ROSAT Extended Sources

Introduction

This tutorial deals with the identification of moderately extended sources in ROSAT data, where "moderate" means more than about 15 arcsec, but less than 5 arcmin. The ASTERIX source searching program PSS produces a measure of a source's extension by finding that Gaussian which, when convolved with the point response, best fits the observed surface brightness distribution. A measure of the significance of the extension is also provided, and an estimate of the error. The problem with this method of extension testing is that a number of different effects can result in the observed spatial distribution of photons varying from the point response.

The model of the instrumental psf is not accurate. This of course is difficult to quantify. As soon as improved calibrations become available they are incorporated in the ASTERIX system, but it is difficult to get useful information on their accuracy.
Attitude errors. The important errors here are the ones which distort the observed photon distribution - simple shifts don't really matter unless you assume a souce position. Non-coherent attitude errors should be well modelled by a Gaussian source convolution and place a lower bound on the extension which is believable. A typical value for the random error in the ROSAT attitude reconstruction is about ? arcsec. Attitude errors which are coherent between sources could be a problem (but see below).
Last, but not least, the source may be extended.

How not to do it!

The naive method of extension testing is to set the PSS parameter EXTEN to value TRUE, and let it run over the entire field looking for extended sources. This is likely to produce misleading results for the first of the reasons listed above, ie. the quality of the psf model. PSS currently only works on 2D data and the usual way to run it is to assume a mean photon energy for the psf. This takes no account of the fact that sources are neither monochromatic nor uniformly so across the field.

Suggested method

The following is the method used by the (mythical) ICL procedure PSS_EXT_TEST, which will be invoked thus,

   ICL> pss_ext_test srclist rawdir+root sig_thresh outlist

where srclist is the input source list generated by PSS, rawdir+root is the combined directory and XRT root name for the observation being processed, sig_thresh is a threshold below which extension testing will not be performed, and outlist is the output source list.

Extract a spectrum from a small area around each source. Background subtract this spectrum using the background spectrum derived from that used to subtract the image.
Run the SPRESP program to create a psf composed of energy dependent psfs weighted by the counts in each spectral bin.
Run PSS on just that source, creating a results file containing single source with the significance of extension. A graph of significance as a function of the Gaussian convolution FWHM is also produced.

Confirmation

A few ways to confirm PSS's results,

Do a radial profile using the image processing system, and superimpose the psf using either the standard option (XRT_PSPC and VARPROFILE), or using the psf file created by the PSS_EXT_TEST procedure. Is the observed profile consistent with PSS's result?
Do other objects in field show extension similar in magnitude (and direction if non-symmetrical). At high galactic latitudes many objects in your ROSAT field will be background quasars, which generally have quite soft spectra. If your unidentified soft sources all show extension then something is probably amiss.
The adventurous could use IMSIM to create simulated copies of their input image and subject them to the same procedure. This would give you another handle on the validity of the significance of the extension.

This page was written by David Allen, is maintained by Richard Beard, and last updated on 6th October 1997. If you have any comments, please contact asterix@star.sr.bham.ac.uk