RI Photometry of 2MASS‐selected Late M and L Dwarfs
ABSTRACT
We report R and I band observations for 204 late M and L dwarfs selected on the basis of Two Micron All Sky Survey 
and photographic red colors, made with a telescope of modest aperture. It is shown that deep surveys covering these red bands can provide data sets that complement 
, permitting a fairly good photometric classification system for L dwarfs. Due primarily to the disappearance of strong TiO opacities, 
reaches a maximum at late M type and turns blueward for subtypes M9 to about L3. Apart from a small plateau at L0–L2, the 
color remains as a monotonic measure of spectral type or temperature over this range, and likewise for M dwarfs. For late L types, both colors probably get redder again, although the accuracy of our data and number of objects do not give us robust conclusions by L6–L8. It is also interesting to look at the dispersions of the 
bands at a given spectral type. It is widely believed that this dispersion is caused by object‐to‐object variations in the amount, location, or other properties of dust or clouds. We find a moderately larger spread for the 
color than for 
.
Received 2005 December 12; accepted 2006 February 13; published 2006 June 5
1. INTRODUCTION
Hundreds of late‐type M and L dwarfs have been discovered within the last several years from analyses of largely photometric surveys: the Deep Near Infrared Survey (DENIS; Delfosse et al. 1997; Phan‐Bao et al. 2001), the Two Micron All Sky Survey (2MASS; Kirkpatrick et al. 1999, 2000), and the Sloan Digital Sky Survey (SDSS; Hawley et al. 2002). While 2MASS is exclusively a near‐infrared survey in the J, H, and 
bands, DENIS includes an I band, while the SDSS CCD photometry features three unique red bands: r, i, and z. For over 600 M dwarfs and several dozen L dwarfs, SDSS r − i colors are reported in Hawley et al. (2002) and Knapp et al. (2004). The r − i colors proved to be of limited use in these papers, as the 1 σ scatter for types later than L0 is large. A brief comparison between SDSS r − i and the Kron‐Cousins R − I colors presented here is made when the appropriate figure for the comparison is presented. In fact, the traditional Kron‐Cousins R and I bands have also hardly been used for observations of very low mass objects, again because these very cool dwarfs emit little flux at the shorter wavelengths. The only paper reporting R and I magnitudes for a significant number of bright L dwarfs is by Dahn et al. (2002).
Other sky surveys such as the ongoing NOAO Deep Wide‐Field Survey (Jannuzi et al. 2000, 2004) combine R and I with the near‐infrared bands. The Pan‐STARRS project team will conduct a deep optical survey of the ∼30,000 deg2 accessible from Hawaii (Kaiser et al. 2005), probably starting in the initial configuration by late 2006 or 2007. A catalog including SDSS‐like r and i magnitudes to approximately 24 mag will be made available to the public. It is planned that as soon as 2012 (C. Claver 2005, private communication), the Large Synoptic Survey Telescope (LSST) will conduct a much deeper, multiepoch survey of the accessible sky at similar optical wavelengths.
Future infrared surveys also featuring very high 
(aperture area times solid angle) will complement Pan‐STAARS and LSST in overlapping fields. A survey already under way with the Wide Field Camera on the UK Infrared Telescope on Mauna Kea, Hawaii, will cover about 7500 deg2 in the Z, Y, J, H, and K bands (Leggett et al. 2005). The VISTA 4 m telescope will be able to image zJHK bands with a 1
65 field for large chunks of the same sky from Chile (Emerson & Sutherland 2002; Dalton et al. 2004). The telescope and camera are scheduled for delivery in 2006.
The main purpose of this paper is to show how deep R and I imaging can be combined with near‐infrared bands in ongoing and future surveys to detect these very low mass dwarfs through late L spectral types. Here we report CCD observations of a significant number of late M and L dwarf candidates selected on the basis of their 
colors from 2MASS. The observing plan and instrumentation are presented in § 2, and the results are discussed in § 3.
