Kink Finder at Belle II

This paper introduces the "Kink Finder," a specialized track-finding algorithm for the Belle II experiment that significantly outperforms standard methods by achieving a 40% reconstruction efficiency for in-flight particle decays and scattering, while also improving track parameter resolution and reducing misidentification rates.

The Gist

Imagine the Belle II experiment as a high-speed, ultra-precise camera capturing the collision of tiny particles. It's like a super-slow-motion movie of the universe's smallest building blocks smashing together. Usually, these particles fly in straight lines (or smooth curves) through the camera's sensors, leaving a clear trail of "footprints" (hits) that computers can trace back to where they started.

But sometimes, things get messy. A particle might suddenly crash into something, break apart, or decay while it's still flying through the detector. This sudden change in direction creates a sharp "kink" in the path, like a runner suddenly tripping and changing direction mid-stride.

The paper you read is about a new software tool called the "Kink Finder" designed to spot these specific accidents. Here is the breakdown in simple terms:

1. The Problem: The "Ghost" and the "Clone"

In the past, the standard software at Belle II had trouble with these kinks. It was like trying to trace a path through a foggy forest where the runner suddenly changed direction.

• The Missed Kink: Often, the software would just see one long, confused line. It would think the particle was one thing (like a Kaon) when it was actually a mother particle that broke into a daughter particle. This led to mistaken identities (misidentification).
• The Clone: Sometimes, the software would get confused and draw two separate lines for the same particle, thinking they were two different people. These are called "clone tracks."
• The Result: The standard software only caught about 11% of these interesting kink events. That's like missing 9 out of 10 car accidents in a parking lot.

2. The Solution: The "Kink Finder" Algorithm

The authors built a new detective, the Kink Finder, specifically trained to look for these sudden turns. Think of it as a specialized traffic cop who doesn't just look for cars driving straight, but specifically hunts for cars that swerved, crashed, or split into two.

The Kink Finder works in two main ways:

• Case A: The Two-Track Kink. Sometimes the software already sees two separate tracks (the "mother" and the "daughter") but doesn't know they are related. The Kink Finder looks at them, realizes they connect at a sharp angle, and says, "Aha! These two are actually one story!" It then stitches them together to figure out exactly where the crash happened.
• Case B: The One-Track Kink. Sometimes the software sees only one messy track because the two particles' footprints are mixed up. The Kink Finder acts like a surgeon, carefully splitting that single messy line back into two clean lines to see what really happened.

3. The Results: A Huge Improvement

The new tool is a game-changer.

• Efficiency: Instead of catching 11% of kinks, the new algorithm catches about 40%. That's nearly four times better!
• Clarity: It fixes the "clones." It stops the software from counting one particle as two, which makes the data much cleaner.
• Accuracy: It measures the speed and direction of the particles much more precisely. Imagine trying to guess the speed of a car that swerved; the old way was a rough guess, but the Kink Finder gives you a precise speedometer reading.
• Identity Crisis: It helps the computer correctly identify what the particles are (e.g., telling a Kaon apart from a Pion), which is crucial for understanding the laws of physics.

4. Why Does This Matter?

Why do we care about particles that crash mid-flight?

• New Physics: These kinks are signatures of rare events. Finding them helps scientists look for "new physics" beyond what we currently know (like dark matter or forbidden decays).
• Precision: By measuring these kinks better, scientists can test fundamental theories of the universe with much higher precision. For example, it helps measure specific properties of how muons (a type of particle) behave.

5. The Catch (and the Future)

The tool isn't perfect yet.

• Speed: It's a bit slow. It takes a little longer to process the data than the standard method, kind of like a detective who solves crimes very accurately but takes a long time to write the report.
• Implementation: Because the Belle II software updates on a strict schedule, this new tool won't be applied to real data for another 1 to 2 years.

In a nutshell: The Belle II experiment has a new, sharper pair of glasses. It can now spot the "twists and turns" in particle paths that were previously invisible or confusing. This allows scientists to see the universe's rarest and most interesting accidents with much greater clarity, opening the door to discovering new secrets of nature.

Technical Summary

1. Problem Statement

The Belle II experiment, operating at the SuperKEKB $e^+e^-$ collider, aims to perform precision flavor physics and search for physics beyond the Standard Model. A critical challenge in tracking charged particles is the reconstruction of "kinks." A kink is the signature of a charged particle that decays or scatters in-flight within the detector material, causing a sudden, significant change in its flight direction.

Current standard track-finding algorithms in Belle II struggle with these signatures in three specific ways:

• Low Reconstruction Efficiency: Standard algorithms often fail to reconstruct the secondary (daughter) track or treat the mother and daughter tracks as a single combined track, leading to a reconstruction efficiency of only ~11% for in-flight decays.
• Misidentification (PID Fake Rates): When a mother particle (e.g., a kaon) decays in-flight, the standard algorithm may reconstruct the hits from the daughter particle (e.g., a pion) as a single track. This leads to the mother particle being misidentified as the daughter (e.g., a kaon misidentified as a pion), increasing Particle Identification (PID) fake rates.
• Clone Tracks: Standard algorithms may accidentally reconstruct a single particle's trajectory as two separate tracks (clones). The intersection of these clones can mimic a kink vertex, creating background noise.
• Resolution Degradation: When kinks are reconstructed without dedicated algorithms, the resolution of track parameters (momentum, vertex position) is poor, hindering precise measurements like Michel parameters.

