FAUST observations of ultraviolet sources in the directions of NGC 4038–39 and 6752

Analysis of ultraviolet (UV) observations with the FAUST shuttle-borne telescope toward the Antennae and NGC 6752 celestial regions resulted in the detection of 46 and 221 candidate sources respectively, for a signal-to-noise ratio of 8. We discuss the source detection process and the identification of UV sources with optical counterparts. Using correlations with existing catalogues, we present reliable identifications for approximately 60 per cent of the sources. We find that most identified objects are B, A and F stars. The remaining identified objects are galaxies, a white dwarf in a binary system, and two K-type stars. Nearly all of the remaining unidentified objects have assigned optical counterparts but, lacking additional information, we give these only as best estimates. With help from new diagnostic diagrams, we suggest that these unclassified objects are main-sequence (or giant) stars within the local spiral arm or halo; or other hot evolved objects within the local spiral arm. We discuss the nature of the objects found and compare our results with those predicted from spectral and Galactic models. shortage in the Antennae field. that normal early A stars, and shown that in most instances it is valid to infer the UV brightness of stars given the B and V colours only.


I N T R O D U C T I O N
The ultraviolet (UV) region of the electromagnetic spectrum is an excellent tool for identifying and quantifying active star formation (SF) regions (Donas et al. 1987). It is also a sensitive probe of evolved hot stellar populations (Burstein et al. 1988;Dorman & O'Connell 1996). Furthermore, UV observations of galaxies are important for the following reasons. First, this wavelength range is shifted into the visible for high-redshift objects, and UV observations are expected to provide templates for evaluating evolution effects. Secondly, the UV contrasts the hot stellar population of a galaxy against the contribution from cooler stars, and should provide new insight on galaxy stellar content. Previous UV observations of galaxies and their current interpretations have been reviewed by O'Connell (1990O'Connell ( , 1992O'Connell ( , 1999.
In this paper we study the population of Galactic UV stars and extragalactic sources detected in UV images of the sky in the directions of the Antennae or NGC 4038-39 l 2868X97Y b 428X46 and the globular cluster NGC 6752 l 3368X50Y b 2258X63X The study of UV sources in this part of the Milky Way may help the understanding of the morphology of the disc and halo regions of our Galaxy.
There is a general lack of information about the nature of UV sources fainter than the completeness limit of the TD-1 all-sky survey conducted during 1972 to 1974. A description of the UV sky survey telescope on TD-1 was presented by Boksenberg et al. (1973). For point sources, the S2/68 experiment on TD-1 provided <300-A Ê broad-band flux measurements at 2740 and <330-A Ê , wide-band photometry at 1565, 1965 and 2365 A Ê . The photometric accuracy over the short-wave band was claimed to be <10 per cent (Jamar et al. 1976).
The results of the TD-1 survey were published in the form of a catalogue (Thompson et al. 1978) that contains approximately 31 000 stars detected with SaN $ 10Y corresponding to a flux density limit of 10 212 erg s 21 cm 22 A 21 or about m 1600 < 9X We use here monochromatic magnitudes defined here as m l 22X5 Â logf l 2 21X75Y where f l is in cgs units. An extended version of the TD-1 catalogue contains 58012 stars (Landsman 1984). In the 1565-A Ê band, the full TD-1 catalogue (henceforth TD1F) is complete to <8.5 mag. Gondhalekar (1990) discussed the TD1F catalogue, and mentions that the experiment is not linear for sources fainter than 10 212 erg s 21 cm 22 A 21 . Thus, although the TD1F contains 58 012 stars, only 47 039 of them are brighter than the linearity limit in at least one band.
The TD-1 catalogue shows only 19 objects 1 for the Antennae field and 50 objects 1 for NGC 6752 field. Since the TD-1 mission there has been no further complete surveys of the UV sky. In particular, for the two fields studied here, only limited regions of the sky have been observed with space platforms such as HST and UIT.
One of the more extended partial surveys was by the FAUST experiment (see below), which operated on board the Space Shuttle (STS) Atlantis in 1992 March and April as part of the ATLAS-1 mission. During this flight, 19 pointed exposures were obtained, of which three were short exposures for pointing checks. The other fixed-pointing images were exposed for 12 to 18 min. In addition, two approximately 308-long scans of the sky were obtained by rolling the STS during the exposure. A catalogue of 4660 FAUST sources (FSC) was published by Bowyer et al. (1995). This was produced using a uniform thresholding algorithm for source detection, and the identifications were obtained from nearest-object correlations against catalogues mostly from SIM-BAD [Set of Identifications, Measurements and Bibliography for Astronomical Data created and maintained by the Centre de Donnee Âs Astronomiques de Strasbourg (CDS), Strasbourg].
The FAUST images are being examined systematically at Tel Aviv University to identify sources using astrophysical criteria and perform, where possible, ground-based follow-up studies. In this context, we have already presented results from eight sky regions: the North Galactic Pole, Brosch et al. (1995); the Virgo cluster region, Brosch et al. (1997); the direction of the Coma cluster, Brosch et al. (1998); four regions of the Fourth Galactic Quadrant, Brosch et al. (2000a); and toward a second area near the North Galactic Pole, Brosch et al. (2000b).
The Antennae and NGC 6752 fields were also observed by the FAUST telescope. The Antennae region is well suited for studies of the UV stellar population and UV-bright galaxies, because of its low extinction. Measurements of H i from Dickey & Lockman (1990) using a rectangular region 88.4 long, covering the FAUST field and centred at l 2868X97 and b 428X46Y give column densities of 3X3±8X0 (mean 4X3 Â 10 20 cm 22 X With a standard extinction law and dust-to-gas ratio, this would imply a Galactic extinction at 1660 A Ê of 0X2±0X5 mag (mean 0.3).
