1TIPS 2011 May
Persistence in the WFC3 IR detector
Knox S. Long
2TIPS 2011 May
Overview
Persistence is a residual image observed in most types of IR arrays
There is physical model for this persistence involving traps that are accessed by free electrons or holes when a pixel is near saturation
The WFC3 IR detectors exhibit persistence from sources that approach or exceed full well
Typically persistence results in signals of 0.3 electrons s-1 , 1000 s after a saturated exposure.
The persistence is primarily a function of the amount of saturation of a pixel and the time since saturation
Persistence decays as a power law of time (g~1)
In most cases, post-processing can remove about 90% of the persistence signal with algorithms that track the history of the stimulus in the preceding 6 hours
3TIPS 2011 May
Basic reason for persistence is understood
-- - --
+ ++++next dark exp.
(small bias reduction) The released charge reduces the bias voltage. persistence
- - -
+ + +
-- - --
+ ++++
high flux signal
(low bias)As signal accumulates the depletion width is reduced. Traps newly exposed to charge can capture some mobile carriers.
Trapped holes
Trapped electrons
- - -
+ + ++ +++
+
-- - --
reset
(large reverse bias)
At “reset” the wider depletion region is restored, but trapped charge stays behind.
Depleted
Mobile electrons
Mobile holes
dark idle
(large reverse bias)All traps have released their charge in depletion region
+ ++++
-- - --
-+P
N
R.Smith, SPIE 7021-22, Marseille 2008-06-24
4TIPS 2011 May
Persistence Examples
Easy case Look for obvious patterns Look for objects that
appear mushy
Hard case Use multidrizzle to find
residuals Subtract the last single
science image from the first
If you suspect persistence, use MAST search to see if there were IR observations that preceded yours
Ex. 1
Ex. 2
http://archive.stsci.edu/hst/history_search.html
5TIPS 2011 May
Cycle 18 Calibration - Omega Cen
3 visits all consisting of an undithered Omega Cen exposure followed by a sequence of darks
Two visits successful, and the third to be repeated in a few weeks
Dark current measured from 234 s to 5800 s after the stimulus
6TIPS 2011 May
Persistence follows a Fermi-like distribution
Note – Not individual pixels but the mean persistence at different stimulus levels. WFC3 nominal saturation value is 70,000 e
Per
sist
ence
(e
s-1)
230 s
5800 s
7TIPS 2011 May
Persistence is ~ a power law function of time
8TIPS 2011 May
Persistence Model
Where q and t are the stimulus (in electrons) and time (in seconds) N is the normalization factor (at 1000 s)
qo is the “Fermi energy” at which the stimulus is “half”
dq is the “Fermi kT” a is the power law slope correction to the Fermi function. g defines the power slope for the decay with time
9TIPS 2011 May
Results (Visit 1)
Global fits good with systematic departure near knee
“Mean value” of persistence 1.43 0.05 e/s
RMS error 0.02 0.0052 e/s
Maximum error 0.04 0.02 e/s
Visit 3 similar
10TIPS 2011 May
Persistence subtraction S/W
HST archive does not currently provide an automated way to subtract persistence from images
However, WFC3 group is providing persistence subtracted images to users on request
Implements the model described above with a set of Python scripts
Provides Persistence image
• External – generated by early visits
• Total – internal + external persistence
Corrected flt file
Original
External Persistence
Total. Persistence
Corrected
11TIPS 2011 May
Open item - Persistence is not completely uniform
Tungsten lamp Dark after Tungsten lamp
Tungsten lamp is brightest in lower right quadrantPersistence is brightest in upper left quadrant
12TIPS 2011 May
Persistence is not completely uniform
Omega Cen Dark after model subtraction
13TIPS 2011 May
Open item – Persistence may be a function of time saturated
• Cycle 18 test – turn the tungsten lamp off at mid-exposure• At 80,000 e, there seems to be a correlation of persistence with hold time• At 140,000 e, the small amount of data in hand is not easily understood
Smith et al. model predicts persistence is a function of time a pixel held at saturation
14TIPS 2011 May
Summary – Persistence is not a virtue, but …
Today IR observations prohibited
after some “bad actors” Substantial Cycle 18
calibration program underway A working model exists for
persistence• Based on prior history of
illumination
A tool exists to mitigate persistence
• Results have been provided to individual users on a case by case basis
• Tool has been tested on a large fraction of the data
Future Ready to run existing
persistence subtraction tool on al the WFC3/IR data and then to advertise the existence of the persistence images
There are additional effects that may be able to improve the model
• Persistence as a function of position
• Add effects including dependency of the decay on illumination and time at saturation