Publications ofthe Astronomical Society ofthe Pacific 99:490-496, June 1987 RADIAL VELOCITIES OF M DWARF STARS GEOFFREY W. MARCY,* VICTORIA LINDSAY,* AND KAREN WILSON Department of Physics and Astronomy, San Francisco State University, San Francisco, California 94132 Received 1987 February 26, revised 1987 March 28 ABSTRACT Radial velocities for 72 M dwarfs have been obtained having internal errors of about 0.1 km s1 and external errors of about 0.4 km s_1. Multiple velocity measurements of ten dMe stars have yielded a set of six which have no stellar companions, providing confirmation that the dMe phenomenon can occur in single stars. These single dMe stars have low space motions indicative of relative youth. Four stars from the entire survey were found to have double-line spectra and two were found to be single-line spectroscopic binaries of low amplitude. The zero point of the velocity scale is found to agree well with that of O. C. Wilson (1967) and differences are noted among other radial-velocity studies. Most of the stars in this study have velocities sufficiently well determined to constitute potential radial-velocity standards. Key words: K-M dwarfs-radial velocities I. Introduction atic difference between their velocities and O. C. 1 Accurate measurements of radial velocities for M Wilson's of —1.7 km s , a discrepancy not explainable by dwarfs are used to address a number of astrophysical the random errors of the two studies. problems such as the age dependence of the kinematics of Thus, it is not known at present which zero point is to the Galaxy (cf. Mayor 1974; Wielen 1974), the existence of be preferred. Here we present velocities of 72 M dwarfs which are internally consistent to better than 0.1 km s-1 substellar companions (Marcy et al. 1986), and the nature of magnetic/chromospheric activity (Bopp and Meredith and which are anchored to an absolute velocity scale 1986; Young, Sadjadi, and Harlan 1987). The most widely determined by high-resolution spectra of a subsample of used set of M-dwarf velocities has been that established the same stars. by O. C. Wilson (1967), both because of the large size of II. Observations the sample and because individual absorption lines were During a two-year period from June 1983 to June 1985, measured, yielding a well-defined zero point. The inter- 1 a major program was carried out at the 100-inch (2.5 m) nal errors in Wilson s study were about 2 km s and the velocities were systematically different from those in the telescope of the Mount Wilson Observatory to obtain 1 highly accurate radial velocities of relatively faint stars. General Catalogue (Wilson 1953) by only 0.5 km s . Recently, Young et al. (1987) have similarly obtained The primary goal of the project was the detection of velocities with an accuracy of 1 to 2 km s1 for 48 M dwarfs single-line spectroscopic binaries of extremely low ampli- in the solar neighborhood. tude. While this principal aim continues (now at Lick Cross-correlation methods, both analogue and digital Observatory), the data in hand permit determination of (Griffin 1967; Simkin 1974), are particularly well suited to absolute velocities for many of the program stars. A sam- M dwarfs owing to their rather complicated spectra and ple of M dwarfs was chosen from the list of Joy and Abt their relative faintness. Cross-correlation velocities have (1974) with the following selection criteria: spectral type been recently obtained for a large number of M dwarfs by later than dM2 (though a few bright dMO stars were Stauffer and Hartmann (1986) and by Bopp and Meredith included), accessibility by the Mount Wilson 100-inch (1986). All of these measurements have quoted internal telescope (dec. < 50°), brighter than V = 11.5, and void of errors of about 1 km s1. Difficulties arise, however, in companions within 10 arc sec. The final sample is not the determination of the zero points of each velocity complete but contains 80% of all of the single M dwarfs system. This problem stems primarily from the lack of later than dM2 and brighter than V = 10.5 that are listed velocity standards, especially for M dwarfs later than Ml. in the Gliese catalogue (Gliese 1969). The coudé spectrograph of the Mount Wilson 100-inch For example, Stauffer and Hartmann (1986) find a system- telescope was used with the 114-inch (2.9 m) camera and a new grating (graciously loaned by the Jet Propulsion Lab- * Guest Observer, Lick Observatory. oratories) to yield a reciprocal dispersion of 1.0 A mm-1. 