GAMMA RAY BURSTS LOCALIZED TO WITHIN 0.25 SQUARE DEGREE. A. CORRELATION WITH EXTRA GALACTIC OBJECTS

W.R. Webber, T.E. Harrison, B.J. McNamara and A. Lopez

Department of Astronomy, New Mexico State University Las Cruces, NM 88003

Abstract

Using two interplanetary burst network catalogues plus additional COMPTEL localizations we have selected þ60 gamma ray bursts that can be localized to ó0.25 square degree. We have correlated þ 30 of these burst locations with the individual locations of 5 classes of catalogued extra-galactic objects in regions of the sky with þ b þ > 25þ where galactic obscuration effects are minimal. These objects include a total of more than 30,000 individual galaxies, QSO's, rich clusters of galaxies, etc. These objects include a significant fraction of the visible matter in the universe as contained in the clusters, as well as a large sample of the most energetic objects in the universe. We find that any correlation between these catalogued objects and the gamma ray bursts appears to be within random chance. This places significant constraints on models where the origin of these bursts is assumed to be extragalactic.

1. Introduction

One difficulty in trying to identify the objects responsible for gamma-ray bursts (GRB) is the relatively large size of the gamma-ray error boxes. For example, the best localization of bursts from the BATSE experiment on CGRO have error boxes þ 2 - 3þ and the average error box is probably a factor þ 2 greater than this (Fishman et al., 1994). This makes optical identification of possible burst sources after the fact nearly impossible and makes the optical correlation of possible groups of objects with these bursts to be very insensitive. For example, Howard et al., 1993, compared various types of catalogued objects, both galactic and extragalactic with the BASTE gamma ray burst error boxes. They conclude that any correlation between the catalogued objects and the bursts was random, given the large uncertainties in the BATSE positions.

A number of well localized GRB error boxes have been deeply imaged to search for interesting stellar and extragalactic objects (see Schaefer 1992, and references therein). Few of these boxes appear to contain unusual extragalactic objects and the error box for GRB790613 (Ricker et al., Harrison et al., 1994) is empty to very faint limits. Thus, for a number of error boxes, the limits on the existence of extragalactic objects within their confines already imposes constraints on the minimum luminosity of the GRB sources, as well as that of its host galaxy (Schaefer 1992, Fenimore et al., 1993). While these searches suggest that normal type galaxies and AGN do not appear to be the hosts for GRBs, only a small number of boxes have been searched to conclusively rule out such scenarios. A catalogue search of known extra-galactic objects versus the smallest GRB error boxes offers a simple way to place useful limits on the presence of objects within the confines of a large number of error boxes.

It is possible to locate þ 60 bursts with a much higher precision than the BASTE error boxes. Using the catalogue of gamma ray bursts measured by the 1st Interplanetary Network and published by Atteia et al., 1984, and an additional catalogue of more recent bursts observed by the 3rd Interplanetary Network (Hurley et al., 1992) as well as several well localized bursts measured by COMPTEL (Hanlon et al., 1993) it is possible to locate þ 60 bursts within error boxes þ 0.25 square degree or less. More than half of these burst locations are known in fact much more accurately - to þ 10 arc min squared. With this kind of accuracy, it is now possible to make meaningful correlation studies between these burst locations and various classes of optical and radio objects - both galactic and extra-galactic. The goal of this study is to search for a correlation between the relatively well localized bursts and various classes of extra galactic objects by calculating the radial distance between the positions of the GRB's and those of the catalogued objects. As an additional check, objects which fell within 0.5ø of a burst position were more closely examined to determine if they were within the error contours of that particular burst. We identify one Abell cluster and two QSO's which are within the current localizations of the GRB error boxes.

We will show through simulations and other calculations that these correlations between extragalactic objects and GRB's are entirely within that expected for random chance.

2. The GRB and Extragalactic Object Data Base

The gamma-ray bursts used in this study are taken from three catalogues. In each case the burst location is required to be known within an error box of 0.25 sq deg. Over half the bursts come from the First Interplanetary burst catalogue (Atteia et al., 1987). Another 16 bursts come from Hurleys catalogue of more recent bursts detected by the 3rd Interplanetary network (IPN) (Hurley et al., 1994). And finally 7 bursts come from the COMPTEL catalogues of bursts (Hanlon et al., 1993). The COMPTEL localizations alone would not satisfy our areal requirement but the inclusion of the IPN annuli for these bursts greatly restricts the final error box, thus meeting our criterion.

