EDP Sciences
Free Access
Issue
A&A
Volume 577, May 2015
Article Number A128
Number of page(s) 19
Section Catalogs and data
DOI https://doi.org/10.1051/0004-6361/201525803
Published online 18 May 2015

© ESO, 2015

1. Introduction

The Calar Alto high-Resolution search for M dwarfs with Exo-earths with Near-infrared and optical Échelle Spectrographs (hereafter CARMENES1) is a next-generation instrument close to completion for the Zeiss 3.5 m Calar Alto telescope, which is located in the Sierra de Los Filabres, Almería, in southern Spain, at a height of about 2200 m (Sánchez et al. 2007, 2008). CARMENES is the name used for the instrument, the consortium of 11 German and Spanish institutions that builds it, and of the scientific project to be carried out during guaranteed time observations. The instrument consists of two separated, highly stable, fibre-fed spectrographs covering the wavelength ranges from 0.55 to 0.95 μm and from 0.95 to 1.70 μm at spectral resolution R ≈ 82 000, each of which shall perform high-accuracy radial-velocity measurements with long-term stability of ~1 m s-1 (Quirrenbach et al. 2010, 2012, 2014, and references therein; Amado et al. 2013). First light is scheduled for the summer of 2015, followed by the commission of the instrument in the second half of that year.

The main scientific objective for CARMENES is the search for very low-mass planets (i.e., super- and exo-earths) orbiting mid- to late-M dwarfs, including a sample of moderately active M-dwarf stars. Dwarf stars of M spectral type have effective temperatures between 2300 and 3900 K (Kirkpatrick et al. 2005; Rajpurohit et al. 2013). For stars with ages greater than that of the Hyades, of about 0.6 Ga, these effective temperatures translate in the main sequence into a mass interval from 0.09 to 0.55 M, approximately (Baraffe et al. 1998; Chabrier et al. 2000; Allard et al. 2011). Of particular interest is the detection of very low-mass planets in the stellar habitable zone, the region around the star within which a planet can support liquid water (Kasting et al. 1993; Joshi et al. 1997; Lammer et al. 2007; Tarter et al. 2007; Scalo et al. 2007). In principle, the lower the mass of a host star, the higher the radial-velocity amplitude velocity induced (i.e., Kstar is proportional to Mplaneta−1/2 (Mstar + Mplanet)−1/2 ≈ (aMstar)−1/2 when MstarMplanet). In addition, the lower luminosity of an M dwarf with respect to a star of earlier spectral type causes its habitable zone to be located very close to the host star, which makes detecting habitable planets around M dwarfs (at ~0.1 au) easier than detections around solar-like stars (at ~1 au).

From transit surveys with the NASA 0.95 m Kepler space observatory, very small planet candidates are found to be relatively more abundant than Jupiter-type candidates as the host stellar mass decreases (Howard et al. 2012; Dressing & Charbonneau 2013, 2015; Kopparapu 2013; Kopparapu et al. 2013). For early-M dwarf stars in the field, some radial-velocity studies have already been carried out (ESO CES, UVES and HARPS by Zechmeister et al. 2009, 2013; CRIRES by Bean et al. 2010; HARPS by Bonfils et al. 2013), but the much-sought value of η, that is, the relative abundance of Earth-type planets in the habitable zone, is as yet only poorly constrained from radial-velocity data (e.g., from Bonfils et al. 2013).

Highly stable, high-resolution spectrographs in the near-infrared currently under construction, such as SPIRou (Artigau et al. 2014), IRD (Kotani et al. 2014), HPF (Mahadevan et al. 2014), and CARMENES, are preferable over visible for targets with spectral types M4 V or later (see Table 1 in Crossfield 2014). This is because the spectral energy distribution of late-M dwarfs approximately peaks at 1.01.2 μm (Reiners et al. 2010), while HARPS and its copy in the northern hemisphere, HARPS-N, cover the wavelength interval from 0.38 to 0.69 μm. That faintness in the optical is quantitatively illustrated with the tabulated V magnitudes of the brightest M dwarfs in the northern hemisphere (HD 79210/GJ 338 A, HD 79211/GJ 338 B, and HD 95735/GJ 411) at 7.57.7 mag, far from the limit of the naked human eye.

The specific advantage of CARMENES is the wide wavelength coverage and high spectral resolution in both visible and near-infrared channels. Simultaneous observation from 0.5 to 1.7 μm is a powerful tool for distinguishing between genuine planet detections and false positives caused by stellar activity, which have plagued planet searches employing spectrographs with a smaller wavelength coverage, especially in the M-type spectral domain (Reiners et al. 2010; Barnes et al. 2011). A substantial amount of guaranteed time for the completion of the key project is also an asset.

A precise knowledge of the targets is critical to ensure that most of the CARMENES guaranteed time is spent on the most promising targets. This selection involves not only a comprehensive data compilation from the literature, but also summarises our observational effort to achieve new low- and high-resolution optical spectroscopy and high-resolution imaging. The present publication on low-resolution spectroscopy is the first paper of a series aimed at describing the selection and characterisation of the CARMENES sample. We have shown some preliminary results at conferences before that described the input catalogue description and selection (Caballero et al. 2013; Morales et al. 2013), low-resolution spectroscopy (Klutsch et al. 2012; Alonso-Floriano et al. 2013a), high-resolution spectroscopy (Alonso-Floriano et al. 2013b; Passegger et al. 2014), resolved multiplicity (Béjar et al. 2012; Cortés-Contreras et al. 2013), X-rays (Lalitha et al. 2012), exploitation of public databases (Montes et al. 2015), or synergies with Kepler K2 (Rodríguez-López et al. 2014). This first item of the CARMENES science-preparation series details the low-resolution optical spectroscopy of M dwarfs with the CAFOS spectrograph at the Zeiss 2.2 m Calar Alto telescope.

2. CARMENES sample

Table 1

Completeness and limiting J-band magnitudes per spectral type for the CARMENES input catalogue.

Table 2

Sources of the CARMENES input catalogue.

To prepare the CARMENES input catalogue with the best targets, we systematically collected all published M dwarfs in the literature that fulfilled two simple criteria:

  • They had to be observable from Calar Alto with target declinations δ> −23 deg (i.e., zenith distances <60 deg, air masses at culmination <2.0).

  • They were selected according to late spectral type and brightness. We only catalogued confirmed dwarf stars with an accurate spectral type determination from spectroscopic data (i.e., not from photometry) between M0.0 V and M9.5 V. Additionally, we only compiled the brightest stars of each spectral type. Our database contains virtually all known M dwarfs that are brighter than the completeness magnitudes shown in Table 1, and most of them brighter than the limiting magnitudes. No target fainter than J = 11.5 mag is in our catalogue.

We started to fill the CARMENES database with the M dwarfs from the Research Consortium on Nearby Stars2, which catalogues all known stars with measured astrometric parallaxes that place them within 10 pc (e.g., Henry et al. 1994; Kirkpatrick et al. 1995; Riedel et al. 2014; Winters et al. 2015). The RECONS stellar compilation was next completed with the Palomar/Michigan State University survey catalogue of nearby stars (PMSU – Reid et al. 1995, 2002; Hawley et al. 1996; Gizis et al. 2002). Afterwards, we gave special attention to the comprehensive proper-motion catalogues of Lépine et al. (2003, 2009, 2013) and Lépine & Gaidos (2011), and the “Meeting the Cool Neighbors” series of papers (Cruz & Reid 2002; Cruz et al. 2003, 2007; Reid et al. 2003, 2004, 2008). Table 2 provides the sources of our information on M dwarfs. Until we start our survey at the end of 2015, we will still include some new, particularly bright, late, single, M dwarfs3.

