Abstract.Although roughly half of all stars are considered to be part of binary or multiple systems, there are only two confirmed cases of RR Lyrae pulsators with companions. One of them is TU Uma [1] -a classical RR Lyrae star in a very eccentric orbit -and the other is OGLE-BLG-RRLYR-02792 [2]. Considering the wealth of well-studied RR Lyrae stars, this number is astoundingly low. Having more RR Lyrae stars in binary systems at hand would be extremely valuable to get independent measurements of the masses. The data from the Kepler mission with their unprecedented precision and the long time span of about four years offer a unique possibility to systematically search for the signatures of binarity in RR Lyrae stars. Using the pulsation as a clock, we studied the variations in the timing of maximum light to hunt for possible binary systems in the sample.
Methods and dataFor our search we made use of the full time span of data provided by the Kepler mission, i.e., about 4 years (Q0-Q17). All the stars in our sample have been observed in long cadence (LC, ∼29 min sampling) during the whole operational time of the mission. Short cadence (SC, ∼1 min sampling) data are available at least during one quarter for each star.About half of all RR Lyrae stars pulsating in the radial fundamental mode (RRab stars) are known to be modulated by the so-called Blazhko effect. This phenomenon can lead to periodic phase or period changes which would be hard to disentangle from any changes due to binarity. Even though this would in principle be possible, we here restricted our search to the 18 non-Blazhko stars in the sample. A detailed analysis of the Q0-Q5 data of these stars has been published by [3], but no search for a light time effect or period fluctuations had been performed.For our study we used the classical O-C technique, where O stands for "observed", and C for "calculated". The time of maximum light is measured for each pulsation cycle, yielding the "O" value. On the other hand, an expected time of maximum, "C", is predicted assuming a constant pulsation period. The difference between the two values is of great diagnostic value, as period changes caused by binarity, continuous period changes due to evolution, and abrupt changes all cause different patterns.To measure the exact time of maximum light from the light curves, a polynomial fit was computed for each pulsation cycle. As most of the data were available in LC only, and RRab stars have sharp maxima, a fit with free parameters showed large scatter. It turned out to be very sensitive to the location of the data points with respect to the light maximum. Fitting a template (i.e., a polynomial with fixed parameters) is a more robust method to accurately determine the time of maximum light even with sparse sampling. We created templates for each star based on the SC data. These templates were then cross-correlated with the LC light curves. The resulting scatter in the O-C diagram is about 1 minute. Using phase diagrams of three periods decreased the scatter to below one...