For a similar reason, NGC 6752 is also well suited for studies of the UV stellar population and UV-bright galaxies. Measurements of H i (Dickey & Lockman 1990) using a rectangular region 88.2 long, covering the FAUST field and centred at l 3368X5 and b 2258X63Y give H i column densities of 7X3±11X5 (mean 8X9 Â 10 20 cm 22 X With a standard extinction law and dust-to-gas ratio, this would imply a Galactic extinction at 1660 A Ê of 0X4±0X7 mag (mean 0.5). Note that N(H i) values from Dickey & Lockman were derived assuming optically thin emission; also, the N(H i) given should be considered a lower limit when the N(H i) is greater than several times 10 20 .
In this paper we describe FAUST imaging, source detection, source astrometry, source photometry, identification of the UV sources with catalogued objects, and observations from the Wise Observatory. We then discuss the stellar content of the fields, comparing observed differential star counts and colours with those predicted from the Bahcall±Soneira model of the Galaxy (Bahcall & Soneira 1980) and modifications therein (Brosch 1991;Bilenko, private communication). We examine FAUST colour±magnitude diagrams to suggest morphological types for unclassified objects. We compare colour±colour plots with those predicted (Shemi et al. 1994, and modifications therein). The colours of metallic-line stars in our fields are investigated, and individual sources of note are discussed. Finally, we discuss the extragalactic content of both fields.

FAU S T O B S E RVAT I O N S T OWA R D S T H E A N T E N N A E A N D N G C 6 7 5 2
FAUST is a wide-field (78.6 field of view) all-reflecting twomirror Wynne camera with a bandpass between 1400 and 1800 A Ê and an angular resolution of 3.5 arcmin. It utilizes a microchannel plate detector with wedge-and-strip anode, which records the position of each detected photon. A description of FAUST and its operation aboard ATLAS-1 on the Space Shuttle was given by Lampton et al. (1993). Details of the image construction and subsequent reductions are given in Bowyer et al. (1993), these include: removing the effects of shuttle drift during the exposure; correcting for airglow, aurora and South Atlantic Anomaly effects; removing optical and detector distortions; and correcting for detector quantum efficiency variations.
The Antennae region was imaged by a single FAUST exposure pointed at 12:01, 218:18. The total exposure time was 863.4 s, and the image covers 67.5 deg 2 . Because of spacecraft motion, the sky covered by FAUST during pointed observations is slightly larger than the instrumental field of view, and is typically 108 in diameter. The NGC 6752 region was also imaged by a single FAUST exposure pointed at 19:02, 260:15. The total exposure time was 1122.1 s, and the image covers 64.0 deg 2 . Shuttle drift results in a non-uniform exposure across both images.
The sky area covered by each image was calculated from the total number of non-zero exposure pixels and by assuming that each pixel is a square of 1.1-arcmin sides. This is an approximation, because electrostatic distortions of the image caused by the detector change slightly the area imaged by each pixel.
Count images of both regions in counts per detector pixel were divided by the corresponding exposure images to produce counts per pixel per second.

Source detection
An automatic detection technique was developed to provide impartial source detection. A typical stellar point spread function of FAUST extends for more than a single pixel, and so a square box three pixels wide, referred to as a source-box, was translated across the FAUST image in single pixel steps. The median of all nine pixels was adopted as the typical source-box count rate. This was compared with the typical neighbourhood background count rate evaluated as follows.
The mean and standard deviation of each of eight square boxes, six pixels wide distributed symmetrically around the source box, were calculated. Sky boxes with any pixels outside the image field were rejected. The median of all eight box means and the median of all box standard deviations were adopted as the most significant background level and typical deviation from this level. Sky boxes for which the mean count rate deviated by more than two typical 1 Excluding a few possible objects within 25 arcmin of each of the FAUST image field boundaries. deviation intervals from the background were rejected. All relevant pixels within the remaining sky boxes were considered at this stage, and their mean and standard deviation were calculated to represent the neighbourhood count rate and the neighbourhood standard deviation s b . As relevant pixels we considered only those with non-zero count rate values. A source was detected if the median pixel value from the source box was $ SaN Â 9 p Â s b Y where SaN is the effective signal-to-noise ratio, adopted here as SaN 8X The size of the sky box, the deviation size for rejection of background boxes and the signal-to-noise level were all set to ensure the maximum number of source detections whilst avoiding spurious noise detections. Such a routine is designed for the detection of faint sources, and therefore when encountering a bright object, it initially detects it at the object periphery and subsequently at its centre. Further processing retains a single entry of these multiple detections. The algorithm identified 46 candidate sources for the Antennae field and 221 candidate sources for the NGC 6752 field: these are indicated in Figs 1 and 2.

Source astrometry
More than 20 bright objects in each field were first identified in predicted UV images, which were created with data from the Hipparcos Input and SAO catalogue stars, together with the algorithm for predicting the UV magnitude using the visual magnitude and spectral type of stars (Shemi et al. 1994, and modifications therein). The pixel coordinates (at the centroid locations of the objects) and corresponding J2000 celestial coordinates were input to the iraf 2 routine ccmap which evaluates a 2D quadratic polynomial surface fit (using a tangential sky projection for the celestial coordinates) that converts image pixel coordinates to celestial J2000 coordinates; it then re-evaluates the celestial J2000 coordinates, using the 2D quadratic polynomial surface fit, for all bright FAUST objects. Those FAUST bright objects whose coordinates showed a deviation larger than 1 arcmin from their catalogue positions were rejected. The 2D quadratic polynomial surface fit was then re-evaluated using the more accurately registered objects. The final accuracy at the last stage of coordinate transformations was better than 1 arcmin. The iraf routine cctrans was then used for each field to transform pixel coordinates to celestial J2000 coordinates for all sources.