As of February 2015, our input catalogue, dubbed CARMENCITA (CARMENES Cool dwarf Information and daTa Archive), contains approximately 2200 M dwarfs. For each target star, we tabulate a number of parameters compiled from the literature or measured by us with new data: accurate astrometry and distance, spectral type, photometry in 20 bands from the ultraviolet to the mid-infrared, rotational, radial, and Galactocentric velocities, Hα emission, X-ray count rates and hardness ratios, close and wide multiplicity data, membership in open clusters and young moving groups, target in other radial-velocity surveys, and exoplanet candidacy (Caballero et al. 2013). The private on-line catalogue, including preparatory science observations (i.e., low- and high-resolution spectroscopy, high-resolution imaging), will become public as a CARMENES legacy.

Of the 2200 stars, we discard all spectroscopic binaries and multiples, and resolved systems with physical or visual companions at less than 5 arcsec to our targets. The size of the CARMENES optical fibres projected on the sky is 1.5 arcsec (Seifert et al. 2012; Quirrenbach et al. 2014), and consequently any companion at less than 5 arcsec may induce real or artificial radial-velocity variations that would contaminate our measurements (Guenther & Wuchterl 2003; Ehrenreich et al. 2010; Guenther & Tal-Or 2010). About 1900 single stars currently remain after discarding all multiple systems.

thumbnail Fig. 1

Six representative CAFOS spectra. From top to bottom, spectra of standard stars with spectral type 1.0 and 0.5 subtypes earlier than the target (black), standard star with the same spectral type as the target (cyan), the target star (red; in this case, J04393+335 – M4.0 V, Simbad name: V583 Aur B), and standard stars with spectral type 0.5 and 1.0 subtypes later than the target (black). We mark activity-, gravity-, and youth-sensitive lines and doublets at the top of the figure (Hγ, Hβ, He i D3, Na i D2 and D1, Hα, Li i, K i, and Na i, from left to right) and molecular absorption bands at the bottom. Note the three first lines of the Balmer series in emission in the spectrum of the target star.

Open with DEXTER

2.1. CAFOS sample

The aim of our low-resolution spectroscopic observations is twofold: (i) to increase the number of bright, late-M dwarfs in CARMENCITA and (ii) to ensure that the compiled spectral types used for the selection and pre-cleaning are correct. With this double objective in mind, we observed the following:

  • High proper-motion M-dwarf candidates from Lépine & Shara (2005) and Lépine & Gaidos (2011) with spectral types with large uncertainties or derived only from photometric colours. Spectral types from VJ colours are not suitable for our purposes (Alonso-Floriano et al. 2013a; Mundt et al. 2013; Lépine et al. 2013V is an average of photographic magnitudes BJ and RF from the Digital Sky Survey; cf. Lépine & Gaidos 2011). In collaboration with Sebastien Lépine, we observed and analysed an extension of the Lépine & Gaidos (2011) catalogue of high proper-motion candidates brighter than J = 10.5 mag. The spectra of stars brighter than J = 9.0 mag were published by Lépine et al. (2013), while most of the remaining fainter ones are published here.

  • M dwarf candidates in nearby young moving groups (e.g., Montes et al. 2001; Zuckermann & Song 2004; da Silva et al. 2009; Shkolnik et al. 2012; Gagné et al. 2014; Klutsch et al. 2014), in multiple systems containing FGK-type primaries that are subjects of metallicity studies (Gliese & Jahreiss 1991; Poveda et al. 1994; Gould & Chanamé 2004; Rojas-Ayala et al. 2012; Terrien et al. 2012; Mann et al. 2013; Montes et al. 2013), in fragile binary systems at the point of disruption by the Galactic gravitational field (Caballero 2012, and references therein), and resulting from new massive virtual-observatory searches (Jiménez-Esteban et al. 2012; Aberasturi et al. 2014). Such a broad diversity of sources allowed us to widen the investigated intervals of age, activity, multiplicity, metallicity, and dynamical evolution.

  • Known M dwarfs with well-determined spectral types from PMSU (see above) and Lépine et al. (2013). The comparison of these two samples with ours was a sanity check for determining the spectral types (see Sect. 4.1).

  • M dwarfs in our input catalogue with uncertain or probably incorrect spectral types based on apparent magnitudes, r′ − J colours, and heliocentric distances, including resolved physical binaries. See some examples in Cortés-Contreras et al. (2014).

  • Numerous standard stars. For an accurate determination of spectral type and class, we also included approximately 50 stars with well-determined spectral types from K3 to M8 for both dwarf (Johnson & Morgan 1953; Kirkpatrick et al. 1991; PMSU) and giant classes (e.g., Moore & Paddock 1950; Ridgway et al. 1980; Jacoby et al. 1984; García 1989; Keenan & McNeil 1989; Kirkpatrick et al. 1991; Sánchez-Blázquez et al. 2006; Jiménez-Esteban et al. 2012).

3. Observations and analysis

3.1. Low-resolution spectroscopic data

Observations were secured with the Calar Alto Focal reductor and Spectrograph (CAFOS) mounted on the Ritchey-Chrétien focus of the Zeiss 2.2 m Calar Alto telescope (Meisenheimer 1994). We obtained more than 900 spectra of 745 targets during 38 nights over four semesters from 2011 November to 2013 April. All observations were carried out in service mode with the G-100 grism, which resulted in a useful wavelength coverage of 42008300 Å at a resolution ℛ ~ 1500. Exposure times ranged from shorter than 1 s to 1 h. The longest exposures were split into up to four sub-exposures. On some occasions, another star fell in the slit aperture (usually the primary of a close multiple system containing our M-dwarf candidate main target). We also added the 13 red dwarfs and giants observed in 2011 March by Jiménez-Esteban et al. (2012), which made a total of 758 targets.

We reduced the spectra using typical tasks within the IRAF environment. The reduction included bias subtraction, flat fielding, removal of sky background, optimal aperture extraction, wavelength calibration (with Hg-Cd-Ar, He, and Rb lamps), and instrumental response correction. For the latter, we repeatedly observed the spectrophotometric standards G 191–B2B (DA0.8), HD 84937 (sdF5), Feige 34 (sdO), BD+25 3941 (B1.5 V), and BD+28 421 (sdO) at different air masses. In the end, we only used the spectra with the highest signal-to-noise ratio of the hot subdwarf Feige 34, which gave the best-behaved instrumental response correction. We extracted all traces in the spectra, including those of other stars in the slit aperture. We did not remove telluric absorption lines from the spectra that were due to the variable meteorological conditions during two years of observation (see Sect. 3.2.2). All our spectra had a signal-to-noise ratio higher than 50 near the Hαλ6562.8 Å line, which together with the wide wavelength coverage allowed us to make a comprehensive analysis and to measure numerous spectral indices and activity indicators.

We list in Table A.1 the 753 observed K and M dwarf and giant candidates according to identification number, our CARMENCITA identifier, discovery name, Gliese or Gliese & Jahreiß number, J2000.0 coordinates and J-band magnitude from the Two-Micron All-Sky Survey (Skrutskie et al. 2006), observation date, and exposure time. The five stars not tabulated are the spectrophotometric standards.

Table 3

Standard and prototype stars.

In Table A.1, our CARMENCITA identifier follows the nomenclature format “Karmn JHHMMm±DDd(X)”, where “Karmn” is the acronym, “m” and “d” in the sequence are the truncated decimal parts of a minute or degree of the corresponding equatorial coordinates for the standard equinox of J2000.0 (IRAS style for right ascension, PKS quasar style for declination), and X is an optional letter (N, S, E, W) to distinguish between physical or visual pairs with the same HHMMm±DDd sequence within CARMENCITA. We use the discovery name for every target, except for M dwarfs with variable names (e.g., EZ PSc, GX And, V428 And) or those that are physical companions to bright stars (e.g., BD–00 109 B, η Cas B, HD 6440 B). We associate for the first time many X-ray events with active M dwarfs (e.g., Lalitha et al. 2012; Montes et al. 2015). In these cases, we use the precovery4Einstein 2E or ROSAT RX/1RXS event identifications instead of the names given by the proper-motion survey that recovered the stars.

