the spot parameters needed to reproduce the RV semi-amplitude are in contradiction with the photometric variability, making the spot scenario unlikely. Conclusions: We qualitatively discuss two possibilities to explain the RV data: a geometry with two accretion columns rotating with the star, and a single accretion flow synchronized with the orbital motion of the hypothetical companion; the second scenario is more consistent with the observed properties of EX Lup. In this scenario, the companion's mass would fall into the brown dwarf desert, which, together with the unusually small separation of 0.06 au would make EX Lup a unique binary system. The companion also has interesting implications on the physical mechanisms responsible for triggering the outburst.