Low-Energy Nuclear Recoil Calibration of XENONnT with a $^{88}$YBe Photoneutron Source

The XENONnT experiment successfully utilized a $^{88}$YBe photoneutron source to calibrate low-energy nuclear recoil light and charge yields in liquid xenon, providing essential data for solar neutrino measurements and searches for light dark matter particles.

The Gist

Imagine you are trying to listen for a tiny, specific whisper in a very loud, noisy room. That is essentially what scientists do when they search for Dark Matter. They use massive tanks of liquid xenon (a heavy, invisible gas turned into a liquid) to catch these "whispers," which are actually tiny particles bumping into the xenon atoms.

However, there's a problem: the "whispers" they are looking for are so quiet that they are right at the edge of what their equipment can hear. To make sure their "ears" (detectors) are working correctly at these very low volumes, they need to practice with a known sound.

This paper describes how the XENONnT team built a special "practice sound" to calibrate their detector. Here is how they did it, broken down into simple steps:

1. The Problem: Listening for the Quietest Whispers

The scientists are looking for two very faint things:

• Dark Matter: A mysterious substance that makes up most of the universe but rarely interacts with normal matter.
• Solar Neutrinos: Tiny particles from the Sun that bounce off xenon atoms.

Both of these create a very small "kick" (called a nuclear recoil) in the xenon atoms. The problem is that these kicks are so weak they are right at the bottom limit of what the detector can see. If the detector isn't perfectly calibrated, they might miss these signals or mistake noise for a signal.

2. The Solution: A "Neutron Flashlight"

To test the detector, they needed something that would create a kick similar to Dark Matter or Solar Neutrinos, but one they could control. They used a special source called 88YBe.

• How it works: Think of this source as a machine that shoots tiny, slow-moving balls (neutrons) at the xenon.
• The Trick: They used a radioactive element (Yttrium) to shoot high-energy light rays (gamma rays) at a block of Beryllium. When the light rays hit the Beryllium, they knock loose a neutron.
• The Result: These neutrons hit the xenon atoms and give them a tiny "kick," creating a signal the detector can see. This is like using a known, gentle tap to test if a microphone is sensitive enough to hear a whisper.

3. Building the "Shielded Box"

The scientists faced a few engineering headaches:

• Too much noise: The source also shoots out a lot of light rays (gamma rays) that are much louder than the neutron kicks. If these hit the detector, they would drown out the signal.
• The Fix: They built a heavy box made of Tungsten (a very dense metal, heavier than lead) to block the loud light rays while letting the tiny neutrons pass through.
• The Air Gap: They also had to build a special air-filled box to push the water out of the way between the source and the detector. If water were there, it would slow down the neutrons too much, changing the "kick" they wanted to measure.

4. The "Noise" in the Room

Even with the shield, there was a lot of background noise.

• The "Accidental" Problem: The detector is so sensitive that sometimes it sees two unrelated things happening at the same time and thinks they are one event. For example, a stray electron might drift up and hit a random flash of light, and the computer thinks, "Aha! A particle hit!"
• The Solution: The team used a computer program (a type of Artificial Intelligence called a Boosted Decision Tree) to learn the difference between a real "kick" and these accidental mix-ups. It's like a bouncer at a club who learns to spot the difference between a real guest and someone just trying to sneak in by looking at their ID and behavior.

5. The Results: Tuning the Microphone

After running the source for about 183 hours, they collected data on 474 valid events (after filtering out the noise).

• What they found: They successfully mapped out exactly how much light and electrical charge the xenon produces when hit by these tiny kicks, even at energies as low as 0.3 keV (which is incredibly small).
• The Comparison: They compared their new measurements to a standard computer model (called NEST) that scientists usually use to predict these things. Their new data matched the model very well.

Why This Matters

Think of this calibration as tuning a musical instrument before a concert.

• Before this, the scientists weren't 100% sure how their "instrument" (the detector) sounded at the very lowest notes.
• Now, they have a precise map of how the detector responds to these tiny kicks.
• This allows them to confidently say, "If we see a signal this small, it is real," which is crucial for finding Dark Matter or measuring those faint solar neutrinos.

In short, the team built a special, shielded neutron generator, used AI to filter out the noise, and successfully "tuned" their giant xenon detector to hear the faintest whispers in the universe.

Technical Summary

1. Problem Statement

Direct detection of dark matter (WIMPs) and solar neutrinos via Coherent Elastic Neutrino-Nucleus Scattering (CE$\nu$NS) requires precise characterization of detector response to low-energy nuclear recoils (NR), specifically in the sub-keV to few-keV range.

• The Challenge: As experiments like XENONnT approach the "neutrino fog," distinguishing rare low-energy signals from background becomes critical. Existing calibration methods often lack the sensitivity to probe the lowest energy thresholds or suffer from high background rates when using photoneutron sources due to intense accompanying gamma radiation.
• Specific Obstacles:
  • Accidental Coincidences (AC): The high activity of gamma-emitting sources creates a high rate of "accidental" pairings between isolated S1 signals (prompt light) and delayed electrons (ionization), mimicking NR events.
  • Energy Threshold: Achieving calibration down to the detector's physical threshold is difficult due to low photon/electron yields at these energies.
  • Source Limitations: Traditional neutron sources (e.g., D-D guns) are complex to operate underground; photoneutron sources are compact but produce overwhelming gamma backgrounds.

