Measuring the Scale Height of L dwarfs from the UKIDSS LAS
125 DR1 L dwarf candidates detected in
SLOAN DR5.
Selection: z'-J>2.0, 0.9<(Y-J)<1.7, 0.5<(J-H)<1.5, 0.5<(H-K)<1.3, MergedClass=-1 (i.e.
a stellar profile). These are z'YJHK detections with the Sloan counterpart located
within 1.2 arcsec of the UKIDSS source. z' detection is defined by psfMagErr_z<0.3 mag.
Note that the z'-J limit is more generous than that of Nicolas Lodieu's search (which
used z'-J>2.4 but only included relatively bright sources with errors <0.1 mag). My limit
may allow some M8 and M9 objects to creep in.
This list will exclude some high proper motion objects. I have used psfMags for Sloan rather
than the default modelMags, since some faint Sloan detections are misidentified as galaxies
and would give erroneous results with modelMags. psfMags and modelMags are the same for "stars"
but different for "galaxies".
425 DR1 L dwarf candidates with no
Sloan DR5 match within 1.2 arcsec.
Selection: These are YJHK detections
which are mostly faint enough that Sloan would not pick them up (see histograms below).
I think all the UKIDSS DR1 data are in the Sloan DR5 footprint but am not certain.
The same criteria as above were used in defining the Y-J, J-H, H-K box.
These lists have not yet been culled of false detections by using cut-out images.
The first list should be reliable. Based on EDR experience it is likely that 10%
of objects in the second list are spurious, eg. binaries and cross talk artifacts.
Background to the scale height problem
Measuring the Galactic scale height of brown dwarfs is one of the original UKIDSS science
goals. It is important to know the answer in order to (i) establish the mass function;
and (ii) determine the spectral type beyond which no more hydrogen burning stars are seen.
We know that all T dwarfs are brown dwarfs and almost all M dwarfs are stars
(except at very young ages). It is thought that old very low mass stars can have early
L type spectra, but probably not later than about L5. Hence early L dwarfs should be
a mixture of old stars and young brown dwarfs that are slowly cooling to later types.
The scale height of Galactic stellar populations is believed to be a function of age.
Star formation is concentrated within a scale height of ~50 pc of the mid-plane.
Random scattering events disperse the population to a scale height of ~400 pc after
a few Gyr. This has been measured for spectral types O-M, eg.Hawkins(1988);
Rana(1987); Bessel & Stringfellow(1993); Pirzkal et al.(2005); Ryan et al.(2005).
A scale height of ~90 pc is found for types O-early F, rising to ~400 pc for K and M
types, which are older. We would therefore expect the measured scale height to decline
from early L-mid-L, as the population becomes dominated by young brown dwarfs rather
than stars. For later types the scale height should slowly rise, as the average age
increases. Detection of this rising trend would help break the age-mass degeneracy
for field brown dwarfs.
The LAS H band limit of H=18.8 corresponds to d=350 pc for L1 types, 260 pc
for L3-L5 types, 115 pc for L7-T2 types and 46 pc for T6-T8 types. Hence UKIDSS
can only address the issue for L and early T dwarfs, not mid- or late T dwarfs.
Observational Approach
In Feb 2006 we put in a SPITZER proposal to determine approximate spectral
types of a sample of 40 L dwarf candidates (types M8-T2) from the EDR
and hence determine photometric distances. This would allow us to solve
this inverse problem by Chi-squared fitting of various model L dwarf
scale height distributions to the data. With hindsight we should probably
have included some Sloan drop outs, since Sloan detections will tend to be
nearby (see J mag histograms below).
The work of Patten et al.(2006) shows that SPITZER/IRAC can easily measure
approximate spectral types via the J-3.6 and K-4.5 colours. The proposal
was rejected. An alternative approach may be to determine approximate spectral
types using optical and infrared colours, eg. i'-z', z'-J and i'-J.
This requires reasonably good photometry of sources with
i'=24.9 (Vega system) or i'=25.3 (AB system) to measure the i'-z' colours of
L9/T0 dwarfs with J=19.0 and i'=24.4-24.9 (Vega system), or
equivalently Cousins I=24.0-24.5
Knapp et al.(2004) showed that i'-z' is correlates well with type from L0
through early T types, but there is a scatter of up to 2.5 sub-types either
side of the mean. z'-J shows more scatter and is less useful.
Dobbie et al.(2002, MNRAS 331, 445) investigated the value of several
I and Z filters for spectral typing, using spectrophotometric modelling.
They found less scatter than Knapp et al.04 - I'm not sure why.
Dobbie et al. found that the I_harris longpass filter on the INT/WFC is
not very useful, since the weakening of TiO bands offsets the reddening
of the underlying continuum at early L types.
Potentially useful options are Sloan i' (770 nm) and Cousins Ic (795 nm).
Unfortunately the INT/WFC appears to lack the more sensitive Ic filter.
In any case the required exposure times simply appear to
be rather long for either INT/WFC or the ESO 2.2-m/WFI, given that a large
sample is needed to solve the problem.
A possible option appears to be a 3 night run on the ESO NTT/SUSI2 high
resolution imager, which has a filter equivalent to Ic. This telescope
can reach Ic=24.1 at 7-sigma in 1270s (according to the ESO ETC, assuming
dark time and 0.8 arcsec seeing). This is equivalent to 10-sigma with an
optimised aperture I believe. The Z filter is similar to z' and the
INT Z(RGO), and would need similar exposure time I think (the ETC does not help
much for Z). Each target would take about 45 minutes to observe at IZ.
In a 3 night run we could observe up to 36 L dwarfs, which would be adequate
for a first stab at the problem.
Is this approach worth trying? The uncertainty in derived spectral types
would be significant, leading to quite a bit of error in distance estimates.
Near IR colours would help a bit to distinguish types though, and the analysis
might be restricted to objects close to the expected colour sequence in Ic-z,
Ic-J, J-H and H-K. Ideally we should do a a Monte Carlo model of the observations....
Histograms of brightness at J band:
Histogram for Sloan z' detections.
Histogram for Sloan z' drop outs.
Author: Phil Lucas
This page last updated on 18th September 2006.