2. OBSERVATIONS
Observations were obtained using the Cassegrain‐focus CCD imager at the Cerro Tololo Inter‐American Observatory (CTIO) 1.5 m telescope. Observing runs were in 1999 March, May, and September and 2000 May. Observations were made with the Kron‐Cousins R and I filters. On each night, standards from Landolt (1992) were observed to determine the nightly zero‐point and extinction corrections. Furthermore, since the CTIO 2MASS telescope was observing on the same nights, we were able to verify photometric conditions using the standards observed by 2MASS every hour. All nonphotometric data were discarded. The Landolt standards are significantly bluer than the late M and L program stars. We therefore included redder dwarfs previously observed at the US Naval Observatory (USNO) to investigate the possibility of color terms. We find no evidence of any color terms between the two observatories. Typical exposure times were 300 s in R and 200 s in I, but these were increased for the faintest L dwarfs. The identification of the targets was achieved by making comparisons to the 2MASS finding charts.
Figure 1 shows the R and I bandpass curves appropriate to the Kron‐Cousins filters, according to Bessell (1986), compared to the spectra of the well‐known L5 dwarf 2MASS J15074769−1627386 (Reid et al. 2000).
Fig. 1.— Smoothed spectra for the bright L5 standard 2MASS J15074769−1627386, along with the bandpass curves for Kron‐Cousins R and I filters (Bessell 1986).
Candidate late M and L dwarf targets were selected from a 1999 analysis of the 2MASS Working Database. The procedure used was that of Gizis et al. (2000), with the exception of the fainter 
limit in this project. As in the Gizis et al. sample, the 
photometry is based on USNO scans of photographic sky survey plates. The photometric criteria were
| 1. |
| ||||
| 2. |
| ||||
| 3. |
| ||||
| 4. |
| ||||
and 
cuts are meant to exclude the earlier M dwarfs, while the near‐IR color‐color cut excludes M giants. As discussed below and in Gizis et al. (2000), the 
cut excludes some M7 dwarfs whose colors are at the blue end of the dispersion for this spectral type while allowing earlier M dwarfs with redder 
measurements to scatter into the sample. Only regions with 
from the Galactic plane were searched. No proper‐motion selections were applied. Because the searches were based on the (at that time) incomplete 2MASS sky coverage, the total area that was searched is, unfortunately, not known, but all objects in the seached regions were observed. In Table 1, the 204 individual observations from all four observing runs are listed in order of right ascension, which is the same as the 2MASS J final point‐source designations. R and I magnitudes are listed first, with errors. The 2MASS 
values for each star from the final data release are also included. Twenty‐one M and L dwarfs with published RI photometry—mostly from Bessell (1991) for late M’s and Dahn et al. (2002) for L’s—and known spectral types were observed as “standards.” These data include several that were observed on more than one telescope run and are listed in Table 2. Fortunately, some 63 more have subsequently been classified by K. L. Cruz, I. N. Reid, and collaborators and are included and referenced in Table 1. As we show, these include all of the likely L dwarfs with spectral types later than L1 in our CTIO sample.
2.1. Two Spectra
Two previously unpublished L dwarfs from this project were observed in 2001 May with the CTIO 4 m telescope and optical spectrograph as part of the observations described in Gizis (2002). The smoothed spectra for 2MASS J21420580−3101162 and 2MASS J21512543−2441001 are shown in Figure 2, again with the bright L5 standard 2MASS J15074769−1627386. Despite the low signal‐to‐noise ratio, it is clear that these are indeed L dwarfs. We applied the Kirkpatrick et al. (1999) system on the basis of the appearance of the spectra and measurements of the band strengths to obtain the spectral type. We tentatively classify 2MASS J21420580−3101162 as L3, while 2MASS J21512543−2441001 has been classified as L6 by K. L. Cruz from a separate spectrum (2005, private communication).
Fig. 2.— Smoothed spectra for the L dwarfs 2MASS J21420580−3101162 and 2MASS J21512543−2441000, bracketing a spectrum of 2MASS J15074769−1627386. Abbreviated labels for these stars are displayed next to each spectrum.
