Scaled Map Transient Analysis Synopsis
Return to Main BAT Transient Monitor Page
For more information contact: Hans Krimm, NASA GSFC / USRA
Please Read Me
This web site provides an overview of Swift/Burst Alert Telescope (BAT) survey results for the convenience of users in monitoring sources, preparing proposals and other work. The complete set of official Swift data products is available via the Swift archive at the HEASARC at NASA's GSFC, as well as the ASI Science Data Center in Italy and the UK Swift Science Data Center. Those sites make available all Swift survey data products, including 80-channel histograms, spacecraft pointing information and calibration files.The FITS and ASCII-formatted data available here are provided primarily for convenience. While the Swift/BAT team hope our instructions and information are adequate for most users, we are not prepared formally to support users of these data. We do, however, appreciate feedback.
We ask that when results from these analyses are used, they be referenced as "Swift/BAT transient monitor results provided by the Swift/BAT team." This team includes all those working on the BAT at the Goddard Space Flight Center and Los Alamos National Laboratory.
These pages contain the Swift Burst Alert Telescope (BAT) hard X-ray transient monitor. Light curves are provided for 488 sources and are updated each time new data from the BAT becomes available. The transmission and processing delay averages six hours, but depends on the timing of Swift data downloads: for instance there are ~8 orbits each day in which the Swift satellite does not pass over its primary ground station at Malindi, Kenya.
Galactic and extragalactic sources are selected for inclusion in the catalog if they are either detected by BAT on a daily basis or have had a significant outburst during the lifetime of Swift. Other bright known galactic sources are also included in the catalog.
Light curves for most sources go back through 12 February, 2005. Sources that have been added to the catalog after December 2006 will have shorter light curves. Work is in place to re-analyze all data and provide light curves for all sources back to 12 February, 2005.
Each source in the table has a link to a page containing two light curves. The top plot (red) is a light curve of the daily average counts for the source. The baseline (zero counts) is shown as a solid line and the average flux is shown as a dotted line. The lower plot (blue) is a light curve of the average flux for each complete Swift pointing. To keep the plots from being too crowded, the time scale for the orbital (Swift pointing) plots is limited to the past twenty days. At the top of each page is a link from which one can download the FITS light curve file from which the plot was made or an ASCII version of the light curve. Points with extremely large errors (more than four times the average statistical error) and large (> 10 sigma) negative fluctuations are not plotted, although all points are included in the light curves. Points which are not included in the plots for these reasons are given a data quality flag of 1 (large negative fluctuation) or 2 (large error bar). Good points have a data quality flag of 0.
The points in the daily plots are the weighted average of all observations starting and ending on that calendar day (from 0 UT to 23:59:59 UT). Since Swift terminates and restarts all observations at midnight UT, data points do not span the day boundary. For the last point (the current day), the daily average is a running average which averages all observations up through the most recent data received and processed. Therefore this data point will change through the day as new data is accumulated. For example if a source has a short, strong outburst early in the day, the weighted average for the day will start out high, but will be reduced as data from times after the outburst are averaged in.
Basic analysis:
The BAT flight software produces "scaled maps" as part of its on-board transient monitor. Scaled maps are maps of the detector array in a single energy band (15-50 keV) and are compressed, or scaled, in such a way as to reduce file size, but not lose any information. Scaled maps are produced as part of the image confirmation for BAT GRB rate triggers and on times scales ranging from 64 seconds to a full observation for the on-board image trigger. Only maps on time scales >= 64 seconds are analyzed in this analysis.
Each scaled map is processed using the standard BAT software tools. Each map is converted to a sky image using BATFFTIMAGE and then compared to a catalog of known sources using BATCELLDETECT . (For complete information on these and other BAT analysis tools please see: Swift Data Analysis site This program also searches for unknown sources at high significance. A further step involves cleaning bright sources from the images using BATCLEAN , a process which allows more faint sources to be detected.
The analysis process also involves significant data verification including rejecting corrupted data and episodes with missing attitude data or invalid star camera solutions. There are also several corrections applied which are discussed below.
Explanation of data fields in output catalogs:
Daily results
| 1 | TIME | [d] | Modified Julian Date |
|---|---|---|---|
| 2 | RATE | [count/cm^2/s] | Rate for object |
| 3 | ERROR | [count/cm^2/s] | Total error (statistical plus systematic added in quadrature) |
| 4 | YEAR | [yr] | Year |
| 5 | DAY | [d] | Day of Year |
| 6 | STAT_ERR | [count/cm^2/s] | Statistical error |
| 7 | SYS_ERR | [count/cm^2/s] | Systematic error |
| 8 | DATA_FLAG | Data quality flag (0= good; 1= large negative fluctuation; 2= large error bar; 3= both) |
Single pointing results
| 1 | TIME | [s] | Swift Mission Elapsed Time (Seconds since 1 Jan 2001) |
|---|---|---|---|
| 2 | RATE | [count/cm^2/s] | Rate for object |
| 3 | ERROR | [count/cm^2/s] | Total error (statistical plus systematic added in quadrature) |
| 4 | YEAR | [yr] | Year |
| 5 | DAY | [d] | Day of Year |
| 1 | MJD | [d] | Modified Julian Date |
| 1 | TIMEDEL | [s] | Exposure of bin |
| 6 | STAT_ERR | [count/cm^2/s] | Statistical error |
| 7 | SYS_ERR | [count/cm^2/s] | Systematic error |
| 7 | PCODEFR | Partial coding fraction: Fraction of BAT detectors exposed to the source | |
| 8 | DATA_FLAG | Data quality flag (0= good; 1= large negative fluctuation; 2= large error bar; 3= both) | |
| 9 | DITHER_FLAG | Spacecraft dither (0= yes; >0= no) |
Corrections and data cuts:
Most systematic errors are corrected in the analysis. Here is a summary of corrections and data cuts applied to the data set:
Data cuts
Corrections
Dither flag
It is a known aspect of coded mask imaging that systematic errors arising from the presence of bright sources in the field of view are spatially correlated. This means that if two or more spacecraft pointings have the same orientation on the sky (to within a few arcminutes) then fluctuations due to systematics (either positive or negative) will tend to accumulate in a particular location in the BAT field of view and hence at a particular equatorial or galactic sky coordinate.
