arXiv:astro-ph/0507164v1 7 Jul 2005
Astronomy & Astrophysics manuscript no. ms2877
(DOI: will be inserted by hand later)
February 5, 2008
Photometric monitoring of the doubly imaged quasar UM 673:
possible evidence for chromatic microlensing⋆
Th. Nakos1,2,3, F. Courbin4, J. Poels2, C. Libbrecht2,3, P. Magain2, J. Surdej2 ⋆⋆, J. Manfroid2, I. Burud5, J. Hjorth6,
L. Germany1, C. Lidman1, G. Meylan4, E. Pompei1, J. Pritchard7,6,1, and I. Saviane1
1European Southern Observatory, Casilla 19001, Santiago, Chile
2Institut d’Astrophysique et de G´ eophysique, Universit´ e de Li` ege, All´ ee du 6 Aoˆ ut 17, Sart Tilman, Bˆ at. B5C, B-4000 Li` ege,
3Royal Observatory of Belgium, Ringlaan 3, B-1180 Brussels, Belgium
4Laboratoire d’Astrophysique, Ecole Polytechnique F´ ed´ erale de Lausanne (EPFL), Observatoire, CH-1290 Sauverny,
5Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
6Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, DK-2100, København Ø, Denmark
7Mount John University Observatory and Department of Physics & Astronomy, University of Canterbury, Private Bag 4800,
Christchurch, New Zealand
Received / Accepted
Abstract. We present the results of two-band CCD photometric monitoring of the gravitationally lensed quasar Q0142−100
(UM 673). The data, obtained at ESO-LaSillawith the 1.54m Danish telescope in theGunni-band (October 1998 − September
1999) and in the Johnson V-band (October 1998 to December 2001), were analyzed using three different photometric methods.
The light-curves obtained with all methods show variations, with a peak-to-peak amplitude of 0.14 magnitude in V. Although
it was not possible to measure the time delay between the two lensed QSO images, the brighter component displays possible
evidence for microlensing: it becomes bluer as it gets brighter, as expected under the assumption of differential magnification
of a quasar accretion disk.
Key words. Gravitational lensing – quasars: individual: Q0142−100 (UM 673)
UM 673 (Surdej et al. 1987) is one of the first gravitational
lens systems (GLs) found in a systematic optical search. It
consists of two lensed quasar images (zs = 2.719) separated
by 2.2′′. Although no direct spectrum of the lensing galaxy
has ever been obtained, CaII and NaI absorption lines, de-
tected in the spectrum of the fainter quasar image B, suggest
that it is at a redshift zl = 0.49 (Surdej et al. 1988; Smette et
al. 1992). Because of its relatively bright apparent magnitude
(mR(A)∼16.5, mR(B)∼19.0) and because of the wide angular
separation between its images, UM 673 is a good target for a
photometric monitoring program. The expected time-delay is
of the order of a few months, which is relatively easy to mea-
⋆Based on observations made with the Danish 1.5m telescope
(ESO, La Silla, Chile) (Proposals: 63.O-0205(A), 64.O-0235(B),
65.O-0214(B), 66.A-0203(B), 67.A-0115(B), 68.A-0109(B)).
⋆⋆Directeur de recherche honoraire du Fonds National de la
Recherche Scientifique (Belgium)
offprintrequeststo: Th.Nakos, e-mail:
sure, providingthat the quasar itself displays significant photo-
UM 673 has been monitored at the European Southern
Observatory (ESO, La Silla, Chile) from 1987 to 1993 with
epochs. Unfortunately, the small field of view (FOV) accessi-
ble with CCDs in the eighties and the subsequent lack of PSF
stars, restricted the effectiveness of the observations to unveil
any weak variation, at the level of a few tenths of a magnitude
(Daulie et al. 1993).
Within the framework of the Hamburg Quasar Monitoring
Program (Borgeest & Schramm 1993), UM 673 has also been
observed from 1988 to 1993 with the 1.23m MPIA telescope,
at Calar Alto Observatory, Spain. In contrast with Daulie et
al. (1993), a preliminary analysis led to the detection of vari-
ations in the light-curve of component A, with a peak-to-peak
amplitude of the order of 0.2 magnitude.
