Spectral Hardness as the Primary Discriminator: Unveiling the Collapsar--Merger Boundary with a Gold-Standard Gamma-Ray Burst Sample

Technical Deep Dive: Spectral Hardness as the Primary Discriminator between Collapsars and Mergers

Overview

This work presents a robust, physically motivated method for classifying gamma-ray bursts (GRBs) into their progenitor populations of collapsars (core-collapse supernovae) and mergers (compact object coalescences). The authors train a Support Vector Machine (SVM) classifier on a "gold-standard" sample of 24 GRBs with confirmed supernova (SN) or kilonova (KN) associations. By isolating the prompt main spike and using intrinsic parameters like rest-frame peak energy ($Ep,i$), isotropic equivalent energy ($E{iso}$), and rest-frame duration ($T_{90,z}$), the SVM uncovers the primary physical discriminator between the two populations: spectral hardness.

Methodology

• The authors constructed a "gold-standard" dataset of 17 collapsar (SN-associated) and 7 merger (KN-associated) GRBs, prioritizing sample purity over quantity.
• They focused on three intrinsic prompt emission parameters: $E{p,i}$, $E{iso}$, and $T_{90,z}$.
• A linear kernel SVM was used to derive the optimal decision boundary separating the two classes.
• For "hybrid" GRBs exhibiting extended emission, the authors explicitly analyzed only the prompt main spike to isolate the central engine properties.

Results

• The SVM classifier achieved 95.8% accuracy in leave-one-out cross-validation.
• The classification index $I{SVM} = 5.01 \log{10} E{p,i} - 1.25 \log{10} E{iso} - 0.34 \log{10} T{90,z} - 12.90$ showed that $E{p,i}$ is the dominant discriminator, with a weight ~5 times higher than $T_{90,z}$.
• This quantitatively demonstrates that spectral hardness and energetics, rather than duration, are the primary physical signatures distinguishing mergers from collapsars.
• The model successfully generalized to an independent validation set of 10 GRBs with secure progenitor identities, including historic events and extreme outliers.

Interpretation

• The SVM boundary implies a "hardness ceiling" for collapsars, while mergers are identified as outliers with excessive hardness relative to their energy budget.
• Removing the soft extended emission component in "hybrid" GRBs causes them to migrate back into the merger region of the $E{p,i}$-$E{iso}$ plane, suggesting the main spike and extended emission originate from distinct physical processes.
• This classification tool can enable high-precision GRB cosmology by excising merger contamination from long-GRB samples, potentially reducing the scatter in luminosity correlations.

Limitations & Uncertainties

• The merger sample is currently limited to $z \lesssim 0.4$, though the redshift-independent rest-frame parameters mitigate this.
• The classifier is calibrated for classical GRBs and does not apply to low-luminosity, shock breakout-driven events.

What Comes Next

• Applying the $I_{SVM}$ classifier to archival and future high-redshift GRB samples to construct a "cosmology-grade" collapsar population.
• Recalibrating luminosity correlations using the purified collapsar sample to reduce systematic errors from merger contamination.
• Exploring the physical implications of the hardness ceiling and the distinct origin of the main spike versus extended emission.