Source photometry
The photon flux from each detected source was measured by integrating all count rates within a simulated round aperture centred on the location of the object and subtracting the local background. The circular aperture of diameter 16 pixels, 2 Image Reduction and Analysis Facility, a general purpose software system for the reduction and analysis of astronomical data. iraf is written and supported by the iraf programming group at the National Optical Astronomy Observatories (NOAO) in Tucson, Arizona. NOAO is operated by the Association of Universities for Research in Astronomy (AURA) Inc. under cooperative agreement with the National Science Foundation. equivalent to 17.8 arcmin, was used for all non-saturated sources; the aperture is shown in Figs 1 and 2. This aperture is much larger than the resolution of FAUST, and it allows for the collection of virtually all the UV photons from a source. The local background was determined using an annulus of 10 pixels wide starting in the region immediately (0.1 pixels) outside the source aperture. The mode rather than the mean of the background pixels was evaluated, as this gives photometry with a smaller error. For very bright (saturated) sources an aperture of diameter 80 pixels (88.8 arcmin) was used with a background annulus of 10 pixels.
FAUST was calibrated before flight in Berkeley, but there was no post-flight calibration of the instrument. Bowyer et al. (1993) used 16 objects that appear on FAUST images and were observed by IUE to establish an in-flight calibration for converting FAUST apparent source brightness counts per second to photon s 21 cm 22 A 21 Y shown as fig. 2 in Bowyer et al. (1994). Using this figure, a relation was derived that converts counts per second (SN), measured for any object in the FAUST image, to UV monochromatic magnitudes (called here F 1660 ) ± whilst assuming that the energy is distributed with constant power per unit wavelength (Brosch et al. 1995): The iraf routine radprof was used to evaluate the count rate for each source (using the aperture size and background annulus described above), and to convert this to a FAUST monochromatic magnitude, F 1660 , using equation (1). The instrumental errors were evaluated for each source from the photon statistics of the source and source background. The systematic errors (resulting from the intrinsic errors of the IUE photometric scale and a systematic error in flux resulting from the laboratory calibration of FAUST) were estimated by Bowyer et al. (1993) to be 15.8 per cent of the measured flux, corresponding to an additional error of 0.17 mag to the monochromatic magnitude used here. The total photometric error was calculated by combining in quadrature the instrumental and systematic errors. Table 1 shows the sources detected in both the Antennae and NGC 6752 fields. The first column is the source number detected, prefixed by FA for Antennae sources, and by FN for NGC 6752 sources. Columns 2 and 3 are the (a , b) coordinates of the source as determined from the astrometric transformation described earlier. Column 4 is the monochromatic flux at 1660 A Ê , and column 5 the total photometric error. Column 6 is the FSC catalogue number. The positional uncertainty is 3 arcmin for sources in the FSC; see Bowyer et al. (1995). Using 2 arcmin (2s ) for our astrometric accuracy, the errors are added in quadrature, and therefore any FSC object within 3.7 arcmin of our source coordinates is considered the same source. Table 1 indicates two possible FSC counterparts for FN22, with FSC 4465 marginally closer. Sources indicated with an asterisk are saturated.
The faintest sources we found for each field are FN32 and FA1. FN32 has a net count rate of 0.046 count s 21 , corresponding to <27 counts collected by FAUST during the exposure, and FA1 measured 0.102 counts s 21 , corresponding to <33 counts collected during the exposure.
The FSC includes 17 sources for the Antennae field and 68  sources for the NGC 6752 field that were not detected by our algorithm. Visual inspection indicated that these were virtually all either noise artefacts or sources too close to the field edges to be picked up by our source detection algorithm. Table 1 includes six sources for the Antennae Field, and 12 sources for the NGC 6752 field detected here but not appearing in the FSC.

Source identification
The identification of each source with a counterpart from a list of catalogue objects within a 2-arcmin radius was based upon the following criteria: blueness, blue magnitude; distance from the    Note. ± Description of morphological type coding for above galaxies is from de Vaucouleurs et al. (1976), Sandage & Tamman (1981), Makenty (1990) and Buta (1995): SA Ordinary spiral; SB Barred spiral; SAB Mixed Spiral; O/a, b, c, dm, m Intermediate spiral classes; the presence of an inner ring structure is indicated by (r) following the spiral subclass, those spirals where the arms spring from the ends of the bar or are traced into the centre are indicated by (s), intermediate cases are designated with (rs); Pseudo-outer rings surrounding the galaxy are indicated by R Â preceding the type with subclass, R Â _1, a type where the outer arms wind 1808 with respect to the bar end and subclass, R Â _2, a type where the arms wind 2708 with respect to the bar ends; pec peculiar; Sy1 Seyfert Type 1. Description of photometric descriptors from de Vaucouleurs et al. (1991): BT total (asymptotic) magnitude in the B band; BTC is the face-on-corrected total magnitude in the B band; BJ is the B magnitude on the Johnson system; UT is the total magnitude in the U band; VT is the total magnitude in the V band; VJ is the V magnitude on the Johnson system; RTC is the face-on-corrected total magnitude in the R band.
