Assessing the impact of the electron ion collider in China on Deeply Virtual Compton Scattering

This paper evaluates the potential of the Electron-Ion Collider in China (EicC) to significantly reduce uncertainties in Compton Form Factors for Deeply Virtual Compton Scattering by employing a neural-network framework fitted to global data and projected EicC pseudo-data.

The Gist

The Big Picture: Taking a 3D X-Ray of the Proton

Imagine a proton not as a solid marble, but as a bustling, three-dimensional city made of tiny particles called quarks and gluons. Scientists want to create a perfect, high-resolution 3D map of this city to understand how it holds together, spins, and moves.

This paper is about a new, powerful tool designed to help draw that map: the Electron-Ion Collider in China (EicC). The authors are essentially running a simulation to predict how much better our "map" will look once this machine starts taking data.

The Challenge: The "Shadow" Problem

To see inside the proton, scientists use a process called Deeply Virtual Compton Scattering (DVCS). Think of this like shining a very bright, high-speed flashlight (an electron) at the proton city and watching how the light bounces off.

However, there's a catch. The light doesn't bounce off the individual buildings (quarks) directly in a way we can easily read. Instead, the information comes back as a complex, blurry signal called a Compton Form Factor (CFF).

• The Analogy: Imagine trying to figure out the layout of a room by looking at the shadows cast on the wall by a complex sculpture. You can see the shadow, but figuring out the exact shape of the sculpture from the shadow alone is incredibly difficult. There are many different shapes that could cast the same shadow. This is the "shadow problem" the paper mentions.

The Solution: A Smart AI Detective

To solve this puzzle, the researchers built a Neural Network (a type of artificial intelligence).

• The Metaphor: Think of the neural network as a super-smart detective who has studied every shadow photo ever taken by other labs (like those in the US and Europe). This detective is flexible and doesn't force the answer into a rigid box; instead, it learns the patterns of the shadows to guess the shape of the sculpture.

The authors used a software package called Gepard to train this detective on all the existing data from around the world. They then asked: "What happens if we feed this detective a massive new set of photos taken by the new Chinese collider?"

The Simulation: What the EicC Will Do

The team simulated what the EicC would see. The EicC is special because it is designed to look at the "sea-quark" region.

• The Analogy: Previous machines were great at mapping the "main streets" of the proton city (where the heavy, valence quarks live). But the "ocean" of the city (the sea of lighter, fleeting quarks) was a foggy, unexplored area. The EicC is like a new submarine designed specifically to dive into that foggy ocean.

They simulated the machine running for a year, accounting for real-world issues like detector efficiency (how good the camera is) and background noise. They generated "pseudo-data"—fake data that looks exactly like what the real machine will produce.

The Results: A Crystal Clear Map

When they fed this new, simulated data into their AI detective, the results were dramatic:

1. Shrinking Uncertainty: The "fog" around the map cleared up significantly. The uncertainty (the error bars) on the measurements dropped sharply.
2. The Sea Quark Breakthrough: The biggest improvement was in the sea-quark region. Before this, the map of the proton's "ocean" was very blurry. After adding the EicC data, the AI could draw these details with much higher precision.
3. Spatial Tomography: Because the data covers a wide range of angles and distances, the scientists can now use a mathematical trick (Fourier transform) to turn the shadow data into a true 3D spatial map. This means they can see exactly where the sea quarks are located inside the proton, not just how many there are.

The Conclusion

The paper concludes that the EicC is a game-changer. Even though the machine hasn't started taking real data yet, the simulation proves that its future measurements will drastically improve our understanding of the proton's internal structure.

The authors also note that their AI method works well as a "closure test"—meaning the AI successfully integrated the new data without breaking, proving the method is robust. However, they also warn that to get the absolute best map, they will eventually need more theoretical help (like data from supercomputers called lattice-QCD) to stabilize the edges of the map where data is still missing.

In short: The paper is a "proof of concept" showing that the new Chinese collider will act like a high-definition lens, turning our blurry, 2D shadows of the proton into a sharp, 3D map, especially for the parts of the proton we currently know the least about.

Technical Summary

Technical Summary: Assessing the Impact of the Electron-Ion Collider in China on Deeply Virtual Compton Scattering

Problem Statement
The three-dimensional structure of hadrons, encoded in Generalized Parton Distributions (GPDs), is primarily probed through Deeply Virtual Compton Scattering (DVCS). However, DVCS observables do not measure GPDs directly; instead, they access Compton Form Factors (CFFs), which are integrals of GPDs weighted by perturbative QCD kernels. Extracting GPDs from CFFs involves a complex deconvolution problem complicated by the "shadow GPD" issue and the high dimensionality of the kinematic space. While global fits to existing data from JLab, COMPASS, HERMES, and HERA have provided quantitative information on leading-twist CFFs, significant uncertainties remain, particularly in the sea-quark region and regarding the real parts of certain CFFs. The Electron-Ion Collider in China (EicC), designed to complement the US EIC and JLab programs, aims to survey the sea-quark regime with unprecedented precision. A critical question remains: how much will the specific DVCS measurements planned at EicC improve the global extraction of CFFs?

