Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/docs/documents/CAL-TN-0020-1-0.ps.gz Äàòà èçìåíåíèÿ: Fri Aug 10 12:42:34 2001 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:17:20 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: optical telescope

First Results from the Monitoring of the RGS
Wavelength Scale
C. Erd
August 7, 2001
Abstract
The wavelength scale was determined by measurements of a few isolated emis­
sion lines across the entire spectrum of RGS. The wavelengths determined by
measurement were compared with wavelengths known from laboratory measure­
ments. No systematic drift was observed as a function of time. The spread of the
determined residuals in wavelengths is consistent with measurement accuracy of
the pointing of XMM­Newton and is \Sigma10 mš A.
1 Measurements
A set of bright coronal targets was chosen and the spectra for first and second order
were extracted by using the SAS. See Table 1 for the list of observations which was
used. This list contains all wavelength calibration observations that could be processed
by the SAS to date. The remainder of the observations are at this point in time either
limited by problems of the ODF, or by processing problems of the SAS. A few additional
observations were included as well, to bridge the gap between dedicated wavelength
observations and to improve the statistics of the sample.
Response matrices where created for each extracted spectrum with SAS 5.1. 1 Sub­
sequently a model consisting of a Gaussian plus a first order polynomial was fitted to a
ranges of PHA channels, limited just around the emission line under investigation. The
ranges in PHA were chosen around clearly visible and isolated emission lines. These lines
were identified and the result of the fit to the data was compared to laboratory mea­
surements. The list of emission lines, their transitions and their laboratory wavelengths
as used in this investigation are listed in Table 2.
The fit was performed by convolving the model through the respective response
matrix with three free parameters for the Gaussian and two free parameters for the
polynomial. The SAS task rgswavelscale was used to perform this convolution, be­
cause it is faster than fitting by XSPEC and it is easier to be executed as part of a
script.
1 This version of the SAS had an error mainly affecting the effective area. Since we are here only
interested in the peak positions of a lines, and by virtue of this analysis, this error has no effect.
1

Target Name Revolution Observation ID Ref. Number
Capella 0043 0119700201 1
Capella 0043 0119700301 2
Capella 0043 0119700601 3
Capella 0054 0121920101 4
YY Gem 0069 0123710101 5
AB Dor 0072 0123720201 6
Procyon 0160 0123940201 7
AB Dor 0162 0123720301 8
LHA 120­N63A 0177 0109990501 9
AB Dor 0205 0134520301 10
Sz 6 0222 0067140201 11
Capella 0232 0134720101 12
Table 1: List of observations used for the monitoring of the wavelength scale. The
reference number of the observation used in this note is given in the last column.
Transition Wavelength in š A
Mg XII Lyff 8.41926
Ne X Lyff 12.132
Fe XVII (27--1) 15.014
Fe XVII (5--1) 16.780
Fe XVII (3--1) 17.05321
O VIII Lyff 18.967
O VII (2--1) r 21.6015
O VII (2--1) f 22.0974
N VII Lyff 24.779
C VI Lyff 33.734
Table 2: List of lines and their wavelength as used during this assessment.
2

2 Discussion of the Results of the Fits
Examples of the fits to first order data for RGS1 and RGS2 are shown in Figures 3 and
4, respectively. The data are plotted as PHA channels converted into wavelengths by
the response matrix (EBOUNDS extension; equivalent to the XSPEC command setplot
wave), and shown as red lines. Similarly, the results of the fit, in PHA space and
converted to wavelengths in the same way are plotted in black.
The following observations can be noted with regards to the fits and with regards to
the selected lines:
ffl Globally it can be noted that the quality of the shape of the LSF as provided by
the convolved model, describes the data very well.
ffl The Mg Lyff line is difficult to fit due to poor statistics in most of the data sets,
and reasonable fits could only be obtained at a limited number of data sets.
ffl The effect of gaps between the CCD's and the effects of bad columns can clearly
be seen in the figures; e.g. at RGS1 (Figure 1) has inter­chip gaps at 17 š A and at
24.75 š A, and bad columns at 19 š A and at 22.18 š A.
This essentially invalidates the use of the N Lyff line for RGS1, as its centroid can­
not be determined accurately with the method which is used in this investigation.
Of course, this does not mean that science cannot be performed with this line, as
usually during scientific modeling of a source data are additionally constrained by
an astronomical model, which provides for the wavelength.
Similarly the Fe XVII emission line at 17.0532 š A at first order of RGS1 lies across
an inter­chip gap. This also results in biased estimates of wavelengths, for the
same reason as above.
The hot column at RGS1 at 19 š A is not present in all data sets. In any case, the
results appear not be limited by this bad column.
For second order spectra, only the Fe XVII line at 16.780 š A is of limited use for
RGS1, as it lies across an inter­chip gap as well.
Note that during the calibration phase there were an sequence of observations
taken with Capella with several large off­axis values (revolution 53, \Sigma5 arcmin,
\Sigma10 arcmin). These data could not yet be analyzed, but will provide useful feed­
back on the accuracy of the wavelength scale for off­axis performance and will also
remove part of the problems due to inter­chip gaps that were experienced with
these on­axis data sets.
ffl The Fe XVII emission line at 17.0532 š A aactually is a duplett which is resolved
in the second order spectra. This makes this line not useful for wavelength deter­
minations for second order (as the lines are too close for a fit with two Gaussians
to be stable when leaving the wavelengths as free parameters), and also its use
for first order wavelength determination should be taken with care. We proceed
under these assumptions and include this line into further analysis. It was used for
3

