Construction and characterisation of the DarkSide-20k veto silicon photo-multiplier tiles

This paper details the successful production, cryogenic testing, and high yield (exceeding 87%) of the silicon photomultiplier-based "Veto Tiles" designed to instrument the Inner Veto volume of the DarkSide-20k experiment for discriminating background radiation in direct dark matter searches.

The Gist

Imagine you are trying to hear a single, tiny whisper in a stadium full of screaming fans. That is essentially what the DarkSide-20k experiment is trying to do. They are looking for Dark Matter, the invisible stuff that makes up most of the universe. The "whisper" is a dark matter particle bumping into an atom in a tank of liquid argon. The "screaming fans" are background noise from cosmic rays and natural radioactivity.

To solve this, the scientists built a massive, ultra-sensitive "ear" (a detector) and surrounded it with a "noise-canceling helmet." This paper is the blueprint and quality report for the silicon sensors inside that helmet.

Here is the story of how they built and tested these sensors, explained simply:

1. The Mission: Catching the Ghost

The DarkSide-20k detector is a giant tank of liquid argon (frozen gas) buried deep underground.

• The Goal: When a dark matter particle hits an argon atom, it creates a tiny flash of light.
• The Problem: Neutrons and cosmic rays also create flashes of light that look exactly like dark matter. If you don't stop them, you'll think you found dark matter when you actually just found a neutron.
• The Solution: They built an "Inner Veto" (a shield) around the main tank. If a neutron tries to sneak in, the shield catches it first and yells, "Ignore this!" The main detector then deletes that event from its records.

2. The Sensors: The "Silicon Eyes"

To see these tiny flashes of light, they needed special eyes. They chose Silicon Photo-Multipliers (SiPMs).

• The Analogy: Think of a traditional light bulb (photomultiplier tube) as a giant, fragile glass vacuum tube. It's big and can be slightly radioactive itself. The SiPMs are like tiny, super-sensitive solar panels made of silicon. They are small, robust, and can detect a single photon (a particle of light) even when frozen in liquid nitrogen.
• The "Veto Tile" (vTile): You can't just stick one sensor on a wall; you need a whole array. They took 24 of these tiny silicon eyes and glued them onto a small circuit board (about the size of a post-it note). They call this a "vTile."
• The "Veto Unit" (vPDU): To cover the whole shield, they took 16 of these vTiles and mounted them onto a larger motherboard. This creates a "Veto Photo-Detector Unit" (vPDU). They needed 120 of these units to cover the entire detector.

3. The Factory: Building a Million Tiny Connections

Building these tiles was like assembling a high-tech LEGO set, but with microscopic precision.

• The Process:
  1. The Board: They started with a blank circuit board.
  2. The Chips: They soldered tiny electronic chips (ASICs) onto the back.
  3. The Eyes: They carefully picked up the 24 silicon sensors and glued them onto the front.
  4. The Wires: This was the hardest part. They had to connect each sensor to the board using gold wires thinner than a human hair. They used a machine to "bond" these wires. If a wire was too loose, the sensor wouldn't work. If it was too tight, it would snap.
• The "Triple-Bagging": Because these sensors are so sensitive to dust and radon gas (which creates background noise), they couldn't just put them in a box. They had to be wrapped in three layers of special vacuum bags with desiccants (like the little "do not eat" packets in shoe boxes) to keep them perfectly clean and dry during transport.

4. The Stress Test: The "Hot and Cold" Exam

You can't just build these sensors and hope they work. They have to survive the extreme conditions of the experiment.

• The Warm Test: First, they tested the tiles at room temperature to make sure the electronics worked.
• The Cold Test: Then, they dunked them in liquid nitrogen (which is -196°C / -320°F). This simulates the environment inside the detector.
  • Why? Electronics behave differently when frozen. Some might crack; others might get too noisy.
  • The Result: They checked if the sensors could still "see" single photons in the cold. They measured the Signal-to-Noise Ratio (SNR). Imagine trying to hear a whisper in a quiet room (good SNR) vs. a loud party (bad SNR). They needed the sensors to be in a very quiet room.
• The Fix: If a tile failed, they didn't just throw it away. They used a special microscope to find the bad sensor, cut off the tiny wire, and replaced just that one "eye" with a new one. This "surgery" saved about 86% of the tiles that would have otherwise been scrapped.

5. The Cleanliness: The "Dust Detective"

The biggest enemy of these sensors isn't just heat or cold; it's dust.