2. Methodology: The Kink Finder Algorithm

The authors developed a dedicated C++/Python algorithm within the Belle II Analysis Software Framework (basf2) to identify, reconstruct, and refine kink topologies. The algorithm operates on two primary scenarios:

A. Case 1: Separately Reconstructed Tracks

When the standard algorithm finds both the mother and daughter tracks individually:

1. Preselection & Pairing: Candidates are selected based on their proximity to the Interaction Point (IP) and the Central Drift Chamber (CDC). The algorithm pairs mother and daughter candidates using six geometric configurations (3D and 2D $r\phi$-plane) to account for charge misassignment or lost hits.
2. Vertex Fitting: A kinematic vertex fit is performed to locate the kink vertex. The algorithm uses the Point of Closest Approach (POCA) of the mother track as an initial seed.
3. Hit Reassignment: An iterative algorithm reassigns hits between the mother and daughter tracks to optimize the fit. It identifies the "kink hit" (closest to the vertex) and shifts hits from one track to the other if it improves the combined $\chi^2$ per degree of freedom ($R_{comb}$).
4. Clone Suppression: To distinguish real kinks from clone tracks, the algorithm attempts to fit a single track using hits from both candidates. A bitmask is generated based on fit quality metrics (e.g., $p$-value, $n.d.f.$) to flag and suppress clones at the analysis level.

B. Case 2: Combined Tracks

When the standard algorithm merges mother and daughter hits into a single track:

1. Candidate Selection: Tracks with poor fit quality (low $p$-value) and specific hit counts are flagged as potential kinks.
2. Track Splitting: The algorithm performs a binary search to find the optimal splitting point along the track. It tests three initial ratios (80/20, 50/50, 20/80) and iteratively refines the split to minimize the deviation of the combined $\chi^2$ metric ($|R_{comb} - 1|$).
3. Refitting: Once split, the resulting mother and daughter tracks are refitted, and the kink vertex is reconstructed.

3. Key Contributions

• Novel Algorithm: The first dedicated kink-finding algorithm implemented for the Belle II experiment, addressing the specific geometry and software constraints (genfit2) that prevent direct porting of algorithms from other experiments like NOMAD.
• Dual-Mode Operation: The ability to handle both separately reconstructed tracks and combined tracks, significantly broadening the scope of detectable in-flight decays.
• Advanced Hit Reassignment: An iterative procedure to correct hit ownership between mother and daughter tracks, improving momentum and vertex resolution.
• Clone Suppression Mechanism: A specific bitmask-based filter to identify and remove track pairs that are actually clones of a single particle, reducing background.

4. Results

The algorithm was validated using Monte Carlo (MC) simulations of $\tau^+\tau^-$ and $B\bar{B}$ events, including beam-induced background.

• Reconstruction Efficiency:
  • The Kink Finder achieves a reconstruction efficiency of ~40% for in-flight decays (kaons and pions).
  • This is a significant improvement over the standard algorithm's ~11%.
  • Specific efficiencies: ~80% for kaon decays and ~50–65% for pion decays (depending on the production process).
• Resolution Improvements:
  • Vertex Resolution: The spatial resolution of the kink vertex is improved to 0.9 cm for kaon decays and 2.4 cm for pion decays.
  • Momentum Resolution: The Kink Finder eliminates biases in transverse momentum and $p_{dm}$ (daughter momentum in mother rest frame) caused by energy loss corrections. For example, a 2–3 MeV/c bias in transverse momentum for kaons is completely removed.
  • PID Fake Rates: Excluding particles identified by the Kink Finder reduces the PID fake rate for pions by 5% (up to 10% for low $p_T$) and for kaons by 1%.
• Clone Suppression: The algorithm suppresses clone tracks mimicking kinks by a factor of two using the combined fit veto.
• Physics Impact: The improved resolution allows for the clear separation of two-body decay peaks (e.g., $K^- \to \pi^- \pi^0$ and $K^- \to \mu^- \bar{\nu}_\mu$) in the daughter momentum distribution, which is crucial for measuring Michel parameters.

5. Significance and Future Outlook

• Physics Reach: The Kink Finder enables the measurement of previously unmeasured physics processes and improves the precision of Standard Model tests, specifically the measurement of the Michel parameter $\xi'$ in $\tau \to \mu \nu \nu$ decays.
• Data Processing: While the algorithm is implemented in the software framework, it has not yet been applied to real data due to the Belle II software release schedule. Full data reprocessing is expected in 1–2 years.
• Future Improvements: The authors identify several areas for optimization:
  • Integrating Graph Neural Networks (GNN) for track finding to improve secondary track efficiency.
  • Implementing Machine Learning to better distinguish genuine kinks from ordinary tracks with poor fit quality.
  • Optimizing the computational speed (currently 13–111 ms per call) by caching Kalman filter results.
  • Utilizing geometric constraints of coupled helices for even better vertex resolution.

In summary, the Kink Finder represents a substantial upgrade to the Belle II tracking capabilities, transforming a previously difficult-to-reconstruct signature into a robust tool for precision physics and background suppression.