THE INTERPLANETARY NETWORK SUPPLEMENT TO THE FERMI GBM CATALOG OF COSMIC GAMMA-RAY BURSTS

This paper presents the latest in a series of catalogs of gamma-ray burst (GRB) localizations obtained by arrival time analysis, or "triangulation" between the spacecraft in the third Interplanetary Network (IPN; Table 1). In the present paper, we present the localization data on 149 bursts which occurred during the period covered by the first, two-year Fermi Gamma-Ray Burst Monitor (GBM) GRB catalog (Paciesas et al. 2012; 2008 July 12 to 2010 July 11). As the composition of the IPN has changed over the years, we present a summary of the instrumentation and techniques in the following section. Section 3 contains the localization data, which are also available on the IPN Web site.²⁶ In Section 4, we discuss the statistics of the localizations.

The triangulation technique is illustrated in Figure 1. When a GRB arrives at two spacecraft with a delay δT, it may be localized to an annulus whose half-angle θ with respect to the vector joining the two spacecraft is given by

$$\begin{equation} \cos \theta =\frac{c \delta T}{D} \end{equation} \tag{ 1 }$$

where c is the speed of light and D is the distance between the two spacecraft. (This assumes that the burst is a plane wave, i.e., that its distance is much greater than D.) The annulus width dθ, and thus one dimension of the resulting error box, is

$$\begin{equation} d \theta =c \sigma (\delta T)/D \sin \theta \end{equation} \tag{ 2 }$$

where σ(δT) is the uncertainty in the time delay. The radius of each annulus and the right ascension and declination of its center are calculated in a heliocentric (i.e., aberration-corrected) frame.

Zoom In Zoom Out Reset image size

The composition of the missions and experiments comprising the interplanetary network changes as old missions are terminated and new missions are introduced. During the period covered in the present catalog, the IPN consisted of Konus-Wind, at distances up to around 5 lt-s from Earth (Aptekar et al. 1995); Mars Odyssey, in orbit around Mars at up to 1250 lt-s from Earth (Hurley et al. 2006); the International Gamma-Ray Laboratory (INTEGRAL), in an eccentric Earth orbit at up to 0.5 lt-s from Earth (Rau et al. 2005); the Mercury Surface, Space Environment, Geochemistry, and Ranging mission (MESSENGER), launched in 2004 August, and in an eccentric orbit around Mercury beginning 2011 March 18, up to 690 lt-s from Earth (Gold et al. 2001); and the Ramaty High Energy Solar Spectroscopic Imager (RHESSI; Smith et al. 2002), Swift (Goldstein et al. 2012), Fermi (Meegan et al. 2009), Suzaku (Takahashi et al. 2007; Yamaoka et al. 2009), and AGILE (Marisaldi et al. 2008; Del Monte et al. 2008; Tavani et al. 2009), all in low Earth orbit.

The detectors in the IPN vary widely in shape, composition, time resolution, and energy range. Also, onboard timekeeping techniques and accuracies are not the same from mission to mission, and spacecraft ephemeris data are given only as predicts for some missions. Since the accuracy of the triangulation technique depends on all these parameters, end-to-end calibrations and sensitivity checks are a constant necessity. For the current IPN, we utilize the following method. For every burst for which the Swift X-Ray Telescope (XRT) detects an X-ray afterglow, we search for GRB detections in all the IPN experiments. If the burst was detected by (1) Odyssey and Konus or by Odyssey and a near-Earth mission, (2) MESSENGER and Konus or by MESSENGER and a near-Earth mission, or (3) Konus and a near-Earth mission, we derive an IPN annulus by triangulation. We then calculate the angle between the annulus center line and the XRT position θ_X, taken from the GCN Circulars, and which we take to be a point source, because its positional uncertainty is much less than the annulus width dθ (Figure 2). dθ is calculated such that the distribution of annulus widths is approximately Gaussian, so the distribution of θ_X/dθ should follow a normal distribution with mean zero and standard deviation 1, if systematic uncertainties are neglected. We have used this procedure so far for 78 MESSENGER/Konus or MESSENGER/near-Earth triangulations, 292 Konus/near-Earth triangulations, and 72 Odyssey/Konus or Odyssey/near-Earth triangulations. We find that for the interplanetary spacecraft a systematic uncertainty equal to roughly 0.75 times the statistical one is required to make the distributions consistent with normal distributions. An example is shown in Figure 3. Systematic uncertainties arise from numerous sources. Some are certainly negligible in some cases, while others may be important. But in almost all cases, it is impossible to assign an accurate number to them. A partial list follows, in no particular order.