Additionally, we always indicate whether the star is a known close binary or triple unresolved in our spectroscopic data (with “AB”, “BC”, “ABC”). There are 75 such close multiple systems in our sample (70 double and 5 triple), of which the widest unresolved pair is J09045+164 AB (BD+16 1895), with ρ ≈ 4.1 arcsec. A comprehensive study on the multiplicity of M dwarfs in CARMENCITA, including spectral-type estimate of companions from magnitude differences in high-resolution imaging data, will appear in a forthcoming paper of this series (preliminary data were presented in Cortés-Contreras et al. 2015).

3.2. Spectral typing

We followed two widely used spectral classification schemes that require an accurate, wide grid of reference stars. The first strategy relies on least-squares minimisation and best-fitting to spectra of the standard stars, while the second scheme uses spectral indices that quantify the strength of the main spectral features in M dwarfs, notably molecular absorption bands (again, preliminary data were presented in Klutsch et al. 2012 – least-squares minimisation – and Alonso-Floriano et al. 2013a – spectral indices).

Before applying the two spectral-typing strategies, we normalised our spectra by dividing by the observed flux at 7400 Å. We also corrected for spatial distortions at the reddest wavelengths (with R ~ 1500, there is no need for a stellar radial-velocity correction). For that, we shifted our spectra until the K iλλ7664.9, 7699.0 Å doublet was placed at the laboratory wavelengths. This correction, often of 12 Å, was critical for the definition of spectral indices, some of which are very narrow.

thumbnail Fig. 2

CAFOS spectra of our prototype stars. From top to bottom, K3 V, K5 V, K7 V, M0.0–7.0 V in steps of 0.5 subtypes, and M8.0 V.

Open with DEXTER

3.2.1. Spectral standard and prototype stars

We list the used standard stars in Table 3 (see also Sect. 3.2.4), most of which were taken from Kirkpatrick et al. (1991) and PMSU (Table 2). Our intention was to provide one prototype star and up to four reference stars per half subtype, but this was not always possible, especially at the latest spectral types. The prototype stars, shown in Fig. 2, are the brightest, least active reference stars that have spectra with the highest signal-to-noise ratio, and that do not deviate significantly from the general trend during fitting. In Table 3, the first star of each subtype is the prototype for that subtype. In the case of standard stars with different reported spectral types in the bibliography (with maximum differences of 0.5 subtypes), we chose the value that gave us less scatter in our fits.

We also used three K dwarfs from Kirkpatrick et al. (1991), not listed in Table 3, to extend our grid of standard stars towards warmer effective temperatures. The three K-dwarf standard stars are HD 50281 (K3 V), 61 Cyg A (K5 V), and η Cas B (K7 V). The standard star LP 609–071 AB is a close binary composed of an M2.5 V star and a faint companion separated by ρ = 0.18 ± 0.02 arcsec (Delfosse et al. 1999). With ΔK = 0.95 ± 0.05 mag, the faint companion flux barely affects the primary spectrum in the optical.

Table 4

Spectral indices used in this work.

thumbnail Fig. 3

Spectral typing of J04393+335 with best-match and methods. Top panel: best-match. CAFOS normalised spectra of the target star and of the standard star that fits best (top) and the difference (bottom). Bottom panel: . Values of χ2 as a function of the spectral type (open red circles). The vertical dashed line marks the spectral type at the lowest χ2 value. Note the logarithmic scale in the Y axis.

Open with DEXTER

3.2.2. Best-match and methods

Before any least-squares minimisation, we discarded five narrow (20–30 Å) wavelength ranges influenced by activity indicators (Hα, Hβ, Hγ, and the Na i doublet – Fraunhofer C, F, and G lines) and by strong telluric lines (O2 band around 7594 Å – Fraunhofer A line). Next, in the full remaining spectral range, we compared the normalised spectrum of every target star with those of all our standard stars in Table 3 and computed a χ2 value for each fit. In the best-match method, we assigned the spectral type of the standard star that best fitted our target spectrum (i.e., with the lowest χ2 value); in the method, we assigned the spectral type that corresponded to the minimum of the curve resulting from the (sixth-order) polynomial fit of all the χ2-spectral type pairs. As expected, the best-match and methods give the same result in most cases. In the representative case shown in Figs. 1 and 3, the target dwarf has an M4.0 V spectral type using both the best match and methods.

3.2.3. Spectral indices

The spectral indices methodology for spectral typing is based on computing flux ratios at certain wavelength intervals in low-resolution spectra (e.g., Kirkpatrick et al. 1991; Reid et al. 1994; Martín et al. 1996, 1999). In the present analysis, we compiled 31 spectral indices defined in the literature to determine spectral types of late-K dwarfs and M dwarfs that occur in the useable wavelength interval of our CAFOS spectra. In general, a spectral index Ii is defined by the ratio of numerator and denominator fluxes (i.e., Ii = Fi,num/Fi,den). Table 4 lists the 31 wavelength intervals of fluxes in the numerator and denominator, Δλnum and Δλden, and corresponding reference for each index. Some flux wavelength intervals are the linear combination of two subintervals (CaH 1, VO-a, VO-b) or, in the case of the VO index, a nonlinear combination. Additionally, there are wavelength intervals of fluxes in the numerator that are either redder and bluer than the one in the denominator, which translates into different slopes in the index-spectral type relations. Of the 31 tabulated indices, nine are related to TiO features, seven to VO, six to CaH, three to the “pseudo-continuum” (i.e., relative absence of features), and the rest to H and neutral metallic lines (Ti, Na).

thumbnail Fig. 4

Nine representative spectral indices as a function of spectral type. Filled (red) circles: standard stars. Open (blue) circles: remaining target stars. Dashed (red) line: fit to straight line, parabola, or cubic polynomial, drawn only in the range of application of the fit. Top left and middle: TiO 2 and TiO 5 (Reid et al. 1995), with negative slopes and useable up to M7 V; top right: PC1 (Martín et al. 1996), a monotonous spectral indicator from K5 V to M8 V; centre left and middle: VO-7912 (Martín et al. 1999) and Color-M (Lépine et al. 2003), two indices very similar to the PC1. Note the logarithmic scale in Color-M; centre right: ratio B (Kirkpatrick et al. 1991), sensitive to several stellar parameters and, thus, with a large scatter in the spectral type relation; bottom left: CaH 3 (Reid et al. 1995), with a slightly larger scatter than pseudo-continuum or titanium oxide indices, due to metallicity; bottom middle: VO (Kirkpatrick et al. 1995), useable only for determining spectral types later than M4 V; bottom right: I2 (Martín & Kun 1996), with a linear range of variation from mid-K to mid-M and a sudden increase (or high dispersion) at late-M.

Open with DEXTER

For every star observed with CAFOS, we computed the stellar numerator and denominator fluxes using an automatic trapezoidal integration procedure. When all indices were available, we plotted all spectral index vs. spectral type diagrams for the standard stars listed in Table 3 and fitted low-order polynomials to the data points. Although some spectral indices allowed linear (e.g., I2, Ratio B) or parabolic fits, most of the fits were to cubic polynomials of the form SpT(i) = a + bi + ci2 + di3, where i was the index. In all cases, we checked our diagrams and fits with those in the original papers and found no significant differences (of less than 0.5 subtypes). We also took special care in defining the range of application of our fits in spectral type. The different shapes of fitting curves, ranges of application, and internal dispersion of the data points are illustrated in Fig. 4.