3. DISCUSSION
3.1. Comparison with Previous RI Photometry
Extensive R and I photometry of M dwarfs on the Kron‐Cousins system has been published in the literature. We thus observed a selection of well‐studied late‐type M dwarfs taken mainly from Bessell (1991). Table 2 lists these in the second and third columns (
and 
, respectively), followed by the reference number listed at the bottom of the table. The only significant published L dwarf photometry is taken from Dahn et al. (2002). These are objects that are bright enough for these authors to obtain trigonometric parallaxes with an optical CCD detector, and they fortunately include one L7, an L7.5, and an L8 on the Kirkpatrick et al. (1999) optical (red) L dwarf system.
In the four panels of Figure 3, the differences 
and 
are plotted against the literature R and I magnitudes and R − I colors. In the top panel, the mean 
, so there appears to be no significant systematic offset for the stars in common with the literature. Likewise, in the third panel, there is no significant evidence for a color term. The residuals for the I band for these cool objects are understandably smaller. The corresponding values are 
. The systematic offset is again negative but is closer to having real significance.
Fig. 3.— Comparison with published RI photometry for 92 of our observations, including several stars with two to three CTIO measurements (Table 2). The differences 
and 
are plotted against magnitudes and colors, as discussed in the text. Each vertical tick mark is 0.1 mag.
3.2. Effect of the J − Ks Color Selection
In Figure 4, 
for the entire program sample is plotted against 
(filled circles), along with a sample of 303 M and late‐type K dwarfs (open circles) compiled by Leggett (1992). The latter is a continuous distribution through spectral type M9. Bessell (1991) showed that 
varies monotonically with spectral type perhaps more cleanly than any other color, and is therefore likely to be more temperature sensitive. When we show a comparison of those dwarfs with known spectral types in § 3.4, it is seen that 
is not sensitive to spectral type through about M8. The abrupt cut due to the 
selection limit in the CTIO 2MASS sample is obvious.
Fig. 4.— Plot of 
colors against 
for the 2MASS/CTIO targets and standards (filled circles) and for an overlapping sample of M dwarfs through M9 (open circles) compiled by Leggett (1992). The sharp edge due to our selection of 
suggests completeness for spectral types M8 and later.
It is obvious from the figure that many candidates were retained in the sample with 
. This is because we started this project based on an initial processing of 2MASS data that differed from the more accurate final processing. This is not bad, however, since it gives us a good sampling of 
objects down to about 0.8. In any case, this demonstrates that our 
color cut begins to be incomplete near 
= 4, due to dispersion in 
at a given 
(and spectral type). For the latter, Figure 4 shows that completeness begins at spectral type M8, or at worst M9.
3.3. The RIJKs Plots
The 2MASS CTIO photometric sample is shown in an 
versus 
two‐color plot in Figure 5 (filled circles). We find this two‐color diagram to be most useful as a proxy for spectral classification. Indeed, 
monotonically increases with later M type until about M8, when it flattens.
It is seen that the selection technique successfully extracted stars of middle M and later spectral types, with a few outliers having very red 
color compared to the dwarf locus. At least two of these have carbon star spectra; they are unlikely to be M or L dwarfs. For the dwarf locus, the relationship between the two colors is monotonic up to about M6 over a range of nearly 2 mag in both colors. At M6 the 
color appears to become nearly constant with 
at an 
color near 2.5, while 
continues to increase monotonically.
In Figure 6, we show the 1 σ photometric errors of the 
color, based on the rms sum of the individual bands. The “R” error dominates in just about every case (Table 1). To avoid a confused figure, we do not show the corresponding horizontal errors in the 
color. The errors in 
are substantial for only a few of the reddest objects in 
. Clearly, the 
color is not well measured for many of the coolest objects, although there are important exceptions. As we show below, the latest object from the program is DENIS J02550359−4700509 (Martín et al. 1999), a bright dwarf with no prior photometry in the R band. It was classified as L8 by Cruz et al. (2003). The 
color measured for this object is more accurate than for any of the standards later than L3, although the errors for the L7 standard DENIS J02052940−1159297 are almost as small.