2. Methodology

A. Source Design and Geometry

The collaboration utilized an $^{88}$YBe photoneutron source, which generates quasi-monoenergetic 152 keV neutrons via the photodisintegration of $^9$Be by 1.84 MeV $\gamma$-rays from $^{88}$Y decay.

• Shielding Optimization: To mitigate the high $\gamma$-ray flux while preserving neutron flux, the source was encased in a tungsten shield (chosen over lead for higher density and lower radiation length within the size constraints of the I-belt system).
• Geometry: The source assembly (55.5 mm $\times$ 29.4 mm) was optimized using Geant4 simulations to maximize the Signal-to-Background ratio (SS NR events vs. ER background). The final design placed the source ~10 cm from the cryostat wall, with an air-filled stainless steel box displacing water to prevent neutron energy broadening.
• Control Run: A "dummy" source using PVC (same geometry, no Beryllium) was used to measure the pure accidental coincidence background rate.

B. Data Acquisition

• Exposure: The source was lowered into the XENONnT neutron veto via the I-belt system.
  • $^{88}$YBe Run: 183 hours of live time (Sept 30 – Oct 9, 2022).
  • $^{88}$Y-PVC Run: 152 hours of live time (Oct 10 onwards) for background modeling.
• Drift Field: Operated at 23 V/cm, matching the standard science run conditions.

C. Event Selection and Background Rejection

The analysis focused on selecting NR events while rejecting the dominant AC background:

1. Multi-Scatter (MS) Inclusion: Unlike previous calibrations that required single-scatter events, this analysis included Multi-Scatter (MS) NR events. Neutrons scatter multiple times in liquid xenon; including MS events significantly lowers the energy threshold by accumulating S1 photons from multiple scatters.
2. Fiducial Volume: An optimized volume was defined to exclude regions near the TPC walls (where charge loss occurs) and specific wire regions, reducing AC events by ~90%.
3. Machine Learning (BDT): A Boosted Decision Tree (BDT) classifier was trained to distinguish between NR and AC events.
   • Signal: Simulated $^{88}$YBe NR events.
   • Background: A data-driven AC model ("Axidence") constructed by resampling isolated S1 and S2 signals from the data.
   • Performance: The BDT cut retained 89% of NR events while rejecting 91% of AC events.
4. Final Sample: After all cuts, 474 events were selected from the $^{88}$YBe data.

D. Yield Extraction

The selected events were fitted using the Appletree framework and the NEST v2 (Noble Element Simulation Technique) model.

• Fit Parameters: The model varied parameters controlling light yield ($\theta, \iota$) and charge yield ($\epsilon, \zeta, \eta$), as well as statistical fluctuations ($F_{ex}, F_i$) and recombination variance ($\xi, \omega$).
• Efficiency Corrections: The analysis accounted for S1 reconstruction efficiency (lower for MS events), event building efficiency (affected by delayed electrons), and event selection efficiency (spatially dependent).
• Background Subtraction: The expected AC background was estimated at 55 $\pm$ 12 events using the data-driven model, which was subtracted during the fit.

3. Key Results

• Lowest Energy Thresholds:
  • Light Yield: Sensitive down to 0.30 $\pm$ 0.02 keV$_{NR}$.
  • Charge Yield: Sensitive down to 0.66 $\pm$ 0.04 keV$_{NR}$.
  • Note: The lower threshold for light yield is attributed to the coherent addition of photons from multiple scatters in MS events, whereas the charge signal is dominated by fewer scatters.
• Yield Measurements:
  • The extracted light and charge yields are consistent with the NEST v2.3.11 model within $\pm$1$\sigma$ uncertainties.
  • The light yield uncertainty is larger than the charge yield uncertainty due to lower light collection efficiency (enhancing statistical fluctuations) and additional free parameters in the NEST light yield model.
• Background Characterization:
  • The data-driven AC model successfully predicted the background rate with ~20% agreement, validating the robustness of the background subtraction.
  • The final dataset contains 474 events, with an estimated 55 $\pm$ 12 residual AC events.

4. Significance and Impact

• Solar Neutrino Physics: This calibration is critical for the measurement of solar $^8$B neutrinos via CE$\nu$NS. The XENONnT experiment recently reported the first indication of these signals; precise low-energy yield calibration is essential to reduce systematic uncertainties in this measurement.
• Light Dark Matter Search: The results enable robust searches for Light WIMPs (masses < 12 GeV/$c^2$). The ability to calibrate down to 0.3 keV$_{NR}$ allows the experiment to probe the "neutrino fog" region where dark matter signals overlap with neutrino backgrounds.
• Methodological Advancement:
  • Demonstrates the viability of using photoneutron sources for low-energy calibration in large-scale TPCs by effectively managing high gamma backgrounds through geometry optimization and machine learning.
  • Establishes a new benchmark for Multi-Scatter event analysis, proving that including MS events significantly lowers the energy threshold compared to single-scatter-only analyses.
  • Validates the data-driven AC modeling technique, which is crucial for future experiments operating in high-background environments.

In conclusion, this work provides the first comprehensive low-energy nuclear recoil calibration for XENONnT, extending the detector's sensitivity to the sub-keV regime and solidifying its capability to explore the frontiers of solar neutrino physics and light dark matter.