• 1.  
  Variations in the clock accuracy from one spacecraft to another. Different spacecraft have different ways of calibrating their clocks and assigning times to the time bins of GRB time histories. We know, for example, that in some cases the GRB timing is subject to uncertainties, even though the spacecraft oscillator is quite accurate.
• 2.  
  Predict timing. In many cases, the time assigned to a GRB is a predicted time, and it is never updated. In other cases, such as Odyssey and MESSENGER, the time is eventually updated using an accurate model for the clock drift; in this study, the updated times have been used for these spacecraft. In other cases, no final clock corrections are applied.
• 3.  
  Different time resolutions. For any given spacecraft pair, the time resolutions can be vastly different, and sometimes one is not an exact multiple of the other. One time history is adjusted to match the time resolution of the other spacecraft in the light curve comparisons. This can be done in a variety of ways, but each is subject to uncertainties. Even in cases where one time resolution is in principle an exact multiple of another, the true values of the bin widths can be slightly different from their nominal values due to different on-board electronics.
• 4.  
  Spacecraft ephemerides. Some ephemerides are predictions, while others are final. In these comparisons, the final ephemerides were used where possible, but they were not always available.
• 5.  
  Different energy responses of the various detectors. In most cases, the GRB light curves are recorded in different energy ranges from one another. Even in those cases where we attempt to match the energy ranges of the detectors (i.e., where the photons are energy-tagged), the detector responses within those ranges are different due to the very different detector designs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is often possible to derive very precise triangulation annuli for bursts detected by Konus and the GBM, even though the distance between the spacecraft is not large. The reasons are first, that the first 1.024 s of triggered Konus data are transmitted with 2 ms time resolution; this is the finest resolution of all the IPN detectors which bin their data. Second, GBM time- and energy-tagged data can be utilized to match Konus' time bins and energy range, minimizing two possible sources of systematic uncertainties. Thus for short-duration or intense GRBs, or bursts with fine time structure, Konus/GBM annulus widths as small as several arcminutes can be obtained (Pal'shin et al. 2013). To verify Konus-GBM triangulations we have derived Konus-GBM triangulation annuli for 52 precisely localized bursts. The 3σ half-widths of these annuli range from 011 to 218 with a mean of 30 and a geometrical mean of 119. Figure 4 shows the distribution of the offsets (in sigma) of the precise positions. The mean offset is 0.09 and the standard deviation is 0.77. The minimum offset is −1.40 and the maximum is 1.69.

Zoom In Zoom Out Reset image size

To calibrate the IPN sensitivity, we use estimates of the peak fluxes, fluences, durations, and E_peak of a large number of GBM bursts.²⁷ These are measured in the 50–300 keV range, and are given in photons cm⁻² s⁻¹ (measured over a 1024 ms period), erg cm⁻², s, and keV, respectively. At the time this catalog was submitted in its final version, there were 1078 GBM bursts with peak flux, fluence, and duration entries, and 482 with E_peak entries. To calculate the IPN sensitivity, we determined (1) whether any other IPN spacecraft also detected the burst, and (2) whether Konus, MESSENGER, or Odyssey detected the burst. Only the latter detections can lead to meaningful triangulations, because of their larger distances from Earth. From these numbers, we calculate the detection probabilities as functions of flux, fluence, duration, and E_peak. The results are shown in Figures 5–8. In each of the four graphs, the detection probabilities (or IPN efficiencies) represent an integral over the other three variables, as well as over duty cycles, and, for all the instruments except Konus, planet blocking. The probabilities of IPN detections are 50% or greater for peak fluxes in the range 1–3 photons cm⁻² s⁻¹ and for fluences in the range 1–3 × 10⁻⁶ erg cm⁻².

