Operational characterization of LAPPD Generation 2: charge sharing, delayed pulses, and dark-count behavior

This paper presents a comprehensive operational characterization of second-generation Large-Area Picosecond Photodetectors (LAPPD Gen 2), detailing their charge-sharing mechanisms, dark-count relaxation behaviors, and resonant cavity effects through experimental measurements and first-principles Monte Carlo simulations.

The Gist

Imagine a giant, high-tech camera sensor the size of a dinner plate, but instead of taking photos of people, it takes "photos" of individual particles of light (photons) traveling at the speed of light. This device is called the LAPPD Gen 2. It's designed to be incredibly fast and precise, helping scientists study mysterious particles like neutrinos.

This paper is like a "user manual" and a "troubleshooting guide" rolled into one. The researchers wanted to understand exactly how this camera behaves when it's turned on, specifically looking at three main things: how signals get mixed up, how the device behaves when it's "dark," and how to predict what it will see using a computer simulation.

Here is a breakdown of their findings using simple analogies:

1. The "Echo Chamber" Effect (Charge Sharing & Cross-Talk)

Think of the LAPPD as a grid of 64 tiny trampolines (pixels) sitting next to each other. When a particle of light hits one trampoline, it bounces up and creates a signal.

• The Problem: The researchers found that when you jump on one trampoline, the vibration doesn't just stay there. It ripples out to the neighbors.
  • Charge Sharing: This is like the trampoline fabric stretching and pulling the neighbor slightly. The signal is shared, but it's a gentle, positive push.
  • Cross-Talk: This is more like a loud "echo" or a negative jolt that happens in the neighboring trampolines because they are electrically connected.
• The Finding: The "echo" is strongest on the trampoline directly next to the one you hit, and it fades away quickly. If you jump on a trampoline, the one two spots away barely feels it.
• The Twist: Usually, engineers hate "echoes" because they mess up the picture. But the authors suggest this might actually be a superpower. Because the echo changes slightly depending on exactly where the light hit, scientists can use these echoes to pinpoint the location of the light even more precisely than the trampoline size would normally allow.

2. The "Ghostly" Noise (Dark Counts & Muons)

When the room is pitch black, a perfect camera should see nothing. But this camera sees "ghosts."

• The "Dark Count" Spike: The researchers turned up the voltage (the power) to the device. Suddenly, the camera started screaming with noise—thousands of fake signals per second. It was like turning up the volume on a radio until it just hissed.
  • The Recovery: When they turned the power back down, the noise didn't stop instantly. It took about half a day for the camera to "cool down" and go quiet again. This suggests the device has an internal "recovery process," like a muscle that stays tense after a workout.
• The Muon Mystery: While testing, they saw huge, loud signals that happened about once or twice a second. They realized these were cosmic rays (muons) from space hitting the device.
  • The Analogy: Imagine a single raindrop hitting a drum (a normal light signal). Now imagine a bowling ball hitting the drum (a cosmic ray). The bowling ball makes a massive boom that echoes for a long time, shaking the whole drum set. The researchers learned to recognize this "bowling ball" sound so they don't mistake it for a real particle they are trying to study.

3. The "Time Travel" Echoes (After-Pulses)

Sometimes, the camera sees a signal, and then a split second later, it sees another signal from the same event.

• The Phenomenon: It's like clapping your hands, and then hearing a second clap 60 or 110 nanoseconds later.
• The Cause: The researchers think this happens because of two things:
  1. Bouncing Back: Electrons hit a wall inside the camera and bounce back (backscatter).
  2. Ion Feedback: Tiny, heavy particles (ions) get created, float around slowly, and then hit the sensor later, creating a delayed "ghost" signal.
• The Simulation: To understand this, the team built a virtual reality video game of the LAPPD. They programmed virtual electrons to fly through the device. The game showed that these "ghost" signals are real physical events caused by electrons bouncing off walls or ions drifting back. The simulation matched their real-world observations pretty well.

4. The "Resonant Cavity" (Radio Interference)

The researchers also tested if the device acts like a radio. They sent radio waves into it.

• The Finding: The device acts a bit like a singing wine glass. If you hum at the right pitch (frequency), the glass vibrates loudly. The LAPPD vibrates (creates electrical noise) at specific radio frequencies (like 180 MHz and 550 MHz). This means if you put this device near a strong radio transmitter, it might get confused and start making noise.

The Big Takeaway

The LAPPD Gen 2 is an amazing, fast detector, but it's not perfect. It has "echoes" (cross-talk), it gets "tired" and noisy when you push the power too hard, and it sees "ghosts" (after-pulses) from bouncing particles.

The paper concludes that to use this device effectively (for example, in future neutrino experiments), scientists need to:

1. Accept the trade-offs: You can't have the lowest noise and the fastest speed at the exact same time; you have to choose what matters most for your specific experiment.
2. Use the "echoes": Instead of fighting the cross-talk, use it to get better location data.
3. Filter the noise: Use computer programs (like the simulation they built) to tell the difference between a real particle, a cosmic ray "bowling ball," and a "ghost" echo.

In short, they mapped out all the quirks and glitches of this high-tech camera so future scientists can use it to see the universe more clearly.

Technical Summary

Technical Summary: Operational Characterization of LAPPD Generation 2

Problem Statement
This study addresses the operational characterization of second-generation Large-Area Picosecond Photodetectors (LAPPD Gen 2), specifically focusing on signal propagation, electronic cross-talk, and dark-count behaviors. While previous models of Microchannel Plates (MCPs) focused primarily on gain and uniform electric fields, they did not adequately address signal spreading between pixels in modern, pixelated readout systems. Understanding these phenomena is critical for improving the accuracy of readout systems in next-generation neutrino observatories (such as CHESS, ANNIE, and THEIA/DUNE), which require high temporal resolution and large detection areas to separate scintillation from Cherenkov photon emissions.

