4. OBJECTS COOLER THAN M STARS

Although we cannot be fully certain of the substellar nature of GD 165B, it deserves mention as the first known object of the new “L” spectral type. Its spectrum was mysterious until recently (Kirkpatrick et al 1999b). It is very red, suggesting that it is very cool, but it does not show the TiO and VO molecular features in the optical and near infrared (NIR) that characterize the M stars. Even before other such objects were finally discovered, work on model atmospheres was showing convincingly that such a spectrum arises because of the onset of photospheric dust formation (Tsuji et al 1996a, Allard 1998).

Dust actually begins to form in mid-M stars. The TiO bands are saturated, then weaken, as one moves to the latest M types. As they are the defining features of the M spectral class, it was suggested by Kirkpatrick (1998) that we really should have another spectral class for cooler objects (which were being called unsatisfactory names like “M10+” or “>>M9”). Martín et al (1997) proposed “L” as an appropriate choice, bearing the same relation to M that A does to B at hotter spectral classes. I should emphasize that not all L stars are BDs, nor are all BDs L stars (and let us agree that “star” in this context is not to be taken literally). Whether a BD is an L star or not depends on both its mass and its age. A BD generally starts in the mid to late M spectral types, then cools through the L spectral class as it ages (eventually becoming a “methane dwarf”). We do not know at which L subclass the minimum main sequence star resides; estimates of its temperature lie in the 1800-2000K range (probably somewhere in the L2-L4 region).

4.1 The Discovery of Field Brown Dwarfs

The discovery of BDs in the field was not very practical until the advent of wide-field CCD cameras or infrared all-sky surveys. Of particular note are the 2MASS and DENIS surveys. These American and European efforts are the first comprehensive, deep looks at the sky in the NIR, and they are producing many faint red objects in the solar neighborhood. Recently they have been joined by the SDSS optical survey, which can detect a similar volume of such objects. BDs lay beyond the sensitivity of older surveys like the Palomar Sky Survey because of their extremely red color and faintness. Even the coolest M subclasses were very sparsely known until recently. Discoveries of BDs in the field were preceded by both cluster and companion BD discoveries. The first announcements were made in 1997, from two very different searches.

One of these was the culmination of a long search for faint red objects with high proper motion (the Calan-ESO survey). A red spectrum of a candidate was obtained in March 1997 (Ruiz et al 1997). This spectrum shows the features now associated with the L dwarfs: broad potassium lines, hydrides, and a lack of TiO bands (Fig. 2). Equally strikingly, it showed the lithium line. As has been discussed above, this guarantees substellar status for all L dwarfs. The team dubbed the object “Kelu-1” (a Chilean native word for “red”).

At about the same time, the DENIS BD team led by Delfosse and Forveille was studying 3 objects that were as red or redder. They obtained NIR spectra of these objects, and showed them also to be L dwarfs (though both these discoveries pre-date the introduction of the “L” terminology). There was a suggestion that the coolest of them might show methane (Delfosse et al 1997), but this has not been confirmed. These objects and Kelu-1 were discussed at the workshop on “Brown Dwarfs and Extrasolar Planets” held in Tenerife in March 1997 (Rebolo et al 1998). This was the first meeting at which the new discoveries of substellar objects were summarized and discussed in detail.

The DENIS objects and Kelu-1 were studied in the optical at high resolution by Martín et al (1997) and Basri et al (1998c). They confirmed the lithium in Kelu-1, and also found lithium in DENIS-P J1228-1547. Lithium detection can be used to place good limits on the mass and age of the objects. They also confirmed that the potassium lines are responsible for the exceptionally strong absorption in these objects. Finally, they found that all of them are rotating rapidly. Lithium in the DENIS object was quickly confirmed by Tinney et al (1997), who also presented the first suite of low resolution optical observations of L stars.

The 2MASS survey was also underway, and soon greatly surpassed the first few objects with a continuing flood of late M and L stars. The early discoveries are summarized by Kirkpatrick et al (1999a), who present a detailed low resolution spectral analysis of 25 objects, and propose a scheme for the L spectral subclasses. Seven of their objects also show lithium (it is still very strong at L5), and so are definite BDs. It is clear that the lithium test works down to the minimum main sequence temperature, below which all objects are automatically BDs. Concerns about whether such very cool objects are still fully convective (probably not) are irrelevant, partly because they are so cool, and partly because they were fully convective at the time they were depleting lithium (when they resembled the Pleiades BDs). A very substantial fraction of the L objects are substellar. The discovery of objects by all-sky surveys has continued apace, and the number of such objects known is rapidly approaching 100. I discuss their numbers further in Section 6.3.

