⚛️ phenomenology

Resolving Local Structure of Distribution Functions in Collider Measurements

This paper demonstrates that specific collider observables linear in target distributions possess a calculable locality set by hard dynamics, enabling the resolution of narrow features in underlying distributions through a baseline-normalized function that minimizes bias from finite data point widths.

Original authors: Cong Li

Published 2026-09-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Cong Li

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

In the high-energy physics laboratories where scientists smash particles together at near-light speeds, the goal is often to reconstruct the invisible architecture of matter. When protons or electrons collide, they shatter into a spray of new particles, but the fundamental building blocks inside them—quarks and gluons—cannot be seen directly. Instead, physicists must infer their properties by measuring the debris that flies out. This process is like trying to understand the shape of a hidden object by looking at the shadow it casts on a wall. The challenge is that the "shadow" is rarely a sharp, perfect outline; it is often a blur. This blurring happens not because the detectors are imperfect, but because the very act of measuring a single point in the experiment actually gathers information from a small, finite region of the underlying distribution. If that region is too wide, the fine details of the hidden object get smoothed over, much like a photograph taken with a slightly out-of-focus lens, making it difficult to spot narrow peaks or sudden dips that might signal new physics.

A researcher, led by Cong Li at Zhejiang Ocean University, has developed a new way to measure exactly how much this blurring occurs before any data is even analyzed. They focused on a specific mathematical framework used to describe how experimental measurements relate to the theoretical distributions of particles. Their work introduces a method to calculate a "locality width," a value that tells physicists how large a region of the underlying distribution is being averaged into a single data point. If this width is zero, the measurement is perfectly sharp, capturing the value of the distribution at a single, precise point. If the width is larger, the measurement represents an average over a small area, and the size of that area determines how much fine structure gets lost. The researcher showed that this width is not a random flaw but a calculable property determined by the specific collision process and the variables chosen to measure it. By calculating this width in advance, scientists can now compare different experimental setups and choose the ones that preserve the sharpest details, ensuring that potential new discoveries hidden in narrow structures are not accidentally smoothed away.

The core of this discovery lies in rethinking how a measurement point connects to the theoretical world. In many collider experiments, a single data point is the result of integrating over many unobserved variables, such as the exact angles or energies of particles that were not detected. This integration acts as a filter, mixing information from different parts of the underlying distribution. The author demonstrated that for most standard measurements, this filter has a finite width, meaning the data point is a weighted average of the distribution over a small range. However, they proved that this averaging effect can be quantified precisely. By defining a "response kernel," which describes how the theoretical distribution is sampled by the experiment, they found that the width of this kernel dictates the resolution. If the kernel is a sharp spike, the extraction is exact; if it is a broad hill, the extraction is an average. Crucially, this width can be calculated using only the known laws of physics and the chosen measurement variables, without needing to know the actual shape of the distribution being studied. This allows researchers to predict the resolution limit of an experiment before it is built or before the data is collected.

The researcher applied this framework to a specific observable involving photon radiation at the proposed Electron-Ion Collider. In this particular setup, they found that the measurement variables completely determine the coordinate of the underlying distribution, effectively collapsing the sampling region to a single point. In this ideal case, the "locality width" is zero, meaning the experiment can extract the distribution value point-by-point without any blurring. This serves as a proof of concept that exact, sharp mapping is possible in principle. For other, more common observables where the width is not zero, the team showed that the error introduced by this blurring depends on two things: the size of the locality width and how quickly the underlying distribution changes in that region. If the distribution has a sharp, narrow feature, a large locality width will wash it out. But if the width is small compared to the feature, the structure remains visible. This provides a clear, quantitative rule for designing experiments: to see fine details, one must choose measurement variables that minimize the locality width.

This approach offers a new tool for selecting the best ways to observe the subatomic world. Traditionally, physicists have balanced factors like the number of events they can collect and the stability of their theoretical calculations when choosing what to measure. Now, they have a third, independent criterion: distribution-space locality. By calculating the locality width for different potential measurements, scientists can identify which ones are best suited to reveal narrow peaks or rapid variations that might indicate new physical phenomena. The study does not claim to solve the problem of extracting distributions for every possible experiment, nor does it replace the complex statistical methods used to analyze data after it is collected. Instead, it provides a pre-analysis diagnostic that helps researchers avoid choosing measurements that inherently blur the very features they are trying to find. The work suggests that by carefully designing the observables to minimize the mixing of information, the next generation of collider experiments could be far more sensitive to the subtle, fine-grained structures of the universe.

Technical Summary: Resolving Local Structure of Distribution Functions in Collider Measurements

Problem Statement
Collider experiments infer unobservable partonic, photonic, and other distributions from differential cross-sections, a process fundamentally treated as an inverse problem. A critical, often overlooked limitation in this extraction is that a single experimental measurement point $X$ typically samples a finite region of the underlying distribution $F$ rather than a single point. This occurs because unobserved degrees of freedom ( $\Phi$ ) are integrated out during the theoretical mapping. Consequently, even with ideal experimental resolution and infinite statistics, the "forward mapping" from the distribution space to the measurement space inherently smooths or blurs narrow features (such as peaks, dips, or rapid variations). While global fits and unfolding techniques are standard tools for recovering underlying distributions, they operate after data collection. There is a need for a pre-inversion diagnostic to determine the degree of "locality" a specific hard-scattering channel and observable possess: to what extent does a measurement point correspond to a single point in distribution space versus a finite region?

