NOTICE: This Legacy journal article was published in Volume 1, May 1992, and has not been updated since publication. Please use the search facility above to find regularly-updated information about this topic elsewhere on the HEASARC site.

New EXOSAT ME Argon Calibration

Frank Haberl

Max Planck Institute (MPE)

The new ME Argon calibration is now available as part of the general XSPEC spectral fitting package. The program VIMAT creates response matrices in XSPEC format for the EXOSAT GSPC and ME detectors. VIMAT uses the new calibration for the ME argon detectors, the ME xenon calibration has not been changed. The new ME calibration is based on seven of the nine Crab observations performed during the EXOSAT mission. Two observations were rejected due to high solar activity. The Crab spectra for the argon detectors A-H were fitted to a powerlaw (photon index -2.1) attenuated by photoelectric absorption (column density 3.1 x 10²¹ H cm^-2) after folding through the response. This article summarizes the changes which have been made to the old calibration to create the new response matrices.

The Updates

The shape of the gain curve has been changed to a polynomial of the form

E = a1 * V + a2 * V² + a3 * V³ (1)

where V is the output of the front-end electronics in mV and E is the energy in keV. The relation between V and the channel number C is still V = (C - 0.5) * 1500/128. With the new gain curves, the fits to the Crab spectra improve above channel ~ 60 and the variations of the coefficients are much smaller than with the old curve which consisted of exponential terms.

The electronic acceptance factors are now the pure rise time veto acceptances. These have been determined from Crab observations with rise time veto on and off. These observations were too short to provide good statistics on the acceptance factors above channel ~ 35. The acceptance factors have therefore been averaged for all the detectors above channel 7 and set to constant above channel 35. The rise time acceptance values are plotted in Figure 1. The additional factors to account for the sinusoidal residuals around 3 keV have been removed. A possible origin of these "wiggles" is a variation of the width of individual energy channels. In the new calibration, the channel boundaries of each individual channel were allowed to vary. The change of the channel width obtained from eqn. 1 was limited to +/- 4%. These "channel width ratios" were determined for each Crab observation and detector separately, but then averaged for each detector because no time variations were seen.

Two detectors, D and G, showed gas leakage during the mission which resulted in bad spectral fits towards the end of the mission. For detector D, fits were improved by a linear decrease in pressure after 1984 day 288 from the nominal value of 2 atm. to a value of 1.85 atm. at the end of the mission. For detector G, the pressure value was reduced to 1.97 atm. after 1985 day 294.

A plot of count rates in the channel range 80-120 versus channels 5-40 shows excess counts in the high channels (Fig. 2). To create Figure 2, all the good quality spectra from the EXOSAT database were used. Only 3 sources (Crab, Cyg X-1 and GX301-2) are strong and hard enough to be detected above channel 80. Their count rate ratios (not shown in Figure 2) lie well above all the others which show a linear relation. A linear fit gives a ratio of

Figure 1

Figure 2

1.2 x10^-3, i.e., 0.1% of the counts are detected above channel 80. This behavior might be explained by a small region of very high gain in the detector. To account for it, a second linear gain curve has been added in the calculation of the response matrix. The average weighting factor for the linear gain curve is 5 x 10^-4 but varies from detector to detector (see Table 1). An example of the gain curves for the first Crab observation of detector A is shown in Figure 3. The linear gain coefficient and a weighting factor were free parameters in the fits to the Crab spectra. The effect of the high gain curve is very small and only improves the fits to the highest channels (above 80). At the highest channels a bump in the residuals is still visible and it is not recommended that one use channels above 100. Also, channel 4 should not be used due to the unknown position of the low level discriminator.

Figure 3

Finally, the effective areas had to be adjusted to give, on average, the same powerlaw normalizations for the Crab spectra as the old calibration. The new values are given in Table 1. The new boundaries, resolution, and efficiencies were calculated for each detector and Crab observation and stored in separate files. These files are used by the program to calculate the detector response matrix. Between the Crab observation boundaries, resolution and efficiencies are linearly interpolated.

                              Table 1


    	   Detector	     Weighting	     Effective areas
			                         cm2 

	      A	             10.0 x 10-4           203.9
	      B	             3.1 x 10-4            193.4
	      C	             0.0                              201.1
	      D	             4.2 x 10-4            203.3
	      E	             6.1 x 10-4            195.0
	      F	             0.94 x 10-4           199.7
 	      G	             5.5 x 10-4            199.0
	      H	             6.1 x 10-4            204.0

To check the new calibration, spectra from various sources have been fitted. Spectra which show a strong iron line could be used to check that the gain is right. Fits to spectra from clusters of galaxies where the line is expected to be around 6.7 keV and spectra from GX301-2 with a line energy at 6.4 keV show no differences in line energy between the old and new calibrations. Spectra from detector C give systematically low line energies with the old calibration. The discrepancy of the line energy increases with time. The new calibration did not solve this problem. As an example, in Table 2 the iron line energies are compared for an observation of M87 on 1984 day 360 using the old and new calibration. Detector C failed on 1985 day 232. A comparison of the Crab spectra shows a drop in the count rates in channels 6-10 some time between the second and third Crab observations. This different shape of the spectra causes too low gain parameters which results in too low line energies for spectra after the first Crab observation. The effect is not understood and spectra from detector C should not be used to determine iron line energies.

                              Table 2


	           Detector	    Line energy (keV)
	                      Old cal.          New cal.

                       A     6.68+/- 0.15     6.71+/- 0.15
                       B     6.66+/- 0.20     6.70+/- 0.20
                       C     6.48+/- 0.25     6.49+/- 0.25
                       D     6.93+/- 0.20     6.75+/- 0.20
                       E     6.91+/- 0.15     6.92+/- 0.15
                       F     6.77+/- 0.15     6.79+/- 0.15
                       G     6.76+/- 0.15     6.73+/- 0.15
                       H     6.65+/- 0.20     6.72+/- 0.20

To summarize, fits to spectra from various sources have shown that the new calibration gives, on average, the same result up to about channel ~ 65. The new calibration fits the Crab spectra better at higher channels, and channels up to ~ 100 can be used in spectra of bright sources.

Proceed to the next article

Return to the previous article

Select another article