Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/docs/documents/CAL-SRN-0302-1-5.pdf
Дата изменения: Fri Oct 25 18:28:54 2013
Дата индексирования: Mon Mar 3 13:09:39 2014
Кодировка:

XMM-Newton CCF Release

Page:

1

XMM-Newton CCF Release Note XMM-CAL-SRN-0302 Calibration of the sp ectral impact of X-Ray Loading (XRL) in EPIC-pn Timing Mo de
Matteo Guainazzi & Michael Smith 2 August 2013

1

CCF comp onents
Name of CCF EPN REJECT 0007.CCF VALDATE 2002-07-07 EVALDATE Blocks changed XRL2PHA XSCS flag NO

This CCF component includes the first public calibration of the X-ray Loading correction (Smith 2004) in EPIC-pn Timing Mo de. It is based on a systematic analysis of a large sample of archival exposures (cf. § 2.2), complementing (and partly superseding) the results of a specific calibration experiment performed on the Crab Nebula (cf. § 2.1). XRL o ccurs when the offset map taken prior to each science exposure is contaminated by a celestial source. This is the case for almost all the exposures in EPIC-pn Timing Mo de (Guainazzi et al., 2012b) taken before the 23rd of May 2012. On that date an changes in operations was intro duced, whereby offset maps are being calculated with the CLOSED optical blo cking filter. This prevents XRL. However, this CCF is valid for al l EPIC-pn Timing Mo de exposure as of the 7th of July 2002 (cf. §4) notwithstanding the filter being used for the calculation of the offset map. SAS do es not apply the XRL correction, if the offset map is calculated with the CLOSED filter.

2
2.1

Changes
The original calibration experiment

Two subsequent observations of the Crab Nebula were performed in September 2011 to test the spectral impact of XRL: one with the offset map calculated with the CLOSED optical filter (Obs.#0611181001), the other with the offset map calculated with the same optical filter as the following science observation (THICK, Obs.#0611181101). XRL o ccurs only in the latter, because the photon sources are blo cked during the integration of the offset map in the former. By comparing the spectra taken during these observations, one

XMM-Newton CCF Release

Page:

2

Figure 1: PHA versus XRL in the [120:450] PHA range for the September 2011 experiment on the Crab Nebula. The dashed line indicates the identity locus. The insets indicate the results of a linear fit of the function P H A=a+b в X RL when b is left free to vary (upper), and when b is fixed to 1 (lower).

can directly estimate the spectral impact of XRL. The X-ray emitting area of the Crab Nebula covers several columns in the EPIC-pn Timing Mo de aperture. This allowed us to estimate the spectral impact of XRL for a wide range of source count rates by analysing spectra extracted in different columns. The results of the corresponding data analysis "spectral shift" (P H A) as a function of XRL columns. (P H A) is defined as the shift to be the offset map in THICK filter to minimise the fo 2 (P H A) = 1 в N (C
i i,CLOSED

summarised in Fig. 1, which shows the for spectra extracted in individual RAWX applied to the spectrum extracted with llowing quantity:
i,THICK 2 (P H A))2/i, CLOSED

-C

where Ci are the exposure-corrected counts, N is the number of spectral channels, and i are the Poissonian errors on each channel according to the Gehrels (1986) prescription. The P H A derived from the above equation were further divided by columndependent gain factors extracted from EPN ADUCONV 0140.CCF. A linear fit of the function: P H A=a+b в X RL yields numbers marginally consistent with a simple linear relation, with large error bars: a=-12 ± 11 ADU; b=1.4 ± 0.7 (1 errors). If one assumes b 1, a systematic "offset" a=-6.9 ± 0.4 ADU is required by the data. The discovery of a non-zero offset is puzzling. In principle, one might expect that the spectral impact of XRL is "purely" linear, i.e. a=0. The origin of the behaviour observed in the Crab Nebula exposure is unknown (as well as the reason for the ubiquitous XRL in EPIC-pn Timing Mo de, by the way). Putting these results in a wider context, the data points corresponding to the Crab Nebula deviate from the tight correlation between

XMM-Newton CCF Release

Page:

3

Figure 2: XRL (in ADU units; 1 ADU = 5 eV) as a function of the 0.710 keV count rate in column for a sample of 100 exposures taken in EPIC-pn Timing Mode. Colours identify of the hardness ratio distribution: HR=(CH -CS )/(CH +CS ), where CH and CS are the 1.510 keV and 0.71.5 keV energy ranges, respectively. The cluster of red points for C R XRL = 1519 are the Crab Nebula results.

the boresight the quartiles counts in the 100 s -1 , and

the net 0.710 keV Count Rate (C R) and the XRL (cf. the cluster of red points for C R100 s-1 in Fig. 2). The determination of the count rate in EPIC-pn Timing Mo de exposures of the Crab Nebula could be affected by multiple FIFO resets (M.Freyberg, private communication). It should be born in mind that the standard EPIC-pn Timing Mo de instrumental set-up is not optimised for sources as bright as the Crab Nebula, that would be heavily piled-up if point-like. There are reasons to seek for an independent confirmation of the results obtained with the experiment performed on it. We have therefore verified the calibration of the XRL spectral impact using a larger, and more diverse source sample.

