Mini-CryoCube Module: Cryogenic Neutrino Detector

• Mini-CryoCube module is a modular cryogenic detector unit in the Ricochet program, featuring high-purity Ge crystals with simultaneous heat and ionization readout at ~10–20 mK.
• It utilizes a three-detector configuration as a commissioned sub-array, validating principles from full CryoCube design for low-threshold and high-resolution performance.
• The design balances event rate, thermal response, and background discrimination, while addressing challenges such as surface-event contamination and noise reduction.

Searching arXiv for the cited Mini-CryoCube and related CryoCube papers to ground the article. The Mini-CryoCube module denotes a modular cryogenic detector unit associated with the Ricochet reactor-neutrino program, and more specifically, in commissioned form, a three-detector germanium sub-array operated at the Institut Laue-Langevin (ILL) for coherent elastic neutrino–nucleus scattering (CE$\nu$NS) studies. In the broader RICOCHET design lineage, the term can also refer to a single fully instrumented Ge detector unit used as a stand-alone prototype or sub-array, built around the same principles as the full CryoCube: simultaneous heat and ionization readout, operation at $\sim 10$–$20$ mK, and low-noise front-end electronics. The commissioned mini-CryoCube reported in Ricochet consists of three $42$ g high-purity Ge cryogenic calorimeters and served as the first deployed element of CryoCube technology under final-site conditions at ILL (Armatol et al., 30 Jul 2025).

1. Definition and position within the CryoCube program

Within Ricochet, CryoCube is the germanium target of a reactor CE$\nu$NS observatory designed to measure low-energy nuclear recoils with simultaneous heat and ionization readout. The 2021 CryoCube optimization study describes the nominal array as 27 detectors arranged as 3 × 3 × 3, with each detector a high-purity Ge crystal of ~38 g instrumented with 1 NTD-Ge thermal sensor and aluminum electrodes operated at 10 mK in order to measure both the ionization and the heat energies arising from a particle interaction (Salagnac et al., 2021). In that framework, a “detector” or “detector module” is effectively 1 Ge crystal, its sensor, electrodes, holder, and dedicated low-noise cryogenic electronics.

The expression Mini-CryoCube module is used in two closely related senses. In the design and optimization literature, it can be understood as one 38 g Ge detector unit or a small sub-array preserving the same electrothermal and background-discrimination principles as the full CryoCube (Salagnac et al., 2021). In the commissioning report, however, the term denotes a realized hardware module of three 42-gram germanium cryogenic calorimeters, mechanically and electrically organized as one independent unit within the staged deployment toward the full array of eighteen detectors arranged in two floors of nine detectors, themselves grouped into six identical three-detector modules called mini-CryoCubes (Armatol et al., 30 Jul 2025).

This dual usage is important because it distinguishes the natural single-crystal modular unit emphasized during R&D from the three-detector installed module used at ILL. A common misconception is therefore to treat “Mini-CryoCube” as a fixed detector geometry independent of context. The literature instead supports a modular hierarchy: the single crystal is the fundamental detector unit, while the commissioned mini-CryoCube is the first site-deployed multi-detector realization (Salagnac et al., 2021).

2. Detector architecture and operating principle

Each commissioned mini-CryoCube detector is a hybrid heat+ionization bolometer based on a 42 g of high-purity Ge crystal in the standard CryoCube geometry with planar electrodes (Armatol et al., 30 Jul 2025). The phonon channel uses a Ge neutron-transmutation-doped (Ge-NTD) thermistor glued to the crystal, operated at $\sim 14$ mK, while the ionization channel uses two Al electrodes, planar configuration, one collecting holes and the other electrons, biased symmetrically at $\pm 2$ V (Armatol et al., 30 Jul 2025). Earlier CryoCube design studies also considered a fully interdigitated (FID) electrode option with dedicated fiducial and veto electrodes, adapted from EDELWEISS technology, as an alternative to the planar geometry (Salagnac et al., 2021).

The detector measures three related quantities. The total phonon energy $E_{ph}$ is written as

$$E_{ph} = E_r + E_{NTL} = E_r + E_r \cdot \mathcal{Q} \frac{qV}{\epsilon},$$

where $E_r$ is the recoil energy, $\sim 10$0 is the electrode bias, $\sim 10$1 eV is the mean energy to produce one electron–hole pair in Ge for electron recoils, and $\sim 10$2 is the ionization yield (Armatol et al., 30 Jul 2025). The ionization energy is

$\sim 10$3

and the reconstructed recoil energy is

$\sim 10$4

The ionization yield,

$\sim 10$5

is the principal event-by-event discriminator: $\sim 10$6 for electron recoils, $\sim 10$7 for keV-scale nuclear recoils, and $\sim 10$8 for heat-only events (Armatol et al., 30 Jul 2025).