Fig. 6.— Plot of 1 σ error bars for the 
colors of the sample (crosses) within the rectangular area of Fig. 5 vs. 
. The standards from Table 2 are shown as filled squares, with errors taken from the literature measurements. The uncertainty in R nearly always dominates the combined 
error for the CTIO targets.
The behavior of these two colors versus spectral type is shown in Figure 7 for 81 observations of our program stars and the photometric/spectroscopic standards (Table 2), including a few repeats with CTIO photometry obtained on 2–3 nights. The region shown is again the area marked with a rectangle in Figure 5. Here the black and cyan symbols indicate M dwarfs, and the number indicates the subtype. For the cyan symbols, one‐half of a spectral type should be added. Thus, a cyan “7” is spectral type M7.5. Likewise, the red symbols and their numbers indicate L dwarfs, with green symbols having that subtype plus 0.5.
Fig. 7.— Plot of 
colors against 
(using published 
magnitudes) for the 81 observations of dwarfs spanning M5–L8 observed to calibrate this photometric data set to spectral type. The panel encompasses the rectangular box in Fig. 5. Most of these spectra of 2MASS sources are reported in the series of papers by K. L. Cruz, I. N. Reid, and collaborators, as documented in Table 1. The black and cyan numbers are M dwarfs of subtype N and N+1/2, respectively, while the red and green are for L dwarfs (see text for details).
Based on what has been learned from spectrophotometry of these objects (e.g., Kirkpatrick et al. 1999, 2000), there is a good understanding of why the 
color stops increasing at late M. The R flux is heavily depressed by TiO absorption for late M dwarfs, but the formation of Ti‐bearing dust beginning at M7–M8 causes this opacity to decrease. This in turn results in a relative increase of the R‐band flux that largely offsets the drop expected from decreasing surface temperature at the later types. Remarkably, the 
colors actually appear to get bluer until about L2–L3. Note that Dahn et al. (2002) found that 
for their stars reaches a maximum near 2.4 for mean values at M7 and then declines to 2.2–2.25 at M9–L0. The color is rising again by type L3. Similar behavior is shown for the SDSS r − i color versus spectral type in § 3.4.
Comparison with the error bars of Figure 6 does indicate that there is evidence for real dispersion in 
among well‐observed program and standard dwarfs as early as M8. This dispersion may be due to variation in the amount and properties of the dust in the atmosphere, as discussed below.
Only at late L type, about L5, does the 
color appear to get dramatically redder. The few well‐observed L7–L8 points show the reddest 
and 
colors.
How good is this diagram as a tool for estimating photometric “spectral types”? We have not tried exploring this with a rigorous analysis, given the paucity of data for the later spectral types.
Another fairly monotonic two‐color diagram (up to a point) is that available to the DENIS project: 
versus 
(Delfosse et al. 1999). This is shown in Figure 8, where we have numbered and color‐coded the objects with spectral classifications as in Figure 7. However, the greater dispersion in 
among the stars of similar L type is apparent. The spread in 
at a given spectral type appears smaller. We investigate the color dispersions in the following paragraphs. For photometric “spectral types,” this diagram is clearly inferior to the previous one. For example, note the wide range in 
for the L2’s. Possible reasons for this are discussed below.
Fig. 8.— Plot of 
colors against 
for the entire data set (small filled circles). However, those with known spectral type are plotted using the numerical and color code from Fig. 7. The dispersion in 
at similar 
or for a given spectral type is noteworthy.
3.4. The 
Dispersions and Implications for Dusty Atmospheres
A valuable check of the statements about color dispersions is the plotting of the four colors versus spectral type in Figures 9 and 10. The dispersions are small among the M dwarfs, but 
in particular shows a large dispersion at the same spectral type for late‐type L dwarfs. (The greater spread for late‐type dwarfs in 
must be attributed primarily to observational error.) The dispersion among late L observations in the 
color appears smaller.
