skip to content
 

HEASARC Staff Scientist Position - Applications are now being accepted for a Staff Scientist with significant experience and interest in the technical aspects of astrophysics research, to work in the High Energy Astrophysics Science Archive Research Center (HEASARC) at NASA Goddard Space Flight Center (GSFC) in Greenbelt, MD. Refer to the AAS Job register for full details.

Goddard Space Flight Center   ASCA Helpdesk
  ASCA What's New
  HEASARC Site Map
  NASA Homepage

HEASARC HOME ASCA HOME ARCHIVE DATA ANALYSIS PROPOSALS & TOOLS STUDENTS / TEACHERS / PUBLIC
ASCA Guest Observer Facility
ABOUT ASCA GOF SERVICES WHAT'S NEW DATA PROCESSING TIMELINES & MISSION INFO RELATED SITES GALLERY

Current Status of the GIS Response Function

--by M. Ishida (ISAS), Y. Ishisaki (U. of Tokyo), Y. Fukazawa (U. of Tokyo), K. Ebisawa (NASA/GSFC) and the GIS team

Introduction

The GIS team has been constructing and updating the GIS response function, based on the extensive pre-launch and in-orbit calibration activities. Here we report on the current status of the GIS response function, including several practical suggestions which may be of use to ASCA users.

The GIS Gain

Before describing the response function, we first report on the GIS gain (i.e., instrumental conversion factor from X-ray energy to pulse height) which is a quantity of fundamental importance in quantitative analysis of the spectrum. The gain can be monitored continuously in reference to radio-isotope 55Fe, which is attached to the edge of the focal plane for calibration purposes and emits X-rays with an energy of 5.89 keV at a counting rate of about 0.3 c/s. The isotope peak will easily be recognized in an X-ray image in which the whole focal plane area is specified.

Gain dependencies

The gain depends on three factors. Firstly, it varies up to +/-10 precent across the detector area, mainly due to non-uniform sensitivity of the photo tube used in the GIS. This effect has been calibrated extensively on the ground, and a slight adjustment has been made based on the in-orbit calibration. This information is summarized in so called "gain map." Secondly the gain depends on temperature of the photo tube, with a typical coefficient of approx. -1 percent/°ree;C as calibrated both on the ground and in orbit using the calibration isotope. Finally, there is a very slight long-term gain change in both S2 and S3; during first several months in orbit, the GIS gain decreased by 2-3 few percent, but the trend has gradually vanished. From April 1993 to April 1994, the S2 and S3 gain decreased by 3.1 percent and 2.8 percent, respectively. There is no noticeable sign of degradation in the GIS performance.

Gain corrections

When accumulating X-ray photons into an X-ray spectrum, the PH (pulse height) of each detected event should be converted into PI (pulse invariant) taking into account the above three factors. This "gain correction" is to be performed by the user separately for each observation, using either MKGAINHIST or temp2gain. NOTE: US Guest Observers do not have to worry much about the gain correction, since the gain correction is carried out for the US GO data at GSFC in the standard processing with MKGAINHIST (for AO1 data) or temp2gain (for AO2 data). Both of these gain-correction procedures assign 5.89 keV to PI=500 channel. The former routine uses the instantaneous peak energy channel of the isotope during the observation, and the gain map released by the GIS team. The latter uses the environmental temperature and the same gain map. In practice, the two procedures have almost the same reliability. We recommend using temp2gain, particularly for bright sources in which the increased dead time prevents accumulation of a sufficient number of radio-isotope X-rays.

Gain accuracy

How accurate is the GIS gain thus determined? We have calibrated the temperature coefficient and the long-term trend together, to an accuracy of +/-0.4 percent. The gain map, on the other hand, is accurate at present to within +/-1 percent in r = 15 arcmin from the detector center. Therefore the GIS gain (both in absolute and relative sense) is currently accurate to about +/- 1 percent.

The gain map is more uncertain near the outer rim of the detector. Therefore, if a spectrum is produced including the calibration isotope region, the isotope peak itself will appear at about PI=495 (5.84 keV) for S2 and PI=472 (5.56 keV) for S3, rather than at the correct channel of PI=500 (5.89 keV). Although it is a good practice for users to confirm the result of the gain correction by analyzing the isotope peaks, they are advised to keep this problem in mind.

