INTERMEDIATE RESOLUTION NEAR-INFRARED SPECTROSCOPY OF 36 LATE M DWARFS

Our sample of targets is listed in Table 1. Furthermore, the table also provides Two Micron All Sky Survey (2MASS) and alternate target names, spectral types (determined in the optical), exposure times, observation dates, J-band magnitudes, and reference to spectral types. The J-band magnitude of the stars vary from ∼7 to 13 mag implying spectrophotometric distances of up to 40 pc (Phan-Bao et al. 2008). Figure 1 shows the distribution of our targets in terms of spectral types.

Zoom In Zoom Out Reset image size

Most of our stars were not known to be binaries with the exception of LP349−25 and 2MASS J2206−2047. LP349−25 was first recognized as a nearby M dwarf (Gizis et al. 2002) but was later revealed as a close binary system with an angular separation of 0125 ± 001 (Forveille et al. 2005). 2MASS J2206−2047 was also found to be a binary system with an angular separation of 0168 ± 0007 (Close et al. 2002).

Our sample of 36 M dwarfs was observed with the near-infrared echelle spectrograph NIRSPEC (McLean et al. 1998) installed at the Keck II 10 m telescope. NIRSPEC is a cryogenic cross-dispersed echelle spectrograph with an ALADDIN-3 InSb 1024 × 1024, 27 μm detector. Our observing program was granted one and half nights for semester 2007A and two nights for semester 2007B, giving us a total of 3.5 nights of NASA Keck time (see Table 1 with log of observations). Observing conditions were good except for the nights of April 30 and October 26, which were affected by computer crashes, and for the nights of December 22–23, which were due to bad weather.

We used the echelle mode with the NIRSPEC-3 (J-band) filter. The echelle angle was fixed at ∼63° with the cross disperser at ∼34°. The entrance slit was set at a width of 0432 and a length of 12''. This setup provided a wavelength range of (∼1.15–1.36 μm) split into 11 echelle orders (m = 66–56). For echelle numbering scheme, we refer to McLean et al. (2007). Echelle orders 66, 57, and 56 are heavily contaminated by telluric lines while no atomic lines are observed in orders 62 and 63. Therefore, for this paper we made use of six echelle orders (m = 65, 64, 61–58) which cover an effective wavelength ranges of (1.1649–1.20011 μm) and (1.24081–1.32370 μm). The nominal dispersion ranged from 0.167 Å at blue wavelengths to 0.191 Å at red wavelengths, and the final resolution elements ranged from 0.55 to 0.70 Å. The resolving power of our spectra ranged from (17,800–22,700) Å (Zapatero Osorio et al. 2006; Lyubchik et al. 2007).

The individual exposure times, based on the J-band magnitude of the targets, ranged from 20 to 300 s per co-add. The targets and the telluric standards (spectral type A stars) were nodded in the A–B format, where A and B are the target positions along the slit that are substantially separated (∼7''), in order to subtract sky background. Flats were taken frequently throughout the observing run. We took spectra of a ThAr lamp immediately after every target to ensure accurate wavelength calibration. In order to remove atmospheric telluric lines, featureless stars in the near-IR (spectral types A0–A2) were observed close to the target both in time and in position of the sky to ensure similar airmass.

