8.1 Persistence in WFC3 IR

Image persistence, a phenomenon commonly observed in HgCdTe IR detectors, is an afterglow of earlier images that in the case of the WFC3 IR detector is present when pixels are exposed to fluence¹ 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 hours.

A very obvious example of persistence is shown in Figure 8.1. An image of a high galactic latitude field, the observation was taken to search for the optical counterpart to a γ-ray burst. However, two visits from separate IR programs had preceded the observation of this field. The afterglow of the bright sources in these dithered observations is clearly visible as 5 point line patterns in Figure 8.1.

Pixels in the WFC3 IR array and other HgCdTe IR detectors are operated as reversed biased diodes. Resets increase the reverse bias. Subsequently, when light falls on the detector, the resulting electron-hole pairs reduce the bias. Persistence is understood to arise from imperfections, or 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 amount of charge accumulated in the diode. A small percentage, of order 1%, of the free charge (WFC3 ISR 2013-07) is captured by the traps 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, however they do not affect the charge that has already been trapped. A detailed theory of persistence has been presented by Smith et al. (2008a, 2008b) and Long et al. (WFC3 ISR 2013-07).

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. This signal decays with time as a power law with a slope of about -1. Thus after 10,000 s, the persistence signal will be about 0.03 e¯/sec, just under 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. The shape of the curve reflects 1) the density of traps in different regions of the pixels and 2) the fact that once the detector is saturated the voltage levels within the diodes do not increases sharply with increasing fluence. The right panel of Figure 8.2 shows the power law decay of the persistence at different fluence levels.

A further complexity to modeling the persistence is that persistence actually depends not just on the total fluence, but on the complete exposure history since the traps have finite trapping times (WFC3 ISR 2013-06, WFC3 ISR 2013-07). 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 because in the second case the traps have more time to capture free electrons before the diode gets reset. While tracking the complete history of each WFC3/IR pixel is a huge task, to a good level of approximation it is possible to model persistence as a function of only total fluence and total exposure time in the stimulus image. This model, which is currently used by the WFC3 team to predict persistence and derive the persistence products (see Section 8.2) is parameterized as:

where t is the time, in seconds, since the end of the stimulus exposure and A and γ are functions of both exposure time and fluence level in the stimulus exposure. We refer to this model as the “A-γ” model (WFC3 ISR 2015-15).

Additionally, clear evidence of spatial variation in persistence across the IR detector has been measured (WFC3 ISR 2015-16). One quadrant (upper left) 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 is available from MAST (Version 3.0.1 of the persistence software). 

Persistence of the magnitude (and impact) seen in Figure 8.1 is rare. In part, this is because the STScI contact scientists check all Phase II submissions to identify any programs that are likely to cause large amounts of persistence so that the mission schedulers can insert 2 WFC3/IR-free orbits before executing another WFC3/IR program. However, this process is only intended to identify the worst cases. A large fraction of WFC3/IR exposures have some saturated pixels and all of these pixels have the potential to generate persistence in a following IR observation. Inhibiting IR observations of 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.

---

¹Fluence here is expressed in electrons. We use this nomenclature, however, with some “abuse-of-notation” regardless of whether or not the pixel's full-well capacity is reached. Therefore when values larger than the typical ~80,000 e¯ full-well capacity for the WFC3/IR channel are reported throughout this Chapter, their meaning is not that of “detected” electrons, but rather that of “electrons that would have been detected for an infinite full-well capacity”. This number is basically proportional to the impinging photon flux multiplied by the exposure time.