The 60 well localized bursts used here are distributed isotopically on the sky thus simulating the known isotropicity of larger catalogues of bursts from BATSE. Since these bursts have been detected by the interplanetary networks and by COMPTEL, which are not as sensitive as BATSE, they represent on average stronger bursts than those seen by BASTE.

The all-sky catalogues we have used in our comparison are the machine-readable forms obtained from the NSSDCA. Any catalogue of extragalactic objects suffers from incompleteness due to the effects of galactic extinction, and due to the observationally defined flux limit of the survey. The determination of the completeness of the surveys used to compile a particular catalogue is difficult. The compilations of QSO's, for example, are from literature searches. Since many diverse sources are used to construct such a catalogue, the areal average of the catalogues is uneven. We therefore emphasize that several of the catalogues used in our comparison suffer from this sort of incompleteness, hampering conclusions reached about the results from burst-source matching.

To address the limitations imposed by galactic extinction we have employed the Third Reference Catalogue of Bright Galaxies (RC3, de Vaucoluers et al., 1991). The RC3 claims to be complete for all galaxies brighter than mB=15.5, with diameters ò 1', and with recession velocities ó 15,000 km s-1. Galaxies having these properties (÷11,000) were sorted into equal area galactic latitude bins to determine the effect of galactic extinction on the number of galaxies detected. We find that this "complete" subset of the RC3 catalogue appears to be unaffected by galactic extinction down to a latitude of þb þ=25ø. This value defines the cutoff-latitude for all of the extragalactic surveys, since they should be likewise affected.

The following additional catalogues were used in our study. "Third Reference Catalogue of Bright Galaxies" (de Vaucouleurs et al 1991, the "RC3"), "A catalog of Rich Clusters of Galaxies" (Abell et al., 1989), "An Optical Catalogue of Radio Sources" (Burbidge & Crowne 1979), "A Catalogue of Quasars and Active Galactic Nuclei" (Veron-Cetty & Veron 1989), and "A Revised and Updated Catalog of Quasi-Stellar Objects" (Hewitt & Burbidge 1993). Two additional catalogues compiled from observations made from the northern hemisphere were also searched: "The Catalog of Galaxies with Ultraviolet Continuum" (Markarian et al., 1967-81), and the "Catalogue of Galaxies and of Clusters of Galaxies" (Zwicky et al., 1961-68, hereafter referred to as the "Zwicky catalogue").

We now discuss the matching of our sample of well-defined GRB locations with catalogued objects. To do this we have written a computer program which finds the catalogued object which is closest to the center of the error box for each GRB. We then counted the number of these closest matches that fell within bins of width 0.5ø in radial separation distance from the GRB. If a particular type of catalogued source was responsible for, or the host of, GRBs, and the catalogue was complete, then most of the bursts should have a match with a catalogued source to a radial error corresponding to the characteristic size of the GRB error box. Since all of the GRB error boxes in our study have areas of <0.25 sq. deg, nearly all of the resulting matches would end up populating the first bin. Because the catalogues are incomplete, however, there will be some GRBs that occasionally would not have a match inside the first bin in this ideal scenario. This effect spreads-out the resulting distribution to the other more radially distant bins. If the catalogue is grossly incomplete, or the sources in the catalogues were not the host sites of GRBs, very few matches would be present in the first bin, and the burst distribution would approach that expected from a random distribution of sources of the same size as the catalogue being investigated. We have used a random distribution of sources of the same size as the appropriate catalogue to simulate such an effect. For each case discussed below, 100 simulations were run, each producing a catalogue of random object positions that were then compared with the burst catalogue. The results from the 100 individual simulations were averaged to determine the likely statistical match to the results from the real burst-catalogue positional matching. Such a technique quickly allows a check to determine if the final matching results for a particular catalogue are consistent with a random distribution.

3. Comparison of GRB locations and Catalogues of Extragalactic Objects

3.1 Comparison with the Third Reference Catalogue of Bright Galaxies.

The RC3 contains some 23,000 galaxies, of this, there is a "complete" portion (see above) which contains approximately 11,000 members. Removing all of those objects with þb þó25ø leaves ÷ 9200 galaxies. The results of our positional matching between the bursts and the RC3 galaxies are presented in Figure 1a. Five RC3 galaxies were located within 0.5ø of a burst position. The average result for 100 random simulations is presented as a dashed histogram in Fig. 1a. (the last bin in this figure, and in those that follow, contains all matches that have radial separations ò 4.5ø). The random simulations predict that we should have found 8 ñ 2.4 (1å) matches inside the first 0.5ø bin. Thus our results for the RC3 are consistent with a random distribution of sources. This does not mean, however, that the 5 galaxies with R ó 0.5ø might not be located with the GRB error boxes, allowing them to be possible hosts. To check this we plotted the positions of the galaxies with respect to the actual error boxes. None of the five galaxies were located inside the interplanetary network error boxes for their respective GRBs. We conclude that normal galaxies, brighter than mB = 15.5, are not the hosts of GRBs.

