Spectroscopic and photometric identification of
candidate nearby dwarfs
The existence of hydrostatic equilibrium in hydrogen-burning stars produces
a well-defined sequence in the luminosity/effective temperature plane -
the main-sequence, originally identified by Hertzsprung and Russell. This correlation
is, of course, driven by mass - higher mass demands more energy support, which requires
higher central pressure and temperature, producing, in turn, a higher surface temperature.
In the observational plane, this relation between physical parameters manifests itself
as a correlation between absolute magnitude and colour/spectral type.
Those observed relations provide the basis for (spectro)photometric parallax surveys:
measure the spectral type or colour of a sample of stars; use those measurements to
estimate the absolute magnitude; determine the distance modulus by comparing the
inferred absolute magnitude with the observed apparent magnitude. This technique
has been widely applied over the last 50+ years, not only as a means of identifying
stars likely to be within the immediate Solar Neighbourhood, but also in larger-scale
surveys of the structure of the Galaxy (for example, Reid & Majewski (1)).
This method is one of those employed in our NStars survey to search for nearby M dwarfs. These pages give a brief potted history of past work in this area and summarise our current efforts.
M dwarf spectra are characterised by the presence of absorption bands due to titanium oxide (TiO), which grow in strength with decreasing temperature. These prominent features are easily detectable, even at very low spectral resolution. Thus, M dwarf surveys could be undertaken using objective prism plates taken with wide-field telescopes, such as the McCormick telescope, used by Vyssotsky in the 1950s (2), and the Burrell Schmidt, used by Sanduleak and Pesch (3) in the 1960s. Accurate absolute magnitude calibration, however, is a more problematic matter, as illustrated in figure 1.3.
Absolute magnitude changes very rapidly as a function of spectral type amongst M dwarfs, as illustrated in Figure 1.3. To compound matters, the spectral classification techniques used to calibrate the blue-green photographic objective prism spectra were generally fairly subjective - classification by eye, rather than by index. Figure 1.3 shows that even with modern spectral types, a given spectral class can span over 1.5 magnitudes at class M; and a small systematic error in spectral classification could lead to a large over- or underestimate of absolute magnitude. As Figure 1.2 shows, the blue-green TiO bands increase dramatically in strength from early- to mid-M, making classification increasingly difficult at those later types. As a result, many of the earlier surveys tended to overestimate spectral type, leading to underestimated luminosities and distances, and substantial overestimates of the local density of M dwarfs - and an apparent explanation of the source of `missing mass' in the Galactic disk (see Reid & Hawley, chapter 7 (4)).
Many (but not all) of these classification problems can be averted by observing M dwarfs at redder wavelengths, where the TiO bands saturate at later spectral types and indices can be devised to measure other molecular features, such as vanadium oxide. This approach to absolute magnitude estimation was used extensively by Reid, Hawley & Gizis (21, 22) in the PMSU survey of M dwarfs in the preliminary version of the third Neareby Star Catalogue (pCNS3). We are employing this type of calibration technique in our NStars program (see further below).
Figure 1.5: The (MV, (B-V)) colour-magnitude diagram: data for nearby stars with accurate parallax measurements (mainly Hipparcos or USNO). Crosses plot stars with photometry from Bessell (13), open triangles mark stars within 8 parsecs of the Sun, photometry from Legggett (14).
Photometric colours offer an advantage over spectal-typing as a temperature (and hence
luminosity) indicator in that photometric indices are continuous, rather than a discrete set of
values. However, not all colour indices are equally useful for all stars. (B-V), for
example, shares many of the same traits as spectral types, with a very steep dependence on
absolute magnitude amongst later-type M dwarfs. This is not surprising, since Morgan and
Keenan's (5) MK system is based on the appearance of the blue-green spectral regions. However,
this behaviour opens up the possibility for similarly severe consequences from small
systematic errors - and those possibilities were realised in the 1970s, where a small
overestimate in (B-V) colours led to a substantial overestimate of the local density of
late-type M dwarfs (Weistrop (6)). Again, M dwarfs seemed to be strong candidates for the
`missing mass' suspected to lurk in the Galactic disk, until more accurate photoelectric
photometry revealed the existence of a small (few hundredths of a magnitude) colour term
in Weistrop's photographic photometry (7, 8).
Figure 1.6: The (MV, (V-I)) colour-magnitude diagram: solid points plot data for L dwarfs (see further below).