4.2 Definition of the L Spectral Class

A good compilation of the temperature scale for all spectral classes can be found in DeJager & Nieuwenhuijzen (1987}. The temperature ranges spanned by the traditional spectral classes are not uniform (reflecting historical ignorance and old observing techniques, as well as diverse effects of temperature on the appearance of different spectral ranges). The OB spectral classes cover large (>10000K) temperature ranges. The A class covers almost 3000K, and the rest are between 1000K and 1500K (the shortest range is for G stars). The M stars span a range of 1500K.

Although the temperature scale attached to late M stars is still not fully settled, there is general agreement that it ends a little above 2000K. This dictates the beginning of the L spectral class. Where to place the cool end of the L class is not obvious from purely spectral considerations. The main optical/NIR spectral characteristics of L stars are the dominance of hydrated molecules, and the strong neutral alkali atomic lines. The Cs I lines are still visible in Gl 229B, and the Na I and K I line wings are a dominant opacity source in the optical spectra. The conversion of CO to CH₄ is similar to the conversion of other oxides to hydrides that happens at the beginning of the L class. It is not even settled whether we should use the CCD red or NIR ranges for spectral classification.

Nevertheless, the community seems agreed that Gl 229B (a “methane dwarf”) deserves yet another spectral class (on the basis of its strikingly different NIR spectrum). Kirkpatrick has suggested spectral class “T” for methane dwarfs, and this has already has received wide usage (Martín et al 1999c prefer “H”). We do not know how close the coolest currently known L dwarfs are to showing methane, nor is its appearance a logically necessary end for the L spectral class (there is still weak TiO in early L stars). Indeed, the appearance of methane depends on which band one is talking about. The strongest (but observationally more difficult) 3.5 micron band is predicted to appear at about 1600K. The 2 micron bands seen in Gl 229B probably appear below 1500K and become very strong by 1200K, where the optical methane bands are just becoming visible.

Delfosse et al (1999) display a sequence of NIR spectra of L stars. Tokunaga and Kobayashi (1999) find a well-behaved color index in the NIR, but neither set of authors define a subclass scheme. Kirkpatrick et al (1999a) provide a classification scheme for L stars founded primarily on the optical appearance or disappearance of various molecules. Based on model predictions about these molecules (but not on detailed model fitting) they suggest that L0 begin just above 2000K and that L8 be at about 1400K. Martín et al (1999c) present another large set of optical observations, and propose a subclass designation similar in temperature to that of Kirkpatrick et al. Theirs is based primarily on optical color band indicies, and its temperature scale is informed by the detailed model fitting of alkali line strengths by Basri et al (2000). They make the more specific suggestion that L0 be at 2200K, and that each subclass be 100K cooler. This means that L9 would occur at 1300K, consistent with the Kirkpatrick et al scale. The two schemes agree on the spectral appearance of L0-L4 objects.

There is disagreement between the 2 groups about the actual temperature of the coolest 2MASS objects, however. Based on the weakening of CrH, Kirkpatrick et al believe their coolest object is about 1400K. Based on fitting of the Cs I and Rb I line profiles, Basri et al assign it a temperature closer to 1700K. An additional fact in favor of the hotter temperature is that methane is not detected in similar DENIS objects (Tokunaga & Kobayashi 1999, Noll et al 1999), while it should be observable at the lower temperature. This is only important because one of the classification schemes will need adjustment to assign the appropriate subclass for the coolest currently known L dwarfs. The community will have to settle this question after a full range of ultra-cool objects is discovered and studied in both the CCD and near-IR spectral ranges, and the models are improved.

4.3 Atmospheres of Very Cool Objects

The behavior of VLM stars and BDs in color-magnitude and color-color plots has been defined both observationally (eg. Leggett et al 1998a), and theoretically (Chabrier, this volume). Since it is well discussed in the latter reference, I concentrate here on the appearance of the spectrum. What distinguishes L stars from M stars is that they are so cool that Ti has been captured in refractory grains, and is not visible in the red in molecular bands (especially at low spectral resolution). The only atomic features visible in the optical are lines of neutral alkaline metals, such as Na and K, as well as the much rarer Cs and Rb (and of course, sometimes Li). In the CCD range commonly observed (650-900 nm), the most striking is the resonance doublet of K at 766,770 nm, which merge together and become a very broad bowl shaped feature covering more than 10 nm as one moves to the cooler L objects (Tinney et al 1998). The NaD lines are an even more spectacular source of opacity, but most spectra do not have the sensitivity to show such broadly depressed flux. Ruiz et al (1997) and Tinney et al (1998) have shown the first comparisons of model atmosphere calculations to low resolution optical spectra of L stars. The models are generally (but not completely) successful. The molecular bands visible in CCD spectra include some VO (in early L stars), and hydrides like FeH, CrH and CaH.

