High Energy Transmission Grating Spectrometer Calibration

This page provides an overview of the HETGS in-flight calibration. It is linked from the HETGS Users web page and from the Chandra Science Center Calibration web page.

In most areas, the HETGS is performing according to pre-flight expectations. For recent developments regarding HETGS calibration, see the work in progress section of this web page. References that may be of interest include:

• New web page  on the comparison of the XMM and Chandra spectra for several cross-calibration observations. Last updated 7/7/06
• Report  on the HETGS effective area, dated 10/14/2005, giving an update of the HEG and HEG efficiencies and a fit to the HRMA Ir-M edge;
• Marshall, Dewey, and Ishibashi (SPIE 2003) on flight calibration results up to August 2003;
• a report (pdf form) about the HETGS effective area, presented at the January 2005 meeting of the Chandra Users' Committee;
• a presentation  on the HETGS effective area, prepared for the Calibration Workshop held on 6-7 November 2002;
• Viewgraphs (PS form)  in PS form of the HETGS calibration overview for the Calibration Review held on 30-31 October 2001;
• other reports at the Chandra Calibration Workshops; and
• a status report (PS form) presented to the Chandra Users Committee meeting in June 2001.

HETGS in-flight performance Summary

• Dispersion relation
• The relative accuracy of wavelengths assigned to events on the basis of the distance from 0th order has been reduced significantly by computing ACIS-S chip rotations and positions to better than 0.05 pixel (see a memo in PS form by Bish Ishibashi). Thus, wavelengths should now be accurate to less than 0.0055A (0.0028A) for MEG (HEG) spectra.
• The absolute accuracy of wavelengths was reduced by eliminating relative errors (above) and taking out a systematic fractional error of 1.34e-4 (or 40 km/s) in the MEG spectra by adjusting the MEG period from 4001.41 A to 4001.95 A. (cf. Dan Dewey's HETGS TechNote from October 2004). This adjustment is similar in magnitude to that determined by Marshall, Dewey, and Ishibashi (SPIE 20003).
• Line response function -- See the HETGS Technotes, where Dan Dewey (MIT/HETGS) has verified that the 0th order images are consistent with models using MARX.  The cores of the lines are used to determine the preliminary E/dE plots of the MEG and HEG results are available from his HAK analysis of an observation of Capella.  See also the study of the line response function by David Huenemoerder and the comparison of profiles from the Line Spread Function library with lines from Capella by Bish Ishibashi. The wings of the lines are under analysis but do not appear to deviate significantly from pre-launch expectations. Some of these results are shown by Marshall, Dewey, and Ishibashi (SPIE 2003).
• Cross dispersion profiles -- The cross dispersion profiles were derived from the HETGS observation of the Crab pulsar for the MEG and the HEG spectra. The cross dispersion enclosed energy fraction (PS form) was computed in rectangular apertures of various sizes as a function of wavelength using an observation of Mk 421 (observation ID 1714). Also plotted in the image is the 1D profile of a point source with minimal pileup. Note that a point source is always narrower than the HETGS cross dispersion profile. See also a May 2004 memo (PS form) by Bish Ishibashi for encircled energy fraction verification using non-randomized data.
• Dispersion angles -- The HEG and MEG dispersion angles were measured relatively early in the calibration phase by Dan Dewey and verified by HLM (MIT/CXC).  Dan's analysis is available.
• Focus is stable (Updated 26 October 2004). The average FWHM of the zeroth order is 38.1 +/- 0.19 microns, giving an angular size of 0.780 +/- 0.004 arcsec. Note that the HETG does not affect the PSF of images so this size should be the same as that obtained without the HETG in place.
• System Effective Area -- Prior to implementation of new HEG and MEG efficiencies, the estimated HETGS relative uncertainties were based on the simple power law fits to the HETGS spectra of the Crab pulsar and Mk 421. The following estimates of the systematic uncertainties of HETGS spectral fluxes were suggested, after correcting for the system effective area using the calibration data products as of September 2005:
  • 10% for 1.5 < E < 6 keV (both MEG and HEG),
  • 20% for 6 < E < 8 keV (HEG only),
  • 20% in the Si-K edge region (1.83-1.84 keV) (both MEG and HEG),
  • 20% for 0.8 < E < 1.5 keV (both MEG and HEG),
  • 30% for 0.5 < E < 0.8 keV (MEG only),
  • 50% in the O-K edge region (0.525-0.57 keV) (MEG only).
  The quoted uncertainties are meant to give the uncertainty on fluxes determined in one band relative to those in another band.  Absolute flux uncertainties have not yet been verified but are expected to be good to a better than 5-10%.  The uncertainties include systematic errors in many instrument components, including the HETGS efficiencies and ACIS filter and QE models, and are meant to reflect the system performance.