The GIS Response Function

The GIS response function consists of two parts; the redistribution matrix file (RMF) and the ancillary response file (ARF).

Redistribution Matrix File (RMF)

The RMF is a matrix relating the X-ray energy to the pulse-height channel. A matrix element Mij is defined as the probability of detecting an X-ray photon of energy i at the pulse-height channel j. This is solely determined by the GIS, with no effect by the XRT. Since det(Mij) = 1, the energy-dependent quantum efficiency of the GIS is not included in the RMF. Fortunately, the RMF depends neither on the X-ray incident position nor on the photon integration region. Therefore there exists in principle only one RMF at a time, which should be the same for the two detectors (S2 and S3), and users do not have to generate it by themselves. The latest RMFs are gis2v3_1.rmf and gis3v3_1.rmf for S2 and S3, respectively. Distinction between S2 and S3 is purely conventional, and these two RMFs are in fact completely identical. Updated versions of the RMF will be supplied now and then by the GIS team.

Ancillary Response File (ARF)

The ARF defines total effective area of the XRT plus GIS as a function of X-ray energy, depending on the position in the field of view used for the observation and the region on the detector used for data accumulation. The concept of total effective area includes various effects such as; effective area of the XRT as a function of energy and off-axis angle; transmission of thermal shields of the XRT and the GIS; quantum efficiency of the GIS detector at a specified position; position-dependent efficiency reduction due to partial shadows cast by the support ribs in the detector entrance window; fraction of photons (as determined by the XRT+GIS point spread function) that falls inside the specified data integration region; and so on. Thus users should compose ARF suitable for their observation with ascaarf. In this procedure, the relation between the pulse height channel and the X-ray energy is taken from RMF.

S2 and S3

While RMF is common between the two GIS detectors, ARF should be produced separately for S2 and S3. This is mainly because the two detectors somewhat differ in relative alignments with respect to the XRTs. However, as the two GIS detectors are almost identical and similarly well understood, it is strongly recommended to use data from both S2 and S3 simultaneously. This can be done in the form of S2/S3 combined fitting. Practically, however, the S2/S3 difference is rather small. Therefore it would be more desirable if the S2 and S3 spectra were averaged channel-by channel, and respective ARFs were similarly averaged, so that the S2 and S3 data could be treated as a single spectrum. Presently this is unavailable in the standard data analysis system, and can be done only in the quick-look data analysis system.

The Crab fit

Figure 1 shows the GIS-S2 spectrum of the Crab Nebula fitted with an absorbed power- law model, using the standard way of gain correction and ARF generation as described above, and the latest version of RMF. The fit is fairly satisfactory, though not perfectly acceptable, in spite of the extremely good photon statistics. The obtained power-law photon index is about 2.1 and the absorbing column density is (3-4) x 10^(21) cm^(-2), somewhat depending on the analysis condition (see below). These values are close to the generally accepted values for the Crab Nebula, 2.08-2.11 and (2.7-3.3) x 10^(21) cm^(-2), respectively.

Figure 1: A power-law fit for the Crab spectrum with the gain determined by the standard procedure.

Limitations and Potential Problems

The 2.2 keV feature

In Fig. 1 the instrumental Au-M edges at about 2.2 keV (in the XRT) and Xe-L edges at about 4.8 keV (in the GIS) are both modeled with a reasonable accuracy, typically within 1-2 percent. The fit residuals show no particular local feature, except a hump at 2.0-2.2 keV. This is due to a slight gain mismatch (within the +/- 1 percent gain calibration uncertainty mentioned above), in combination with the sharp decrease in the XRT reflectivity across the Au-M edges. Thus an emission-line like feature could arise around 2.2 keV up to an equivalent width of 20-30 eV. In this particular case, the residual feature is almost removed by artificially raising the gain by 0.8 percent (which can be done with the gain command in XSPEC). Users are thus cautioned not to confuse this instrumental structure with real emission lines from H-like Silicon or nearly neutral Sulfur, both appearing in this energy range. This is particularly important when analyzing spectra with high statistics and those integrated near the detector rim where the gain map is less accurate.

