Putting the electrons back where they belong

Transcription

Putting the electrons back where they belong
THE NEW CALACS!
Pixel-based CTE Correction of ACS/WFC:!
CTE Losses for the Smallest Pixel Packets!
Jay ANDERSON and the ACS Team!
Space Telescope Science Institute, Baltimore MD!
ABSTRACT!
In our recent PASP article (Anderson & Bedin 2010, AB10), we show that it is possible to use warm pixels (WPs) in dark exposures
as delta-function experiments to probe how electrons are trapped and released as charge is transferred down the detector. The
model has shown considerable success in correcting science images for CTE (charge-transfer efficiency) blurring. Unfortunately, the
standard 1000s dark exposures have so many bright WPs that the trails of the fainter WPs (those with less than about 50 e-) overlap,
and it is not possible to calibrate the trail profiles. Here we illustrate how shorter darks can be used to fill in this missing information.!
Putting the electrons back where they belong!
INTRODUCTION!
THE NEED FOR DARKER DARKS!
 Massey et al. (2010) developed a pixel-based correction for
CTE from a study of WP trails in COSMOS science images.
Unfortunately the high background of the images (65 e-) did
not allow charge-transfer losses of small electron packets to
be explored.!
 Anderson & Bedin (2010, AB10) pushed this technique
further, studying the WPs in the standard 1000s dark
exposures; this enabled them to study much smaller electron
packets on much lower backgrounds.!
 The AB10 model is based on a purely empirical!
parametrization of how charge is trapped and released as it
is shuffled down the detector.!
 The delta-function WPs allow us to construct a forward
model of the charge-transfer process.!
 Unfortunately, the standard dark exposures are not completely dark. The image on the left below shows that the standard ~1000s darks are crowded
with WPs of more than 50 e-. All five images show a region far from the readout register and have a grayscale from -5 e- to 50 e-.!
 By taking a variety of shorter darks, we can isolate the fainter WPs and directly probe the interplay between charge-packet size and CTE losses.!
lost from
WP
gained by downstream pixels
some electrons are
a WP generated
delayed by traps that are the resulting during 1040s distribution
encountered dark exposure
of charge
during charge transfer
through the pixel grid
survive
intact to readout
register
net change
1040s dark
(corrected)
1040s dark
(uncorrected)
THE STRATEGY!
 With a forward model in hand, we can then construct a
reverse model to predict the original distribution of pixel
values from an observed distribution. This involves an
iterative deconvolution.!
 The figures below show a portion of a 1040s dark
exposure far from the readout register. The grayscale is
linear from 0 to 500 e-.!
Observed dark
CTE-corrected dark
 The same correction can be applied to science images.
We find that photometry and astrometry are generally
restored, although some noise amplification cannot be
avoided. The grayscale below is linear from -10 to 50 e-.!
Observed science image
CTE-Corrected science image
WP  The CTE-corrected long darks give us a robust estimate of the dark rate
in each WP. This allows us to estimate, for darks with various exposure
times, the number of electrons contained in each WP before the charge
began its shuffle down the detector. We can then compare the observed
number of electrons against the expected number to evaluate absolute
CTE losses for extremely small charge packets.!
 The table to the right documents the surviving charge for the specific
WPs identified above.!
 The plot below the table shows the results graphically, along with the
overall trend seen for many tens of thousands of WPs.!
THE BAD NEWS!
 CTE losses are truly pathological for small e- packets on low
backgrounds. WPs that start their journey at the top of an empty chip
with less than 20 e- are often blurred completely out of existence during
read out. !
 Such devastating losses could never be reconstructed with a pixelbased scheme.!
THE GOOD NEWS!
100s dark
(uncorrected)
339s dark
(uncorrected)
1040s (cor) 33s dark
(uncorrected)
Rate 1040s (obs/exp) 339s (obs/exp) -­‐
(e /s) 100s (obs/
exp) 33s (obs/exp) 1 187 0.18 133±12 / 187 33±6 / 61 8±5 / 18 5±4 / 6 2 559 0.54 380 / 559 88 / 182 14±6 / 56 2±4 / 18 3 605 0.58 416 / 605 99±11 / 197 16±6 / 58 7±5 / 19 4 643 0.62 443 / 643 141 / 210 27 / 64 6±4 / 20 5 1281 1.27 1005 / 1281 260 / 430 41 / 128 5±4 / 42 6 4971 4.78 4349 / 4971 1240 / 1620 278 / 497 62±9 / 158 7 37241 35.81 35369 / 37241 12165 / 12140 3408 / 3724 1088 / 1182 100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
Surviving Electrons
traps
within
traversed
pixels
 Virtually no ACS science observations have such low backgrounds. !
 ACS observes most efficiently in the visible bands with wide-band !
10
filters and long exposure times, all of which combine to give backgrounds !
greater than 65 e- per pixel in all but a handful of prime science exposures. !
These high typical backgrounds shield even faint sources from extreme !
CTE losses: most sources retain more than 60% of their original counts.!
 This study has allowed us to develop an exquisite model of how the traps in the pixels
impact charge packets of various sizes, from very small to very big. With such a wellcalibrated model, we can then infer what fraction of electrons should survive the trip to the
readout register for sources of different size on different backgrounds. With a better
model, we can do a better job reconstructing the original pixel distribution.!
Observed trend
for all WPs
20
50
100
200
500
1000
2000
5000
10,000
20,000
WP pixel value at top of chip (e-) REFERENCES!
 Anderson, J. & Bedin, L. R. 2010 PASP 122 1035–1064!
 Massey, R. et al. 2010 MNRAS 401 371 !