Some indices are sensitive not only to spectral type (i.e., effective temperature), but also to surface gravity (e.g., I2, Ratio A, CaH Narr, Ratio C, Na-8190 – Sect. 4.2), metallicity (e.g., CaOH, CaH 1, CaH 2, CaH 3 – Sect. 4.3), or activity (Hα – Sect. 4.4). We identified the spectral indices with the widest range of application and least scatter. Table 5 lists the coefficients of the cubic polynomial fits of the five spectral indices that we eventually chose for spectral typing (note the logarithmic scale of the Color-M index). The spectral index vs. spectral type diagrams of VO-7912 and Color-M are very similar to those of PC, TiO 2, and TiO 5, shown in Fig. 4. All of them are valid from K7 V to M7 V (to M8 V in the case of PC1 and Color-M), while the dispersion of the fits is of about 0.5 subtypes. The TiO 5 has been a widely used index for spectral typing (Reid et al. 1995; Gizis 1997; Seeliger et al. 2011; Lépine et al. 2013) but, to our knowledge, we propose here for the first time to use it with a nonlinear fit.

In Table A.2, we list the values of the five spectral-typing indices of all CAFOS stars together with the CaH 2 and CaH 3 indices that are used to compute the ζ metallicity index (Sect. 4.3) and the pseudo-equivalent width of the Hα line (Sect. 4.4).

Table 5

Coefficients and standard deviation (in spectral subtypes) of the cubic polynomial fits of the five spectral-typing indices.

3.2.4. Adopted spectral types

After applying the best-match and methods and using the five spectral-index-type relations in Table 5, we obtained seven complementary spectral-type determinations for each star. In Table A.3, we assigned a value between 0.0 and 8.0 in steps of 0.5 to each M spectral subtype for all stars of dwarf luminosity class (there are no stars later than M8.0 V in our CAFOS sample). In addition, we used the values –2.0 and –1.0 for referring to K5 V and K7 V spectral types (there are no K6, 8, 9 spectral types in the standard K-dwarf classification – Johnson & Morgan 1953; Keenan & McNeil 1989). In some cases, we were able to identify dwarfs earlier than K5 V with the best-match and methods. For giant stars, we only provide a visual estimation (K III, M III) based on the spectral types of well-known giant standard stars observed with CAFOS.

For all late-K and M dwarfs, we calculated one single spectral type per star based on the information provided by the seven individually determined spectral indices. We used a median of the seven values for cases where they were identical within 0.5 subtypes. To avoid any bias, we carefully checked the original spectra if the spectral types from the best-match and methods and from the spectral indices were different by 0.5 subtypes, or if any spectral type deviated by 1.0 subtypes or more (which occurred very rarely). In these cases, we adopted the spectral type of the closest (visually and in χ2) standard star. The uncertainty of the adopted spectral types is 0.5 subtype, except for some odd spectra indicated with a colon (probably young dwarfs of low gravity or subdwarfs of low metallicity; see below).

4. Results and discussion

4.1. Spectral types

thumbnail Fig. 5

Spectral type comparison between our results and those from PMSU (top panel), Lépine et al. (2013; middle panel), and Lépine & Gaidos (2011; from VJ photometry, bottom panel). The larger a circle, the greater the number of stars on a data point. Dotted lines indicate the one-to-one relationship.

Open with DEXTER

Among the 753 investigated stars with adopted spectral types from CAFOS data, there were 23 late-K dwarfs at the K/M boundary (K7 V), 21 early- and intermediate-K dwarfs (K05 V), 22 M-type giants (M III), three K-type giants (K III), and one star without class determination (i.e., J04313+241 AB). This left 683 M-type dwarfs (and subdwarfs) in our CAFOS sample.

As shown in Table A.3, we searched for previous spectral type determinations in the literature for the 753 investigated stars (small “m” and “k” denote spectral types estimated from photometry). Taking previous determinations and estimations into account, we derived spectral types from spectra for the first time for 305 stars, and revised typing for most of the remaining 448 stars.

The agreement in spectral typing with previous large spectroscopic surveys of M dwarfs is shown in Fig. 5. The standard deviations of the differences between the spectral types derived by us and by PMSU (with a narrower wavelength interval; 100 stars in common) and by Lépine et al. (2013; 95 stars in common) were 0.55 and 0.38 subtypes, respectively, which are of the order of our internal uncertainty (0.5 subtypes). The standard deviation of the differences between our spectral types and those estimated from photometry by Lépine & Gaidos (2011; 576 stars in common) is larger, of up to 1.32 subtypes. The bias towards later spectral types in Lépine & Gaidos (2011) and the scatter of the spectral type differences is obvious from the bottom panel in Fig. 5. In particular, we measured maximum differences of up to 7 subtypes, by which some late-M dwarf candidates become actual K dwarfs (probably due to the use by Lépine & Gaidos of BJ and RF from photographic plates for the spectral type estimation; see references in Sect. 2). However, over 93% of the compared stars have disagreements lower than or equal to 2 subtypes. We emphasize that our CARMENCITA data base is very homogeneous because more than 95% of the spectral type determinations come from either PMSU, Lépine et al. (2013), or our CAFOS data, which are consistent with each other, as shown above.

Of the 683 CAFOS M-type dwarfs (and subdwarfs), 414 and 106 M dwarfs satisfy our criteria in Table 1 of restrictive J-band spectral type limiting and completeness, respectively. In total, 261 dwarfs have spectral type M4.0 V or later. The brightest, latest of them are being followed-up with high-resolution spectrographs and imagers and with data from the bibliography to identify the most suitable targets for CARMENES (no physical or visual companions at less than 5 arcsec, low vsini; see forthcoming papers of this series). Furthermore, there are 61 relatively bright (J< 10.9 mag) CAFOS stars with spectral types between M5.0 V and M8.0 V that are also suitable targets for any other near-infrared radial-velocity monitoring programmes with the instruments mentioned above (i.e., HPF, SPIRou, IRD).

4.2. Gravity

Table 6

Giant stars observed with CAFOS.

Table 6 lists the 25 giants observed with CAFOS. Of these, 17 stars have previously been tabulated as M giant standard stars (e.g., Keenan & McNeil 1989; García 1989; Kirkpatrick et al. 1991; Sánchez-Blázquez et al. 2006). They are bright (J 5.0 mag; down to –1.0 mag in the case of Mirach, β And) and show the low-gravity spectral features typically found in M giants: faint alkali lines (K i λλ7665, 7699 Å and Na iλλ8183, 8195 Å), a tooth-shaped feature produced by MgH/TiO blend near 4770 Å, and a decrease of CaH in the A-band at 69086946 Å with the increase of luminosity (Kirkpatrick et al. 1991; Martín et al. 1999; Riddick et al. 2007; Gray & Corbally 2009). Two other stars, V428 And and HD 236547, are well-known K giant standard stars (Jacoby et al. 1984; García 1989; Kirkpatrick et al. 1991).

Of the other six giant stars in Table 6, three have J-band magnitudes of 6.16.2 mag and were identified by Jiménez-Esteban et al. (2012) as some of the reddest Tycho-2 stars with proper motions μ> 50 mas a-1, namely Ruber 7, TYC 6238–480–1, and Ruber 8 (which seems to be also one of the brightest metal-poor M giants ever identified). The remaining three giant stars, with faint J-band magnitudes between 8.2 and 10.0 mag, are listed below.

  • J01012+571 (1RXS J010112.8+570839). It is a previously unknown distant M giant close to the Galactic plane (b = −5.7 deg). It was serendipitously identified in an unpublished photometric survey by one of us (J.A.C.), and was observed with CAFOS because of its very red optical and near-infrared colours and possible association with an X-ray event catalogued by ROSAT at a separation of only 6.4 arcsec.

  • J03319+492 (TYC 3320–337–1). From photographic magnitudes, Pickles & Depagne (2010) and Lépine & Gaidos (2011) classified it as an M1.9 and M3 dwarf, respectively. However, it appears to be an early-K giant with a significant proper motion of 56 mas a-1. It is not possible to separate it from the main sequence in a reduced proper-motion diagram.

