Reproducibility of the GIS Non X-Ray Background
Y. Ishisaki^1, Y. Ueda^2, H. Kubo^2,
Y. Ikebe^3, K. Makishima^4, and the GIS team
1 Department of Physics, Tokyo Metropolitan University,
2 Institute of Space and Astronautical Science,
3 Institute of Physical and Chemical Research (RIKEN),
4 Department of Physics, University of Tokyo
ishisaki@miranda.phys.s.u-tokyo.ac.jp
1 Introduction
Thus far, the characteristics of the GIS non X-ray background (hereafter NXB) were reported by H. Kubo et al. (ASCA News No. 2) and by Y. Ikebe et al. (ASCA News No. 3). In this article, we report more detailed properties of the GIS NXB, particularly about its time variation, and reproducibility of the GIS NXB with several methods of NXB estimation. This investigation of the GIS NXB has been done through the spectral study of the Cosmic X-ray background (hereafter CXB). Methods developed here to reject or estimate the GIS NXB were actually utilized for the Cosmic X-ray background analyses (Y. Ishisaki 1996; Y. Ueda 1996). See also K. Makishima et al. (1996), which contains some part of this article as well as most of the previous two.
We utilized the night-earth data, i.e., events obtained when the XRT was
pointing to the night-side of the Earth, following the previous NXB studies.
Because there is some contamination of solar X-rays when the day-side of the
Earth is visible from the satellite, even when it is outside of the GIS f.o.v.,
we further set a condition that sun height from the rim of the Earth
(the Sun locates behind the Earth). Results shown in this article are based on
the night-earth observations from June 1993 to June 1995, which amount to 1,400
ks (figure 1, 2). In figure 2 we find a long-term NXB increase by ~15% in two
years, which is probably caused by a gradual build-up of long-term decay
radioisotopes.
2 Flare-events and the GIS monitor counts
As reported previously in the ASCANews No. 2 and 3, time variation of the GIS NXB is predominantly correlated with geomagnetic cut-off rigidity (hereafter COR) for cosmic rays along the satellite orbit. Although COR is a good indicator of NXB, COR is merely one of many factors that affect NXB. Therefore, we utilized the GIS monitor counts to estimate NXB. Among nine GIS monitor count items (ref. T. Ohashi et al. 1996), we selected H0 (too large pulse-height and too short rise-time) and H2 (too large pulse-height and too large rise-time) as the NXB indicator, because they are almost free from signal X-rays in the f.o.v., even during the on-source observations. H0 and H2 have adequate statistics of 10--30 c s-1 sensor -1 when they are summed up. Since GIS2 and GIS3 show similar variations, we still sum the counts of GIS2 and GIS3 in order to reduce statistical errors. Hereafter, we call this summed up counts "H02". Figure 3 shows a correlation between COR and H02, which has almost the same distribution as COR vs. LDHIT. As shown in figure 5, H02 and COR also correlate well to the night-earth data counts inside 17 mm radius (hereafter NTE). However, there are several issues on these correlations.
The first problem is a secondary branch below COR ~ 10 GeV c-1 on the correlation, as shown in figure 3. This is due probably to a slight inaccuracy of our COR-map. If we plot the satellite positions on the Earth, those periods which fall on the secondary branch appear only at the north east region of the map above the north Atlantic Ocean (dots in figure 4). Since the COR-map which we use does not refer to the satellite height, this discrepancy may change in the future when the satellite altitude decreases.
Figure 1 Integrated spectra of the day-earth (smooth line), the CXB
(crosses), and the night-earth (filled circles) observations taken by the
GIS2+3 on the whole detector area (
25 mm without ^55 Fe calibration isotope). Each of the exposure time
is given in the parentheses. The CXB and the day-earth spectra contains NXB.
Possible origin of the line feature with the energy in keV are also given in
the figure. The flare-cut (table 1) is NOT applied.
Figure 2 Long term (June 1993--June 1995) variations of the GIS2+GIS3 counting rate during the night-earth observations in the 0.6--1.0 (open circle), 1.0--2.0 (diamond), 2.0--4.0 (filled circle), and 4.0--8.0 (crosses) keV energy bands, integrated within the radius of 20 mm from the optical axis, where x-axis is time in unit of months. A solid line on the data of 4.0--8.0 keV band shows the best fit line of y = 0.00257x + 0.414. The flare-cut (table 1) is applied.