The detector mass scale is deliberate. The optimization study reports that 30 g-scale detectors with ~20 eV RMS resolution are at the state-of-the-art in terms of CENNS rate at surface, whereas larger masses such as ∼300 g would increase per-crystal CENNS rate by a factor ~2.5 but at the cost of higher energy thresholds and slower response, which are detrimental for Ricochet’s goals (Salagnac et al., 2021). This suggests that the Mini-CryoCube mass scale is not merely a convenience of prototyping but a design compromise between event rate, thermal response, and low-threshold discrimination.

3. Cryogenics, readout chain, and module infrastructure

The commissioned mini-CryoCube operated on the 10 mK plate of a CryoConcept Hexa-Dry 200 dilution refrigerator, with a first validation run reaching 8.6 mK without detectors. With detectors installed, the reported operating ranges were 13.5–14.4 mK in RUN013 and 12.0–15.6 mK in RUN014, with most data around 14–15 mK (Armatol et al., 30 Jul 2025). The 2021 optimization study similarly states that the CryoCube stage is regulated at about 15 mK, with prototype characterization at ~20 mK and holder-noise comparisons at 16 mK with an NTD resistance of 1.3 MΩ (Salagnac et al., 2021).

The phonon readout in the commissioned module used Ge-NTD thermistors with bias current: 0.5 nA, optimized at ∼14.2 mK for baseline resolution, and a first amplification stage: JFETs located at the 50 K stage through a thermal decoupler (Armatol et al., 30 Jul 2025). The ionization readout used cryogenic high electron mobility transistors (HEMTs) at ∼1 K, mounted above the detectors on titanium alloy structures; the two ionization signals were digitized and combined offline into a difference channel $\sim 10$9 and a sum channel $20$0 (Armatol et al., 30 Jul 2025). More generally, the front-end development paper specifies HEMTs optimized to achieve ultra-low noise performance at 1K with a dissipation as low as 15 $20$1W per channel, and its noise model predicts that 10 eV heat and 20 eVee RMS baseline resolutions are feasible with a high dynamic range for deposited energy up to 10 MeV (Baulieu et al., 2021).

Data acquisition in the commissioned system used a 20-bit DAQ system with continuous sampling at 100 kHz, with offline downsampling to 50 kHz and matched-filter processing (Armatol et al., 30 Jul 2025). The HEMT front-end architecture was developed specifically for a large multi-channel CryoCube installation in which the 1 kg Ge CryoCube will consist of 27 Ge crystals, all mounted at the 10 mK stage, with the 1 K stage housing the first-stage HEMT amplifiers very close to the detector stage (Baulieu et al., 2021).

At the subsystem level, the modular cryogenic interconnect work provides a complementary architectural model for compact cryogenic modules. That platform provides 74 filtered dc-bias connections, 32 control and readout connections with a bandwidth above 5 GHz, and 4 microwave feed lines that allow operation to above 10 GHz, using an rf interposer that decouples a simple device board from a reusable multilayer signal board (Colless et al., 2014). A plausible implication is that such an interposer-based, reusable-core architecture is technically aligned with the modular logic of Mini-CryoCube deployment, even though the Ricochet detector papers focus on detector, cryostat, and low-noise readout performance rather than on a dedicated backplane module.

4. Mechanical layout, shielding, and auxiliary systems

Mechanically, the three detectors of the commissioned mini-CryoCube were mounted on a copper holder on the 10 mK plate, and the assembly was covered by a cylindrical copper “hat” for radiative shielding and mechanical protection (Armatol et al., 30 Jul 2025). Above the detector assembly, at ∼1 K, sat the HEMT stage on titanium x- and z-shaped supports, connected by Kapton flex PCBs (Armatol et al., 30 Jul 2025). The low-background and low-vibration character of the holder is central to CryoCube performance more generally: the optimization study reports a redesigned holder using 3 sapphire balls on the bottom, 3 on the top, 3 on the sides, all held by chrysocale/bronze clamps with adjustable rigidity, and shows that at 16 mK, with the pulse tube on and no suspension, the new holder reduces vibration-induced noise at 1 Hz by two orders of magnitude in the heat-channel power spectral density (Salagnac et al., 2021).