  • J23266+453 (2MASS J23263798+4521054). Our intention was to observe BD+44 4419 B (G 216–43), an M4.5 dwarf of roughly the same V-band magnitude (10.3 vs. 10.9 mag). Unfortunately, we incorrectly observed instead a background giant at a separation of about 20 arcsec.

In a Ratio C vs. PC1 index-index diagram as the one shown in Fig. 6, where Ratio C is highly sensitive to gravity and PC1 is an effective temperature proxy (PC1 was indeed one of the five indices used for deriving spectral types), all giants are below the dashed line at Ratio C = 1.07. There is only one star not classified as a giant that lies below that empirical boundary. It is J04313+241 (V927 Tau AB), a T Tauri star for which we did not provide a luminosity class in Sect. 4.1. We discuss this in detail in Sect. 4.4. Ratio C, which contains the sodium doublet at 8193, 8195 Å, can also be used as a youth indicator (e.g., Schlieder et al. 2012b).

thumbnail Fig. 6

Ratio C vs. PC1 index-index diagram. The different symbols represent field dwarfs (small dots, blue), giants (open squares, dark blue), Taurus stars (filled stars, red), and other young stars (open stars, red). All giants are below the dashed line at Ratio C = 1.07. The dashed line is the empirical border of the giant star region.

Open with DEXTER

4.3. Metallicity

In F-, G-, and K-type stars whose photospheric continua are well-defined in high-resolution spectra, stellar metallicity is computed through spectral synthesis (McWilliam 1990; Valenti & Piskunov 1996; González 1997; González Hernández et al. 2004; Valenti & Fischer 2005; Recio-Blanco et al. 2006) or measuring equivalent widths, especially of iron lines (Sousa et al. 2008, 2011; Magrini et al. 2010; Adibekyan et al. 2012; Tabernero et al. 2012; Bensby et al. 2014). However, it is not possible to measure a photospheric continuum in M-type stars and, thus, their metallicity is studied through other techniques. Since the first determinations from broad-band photometry by Stauffer & Hartmann (1986), there have been three main observational techniques employed to determine metallicity in M dwarfs:

  • Photometry calibrated with M dwarfs in physical double and multiple systems with warmer companions, typically F, G, K dwarfs, of known metallicity (Bonfils et al. 2005; Casagrande et al. 2008; Schlaufman & Laughlin 2010; Neves et al. 2012).

  • Low-resolution spectroscopy, also calibrated with M dwarfs with earlier primaries, in the optical (Dhital et al. 2012), in the near infrared (Rojas-Ayala et al. 2010, 2012; Terrien et al. 2012; Mann et al. 2014; Newton et al. 2014), or in both wavelength ranges (Mann et al. 2013, 2015).

  • High-resolution spectroscopy in the optical (from spectral synthesis: Woolf & Wallerstein 2005; from spectral indices: Woolf & Wallerstein 2006; and Bean et al. 2006; from the measurement of pseudo-equivalent widths: Neves et al. 2013, 2014), in the near-infrared (Shulyak et al. 2011, in the Y band; Önehag et al. 2012, in the J band; Tsuji & Nakajima 2014, in the K band), or in the optical and near-infrared simultaneously (Gaidos & Mann 2014). The novel mid-resolution spectroscopy study in the optical aided with spectral synthesis by Zboril & Byrne (1998) also belongs in this item.

For the 753 CAFOS stars, we computed the ζTIO/CaH metallicity parameter (denoted ζ for short) described by Lépine et al. (2007): (1)where [TiO 5] Z = 0.571−1.697CaH + 1.841CaH2−0.454CaH3 is a third-order fit of CaH = CaH 2 + CaH 3 for our standard stars, and TiO 5, CaH 2, and CaH 3 are the spectral indices of Reid et al. (1995) (see also Gizis & Reid 1997; and Lépine et al. 2003). The ζ index is correlated with metallicity in metal-poor M subdwarfs (Woolf et al. 2009) and metal-rich dwarfs (Lépine et al. 2007; Mann et al. 2014). For completeness, we also tabulate the ζ index for our 25 giants, but they are not useful for a comparison. We made the same assumption of standard stars having solar metallicity (ζ 1) as in Lépine et al. (2007), which was later justified by the small dispersion of the data points.

thumbnail Fig. 7

CaH 2 + CaH 3 vs. TIO 5 index-index diagram of our CAFOS stars, after discarding giants. The spectral types at the top are indicative, but follow the TiO 5 fit given in Table 5. The solid and dash-dotted lines are iso-metallicity curves of the ζ index.

Open with DEXTER

We looked for M dwarfs (and subdwarfs) in our sample with abnormal metallicity, which could be spotted in a CaH 2 + CaH 3 vs. TIO 5 index-index diagram as in Fig. 7. Lépine et al. (2007) defined the classes subdwarf (sd), 0.5 <ζ< 0.825, and extreme subdwarf (esd), ζ< 0.5. All our non-giant stars except two have ζ values greater than 0.825, which is the empirical boundary between dwarfs and subdwarfs. The spectra of the two exceptions show shallower molecular bands and lines than M dwarfs of the same spectral type.

One of our two subdwarf candidates is J19346+045 (sdM1:, ζ = 0.775 – HD 184489). Some authors have reported features of low metallicity (e.g., Maldonado et al. 2010), but none had classified it as a subdwarf (but see Sandage & Kowal 1986). Its low effective temperature has prevented spectral synthesis analyses on high-resolution spectra.

The other new subdwarf candidate is J16354–039 (sdM0:, ζ = 0.664 – HD 149414 B, BD–03 3968B). Giclas et al. (1959) discovered it and associated it with the G5 Ve single-line spectroscopic binary HD 149414 Aa,Ab. Afterwards, its membership in the very wide system has been investigated by Poveda et al. (1994), Tokovinin (2008), and Dhital et al. (2010), for example, and confirmed and quantified by Caballero (2009). The projected physical separation between Aa,Ab and B amounts to 53 000 au (about a quarter of a parsec). Remarkably, the primary is a halo binary of low metallicity ([Fe/H] ~ –1.4 – Strom & Strom 1967; Sandage 1969; Cayrel de Strobel et al. 1997; Holmberg et al. 2009). This explains the low ζ metallicity index of J16354–039 for its spectral type and the wide separation of the system (due to gravitational disruption by the Galactic gravitational potential or to common origin and ejection from the same cluster; cf. Caballero 2009, and references therein).

In addition, J12025+084 (M1.5 V; ζ = 0.898 – LHS 320) was classified by Gizis (1997) as an sdM2.0 star and was investigated extensively afterwards with high-resolution imagers (Gizis & Reid 2000; Riaz et al. 2008; Jao et al 2009; Lodieu et al. 2009). However, we failed to detect any subdwarf signpost in our high signal-to-noise spectrum, which is partly consistent with the metallicity [ Fe/H ] = −0.6 ± 0.3 measured by Rajpurohit et al. (2014).

No CAFOS star showed a very high metallicity index greater than ζ = 1.5. In spite of the dispersion of the ζ index around unity, we considered that all our 726 dwarfs (753 stars in total minus the 25 giants and the two subdwarfs) approximately have solar metallicity ([ Fe/H ] ≈ 0.0). This assumption is relevant for instance to derive the mass from absolute magnitudes, the spectral types, and theoretical models that need metallicity as an input.

4.4. Activity

Chromospheric activity is one of the main relevant parameters for exoplanet detection around M dwarfs. The heterogeneities on the stellar surface of the almost-fully convective, rotating, M dwarfs, such as dark spots, may induce spurious radial-velocity variations at visible wavelengths (Bonfils et al. 2007; Reiners et al. 2010; Barnes et al. 2011; Andersen & Korhonen 2015; Robertson et al. 2015). Near-infrared observations are expected to improve the precision of radial-velocity measurements with respect to the visible for stars cooler than M3, and CARMENES will cover the wavelength range from 0.55 to 1.70 μm. In spite of this, we plan to identify the least active stars for our exoplanet search. Moreover, several authors have identified significant differences between colours and spectral indices of active and inactive stars of similar properties that may affect the spectral typing of M dwarfs (Stauffer & Hartmann 1986; Hawley et al. 1996; Bochanski et al. 2007; Morales et al. 2008).

