ASTAROTH: Cryogenic SiPM NaI(Tl) Detector

• ASTAROTH is a dark matter direct-detection program employing NaI(Tl) scintillators doped with thallium and advanced cryogenic SiPM arrays instead of PMTs.
• The design targets sub-keV energy thresholds and reduced instrumental backgrounds to perform a model-independent test of DAMA/LIBRA’s annual modulation claim.
• Prototype measurements validate improved photon detection efficiency, enhanced light yield, and a scalable detector architecture for future multi-kg modules.

Searching arXiv for ASTAROTH-related papers to ground the article in the relevant literature. ASTAROTH is a direct-detection dark-matter research and development program centered on sodium iodide scintillators doped with thallium, NaI(Tl), and on the replacement of conventional photomultiplier tubes with cryogenic silicon photomultiplier matrices. In the project literature, the name has been expanded both as “All SensiTive ARray with lOw Threshold” and as “A Novel Detector for Dark Matter Direct Detection Using Cryogenic SiPMs.” Its defining objective is a same-target test of the DAMA/LIBRA annual-modulation claim using NaI(Tl), while pushing energy threshold and instrumental background below the regime typically accessible to PMT-based NaI(Tl) arrays (D'Angelo et al., 2022).

1. Scientific target and experimental rationale

ASTAROTH is motivated by the long-standing DAMA observation of an annual modulation in NaI(Tl) single-hit event rates in the $1$–$6\ \mathrm{keVee}$ interval. In the project description, that modulation is parameterized as

$$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$

with $T \approx 1$ year and $t_0$ the phase of the expected maximum. The explicit ASTAROTH goal is to enable a NaI(Tl) detector with sufficiently low threshold and background to test $S_m$ in the $1$–$6\ \mathrm{keVee}$ region and, crucially, to access sub-$\mathrm{keVee}$ energies (D'Angelo et al., 2022).

The emphasis on NaI(Tl) is methodological rather than generic. DAMA’s claim is formulated in that target material, and the project literature argues that a model-independent test therefore requires NaI(Tl) with equal or better low-energy sensitivity and backgrounds. This target-specific framing also explains ASTAROTH’s focus on instrumental rather than only astrophysical systematics.

The project identifies three PMT-driven limitations in current NaI(Tl) arrays. First, PMTs have modest quantum efficiency at the Tl emission peak, approximately $400$–$6\ \mathrm{keVee}$0. Second, low-energy triggering is compromised by high PMT noise, including “after-glow” and other low-energy nuisance populations whose rate in the region of interest is described as about an order of magnitude higher than the true scintillation rate. Third, PMT materials contribute radiogenic background. ASTAROTH’s central design premise is that these limitations can be attacked simultaneously by using cryogenic SiPMs, which offer higher photon detection efficiency, substantially lower dark count rate at low temperature, and intrinsic sensor radiopurity because they are essentially silicon (D'Angelo et al., 2022).

2. Detector concept and operating principles

The baseline ASTAROTH concept is an all-active NaI(Tl) geometry based on cubic crystals. The reference crystal is a cube of $6\ \mathrm{keVee}$1 edge, with all six faces read out by SiPM arrays in order to maximize light collection and spatial uniformity. Future scaling discussed in the project envisions larger cubes of $6\ \mathrm{keVee}$2–$6\ \mathrm{keVee}$3 edge while remaining within modern ultra-pure crystal manufacturing capabilities (D'Angelo et al., 2022).

The underlying threshold logic is expressed in the project with the relation

$6\ \mathrm{keVee}$4

where $6\ \mathrm{keVee}$5 is the crystal light yield in photons per $6\ \mathrm{keVee}$6, $6\ \mathrm{keVee}$7 the deposited energy in $6\ \mathrm{keVee}$8, $6\ \mathrm{keVee}$9 the optical transport and coupling factor, and $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$0 the SiPM photon detection efficiency. For a trigger that requires a minimum number of photoelectrons $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$1 in a window $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$2, the threshold follows as

$$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$3

This is the formal basis for the project’s repeated emphasis on all-face coverage, improved optical coupling, and cryogenic SiPM operation (D'Angelo et al., 2022).