Fig. 9.— Plot of 
and 
colors against spectral type for the full range of the latter. The numerical code is 0 for type M0 and 10 for type L0. The Leggett (1992) K–M sample is plotted as red open circles, and this data set as filled black circles. The mean values of SDSS 
are plotted as blue squares, with the mean errors as given in Table 3 of Hawley et al. (2002). Observational error widens the dispersion in 
for the late L spectral types, although the errors from this data set are generally smaller than the SDSS errors.
To take a more careful look at the dispersions, we need to plot the colors using the same vertical scale. In Figure 11, the 
color combinations are plotted in this manner. The dispersion at 
seems to jump at about L2 and is modestly greater than that of 
. In both cases, the dispersions are not due to observational error. Moreover, they co‐add to make the dispersion in 
a bit greater still.
Since the spectral type is based on the relative strengths of atomic and molecular absorption features, the dispersion in color at a given type is generally attributed to varying amounts and properties of dust in the atmosphere from object to object of a given spectral type (Ackerman & Marley 2001; Tsuji 2002; Marley et al. 2002). In Marley et al., the argument was made that for late L dwarfs, the shorter wavelength bands, such as SDSS i and z, are influenced more by the wings of K i and other alkali gas opacities, which should generally cause optical depth of unity to be reached above the cloud layer.
The effect of the cloud on the emitted radiation field can vary greatly with wavelength. Figure 12 (courtesy of A. Burrows) presents brightness temperatures versus wavelength from models in Burrows et al. (2006), where a cloud layer is assumed to sit at a plausible temperature just below 1700 K. The brightness temperature is defined by Burrows to be the temperature at which τ near unity is reached in the atmosphere at the given wavelength.
Fig. 12.— Brightness temperatures as a function of wavelength for synthetic spectra from model atmospheres of Burrows et al. (2006) and temperatures, starting at the top, of 1400 K (black), 1500 K (red), 1600 K (blue), and 1700 K (green) at solar composition. A cloud or dust layer was placed in the atmosphere just below 1700 K. Strong gaseous opacities above this physical level in the atmosphere cause all of the H‐ and K‐band radiation, and some of the radiation at I and J, to be emitted above the cloud. (Courtesy of A. Burrows.)
From top to bottom (with successive colors black, red, blue, and green), the stellar 
values range from 1400 to 1700 K (in 100 K steps) in the mid to late L range. Strong gaseous opacity makes the entire 
band, and probably also H, opaque at levels above where the cloud sits. On the other hand, while I is affected by gaseous alkali opacities, especially the K i wind, some of the radiation field of the band is formed nearer the cloud level, and the same is true for J.
Does this mean that the 
band is less affected by the cloud cover than the other bands? Not necessarily. Marley et al. (2002) show explicitly how the presence of the cloud affects the entire temperature profile of the atmosphere, including that physically above the cloud layer. The Burrows et al. group’s models demonstrate the same thing. This is a complicated story. Our observational work does suggest that 
has a larger dispersion than the other colors. It is our hope that the theorists can sort out why this is true.
4. SUMMARY
L dwarfs are feeble emitters in the I and especially the R band. Nonetheless, deep surveys that cover these bands can provide data sets that complement 
, enabling a fairly good photometric classification system. 
turns blueward at about M9 for subtypes M9 to about L3. Except for the flat “kink” near L0–L2, the 
color remains a monotonic measure of spectral type or temperature over this range, and likewise for M dwarfs. For late L types, both colors are probably getting redder, although the accuracy of our data and the limited number of objects do not allow us to make robust conclusions by L6–L8.
It is also interesting to look at the dispersions of the 
bands at a given spectral type, which should be a proxy primarily for a dwarf’s 
. It is widely believed that this dispersion is caused by object‐to‐object variations in the amount, location, or other properties of dust or clouds. At mid to late L types, one might expect the 
flux to be emitted above the cloud deck and therefore show less dispersion. However, this does not seem to be the case.
