Radial velocities were measured using the wavelength shift of the line
centers of the OI triplet at 7771.94, 7774.17 and 7775.39 Å. These lines
were selected since, firstly, they are unblended; secondly, at the temperatures
of RR Lyraes, their strength increases as the effective temperature increases
thus the lines can be used at all phases of the pulsation cycle in both ``ab''
and ``c'' type stars; and, thirdly, because oxygen is over-abundant in the more
metal-poor RR Lyraes (Clementini et al. 1995;
Fernley & Barnes 1996), the
lines could be seen in the spectra of all the stars. Despite this there were
a number of spectra where, due to a combination of low metallicity in the
star and poor S/N in the spectra, we were forced to measure the radial
velocity from the wavelength shift of the line centre of H
at
6562.81 Å.
On each night we observed between two and four IAU radial velocity standards (see Table 1 for the complete list of standard stars).
Star (HR) | Radial velocity (km  ) |
33 | 14.7 |
| 1101 | 27.9 |
| 3145 | 71.0 |
| 4540 | 5.0 |
| 5694 | 53.5 |
| 7560 | 0.1 |
| 8969 | 5.3 |
|
Comparing our velocities, derived from both H and OI triplet,
with the published values showed differences that were always 
km with an
rms value,
over the four observing runs, of 1.1 km .
As mentioned in the previous section, our goal was to obtain three
well-phased spectra per RR Lyrae and then to derive the systemic velocity
by fitting these three measurements to the ``standard" RRab Lyrae radial
velocity curve given by Liu (1991). A free parameter in the Liu curve
is the amplitude. Liu found 22 RRab Lyraes with published radial velocity
curves and these have a mean amplitude of 61.5 km with an rms
scatter
of 8.4 km . Liu shows that there is a correlation between light curve
amplitude and velocity amplitude; however, since the light curve amplitude
is not known for many of the stars in our sample, we have used the mean
amplitude of 61.5 km to construct the ``standard" curve. Amongst the stars
we observed there are 8 RRab Lyraes and 2 RRc Lyraes with good quality radial
velocity curves available in the literature. Comparing our values of the
systemic velocity with the literature values for the 8 RRab Lyraes showed a
mean difference of 0.2 km and an rms difference of 3.7 km . For the
RRc Lyraes, where the velocity amplitude is smaller and the velocity curve
more symmetric, we determined the systemic velocity by simply taking a mean
of our three measurements. Again, comparing our systemic velocities with the
literature values showed, for the two RRc Lyraes, a mean difference of
-0.3 km and an rms difference of 2.3 km .
Thus for those stars where we have three well-phased radial
velocities, measured from the OI triplet, a realistic 1 
error in
the systemic velocity is typically 3 km . For those stars with only two
measurements (the minimum number for any of the stars) or those stars where
some of the measurements used H, a higher error is appropriate. For
the case where H was used in the measurement this higher error arises
from several sources, principally the lower S/N of the spectra which reduces
the accuracy of the measurement. In addition, there are problems due to the
presence of emission in H at certain phases of the pulsation (e.g.
Preston & Paczynski 1964;
Gillet & Crowe 1988), which distorts the
measurement of the line centre and also the larger radial velocity amplitude
obtained from H (Oke et al. 1962), which degrades the fitting to
the ``standard" curve. In these cases we assume an error of 
km .
The list of measured RR Lyraes and their systemic velocities is given in the Appendix (available electronically).
Because the Horizontal Branch is a relatively short-lived phase of stellar evolution most companion stars will be much fainter than the RR Lyrae itself, i.e. the companions will either be low-mass main sequence stars or white dwarfs. In general therefore, the only methods available to detect companions are to look for variations in either the time of maximum light or the systemic velocity. In Table 2 we list those RR Lyraes which show significant differences between our values and previous values of the systemic velocity. Examination of the spectra of these stars showed no evidence of line doubling for any of them.
CI And
Systemic velocity km
RR Lyrae This work Previous work Ref.
(see Appendix) 
11
DM Cyg 
15
11
BK Dra 
11
XX Hya 
18
11
ST Leo 
6
16
15
11
CN Lyr 
11
TU UMa 
1
10
13
14
The only one of these stars previously suspected to be a member of a binary is TU UMa. Saha & White (1990), using published times of maximum light, calculated the orbital parameters for TU UMa and our value of the systemic velocity is not inconsistent with their prediction. For the other stars, a literature search showed there were insufficient published times of maximum light to attempt an orbital solution and clearly it would be of value to place the stars in a long-term observing programme in order to acquire these data.