source, the measured source F 1660 2 B magnitude; and, if known, the candidate counterpart spectral class, luminosity class and B 2 V colour. The bluer, brighter and closer the object, the more likely it is to be the counterpart. In addition, where some or all of the spectral, luminosity and B 2 V colour are available, the object is more likely to be the counterpart if its predicted F 1660 magnitude (using the colour relationships of Shemi et al. 1994, and modifications therein) approximates the measured F 1660 magnitude. Finally, the measured source F 1660 2 B magnitude of the object must fall within the expected range # 26X0 of known astrophysical objects. The majority of the bright sources were identified using the main part of the Tycho catalogue (Hog et al. 1998) for stellar sources and the NASAaIPAC Extragalactic Data base (NED) for extragalactic sources. Fainter sources were identified using the US Naval Observatory two-colour catalogue, USNO-A2.0 (Monet et al. 1998), and the COSMOS 3 (Yentis et al. 1992) catalogue. The USNO-A2.0 catalogue, containing both stellar and non-stellar sources, is based upon a re-reduction of the Precision Measuring Machine (PMM) scans from the Palomar Observatory Sky Survey I (POSS-I) O and E plates and the UK Science Research Council SRC-J survey plates and European Southern Observatory ESO-R plates. The COSMOS catalogue data, also containing both stellar and non-stellar sources, consist of a set of scans on the SRC-J plates for all objects detected to B J (where J refers to the J plate) <21.5 mag, but with a smaller machine scanning aperture than USNO-A2.0 and a better photometric accuracy.
The USNO-A2.0 photometric accuracy is reported to be 0.15mag internal error and up to 0.5 mag due to systematic effects. The absolute values of the blue and red magnitudes are not accurate, but a local (radius 2 arcmin) comparison of objects for each UV source gives the bluest and brightest candidate which, when crosscorrelated with corresponding COSMOS objects and visual inspection of images from the Digitized Sky Survey, yields the best counterpart.
Additional catalogue searching was undertaken with the Deep Near Infra-red Survey of the Southern sky, DENIS (Epchtein et al. 1999), the HEArsarc data base (viz., FAUST, Palomar Green, ROSAT, McCook and Sion Catalogue of White Dwarf Stars, TD-1, IRAS and all the major multicatalogues), the Multi-Archive at the Space Telescope, MAST (the Multimission Archive at StscI being developed to support a variety of astronomical data archives, with the primary focus on scientifically related data sets in the optical, UV, and near-infrared parts of the spectrum; MAST is funded by NASA's Office of Space Science through a grant from NASA and other grants and contracts) the HST, ASTRO, IUE, Copernicus, EUVE catalogues, the Washington Visual Double Star Catalogue (Worley & Douglass 1997) and the TD1F Catalogue. Table 2 gives the identification of UV Sources: repeating the source identifier from Table 1 in column 1; a catalogue identifier 4 in column 2; spectral type/luminosity class (if known) for stars obtained from SIMBAD or morphological type for galaxies from NED in column 3; the F 1660 magnitude in column 4; the apparent B magnitude in column 5; the apparent V magnitude in column 6; the apparent R magnitude in column 7; F 1660 2 B in column 8; and B 2 V in column 9 ± where B 2 V is not available, B 2 R from the USNO-A2.0 catalogue is given and is indicated by the symbol *u. *uu indicates that the B value from the USNO-A2.0 catalogue was used to evaluate F 1660 2 BY as there was no COSMOS counterpart. BJ indicates the Cosmos B J magnitude.
The summary of extragalactic detections is given in Table 3: column 1 is the source identifier as before; column 2 gives the catalogue identifier; column 3 the morphological type; columns 4±7 the apparent F 1660 , U, B and V magnitudes; and columns 8, 9 and 10 the F 1660 2 BY B 2 V and B 2 R colours respectively.

Wise observations
Follow-up observations were made with the 1-m telescope at the Wise Observatory (WO). Observations were possible only for the Antennae Field, as NGC 6752 is located too far south to be observable from Israel. The seeing conditions at the WO are <2 arcsec. Two source locations indicated by an asterisk in column 1 in Table 1 were examined with CCD imaging, and one source indicated by an`s' in column 1 in Table 1 was examined with spectroscopy. The additional information obtained from the WO assisted in the identification and confirmation of catalogue counterparts.
Some Antennae sources which could not be safely identified with known objects were observed with the WO Tektronix 1024 Â 1024 CCD in imaging mode. The pixel scale is 0.7 arcsec pixel 21 and the overall field of view of the CCD is 11X95 Â 11X95 arcmin 2 X Two source locations were observed during the period 1999 April 22±24 with B, V and R filters. Flat-field and bias measurements were taken and used in the reduction process to remove instrument background. The reduction also included corrections for atmospheric extinction. Photometry was performed using the iraf radprof package. Results are discussed in Section 3.2.1 on individual sources, FA9 and FA11.
The Faint-Object Spectrograph and Camera (FOSC) was used for spectroscopic observations of FA36 (from Table 1), a multiple star system, on 1999 April 30. The FOSC in spectroscopic mode consists of a focal-reducer camera with a collimated beam section and a 10-arcsec-wide long-slit focal plane entrance aperture. A grism with 600 line mm 21 was inserted into the collimated beam section in order to disperse the incoming light; and a wedged window as a beam steerer was used to centre the desired spectral region on to a Tektronix back-illuminated CCD chip with Metachrome II coating to enhance its blue±ultraviolet response. The CCD has a 1024 Â 1024 pixel format, but for spectral observations only the relevant part of the chip containing the spectrum of the object and that of the nearby sky was read out. The FOSC spectra, in configuration chosen for the two FA36 candidates, cover the region from ,4000 to ,7800 A Ê with 3.75 A Ê per pixel, resulting in ,8-A Ê resolution.
During the spectroscopic observations the spectrograph's 10arcsec-wide long slit was aligned at PA 1678 in order to include simultaneously both objects. In this manner a maximum separation between the spectra of two objects is achieved, minimizing the possibility of light from one object contaminating the other. The spectra were extracted after debiasing, calibration and flat-fielding. Reduction of all spectra was conducted using the iraf specred package. Since no spectroscopic standards were observed when the spectra of the stars were taken, the spectrum was flux-calibrated using the WO standard coefficients. These generally do not change significantly from night to night. Results are discussed in Section 3.2.1 on individual sources, FA36. Table 4 gives the summary of results with respect to source type. The number of unknown sources in the Antennae is ,28 per cent compared to ,55 per cent in NGC 6752. We noted earlier that the exposure for NGC 6752 was longer than that for the Antennae; consequently, more faint sources were detected, and most of these are unclassified.

