GRBs positions can be determined from three types of Swift XRT data. In the order that they become available after a trigger, these are: image mode, prompt-ground and ground. Positions from the latter two types can sometimes be improved by UVOT-enhancement or X-ray astrometry (ground data only). An overview of these data types, positions and improvements is available on the introduction page, here we describe them in more detail. The data types will be reviewed in the chronological order in which they are available, then the improvements will be explained.
The absolute sky co-ordinates of standard XRT data are subject to a systematic error of 3.5" (90% confidence), arising from the spacecraft's star trackers. This error is not inclued when the UVOT enhancement or X-ray astrometry processed are applied. These methods calculate the spacecraft pointing without reference to the star trackers.
NOTE In the text below, the "initial circular" refers to the GCN circular produced by the Swift team shortly after the trigger, announcing the detection of the burst.
The automated positions assume that the brightest object in the BAT error circle is the GRB: this is generally true. However, if there is a bright serendipitous source within the BAT error circle it may initially be mistaken for the GRB. This is very rare, and when it happens the Swift team endeavour to issue a correction at the earliest opportunity.
If the GRB lies close to the bad columns, the position determined is less reliable than normal, unless the centroiding algorithm is aware of the bad columns. The PSF fitting used with the prompt-ground data and by the UVOT-enhancement process does account for the bad columns, however the barycentric fit used in other positions does not.
When Swift slews to a new GRB the XRT takes a 0.1 s exposure in image mode, and the onboard software searches this image for a source. If successful, it performs a barycentric centroid, and then telemeters the image and the position to the ground. A GCN XRT position Notice is automatically produced. Should the onboard software be unable to find a source in the initial image, the process is repeated on up to two 2.5-s images. If it still fails, a Nack-Position GCN notice is sent. For full details of the onboard centroiding process, see Hill 2003.
The images are also telemetered to the ground. If no position was found automatically onboard, XRT Burst Support scientist (XBS) will combine these images and search for a source. If one is found a barycentric centroid is carried out. If no prompt-ground data (see below) are received, this position will be incorporated into the initial circular and a GCN position notice.
If the XRT enters Photon Counting (PC) mode during its first snapshot of the new GRB, any single pixel (grade 0) events above 0.5 keV that occur within the central 200x200 pixel region of the XRT are telemetered periodically to the ground. The XBS can then use this "SPER" data (which usually contains many more counts than the short images) to search for a source, and determine the position using a barycentric fit. As of mid-May 2007, this process has been automated, and the position is calculated using a PSF-fit which is aware of the bad columns on the CCD. This reduces the boresight uncertainty from 3.5" to 2.9". Full details of this are given on the SPER processing page. From February 2008 onwards, the UVOT enhancement (below) is also applied to SPER data where possible.
Should a position be found in the SPER data, this will be used in the initial circular. The SPER position will also be distributed as a GCN Notice. All positions derived from SPER data are published online on the SPER results page.
In the minutes post-trigger only limited data products are telemetered to the ground. The full dataset is downlinked to the Malindi groundstation, passes occur every few hours. Once the data are on the ground and processed, the XBS can use all available PC mode data to determine the GRB position. This is usually done using a barycentric fit and can be superior to previous positions since it comprises more data (typically several kiloseconds of exposure, as opposed to a few hundred seconds for the SPER data. It also contains single- and multi-pixel events). Although the barycentric algorithm does not correct for the bad columns on the CCD, the ground position is usually calculated from several snapshots of the target and Swift uses offset pointings to help keep the GRB (once identified) away from the bad columns. Thus the effect of bad columns on the ground positions is minimised. This position will be reported in the refined circular, unless a UVOT-enhanced position is available. The UVOT enhancement process is applied to ground data when possible, and ground positions can also be improved using X-ray astrometry.
If both XRT and UVOT data covering approximately the same time interval are available, the UVOT data can be used to astrometrically correct the X-ray position. Stars in the UVOT field of view are matched with the USNO-B1 catalogue, allowing the UVOT pointing to be calculated more accurately than is done onboard. Using the known mapping between the XRT and UVOT V-filter, the XRT pointing can thus be enhanced, correcting the position and reducing the systematic error.
This correction can be calculated for each UVOT exposure (in the White, V and B filters), provided PC mode X-ray data are taken which overlap in time with the UVOT exposure. If no such overlaps exists, or if the X-ray source cannot be found in the XRT data, no enhanced position is produced. The exact implementation of the method depends on whether it is applied to prompt ground or ground data. The prompt ground and enhanced ground documentation pages give full details.
Enhanced prompt-ground positions are distributed as described in the prompt ground positions section above. When an enhanced ground position is first produced for a GRB, a GCN Circular is automatically dispatched, to advise the community.
Note: The UVOT enhancement process was first applied in February 2007. In February 2008 a new version of the software, and a new catalogue of positions, was released. The new version provides more accurate positions with smaller error radii, and finds positions for a greater proportion of GRBs. The catalogue was revised again in January 2009, using new fitting techniques which correct for pile-up on the XRT. The most recent changes are detailed in the Changelog for the position enhancement..
If serendipitous X-ray sources are found in the X-ray field of view, these can be compared with ground catalogues for optical counterparts. These matches can then be used to astrometrically correct the X-ray position and reduce the systematic error. These positions are occasionally manually produced by the XRT team, and reported in the refined circular, or in a special circular. These positions are also automatically produced by Nat Butler, and are available on his website.