2.2

The XRL calibration embedded in this CCF

We have analysed a sample of 26 EPIC-pn Timing Mo de exposures, where the number of shifted electrons, Ne in the 0.710 keV energy band is lower then 50. Ne is defined as Ne = Np
p i=1 Ei s в Texp в 3.6

N

ixel

where Ei is the energy of the i-th photon, Npixels is the number of pixels of the column whence each spectrum was extracted, Np is the number of detected photons, Texp is the exposure time and the factor 3.6 (in eV) represents the energy required to pro duce an electron-hole pair. It has been shown (Guainazzi 2013) that the energy scale is not affected by rate-dependent effects below this threshold. This sample can be therefore used

XMM-Newton CCF Release

Page:

4

Figure 3: Histograms of the energy scale deviations at the Au edge (2.3 keV) for a sample of weak (Ne <50) sources observed in EPIC-pn Timing Mode. Red histogram: a = 0; blue histogram: a = -6.9. The large points represent the median of the histogram values, and of their 1 errors: -11 ± 11 eV, and 24 ± 10 eV, respectively.

to estimate the accuracy of the XRL correction alone1. Sources with strong emission lines in the soft X-ray band were excluded from the sample, to ensure that the spectra were featureless in the 1.53 keV energy band. Spectra were extracted from a box in RAWX co ordinates 9 pixels wide around the boresight column. We used standard data reduction and screening criteria as recommended, for instance, in Guainazzi et al. (2012a) In order to estimate the accuracy of the energy scale, we made use of the steep effective area gradient at the Si (1.8 keV) and Au ( 2.3 keV) edges. We fit the spectra with a simple phenomenological mo del: the combination of a power-law and an accretion disk blackbo dy2 , seen through a screen of photo electrically absorbing neutral gas. A shift to the energy scale was applied during the fitting pro cedure (through the command gain fit in Xspec) to estimate the average deviation between the energy scale in the data and in the response. We tested on real data that such a pro cedure, while not entirely rigorous, yields comparable results (within the statistical accuracy of the data) as the direct shift of the PI energy scale in the event lists. The results are summarised in Fig. 3 The reconstruction of the energy scale is more accurate, and consistent at the 1 level with the nominal, when no offset term (a = 0) is included in the XRL correction. As this parameter choice reflects the expected behaviour of the camera, we implemented in this CCF the following XRL correction parameters: a = 0.0, b = 1.0. As we do not have an explanation for the residual -10 eV systematic error in the energy scale, and this level of uncertainties is consistent with the systematics
1 This assumes that the sp ecial gain correction in EPIC-pn Timing Mo de (default option withgaintiming=yes in epchain) is correct. Guainazzi (2013) shows that this option indeed yields the most accurate energy scale in this mode. 2 The choice of this comp onent is driven by the fact that most of the sources are X-ray binaries in soft state.

XMM-Newton CCF Release

Page:

5

asso ciated to the energy scale in Full Frame (the reference mo de for the gain correction), we have not attempted at further correcting it.

3

Scientific Impact of this Up date

The XRL correction embedded in this CCF constituent go es together with an updated calibration of the Rate-Dependent CTI, presented in an accompanying CCF Release Note (Guainazzi 2013). Their goal is achieving an accurate calibration of the energy scale, once the ubiquitous XRL in this mo de is properly subtracted. This CCF do es not affect the energy scale of any other EPIC-pn instrument mo des.

4

Estimated Scientific Quality

With this calibration, we expect that the average accuracy of the energy reconstruction in EPIC-pn Timing Mo de is better than 25 eV on the whole calibrated energy bandpass (0.710 keV), and for all level of count rates up to the threshold level (800 s-1 ). The following caveats, however, apply: · before Rev.$472 (started on the 7th of July 2002) no offset maps were taken prior to the science exposure. In this case SAS cannot correct for the (unknown) level of XRL, and do es not attempt at it. For these cases, the accuracy of the energy scale quoted in this do cument do es not hold. Applying the RDCTI correction as in EPN CTI 0028.CCF and earlier may still allow users to recover an accuracy of the energy scale better than 30 eV for energies lower than 4 keV, and better than 50 eV for energies higher than 4 keV. epevents issues a warning during the reduction of ODFs lacking an offset map. We plan to implement a statistical XRL correction for these ODFs (based on the correlation shown in Fig. 3) in a future SAS version · the accuracy of the energy reconstruction in EPIC-pn Timing Mo de exposures of the Crab Nebula could be affected by a systematic error of 30-40 eV. Users are recommended not to use EPIC-pn Timing Mo de for spectroscopic measurements on this (or a comparably bright!) source. Two specific exposures in Burst Mo de were designed for this purpose: Obs.#0160960401, and Obs.#0160960601. Readers are referred to Kirsch et al. (2006) and Weisskopf et al. (2010) for a more extensive discussion on EPIC spectroscopy on the Crab Nebula · The XRL correction is not the default in the SASv13.0.1. It has to be explicitly activated in, e.g., epchain by setting the following parameters: runepreject=yes, withxrlcorrection=yes. The SAS task performing the XRL correction is epreject.

XMM-Newton CCF Release

Page:

6

5

Test pro cedures and results

The results of the tests performed with a combination of this CCF and of the updated calibration of the RDCTI are discussed in an accompanying RN (Guainazzi 2013).

6

Exp ected up dates

None.

7

References

Gehrels, 1986, ApJ, 303, 386 Guainazzi M., 2013, XMM-SOC-CAL-SRN-0304 (available at: http://xmm2.esac.esa.int/docs/documents/CAL-SRN-0304-1-0.ps.gz) Guainazzi M., et al., 2012a, XMM-SOC-CAL-TN-0018 (available at: http://xmm2.esac.esa.int/docs/documents/CAL-TN-0018.pdf) Guainazzi M., et al., 2012b, XMM-SOC-CAL-TN-0083 (available at: http://xmm2.esac.esa.int/docs/documents/CAL-TN-0083.pdf) Kirsch M., et al., 2006, A&A, 453, 173 Smith M., 2004, XMM-SOC-CAL-TN-0050 (available at: http://xmm2.esac.esa.int/docs/documents/CAL-TN-0050-1-0.ps.gz) Weisskopf M., et al., 2010, ApJ, 713, 912