🔭 astrophysics

An internal shock model calibrated with real gamma-ray burst light curves using a genetic algorithm

This paper employs a genetic algorithm to calibrate an internal shock model against real gamma-ray burst light curves from multiple catalogs, successfully reproducing key observational properties and revealing that the central engine's shell ejection follows a generalized Zipf distribution while emission times follow a negative exponential distribution.

Original authors: Manuele Maistrello, Cristiano Guidorzi, Shiho Kobayashi, Romain Maccary

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Manuele Maistrello, Cristiano Guidorzi, Shiho Kobayashi, Romain Maccary

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 is a cosmic fireworks factory, and Gamma-Ray Bursts (GRBs) are its most spectacular, chaotic explosions. For decades, astronomers have been trying to figure out exactly how the "central engine" inside these bursts works to create such wild, flickering light. One leading theory is the Internal Shock model. Think of it like a high-speed train race: the engine spits out multiple "shells" of matter at different speeds. The fast ones catch up to the slow ones, crash into them, and the resulting collision creates a flash of light.

But here's the problem: while the idea makes sense, nobody knew exactly how to tune the engine's settings to match the messy, real-life light curves (the graphs of brightness over time) we see from actual bursts. Until now.

The Great Cosmic Tuning

In this study, the authors didn't just guess the settings. They built a digital simulator and used a Genetic Algorithm—a type of computer program that works like evolution. Imagine a robot trying to solve a puzzle by making thousands of random guesses, keeping the ones that look most like the real thing, and "breeding" them to make better guesses.

They fed this robot data from three massive treasure troves of real GRBs:

BATSE (from the 1990s, looking at 25–2000 keV energy).
Swift/BAT (2005–2023, looking at 15–150 keV).
Fermi/GBM (looking at 8–1000 keV).

The robot's job was to adjust the engine's knobs until the simulated explosions looked statistically identical to the real ones. It had to match six different "metrics," including how long the bursts lasted, how bright they got, how many peaks they had, and even the subtle "fuzziness" of their light curves.

What the Engine Actually Looks Like

After the genetic algorithm did its work, the researchers found the "perfect" settings for the engine. Here is what the simulations revealed about how this cosmic engine behaves:

1. The Shell Count: A "Zipf" Law
The number of shells (the "fireworks") ejected in a single burst isn't random in a simple way. Instead, it follows a generalised Zipf distribution.

The Analogy: Think of this like earthquakes. In any given region, you have many tiny tremors, fewer medium ones, and very few massive quakes. The paper found that GRBs follow this same "heavy-tailed" rule. Most bursts have a small number of shells, but a few have a huge number.
The Math: The distribution follows a power law with an index ( $\alpha$ ) very close to 2. This is the same number that describes the frequency of earthquakes (the Gutenberg-Richter law). This suggests the central engine might be a "self-organised critical" system, where tiny changes can trigger massive cascades of activity.

2. The Timing: A Stochastic Drop
When the engine spits out these shells, it doesn't do it at a steady rhythm like a metronome. The simulations show the shells are emitted at times that follow a negative exponential distribution.

The Analogy: Imagine a radioactive atom decaying. You can't predict exactly when one specific atom will decay, but you know the probability of it happening per second is constant. The engine works the same way: every shell has the same, independent chance of being ejected in any given moment.
The Result: This means the engine starts off frantic and gradually calms down, with the time between shells getting longer on average. The "e-folding time" ( $\tau$ ) for this process varies from burst to burst, but the paper found the average $\log \tau$ values to be around 3.53 for Swift data, 3.14 for Fermi, and 2.76 for BATSE.

3. The Speed and Size
The shells aren't all the same size or speed. The model assumes lighter shells move faster. The simulations suggest the fastest shells travel at Lorentz factors ( $\gamma$ ) ranging from about 30 to 832 (depending on the dataset), while the slowest are around 23 to 33.

What the Model Doesn't Explain (The "No-Go" Zones)

It's important to know what this model doesn't do. The paper explicitly notes that while the model is great at matching the shape and timing of the light, it is a simplified version of reality.

