⚛️ nuclear theory

Path-length dependence of parton energy loss across collision systems: a Bayesian analysis of charged-particle RAA, consistent with a universal exponent from O+O to Pb+Pb

This study employs a Bayesian analysis of charged-particle nuclear modification factors across four collision systems (O+O, Ne+Ne, Xe+Xe, and Pb+Pb) to determine that parton energy loss in the quark-gluon plasma scales with an effective path-length exponent of $n \approx 1.78$ , decisively favoring a radiative mechanism over collisional or strong-coupling scenarios and confirming a universal energy-loss regime from small to large systems.

Original authors: Fouad A. Majeed, Hussein Ali Hussein Al Naffakh, Sarah M. Obaid, Muntaha Abdullah Reishaan

Published 2026-07-29

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Fouad A. Majeed, Hussein Ali Hussein Al Naffakh, Sarah M. Obaid, Muntaha Abdullah Reishaan

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 kitchen. Usually, the ingredients (protons and neutrons) sit quietly in their own little bowls. But sometimes, scientists smash these bowls together at speeds close to the speed of light. When they do, the ingredients melt into a super-hot, super-dense soup called the quark-gluon plasma (QGP). Think of this soup as the "primordial stew" that existed just after the Big Bang, before atoms even had a chance to form.

Inside this soup, tiny particles called partons (which are like the individual grains of salt or pepper in our stew) try to zoom through. But the soup is thick and sticky. As the partons race through, they crash into the soup's molecules and lose energy, slowing down. This process is called energy loss. Scientists have long debated how these particles lose that energy. Do they lose it slowly, like a runner getting tired step-by-step? Do they lose it faster, like a runner tripping over a hurdle every few steps? Or do they lose it in a wild, explosive burst? The answer to this question tells us the secret rules of how this cosmic soup behaves.

Now, enter a team of physicists who decided to settle this debate by looking at the size of the "soup bowls." They didn't just look at one giant bowl; they looked at four different sizes, ranging from a tiny drop (made of Oxygen) to a massive pot (made of Lead). By comparing how much the particles slowed down in the tiny drops versus the giant pots, they could figure out the exact recipe for how energy is lost.

The Great Soup Size Experiment

The researchers, led by Fouad A. Majeed and Hussein Ali Hussein Al Naffakh, took data from the Large Hadron Collider (LHC), the world's biggest particle accelerator. They looked at collisions involving four different types of atomic nuclei: Oxygen (O), Neon (Ne), Xenon (Xe), and Lead (Pb). These range from very light (Oxygen has 16 particles) to very heavy (Lead has 208 particles).

Their goal was simple but tricky: measure how the "slowing down" of particles changed as the size of the soup changed. They used a sophisticated computer method called Bayesian analysis—think of it as a super-smart detective that weighs all the clues (the data) against different theories to see which one fits best. They also used a special kind of artificial intelligence to double-check their work, ensuring they didn't miss anything.

The Big Discovery: It's a Radiative Splash!

The team found a clear pattern. As the soup got bigger, the particles lost energy much faster than a simple "step-by-step" slowdown would predict.

They measured a specific number, called an exponent, which describes how the energy loss scales with the size of the system. Their result was 1.78 ± 0.15.

Here is what that number means in plain English:

The "Step-by-Step" Theory (Collisional): This theory suggests energy loss grows in a straight line with size (like walking 1 step, then 2, then 3). The paper's data rules this out. The number 1.78 is way too high for this to be true.
The "Explosive" Theory (Strong Coupling): This theory suggests energy loss grows incredibly fast (like a cube, 1, 8, 27). The paper's data rules this out as well. The number 1.78 is too low for this wild explosion.
The "Splash" Theory (Radiative): This theory suggests energy loss grows with the square of the size (like 1, 4, 9). This is the winner! The measured number of 1.78 is very close to 2.

The authors are extremely confident in this conclusion. They ran their analysis 160 different ways, changing the settings and assumptions, and the result stayed stubbornly close to 2. They even used a "leave-one-out" test, removing one type of soup at a time, and the pattern held up every single time.