Stars
Tables 5 and 6 bin sources by both F 1660 magnitude and spectral type for the Antennae and NGC 6752 fields respectively. Assuming that all unclassified objects are stars, the proportion of B, A and F stars is 67 per cent of the stellar population for the Antennae, and 46 per cent of the stellar population for NGC 6752.
Both tables indicate a notable deficiency in hot evolved stars (HES); only one identified HES object (the hot white dwarf REJ1925 2 563 was found in the NGC 6752 field. This is far fewer than in other FAUST fields: 11 HES in the Coma cluster (Brosch et al. 1998), 10 HES in the North Galactic Pole (Brosch et al. 1995), and seven HES in the Virgo cluster (Brosch et al. 1997). As we will discuss later, Figs 7±10 indicate that many of the unclassified objects in Tables 5 and 6 may be HES. This can only be verified by follow-up spectroscopic observations of all these objects.
A comparison with the predicted differential star density from our UV model (Brosch 1991, and modifications therein;Bilenko, private communication) for the Antennae and NGC 6752 regions is given in Figs 3 and 4. The UV model does not include subdwarfs, horizontal branch stars, post-asymptotic giant branch or post-horizontal branch stars. Fig. 3 shows a marked lack of stars for all F 1660 magnitudes .7.0 (e.g., at F 1660 9X5Y there is a .4s difference between the model and the observations). The agreement in Fig. 4 is very good up until F 1660 10X0Y when the expected drop in detected stars starts to occur as the faintness detection limit of FAUST is approached. To evaluate the detected stellar densities in Figs 3 and 4, we took the effective exposure of each FAUST field as all pixels with a value .25 per cent of the peak exposure, where objects can be reliably found, giving a field of 42.8 deg 2 for the Antennae, and 43.6 deg 2 for NGC 6752. Using these field areas and the total number of stars from Tables 5 and 6, the UV stellar density was evaluated as 0.85 star deg 22 for the Antennae and 4.45 star deg 22 for NGC 6752.
Figs 5 and 6 show the F 1660 2 B colour distribution for the Antennae and NGC 6752 fields respectively, with along the predicted curves from our UV model. The predicted curves were generated using a limiting faintness of F 1660 12 magY where the FAUST source detection begins to drop sharply. The curves were then scaled to the peak stellar count/colour bin for each figure. In Fig. 5 the x 2 probability for the predicted curve describing the data, over the region F 1660 2 B $ 23X0Y is 51 per cent, indicating a very good fit. The discrepancy between the predicted curve and data for F 1660 2 B , 23X0 is possibly due to those hot evolved objects (e.g., ZAHB, EHB, post-EHB and subdwarf stars) that our UV model does not include. In Fig. 6 the discrepancy starts already at F 1660 2 B , 0X Amongst our detections in the NGC 6752 field, there are many more unclassified objects in the colour range 23X0 # F 1660 , 0X0 compared to the detections in the Antennae field ± see Fig. 10 compared to Fig. 8. These unclassified objects could include many of the afore-mentioned hot evolved objects and hence explain the discrepancy starting at a redder colour. We also note that in Fig. 8, for the region F 1660 2 B $ 0X0Y the predicted curve only roughly (within 2s ) describes the data; the reason for this is not known at present.
We have plotted colour±magnitude diagrams for both fields: Antennae (Figs 7 and 8) and NGC 6752 (Figs 9 and 10), and use them as diagnostic diagrams to establish the likely types of stars that are FAUST sources. We have also plotted in Figs 7 and 9 the main sequences for different distances. These have been evaluated using data from Fanelli et al. (1992). Specifically, we calculated M 1700 and M 1700 2B from M V , 1700 2 V and B 2 V given in their tables 5 and 7. The M 1700 band defined there is very similar to the FAUST sensitivity band, so a direct comparison is valid. In Figs 8 and 10, we have also plotted the canonical zero-age horizontal branch (ZAHB) (for a scaled-solar metallicity of Z 5 Â 10 24 and a main-sequence helium abundance of Y 0X23 at varying distances. The ZAHB was derived from fig. 2 of Landsman et al. (1996) and the B 2 V data (table 7) (1979) extinction law for EB 2 V 0X3X This represents a moderate amount of extinction to the objects, as shown in the Introduction. The very bright objects labelled 29, 32, 38 and 46, the faint red object labelled 1, and the blue object labelled 25 in Fig. 7 are discussed in Section 3.2.1 on individual sources. In a similar manner the very bright identified object labelled 178 in Fig. 9 and the very blue object labelled 151 in Figs 9 and 10 are also discussed in that section.
We now discuss the nature of the unclassified sources in Figs 7±10. For both the Antennae and NGC 6752 the view direction is through the nearby disc and Galactic halo. The unclassified objects in the Antennae image (referring to Figs 7 and 8) with F 1660 2 B . 24X0Y could be ZAHB stars closer than 2 kpc, white dwarfs within 500 pc, and main-sequence (or giant) giants at distances .1 kpc. For F 1660 2 B , 24X0Y subdwarfs are the only known candidates. The unclassified objects in the NGC 6752 image (referring to Figs 9 and 10) with F 1660 2 B . 24X0 could be: ZAHB stars ,4 kpc, white dwarfs within 500 pc, and main-sequence (or giant) stars more distant than 300 pc. For F 1660 2 B , 24X0Y subdwarfs are the only known candidates. We also note from fig. 3 of Landsman et al. (1996) that EHB and post-EHB stars are possible candidates for unclassified objects with 25 , F 1660 2 B , 0X The diagnostic diagrams in Figs 7±10 allow a tentative assignment of spectral types to the unclassified sources. These consist of hot white dwarfs within a few hundred parsecs from the Sun, ZAHB/EHB stars within 4 kpc, and main-sequence stars within 2 kpc. The direction of NGC 6752 seems to be particularly rich in unclassified sources. As the F 1660 -bright objects do not appear to concentrate particularly near the globular cluster, the UV objects are probably not cluster members. This is probably some disc population of WD or ZAHB/EHB stars, which shows up more prominently due to the lower Galactic latitude of this pointing direction.