To mitigate this effect, starting on September 17, 2005, the Swift mission operations team instituted a procedure known as "roll dithering." This means that in successive pointings at the same target (same field center), the spacecraft roll will be changed to a value within plus or minus one degree of the original value. This is small enough so that it does not affect the narrow-field instruments (NFIs) and ensures that systematic errors do not accumulate in BAT images.
The roll dither procedure is done for most pointings. However, there are certain situations in which roll dithering is not done. First, since the dithering is commanded, there is no dithering for automatic targets (ATs): GRBs or on-board fast transient response. Similarly there is not dithering for targets of opportunity (ToOs). There are also other times when it is decided not to do the dithering, either because the precise orientation of source in the UVOT or XRT field is required, or for NFI calibration purposes. Finally, for part of 2005 and early 2006, the dithering commands were generated by hand, and sometimes this step was forgotten in calculating the daily observing schedule.
The consequences for BAT transients of the "no dither" times is that it is possible that small systematic errors in individual pointings will add up when the daily averages are made. For example a 2 standard deviation positive fluctuation in multiple single pointings at the same sky location would accumulate and lead to a 7 or higher sigma positive point in the daily averages. This means that the significance of some positive daily average points is much higher than it should be (or conversely the systematic error bars are underestimated). This is only a problem for individual daily averages, since the spacecraft roll is always changed each day. In other words, a positive source increase seen on two or more days cannot be attributed to this systematic.
Starting on 6 December, 2007, we have added a "dither_flag" field to the FITS and ASCII orbital light curves. A non-zero value of the flag indicates that roll dithering was not in place for this pointing. Although the adverse effects of no dithering are only seen in daily averages, the dither flag itself has meaning only for an individual pointing since most sources are seen in the BAT field of view for multiple different targets and hence very different spacecraft orientations.
Bottom line: When you observe a high positive (or negative) fluctuation for a source on a single day, please treat it with caution. Look at the dither flag in the orbital light curves for that day and if most of the pointings including the source are "no dither," then it is likely that the positive fluctuation is not physical.
Further caution: The dither_flag is currently only filled in for new data (after 6 December 2007). When the data is reprocessed starting in mid-December, the dither flag will be present for all light curves.
Error bars:
The error bars are a combination of statistical errors and systematic errors. Most systematic errors are corrected in the analysis (see above) and the remaining systematics are accounted for in two systematic error terms as described below.
Systematic error derived from blank sky positions
In order to understand residual systematics in the distributions of counts from catalog sources, the BAT transient monitor catalog includes 106 "blank" points in the sky, randomly distributed across the sky and chosen to be at least 10 arcminutes from any reported X-ray source. The light curves from these blank sky positions can be seen here .Since there are no sources in these locations the distribution of significances of counts from these locations should follow a Gaussian distribution with zero mean and width of unity. As seen here , there are no systematic biases toward either high or low significance, however, the width of the significance histogram is larger than one, which indicates that the statistical errors underestimate the true distribution of errors. The statistical errors must therefore be increased by a systematic factor which makes the width of the distribution unity. This correction was found to be 10% for the orbital data and 23% for daily averages. It is as expected that systematic errors increase as the monitor duration increases.
This correction is a multiplicative factor which increases all statistical error values in the transient monitor (included in the STAT_ERR column in the FITS and ASCII light curves.
Systematic error derived from the Crab light curves
This systematic error was derived from an empirical analysis of the Crab light curve in which it was found that there was more scatter in the data points than could be explained by statistical variations alone. Since the data presented here are not corrected using the BAT response matrix, these errors are expected to affect the measured flux by ~10%. When averaging the Crab data above 0.2 counts/cm2/sec, it was found that the residual scatter in the orbit light curve had a standard deviation of 4.6%, and in the daily light curve 3.5%. This value was applied to all light curves, but only makes an important contribution to bright sources.This correction is found in the FITS and ASCII light curves in the SYS_ERR column and is derived by multiplying the rate found in the transient monitor by the correction factors listed in the previous paragraph.
The full error reported in the ERROR columns and represented on the light curve plots is thus the addition in quadrature of STAT_ERR and SYS_ERR.
It is important to note that there are strong spatial correlations in the BAT observations of a given source which can place the Crab in the same location in the field of view for many days at a time. Swift is not a scanning survey instrument. Its observing program is driven by the random location of gamma-ray bursts on the sky, and gamma-ray burst afterglows are typically observed for many days. Thus in any given ~week long interval, the Crab is likely to be at the same location in the BAT field of view and the same systematics will apply. This is a likely cause for the coherent structure one can see in the light curve.
Future Work:
- I have attempted to list each source by its most common name as determined by the first name listed for each source in the Simbad database. This process is not perfect and there is no unanimity in the community as to the standard name for many sources. I plan to add cross-references to alternative names for catalog sources.
- Reanalysis of all the data back to 12 February 2005 will take place over the next several months. At that point all light curves will be current back to this date. This reanalysis will include all data corrections and filters which are currently in place.
Hans A. Krimm
If you have a question about Swift, please contact us via the Feedback form.