A novel method for lepton energy calibration at Hadron Collider Experiments

This paper presents a novel method for improving lepton energy calibration at hadron collider experiments by introducing additional parameterization terms and reducing parameter correlations through kinematic separation of $Z/\gamma^* \to \ell^+\ell^-$ samples, thereby achieving higher precision while remaining faster and simpler than detailed detector simulations.

The Gist

Imagine you are trying to measure the speed of cars on a highway using a radar gun. But there's a problem: your radar gun is slightly broken. Sometimes it reads 5 mph too slow, and other times it reads 10 mph too fast depending on how far away the car is.

In the world of particle physics, scientists at the Large Hadron Collider (LHC) face a similar problem. They smash protons together to create particles like electrons and muons (let's call them "leptons"). To understand what happened in the crash, they need to know the exact energy of these leptons. But their detectors aren't perfect; they have "biases" or errors that make the energy readings slightly off.

The Old Way: The "One-Size-Fits-All" Fix

For a long time, scientists used a classic method to fix this. They would look at a very specific event: a "Z boson" decaying into two leptons. They know exactly how heavy a Z boson is (like knowing a standard car weighs exactly 2,000 lbs).

If their detector says the two leptons weigh 1,900 lbs combined, they know the detector is under-reporting. So, they apply a single magic multiplier (let's call it $k$) to fix everything.

• Analogy: Imagine your scale says you weigh 190 lbs when you actually weigh 200 lbs. You just multiply every future reading by 1.05. Done!

The Problem: This "single multiplier" works okay if the error is consistent. But in reality, the error changes.

• If a particle has low energy, the error might be a fixed amount (like a static noise of +5 lbs).
• If a particle has high energy, that same +5 lbs noise becomes a tiny percentage, but the scale might also stretch or shrink differently.

The old method ignored the "fixed noise" (the offset). It assumed the error was just a scaling issue. This meant that for very low-energy particles, the correction was wildly inaccurate, and for very high-energy ones, it was also slightly off. It was like trying to fix a crooked picture frame by only stretching the canvas, ignoring that the frame itself was bent.

The New Method: The "Multi-Tool" Approach

The authors of this paper propose a smarter way. Instead of just one magic number ($k$), they use two numbers:

1. $k$ (The Multiplier): To fix the stretching/shrinking.
2. $b$ (The Offset): To fix the static noise or "zero-point" error.

The Challenge: You can't just guess two numbers. If you have one equation (the known weight of the Z boson), you can only solve for one unknown. If you try to solve for two, the math gets messy and the answers bounce around wildly (like trying to find the intersection of two lines that are almost parallel).

The Solution: The "Traffic Jam" Strategy

Here is where the paper gets creative. The authors realized they could split the traffic of particles into different groups based on how they are moving.

1. Separate the Crowd: They look at the angle between the two leptons.

   • Group A: Leptons flying apart at wide angles (like cars driving in opposite lanes). These usually come from "slower" Z bosons.
   • Group B: Leptons flying close together (like cars in a tight convoy). These come from "fast" Z bosons that got a boost.
   • Group C: Everything in between.

2. Create Multiple "Checkpoints": By splitting the data this way, they create different "sub-samples." Each sub-sample has a different average energy.

   • Analogy: Imagine you have a broken ruler. Instead of just measuring one object, you measure a pencil, a book, and a table. Because the objects are different sizes, the way the ruler is broken (is it stretched? is the zero mark shifted?) affects each measurement differently. By looking at all three, you can figure out exactly how the ruler is broken.

3. The Math Trick: The authors developed a clever way to link the two numbers ($k$ and $b$) so they don't fight each other. They realized that for particles in the same "lane" (same energy region), the relationship between the multiplier and the offset is predictable. This allows them to solve for both numbers simultaneously without the math getting confused.

Why This Matters

• Precision: The old method was like guessing the temperature with a thermometer that was off by a few degrees. The new method is like using a high-precision lab sensor. They can now calibrate the energy to within 0.01% (or even better with more data).
• Speed: Usually, to fix these errors, scientists have to run massive, slow computer simulations of the entire detector to see where the errors come from. This new method uses the actual data (the Z bosons) as the ruler. It's much faster and easier.
• Universality: It works for particles going in any direction (even the tricky "forward" ones) and for different types of particles (electrons and muons), even if the muons behave slightly differently based on their electric charge.

The Bottom Line

Think of the old method as trying to tune a piano by only listening to one note. The new method listens to the whole chord, separates the notes based on how they are played, and uses that rich information to tune every single key perfectly.

This allows physicists to see the "true" energy of particles with incredible clarity, leading to better discoveries about the fundamental laws of the universe.

Technical Summary

1. Problem Statement

In hadron collider experiments (e.g., ATLAS, CMS), precise measurement of electron and muon energy scales is critical for physics analyses. The standard calibration method relies on $Z/\gamma^* \to \ell^+\ell^-$ events, using the invariant mass of the dilepton system to constrain the lepton energy scale.

Limitations of the Classic Method:

• Imperfect Parameterization: The classic method typically uses a single scaling factor ($k$) to correct observed energy ($E_{obs}$) to true energy ($E_{true}$), assuming a linear relationship $E_{corr} = k \cdot E_{obs}$.
• Offset Bias: In reality, energy measurements often suffer from an offset bias ($b$) due to pile-up, noise, or reconstruction inefficiencies, leading to a relationship $E_{true} = b + k \cdot E_{obs}$.
• Energy-Dependent Bias: If the calibration ignores the offset $b$ and uses only a scaling factor $k'$, the resulting correction introduces an energy-dependent bias:
  $$\frac{E_{corr} - E_{true}}{E_{obs}} \approx \frac{b}{E(E_{obs})} \left( 1 - \frac{E(E_{obs})}{E_{obs}} \right)$$
  This bias is significant for leptons with energies far from the mean energy of the calibration sample.
• Under-constrained System: Introducing both $k$ and $b$ parameters creates two unknowns per kinematic region. However, the classic method provides only one constraint (the Z boson mass peak) per sample, making it impossible to uniquely determine both parameters without additional assumptions or complex simulations.

