The model which is used to describe the energy response of the pn-CCD is the so-called partial event model (see []). The model accounts for the effects of incomplete charge collection close to the detector surface. An efficiency function CCE is defined that reflects the portion of charge collected if a photon is absorbed in a certain depth. This function is folded with the probability of absorption in that depth which is given by the absorption law. The resulting analytical function for the spectral shape is then folded with the Gaussian noise distribution given by the Fano noise and the system noise (electronic and transfer noise).
Different processes are responsible for different spectral features that are visible in the detected channel distribution of infalling monochromatic X-rays. Absorption in the oxide causes detection of very few electrons, resulting in a flat shelf that extends down to the detector noise peak. If the photon is absorbed in the silicon, but close to the silicon surface, only a part of the generated electrons will reach the readout node, causing a shoulder at the low-energy side of the main peak. The variation of this portion with absorption depth is described by the CCE. If the absorption takes place deep in the detector all generated electrons will be detected. The relative strength of the flat shelf, shoulder and main peak is dependent on energy.
Monochromatic synchrotron radiation at 15 different energies ranging from 0.150 keV to 15 keV was used to calibrate the energy response of the pn-CCD detector. The resulting channel spectra were fit with the partial event model and the model parameters derived as function of energy. The detector response matrix is then filled using the partial event model with the parameters derived from the ground calibrations.
