M dwarfs, L dwarfs and T dwarfs

Prev NStars home page --- | --- PMSU home page Next

M dwarf spectral classification

L dwarf characteristics - photometry

L dwarf characteristics - temperature & mass

L dwarf and T dwarf spectral classification

3. Classifying L dwarfs - optical spectra

The first question which usually comes up in discussing L dwarfs is, "Why L?". The following table (adapted from Kirkpatrick et al, 1999) gives as much of an answer to that question as is possible. Basically, there are relatively few letter left which can be used without ambiguity:
Letter Status Comments
A in use Harvard/MK spectral class
B/td> in use Harvard/Mk spectral class
C in use MK spectral class, carbon stars
D ambiguous confusion with white dwarfs, DA, DB, DC, DZ
E ambiguous elliptical galaxies
F in use Harvard/MK spectral class
G in use Harvard/MK spectral class
H available
I problematic confusion with irregular galaxies, Io vs I0
J in use > carbon star spectral class
K in use Harvard/MK spectral class
L available
M in use Harvard/MK standard spectral class
N in use carbon star spectral class
O in use Harvard/MK spectral class
P problematic possible confusion with P Cygni or planets
Q problematic confusion with QSOs
R in use carbon star spectral class
S in use AGB star spectral class
T available
U problematic confusion with UV sources
V available possible confusion between V0 and VO
W problematic possible confusion with Wolf-Rayet stars
X problematic association with X-ray sources
Y avaiable
Z problematic too much of a sense of finality
The arbitrary part comes in that `L' was selected as being closest to `M' in the alphabet. `T' has been adopted as the designation for the even cooler methane dwarfs (see further below). Note that neither of these designations has IAU sanction, in keeping with the classifications in the MK system.

Figure L2.1: The main atomic and molecular features in L dwarfs. (see K99, K00)

The original L dwarf classification scheme was defined based on spectra covering the red and far-red regions of the energy distribution, 6300 - 10200 A. Figure L2.1 shows representative spectra from that grid, identifying the main atomic and molecular features. Even at spectral class M9, TiO bands are prominent in the spectrum at these wavelengths, although they have been declining in strength from ~M7. Those bands become progressively weaker and most disappear by L3, although the 8432 Angstrom band is still detectable at L6. Vanadium oxide is also prominent in the latest-type M dwarfs, at ~7300 and ~7800 Angstroms, and the latter band still evident, although weak, at spectral type L3. But VO also disappears from the spectrum by L5, leaving the metal hydride bands (mainly FeH, but also CaH and CrH) as the dominent molecular features. Those bands reach maximum strength at mid-L types, before weakening in the latest L dwarfs, where H2O provides the most prominent molecular absorption. at ~9300 Angstroms.

At the same time as the molecular bands decrease in strength, a number of atomic lines become significantly more prominent. The resonance lines due to Rubidium (7800, 7948 A) and Caesium (8521, 8943 A) are discernible in the early L dwarfs, and grow in strength throughout the sequence. Lithium I (6708 A) is also evident in L dwarfs (and M dwarfs) with masses below 0.065 Msun (see the following page for a more extensive discussion).

Figure L2.2: The KI 7665/7699 resonance doublt in ultracool dwarfs

The most impressive changes, however, are associated with the 7665/7699 K I resonance doublet. The lines are relatively narrow in mid-type M dwarfs, although more pronoinced than in giants of similar spectral type. The lines broaden with later spectral type through spectral type M7/7.5, but become narrower, with lower equivalent widths, at spectral types M8 and M9. However, that trend reverses between M9.5 and L0, and the lines broaden substantially until by spectral type L5 the separate cores are indistinguishable, and the doublets merge to form a broad trough, ~600 A wide. By spectral type L8, the atomic feature due to potassium is ~1000 A wide.

Figure L2.2: Ultracool dwarfs at blue/green wavelenths (from Reid et al, 2000).

