Skip to main content

Machine Learning Challenges Earlier Concepts about Formation of Long and Short Duration Gamma-Ray Bursts

Machine learning algorithms have revealed unexpected complexity in Gamma-Ray Bursts (GRBs), challenging earlier concepts that Long-duration GRBs, originate from the collapse of massive stars and short-duration GRBs, arise from compact binary mergers. Two distinct populations of compact binary merger (associated with kilonovae) GRBs identified through the algorithms, showed that the two clusters of kilonova-associated GRBs have similar compactness but differ in their duration. This new unexpected complexity provides new insights into the nature of these powerful cosmic explosions.

GRBs are one of the most luminous transient astrophysical phenomena in the Universe. These events are traditionally classified as long or short, based on the duration of their prompt emission. Long-duration GRBs, with duration greater than 2 seconds, are thought to originate from the collapse of massive stars. In contrast, short-duration GRBs, with duration of less than 2 seconds, are thought to arise from compact binary mergers. However, recent observations have challenged this traditional classification, with some long-duration GRBs found to have originated from the compact binary merger (associated with kilonovae) and some short-duration GRBs are suspected to originate from the death of massive stars (associated with supernovae).

To solve the conundrum, a recent study published in The Astrophysical Journal Letters by a team of researchers from Aryabhatta Research Institute of Observational Sciences (ARIES), Nainital, an autonomous institute of the Department of Science & Technology (DST), Govt. of India, and Chennai Mathematical Institute (CMI), Chennai, employed prompt emission light curves of GRBs and machine learning techniques to find evidence for two distinct populations of compact binary merger (associated with kilonovae).  

The algorithms identified five distinct clusters of GRBs, with the kilonova-associated GRBs located within two separate clusters. This suggests that these kilonova-associated GRBs having similar compactness but different duration may have been produced by different progenitor systems or due to subclasses of neutron star-neutron star (NS-NS) and/or neutron star–black hole mergers (NS-BH). 

kilonova-associated Gamma-Ray Bursts (GRBs)

Figure 1: The locations of kilonova-associated Gamma-Ray Bursts (GRBs) on a two-dimensional map obtained using PCA-tSNE and PCA-UMAP. The coral-coloured labels represent GRBs with confirmed kilonova association, while the turquoise-coloured labels represent GRBs with probable kilonova candidates. The kilonova-associated GRBs with long and short duration occupy two locations on the map (right and left corners). The filled-grey squares show the location of GRB-SNe on the map. The map is color-coded with respect to the five clusters identified by an algorithm.

Overall, the study demonstrates the power of machine learning techniques for uncovering unexpected complexity within astronomical populations, with the discovery of two distinct kilonova-associated GRB populations and could have important implications for future gravitational wave observations. This could also help understand the different clusters of GRBs in more detail and potentially provide new insights into the nature of these powerful cosmic explosions.

Publication Link: DOI 10.3847/2041-8213/acd4c4

For more details, contact dimple[at]aries[dot]res[dot]in, kuntal[at]aries[dot]res[dot]in, kgarun[at]cmi[dot]ac[dot]in

This story is also available on DST website.