We provide a selection of response matrices which may be useful for
performing simulations with XSPEC. The user should however take note of some
important caveats regarding their validity:

1) The matrices were produced using the XMM Science Simulator v2.0. The data
used in this version of the simulator, while broadly correct (especially
for mirror area), remained unverified by the ground calibration campaign.
Examples where optimistic results might be erroneous would be in the analysis
of hydrogen column densities and soft spectra, where we remain unsure about 
the effects of the filters and low energy CCD performance. Similarly at the
highest energies we have not yet a detailed knowledge of effective areas after
all appropriate event processing is performed, so high temperature/hard power
law estimates obtained from XSPEC simulations should be examined with 
scepticism.

2) The data are not even valid vis a vis the latest release of SciSIM, so that
the user should take great care in attempting to analyse data produced by the 
latest version of SciSIM using THESE matrices. There are some different detector 
and mirror efficiencies modelled in the latest SciSIM, and CTI effects in the 
CCD simulation are different, so systematic errors will inevitably result. Also 
compared with the effective area files used in the current W3PIMMS 
for example, one might expect to see of order 10% discrepancies.

3) The data for the matrices were generated by SciSIM but then subject to some
emulation of ground processing, such as event selection, and also charge
transfer efficiency correction. We have included events with the largest 
acceptable size in terms of number of pixels hit. In reality the observer 
might choose only single and bi-pixel events for spectral analysis. In such a 
case the effective areas could be substantially lower. These features are all 
expected to be significantly different in their final  form, so the user 
should again not treat these matrices as anything but indicative of final 
responses.

4) The matrices carry no explicit representation of PILE_UP effects. If the user 
is considering simulating a point source with relatively high flux, then
s/he should check with the W3PIMMS software to get an estimate of the pile-up
expected, or perform some experiments with the latest SciSIM tool to judge
the amount of pile-up which might occur. Unless we are eventually able to
calibrate and recover these effects in analysis, the conventional fitting
with fixed matrices will lead to inaccuracies beyond the systematic 
calibration uncertainties. See also the descriptions in the User Handbook

5) Very limited testing has been performed with these matrices, beyond 
checking that they can be read in with XSPEC, and that broadly correct spectra
result, in terms of count rates, in a very limited set of conditions. We would
welcome feedback on their use.

INSTRUMENT		EPIC PN			EPIC MOS

RMF File		epn_rmf_all.fits	emos_rmf_all.fits

ARF Files		epn_thin_arf.fits	emos_arf_thin_all.fits
			epn_med_arf.fits	emos_arf_med_all.fits
			epn_thick_arf.fits	emos_arf_thick_all.fits

There is a set of related .ascii files generated by FDUMPing the headers of the 
files, so you can see the contents. 

Generation was performed by running SciSIM with a large number of monochromatic 
input stimulations. The outputs were spline interpolated to the resulting 
energy grid. Because of the finite input rate adopted to ensure the simulation 
did not extend for a semi-infinite time, the astute user would see that 
(for example by inputting a Gauss model in XSPEC) that a small amount of 
pile-up is evident in the output spectrum. This will give some representation 
of real pile-up in that counts are moved from lower to higher energies, 
but it will be stystematically wrong. Also note that due to the automatic 
nature of the generation, it means that some output channels at up to twice
the maximum input PHA channel could have been stimulated, and this has lead to 
a very large RMF matrix.

The ARF files are generated with 2381 energy bin rows
The RMF file for PN is 8460 output bins, and for the MOS, because of its lower
effective area at high energies the pile-up effects limited the ouput bins to
6089.

FUTURE RELEASES
We are still awaiting the results of the analysis of the instrument ground
calibration campaign from EPIC, and these may require a modification of the
response files we adopt before launch. 

The various deficiencies described above also are being addressed in the eventual
calibration database by providing response generation tasks in the SAS. 
These are intended to take the user-selected data set, and provide a matched 
effective area for the vignetting and CCD efficiency at the source location, 
to account for the event selection criteria applied etc.. The response 
should also include some first order corrections for the amount of pile-up 
observed in the source and corrections for charge transfer losses. It will 
be appreciated that this is a necessary yet  complex task which is still 
under development, so we cannot aim to provide substantive updates before 
the AO-1 closure, and indeed one would expect that eventual in-orbit 
experience will require substantial revisions to our knowledge.

D Lumb, 31 Jan 1999