Sinachopoulos et al. (2001) reported variation of up to 0.5
magnitude over several years. Nevertheless, their data were ir-
regularly distributed and the small size of the telescopes used,
2Nakos et al.: Photometric monitoring of the lensed quasar UM 673
Fig.1. Part of oneGunni DFOSC CCD frameof UM 673. The
field ofviewis 5′×5′.Two referencestars areindicated(S1and
S2). Star S1, whose absolute photometryis well known (Nakos
et al. 2003) has been used to normalize the CCD frames. Star
S2 was taken as a comparison star.
under moderate seeing conditions, only allowed the measure-
ment of the total flux of the gravitational lens system.
The latest work on UM 673 (Wisotzki et al. 2004) re-
ported on spectra obtained in September 2002. The data were
taken with the 3.5m telescope at Calar Alto (Spain), using the
Potsdam Multi-ApertureSpectrophotometer(PMAS). By mea-
suring the emission line to continuum flux ratios for the two
quasar images, they concludedthat, during the period of obser-
vations, no microlensing signature was detectable.
UM 673 was part of a photometric monitoring program
carried out at ESO-La Silla, using the 1.54m Danish telescope.
In this monitoring, the photometric variability of several GL
systems was first tested for a few months and the object was
either followed up further in order to measure the time-delay
(see for example Burud et al. 2002a), or stopped due to insuf-
ficient photometric variability. UM 673 is one of the objects
that was stopped. However, it turned out to be interesting,
after a more detailed analysis of the data was carried out.
Both components show variability on long time-scales, with
peak-to-peak variations of the order of 0.14 and 0.08 mag-
nitude, in V and i respectively. While further (spectroscopic)
observations will be needed to discriminate between intrinsic
and microlensing induced variability, it is argued, on the basis
of color information, that at least part of the variability, on
time-scales of a few months, is likely due to microlensing.
Fig.2. Upper panel: a 25′′× 25′′region extracted from one
of the 63 Gunni CCD frames of UM 673. The seeing is 1.2′′
and the pixel scale is 0.39′′. Lower panel: simultaneous de-
convolution of the 63 frames, with a final resolution of 0.585′′
and a pixel size of 0.195′′. Due to the PSF variations in the
DFOSC field, minor artifacts were produced in the vicinity of
the brighterquasarimage.Theyrepresentonly0.5%of its peak
intensity. The lensing galaxy is marginally seen between the
two QSO components,close to componentB. North is up, East
is to the left. A logarithmic intensity scale has been used to
display the images.
The observations of UM 673 were carried out on a weekly ba-
sis using the Danish 1.54m telescope at La Silla (Chile); at
least three CCD frames were taken each night for each filter.
In total, 63 CCD frames were obtained in the Gunni filter, cor-
responding to 18 observing nights, from October 3, 1998 to
September 29, 1999. In the Johnson V-band, 91 CCD frames
were obtained, corresponding to 23 nights. 18 out of these 23
V-band epochs were taken during the same nights as the 18
Nakos et al.: Photometric monitoring of the lensed quasar UM 6733
Gunni CCD frames. The remaining 5 V-band epochs were ob-
tained between August 29 to December 15, 2001.
The CCD mounted on DFOSC (Danish Faint Object
Spectrograph Camera) until September 2000 was a backside
illuminated LORAL/LESSER chip, with a pixel scale of 0.′′39
pixel−1and a field of view of 13.7 × 13.7 arcmin2. This de-
tector was then replaced by an EEV/MAT chip, with the same
pixelscale and FOV. The exposuretime forthe individualCCD
frame (> 3 per epoch) in Gunni was 250 seconds. For the V-
band data, the integration time was varying between 150 and
of the frames having an exposure time of 250 seconds. No data
were lost due to bad weather conditions, thanks to the flexible
a part of a typical Gunni CCD frame. UM 673 can be identi-
fied on the north-east part of the field. Two neighboring stars,
labeled S1 and S2, extensively used during the data reduction,
are also indicated.
