R. M. Kippen,
V. Connaughton,
G. N. Pendleton,
S. D. Barthelmy,
P. Woods, M. S. Briggs, G. J. Fishman,
C. Kouveliotou,
C. R. Robinson, and C. A. Meegan
--
University of Alabama in Huntsville, AL 35899
NAS/NRC NASA/Marshall Space Flight Center, Huntsville, AL 35812
USRA NASA/Goddard Space Flight Center, Greenbelt, MD 20771
NASA/Marshall Space Flight Center, Huntsville, AL 35812
USRA NASA/Marshall Space Flight Center, Huntsville, AL 35812
The recently developed BATSE Rapid Burst Response (RBR) system
provides gamma-ray burst locations accurate to
2
-3 for several bursts per month within
15-35 minutes of onset. Combined with wide-field or scanning
instruments, this system offers the chance of detecting prompt
x-ray/optical/radio afterglow emission from many bursts. We discuss
the operation, performance and status of the BATSE RBR system as well
as results from coordinated multi-wavelength follow-up observations.
The combination of rapid and accurate gamma-ray burst (GRB) locations
has long been recognized as the fundamental key to solving the
burst counterpart problem. Unfortunately, quick and precise
measurement of GRB locations is a difficult technical challenge. For
the past four years the BACODINE system has been attempting to solve
this problem by providing roughly-determined (accurate to
6-12; 1
) BATSE burst locations in
near real-time (typically < 10 seconds) to a small network of
dedicated wide-field optical/radio counterpart observers
[1]. The BACODINE system sacrifices location accuracy
in favor of speed through its use of small amounts of data, automated
background estimation and a simplified burst location algorithm.
Unfortunately, BACODINE locations have proved to be too imprecise to
allow the sensitive observations required to detect faint GRB
afterglows.
To facilitate more sensitive counterpart searches on somewhat longer time-scales, we have developed the BATSE Rapid Burst Response (RBR) system. BATSE RBR takes a different approach from BACODINE--better location accuracy at the cost of longer delay. This is accomplished using a larger sample of BATSE data, interactive background estimation, and a much more comprehensive burst location algorithm. The system operates in parallel with normal BACODINE processing and the locations are publicly distributed via the GRB Coordinates Network (GCN [1]). The BATSE RBR approach, favoring precise locations over real-time delay, has significant merit given that the recently discovered GRB afterglows are faint, but can last from several hours, to days [2, 3, 4, 5]. Below, we discuss the system's operation and performance.
The BATSE RBR system is a combination of the existing real-time BACODINE system and the burst processing software of the BATSE instrument team. By using aspects of both, we are able to provide BATSE GRB locations as accurate as possible, with minimal delay. The system operates as follows.
Upon recognition of a transient event trigger, up to 500 s of buffered, real-time BATSE telemetry are accumulated and then automatically downloaded by the BACODINE computer at Goddard Space Flight Center to the BATSE team headquarters at Marshall Space Flight Center. Concurrently, on-duty BATSE team scientists are alerted via e-mail and alphanumeric pagers (24 hours), and begin to manually reduce the data as soon as they arrive. BATSE Large Area Detector (LAD) count rate data, with 1024 ms temporal resolution, in four discriminator energy channels (DISCLA), are interactively used to estimate source and background intervals that optimize the signal-to-noise ratio. Interactive processing also allows the efficient rejection of non-GRB events, and events with large location errors. The background-subtracted count rates in each BATSE LAD are used as input to the LOCBURST [6] burst location algorithm. The LOCBURST algorithm has a considerable history of development over the course of the BATSE mission. It uses detailed models of the detector response and atmospheric scattering to determine the GRB location that best fits the observed count rates. Upon completion, the burst location and its statistical uncertainty are transferred to the BACODINE computer, from which they are distributed to the GCN network of multiwavelength observers. A special GCN burst alert notice type ``LOCBURST'' distinguishes the RBR locations from BACODINE locations. Under normal conditions, the entire process requires only 15-35 minutes from trigger to completion.
To reduce the burden on duty scientists, only the brightest bursts are
processed. The current threshold for RBR notification results in
2 LOCBURST alerts per week. Weaker bursts are not a significant
loss since they have larger location errors.
Figure 1: BATSE RBR location errors. (a) Cumulative distributions
of 73 BATSE/IPN offset angles compared with best-fit
RBR two-component systematic error model. (b) Total
error histograms (referenced to the left axis) for the same 73
bursts using the best-fit systematic error model. Smooth curves
(referenced to right axis) show the relation between total error
and statistical error in this model.