Methodology
The author proposes a framework based on factorized observables where the differential cross-section can be written linearly in a target distribution $F$ :
$\frac{d\sigma}{dX} = \int d\Phi \, H(X, \Phi) F[\xi(X, \Phi)]$
Here, $X$ represents measured variables, $\Phi$ represents unobserved/integrated-out degrees of freedom, $H$ is the calculable theoretical weight, and $\xi$ is the coordinate entering $F$ .

The core methodology involves defining a theory baseline $B(X) = \int d\Phi \, H(X, \Phi)$ and a normalized direct-extraction quantity:
$F_{\text{dir}}(X) = \frac{1}{B(X)} \frac{d\sigma}{dX}$
The author constructs a distribution-space sampling kernel (or response kernel) $\rho_X(\zeta)$ , defined as the probability density of the distribution coordinate $\zeta$ induced by the unobserved variables $\Phi$ at a fixed $X$ :
$\rho_X(\zeta) \equiv \int d\Phi \, p_X(\Phi) \delta[\zeta - \xi(X, \Phi)]$
where $p_X(\Phi) = H(X, \Phi)/B(X)$ is a normalized weight. This allows $F_{\text{dir}}(X)$ to be expressed as a weighted average:
$F_{\text{dir}}(X) = \int d\zeta \, \rho_X(\zeta) F(\zeta)$

Key Contributions and Results

Quantification of Locality:
The framework converts the qualitative question of whether a measurement probes a point or a region into a calculable physical quantity.
- Exact Locality: If $\xi(X, \Phi)$ is independent of $\Phi$ (i.e., $\xi(X, \Phi) = \xi_X$ ), the kernel collapses to a delta function $\rho_X(\zeta) = \delta(\zeta - \xi_X)$ . In this case, $F_{\text{dir}}(X) = F(\xi_X)$ , allowing point-by-point extraction.
- Finite Locality: For general observables, $\rho_X$ has a finite width. The author defines the locality width (variance) $\Delta\xi_X^2$ and the covariance matrix $\Sigma$ for multidimensional cases. These metrics quantify the spread of the distribution-space region sampled by a fixed $X$ .
Error Analysis and Bias Control:
By expanding the target distribution $F$ around the weighted center $\bar{\xi}_X$ , the author derives that the leading bias in direct extraction is controlled by the product of the locality width squared and the local curvature of the distribution:
$|F_{\text{dir}}(X) - F(\bar{\xi}_X)| \leq \frac{\Delta\xi_X^2}{2} \sup |F''(\xi)|$
This establishes a rigorous condition for "approximately local" observables: the induced locality width $\Delta\xi_X$ must be significantly smaller than the characteristic width $w$ of the physical structure being resolved ( $\Delta\xi_X \ll w$ ). If this condition is met, the integration over unobserved degrees of freedom does not introduce additional smoothing of the same order as the structure itself.
Observable Design Criterion:
The paper introduces "distribution-space locality" as a new, independent criterion for designing and selecting collider observables. Traditionally, observables are chosen based on cross-section size, statistical sensitivity, and detector resolution. The author argues that minimizing the response kernel's width (via kinematic variable selection or event cuts) is essential for preserving fine structures. They demonstrate that for a Gaussian kernel with width $\Delta_X$ probing a structure of width $w$ , the effective width becomes $\sqrt{w^2 + \Delta_X^2}$ , and the peak amplitude is suppressed if $\Delta_X \gtrsim w$ .
Exact Realization Case Study:
The author applies this framework to a previously studied Electron-Ion Collider (EIC) photon-radiation observable. In this specific process, the measured variables (photon rapidity and transverse momentum) completely determine the coordinate entering the background photon distribution, independent of unobserved kinematics. This results in a zero-width kernel ( $\Sigma = 0$ ), realizing the exact direct extraction limit where $F_{\text{dir}}(X)$ yields the exact local value of the distribution.

Significance and Claims
The paper claims to provide a unified criterion for selecting hard-scattering channels and observables that better preserve local information before data inversion (fitting or unfolding) is attempted. By calculating the response kernel $\rho_X$ and its covariance $\Sigma$ using only known theoretical weights and kinematics (without prior knowledge of the unknown distribution $F$ ), researchers can:

Compare different processes and observables to identify those that minimize distribution-space smearing.
Prioritize measurement strategies that retain unexpected fine structures, which may signal new physics.
Rigorously estimate the extraction bias for a given observable based on the local curvature of the target distribution.

The author emphasizes that this framework is complementary to existing inversion methods. It does not replace global fits but addresses the fundamental resolution limit imposed by the theoretical forward mapping itself. The EIC photon-radiation example serves as a proof-of-concept for the zero-width limit, while the general framework provides the tools to quantify and manage non-locality in more complex, realistic scenarios.

Technical Summary: Resolving Local Structure of Distribution Functions in Collider Measurements

More like this