The Swift-XRT GRB Catalogue was originally published by Evans et al. (2009) (hereafter ‘E09’) and contained Swift-XRT results for 327 GRBs. This web-site provides a continually updated version of the data given in that paper. Specifically on these pages we present for each GRB detected by the Swift-XRT:
which corresponds to Tables 5—11 of E09. We also here provide some figures showing the aggregate properties of the XRT GRB sample, comprising probability density functions of the parameters in the above fits, and scatter plots of the temporal vs spectral indices for each phase of each GRB, with regions 'permitted' by theory marked; these correspond to Figures 5, 6, 7a, 8a and 10 of the earlier paper. Furthermore, for each individual GRB we present each of these figures with the parameter space covered by that GRB marked on, showing how any given burst compares with the sample as a whole.
The data can be accessed via the Index table, which tabulates the results for all GRBs, and via individual object pages, accessible via the index table, which present detailed results for a given object.
The analysis techniques used in this online resource are largely the same as in E09 with any differences detailed in the relevant sections below.
This website is automatically kept up to date after every receipt of GRB data from Swift, however when a new GRB is detected it will not be added to this catalogue until 24 hours after the trigger, as details such as light curve shape can be unclear at early times.
As with all UKSSDC products, these data may be freely used by anyone, provided appropriate credit is given. Where these data are used in any publication, Evans et al. (2009, MNRAS, 397, 1177) should be cited. In addition to this if the enhanced positions are used, Goad et al. (2007, A&A, 476, 1401) should also be cited, and where the light curves are used, Evans et al. (2007, A&A, 469, 379) should be cited. In the acknowledgments section, please use the following text: This work made use of data supplied by the UK Swift Science Data Centre at the University of Leicester.
While this process is fully automated, imperfections in the system are inevitable and from time to time, manual corrections may be made; these are invisible to the end user.
This procedure has been completely rewritten since E09 was published
The automated flare identification process is an iterative procedure. The underlying light curve is modeled with the functional form of Willingale et al. (2007); if the XRT data start within 2 ks of the trigger and contain at least 10 bins (true for the majority of bursts) then two components are used, to reproduce the prompt and afterglow components. The time at which both of these peak is frozen at 10-5 s, i.e. a generic point before our observations begin. From this point the following process is followed:
As is implicit in the above, this iteration continues until no new flares are found or old flares extended.
The significance of a flare is defined as the sum of (Data-model)/Errdata) across the time-range of interest. For a candidate to be confirmed as a real flare it must pass one of the following criteria:
These criteria were derived largely by trial and improvement. Examining all GRB light curves up to 2011 September by eye suggests that this procedure has both false positive and false negative rates of around 3%. Note that flare identification can be a rather subjective process, thus these values are indicative, rather than precise.
The light curve fitting process will be summarised below, but is largely as documented in E09 with two exceptions: the threshold for adding a break is 4-σ and we do not integrate the model across the bin. The former occurs because, with a large number of light curves and fit attempts, the false positive rate was too high when accepting breaks at the 3-σ level. The latter is purely a matter of computational efficiency.
For each light curve, once the times of flares have been excluded, an unbroken power-law is fitted to the data. Breaks in the power-law are then added, one at a time and the curve refitted. This is continued until the light curve contains 5 breaks, the addition of an extra break would leave no degrees of freedom in the fit, or 2 successive breaks are added, but not justified by the improvement to the fit.
To add a break, the current fit is compared to the data, and the χ2 contribution of each run of data above or below the model is determined. The new break is then added at the start of the run with the largest contribution, however like all the other parameters, it is free to vary. Once the new model is fitted an f-test is used to compare the new fit with the previous fit: the new break is only considered justified if the probability of the improvement being chance (i.e. a result of 2 new free parameters) is <6.2 × 10-5, i.e. the break is required at the 4-σ level. Even if the break is not necessary, a subsequent fit with an extra break is carried out, but is compared to the last accepted fit. This is because sometimes the f-test in the first instance simply compares two bad fits. (For example, consider a light curve which is well described by a doubly-broken power-law. In some cases the unbroken power-law and single-broken power-law both give poor fits, and the inclusion of the first break is not justified. However, the addition of the second break suddenly gives a much better fit, and the 2-break fit is justified at the 4-σ level compared to the 0 break fit.)
Each light curve is automatically classified, according to the scheme in E09 This groups light curves into one of 5 categories:
The first four of these are illustrated in Fig. 9 of E09 reproduced to the right (larger version); the "oddball" category being a catch-all for any burst which does not fit neatly into the other groups. Classification in groups b—d is self-explanatory. As in our paper, a burst is considered canonical if it shows a flattening break, following by a later steepening break, however whereas in our paper these breaks had to change α by at least 0.5, in this work we relaxed this threshold to 0.3. Note that the fourth phase of the canonical light curve is not necessary for a burst to be classified as canonical.
