Photoproduction in general-purpose event generators

This paper systematically compares HERWIG, PYTHIA, and SHERPA event generators for jet photoproduction in $e^+e^-$ and $ep$ collisions, finding that while all provide decent descriptions of LEP and HERA data, achieving precision predictions for the future EIC requires a global refit of photon parton distributions and dedicated measurements to constrain non-perturbative parameters.

The Gist

Imagine you are trying to predict the outcome of a massive, chaotic traffic jam caused by two high-speed cars crashing into each other. But instead of cars, these are subatomic particles (electrons and protons) smashing together at nearly the speed of light.

This paper is a "taste test" of three different simulation software programs (called event generators: HERWIG, PYTHIA, and SHERPA) that physicists use to predict what happens after these crashes. Specifically, the authors are looking at a special type of crash called "photoproduction," where a photon (a particle of light) acts like a tiny, invisible bomb that explodes into a shower of other particles.

Here is the breakdown of the paper using simple analogies:

1. The Goal: Predicting the Chaos

When these particles collide, they don't just bounce off; they explode into a spray of new particles (jets, hadrons, etc.). Physicists need to know exactly what this spray looks like to understand the laws of physics.

• The Problem: The math for the initial explosion (the "hard process") is well understood. But what happens after the explosion—how the particles slow down, stick together, and form new shapes—is messy and hard to calculate. This is called "non-perturbative" physics.
• The Solution: The three software programs (HERWIG, PYTHIA, SHERPA) act like different chefs trying to cook the same recipe. They all start with the same ingredients (the laws of physics), but they use different techniques to handle the messy parts.

2. The "Special Ingredient": The Photon

Usually, we smash protons into protons. But in this study, they are smashing electrons into protons. The electron shoots out a photon (light).

• The Twist: Sometimes, this photon is just a pure flash of light (Direct). But often, the photon acts like a "ghost" that briefly turns into a cloud of quarks and gluons (Resolved) before hitting the proton.
• The Analogy: Imagine throwing a rock (the proton) at a target.
  • Direct: You throw a rock at a target. Simple.
  • Resolved: You throw a rock, but the rock is actually a hollow shell filled with smaller pebbles. When it hits, the shell breaks, and all the pebbles scatter. This is much harder to predict because you have to guess how many pebbles were inside and how they were packed.

3. The Three Chefs (The Generators)

The authors compared how these three programs handle the "pebble scattering" (the resolved photon).

• PYTHIA: The veteran chef. It has been cooking for decades. It uses a specific recipe for the "pebbles" inside the photon and includes a lot of extra "side effects" (like particles bumping into each other after the main crash).
• SHERPA: The high-tech chef. It uses very precise, modern math (Next-to-Leading Order) to calculate the initial explosion. It's like using a supercomputer to measure the rock's trajectory.
• HERWIG: The traditionalist. It focuses on accuracy in the way particles spin and branch out, but it currently has some technical limitations in handling the "pebble" aspect of the photon.

4. The Step-by-Step Investigation

The authors didn't just compare the final dishes; they took the recipes apart to see where the differences came from. They added ingredients one by one:

1. The Remnants: What's left over after the main crash? (Like the crust of the bread).
2. The Shower: How do particles spray out? (Like steam rising).
3. The Underlying Event: Do other particles bump into each other by accident? (Like bystanders getting hit by debris).
4. Hadronization: How do the invisible particles turn into visible matter? (Like steam condensing into water droplets).

The Findings:

• The "Peek-a-Boo" Effect: One specific rule in PYTHIA (where the photon splits into quarks) changes the results significantly, making the spray look different at the edges.
• The "Crowded Room" Effect: The "Underlying Event" (MPIs) adds a lot of extra particles, especially in the resolved photon crashes. PYTHIA adds more of these than SHERPA.
• The Result: All three chefs make a dish that is "edible" (matches experimental data from past experiments like LEP and HERA), but they taste slightly different. SHERPA (with its high-tech math) and PYTHIA (with its tuned settings) did the best job matching the real-world data.

5. Why Does This Matter? (The Future)

The paper concludes by looking ahead to a new, massive particle collider called the EIC (Electron-Ion Collider) in the US.