• The Problem: A speck of dust on a sensor can be radioactive. It acts like a tiny, fake light source, tricking the detector.
• The Solution: They scanned every single tile with a super-high-resolution camera (like a microscope camera) to count every speck of dust. If a tile had too much dust, it was cleaned or rejected. They kept the tiles in nitrogen-purged cabinets (like a sterile hospital room) whenever they weren't being worked on.

6. The Result: A Resounding Success

The paper reports on the final numbers:

• Goal: They needed to build enough tiles to fill 120 units.
• Success Rate: They achieved an 87% yield. This means that out of every 100 tiles they started building, 87 made it to the finish line working perfectly. This is better than their 80% goal.
• Total Count: They successfully built 1,920 working tiles (plus extras for spares).

The Big Picture

This paper is essentially a "receipt" proving that the DarkSide-20k team successfully built the most advanced, ultra-clean, cryogenic light sensors ever assembled. By proving these sensors work perfectly in the cold and are free of radioactive dust, they have paved the way for the experiment to start hunting for dark matter in 2029.

If the sensors are the "ears," this paper confirms that the ears are perfectly tuned, dust-free, and ready to listen for the universe's biggest secret.

Technical Summary

1. Problem Statement

The DarkSide-20k experiment aims to detect dark matter by observing keV-scale energy deposits from Weakly Interacting Massive Particles (WIMPs) scattering off nuclei in 51 tonnes of underground liquid argon (UAr). A critical challenge in such experiments is background suppression, specifically distinguishing rare dark matter signals from background events caused by radiogenic neutrons and cosmic muons.

To achieve this, the experiment utilizes an Inner Veto (IV) volume surrounding the central Time Projection Chamber (TPC). The IV must detect neutrons via their capture on hydrogen (in the PMMA shell) and argon, which release gamma rays. Detecting these gamma-induced scintillation light signals requires a photodetector system that is:

• Highly sensitive: Capable of detecting single photoelectrons (SPE) at cryogenic temperatures (~87 K).
• Low-radioactivity: To avoid introducing new background sources.
• Robust: Capable of operating reliably in liquid argon for a decade.
• Scalable: The system requires instrumenting a large volume with 120 Photo-Detector Units (vPDUs), each comprising 16 sensor tiles.

The paper addresses the design, construction, and rigorous validation of the "Veto Tiles" (vTiles)—arrays of 24 Silicon Photo-Multipliers (SiPMs) integrated onto printed circuit boards (PCBs)—which serve as the fundamental building blocks of the DarkSide-20k veto system.

2. Methodology

Detector Design and Architecture

• SiPM Technology: The system uses custom NUV-HD-Cryo SiPMs (94,904 Single Photon Avalanche Diodes per chip, $11.7 \times 7.9$ mm$^2$) developed with Fondazione Bruno Kessler (FBK).
• vTile Configuration: Each vTile consists of a $5 \times 5$ cm PCB hosting 24 SiPMs.
  • Electrical Topology: The SiPMs are arranged in a "2s-3p" configuration (3 parallel branches of 2 SiPMs in series) per quadrant. This reduces capacitance to preserve single-photon resolution while maintaining a large active area.
  • Readout: The 24 SiPMs are grouped into 4 quadrants. Each quadrant feeds into a custom Application-Specific Integrated Circuit (ASIC) with a Trans-Impedance Amplifier (TIA). The outputs are summed to provide a single signal per vTile.
• vPDU Assembly: 16 vTiles are mounted on a Veto Motherboard (vMB) to form a Veto Photo-Detector Unit (vPDU). Each vPDU provides four summed output signals (one per quadrant), covering an active area of $\approx 10 \times 10$ cm.

Production and Quality Assurance (QA/QC)

Production was distributed across institutes in the UK and Poland, involving strict Quality Assurance (process-oriented) and Quality Control (product-oriented) protocols:

• Database Tracking: A PostgreSQL-based database tracks every component (SiPM wafers, ASICs, PCBs) via unique QR codes, recording test results and shipment history.
• Component Metrology:
  • SiPMs: Characterized at cryogenic temperatures for breakdown voltage ($27.2 \pm 1.0$ V), quenching resistance, and leakage current. Visual inspections rejected wafers with gross physical flaws (pad defects, metal breaks).
  • ASICs: Tested for current draw and charge injection response to ensure consistent gain and timing.
• Assembly Process:
  • Back-side Population: Electronic components were soldered using lead-free solder paste in an ISO-6 cleanroom.
  • Front-side Population: SiPMs were die-attached using indium solder (lower reflow temperature to protect back-side components) and wire-bonded with Al-1%Si wire.
  • Transport: Components were triple-bagged (ESD moisture barrier + radon-impermeable polyethylene/polyamide) and stored in dry nitrogen to prevent radon progeny contamination.

Testing Protocols

Each vTile underwent comprehensive testing at both room temperature ("warm") and cryogenic temperatures ("cold," liquid nitrogen):

1. Electrical Checks: I-V curves to verify breakdown voltage and detect "double breakdowns" (indicating shorts or defects).
2. Performance Metrics:
   • Signal-to-Noise Ratio (SNR): Measured using a 405 nm laser. Defined as the separation between 1-PE and 2-PE peaks relative to the RMS noise.
   • Pulse Count Rate (PCR): Measured in the pre-trigger region to assess uncorrelated noise (Dark Count Rate + environmental radiation).
   • Gain: Verified via 1-PE amplitude.
3. Radiopurity: Components were assayed via ICP-MS, HPGe, and polonium extraction to quantify Uranium/Thorium chains and $^{40}$K. Dust levels were monitored via high-resolution automated imaging.

3. Key Contributions

• Scalable SiPM Array Design: Successfully demonstrated the integration of 24 SiPMs into a single tile with a custom "2s-3p" topology and ASIC-based readout, optimized for low noise and high optical photon detection efficiency ($\approx 46\%$ at 420 nm).
• Robust Cryogenic Assembly: Developed a reliable manufacturing workflow involving indium soldering and wire bonding that withstands thermal cycling from room temperature to liquid nitrogen without mechanical failure or performance degradation.
• Rigorous QA/QC Framework: Established a production pipeline with a total yield of 87% (exceeding the 80% requirement), incorporating repair procedures (e.g., replacing single defective SiPMs on vTiles with double breakdowns) that salvaged over 86% of initially failing units.
• Radiopurity Validation: Confirmed that the assembled vTiles meet stringent radioactivity limits, contributing only $\approx 4\%$ to the total neutron background and a negligible gamma rate to the TPC.

4. Results

• Production Yield: The final production yield for vTiles was 87%, resulting in 1,920 qualified vTiles (plus 6% spares) to populate 120 vPDUs.
• Performance at Cryogenic Temperatures (69 V bias):
  • Breakdown Voltage: Nominal value of 55 V (consistent with $2 \times$ single SiPM breakdown).
  • SNR: Average SNR of 9.3, with a requirement of $>8$. This ensures a SNR $>4$ at the vPDU quadrant level.
  • 1-PE Amplitude: Consistent with the expected board gain of 3.9 mV (accepted range: 3.4–4.8 mV).
  • Noise (PCR): Average Pulse Count Rate of 0.38 Hz/mm$^2$, well below the rejection threshold of 1.8 Hz/mm$^2$.
  • RMS Noise: Average of 300–450 $\mu$V.
• Radiopurity:
  • The total neutron background contribution from the 120 vPDUs is estimated at 5.6 $\times$ 10$^{-3}$ of the surviving neutron fraction.
  • The induced gamma rate in the Inner Veto is 25 Hz, well below the 50 Hz limit required to minimize detector dead time.
• Dust Control: Dust levels remained constant throughout production, averaging <200 counts/cm$^2$ (approx. 400 ng/cm$^2$), confirming the effectiveness of the cleanroom and triple-bagging protocols.

5. Significance

This paper validates the technical feasibility of using large-scale, custom SiPM arrays for the DarkSide-20k experiment. The successful construction and characterization of the vTiles demonstrate that:

1. SiPMs are a viable alternative to traditional photomultiplier tubes (PMTs) for large-volume liquid argon detectors, offering lower radioactivity and high photon detection efficiency.
2. Complex cryogenic electronics can be mass-produced with high yields and reliability, even with strict radiopurity constraints.
3. The veto system is ready for deployment, providing the necessary background rejection capabilities to achieve DarkSide-20k's projected sensitivity to the spin-independent dark matter-nucleon cross-section of $10^{-48}$ cm$^2$.

The robustness of the design and the high production yield ensure that the DarkSide-20k experiment will be fully instrumented and ready to begin its decade-long search for dark matter starting in 2029.