The Smallest Drop Matters Most

One of the most exciting parts of the study is that they didn't just look at the giant Lead pots. They looked at the tiny Oxygen drops, which are so small that scientists weren't even sure if a "soup" could form there at all.

Using their new method, the team found decisive evidence that even in the tiniest Oxygen-Oxygen collisions, the particles do lose energy. The statistical proof for this is incredibly strong (a value of 110, which is huge in the world of statistics). This means the "soup" forms even in the smallest collisions, and the particles get "quenched" (cooled down) just like they do in the big ones.

What They Can't Say (Yet)

The paper is very honest about what it doesn't know. The size of the soup and the density of the soup are linked; bigger bowls usually mean denser soup. Because of this, the team couldn't perfectly separate "how much the soup slowed the particle" from "how thick the soup was." They call this a "degeneracy."

So, while they know the overall effect is a "splash" (radiative loss), they can't yet say exactly how much of that is due to the path length versus the density without more experiments. However, they have made a bold prediction for the future: if the Large Hadron Collider runs collisions with Argon and Krypton (sizes in between Oxygen and Lead), the particles should slow down by a specific amount (about 0.55 for Argon and 0.41 for Krypton). If future experiments match these numbers, it will confirm that their "universal splash" theory is correct.

The Bottom Line

This paper acts like a precise ruler for the universe's hottest soup. By measuring how particles slow down in bowls of different sizes, the team proved that the energy loss follows a "radiative" rule (growing with the square of the size), effectively ruling out the simpler "step-by-step" idea and the wild "explosive" idea. They showed that even the tiniest drops of this cosmic soup are enough to stop high-speed particles in their tracks, and they've left a clear map for other scientists to verify their findings with future experiments.

Technical Summary: Path-Length Dependence of Parton Energy Loss Across Collision Systems

Problem Statement
The mechanism governing parton energy loss in the quark–gluon plasma (QGP) is encoded in how the average energy loss $\langle \Delta E \rangle$ scales with the in-medium path length $L$ . Theoretical frameworks predict distinct scaling behaviors: collisional energy loss scales linearly ( $\langle \Delta E \rangle \propto L$ ), medium-induced radiative loss scales quadratically ( $\langle \Delta E \rangle \propto L^2$ ), and strong-coupling (holographic) models predict cubic scaling ( $\langle \Delta E \rangle \propto L^3$ ). Historically, extracting this scaling exponent has been complicated by analyzing centrality variations within a single large system (e.g., Pb+Pb), which introduces biases from centrality selection and Glauber-based normalization ( $T_{AA}$ ) in peripheral events. The emergence of light-ion collision data at the LHC (O+O, Ne+Ne) offers a complementary lever: varying the system size ( $A$ ) at minimum bias to probe path lengths comparable to peripheral heavy-ion collisions, thereby isolating the system-size dependence of energy loss.

Methodology
The authors perform a Bayesian analysis of charged-particle nuclear modification factors ( $R_{AA}$ ) across four collision systems: O+O, Ne+Ne, Xe+Xe, and Pb+Pb, spanning mass numbers $A=16$ to $208$.

Forward Model: A parametric model relates $R_{AA}$ to the fractional energy loss $\Delta p_T$ . The model assumes $\Delta p_T \propto \rho G^n$ , where $\rho$ is the relative medium density, $G$ is a dimensionless geometry ratio, and $n$ is the system-size exponent to be determined. The spectral index $a(p_T)$ is fixed using data-driven measurements from $pp$ collisions, with $x_T$ -scaling applied for different collision energies.
Geometry: Collision geometry is computed using two independent Glauber implementations: an optical Glauber model and a custom Monte-Carlo Glauber model (sampling nucleon positions with hard-core exclusion). Key geometric proxies include $A^{1/3}$ , $\langle N_{part} \rangle^{1/3}$ , and the average exit path length $\langle L \rangle$ .
Statistical Framework:
- Likelihood: A correlated Gaussian likelihood is used, incorporating a bin-to-bin covariance matrix. Systematic uncertainties are modeled with a correlation length $\xi = 4$ bins to account for partial correlations, avoiding the over-constraint of fully correlated systematics.
- Inference: The posterior distribution for parameters ( $\kappa$ , $n$ , $\beta$ ) is sampled using the affine-invariant ensemble sampler (EMCEE).
- Model Selection: Nested sampling (DYNESTY) computes the Bayesian evidence to compare models with fixed exponents ( $n=1, 2, 3$ ).
- Validation: The framework includes a "simulation-based inference" (SBI) pillar using a normalizing-flow neural posterior estimator, validated via simulation-based calibration (SBC) and leave-one-system-out (LOSO) cross-validation. A Gaussian process emulator is also trained to predict held-out systems.
Robustness: The analysis is tested across 160 variants involving different geometry proxies, covariance lengths, spectral baselines, and priors.

