Although uniform surveys are the most useful from the point of view of studying stellar populations, our main concern in this work was to find a series of reliable kinematic probes to investigate the complex dynamical interaction history of the Clouds; determine the mass and velocity structure of the LMC; and to investigate tidal disturbances around the outer halo of the Clouds - subjects for future papers. The magnitude and colour selection limits imposed have produced a sample of cool AGB carbon stars which are ideal probes to investigate the kinematics and spatial extent of the dominant intermediate age Cloud populations. Since all candidates in the outer halos were observed, the spatial extent of cool AGB Cloud carbon stars is well defined (cf. Fig. 1 (click here)). Toward the central parts of the LMC and SMC there were too many candidates to follow up all of them. In these cases we selected subsets simply to give a dense enough sample to answer the dynamical questions above.
The complex interaction history of the Magellanic Clouds-Galaxy system poses interesting challenges to kinematic surveys. Relic signatures of previous interactions are expected to produce subtle spatially systematic and random velocity feature, at levels that require velocity probes with precision around 5 km/s. All radial velocity surveys are to some extent plagued by the same problems that contrive to make this target difficult to attain. For example: random errors from poor signal:to:noise spectra give unreliable individual velocity measures; atmospheric disturbances cause short timescale velocity variability; binary stars cause much longer timescale variability; different reference templates give different systematic offsets; alternative reduction methods often yield different results; spectra taken at different epochs exhibit peculiar systematics; and so on. Our survey is no exception and we have investigated these concerns in several ways described in Sect. 2 and summarised below.
Random errors are relatively easy to deal with and can be characterised by sufficient repeat measurements, as in Table 19 (click here). Velocity variability turned out to be more of problem than we had anticipated, with 10% and possibly as high as 20% of the sample showing evidence for both atmospheric variation and longer timescale variations indicative of of a substantial binary fraction. Long term monitoring will be necessary to address this issue further. The variable fraction is not dissimilar to the results found in many studies of velocity dispersion of giant branch members of Galactic dwarf spheroidal systems (e.g. Hargreaves et al. 1994 and references therein).
Table 19: Velocity repeatibility for LMC/SMC periphery
carbon stars
Systematic effects are much harder to quantify and we expended a considerable amount of effort to both minimise and characterise them. In the end we opted to use a single reference template for all reductions. Primarily this was possible because most of the carbon stars are of similar spectral type due to the colour selection imposed. They have compatible absorption line features (particuarly the CN bands) and the spectral rectification, prior to cross correlation, removes any "continuum'' variations. Consequently a single template is adquate to determine velocities to 5 km/s for all the carbon stars analysed. Although only one RV template was used the internal systematic errors were minimised to the level of 2 km/s through repeat observations and pairwise matching of 75 stars. A comparison with Hardy's radial velocities showed an external systematic offset of 6 km/s, consistent with our estimates of the velocity errors.
Most of the weighting in the cross-correlation template comes from the CN
bands which are strongly present in all the carbon star types in our sample
apart from the Wk C/ M stars. This mitigates against template-induced
systmatic effects. Tests involving splitting the template into different
wavelength portions, e.g. using only the CaII triplet - clearly present in
most objects - and comparing derived velocties with respect to CN-only
templates, leads to systematic differences of order 
. It is
not clear how much of these small systematic shifts are due to the template,
to the reduction algorithm, or to subtle atmospheric disturbances in the stars.
However, since the measured systematic changes are well
below 5 km/s and the random errors from repeatibility measurements are
of order 5 km/s we feel justified in describing the average accuracy of our
measurements as 5 km/s. The question of systematic velocity differences
between varying carbon star type motivated the inclusion in our
spectroscopy of the large number of WORC, Blanco et al. (1980),
and Hardy et al. (1989) stars.
We have carried out many further tests (colour cuts, line strength cuts etc.)
to look for systematic differences between perceived velocity structure for
the different carbon star samples/types. Although these tests are necessarily
of a coarser nature and might involve real kinematic population differences
in no cases have we seen any evidence for significant trends.
A search of the literature to confirm the radial velocities of the
globular clusters (Table 18) has been nearly fruitless. There are
available quite a lot of data on the LMC clusters but very few radial
velocities have been published for the neglected SMC clusters. We found
three radial velocities for NGC 121: two observed by Hesser et al.
(1986) and one by Zinn & West (1984) and a fourth one
quoted by Cohen (1983). The weighted mean
of these values is 
. This velocity is
compatible with our velocity but Suntzeff et al. (1986) have
published the radial velocity of one star presumed to be in NGC 121, its
velocity of 
is far from our estimate.
Zinn & West (1984) have also published
the radial velocity of NGC 419, their value 
agrees well with
our two radial velocities.
Morgan & Hatzidimitriou (1995) have recently published the
results of a survey for carbon stars in and around the SMC, selected using
visual inspection of UKST IIIaJ objective prism plates. Several of their
fields overlap our survey region and hence provide an external check on our
carbon star selection criteria, subject to a few caveats. For the area in
common to both surveys 39 of our stars were also
identified by Morgan and Hatzidimitriou. Half a dozen were not. Most of these
ones are wkC stars which may be late-type M giants as we explained in 2.2.
This is one of several likely reasons for these differences but the main
two are:
firstly our colour boundary is specific to finding cool carbon stars near
the AGB tip with a 
colour > 2.4. The reddest carbon stars
in our sample are fainter than 19th magnitude on a IIIaJ objective prism
plate and if visible at all, would appear as no more than a short stubby red
object indistinguishable from late-type M stars. Secondly, our colour
criteria will not in general be sensitive to CH-type Carbon stars, which
are themselves readily detectable in IIIaJ prism surveys due to the strong
bands blueward of the emulsion cutoff. We are therefore not
surprised that the prism survey finds objects not in our sample and vice
versa, that some of our redder and fainter carbon stars are not in the
prism sample. Thirty nine stars are not enough to make a good comparison,
when our survey will cover a larger area around the SMC a better comparison
will be possible.