3.2 Comparison with the Catalog of Rich Clusters of Galaxies

This all sky catalogue of rich galaxy clusters contains 4073 members. There are 3853 clusters with þb þ > 25ø. The results of the burst-cluster matching are presented in Fig. 1b. We found five matches within the bin for radial distance ó 0.5ø. A run of 100 random simulations with 3853 objects finds that 4 ñ 1.8 matches should result. Thus, the results from the catalogue-burst matching are consistent within the 1å error bar of the random distribution simulations. Because galaxy clusters are extended objects, it is more difficult to ascertain whether any of the 5 matches we found are correlated with the actual clusters. To determine the boundaries of the galaxy clusters we have used the boundaries defined in the Zwicky catalogue of clusters. This boundary represents the locus where the density of the cluster galaxies falls to twice that of the local field density of galaxies. Fortunately, all five clusters where matches occurred were contained in the Zwicky catalogue. When the burst position is plotted with respect to the cluster boundary, we find that two of the bursts fall well outside the boundaries of the cluster, two bursts fall on the cluster boundary, and one burst is consistent with the position of a cluster (Abell 2503 with GRB910814). We conclude that this match, as well as the two marginal cases, can be explained by random chance, and the rich clusters contained within this catalog are not the likely hosts of GRBs.

3.3 Comparison to the Optical Catalogue of Radio Sources

This catalogue contains all known radio galaxies (identified up to 1979) that have Lradio > 1041 erg - s-1. Of the 495 objects in this catalogue, 380 have þb þ > 25ø. One radio galaxy-GRB match was found within the 0.5ø bin. This result is consistent with the random simulation test (Fig. 1c). We have plotted the position of the radio galaxy with respect to the position of the GRB and find that it is located well outside the error box. We conclude that strong radio galaxies contained in this catalogue are unlikely GRB hosts.

3.4 Comparison to the Catalogue of Quasars and Active Galactic Nuclei

This catalogue is divided into three separate catalogs: 1) QSOs, 2) AGNs (e.g. Seyfert galaxies), and 3) BL Lac objects. Comparisons were run for all three types of objects. This comparison included 4234 QSOs, of which 4004 met our galactic latitude criterion. We found two QSOs within a 0.5ø radius of a burst localization (Fig. 1d). Both of these objects were found to be located inside their respective GRB error boxes. The first, QSO 0116-288, was discovered in a deep optical search of the GRB791119 error box (Pederson 1983). The second, the match of QSO 2219-420 with GRB790419, does not appear to have been previously recognized. The simulations for the QSO catalogue would predict 3.73 ñ 1.78 matches within the first bin. Thus, the occurrence of two QSOs within the first bin appears to be random. We will discuss this conclusion more thoroughly below. There are 1108 AGN of which 997 are at þb þ > 25ø. No AGN had a burst match within the first bin (Fig. 1e), and the remaining distribution is completely consistent with that expected by chance. Of the 84 BL Lac objects with þb þ > 25ø, no matches were found (Fig. 1f). Again, our results are consistent with the expectation of a random distribution of 84 sources.

3.5 Comparison to the Revised and Updated Catalog of Quasi-Stellar Objects

There are 7315 objects in this catalogue, including 90 BL Lac objects. Imposition of our latitude criterion reduces this to 6220 objects. We found the same two QSOs within the first bin (Fig. 1g) as were found in the previous QSO catalogue. The simulations for this particular catalogue would predict that 5.3 ñ 2.2 objects should have matches within the first bin. Again, the occurrence of two QSOs within GRB error boxes appears to be within the statistical probability. We discuss this in more detail below.

3.6 Comparison to Markarian Galaxies and Zwicky Clusters

Both of these catalogs cover the portion of the sky accessible from the northern hemisphere. This results in a complicated galactic longitudinal and latitudinal distribution of sources. Because of this, we imposed no limits on the catalogues, using them in their entirety. This meant that we also used the entire 60 burst catalogue for matching. The Zwicky catalogue contains 9134 rich clusters of galaxies, while the Markarian catalogue contains 1525 galaxies. In neither case was there evidence for a correlation between objects in these two catalogues and GRBs, and the result was within that expected for random distributions with the same number of sources as contained in the two catalogues.