4.4.1. Hα emission

thumbnail Fig. 8

Pseudo-equivalent width of Hα line vs. spectral type diagram. Measurement errors are 0.5 subtypes for the spectral type and are of the order of 10% with a minimum of 0.1 Å for pEW(Hα) (but scatter due to variability is probably larger than 10%). Dotted, dash-dotted, and dashed lines indicate the boundaries between chromospheric and accretion emission for different authors. Giants, plotted with open squares, have filled Hα lines or in absorption. None of the young stars, plotted with open and filled stars, show accretion emission. The four stars at the accretion boundary discussed in the text are encircled.

Open with DEXTER

The Hα index (Reid et al. 1995) was one of the 31 indices measured on our CAFOS spectra (Table 4). For a better reliability on the activity determination and comparison with other works, we also measured the pseudo-equivalent width of the Hα line, pEW(Hα), of all the CAFOS stars (Table A.2). We here used pEW(Hα) as the proxy for activity (we used the pseudo-equivalent width of the line measured with respect to a local pseudo-continuum instead of the equivalent width because in M – and L, T, and Y – dwarfs the spectral continuum is not observable – e.g. Tsuji & Nakajima 2014).

We plot in Fig. 8 the pEW(Hα) vs. spectral type diagram for the whole sample. M dwarfs with late spectral types tend to show the Hα line in (strong) emission more often than earlier stars (see e.g. Hawley et al. 1996; West et al. 2004; or Reiners et al. 2013). There is a significant number of M4.0 V stars and later, however, that show very low Hα emission below 5 Å (in absolute values).

A few stars stand out in the pEW(Hα) vs. spectral type diagram in Fig. 8. Four of them lie at the boundary between chromospheric and accretion emission, as defined by Barrado y Navascués & Martín (2003), White & Basri (2003), and Fang (2010). Their high activity led us to investigate them in detail.

  • J04290+186 (M2.5 V, pEW(Hα) = –11.9 Å, V1103 Tau). It is a member of the 600 Ma-old Hyades cluster (Johnson et al. 1962; Griffin et al. 1988; Stauffer et al. 1991; Reid 1992).

  • J04544+650 (M4.0 V, pEW(Hα) = –13.9 Å, 1RXS J045430.9+650451). It is an anonymous Tycho-2 star (TYC 4087–1172–1; Lépine & Gaidos 2011) that we cross-matched with an aperiodic, variable, X-ray source identified by Fuhrmeister & Schmitt (2003). This X-ray variability and the presence of He i λ5875.6 Å in emission indicates that J04544+650 was flaring during our observations.

  • J01567+305 (M4.5 V, pEW(Hα) = –16.0 ± 0.4 Å, NLTT 6496, Koenigstuhl 4 A). It forms a loosely bound common-proper-motion pair together with the M6.5: V star NLTT 6491 (Koenigstuhl 4 B), and is associated with an X-ray source (Caballero 2012). Interestingly, Aberasturi et al. (2014) collected low-resolution spectroscopy for J01567+305 just two months earlier, for which they determined a spectral type identical to ours within the uncertainties, but measured pEW(Hα) = –9.3 ± 0.3 Å, which is significantly lower than our measurement. Our CAFOS spectrum also shows He i in emission, so the mid-M dwarf likely underwent a flare during our observations.

  • J07523+162 (M6.0 V, pEW(Hα) = 25.4 Å, LP 423−031). It has also been classified as a single M7 Ve star from optical spectra (Cruz et al. 2003; Reid et al. 2003; Gatewood & Coban 2009; Reiners & Basri 2009), but as an M6 V with surface gravity consistent with normal field dwarfs from near-infrared spectra (Allers & Liu 2013). From high-resolution spectroscopy (pEW(Hα) = –22.3 Å) and ROSAT X-ray count rates, Shkolnik et al. (2009) assigned J07523+162 an age of about 100 Ma, younger than the Pleiades. However, Reiners & Basri (2010) observed flaring activity in a J07523+162 spectrum (pEW(Hα) = 44.4 Å) and Gagné et al. (2014) and Klutsch et al. (2014) were not able to determine membership in any known stellar kinematic group.

There is an additional fifth active dwarf that stands out among the remaining stars in Fig. 8. It is J03332+462 (M0.0 V, pEW(Hα) = –3.4 Å, V577 Per B), a confirmed member of the ~70 Ma-old AB Doradus moving group (Zuckerman et al. 2004; da Silva et al. 2009; Schlieder et al. 2012a). Its relatively bright primary at about 9 arcsec is a young K2 V star with strong ultraviolet and X-ray emission and lithium in absorption (Pounds et al. 1993; Jeffries 1995; Montes et al. 2001; Zuckerman & Song 2004; Xing & Xing 2012).

4.4.2. Young (and very young) stars

Table 7

Reported young stars in our sample.

The identification of one open cluster member, one moving group member, and one purported young star in the field among five M dwarfs led us to examine the bibliography for other young star candidates in our CAFOS sample. The result of this bibliographic search is summarised in Table 7. In total, 49 spectroscopically investigated stars in this work have been reported to belong to the Taurus-Auriga star-forming region (~110 Ma, three stars), β Pictoris moving group (~1222 Ma, five stars), Carina or Columba associations (~1550 Ma, two stars), Argus association (~40 Ma, one star), AB Doradus moving group (~70120 Ma, five stars), Pleiades cluster (~120 Ma, one star with a relatively early K5 V spectral type), IC 2391 supercluster (~100200 Ma, one star), Hercules-Lyra moving group (~200300 Ma, four stars note the question marks), Castor moving group (~200300 Ma, six stars), Ursa Major moving group (~300500 Ma, one star), and Hyades cluster and supercluster (~600 Ma, 14 and 2 stars, respectively), and four to the young (τ ≲ 600 Ma) field star population in the solar neighbourhood. See Zuckerman & Song (2004) and Torres et al. (2008) for reviews on young moving groups.

The actual existence of some of the entities above (e.g., Hercules-Lyra and Castor moving groups, and IC 2391 and Hyades superclusters) is questioned by several authors. Three of the five stars with the lowest Hα emission for their spectral type belong to the hypothetical Castor moving group (Barrado y Navascués 1998; Montes et al. 2001; Ribas 2003; Caballero 2010; Mamajek 2013; Zuckerman et al. 2013), and the other two to the Hyades (super-) cluster (van Altena 1966; Hanson 1975; Legget & Hawkins 1988; Hawley et al. 1996; Stauffer et al. 1997; Montes et al. 2001; Klutsch et al. 2014). However, the extreme youth of some targets is confirmed by detection of lithium in absorption, X-ray in emission, and common proper-motion to bona fide primaries in nearby young moving groups.

Eleven of the 16 Hyads are known to be binaries. The relatively large number and (apparent) high binary frequency is a natural consequence of the Malmquist bias, which leads to the preferential detection of intrinsically bright objects. Equal-mass binaries are brighter than single stars of the same spectral type (by up to 0.75 mag) and, thus, the frequency of binarity in our magnitude-limited sample is higher than in a bias-free, volume-limited sample. While most of our targets lie at 2030 pc (Cortés-Contreras et al. 2015), the overbrightness of binary Hyads makes them to look as if they were located roughly at 30 pc instead at the nominal distance of the Hyades at about 46 pc. We suggest to investigate the actual multiplicity status of the five remaining single M dwarfs with a mid-resolution spectroscopic monitoring.