The data reduction procedure in this paper is similar to that provided in Zapatero Osorio et al. (2006). The bias and dark current subtraction, flat fielding, wavelength calibration, and telluric lines removal are carried out using the echelle package within the Image Reduction and Analysis Facility (IRAF)¹² (Tody et al. 1993; Tody 1986). The flats taken at the same instrumental setup are used to flat-field the images. The sky subtraction on the spectra is performed by subtracting the two nodded images. The spectra are then wavelength calibrated using the emission lines of ThAr lamp. The lines are identified prior to wavelength calibration using the National Institute of Standards and Technology (NIST)¹³ line database. The fits to these lines are made using a third-order Legendre polynomial along the dispersion axis and second-order perpendicular to it. The mean rms for the fit is ∼0.016 Å or 0.38 km s⁻¹ when calculated from the central wavelength of 12485 Å. The wavelength-calibrated spectra is then normalized using the continuum package in IRAF. The reduction of telluric standards is performed in an identical manner. Furthermore, a strong hydrogen line at 1.282 μm and few weaker intrinsic stellar lines present in the telluric standard are removed manually by fitting the line profile using the task splot in IRAF. The wavelength calibrated and normalized target spectra are then divided by the normalized wavelength calibrated telluric standard. The output spectra, A–B, are then free of telluric lines. The whole process is repeated on the other subtracted nodded image, B–A. In order to increase the signal-to-noise ratio (S/N), the spectra of nodded target (i.e., A–B and B–A) are combined using the scombine package within IRAF.

Since FeH and H₂O molecular lines are present all over our spectral range, we present an approximate S/N values based on photon noise calculations. Using the instrument gain factor and normalized photon counts, as exposure times of the targets varied, we compute the S/N for all of our targets. These values are listed in Column 6 of Table 2.

Figure 2 shows a sample of 10 targets that were reduced using the method described above. This figure illustrates the spectra of these targets as seen in six orders. For clarity the stellar spectra are stacked and separated by a constant value. The vertical axis on the right of each plot notes the spectral type. The strong and weak neutral atomic lines are indicated by dotted lines in the figure.

Zoom In Zoom Out Reset image size

The projected rotational velocities (v sin i) presented in this paper are measured by cross-correlating the spectra of our targets with the spectra of slowly rotating template of similar spectral type and observed with the same instrumental configuration. Procedures used to calculate projected rotational velocities are described in the literature (Tinney & Reid 1998; Mohanty & Basri 2003; White & Basri 2003; Bailer-Jones 2004). With the assumption that the line profile is primarily dominated by rotation, projected rotational velocities are calculated by measuring the width of the function obtained from the cross-correlation of the target's spectrum with a slowly rotating template.

To calibrate this relationship, we artificially broaden the template spectrum for a range of velocities (10–50 km s⁻¹, in steps of 5 km s⁻¹ and limb-darkening parameter, = 0.6; Claret 1998). We employ the line rotation profile given by Gray (2005). A spectral order in a velocity broadened template is cross-correlated with the same order in the original template. A Gaussian is fit to the cross-correlation function (CCF) and its width is measured. This width is associated to the velocity by which the original template was broadened for a given order. The process is repeated for the range of velocities and orders. A correlation table between the widths and the velocities is thus created. As a final step, the CCF width measured by cross-correlating a target and the original template is associated with a projected rotational velocity from the table. The process is repeated for all orders. The mean of the associated velocities is taken to be the projected rotational velocity of the star. The errors in the measurement are calculated from the standard deviation of the mean. In our study we employed six orders. Table 2 lists the v sin i of our targets. Furthermore, Table 2 also lists the targets by their 2MASS and alternate names, spectral types, literature v sin i, and their references.

Ideally, non-rotating stars should be used. However, all stars rotate and the utilization of synthetic spectra is outside the scope of this paper. Hence, we have used templates that have published v sin i values well below the resolution of our spectra, which is estimated to be ∼12 km s⁻¹. For all our targets for which we find v sin i at or below 12 km s⁻¹, we report an upper limit.

As our sample consists of targets with a range of spectral types, there is a possibility of spectral mismatch when computing projected rotational velocities with a single template. To see whether spectral type and v sin i of the template affect the calculations of projected rotational velocities, we considered two slowly rotating stars (GJ406 and vB10). The reported projected rotational velocities of these two stars are: v sin i < 2.5 km s⁻¹ (Delfosse et al. 1998) and v sin i = 6.5 km s⁻¹ (Mohanty & Basri 2003), respectively. We measured v sin i of all stars from our sample using these two templates. Figure 3 illustrates this effort. The measured v sin i are within each other errors. However, we obtain higher measurement precision when the target's spectral type is closer to that of a template. Hence we have employed GJ406 (M6.0) as a template to compute v sin i for stars with spectral types M5.0–M6.5 and vB10 (M8.0) for M7.0–M9.5.