The FAUST sample allows the evaluation of the accuracy of inferring the UV brightness of stars while relying on optical colours only. Figs 11 and 12 are F 1660 2 V versus B 2 V colour plots for the Antennae and NGC 6752 images respectively. All classified stars with known luminosities are plotted; in addition, four very B 2 V red objects (FN83, FN87, FN165 and FN184) with unknown luminosities have been added for discussion in Section 3.2.1 on individual sources. The curves are the modified relationships of Shemi et al. (1994) which were evaluated using IUE spectral data sets, data from the Hipparcos Input Catalogue, and measurements from the TD-1 satellite. Both figures demonstrate that the above method of inference is reliable, considering the uncertainties in the quadratic relationships and any extinction associated with each star. The method is applicable to all luminosity classes and spectral types (B, A, F and K) detected by FAUST; only those numbered stars (4,37 and 46 in Fig. 11,and 9,83,87,88,123,165,169 and 184 in Fig. 12) are significantly distant from the predicted curves, and are discussed in Section 3.2.1. Fig. 12 shows that most sources in the NGC 6752 image lie to the right of the theoretical relationships. This is true for all luminosity classes, and indicates a general behaviour of UV       sources in this region. We plotted in the figures the influence of a moderate amount of extinction: this shows that the UV sources in this field could be reddened by dust. We further tested the proposition of Brosch et al. (2000a) that the UV properties of A stars can be used to identify possible candidate Am or Ap stars. Fig. 13 shows a F 1660 2 B versus B 2 V colour plot for all A stars in both FAUST images. Metallic A and F stars (both indicated by filled triangles) are clearly UV-faint when compared to early A stars (open triangles). Of the nine`m' stars, two are F stars. The UV deficiency is probably due to line blanketing of their spectral energy distributions. This deficiency has been noted before by van Dijk et al. (1978), mostly in regard to Ap stars. Our results agree with Brosch et al. (2000a) that UV faintness in loosely classified A stars is an indicator of possible metallicity. Fig. 13 also includes an Ap star, two nebular B stars, and a BII±III emission star (FN178 saturated in the FAUST image).

Individual sources
Here we discuss individual sources that merit attention due to their unusual location in the colour±magnitude diagrams of Figs 7±10 and/or the colour±colour diagrams of Figs 11 and 12. In addition, objects observed at the WO are discussed.

Stars
FA1 This object appears extremely dim and red in Figs 7 and 8. This can be explained by the very large error in F 1660 14X12 X16X FA4 In Fig. 11 this object it appears redder in both colours than predicted. This could be a result of reddening by dust of an A star, as indicated by the spectroscopic typing. Figure 7. UV colour±magnitude diagram for objects identified as stars in the Antennae image with the main-sequence branch at different distances derived from Fanelli et al. (1992). Fiducial error bars are indicated at the lower-left corner for classified and unclassified stars. The giant-branch curves are not shown, as they do not differ notably in location from the main-sequence branch. The colour±magnitude shift due to a typical colour excess of EB 2 V 0X3 is shown as a vector. Figure 8. UV colour±magnitude diagram for objects identified as stars in the Antennae image with WD cooling curves at different distances and masses from Wood (1995) and zero-age horizontal-branch (ZAHB) stars at different distances from Landsman et al. (1996). Fiducial error bars are indicated at the lowerleft corner for classified and unclassified stars. The colour±magnitude shift due to a typical colour excess of EB 2 V 0X3 is shown as a vector.
FA9 This is the second brightest object, F 1660 8X84^0X17Y in the Antennae image without a reliable identification. The best counterpart USNOA-2.0 0675-11972813, given in Table 2, was verified with UBV observations at the WO: USNOA-2.0 0675-11972813 is the bluest source present in the U 2 B versus B 2 V plot. The F 1660 2 B of this source is 24X37^0X23 and F 1660 is 8X84^0X17X From our discussions of Figs 7 and 8, this would indicate a subdwarf, a very distant (.15 kpc) mainsequence or giant B star in the Galactic halo or a ZAHB/EHB star at under 1 kpc. Follow-up spectral observations are essential to determine its nature.
FA11 The best counterpart USNOA-2.0 0675-11794652 was observed in U, B and V at the WO: USNOA-2.0 0675-11794652 was found to be the bluest source present. F 1660 2 B 20X910 X29 and F 1660 11X19^0X25X Possible types of counterparts are a ZAHB/EHB at between 1±2 kpcY a main-sequence or giant star at <2 kpc, or a very nearby hot white dwarf at <10 pc. The last possibility is rather unlikely, as it would have implied a measurable parallax. Follow-up spectral observations are essential to determine its nature.
FA25 Fig. 7 places this object at <2 kpc, the most distant Tycho counterpart in the Antennae field.
