The principle of sideward depletion in high resistivity silicon is the basis of a large variety of novel silicon detectors, such as silicon drift detectors, controlled drift detectors, active pixel sensors -- and pn-CCDs.

The angular resolution of the XMM X-ray telescope in front of the pn-CCD camera (mirror flight model 2, FM2) is 15 arcsec half energy width (HEW) at 1.5 keV and 8 keV. This translates to 540 $\mu$ m position resolution required in the focal plane. For a given telescope performance the concept of sideward depletion allows for an optimum adaption of the pixel size to the X-ray optics, varying from 30 $\mu$ m up to 300 $\mu$ m. The FWHM of the point spread function (PSF) is 6.6 arcsec. A pixel size of 150 $\mu{\rm m}\times 150~\mu$ m (4.1 arcsec) was chosen, with a position resolution of 120 $\mu$ m, resulting in an equivalent angular resolving capability of 3.3 arcsec. The energy response is higher than 90% at 10 keV because of the sensitive thickness of 300 $\mu$ m. The low-energy response is given by the very shallow implant of the p⁺ back contact; the effective ``dead'' layer is of the order of 300 Å (Hartmann et al. 1997). High time resolution is a consequence of the parallel readout of 64 channels per subunit; in total 768 channels for the entire camera. High radiation hardness is built in by avoiding active MOS structures and by the fast transfer of the charge in a depth of more than 10 $\mu$ m below the surface. For low energy protons, imaged through the X-ray optics (Aschenbach 2001) the pn-CCD is ``self shielding'', because the ionizing radiation has to propagate through 290 $\mu$ m of silicon before damaging the transfer channel and decreasing charge transfer efficiency (CTE). As there is only a negligible transmission of protons through the X-ray optics above 500 keV, there is no problem for the pn-CCD with low energy protons at all. Measurements in a proton accelerator by Meidinger et al. (2000) with up to 2 10⁹ 10 MeV protons per cm², equivalent to 4 times the expected 10 year XMM irradiation in space, only showed a degradation of about 30 eV in the FWHM of the MnK $_{\alpha }$ line. Kendziorra et al. (2000) tested the pn-CCDs with low energy proton up to 1.4 10⁹ protons per cm². No change of the detector's properties was seen. This proton irradiation at energies between 1 keV and 300 keV with prominent peaks at 70 keV and 170 keV, was a factor of 1000 above the expected low energy proton flux in orbit. Up to now, no measurable degradation due to radiation damage was found in orbit.

$\begin{figure} \includegraphics[width=8.6cm,clip]{XMM35r_ccdinq.eps}\end{figure}$

Figure 1: Inside the pn-CCD. The X-rays hit the device from the backside (bottom). The charges are collected in the electron potential minimum 10 $\mu$ m from the surface having the pixel structure. After integration, they are transferred to the on-chip amplifier. Each CCD column is terminated by an on-chip JFET amplifier

2.2 The basic principles of pn-CCDs

The schematic view into the pn-CCD in Fig. 1 already introduces intuitively the advantages of the concept: X-rays hit the detector from the rear side. In case of an X-ray interaction with the silicon atoms, electrons and holes are generated. The average energy required to form an electron-hole pair is 3.7 eV at $-90~^\circ$ C. The strong electric fields in the pn-CCD detector separate the electrons and holes before they recombine. Signal charges (in our case electrons), are drifted to the potential minimum and stored under the transfer registers. The positively charged holes move to the negatively biased back side, where they are ``absorbed". The electrons, captured in the potential wells 10 $\mu$ m below the surface can be tansferred towards the readout nodes upon command, conserving the local charge distribution patterns from the ionization process. As can be seen in Fig. 1, each CCD line is terminated by a readout amplifier. The focal plane layout is depicted in Fig. 2. Four individual quadrants each having 3 pn-CCD subunits with a format of $200 \times 64$ pixel are operated in parallel.

$\begin{figure} \includegraphics[width=8.8cm]{XMM35_border_scheme.eps}\end{figure}$

Figure 2: Overview of the internal boundaries of the pn-CCD focal plane. The division of the focal plane in subunits was made because of redundancy reasons. The focal point of the X-ray telescope is in CCD0, quadrant 1. About 97% of the telescopes field of view is covered by the focal plane. About 6 cm² of the CCD's sensitive area are outside the field of view and is used for background studies. The world's largest X-ray CCD with a sensitive area of 36 cm² was fabricated in the MPI- semiconductor laboratory

$\begin{figure} \includegraphics[width=8.8cm]{XMM35_win-modes.eps}\end{figure}$

Figure 3: Operating modes of the pn-CCD camera a) Full frame and extended full frame mode, b) Large window mode, c) small window mode and d) timing mode and burst mode

**Table 1:** Parameters of the six standard readout modes as implemented in-orbit. The energy flux in the last column^* was derived in the energy band from 0.1 keV to 10 keV with the assumption of an unabsorbed power law with photon index 2.0, observed with the thin filter
mode	field of view (FoV)	time resolution	out of time	life time	brightest point source
		resolution	(OOT) events	with OOT events	for XMM
	in pixel format	in ms	in %	in %	in counts per sec
	in arcmin				in erg cm^-2 s^-1 ^*
full frame	398 $\times$ 384	73.3	6.2	99.9	6
(1)	27.2 $\times$ 26.2				8.1 10^-12
extended full frame	398 $\times$ 384	199.2	2.3	100	for extended sources
(2)	27.2 $\times$ 26.2				only
large window	198 $\times$ 384	47.7	0.15	94.9	9
(3)	13.5 $\times$ 26.2				1.2 10^-11
small window	63 $\times$ 64	5.7	1.1	71.0	104
(4)	4.3 $\times$ 4.4				1.4 10^-10
timing	199 $\times$ 64	0.03	100	99.5	4000
(5)	13.6 $\times$ 4.4				5.9 10^-9
burst	20 $\times$ 64	0.007	depends on	3.0	60000
(6)	1.4 $\times$ 4.4		PSF		8.1 10^-8

The spatially uniform detector quality over the entire field of view is realized by the monolithic fabrication of 12 individually operated $3\times 1$ cm² pn-CCDs on a single wafer (see Fig. 2). No inhomogenities were observed in the tested energy range from 700 eV up to 8 keV, the measured flatness of the homogeneity measurements was always limited by Poisson statistics. Figure 2 shows the insensitive or partially sensitive gaps in between the different CCDs and quadrants. As all CCDs are monolithically integrated on a single 4 inch wafer, the relative adjustment of the chips, i.e. all pixels, is known with a precision of better than 1 $\mu$ m.

The improvement of the energy resolution of the detector requires cooling, to suppress the thermally generated leakage current. We have chosen a temperature of $-90~^\circ$ C, reducing the leakage current to less than 0.1 e^- per pixel and per readout cycle of 73 ms. Taking into account the residual partial pressure inside the camera of less than 10^-5 mbar, formation of monolayers of e.g. ice on the radiation entrance window should not occur.

2.3 Operating modes

CCDs have originally been designed for photon intensity imaging, not single photon counting in a spectroscopic mode. To make CCDs useful for X-ray imaging and spectroscopic applications simultaneously, they must be operated such that only one X-ray photon hits the detector without an overlap in time and position of another photon. The design of the readout modes was driven by the assumption that the local photon flux should be below $\frac{1}{40}$ events per pixel and integration time. To adapt the X-ray camera readout mode to the point source brightness, the integration time of the CCD camera can be shortened - by reducing the area to be read out. Under the cost of sensitive area the photon flux can be increased. Figure 3 shows which part of the CCD array is read out in the different modes. Table 1 contains the most important parameters of the pn-CCD readout modes. A detailed summary was given by Kuster et al. (1999).

2.3.1 The full frame and the extended full frame mode

2.3.2 The large and small window mode

The window modes as shown in Fig. 3 reduce the field of view to reduce the number of out-of-time events and to improve the time resolution and finally increase the pile-up limit for bright sources. In the large window mode the inner half of the CCD is used for imaging, then rapidly transferred towards the readout node (720 ns per CCD line) and eventually read out, similar to the full frame mode. During the fast transfer the image outside the defined FoV is automatically cleared. The time resolution is lowered to 47.7 ms and the amount of out-of-time events drop below 0.2% in the large window mode. The small window mode operates comparably. The difference is that the field of view is further reduced to 63 $\times$ 64 pixels and only one quadrant is operated instead of four. The time resolution drops to 5.7 ms and out-of-time event contribution is 1.1%. The small window mode only uses CCD0 of quadrant 1, i.e. the focal CCD.

$\begin{figure} \includegraphics[width=8.5cm,clip]{XMM35_struktur1.eps}\end{figure}$

Figure 4: Mechanical structure of the pn-CCD camera system. X-rays imaged through the telescope enter the detector from the bottom

2.3.3 The timing and burst mode

The burst mode rapidly transfers 179 pixels and then reads the content of CCD0 in the conventional way. This allows for a 7 $\mu$ s time resolution and up to 60000 counts per second in the PSF. After each read out, the entire CCD is cleared from signals. The duty cycle (life time) in this mode is only 3%. The strongest sources can be observed in that mode.

2.4 The mechanical and thermal concept of the pn-CCD camera system

The camera housing is mainly made out of aluminum, the average integrated equivalent aluminum thickness, shielding the CCD from cosmic ionizing radiation is roughly 3 cm. The aluminum (AlZnMgCu1,5) contains Si, Fe, Cu, Mn, Mg, Cr, Zn and Ti. In total, these ``trace" elements represent about 10% of the total mass. Figure 4 shows a cross section through the pn-CCD camera system. The main components are the radiator, cold finger, proton shield and the printed circuit board with the integrated preamplifiers (CMX and TMX) and the pn-CCD, mounted in an invar ring. The interconnections between the CCD and the surrounding electronics on the PC board are wedge-bonded. About 900 bonds were required, all individually coated, to improve their mechanical stability.

The invar consists mainly of Ni, Mn, Si, C and Fe. The PC board contains beside its Mo core Cu lines as metallization layer. A spider type support structure, smoothly pressed onto the PC board and on the CCD wafer act as a mechanical stabilization of the main components of the camera head.

Between stand-off cone and stand-off base a filter wheel is implemented with 6 filter positions: Four positions carry filters of different thicknesses (see Table 2), one position is open and the closed position is realized by a 1 mm thick aluminum plate, to block ionizing radiation imaged through the mirror system. Four positions are equipped with thick, medium and two thin filters as specified in Table 2. A calibration fluorescent source (AlK $_{\alpha }$ and MnK $_{\alpha }$ and MnK $_{\beta }$ , 1.487 keV and 5.894 keV and 6.489 keV resp.) can illuminate the CCD through the filter wheel upon command. The total count rate of the calibration source on the detector is of the order of 100 counts per second, the half-life of ⁵⁵Fe is 2.7 years. The spatial distribution of the X-rays from the calibration source is inhomogeneous over the field of view.

The actual planning foresees an operating temperature of the pn-CCD of $-90~^\circ$ C during the whole mission. An active temperature control stabilizes the chip temperature to better than 0.1 K. The temperature is measured on the CCD directly and on the invar ring. The thermal design was made to achieve CCD temperatures as low as $-140~^\circ$ C while dissipating 0.9 W of power in the focal plane.

**Table 2:** Filter properties (Sn = Tin, PP = Polypropylene, PI = Polyimide, Al = Aluminium)
filter	layer 1	layer 2	layer 3	layer 4
	$\mu$ g cm^-2	$\mu$ g cm^-2	$\mu$ g cm^-2	$\mu$ g cm^-2
open	-	-	-	-
position 1
closed	Al	-	-	-
position 2	270200	-	-	-
2 $\times$ thin	Al	PI	-	-
position 3, 4	10.8	22.4	-	-
medium	Al	PI	-	-
position 5	21.6	22.4	-	-
thick	Sn	Al	PP	Al
position 6	18	28	27.5	28

The electronic concept was designed with a high degree of redundancy. The four quadrants of the CCD wafer are operated and controlled seperately. In addition the three CCDs of one quadrant can be electrically adjusted almost independently. The relevant supply voltages and currents of each CCD are programmable from ground. This enables the instrument team to modify operating conditions in case of performance degradation, if needed.