🔭 astrophysics

Short gamma-ray burst progenitors have short delay times

Using a hierarchical Bayesian framework to analyze a large sample of short gamma-ray bursts, this study finds that their progenitors have significantly shorter delay times (ranging from 10 to 900 Myr) than previously thought, attributing earlier estimates of gigayear-scale delays to incorrect treatments of selection effects.

Original authors: Matteo Pracchia, Om Sharan Salafia

Published 2026-08-06

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Matteo Pracchia, Om Sharan Salafia

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, chaotic construction site where stars are born, live, and die in spectacular fashion. Sometimes, when two dead stars—specifically neutron stars, which are city-sized balls of super-dense matter—get stuck in a cosmic dance, they spiral toward each other and crash. This collision is so violent it creates a "short gamma-ray burst" (SGRB), a blinding flash of high-energy light that lasts for less than two seconds. Astronomers have long known these flashes happen, but a big mystery remained: how long does it take for these two stars to meet after they are first born?

Think of it like a cosmic dating app. When two massive stars are born as a pair, they don't always crash immediately. They might orbit each other for millions or even billions of years before gravity finally pulls them together for the final, explosive embrace. This waiting period is called the "delay time." If the delay is short, the bursts happen right after a burst of star formation. If the delay is long, the bursts happen long after the stars were born, like a ghost story told generations later. Knowing the length of this delay helps scientists understand how the universe evolves and whether these explosions are the main factories creating heavy elements like gold and platinum.

For years, many scientists believed these delays were very long, perhaps taking billions of years, similar to how long it takes for a human to grow old. However, a new study by M. Pracchia and O. S. Salafia suggests this old idea might be wrong. By using a sophisticated statistical method called "hierarchical Bayesian inference"—which is essentially a super-smart way of weighing all the evidence together while accounting for the fact that our telescopes might miss the faintest flashes—the authors re-examined a large collection of these short bursts. They found that the "dating" between these neutron stars is much faster than previously thought. Instead of waiting billions of years, the average delay time is likely between 10 million and 900 million years, with a minimum wait of less than 350 million years.

The authors argue that previous studies which found longer delays made a mistake in how they handled the "selection effects." Imagine trying to guess the average height of everyone in a city, but you only measure people who can reach the top shelf in a grocery store; you would wrongly conclude everyone is tall. Similarly, the researchers show that earlier studies didn't properly account for the fact that telescopes are better at spotting bright, nearby bursts than faint, distant ones. Once they corrected for this bias, the data pointed clearly to a population of short gamma-ray bursts that happen relatively quickly after their parent stars are born.

Interestingly, the study tested two different ways of describing the brightness of these bursts: one based on a simple mathematical curve (the "empirical luminosity function") and another based on a complex model of how the explosion's jet is shaped (the "quasi-universal structured jet"). Despite using these different "lenses" to look at the data, both models agreed on the same conclusion: the delay times are short. The authors also compared their findings to the rate of neutron star mergers detected by gravitational wave observatories (like the ripples in space-time caused by the crash itself). Their calculated rate of these bursts fits well with the gravitational wave data, further supporting the idea that these events happen quickly after the stars form.

In short, this paper suggests that the universe is a much more impatient place than we thought. When two neutron stars are born, they don't wait eons to collide; they often meet within a few hundred million years. This discovery helps refine our understanding of how heavy elements are scattered across the cosmos and reminds us that even in the vastness of space, some things happen on a surprisingly human timescale.

Problem Statement
Short gamma-ray bursts (SGRBs) are widely believed to originate from the coalescence of compact binary systems, specifically binary neutron stars (BNS) or neutron star-black hole (NSBH) mergers. A critical parameter in understanding the evolution of the SGRB population is the delay time distribution (DTD), which describes the time interval between the formation of the progenitor massive star binary and the eventual merger. Previous studies have yielded conflicting results regarding the shape of the DTD and the magnitude of the delay times. Some analyses suggest long delay times on the order of gigayears (Gyr), implying a significant evolution of the cosmic rate density that departs from the cosmic star formation history (CSFH). Conversely, other theoretical considerations and local population studies suggest the existence of a "fast" merging channel with delays of only a few hundred megayears (Myr). The authors note that discrepancies in previous findings may stem from the incorrect treatment of selection effects and the specific luminosity function models employed.