It doesn't explain the "Low-Luminosity" crowd: The model successfully reproduces the bright, standard GRBs. However, it struggles to account for the abundance of "low-luminosity" GRBs (those with $L_{\gamma,iso} < 10^{50}$ erg s $^{-1}$ ). The authors suggest these faint bursts might have a totally different origin, perhaps coming from shock-breakout radiation or being viewed from the side (off-axis), rather than being standard engines running on low power.
It's a simulation, not a microscope: The model treats the collisions as "fully inelastic" and assumes all the energy from the crash turns instantly into light. In reality, the physics of how electrons actually radiate that energy (the "microphysics") is complex and not fully included here. The paper admits that the actual efficiency of turning kinetic energy into gamma-rays is likely lower than the model assumes, but the timing and morphology still work.

The Bottom Line

This paper doesn't claim to have solved the mystery of the central engine once and for all. Instead, it provides a highly calibrated, predictive tool. By using a genetic algorithm to tune the Internal Shock model against thousands of real bursts, the authors have shown that a relatively simple physical picture—shells crashing into each other—can reproduce the complex, chaotic dance of real GRB light curves.

The model suggests the engine is a stochastic, earthquake-like machine that spits out shells at random intervals, governed by a heavy-tailed distribution. While it can't yet explain the faintest bursts or the nitty-gritty details of particle physics, it gives us a solid, mathematically tuned blueprint for how the universe's most violent explosions might be choreographed. And because the code is open, future missions can use this "engine" to predict what new telescopes might see next.

Technical Summary: An Internal Shock Model Calibrated with Real Gamma-Ray Burst Light Curves Using a Genetic Algorithm

Problem Statement
The origin of prompt emission in Gamma-Ray Bursts (GRBs) remains a fundamental open question. While the internal shock (IS) model is a leading mechanism proposed to explain the dissipation of relativistic ejecta kinetic energy into gamma-rays, its parameters have not been fully optimized to reproduce the diverse morphological properties observed in real GRB light curves (LCs). Previous studies utilizing stochastic pulse-avalanche models lacked a direct physical connection to the underlying dissipation mechanism. A key open question is whether a physically motivated model, such as the IS scenario, can reproduce observed temporal properties when directly calibrated against real data using modern machine-learning techniques.

Methodology
The authors employed a machine-learning framework utilizing a genetic algorithm (GA) to optimize the parameters of a specific IS model (based on Kobayashi et al. 1997, hereafter K97). The optimization process involved the following steps:

Dataset: The study utilized three complementary GRB catalogues: 1,389 GRBs from Swift/BAT (2005–2023), 1,120 from Fermi/GBM, and 1,389 from CGRO/BATSE. Only long-duration GRBs ( $T_{90} > 2$ s) with sufficient signal-to-noise ratios were retained. Events with evidence of compact-object merger origins (e.g., GRB 211211A) were excluded to maintain sample homogeneity.
Model Modifications: The authors implemented a version of the K97 model with two significant modifications to the original formulation:
- Shell Emission Times: Instead of a constant emission rate, the model assumes a temporally declining shell emission rate. For a given GRB, emission times follow a negative exponential distribution, implying a stochastic process where each shell has an identical, independent, constant probability of ejection per unit time. The characteristic timescale $\tau$ varies per burst, drawn from a log-normal distribution.
- Lorentz Factor Sampling: The model introduces burst-to-burst variability in the Lorentz factor contrast. For each GRB, minimum and maximum Lorentz factors are drawn, and individual shell Lorentz factors are distributed uniformly in logarithmic space between these bounds.
- Shell Count Distribution: The number of shells ( $n$ ) per GRB is sampled from a generalized Zipf (Zipf-Mandelbrot) distribution, a heavy-tailed distribution.
Optimization Metrics: The GA minimized a total loss function based on six independent metrics comparing simulated LCs against observed data:
- Average post-peak time profile ( $\langle F/F_p \rangle$ ).
- Average third moment of the post-peak profile ( $\langle (F/F_p)^3 \rangle$ ).
- Average autocorrelation function (ACF).
- Distribution of durations ( $T_{20\%}$ ).
- Distribution of signal-to-noise ratios (S/N).
- Distribution of the number of peaks per GRB ( $N_p$ ).
Simulation Pipeline: The model generates synthetic LCs by simulating shell collisions, calculating internal energy dissipation, and converting this to bolometric luminosity. Redshifts are assigned based on a GRB formation rate model (Ghirlanda & Salvaterra 2022). The bolometric LC is then converted to observed flux within specific instrumental passbands (BATSE, Swift, Fermi) using Band function spectra and k-corrections, finally adding realistic background noise.

