The heart of the veto system is a device called a vTile, a five-centimetre-square printed circuit board carrying 24 silicon photo-multipliers, or SiPMs. These sensors are remarkable pieces of engineering: each one packs 94,904 single photon avalanche diodes into an area of just 11.7 by 7.9 millimetres, and each can register the arrival of a single photon even when chilled to cryogenic temperatures. Unlike the vacuum-tube photomultipliers used in earlier generations of detectors, SiPMs are solid-state devices that can be manufactured with extremely low levels of natural radioactivity, a crucial property when the faintest gamma ray or neutron can masquerade as a dark matter signal.

The scale of the production effort is staggering. Around 1,400 eight-inch silicon wafers, each containing 268 SiPMs, were fabricated by LFoundry and tested in an ISO-6 cleanroom at the Nuova Officina Assergi facility at Italy’s Gran Sasso National Laboratory. Every accessible sensor on every wafer was probed at liquid nitrogen temperature, with strict acceptance criteria on breakdown voltage, quenching resistance and leakage current. More than 93 percent of the sensors passed. The accepted devices were then diced from their wafers and shipped to assembly sites in the United Kingdom, where they were mounted onto circuit boards and connected with aluminium wire bonds just 25 micrometres thick.

The electrical design of each vTile reflects a careful trade-off between performance and simplicity. The 24 SiPMs are arranged in four quadrants, each containing three parallel branches of two sensors wired in series, a configuration chosen to keep the overall capacitance low enough to preserve the ability to resolve single photons. A precision network of resistors divides the input bias voltage so that every sensor receives an even share, and a custom application-specific integrated circuit amplifies each quadrant’s signal before the outputs are summed. Sixteen vTiles are then mounted on a larger motherboard to form a Veto Photo-Detector Unit, whose four output channels each sum 96 SiPMs covering roughly 100 square centimetres. This aggressive summing slashes the number of data channels and, just as importantly, reduces the mass of radioactive cabling that would otherwise sit inside the detector.

Every single tile was subjected to a punishing battery of tests at both room temperature and, in a bath of liquid nitrogen, at around 77 kelvin. Physicists at the University of Oxford and STFC’s Interconnect facility built identical test stands that lower four tiles at a time into dewar flasks, illuminating them with pulses from a 405-nanometre laser. The team measured how the sensors respond to individual photons, recording the voltage of the single and double photoelectron peaks and comparing them to the electronic noise floor. The requirement was a signal-to-noise ratio above 8 for each tile, a threshold that guarantees the assembled detector units can reliably distinguish genuine light signals from background fluctuations.

Not everything survived the ordeal, and the failure modes proved scientifically interesting in their own right. Some tiles displayed a puzzling double breakdown in their current-voltage curves, a signature that could indicate mechanical damage to a sensor or a deformed wire bond shorting against a silicon surface. Rather than discarding these units, the collaboration developed a diagnostic procedure that biases each of the 24 SiPMs individually to identify the single faulty device, then removes it, cleans the solder and bonds in a replacement. This rework rescued more than 86 percent of the affected tiles, saving precious silicon and keeping the production on schedule.

Radioactive cleanliness demanded almost obsessive attention throughout. Every component, from the capacitors to the circuit boards themselves, was assayed using three independent techniques: mass spectrometry, high-purity germanium gamma spectroscopy and polonium extraction. The dominant neutron background traces to the lower uranium chain in the capacitors, followed by the boards, whose sheer mass makes them significant despite modest activity levels. Because ordinary dust carries radioactive contaminants and radon daughters can plate onto surfaces, all assembly took place in cleanrooms with radon levels around 2 becquerels per cubic metre, and parts were triple-bagged in vacuum-sealed barriers for transport. Automated scanners photographed each tile at 6400 dots per inch before and after every production stage, using image analysis algorithms to count and characterise every particle of dust on the silicon surfaces.

The results validate the entire production line. A fully populated tile was radio-assayed directly and compared with the sum of its individual components, and the two agreed, confirming that the assembly process itself added negligible radioactivity. Based on these measurements, the 120 detector units destined for the Inner Veto will contribute only about 4 percent of the total neutron background and a gamma rate of 25 hertz, comfortably within the specifications needed to keep the experiment essentially free of instrumental backgrounds for a decade of operation. Dust levels held steady at under 200 counts per square centimetre across all testing and integration phases.