At d ~ 140 pc, the three Taurus stars in Table 7 are not in the solar neighbourhood. Since they are still on the Hayashi track of contraction, their radii are larger than those of dwarfs of the same effective temperature. As a result, they are also much more luminous, which explains why we were able to observe them even though they are located an order of magnitude farther away than the rest of our dwarf targets. As expected from their extreme youth, the three T Tauri stars have Hα emissions in the highest quartile (pEW(Hα)s between –9 and –11 Å, and spectral types between M4.0 and M4.5) and have been investigated spectroscopically earlier (Herbig & Rao 1972; Mathieu 1994; Wichmann et al. 1996; Kenyon et al. 1998; Scelsi et al. 2008; Sestito et al. 2008). The three of them displayed not only Hα in emission, but also Hβ and Hγ (we used one of them, J04393+335, in Fig. 3 to illustrate best the discarded wavelength ranges that are contaminated by activity in Sect. 3.2.2).

thumbnail Fig. 9

Spectral-typing indices TiO 2, TiO 5 (top), PC1, VO-7912 (middle), Color-M, and Ratio B (bottom) as a function of Hα pseudo-equivalent width for spectral types M3.0 V, M3.5 V, M4.0 V, and M4.5 V, from top to bottom. For clarity, the indices are offset in steps of 0.1 in the vertical axis. Solid lines are linear fits of the indices as a function of pEWH(α). Dashed lines indicate the mean and ± 1σ index at quiescence (pEWH(α) > –1 Å).

Open with DEXTER

A large radius also translates into low gravity. Indeed, the brightest of the trio of T Tauri stars, J04313+241 AB (V927 Tau AB, J = 9.73 mag) was the only non-giant target with spectral index Ratio C < 1.07 (Fig. 6) and the only one to which we did not assign a luminosity class in Table A.3. Its optical spectrum is intermediate between those of giants and dwarfs of the same spectral type (M4.0:). Something similar is true for the other two T Tauri stars, which also have very low Ratio C indices for their spectral type (but all giant stars in our sample display Hα in absorption). Although T Tauri stars are not natural targets for radial-velocity searches of low-mass exoplanets and none of the trio satisfies our criteria to be considered in the CARMENES sample (Table 1), a monitoring of bright, young, M dwarfs could shed light on the process of exoplanet formation (e.g., Crockett et al. 2012). Furthermore, young, nearby, very late stars are also ideal targets for direct-imaging surveys for Jupiter-like planetary companions at wide separations (Masciadri et al. 2005; Daemgen et al. 2007; Chauvin et al. 2010; Biller et al. 2013; Delorme et al. 2013, and references therein). Some of these targets are J13143+733 AB (NLTT 33370, M6.0 V in AB Doradus) and J09328+269 (DX Leo B, M5.5 V in Hercules-Lyra).

4.4.3. Effect of activity on spectral typing

As pointed out above, chromospheric activity could affect spectral typing. To study the validity of our results, we plot in Fig. 9 four of the five indices that we used for spectral typing as a function of pEW(Hα). We investigated the four spectral type intervals with the largest number of stars (in parenthesis): M3.0 V (72), M3.5 V (134), M4.0 V (113), and M4.5 V (84). Grouping by spectral type minimised the natural variation of the spectral index with effective temperature. The effect of activity on the TiO 5 and PC1 indices is not significant. However, strong activity in the largest quartile of pEW(Hα) has an appreciable effect on the indices VO-7912 and Color-M, but they fortuitously compensate each other because of the opposite slopes in their index vs. pEW(Hα) relations. The effect of activity on the TiO 2 index is more appreciable in the top left panel of Fig. 9, which agrees with the results shown by Hawley et al. (1996), who found that active M dwarfs tended to have lower values of TiO 2 (more absorption) than inactive dwarfs with the same TiO 5 indices (spectral types). However, this level of activity translates into a variation in spectral type of less than 0.8 subtypes for the most active stars, after using the coefficients in Table 5. In the end, the counter-weighting combination of five spectral indices and, especially, the use of the χ2 and best-match methods guarantee that our adopted spectral types are free from the effect of chromospheric activity (for the investigated interval of pEW(Hα)).

Our results on quantifying the variation of some spectral types as a function of activity are seemingly in contrast to some previous works, such as Morales et al. (2008). However, a direct comparison should be avoided because they grouped the stars by absolute magnitude, for which a determination of the distance is needed. A specific work on activity in M dwarfs will be another item of this series of papers on the science preparation of the CARMENES sample. It will be supported on one hand by new measurements of the emission of Hα, Hβ, and the Ca ii H & K doublet and near-infrared triplet from high-resolution spectra and on the other hand by a comprehensive parallax distance compilation and accurate spectro-photometric distance determination.

5. Summary

CARMENES, the new spectrograph at the 3.5 m Calar Alto telescope, will spectroscopically monitor a sample of M dwarfs to detect exoplanets with the radial-velocity method. We are selecting the best planet host star candidates. For that, we are compiling a comprehensive list of dwarf stars coming from existing spectroscopic and photometric catalogues, as well as from late-type star studies. Currently, we are gathering all available information and determine fundamental properties from observations for approximately 2200 targets. Here we presented the first paper of a series that explains in detail the characterisation of our sample of targets for the CARMENES survey. This paper detailed optical low-resolution spectroscopy.

One of the key stellar astrophysical parameters that we need for each target is its spectral type. From the spectral type we estimate the stellar mass and infer planet detectability thresholds, and we ensure that we collate an even sampling of early-, mid-, and late-M dwarfs. Here, we undertook low-resolution spectroscopic observations of 753 targets with the CAFOS spectrograph on the 2.2 m Calar Alto telescope. This CAFOS sample contained M-dwarf candidates with poorly constrained spectral types, cool stars in multiple systems, and numerous comparison and standard stars. We classified our targets using both least-squares fitting techniques and 31 spectral indices, of which we chose five indices with small dispersion to empirically calibrate spectral types (TiO 2, TiO 5, PC1, VO-7912 and Color-M).

Additionally, we investigated the relation of spectral indices with surface gravity. We classified 25 of the observed targets as giant stars using the CaH series of related spectral indices, which are useful indicators to segregate giant stars from dwarfs. Metallicity was estimated through the ζ parameters (Lépine et al. 2007). We concluded that all our field dwarf stars except two new subdwarf candidates have solar metallicity. We identified 49 late-type stars as young dwarfs in star-forming regions or moving groups already reported in the bibliography.

Finally, we also computed stellar activity indicators. Stellar activity is a fundamental property for the CARMENES survey because activity features, such as photospheric spots, can mimic the signature of exoplanets or increase the stellar intrinsic jitter that can mask real exoplanet signals. We computed the pseudo-equivalent width of the Hα line of each target as an activity indicator, and analysed the effect of activity on spectral typing through indices. Although we have identified significant trends for some indices, the spectral type variation due to stellar activity is below one subtype level.

In summary, from the 753 targets that we observed with CAFOS, we obtained for the first time spectral types for 305 stars and improved it for 448 stars. We estimated gravity, metallicity, and activity indices for all targets. We identified 683 M dwarfs, of which 520 fulfill the CARMENES requirements and, therefore, will be included in the list of input targets. A more detailed investigation of these targets with high-resolution spectroscopic and imaging observations to select the best candidates for the CARMENES survey will produce the largest compilation of fully characterised M-type stars.


1

http://carmenes.caha.es – Pronunciation: /kαr’·mεn·εs/

3

Please contribute to the comprehensiveness of our input catalogue by sending an e-mail with suggestions to José A. Caballero,e-mail:e-mail: caballero@cab.inta-csic.es.

4

“Pre-discovery recovery”.