As with light curve fitting, the automatic classification can occasionally be incorrect, usually due to a 'curved' decay slope resulting in several power-law segments being required to fit a single phase in the simple schema depicted in the figure to the right. These will be corrected manually.
For each phase of the power-law decay, as determined by the light curve fitting, a spectrum is determined and fitted with an absorbed power-law.
The absorber contains two components: the first is fixed to the Galactic value, the second is free to vary. Where a spectroscopic
redshift of the GRB has been published, the second component is a
zphabs, with its redshift fixed at this value, otherwise,
like the Galactic component, it is a
phabs. Note that the fit results tabulated are the photon indices,
N(E)∝E-Γ. In the figures however, for comparison with theory, we use the energy index,
F(E)∝E-β, where β=Γ-1.
Times of flaring are not excluded from these fits. Figs. 7c, 8c of E09, reproduced to the right, showed the distribution of the change in column density or spectral index caused by removing times of flaring. This is slightly biased as not all GRBs have flares, but it is clear that, in the majority of cases, the effect of flares on the spectral fits is not significant. Users can of course build spectra for customised time regions via the "Add time-sliced spectrum" link on the spectrum repository page for the burst of interest.
All of the spectral files, ready for use in
xspec, can be downloaded for each spectrum. For
details of these files please see the spectrum
The Probability Density Functions, or PDFs, show the probability density distribution of a given fitted parameter. On this website we present 4 PDFs, corresponding to Figs. 5, 6, 7a and 8a of E09. These show, respectively, the distribution of the light curve power-law decay index (α), the light curve break times (Tb), the intrinsic column density (NH) and the spectral energy index (β). These latter two distributions are derived only from the time-averaged spectral fits.
On the individual burst pages these PDFs also show the values for that specific burst, enabling easy comparison of a specific GRB with the population as a whole. More information is given in the target page section below.
The Fireball model for GRB afterglow can be expressed in a series of closure relationship, relating the temporal decay (α) at a given time to the contemporaneous spectral energy index (β). For each phase of each the light curve types a—d we provide a scatter-plot of (α, β), corresponding to Fig. 10 in E09 (except that the red and black points of panels e) and f) in that paper are here separated into individual plots). As in the paper, the regions permitted by the standard closure relationships are marked by the grey band (slow cooling, pre jet-break), blue band (post jet-break), and dark grey lines (fast cooling). These all assume no energy injection, however they mark the lower limit of the region permitted if energy injection is included. i.e. the addition of energy injection permits points with higher β for a given α
As with the PDFs, on the individual burst pages the closure relationship figures also appear, with the specific GRB also marked on, to allow comparison of that burst with the sample at large. More information is given in the target page section below.
The catalogue index page contains access to the individual burst pages, and the aggregated information pertaining to the catalogue of bursts as a whole. This is split into two parts, the table, and the figures. Additionally, the index page provides an interface to access individual burst pages. This can be done either by clicking on a GRB name in the table, or by entering the burst name or trigger number in the search box in the top section of the page.
The table and figure sections can be hidden (or re-shown) by using the checkboxes towards the top of the page. Additionally, each section begins with a link to hide the section, which is replaced by an option to show the section, if it is hidden.
The figures presented on the index page have already been described, here the table is discussed.
The results table contains, in a single place, Tables 5—11 of E09. All errors in this table are given at the 90% confidence level. The default view of the table gives some summary information for each burst. This can be adjusted as described in a moment. The GRB name links to the results page specific to that burst. If a spectroscopic redshift for the burst is known, this serves as a link to the publication announcing the redshift. Redshifts are entered manually, thus there may be some delay between publication of a value and its appearance in this table.
Note that the table takes a few moments to load after changes are made. For this reason, it does not reload when the options are changed; instead you must click on one of the "Update Table" buttons.
This panel is the only one which does not affect the results table, instead affecting how you select columns (it is coloured red to highlight the fact that it is different from the other panels). The options in this box determine the detail level of the table controls that are shown. By default only the "Basic options" control panel is shown, but this can be disabled/other controls enabled. Those panels will be detailed below. Note that there is some overlap between the controls, as will be explained below, however the checkboxes are "aware" of this and will automatically update accordingly.
This panel contains a single control which determines how the table is presented. The default option is as an HTML table. This can be replaced either with "ASCII (inline)", in which case the HTML table in the web page is replaced with preformatted text embedded within the web-page; or "ASCII (new page)" in which case when you click on "Update Table" you are redirected to an ASCII version of the table. The ASCII data are delimited by pipes (|). The HTML and inline ASCII tables can easily become much wider than the webpage, in which case they will overflow and your browser should provide scrollbars.
This panel provides high-level control over which columns are included in the table. Naturally, changing these options affects the columns over which finer control can be gained using the panels described below. If those panels are also visible, any changes made to the boxes in this panel will immediately alter the selections in those, more detailed, panels. For example, if you select "Time-averaged data" and the "Time-averaged fields" panel is visible, then all of its boxes will be checked/unchecked accordingly. By default only the summary data are selected. The options here are fairly self-explanatory, and each of them has a corresponding detailed control panel of its own, as described in the following sections.
This panel allows you to select which of the summary fields are included in the results table. The options are:
phabscomponent with its density fixed at this value.
phabscomponent or, if the redshift is known, a
zphabscomponent with its redshift fixed.
This panel allows you to select which results of the time-averaged spectral fits are included in the table. As for the time-resolved spectral data, these are split into the two XRT operating modes: Windowed Timing (WT) and Photon Counting (PC). Note that changing the "Time-averaged spectral data" option in the "Basic Options" box will cause all of the items in this panel to be selected/deselected. The columns here are self-explanatory.
This is the first of the controls affecting time-resolved data, and like the "Basic Options" panel, it offers high-level controls: selecting/deselecting options in here also changes the options which are controlled in details in the "Light curve data" and "Time-resolved spectral data" boxes. By (de)selecting a "phase" in this box you (de)select all of the results associated with that phase of the afterglow. Phases are defined by the light curve fit, with the breaks in the power-law delimiting the phases. For each phase selected in this box, the decay index; break time; column density, photon index and flux (WT and PC); and the associated uncertainties, will be included in the table.
This section allows you to select individually which of the decay indices, break times, and associated errors are reported in the table.
This section allows you to select individually which of the various spectral fit components and uncertainties for each light curve phase are reported in the table.
Each GRB detected by the XRT has its own page giving a detailed view of that GRB. This page contains useful links, some summary information and navigation tools, and then the main results, which are divided split into three sections:
The navigation tools allow you to jump directly to one of these sections, or to hide some of the sections to make the page easier to read.
The summary box contains the best-known XRT position of the GRB (an enhanced position, if possible, otherwise a PSF-fitted position), its corresponding Galactic coordinates, the number of breaks in the best-fitting light curve, the light curve classification, and details of the time-averaged spectral fit. For the latter the PC-mode data are shown by default, unless no such spectrum exists in which case the WT mode values are given.
This section contains details of the light curve fit. Errors are given at the 90% level. The figure shows by default the best fit, that is, the fit determined by testing a range of power-law fits, and requiring a 4-σ improvement in the fit (according to an f-test) for a new break to be kept (see light curve fitting for details). Any times which were identified as flares and excluded from the fit are marked with cross-hatched boxes, and dashed vertical lines indicate where the power-law breaks are.
Beneath the plot some details are given, and two tables appear. The left-hand table details the parameters of the light curve fit. The right-hand table gives some information about each fit attempted, indicating why the best fit was so chosen. Clicking on the number of breaks in the relevant column of this table causes the main light curve figure, and the fit results table, to update, showing the fit and parameters for the model selected.
This section contains spectra generated for each phase of the light curve, as delimited by the break times in the best fit. Note that spectra are only created based on the best light-curve fit. Times of flaring were not removed from these spectra (see above). The format of this section is identical to that of the spectrum repository, and includes figures, results tables, and a file to download with all of the spectral data. Spectral fit errors are given at the 90% confidence level.
The last section of the GRB-specific results page shows the figures collating the results for all GRBs (as shown on the index page) but with the specific burst also marked independently, allowing for easy comparison of this burst with the population as a whole. For the probability density functions, the values for this burst are marked by solid vertical lines, with dashed vertical lines indicating the 1-σ bounds on the parameter. For the temporal indices and break times, where there is more than one value, each value is marked by a different colour, following the colour scheme described below. For the spectral parameters (note that these are time-averaged spectral values), WT mode data are given in blue, and PC in red, following the normal XRT convention.
For the closure relationships, the values for this specific burst are marked in colour; where multiple points are present, from different spectral modes or because several light curve phases correspond to the same closure relationship plot (e.g. sometimes the steep-decay phase of a canonical light curve contains a break and thus two 'phases' of the light curve and hence 2 spectra) the colours follow the colour scheme described below. The colours are ordered by light curve phase, and within that, by mode (i.e. WT1, PC1, WT2, PC2 etc.). Unless the light curve is classified as an oddball, only the closure relationship pertaining to the specific class to which this burst belongs, are shown. For odd-balls, each closure-relationship plot is shown, with all of the light curve phases shown on each plot, following the colour scheme just described.
For the plots in the "Comparison" section of the GRB-specific pages, where multiple values for that GRB are marked on a given plot, values are coloured in the following order (the QDP colour scheme:)