Rotation, activity, and lithium abundance in cool binary stars^††thanks: Based on data obtained with the STELLA robotic telescopes in Tenerife, an AIP facility jointly operated with IAC, and the Automatic Photoelectric Telescopes in Arizona, jointly operated with Fairborn Observatory.

Time-series high-resolution echelle spectroscopy was taken with the 1.2 m STELLA-I telescope between June 2006 and May 2012. Most spectra were exposed just long enough to measure a precise radial velocity and had S/N of between 40–80:1 but several spectra per target were exposed to reach S/N well above 100:1. A total of 6,609 spectra for a total of 60 stars were obtained over the course of approximately six years. STELLA-I is a fully robotic telescope that, together with STELLA-II, makes up the STELLA observatory at the Izãna ridge on Tenerife in the Canary islands (Strassmeier et al. [2004], [2010]). The fiber-fed STELLA Echelle Spectrograph (SES) is the telescope’s only instrument. It is a white-pupil design with an R2 grating with two off-axis collimators, a prism cross disperser and a folded Schmidt camera with an E2V 2k$\times$2k CCD as the detector. All spectra have a fixed format on the CCD and cover the wavelength range from 388–882 nm with increasing inter-order gaps near the red end starting at 734 nm towards 882 nm. The resolving power is $R$=55,000 corresponding to a spectral resolution of 0.12 Å at 650 nm (3-pixel sampling). An example spectrum is shown in Fig. 1. We note that the SES received a major upgrade in summer 2012 with a new cross disperser, a new optical camera, and a new CCD. A bit earlier, the SES fiber was moved to the prime focus of the second STELLA telescope in 2011. Further details of the performance of the system were reported by Weber et al. ([2012]) and Granzer et al. ([2010]).

SES spectra are automatically reduced using the IRAF¹¹1The Image Reduction and Analysis Facility is hosted by the National Optical Astronomy Observatories in Tucson, Arizona at URL iraf.noao.edu.-based STELLA data-reduction pipeline (Weber et al. [2011]). The images were corrected for bad pixels and cosmic-ray impacts. Bias levels were removed by subtracting the average overscan from each image followed by the subtraction of the mean of the (already overscan subtracted) master bias frame. The target spectra were flattened by dividing by a nightly master flat which has been normalized to unity. The nightly master flat itself is constructed from around 50 individual flats observed during dusk, dawn, and around midnight. After removal of the scattered light, the one-dimensional spectra were extracted with the standard IRAF optimal extraction routine (Horne [1986]). The blaze function was then removed from the target spectra, followed by a wavelength calibration using consecutively recorded Th-Ar spectra. Finally, the extracted spectral orders were continuum normalized by dividing them with a flux-normalized synthetic spectrum of the same spectral classification as the target in question.

Johnson-Cousins $V(I{)}_C$ and/or Strömgren $by$ photometry of 56 of the targets were obtained with the Amadeus (T7) and/or the Wolfgang (T6) automatic photoelectric telescope (APT) at Fairborn Observatory in southern Arizona, respectively. These 56 stars are listed in Table 6 along with the analysis results. A total of 16,591 observations in either $V(I{)}_C$ or $by$ pairs were obtained. Typically, one APT observation consists of three ten-second (30 s for Strömgren) integrations on the variable, four integrations on the comparison star, two integrations on the check star, and two integrations on the sky. A 30\arcsec diaphragm was used for all targets. The standard error of a nightly mean for Amadeus from the overall seasonal mean changed over the past decade but was typically 4–6 mmag in $V$ and 6–8 mmag in $I_C$ for the brightness range of stars in this paper. The observing seasons between 2007–2009 showed increasing scatter due to a slowly but systematic malfunction of the acquisition CCD camera. By late 2009 it got so bad that we had to exchange the entire camera plus its CCD and after that, starting with HJD 2,455,143, the scatter was back to the original values quoted above. The standard error of a nightly mean for Wolfgang was between 1.2–3 mmag, depending on system brightness. For further details we refer to Strassmeier et al. ([1997]) and Granzer et al. ([2001]).

From concurrent observations of Johnson standards in $V(I{)}_C$, we also deduce an all-sky solution and apply it to the differential values whenever feasible. Its accuracy never significantly exceeds 0\fm01 in $V$ though. In case $B$ data were taken, $B-V$ can be determined only relative to the comparison star because our all-sky standards were not observed in the $B$ band due to time constraints. Absolute errors are typically around 0\fm01 for $\Delta (V-I_C)$ except for the time period 2007–2009, as mentioned above, and was then likely 0\fm02.

Radial velocities were determined from an order-by-order cross correlation with a synthetic template spectrum from an ATLAS9 atmosphere (Kurucz [1993]) that roughly matches the target spectral classification. A two-dimensional cross correlation is performed in case the target is double lined. Each of 19 selected echelle orders gives one relative radial velocity. These velocities are then weighted according to the spectral region depending on the amount of available spectral lines and are then averaged. The true internal error is likely close to the rms of these relative velocities. Numerical simulations of the cross correlation errors from synthetic spectra and various two-Gaussian fits to its peaks are discussed in the paper by Weber & Strassmeier ([2011]). Note that the external rms values were significantly larger during the initial year of STELLA operation in 2006/07 ($\approx$120 m s${}^{-1}$), compared to thereafter ($\approx$30 m s${}^{-1}$). The final radial velocities of our program stars are barycentric and corrected for Earth rotation. No gravitational redshift corrections are applied.