3. Data reduction techniques
We performed the photometric reduction and analysis using
three different numerical methods to treat with photometry
of blended objects. All three methods were developed at the
Institute of Astrophysics and Geophysics of the University of
Li` ege. The first one, the so-called MCS, is based on decon-
volution (Magain, Courbin & Sohy 1998). It has been exten-
sively used to obtain the light-curves of many lensed quasars
(e.g. Burud et al. 2002a,b). The second method, Differential
Imaging, is based on a matching kernel technique similar to
that developed by Phillips & Davis (1995). The last method,
General, is a profile fitting technique that has been used to ana-
lyse data of lensed quasars, e.g., by Remy (1996) and Courbin
et al. (1995).
Prior to the photometricanalysis itself, all frames were pre-
processed, i.e. bias-subtracted, flat-fielded and rescaled to the
same pixel-coordinatereference system, using standard IRAF1
procedures. Sky subtraction was performed using SExtractor
(Bertin 1996), which can treat complicated sky structures, by
removing the objects from the data and by estimating the sky
backgroundthrougha gridofcells adaptedto theobjectdensity
of the field. In the case of UM 673, the process resulted in a
very accurate sky subtraction, at the percent level, due to the
low stellar density and to the small spatial variations of the sky
value in the original data.
The PSF varies significantly across the DFOSC field of
view and the observed variation changes from one frame to
another. Because of the very low number of stars suitable for
the PSF determination, it was not possible to properly take this
effect into account. This is probably the major cause of photo-
metric uncertainties in our results.
1Image Reduction and Analysis Facility — is written and sup-
ported at the National Optical Astronomy Observatories (NOAO)
3.1. Image deconvolution
The main difference between the MCS (Magain, Courbin &
Sohy 1998) method and most others is the simultaneous de-
convolution capability of the algorithm: the deconvolved im-
age is computed by minimizing the χ2between all the indi-
vidual frames and a unique model image. The individual χ2
corresponding to each image are minimized at the same time,
hence the name of simultaneous deconvolution. All the availa-
ble data are involved in this simultaneous fit, including those
taken under mediocre observing conditions.
The deconvolved image is decomposed into a sum of ana-
lytical (Gaussian) point sources and a numerical deconvolved
ing galaxy or even the quasar host galaxy, if detected at all.
In addition to the deconvolved image, the procedure returns
the peak intensity and astrometric center of all point sources.
The intensities are allowed to vary from one image to the next,
hence leading to light-curves. The position of the point sources
is forced to be the same in all images. The astrometry of the
images with poor seeing is therefore constrained by the ones
with better seeing.
The signal-to-noise of the final image is thus constrained
by the whole dataset. It also has an improvedspatial resolution,
sampling and depth. Fig. 2 compares a simultaneously decon-
volved image with one of the original data, in the Gunni filter.
The chosen resolution is 0.585′′and the improved pixel size is
0.195′′. The “ring” at the wings of the brighter component is
due to strong PSF variations in the DFOSC field (see also sec-
tion3.2).Its fluxcorrespondsto ≈0.5%ofcomponent’sApeak
flux, for the Gunni-band. For the V-band this effect is weaker,
of the order of 0.1%.
Four stars in the vicinity of UM 673 were used to construct
the PSF. A 64×64 pixel sub-frame,centered on UM673, is ex-
tracted from each DFOSC frame. Prior to deconvolution, this
sub-frame is flux-normalized,by division of the integrated flux
of the non-variable star S1 (see Fig. 1). This cancels the effect
of variable airmass and sky transparency. Star S2 is used as a
comparison, in order to check the relative photometry, to con-
trol the errors and to quantify the effect of the PSF variations
across the field. Both stars have been found to be photomet-
rically stable (Sinachopoulos et al. 2001; Nakos et al. 2003).
Unfortunately, no other bright, isolated star is present in the
field of view.
