Pushing the Limits of Pulse Shape Discrimination in a Large Liquid Xenon Detector

This paper presents an optimized analysis framework for Pulse Shape Discrimination (PSD) in the LUX-ZEPLIN (LZ) experiment that, when combined with charge-to-light measurements into a two-factor discrimination method, significantly reduces electronic recoil background leakage and false positive rates in the search for dark matter.

The Gist

Imagine you are trying to find a single, specific whisper in a room filled with thousands of people talking, coughing, and shuffling their feet. That is essentially what the LUX-ZEPLIN (LZ) experiment is trying to do.

The "whisper" is a WIMP (Weakly Interacting Massive Particle), a hypothetical ghost-like particle that makes up Dark Matter. The "noise" is background radiation from rocks, pipes, and even the air around the detector.

This paper is about a new, super-sharp ear that the LZ team has built to help them hear that whisper more clearly. They call it Pulse Shape Discrimination (PSD).

Here is the story of how they did it, explained simply:

1. The Setting: A Giant Fish Tank in a Mine

The LZ detector is a massive tank filled with 7 tons of liquid xenon, sitting 4,850 feet underground in a South Dakota mine. It's shielded by water and rock to block out cosmic rays.

When a particle hits the xenon, it creates a tiny flash of light (like a firefly blinking) and a tiny electrical charge.

• The Good Signal (WIMP): If a WIMP hits a xenon nucleus (the heavy center of the atom), it creates a specific type of flash.
• The Bad Noise (Background): If a gamma ray or electron (common background noise) hits a xenon electron, it creates a slightly different flash.

2. The Old Way: Measuring Brightness vs. Charge

For a long time, scientists used a method called Charge-to-Light discrimination.

• The Analogy: Imagine you are judging a race. You look at two runners. One is tall and thin (the WIMP), and one is short and wide (the background noise). You can tell them apart by their height-to-width ratio.
• The Problem: Sometimes, the background noise tries to look like the WIMP. It gets "tall and thin" enough to sneak past the guard. The team needed a better way to catch these impostors.

3. The New Trick: Listening to the "Shape" of the Flash

This is where Pulse Shape Discrimination (PSD) comes in.

Instead of just measuring how much light there is, the team started measuring how the light behaves over time.

• The Analogy: Imagine two people clapping their hands.
  • Person A (The WIMP/Nuclear Recoil): Claps once, sharp and fast. Clap! It's over quickly.
  • Person B (The Background/Electron Recoil): Claps, but their hand lingers a bit, or they do a little echo. Clap... (fizzle)...

In liquid xenon, the "WIMP" flash is slightly faster and sharper than the "Background" flash. The difference is tiny—like the difference between a millisecond and a few nanoseconds—but it's there.

4. The Challenge: The "Echo" in the Room

The problem is that the detector is huge. Light bounces off the walls, the liquid surface, and the glass. It's like shouting in a cathedral; the sound bounces around, making it hard to tell exactly when the original shout happened.

To fix this, the team built a super-precise timing system:

• The N-Photon Model: They developed a mathematical trick to look at the raw data and say, "Okay, that big blip on the screen wasn't one big flash; it was actually three tiny photons arriving at slightly different times."
• The Map: They mapped out exactly how long light takes to travel from the bottom of the tank to the top, correcting for every inch of depth. This is like knowing exactly how long it takes for a sound to echo in every corner of the cathedral so you can ignore the echo and hear the original voice.

5. The Result: A Two-Factor Security Check

Once they mastered the "shape" of the flash, they combined it with the old "size" method (Charge-to-Light) to create a Two-Factor Discriminator (TFD).

• The Analogy: Think of a high-security bank vault.
  • Old Method: You just showed your ID card (Charge-to-Light). Sometimes, a clever forger could make a fake ID that looked good enough.
  • New Method (TFD): Now, you have to show your ID AND you have to solve a quick math puzzle (Pulse Shape). Even if a forger has a fake ID, they probably can't solve the puzzle.

6. Why This Matters

The paper shows that this new method is incredibly effective:

• Catching the Impostors: It reduced the number of background "noise" events that look like WIMPs by about half for larger signals.
• The "Double Electron Capture" Problem: There is a specific type of radioactive decay (from Xenon-124) that is very tricky because it mimics WIMPs perfectly in the old system. The new PSD method acts like a specialized filter that catches these specific impostors, which the old system missed.
• The Future: This proves that even in a giant detector like LZ, we can listen to the "shape" of the light to find the universe's darkest secrets.

Summary

The LZ team didn't just build a bigger net; they built a smarter net. By listening to the rhythm and shape of the light flashes, rather than just counting them, they can now tell the difference between a real Dark Matter signal and background noise with much higher confidence. It's like going from trying to hear a whisper in a noisy room by just turning up the volume, to finally putting on noise-canceling headphones that know exactly what the whisper sounds like.

Technical Summary

1. Problem Statement

The primary challenge in direct dark matter detection experiments, such as LUX-ZEPLIN (LZ), is distinguishing Weakly Interacting Massive Particle (WIMP) signals from background events.

• Background Sources: The dominant backgrounds are Electronic Recoils (ER) caused by gamma and beta radiation from residual radioactivity in detector materials. WIMPs are expected to produce Nuclear Recoils (NR).
• Current Limitations: The standard discrimination method in Liquid Xenon (LXe) is Charge-to-Light (CTL) discrimination (the ratio of ionization S2 to scintillation S1). While effective, further improvements are needed to enhance signal acceptance and lower background leakage, particularly for specific backgrounds like $^{124}$Xe double electron capture (DEC), which has a suppressed charge yield and can leak into the NR band.
• The Challenge in LXe: Pulse Shape Discrimination (PSD) is highly effective in Liquid Argon due to large differences in singlet and triplet state lifetimes. In LXe, the lifetime difference is small ($\tau_1 \approx 2\text{--}4$ ns vs. $\tau_3 \approx 21\text{--}28$ ns), making PSD significantly more difficult and requiring high-precision photon timing reconstruction.

