Surface background of the BULLKID detector array operated with moderate shielding

The BULLKID detector array, operated with moderate shielding in a surface laboratory, demonstrated background levels consistent with simulations down to 600 eV over a 290-hour exposure, while revealing an unexplained rise in the 225–600 eV range that differs from low-energy excesses observed in other cryogenic experiments.

The Gist

Imagine you are trying to hear a single, tiny whisper in a very loud, noisy room. That is essentially what the scientists behind the BULLKID experiment are trying to do. They are building a super-sensitive "ear" to listen for the faintest possible signals from the universe: tiny particles of Dark Matter or neutrinos bouncing off atoms.

Here is a breakdown of their latest experiment, explained simply:

1. The Detective and the Room

The "detective" is a device called BULLKID. It's a flat silicon wafer (like a giant computer chip) cut into 60 tiny little cubes (dice). Each cube is a sensor. When a particle hits one of these cubes, it creates a tiny vibration (a sound wave in the solid material), which the sensor detects.

The "room" is a laboratory on the surface of the Earth (at a university in Rome). This is a problem because the surface is full of background noise—radiation from the ground, cosmic rays from space, and even the natural radioactivity in the walls. To make the room quieter, the scientists built a fortress around their detector:

• The Inner Shield: A thick lead pot (like a heavy metal bucket) that holds the detector.
• The Outer Shield: A massive castle made of lead bricks surrounding the entire machine, weighing about 170 kg (roughly the weight of a small car).

2. The "Self-Veto" Trick

Here is the clever part. The detector isn't just one block; it's an array of 60 cubes.

• If a particle hits the center cube, the sound wave stays mostly there.
• If a particle hits a neighbor cube, the sound wave "leaks" over to the center cube, but it sounds different.

The scientists use the center cubes as the "main ears" and the surrounding cubes as "veto guards." If the main ear hears a sound and the neighbor ears hear a matching sound at the same time, they know it was just a leaky neighbor and ignore it. They only count the "pure" whispers that happen right in the center.

3. The Results: A Flat Line with a Bump

The team ran the experiment for about 12 days (290 hours) and listened to the background noise.

• The Good News: For most of the energy range (from 2 keV down to 600 eV), the noise level was exactly what they predicted with their computer simulations. It was a nice, flat line. This proves their "fortress" works and their "self-veto" trick is effective. They successfully reduced the background noise by a factor of 29 compared to when they had no shields at all.
• The Bad News (The Mystery): When they looked at the very lowest energies (between 225 eV and 600 eV), the noise didn't stay flat. Instead, it started to rise sharply, like a hill getting steeper as you go down. The count rate jumped up to 7 times higher than expected right at the bottom.

4. Is it the "Low Energy Excess"?

Other experiments in the field have seen a similar "hill" of noise at low energies, which they call the "Low Energy Excess" (LEE). Some scientists think this might be a new type of particle or a glitch in the detector.

The BULLKID team investigated if they had the same problem:

• Is it a glitch? They changed the settings and the number of sensors. The "hill" stayed the same.
• Is it a time-based decay? Other experiments saw this noise fade away over days or weeks. The BULLKID team watched for a month, and the noise did not fade. It stayed constant.

Conclusion: The "hill" they found looks like the Low Energy Excess, but it behaves differently. It's likely a different kind of mystery, not the same one other teams are seeing.

5. What's Next?

The scientists are not panicking; they are just curious.

• They suspect the "hill" might be caused by something they haven't simulated yet, perhaps related to how gamma rays interact with the lead shield.
• Their plan is to build an even stronger shield and eventually move the detector underground (to the Gran Sasso National Laboratory in Italy). Being deep underground is like moving the detective into a soundproof basement, which should silence the background noise enough to see if this mysterious "hill" disappears or if it's truly a new signal.

In short: The BULLKID detector is working beautifully as a shielded, self-checking system. It successfully filtered out most of the universe's noise, but it found one small, unexplained "bump" in the quietest part of the spectrum that needs more investigation.

Technical Summary

1. Problem Statement

The direct detection of light Dark Matter (WIMPs with mass $\lesssim$ GeV/c$^2$) and Coherent Elastic Neutrino–Nucleus Scattering (CE$\nu$NS) requires cryogenic detectors with extremely low energy thresholds (tens to hundreds of eV). A major challenge in this field is the "Low Energy Excess" (LEE), an unexplained rise in background events near the energy threshold observed in various cryogenic experiments (e.g., CRESST, SuperCDMS). This excess complicates the search for rare signals.

The BULLKID-DM collaboration aims to address this by developing a detector with a fully active target (no inert materials) capable of self-vetoing background events. While a previous prototype demonstrated a flat background down to 160 eV without shielding, it operated in an unshielded environment. To validate its performance against the current experimental landscape, the detector needed to be tested under moderate shielding to achieve background levels comparable to or lower than existing experiments, while investigating the origin of the LEE.