Weistrop's (B, V) survey marked the first use of large-scale, quantitative photometry of 48-inch Schmidt plate material as a tool is searching for nearby M dwarfs. Previous studies, such as those undertaken by Haro and Luyten, had used multiple exposures on a single plate to search for extreme-coloured stars (different filters) or flare stars (same filter), while Luyten had undertaken extensive surveys for stars with significant proper motions (see our discussion of proper motion selection ). At the time of Weistrop's survey, large-scale surveys were limited to relatively blue wavelengths by the limited sensitivity of photographic emulsions at red and far-red wavelengths - the 103aE (`R') plates taken as part of the first Palomar survey (POSS I), centred at H-alpha, were the reddest available. However, with the development of first I-N, and then IVN emulsion, it became possible to extend surveys to the I-band, at 7500-9000 Angstroms. This provided both greater sensitivity and access to longer-baseline colour indices, with consequently more accurate absolute magnitude calibration. Reid & Gilmore were the first exploit these new possibilities, with surveys of several southern fields using plates from the UK Schmidt telescope (11, 12). The relatively small numbers of very late-type dwarfs discovered by those surveys laid to rest the possibility that M dwarfs could make a substantial contribution to any local dark matter.
(V-I) colours offer the advantage of a long baseline in wavelength, but also demand accurate photometry at visual wavelengths, where later-type M dwarfs become increasingly faint. This problem can be ameliorated to some extent by using the R passband (~6000-7000 A) in preference to V - the only option, in fact, for photographic detectors. Several wide-field surveys were undertaken in the late 1980s and early 1990s, notably by Hawkins & Bessell (15) and Tinney (16), using IIIaF plate material from the northern and southern 48-inch Schmidts, and by Kirkpatrick et al (17), using CCD data from the KPNO transit survey. As figure 7 shows, (R-I) colours become less useful as photometric parallax indicators at the latest spectral types. The effect is less pronounced in the (narrower) photographic R passband.
Figure 1.8: Pitfalls in photometric parallax derivation: a variety of calibrations of the (MV, (V-I)) relation. A mismatch in the adopted calibration leads to systematic errors in the derived absolute magnitude, distance and local density. Note the significant change in slope at (V-I)~2.9 masgnitudes and the reversal in (V-I) at the faintest luminosities.
Photometric parallax analyses are not without pitfalls, however. In particular, the accuracy of the distance derivation rests on the accuracy of the mean relation used to represent the main-sequence, and on the dispersion in absolute magnitude about that mean relation. An error in the calibration will produce a systematic error in the derived MV, and hence systematic errors in both distance and the local density as function of MV (the luminosity function); dispersion about the mean relation produces both random uncertainties in MV, and systematic errors in statistical quantities, since the larger sampling volume at r > rlim than r < rlim leads to more stars scattering into than out of a given sample.
These potential problems would not be of concern if the main sequence followed a simple, well-behaved mathematical relation in the colour-magnitude diagram. Unfortunately, such is not the case, and matching the observations is particularly difficult in certain regions where the main sequence exhibits rapid changes in slope. Figure 1.8 highlights one region, around (V-I)~2.9; there is a clear, and fairly abrupt, change in slope, also evident at other passbands (eg at MJ ~ 8 in Figure 1.12, below). Note the exceptionally small dispersion in the main-sequence below the "step", at (V-I) > 3.2 mag. These low-mass dwarfs are supported predominantly by degeneracy pressure and all have similar radii (~10% larger than Jupiter).
Figure 1.8 plots several attempts to reproduce the (MV, (V-I)) relation at these colours: most do a poor job of matching the observations (the Reid/Gilmore relation manages to track the dip at (V-I)=2.9 fairly well, but does a fairly woeful job at redder colours and fainter magnitudes). The net result is that stars are systematically assigned the wrong absolute magnitude and distance, tending to push too many stars into the MV~12, producing a sharp peak in the luminosity function (see (21)). These effects need to be given careful attention in photometric parallax analyses. The composite relation plotted in Figure 1.8 is
For more extensive data on the photometric properties of late-type dwarfs, see this page, which includes links to the datafiles used to make the above plots.
The most effective method of detecting most types of celestial object is to look where they are brightest. M dwarfs emit most of their energy at near-infrared wavelengths, a fact emphasised by the spectral energy distribution of Wolf 359 - at spectral type M6, a relatively warm M dwarf. Until the 1990s, however, infrared detectors were single-element devices, eminently capable of chopped observations of individual sources, but ill-suited to large-scale scans. Prior to the 1990s, Neugebauer and Leighton's (18) Mt. Wilson Two-Micron Sky Survey (TMSS) was the only wide-field survey at those wavelengths, covering the sky to a southern declination of -30o at a resolution of a few arcminutes and with a limiting magnitude of K=3.