Perhaps the most striking number in the paper is the production yield. The collaboration needed an 80 percent cumulative yield across all manufacturing steps to have enough working tiles, and it achieved 87 percent, corresponding to 1,920 qualified vTiles filling 120 detector units, with a further 6 percent held as spares. For a technology being deployed at this scale for the first time, that figure represents a genuine industrial achievement, comparable to the yields of mature commercial semiconductor production lines rather than a first-of-its-kind scientific instrument.

When DarkSide-20k begins taking data around 2029, it will operate the largest SiPM-based detector ever built, with a projected sensitivity to dark matter interactions down to cross-sections of 10 to the minus 48 square centimetres for a 100 gigaelectronvolt particle. The Inner Veto will surround the central time projection chamber with nearly complete four-pi coverage, tagging neutrons through their capture on hydrogen and argon nuclei and rejecting any event with coincident light. If a dark matter particle finally reveals itself in the argon below, it will be thanks in no small part to these meticulously tested squares of silicon, each one a tiny sentinel standing guard against the universe’s most persistent background noise.

Subject of Research: Production and cryogenic characterisation of silicon photo-multiplier veto tiles for the DarkSide-20k dark matter experiment

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

Article References: DarkSide-20k Collaboration, Acerbi, F., Adhikari, P., Agnes, P., Ahmad, I., Albergo, S., Albuquerque, I. F. M., Alexander, T., Alton, A. K., Amaudruz, P.-A., Anastasi, G. A., Angiolilli, M., Aprile, E., Auty, D. J., Pernas, M. A., Azzolini, O., Back, H. O., Balmforth, Z., Olmedo, A. I. B., … Zuzel, G. (2026). Construction and characterisation of the DarkSide-20k veto silicon photo-multiplier tiles. The European Physical Journal C, 86(9), Article 1089. https://doi.org/10.1140/epjc/s10052-026-16217-7

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16217-7

Keywords: DarkSide-20k, dark matter, silicon photo-multipliers, liquid argon, neutron veto, cryogenic detectors, Gran Sasso, radiopurity, particle physics, detector instrumentation, wire bonding, quality control

The DarkSide-50 detector is a dual-phase time projection chamber filled with 46.4 kilograms of argon harvested from deep underground wells. This underground argon is precious because its concentration of the radioactive isotope argon-39 is at least 1400 times lower than that of ordinary atmospheric argon, dramatically suppressing one of the most stubborn backgrounds in rare-event searches. When a particle interacts inside the cylindrical target, it produces a prompt flash of scintillation light, called S1, and liberates ionization electrons that drift upward through the liquid under a 200 volts-per-centimeter electric field. At the liquid surface, a stainless steel grid extracts these electrons into a thin layer of argon gas, where they generate a second, amplified flash of electroluminescence known as S2.

Each extracted electron produces roughly 23 detected photoelectrons, making the S2 signal extraordinarily sensitive. This sensitivity underpins the so-called S2-only analysis strategy, in which researchers hunt for dark matter collisions so feeble that they generate ionization electrons but too little scintillation light to register an S1. Because the S2 channel can be read with near-perfect efficiency, it opens a window onto dark matter particles with masses below about 10 gigavolts per square of the speed of light, a regime where conventional analyses lose their grip. The same technique is proposed for detecting coherent neutrino scattering from supernovae and reactors, making its reliability a matter of broad interest across astroparticle physics.

But every experiment that has run in this mode, whether filled with xenon or argon, has been plagued by an excess of tiny signals corresponding to just a few extracted electrons. These spurious electrons, or SEs, appear at rates that no model of radioactive background can explain. Their presence forced DarkSide-50’s earlier light dark matter searches to discard their lowest-energy data, capping their sensitivity at WIMP masses above roughly 1.2 gigavolts per square of the speed of light. Until now, no one had systematically characterized these events in liquid argon, leaving theorists to extrapolate from xenon experiments such as XENON1T and LUX.

The new analysis draws on nearly 1000 days of data collected between April 2015 and February 2018. The collaboration classified every triggered event by its pulse structure and isolated those containing a single S2-like pulse with fewer than four extracted electrons. The resulting time series revealed something striking: while ordinary background categories remained rock-steady over three years of operation, the spurious-electron rate decayed rapidly during the first 200 days after the detector was filled, with a characteristic timescale of about 65 days, before settling into a much slower decline. Even more telling, when the hot getter that purifies the recirculating argon gas was bypassed for five days of maintenance, the spurious-electron rate jumped sharply and then fell back with a 36-hour time constant after the getter was restored, matching the expected purification speed of the circulation loop.