Figure 3 Correlation between COR and H02 counts during the
night-earth observations, where x-axis is COR in GeV c-1.
H02 is integrated for every 32 s. A dashed line means a function
derived from fitting. Two solid lines below and above show f(x)
scaled by 0.9 and 1.5 times, respectively. The flare-cut (table 1) is NOT
applied.
Figure 4 Distribution map for three types of phenomena on the Earth surface. The mark "." represents a position when H02 is less than the lower solid line in figure 3, where the COR-map is thought to be inaccurate. The mark "." represents a position when H02 is greater than the upper solid line in figure 3 (the hard-flares). The mark "+" represents a position when H02 < 1000 c/32 s = 31.25 c/s and NTE > 15 c/32 s = 0.46875 c/s in figure5a (the soft-flares). Two regions surrounded by solid lines define the dangerous areas where we should be cautious of the hard- and the soft-flares.
Figure 5 (a) A correlation between H02 and NTE, and (b) a correlation between COR and NTE. H02 and NTE are integrated for every 32 s. NTE is integrated in r <= 17 mm in the 0.3--11 keV energy band. Crosses superposed on the figure show averages (upper) and standard deviations (lower) in each H02 or COR bins. The flare-cut (table1) is NOT applied.
The next problem is a large scatter around COR ~ 11--12 GeV c-1 in the COR vs. H02 plot, which is also seen in COR vs. NTE (figure 5b). We call this phenomenon "hard-flares". If we plot the satellite positions on the Earth, the hard-flare periods are concentrated in the region peripheral to the south Atlantic anomaly (SAA; filled circles in figure 4). The SAA is located up in the sky around Brazil, where protons of E<100 MeV are confined by terrestrial magnetic fields. The particle background in SAA is extremely high, so that the high voltage of the GIS is turned off in SAA. The hard-flares are probably caused by particles trapped in SAA. These events can be eliminated by rejecting time intervals when H02 > 1.5 x f(COR) in figure 3.
After eliminating the hard-flares, we encounter yet another issue, i.e., the component which scatters upwards around H02~20--40 in figure 5a. We call this phenomenon "soft-flares". During the soft-flares, the NXB counts increase without an increase of H02. The significant rise of NXB counts is due to the soft flares. More caution should be used for the soft-flares than the hard-flares, because they cause mis-estimation of the NXB counts. Furthermore, the spectrum during the soft-flares shows a harder profile than the average spectrum for the entire night-earth observations. If we plot the satellite positions on the Earth, the soft-flare periods are concentrated in two regions: the south west to SAA and above Hawaii (crosses in figure 4). Although the soft-flares could be eliminated by masking these regions, the associated loss in exposure would be too large. We found that RBMCNT, i.e., a counting rate of the radiation belt monitor (RBM; ref. T. Ohashi et al. 1996), can reject the soft-flares efficiently. As shown in figure 6, RBMCNT distribution during the soft-flares (thick line) is almost flat, in contrast to the distribution for the entire night-earth observations (thin line) which is concentrated towards RBMCNT < 25 c/s. Therefore we reject the time intervals when RBMCNT > 300 c/16 s = 18.75 c/s everywhere, and especially RBMCNT > 100 c/16 s = 6.25 c/s in the dangerous areas (solid line in figure 4). We add the RBMCNT distribution map on the Earth in the appendix of this article for reference.
The frequency of the soft-flares has a slight correlation with the angle
of the terrestrial magnetic field to the XRT direction. The soft-flares seem
to happen more frequently when
~ 60--120 deg.
It is possible that the physical origin of the soft-flares is electrons which
come into the GIS from the direction of the XRT.
From these investigations, we have arrived at the conditions to reject the hard- and soft-flares as summarized in table 1. To avoid high NXB periods, we usually restrict H02 in the range of 15--45 c/s, too. We call the procedure to apply these conditions "flare-cut". Correlations between H02 vs. NTE and between COR vs. NTE after the full flare-cut are shown in figure 7. Compared with figure 5, the humps around H02 ~20--40 and COR ~ 10--12 have almost disappeared. The standard deviation of NTE in each bin is approximately equal to the square root of the average NTE of that bin, which means the Poisson statistics. Note that we must multiply 32 s to convert counting rate (c/s) to the original counts. This means that the NXB counts can be successfully estimated by H02.