2. Methodology

Experimental Setup:

• Detector: The BULLKID prototype consists of a 3-inch silicon wafer carved into 60 cubic dice (0.34 g each), totaling ~20 g of instrumented mass. Each die is read out by a multiplexed Kinetic Inductance Detector (KID).
• Shielding: The experiment was conducted on the surface at Sapienza University of Rome with a two-tier shielding system:
  • Inner Shield: A lead pot (4.4 mm top/bottom, 7.0 mm sides) providing 4$\pi$ coverage, limited to ~2 kg due to cryostat payload constraints.
  • Outer Shield: A "castle" of lead bricks (~170 kg) surrounding the cryostat (25 mm thickness).
• Cryogenics: The system was cooled to ~20 mK using a dry $^3$He/$^4$He dilution refrigerator.
• Readout: Microwave resonators were excited and read by an Ettus X300 board. Data was acquired via a matched filter with a threshold of 4.5$\sigma$.

Data Acquisition Strategy:

• Fiducialization: The analysis focused on 15 KIDs (2 central "main" dice and their neighbors). Events were triggered by the main dice (KID-47 or KID-49), while surrounding dice acted as a veto to reject events originating outside the target dice (e.g., phonon leakage from neighbors).
• Exposure: The detector operated for 290 live hours.
• Calibration: Energy calibration was performed using:
  • A 400 nm LED system (room temperature) for low-energy response.
  • Natural lead-induced X-rays and $\gamma$-rays (from $^{210}$Pb contamination in the shield) for high-energy cross-calibration.

Simulation:

• Backgrounds were modeled using Geant4 (Shielding physics list).
• Sources included: Environmental $\gamma$-rays, fast neutrons, cosmic muons, and $^{210}$Pb contamination in the inner lead shield.
• The simulation incorporated the specific geometry of the lead castle, cryostat, and building structure to model muon flux.

3. Key Contributions

• First Surface Operation with Shielding: This is the first demonstration of the BULLKID array operating with moderate shielding, reducing the background by a factor of 29 compared to the unshielded configuration.
• Advanced Event Discrimination: The paper details an improved analysis algorithm using cluster event selection. By analyzing phonon leakage patterns across the KID array and applying a multidimensional cut (variable $\Omega$), the team achieved high rejection of background events while maintaining >45% efficiency for signal-like events near the threshold.
• Characterization of the Low Energy Rise: The study rigorously investigates the unexplained rise in counts below 600 eV, comparing it to the "Low Energy Excess" (LEE) seen in other experiments.

4. Results

High Energy Spectrum (>2 keV):

• The measured spectrum up to 85 keV is consistent with Geant4 simulations.
• The spectrum displays characteristic lead-induced X-ray and $\gamma$-ray peaks (e.g., 46.6 keV from $^{210}$Pb).
• Discrepancy: The simulation overestimates the background by up to 30% in the high-energy region. The authors attribute this likely to systematic uncertainties in the assumed external gamma flux or geometric offsets in the shield.

Low Energy Spectrum (<2 keV):

• Flat Region: Between 600 eV and 2 keV, the background is flat and compatible with simulations.
  • Measured rate: $(6.8 \pm 0.4 \text{ stat.} \pm 0.1 \text{ syst.}) \times 10^4$ counts / keV kg days.
  • This is consistent with the Monte Carlo prediction of $(8.2 \pm 0.3) \times 10^4$ DRU.
• Analysis Threshold: The analysis threshold was set at 225 eV (approx. 6$\times$ the baseline resolution).
• The Anomaly: Between 225 eV and 600 eV, the measured spectrum shows a progressive rise in count rate, reaching $5 \times 10^5$ DRU at the threshold (a factor of ~7 higher than the flat region).
  • This rise is not explained by noise false positives (verified via negative triggers).
  • It is not present in the Geant4 simulation.

Investigation of the Low Energy Rise:
The authors tested if this rise was the same phenomenon as the LEE reported by other experiments:

1. Operating Conditions: Changing the number of active sensors (15 to 18) and bias power did not alter the rise, ruling out simple SNR or readout artifacts.
2. Time Decay: Unlike the LEE in other experiments (which often decays over days/weeks), the BULLKID rise showed no time-dependent decay over the measurement period.
3. Comparison: When compared to other experiments (CRESST, SuperCDMS, NUCLEUS), the BULLKID rise is smaller in magnitude relative to the background above 1 keV and does not follow the same mass-scaling trends.

5. Significance and Conclusion

• Validation of Technology: The BULLKID detector successfully demonstrated its ability to operate with moderate shielding, achieving a background level of $\sim 7 \times 10^4$ DRU in the 0.6–2 keV range, which is competitive with state-of-the-art surface experiments.
• Nature of the Excess: The unexplained rise below 600 eV appears to be of a different nature than the LEE observed in other cryogenic experiments, given the lack of time decay and different spectral shape.
• Future Outlook:
  • The collaboration plans to extend simulations to investigate a potential particle origin for the rise.
  • Future runs will include increased shielding above ground and a move to an underground laboratory (Gran Sasso) to target background levels of $\sim 10^4$ DRU or lower. This will determine if the discrepancy persists at lower background levels and help isolate the source of the low-energy rise.

In summary, the BULLKID prototype has proven its viability for light Dark Matter searches, successfully mitigating background to near-simulation levels in the keV range, while identifying a distinct low-energy anomaly that requires further investigation in deeper underground environments.