The pattern of fields in Fig. 1 resulted from maximising area coverage, at the same time minimising overlap (i.e. by choosing alternate fields) with the additional requirement that a short exposure V plate existed. In all, 19 fields make up this survey and Table 1 gives details of the photographic material.
|Figure 1: The 19 Schmidt fields in the galactic centre survey. Galactic longitude is with respect to the galactic centre; field numbers are standard UKSTU field designations described in [36, Tritton (1983)]|
1 Standard United Kingdom Schmidt Telescope field numbers are defined in [36, Tritton (1983);] see also the UKSTU online catalogue at http://www.roe.ac.uk/
2 Indicates if photographic material has been scanned on SuperCOSMOS, processed and blue objects selected.
3Indicates if spectra have been obtained, and if so, in which wavelength regions; b - 3860-4640 Å, r - 5400-6900 Å.
The photographs were scanned using the precision microdensitometer SuperCOSMOS (see, for example, [9, Hambly et al. 1998).] To date, plate pairs for the 14 fields nearest the galactic centre have been measured (see Table 1). SuperCOSMOS digitises each plate at m (0.67)resolution, 15-bit grey levels in 2.5 hours; offline software then processes the pixel map into a parameterised image list for all images detected above a preset threshold. Every image has 32 parameters including celestial co-ordinates, shape and instrumental magnitude. Deblending software [1, (Beard et al. 1990)] rethresholds the pixel data during processing to detect and unscramble merged objects which is vital at these low galactic latitudes. Global astrometry over each field is achieved via Tycho-ACT catalogue standards [37, (Urban et al. 1998).] The accuracy of the resulting celestial co-ordinates is typically 0.25.
Plate pairs were reduced as follows. The individual image catalogues were paired using a search radius of 3. Poor quality images, parent images of deblends and elliptical images were all excluded from further analysis. An approximate calibration of the V-band instrumental magnitudes was obtained by fitting instrumental magnitudes versus V photometry from the Tycho stars in each field. In the case of field 456, an I-band photometric sequence was compiled using photometry presented in [2, Blanco & Blanco (1984);] [40, Walker & Mack (1986)] and [35, Szymanski et al. (1996).] No independent data were available for calibration of the U-band photographic instrumental magnitudes; hence in order to make the blue object selection in each field, we used the raw u-v (or u-i in the case of field 456) colour index as an indicator of spectral type. In general, the instrumental magnitudes measured from photographic plates are non-linear with respect to true apparent magnitudes due to difficulties in intensity calibration and the limited dynamic range during measurement. This results in raw instrumental colours (u-v, u-i) having, in general, zero-point offsets, which are functions of magnitude; in addition, photometric errors increase with object faintness. These in turn result in colour-magnitude diagrams in which the locus of the modal colour of the dominant F/G-type dwarfs is not vertical (i.e. not at constant colour) and in which the scatter about that locus increases with magnitude. Any selection of blue objects must take these into account, cf. the blue object selection procedure detailed by [34, Stobie et al. (1997).] Here, we adjusted the raw colours as a function of magnitude so that the locus of the modal stellar colour minus the width of the distribution (estimated on it's blue side) was at zero colour. This then allowed us to select the 80 bluest stars in each field in the range 11.5<V<16 regardless of magnitude and in a consistent manner from field to field (the number of targets was chosen to be compatible with the FLAIR spectrograph - see Sect. 2.2). A typical pseudo colour-magnitude scatter plot is shown in Fig. 2 for field 393. However, note that these adjusted colour indices (also listed in Tables 2 to 5) are not true colours -- they merely reflect relative colour between stars of similar magnitude.
|Figure 2: A pseudo colour-magnitude scatter plot for field 393 (see text). The 100 bluest candidates are circled, while only 1000 other stars are plotted to illustrate the general scatter in this diagram|
The relative accuracy of these colours is estimated to be at best ; however position-dependent systematics of several tenths of a magnitude or more are doubtless present. Moreover, crowding increases the scatter in the photometry, and there will inevitably be some cool stellar contamination in these selections. Such erroneously blue objects are of course weeded out via low-resolution spectroscopy, as discussed below. Note also that because the fields are very crowded and because the deblending algorithm is limited by the seeing on the plates (typically 2) many overlapping images will be discarded and completeness is likely to be significantly compromised -- for example, [1, Beard et al. (1990)] estimate completeness of 70% at . However, for our ultimate goal of identifying early-type stellar candidates in the region of the galactic centre for abundance studies, completeness is not a major consideration. Table 1 lists the UKST fields for which we have obtained U-material and indicates which fields have already been scanned. The photometric data are available from the authors on request for the 3 fields which are presented here.