At shorter wavelengths, the same type of changes are evident. TiO absorption has essentially disappeared by spectral type L2, with metal hydrides (CaH, MgH) and CaOH providing the main molecular features. The sodium D lines, another resonance doublet, mimics (and anticipates) the behaviour of the KI doublet by brodening from a normal atomic feature at spectral type M9 to a 1400A wide shallow trough at spectral type L5. As an aside, the combination of absorption features present in the mid-type L dwarfs would lead to those dwarfs having a `true colour', as observed with the naked eye at appropriate light levels, of a somewhat purplish hue - approximately lilac (thus providing a post-hoc justification for L, were such necessary).

The origin of these spectral changes lies with the formation of a variety of solid condensates (i.e. dust particles) as the dwarf atmosphere cools below ~2600K (or ~spectral type M7). Tsuji et al (1996) originally suggested the possibility of this process, pointing out that backwarming from dust could help explain a longstanding discrepancy between the predicted (deep) and observed (shallow) depth of the near-infrared H2O bands. Almost contemporaneously, Fegley & Lodders (1996) pointed out corresponding changes in the atmospheric composition as materials are removed from gas to solid phase. More extensive chemical equilibrium calculations have since been made by Lodders (1999), Burrows & Sharp (1999), Hauschildt et al (2000) and, most recently, Lodders & Fegley (2002). Some of the major transformations are the following:
Species Condensate Temperature of Spectral class
full depletion of disappearance
TiO CaTiO3 (perovskite) 2300-2000K ~L2
VO solid VO 2300-2000K ~L2
CrH metallic CrH ~1400K >L8
CO C bound in CH4 1200-1500K L/T transition
K, Rb, Cs, Li KCl, RbCl, CsCl, LiCl <1200K mid-T?
In addition, silicates such as enstatite (Mg2SiO3) and fosterite (Mg2SiO4) are predicted to form at temperatures between 2600 and 1900 K (late M to early L). The formation of those grains depletes both Mg and oxygen from the gas phase (where H2O remains the dominant molecule) and changes the temperature structure in the atmosphere.

Figure L2.4: The L dwarf sequence - earlier spectral types

Removing increasing numbers of these molecules from the gas phase leads to decreased opacity, and correspondingly weaker absorption features in the optical spectra. Thus, the gradual reduction in prominence of the TiO bands from M7 to L3 (Figure L2.4) and their eventual disappearance at mid-L types reflects the reduced abundance of this molecule with decreasing temperature. The later disappearance of VO also meshes with this scheme. Dust formation probably also accounts for the initial decrease in strength of the K I lines - initially, dust contributes a scattering layer at relively high levels in the atmospheres of late-type M dwarfs, but in lower temperature atmospheres (later spectral types) the dust particles either `rain out' to greater depths (below the photosphere) or form larger particles, in either case reducing scattering at optical wavelengths.
Notice that the original L dwarf spectral type sequence was defined solely on spectral morphology. The expectation at the time was that this would result in a scale which is correlated primarily with effective temperature, but the detailed theoretical analysis justifying that expectation is independent of the definition of the classification system.

Figure L2.5: the L dwarf sequence - later spectral types

By spectral types mid- the late-L, almost all of the gas-phase TiO and VO molecules have been depleted onto grains. In addition, the scarcity of free electrons leads to a reduced abundance of H-, and further reductions in the level of continuum opacity. This results in a substantial reduction in the opacity at optical wavelengths, and a much smoother energy distribution at these later types. In these highly transparent atmospheres, the `photosphere' lies at substantial physical depths, leading to high column densities for the surviving alkali elements (Na, K, Rb, Cs, Li) and the strong absorption lines evident in Figure L2.5. The relative strength of those lines reflects the relative abundances of the different species, while the extremely broad profiles of the Na D-lines and, at later types, the KI 7665/7699 doublet are due to pressure (van der Waal's) broadening, as in degenerate white dwarfs. The continued presence of metal hydrides is analogous to the situation in cool, metal-poor extreme subdwarfs: few metallic atoms, but lots of H.

All of these issues are discussed in more depth in K99, K00 and Reid et al (2000), and in the theoretical papers refernced above. A more detailed description of the observed spectral changes as a function of spectral class is given here (Table 6 from K99). Optical spectra of a representative sample of late-M and L dwarfs are available on this web page.

