Background characteristics of the GIS--by H. Kubo, Y. Ikebe, K. Makishima & the GIS Team, University of Tokyo
IntroductionThe GIS (Gas Imaging Spectrometer) experiment has been designed to achieve a very low and stable level of non X-ray background (NXB) over the wide field of view of 25 arcmin radius and the wide energy band of 0.7-10 keV. This is essential when investigating extended X-ray emission with low surface brightness, e.g., outer regions of clusters of galaxies.
We first note that the GIS onboard background rejection is carried out via rise-time discrimination and spread discrimination. The onboard rise-time acceptance window is set to be rather loose, which can be tightened further on ground in the course of data analysis. This can be done via an xselect command called "gisclean." By applying this processing, the NXB count rate can be reduced by about 15 percent on the average, and almost by as much as 50 percent above 5 keV. We therefore recommend all the users to use "gisclean." However it is cautioned that the high-time-resolution mode of the GIS sacrifices the rise-time information in the telemetry data, so that the additional NXB reduction becomes unavailable.
Using the PV phase data, we have investigated the NXB characteristics of the GIS in detail. All the results below apply to the data obtained after 1993 May 28, when the spread discriminator was enabled. We have also applied the tighter rise-time window to all the results in this report. Since the two GIS detectors (S2 and S3) exhibit very similar NXB characteristics, we do not specify the detector below unless otherwise stated. Throughout this report, we exclude data acquired with the XRT elevation angle lower than 5 degrees above the local horizon, when contamination from atmospheric components becomes significant.
The average NXB spectrumFigure 1 shows the average NXB spectrum, obtained when the XRT was pointing at the night Earth. The data have been averaged within 17 arcmin radius of the center of the two detectors, and over satellite orbital phases where the cutoff rigidity (COR) for cosmic rays is larger than 10 GeV/c. The cm2 in the count rate unit refers to the detector area, not to the XRT area. Thus the in-orbit NXB level is about 1x10^(-3) c/s/cm^2/keV below 1 keV, and (4-7)x10^(-4 )c/s/cm^2 keV above 1.5 keV. This is 2-3 times larger than that measured on ground, and very close to what we expected to achieve.
Figure 1: Typical GIS spectra accumulated within 17 arcmin of the detector center, discarding data acquired with low (<10 GeV/c) cutoff rigidity and those acquired with low (<5 deg) elevation above the local horizon. The "night Earth," "blank sky" and "day Earth" data correspond to NXB, NXB plus CXB, and NXB plus scattered solar X-rays, respectively.
The NXB spectrum exhibits several instrumental features, including Cu-K line at 8.03 keV, a hump around 5 keV due to Xe L-edges, Sn-L line at about 3.4 keV, a possible blend of Mg-K (1.26 keV) and Al-K (1.49 keV) lines, and so on. All these lines and edges arise from materials used for the detector.
Figure 1 also shows two more GIS spectra, obtained during the blank sky pointing and during the sunlit Earth pointing. The former data consist of the NXB plus cosmic X-ray background (CXB), while the latter consist of the NXB and solar X-rays scattered off the atmosphere. Mg-K, Si-K and S-K lines in the sunlit Earth spectrum is of solar origin (ionized species), while the Ar-K line (neutral) is due to atmospheric fluorescence. A close agreement of the night Earth and day Earth spectra above 5 keV visualizes the NXB stability.
Position dependence of the NXBThe NXB spectrum above 1 keV is fairly constant both in shape and normalization as a function of the detector position, up to about 12 arcmin from the detector center. Beyond this radius, the NXB count rate starts increasing towards the detector rim as illustrated in Figure 2. As shown in Figure 3, this increase is mostly limited to a spectral range below 1 keV. Therefore, we may presume that there are approximately two NXB components:
- a "hard" component, which has a flat spectral shape and an almost constant intensity over the whole detector area, and
- a "soft" component, which rises below 1 keV and with a normalization that significantly increases towards the detector rim. Physical origins of these two components are yet to be identified.
