Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/docs/documents/CAL-SRN-0248-1-0.ps.gz Äàòà èçìåíåíèÿ: Mon Feb 2 18:09:48 2009 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:54:42 2012 Êîäèðîâêà: IBM-866 Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ï ï

XMMíNewton CCF Release Note
XMMíCCFíRELíXXX
Rateídependent CTI correction for EPICípn Timing Modes
M. Guainazzi, M. Kirsch, M.DÒÐaz Trigo
December 2, 2008
1 CCF components
Name of CCF VALDATE List of Blocks
changed
Change in CAL HB
EPN CTI 0020 2000í01í01T00:00:00 BURST GAIN NO
EPN CTI 0020 2000í01í01T00:00:00 RATE DEPENDENT CTI NO
2 Changes
2.1 Rateídependent CTI correction
As of SASv8.0, a new task (epfast) has been introduced, aiming at correcting rateídependent CTI
e#ects in pn event lists of exposures taken in Timing and Burst Modes. Readers are referred to the
SAS documentation for a full description of the task and its functionalities. It su#ces here to say
that epfast corrects the energy of each individual photon on the basis of the source+background
count rate measured at the photon arrival time.
The correction is formally expressed as a linear ``gain'' factor, G corr
G corr = a 0 # X a 1 + a 2
where X is the number of shifted electrons per pixel per second, and the a i are numerical coe#cients.
The new extension of the pn CTI CCF contains the values of the a i coe#cients for Timing and
Burst Mode.
1

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 2
Rate dependent CTI í pn timing mode
1 10 100 1000
Shifted electrons/pixel/sec
0.85
0.90
0.95
1.00
1.05
Gain
0060740901
0061140101
0061140201
0084020401
0084020601
0085290301
0087350101
0087350601
0087350801
0090340201
0099280101
0109090101
0111040101
0111061601
0112900101
0116690801
0117510201
0122340101
0124930301
0127720501
0134120101
0137551001
0144090101
0148220201
0148220301
0149810101
0150498701
0153750401
0153950301
0154750301
0157960101
0160960201
fit: y=a0(x a1 )+a2
a0=0.006
a1=0.190
a2=0.984
Rateídependent CTI í pn burst mode
100 1000 10000
Shifted electrons/pixel/sec
0.85
0.90
0.95
1.00
1.05
Gain
0112920701
0112920901
0112921201
0112921401
0112921501
0112921601
0112990101
0112990501
0117510201
0118700201
0135730701
0136140301
0153750201
0153750301
0155762501
0155762601
0160960401
0160960601
0160960801
0160960901
0160961101
0160961301
0202401101
0202401201
0311590501
0312790201
0312790401
0410581701
0412590301
0412590701
fit: y=a0(x a1 )+a2
a0=0.047
a1=0.121
a2=0.854
Figure 1: G corr versus N e relation for the sample EPICípn exposures in Timing (left panel and
Burst (right panel Modes, respectively. The solid line indicates the best fit with the functional form:
a 0 # N a1
e + a 2 .
2.2 Calibration results
The a i coe#cient have been calibrated according to the following procedure:
. a sample of 36 exposures in EPICípn Burst Mode and 42 exposures in EPICípn Timing Mode
has been selected on the basis that the backgroundísubtracted net source light curve was
statistically consistent with being constant
. for each of the sample source, spectra have been extracted from each of the four columns
surrounding the boresight column (this included)
. each spectrum was fit in the 1.5--3 keV energy band with a simple continuum model: powerí
law+black body corrected for photoelectric absorption. A constant gain shift G corr was applied
to the spectral model (through the gain function in Xspec) and calculated for each spectrum
under the condition to minimise the # 2
. for each spectrum, the number of equivalent shifted electrons N e was calculated, according to
the following formula:
N e = # Np
i=1 E i
N pixels ½ T exp ½ 3.6
where E i is the energy of the iíth photon, N pixels is the number of pixels of the column whence
each spectrum was extracted, N p is the number of detected photons, T exp is the exposure time
and the factor 3.6 (in eV) represents the energy required to produce an electroníhole pair.
The relation between G corr and N e was fit with the functional form: a 0 # N a1
e
+ a 2 (see Fig. 1).
The results of the above procedure for the selected sample are shown in Tab. 1.

