Presentazione di PowerPoint

1
An Intermediate Redshift Supernova Search at
ESO Reduction Tools and Efficiency Tests
M. Riello, G. Altavilla, E. Cappellaro, S.
Benetti, A. Pastorello , F. Patat, M. Prevedello,
M. Turatto, L. Zampieri
We used the Wide Field Imager at ESO 2.2m
telescope (La Silla, Chile) to obtain deep V, R
images of 21 fields, covering a total useful area
of 5.1 deg2. For the spectroscopic
classification we took one or more spectra for
the best SN candidates with FORS1/2 at the ESO
VLT which provides a sufficient resolution (10Å)
and appropriate wavelength range (6000-11000Å).
We want to emphasize that only the brightest
candidates where checked at the VLT for the lack
of enough observing time.
SNe CANDIDATES SELECTION Variable sources were
identified by running Sextractor on the
difference image. This gave a list which was
heavily contaminated by spurious detections due
mainly to residuals of bright saturated stars,
cosmic rays and detector cosmetic defects. In a
typical frame more than 1000 sources were found.
To improve our candidate lists we developed an
automatic procedure which attributes a score to
each detection based on several parameters (see
below) measured on the difference image, but also
on the new and reference ones. The scores were
calibrated through artificial stars experiments
and our experience with WFI frames in order that
good candidates had always a positive score. This
reduced the list to about 80-130 detections. All
candidates with positive score were verified by
direct inspection.
ISIS (by Alard Lupton) PSF MATCH WITH
SPACE-VARYING KERNEL IMAGE SUBTRACTION
IMAGE STACKING GEOMETRIC REGISTRATION IMAGE
COMBINATION REFINED ASTROMETRIC SOLUTION
PRE-REDUCTION The pre-reduction (bias, flat
fielding etc.) was performed using standard
MSCRED tasks. Whenever it was possible, standard
flat fielding was improved by constructing a
Super Flat using both the skyflats and the
science frames. The three images were registered
to a common geometrical grid, properly scaled in
intensity and finally stacked. Finally to the
stacked image was attached an accurate
astrometric calibration using on line reference
catalogues (USNO).
IMAGE SUBTRACTION The new epoch and the reference
one were first matched and trimmed in order to
overlap. ISIS software by Alard Lupton (1998)
was then used to perform the image subtraction
the reference image was convoluted with a kernel
(spatially variable in general) in order to match
as close as possible the other image. This
convoluted image was then subtracted to the
other. Best results are obtained if images have a
comparable seeing.

• SCORE ALGORITHM
• Any object located near saturated stars is
  deleted from the list.
• 60 points are initially assigned to all detected
  object.
• 60 points are gained if the object is a star
  inside a galaxy.
• 30 points are lost if the FWHM is outside the
  range ltFWHMgt /- 40.
• Further 30 points are lost if the FWHM is outside
  the range ltFWHMgt/- 80.
• 60 points are lost if the object is near a bright
  star.
• 30 points are lost if the object is faint and
  near the nucleus of a galaxy.
• 60 points are lost if there is no counterpart in
  the new image.
• 30 points are lost if the difference between a
  fixed aperture magnitude and seeing scaled
  aperture magnitude is greater than 0.5 mag.
• Further 30 points are lost if the difference
  between a fixed aperture magnitude and seeing
  scaled aperture magnitude is greater than 1.0 mag.