This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aerospeace and Space Administration and the National Science Foundation. We are most grateful to David Monet for personally matching all of the 2MASS and USNO scans. We thank Adam Burrows for providing Figure 12, Davy Kirkpatrick for finding positions and designations for a few difficult 2MASS sources, and Kelle Cruz for cross‐checking our observed targets with the spectroscopic results of the NStars project, published in the series of papers “Meeting the Cool Neighbors” and supported by NASA JPL grant 961040NSF. This research was supported by this same grant. Helpful discussions with Michael Cushing are acknowledged.
REFERENCES
- Ackerman, A. S., & Marley, M. S. 2001, ApJ, 556, 872
- Bessell, M. S. 1986, PASP, 98, 1303
- ———. 1991, AJ, 101, 662
- Burrows, A., Sudarsky, D., & Hubeny, I. 2006, ApJ, 640, 1063
- Cruz, K. L. 2004, Ph.D. thesis, Univ. Pennsylvania
- Cruz, K. L., Reid, I. N., Liebert, J., Kirkpatrick, J. D., & Lowrance, P. J. 2003, AJ, 126, 2421
- Dahn, C. C., et al. 2002, AJ, 124, 1170
- Dalton, G. B., et al. 2004, Proc. SPIE, 5492, 988
- Dawson, P. C., & Forbes, D. 1989, PASP, 101, 614
- Delfosse, X., Tinney, C. G., Forveille, T., Epchtein, N., Borsenberger, J., Fouqué, P., Kimeswenger, S., & Tiphène, D. 1999, A&AS, 135, 41
- Delfosse, X., et al. 1997, A&A, 327, L25
- Emerson, J. P., & Sutherland, W. 2002, Proc. SPIE, 4836, 35
- Gizis, J. E. 2002, ApJ, 575, 484
- Gizis, J. E., Monet, D. G., Reid, I. N., Kirkpatrick, J. D., Liebert, J., & Williams, R. J. 2000, AJ, 120, 1085
- Hawley, S. L., et al. 2002, AJ, 123, 3409
- Jannuzi, B. T., Dey, A., Tiede, G. P., & Brown, M. J. I. 2000, BAAS, 197, 77.09
- Jannuzi, B. T., et al. 2004, BAAS, 205, 81.06
- Kaiser, N., & the Pan‐STARRS Collaboration. 2005, BAAS, 206, 24.02
- Kirkpatrick, J. D., Henry, T. J., & McCarthy, D. W., Jr. 1991, ApJS, 77, 417
- Kirkpatrick, J. D., et al. 1999, ApJ, 519, 802
- ———. 2000, AJ, 120, 447
- Knapp, G. R., et al. 2004, AJ, 127, 3553
- Landolt, A. U. 1992, AJ, 104, 340
- Leggett, S. K. 1992, ApJS, 82, 351
- Leggett, S. K., the WFCAM Team, & the UKIDSS Cool Dwarf Group. 2005, in Proc. 13th Cambridge Workshop on Cool Stars, Stellar Systems, and the Sun, Vol. 2, ed. F. Favata, G. Husain, & B. Battrick (Noordwijk: ESA), 751
- Marley, M. S., Seager, S., Saumon, D., Lodders, K., Ackerman, A. S., Freedman, R. S., & Fan, X. 2002, ApJ, 568, 335
- Martín, E. L., Delfosse, X., Basri, G., Goldman, B., Forveille, T., & Zapatero Osorio, M. R. 1999, AJ, 118, 2466
- Phan‐Bao, N., et al. 2001, A&A, 380, 590
- Reid, I. N., & Cruz, K. L. 2002, AJ, 123, 2806
- Reid, I. N., et al. 2000, AJ, 119, 369
- Tsuji, T. 2002, ApJ, 575, 264
-
1 Visiting Astronomer, Cerro Tololo Inter‐American Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation.