Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/docs/documents/CAL-TN-0018-2-7.pdf Äàòà èçìåíåíèÿ: Mon Aug 11 13:07:12 2008 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:06:06 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: annular solar eclipse

XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -1­ 2.7 ainazzi 08.08.08

EPIC status of calibration and data analysis
Matteo Guainazzi with the inputs from the whole EPIC Consortium please send all comments to Matteo.Guainazzi@sciops.esa.int This document reflects the status of the calibration of the EPIC camera as implemented in SAS 8.0 with all available CCFs at 30/06/2008, unless otherwise specified. Furthermore the outlook is considered for improvements of calibration, which at the moment can be expected for the next SAS release.

Contents:
1 Calibration Overview ......................................................................................................................... 1
1.0 Summary ................................................................................................................................................... 2 1.1 Imaging ..................................................................................................................................................... 5
1.1.1 Astrometry ............................................................................................................................................................. 5 1.1.2 Point Spread Function and Encircled Energy ....................................................................................................... 6

1.2 Effective Area ........................................................................................................................................... 8
1.2.1 1.2.2 1.2.3 1.2.4 Mirror collecting area ............................................................................................................................................ 9 Filter transmission ................................................................................................................................................. 9 CCD Quantum efficiency .................................................................................................................................... 10 Vignetting ............................................................................................................................................................ 11

1.3 Energy Redistribution ........................................................................................................................... 14
1.3.1 MOS ..................................................................................................................................................................... 14 1.3.2 PN ........................................................................................................................................................................ 15

1.4 CTI/Gain ................................................................................................................................................. 16 1.5 Background............................................................................................................................................. 20 1.6 Timing ..................................................................................................................................................... 21 1.7 Statistical flux comparison .................................................................................................................... 25 1.8 Cross Calibration with RGS and other Satellites ................................................................................ 25 1.9 Unexpected events ................................................................................................................................. 25

2 Data Analysis ................................................................................................................................. 26
2.1 New features in SAS .............................................................................................................................. 26
2.1.1 2.1.2 2.1.3 2.1.4 2.1.5 2.1.6 2.1.7 2.1.8 2.1.9 SA SA SA SA SA SA SA SA SA S S S S S S S S S 8.0: ............................................................................................................................................................... 7.1: ............................................................................................................................................................... 7.0: ............................................................................................................................................................... 6.5 ................................................................................................................................................................ 6.1: ............................................................................................................................................................... 6.0.0: ............................................................................................................................................................ 5.4.1: ............................................................................................................................................................ 5.3.3: ............................................................................................................................................................ 5.3.0: ............................................................................................................................................................ 26 26 26 27 27 27 27 28 28

2.2 Data Analysis .......................................................................................................................................... 28
2.2.1 MOS ..................................................................................................................................................................... 29 2.2.2 PN ....................................................................................................................................................................... 29


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -2­ 2.7 ainazzi 08.08.08

1

Calibration Overview

This section gives a short overview of the status of the calibration of the EPIC instruments MOS1, MOS2 and PN, operating on-board the XMM-Newton observatory. It summarises the quality of the calibration to the extent that this may influence the scientific interpretation of the results. The instrument calibration is based on a physical model of the various components including mirror response, filter transmission and detector response (energy redistribution, gain, Charge Transfer Efficiency, CTE; in the following: CTI1-CTE). During ground calibration various components were calibrated and the physical model for each component was optimised. These models were verified in flight and are, where relevant, continuously monitored (e.g. contamination of the detector, changes in gain and CTI due to radiation damage). Corrections, which are needed for these time-variable changes, will be applied to the Current Calibration Files (CCFs) and/or the processing software. For more detailed information see the release notes of the CCFs http://xmm2.esac.esa.int/external/xmm_sw_cal/calib/rel_notes/index.shtml at:

Blue coloured text gives distilled information or html links. Red (Purple) coloured text marks current high priority (low priority) problems.

1.0

Summary

We give in the next table a summary of the status of the on-axis calibration:

Effect Accuracy of the XMM-Newton frame with reference to optical frame Relative Astrometry Absolute Astrometry Point Spread Function (PSF) Relative Effective Area Absolute Effective Area Absolute Energy scale Relative Timing Absolute Timing
1

Max. Error 1''(r.m.s.) 1.5''(r.m.s) 2.0''(r.m.s) 2% ±7% ± 10 % ± 10 eV P/P<10 100 µs
-8

Energy dependent NO NO NO Y ES Y ES Y ES Y ES NO NO

Off axis angle dependent NO YES YES YES YES YES YES NO NO

Summary of important improvements since the 2.6 issue (August 2007) of the document: · 2-D elliptical PSF parameterization: The EPIC ON- and OFF-Axis 2D-PSF is based on simulations from SCISIM. Differences between simulations and real PSF may lead to problems in source detection algorithms and flux determinations. A first version of an off-axis and energy-dependent 2-D ("elliptical") parameterization based on the analysis of stacked image of a large sample of sources has been included in a new CCF extension of the PSF CCFs with SASv8.0. It will allow an improved determination of fluxes, especially for off-axis sources thanks to a better estimation of the Encircled Energy Fraction by arfgen. Further improvement of the parameterization, including an accurate treatment of the azimuthal dependence of the PSF envelope and of the spikes due to the optics support structure is foreseen for future SAS versions.

1

This number represents the maximum inaccuracy of the Encircled Energy Fraction for typical source extraction radii


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre · XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -3­ 2.7 ainazzi 08.08.08

pn time jumps: an improved algorithm to detect and correct sporadic "jumps" in the flow of the photon arrival times has been implemented with SASv8.0. It is based on: a) a more accurate determination of the frame times for all pn modes; b) correction of frame time drifts due to temperature variation and ageing of the on-board clocking (Freyberg et al. 2005, MPE Report, Vol.288, 159). The distribution of the percentage of exposures affected by residual uncorrected time jumps for each pn instrumental mode is given by the following table: pn instrument mode Full Frame Extended Full Frame Small Window Large Window Timing Burst Total Pre SASv8.0 5.8% 20.8% 37.2% 39.0% 12.0% 52.9% 15.6% Post SASv8.0 1.9% 2.3% 6.6% 1.4% 1.1% 3.2% 2.2%

Summary of important ongoing calibration topics: HIGH PRIORITY TASKS · PN-redistribution below 500 eV: Spectra of the supernova remnant (SNR) N132D taken with thin filter show a problem in modelling the CVI line near 368 eV (see Fig1-0a). The problem is not visible in spectra with thick filter neither in spectra from the other calibration SNR 1E0102.27219. In the latter cases the CVI line is strongly suppressed by the filter or is intrinsically much weaker which indicates that the problem arises by improper re-distribution modelling around 370 eV and not due re-distribution from higher energies down to 370 eV. A comparison of RGS and pn spectra of Zeta Puppis confirms this. MOS redistribution: The MOS cameras are suffering strong spatial dependent redistribution effects due to a patch in the focal position of the CCD. The current understanding of the extension of the patch is still crude and only modelled with 3 circular 2-D regions (1: on patch, 2: patch wings, 3: off patch). The Figure 1-0a: pn improper redistribution at 350 eV detailed determination and remodelling of the patch spatial structures is a medium and long term goal for EPIC targeted for the calibration efforts of 2008-2010. PN CTI rate dependency for count predicts the CTI losses. This can result features. For details see G. Sala et al., effect for spectra in Burst mode and scientific validation with SAS 8.0 rates > 200-300 cts/s: For very bright sources CTI correction over in an up to 2% apparent gain shift most visible for narrow-band spectral ESA proceedings SP-604, astro-ph/0511762. Figure 1-0b describes the Timing Mode. An implementation of this correction is undergoing

·

·


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -4­ 2.7 ainazzi 08.08.08

Figure 1-0b: Gain shifts applied to reduce systematic residuals around the Si (1.8 keV) - and Au-edge (2.2 keV) for a sample of EPIC pn observation of non-variable targets in Timing (left) and Burst (right) Mode. The X-axis represents the number of shifted electrons, a proxy for the total count rate. The continuous lines represent the best-fit to the data with a function: Y=a0Xa1+a2 · In light of the recent calibration adjustments described elsewhere, the EPIC-MOS, EPIC-pn and RGS instruments now return almost energy-independent flux differences for broadband sources (see the crosscalibration status document, available at: http://xmm2.esac.esa.int/docs/documents/CAL-TN-00525-0.ps.gz). The reason for this difference is currently a focus of calibration investigation by reviewing all components of the effective area of each camera system. MOS give in average ~5-7 % higher fluxes than pn. Analysis of a large sample of off-axis 2XMM sources (Mateos et al., in preparation; see also the "EPIC Calibration Status" presentation at the 2008 Users' Group Meeting, available at: http://xmm.esac.esa.int/external/xmm_user_support/usersgroup/20080506/index.shtml) has shown that the relative flux difference between the pn and the MOS camera in the hardest energy band (4.512 keV) is a function of the azimuthal angle in detector coordinates. The dynamical range is significantly larger along the RGS dispersion angle than along the perpendicular direction. This may be due to an inaccurate calibration of the RGA obscuration. Investigation is ongoing.

·

LOW PRIORITY TASKS: · PSF core: The improved King function of the PSF is a good, but not perfect, fit to the PSF of the telescopes, as the core of the PSF is very slightly underestimated. This is visible more in the MOS than in the pn (as the MOS detector pixels are much smaller than the pn pixels). This can produce an error in the enclosed energy of at most ~2 %, depending on instrument, energy and extraction radius. Work is currently underway to model the PSF as a combination of a King function plus a Gaussian function (the latter to model the slight excess at the core). Correcting this (at most) 2 % effect in the PSF core is a low priority task. PN spectroscopy with double events: Analysis of the pattern distributions in EPIC-pn has shown that charge is more often observed to be split along the readout direction than perpendicular to it, even after the effect of reemission has been taken into account. These additional `false doubles' show peculiarities in their pulse height distribution, which may introduce a systematic error in the energy. Thus, in particular for sources with bright emission lines, it is recommended to restrict the spectral analysis to single pixel events.

·


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -5­ 2.7 ainazzi 08.08.08

1.1

Imaging

1.1.1 Astrometry Astrometry means: The precision with which astronomical coordinates can be assigned to source images in the EPIC focal plane. We distinguish between: Absolute Astrometry (precision relative to optical coordinates), Relative Astrometry within a camera (precision within one camera after correction for systematic offsets due to spacecraft mispointing) and Relative Astrometry between cameras (positions in one camera relative to another one) The XMM-Newton absolute astrometric accuracy is limited by the precision of the Attitude Measurement System. Fig. 1-1 shows that the shift from the XMMNewton-EPIC to the optical frame is on average 0'' with a standard deviation of less than 0.8'' per axis. Hence the shift between a standard XMM-Newton observation and the optical frame is considered to be 1" (r.m.s). The relative astrometry within each camera is accurate to 1.5" for all cameras and over the full field of view. The MOS metrology has been revised with SAS6.0, searching for systematics in the offsets of MOS peripherical CCDs with respect to the central CCD by using observations on rich stellar fields. CCD offsets of up to 2.7" have been corrected in the MOS LINCOORD CCF issue 17. With this new CCF the MOS relative astrometry accuracy has been assessed to be 1.5" (r.m.s.) while it is as good as ~1.0" (r.m.s.) for the EPIC-pn. Among all three EPIC cameras the relative astrometry is also estimated to be better than 1.5" across the whole field of view. This leads to an absolute astrometric accuracy of any source in the XMMNewton field-of-view of ~ 2" (r.m.s.). Of course this does not include the statistical uncertainty of the actual measurement especially important for faint sources. Figure 1-1: Histogram of the distribution of offsets for each EPIC camera with respect to the 2MASS reference frame projected on to the two spacecraft axis. Top: MOS1, Middle: MOS2, Lower: pn (Details at XMM-SOC-CCF-REL-168) Note that for faint MOS sources near the detection limit the statistical accuracy of the measurement limits the 90 % confidence contours to 24".

A possible residual in the position angle rotation (Euler angle) of the order of 0.1 deg is under investigation. This could lead to an additional uncertainty of up to 1.5'' at the edge of the XMM-Newton field of view.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -6­ 2.7 ainazzi 08.08.08

1.1.2 Point Spread Function and Encircled Energy Point Spread Function: The spatial distribution of light in the focal plane in response to an observed (monochromatic) point source. The PSF integrates to 1 over the infinite focal plane. Encircled Energy: The fraction of the energy of a point source collected within a certain radius. 1D-PSF: Though the shape of the PSF is quite complex, the radially averaged profile can be adequately represented by an analytic function - a King function - whose parameters, core radius r0 and index, are themselves functions of energy and off-axis angle:

PSF

=

r A 1 + r 0



2



-

It is worth noting that both this function and its integral are analytic. An example of a surface brightness radial profile plus a fitted King profile is shown in Fig.1-2. Earlier work (EPIC-MCT-TN-011, EPIC-MCT-TN-012, XMM-CCF-REL-116) used many bright point sources both on and off axis to determine the energy dependent PSF. This resulted in a linear dependency of r0 and with energy and off-axis angle. It is shown in XMM-CCF-REL-0167 that this linear dependency is not valid - the dependencies of r0 and are seen to be flatter (almost constant) with energy (at least out to ~ 8-10 keV). Thereafter the dependencies rapidly turn steeper.

Figure 1-2: Surface brightness radial profiles (crosses) plus fitted King profiles (lines)): MCG-0630-15 Rev. 303 pn at 6 keV.

Figure 1-3: r0 ­ Energy (top) and ­ Energy (bottom) dependencies for the MOS1, MOS2 and pn on-axis PSFs.

Two threads of analysis using data from very long and clean Small Window mode observations of very bright non piledup sources were followed: One involved the forming of narrow-energy-band images, and fitting the surface brightness radial profiles with a King function to obtain r0 and as a function of energy. A second analysis thread involved the extraction of spectra from narrow annuli around point sources, and once Ancillary Response Files (ARF) had been generated (this involving the actual form of the PSF), the spectra were fitted with standard spectral models, and to see how (if at all) the spectral parameters obtained varied


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -7­ 2.7 ainazzi 08.08.08

with extraction radius. This whole process was repeated for several sets of PSF parameters (including those obtained from the surface brightness radial profile fitting described above). The new PSFs were used in the analysis of spectra extracted from narrow annuli around a number of bright point sources, as described above. A point source should show no variation in fitted spectral parameter whether the spectrum is extracted from the very centre of the distribution or from the wings. Usage of previous PSFs results, however, in a very wide range of spectral parameters for different radii. Usage of the new PSFs gives rise to a significantly improved situation, with the fitted normalization and power-law index remaining constant and 'flat' with radius. A major problem with the previous parameterisation was its inability to produce consistent spectral fits for annular extraction regions such as those for instance used for the analysis of piled-up sources. Hence MCG-6-30-15 (from Rev 302) has been extracted from annuli of 5-40", 10-50" and 15-60" and fits have been compared to those obtained for a circular extraction (0-30"). This has been performed using the old and the new PSFs, and the results are presented in Fig. 1-4. Whereas usage of the old PSFs results in a per instrument normalization variation of up to 40%, and changes in the fitted spectral slope of 0.2, the new PSFs give rise to normalization variations of ~ 5% and a spectral slope change of at

Figure 1-4: Plots showing how the fitted normalization (top) and power-law index (bottom) vary as a function of extraction region (left to right: 0-30" circle, 5-40" annulus, 10-50" annulus, 15-60" annulus), using the old CCF PSFs (left) and the new CCF PSFs (right panel) for the MCG-06-30-15 Rev.302 data.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre most 0.03. Note that the King function is a good but not perfect fit to the PSF of the telescopes, as the core of the PSF is very slightly underestimated. This affects the MOS more than the pn (as the MOS detector pixels are much smaller than the pn pixels). This can produce an error in the enclosed energy of at most ~2 %, depending on instrument, energy and extraction radius. Work is currently underway to model the PSF as a combination of a King function plus a Gaussian function (the latter to model the slight excess at the core). Correcting this (at most) 2% effect in the PSF core is a low priority task. 1D-Off-axis PSF As yet, no sources bright enough for this new type of analysis to be performed off-axis have been observed. Therefore, the general off-axis results of previous work (EPIC-MCT-TN-011, EPIC-MCT-TN-012, and XMM-CCF-REL-116) have been used to transform the new on-axis parameters presented here to projected off-axis values. How the PSF varies off-axis is only known reasonably well for small off-axis values. The following two plots indicate the reliability of the PSF at several off-axis locations: XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -8­ 2.7 ainazzi 08.08.08

Figure 1-5: Reliability of PSF

In the green region: For low energies and nearly on-axis positions a large quantity of data is available and in general statistics for these measurements are good. Error bars are in general small (/r0 (or ) <~ 1 %) and the r0 and evaluations for these curves are not very "far" from the final fit (r/ r < 5% is the worst case). In the yellow regions: for some off-axis angles few measurements are available (< 2-3) at the different energies. In general these measurements have large uncertainties (/rc0 ~ 10%) and the r0 and parameters for these sources sometimes are "far" from the final best-fit values (r /r can be as large as 10-20%). In the red region: no calibration data are available, and best-fit values coming from the model are an extrapolation, not a real calibration. Addendum: Off-axis PSF MOS speciality While the EPIC-pn PSF is azimuthally symmetric, the placing of the CCDs in the MOS cameras to follow the focal plane results in a chip-to-chip variation in the MOS PSF (Saxton, R.D., Denby, M., Griffiths, R.G., Neumann, D.M., 2003,


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: -9­ 2.7 ainazzi 08.08.08

Astron. Nach., 324, 138.). This is not currently modelled in the SAS calibration but will result in an azimuthal variation in the encircled energy fraction, which is dependent on the extraction radius and the off-axis angle. This variation is not yet quantified but is estimated to be ± 4% for a source circle of radius 25" at an off-axis angle of 7'.

1.2

Effective Area

Effective Area means: the collecting area of the optical elements folded with the energy-dependent sensitivity of the detector systems of the EPIC cameras. For on-axis sources an internal accuracy of better than 5% in the determination of the total effective area is reached over the spectral range from 0.4 - 12 keV for each instrument separately. The cross calibration between MOS1 and MOS2 agrees within 5 %. The MOS/PN global flux normalization agrees to 7 % from 0.4 keV to 12 keV. A statistical assessment of the relative flux cross-calibration among the X-ray cameras on board XMM-Newton is available in the cross-calibration preview tool.

1.2.1 Mirror collecting area Mirror collecting area means: the face-on area of the mirror system that reflects X-rays to the focal region. The mirror collecting area has been measured on ground and verified in orbit. However, XMM-Newton EPIC-pn observations with very high statistics still showed residuals above 6 keV. The mirror effective area has been changed in agreement with simulations of the multiple mirror layers in order to minimize these residuals. For further details see (XMM-CCF-REL-205.)

1.2.2 Filter transmission Filter transmission means: the fraction of incident X-ray photons that pass through the filter The filter transmission has been measured on ground. The following plot shows filter transmission curves for all the cameras. For the thick filter the green curve shows the single transmission function currently used in the CCFs for MOS1, MOS2 and PN.

Figure 1-6: Filter transmissions in CCF and ground calibration filter measurement data points. green - thick, red - medium, blue- thin (Thin1 & Thin2 have the same CCF) ground: squares-PN, star-M1, diamond M2, CCF: dashed lines


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre 1.2.3 CCD Quantum efficiency XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 10 ­ 2.7 ainazzi 08.08.08

Quantum Efficiency (QE) means: the fraction of incident photons on the detector that generate an event in the CCD. Recent recalibration of the MOS QE (XMM-CCF-REL-235) led to an increase of the surface layers of C, N and O. Nitrogen and Oxygen are constituents of the surface layers of the CCDs and the level of increase of these depths is within the accuracy of physical measurements of the CCD structure. Carbon is not a natural layer on the CCD. It has been added to the QE model primarily because an additional edge at this energy is most compatible with the existing redistribution function. There is no evidence for a significant change in these additional layers since launch and the new QE model is in

Figure 1-7: Upper panel: MOS-QE ground calibration measurements (data points), old QE model (straight line), and new QE model (dashed line). Lower panel: pn-QE ground measurements and


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 11 ­ 2.7 ainazzi 08.08.08

fact marginally in better agreement with the ground calibration measurements. Ground calibration measurements have shown that the quantum efficiency of MOS CCDs is spatially uniform above 400 eV. Below this energy spatial variations within a CCD are seen as patches in the outer parts of the CCDs where the response is degraded. This inhomogeneity is currently not taken into account by the SAS.

Figure 1-8: QE spatial inhomogeneities within MOS1 appearing only at very low energies (left 150 eV, right 400 eV).

1.2.4 Vignetting Vignetting means: reduction in the effective area with radial distance from the telescope's axis. The telescope vignetting is well determined for off-axis angles of up to more than 10'. One of the most important outstanding problems of the calibration was an offset of around 1' in the telescope axis from nominal. This did not affect the astrometry but could have been the reason for some of the flux discrepancies between MOS and PN caused by the vignetting correction that was not adapted to this offset in SAS versions earlier than 6.0.0. The offset was determined and implemented in the corresponding CCF (XMM_MISCDATA_0020). As of SAS 6.0.0 and the current available CCF the corrected optical axis position improves the vignetting correction. However the vignetting correction itself has not changed at all, the only difference is that it is now applied for correct off axis angles that could not be calculated correctly before due to the wrong information for the optical axis. This improves differences in flux for off axis sources for each camera from ± 14 % to ± 5 %. Detailed information can be found in XMM-CALTN-54 and XMM-CCF-REL-0156. Figure 1-9 shows the average vignetting measured over 4 azimuths at an off-axis angle of around 10' derived from Calibration measurements of the SNR G21.5-09


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 12 ­ 2.7 ainazzi 08.08.08

Figure 1-9: Average vignetting measured over 4 azimuths at an off-axis angle of around 10' derived from calibration measurements of the SNR G21.5-09


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 13 ­ 2.7 ainazzi 08.08.08

1.3

Energy Redistribution

Energy Redistribution means: The energy profile recorded by the detector system in response to a monochromatic input. This includes the spreading of the recorded energy due to the statistical nature of charge collection (i.e. the energy resolution) and also charge loss effects, which distort the profile. The energy resolution is monitored by the onboard calibration source, which produces strong lines Al-K (1.487 keV) and Mn-K (5.893 keV) lines. The resolution of the MOS cameras degraded significantly with time up to the epoch where the cameras were cooled (Figure 1-12). After this epoch the degradation in energy resolution has been small. For the EPIC-pn camera the line width of Mn-K increased by 4% since launch; for Al-K, the increase is only 2%. 1.3.1 MOS The low energy redistribution function (RMF) of the MOS CCDs has a complex shape, in that the main photo peak has a secondary component (a shoulder) which relatively increases with decreasing energy, until, at the very lowest energies, it is the dominant component. Observations of non- or weakly-varying sources, such as the SNR 1ES0102-72 and the O star Zeta Puppis, have shown that the shape of the in-flight redistribution function changes both spatially across the detector (the change from the on-ground calibration being most pronounced at the boresight) and also temporarily with observation epoch (Read et al. 2005, ESA-SP 604). The form of this was such that the original shoulder evolved into a flatter `shelf', of lower amplitude, but extending to lower energies. The bulk of the spatial change occurs within 40 arcseconds of the boresight, a region which we refer to as the "patch". A new method of deriving the RMF was constructed to account for this spatial and temporal variation. This has been incorporated into the Science Analysis Software (the SAS), such that as of SAS v6.5.0, there are three RMF regions on each of the two MOS detectors; a "patch core" region, a "patch wings" region and an "outside patch" region (see Fig. 1-10). This, in combination with the 10 temporal epochs now considered in the SAS, gives rise to a total of 60 MOS RMFs in the current calibration files. For a source extracted close to the patch (e.g., as in figure with patch), rmfgen automatically constructs a PSF-(default) or flat-weighted average RMF from the three region-defined RMFs (making use, of course, of the calibration files from the correct epoch).

Figure 1-10: The 3 different RMF regions The difference between SAS v6.1.0 and the SAS v6.5.0 is shown in Figure 1-11 where an RGS fluxed model of this linerich star is folded through the redistribution functions derived from the old and new SAS versions (XMM-CCF-REL-

Figure 1-11: zeta Puppis with old SAS 6.1 (black) and as of SAS 6.5 (red)


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 14 ­ 2.7 ainazzi 08.08.08

192, XMM-CCF-REL-196). Since SAS 6.5, rmfgen produces RMF following the above method.

Figure 1-12: Energy resolution MOS1-CCD 1 (red), MOS2 ­CDD1 (green) and pn-CCD4 (black). The drop after rev 532 is due to the cooling of the cameras.

1.3.2 PN The energy redistribution is mode-dependent for the PN camera. That means that different response matrices are needed for the different modes. Since SAS 5.4.1 rmfgen supports all modes. Also, ready-made matrices can be used, which can be obtained at: http://xmm2.esac.esa.int/external/xmm_sw_cal/calib/epic_files.shtml. The response is mainly studied for on-axis sources. The response matrices are identical for all CCDs (note that for SW, TIMING and BURST only the CCD containing the focal point is used); the only dependence of the redistribution (i.e. the energy resolution) is the RAWY dependence (see 2.2.2). Other off-axis (radial) dependencies do not exist in the redistribution, vignetting is part of the effective areas computed by arfgen. For SAS 6.0 the re-distribution was strongly increased for energies below the O-edge (0.53 keV) in order to reduce excesses seen at low energies. In particular for the isolated neutron star RX J1856.5-3754, a soft excess in the EPIC-pn spectrum resulted in an absorption column density of almost zero in comparison to NH = 91019 cm-2 obtained from a 500 ksec Chandra LETGS high resolution spectrum. At that time re-distribution parameters were adjusted to match the column density value obtained with LETGS. Subsequent analysis of EPIC-pn spectra of Zeta Puppis (with SAS 6.0), however, revealed large residuals around 430 eV (N lines), which clearly showed that the redistribution was modelled incorrectly. The original re-distribution parameters model the N lines in the Zeta Puppis spectrum better. Therefore, values near the old (from ground calibration) ones were adopted for SAS 6.1. As a result the column density derived from the EPIC-pn spectrum of RX J1856.5-3754 was again slightly lower (71019 cm-2) than the LETGS value. Based on the analysis of a set of Blazar spectra, which are expected to be fit well with single power-law models in 0.1-2 keV, and showed a consistent set of residuals, we performed an energy-dependent re-working of the RMF flattening


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 15 ­ 2.7 ainazzi 08.08.08

these residuals. The new RMFs improve the fit to other continuum sources such as RX J1856.5-3754 and bring the PN, MOS and LETGS into better agreement (XMM-CCF-REL-189).

1.4

CTI/Gain

CTI means: Charge Transfer Inefficiency is the imperfect transfer of charge as it is transported through the CCD to the output amplifiers during read-out. Gain means: Gain is the conversion (amplification) of the charge signal deposited by a detected photon from ADU (Analogue to digital unit) charge into energy (electron-volts).

Figure 1-13: MOS1-CCD 1 (red), MOS2 ­CCD1 (green) and pn-CCD4 (black) calibration source line positions at Al-K (upper) Mn-K (lower). Note that the pn Al line is not at the correct place for pn. This is taken into account later through the response matrix. For the instruments in normal conditions, the Gain and CTI are currently known to the extent that the line energy can be determined with an uncertainty of 5 eV over the full energy range for all MOS imaging modes and with an uncertainty of 10 eV over the full energy range for all pn imaging modes. The relative accuracy of the Timing modes compared to the imaging modes is better than 0.5% over the full energy range (see Fig.1.14). The situation is different for abnormal conditions like the eclipse seasons or during solar flares. Here the gain at the beginning of an orbit (i.e. during calibration observations) can vary significantly from the gain for the remainder of the orbit, due to temperature variations in the on-board electronics.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 16 ­ 2.7 ainazzi 08.08.08

For the MOS, no correction has been achieved yet for this, but the effect of the EMAE (EPIC MOS Analogue Electronics) temperature excursions during scientific observations are usually small (MOS1: < 10eV at Mn-K and < 5eV at Al-K). For the PN, deviations in the gain occurred during the first eclipse season (revolutions 60-80) and during

Figure 1-14: Upper: CasA spectra in different PN-Modes show the accuracy of the energy calibration. (Count rate spectra are scaled differently for clarity). Lower: Relative percentage difference between the energies of the emission lines measured in CasA Obs.#0165560101 (pn in Full Frame, MOS cameras in Timing Mode). In the MOS2 observation the ~7.6 keV line is not detected.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 17 ­ 2.7 ainazzi 08.08.08

occasions when the RGS was first turned off to diagnose a CCD chain failure (revolutions 136-146, and 149-150), when the platform temperature changed significantly and 10-30 eV discrepancies were occasionally observed. The PN team has established a good correlation to implement a temperature dependent gain correction (see XMM-CCF-REL-223), which is available as of SASv7.1.2.

1.5

Background

The XMM-Newton observatory provides unrivalled capabilities for detecting low surface brightness emission features from extended and diffuse galactic and extragalactic sources, by virtue of the large field of view of the X-ray telescopes and the high throughput yielded by the heavily nested telescope mirrors. In order to exploit the excellent EPIC data from extended objects, the EPIC background, now known to be higher than estimated pre-launch, has to be understood thoroughly. Detailed information on the treatment of the EPIC BG is provided on an especially dedicated web page at the XMMSOC portal at: http://xmm.esac.esa.int/external/xmm_sw_cal/background/index.shtml where several dedicated tools and files are available for use in dealing with the EPIC background. There are several different components to the EPIC background: 1. Photons: · The astrophysical background, dominated by thermal emission at lower energies (E < 1 keV) and a power law at higher energies (primarily from unresolved cosmological sources). This background varies over the sky at lower energies. Solar wind charge exchange. Single reflections from outside the field of view, out-of-time events etc. Soft proton flares with spectral variations from flare to flare. For weak sources the only option is to select quiet time periods from the data stream for analysis. To identify intervals of flaring background the observer should generate a light curve of high energy (E > 10 keV) single pixel (PATTERN = 0) events. To identify good time intervals use the selection criteria: · · MOS: < 0.35 cts/s (#XMMEA_EM && (PI>10000) && (PATTERN==0) on the full FOV PN: < 0.4 cts/s (#XMMEA_EP && (PI>10000) && (PI<12000) && (PATTERN==0) on the full FOV

· · 2. ·

Particles:

Note: These selection criteria assume that sources above 10 keV do not contribute significantly to the overall intensity. · 3. Internal (cosmic-ray induced) background, created directly by particles penetrating the CCDs and indirectly by the fluorescence of satellite material to which the detectors are exposed.

Electronic Noise: Bright pixels and columns, readout noise, etc.

The following two images show the strong metal line features that make the background subtraction complex, especially for extended sources.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 18 ­ 2.7 ainazzi 08.08.08

Figure 1-15: Upper: MOS-Al-K at 1.48 keV (left) and PN Cu-K at 8.04 keV (right) internal background caused by X-ray fluorescence lines correlated with the structures of the electronic board (pn-Cu-K) and the more distant camera itself (MOS Al-K) Lower: Background spectrum for the MOS1 camera (left) during an observation with the filter wheel in the closed position. The prominent features around 1.5 and 1.7 keV are respectively Al-K and Si-K fluorescence lines. The rise of the spectrum below 0.5 keV is due to the detector noise. Background spectrum of singles only for the pn camera (right) during an observation with the filter wheel in the CLOSED position in the energy range 0.2-18 keV. The prominent features around 1.5 keV are Al-K, at 5.5 keV Cr-K, at 8 keV Ni-K, Cu-K, Zn-K and at 17.5 keV (only in doubles) Mo-K fluorescence lines. The rise of the spectrum below 0.3 keV is due to the detector noise. The relative line strengths depend on the (variable) incident particle spectrum.


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 19 ­ 2.7 ainazzi 08.08.08

1.6

Timing

Absolute timing means: Locating events in time with reference to standard time defined by atomic clocks or other satellites. Relative timing means: The capacity to measure time intervals and periodicity reliably The release of SAS 5.3.3 included for the first time all necessary components to support timing analysis with outstanding time resolutions down to 29 and 7 microseconds for PN Timing and Burst modes respectively. Tests on data from the Crab pulsar taken during XMM-Newton's performance verification campaign in early 2000 indicate that the relative deviation in the observed pulse period with respect to the most accurate radio data available (P/P) is now considerably less than 3·10-8 (1=1·10-8), with an absolute timing accuracy of < 100 microseconds. For the Crab pulsar the new results conform to estimates of the theoretically attainable accuracy with XMM-Newton and the expected statistical errors. Further investigations of periodicity of other objects are currently underway. The onboard time is a running counter, kept by the CDMU and synchronised with all the Data Handling units (48 bits, resolution (LSbit) is 1/65536 seconds). The event time is a Time in a counter internal to the EPEA that timestamps each frame. This timer is reset to 0 at the beginning of each observation, at a sharp (integer) second of the On Board Time. Because of an improvement in converting from the onboard time to the event time as of SAS 5.3.3 the user should make sure that the data are fully reprocessed with the SAS higher than Version 5.3.3 and not only with the barycen task of SAS 5.3.3 in order to achieve the highest timing accuracy.

Figure 1-16: Upper: Relative timing accuracy in (PX-Ray-PRadio)/ PRadio for the Crab as function of mission time. Lower: absolute timing with respect to Jodrell Bank radio ephemeris


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 20 ­ 2.7 ainazzi 08.08.08

1.7

Statistical flux comparison

Using 118 calibration sources fitting the 3 EPIC cameras separately in 8 energy bands one can statistically evaluate the cross calibration situation for SAS 7.1 between the EPIC cameras.

Figure 1-18; Flux distribution in various energy bands for all EPIC cameras. Fluxes are normalized to the fluxes of a separately performed joint fit to all data sets in each energy band. The numbers in each panel indicates the distribution mean (standard deviation). EPIC MOS cameras give about ~ + 5-7 % more flux then the pn, but consistently over all energy bands. In earlier versions MOS showed a flux deficit at low energies and flux excess at higher energies. We consider this fact as a big improvement in the consistency of the flux and spectral calibration of EPIC!


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 21 ­ 2.7 ainazzi 08.08.08

Figure 1-18 shows flux distributions in various energy bands for all EPIC cameras, where the fluxes are normalized to the fluxes of a separately performed joint fit to all data sets.

1.8

Cross Calibration with RGS and other Satellites

These sections have been removed since a dedicated Cross-Calibration Document is available at the XMM-Newton Science Operations Centre web portal (XMM-SOC-CAL-TN-0052).

1.9

Unexpected events

EPIC suffered 4 events so far in the mission, which were most probably micro meteoroid impacts along the boresight. The last ones in rev 961 (March 05) caused the loss of MOS1 CCD6 and a new hot column passing very close to the MOS1 boresight. After a sudden optical flash, bright hot pixels appeared. The events are interpreted as a dust micrometeoroid scattered off the mirror surface under grazing incidence and reaching the focal plane detector. The typical size of those particles has been estimated to be < 1 micron. Their origin is most probably from interplanetary (or interstellar) dust but not linked to a meteor shower (higher sizes/masses). (See Abbey et al., ESA proceedings SP-604, Kirsch et al., 2005, Proc. SPIE 5898) The EPIC-pn team (MPE) did an experimental verification of a micrometeoroid impact in a dust accelerator. This demonstrated that particles of micron size could be deflected by grazing incident optics onto the focal plane of CCD (see N. Meidinger, 2002, Proc. SPIE 4851, 243-254) The EPIC-MOS team (Leicester) has a joint programme with ESTEC to investigate Figure 1-19: the effect of impacts on Silicon detectors and micro channel plates. Exposure maps corresponding to exposures before (upper) and A new hot column appeared on MOS1 CCD1 due to the last meteoroid event on after (lower) the loss of MOS1 MOS1 CCD1 at position (RAWX,RAWY)=(318,572), approximately, in the SAS CCD6 calibrated event list coordinate system. This new defect is leaking into the whole column 318. As a consequence, the offset of the whole column is raised by about 20 ADUs, therefore generating a lot of noise events at low energy above the low energy threshold, and the whole column is identified as bad by embadpixfind and masked out in the calibrated event list. This new hot column passes 3 pixels away from the nominal target position on CCD1 when the EPIC-pn is the prime instrument, and 9 pixels when RGS is prime instrument, therefore, in the first case, affecting a significant fraction of the on-axis source PSF. Starting at revolution 1044, the increased noise of the hot column is suppressed on-board. Real events hitting this column suffer a raised baseline from the leaking hot pixels, and therefore the offset correction restores the correct energy. The column is usable for science again. However, recent measurements indicate that the offset evolves with time. An evaluation of the scientific impact of this finding is currently ongoing. Moreover, the post-impact offset value in MOS1 Timing Mode is far too large, for a meaningful correction to be possible. Users are advised to discard the affected column (which in Timing Mode calibrated event lists corresponds to RAWX=320) and the adjacent ones from the accumulation of any scientific products.

2

Data Analysis

This section provides an overview of the major calibration/SAS improvements on time frames of SAS releases. Also included is a guideline of how to work with the different modes of the cameras.

2.1
2.1.1 ·

New features in SAS
SAS 8.0: improved pn time jump detection algorithm


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre · · 2.1.2 · XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 22 ­ 2.7 ainazzi 08.08.08

pn rate-dependent CTI correction (default off) 2-D elliptical PSF parameterization SAS 7.1: Refinement of the MOS QE improving the cross calibration between pn and MOS. The MOS cameras will return now an energy independent flux difference with respect to the pn of ~ + 5-7 %. Before MOS fluxes have been lower than pn at low energies (0.2-0.8 keV) and higher at high energies (2-10 keV). SAS 7.0: Column dependent CTI/Gain correction, improving line widths by up to ~15% for some cases. Improved EPIC source detection and parameterisation tasks, especially related to the detection of extended sources. Robustness and efficiency have been increased, to make possible the 2XMM Catalogue derivation, which is in final preparation at the time of this release. Inclusion of PSF correction for EPIC timing and burst modes in response matrix generation. Support for arbitrary uniform binning of MOS and PN spectra. pn FIFO reset correction corrects exposure time for saturation of the on chip amplifier stage (< 5 % more flux) pn quadrant box temperature correction (see CAL-SRN-223) SAS 6.5: rmfgen: The EPIC MOS response calibration has been improved, including modelling of spatial and temporal response dependencies. This, together with calibration improvements in the newly released PN calibration files, has as a result a much better cross-calibration among the EPIC instruments. EPIC source detection tasks have been upgraded to be more robust and efficient, with correct detection likelihoods. EPIC MOS badpixel finding task has been upgraded. EPIC metatasks emproc and epproc include now all the functionalities and filtering possibilities present in the PERL scripts emchain and epchain. SAS 6.1: arfgen Timing mode: the effective area in pn Timing and Burst mode produced by arfgen was not correctly taking the Y-extent of the extraction region (which needs to be a box in the fast pn modes) into account. With SAS 6.1 spectral normalization is now treated in the right way by renormalization with the extraction region extent in Y-direction. epevents-6.41 [SAS-6.1] has default setting "withgaineff=N" while in the DT it is already "withgaineff=Y" (correspondingly for epchain-8.53 in SAS-6.1]. After the deadline for SAS-6.1 the tests for the setting "withgaineff=Y" were finished and successful. It is highly recommended to use the setting "withgaineff=Y" SAS 6.0.0 The new SAS task epreject corrects shifts in the energy scale of specific pixels due to high-energy particles hitting the EPIC PN detector during offset map calculation and suppresses the detector noise at low energies by statistically flagging events based on the known noise properties of the lowest energy channels. In the case of timing mode data, flagging of soft flare events may be performed. An additional function is added to emevents that performs a blanking of bad-energy columns and handles those now correctly as dead areas. emevents is now doing filtering & removal of flickering pixels decreasing significantly the noise at lowenergies. The new SAS tasks ebadpixupdate allows operations on bad pixels at the level of the calibrated events list

2.1.3 · ·

· · · · 2.1.4 ·

· · · 2.1.5 ·

·

2.1.6 ·

· · ·


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre · XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 23 ­ 2.7 ainazzi 08.08.08

epatplot: The optional input of a background event list now allows the determination of background-subtracted pattern fractions. This is useful, e.g., in the case of extended source analysis or close to spectral background features. badpixfind: The task now permits bad pixel searching on calibrated multi-chip (i.e. final pipeline product) event list. Previous versions only operated on raw event sets. SAS 5.4.1 MOS CTI correction has been modified in order to compensate the stepwise degradation after solar flares Degradation of the MOS intrinsic spectral resolution due to the larger noise component of the degraded CTI has been modelled. Refined calculation of the MOS gain as a function of observation epoch has been implemented. epatplot now also indicates the distribution and amount of invalid PN patterns to ease flux comparison in the case of pattern pile-up. New PN quantum efficiency function, which is based on measurements of the thickness of the SiO2 layer on top of the CCD (see Figure 2-1). SAS 5.3.0 The MOS CTI correction has been improved to take into account the changes, which have occurred since, launch. A new task, evigweight, assigns a vignetting correction to each individual event. This allows extraction of vignetting-corrected images and spectra directly. A new task, epatplot, is available to identify pile up. The PN background rejection and CTI correction have been significantly improved. The tasks rmfgen and arfgen now reproduce the canned matrices to within 1%. Also, the full range of event patterns is supported. Further, rmfgen supports the major observing modes. SAS 5.3.3 The latest CTI correction for PN SW and LW mode has been implemented. EPIC-PN Timing and Burst mode response files are available Refined spectral redistribution for PN (concerns all readout modes) at energies below 600 eV. The redistribution was adjusted to achieve agreement in column density derived from PN and Chandra LETG spectra of RXJ1856.3-3754. calpnalgo now receives all quantities from CCFs. The SW/LW mode CTI-correction function for PN has been modified. arfgen and rmfgen now support all modes (incl. TI and BU modes).