The installed module was embedded in a heavy shielding environment. The external passive shielding totaled ∼22 t and consisted of a 35 cm borated high-density polyethylene (HDPE) inner layer and 20 cm lead outer layer around the detector region, with 20 cm lead and 35 cm HDPE on top, all surrounded by a soft-iron shell (Armatol et al., 30 Jul 2025). An internal shielding assembly above the detectors, installed between RUN013 and RUN014, used a cylindrical plug on the 1 K plate with two 4.25 cm lead disks, three 1.5 cm copper disks, and 16 alternating layers of ~3.5 cm HDPE disks and 1 cm copper disks, plus 8 mm HDPE sheets around the 1 K, 4 K, 50 K copper cans at the height of the plug (Armatol et al., 30 Jul 2025).

The external active muon veto comprised 34 plastic scintillator panels (200×50×3 cm³) arranged in 17 inner/outer panel pairs, with 6 pairs on top and 11 pairs around the sides, each read out with a PMT (Armatol et al., 30 Jul 2025). The hardware trigger required coincidence of at least one inner and one outer panel above thresholds of ∼3 MeV for top panels and ∼4 MeV for side panels. The veto had panel efficiencies >98%, monitored to better than 1% stability, and shared a common clock with the bolometers, with timing resolution ~1.8 μs initially, improved to ~1.0 μs after firmware upgrade (Armatol et al., 30 Jul 2025).

A further auxiliary system was the laser system, with one optical fiber per detector and room-temperature IR laser diodes at 1590 nm (0.780 eV) and 1650 nm (0.751 eV) for linearity studies, absolute efficiency measurements, and regular neutralization of trapped charge (Armatol et al., 30 Jul 2025). These features show that the commissioned mini-CryoCube was not merely a detector stack but a complete low-temperature measurement module integrated with cryostat, shielding, veto, calibration, and charge-management subsystems.

5. Performance, calibration, and event reconstruction

The commissioning report defines baseline resolution as the standard deviation (1σ) of the fitted pulse amplitudes in noise-only traces using the same matched-filter procedure as for real events (Armatol et al., 30 Jul 2025). After vibration mitigation and optimal configuration, the ionization baseline reached $20$2 eV$20$3 for both RED167 and RED237, while the phonon baseline reached $20$4–66 eV$20$5 for RED167 and $20$6–108 eV$20$7 for RED237, all at 4 V total bias across the electrodes ($20$8 V) (Armatol et al., 30 Jul 2025). Representative values from reactor-off $20$9Cf calibration include 42.2(0.4) eV$42$0, 72(1) eV$42$1 for RED167 and 40.1(0.1) eV$42$2, 79(1) eV$42$3 for RED237 (Armatol et al., 30 Jul 2025).

The R&D target preceding commissioning was more ambitious. For ~33.4 g Ge prototypes, the heat-channel optimization study reported an average baseline heat resolution: 22 eV RMS over five detectors and a best resolution: 17 eV RMS using AC JFET electronics at 400 Hz, with a model prediction of ~13.2 eV RMS for the same electronics and a target: 10 eV RMS baseline resolution for the heat channel with HEMT front-end electronics (Salagnac et al., 2021). The HEMT front-end paper correspondingly predicts 10 eV heat and 20 eVee RMS baseline resolutions as feasible (Baulieu et al., 2021). The difference between these design projections and the commissioned mini-CryoCube performance reflects real site conditions, including vibration, crosstalk, and pile-up.

Calibration relied on internal cosmogenic activation of germanium. The ionization scale used $42$4Ge electron-capture lines at 10.37 keV and 1.3 keV, and an additional $42$5Ar gamma line at 1293.6 keV$42$6 during reactor-on periods (Armatol et al., 30 Jul 2025). Ionization response was found to be linear to better than 1% between 1.3 keV and 1293.6 keV, with RED167 nonlinearity −0.4(5)% at 1.3 keV and −0.35(2)% at 1293.6 keV, and RED237 nonlinearity +0.14(16)% at 1.3 keV and −0.38(2)% at 1293.6 keV (Armatol et al., 30 Jul 2025). The phonon channel used direct calibration to 10.37 keV when visible above noise and otherwise was anchored to ionization through the NTL relation

$42$7

Triggering was performed in software on the phonon channel using a matched filter. The algorithm scanned 15-minute data segments, picking the largest matched-filter sample, excising a 1 s dead-time window around it, and repeating until no significant peaks remained (Armatol et al., 30 Jul 2025). For each triggered event, the channels $42$8, $42$9, and $\nu$0 were fitted in the frequency domain with template shapes, minimizing total $\nu$1, and a common event time parameter was scanned in steps of 2 μs around the trigger time (Armatol et al., 30 Jul 2025). Data-quality cuts included an ionization baseline cut removing 1–2% of data and a phonon low-frequency $\nu$2 cut tuned to retain >93% efficiency for the 1.3 and 10.37 keV lines while removing ~7% of 10.37 keV events (Armatol et al., 30 Jul 2025).

