Skip navigation tools.

• Home
• Quick Look
• Data Archive
• Data Analysis

• About Swift
• GCN
• Swift Results
• Links

---

Swift-XRT enhanced ground positions

Introduction

The positions determined by the XRT can be improved in both accuracy and precision by using the UVOT to accurately determine the spacecraft pointing. The sky co-ordinates of standard XRT data are subject to a systematic error of 3.5" (90% confidence). By matching the UVOT field of view with the USNO-B1 catalogue and using the known mapping between the UVOT detector and the XRT detector, we can correct the XRT sky co-ordinates, reducing the systematic error component to 1.4". Using this technique, we obtain positions for 90% of Swift-detected GRBs with the present operational setup. The 90% confidence error radius is <2" for 90% of these.

See Goad et al. (2007, A&A, 476, 1401) and Evans et al. (2009, MNRAS, 2009, 397, 1177) for full details.

We have verified that this process is still accurate following the gyro anomaly of 2007 August. See Verification for details.

Usage

When quoting enhanced XRT positions of GRBs (even if those positions were given in a GCN Circular) please cite Goad et al. (2007, A&A, 476, 1401) and Evans et al. (2009, MNRAS, 2009, 397, 1177). If you use these positions in a paper, please include the following text in the acknowledgements section: This work made use of data supplied by the UK Swift Science Data Centre at the University of Leicester.

Algorithm

The basic principle is that the GRB is located within the XRT's field of view, its position is transformed to the corresponding position on UVOT and then the UVOT field of view is matched to the USNO-B1 catalogue to provide the correct 'enhanced' position. Since the exact direction in which the spacecraft is pointing will change from one snapshot to the next this process has to be performed for each snapshot separately. Further, the UVOT observes through various filters with different boresights, so the process must be applied separately to the different filters. The data are thus split into 'overlaps' (where XRT and UVOT data are both obtained), a position calculated for each overlap, and then these positions combined.

The first step is thus to split the observations into overlaps. We take any observation which began within 12 hours of the first observation, and create one overlap for each valid UVOT exposure which is at least partly covered by the XRT operating in Photon Counting (PC mode). We take as much XRT data as possible without combining snapshots or extending into another valid UVOT exposure. Fig. 1 (below) demonstrates this. A 'valid' UVOT exposure is one taken through the V, blue or white filters. We do not use the UV filters since there are fewer stars and the astrometric corrections can become unreliable.

Fig 1. An example of how overlaps are defined. In the first snapshot, there are two overlaps. The first XRT exposure starts with the snapshot, and ends midway between the UVOT exposures. The second XRT exposure starts immediately, and continues to the end of the snapshot.
For the second snapshot there is only one valid UVOT image, so XRT data are taken from the entire snapshot.

Once the overlaps have been defined, the algorithm can be summarised thus:

• For each overlap:
  1. Create a sky-coordinate X-ray image and exposure map, and search it for sources. Centroid on these sources using a PSF fit which uses exposure maps to correct for bad columns and hot pixels. If an XRT position of the burst is already known, choose the source closest to this position. Otherwise, choose the brightest source in the BAT error circle for the first three overlaps, and thereafter choose the source nearest to the weighted mean of these 3 positions.
  2. If the position thus found is based on less than 10 X-ray photons, reject it and move to the next overlap.
  3. If the position is more than 20" from the known XRT position/mean of the first 3 positions, reject it and move to the next overlap.
  4. Convert the sky position to the equivalent XRT detector position. Then, using the known alignment of the XRT and UVOT, convert from XRT co-ordinates to UVOT co-ordinates, and thence to UVOT sky co-ordinates.
  5. Match the contents of the UVOT field of view with the USNO-B1 catalogue to determine the transformation from UVOT sky co-ordinates into RA and Dec relative to USNO-B1.
  6. If this transformation is >10" Swift may have been moving significantly during the overlap. Reject it and move to the next one.
  7. Use the transformation to convert the XRT GRB position into RA and Dec relative to USNO-B1.
• Calculate the weighted mean of all the positions found from the above.
• If any of these positions lie more than 3-sigma from the weighted mean, reject them as outliers and recalculate the weighted mean.
• Create the detailed results web page, and add an entry on the results index page.

For each overlap, when the UVOT field stars are matched with the USNO-B1 catalogue, a residual error is returned. This is added in quadrature to the error in the PSF fit, to give the error in the overlap. These errors are then propagated through the weighted mean calculation to give an error on the final position (90% confidence). A systematic error of 1.5" is then also added in quadrature; this arises primarily from the uncertainty in the XRT to UVOT transformation.

Warning: Very occasionally there may be a serendipitous source within the BAT GRB error circle which is brighter than the GRB even in the first images. In such cases the enhanced position will likely be calculated for the wrong source. As soon as this is known, the process will be manually overridden and the position recalculated for the real afterglow.

Results Index Page

The results index page contains a short summary for each GRB observed by Swift. The first column gives the name, and the target ID. Assuming some data were available, the name is a link to a more detailed result page, described below. The second column is a link to the UVOT filter sequence for this burst, which shows the time and duration of the exposures in each filter. The next three columns give the position and error which the enhancement process produced.

Sometimes the process fails to find an enhanced position. When this happens the last 3 columns of the index table will report this. This can happen for one of three reasons: there were no snapshots with both PC mode XRT data and V, blue or white UVOT images, the UVOT data could not be corrected with respect to the USNO-B1 catalogue, or all of the overlaps were rejected for the reasons described in the "Algorithm" section. In the latter case the results table will list how many overlaps were rejected because the source has less than 10 counts, was an outlier (described above), was not detected, or was >20" away from the expected position, so was likely to be a serendipitous source (see example image below). A detailed results page will still be produced and linked to via the burst name, and this page gives more information on why no position was found.

Detailed results pages

For each burst for which a UVOT-enhanced position was found, a detailed results page is produced. This can be accessed either directly, via the URL: http://www.swift.ac.uk/xrt_positions/xxxxxxxx where xxxxxxxx is the target ID of the burst (i.e. the trigger number padded to 8 digits with leading zeroes), or by clicking on the burst name on the main results page.

The first table on this page reports in bold text the enhanced position and error, and also indicates on how many overlaps and how much exposure time this position is based. The second table then reports how many overlaps there were in total, and the overall exposure time, before detailing how many overlaps were discarded and why. The reasons for discarding overlaps are summarised below. See also the "Method" section, above.

• No source found: No XRT source could be found in the overlap.
• Low-count fit: A source was found, but contained fewer than 10 photons. Such sources are deemed questionable, so discarded.
• Spurious fit: A source was found, but it is more than 40" from the unenhanced XRT position/mean position of the brightest source in the first 3 images.
• Bad astrometric correction: The astrometric correction moved the source by >10".
• Outlier: The source lay outside the 3-sigma error in the weighted mean.

Finally, there are two images. The first shows a summed, aspect corrected UVOT image centred on the position of the burst, with the enhanced XRT position and error drawn in magenta. The second image has a larger field of view, and shows the positions returned by each overlap. Any which contributed to the final enhanced position are shown in green, while those rejected are shown in red if they contained fewer than 10 photons, and yellow for any other reason.

Position Accuracy Verification

In order to confirm the accuracy of the positions and their errors, we compared the positions produced with the positions of Swift-detected optical GRB counterparts. For 59 bursts with both UVOT and enhanced-XRT positions, we calculated the angular offset between the two positions, and divided it by the uncertainties in the positions (added in quadrature). If this ratio is <=1, the two positions agree. We found that an extra systematic error of 0.4" (in quadrature), was necessary to achieve 90% agreement with the UVOT, thus the overall systematic error is 1.36" The plot below shows the cumulative proportion of GRBs plotted against this ratio; as can be seen, 90% have a ratio <=1.

There was a small possibility that, during the gyro anomaly of 2007 August, the XRT to UVOT map may have shifted slightly. However, using calibration and GRB observations, taken since the return to normal operations, we have confirmed that this was not the case, and the positions and their accuracy are still valid.

Changelog

• 2009 Jun 5: A new set of UVOT teldef files have been released as part of the CALDB and this affects the positions to some degree; improving them. All of the positions have accordingly been rebuilt and the error calibration verified. There is some evidence that, from 2008, our error may be underestimated by around 5%, however at present the statistical evidence is marginal. We continue to monitor this situation.
• 2009 Jan 16: The improved positions, detailed in Evans et al. (2009, MNRAS, 2009, 397, 1177) were uploaded. The major improvement is the use of piled-up PSF profiles to fit bright sources (see example below). The old positions are still available.

GRB 080319C demonstrates the improvement for bright XRT sources yielded by using piled-up PSF profiles where appropriate. Note that this is an extreme case. The images below show a UVOT image of GRB 080319C -- the optical afterglow is obvious. The magenta circle shows the enhanced XRT position. In the left panel (version 2.0 of the software, no pile-up awareness) the XRT position is clearly offset. In the right panel (version 2.1, with pile-up awareness), the position agrees with the UVOT.