⚛️ phenomenology

Investigating the most active pp collisions (top 0.1%) using the tools developed by experiments at the LHC

This paper analyzes simulated LHC proton-proton collisions to identify the most active events (top 0.1%) using various estimators, concluding that flattenicity is the optimal tool for selecting these events with minimal bias on particle ratios and collision hardness.

Original authors: Jesús Eduardo Muñoz Méndez, Antonio Ortiz

Published 2026-03-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jesús Eduardo Muñoz Méndez, Antonio Ortiz

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 you are at a massive, chaotic concert where millions of people (particles) are crashing into each other. Usually, these crashes are messy and random. But recently, scientists at the Large Hadron Collider (LHC) noticed something weird: in the most crowded, energetic crashes, the particles started behaving like a fluid, moving together in a coordinated dance, almost like a drop of water. This is strange because these are just two tiny protons colliding, not the huge, heavy nuclei usually needed to create such "liquid" behavior.

The big question is: What causes this? Is it because there are simply more people in the room (high multiplicity), or is it because the people are pushing harder against each other (harder collisions)?

This paper is like a detective story where the authors try to figure out the best way to pick the "most active" crashes from their data to solve this mystery. They tested six different "filters" or "lenses" to see which one gives the cleanest picture without tricking the results.

Here is a breakdown of their investigation using simple analogies:

1. The Problem: Choosing the "VIP" Crashes

To study these special fluid-like crashes, scientists need to pick the top 0.1% of the most energetic events. Think of it like trying to find the wildest mosh pits at a concert.

The Old Way: Most scientists just counted how many people were in the pit (Multiplicity). If the number was high, they assumed it was a "hard" crash.
The Flaw: The authors realized that just counting people might be misleading. A crowd could be huge just because the music was loud (a "hard" collision), or it could be huge because the room was just packed (a "soft" but dense collision). If you pick based on crowd size, you might accidentally only pick the loudest, hardest crashes, skewing your data.

2. The Six "Lenses" (Event Estimators)

The authors simulated 6 billion collisions using a computer program (PYTHIA) and tried sorting them using six different methods to see which one was the fairest:

Mid-Rapidity Multiplicity ( $N_{ch}$ ): Counting people in the center of the room.
Forward Multiplicity ( $V_0M$ ): Counting people at the very front and back of the room.
Sphericity & Spherocity: Measuring the shape of the crash. Did the particles fly out in a perfect circle (isotropic/liquid-like) or shoot out in a straight line like a pencil (jet-like)?
Relative Transverse Activity (RT): Looking at the crowd away from the loudest noise to see if the background is busy.
Flattenicity: A new, fancy tool that looks at how "flat" or spread out the crowd is across the whole room (both front-to-back and side-to-side).

3. The Investigation: What Did They Find?

The "Biased" Lenses (The Distorted Mirrors)

Some of the old methods turned out to be biased.

The "Hard" Bias: If you picked crashes based on the number of people in the center ( $N_{ch}$ ) or the "RT" method, you were mostly picking collisions where the particles were hitting each other incredibly hard. It's like picking the wildest mosh pits only by looking for people jumping the highest. You miss the ones that are just densely packed but moving smoothly.
The "Shape" Bias: The "Sphericity" and "Spherocity" methods were too picky. They filtered out almost all the high-energy "jets" (fast particles). It's like trying to study a concert by only looking at the people standing still; you miss the musicians playing the guitars. This made it impossible to study how jets get "quenched" (dampened) in these collisions.

The "Unbiased" Lenses (The Clear Windows)

The authors found two methods that were much fairer:

Forward Multiplicity ( $V_0M$ ): Counting people at the edges of the room helped avoid the bias of the center, but it still had some quirks.
Flattenicity (The Star of the Show): This new method was the winner. It looked at the whole picture.
- Analogy: Imagine trying to find the most chaotic room in a building. If you just count people in the hallway, you might miss the chaos in the kitchen. Flattenicity is like a security camera that scans the entire building, checking both how many people are there and how they are distributed.
- The Result: When they used Flattenicity, the data looked very similar to a "random" sample of collisions. It didn't accidentally pick only the "hard" crashes, nor did it filter out the "jets." It gave a balanced view.

4. Why Does This Matter?

The goal of this research is to understand if these tiny proton collisions are creating a tiny drop of Quark-Gluon Plasma (QGP)—the super-hot, super-dense soup that existed right after the Big Bang.