For the blue candidates, identified from the photographic photometry, low resolution spectroscopy was undertaken using the FLAIR instrument on the UK Schmidt Telescope at Siding Springs. Eight nights were allocated in June 1997 but due to weather conditions observations were only possible on three nights; consequently only three of the 14 fields measured and processed with SuperCOSMOS were observed. The FLAIR system allows the simultaneous observation of up to 90 objects (from which target and sky fibres need to be allocated) with the m diameter fibres (or up to 150 objects using the m fibres) subject to limitations imposed by the finite size of the fibre ferrules and prisms. For the observations discussed here, the m fibre bundle was used and between 30 and 50 fibres were positioned on stellar targets with typically 10 sky positions. The prioritisation of targets was primarily based on their colours, with the very blue targets only being excluded if proximity to a bluer star made this necessary. Spectra, covering the wavelength region from approximately 3860 to 4640 Å, were obtained using the 1200B grating for all three fields at a spectral resolution (full-width-half-maximum) of approximately 2.7 Å. For Field 456, further red spectroscopy covering approximately 5400 to 6900 Å at a resolution of approximately 5.9 Å was obtained using the 600R grating. The FLAIR CCD is an EEV thinned and back-illuminated device, with pixels in a array (with the long axis along the spectral direction). Further details of the FLAIR instrument can be found in [19, Parker & Watson (1994).]
Stellar exposures were split into 1800 s segments, in order to
reduce the impact of cosmic-ray events. Tracking of the Schmidt
was typically good across approximately 4 hr either side of the
meridian, but weather prevented the full exposure time
being utilised. A total of s exposures were
obtained for Fields 393, 454 and 456 in the blue region, plus
for the last field s in the red. The seeing was
generally poor, varying from 3 upwards on the three nights.
The target exposures were bracketed with arc frames of
Hg+Cd+Rb (for the blue region) and Hg+Cd+Ne (for the red region), and
dome flat fields where taken for each setting (twilight flats, while
preferable were not possible due to weather constraints).
Preliminary data reduction was undertaken using the AAO's
FLAIR package within the IRAF environment
[6, (Drinkwater & Holman 1996).] As no structure was found in the bias frames,
bias subtraction was achieved using the median integer
bias value for the on-sky and the
flat field frames. The target data frames were combined to remove
cosmic-ray events using suitable clipping algorithms, in consecutive
groups of 2-3 frames to ensure the signal levels in each were similar.
The difference between frames combined with and without sigma-clipping
showed the expected random patterns, indicating
that rejection of real signal had not occurred.
An aperture identification file, created when the
fibres where placed on each object, was used to assign
slit positions to either an object or sky.
The IRAF package
[38, (Valdes 1992)] was used to
extract the 1D spectra from each combined image, correct for the
fibre-to-fibre variation in throughput (from the flat-field response frames),
sky subtract, and then wavelength calibrate the final product.
No traditional "flat-fielding'' which removes the pixel-pixel CCD variations
was performed, in line with normal fibre reduction procedures.
Only five arc lines were visible in the blue region of the spectrum,
but the pixel-wavelength relation is quite linear in the FLAIR
spectrograph, and produced a RMS of 0.06 Å.
In the red, the rich neon spectrum (26 lines) produced
a RMS of 0.07 Å.
The spectra from the blue stellar candidates are summarised in Tables 2 to 5 for the three fields observed. In the first are listed co-ordinates, magnitudes and colours for those targets for which no fibres were placed (due to crowding or to a relatively low priority based on colour). The last three tables list for each field the corresponding information for stars which were observed spectroscopically. A complete library of the spectral data is available on request from the authors.
Copyright The European Southern Observatory (ESO)