Comparing MEG Spectra to HEG spectra


The HETGS has two different grating types that disperse into two independent spectra. The medium energy gratings (MEGs) have an energy range of about 0.4-7 keV, depending on the observation parameters, and the high energy gratings (HEGs) have an energy range of about 0.8 to 10 keV. Because they are built into the same structure, the MEG and HEG spectra are obtained simultaneously, facilitating cross calibration even for variable sources. This is an update of the HETG flight calibration paper, which contained some results comparing ACIS-S quantum efficiencies (QEs). A preliminary version of the ACIS-S QE analysis is also available.

Data used in this analysis

For most observations (see Table 1), the target’s zeroth order was offset by 20” and the SIM was shifted by -3 mm; exceptions are given in the table. The ACIS focal plane temperature was reduced in the middle of January, 2000, so two observations were obtained at -110°, while the focal plane was at -120° for the remainder.

Obs ID

Obs Date

Exp. Time

PL Norm

PL gamma

3C 273 459


38600 0.025 1.66 a
3C 273 2463


26695 0.021 1.62  
3C 273 3456


24531 0.016 1.59 b
3C 273 3457


24849 0.013 1.37 c
3C 273 3573


29680 0.016 1.62  
3C 273 4430


27750 0.024 1.61  
3C 273 5169


29863 0.014 1.52  
PKS 2155-304 337


38666 0.037 2.67 a
PKS 2155-304 1705


25508 0.035 2.45  
PKS 2155-304 1014


25508 0.037 2.32  
PKS 2155-304 3167


29653 0.045 2.52  
PKS 2155-304 3706


27713 0.016 2.45  
PKS 2155-304 3708


26624 0.020 2.49 d
Notes a T_ACIS = -110, y_offset = 0, SIM_Z = 0 mm        
  b SIM_Z = -5.6 mm        
  c SIM_Z = -8.1 mm        
  d SIM_Z = 4 mm        

Table 1: Observations used in this analysis.

3C 273 varied by ± 35% about a mean normalization of 0.018 ph/cm
2/s/keV (at 1 keV) while PKS 2155-304 varied by up ± 50% about a mean of 0.032 ph/cm2/s/keV. The photon indices varied by ± 0.20 about means of 1.570 and 2.483 for 3C 273 and PKS 2155-304, respectively.

Correcting for CCD QE differences

The dispersion relations differ by a factor of two, so photons of a given energy will be detected at different locations in the focal surface. There are two different CCD types in the detector system and the quantum efficiencies (QEs) are different, so before comparing fluxes at a given energy, the relative QEs of the detectors must be verified. A HETG flight calibration report showed that there were up to 15% discrepancies between the FI and BI CCDs, which could be corrected empirically using with a correction function. The BI QEs have now been updated so this correction is no longer necessary. Fig. 1 shows that the ratios of the BI to FI counts are consistent with the new QE models to within about 5%.

Fig. 1: The ratio of data from BI CCDs to data from FI CCDs by comparing +1 and -1 orders in LETGS (diamonds and triangles) and HETGS data (crosses and + signs). The values represent the correction to the BI QEs that would be needed in order to obtain agreement between the FI and BI data. The dashed line is a polynomial fit to the data that deviates from expectations by no more than 5% over the 2-40 Angstrom range, showing that the new BI QEs correct for previous systematic errors (see the LETGS calibration preliminary report).

Correcting for HRMA shell differences

The HEG and MEG spectra involve separate independent portions of the high resolution mirror assembly (HRMA). The HEGs are mounted behind the inner two shells ("4" and "6") while the MEGs are mounted behind the outer two ("1" and "3"). Hence the calibration of the HRMA shell effective areas will also effect the HEG-MEG spectrometers relative calibration when in-flight calibration observations are used. The analysis here includes an update of the Ir constants and a nominal 20 Angstrom overlayer on the mirror surfaces. The HETGS data were used to determine this overlayer thickness. See the presentation by Diab Jerius at the Chandra Calibration Workshop for modeling of the overlayer. In a preliminary report by Herman Marshall, presented at the Chandra Users' Committee meeting in January 2005, the HRMA overlayer thickness was estimated to be 17 +/- 5 Angstroms. In a more recent analysis using updated HRMA reflectivities provided by Diab Jerius, the overlayer thickness came out to be 20 +/- 5 Angstroms. A web page is in preparation with the final results and a new set of HRMA reflectivities will be released.

Comparing derived spectral parameters

Once the BI fluxes agree with FI fluxes for the same grating, one may compare the spectral indices obtained by the MEG to those obtained from the HEG data. Fig. 2 shows a comparison of spectral fits to MEG and HEG data, plotting the HEG-MEG fit parameter differences as a function of the HETGS best fit. A simple power law model was fit independently to the MEG and HEG spectra of 15 observations of 4 sources: Mk 421, PKS 2155-304, 3C 273, and 1H1821-643. The column densities are all small enough to be ignored in HETGS data and were fixed to Galactic N_H values: 1.45e20, 1.2e20, 1.7e20, and 3.8e20, respectively.

The results indicate that there are systematic differences between the MEG and HEG spectral parameters of order 8% in the spectral normalization, A, in the sense that the HEG gives slightly larger values. Similarly, the photon indices derived from the HEG data are steeper by about 0.1 than those derived solely from MEG data. The statistical uncertainties on the photon indices are estimated to be about 0.05 for spectra with A ~ 0.02 ph/cm
2/s/keV. There may be a trend that the systematic differences are largest for the steepest spectra. The systematic differences are not likely to be related to any intrinsic source properties because the same biases are seen in spectra that have positive curvature (such as 3C 273, which is flatter at high energies) as in those with negative curvature (such as PKS 2155-304, which is steeper at high energies). Uncertainties in the ISM absorption on these spectra are generally small.

Fig. 2: A comparison of spectral fits to MEG and HEG data. Each symbol represents the difference or fractional difference in the fit value (vertical axis) against the fit quantity — normalization at 1 keV or photon index — for an observation of a source with a nearly featureless power law spectrum.

Deriving a correction to the grating efficiencies

Comparing MEG and HEG fluxes for these sources in adaptively binned spectra, shows that only mild corrections to the grating efficiencies would be needed between 1.7 and 17 Angstroms. The analysis does not provide a way to indicate whether the MEG or the HEG efficiencies are in error but the correction is calculated as if the HEG efficiencies are correct so that the values give the corrections to the MEG efficiencies that would bring MEG fluxes into agreement with the HEG fluxes.

There are significant differences above 15 Angstroms (below 0.83 keV). The deviation below 0.8 keV was unexpected and requires further investigation. The HEG data disperse to the edge of the detector at 18 Angstroms so there are no data to test how the MEG and HEG agree below 0.7 keV. Beyond 17.5 Angstroms, the correction is not reliable. A correction file in text format is available here.

Fig. 3: The ratio of MEG and HEG fluxes as a function of wavelength. The data points give the correction to the MEG efficiencies needed to obtain agreement between the HEG and MEG fluxes for the 13 observations used in the analysis. The dashed line is a polynomial fit to the data between 1.7 and 18 Angstroms.


Herman Marshall
hermanm @
Last updated April 27, 2005