The BATSE RBR system uses the most current version of LOCBURST--the same algorithm that is employed for the newest (since June 1997) burst locations in the ``current'' BATSE GRB catalog [7]. This version of LOCBURST differs from that used for earlier catalogs in that it incorporates a more accurate energy calibration of the LAD discriminator. The BATSE RBR locations also differ from cataloged results because RBR cannot use the full BATSE dataset (most burst data are stored in memory and are thus not available in the real-time telemetry at the time of a burst). We have therefore independently assessed the BATSE RBR location errors.
As a comparison sample, we use 73 bursts (detected between Feb. 1995
and Sept. 1997) that have been independently located to single annuli
through BATSE/Ulysses Interplanetary Network (IPN) timing
analysis [8]. While only 7 of these bursts were processed
in ``real-time'' with the RBR system, they are typical, in brightness,
of the majority of RBR events. Since the IPN annuli are precisely
determined (
0.1 in width), the shortest angular
separation 
between a BATSE location and an IPN annulus yields
an estimate of the total BATSE location error (a combination of the
measured statistical error and an unknown systematic contribution).
Figure 1a shows the cumulative distribution of for
locations from RBR, the official BATSE catalog, and BACODINE. The RBR
locations are statistically indistinguishable from the official
catalog results, and a factor of 3.5 more accurate than
BACODINE.
The BATSE team has recently significantly improved the
characterization of systematic burst location errors from a single
Gaussian distribution (
), to the
weighted sum of two Gaussians, with differing widths [9].
A maximum-likelihood fit of this improved ``core-plus-tail'' model to
the 73 RBR/IPN bursts is shown in Figure 1a. The best fit
model, which is statistically consistent with fits to larger data
sets, has a 1.9 systematic error ``core'' component with 78%
probability and a 5.3 ``tail'' component with 22%
probability. Using these parameters, we can compute the total error
of a burst, given its measured statistical error radius. The
cumulative distribution of total errors for the 73 test bursts is
shown in Figure 1b, which indicates that 85% of the
RBR bursts have total location errors in the range
2-3 (68% confidence level). Also shown in this
figure is the relation between statistical and total error. For
example, RBR bursts with statistical errors 
have 68%
confidence total errors 
.
Figure 2: Equatorial coordinate (J2000) sky maps of the five BATSE RBR
events scanned by RXTE-PCA. Each map shows the RBR localization
(solid lines; 50%, 68%, and 90% confidence), the PCA
scan region (dashed lines), and the preliminary
IPN annulus (dotted lines [11]). For GRB 970807, COMPTEL
location contours (1,2,3 [12]) are shown and for
GRB 970616, the PCA transient x-ray source location is shown
(boxed X).
The BATSE RBR system has been operational since late April 1997--
providing burst location alerts at the rate of 2 per week.
Wide-field, ground-based telescopes have observed some of the
locations within several hours, but no optical afterglow counterparts
have been discovered (see these proceedings).
The most promising use of BATSE RBR locations lies in a coordinated
effort with the Rossi X-ray Timing Explorer's Proportional
Counter Array instrument (RXTE-PCA). In this effort, RXTE is
re-pointed and the PCA (field-of-view 1) is used to
scan the most precise RBR localizations for fading x-ray counterparts
(2-10 keV) within 3 hours [10]. This has been done
successfully for five GRBs. The BATSE RBR localizations of these
events (derived from the systematic error model discussed above) are
shown in Figure 2.
The nominal PCA GRB follow-up strategy is to scan a 
-6 region, centered on the RBR location, in a
40-60 minute observation. This is sufficient to detect 0.5 mCrab
sources with a significance >3. In one case (GRB 970807), a
smaller scan region was used due to poor observing efficiency and a
scheduling mistake. The nominal PCA scan area covers most of the
typical RBR 68% confidence region, but this region will not always
contain the GRB source (e.g., GRB 970925, which has an unusually large
location error). In the case of GRB 970616, a previously unknown
transient x-ray source was detected by the PCA in the RBR location
scan. The 0.5 mCrab source was localized by the PCA to
18
and the location was later found to be consistent
with IPN triangulation. A large-scale search for afterglow emission
at other wavelengths was undertaken--resulting in several potentially
interesting x-ray/optical sources. It is still uncertain which, if
any, is the counterpart to GRB 970616.
The BATSE/RXTE effort will continue to search for x-ray afterglows
from the brightest (best localized) bursts at the rate of 1 per
month. In addition, several new optical instruments show the promise
of being able to scan RBR error boxes with improved sensitivity within
a few hours.
The BATSE Rapid Burst Response System
This document was generated using the LaTeX2HTML translator Version 96.1-h (September 30, 1996) Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.
The command line arguments were:
latex2html -no_navigation -split 0 BRBR.tex.
The translation was initiated by R. Marc Kippen on Wed Jan 21 18:01:33 CST 1998