The Gunni and V fluxes of the two quasar components are
converted to standard Johnson-Cousins magnitudes using the
literature values of the reference star S2 (Nakos et al. 2003).
We notice a slight shift between the V magnitude of S2 and the
one published in the literature, probably due to differences in
the apertures used to carry out the flux normalization relative
to S1. The corrected value of the Cousins magnitude of S2 is
0.04 fainter than in Nakos et al. (2003), i.e. V = 17.74 ± 0.02
and (V − I) = 2.28.
3.2. Differential imaging
The basic concept of this method, implemented in a new code
(Python/IRAF/C++), contains elements taken from the IRAF
4Nakos et al.: Photometric monitoring of the lensed quasar UM 673
Fig.3. Cousins I-band light-curves, for component A (top), component B (middle), and for the reference star S2 (bottom), as
obtained from left to right with 1- the MCS deconvolution, 2- the Differential Imaging method, and 3- the PSF fitting method.
The dispersion between the photometric points of star S2 is indicated in each panel. The Julian Date 2451091 corresponds to
1998, October 3.
Fig.4. Cousins V-band light-curves, spanning a period of 400 days (same symbols and organization as in Fig. 3). Each point
corresponds to observations taken a few minutes before or after the I-band data shown in Fig. 3.
Nakos et al.: Photometric monitoring of the lensed quasar UM 6735
Fig.5. Observations in the V-band were carried out for a longer time than the Gunni ones, extending the light-curve up to 1200
days (same symbols and organization as in Fig. 3).
task psfmatch (Phillips & Davis 1995). It is based on the com-
parison of a set of images, which have been registered and
transformed so that their PSF and photometry perfectly match.
First, the best seeing and S/N image is identified and la-
beled as the reference image, all the other frames being called
input frames in the following. All the frames are rescaled so
that star S1 has a constant flux in all images.
The reference frame is subsequently degraded to match the
resolution of the other frames. This is done by convolving the
reference frame with a transformation kernel computed in the
Fourierspace. Each frame is then subtracted from the degraded
reference, hence providing a residual frame showing only rela-
tive flux variations. Following this procedure, non-variable ob-
jects are removed, i.e. in the present case the contamination by
the lensing galaxy is canceled.
Importantly, while this method provides accurate relative
photometry, it does not give any information on the absolute
photometric zero-point. This information is obtained by using
the photometry from the PSF fitting method (see section 3.3)
as a cross calibration. Since the PSF fitting does not take the
lensing galaxyinto account,slight shifts are possible on the ab-
solute scales adopted for the light-curves produced with MCS
and with the differential imaging or PSF fitting method.
Itis possiblein theimplementationofthecodetouse acon-
volution kernel that changes across the field of view. However,
due to the very low number of useful PSF stars in the field of
UM 673, a fixed kernel was used in the present application.
Trial and error showed that the best kernel was obtained when
star S1 alone was used for the computation. It is indeed the
bright star closest to UM 673, for which the PSF variations are
expected to be the smallest.
3.3. PSF fitting
“General” is a profile fitting technique that has already been
applied to several gravitational lens systems (e.g., Burud et
al. 1998; Østensen et al. 1997; Courbin et al. 1995). A set of
field stars is used to construct a numerical profile to be fitted to
that represent extended objects. The advantage of this method
is that it has few free parameters: the relative position and in-
tensities of the quasar components, and the shape parameters
of the lens (ellipticity, position angle). All parameters of the
point sources and lensing galaxy are fitted simultaneously to
6Nakos et al.: Photometric monitoring of the lensed quasar UM 673
Fig.6. Left Panel: difference I-band light-curve between the aperture photometry and the magnitude corresponding to the sum
of the fluxes of the two QSO components (A+B). The upper panel corresponds to the MCS algorithm, the middle panel is for the
Differential Imaging, and the lower panel is for the PSF Fitting method. Right Panel: idem for the V-band.