Key Results

Effective Exponent: The analysis extracts an effective system-size exponent of $n = 1.78 \pm 0.15 \text{ (stat)} \pm 0.05 \text{ (syst)}$ for the $\langle N_{part} \rangle^{1/3}$ geometry.
Model Selection: Bayesian evidence decisively favors the radiative scaling ( $n=2$ ) over the collisional ( $n=1$ ) and strong-coupling ( $n=3$ ) hypotheses. The log-evidence differences are $2\Delta \ln Z \approx -29$ (for $n=1$ ) and $-48$ (for $n=3$ ), both exceeding the "decisive" threshold.
Microscopic Bounds: By analyzing the mapping between microscopic and effective exponents in the presence of energy-loss fluctuations, the authors establish that fluctuations can only lower the effective exponent below the microscopic one. Consequently, the measurement places a lower bound on the microscopic exponent ( $n_{micro} \ge 1.78$ ), effectively excluding purely collisional energy loss ( $n_{micro}=1$ ) at fixed geometry.
Universality: A Bayes-factor test finds no evidence for a change in the energy-loss regime between small (O+O, Ne+Ne) and large (Xe+Xe, Pb+Pb) systems, supporting a universal exponent across the full range.
O+O Suppression: The framework provides decisive Bayesian evidence ( $2\Delta \ln Z \approx 110$ ) for non-zero parton energy loss in O+O alone, quantifying the onset of suppression in the smallest system.
Transport Coefficient: The extracted energy-loss magnitude corresponds to an effective jet transport coefficient $\hat{q}/T^3 \approx 2\text{--}5$ , consistent with independent JETSCAPE determinations.

Significance and Claims
The paper claims to provide the first statistically calibrated, system-size-based extraction of the path-length exponent for parton energy loss. Its primary contributions are:

A Robust Effective Exponent: It establishes $n \approx 1.78$ as a robust, geometry-level observable that is independent of specific transport implementations, serving as a benchmark for future multi-stage energy-loss calculations.
Exclusion of Collisional Mechanisms: Through the fluctuation bound, it demonstrates that purely collisional energy loss is incompatible with the observed system-size trend, even when accounting for fluctuations.
Methodological Rigor: The work introduces a comprehensive Bayesian framework that propagates correlated uncertainties, validates interval coverage via closure tests, and employs simulation-based inference to cross-check results, ensuring the statistical defensibility of the extraction.
Predictive Power: The framework yields falsifiable predictions for future light-ion measurements (Ar+Ar and Kr+Kr), predicting $R_{AA}(10 \text{ GeV}) \approx 0.55 \pm 0.03$ and $0.41 \pm 0.03$ , respectively.

Limitations and Scope
The authors explicitly state that the result is an effective exponent, not a direct measurement of the microscopic path-length dependence. The analysis is limited by the degeneracy between medium density and path length in a minimum-bias system-size scan (both scale as $\langle N_{part} \rangle^{1/3}$ ), meaning the extracted $n$ combines both effects. The paper does not claim to resolve the microscopic mechanism (e.g., distinguishing between BDMPS-Z and finite-opacity corrections) without further inputs like centrality-differential data or azimuthal anisotropy ( $v_2$ ) measurements. The conclusions are strictly bounded to the four systems analyzed and the specific forward model assumptions employed.