Zoom In Zoom Out Reset image size

Absolute radial velocities are determined by cross-correlating the target spectra with the radial velocity template spectrum. The telluric-free-normalized spectra of each object and vB10 are cross-correlated using the fxcor package in IRAF. As most target spectra have high S/N, the peak for each CCF is narrow and well defined. A Gaussian is fit to the CCF peak and the radial velocity is determined.

As our sample spans a wide range of spectral types, we tested the same two templates, GJ406 and vB10, as we did in the previous section. GJ406 is known to have a low projected rotational and absolute radial velocity. However, vB10 has been extensively used in the literature as a radial velocity template with a reported velocity of ∼35 km s⁻¹ (Tinney & Reid 1998; Martín 1999; Basri & Reiners 2006; Zapatero Osorio et al. 2007). Unlike GJ406, vB10 is known to flare occasionally (Linsky 1995). However, radial velocity measurements of vB10 over the last 12 years have shown no significant variation. vB10 had been suspected of harboring a gas-giant planet (Pravdo & Shaklan 2009; Zapatero Osorio et al. 2009) but high-precision radial velocity monitoring (Bean et al. 2010; Anglada-Escudé et al. 2010; Rodler et al. 2011) has refuted such a claim. Thus, we consider GJ406 (19.0 km s⁻¹; Mohanty & Basri 2003) and vB10 (34.37 ± 0.3 km s⁻¹; Zapatero Osorio et al. 2009) as radial velocity templates.

The systematic errors or zero-point shifts that are likely to be present in the wavelength calibration of our data affect the measurement of absolute radial velocities. The observed velocity shifts in the telluric lines are ∼±10 m s⁻¹ (Seifahrt & Käufl 2008), which are less than 10 times the velocity precision we can achieve with NIRSPEC (Zapatero Osorio et al. 2009). Therefore, at these precisions, telluric lines appear to be stationary. These lines, present in the target spectrum, become excellent calibration tool for measuring instrumental zero-point shifts. We cross-correlate the heavily contaminated telluric orders in the target spectrum (e.g., orders 59 and 58) with a synthetic telluric spectrum (Rothman et al. 2009) at the same resolution. The velocity shifts between the two spectra give the zero point. The value is then subtracted from the radial velocity of the target. The radial velocities are corrected for barycentric motion. This procedure is repeated for each of the six orders. After applying the zero-point correction to the measured velocity in each order, the mean of radial velocities values is taken as the heliocentric radial velocity for the target while the standard deviation of the mean is estimated as the error in determination of radial velocities. We find that (Figure 3, left panel) both templates, GJ406 and vB10, yield measurements that are within each other errors. Figure 4 (left panel) compares absolute radial velocities of targets from our sample with those in the literature. We find that the radial velocity measurement precision is better for a target of spectral type that is closer to the template. Therefore, we adopt GJ406 (M6.0) as a template for stars with spectral types M5.0–M6.5 and vB10 (M8.0) for M7.0–M9.5. Table 3 lists absolute radial velocity for our sample. In addition, for completeness, the table also lists trigonometric parallaxes and spectrophotometric distances to the stars. Literature radial velocities are also listed.

Zoom In Zoom Out Reset image size

We employed the Vienna Atomic Line Database (Kupka et al. 2000) to try to identify new lines in the spectra. We set the following stringent criteria to search for new lines: (1) lines with oscillator strengths (log (gf) < −2.0) were excluded to ensure that the most probable transitions are chosen; (2) if a line from the spectra is within 0.6 Å (minimum separation that can be resolved) of a line in the database, then that wavelength value is selected; and (3) we compare the positions of all lines in the target spectra with those positions of lines in the synthetic spectra that have been generated using the WITA6 program (Pavlenko 2000) and the NextGen model atmosphere structures (Hauschildt et al. 1999) for cool stars (Lyubchik et al. 2004). From this procedure, we conclude that new lines other than those identified by McLean et al. (2007) were identified.