4. Classifying L dwarfs - near-infrared spectra

Late-type dwarfs produce most energy, and are brightest, at near-infrared wavelengths. It therefore makes sense to examine the near-infrared spectra of these objects to determine whether suitable spectral classification and/or luminosity indicators are present at those wavelengths. This question has been examined in several studies, notably by Leggett et al. (2000), Reid et al. (2001), McLean et al (2000, 2002) Testi et al (2001) and Geballe et al (2002). Most of the examples shown below are taken from the second-cited study.

Figure L2.6: Ultracool dwarfs at near-infrared wavelengths - earlier spectral types. The hatched areas span regions of the spectrum which are adversely affected by terrestrial water absorption (spectra from Reid et al, 2001).

The most prominent features at near-infrared wavelengths are broad absorption bands at 1.4, 1.9, 2.7 and. to a lesser extent, 1.1 microns, produced by rotational-vibrational transitions in the H2O molecule (see Auman, 1967). These bands appear in early M dwarfs and grow in strength with increasing spectral type and decreasing temperature. The fact that water vapour in Earth's atmosphere also contributes substantial absorption at these wavelengths hampers analysis to some extent, even from a high, dry site like Mauna Kea. However, the higher temperatures in the stellar atmospheres means that the associated steam bands are broader than terrestrial absorption, so the wings are accessible for measurement. The first detailed models were computed by Mould (1976) and, as noted above, both narrowband photometry (Persson et al, 1977) and spectrophotometry (Reid & Gilmore, 1984; Berriman & Reid, 1987) showed that the models tended to overestimate the depth of the bands as a function of temperature. Re-radiation by dust, heating the upper levels of the atmosphere, is now generally regarded as the most likely source of this discrepancy. In any event, the water bands continue to increase in strength through spectral type L.

Figure L2.7: Ultracool dwarfs at near-infrared wavelengths - later spectral types

With the availability of near-infrared arrays in the 1990s, more detailed spectroscopy became possible, and Jones et al (1994, 1996), in particular, analysed some of the atomic lines evident in the shorter-wavelength J passband (centred on 1.2 microns). Those features remain fairly constant in strength through late-M to mid-L, becoming weaker in the latest L dwarfs. The H-band (centred on 1.6 microns) is almost featureless in M and L dwarfs. The most prominent feature in the 2.2 micron K-band is the molecular bandhead at 2,3 microns due to overtone absorption by the CO molecule. Figure L2.7-2.9 provide more detailed maps of the different wavelength regions.

Figure L2.7: Ultracool dwarfs at near-infrared wavelengths - J-band spectra. The spectra are ordered by spectral type: LP 412-31 (M8); TVLM (M8.5); LHS2065 (M9); 2M0345 (L0); 2M0746 (L0.5); 2M0829, 2M1029, Kelu 1 (L2); Denis1058 (L3); 2M0036 (L3.5); GD 165B (L4); 2M1112 (L4.5); Denis1228, SDSS0539 (L5); Denis0205 (L7); 2M0825 (L7.5); and 2M0310 (L8). 2M0746, Denis1228 and Denis0205 are known to be near equal-mass/equal-temperature binaries.

The main features evident in the J passband are atomic lines due to neutral sodium (1.14 microns) and potassium (two doublets at 1.169/1.177 microns and 1.244/1.253 microns). The Na I line has an equivalent width of ~20 Angstroms, and shows relatively little variation in strength with increasing spectral type from M8 to at least L5. The line probably weakens at later types, although this may partly be due to increased absorption by the nearby 1.1-micron H2O band (intrinsic, not terrestrial). The 1.17-micron potassium lines have individual equivalent widths of ~5 A at M8, rising to ~10 A at L5, before dropping sharply in strength in the later L dwarfs. The 1.25-micron doublet is somewhat stronger, but shows identical behaviour.

Molecular absorption due to FeH is present in the J passband, but water is the strongest molecular feature. The H2Oa index measures the depth of the blueward wing of the 1.4-micron steam band. As discussed further below, this index is well correlated with spectral type, increasing in depth with decreasing temperature.