The same design logic appears in the project’s conceptual signal-to-noise estimate,

$$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$4

which makes explicit why the steep reduction of SiPM dark count rate in the $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$5–$$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$6 range is structurally important. In ASTAROTH’s formulation, the dominant low-energy nuisance population shifts away from PMT noise pulses, enabling cleaner spectra in the DAMA region of interest and potentially below $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$7 (D'Angelo et al., 2022).

The project’s first prototype differs from the six-face baseline design. It used a cylindrical NaI(Tl) crystal of $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$8 diameter and $$R(t) = R_0 + S_m \cos\!\left(\frac{2\pi}{T}\,(t - t_0)\right),$$9 height, with mass approximately $T \approx 1$0, coupled to a single $T \approx 1$1 SiPM matrix while the other faces were wrapped in PTFE. This was explicitly a proof-of-concept configuration rather than the final architectural endpoint (Martinenghi et al., 29 Jul 2025).

Configuration	Geometry	Readout concept
Baseline ASTAROTH design	$T \approx 1$2 cubic NaI(Tl)	SiPM arrays on all six faces
Future scaling concept	$T \approx 1$3–$T \approx 1$4 cubes	Same all-active readout principle
First prototype	$T \approx 1$5 cylinder, $T \approx 1$6	One $T \approx 1$7 SiPM matrix

This progression suggests that ASTAROTH should be understood less as a single detector module than as a staged detector architecture: an initial single-face proof of viability, a six-face low-threshold design, and a longer-term route to multi-kg NaI(Tl) modules.

3. Cryogenic implementation and veto architecture

ASTAROTH operates in a tunable cryogenic interval from $T \approx 1$8 to $T \approx 1$9, with the lower bound set by the liquid-argon bath and the upper bound chosen because SiPM dark count rate begins to approach PMT-like noise around that temperature. The cryostat is a double-walled, vacuum-insulated assembly built around concentric OFHC copper cylinders and a concentric double-walled AISI 316L stainless-steel chimney. A controlled thermal bridge in stainless steel delivers cooling to the detector volume, while a resistive heater establishes the operating setpoint anywhere between bath temperature and $t_0$0 (Alessandria et al., 27 Nov 2025).

The inner volume is filled with dry helium gas. In the design literature this gas plays three roles: it provides thermal conductance to the payload, suppresses condensation, and supports a uniform, tunable gas-conducted cooling mode rather than direct immersion of NaI(Tl). Anti-convection disks along the chimney promote stratification and suppress convective cycles. The operational targets are a temperature range of $t_0$1–$t_0$2, stability of $t_0$3 during data taking, a crystal temporal cool-down gradient below $t_0$4, and a spatial gradient below $t_0$5 across a crystal during cool-down (Alessandria et al., 27 Nov 2025).

Thermal and structural validation are integral to the detector concept because NaI(Tl) is hygroscopic and mechanically delicate under cryogenic cycling. In the design phase, finite-element analysis showed inner-volume temperature uniformity at the level of $t_0$6, while the later multiphysics validation established that the chamber could be tuned across the full target range and remain stable within $t_0$7. At the highest operating temperature, the steepest temperature-gradient region exhibits stresses that slightly exceed the yield strength of copper in a localized strain-hardened condition; nevertheless, after construction the cryostat underwent more than $t_0$8 cooling cycles with no signs of degradation (D'Angelo et al., 2022, Alessandria et al., 27 Nov 2025).

The outer cryogen is intended to be liquid argon not only as a coolant but also as an active veto. The bath can be instrumented with SiPM arrays or PMTs to tag coincident high-energy gammas associated with backgrounds such as $t_0$9 and $S_m$0. The project literature presents this as a scalable and underground-laboratory-compatible alternative to organic liquid scintillator vetoes, which are often disfavored by safety and environmental constraints (D'Angelo et al., 2022).