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Every cosmic burst detected by the GBM was searched for in the IPN data; GBM localizations were used to calculate arrival time windows for Odyssey and MESSENGER, but the total crossing time windows defined by light-travel times were examined in all cases. The resulting detections are given in Table 2. (Note that this table supersedes the information in Table 2 of Paciesas et al. (2012), which is incomplete.) Konus and Suzaku can detect bursts in both triggered (2–64 ms time resolution) and an untriggered (1–3 s time resolution) modes; both modes are counted as detections in this table. Also, detections by several instruments which are not part of the IPN have been reported in the table, namely the Fermi LAT, Monitor of All-sky X-ray Image (MAXI), and Rossi X-Ray Timing Explorer (RXTE).

Two events in Table 2 were detected by the GBM, but did not appear in the GBM catalog. The origin of GRB 091013 was classified as "uncertain." However, the cosmic nature of this event is confirmed by Konus. GRB 100501 was detected by numerous IPN spacecraft, including the GBM. However, in that case, the actual GBM trigger was caused by a terrestrial gamma flash.

Whenever Konus (in triggered, high time resolution mode), Odyssey, or MESSENGER detected the burst, we calculated one or more triangulation annuli. The annuli are given in Table 3, and figures may be found in Figure 9. In general, the annuli obtained by triangulations are small circles on the celestial sphere, so their curvature, even across a relatively small GBM error circle, may not be negligible, so that a simple, four-corner error box cannot always be defined accurately. For this reason, we do not cite the intersection points of the annuli with the error circles. A prescription for deriving these points, however, may be found in Hurley et al. (1999a).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (149 panels)

Tarball of figure set images

When three widely separated experiments observe a burst, the result is two annuli which generally intersect to define two small error boxes. The proximity to the GBM error circle may be used to distinguish the correct one. When Konus, Odyssey, MESSENGER, and a near-Earth spacecraft (including INTEGRAL) detect a burst, the position is over-determined. In these cases, a goodness-of-fit can be derived for the localization, and an error ellipse can be generated (Hurley et al. 2000a). Although we utilize this procedure whenever possible, we do not quote the localizations as error ellipses in this catalog, because, like the annuli, their curvature can render a simple parameterization inaccurate. A number of degenerate cases can occur in a three-spacecraft triangulation; they are discussed in Hurley et al. (2011b).

When Konus and an interplanetary or near-Earth spacecraft observe a burst, it is often possible to define a long, narrow error box from Konus' determination of the burst's ecliptic latitude. This is derived from a comparison of the count rates on the two Konus detectors, and its accuracy is generally of the order of ±10°. A study of over 1800 Konus events indicates that the ecliptic latitude limits determined in this way can be considered to be an ∼95% confidence band. Systematic uncertainties usually prevent a more accurate determination.

IPN annulus widths are often comparable to, or smaller than, Fermi LAT error circle radii, and can therefore reduce the areas of LAT localizations. An example is GRB 090323 (Ohno et al. 2009; Hurley et al. 2009), for which a Swift ToO observation led to the discovery of an XRT (Kennea et al. 2009), optical (Updike et al. 2009), and radio (Harrison et al. 2009) counterpart.

The 21 columns in Table 3 give (1) the date of the burst, in yymmdd format; this contains a link to a figure on the IPN Web site showing the annulus or error box and the GBM error circle, (2) the Universal Time of the burst at Earth, (3) the GBM right ascension of the center of the error circle (J2000), in degrees, (4) the GBM declination of the center of the error circle (J2000), in degrees, (5) the 1σ statistical GBM error circle radius, in degrees, (6) the right ascension of the center of the first IPN annulus, epoch J2000, in the heliocentric frame, in degrees (7) the declination of the center of the first IPN annulus, epoch J2000, in the heliocentric frame, in degrees, (8) the angular radius of the first IPN annulus, in the heliocentric frame, in degrees, (9) the half-width of the first IPN annulus, in degrees; the 3σ confidence annulus is given by R_IPN1 ± δ R_IPN1, (10) the right ascension of the center of the second IPN annulus, epoch J2000, in the heliocentric frame, in degrees, (11) the declination of the center of the second IPN annulus, epoch J2000, in the heliocentric frame, in degrees, (12) the angular radius of the second IPN annulus, in the heliocentric frame, in degrees, (13) the half-width of the second IPN annulus, in degrees; the 3σ confidence annulus is given by R_IPN2 ± δR_IPN2, (14) and (15) the Konus ecliptic latitude band, in degrees, (16)–(18) the right ascension, declination, and angular radius of the Earth or Mars, if the planet blocks part of the localization, in degrees, and (19)–(21) any other localization information, in right ascension, declination, and angular radius, in degrees.