· 2.1.7 · · · · · 2.1.8 · · · · · 2.1.9 · · ·

· · ·

2.2

Data Analysis

In the next section some general recommendations for data analysis are provided. This includes: · · · · Where should data be taken from the CCD Which energy range should be used Which pattern range should be used Which response matrix should be used


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 24 ­ 2.7 ainazzi 08.08.08

For detailed guidelines on XMM data analysis please use the SAS Users' Guide at http://xmm.esac.esa.int/external/xmm_sw_cal/sas_frame.shtml

2.2.1 MOS Imaging modes Source region: where appropriate Background region: · · For point sources: Background can be extracted from the same observation, from another region of the same CCD, off-axis, away from source counts. For extended sources: This is more complicated. Please have a look at explanatory notes available on the XMM web site at http://xmm.esac.esa.int/ccf/epic/#background.

Energy range: 0.1-10.0 keV, however, both limits depend on the aims of the analysis. In general the user should use patterns 0-12. (see XMM-Newton Users Handbook section 3.3.11 available at http://xmm.esac.esa.int/external/xmm_user_support/documentation/index.shtml). However, pattern 0 events can be used to minimise the effects (e.g. spectral distortion) of pile-up. Pattern 0 events can also be used for observations in which the best-possible spectral resolution is crucial and the corresponding loss of counts is not important. In addition the user should only use events flagged as "good" by using (#XMMEA_EM) in the selection expression window of xmmselect. When analysing spectra the user should use effective area files produced by the SAS task arfgen in conjunction with redistribution matrices produced with the SAS task rmfgen. Timing modes Source region: where appropriate Background region: background subtraction will not usually be an issue for sources observed in timing mode. However because the timing strip is only 100 pixels wide, background regions should be taken from the outer CCDs, which are collecting data in imaging mode. Energy range: 0.3-10.0 keV (this avoids contamination by low energy electronic noise events below 300 eV). Pattern 0 only for the source. Pattern 0, 1, 3 for the BG region in outer CCDs. As for imaging mode, canned redistribution matrices valid for timing mode are available and should be used with ARF files produced by the SAS. 2.2.2 PN Imaging modes Source region: where appropriate Background region: · For point sources: From the same observation but away from the source. Ideally the region should have the same distance to the readout node (RAWY) as the source region. This ensures that similar low-energy noise is subtracted, because this increases towards the readout-node. Do not use the columns passing the source to avoid out-of-time events from the source, i.e. do not use an annulus around the source region. Note that since SASv6.0 it is possible to suppress the low-energy noise with the task epreject. For extended sources: This is more complicated. Please have a look at explanatory notes available on the XMM web site at http://xmm.esac.esa.int/ccf/epic/#background.

·

Energy range: 0.13 keV - 15 keV For imaging purposes pattern 0-12 can in principle be used. However, since doubles (1-4), triples (5-8) and quadruples (9-12) (see XMM-Newton Users Handbook section 3.3.11 available at http://xmm.esac.esa.int/external/xmm_user_support/documentation/index.shtml) are only created above twice, three and four times the low-energy threshold, respectively, cleanest images are produced by excluding the energy range just above


XMM-EPIC status of calibration and data analysis XMM-Newton Science Operations Centre XMM-SOC-CAL-TN-0018 Page: Issue: M.Gu Date: - 25 ­ 2.7 ainazzi 08.08.08

the thresholds. E.g. to produce a 0.2 - 10 keV image one may select singles from the whole energy band and doubles only from 0.4 keV. FLAG == 0 omits parts of the detector area like border pixels, columns with higher offset, etc. This may be not desired (and is unnecessary) in the case of broadband images. For spectral analysis, response matrices are available only for singles, doubles and singles+doubles. Higher order pattern types are of low statistical significance, have degraded spectral resolution and are therefore not useful. Best spectral resolution is reached by selecting singles for the spectrum. FLAG == 0 should be used for high accuracy to exclude border pixels (and columns with higher offset) for which the pattern type and the total energy is known with significantly lower precision. At high energies the fraction of doubles is however almost as high as that of singles and to include doubles is recommended to increase the statistics. If a sufficient number of counts are available single- and doublespectra can be created separately and fitted simultaneously in XSPEC (with all parameters including norm linked together). One exception is the timing mode (see below). To choose the valid energy band for the spectral fit it is highly recommended using the task epatplot. It uses as input a spatially selected (source region) event file and plots the fractions of the various patterns as function of energy. Spectral analysis should only be done in the energy band(s) where single- (and double-) fractions match the expected curves. In some observations the low-energy noise can be high, restricting the useful band at low energies. Deviations at medium energies indicate pile-up (more doubles than expected) and in such cases the inner part of the PSF in the source region should be excluded for spectral analysis. (More information on the topic of the pile-up is available in the XMM-Newton Users Handbook at http://xmm.esac.esa.int/external/xmm_user_support/documentation/index.shtml section 3.3.9). The user should use ARF files produced by the latest version of the SAS task arfgen in conjunction with canned redistribution matrices (which are compatible to the CTI correction used in SAS 5.3.3 or above) or produce those redistribution matrices with the SAS task rmfgen. For each readout mode of the PN a set of RMF files is available (for singles, doubles, singles+doubles. Except timing mode where only singles+doubles must be used). The CTI causes a dependence of spectral resolution with distance to the readout node. Therefore the 200 lines of each CCD are divided into areas of 20 lines each (Y0 at readout, Y9 at opposite side, which includes the nominal focus point) and for each area an RMF file is available Timing and Burst modes Source region: Columns around source position Background region: Columns away from source The RAWY coordinate is related to a fine-time2. Selection on RAWY will therefore exclude certain time periods. For timing mode no such selection on RAWY should be made and whole columns of a source free RAWX region should be used; in burst mode use RAWY < 160 to avoid direct illumination by the source. For timing mode only singles+doubles should be selected for a spectrum and the fit restricted to energies > 0.5 keV to avoid the increased noise. For burst mode energies > 0.4 keV can be used with combinations of singles/doubles as in the window modes. Note that detailed instructions on the analysis of Burst mode data and their caveats can be found in XMM-SOC-CALTN-0069.

2 fine-time: in Timing mode a higher time resolution than the frame time can be reached since the RAWY position gives additional timing information due to the special read out in timing mode (for details see: E. Kendziorra et al, PN-CCD camera for XMM: performance of high time resolution/bright source operating modes, Proc. SPIE, 3114, 1997)