Near-infrared sky surveys have only become feasible within the last half decade, with the development of sensitive array detectors and customised imaging cameras. Even so, with only 256-square (or at most 512-square) arrays, the field of view covered by a single snapshot is limited to a few square arcminutes, and limiting sensitivity must be traded off against areal coverage. Thus, DENIS (19) surveyed the southern skies (declination -88 to +2) in 3 years, but to limiting magnitudes of only I~17, J~15 and K~14, while 2MASS (20) surveyed the whole sky (with two slightly larger telescopes) in ~4 years, reaching J~16, H~15 and K~14.5 magnitudes. (Full details on the latter project are given here and href="http://www.ipac.caltech.edu/2mass/"> here. ) These modest limiting magnitudes still represent a gain of ~105 in sensitivity over the TMSS, so it is little surprise that both surveys have revealed a plethora of interesting new objects.
Amongst the new types of sources identified by 2MASS and DENIS, the most interesting for present purposes are the extremely cool low-mass stars and brown dwarfs. These clearly extend beyond the latest-type stars in the conventional M sequence; TiO bands become increasingly weaker and eventually disappear, while methane eventually replaces H2O as the most prominent feature in the 1 to 2.5 micron near infrared. Since TiO is the defining feature of class M, this behaviour has necessitated the introduction of two new spectral classes - class L and class T, where the T dwarfs are characterised by the presence of strong CH4 absorption at 2 microns. This range of behaviour is explained most simply as the effects of decreasing temperature: as the atmosphere cools, dust forms, depleting first the TiO, then the VO, gas phase abundance both directly (forming perovskite etc) and indirectly by binding oxygen in silicate grains. That removes the sources of opacity, weakening the TiO and VO bands, but leaving the metal hydrides relatively unaffected. At the same time, the increased transparency of the atmosphere leads to increasingly strong alkaline absorption lines (Na, K, Cs, Rb and Li) due to the increased column to the "photosphere"; the lines of the more abunadant elements, Na and K, become significantly pressure-broadened in the later-type L dwarfs and the T dwarfs. The presence of methane in T dwarfs indicates atmospheric temperatures well below 1400K, identifying them as brown dwarfs, not hydrogen-burning stars; lithium absorption in L dwarfs tags objects with masses of elss than 0.06 MSun. The overall behaviour is discussed in more detail elsewhere - see these pages - but these sources clearly extending the local census to dwarfs well below the hydrogen-burning limit.
The main goal of our present survey is to use the new resource provided by the 2MASS near-infrared data to identify late-type dwarfs within 20 parsecs of the Sun. Our photometric searches use two main strategies, both based on photometric parallax.
The second incremental 2MASS data release provides JHK photometry for 162,213,354 point sources within ~19,680 square degrees, or 47.7% of the sky. Those data, in isolation, are sufficient to identify the latest-type dwarfs in the Solar Neighbourhood. Figure 1.11 plots the (MJ, (J-K)) colour-magnitude diagram outlined by stars and brown dwarfs with well-determined trigonometric parallaxes. Main-sequence dwarfs with spectral types between ~K7 and M7 define a broad, nearly vertical sequence, centred on (J-K)~0.9; this stems from decreasing (J-H) (due to increased H- opacity) in concert with increasing (H-K) at these temperatures (see the (J-H)/(H-K) two-colour diagram). Later-type, ultracool dwarfs, however, have distinctly redder colours, defining a sequence extending to (J-K)~2. Thus, a simple technique for identifying nearby ultracool dwarfs is to determine a lower boundary to the latter sequence, shift by distance modulus of +1.5 magnitudes (corresponding to 20 parsecs), and select sources with J < Jlim. This technique starts to fail for the earliest T dwarfs, where the onset of methane absorption leads to significant decreased flux at H and K, and relatively blue colours. Kelle Cruz is using this method to sift through the published 2MASS scans. We have accumulated a target list of ~2500 candidates to date.
Figure 1.11 shows that (J-K) is a very poor photometric distance indicator for early- and mid-type M dwarfs. (H-K) offers an alternative, but the relatively small baseline leads to substantial dispersion about the mean relation, and, consequent, significant uncertainties in the estimated distance. Optical-to-infrared colours offer a better option, as illustrated in Figure 1.12, while also clearly segregating T dwarfs from all other stellar sources. Adam Burgasser has pioneered 2MASS-based T dwarf surveys (nn), and he is continuing to concentrate on those sources in this project. Others (Cruz, Reid, Liebert, Levine) are searching for nearby M dwarfs, combining the 2MASS data with optical scans of POSS II and UKST R (IIIaF) and I (IVN) Schmidt plates. Since dwarfs later than type M8 should be identified from (J-K) colours alone, this aspect of the survey can be limited to bright sources, with apparent magnitude J < 12.5. Such bright sources have 2MASS photometry accurate to 1-2%, so we can use the (J-H)/(H-K) two-colour diagram to isolate sources with dwarf-like colours. So far, we have over 10,000 candidates to sift through for this aspect of the project.
Further details on the various photometric projects are given on the following pages: