RESPONSABILIDADES GENERALES DE LOS MIEMBROS DE LA COMUNIDAD EDUCATIVA JAVERIANA
CONSENTIMIENTO INFORMADO
Low energy spectral residuals in Windowed Timing Mode
If the source is very absorbed in the 0.3 − 2 keV range, spectrum may show an artificial bump in the 0.4 − 1 keV range and a turn-up at very low energy range. The bump may disappear if only grade 0 photons are used instead of typical grades 0-2. This option is implemented in scripts. The turn-up doesn’t disappear and the energy where it starts changed over years. It is important to keep these effects in mind because the artifacts resemble thermal spectral component.
Position-dependent WT RMFs
The reason why WT image is compressed in one dimension is because 10 rows are red at a time. The merged ”big” row where the source lies means that the source could be at any of the 10 rows which were merged. Depending where it is can produce different WT images (due to multiple pixel events being split between merged rows) and then spectra. So, position dependent RMFs may be used. There are 3 of them for each ordinary RMF. It cannot be precisely determined where the source actually lies and which one to use. So, one strategy is to use all of them and select the one where the fit statistics is best (lowest χ2). Second strategy is to do fitting with all three and find an average value for 3 different sets of parameters, and combine all the confidence intervals of 3 sets of parameters (errors). In other words, treat it as an additional systematic error.
Burst position on the XRT detector
The position of the burst read in the header of the Level 2 data file doesn’t always correspond to the brightest pixel in the WT image. Additionally, the brightest pixel may change during course of several seconds or tens of seconds. This is due to uncertainty of XRT pointing, so the same sky position (RA and Dec) may ”drift” on the XRT detector plane as the satellite sways. The sky position should be chosen where the brightest pixel is.
Binning BAT data
By default BAT data are binned into 80 bins in the batbinevt, and later, bins correspond- ing to range below and above 15 − 150 keV are marked as bad in grppha. If the binning is custom and corresponds to 15 − 150 keV, then during fitting values corresponding to first 2 or 3 BAT bins will have unusual lower value. It is not clear what causes this and unlike previous problems it is not officially recognized. A way to overcome this is to have additional 3 bins from 10 − 15, 16 keV and then set them as bad in the grppha.
Additionally, default 80 bins are linearly equally spaced. It is more useful to have them logarithmically equally spaced and the number of such bins can be entered in the script (not counting first 3 bad bins).
Intrinsic column density
Intrinsic column density is an unknown factor. The value may be taken from Swift automatic analysis. Two values are from WT data and later PC data. Value from WT data is more precise but it may be wrong due to spectrum in XRT range being different than power-law (which was used to obtain column density) and varying. The PC data value is less precise but the spectrum in this later time is much more constant and powerlaw-like. Intrinsic column density may be left as a free parameter during fitting. Since it has to have the same value for different time bins, all the spectra may be fit simultaneously in XSpec with column density being the same parameter for all time bins. Value of intrinsic density is important if the thermal emission is in the lower band of the XRT range. Otherwise value(s) from automatic analysis may be taken and kept constant.
XRT-BAT normalization constant
These two instruments may have some unaccounted instrumental error so additional parameter during fitting may be a normalization constant. Since it has to be the same for all time bins (and at least very similar for other bursts), all the spectra may be fitted in XSpec with normalization parameter being the same for all time bins. Both instruments are on the same satellite, pointing in the same direction. In literature it was often found that normalization constant was close to one, and was kept as such during fitting. Ignoring it should not produce relevant effects.
Number of energy bins for XRT and BAT
XRT data should be binned minimum to 20 counts per bin. Depending on the intensity of the burst and excluded central pixels this may give different number of bins. BAT data should be binned in such a way to have at least half number of bins with errors smaller than the value. If XRT or BAT has many more bins than the other, then that one will affect
χ2 statistics much more and other instrument wont play much role. So, number of bins
should be similar. This goes for the underlying component which can be power-law or cutoff-powerlaw, or Band function. If the thermal component is significantly in the range of one instrument, XRT for example, than XRT should have good resolution in order to follow the bump of the thermal component.
Another thing is as binning change within some excepted intervals, the values ob- tained from the fit will change. In some cases this may produce significant differences in parameter values, errors, and comparisons between models.
Final remark
All the above mentioned should be kept in mind when reading results of from the fit. Small errors of the parameter may not mean it is likely close to that value. There may be underlying uncertainties which are beyond fitting in XSpec. Even if all the above un- certainties didn’t exist, it is hard to claim detection of the thermal emission. Black body spectrum is distinguished by a steep index of +1 before the peak which then falls of ex- ponentially. Even if thermal peak isn’t deformed by mildly relativistic motion, or lack of complete thermalization, to detect it would require a black body peak to be significantly above underlying component. In vast majority of cases this is not true and at best adding a black body component is done to detect curvature of underlying component or a bump. Then, assuming it is a black body, the evolution of its temperature, luminosity and cal- culated radius may be followed. From there it may be discuses weather the results have physical sense. For example, radius should start from smaller value and only increase with time (also relativistic correction should be taken into account).