If temperatures drop below 4000 K, molecules form in the atmosphere of M dwarfs (Burrows & Volobuyev 2003). Unlike atoms, molecules have a large number of energy levels due to their rotational and vibrational states. Such energy levels lead to band structures where numerous individual transitions take place (Tennyson et al. 2007). These individual transition features dominate the M dwarf's spectra at infrared wavelengths. This severe blanketing of lines hinders identification of the continuum. Therefore, the traditional method of calculating equivalent widths (EWs) cannot be utilized. Instead, we search for the pseudo-continuum which is formed by molecular absorptions (Martín et al. 1996; Pavlenko et al. 1995). EW measurements at these pseudo-continuum are termed p-EWs (Zapatero Osorio et al. 2002).

For the determination of the flux envelope or pseudo-continuum, we use the relation $\sigma = 100\times {\rm {\rm S/N}}^{-1}$. Such an envelope is dependent on the S/N of the spectra. We term such a continuum a pseudo-continuum. Once the pseudo-continuum is determined, the p-EW is measured by direct integration of the observed line in the spectra. We calculate the uncertainty in p-EW by using the relation given by Cayrel et al. (1988), which estimates the uncertainty of p-EW as a function of spectral quality (Stetson & Pancino 2008). As the true continuum cannot be determined precisely, we obtain a lower limit to the estimated uncertainty in p-EW measurements.

We have identified 13 targets in our sample with v sin i below our detection threshold (∼12 km s⁻¹), while four stars (LP44−1627, LP349−25, 2MJ0004−1709, and 2MASSJ1835+3259) show velocities greater than 30 km s⁻¹. Figure 4 (right panel) compares v sin i values of our sample with those in literature (Stauffer & Hartmann 1986; Marcy & Chen 1992; Delfosse et al. 1998; Gizis et al. 2002; Mohanty & Basri 2003; Bailer-Jones 2004; Fuhrmeister & Schmitt 2004; Jones et al. 2005; Reiners 2007; West et al. 2008; Reiners & Basri 2010; Jenkins et al. 2009). Our results indicate low residuals for stars with projected rotational velocities below 30 km s⁻¹. Although the three fast rotating stars show larger velocity dispersion, their measurements are within 3σ of the literature values.

The measurement of rotational velocity of the two components of LP349−25 using NIRPSEC and LGS AO (Konopacky et al. 2012) in conjunction suggests that LP349−25A (M8.0) has a velocity of 55 ± 2 km s⁻¹ while the fainter LP349−25B has a velocity of 83 ± 3 km s⁻¹. Our measurement for the combined spectra is closer to the primary than the secondary. As the secondary is faint, the projected rotational velocity is dominated by the primary. This is particularly evident in works of Reiners et al. (2010) where, using R ∼ 31,000, their measured projected rotational velocity is also closer to that of the primary. The same scenario is observed for the case of 2M2206−2047, where the projected rotational velocities of the two components, A and B, are 19 ± 2.0 km s⁻¹ and 21 ± 2.0 km s⁻¹, respectively. We find the velocity of the combined spectrum to be 22.2 ± 2.0 km s⁻¹. However, as both components are M8.0, their contribution is similar.

Figure 2 plots a sample of our reduced spectra by NIRSPEC echelle orders with order 65 in the upper left panel and order 58 in the lower right panel. Atomic absorption lines are identified by dashed lines. Spectral types are listed on the right side of each plot. These spectra are similar to those shown in McLean et al. (2007); however, with larger sample size each panel shows a continuous transition from M5.0 to M9.5 in steps of half spectral type. Tables 4–6 list the p-EW of 12 neutral lines.

