3 Instrument performance

The pn-CCD camera system was submitted to an intensive ground calibration programme. A detector response matrix was filled with measured data and modelled interpolations, where no data were available. Within the measurement accuracy the in-flight performance does not deviate from on-ground calibrations under the same operating conditions (Briel et al. 2000).

3.1 Quantum efficiency

The fully depleted 300 $\mu$m of silicon determines the detection efficiency on the high energy end, while the quality of the radiation entrance window is responsible for the low energy response. Figure 5 shows the result of the absolute quantum efficiency calibration at PTB (BESSY synchrotron in Berlin) and LURE (synchrotron in Orsay, Paris). All measurements were made under conditions comparable to space operation. The drop of quantum efficiency (QE) at the lowest energies is caused by the properties of the silicon L-edge. The absorption length of X-rays in silicon at e.g. 150 eV is only 30 nm. A thin oxide layer of the order of 20 nm already absorbs about one half of the incident photons. The drop of about 5% of QE at 528 eV is due to the additional absorption in the SiO₂ passivation on the detector surface. The other prominent feature in Fig. 5 is the typical X-ray absorbtion fine structure (XAFS) behaviour around the silicon K edge at 1.838 keV, enlarged in the inset of Fig. 5. At higher energies the solid line nicely fits the photon absorption data for 300 $\mu$m of silicon. The solid line is a fit to the measured data with a depletion thickness of 298 $\mu$m. The QE is not supposed to change during the XMM lifetime under nominal conditions.

3.2 Energy resolution

Figure 6: Calibration spectrum with the internal radioactive source including the background with the filter wheel in closed position. The continuous background below the Mn lines arises mainly from photoelectrons stimulated from the 55Fe source in the Al target. The iron K$_{\alpha }$ line between MnK$_{\alpha }$ and MnK$_{\beta }$ is not resolved

The energy resolution is mainly determined by the statistical fluctuations of the ionization process (Fano noise), the charge transfer properties of the CCD and the electronic noise of the readout node. Figure 6 shows about 31 hours of in-orbit data with the internal calibration source in the ``closed-cal'' position from june 2000. The signals selected are only those which hit the CCD in the last 20 lines of the 12 CCDs, the area, which also contains the focal point. X-ray events from this region have undergone the maximum number of charge transfers and therefore the highest charge losses. The AlK$_{\alpha }$, the MnK$_{\alpha }$ and MnK$_{\beta }$ and the MnK$_{\alpha }$ escape peak are clearly visible. The CuK$_{\alpha }$ and CuK$_{\beta }$ peaks are fluorescence lines from the printed circuit board, generated by ionizing particles traversing the whole pn-CCD camera. The other fluorescence lines (e.g. KK$_{\alpha }$, TiK$_{\alpha }$, VK$_{\alpha }$, CrK$_{\alpha }$, FeK$_{\alpha }$, NiK$_{\alpha }$, ZnK$_{\alpha }$) and others are trace elements in the aluminum structure of the camera and the invar ring holding the pn-CCD wafer. The energy resolution in the full frame mode is extracted from the internal calibration source including all kind of X-ray background. Over the first 9 months the peak position and the FWHM are shown in Figs. 7 and 8. At MnK$_{\alpha }$ the FWHM is 161 eV in the focal point, it is 152 eV averaged over the whole CCD and is 140 eV close to the readout nodes. The energy resolution improves in the extended full frame mode to 148 eV (FWHM) averaged over the entire chip. The AlK$_{\alpha }$ resolution is 111 eV (FWHM) for the full frame and 105 eV in the extended full frame mode. Due to the heavy overlap of many lines and because of the underlying continuous background the energy resolution is slightly better for monochromatic radiation.

3.3 Instrument stability

The long term instrument stability is checked routinely in terms of housekeeping data from all relevant camera parameters and by analysing the spectroscopic performance of the on-board calibration source. Figure 7 shows the variation of the MnK$_{\alpha }$ peak position as a function of time after launch. Within less than 1 ADU count (5 eV) all measurment points are compatible with the pre-launch data.

The strong solar flare on July-14 did not leave any measurable damage in the pn-CCD camera.

The variation of the peak position (see Fig. 7), e.g. at the day 200 and 220 in the year 2000, was due to a strong temperature drop (approximately 10 K) of electronic boxes outside of the camera housing, causing gain changes and shifts in the analog-to-digital conversion. The peak shift is correlated with satellite commands influencing the thermal budget. On ground, the change of the peak position is corrected. As can be seen in Fig. 8 the peak shift had no effect on the energy resolution. The error bars in Fig. 8 comprise only the statistical errors, systematics are not included. The larger error bars in CCD 10, 11 and 12 reflect the fact, that this area of the focal plane was only poorly illuminated by the calibration source. The variance of the FWHM at the MnK line including 36 extended calibration measurements over the last 9 months is only 1.8 eV. A peak shift due to CTE changes because of possible radiation damage is less than 5 eV.