Figure L2.8: Ultracool dwarfs at near-infrared wavelengths - H-band spectra

The H-band offers little in the way of obvious atomic or molecular features. Jones et al (1994) identify a KI doublet at 1.5167/1.5172 microns in their M dwarf spectra, and there are features in several of the mid-type L dwarfs (eg 2M0036) at 1.58, 1.613 and 1.627 microns. Reid et al (2001) originally suggested that those might have a molecular origin, and Wallace & Hinkle subsequently identified the absorption as due to FeH. Both the redward wing of the 1.4-micron steam band and the blueward wing of the 1.85-micron band encroach on the terrestrial window, and the H2Ob and H2Oc indices measure those features, respectively. As with H2Oa, the depth is well correlated with increasing spectral type, particularly for H2Ob, as discussed further below.

Figure L2.9: Ultracool dwarfs at near-infrared wavelengths - K-band spectra

Besides the CO bandhead and the redward wing of the 1.85-micron steam band, the K-band includes several atomic lines, notably the Na I doublet at 2.3062/2.2090 microns. Jones et al (1994) also identify Ca I lines at 1.9272, 1.95, 1.987 and 1.992 microns in M dwarf spectra. The NaI line shows the most interesting behaviour of these species. The line is clearly present in the late-type M dwarfs plotted in Figure L2.7, but appears absent in the L2 dwarfs Kelu 1 and 2M1029. A feature re-appears at that wavelength in the latest-type dwarfs, notably 2M0825 and 2M0310. Nakajima et al (2002) have suggested that this 2.2-micron feature is not sodium, but is rather due to methane absorption - a weak precursor of the characteristic identifier of T dwarfs (see the following section).

The suggestion that methane is present in absorption in the K-band in L dwarfs is not unprecedented - Delfosse et al. (1997) originally suggested that methane absorption might account for the steep slope observed longward of 2.15 microns in Denis0205, the latest-type dwarf in their sample. Inded, methane is expected to be present in the lower-temperature upper reaches of the L dwarf atmoisphere, and 3.3-micron CH4 absorption has been observed in even mid-type L dwarfs (Noll et al, 2001). However, the latter feature is the fundamental band, while the 2.2 (and 1.6) micron features are overtone bands. Moreover, Tokunaga & Kobayashi (1999) demonstrated that the overall shape of the Denis0205 spectrum is better explained by H2 absorption. Nonetheless, more recent evidence is accumulating that some of the narrower features (at least those in the K-band window) identified by Nakajima et al may indeed be due to methane (McLean et al, 2002).

Figure L2.10: The correlation between spectral type and the 4 H2O indices defined in Figures L2.7-2.9

Figure 2.7 to 2.9 identify four narrowband indices designed to measure the depth of water absorption in these late type dwarfs. All four are constructed as flux ratios,

H2Oa = F(1.34)/F(1.29)
H2Ob = F(1.48)/F(1.60)
H2Oc = F(1.80)/F(1.70)
H2Od = F(2.0)/F(2.16)
where F(1.34) represents the average flux in a band, with +/-0.01 microns, centred on 1.34 microns. Figure L2.10 shows the correlation between these measured indices and spectral type. Clearly, the two indices measuring the depth of the 1.4-micron band offer the best shortcut to spectral type estimation. Testi et al (2001) and Burgasser et al (2002) devise similar indices, some of which show less scatter than these. However, Burgasser et al point out that H2Ob index shows a remarkably linear relation with spectral type
SpT = (12.6+/-0.9) - (26.7+/-0.6) H2Ob
where SpT = 0 at spectral type T0, SpT = -4 at L5, and the relation is valid for spectral types between M5 (-14) and T8 (8).

All of the near-infrared spectra of late-M and L dwarfs plotted here are available at the foot of this web page.

5. The L-dwarf/T-dwarf transition

As brown dwarfs cool through spectral types M and L, their spectral energy distribution change as the composition of the atmosphere changes. The most dramatic change at near-infrared wavelengths occurs at temperatures between 1400 and 1200K, where methane takes over from carbon monoxide as the dominant repository of carbon. CH4 has extremely broad absorption bands in the near-infrared region of the spectrum, which, as is now well known, modify significantly the near-infrared broadband colours.

