# Current caveats for MARX¶

## General Caveats for Monte-Carlo simulations¶

marx is a Monte-Carlo tool, that means it uses random numbers when simulating the X-ray flux and the response for Chandra. For example, if the source spectrum indicates a constant flux between 1 and 10 keV, marx will create photons randomly from this interval. If only a small number of photons is used in a simulation, it is possible (though not very likely) that a lot more photons are softer than 5 keV. For a large number of photons, the result will eventually converge to a constant spectrum.

Random numbers are used for all interactions in marx, not only when selecting the energy of the photons, but also to see where in the aperture it enters Chandra, how it bounces of the mirrors, how it defracts from the gratings, etc. Thus, even if the physical model for a source is perfect, the simulated spectrum and a Chandra observation will differ.

Furthermore, this means that two marx simulations with the same input parameters will give different results, unless they use the same sequence of random numbers (set the parameter RandomSeed to a positive number to obtain the same sequence of random numbers).

## Spatial Dependence of the Quantum Efficiency¶

The current version of marx does not incorporate the QE uniformity maps and bad-pixel files that give rise to spatial variation in the QE. Consequently, exposure maps and ARFs created by default in ciao will be inconsistent with MARX simulations. For simulated observations on ACIS-S3, this difference should be small since spatial variations in the QE are relatively small on this CCD. However, simulated ACIS-I observations will be effected to a greater degree due to the larger QE variations produced by CTI effects.

For users of CIAO 2.3 and higher, the tools mkarf and mkinstmap include the ability to turn off the QE uniformity maps through the use of ardlib qualifiers. For example, a calling sequence like:

unix%  mkarf detsubsys="ACIS-7;uniform;bpmask=0" ......


will produce an ARF on ACIS-7 (ACIS-S3) but with bad pixel processing disabled (bpmask=0) and without the effects of the CALDB non-uniformity files included. The resulting ARF will be consistent with a marx simulation. A similar call to mkinstmap can be used in conjunction with mkexpmap to create exposure maps appropriate for marx simulations. This technique is illustrated in Creating CIAO-based ARFs and RMFs for MARX Simulations.

## ACIS Response Functions¶

CIAO includes a couple of different tools for creating ACIS response matrices (RMFs): mkacisrmf and mkrmf. The mkacisrmf tool is designed for the analysis of CTI corrected data, whereas mkrmf creates an RMF for non-CTI corrected data. The response algorithm implemented in marx is based upon the calibration data used by mkrmf. Hence the PHAs generated by marx for the ACIS detector represent non-CTI corrected values and as such are consistent with the responses generated by mkrmf but not with mkacisrmf. Consequently, users should continue to use the mkrmf to create RMFs that are consistent with their marx simulations. More information about using mkrmf in the context of a marx simulation may be found in Creating CIAO-based ARFs and RMFs for MARX Simulations.

Alternatively, marxrsp may be used to apply any RMF to a marx simulation with the caveat that the mapping from photon energy to PHA does not vary over the detector.

## Mismatch between the FEF-based response and the CALDB order-sorting tables¶

As mentioned above, marx generates non-CTI corrected PHA values. This is accomplished by mapping the incident photon energy to a PHA value using a probability from derived from the most recent non-CTI calibration data (CALDB acisD2000-01-29fef_phaN0005.fits). The CIAO tool tg_resolve_events assigns a diffracted order to each event by comparing the event’s ACIS energy to its dispersion coordinate. The ACIS energy window for a particular order is tabulated in a CALDB order-sorting table (OSIP).

For non-CTI corrected data, the CALDB order sorting table (acisD2000-01-29osipN0006.fits) was computed using a much older version of the non-CTI response data (acisD2000-01-29fef_phaN0002.fits). These files (acisD2000-01-29fef_phaN0002.fits vs acisD2000-01-29fef_phaN0005.fits) differ mainly in the region around the Si K edge (~1.8 keV). As such, a comparison of a marx spectrum with the expected spectrum of the input model will show strong systematic residuals near 1.8 keV.

## ISIS Pileup Fitting Kernel¶

The default parameters for the pileup fitting kernel in ISIS, but also in Sherpa and XSpec have been calibrated for point source extractions. Specifically, the values correspond to a circular extraction region 4 ACIS pixels in radius. Although marx can be used to include the effects of photon pileup for any arbitrary spatial and spectral source model, the fitting kernel may need to be adjusted for larger extraction regions. In particular, the psffrac parameter represents the fraction of the Chandra PSF contained within the extraction region and may need to be increased for larger regions. Note, however, that for real data, larger extraction regions will include a higher fraction of unpiled background photons complicating the fitting of the piled source spectrum. As such, it is recommended that this value be allowed to vary during the spectral fit. See the ISIS manual for more discussion of the pileup fitting kernel.

## Chandra Aimpoint Drift¶

marx does not currently take into account of temporal drift in Chandra’s HRMA aimpoint. Fortunately the effect of the drift is generally negligible and should not be a concern for Chandra proposers.