FA29 This object is saturated in the FAUST image of the Antennae Field, and so its true location in Figs 7 and 8 should be brighter and bluer than shown. The counterpart is a spectroscopic binary and the parallax measurement quoted in SIMBAD places it at 538 379 2158 pcX FA32 This object is saturated in the FAUST image of the Antennae Field, and so its true location in Figs 7 and 8 should be brighter and bluer than shown. It is a variable star, and the parallax measurement from SIMBAD places it at 51^2 pcX FA36 This is the brightest object, F 1660 8X36^0X17Y in the Antennae image lacking a Tycho identification. The best counterpart was found to be USNOA-2.0 0675-11972813. Examination of the SIMBAD data base revealed a double or multiple star system listing a B9V star HD 105604 with B and V 9X7 and an unclassified object BD-16 3399B with a V magnitude of 11.6. Both stars are also identified in the Washington Double Star Catalogue WDS J1209X6 2 1734 that indicates the presence of a triple system. We examined the two SIMBAD listed stars at the WO, and confirmed that HD 105604 is a B9V star. The second candidate is probably an F or G star. HD 105604 is therefore the hotter of the two objects (as well as the brighter in V), and is thus identified as the correct counterpart UV object.
FA37 In Fig. 11, this object appears F 1660 2 B bluer and B 2 V redder than predicted. This could be the result of dust reddening of an A star, as indicated by the assigned spectral type. The parallax measurement from SIMBAD indicates a distance of 240 87 251 pcX FA38 The parallax measurement from SIMBAD indicates a distance of 166 26 219 pcY confirming its location in Fig. 7 as a nearby source.
FA46 This is a variable star, and the parallax measurement from SIMBAD indicates a distance of 88^6 pcY confirming its location in Fig. 7 as a nearby source.
FN9 The colours are F 1660 2 B 20X12^0X17 and B 2 V 20X19^0X01X The F 1660 2 B colour is much redder than Figure 10. UV colour±magnitude diagram for objects identified as stars in the NGC 6752 image with WD cooling curves at different distances and masses from Wood (1995) and zero-age horizontal-branch (ZAHB) stars at different distances from Landsman et al. (1996). Fiducial error bars are indicated at the lower-left corner for classified and unclassified stars. The colour±magnitude shift due to a typical colour excess of EB 2 V 0X3 is shown as a vector. Figure 9. UV colour±magnitude diagram for objects identified as stars in the NGC 6752 image with the main-sequence branch at different distances derived from Fanelli et al. (1992). Fiducial error bars are indicated at the lower-left corner. The colour±magnitude shift due to a typical colour excess of EB 2 V 0X3 is shown as a vector. predicted in Fig. 12. SIMBAD gives a radial velocity of 151 km s 21 , and the distance from the parallax is 244 59 240 pcX The motion perpendicular to the line of sight, evaluated from proper motion measurements in SIMBAD and the parallax distance, is <39 km s 21 . The spatial velocity is thus <156 km s 21 , a fairly high-velocity star. It has a companion star HD 180183B, a K0V star, with V 12X0 FN78 This is a double or multiple star; on sky projection it is within 1 arcmin of the globular cluster NGC 6752 and outshines it at F 1660 .
FN83 The colours are F 1660 2 B 20X91^0X18 and B 2 V 0X46^0X08Y much redder than predicted (Shemi et al. 1994, and modifications therein) for an A7 star of any luminosity class. This could be the result of very high dust extinction, EB 2 V < 0X5X FN87 The colours are F 1660 2 B 1X31^0X19 and B 2 V 0X45^0X04X These colours are redder than predicted for an F0 star of any luminosity class. This could be the result of high dust extinction.
FN88 In Fig. 12 this object appears much redder than predicted. This could be the result of high dust extinction.
FN123 The distance derived from the SIMBAD parallax is 260 80 240 pcX Its F 1660 2 B 1X28^0X17 and B 2 V 0X670 .01. The F 1660 2 B colour is much bluer than predicted by its type. Brosch (2000a) indicates that K stars are expected to have F 1660 2 B < 3X2±7X4Y but some stars in this class are more UVbright than this, possibly because of some form of coronal activity (see, e.g., Jordan et al. 1987), or because of the presence of a hot secondary such as a hot white dwarf.
FN151 This source is identified with a star spectrally classified as B. If main-sequence, it should be at <30 kpc ± putting it far beyond the accepted Galactic boundaries. If, on the other hand, we assume it to be a WD (perhaps a DA dwarf), it could be as close as <100 pc.  FN165 The colours are F 1660 2 B 0X43^0X20 and B 2 V 0X40^0X05Y much redder than predicted for an F2 star of any luminosity class. This could be the result of high dust reddening.
FN169 In Fig. 12 this object appears much redder than predicted. This could be the result of high reddening.
FN178 This object is saturated in the FAUST image of the Antennae field, and so its true location in Figs 9 and 10 should be brighter and bluer than shown.
FN184 The colours are F 1660 2 B 20X97^0X17 and B 2 V 0X31^0X04Y much redder than predicted for an A3 star of any luminosity class. This could be the result of very high dust extinction.
For all those spectrally unclassified sources in Table 2, we have already discussed the nature of possible counterparts with reference to Figs 7±10.

Galaxies
We detected a total of 11 galaxies: 10 are known objects, and one is the best candidate from the USNO-A 2.0 and COSMOS Catalogues. Two galaxies show forms of nuclear activity; one is a Seyfert, and the other is a LINER. The origin of the UV emission in the nine other objects is presumably general star formation. This is supported by the high fraction of barred galaxies among the 10 objects. An extensive discussion of galaxies observed by FAUST was given by Deharveng et al. (1994). We now discuss each object listed in Table 3. All data are from the NED data base.
FA28 This is the spiral [SA(s)c] NGC 3956 with a heliocentric radial velocity of 1645^5 km s 21 X It has patchy arms with many knots.
FA31 This is the barred spiral [SB(s)dm] NGC 4027 with a heliocentric radial velocity of 1671^6 km s 21 X It is an asymmetric spiral with very bright knotty arms.
