ESA ESTEC Electrical Engineering Electromagnetics and Space Environment Home  



Ray-tracing Monte Carlo simulations of the ISOCAM LW CCD detector

One of the two ISOCAM detectors (Long-Wave, or LW) contains a 32 x 32 pixel Si:Ga CCD with dimensions of 3.2 x 3.2 x 0.5 mm3. The sensitive volume of this CCD is continually traversed by high-energy cosmic rays (mainly protons, but also heavier ions) penetrating the spacecraft, depositing charge along their track, and occasionally creating secondary particle showers both within and outside of the instrument.

Along with a number of other possible interference mechanisms, these primary and secondary sources can briefly illuminate a pixel or a group of pixels in the CCD, overwhelming the photon response from the real observation target. As a consequence, spurious "glitches" in the detector response are seen.

The simplest way to simulate this phenomenon is to assume an isotropic source distribution for the incident cosmic particles, ignore physical interactions in the spacecraft and in the instrument, and follow the purely mathematical, straight tracks of the impinging particles through the CCD. Such non-interactive rays will lead to a certain track length distribution, that in turn can be translated into an affected pixel number distribution using the so-called "taxi metric". In this approach, it is as a first approximation assumed that only those pixels directly coinciding with the track are affected (in reality, especially for the heavier incident ions, the amount of charge liberated is so large that it can spread to the surrounding pixels). The simulated distribution can consequently be compared with real glitch data.

We have done such a ray-tracing simulation using the GEANT-4 Monte Carlo package from CERN. The plot below shows a comparison between the simulation results for 105 incident particle entries (dashed line) and the CGLITCH glitch data set we obtained from Vilspa (solid line).

Please note the following comments on these curves.

  1. There is difference in the low-number (1-3 pixels affected) part of the distribution between the data and the simulation. Possible, but at this point not certain, explanations for this difference can be cosmic ray-induced low-energy secondary particles not considered in the simulation; fake counts in the ISOCAM deglitching algorithms; and/or counts of a-particles emitted from the Thorium coating in the lens system of the instrument. In decaying, Th232 produces a-particles of energy ~4 MeV, which will be stopped in less than 20 mm in Silicon. Since one CCD pixel has a cross-sectional area of 10 x 10 mm2, it is conceivable that an excess of glitches with a low number of pixels affected could be produced via this mechanism.
  2. Based on the Monte Carlo simulation, the mean number of pixels affected was calculated to be ~8.4, whereas the CGLITCH data set yields a value of ~9.0. However, these figures cannot be directly compared, since the simulation gives the number of pixels affected for each individual ray/particle, whereas the CGLITCH data contains the number of pixels affected per given integration time. In that integration time more than one particle may have penetrated the CCD, such that the total number of pixels affected is in this case rather the sum of pixels affected per each incident particle. Moreover, secondary particle production has not been considered in the simulation. The normalisation between the two curves is here chosen to be at 10 pixels affected.
  3. There is a knee in the simulated distribution that is not visible in the CGLITCH data. The knee is a result of the fact that the chord length distribution within a rectangular volume, although being continuous, is not a smooth one. It is unclear why this feature is not visible in the data: again, it is possible that secondary particle production smoothes the distribution. An important aspect is also the geometry of the local shielding around the CCD. This may have an effect on the directionality of the incoming radiation, and hence on the particle track length distribution.
  4. Since the simulated distribution is purely a mathematical one, it ends at 64 pixels affected. This corresponds to the maximum 2-dimensional particle track length from one corner of the CCD to the diagonally opposite corner (the z-coordinate, or the depth, of the entrance and exit points is not in this case relevant).
  5. The measured distribution naturally does not stop at 64 pixels, but continues up the maximum of 1024 pixels. There can be several mechanisms responsible for cases where several hundreds of pixels are simultaneously affected, most plausible one being a cosmic ray-induced nuclear event in the immediate vicinity of the CCD, particularly in the high-Z shielding surrounding the detector. In such a case, an extensive shower of secondary particles can be created, covering large portions, or the whole, of the CCD. An energetic primary heavy ion may also deposit enough energy to saturate the device. Note, however, that these events are several orders of magnitude less frequent than the ones where a few or some tens of pixels are involved.

For questions, comments, and further information, contact P. Nieminen at ESA/ESTEC.



Copyright 2000-2007   European Space Agency. All rights reserved
Please report bugs and send comments by clicking