Acknowledgments

CARMENES is funded by the German Max-Planck-Gesellschaft (MPG), the Spanish Consejo Superior de Investigaciones Científicas (CSIC), the European Union through European Regional Fund (FEDER/ERF), the Spanish Ministry of Economy and Competitiveness, the state of Baden-Württemberg, the German Science Foundation (DFG), and the Junta de Andalucía, with additional contributions by the members of the CARMENES Consortium (Max-Planck-Institut für Astronomie, Instituto de Astrofísica de Andalucía, Landessternwarte Königstuhl, Institut de Ciències de l’Espai, Institut für Astrophysik Göttingen, Universidad Complutense de Madrid, Thüringer Landessternwarte Tautenburg, Instituto de Astrofísica de Canarias, Hamburger Sternwarte, Centro de Astrobiología, and the Centro Astronómico Hispano-Alemán). Financial support was also provided by the Universidad Complutense de Madrid, the Comunidad Autónoma de Madrid, the Spanish Ministerios de Ciencia e Innovación and of Economía y Competitividad, and the Fondo Europeo de Desarrollo Regional (FEDER/ERF) under grants AP2009-0187, SP2009/ESP-1496, AYA2011-30147-C03-01, -02, and -03, AYA2012-39612-C03-01, and ESP2013-48391-C4-1-R. Based on observations collected at the Centro Astronómico Hispano Alemán (CAHA) at Calar Alto, operated jointly by the Max–Planck Institut für Astronomie and the Instituto de Astrofísica de Andalucía. This research made use of the SIMBAD, operated at Centre de Données astronomiques de Strasbourg, France, the NASA’s Astrophysics Data System, the RECONS project database (http://www.recons.org), the M, L, T, and Y dwarf compendium housed at http://dwarfarchives.org maintained by C. Gelino, J. D. Kirkpatrick and A. Burgasser, and the Washington Double Star Catalog maintained at the US Naval Observatory. We thank S. Lépine and E. Gaidos for sharing unpublished data with us, J. I. González-Hernández, E. W. Guenther, A. Hatzes, and M. R. Zapatero Osorio of the CARMENES Consortium for helpful comments, and the anonymous referee for the quick and encouraging report.

References

Appendix A: Long tables

Table A.1

Observed stars: identification, common name, Gliese number, 2MASS coordinates and J magnitude, observing date, and exposure time.

Table A.2

Seven representative spectral indices, ζ metallicity index, and Hα pseudo-equivalent width.

Table A.3

Spectral types of observed stars.

All Tables

Table 1

Completeness and limiting J-band magnitudes per spectral type for the CARMENES input catalogue.

Table 2

Sources of the CARMENES input catalogue.

Table 3

Standard and prototype stars.

Table 4

Spectral indices used in this work.

Table 5

Coefficients and standard deviation (in spectral subtypes) of the cubic polynomial fits of the five spectral-typing indices.

Table 6

Giant stars observed with CAFOS.

Table 7

Reported young stars in our sample.

Table A.1

Observed stars: identification, common name, Gliese number, 2MASS coordinates and J magnitude, observing date, and exposure time.

Table A.2

Seven representative spectral indices, ζ metallicity index, and Hα pseudo-equivalent width.

Table A.3

Spectral types of observed stars.

All Figures

thumbnail Fig. 1

Six representative CAFOS spectra. From top to bottom, spectra of standard stars with spectral type 1.0 and 0.5 subtypes earlier than the target (black), standard star with the same spectral type as the target (cyan), the target star (red; in this case, J04393+335 – M4.0 V, Simbad name: V583 Aur B), and standard stars with spectral type 0.5 and 1.0 subtypes later than the target (black). We mark activity-, gravity-, and youth-sensitive lines and doublets at the top of the figure (Hγ, Hβ, He i D3, Na i D2 and D1, Hα, Li i, K i, and Na i, from left to right) and molecular absorption bands at the bottom. Note the three first lines of the Balmer series in emission in the spectrum of the target star.

Open with DEXTER
In the text
thumbnail Fig. 2

CAFOS spectra of our prototype stars. From top to bottom, K3 V, K5 V, K7 V, M0.0–7.0 V in steps of 0.5 subtypes, and M8.0 V.

Open with DEXTER
In the text
thumbnail Fig. 3

Spectral typing of J04393+335 with best-match and methods. Top panel: best-match. CAFOS normalised spectra of the target star and of the standard star that fits best (top) and the difference (bottom). Bottom panel: . Values of χ2 as a function of the spectral type (open red circles). The vertical dashed line marks the spectral type at the lowest χ2 value. Note the logarithmic scale in the Y axis.

Open with DEXTER
In the text
thumbnail Fig. 4

Nine representative spectral indices as a function of spectral type. Filled (red) circles: standard stars. Open (blue) circles: remaining target stars. Dashed (red) line: fit to straight line, parabola, or cubic polynomial, drawn only in the range of application of the fit. Top left and middle: TiO 2 and TiO 5 (Reid et al. 1995), with negative slopes and useable up to M7 V; top right: PC1 (Martín et al. 1996), a monotonous spectral indicator from K5 V to M8 V; centre left and middle: VO-7912 (Martín et al. 1999) and Color-M (Lépine et al. 2003), two indices very similar to the PC1. Note the logarithmic scale in Color-M; centre right: ratio B (Kirkpatrick et al. 1991), sensitive to several stellar parameters and, thus, with a large scatter in the spectral type relation; bottom left: CaH 3 (Reid et al. 1995), with a slightly larger scatter than pseudo-continuum or titanium oxide indices, due to metallicity; bottom middle: VO (Kirkpatrick et al. 1995), useable only for determining spectral types later than M4 V; bottom right: I2 (Martín & Kun 1996), with a linear range of variation from mid-K to mid-M and a sudden increase (or high dispersion) at late-M.

Open with DEXTER
In the text
thumbnail Fig. 5

Spectral type comparison between our results and those from PMSU (top panel), Lépine et al. (2013; middle panel), and Lépine & Gaidos (2011; from VJ photometry, bottom panel). The larger a circle, the greater the number of stars on a data point. Dotted lines indicate the one-to-one relationship.

Open with DEXTER
In the text
thumbnail Fig. 6

Ratio C vs. PC1 index-index diagram. The different symbols represent field dwarfs (small dots, blue), giants (open squares, dark blue), Taurus stars (filled stars, red), and other young stars (open stars, red). All giants are below the dashed line at Ratio C = 1.07. The dashed line is the empirical border of the giant star region.

Open with DEXTER
In the text
thumbnail Fig. 7

CaH 2 + CaH 3 vs. TIO 5 index-index diagram of our CAFOS stars, after discarding giants. The spectral types at the top are indicative, but follow the TiO 5 fit given in Table 5. The solid and dash-dotted lines are iso-metallicity curves of the ζ index.

Open with DEXTER
In the text
thumbnail Fig. 8

Pseudo-equivalent width of Hα line vs. spectral type diagram. Measurement errors are 0.5 subtypes for the spectral type and are of the order of 10% with a minimum of 0.1 Å for pEW(Hα) (but scatter due to variability is probably larger than 10%). Dotted, dash-dotted, and dashed lines indicate the boundaries between chromospheric and accretion emission for different authors. Giants, plotted with open squares, have filled Hα lines or in absorption. None of the young stars, plotted with open and filled stars, show accretion emission. The four stars at the accretion boundary discussed in the text are encircled.

Open with DEXTER
In the text
thumbnail Fig. 9

Spectral-typing indices TiO 2, TiO 5 (top), PC1, VO-7912 (middle), Color-M, and Ratio B (bottom) as a function of Hα pseudo-equivalent width for spectral types M3.0 V, M3.5 V, M4.0 V, and M4.5 V, from top to bottom. For clarity, the indices are offset in steps of 0.1 in the vertical axis. Solid lines are linear fits of the indices as a function of pEWH(α). Dashed lines indicate the mean and ± 1σ index at quiescence (pEWH(α) > –1 Å).

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.