The stars used to build the PSF for each frame are the ones
labeled S1-S4 in Sinachopoulos et al. (2001) and Nakos et
al. (2003). As for the other two methods, the flux of star S1
is taken as a reference to normalize all epochs.
The application of the method to the present data set was
not straightforward, due to the large pixel size of the detector
that hampers the accurate determination of the lens position.
We therefore model UM 673 as two point sources only. The
consequence is that the flux we measure for component B is
contaminated by the lensing galaxy, even though the residuals
after the simultaneous fit of the two PSFs to the quasar im-
ages are good. However, even with this contamination, the to-
tal summed fluxes of the two quasar images remain compatible
with thefluxmeasuredthrougha largeapertureapplieddirectly
to the data (see Fig. 6).
The photometry of UM 673A and B, and of the reference star
S2, is displayed in Figs. 3-5. Between 3 and 6 frames were
obtained for each observing epoch and each filter. The values
plotted in the figures are the mean values for each night. The
1σ error bar for each epoch is taken as the error on this mean
value, i.e. the standard deviation between the magnitudes mea-
where N is the number of frames. The systematic errors due to
the zero-point corrections, which are of the order of 0.02 mag-
nitude, are not incorporated in the error bars.
On short time-scales, where both V and I-band data are
available (see Figs. 3 & 4), all methods agree that component
A shows a peak-to-peak variability of about 0.08 magnitude in
the I-band, and about 0.14 magnitude in the Johnson V-band.
The time scale of the variability is of the order of 350 days, for
bothfilters. ForcomponentB, wheretheuncertaintyin thepho-
tometry is larger, the short time-scale variability is at the limit
of detection, for both filters. However, the five points around
JD∼1100, placed more than 3σ away from the points between
JD∼0 and JD∼350, clearly show that, on larger time-scales,
component B has also undergone photometric variations (see
The comparison star S2 is used as an independent check
of each photometric method. It is found that the dispersions
between the photometric points (i.e. between all the epochs, as
for each photometric method.
5.1. Comparison of the methods
We have analyzed a two-band photometric data set of the
gravitational lens system UM 673 (18 epochs in Gunni, 23
epochs for the Johnson V-band), using three different photo-
metric methods. All methods clearly reveal photometric varia-
tions, both over short (i.e. ≈ 350 days) and long (i.e. ≈ 1200
days) time scales. The variability of the brighter component A
is well measured by all three methods,but is not as clear forthe
fainter image B.
Thethreephotometricmethodsareappliedto the datainde-
pendently of any prior knowledge on the geometry of UM 673,
e.g., as can be done from HST images (Leh´ ar et al. 2000). This
choice allows us to estimate the relative merits and drawbacks
of the methods,especially with respect to the systematic errors,
due to the PSF variations or to the design of the methods them-
Nakos et al.: Photometric monitoring of the lensed quasar UM 6737
Fig.7. Color (V-I) dependence of the photometric variations, for components A, B and for the reference star S2, as a function of
the Johnson V magnitudes. A clear trend is visible for the quasar image A, becoming bluer when it is getting brighter.
selves. The only internal calibrationapplied is that the absolute
flux ratio of the quasar images, as derived from the differential
imaging,is determinedby applyingthe absolute flux ratio from
the PSF fitting method. The lensing galaxy is not taken into ac-
count in the PSF fitting. Our cross-calibration of one method
where the lens is faint. However, it has an effect on the I-band
light-curves:Fig. 3 shows a systematic offset between the QSO
light-curves obtained with MCS and the other two methods.
The shift is in the opposite direction between images A and B.
This effect is not present in the curves of the reference star S2
and is probably due to the contamination of the lens affecting
the different methods in different ways. A zero-point miscal-
culation is excluded, since this would have resulted in the two
curves moving in the same direction.
PSF variation or distortiondoes not have the same effect on
the different methods. An outlier is seen, for example, in the
I-band light-curve of star S2 when using the MCS method (I =
15.53), and corresponds to a set of images with bad focusing.