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 3
Table 1: a i parameters in the EPNíCTI CCF version #20
Mode a 0 a 1 a 2
Timing 6.369 ½ 10 -3 1.900 ½ 10 -1 9.836 ½ 10-1
Burst 4.700 ½ 10 -2 1.210 ½ 10 -1 8.540 ½ 10 -1
Table 2: Parameters of the special burst gain correction (extension BURST GAIN) in EPNíCTI CCF
versions#19 and #20
CCF #19 CCF#20
A 1 1.0394 1.0414
A 2 1.0075 1.0075
E 0 4.7286 ½ 10 3 1.0061 ½ 10 3
#E 6.6123 ½ 10 2 6.6123 ½ 10 2
The application of the rateídependent EPICípn CTI to the sample of Burst Mode exposures
shows that the best results are obtained if the special Burst Mode gain values calibrated on ground
are applied. This function represents a relative gain (G rel ) ``tuning function'', calibrated by comí
paring observations of the Crab in Small Window and Burst modes (Kirsch, 2003, Ph.D Thesis,
UniversitØat TØubingen, Der Andere Verlag). The fitting function has the following functional form:
G rel = A 1 +A 2
1 + e (E-E0 )/#E
+A 2
Later recalibration of this CTI contribution did not take into account the rateídependent e#ect.
The introduction of the rateídependent CTI correction requires therefore that the parameters of the
BURST GAIN extension are changed to the values as listed in Tab. 2
3 Scientific impact of this update
The main aim of the rateídependent CTI is to improve the quality of spectral fitting on bright
sources in EPICípn Fast Modes by improving the accuracy of energy reconstruction. As shown in
Sect. 5 of this document, the application of the new CCF yields improvements both in the systematic
residuals around the Silicon and Gold edges (#2 keV) and in the agreement between the measured
and laboratory energy of narrowíband absorption features.
4 Estimated scientific quality
Systematic residuals around the Silicon and Gold edges are #5% for the overwhelming majority
of the sources in the testbed sample. The accuracy of the energy reconstruction for narrowíband
absorption features in the iron regime is now # 1%.

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 4
100
200
500
Cts/sec/keV
Obs.0084020601 - Serpens X-1 - 308.3 cts/s
1.6 1.8 2 2.2 2.4
0.95
1
1.05
Data/model
ratio
Energy (keV)
1000
200
500
Cts/sec/keV
Obs.0153750401 - Crab - 302.2 cts/s
1.6 1.8 2 2.2 2.4
0.95
1
1.05
Data/model
ratio
Energy (keV)
100
200
500
Cts/sec/keV
Obs.0090340201 - U1735-44 - 302.1 cts/s
1.6 1.8 2 2.2 2.4
0.95
1
1.05
Data/model
ratio
Energy (keV)
100
200
500
Cts/sec/keV
Obs.0148220301 - GX 339-4 - 254.7 cts/s
1.6 1.8 2 2.2 2.4
0.95
1
1.05
Data/model
ratio
Energy (keV)
Figure 2: Systematic residuals around the Silicon and Gold edges with EPICípn CTI #20 in the
four brightest observations (except Circinus Xí1) belonging to the Timing Mode testbed sample.
5 Test procedure and results
5.1 Timing Mode
In Fig. 2 we show the residuals against the baseline model around the Silicon and Gold instrumental
edges, once epfast coupled with the version #20 of the EPICípn CTI CCF is applied to the event
list of the four brightest sources (except Circinus Xí1) in our Timing Mode testbed sample. In all
cases the maximum systematic uncertainties are within ‘2.5%.
In Fig. 3 we compare the count spectra of 4 recent observations of GX13+1 without (left panel)
and with (right panel) the epfast correction coupled with CCF#20. The visual inspection already
suggests that the resonant Fe xxv and Fe xxvi absorption features are closer to their nominal
laboratory energies in the latter spectra. This is confirmed by a quantitative analysis (Tab. 3). The
di#erence against the nominal laboratory energy decreases from 1.5% to 0.7%.
Users shall be aware that the dateídependent CTI correction does not produce the same level of
improvements to all spectra of our testbed sample. The reasons of this uneven behaviour are still uní
der investigation, but are probably due to a combination of the intrinsically stochastic nature of this
correction and to additional residual e#ects which are not currently taken into account by epfast and
the associated CCF. Examples of observations where residuals #5% are still present once the rateí