If we use the biased lenses (like just counting people), we might think the "soup" is forming because we are only looking at the hardest crashes.
If we use Flattenicity, we get a truer picture. The authors suggest that Flattenicity is the best tool to see if the "fluid" behavior is real or just an illusion caused by how we pick our data.

The Bottom Line

The paper concludes that Flattenicity is the best tool for the job. It's the least biased way to select the most active proton collisions. It allows scientists to look for "jet quenching" (a sign of the Quark-Gluon Plasma) without the results being skewed by the method they used to choose the events.

In short: Stop just counting the crowd; look at how the crowd is spread out. That's the key to unlocking the secrets of the smallest, most energetic crashes in the universe.

1. Problem Statement

Recent data from the Large Hadron Collider (LHC) has revealed unexpected "heavy-ion-like" phenomena in small collision systems (proton-proton and proton-lead), such as collective flow and strangeness enhancement. While these effects were initially observed in high-multiplicity events, recent findings suggest that multiplicity alone may not be the primary driver.

A critical challenge in analyzing these phenomena is event selection bias. Standard methods for selecting high-activity events (e.g., imposing a threshold on charged particle multiplicity) often introduce biases:

Hard-process bias: Selecting high multiplicity often inadvertently selects events with harder parton-parton scattering, skewing the sample away from the "average" collision.
Neutral-to-Charged bias: Measuring event activity and observables in the same pseudorapidity region can distort particle ratios (e.g., charged-to-neutral kaons) due to isospin symmetry violations in the selection.
Jet quenching ambiguity: To search for jet quenching (a signature of Quark-Gluon Plasma) in small systems, one needs an event classifier that does not artificially suppress or enhance jet fragmentation.

The paper aims to evaluate different event activity classifiers to determine which method best selects the top 0.1% most active collisions while minimizing these biases.

2. Methodology

The authors utilized PYTHIA 8 (version 8.312, Monash tune) to simulate 6 billion proton-proton collisions at a center-of-mass energy of $\sqrt{s} = 13$ TeV. The simulation corresponds to an integrated luminosity of approximately 70 nb $^{-1}$ and excludes pile-up effects.

The study compared six different event classifiers, selecting the top 0.1% of events based on each:

Mid-pseudorapidity Multiplicity ( $N_{ch}$ ): Count of primary charged particles in $|\eta| < 0.8$ .
Forward Multiplicity ( $V0M$ ): Count of primary charged particles in the forward regions covered by the ALICE V0 detector ( $3.4 < |\eta| < 5.1$ ).
Transverse Sphericity ( $S_T$ ): A measure of event isotropy in the transverse plane based on the eigenvalues of the momentum tensor.
Transverse Spherocity ( $S_0$ ): A measure minimizing the ratio of transverse momentum perpendicular to an axis, sensitive to jet-like vs. isotropic topologies.
Relative Transverse Activity ( $R_T$ ): The ratio of transverse region multiplicity to the average, designed to be less sensitive to hard processes by removing the leading jet.
Flattenicity ( $1 - \rho_{nch}$ ): A new observable measuring the fluctuation of particle density in "cells" of the detector (combining $\eta$ and $\phi$ information), inspired by future ALICE 3 capabilities.

Analysis Techniques:

Bias Analysis: Examined the ratio of charged-to-neutral kaons ( $K^\pm/K^0_S$ ) and the average transverse momentum ( $\langle p_T \rangle$ ) as a function of the number of multi-parton interactions ( $N_{mpi}$ ).
Jet Reconstruction: Used FastJet (anti- $k_T$ algorithm, $R=0.4$ ) to reconstruct recoil jets associated with leading particles ( $10 < p_T < 30$ GeV/c). The distribution of recoil jets was analyzed to detect autocorrelations or biases in jet acoplanarity.
Phase Space Correlation: Mapped the selected events in the $\hat{p}_T$ (average parton scattering transverse momentum) vs. $N_{mpi}$ plane to understand the physical characteristics of the selected samples.