490 © Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System M DWARF STARS 491 The intensified Reticon detector (Shectman 1981) was velocities obtained on that night. Typically, 20 stars are employed because it produces essentially no noise itself observed per night, so that this nightly correction should and is therefore ideal for the anticipated low signal-to- have an error of about a/V2Ö, where σ is the internal noise applications. This setup yields a pixel spacing of error. Tests of repeatability during a night show that σ is about 0.010 A corresponding to 500 m s1 at λ = 6000 Α. about 150 m s-1, yielding nightly systematic errors of less Several significant structural modifications were made to than about 40 m s1. In effect, the program stars them- the spectrograph to ensure stability of the optics during selves provide the nightly reference velocity scale (with a the course of a night. Careful steps were taken to ensure zero point to be discussed below) and stars found to be that both the optical axis and/-ratio of the telescope were variable are not used as references. matched by those of all calibration sources. All observa- The determination of the zero point of the velocity scale tions were made within three hours of the meridian. has been carried out with extremely high-resolution spec- As the major source of error was anticipated to be the tra of a subset of the program stars, enabling measure- nonuniform illumination of the slit (cf. Griffin and Griffin ment of individual absorption lines. The spectra were 1973), the slit width was chosen to be 0.35 arc sec. This, of obtained with the echelle spectrograph and image-tube course, did not eliminate that source of error. Indeed, scanner located at the coudé focus of the Lick Observa- experience showed that the poorest velocities occurred tory 120-inch (3-m) telescope (Soderblom 1982; Robinson on nights of sub-arc-second seeing, a common problem at and Wampler 1972). The spectrograph was used in single- Mount Wilson. pass mode and was fed with the 0.6-m coudé auxiliary The observing procedure involves obtaining a spec- telescope. The spectra cover 13 angstroms with a recipro- trum of the comparison source, a thorium hollow-cathode cal dispersion of 0.4 A mnT1 and are centered at λ6496, a lamp, both before and after each stellar spectrum. The region containing four deep, narrow, and relatively un- typical exposure time for a star with V = 11 is 30 minutes. blended lines. The spectra were reduced using the stan- These comparison spectra are used to monitor the zero dard reduction package written by M. Hartoog. point of the wavelength scale during the night. At the The four lines used are; X6493.782 (Cal), λ6494.985 beginning and end of each night long exposures are made (Fei), X6498.943 (Fei), and X6499.650 (Caí). These of the thorium source to determine the shape of the adopted wavelengths were taken from Kurucz and dispersion function in detail. Peytremann (1975) and are systematically different from Data reduction involves correcting the raw spectra for those given by Moore, Minnaert, and Houtgast (1966) by varying pixel sensitivity and the determination of a wave- an amount (latter — former) = 0.0043 A, implying a possi- length scale using a fifth-order Legendre polynomial fit to ble systematic error in the final stellar velocities of 0.20 the thorium lines. All spectra (stellar and "book-end" km s1. Velocities were obtained by determining the first thoriums) are rebinned in wavelength, and the small moment of each absorption line. Of the four lines wavelength drifts of the book-end comparison spectra are X6498.943 became unusable in some of the fainter and determined. These drifts are used to correct the zero later-type M dwarfs and was rejected. point of the wavelength scale of the stellar spectra. The In Table I we identify (by Gliese number) the M dwarfs observed drift per hour, typically less than 100 m s-1, is observed with the echelle. In the second column is given effectively removed with this procedure. the spectral type from Joy and Abt (1974), in the third The Doppler shift of each stellar spectrum is now deter- column is given the UT date of observation, and in the mined relative to a template which is a composite of 12 fourth column is shown the uncertainty in the mean stars in the sample and is designed to be representative of velocity of the four absorption lines used in the determi- the range of M dwarf subclasses among program stars. nation. This uncertainty typically lies between 0.2 km s-1 The resulting cross-correlation function is fit to a parabola and 0.6 km s_1 and reflects not only random errors owing near the peak, yielding the velocity of the program star to photon statistics but also effects due to blends which relative to the template. Much of the cross-correlation analysis was built upon existing routines kindly provided TABLE I by Alan Dressier.