WFC3 Data Handbook

8.1	Image Persistence

Image persistence is a phenomenon commonly observed in HgCdTe IR detectors. It is an afterglow of earlier images that in the case of the WFC3 IR detector is present when pixels are exposed to fluence levels greater than about 40,000 electrons. In cases where portions of the detector are heavily saturated in the initial image, the afterglow can be detectable at levels comparable to the background for several orbits.

A very obvious example of persistence is shown in Figure 8.1. The image shown is of a high galactic latitude field; the observation was taken to search for the optical counterpart to a γ-ray burst. However, two observations of a bright field near the galactic plane had preceded the image shown here by 2 to 5 hours, and the imprint of those dithered observations is clearly visible in current exposure.

Figure 8.1: Persistence in an IR image.

Example of an IR image, ia21h2e9q, in which persistence due to earlier visits is very obvious. The amplitude of the persistence for the brightest pixels in the diagonal trails of stars is about 0.1 electrons/sec The image is plotted on a linear scale from 0.7 to 0.9 electrons/sec

The pixels in the WFC3 IR array and other HgCdTe IR detectors are operated as reversed biased diodes. Resets increase the reverse bias. Electron-hole pairs created when light falls on the detector reduce the bias. Persistence is understood to arise from imperfections, traps, within the detector pixels that are exposed to free charge as the bias is reduced. The number of traps exposed is determined by the fluence, the total amount free charge released in the exposure. A small percentage of order 1% of the free charge (Long, Baggett & MacKenty 2013b) is captured on the imperfections and released later, creating the afterimages known as persistence. Resets prevent more charge from being trapped, because they remove free charge from the location of the imperfections, but do not affect the charge that has already been trapped. A detailed theory of persistence has been presented by Smith et al. (2008a and b).

Different HgCdTe IR detectors have different persistence characteristics. In WFC3, a pixel exposed to an effective fluence level of 105 electrons produces a signal of about 0.3 e-/sec 1000 s after the exposure. The signal decays with time as a power law with a slope of about -1. Thus at 10,000 s, the flux will be about 0.03 e-/sec, compared to the dark current of 0.048 e-/sec (median). As shown in the left panel of Figure 8.2, the amount of persistence in the WFC3 IR detector depends strongly on the fluence of the earlier exposure. This shape of the curve reflects the density of traps in different regions of the pixels (and the fact that once the detector is saturated the voltage levels within the diodes do not change much with increasing fluence). The right panel of Figure 8.2 shows the power law decay of the persistence at different fluence levels.

Figure 8.2: persistence as a function of stimulus and time

a) Left. Persistence as a function of stimulus in a series of darks after an observation of the globular cluster Omega Cen. b) Right. The persistence decay as a function of time. (Long et al 2012)

Persistence can be characterized approximately in terms of the following equation:

where x and t are the fluence and the time since the end of the exposure causing the persistence. The other parameters and their nominal values are provided in Table 7.1 (see also Long et al 2012). An improved model of persistence incorporates the effect of exposure time, and is described in WFC3 ISR 2015-15. The new parametrizations have been incorporated into the software package used to predict persistence in all WFC3 IR observations.

Table 8.1: Model variables and typical values (as currently used in MAST).


Variable	Definitions	Typical Value
No	The position-dependent normalization factor in e s\ :superscript:`-1`\ at 1000 s	0.30
xo	The mid-point of the sharp rise in persistence measure in e-	90,000
δ x	Variable describing the sharpness of the rise	20,000
α	Power law exponent describing the rise in rate above saturation	0.17
γ	Power law exponent characterizing the decay in persistence with time	1

Additionally, clear evidence of spatial variation in persistence across the IR detector has been measured. One quadrant has a higher persistence amplitude than the other three. The shapes of the power law exponents also appear to differ between quadrants. Using a correction flat provides a factor of two reduction in the peak to peak uncertainties. This flat is incorporated into the persistence prediction software and available from MAST (Version 3.0.1 of the persistence software). A full description is in WFC3 ISR 2015-16

The WFC3 team has developed a series of tools which allows us to predict the amount of persistence in IR images obtained with WFC3. It currently uses this formula. The model is not a complete description of persistence in WFC3. It ignores the differences in the power slopes evident in Figure 8.2. It also ignores the fact that persistence actually depends not just on the total fluence, but on the complete exposure history. The traps have finite trapping times (see Long, Baggett & MacKenty 2013a & 2013b). A short exposure of a source that results in a fluence of 105 electrons produces less persistence than a longer exposure of a fainter source that reaches the same fluence level. Progress toward a better model of persistence in WFC3 is described by Long (2013), and this may be incorporated into to the tools that the WFC3 team has developed for modeling persistence. But at present the model being used for predicting persistence in an image is the one described above.

Figure 8.3: MAST search screen used to retrieve information about persistence.

Persistence of the magnitude (and importance) seen in Figure 8.1 is rare. This is in part because contact scientists check phase II submissions to identify programs that are likely to cause large amounts of persistence and mission planners inhibit WFC3 IR observations for 2 orbits after such observations. However, this process is only intended to identify only the worst cases of persistence and the process is not error free. A large proportion of the exposures taken with WFC3 have some saturated pixels and all of these pixels have the potential to generate persistence in the next observation in the schedule. Inhibiting IR observations after all exposures that could generate persistence would make it impossible to schedule the large numbers of IR observations that are carried out with HST, and in most cases, small amounts of persistence do not affect the science quality of the data, as long as observers and data analyzers take time to examine their IR images for persistence.