0
0.2
0.4
0.6
x 10 ­3
8.4 8.5
0.002
0.004
0.006
12 12.1 12.2
0.005
0.01
0.015
0.02
14.9 15 15.1
0.005
0.01
16.7 16.8 16.9
0
0.005
0.01
0.015
17 17.1 17.2
0
0.0025
0.005
0.0075
0.01
18.816 18.917 19.018 19.119
0
0.05
0.1
0.15
0.2
x 10 ­2
24.5 25
0.001
0.002
0.003
33.6 33.7 33.8
0
0.002
0.004
21.25 21.5
0119700301001: Order ­1 01/08/06 15.10
Wavelength in A Wavelength in A Wavelength in A
Wavelength in A Wavelength in A Wavelength in A
Wavelength in A Wavelength in A Wavelength in A
Wavelength in A
0
0.001
0.002
0.003
0.004
22 22.2 22.4
Figure 1: Results of detailed fits to parts of the first order spectra of RGS1. PHA
channels of the data and of the fitted function after convolution with the response
matrix are converted to wavelength by using the EBOUNDS extension of the response
matrix. 4

­0.25
0
0.25
0.5
0.75
x 10 ­3
8.4 8.5
0.002
0.004
0.006
12 12.1 12.2
0.005
0.01
0.015
0.02
14.9 15 15.1
0.005
0.01
16.7 16.8 16.9
0.005
0.01
0.015
0.02
17 17.2
0
0.005
0.01
0.015
18.821 18.922 19.023 19.124
0
0.001
0.002
0.003
0.004
24.5 25
0119700201002: Order ­1 01/08/06 15.11
Wavelength in A Wavelength in A Wavelength in A
Wavelength in A Wavelength in A Wavelength in A
Wavelength in A Wavelength in A
0
0.001
0.002
0.003
33.6 33.8
Figure 2: As Figure 1, but for RGS2.
5

some of the second order spectra as well, based on visual checks where the result
of the fit was reasonable close to the center of the duplett. These data points were
included to provide at least some comparative observations.
3 Performance of the Accuracy of the Wavelength
Scale
The residuals of the determinations of the wavelengths of the selected lines are plotted
after subtraction of the nominal wavelengths in Figures 3 and 4 for RGS1 and RGS2
respectively. Each frame shows the residuals determined for one emission line plotted as
a function of observation index. The key of the observation index is listed at the bottom
right of the figures and also in Table 1. First order residuals are plotted as circles, second
order residuals by squares. The first and second order data are slightly offset along the
abscissa in order to avoid overlapping of error bars. Unless drawn otherwise, the errors
are smaller than the size of the symbols.
For some observations more than one residual is plotted in cases where the obser­
vation consisted of more than one exposure, or in some cases a few measurements were
repeated to provide for an indication of the systematic accuracy of the method.
A clear correlation of residuals of different lines for the same observations can be
seen. Too few observations are however sampled to establish a clear drift over time.
This needs to be performed by addition of further data sets, which could not yet be
processed.
The wavelength scale of observation #5 appears to be consistently shifted towards
longer wavelengths. This is probably due to a property of the source, as this is an
eclipsing binary and the wavelength scale may be shifted due to red­shift at the source.
It was noted before that the data for RGS1 at 17.05 š A and at 24.78 š A should be
considered unreliable, and it can be clearly seen that these data show a spread with a
larger amplitude. For the same reason the second order data of RGS1 of Fe XVII at
16.78 š A show a large spread, too.
It is interesting that the Fe XVII lines at 17.05 š A appear to be systematically shifted
in RGS2 towards longer wavelengths by about 20 mš A. This is probably a consequence
of the nature of these lines and of the fact that they are just about to be resolved by
RGS2, due to its more symmetric LSF, which is also narrower. These lines are at an
inter­chip gap in the first order spectrum of RGS1.
It should also be noted that the attitude history files (AHF) are important for the
accuracy of the wavelength scale, as they provide the actual pointing of the instruments'
bore­sight on the celestial sphere, from which the angle of incidence on the gratings is
calculated, together with the coordinates of the target. Uncertainties in the angles of
incidence result in uncertainties of the wavelength scale. The accuracy of the AHF is a
function of the threshold at which a pointing offset is noted as additional record in the
file. During the cause of the mission this threshold was changed such that that more
recent AHF's have higher accuracy pointing information. This change was performed in
between revolution 83 and 84, such that a threshold of 2.5 arcsec pointing offset was used
6