Figure 2: Radial dependence of the NXB count rate across the GIS detector area, sorted according to the cutoff rigidity. Plate scale is about 1 arcmin for 1 mm.
Figure 3: The NXB spectra near the central area and towards the rim of the GIS detector. Difference is significant below 1 keV.
In contrast, the CXB intensity is strongest at the XRT axis, because of the vignetting. The observed CXB intensity at r=20 arcmin off-axis is roughly 2/3 of the on-axis CXB intensity. When integrated over the whole energy range and for COR > 10 GeV/c, the CXB to NXB intensity ratio is about 3:1 on axis, about 2:1 at r=12 arcmin, and about 1:1 at r=20 arcmin. The overall (NXB+CXB) background brightness is therefore constant within about 20 percent across the entire detector area.
There seems to be some azimuthal dependence in the NXB intensity, particularly in the extreme rim region. Our preliminary analysis indicates that such an enhanced NXB count rate is most prominent in the S2 rim with smaller DET-Y values, and in the S3 rim with larger DET-Y values. Caution is needed when analyzing faint signals in these outermost regions. This issue is currently being investigated, and will be reported in more detail in the future.
Time variability of the NXBThe background count rate during the blank-sky pointing, summed over the entire GIS detector area and over the entire energy range, is 0.45-0.55 c/s per detector. This includes about 0.3 c/s of Fe-55 calibration isotope and about 0.085 c/s of CXB.
As already indicated in Figure 2, the NXB variability is predominantly correlated with COR (cutoff rigidity) along the satellite orbit. For COR > 7 GeV/c, the NXB intensity in c/s/cm^2 at a given value of COR may be approximately expressed as
f(r,E;COR) = f(r,E;COR=10) x [1 - ((COR-10)/20)]where r is the radius measured from the detector center, and E is energy. However, the "soft" component may have somewhat stronger dependence on the COR than the "hard" component, as evidenced by Figure 4. More detailed study of the NXB variability and reproducibility is under progress.
Figure 4: Comparison of the NXB spectra acquired under different ranges of cutoff rigidity.
Background-subtraction schemeThe currently available simplest scheme of background subtraction is to utilize the blank-sky pointing data, which are sorted into broad COR ranges and are made publicly available. Each observer may compute the COR distribution of his or her on-source spectrum, and produce a background spectrum (including the CXB) from the publicly available blank-sky data so that the COR distribution becomes approximately the same between the on-source and off-source spectra.
For typical observations, a significant fraction of exposure is made over a COR range of 10-14 GeV/c with a fairly fixed distribution. Therefore, for the simplest background subtraction, one may limit the COR range to > 10 GeV/c in the data analysis, and simply subtract the blank-sky data accumulated for COR > 10. This is expected to give a NXB reproducibility significantly better than 10 precent, implying that the NXB uncertainty is less than several times 10^(-5) c/s/cm^2/keV. This criterion, when combined with the elevation angle cut (< 5 degrees), leads to a loss of about 25 percent of the useable data.
As a demonstration of the GIS background subtraction, we present in ASCA Gallery of this issue two GIS spectra for extended diffuse emission, one for the Galactic ridge emission (in which only the NXB has been subtracted) and the other for the Virgo cluster ICM emission (in which both NXB and CXB have been subtracted). Both spectra have been obtained via data accumulation over almost the entire GIS field of view, and employing the simplest background subtraction method mentioned above.
We continue investigations of the NXB characteristics in order to further improve the NXB reproducibility. We also plan to construct a reliable model NXB. The model, given the information on satellite orbit, attitude, environmental parameters (e.g. temperature) and so on, will predict a NXB count rate at a specified region on the detector over a specified energy interval. This will eventually allow every ASCA user to subtract NXB and/or CXB, either from the spectra or from the images, in a self-contained way.
Proceed to the next article Return to the previous article