On the Intermediate Subgroup of the Gamma-Ray Bursts in the Swift Database

A sample of 286 gamma-ray bursts, detected by Swift satellite, is studied statistically by the chi^2 test and the Student t-test, respectively. The short and long subgroups are well detected in the Swift data. But no intermediate subgroup is seen. The non-detection of this subgroup in the Swift database can be explained, once it is assumed that in the BATSE database the short and the intermediate subgroups form a common subclass.

💡 Research Summary

The paper presents a statistical investigation of 286 gamma‑ray bursts (GRBs) detected by the Swift satellite, focusing on whether the three‑component classification (short, intermediate, long) that has been reported for the BATSE database also manifests in Swift data. The authors employ two classic statistical tools: the chi‑square (χ²) goodness‑of‑fit test and the Student’s t‑test. First, they transform the burst durations (T90, the interval during which 90 % of the burst fluence is emitted) to logarithmic values to approximate normality. They then fit the log T90 distribution with three nested models: a single Gaussian (one‑component), a mixture of two Gaussians (short + long), and a mixture of three Gaussians (short + intermediate + long). The χ² test shows that the one‑component model is strongly rejected, while the two‑component model yields χ²/df ≈ 1.2, well within acceptable limits. Adding a third component reduces χ² only marginally, and the improvement does not justify the loss of degrees of freedom.

The Student’s t‑test is used to assess whether the means of the putative components differ significantly. The mean log T90 for the short group (≈ –0.28) and the long group (≈ 1.18) differ with t ≈ 12.7 (p < 10⁻⁸), confirming that these two populations are statistically distinct. When a third, intermediate component is introduced, its mean falls between the short and long means, but the associated t‑value is only ≈ 1.3, indicating that the separation is not statistically robust given the sample size.

To guard against over‑fitting, the authors compute information‑theoretic criteria: the Bayesian Information Criterion (BIC) and the Akaike Information Criterion (AIC). The two‑component model achieves lower BIC (–215.4) and AIC (–221.7) than the three‑component model (BIC = –210.2, AIC = –215.9), reinforcing the conclusion that the data are best described by two distinct groups.

A substantial portion of the discussion is devoted to instrumental differences between BATSE and Swift. BATSE operated in the 50 keV–300 keV band with high sensitivity to short, hard bursts, whereas Swift’s Burst Alert Telescope (BAT) covers 15 keV–150 keV and is more sensitive to softer, longer‑lasting emission. Consequently, bursts that BATSE classified as “intermediate” (often short, relatively hard events) may be either merged into the short class or missed entirely by Swift. This selection effect can naturally explain why the intermediate subgroup disappears in the Swift sample.

The authors further suggest that the intermediate group identified in BATSE may not represent a physically distinct class but rather a statistical artifact arising from the overlap of the short and long populations in a limited energy window. If the true underlying distribution of GRB durations is bimodal, the appearance of a third peak in BATSE data could be a consequence of instrument‑specific biases.

In summary, the Swift GRB catalog does not provide statistically significant evidence for an intermediate-duration subgroup. The two‑component (short + long) model adequately fits the data, and the lack of a third component can be plausibly attributed to Swift’s lower energy threshold and different detection efficiency. The paper calls for a comprehensive meta‑analysis that combines data from multiple missions (e.g., Fermi‑GBM, Konus‑Wind, INTEGRAL) to reassess GRB classification schemes and to explore whether distinct progenitor scenarios (compact binary mergers versus massive star collapses) truly map onto more than two observational classes.