Positional dependence

Both the XRT effective area and the XRT point spread function vary as a function of the off-axis angle. In addition, thickness of the 10-micron Beryllium foil used for the GIS entrance window depends slightly on position. Therefore the raw PI spectrum depends on the focal plane position used for the observation. This effect is in principle compensated by the ARF, which takes these effects into account so that the spectral fitting results should not depend on the focal plane position. We have confirmed this property by the Crab data obtained at various locations in the field of view. Over an off-axis angle range of 0-15 arcmin, the derived Crab index scatters only by about +/-0.03, the absorbing column density by +/-0.3 x 10^(21) cm^(-2), and the inferred Crab flux by +/-5 percent.

The ARF generation is expected to be significantly complicated for an extended source, because one would have to calculate different ARFs corresponding to different regions of the source, and combine them with relative weights by referring to the brightness distribution. We suppose it is still too premature to report on such image-spectrum coupled analysis of the ASCA data.

Data integration radius

Even for the same observation, the derived spectrum differs slightly depending on the radius of data integration region employed. This is because the XRT point spread function is slightly energy dependent, and the GIS position resolution improves towards higher energies. ARF is designed to compensate this effect as well, to yield constant set of spectral parameters without regard to the data integration radius. We have studied this issue again using the Crab data, to find that some dependence still remains.

For example, when we integrate a particular set of Crab data within a radius of 6 arcmin, the best fit photon index becomes 2.10 for both S2 and S3, while the hydrogen column density turns out to be 3.0 and 3.2 x 10^(21) cm^(-2) for S2 and S3, respectively. An integration radius of 3 arcmin for the same observation leads to a hydrogen column density larger by 0.5 x 10^(21) cm^(-2) for both S2 and S3, although the photon index remains the same. For 3C273, the photon index increases by about 0.02 when the integration radius is changed from 6 to 3 arcmin.

Therefore, for point sources as well as extended sources with <eq; 3 arcmin size (like the Crab), users are recommended to use 6 arcmin as a standard integration radius because the XRT encircled energy function is normalized at this radius. This is also large enough to absorb the remaining uncertainty in the XRT+GIS point spread function, as well as the approx. 1 arcmin uncertainty in the attitude control.

Absolute flux

Another aspect of the response function is the accuracy with which absolute fluxes of X- ray sources are measured. Currently, the 2-10 keV Crab flux we obtain via the standard analysis of the GIS data is 1.7(+/-0.1) x 10^(-8) erg s^(-1) cm^(-2). This is some 15 percent lower than the generally accepted value, e.g., 2.0 x 10^(-8) erg s^(-1) cm^(-2) in 2-10 keV measured with Ginga. Therefore, the absolute flux measurements may be subject to systematic uncertainty of this order. Users are reminded of this limitation when trying to derive absolute fluxes and luminosities based on the GIS data.

Low energy response

The low-energy cutoff of the GIS is determined at about 0.7 keV, primarily by the 10- micron thick Beryllium window, the effective thickness of which has been measured accurately on ground. However there exist some signals even in the pulse height region below 0.7 keV, because the GIS energy resolution function for soft X-rays exhibits a significant tail towards lower pulse heights. This effect requires a very careful tuning when we construct the response function.

We presume that the low-energy-tail effect is slightly underestimated in the latest response function, as shown in Fig. 2. This is a fit to the spectrum of an X-ray pulsar EXO2030+375, which is subject to a considerable interstellar absorption. A sort of artificial soft excess below 0.9 keV in the fit residuals demonstrates the current limitation in our understanding of the GIS soft X-ray response. We estimate that the current uncertainty translates into about (2-4) x 10^(20) cm^(-2) in terms of absorbing hydrogen column density. Therefore, users are advised to be aware of this limitation.

Figure 2: Spectral fit for EXO2030+375 by a power-law with absorption. The best fit NH is 2 x 10^(22) cm^(-2).

Summary and Future Prospects

As described so far, we believe that the XRT+GIS calibration activities are progressing well and are converging reasonably. We will continue in our activity of improving the calibration accuracy in all the aspects discussed in this report, and keep updating the response function on a reasonable time scale. Meantime we would like every user to be aware of the limitations described above when trying to derive scientific results from the GIS data.

Next Proceed to the next article Previous Return to the previous article

Contents Select another article