The GBM data have been taken from the HEASARC online catalog,²⁸ if the localization source was "Fermi, GBM." For bursts with other localization sources, the "human-in-the-loop" localization was used (V. Connaughton 2012, private communication). GBM localizations are subject to change, and are given here for convenience only. The latest online catalog should be considered to be the most authoritative source of the up-to-date GBM data. The data in Table 3 are also available electronically.²⁹

There are 491 bursts in the GBM catalog (Paciesas et al. 2012). Of these, 427 (87%) were observed by at least one other IPN spacecraft. They are listed in Table 2, and the number of bursts observed by each IPN spacecraft is compiled in Table 4. Those events which were not observed by an IPN spacecraft had fluences between 4.5 × 10⁻⁸ and 9.5 × 10⁻⁶ erg cm⁻², peak fluxes between 0.33  and  8.8 photons cm⁻² s⁻¹, and durations between 0.13 and 218 s, as measured by the GBM (Goldstein et al. 2012; Paciesas et al. 2012). For 149 of them, it was possible to improve the localizations by triangulation. The minimum and maximum 3σ IPN annulus half-widths were 7.40 × 10⁻³ and 319, and the average was 18. The IPN error boxes have 3σ areas between about 1 arcmin² and 110 deg². Each IPN localization was compared to its corresponding GBM error circle, as given in the online catalog.³⁰ In that catalog, the GBM localizations have been approximated as circles, with 1σ (statistical only) radii. Assuming that they are described by a two-dimensional normal distribution, we would expect 87% of the 3σ IPN localizations to agree with them (i.e., to have some intersection with them). We find only 52% agreement.

If a GBM systematic uncertainty of 6° is assumed, and added in quadrature to the statistical uncertainty, we find the expected 87% agreement. If that radius is then multiplied by three, the agreement becomes 98% (3 events with discrepant localizations), so that this can be taken as an approximation to a 3σ GBM confidence region for this particular GRB sample. A more detailed analysis of systematics is given in V. Connaughton et al. (2013, in preparation). Comparing each IPN area with its corresponding 3σ GBM area, as approximated above, we find an average reduction in area of a factor of 180.

The Fermi GBM has proven to be a worthy successor to BATSE. It detects about 245 GRBs yr⁻¹ and distributes their coordinates almost instantaneously to a wide astronomical community. The nine-spacecraft IPN is a good complement to it, just as it was to BATSE. It detects a total of about 325 bursts yr⁻¹ (18 yr⁻¹ are short-duration, hard spectrum GRBs; see Pal'shin et al. 2013), has virtually no planet blocking or duty cycle restrictions when all the spacecraft are considered, and it is capable of good localization accuracy at the cost of longer delays. There are many ground-based experiments, both electromagnetic and non-electromagnetic, which can take advantage of the smaller IPN error boxes, and for which delays are not an issue. In this sense, the GBM and the IPN both expand the reach of Swift, by localizing bursts which Swift cannot. For example, a search for gravitational radiation is in progress which utilizes the IPN data on over 500 GRBs, the most extensive such search to date; another search has begun for neutrinos, using IceCube data and almost 1000 IPN events.

This catalog represents the first installment of the IPN supplements to the GBM burst catalogs. Work is proceeding on the localization of IPN bursts observed during the third and fourth years of GBM operation. Data on some of these events may be found at the IPN Web site.³¹

Support for the IPN was provided by NASA grants NNX09AU03G (Fermi), NNX08AC90G and NNX08AX95G (INTEGRAL), NNX08AN23G and NNX09AO97G (Swift), NNX08AZ85G and NNX09AV61G (Suzaku), NNX07AR71G (MESSENGER), and JPL Contracts 1282043 and Y503559 (Odyssey). The Konus-Wind experiment is supported by a Russian Space Agency contract and RFBR grant 12-02-00032-a. This research has made use of data and/or software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC), which is a service of the Astrophysics Science Division at NASA/GSFC and the High Energy Astrophysics Division of the Smithsonian Astrophysical Observatory.