­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
RGS1
8.41926 A
Residuals
in
mA
12.132 A 15.014 A
16.78 A
Residuals
in
mA
17.0532 A 18.967 A
21.6015 A
Residuals
in
mA
22.0974 A 24.779 A
33.734 A
Residuals
in
mA
1: Capella 0043 0119700201
2: Capella 0043 0119700301
3: Capella 0043 0119700601
4: Capella 0054 0121920101
5: YY Gem 0069 0123710101
6: AB Dor 0072 0123720201
7: Procyon 0160 0123940201
8: AB Dor 0162 0123720301
9: LHA 120­N63A 0177 0109990501
10: AB Dor 0205 0134520301
11: Sz 6 0222 0067140201
12: Capella 0232 0134720101
Figure 3: Residuals of the measured wavelengths subtracted by the nominal wavelengths
for RGS1. See text for details.
7

­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
­40
­20
0
20
40
5 10
RGS2
8.41926 A
Residuals
in
mA
12.132 A 15.014 A
16.78 A
Residuals
in
mA
17.0532 A 18.967 A
21.6015 A
Residuals
in
mA
22.0974 A 24.779 A
33.734 A
Residuals
in
mA
1: Capella 0043 0119700201
2: Capella 0043 0119700301
3: Capella 0043 0119700601
4: Capella 0054 0121920101
5: YY Gem 0069 0123710101
6: AB Dor 0072 0123720201
7: Procyon 0160 0123940201
8: AB Dor 0162 0123720301
9: LHA 120­N63A 0177 0109990501
10: AB Dor 0205 0134520301
11: Sz 6 0222 0067140201
12: Capella 0232 0134720101
Figure 4: Residuals of the measured wavelengths subtracted by the nominal wavelengths
for RGS2. See text for details. Note that no measurements could be performed for the
O VII lines at 21.6 š A and at 22.1 š A, because they are dispersed onto the CCD whose
readout electronics failed very early in the mission. 8

for AHF's up to and including revolution 83, and a threshold of 1 arcsec since the begin
of revolution 84. Furthermore AHF's for the following revolutions were re­processed
with a threshold of 1 arcsec: 42, 56, 60, 67, 69, 70, 72, 75, 81.
With the selection of data sets performed for this analysis, the change of accuracy of
attitude information occurs between data sets #4 and #5. The graphs in Figures 3 and
4 appear to show slightly better coherence of the residuals for observations with lower
threshold pointing information data.
In the near future a further possible improvement will be performed, in that raw
attitude information will be made available and will be ingested into the SAS such that
pointing drifts can be interpolated, providing smoother attitude changes with time,
which possibly is more accurate.
Finally the observed overall amplitude of the residuals of the wavelength scale is
consistent with an absolute measurement accuracy of the attitude of \Sigma5 arcsec, which
was also reported in calibrations of the optical monitor [1]. Implying that this is the
inaccuracy of the angle of incidence onto the grating plates, this relates to a inaccuracy
of the wavelength scale of \Sigma10 mš A applicable offset for the entire RGS bandwidth.
4 Summary
In summary it is confirmed that the accuracy of the wavelength scale at any given
moment is consistent with a peak­to­peak variation of \Sigma10 mš A peak­to­peak, which is
also consistent with systematic pointing errors as observed independently by OM.
5 Future Improvements
Data from more observations need to be added to this analysis, to shorten the gaps
in time between the current sample. This would possibly allow for more accurate con­
straints for seasonal drifts of the wavelength scale, possibly being caused by satellite
mechanical instabilities of systematic inaccuracies due to the pointing or attitude mea­
surement.
The off­axis performance of the wavelength scale should be included by processing
data from rev 53.
References
[1] XMM­SOC­CAL­TN­0019, Calibration of the XMM­Newton Optical/UV Monitor
Telescope (OM), available at
http://xmm.vilspa.esa.es/docs/documents/CAL­TN­0019­0­0.html
9