Table 1: Summary of the flare-cut conditions _______________________________________________________________________________ Condition Reason _______________________________________________________________________________ 15 c/s <= H02 <= 45 c/s to avoid high NXB H02 <= 1.5 x f(COR)c/s for hard-flares RBMCNT <= 300 c/16 s = 18.75 c/s everywhere for soft-flares RBMCNT <= 100 c/16 s = 6.25 c/s in the dangerous areas (figure 4) for soft-flares _______________________________________________________________________________
Figure 6 RBMCNT distribution for the entire night-earth observations (thin solid line) and during the soft-flares (thick line).
Compared with figure 5, the humps around H02 ~ 20--40 and COR ~ 10--12 have almost disappeared. The standard deviation of NTE in each bin are approximately equal to the square root of the average NTE of that bin, which means the Poisson statistics. This means that the NXB counts can be successfully estimated by H02.
3 NXB Reproducibility in the GIS
When analyzing the GIS data, either spectrally or spatially, we usually subtract background derived from blank-sky observations. In such cases, the background reproducibility is limited by two factors; the field-to-field fluctuation in the CXB brightness, and the NXB variation. To generate a NXB spectrum or image to be subtracted from the raw CXB data (which includes NXB), we integrate the night-earth data which are screened on the same conditions as the raw CXB events, at the same detector region (for a spectrum) or in the same energy band (for an image). In fact, the NXB count rate can vary by ~ 40% in the range of COR > 6 GeV c-1 (figure7). If the exposure time is long, e.g. 100 ks, this variation is averaged to a reasonable level of ~ 10% (table2) depending on the length of the observation. It would be better to take into account the NXB variation by some methods. As seen in the previous section, H02 can be a good indicator of NXB. Thus we choose H02 to predict the NXB variation. Below, we explain how to generate the expected NXB spectrum for a specified on-source observation.
Figure 7 Almost the same as figure 5 except that the flare-cut (table 1) is applied.
Figure 8 H02 sorted spectra (left panel) and radial profiles
(right panel) of the GIS counts during the night-earth observations in the
0.6--7.0 keV energy band. The mark `+' represents the data when H02 is in the
range 50--80 c/s, '
' 40--45 c/s,
'
' 30--35 c/s, and '
'
20--25 c/s. H02 are integrated for every 16 s. The flare-cut (table 1)
is applied.
We sort the night-earth data according to the level of H02 counts
estimated in every 16 seconds and by H02 in every 5~c/s step, i.e.,
15--20~c/s, 20--25~c/s, 25--30~c/s, etc. H02 is counted in the 16 s
time interval including the arrival time when the corresponding event is
detected. Figure 9 shows the night-earth spectra and radial profiles
corresponding to five H02 intervals. As H02 increases, the
spectrum gets softer and the slope of the radial profile gets steeper towards
the detector rim. It is due probably to the increase of the
soft-component of NXB described in H. Kubo et al. (ASCA News No. 2).
We usually restrict H02 to the range of 15--45 c/s, where the influence
of the soft-component is not severe. Figure 8 shows distribution of
exposure time out of the entire night-earth observations, sorted as a function
of H02. We denote each exposure time
,
and each spectrum
(PI),
both of which are counted or created on the same condition and in the same
detector region as the on-source observations. As described above, the sorting
steps of H02 is 5 c/s. When
is the exposure time in the
i-th
H02 bin during the on-source observation, the NXB spectrum contained in
the on-source data is estimated as:
(1)
We call this method "H02-sorting method".
Figure 9 H02 distribution for the entire night-earth observations. Net exposure time (sec) in each 5~c/s bin of H02 is given in the figure. The flare-cut (table 1) is applied.