CANDIDATES DATABASE We developed a MySQL database
with a web user interface containing all the
reliable variable sources. To all selected
candidates we attached a provisional label SN
for candidates which looked like SN AGN VAR for
variable stars and MOV for minor planets. At each
new run, we can check if a given candidate was
already detected on previous epochs with a simple
database query. This helps to clean the list of
SN candidates from AGN and variable stars
contamination. On the right it is shown some of
the typical variable sources found on the
difference image and three of our confirmed SNe.
Each image is made of six stamps the upper three
are the three dithered images on the bottom at
left is the stacked image, at the center the
reference image and at the right the difference
image.
1.1 Laurentia Minor Planet
1.2 Variable Star
1.3 Typical Bad Detection
1.4 SN 1999gu
1.8 SN 2001ge
ARTIFICIAL STAR EXPERIMENTS In order to
accurately estimate the control time, which is
needed to derive SN rates (see the poster by
Altavilla et al.), we need to determine the SN
detection efficiency by constructing the
completeness curve as a function of magnitude. To
build this curves we performed a number of
artificial stars experiments artificial SNe
with the proper PSF and with different magnitudes
were added to a given image, which was then
searched using the standard recipe described
above. Simulated SNe were added to each galaxy of
the frame (assuming that a galaxy is a
Sextractor object with stellarity index lower
than 0.2). The distances from the host galaxy
nucleus were distributed randomly assuming a
gaussian distribution with s equal to the HWHM of
the galaxy profile. Also the position angles were
distributed at random in the range 0360
degrees. The PSF for the artificial SN was
determined for each chip of the mosaic using
Daophot IRAF package.
ARTIFICIAL STAR EXPERIMENTS THE ROLE OF
SEEING To check the dependence of the SN
detection efficiency from the seeing we run
several artificial star experiments on images of
the same field taken at seven different epochs
and with different seeing (0.65 1.32 arcsec).
The upper panels of the first two figures show
the magnitude distribution for all the sources
detected in the field (black), for the galaxies
(green) and the stars (red). The lower panels
show the detected sources in the plane (magnitude
Sextractor Stellarity Index) it appears
clearly that there is a limiting magnitude beyond
which Sextractor is not able to safely
distinguish between stellar and diffuse objects
and that this magnitude strongly depends on the
seeing condition. The third figure shows the
detection efficiency for the seven different
epochs and for each chip of the mosaic while the
last figure shows the mean detection efficiency
for each epoch, with the error bars showing the
dispersion due to the differences between the
chips. It turns out that an improvement of the
seeing of 0.67 arcsec (from 1.32 to 0.65 arcsec)
increases the limiting detection magnitude of
about 1.5 mag. The efficiency drops abruptly to
zero in about 0.5 mag this helps defining the
limiting magnitude because changing the limiting
magnitude by few tenths has a negligible effect
on the control time (see poster by Altavilla et
al.). We defined as limiting magnitude the point
where the detection efficiency is 90.
ARTIFICIAL STAR EXPERIMENTS THE ROLE OF
SEXTRACTOR THRESHOLDS To check the dependence of
the detection efficiency on the threshold used
with Sextractor to identify the SN candidates on
the difference image, we run four experiments on
the epoch with 0.90 arcsec seeing using the
following thresholds 1.5s (black), 2.0s (red),
3.0s (green), 4.0s (blue). The different
detection efficiencies obtained are shown on the
first figure on the right. As expected, the
limiting magnitude is fainter for a lower
threshold, the range spanned by our experiment is
about 1 mag. However, using a lower threshold
increases significantly the number of spurious
detections. The histograms of the right panel
show the magnitude distribution of the sources
detected by Sextractor using different
thresholds 4.0s solid histogram, 3.0s double
dashed histogram, 2.0s single dashed histogram,
1.5s empty histogram. It clearly appears that the
number of bright objects does not depend on the
threshold used while at fainter magnitudes the
number of detected objects depends significantly
on the threshold used.
CONCLUSIONS The reduction recipe, the tools for
SN search and for data mining presented here
provided a powerful framework for the
determination of the SN rate at intermediate
redshifts (see the poster An Intermediate
Redshift Supernova Search Preliminary Results
by Altavilla et al.). The artificial star
experiments performed have shown that an
improvement of the seeing of 0.67 arcsec (from
1.32 to 0.65 arcsec) increases the limiting
detection magnitude of about 1.5 mag, while
changing Sextractor threshold from 4s to 1.5s
results in a detection limit 1 mag fainter.
These tools will be enhanced in sight of the use
with data that will be available from the
forthcoming generation of wide field imagers
OmegaCAM_at_VST, WFI_at_LBT.