Methodology
The authors employ a hierarchical Bayesian inference framework to analyze a large sample of SGRBs. The methodology involves the following key components:

Statistical Framework: Following the approach of Mandel et al. (2019) and Salafia et al. (2023, S23), the authors model the population using hyper-parameters ( $\lambda'_{pop}$ ) that define the intrinsic distributions of source parameters. The posterior probability distribution is calculated by combining the likelihood of the observed data with a selection function ( $P_{det}$ ) that accounts for detection thresholds and instrumental biases. The framework is updated to include the observed number of events via a Poissonian term to constrain the local rate density ( $R_0$ ).
DTD Models: Two distinct models for the delay time distribution are tested:
1. A power-law distribution with a minimum time delay cut-off ( $P(\tau_d) \propto \tau_d^{-\alpha_\tau}$ for $\tau_d \ge \tau_{min}^d$ ).
2. A log-normal distribution defined by a median ( $\mu_\tau$ ) and dispersion ( $\sigma_\tau$ ).
Luminosity Function Models: To test the robustness of the results against assumptions about intrinsic brightness, two models are utilized:
1. Empirical Luminosity Function (ELF): A doubly broken power-law function extending previous empirical models to lower luminosities, incorporating constraints from GRB 170817A.
2. Quasi-Universal Structured Jet (QUSJ): A physically motivated model where luminosity and peak photon energy are correlated via the viewing angle relative to a structured jet core.
Data Samples: Two primary samples are constructed from Fermi/GBM and Swift/BAT data:
1. An observer-frame sample (210 events) selected based on flux completeness and spectral peak energy cuts.
2. A rest-frame sample (18 events, 16 with redshifts) selected to maximize redshift completeness.
Reproduction of Previous Work: To isolate the source of discrepancies with prior literature, the authors reproduce the analysis of Wanderman & Piran (2015, W15) using the same sample and selection cuts but applying their hierarchical Bayesian inference method (referred to as W15*).

Key Results
The analysis yields consistent results across both luminosity function models (ELF and QUSJ) and both DTD parametrizations (power-law and log-normal):

Short Delay Times: The study finds average delay times ( $\langle \tau_d \rangle$ ) in the range of $10 \lesssim \langle \tau_d \rangle / \text{Myr} \lesssim 900$ . Specifically, median values are found to be approximately $77$ Myr (ELF, power-law) and $105$ Myr (QUSJ, power-law), with 90% credible intervals extending up to roughly 1 Gyr.
Minimum Delay: The minimum delay time is constrained to $\tau_{d,min} \lesssim 370$ Myr (90% credible limit).
Power-Law Index: For the power-law DTD, the index is found to be steep, with $\alpha_\tau > 1.5$ , indicating a preference for shorter coalescence times.
Rate Density Evolution: The inferred cosmic rate density evolution closely follows the cosmic star formation history (CSFH), contrasting with models that predict a significant lag or evolution driven by long delays.
Local Rate Density: The inferred local rate densities ( $R_0$ ) for the ELF (power-law) and QUSJ (log-normal) models are largely consistent with the BNS merger rate density constraints derived from gravitational wave observations (GWTC-4).
Discrepancy with W15: When the authors reproduce the W15 analysis (W15*) using their own inference framework, they obtain significantly longer delay times ( $\langle \tau_d \rangle \approx 3-4$ Gyr). This confirms that the difference in results is not due to the data itself but rather the methodology.

Significance and Claims
The paper claims that the primary cause of the longer delay times inferred in previous studies is an incorrect treatment of selection effects. By utilizing a rigorous hierarchical Bayesian framework that explicitly models the selection function and incorporates a broader range of luminosity functions, the authors demonstrate that the SGRB population is dominated by progenitors with short delay times.

The authors conclude that SGRBs are primarily associated with a "fast" merging channel of compact binaries, with average delays of less than a gigayear. This finding aligns the SGRB population evolution more closely with the cosmic star formation history and supports the existence of a dominant fast-merging BNS population, consistent with some local Galactic BNS studies and chemical enrichment constraints. The study emphasizes that robust inference of population parameters requires careful modeling of detection biases and the intrinsic luminosity distribution.

More like this