In the near infrared, steam bands become increasingly strong (Fig. 3), along with H₂ and CO (Allard et al 1998). A good compilation of NIR spectra can be found in Delfosse et al (1999). A few atomic lines are seen, particularly due to Na I. The ordering of objects by temperature as deduced from NIR spectra agrees well with that from optical spectra. A detailed discussion of a spectrum and modeling for an L star is in Kirkpatrick et al (1999b). The best fitting models there, as well as in Leggett et al (1998a,b) include both dust formation and dust opacities (although the distribution of grain shapes and sizes is unknown). These do much better, in particular, than models in which dust formation has not been considered. Dust is known to play a strong role even in the late M stars (Tsuji et al 1996; Jones & Tsuji 1997; Allard et al 1998).

From the first observation of strong alkali lines in a cool dwarf, Basri & Marcy (1995) suggested they could be important spectral diagnostics for very cool stars. They had already been observed in very cool giants, and modelers were aware of their potential utility. It is now clear that Cs I resonance lines can serve as a spectral diagnostic with simple behavior throughout the L spectral range (Fig. 4) and extending to the methane dwarfs. One scenario which is fairly successful in modeling the optical line profiles allows the dust to form (and deplete elements like Ti from the molecular source list), but does not use the dust opacity which might result from that. The physical situation mimicked by such “cleared dust” models is condensation of dust followed by gravitational settling of the grains below the line formation region. Tinney et al (1998) find that low resolution optical spectra are better fit by cleared dust models. Basri et al (2000) show that such models are also successful in explaining observations of the Cs I and Rb I line profiles. They derive a temperature scale for the L stars using these models.

The influence of atmospheric convection, cloud formation, and particle suspensions remains to be properly treated (see also Section 4.4.1). It is very likely that the discrepancies above arise because we do not yet understand the formation and disposition of dust in L star atmospheres. One possibility is that the dust in the upper cooler layers condenses to large enough size to settle down to where it still influences the infrared but not the optical. Such a model has been discussed by Tsuji et al (1999) in the context of Gl 229B, but it may well apply to warmer objects. It is worth recalling that there is a range of temperatures in the atmospheres of these objects; in particular they are substantially cooler than the effective temperature in the upper layers. The Lyon group is also working on new “settled dust” models. As we discover more very cool objects, this will be an active area of investigation for the next several years.

4.4 The “Methane” Dwarfs

At extremely cool temperatures (<1400K, which only BDs can attain), methane becomes an increasingly important form for carbon molecules. The infrared spectral energy distribution for Gl 229B was first presented by Nakajima et al (1995). While such objects are very red in a color like I-J, they become bluer in colors like J-K because of methane absorption in the K band (see Chabrier, this volume). There has been a good deal of follow-up work on Gl 229B: Oppenheimer et al (1995, 1998), Matthews et al (1996), Geballe et al (1996), Golimowski et al (1998), Schulz et al (1998), and Leggett et al (1999).

After the discovery of Gl 229B, there was a gap of four years before the next methane dwarf was announced. This is largely due to the faintness of such objects; the current surveys can only see them out to about 10 pc. In fact, most BDs in the Galaxy should be methane dwarfs, since they will cool to the required low temperatures within a few gigayears. There was a burst of discoveries in 1999 (announced at the June AAS meeting). The SDSS team (Strauss et al 1999) found two and the 2MASS team (Burgasser et al 1999) announced four. At first glance they appear very similar to Gl 229B in the NIR, though more careful analysis implies they lie between 1000-1300K. These are all apparently single objects in the field (though of course some could be close BD pairs not yet resolved).