FA33 This corresponds to both Antennae galaxies, NGC 4038 and 4039. Both have very long tidal tails. NGC 4038 is a barred spiral [SB(s)mpec] with a heliocentric radial velocity of 16421 2 km s 21 X The northern part of the Antennae, it has very bright knotty areas. NGC 4039 is a barred spiral [SA(s)mpec] with a heliocentric radial velocity of 1641^9 km s 21 X The southern part of the Antennae, it shows both loops and plumes, presumably the result of tidal interactions.
FN30 This is the Seyfert [Sc, Sy1] ESO 141-G055 with a heliocentric radial velocity of 10 793^90 km s 21 X Marginal evidence for spectral flattening has been found at high energies. It was observed by the HST Goddard High Resolution Spectrograph (GHRS), and also by the IUE. The F 1660 2B colour of 22.96 is the bluest of the FAUST galaxies detected here. There were seven observations by the IUE over the period 1980 October 2 to 1990 September 14; the variation in F 1660 was 11X13^0X05 to 11X88^0X10X This compares with the FAUST measurement of F 1660 11X40^X09; this suggests that all the F 1660 emission originates from the nucleus.
FN36 This is the barred spiral R 2 1SBsrOaa NGC 6782 with a heliocentric radial velocity of 3736^37 km s 21 X It has a double bar structure and a very bright nucleus. The outer arms form an R 2 1 outer pseudoring in blue light. FN42 This is the spiral RSArb NGC 6753 with a heliocentric radial velocity of 3124^26 km s 21 X The ellipticity and PA profiles of this galaxy show several small peaks related probably to the presence of star forming regions. H ii regions and diffuse emission concentrate in the nuclear ring-lens.
FN58 This corresponds to the interacting pair of galaxies NGC 6769 and 6770. There is a bridge between the two galaxies. NGC 6769 is a spiral [(R Â )SA(r)b] with a heliocentric radial velocity of 3686^46 km s 21 X There is a dark lane in the disc and an H ii region evident. NGC 6770 is a barred spiral [SB(rs)b] with a heliocentric radial velocity of 3813 km s 21 . There are streamers, plumes, loops and one arm with H ii regions. There are two dark lanes.
FN140 This is the mixed spiral [SAB(r)bc] LINER NGC 6744 with a heliocentric radial velocity of 841^5 km s 21 X The region dominated by the bar is very extended and quite clean of gas. Many knots are evident. Two arms in particular contain a bunch of adjoining giant H ii regions where most of the massive star formation must currently be taking place. NGC 6744, with F 1660 8X95Y is the brightest of the FAUST galaxies observed in the NGC 6752 and the Antennae fields.
FN147 This is not identified with any known galaxy; however, the USNOA-2.0 catalogue gives USN0A-2.0 0225-29766917 as the bluest object in this field with B 2 R 0X1X Examination of the Digitized Sky Survey image shows USN0A-2.0 0225-29766917 to be an elliptical galaxy with dimensions 50 arcsec by 34 arcsec. There is a galactic triplet, IC 4803 at 1.3 arcmin from USN0A-2.0 0225-29766917, which may contribute to the F 1660 flux. FN147 is the bluest object among our FAUST-detected galaxies.

C O N C L U S I O N S
We have analysed two UV images from the FAUST experiment: the Antennae region covering 67.5 deg 2 , and NGC 6752 region covering 64.0 deg 2 . A total of 227 sources were detected, of which approximately 60 per cent have reliable identifications: most are B, A and F stars; the remaining objects are 11 galaxies, two K stars and a white dwarf in a binary system. The remaining <40 per Figure 13. F 1660 2 B versus B 2 V colour comparison plot between regular A stars and metallic (and other non-normal) stars from both FAUST images. The colour±colour shift due to a typical colour excess of EB 2 V 0X3 is shown as a vector. cent have assigned optical counterparts from the USNO-A2.0 catalogue, and these objects could plausibly be white dwarfs, subdwarfs, horizontal branch stars or main-sequence (or giant) stars. Only one bright F 1660 9X57 object has no assigned optical counterpart. A comparison of the star counts with the predictions of a model for the galactic UV stellar populations shows good agreement for the NGC 6752 field but there is a marked shortage of stellar objects in the Antennae field. We have demonstrated that metallic A stars are UV-faint compared to normal early A stars, and shown that in most instances it is valid to infer the UV brightness of stars given the B and V colours only. Follow-up spectroscopic measurements of the <40 per cent unclassified objects are essential to determine their true nature. The entire sample of UV measured objects from the FAUST flight will allow firmer statistical conclusions on the distribution of UV bright stars and possibly the identification of additional peculiar UV sources.

AC K N OW L E D G M E N T S
The UV astronomy effort at Tel Aviv is supported by special grants to develop a space UV astronomy experiment (TAUVEX) from the Ministry of Science and from the Austrian Friends of Tel Aviv University. This research has made use of the NASA/IPAC Extragalactic Data base (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This research has also made use of High Energy Science Archive Research Center (HEArsarc) data base provided by NASA's Goddard Space Flight Center. In addition, this research has also made use of the COSMOS/UKST Southern Sky Catalogue supplied by the Anglo-Australian Observatory. We also appreciate the assistance of Eran O. Ofek for his assistance in the processing of WO photometric data. JD thanks both the Leverhulme Foundation of the United Kingdom, and Roger Smith in association with the Adult Education HELP facility at Eastfield School, Hull, UK for initiating his post-doctoral career in UV astronomy. NB acknowledges support from the US-Israel Binational Science Foundation and from the Austrian Friends of Tel Aviv University. EA was supported partly by a grant from the Israel Ministry of Science, Culture and Sport to develop TAUVEX. JD was supported partly by an internal grant of Tel Aviv University. JD is also grateful to the referee for his helpful comments.