4. Discussion

We have compared the positions of 30 well localized GRBs with nearly 25,000 extragalactic objects (not all independent) contained in five all-sky catalogs, and 60 bursts with small error boxes with 10659 objects from two northern-sky catalogues. In none of the cases were the results inconsistent with that expected from random distributions of sources of the same sizes as the catalogues being investigated. While all of these catalogues suffer from incompleteness at some level, it is striking that such a considerable subset of the local visible mass of the universe does not appear to be the host sites of even a modest percentage of GRBs. This result is not totally unexpected: the deep optical searches of individual GRB error boxes discussed in the introduction have shown that no bright extragalactic objects are present in several small GRB error boxes. To accurately constrain the luminosity/distance limits at which the GRB hosts are present would require a detailed knowledge of the completeness of each catalogue. The data to make this estimate, however, are not yet available. Only the RC3 catalogue of normal galaxies can be claimed to be complete in any sense, and studies such as those by Fenimore et al., (1993) have already shown that normal-sized galaxies are unlikely hosts for GRBs.

One result from our investigation is that while the burst- source matches are consistent with random distributions, in each case the number of real matches in the first bin was always lower than that predicted from the random distribution. One simple explanation for this is that the catalogued sources do not have a completely random distribution, while the bursts do. Such a conclusion basically confirms the calculation of the GRB autocorrelation function by Hartmann & Blumenthal (1989). They show that the autocorrelation functions of well-localized bursts from the Atteia et al., and Mazets et al., (1981) GRB catalogues are consistent with zero. For normal galaxies which are clustered on smaller spatial scales, this implies that the edge of the distribution of these galaxies must extend to ÷ 140 Mpc to produce an autocorrelation function that agrees within the limits imposed by that of the GRBs if these objects are to be GRB hosts. If a group of even more spatially correlated objects are presumed to be GRB hosts (such as Abell galaxy clusters), an even greater distance must be surveyed to produce an autocorrelation function consistent within the observed GRB limits. Thus, it appears that "bright" GRBs have a more isotropic distribution than local, surveyed forms of visible matter.

Interestingly, Shanks et al., (1987) have shown that on large scales, the correlation function for QSOs is zero. We have found two cases where a QSO lies within the confines of a GRB error box. As noted above, this is consistent with chance. An additional test of that conclusion combines the observed surface density of QSOs with the areal coverage of the GRB error boxes used in this survey. Both of the QSOs found within the error boxes had apparent magnitudes of m ÷ 20. Boyle et al., (1988) find that there should be about 10 such QSOs per square degree of sky. A more recent survey by He & Chen (1993) finds a QSO density of 4.5 per square degree of sky at the same magnitude. The total areal coverage of all 30 GRBs used in this survey was 0.81 sq. deg. The more conservative QSO density from above implies that 4 GRB error boxes in our survey should have contained QSOs with m ó 20.0. The fact that two QSO-burst associations exist is therefore not surprising. An additional calculation using Fig. 2 from Boyle et al., which shows the distribution of QSO magnitudes, suggests that each of the 30 GRB error boxes in our sample should contain a QSO with mB ó 23 (assuming AV = 0.0).

If the two QSOs found in our survey, QSO 2219-420 (z = 1.3) and QSO 01116-288 (z = 0.80), were in fact the sites of their respective GRBs, then the total gamma-ray energy released by the two GRB events (assuming an isotropic outburst, and using a standard cosmological model) was Eþ > 1053 erg. As discussed by Shaefer (1992, and references therein), such an energy is greater than the binding energy of a neutron star, making extragalactic models which use this binding energy as the source of GRB's unlikely.

5. Conclusions

We have searched a number of catalogues of extragalactic objects containing > 35,000 sources for correlations with well- localized GRBs. None of these catalogued objects appears to be the hosts sites of GRBs to the completeness limits of each of the catalogues. We have found two QSOs which lie within GRB error boxes, and the one Abell cluster. We show that for a random distribution at least four of the 30 GRB error boxes could contain such QSOs. Unless the luminosity of a GRB is radiated in a highly anisotropic manner, however, it seems that such a population of QSOs is much to distant to be the host for stellar-sized cataclysms that might produce GRBs.

Overall we believe that our essentially null result with regard to the association of GRB and catalogued extragalactic objects means that indeed if GRB are extragalactic in origin they must come from very distant known objects or from objects not yet included in the catalogued visible matter.

This work was supported by the NASA Compton Gamma-Ray Observatory Guest Investigator Program under NASA Grant NAG-5-1516. The authors also thank the Natural Space Science Data Center for providing the catalogue data.

References