4.4.1 The Atmosphere of Gl 229B

The first analysis of the spectrum of Gl 229B was by Oppenheimer et al (1995). This was followed by more detailed papers (Allard et al 1996; Marley et al 1996). They, along with Oppenheimer et al (1998), generally conclude that it is not matched by an atmosphere containing dust, but the dust free models fit. The effective temperature of Gl 229B is about 900K. Features due to H₂O dominate the spectrum (Fig. 5). Methane (CH₄) is now also a dominant producer of molecular absorption, particularly in the K band (and presumably at 3.5 microns as well). CO is also seen (Noll et al 1997; Oppenheimer et al 1998), and that is surprising for such a cool object. This has been interpreted to mean that there is some convective overshoot that passes through the subphotospheric radiative zone predicted by models (eg. Burrows et al 1997), and brings up species from the hotter interior. The chemical equilibrium of species is quite complicated in the methane dwarfs. It has been discussed with varying degrees of sophistication by Fegley & Lodders (1996), Burrows et al (1997), Lodders (1999), and Griffith & Yelle (1999), among others (see also Chabrier, this volume). It is important that calculations be done in the context of self-consistent radiative/convective equilibrium models, or the temperature structure and mixing will be incorrectly treated and lead to misleading results.

The presence of strong alkali lines (eg. Cs I, Oppenheimer et al 1998) is indicative that they have either not yet formed molecules or are dredged up from below. The optical colors of Gl 229B seemed to require some sort of broad-band opacity in excess of the dust free models, which are substantially brighter than observed (eg. Golimowski et al 1998). This was taken to indicate that a proper treatment of dust, hazes, and aerosols in the atmosphere might be important (Griffith et al 1998; Burrows & Sharp 1999). Recently however, Burrows et al (1999) and Tsuji et al (1999) have suggested that the missing opacity in the 700-950 nm range is actually just the enormous damping wings of K I and Na I (apparently not treated properly in the initial calculations).

Tsuji et al also reconsider the question of where the dust might be, and show that hybrid models with the dust settled below a certain (currently arbitrary) layer, do a better job of matching the spectrum. Basri et al (2000) were led independently to a similar suggestion for the L stars, so this issue will be important to pursue. Optical flux is blocked by the dust in the inner photosphere (where it is cool enough to form dust but not hot enough to evaporate it), and reprocessed to the infrared. The dust is more transparent at longer wavelengths, of course. Then above a certain layer the grains become large enough to settle out, and the optical opacity is freed of the dust (above the infrared photosphere but in the optical line forming region).

Gl 229B provided us the first opportunity to test our understanding of atmospheres intermediate between stars and the giant planets in our Solar System. Because methane dwarfs are brighter than cold planets, it is likely that the first extrasolar planets whose spectra are recorded will be in this temperature range (planets begin as L stars when very young). The discovery of Gl 229B has stimulated a resurgence in the work on opacities, chemistry, and the atmospheric structure of such objects. It is clear that the discovery of more methane dwarfs covering a range of temperatures will now greatly advance this effort.

4.5 Rotation and Activity in Very Low Mass Objects

It is now possible to draw the first conclusions about the nature of magnetic activity and angular momentum evolution for objects near and below the substellar boundary. Among convective solar-type stars, there is a well-known connection between the rotation of an object and the amount of magnetic activity at its surface. The more rapid the rotation, the more active the object, leading to emission in spectral lines like CaII K or Ha, or in coronal X-rays. This in turn leads to a magnetized wind from the object that carries away angular momentum and spins it down (reducing the level of activity). The field is generated by a dynamo, which in solar-type stars is thought to arise primarily at the bottom of the convective zone. Recent thinking is that the non-cyclical half of the Sun’s flux might arise in a turbulent dynamo throughout the convective zone (Title & Schrijver 1998). The fraction contributed by the turbulent dynamo probably increases with the depth of the convection zone, until it takes over when the star becomes fully convective. That would explain why there is no obvious change in stellar activity passing through early M stars (Giampapa et al 1996).

The first indications that something else might happen near the substellar boundary came from observation of an M9.5 star at high spectral resolution by Basri and Marcy (1995). They found that this (old field) star, BRI 0021, has an amazingly high spin rate, and virtually no Ha emission (although some has been sporadically seen since). This suggests that it had never had much magnetic braking, and that the connection between rotation and activity does not apply to VLMS. Delfosse et al (1998) surveyed a complete sample of nearby early and mid M stars, and found that the fraction of fast (>5 km/s) rotators is quite low until M4 or so (the boundary for fully convective stars), and then begins to increase rapidly. Basri et al (1996, 2000) and Tinney & Reid (1998) have found that rapid rotation becomes ubiquitous later than M7 or so (despite the effect of equatorial inclination on vsini). These rapid rotators are characterized by moderate to very weak Ha emission, and all the rotators above 20 km/s have weak Ha emission (less than 5A equivalent width).

