2. Observations and data reduction
2.1. UV spectroscopy
The UV spectra of ROA 5701 and Barnard 29 were obtained with the Goddard High Resolution Spectrograph onboard the Hubble Space Telescope, equipped with the G200M grating (1860 - 1906 Å, 0.07 Å resolution) and using the large science aperture. This spectral region was chosen because the strongest Fe III absorption lines are expected there as judged from the high resolution IUE spectra of (Napiwotzki et al., 1994), which shows an optical spectrum similar to ROA 5701 and Barnard 29 . The exposure times were 4711 s for ROA 5701 (observed on August 3rd, 1996) and 4570 s for Barnard 29 (observed on November 30th, 1996). We did not use the FP split option as we did not expect any lines strong enough to allow a correct alignment of the individual spectra by correlation.
After the standard pipeline reduction we co-added the flux of the individual spectra, which were on identical wavelength scales. The resulting spectra were converted to MIDAS bdf-format and interpolated to a step size of 0.02 Å. They were then corrected for Doppler shifts using the heliocentric radial velocities of the clusters (+232 km/sec, Cen; -246 km/sec, M 13) and the corrections for heliocentric velocities appropriate for the observation dates. To allow a better definition of the continuum we smoothed the spectra with a 3 pixel wide box average filter. The continuum was then defined by eye and we estimate that the error of the normalization lies between 5% and 10%.
Fig. 1 shows a section of the GHRS spectra of ROA 5701 and Barnard 29 compared to the IUE data of 1, a field post-AGB star with and similar to ROA 5701 and Barnard 29 . The abundance noted for in Fig. 1 was derived from equivalent width measurements (Napiwotzki et al., 1994) for a microturbulent velocity of 10 km/s (which is the value Conlon et al., 1994, derived for Barnard 29 .) The iron abundance of is comparable to that of M 13 and Cen. The much weaker iron lines in the GHRS spectra therefore suggest that Barnard 29 and ROA 5701 show significant iron depletions.
2.2. Optical high resolution spectra
ROA 5701 was observed with the ESO CAssegrain echelle SPECtrograph (CASPEC) at the 3.6m telescope at La Silla, Chile, on May 24, 1988. Two spectra of 1 hour integration time each were obtained. The spectra were binned during read out in order to improve the S/N ratio. Reduction of the data proceeded in two steps: first the ESO-MIDAS software (Ponz & Brinks, 1986) in Garching was used for wavelength calibration and extraction of the echelle orders. The background correction and flat fielding were done separately using a computer program written by G. Jonas (Kiel, see Heber et al., 1988). We then merged the orders of the CASPEC spectra and rebinned them to a common wavelength scale. The spectra have a resolution of 0.3 Å .
2.3. Equivalent widths
We always used the normalized spectra to measure equivalent widths. The measurement in the optical spectra was straightforward since the lines are isolated and well defined and the spectra have a good S/N. In the UV, however, the lines are more crowded and the S/N is lower. Therefore, three different methods were used: i) direct integration without any fit of the line shape using a global continuum (assuming that the overall continuum definition is more reliable than a local one due to the low S/N), ii) same as i), but for a local continuum, and iii) fitting Gaussians to the absorption line profiles (using a locally defined continuum). Method iii) could not be used for ROA 5701 because the lines were too weak.
For the GHRS data the equivalent widths measured for a global continuum (which were used for the abundance determinations) were on average larger than those measured for a locally defined continuum. This offset leads to a difference in the mean iron abundance of 0.08 dex for ROA 5701 and 0.03 dex for Barnard 29 .
For the CASPEC data the equivalent widths measured for a global continuum resulted in abundances larger than those determined for a local continuum by 0.12 dex for O, 0.02 dex for N, and 0.14 dex for Si. As the CASPEC spectra showed small-scale continuum variations that are difficult to correct for by a global continuum fit we decided to keep the abundances derived from the "local continuum" equivalent widths.
© European Southern Observatory (ESO) 1998
Online publication: October 21, 1998