Key Contributions

First Systematic Optimization: This work presents the first systematic optimization of an IS model directly against multiple statistical properties of observed GRB LCs across three different instruments.
Physical Constraints on Central Engine: By replacing a purely stochastic toy model with a physical IS framework, the study constrains the fundamental parameters governing the central engine's dynamics (e.g., shell emission statistics and Lorentz factor distributions).
Discovery of Statistical Laws: The optimization reveals that the number of shells per GRB follows a generalized Zipf distribution, and rest-frame emission times follow a negative exponential distribution.

Results
The calibrated IS model successfully reproduces a wide range of observed GRB LC properties:

Temporal Profiles: The model accurately reproduces the average post-peak temporal profile, the root-mean-square deviation, the third-moment profile, and the average autocorrelation function.
Statistical Distributions: The model recovers the observed distributions for duration, S/N, peak count per burst, peak flux, and fluence.
Parameter Consistency: The best-fitting parameters (e.g., Zipf index $\alpha \approx 2$ , shift $S \approx 1-3$ , and log-normal parameters for emission times) are broadly consistent across the Swift, Fermi, and BATSE datasets.
Intrinsic Properties: The model predicts isotropic-equivalent peak luminosities ( $L_{\gamma,iso}$ ) peaking at $10^{52}–10^{53}$ erg s $^{-1}$ and total energies ( $E_{\gamma,iso}$ ) centered around $10^{53}$ erg, consistent with observations. The derived local GRB rate is $\sim 0.13–0.15$ Gpc $^{-3}$ yr $^{-1}$ .
Luminosity Function: The model yields a broken power-law luminosity function with a break near $10^{52}$ erg s $^{-1}$ . However, it predicts a lower volumetric rate of low-luminosity GRBs (ll-GRBs) compared to some observational estimates, suggesting ll-GRBs may have a distinct origin or geometric configuration not captured by the on-axis IS model.
Limitations: The model shows poorer agreement with the BATSE peak-number distribution, potentially due to instrumental response differences. It also simplifies radiation physics by neglecting microphysical shock parameters and spectral evolution due to latitude-dependent Doppler boosting.

Significance and Claims
The authors claim that despite its simplified treatment of radiation physics, the optimized IS model successfully reproduces the diverse temporal variability of GRB LCs. The study provides key insights into the central engine's activity:

Stochastic Ejection: The negative exponential distribution of emission times suggests a stochastic process where shells are ejected with a constant, independent probability per unit time, analogous to radioactive decay or the accretion of disc fragments.
Self-Organized Criticality: The heavy-tailed Zipf distribution of the number of shells mirrors the Gutenberg-Richter law for earthquakes, suggesting the central engine may exhibit self-organized critical (SOC) behavior.
Predictive Tool: The optimized model serves as a predictive tool for GRB populations expected to be detected by future missions, capable of simulating LC morphologies at any redshift to test onboard trigger algorithms.

The paper concludes that while the IS model requires further development to include microphysics and spectral evolution, this optimized formulation represents a significant step in quantitatively linking the central engine's activity to observed GRB light curves.

The Great Cosmic Tuning

What the Engine Actually Looks Like

What the Model Doesn't Explain (The "No-Go" Zones)

The Bottom Line

More like this