The K i doublet at 11690 Å and the K i triplet at 11771 Å in order 65 (Figure 2, top left panel) show increasing line width with later spectral type. By M9.5 the width of the K i triplet increases to the extent that it blends into a single line. The widening of lines is induced by collisional broadening with H₂ molecules in the atmospheres of M dwarfs (McLean et al. 2007). We also find that the K i lines in order 61 have a similar trend to those in order 65. These results shown in Figure 5 agree with the trend found by McLean et al. (2007) and Cushing et al. (2005). A weak Fe i line in order 65 is observed at longer wavelengths. This line weakens with later spectral type and by M9.5 it is almost indiscernible.

Zoom In Zoom Out Reset image size

The modeled spectra of late M dwarfs by Lyubchik et al. (2004) and Rice et al. (2010) indicate a presence of weaker Fe i and Ti i lines in order 65. We identified their positions, but found them to be heavily blended and hence they are not clearly distinguishable in our spectra. Better spectral resolution is required in order to distinguish these lines.

The Fe i lines (doublet 11886 Å, 11887 Å and a singlet 11976 Å) dominate order 64 (Figure 2, top right panel). At M5.0, Fe i lines are seen as narrow sharp absorption lines. However, with later spectral type, the Fe i line at the shorter wavelength broadens and blends with surrounding lines. Our results agree with those in Cushing et al. (2005) within the errors. A quantitative comparison is discussed in the next section.

Order 64 also contains the Mg i line at 11828 Å and Ti i line at 11893 Å. Both lines are weak and weaken with later spectral type. The Ti i line disappears around M9 while the Mg i line becomes indistinguishable around M8.5. The Mn i line (12899 Å) in order 59 (Figure 2, bottom left panel) maintains its strength with later spectral type while the two Ti i lines are observed to weaken with later spectral types. As seen in the figure, the Mn i line maintains its sharp and narrow feature with later spectral type.

The two Al i lines (13127 Å and 13150 Å) in order 58 (Figure 2, bottom right panel) weaken with later spectral types. The Na i line at 12682 Å (Figure 2, middle right panel) also shows a decline in strength with later spectral type. By M8.5 the Na i is completely indistinguishable.

As a trend of increasing width with later spectral type is qualitatively observed in the neutral atomic lines in Figure 2, a quantitative comparison is desired. We calculate the p-EWs of neutral atomic lines shown in Figure 2. Figures 5–7 plot the computed p-EWs for our targets with respect to the spectral types.

Comparing our data (open circles) with those from Cushing et al. (2005) in Figures 5 and 6, we see that Cushing and colleagues' values (filled diamonds) are within our expected errors. For clarity, in Figures 5 and 6, Cushing et al. (2005) values have been shifted to the right by 0.1 spectral types. However, p-EW values by McLean et al. (2007) for two late M dwarfs do not match with ours or with those of Cushing et al. (2005). These values are not plotted as they are well beyond the range of the plot.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In each panel of Figures 5–7, a linear regression fit to the data is shown as a short-dash line. The linear relation including its y-offset and slope are listed in Table 7 for the 12 neutral atomic lines. These relations quantitatively illustrate the variation of each atomic line with respect to spectral type. The top two panels of Figure 5 show an increasing trend of p-EWs with later spectral type for K i lines. Figure 5 compares the p-EW of our sample with those from Cushing et al. (2005). We find that our p-EW measurements agree with theirs.

The p-EW calculations of Fe i lines in Figure 6 (top panel) suggest small variation with spectral type. However, the Al i line shows a decreasing trend. Mn i line along with two Ti i lines are observed in order 59. As seen in Figure 2, the Mn i line maintains its narrow sharp feature with later spectral type. This line has been studied in greater detail by Lyubchik et al. (2007). They model the Mn i line in five ultracool dwarfs ranging from M6 to L0. Using the WITA6 program (Pavlenko 2000) for the NextGen model structure they have been able to fit the Mn i line, and their p-EW calculations carried out using the DECH20 package on three dwarfs between M6 and M9 indicate a variation of 2 mÅ. Our results confirm that the Mn i line is not sensitive to variation in spectral type.