To decompose the complex configuration into discrete sources, we employed the DAOPHOT II software package as implemented within ESO-MIDAS. For the optical images, a numerical composite PSF was built for each image using three bright stars at distance to the QSO. In the K band data no bright PSF star was available within the small field of view, and a purely analytical PSF was estimated from the only two isolated stars in the field. The QSO was then modelled by the superposition of five point sources, with positions and PSF scaling factors simultaneously optimised (routine ALLSTAR). Fitting five components gave always a significantly better fit, in terms of residual , than with four, even in the B band image where C and D are both very faint. The fitting results are collected in Tables 1 and 2. Positions are measured relative to A1, which was arbitrarily taken as reference point.
Table 1. Differential astrometry for HE 0230-2130, based on the I and K band images.
Table 2. Differential photometry of HE 0230-2130. The first row gives the total magnitude from aperture photometry, all other entries were computed relative to these values.
For components A1, A2, and B, the positions measured in the four BRIK images are very consistent, and the scatter between the photometric bands reflects the measurement error. For C and D, image positions are consistent only between I and K; at shorter wavelengths, the fitted centroids approach each other as illustrated in Fig. 2. This could be related to different intrinsic colours of the objects, but could also be an artefact of low S/N and the somewhat poorer seeing in the shorter wavelength data.
Differential PSF photometry of the QSO components relative to A1 was available from the ALLSTAR analysis (Table 2). Flux calibration was established separately using simple aperture photometry (aperture diameter was for BRI and for ). Standard stars from the RU 149 field (Landolt 1992) served to obtain colour terms and photometric zeropoints in the optical, while SJ 9106 (Persson et al. 1998) was used for the NIR photometry.
The total magnitudes thus measured are listed in the first line of Table 2, with uncertainty estimates as given in the DAOPHOT output. We have also determined aperture BRI magnitudes for 14 nearby stars in the field that may be useful to serve as reference stars in future monitoring. A list with these measurements is available on request. 1
The relative photometry confirms that A1, A2, and B have very similar optical-NIR colours although B appears slightly redder than A1 and A2. C and D are much redder, on the other hand, so neither can correspond to a single unobscured fourth QSO image. Because of the apparent positional shift between the bands, we computed a second model with fixed positions imposed from the I band image, fitting only the PSF scaling factors. The resulting colours are slightly bluer for components B and C, and even much redder for D. However, inspection of the PSF-subtracted images indicated that the fit quality of these restricted models was much poorer, leaving residuals significant on the 2-3 level.
Table 3. Colours of the components in HE 0230-2130, based on the PSF magnitudes from Table 2.
3.3. Spectroscopic properties
While in double QSOs there is always the possibility that a true binary system is being observed, a configuration like that seen in HE 0230-2130 is almost certainly best explained as a lensed system, even without spectroscopic evidence. Although we do not yet have spectra of all components, the available data allow nevertheless to confirm the lens hypothesis beyond all reasonable doubt:
(1) The total spectrum (Fig. 3) contains no trace of absorption features that would be expected if A1 was a star or a galaxy. We conclude that A1 and A2 have most probably very similar spectra, given the broad-band colours.
(2) A2 and B have both very similar QSO spectra, apart from the slit loss effects. Fig. 4 shows that the emission line centroids agree within the measurement accuracy, the line widths are equal, and also the strong `associated' () C IV absorption system is clearly present in both components.
A curious feature, however, are the significant residuals detected in the difference spectrum A2-B(scaled), indicating non-identical emission line profiles and/or equivalent widths. Whether this might be due to differences in intervening line absorption along the lines of sight, or due to selective continuum microlensing such as proposed in HE 1104-1805 (Wisotzki et al. 1993) remains to be explored; the current data do not permit a more detailed analysis.
3.4. Nearby galaxies
Visual inspection of the available images (see Fig. 5) shows a number of faint galaxies in the vicinity of the QSO, but most of these objects are very faint and at the limit of the present data (, , ). The surface density of faint objects around the QSO (within radius) was found to be enhanced by a factor of 2-3 compared to the surrounding regions. As star-galaxy separation was not reliable for such faint sources, every detected object was counted, so the overdensity is a conservative estimate. However, the signal is significant only on the 2 level, using Poisson statistics.
Most of the objects within from the QSO are detected in RI (and expectedly not in B). The and colours show relatively little dispersion and are consistent with galaxy colours at low to intermediate redshifts, (cf. Fukugita et al. 1995). We conclude that there is some evidence for a cluster near the line of sight to HE 0230-2130, but that this has to be confirmed by deeper imaging and spectroscopy.
© European Southern Observatory (ESO) 1999
Online publication: July 26, 1999