• The Challenge: The EIC will be a "super-microscope" that will see these collisions with incredible detail.
• The Warning: If the chefs (the software) aren't perfect, the EIC data will be confusing. We need to know exactly how the "pebbles" inside the photon are packed.
• The Call to Action: The authors say we need to:
  1. Update the Recipes: The current maps of the "pebbles" inside the photon are old (from 20 years ago). We need to redraw them with modern math.
  2. Taste Test More: We need to feed the software more real data so it can learn the correct settings for the "messy" parts of the crash.

Summary

Think of this paper as a quality control report for the software that predicts particle collisions. The authors found that while the current software is good, it's not perfect. To prepare for the next giant leap in physics (the EIC), we need to update our "maps" of how light turns into matter and make sure our simulation chefs are all singing from the same songbook. Without this, we might misinterpret the new discoveries waiting at the EIC.

Technical Summary

1. Problem Statement

The paper addresses a critical gap in the theoretical modeling of jet photoproduction processes in electron-proton ($ep$) and electron-positron ($e^+e^-$) collisions. While general-purpose Monte Carlo (MC) event generators (HERWIG, PYTHIA, SHERPA) are highly optimized for proton-proton ($pp$) collisions at the LHC, their performance in photoproduction regimes is less scrutinized.

• Context: Photoproduction occurs when the exchanged photon is nearly on-shell ($Q^2 \to 0$), allowing it to fluctuate into a hadronic state. This leads to "resolved" photon interactions where the photon acts as a source of partons, similar to hadron-hadron collisions.
• The Challenge: Photoproduction probes lower energy scales than typical LHC processes, making it highly sensitive to non-perturbative corrections (hadronization, beam remnants, multiparton interactions).
• Motivation: With the upcoming Electron-Ion Collider (EIC), precision predictions for photoproduction are essential. However, current photon Parton Distribution Functions (PDFs) are outdated (from ~20 years ago), and there is a lack of systematic understanding of how different generators handle the specific modeling of resolved photons, photon fluxes, and non-perturbative effects.

2. Methodology

The authors conducted a systematic, step-by-step comparison of three leading general-purpose event generators: HERWIG 7, PYTHIA 8, and SHERPA 3.

• Process Scope: The study covers jet photoproduction in $ep$ (EIC kinematics) and $e^+e^-$ (LEP kinematics) collisions, distinguishing between direct (photon interacts as a point-like particle) and resolved (photon interacts via its partonic structure) components.
• Step-by-Step Disentanglement: To isolate the sources of discrepancies, the authors built up the event generation process incrementally:
  1. Fixed-order hard process (LO/NLO).
  2. Addition of beam remnants.
  3. Inclusion of primordial $k_\perp$.
  4. Parton showers (Initial State Radiation - ISR, Final State Radiation - FSR).
  5. Multiparton Interactions (MPIs).
  6. Hadronization.
• Key Observables:
  • $x_{\gamma}^{obs}$: The fraction of the photon's momentum participating in the hard scatter. This is the primary observable for separating direct vs. resolved contributions.
  • Jet transverse energy ($E_T$) and pseudorapidity ($\eta$).
  • Event shapes (Transverse Thrust $T_\perp$, Transverse Sphericity $S_\perp$).
  • Charged particle multiplicity.
• Data Comparison: Predictions were validated against historical experimental data from OPAL (LEP) and ZEUS (HERA).
• Future Predictions: The study generated predictions for the kinematic reach of the future EIC ($E_e = 18$ GeV, $E_p = 275$ GeV).

3. Key Contributions & Technical Differences

The paper provides a comprehensive breakdown of the architectural and input differences between the generators:

• Photon Flux & PDFs:
  • PYTHIA: Uses CJKL PDFs and includes a $\gamma \to q\bar{q}$ splitting term in the ISR algorithm (a unique feature that significantly alters the resolved cross-section). It uses a Leading Log (LL) flux with a specific tune for $p_{T,0}$.
  • SHERPA: Uses SAS2M PDFs and includes non-logarithmic (NL) corrections to the photon flux by default. It operates at MC@NLO accuracy for photoproduction.
  • HERWIG: Uses SAS PDFs. Notably, it currently lacks MPI modeling for resolved photons due to technical constraints and does not include the $\gamma \to q\bar{q}$ splitting in the same manner as PYTHIA.
• $\alpha_S$ Handling: The generators use different values and running orders for the strong coupling constant ($\alpha_S$), leading to inherent normalization differences that are not fully consistent with modern proton PDF sets.
• Remnant Modeling: The treatment of beam remnants (the leftover part of the photon/proton after a parton is extracted) differs significantly, affecting the kinematic distribution of the final state.