Figure L2.11: The optical/near-infrared spectrum of Gl 229B (from Oppenheimer et al, (1995).

Figure L2.11 plots the optical and near-infrared spectrum of Gl 229B, the prototype T dwarf. Strong methane absorption at 1.6 and 2.2 microns removes 50 to 60% of the flux from the H and K passbands, resulting in near-infrared colours similar to those of A stars, rather than late-type dwarfs (see the CMD page and L dwarf photometric properties for more details). Gl 229B analogues have (J-K)~0+/-0.1 magnitude, while the latest-type L dwarfs have (J-K)~2. In Gl 229B, the CH4 bands are saturated. Dwarfs with unsaturated CH4 absorption should have intermediate colours; finding such dwarfs not only helps map out the chemical changes in cool brown dwarf atmospheres, but their frequency relative to late-L and Gl229B-like T dwarfs allows an estimate of the rapidity of the L/T transition.

Figure L2.12: The L/T transition:

2MASS is not the ideal hunting ground for sources with intermediate near-infrared colours - particularly when those colours prove to be most similar to late-K giants. On the other hand, the optical/near-infrared colours of these objects (and Gl 229B) remain extremely red, allowing their identification in deep optical/far-red surveys. Thus, the Sloan Digital Sky Survey (SDSS) was responsible for the discovery of both the first isolated T dwarfs(SDSS1624+0029; Strauss et al, 1999) and the first L/T transition objects (Leggett et al, 2000). Figure L2.12 compares the near-infrared spectrum of one of these objects, the T3 dwarf SDSS1021-0304, against a late-L and mid-T dwarf. CO is still weakly discernible in the SDSS1021, but methane is clearly starting to dominate the H and K bands. Note that the 1.1-micron water band shows a clear progression of strength from detectable at L7.5 to saturated at T6.5. Reliable statistics are not yet available for the relative surface density of early-T and late-L dwarfs, but the indications are that the transition shown in Figure L2.12 is relatively rapid, spanning less than 200K in temperature.

A series of flux ratios designed to measure the strengths of the more prominent near-infrared features, notably CH4 and H2O have been devised independently by Burgasser et al (2002) and Geballe et al (2002). The two T dwarf spectral classification schemese derived from those indices are almsot identical. However, as more and more near-infrared data are being obtained, it is becoming clear that the optical and near-infrared schemes can arrive a significantly different classifications in the late-L/early-T regime. Specifically, some of the early-type T dwarfs (unsaturated, but detectable, methane) have optical spectra characteristic of L6/L7 dwarfs. This suggests that a econd parameter besides temperature (gravity/mass?, weather?) plays a role in the onset of methane absorption.

Figure L2.13: The 0.6-5 micron spectrum of the T dwarf, Gl 229B

Finally, Figure L2.13 shows both the near-infrared and mid-infrared energy distribution of the prototype T dwarfs, Gl 229B. The strong absorption due to the fundamental band of CH4 at 3.3 microns is obvious, as is CO absorption near 5 microns. The gaps in the spectrum reflect regions of high opacity due to terrestrial water and CO2. The latter molecule is not present in significant quantities in cool dwarf atmospheres, and theoretical models (eg Burrows et al, 1997) predict that T dwarfs should emit a substantial fraction of their energy at 4.5+/-0.7 microns. Indeed, this wavelength (and to a lesser extent to 1.2 micron J passband) appears to be the optimum choice for searching for extremely cool (T < 500K) brown dwarfs. SIRTF should provide observational confirmation of this prediction by obtaining low resolution spectrophotometry at mid-infrared wavelengths for a number of late-M, L and T dwarfs.


Prev M dwarf classification --- | --- L dwarf photometric characteristics Next

Prev M dwarf Spectral classification --- | --- L dwarf photometric properties Next

Index NStars home page

Index PMSU main page

Index INR home page

page by Neill Reid, last updated 20/03/2002