Absolute trigger-plus-reconstruction efficiency was measured with laser events. In the recoil-energy region 2–7 keV, the plateau efficiency was ∼61(9)% for RED167 and ∼63(9)% for RED237, with “half-plateau” efficiency around 0.2 keV$\nu$3 (Armatol et al., 30 Jul 2025). A continuous injection test with ~15 keV$\nu$4 pulses at 1/60 Hz showed that reactor-on efficiency was reduced by ≈15% relative to reactor-off because of pile-up from high gamma rates, especially $\nu$5Ar (Armatol et al., 30 Jul 2025).

6. Backgrounds, limitations, and development trajectory

The background analysis in commissioning concentrated on the recoil-energy interval 2–7 keV, where event populations can be separated in the $\nu$6 plane into electron recoils, nuclear recoils, heat-only events, and surface events (Armatol et al., 30 Jul 2025). The analysis defined an ER band with $\nu$7, an NR band centered on $\nu$8 with width derived from $\nu$9Cf calibration, and a HO region selected by $\sim 14$0 keV$\sim 14$1, approximately 5σ of the ionization baseline (Armatol et al., 30 Jul 2025).

Muon-coincident and anticoincident categories were formed using a $\sim 14$2 μs coincidence window relative to veto triggers (Armatol et al., 30 Jul 2025). In the reactor-off, muon-coincident sample, the likelihood-fitted nuclear-recoil rate in 2–7 keV was 14(3) DRU for RED167 and 13(3) DRU for RED237; in the abstracted commissioning summary, the module reports a nuclear recoil rate of 15(2) events/(kg day keV) during the reactor-off period selecting events in coincidence with muon veto signals, in agreement with the cosmogenic neutron rate calculated from GEANT4 simulations (Armatol et al., 30 Jul 2025). In the reactor-on, anticoincident sample, the combined 90% C.L. upper limit on the nuclear recoil background was reported as < 9 events/(kg day keV) in that energy region, compatible with a GEANT4 model calculation corresponding to a total rate of 5 events/(kg day keV) (Armatol et al., 30 Jul 2025).

The commissioning campaign also identified several limiting backgrounds. Heat-only rates in 2–7 keV were ≈600–650 DRU in reactor-off anticoincidence and ≈700–800 DRU in reactor-on anticoincidence, while ER rates rose from ≈120–135 DRU to ≈1100–1200 DRU when the reactor was on, attributed mainly to $\sim 14$3Ar $\sim 14$4-ray background (Armatol et al., 30 Jul 2025). The principal limitation, however, was surface event contamination. These are electron-recoil interactions near detector or holder surfaces with incomplete charge collection, producing events with intermediate ionization yield that overlap the NR band (Armatol et al., 30 Jul 2025).

This limitation had been anticipated in earlier CryoCube detector-design studies. The planar (PL) design offers ~99.2% fiducial volume from simulations and >98% from data with 3σ charge-conservation cuts, but has no dedicated surface-veto structure; the fully interdigitated (FID) design offers active surface tagging but at the cost of reduced fiducial fraction, ~70% from simulations and ~62% from data, and a larger capacitance of ≈18 pF / 16 pF (Salagnac et al., 2021). Incomplete charge collection at 10.37 keV is <10% for PL with <1% leakage into the NR band, versus <1% incomplete collection and <0.3% leakage for FID (Salagnac et al., 2021). The commissioning paper therefore concludes that surface-event contamination is being addressed with upgraded detectors, specifically including FID detectors, along with copper holder etching / cleaning, a future cryogenic muon veto close to the detectors, and shielding improvements (Armatol et al., 30 Jul 2025).

From the standpoint of CE$\sim 14$5NS sensitivity, the mini-CryoCube did not present a full sensitivity curve, but it demonstrated that the cryogenic technology can reach the required sub-100 eV baseline resolutions at the ILL site, validated the shielding and muon-veto design against GEANT4 predictions for cosmogenic neutrons, and isolated surface events and heat-only events as the main limitations (Armatol et al., 30 Jul 2025). Earlier scaling estimates for a single 38 g detector imply ~0.49 events/day at ILL for a 50 eV threshold, assuming full shielding and background control, whereas the full ~1 kg CryoCube was expected to yield ~12.8 CENNS events/day in total (Salagnac et al., 2021). This suggests that the primary scientific role of a Mini-CryoCube module is system validation—thresholds, discrimination, noise, and background characterization—rather than immediate maximization of CE$\sim 14$6NS significance.