3. Key Contributions

Systematic Comparison: Provides a comprehensive comparison of existing and novel event classifiers specifically for the top 0.1% most active events, a regime crucial for studying collective effects.
Introduction of Flattenicity: Demonstrates the utility of Flattenicity as a superior event classifier for small systems, highlighting its ability to utilize both pseudorapidity and azimuthal information to reduce biases.
Quantification of Biases: Quantifies how different classifiers distort the neutral-to-charged particle ratio and the hardness of the collision, providing a guide for future experimental analyses.

4. Key Results

A. Neutral-to-Charged Particle Ratio Bias

Biased Estimators: $N_{ch}$ , $R_T$ , $S_T$ , and $S_0$ showed significant deviations from unity in the $K^\pm/K^0_S$ ratio. Specifically, $N_{ch}$ and $R_T$ showed ratios up to 15% above unity, while $S_T$ and $S_0$ showed a rapid drop at high $p_T$ due to implicit cuts on high- $p_T$ charged particles.
Unbiased Estimators: Flattenicity and $V0M$ (forward multiplicity) maintained a ratio consistent with unity across the entire $p_T$ spectrum. This is because they calculate activity in a region distinct from the measurement region (or use a 2D cell structure), avoiding self-correlation.

B. Bias Towards Hard Physics (Jet Fragmentation)

Hard Bias: $N_{ch}$ , $R_T$ , and $V0M$ selected events that were "harder-than-average," showing an enhancement in high- $p_T$ particle yields compared to minimum bias.
Soft/Suppression Bias: $S_T$ and $S_0$ selected events with suppressed high- $p_T$ yields (jet-free events) because their definitions favor isotropic topologies, effectively filtering out jet-like structures.
Flattenicity Performance: The Flattenicity estimator produced a $p_T$ spectrum ratio that was nearly flat and consistent with the unbiased case (selection based on $N_{mpi}$ ). This suggests it does not artificially bias jet fragmentation.

C. Jet Reconstruction and Autocorrelation

Recoil Jet Statistics: For the top 0.1% class, $S_T$ $S_{T}$ and $S_0$ $S_{0}$ yielded almost no recoil jets (due to the lack of high- $p_T$ $p_{T}$ particles). In contrast, $N_{ch}$ $N_{c h}$ , $R_T$ $R_{T}$ , and $V0M$ $V 0 M$ yielded high numbers of recoil jets but with significant autocorrelation:
- $N_{ch}$ and $R_T$ showed enhancements in the $|\eta| < 0.8$ region (where they are calculated).
- $V0M$ showed asymmetry favoring the V0C or V0A detector regions.
Flattenicity Advantage: Events selected by Flattenicity showed a recoil jet pseudorapidity distribution symmetric and similar to the unbiased (minimum bias) case, with no significant autocorrelation structures.

D. Phase Space Coverage ( $\hat{p}_T$ vs. $N_{mpi}$ )

While all classifiers selected the top 0.1% of the cross-section, they sampled different physical regions.
Overlap: The overlap between most classifiers (except $S_T$ ) was extremely narrow, centered around $N_{mpi} \approx 17$ and $\hat{p}_T \approx 10$ GeV/c.
Dispersion: $N_{ch}$ , $R_T$ , and $V0M$ had broad dispersion in $\hat{p}_T$ , making them susceptible to hard-process bias.
Flattenicity: Exhibited the narrowest dispersion in $\hat{p}_T$ ( $6 < \hat{p}_T < 17$ GeV/c) while maintaining a decent $N_{mpi}$ reach, closely mimicking the "unbiased" selection.

5. Significance and Conclusion

The paper concludes that Flattenicity is the most robust event activity estimator for studying the most active pp collisions.

Reduced Bias: It minimizes bias in neutral-to-charged particle ratios and avoids the "harder-than-average" bias associated with multiplicity-based selectors.
Jet Quenching Searches: Its lack of autocorrelation and preservation of jet fragmentation patterns make it the ideal tool for searching for jet quenching in small collision systems, a key signature of the Quark-Gluon Plasma.
Future Outlook: The authors suggest that standard multiplicity-based classifiers may be hiding subtle physical effects. They recommend using Flattenicity for future analyses of LHC Run 3 and 4 data, and highlight the importance of future detectors like ALICE 3 (with extended pseudorapidity coverage) to further refine these measurements.

In summary, the study provides a critical methodological framework for interpreting LHC data, arguing that the choice of event classifier fundamentally alters the physical picture of high-activity pp collisions, and that Flattenicity offers the least biased window into these phenomena.