  The ACIS team has provided QE models that eliminate most of the discrepancies between the BI and FI chips, as documented in a preliminary report (PS form) on the LETG/ACIS calibration and by Marshall, Dewey, and Ishibashi (SPIE 2003). There is also a significant contamination buildup which is also observed in direct ACIS spectra. This contamination was first detected in LETG/ACIS data but clearly affects all ACIS observations. See also the paper by Marshall et al. (SPIE 2003b), which provides details about the composition of the contaminant.

  A web page documents some work on comparing the MEG and HEG spectra as of April 2005. A more refined report  on the HETGS effective area, dated 10/14/2005, gives an update of the HEG and HEG efficiencies and a fit to the HRMA Ir-M edge. Using these updated HEG and MEG efficiencies as well as the suggested correction to the Si-K and Ir-M edges, fractional residuals should now be less than 5% from 0.5 to 8 keV.

Work in progress (Updated 10/14/05)

1. The effects of pileup are examined in an analysis by Dan Dewey and are apparent in this plot comparing the MEG and HEG spectra of the Crab pulsar.  (The dashed line is the MEG spectrum, which shows a deviation from the HEG spectrum near 2.1 keV, which is associated with the EA increase due to the mirror Ir M edges.)   The HETGS calibration observation of Cyg X-2 is very heavily affected by pileup, to the extent that the continuum is very difficult to measure.  From the Crab and Cyg X-2 observations, we surmise that pileup has a greater effect on HETGS spectra than would be predicted from naive models and that the spectral region that is most often affected is where the MEG EA peaks near 2 keV.

   Pileup correction may be considered to be a calibration task. At present, there is a method for correcting for pileup within ISIS and the jdpileup function in sherpa (see the pileup model thread). A simple pileup correction method seems to be sufficient for calibration tasks, which does not involve iteration or forward folding. We will be assessing the accuracy of pileup correction methods with calibration data.

2. We will be testing the models of the high order efficiencies of the HETG using data from 3C 273 and Cyg X-2, among other sources. This work is on hold while we complete analysis of the first order efficiency, which has higher priority.

3. There are a few stages in grating data analysis that involve data filtering or corrections which have not yet been examined as part of the calibration program. Issues that may have an effect on the reliability of derived spectra include:
   • order sorting -- using a very wide order sorting selection window will yield the best reliability at the expense of statistical uncertainty (see Dave Huenemoerder's presentation at the 2002 Chandra Calibration Workshop for a discussion of order sorting in ciao);
   • destreaking -- the streaks in the S4 chip are flagged and then removable using the destreak tool (see the "destreaking ACIS data" web page) but for very bright sources some real events may be removed so the threshold could be raised to reduce this loss;
   • removing "afterglow" events -- the afterglow events are isolated in time and chip location and can be interpreted incorrectly as sources but are generally not a problem for grating spectra, so one should ordinarily eliminate this step of the ACIS processing or reset the afterglow status bit for grating data analysis using this thread; and
   • continuous clocking mode -- this mode has been used often to observe bright sources but order sorting and background selection are sufficiently different that the calibration of the resultant spectra is not readily comparable to that of spectra taken in standard timed exposure (TE) mode.