This point is also an outlier in the PSF fitting light-curve and
reveals a higher sensitivity of these methods to imperfect PSF
The systematic errors are consistent among the methods,
but show a slight filter-dependence. In Fig. 6 we present the
difference curve between the total light, as measured by direct
aperture photometry of UM 673, and the sum of the fluxes of
the two QSO components, as measured by the three methods.
For the I-band (left panel), the points lie systematically above
the mean value < ∆mag >= 0, indicating that all methods
overestimate the sum of the fluxes (A+B) at the level of 2%.
For the V-band however (right panel), the spread of the points
around the mean value < ∆mag >= 0 is more homogeneous,
leaving a hint that the presence of the galaxy in the I-band has
an effect, although minor, in the photometric results obtained
by the three methods.
5.2. The time delay
The parts of the light-curves that are the best sampled cor-
respond to the 15 first measurements, spanning 120 days.
Consideringonlythis partofthe curvesandneglectingpossible
microlensing events, we have attempted to estimate the time
delay between the quasar images. Unfortunately the peak-to-
peak amplitude of the photometric variation is only 0.08 mag-
nitude in the V-bandand 0.05magnitudein I, makingit impos-
8Nakos et al.: Photometric monitoring of the lensed quasar UM 673
sible to estimate the time delay. This is also the reason why it
was decided to stop the observationsof UM 673 after 3 months
of monitoring, giving priority to other, more variable objects.
UM 673 shows larger photometric variations over longer
time-scales, but these variations do not match at all after shift-
ing the curve of component B relative to that of component A.
These long term variations are sampled, in the V-band light-
curves, with only three points at JD≈350 days and with five
points at JD≈1100 days. The two groups of points span 150
days each. The fact that the light-curves do not match on long
time scales has two possible interpretations: 1- the time delay
is longer than the length of the groups of points at JD≈350
days and JD≈1100 days, i.e. ∆t > 150 days or, 2- the varia-
tions are largely dominatedby microlensing.We will see in the
following that there is possible evidence supporting the latter
interpretation rather than the former one, also given that pre-
dicted time delays for UM 673 are of the order of a few months
(Surdej et al. 1988).
5.3. Evidence for microlensing
Although the large gaps in the light-curves of UM 673 do not
tionsobservedoverlongtimescales areindicativeof chromatic
We have plotted in Fig. 7 the V − I color index as a func-
tion of the V magnitude, for the two quasar images and for
star S2, as derived with the three photometric methods. Since
the error bars on the V-band photometry are comparable with
the error bars on the I-band points, the distributions of points
in Fig. 7 should be elliptical, with a long-axis parallel to the
the case of no correlation between color and magnitude. While
this is the observed situation for star S2, the plot for compo-
nent A of the quasar displays a clear linear trend, indicating
that it is becoming bluer as it gets brighter, in agreement with
microlensing scenarios where a compact quasar accretion disk
is (micro)lensedby a network of caustics. Microlensing prefer-
entially amplifies the most central and bluer parts of the source
be seen in the continuum of a quasar spectrum (e.g. Wisotzki
et al. 1993; Courbin et al. 2000).
This result is in apparent contradiction to the results of
Wisotzki et al. (2004), who exclude the microlensing scenario
from integral field spectroscopy, but their observations were
performed in 2002. This is three years after the latest of the
epochs of our light-curves where both V and I-band observa-
tions are available.
The reliability of the trend we observe in Fig. 7 can be
tested by comparing the V-I index obtained using one method,
to the index measured by using the other two methods. If all
methods are sensitive in the same way to a real V − I trend as a
functionof the V magnitude,then the V −I colors measured by
two different methods should correlate well. If there is no real
trend, the V − I measurements are dominated by photon noise,
and no correlation should be found. Doing this test reveals a
strong correlation between the results in the three methods for
√2 times larger than the (horizontal) short axis, in
quasar A, but not for quasar B or for the reference star S2.
The color-dependentvariations seen in the quasar image A can
therefore be safely considered as real.
Although the case of microlensing being present for both
components would be rather unusual, the color dependence on
the V-band magnitude for component B, shown by differential
imaging (Fig. 7), is quite puzzling and requires confirmation.