4. Results

A. Step-by-Step Analysis (EIC Kinematics)

• Hard Process: At the fixed-order level, PYTHIA and SHERPA agree perfectly.
• Remnants & Primordial $k_\perp$: Adding remnants reduces the cross-section at high $x_{\gamma}^{obs}$. SHERPA shows a more pronounced effect than PYTHIA.
• ISR & $\gamma \to q\bar{q}$ Splitting: In PYTHIA, the inclusion of the $\gamma \to q\bar{q}$ splitting in the ISR shifts events to higher $x_{\gamma}^{obs}$, significantly increasing the cross-section for $x_{\gamma}^{obs} > 0.8$. This effect is absent in SHERPA and HERWIG.
• MPIs: Multiparton interactions increase the cross-section at low $x_{\gamma}^{obs}$ by roughly 50% for resolved events. PYTHIA shows a slightly stronger MPI effect than SHERPA.
• Hadronization: Pushes events to lower $x_{\gamma}^{obs}$. The separation between PYTHIA and SHERPA persists even after hadronization.

B. Comparison with Experimental Data (LEP & HERA)

• Overall Agreement: All generators provide a decent description of data within uncertainties.
• SHERPA (MC@NLO): Provides the best description of the data, particularly for resolved processes, with significantly reduced scale uncertainties compared to LO.
• PYTHIA (LO): Tends to overshoot data for direct processes but agrees well for resolved-resolved cases. The inclusion of the $\gamma \to q\bar{q}$ splitting helps describe certain kinematic regions but introduces a normalization shift relative to SHERPA.
• HERWIG (LO): Generally sits between PYTHIA and SHERPA. Its lack of MPIs for resolved photons leads to a softer description of low-$x_{\gamma}^{obs}$ regions compared to the others.
• Discrepancies: A consistent undershoot is observed around $x_{\gamma}^{obs} \approx 0.8$ (the transition region) across all models, suggesting issues with photon PDFs or fragmentation tuning.

C. EIC Predictions

• Cross-Sections: SHERPA-LO yields the smallest cross-sections, while PYTHIA-LO yields the largest. The NLO K-factor is approximately 50%.
• Event Shapes: PYTHIA predicts slightly more isotropic events (higher multiplicity, lower sphericity) than SHERPA, attributed to its more active MPI modeling.
• Multiplicity: Large disagreements exist in charged-particle multiplicity predictions, with SHERPA generating fewer hadrons than PYTHIA or HERWIG. This highlights that non-perturbative modeling (hadronization/MPI) dominates these observables.

5. Significance and Conclusions

The paper concludes that while current generators can describe photoproduction data within uncertainties, precision phenomenology for the EIC requires significant improvements:

1. Photon PDF Refit: The current photon PDFs are outdated (determined >20 years ago), lack modern uncertainty estimation, and show large discrepancies in the gluon content at small $x$. A global refit using state-of-the-art methodology is a prerequisite.
2. Non-Perturbative Tuning: The large variations in MPI and hadronization modeling between generators indicate a need for dedicated experimental measurements ported to the RIVET framework to constrain these parameters.
3. Theoretical Consistency: The inconsistent handling of $\alpha_S$ and photon flux corrections (NL terms) across generators introduces systematic normalization offsets that must be resolved for high-precision physics.
4. EIC Readiness: Photoproduction is the dominant mechanism for hadronic final states at the EIC. Understanding the transition between Deep Inelastic Scattering (DIS) and photoproduction ($Q^2 \approx 1 \text{ GeV}^2$) is critical for the success of the collider's physics program.

In summary, this work serves as a foundational benchmark for the EIC community, identifying specific modeling gaps in HERWIG, PYTHIA, and SHERPA and outlining the necessary theoretical and experimental steps to achieve precision photoproduction phenomenology.