Most of the DENIS and 2MASS L dwarfs show no Ha emission. There are a few earlier than L4 that show a little Ha emission (Leibert et al 2000), but the implied surface fluxes are extremely low. Because of the extremely cool photospheres, Ha can only show up if there is chromospheric or coronal heating. It is also the case that a given value of emission equivalent width (say 5A) represents a dramatically weakening surface flux as we move into the late M and L stars. The continuum, which defines the normalization of equivalent width, is dropping very quickly with temperature (Ha now occurs in the Wien part of the Planck function). There cannot be a corona in the stars showing no Ha, because it would create a chromosphere by photoionization (Cram 1982) that would easily show up. Basri et al (2000) find that most of the L dwarfs have vsini corresponding to rotation periods of at most a few hours. Thus it is quite clear that for older BDs and VLMS, the usual rotation-activity connection is completely broken, and may even be reversed (since the late M stars showing stronger emission tend to be the slower rotators).

There are several possible explanations for these results. One is that the ionization levels in the photosphere may have become so low that there is insufficient conductivity to allow coupling of the magnetic field to the gas. Then gas motions do not twist up the fields, and there is no dissipation to heat the upper atmosphere. This has to be true even in the face of ambipolar diffusion, which couples small numbers of ions to the neutrals fairly effectively (as in T Tauri disks). The alkali metals that are the last suppliers of electrons are becoming quite neutral in the L stars. A possible counter-example to this hypothesis is provided by the detection of (non-flaring) Ha emission in a methane dwarf (Liebert et al 2000).

All low mass objects should have turbulent dynamos, which are driven by convective motions. Rotation can enhance production of fields, and the amplitude of convective velocities also does. But convective overturn times scale with luminosity in these objects. At the bottom of the main sequence they can increase to months, while typical spin periods are dropping to hours. The traditional rotation-activity connection may arise because activity increases with decreasing Rossby number (the ratio of rotation period to convective overturn time). Activity levels increases steadily from a Rossby number of unity down to 0.1. They saturate between 0.1 and 0.01, with a hint of a downturn at 0.01 (Randich 1998). The BDs have Rossby numbers in the range from 0.01 to 0.001. I speculate that the dynamo may be unable to operate efficiently at such low levels, perhaps because rotation organizes the flows too much. A possible counter-example to this hypothesis is provided by the very rapid rotator Kelu-1, which exhibits a persistent (though very weak) Ha emission line.

A related possibility is that the field is not actually quenched by rapid rotation, but instead takes on a relatively stable, large-scale character (see Chabrier, this volume) like that of Jupiter. In that case, the field might be sufficiently quiet (especially in conjunction with the low atmospheric conductivity) that it does not suffer the dissipative configurations that power stellar activity. To the extent that acoustic or magneto-acoustic heating play a role, the low convective velocities in these objects will reduce it. Thus, the objects might still have strong fields, but no stellar activity.

This could be tested in principle using Zeeman diagnostics. Valenti et al (1999) have suggested using FeH for objects in this temperature range, and shown that it can work in late M stars. Occasional flaring does occur on some of these objects. Flares have been seen in objects which seem otherwise quite quiescent, such as VB10 (Linsky et al 1995) and 2MASSW J1145572+231730 (Leibert et al 1999). Another possibility is to search for rotational periodicities (photometrically or spectroscopically). These traditionally indicate the presence of magnetic spots. Some very cool objects have shown such behavior (Martín et al 1996, Bailer-Jones & Mundt 1999), but many have not. A possible complication arises if dust clouds condense inhomogeneously in the atmospheres of these objects. One might then detect rotational modulation due to “weather” (Basri et al 1998, Tinney & Tolley 1999). There is no confirmation this has been seen yet; one will have to very carefully distinguish between the two possible sources of variability (spots or clouds) by showing that opacity rather than temperature is the cause (they will cause different effects in different spectral features).

The only BDs which seem to show strong magnetic activity are the very young ones (eg. Neuhauser et al 1999 for X-rays; many examples of Ha emission in SFR and young clusters). These are sufficiently luminous objects that are hot enough and/or perhaps not rotating too fast. In the youngest cases, there may be an added contribution due to accretion phenomena. They all eventually become relatively inactive as the convection weakens and the atmosphere cools. Apparently most objects near or below the substellar boundary are rapid rotators because they have not experienced much magnetic braking.