Background subtraction issues
The usual method for correcting a spectrum from an imaging detector
for background is to extract a spectrum from another part of the
image and subtract that from the source spectrum. This assumes that
the source region has the same background spectrum as the background
region. However, instrument performance often varies across the face
of the detector either due to mirror vignetting or changes in detector
efficiency. There are three possible strategies for dealing with this.
- Instead of using a different region of the same image take background
from the same region of the detector for a different image, for instance
a deep field observation. This has the advantage that it eliminates
the effects of any changes in instrument performance over the field. It
has the disadvantage that it assumes that the X-ray background is the
same in the source and background images and it assumes that the
non-X-ray background does not vary over the time between the source
and background observations.
- A second alternative is to use background from the source image but
correct the background spectrum for the position-dependent effects.
This depends on knowledge of the spectrum of the background, both
X-ray and non-X-ray, and requires considerable care because the
X-ray background may have to be corrected by a different amount than
the non-X-ray background. It is also difficult to know how to take
this operation into account in the error analysis.
- The final possibility is to use background from the source image and to
fit simultaneously the source and background spectra. The background
spectrum is fit using a model for the X-ray and non-X-ray backgrounds
and the source spectrum is fit using a the background model plus whatever
model is required for the source. The parameters for the background model
are linked between the source and background spectra. This method has
the advantage that it is statistically well-defined but it is more
cumbersome. Again it is necessary to be careful about variations of
the X-ray and non-X-ray backgrounds across the image.
The ROSAT satellite has some vignetting of the X-ray background and
also a radial dependence of the non-X-ray background. The ASCA satellite
has strong vignetting of the X-ray background and little spatial
dependence of the non-X-ray background (except at the edges of the GIS).
ASCA also has X-rays scattered into the image from outside the field-of-view
through reflections off only one of the pair of foils or off the back of
a foil. The extra X-rays mitigate the vignetting effect of the telescope
and make the X-ray background flatter across the detector.
Keith Arnaud
(kaa@genji.gsfc.nasa.gov)