Simulation Study on the Discrimination of $0νββ$ Events from Single-Electron Events Using Orthogonal-Strip HPGe Detectors

This study demonstrates through simulation that orthogonal-strip HPGe detectors, analyzed via a hybrid Geant4-CNN framework, can effectively distinguish neutrinoless double beta decay events from single-electron backgrounds, providing quantitative design guidelines on how strip pitch and crystal thickness impact background rejection efficiency.

The Gist

Imagine you are a detective trying to solve a very rare crime: a Neutrinoless Double Beta Decay. This is a ghostly event where an atom spontaneously splits into two electrons at the exact same time. Finding this event is like finding a specific grain of sand on a beach, but the beach is full of "fake" grains (background noise) that look almost identical.

The problem? The "fake" grains are usually just single electrons from other radioactive sources. They look so similar to the "real" double-electron crime that standard detectors can't tell them apart.

This paper is about building a super-powered microscope (a special type of Germanium detector) and teaching a smart AI to spot the tiny differences between the real crime and the fake ones.

Here is the story of how they did it, broken down into simple concepts:

1. The Crime Scene: The "Orthogonal-Strip" Detector

Imagine a block of ultra-pure Germanium crystal. In old detectors, this block was like a single, giant bucket that just caught the energy of any particle hitting it. It knew how much energy hit, but not where.

The researchers built a new kind of detector. Think of it like a high-tech waffle iron or a garden trellis.

• They sliced the top surface into thin strips running North-South.
• They sliced the bottom surface into thin strips running East-West.
• This creates a grid (an "orthogonal" grid).

When a particle hits the crystal, it doesn't just light up the whole bucket; it lights up specific little squares on this grid. This gives the detector 3D vision, allowing it to see the shape of the path the particle took.

2. The Clue: The Shape of the Path

Here is the secret sauce:

• The Fake Crime (Single Electron): A single electron travels through the crystal like a single snake. It leaves one long, continuous trail.
• The Real Crime (Double Beta Decay): Two electrons are born at the same spot and fly off in different directions. They look like two snakes starting from the same point and slithering away.

If you have a high-resolution grid, you can see the "two snakes" vs. the "one snake." But there's a catch: as the electrons move, they spread out like a cloud of smoke. If the grid strips are too wide, the smoke blurs the two snakes together, making them look like one big blob.

3. The Simulation: A Digital Twin

Building a real detector is expensive and slow. So, the team built a virtual world inside a computer.

• They used a program called Geant4 to simulate how particles crash into the crystal.
• They created a clever "hybrid" math model to simulate how the "smoke" (the charge cloud) spreads out as it drifts toward the strips.
• This allowed them to generate thousands of fake "crimes" and "fakes" to train their system without needing a physical lab.

4. The Detective: The AI (CNN)

They didn't just look at the data with human eyes; they trained a Convolutional Neural Network (CNN). Think of this AI as a super-smart art critic.

• The AI looked at the "energy maps" (the pattern of light on the strips).
• It learned to recognize the "Two-Blob" signature of the real crime versus the "One-Blob" signature of the fake.
• It's like teaching a dog to distinguish between a Golden Retriever and a Labrador by looking at their ears, rather than just guessing.

5. The Findings: Tuning the Microscope

The researchers tested different settings to see what made the AI the best detective:

• The Strip Spacing (Pitch):

  • Analogy: Imagine looking at a picture through a window with bars. If the bars are wide apart (0.5 mm), you can't see the details. If the bars are very close together (0.1 mm), you see everything clearly.
  • Result: The closer the strips, the better the AI could tell the difference. When the strips were too far apart, the "two snakes" blurred into one, and the AI got confused.

• The Crystal Thickness:

  • Analogy: Imagine a room. If the room is too small, you can't see the whole path of the snakes. If the room is huge, the smoke spreads out so much that the snakes look like a foggy cloud.
  • Result: They found a "Goldilocks" zone. A crystal that is 20 mm thick was perfect. It was thick enough to catch the energy efficiently, but not so thick that the "smoke" blurred the clues.

The Bottom Line

This paper proves that by using a grid-like detector and a smart AI, we can filter out the "fake" background noise and find the rare "double electron" events.

They found that if you build a detector with 20 mm thick crystals and very fine strips (0.25 mm), you can reject about 79% of the fake background events while keeping 80% of the real signals.

Why does this matter?
Finding Neutrinoless Double Beta Decay would prove that neutrinos are their own antiparticles (Majorana particles) and help us understand why the universe exists. This research gives engineers a blueprint for building the best possible "microscope" to find this cosmic secret.

Technical Summary

1. Problem Statement

The search for neutrinoless double beta decay ($0\nu\beta\beta$) in $^{76}\text{Ge}$ is a critical endeavor for determining the absolute scale of neutrino mass and confirming the Majorana nature of neutrinos. However, the extreme rarity of these events requires ultra-low background conditions.

• The Challenge: A significant background source consists of single-electron events (from cosmogenic beta decays or external gamma-ray Compton scattering) that deposit energy near the $Q$-value of $^{76}\text{Ge}$ (2039 keV).
• Limitation of Current Detectors: Conventional single-electrode High-Purity Germanium (HPGe) detectors rely on Pulse Shape Analysis (PSA) to distinguish Single-Site Events (SSEs) from Multi-Site Events (MSEs). However, both $0\nu\beta\beta$ events (two electrons) and single-electron backgrounds often appear as SSEs in these detectors due to a lack of spatial resolution, making them indistinguishable.
• The Opportunity: Theoretically, $0\nu\beta\beta$ events produce a track topology with two Bragg peaks (two electron tracks), while single-electron events produce a single track. Exploiting this topological difference requires detectors with high spatial resolution, which standard HPGe detectors lack.