Methodology
To address this, the authors developed a comprehensive analysis framework utilizing the Gepard software package and a flexible neural network (NN) architecture.

1. Simulation and Pseudo-Data Generation:

   • DVCS events for the EicC (3.5 GeV electron beam on a 20 GeV proton beam) were generated using the MILOU Monte Carlo generator.
   • Unlike previous studies, this analysis incorporated realistic detector acceptance and efficiency based on the EicC Conceptual Design Report (CDR). This included a fast simulation framework accounting for beam crossing angles, recoil-proton detection via optimized Roman pots, and two operational modes: high-luminosity ($4 \times 10^{33} \text{ cm}^{-2}\text{s}^{-1}$) and small-angular-divergence ($1.1 \times 10^{33} \text{ cm}^{-2}\text{s}^{-1}$).
   • Pseudo-data were generated for unpolarized cross sections and all measurable spin asymmetries ($A_{LU}, A_{UL}, A_{UT}, A_{LL}, A_{LT}$) across 69 kinematic bins in $(x_B, Q^2, |t|)$, subdivided into 18 azimuthal angle ($\phi$) bins.
   • Statistical uncertainties were estimated using a likelihood-based estimator, while systematic uncertainties (radiative corrections, $\pi^0$ background, etc.) were noted but not fully propagated to the final error bands to isolate the statistical impact of the new data.

2. Neural Network Architecture:

   • The CFFs ($H, E, \tilde{H}, \tilde{E}$) were parameterized using independent neural networks for their real and imaginary parts to ensure complete independence.
   • The architecture consists of three input neurons (linearized and normalized $\xi, t, Q^2$), four hidden layers with 90 neurons each, and a single output neuron. The activation function is a modified exponential linear unit ($f(x) = 0$ for $x<0$, $e^x-1$ for $x>0$).
   • Training utilized the Adam optimizer to minimize an uncertainty-weighted Huber loss function.
   • The replica method (bootstrapping) was employed to propagate statistical uncertainties from the data to the extracted CFFs, generating a large ensemble of NN replicas.

3. Global Fit Strategy:

   • The framework was fitted to the combined worldwide dataset (JLab, COMPASS, HERMES, ZEUS, H1) with kinematic cuts ($Q^2 > 1.5 \text{ GeV}^2$, $-t/Q^2 < 0.2$) to suppress higher-twist contributions.
   • The real parts of $\tilde{E}$ and $\tilde{H}$ were assumed to vanish, following previous analyses, due to a lack of constraints in current data.
   • The EicC pseudo-data (specifically the asymmetries) were then added to the training set to assess the reduction in CFF uncertainties.

Key Contributions and Results

• Realistic Detector Modeling: The study advances previous impact assessments by incorporating realistic EicC detector efficiencies and acceptance, rather than assuming ideal conditions.
• Uncertainty Reduction: The inclusion of EicC pseudo-data leads to a significant reduction in the uncertainties of all leading-order CFFs.
  • Sea-Quark Region: The most dramatic improvement is observed in the sea-quark region ($x_B \sim 10^{-4}$ to $10^{-2}$). The uncertainties for CFFs in this domain are substantially narrowed, as illustrated by the transition from broad blue bands to tight light-red bands in the results figures.
  • Kinematic Reach: The analysis demonstrates that EicC data will allow for precise extraction of the $-t$ dependence of CFFs up to $1.0 \text{ GeV}^2$ and the $Q^2$ dependence up to $20 \text{ GeV}^2$ in the sea domain.
  • Closure Test: The global fit including EicC data remains consistent with the fit using only existing data in regions where EicC data overlaps with current measurements, serving as a non-trivial validation of the NN methodology.
• Limitations: The study notes that the real part of $H$ ($\text{Re}H$) is not expected to see a substantial reduction in uncertainty from EicC unpolarized cross-section data alone, as suggested by the dominance of the Bethe-Heitler process in the cross section. Furthermore, the extraction relies on the assumption of vanishing real parts for $\tilde{E}$ and $\tilde{H}$.

Significance and Claims
The paper claims that the EicC is uniquely positioned to fill the kinematic gap between JLab and the US EIC, specifically targeting the sea-quark regime. The primary significance of this work is demonstrating that EicC measurements of spin asymmetries will provide a powerful constraint on the extraction of GPDs, particularly for sea quarks, which are currently poorly constrained.

The authors modestly frame their work as a step toward the reliable extraction of GPDs at Leading Order (LO) using flexible parametrizations, analogous to how Parton Distribution Functions (PDFs) are extracted. They emphasize that while the NN framework successfully reduces uncertainties and validates the extraction strategy, achieving a fully stable global fit requires:

1. Additional theoretical input (e.g., improved lattice-QCD constraints).
2. Future extensions to Next-to-Leading Order (NLO) and Next-to-Next-to-Leading Order (NNLO) frameworks, which remain a major challenge.
3. The inclusion of higher-twist contributions in future fits.

Ultimately, the study concludes that the EicC will enable a more precise "tomography" of the nucleon's internal structure, specifically regarding the spatial distribution of sea quarks, provided that the experimental data is combined with the flexible NN extraction methods presented here.