For a description of how the spectra are created, please see Evans et al. (2009, MNRAS, 397, 1177) which explains the process in detail. An overview of the algorithm is also given online.
When a spectrum or series of spectra are created, they are made available through a web page.
The automated fit performed in xspec
is an absorbed power-law which may not be
appropriate for your source, however you can download all of the appropriate files to
fit the spectrum yourself.
The results web page contains a section for each time interval you requested, at the top of the page is a series of links to the relevant spectra. Each section begins with a title line gives the spectrum name, and the time range covered by the spectrum. Then follows a link to download the spectral data. The download is of a gzipped tar file, containing the following files:
^{1} This is the name of the spectrum. If you did not specify a time region, it
will be interval0
, otherwise it will be the name you supplied.
^{2} i.e. the filename will either contain "pc" or "wt", corresponding to the XRT mode of the spectrum. If both modes were available, there will be 2 files.
After the tar file there is a series of links to postscript and gif mode
plots for each spectrum, and then a single plot is shown. If WT and PC mode
spectra were available, a single plot showing both spectra and their
(independent) models and residuals is shown, otherwise the only available
mode is shown. If the automatic xspec
modelling was not able to
fit the spectrum, the plot will show the spectrum alone, with no model or
residuals panel.
Below the plot the automatic fit results for each mode are shown, if available. Above the tables is a header giving the mode, and also the "Mean photon arrival" time for that spectrum. This is simply the mean time of all the events in the spectrum, and is not always intuitive: if the GRB is piled up for part of the interval of interest, the annular source region used at these times will reduce the number of photons arriving in this interval, biasing the spectra to later, fainter times: the opposite of what one would naïvely expect. It should also be noted that, if the spectra evolve during the time interval, the best fit will be misleading, being an average of a varying spectrum. However, the mean photon arrival time will give some indication as to which times dominate the spectrum.
The spectral-fit values themselves are given in the tables, with the 90% confidence errors (i.e. the maximum range a parameter can cover before the fit statistic increases by 2.706 compared to its best-fit value). It is possible that the automatic fit may have found a local minimum and we thus recommend you examine the plot and residuals and, if in any doubt about the automatic fit results, download the tar file and fit the spectra yourself.
Spectra are fitted in xspec
using the command
statistic cstat
, however since the background spectrum
is included, xspec
automatically selects
the w-statistic (see The
XSPEC statistics manual under For Poisson data with Poisson background (cstat)).