That spike pointed the finger at chemical impurities. The team therefore measured two independent purity metrics: the free electron lifetime, which quantifies how long electrons survive drifting through the liquid before being captured by electronegative contaminants such as oxygen, and the lifetime of the long-lived triplet component of the argon scintillation, which is sensitive to nitrogen contamination. Surprisingly, neither metric degraded during the getter-off period, and neither correlated with the spurious-electron rate over the full dataset. Whatever impurities are capturing and later releasing electrons, they are evidently not the same species that dominate electron attachment or scintillation quenching, hinting at a subtle population of contaminants that standard purity monitors cannot see.

To pin down the mechanism, the collaboration searched for temporal coincidences between spurious electrons and ordinary particle interactions, treating the latter as potential parent events. The distribution of time delays between parents and SEs resolved cleanly into two exponential components with decay constants of about 5 milliseconds and 40 to 80 milliseconds, plus a third component with a roughly 16-millisecond constant that appeared only while the getter was off. At the start of the dataset, these correlated components accounted for about 70 percent of all spurious electrons, falling to about 30 percent after 200 days as the responsible impurities were gradually removed. A longer-lived component extending beyond one second was also visible, and the team found that the apparently uncorrelated SE rate scales linearly with the total ionization activity in the detector, suggesting that even these events may trace back to parents whose delayed signatures stretch over seconds.

The spatial and energetic fingerprints reinforced the impurity picture. Spurious electrons tended to appear within about 5 centimeters of their parent’s reconstructed horizontal position, and the probability of an SE following a parent grew linearly with both the parent’s S2 charge and its electron drift time. Normalizing for the parent’s charge, the collaboration derived a trapping probability of about 1.74 times ten to the minus eight per electron per millimeter of drift, a tiny but consequential efficiency. Notably, the number of electrons lost to genuine capture by electronegative impurities is far larger than the number that resurface as spurious signals, meaning only a small fraction of captured electrons ever escape their traps.

The multi-electron events yielded perhaps the most unexpected clue. The distribution of electron multiplicity in SE events follows a Poisson law with a mean of just 0.062 extra electrons per event, independent of impurity species, parent energy, and drift time. The team proposes that the 128-nanometer ultraviolet photons of the electroluminescence process themselves photo-ionize the stainless steel extraction grid, whose measured photoelectric yield at these wavelengths is around one percent. A simple calculation predicts a 2 to 4 percent chance that a single-electron event liberates a secondary electron from the grid, beautifully consistent with the observed mean. Pulse-shape analysis supports this: two- and three-electron events show light arriving at slightly different times, exactly as expected if secondary electrons reach the gas pocket about a microsecond after the photons from the first electron strike the metal.

The implications reach well beyond DarkSide-50, which has since completed its mission. Its successor, DarkSide-20k, and other next-generation argon and xenon detectors plan to push S2-only analyses to even lower thresholds in pursuit of dark matter particles with sub-GeV masses. If the impurity-driven components can be eliminated through tighter purification and material control, and if the grid photo-ionization can be suppressed by coating or redesigning the extraction electrode, the spurious-electron floor that currently limits these searches could drop dramatically. The DarkSide-50 results provide the first quantitative roadmap for that effort, turning a mysterious background into a characterized, and potentially conquerable, phenomenon.

Subject of Research: Spurious-electron background signals in the dual-phase liquid argon time projection chamber of the DarkSide-50 dark matter experiment

Article Title: Characterization of spurious-electron signals in the double-phase argon TPC of the DarkSide-50 experiment

Article References: DarkSide-50 Collaboration, Agnes, P., Albuquerque, I. F., Alexander, T., Alton, A. K., Ave, M., Back, H. O., Batignani, G., Berzin, E., Biery, K., Bocci, V., Bonivento, W. M., Bottino, B., Bussino, S., Cadeddu, M., Cadoni, M., Calaprice, F., Caminata, A., Campos, M. D., … Zuzel, G. (2026). Characterization of spurious-electron signals in the double-phase argon TPC of the DarkSide-50 experiment. The European Physical Journal C, 86(9), Article 1099. https://doi.org/10.1140/epjc/s10052-026-16193-y

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16193-y

Keywords: DarkSide-50, dark matter, liquid argon, time projection chamber, spurious electrons, WIMP, low-mass dark matter, ionization signal, Gran Sasso, electronegative impurities, photo-ionization, S2-only analysis