Given the fact that none of the two other methods confirms the
required, or an extensive high SNR spectrophotometric moni-
toring of UM 673.
If the variation we have observed is not due to microlens-
ing, the peak-to-peakvariationdetectedoverthree yearsis only
0.14 magnitude in the V-band. The photometric error on the
individual points is about 0.01 magnitude for all three meth-
ods used in this paper and the visibility of UM 673 is about
6 months across the year. According to the simulations by
Eigenbrod et al. (2005), the minimum sampling to adopt in or-
der to derive the time-delay with an accuracy better than 2%
in less than 2 years is of one point every three days, making
UM 673 a very “expensive” object in terms of telescope time,
ations detected in the present photometricmonitoring,UM 673
may however turn out to be a very interesting object for mi-
Acknowledgements. We thank IJAF and ESO for granting us ob-
serving time for this project on a flexible basis. Th.N. acknowl-
edges support from the project “Chercheurs Suppl´ ementaires aux
´Etablissements Scientifiques F´ ed´ eraux”. Part of the research was also
performed in the framework of the IUAP P5/36 project, supported by
the DWTC/SSTC Belgian Federal services. Th.N. also acknowledges
C. Abajas Bustillo for the interesting discussions. FC acknowledges
financial support Pˆ ole d’Attraction Interuniversitaire, P4/05 (OSTC,
Belgium). J.S.wishestoacknowledge support fromtheBelgianOSTC
PRODEX program “Gravitational Lensing”.
Bertin, E. 1996, A&AS, 117, 393
Borgeest, U.&Schramm, K.J.1993, inProc.1st Megaphot Workshop,
ed. U. Borgeest, K.J. Schramm & J. von Linde, 105
Burud, I., Stabell, R., Magain, P., et al. 1998, A&A, 339, 701
Burud, I., Courbin, F., Magain, P., et al. 2002a, A&A, 383, 71
Burud, I., Hjorth, J., Courbin, F., et al. 2002b, A&A, 391, 481
Courbin, F., Magain, P., Remy, M., et al. 1995, A&A, 303, 1
Courbin, F., Lidman, C., Meylan, G., et al. 2000, A&A, 360, 853
Daulie, G., Hainaut, O., Hutsem´ ekers, D., et al. 1993, Gravitational
lenses in the universe. In 31st Li` ege International Astrophysical
Colloquium, ed. J. Surdej et al., 181
Eigenbrod, A., Courbin, F., Vuissoz, C., et al. 2005, A&A, 436, 25
Leh´ ar J., Falco, E.E., Kochanek, C.S., et al. 2000, ApJ, 536, 584
Magain P., Courbin F., & Sohy S. 1998, ApJ, 494, 452
Nakos, Th., Ofek, E.O., Boumis, P., et al. 2003, A&A, 402, 1157
Østensen, R., Remy, M., Lindblad, P.O., et al. 1997, A&AS, 126, 393
Phillips, A.C. & Davis, L.E. 1995, Astronomical Data Analysis
Software and Systems IV, in ASP Conference Series, Vol. 77, ed.
R.A. Shaw, H.E. Payne, & J.J.E. Hayes, 297
Remy, M. 1996, PhD Thesis, Li` ege University
Sinachopoulos, D., Nakos, Th., Boumis, P., et al. 2001, AJ, 122, 1692
Smette, A., Surdej, J., Shaver, P.A., et al. 1992, ApJ, 389, 39
Nakos et al.: Photometric monitoring of the lensed quasar UM 6739 Download full-text
Surdej, J., Swings, J.-P., Magain, P., et al. 1987, Nature, 329, 695
Surdej, J., Magain, P., Swings, J.-P., et al. 1988, A&A, 198, 49
Wisotzki, L., K¨ ohler, T., Kayser, R., et al. 1993, A&A, 278, L15
Wisotzki, L., Becker, T., Christensen, L., et al. 2004, Astron. Nachr.,