Table 2 lists our standard-star measurements. A total of 22 radial-velocity standards were monitored since mid 2006. Their velocities were derived similarly from cross correlations with a synthetic template spectrum that fitted the star’s classification. Among these standards are often observed bright giant stars like $\beta$ Oph, $\alpha$ Ari, $\beta$ Gem, $\alpha$ Tau, 16 Vir, a.o., which are all IAU radial-velocity standards (see Scarfe et al. [1990]). These spectra are also used to identify two epochs in our data when all stars on the observing menu showed a constant offset in velocity. These epochs are identified with two of our hardware maintenance episodes where the fiber injection unit had been readjusted. The average offsets determined from above standard stars were $-0.244$ km s${}^{-1}$ and $-0.366$ km s${}^{-1}$ for the epochs 2,453,902–4,085 and 2,454,192–435, respectively. These corrections were applied to all data in this paper.

The preliminary zero point determination in Strassmeier et al. ([2010]) from four standard stars ($\beta$ Oph, $\alpha$ Cas, $\beta$ Gem, $\beta$ Cet) showed an offset of the STELLA spectra by +0.46 km s${}^{-1}$ with respect to the CORAVEL system defined by Udry et al. ([1999]). In the present paper, we revise this value to +0.503 km s${}^{-1}$ from 22 standard stars. The weights applied for this average shift are the total number of spectra per radial-velocity standard. Note that an absolute heliocentric zero point for STELLA has not been determined yet, e.g. by using asteroids. All velocities in the present paper are on the STELLA system.

Table 3 lists the stars in our sample that were found to exhibit no orbital-induced radial velocity variations, i.e. are presumably single stars. These stars are mostly still very active stars and further investigation with the present time-series data are forthcoming. Several of them, e.g. SAO 150676 or HD 76799, show periodic radial-velocity variations with amplitudes of the order of 1 km s${}^{-1}$, which we interpret to be due to stellar rotation. Few of our single-star survey targets in paper I were also revisited by Griffin ([2010]), most notably HD 136655 and HD 184591, for which we recommend to read his “Notes on the six constant-velocity stars” for their observational history.

For single-lined (SB1) as well as double-lined (SB2) binaries we solve for the components using the general least-square fitting algorithm MPFIT (Markwardt [2009]). For solutions with non-zero eccentricity we use the prescription from Danby & Burkardt ([1983]) to calculate the eccentric anomaly. Usually, we first determine the orbit for both components separately and then combine the two and give rms values for both. Eclipsing binaries were treated slightly separately due to the Rossiter-McLaughlin effect in the radial-velocity curve. Some of the velocities around the two conjunctions were discarded by applying 3-$\sigma$ clipping. Most of the orbits are sampled nearly perfectly due to robotic scheduling (Granzer et al. [2001]) but the unexpected high eccentricity in three systems left some phase gaps during periastron. The STELLA time coverage does not exceed 5.9 years, typical values are around 3.3 years. Although partly compensated for by the high velocity precision and the dense sampling, systems with periods longer than, say, 200 days, have comparably uncertain orbital periods. Whenever there were literature data for the eight systems with orbital periods in excess of $\approx$100 d that expanded the total time range and improved the orbit, these are included in the period determination. This was the case for HIP 999, HD 105575(ab)c and HD 202109. Otherwise, the orbits are computed from STELLA radial velocities alone. The resulting elements and their errors are summarized in Table 4 and Table 5 for the single-lined and the double-lined systems, respectively. Assuming that the fit is of good quality, we derive the element uncertainties by scaling the formal one-sigma errors from the covariance matrix using the measured ${\chi }^2$ values. $T_{per}$ is a time of periastron or, if the orbit is circular, a time of maximum positive radial velocity. Note that we give orbital periods as observed and not corrected for the rest frame of the system.

The computed radial-velocity curves are compared with our observed velocities in the series of panels in Fig. 2 and Fig. 3 for the single-lined orbits and the double-lined orbits, respectively. Note that two of the four triple systems, HIP 999 and HD 105575, appear in both figures. The latter system is even triple lined. The third triple system, HD 237944, is also triple lined and consists of an SB2 in a short orbit but with the third component in a wide thousands-of-years orbit around the close pair. The fourth triple system is HD 45762 where the secondary is again an SB1 but with a period exactly 1/3 of the wide SB2 pair. HD 106855 is a close visual double with its primary A-component to appear double lined, its bona-fide more massive Ab component being again an SB1 for itself. HD 16884 appears to be a quadruple system with two SB1 pairs in a bound 106.65-d orbit. More comments to individual systems are collected in Sect. 5. Note that these comments do not intend to give a comprehensive overview of CDS/Simbad-collected literature but are meant to draw attention to some specific system features.

At the time of the start of our monitoring program in 2006, a total of 17 stars were already either known binaries or had even an orbit computed, or at least had an orbital period known. However, we accumulated enough observations for all systems to allow for orbital solutions just using our own data for a certain epoch. This was motivated not only by consistency arguments but also by detections of orbital-period variations in active binaries and particularly in one of our targets (FF UMa; Gálvez et al. [2007])²²2See our individual notes for this star though.. Orbital-period variations of magnetically active binaries are known for a long time (e.g. Hall & Kreiner [1973]; see also Lanza et al. [2006]) and were partly explained by mild mass exchange and mass loss due to a stellar wind. However, Applegate ([1992]) proposed a connection between the gravitational quadrupole moment and the shape of a magnetically active component in the system as it goes through its magnetic activity cycle. Therefore, it appears rewarding to redetermine the orbits of such binaries from time to time.