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 5
6.5 7 7.5
20
40
60
80
Cts/s/keV
Energy (keV)
Black=01, Red=02, Blue=03, Green=05
GX13+1 - No epfast (SASv8.0)
Fe XXVI
Fe XXV
6.5 7 7.5
20
40
60
80
Cts/s/keV
Energy (keV)
Black=01, Red=02, Blue=03, Green=05
GX13+1 - epfast (SASv8.0)
Fe XXVI
Fe XXV
Figure 3: Count spectra of four recent observations of GX13+1 once the rateídependent EPICípn
CTI correction (right panel) is compared against the notícorrected spectra (left panel). The vertical
dashed lines indicate the laboratory energy of the resonant Fe xxv and Fe xxvi absorption lines.
Table 3: Centroid energies in keV of the Fe xxv and Fe xxvi resonant absorption features in four
recent observations of GX13+1. Errors are 90% confidence level for one interesting parameter. A
systematic error of 10 eV in included.
Observation# Fe xxv (6.70 keV) Fe xxvi (6.96 keV)
no epfast epfast no epfast epfast
0505480101 6.81‘ 0.06
0.01 6.75‘ 0.01
0.06 7.07 ‘ 0.01 7.02 ‘ 0.01
0505480201 ... ... 7.08 ‘ 0.01 7.02 ‘ 0.01
0505480301 6.81‘ 0.07
0.05 6.75 ‘ 0.03 7.06 ‘ 0.01 7.01 ‘ 0.01
0505480501 6.78‘ 0.03
0.01 6.746 ‘ 0.016 7.06‘ 0.04
0.01 7.02 ‘ 0.01

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 6
dependent CTI correction is performed are: Obs.#0122330801 (Crab Pulsar), Obs.#0148220201
(GX339í4), Obs.#0124930301 (PKS2155í304).
5.2 Burst Mode
In Fig. 4 we show four examples of improvements in the systematic residuals around the Silicon
and Gold instrumental edges, due to the application of epfast coupled with the version #20 of the
EPICípn CTI CCF to some of the brightest objects in the Burst Mode exposure sample. In all cases
the maximum systematic uncertainty decreases from #5% to #2%.
Users shall be aware that the dateídependent CTI correction does not produce the same level of
improvements to all spectra of our testbed sample. The reasons of this uneven behaviour are still
under investigation, but are probably due to a combination of the intrinsically stochastic nature
of this correction and to additional residual e#ects which are not currently taken into account by
epfast and the associated CCF. Examples of observations where residuals #5% are still present once
the rateídependent CTI correction is performed are: Obs#202401201 (Cyg Xí1), Obs.#0412590301
(Crab), Obs.#0160960401 (Crab), Obs.#0093562701 (GX339í3).
In Fig. 5 we show the count spectrum extracted from the pn exposure in Obs.#0155762501. The
target in this observation (GROJ1655í40; DÒÐaz Trigo et al., 2007, A&A, 462, 657) is known to show
a prominent resonant absorption feature associated with Fe xxvi. The bestífit energy with SASv8.0
CCF#19 is E = 6.93‘0.02 keV, whereas the best fit energy with CCF#20 is E = 6.982‘0.017 keV
(both errors are purely statistics at 90% confidence level for 1 interesting parameter). Measurements
of the Fexxiv with the RGS and of several lines with the Chandra gratings suggest a blueíshift
corresponding to an outflow velocity comprised between 400 and 1000 km s -1 . Such a shift should
bring the centroid energy of the Fexxvi transition to the energy range 6.969--6.983 keV, in excellent
agreement with the measurement yielded by CCF#20.
6 Expected Updates
Further studies will concentrate on those sources, where still residuals #5% are seen around the
Silicon and Gold edges.

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 7
1000
2000
5000
Cts/sec/keV
Obs.0155762501 - GRO J1655-40 - 350.2 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
2000
5000
Cts/sec/keV
Obs.0155762501 - GRO J1655-40 - 350.2 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
2000
5000
Cts/sec/keV
Obs.0112921501 - GRO J1655-40 - 284.4 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
2000
5000
Cts/sec/keV
Obs.0112921501 - GRO J1655-40 - 284.4 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
500
2000
Cts/sec/keV
Obs.0118700201 - Crab Nebula - 255.2 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
500
2000
Cts/sec/keV
Obs.0118700201 - Crab Nebula - 255.2 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
500
2000
Cts/sec/keV
Obs.0153750301 - Crab - 172.3 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
1000
500
2000
Cts/sec/keV
Obs.0153750301 - Crab - 172.3 cts/s
1.5 2 2.5
0.90.95
1
1.051.1
Data/model
ratio
Energy (keV)
Figure 4: Systematic residuals around the Silicon and Gold edges with EPICípn CTI version #19
(left panel) and #20 (right panel) in four observations belonging to the Burst Mode testbed sample.

XMMíNewton CCF Release XMMíCCFíRELíXXX Page: 8
6.5 7 7.5
50
100
150
Cts
s
-1
keV
-1
Energy (keV)
Black: CCF#20; Red: CCF#19
Obs.0155762501 - GROJ1655-40
:
Figure 5: Count spectrum of GROJ1655í40 (Obs.#0155762501) around the Fe xxvi resonant
absorption feature. The black histogram is the spectrum extracted with EPICípn CTI CCF#20, the
red histogram with CCF#19. The vertical dashed line marks the nominal laboratory energy of the
Fe xxvi line (6.96 keV).