3.4 Instrument background

As can be seen in Fig. 6 two other features of the spectrum need some explanation. (a) The continuous background from the lowest energies up to the MnK lines. This background is due to photo electrons from the Al fluorescence target, excited from the MnK X-rays from the calibration source. Because of the very thin radiation entrance window of the pn-CCD the low energy electrons can be clearly detected with high QE. This property is equally responsible for the high QE response for soft X-rays. (b) The flat background distribution for the highest energies arises from Compton electrons, generated by X- and gamma rays.

Figure 7: Longterm stability of the pn-EPIC camera system monitored with the internal calibration source. Within a precision of 5 eV no CTE related peak shifts were observed, given nominal operation temperatures in the pn-CCD camera system

Figure 8: The full width at half maximum of the internal calibration source. The larger errors in CCD's 10, 11 and 12 are due to the inhomogeneous focal plane illumination resulting in a reduced count rate

Another source of instrument background is caused by highly ionizing particles, being imaged by means of grazing incidence reflection through the X-ray telescope. They can be light and heavy ions as well as highly ionizing low energy protons (Aschenbach 2001). The duration of the flares can be of the order of minutes up to hours, their occurence is unpredictable (see e.g. Briel 2000). In case the detectors register a significant increase of counts, the filter wheel is put in closed position, but the EPIC instruments remain operational.

The pn-CCD camera has the option to lower the gain of the signal processing electronics by a factor of 20 to increase the dynamic range in the so-called low-gain mode above 300 keV. This mode is very useful to study background phenomena. Figure 9 shows an example of such a measurement. During an observation in low-gain a sudden increase in count rate by a factor of 2.5 occurred in the CCD cameras without getting notice from the radiation monitor. The above threshold counter indicated an increased number of particles. The result of the analysis of those ``soft proton flares" were summarized by Strüder et al. (2000):

1.

The energy distribution of the protons has its maximum at the lowest measured energies at 1 keV with an exponential attenuation of 4 orders of magnitude after 250 keV. In this measurement, 50 keV of the proton's energy was already absorbed by the thick filter;

2.

The protons show a clear vignetting proofing that they have been imaged through the telescope. We do not find the low energy protons in the 6 cm² sensitive area outside the field of view. In contrast to minimum ionizing particles, the protons mainly produce single pixel events;

3.

The protons lose typically 50 keV of energy in the thick filters and about 20 keV in the thin filters;

4.

The soft proton flares heavily load the observational background (and the satellite telemetry) and therefore disturb and limit the observations, but they do not damage the instrument.

The lower (red) curve in Fig. 9 exactly represents the simulated spectrum of isotropically distributed mips crossing the pn-CCD's pixel structure. The difference of the lower and upper spectrum yields the low energy proton spectrum.

Figure 9: Measured background spectra (thick filter) with a reduced gain (factor 20) during normal background, mainly mips (red line) and during a ``soft proton flare" (black line). The upper panel shows the ``light curve'' of the observation with an increased background in the last 500 s. The lower left picture shows the event pattern of the high and low background measurements. The lower right diagram shows the difference in energy deposition for the low and high background events

Minimum ionizing particles (mips) traverse the detector and leave about 80 electron-hole pairs per $\mu$m track length in the silicon. In a Monte-Carlo simulation we assumed a 4$\pi$ isotropic distribution of the mips, which nicely fits the measured data: The average energy deposition in one pixel is in the order of 50 keV and the average number of pixels involved in a mip track is about 10 (Strüder 2000). The most probable track length in one single pixel is 150 $\mu$m. The onboard processor is able to remove almost 100% of the mips. The processing power onboard is not sufficient to remove them all. Post-processing on ground then rejects 100% of the mips.

The instrument background was determined by measurements with the filter wheel closed. In the energy band of 2 keV to 10 keV 4.5 10^-4 single events per sec, per keV and per arcmin² were measured. For a circle with a radius of 7 arcsec - i.e. the half energy width of the telescope system - this reduces to 1.5 10^-4 per sec. The cosmic X-ray background within the 7 arcsec radius is 2.3 10^-5 singles per sec. A circle with a radius of 7 arcsec characterizes the half energy width of the telescope and therefore gives a background estimate for a point source.

The analysis of the radiation background of all kinds will be an ongoing activity because of the lack of precise predictability, time transients and missing knowledge about the composition, energy profile, spatial distribution in the XMM orbit.