2. Methodology

The paper outlines a comprehensive framework to optimize PSD in the LZ detector using calibration data and NEST (Noble Element Simulation Technique) simulations.

A. Photon Timing Reconstruction

To overcome the small lifetime differences in LXe, the authors developed a high-precision analysis framework:

• Relative Timing Calibration: Channel-to-channel timing offsets between the 494 PMTs were corrected using LED calibration data, achieving an accuracy better than 1 ns.
• N-Photon Model: A statistical model was developed to reconstruct individual photon arrival times from overlapping waveforms. It fits waveforms as a sum of $N$ Single Photoelectron (SPE) templates.
  • The model uses Bayes' Theorem to determine the most likely number of photons ($N$) in a waveform (typically $N \le 3$).
  • It accounts for Double Photoelectron (DPE) effects and rejects cases where two reconstructed photons are separated by less than 1 ns to avoid bias.

B. Discriminator Optimization

• Prompt Fraction (PF): The discriminator is defined as the ratio of photons detected in a prompt time window ($t_0$ to $t_1$) to the total photons in the pulse.
  • $PF = \frac{\sum_{t_0}^{t_1} S1(t)}{\sum_{t_2}^{t_3} S1(t)}$
• Optimization: Using CH$_3$T (ER) and DD (NR) calibration data, the parameters $t_0$ and $t_1$ were optimized using a raster scan followed by the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) algorithm. The goal was to minimize ER leakage while maintaining 50% NR acceptance.
• Position Correction ($z$-correction): Photon transit times vary with the interaction height ($z$) due to light propagation paths. A linear correction $\Delta PF(z)$ was derived from CH$_3$T data across the fiducial volume to ensure uniform discrimination performance.

C. Two-Factor Discrimination (TFD)

To maximize rejection power, PSD was combined with the standard CTL method into a Two-Factor Discriminator (TFD):

• Formula: $TFD = -1 \times CTL + w(S1_c) \times PF_{cor}$
• Weighting: The weight $w(S1_c)$ was optimized as a linear function of the corrected S1 charge ($S1_c$) to maximize the separation (Z-score) between ER and NR distributions.

D. Simulation and Validation

• NEST Tuning: The NEST simulation framework was tuned to match LZ calibration data, specifically adjusting the recombination time constant ($\tau_r$) and photon transit models.
• Background Study: The tuned NEST model was used to simulate $^{124}$Xe DEC events to evaluate PSD performance on this specific background.

3. Key Results

A. PSD Performance

• ER Leakage: For S1 pulses between 20 and 80 photoelectrons (phd), the optimized PSD achieves an ER leakage of 15% to 35% at 50% NR acceptance.
• Dependence on Signal Size: Leakage decreases as the number of photons increases. For the 70–80 phd bin, ER leakage is 14.4 ± 1.0%.
• NEST Comparison: NEST simulations predict slightly lower leakage than data, confirming that signal processing uncertainties (not included in NEST) contribute to the observed leakage.

B. Specific Background Rejection ($^{124}$Xe DEC)

• For $^{124}$Xe double electron capture decays, which are difficult to reject with CTL alone due to low charge yield:
  • Below 70 phd, PSD offers limited improvement over CTL.
  • Above 95 phd, PSD outperforms CTL, providing a crucial handle for suppressing this background in future high-sensitivity runs.

C. Two-Factor Discrimination (TFD)

• Combining CTL and PSD significantly improves separation.
• False Positive Rate Reduction: The TFD method reduces the false positive rate (ER leakage) by up to a factor of two compared to CTL alone, particularly for large S1 pulses (70–80 phd).
• Z-score Improvement: The separation between ER and NR medians (normalized by ER width) increased from ~3.65$\sigma$ (CTL) to ~3.83$\sigma$ (TFD) in the 70–80 phd bin.

D. Application to WS2024 Data

• Applied to the WIMP Search (WS2024) dataset (329 radon-tagged events):
  • PSD: All 4 events that fell within the CTL NR band were classified as ER events by PSD (falling outside the PSD NR acceptance region).
  • TFD: Using TFD, the number of events in the NR band was reduced from 4 (using CTL) to 2.
  • These results are consistent with LZ's null result for WIMP detection.

4. Significance and Impact

• Feasibility in Large Detectors: This work demonstrates that PSD is a viable and powerful tool even in large-scale LXe detectors where photon statistics and timing resolution are challenging.
• Complementary Tool: PSD provides an independent discrimination dimension that is orthogonal to CTL, offering robustness against systematic uncertainties in charge yield modeling.
• Future Detectors: The success of PSD in LZ suggests it will be a critical component for next-generation detectors (e.g., LZ upgrades or future multi-tonne experiments), especially under lower drift-field conditions where recombination effects (and thus timing differences) are enhanced.
• Background Suppression: The ability to suppress specific backgrounds like $^{124}$Xe DEC using PSD is vital for pushing the sensitivity limits of dark matter searches to the "neutrino fog."

In conclusion, the paper establishes a rigorous framework for photon timing and pulse shape analysis in LZ, proving that combining PSD with traditional charge-to-light discrimination significantly enhances background rejection capabilities, thereby pushing the sensitivity limits of liquid xenon dark matter experiments.