2. Methodology

The authors propose a novel method that allows for the simultaneous determination of both the scaling factor ($k$) and the offset ($b$) by introducing multiple independent constraints and reducing parameter correlations.

A. Multiple Mass Constraints via Kinematic Separation
Instead of treating all $Z \to \ell^+\ell^-$ events as a single sample, the method separates them into subsamples based on the opening angle between the two leptons (which correlates with the Z boson boost).

• Region Definition: The detector pseudorapidity ($\eta$) is divided into three distinct regions: Low ($L$), Medium ($M$), and High ($H$).
• Subsample Formation: Events are categorized by the $\eta$ regions of the two leptons (e.g., $LL$, $LM$, $LH$, $MM$, $MH$, $HH$).
• Kinematic Diversity: Leptons in a specific $\eta$ region (e.g., $L$) have different mean energies depending on the $\eta$ of the other lepton. For example, an $L$-region lepton paired with an $H$-region lepton ($LH$) has a different mean energy than one paired with another $L$-region lepton ($LL$).
• Constraint Counting: With 3 regions, there are 6 subsamples providing 6 mass constraints. This is sufficient to solve for the 6 unknown parameters ($k_L, b_L, k_M, b_M, k_H, b_H$).

B. Reducing Parameter Correlation
A direct fit for $k$ and $b$ simultaneously often fails due to strong correlations between the parameters, leading to unstable $\chi^2$ minimization.

• The Solution: The authors derive a relationship between $k$ and $b$ using events where both leptons are in the same $\eta$ region (e.g., $LL$).
• The $\epsilon$ Parameter: By analyzing the covariance between the mass and energy spectra, they express $b$ as a function of $k$ and a small correction term $\epsilon$:
  $$b = -E(E_{obs}) \cdot k + E(E_{obs}) \cdot \left[ \frac{E(M_{true})}{E(M_{obs})} + \epsilon \right]$$
• Fitting Strategy: Instead of fitting $k$ and $b$ directly, the method fits for $k$ and $\epsilon$. Since $\epsilon$ is physically small (order of $10^{-3}$), its fit range is tightly constrained, effectively decoupling it from $k$ and stabilizing the fit.

C. Alternative Scenarios

• Forward Leptons: For high-$\eta$ regions where double-forward events are rare, the method uses a "central-forward" topology (one lepton in a calibrated central region, one in the forward region). By varying the central region subsamples, multiple lines of $b_F$ vs. $k_F$ are generated, and their intersection determines the unique solution.
• Muon Calibration: For muons, charge-dependent misalignment introduces $k^+$, $b^+$, $k^-$, and $b^-$. The method adds constraints based on the ratio of positive to negative muon energies ($R = E_{\mu^+}/E_{\mu^-}$), which cancels out PDF and QCD effects, providing the necessary extra constraints to solve for the 12 parameters.

3. Key Contributions

1. Dual-Parameter Calibration: The method successfully determines both the scaling factor ($k$) and the energy offset ($b$) simultaneously, eliminating the energy-dependent bias inherent in single-parameter calibrations.
2. Data-Driven Efficiency: Unlike methods relying on detailed detector simulations to estimate biases, this approach is purely data-driven, using only the reconstructed dilepton mass spectrum.
3. Correlation Reduction Technique: The introduction of the $\epsilon$ parameter and the linear relationship derivation allows for a stable, fast fit without requiring advanced, time-consuming minimization algorithms.
4. Robustness: The method is shown to be insensitive to Parton Distribution Function (PDF) uncertainties and QCD calculation differences, as these affect energy and mass differently, but the mass constraint isolates the scale.

4. Results

The method was validated using Pythia-generated samples equivalent to 35 fb$^{-1}$ of LHC data at 13 TeV (72 million events) and larger samples (1000 million events).

• Precision:
  • The relative uncertainty on the determined parameters ($\delta k$ and $\delta b$) is approximately 0.2% for the 35 fb$^{-1}$ sample.
  • With larger statistics (1000M events), the relative uncertainty drops to < 0.05%.
• Bias Reduction:
  • The residual energy-dependent bias ($\Delta E / E$) is reduced from the classic method's $\sim 1\%$ (due to uncorrected offsets) to < 0.01% (dominated by statistical fluctuations of $\delta k$).
• PDF Independence: Tests using different PDF sets (NNPDF3.1 NLO vs. LO) showed that the difference in determined $k$ and $b$ values was negligible ($< 0.0006$), confirming the method's independence from theoretical modeling.
• Forward/Muon Tests: The extended methods for forward leptons and charge-dependent muon calibration successfully recovered injected parameters with uncertainties consistent with statistical expectations.

5. Significance

This paper presents a significant advancement in experimental particle physics calibration techniques.

• Physics Impact: By reducing the energy-dependent bias to negligible levels, this method improves the precision of measurements involving high-energy leptons (e.g., $W$ boson mass, Higgs boson decays, and searches for new physics), where the classic method's bias would otherwise be a limiting systematic uncertainty.
• Operational Efficiency: The method is computationally efficient ("much faster and easier") compared to running full detector simulations or complex iterative fitting procedures. It can be applied directly to large datasets as they are collected.
• Scalability: As luminosity increases at the LHC, the statistical precision of this method improves, whereas the systematic bias of the classic method remains constant. This makes the new method essential for the High-Luminosity LHC era.