To check the reproducibility of NXB by the H02-sorting method, we split
the entire night-earth data into 122
x 10 ks intervals. Then we compared the counting rate of each interval with that
estimated by H02-sorting method using the entire night-earth data. The
results are shown in figure 10. Crosses in the upper panel show the actual
counting rate of each interval, and the histogram show the prediction. It
seems that 10 ks-integrated NTE varies more than the prediction. The middle
panel shows a ratio of divided NXB to estimated one (crosses). There is a
general trend of increasing counting rate with time, which means that the
long-term NXB increase which is seen in figure 2 cannot be reproduced by the
H02-sorting method. As seen in the lower panel of figure 10, the mean
value of observed/predicted ratio is 0.9983 and very close to 1.0, which is
rather natural because their mother groups are the same. The standard
deviation of the distribution is 0.07429, which mean 7.4% rms variation of the
NXB counts. This value includes a Poission error of 3.4%; therefore,
systematic error of the H02-sorting method for 10 ks observation is
thought to be 6.7%. If we use the average NTE instead of the
H02-sorting method prediction, the systematic error increases to 7.8%.
Thus the H02-sorting method reduces the systematic errors by about 1% level.
Since the
/d.o.f
of the line fitted to the ratio shown in middle panel is 412.4/121 = 3.4, the
systematic error will be reduced to
,
if the prediction of H02-sorting method are corrected for the long-term
trend.
This long-term NXB increase of ~ 15% in two years can be a serious problem for a faint source analysis as well as for an extended source analysis (extreme case is the CXB). No compensation of this long-term trend results in a NXB under-subtraction for recent observations especially in harder energy bands, which may cause an artificial hard tail. In fact, it can make ~ 0.1 difference of the CXB photon indices between two observations at intervals of two years. We, therefore, adjusted the normalization of the NXB image or spectrum estimated by the H02-sorting method with a formula:
scaling factor.(ascatime - 45878400), (2)
where ascatime represents the time in second from 00:00:00 (UT) on 1 January 1993, and is calculated from the mean observation time.
The systematic errors described above are summarized in table 2. We conclude that we can achieve a ~ 3.5% rms reproducibility in regard to NXB count rate, by a combined use of the standard background rejection, the flare-cut procedure, the H02-sorting method, and the long-term NXB trend correction. In the following, we quote 3.5% (rms) as a typical systematic uncertainty associated with the NXB estimation.
_______________________________________________________________________________ data NXB 0.7-7 KeVrms estimation error screening subtraction counting rate time ________________ condition method (c s^-1 cm^-2keV^-1) fraction 10 ks 40ks _______________________________________________________________________________ COR >= 6 GeV c^-1 average NTE 4.01 x 10^-4 92% 16% 14% flare-cut average NTE 3.59 x 10^-4 74% 7.8% 5.6% flare-cut H02-sorting method 3.59 x 10^-4 74% 6.7% 5.3% _______________________________________________________________________________
4 Summary
The night-earth data from June 1993 to June 1995 shows a long-term NXB increase by ~ 15%. We found NXB correlates well with H02, if hard- and soft-flares are carefully excluded. We can estimate the NXB counts with sim 15% error, when naive NXB rejection of COR >= 6 GeV c-1 and an estimate by averaging night-earth data are applied. We can achieve 3.5% reproducibility at best with flare-cut, H02-sorting, and long-term trend correction.
Figure 10 Figures to check the NXB reproducibility by H02-sorting method. Upper panel shows the actual NTE (crosses) and the H02-sorting method prediction (solid line), in the 0.7--7.0 keV energy band, for every 10 ks exposure. Middle panel shows ratios of the NTE to the prediction. The solid line shows the best fit line, and fitting results are indicated in the figure inset. Lower panel shows a distribution of the ratios. The mean value and the standard deviation for the distribution are indicated in the figure.
5 References
Kubo, H. et al., 1994, ASCA News No. 2, p14
Ikebe, Y. et al., 1995, ASCA News No. 3, p13
Ishisaki, Y. et al., 1995, ASCA News No. 3, p19
Ishisaki, Y., 1996, PhD thesis, the University of Tokyo
Ueda, Y., 1996, PhD thesis, the University of Tokyo
Ohashi, T. et al., 1996, Publ. Astron. Soc. Japan 48, 157--170
Makishima, K. et al., 1996, Publ. Astron. Soc. Japan 48,
171--189
Proceed to the next article
Return to the previous article
Select another article