4.3 Glitches

Glitches are the result of an energy deposit from charged particles on the detector. This energy deposit is spatially localised on the detector and it takes a certain period of time for the detector to recover from it. These glitches were one of the main limitations in the sensitivity of the ISOCAM LW detector. Several papers describe the nature of the detected gliches and the glitch rates (Claret & Dzitko 1998, [15]; Dzitko et al. 2000, [30]; Claret et al. 2000, [17]); and Claret & Dzitko 2001, [16]). The active zone of SW pixels is very thin (less than 10 $\mu$m) so that the SW detector has a low susceptibility to radiation effects. The LW detector, instead, has a 100 $\mu$m pitch and a much larger thickness of 500 $\mu$m leading to a much larger amount of deposited energy. Below we will discuss the main results and conclusions of the investigation of glitches affecting the LW detector. Responsivity variations were a major challenge in maximising the sensitivity of the LW array. They could be induced by strong changes in the incident flux or by glitches. Glitch induced variations were manifested in two rather distinct problems:

1. a systematic responsivity variation after the perigee passage due to the very high radiation dose from trapped particles in the van Allen belts, and
2. (sometimes strong) responsivity variations due to the impact of individual galactic cosmic ray particles, all along the ISO orbit.

Extensive radiation tests were performed on the ground before launch using gamma-ray, proton and heavy ion accelerator beams to simulate the conditions in the van Allen belts. High ionising radiation flux induced a responsivity increase which relaxed in a few hours. This effect was minimised if the photoconductor was under bias and exposed to a high infrared flux. Thus, in-orbit, during the perigee passage, since the experiment was switched off, a specific power supply kept the necessary bias voltage on the photoconductors, and the camera was left open to light to permit detector curing by the background infrared flux. Outside the van Allen belts the main responsive perturbation came from galactic protons and $\alpha$-particles. In addition to these external particles, the anti-reflection coating of the lenses contained radioactive thorium, which generated a dose of low energy $\alpha$-particles. This flux depended on the solid angle of the lens as viewed from the array, and had a maximum for the 12 $^{\prime \prime }$ lens. Less frequent, but more disruptive, were incoming heavy ions. Each ion typically affected about 50 pixels, and generated a glitch followed by a decrease of responsivity. In-orbit, the typical glitch rate and related numbers for ISOCAM detectors outside the van Allen belts were:

• Average glitch rate = 1 glitch/second
• Average number of pixels hit per glitch = 8 pixels/glitch
• The sensitivity loss was $\approx$ 2%, $\approx$ 5% and $\approx$ 10% for images of 2, 5 and 10 seconds on-chip integration times respectively (but the actual sensitivity loss could be higher due to glitch tails).

There was no clear variation in the glitch rate along an orbit, except just after or before the passage through the radiation belts (i.e. at the beginning or at the end of the scientific window). The glitch rate only slightly increased over the ISO operations (1995-1998). Solar activity could cause the variations observed as confirmed by the solar flare event which took place on revolution 722 in which the glitch rate became extremely high (with an increase by more than a factor of 7). Three main families of glitches were defined by Claret & Dzitko 1998, [15]. Examples of these can be found in Figure 4.6.

• Type-A, called common glitches (more than 80%). These are quite easy to detect and remove from the data.
• Type-B, called faders; the pixel value decreases until a stabilised value is reached.
• Type-C, called dippers. For this type of glitches the pixel value first decreases below the stabilised value, and then increases until the stabilised value is reached.

The interpretation given in Claret & Dzitko 2001, [16] is that common glitches are induced by galactic protons and electrons, faders are induced by light galactic ions, and dippers are caused by particles providing higher linear energy transfer, such as heavy galactic ions. The CAM Interactive Analysis package (see the ISOCAM Interactive Analysis User's Manual, [28]) contains several glitch removal methods. The different methods and their performance are discussed in Claret & Dzitko 1998, [15]. They give very good results for common glitches (type-A), but other type of glitches are more difficult to remove from the data. The temporal profiles of faders and dippers have some similarities with the temporal variations which are observed after a strong change of the incident flux, which also leads to a transient behaviour of the detector (Section 4.4). For example the gain variation of dippers (type-C glitches; see Figure 4.6) could be interpreted as an increasing signal after a flux change by a transient correction algorithm, leading to a false source detection. There is no method fully reliable and several methods should be used successively in order to get a nearly 100% glitch rejection (Ott et al. 2000, [48]), especially when the data contain strong glitches.

Figure 4.6: The three main glitch families as defined in Claret & Dzitko 1998, [15]. Type-A (common) glitches have a decay time roughly as short as the rise time. For type-B (fader) glitches, the decay time is much longer than the rise time and has an exponential profile. For type-C (dipper) glitches, the detector gain is affected and the nominal sensitivity is recovered only after several readouts. Type-C glitches do not necessarily have the largest amplitudes.