Methodology
The authors employed a combination of experimental measurements and first-principles Monte Carlo simulations:

• Experimental Setup: A LAPPD Gen 2 device (featuring a bialkali photocathode, two 203 mm × 203 mm MCPs, and a resistive anode coupled to an 8 × 8 pixelated readout board) was placed in a light-tight enclosure. A 450 nm picosecond pulsed laser was used to scan specific pixels (e.g., E5) while a Tektronix MSO64B oscilloscope (25 GSa/s) recorded signals from the target and neighboring pixels.
• Cross-Talk and Resonance Analysis: The team measured charge sharing by integrating the area under photoelectron waveforms on neighboring pixels relative to the source. Additionally, they injected electrical sine-wave pulses into the readout board to characterize the device's behavior as a resonant cavity.
• Dark Count and High-Voltage Stability: The dark count rate was monitored over 300 minutes while systematically adjusting high-voltage settings on the top MCP. Current instabilities were tracked using a digital ammeter.
• FastFrame Acquisition: To capture rare events and after-pulses, the oscilloscope was operated in FastFrame mode, collecting over 10,000 waveforms per channel to identify delayed signals (e.g., at 60 ns and 110 ns).
• Simulation: A 2D Python-based Monte Carlo simulation was developed to model electron trajectories, secondary emission, back-scattering, and ion feedback. The model utilized Gaussian distributions for initial photoelectron positions and energies, solving for time-of-flight and radial displacement on the anode.

Key Results

1. Charge Sharing and Cross-Talk:

   • Significant charge sharing was observed on perpendicular direct-neighboring pixels, while diagonal/off-diagonal pixels showed less sharing.
   • Electronic cross-talk fell below 1% beyond a one-pixel neighbor distance.
   • The cross-talk signal on diagonal pixels (e.g., F6 from E5) exhibited an underdamped harmonic oscillator shape.
   • The authors note that while cross-talk is often undesirable, in LAPPDs it can be advantageous for event reconstruction, as the specific cross-talk response pattern varies with the photon's precise impact location.

2. Resonant Cavity Behavior:

   • When electrical pulses were injected, the LAPPD demonstrated frequency-dependent cross-talk amplitude.
   • Resonant peaks were identified at 180 MHz and 550 MHz for all pixels, with specific offsets observed in the 350–450 MHz and 750–800 MHz ranges. This suggests susceptibility to radio-frequency (RF) interference.

3. Muon Detection:

   • The team observed large, spatially distributed events at a rate of ~1–2 Hz, consistent with cosmic-ray muon interactions.
   • These events displayed a large initial peak (700–800 mV) followed by delayed peaks, attributed to charge deposition within the MCP geometry and electron backscattering.

4. Dark Count Rate and Voltage Stability:

   • Increasing the entry and exit voltages of the top MCP caused an instantaneous spike in the dark count rate (from ~200 Hz to ~28 kHz) and a current spike.
   • The recovery followed a tri-exponential decay with fast (0.80 min), intermediate (7.79 min), and slow (78.7 min) relaxation timescales.
   • No such spike occurred when adjusting the bottom MCP voltages, indicating the instability is specific to the top MCP configuration. The authors hypothesize this is driven by thermal emissions or field emission rather than purely electronic transients.

5. Pulse Classification and After-Pulses:

   • FastFrame analysis revealed distinct clusters of delayed signals at approximately 60 ns and 110 ns.
   • A pulse classification scheme categorized events based on timing windows. "Class 1" (primary pulse) dominated, while coincident multi-window events were rare.
   • Delayed peaks were more frequent in B pads than A pads, suggesting potential cross-talk influence from the A side.
   • The amplitude of delayed peaks did not scale strongly with voltage, suggesting that if ion feedback is the cause, the process is limited by factors other than simple ionization probability.

6. Simulation Findings:

   • The Monte Carlo simulation successfully reproduced the Single Photoelectron (SPE) pulse arrival (~4 ns) and the radial spread of the electron cloud (biased slightly left due to chevron channel geometry).
   • The model identified anode and MCP back-scattering as sources of signal tails and minor secondary peaks.
   • Simulated ion feedback generated after-pulses in the 15–30 ns range, though the authors acknowledge that positive ion feedback alone may not fully explain the 60 ns and 110 ns peaks observed experimentally.

Significance and Claims
The paper claims to provide a comprehensive operational model for LAPPD Gen 2, filling gaps left by earlier, simpler MCP models. The authors emphasize that:

• Intrinsic Cross-Talk: Cross-talk is an inherent feature of the capacitive coupling design that can be leveraged for improved position reconstruction rather than solely treated as noise.
• Voltage Trade-offs: Optimizing for a single metric (e.g., minimizing dark count rates) often degrades other characteristics like gain or timing resolution. There is no universally optimal operating point; trade-offs must be defined by the specific application (e.g., scintillation vs. Cherenkov detection).
• Background Mitigation: Understanding the mechanisms behind dark count spikes and muon-induced signals is essential for calibrating data acquisition systems to filter unwanted background in low-light, sparse-photon environments typical of neutrino experiments.
• Future Directions: The authors suggest that FPGA-resident machine learning could be utilized for real-time feature extraction to handle the complex pulse shapes and after-pulse characteristics identified in this study.

The study concludes that while the simulation aligns reasonably well with experimental pulse classifications, the exact physical mechanisms driving the specific 60 ns and 110 ns delayed peaks remain an area for further investigation.