2. Methodology

The authors developed a comprehensive simulation framework to evaluate the potential of Orthogonal-Strip HPGe Detectors (featuring perpendicular strip electrodes on opposite surfaces) to resolve these topological differences.

A. Simulation Framework

1. Particle Interaction (Geant4):

   • Simulated $0\nu\beta\beta$ decay using the BxDecay0 package, generating two electrons with an opening angle peaking at $120^\circ$.
   • Simulated single-electron backgrounds with an initial kinetic energy of 2039 keV.
   • Used G4EmStandardPhysics_option4 with a 1 $\mu$m production range cut to track electron trajectories precisely.
   • Output: Discrete interaction points $(x, y, z, \Delta E)$ representing energy deposition.

2. Charge Cloud Modeling (Hybrid Approach):

   • To balance accuracy and computational speed, a hybrid numerical-analytical approach was developed:
     • Numerical: Used SolidStateDetectors.jl to simulate the drift and mutual Coulomb repulsion between macro-particles (clusters of charge) generated from Geant4 data.
     • Analytical: Modeled the internal expansion of each macro-particle due to thermal diffusion using Gaussian profiles.
   • Charge Collection: Calculated the energy deposited on each strip by integrating the superimposed Gaussian charge density over the strip boundaries.
   • Validation: The hybrid model was validated against full numerical simulations, showing residuals within 10 keV and reducing computation time from hours to seconds per event.

B. Event Discrimination (Deep Learning)

• Input Data: Energy profiles from the top (anode) and bottom (cathode) strip arrays were extracted as 1D vectors ($E_{top}$ and $E_{bottom}$).
• Preprocessing: Vectors were padded to a fixed length of 32 elements to standardize input.
• Model Architecture: A Dual-Branch Convolutional Neural Network (CNN) was implemented:
  • Two parallel branches processed the top and bottom energy vectors independently.
  • Each branch contained three 1D Convolutional layers (filters: 16, 32, 64; kernel sizes: 7, 5, 3) with ReLU activation and Batch Normalization.
  • Features were aggregated via Global Average Pooling, concatenated, and passed through a fully connected layer with Dropout.
  • Output: A sigmoid classification score distinguishing $0\nu\beta\beta$ (two-blob topology) from single-electron events (single-blob topology).

3. Key Contributions

1. Novel Detector Application: First systematic investigation of orthogonal-strip HPGe detectors specifically for $0\nu\beta\beta$ background suppression, moving beyond traditional PSA methods.
2. Efficient Charge Cloud Modeling: Development of a validated hybrid numerical-analytical model that accurately captures Coulomb repulsion and diffusion effects while remaining computationally feasible for large-scale Monte Carlo studies.
3. Topology-Based Classification: Demonstration that a dual-branch CNN can effectively extract subtle topological features (two-blob vs. single-blob) from orthogonal strip energy profiles to discriminate signal from background.
4. Quantitative Design Guidelines: Provided specific, data-driven recommendations for detector geometry (strip pitch and crystal thickness) to optimize experimental sensitivity.

4. Key Results

The study evaluated the impact of two critical geometric parameters: Strip Pitch and Crystal Thickness.

• Baseline Performance:

  • For a 15 mm thick crystal with 0.25 mm strip pitch, the CNN achieved an AUC of 0.840.
  • At a fixed signal efficiency of 80%, the background rejection rate was 74.6%.

• Impact of Strip Pitch (Spatial Resolution):

  • Fine Pitch (0.1 mm): AUC = 0.870; Background rejection = 79.5%. The two-blob topology is clearly resolved.
  • Coarse Pitch (0.5 mm): AUC dropped to 0.755; Background rejection fell to 59.0%.
  • Conclusion: High-granularity segmentation is critical. As pitch increases, charge cloud broadening blurs the topological features, significantly degrading discrimination power.

• Impact of Crystal Thickness (Trade-off):

  • Thin Crystals (<15 mm): Good topological resolution but lower Full-Energy Peak (FEP) efficiency.
  • Thick Crystals (>30 mm): High FEP efficiency, but long drift times cause significant charge cloud broadening, obscuring the two-blob structure and making discrimination difficult.
  • Optimization: Using a Figure of Merit (FOM) based on half-life sensitivity ($T_{1/2} \propto \epsilon_{fep} / \sqrt{1-\eta_{rej}}$), the study identified 20 mm as the optimal crystal thickness for a 0.25 mm pitch detector, balancing FEP efficiency and background rejection.

5. Significance

This work provides essential quantitative guidance for the next generation of $^{76}\text{Ge}$ $0\nu\beta\beta$ experiments.

• Feasibility: It proves that orthogonal-strip HPGe detectors can effectively suppress single-electron backgrounds by exploiting track topology, a capability previously unavailable in standard detectors.
• Design Optimization: The study establishes that high spatial resolution (fine strip pitch) is the dominant factor for success, while crystal thickness must be carefully optimized (around 20 mm) to avoid smearing topological features.
• Future Impact: These findings support the design of advanced detectors that could significantly enhance the sensitivity of future neutrinoless double beta decay searches, potentially leading to a discovery of the Majorana nature of neutrinos.