4. Photosensors, optics, and front-end readout

ASTAROTH’s sensor choice is driven by the spectral and noise properties of NaI(Tl). The scintillator yields approximately $S_m$1–$S_m$2 photons/$S_m$3 at room temperature and emits near $S_m$4. In PMT-based modules this typically translates into roughly $S_m$5–$S_m$6 photoelectrons/$S_m$7, whereas the ASTAROTH design goal, supported by simulations and preliminary optical studies, is to exceed $S_m$8 with all-face SiPM coverage (D'Angelo et al., 2022).

The project has characterized multiple cryogenic SiPM implementations. Early design work discussed FBK NUV-HD-Cryo arrays with $S_m$9 microcells and Hamamatsu S13361-6050 arrays with $1$0 microcells, both in $1$1 formats. The first large-area prototype used an FBK NUV-HD-cryo matrix consisting of $1$2 devices in an $1$3 arrangement, each die $1$4, with $1$5 microcell pitch and total active area $1$6. At $1$7 and approximately $1$8 excess bias, the reported device-level figures are $1$9 PDE at $6\ \mathrm{keVee}$0, dark count rate below $6\ \mathrm{keVee}$1, afterpulsing probability below $6\ \mathrm{keVee}$2, and breakdown-voltage spread below $6\ \mathrm{keVee}$3 across the array (Martinenghi et al., 29 Jul 2025).

The readout is cryogenic and hierarchical. In the prototype, a cold front-end mounted directly behind the SiPM PCB first summed groups of four devices in transimpedance amplifiers, then formed four quadrant outputs, and finally drove shielded differential lines to a warm board and a CAEN V1730 digitizer with $6\ \mathrm{keVee}$4 channels, $6\ \mathrm{keVee}$5-bit depth, and $6\ \mathrm{keVee}$6. Measured single-photoelectron pulse amplitude was approximately $6\ \mathrm{keVee}$7, with pedestal RMS around $6\ \mathrm{keVee}$8, and laser calibration displayed well-separated $6\ \mathrm{keVee}$9 photoelectron peaks with linearity $\mathrm{keVee}$0 (Martinenghi et al., 28 Oct 2025).

Optical coupling remains one of the project’s principal technical bottlenecks. Because NaI(Tl) is highly hygroscopic, ASTAROTH developed gas-tight encapsulations transparent on all six faces and designed to survive repeated thermal cycles. The baseline solutions are a fused-silica case, co-developed with Hilger Crystals, and an acrylic case developed with the Alberta University group. In the first prototype, the crystal was sealed in an airtight fused-silica shell filled with dry neon at $\mathrm{keVee}$1, with a uniform $\mathrm{keVee}$2 gap to accommodate differential thermal contraction. A Geant4-based optical Monte Carlo estimated that only about $\mathrm{keVee}$3 of scintillation photons reached the sensors because of total internal reflection at the crystal–neon interface. An epoxy-resin-based encapsulation and coupling scheme is therefore under development to mitigate these losses (D'Angelo et al., 2022, Martinenghi et al., 28 Oct 2025).

5. Prototype measurements and reported performance

The first cryogenic operation of NaI(Tl) read by SiPMs is a defining milestone of ASTAROTH. The project reports the first operation of a NaI(Tl) crystal read by SiPMs in cryogenic conditions, with separate low-temperature characterizations of both crystal and SiPM response completed before commissioning at INFN LASA (D'Angelo et al., 2022).

Prototype performance was established through laser calibration and gamma-source measurements. For single-photon calibration, a $\mathrm{keVee}$4 picosecond laser at repetition rate below $\mathrm{keVee}$5 was used at $\mathrm{keVee}$6, with about $\mathrm{keVee}$7 pulses recorded in a light-tight Faraday cage. Charge integration over $\mathrm{keVee}$8 captured more than $\mathrm{keVee}$9 of the pulse area, and the charge-per-photoelectron factors in the quadrant sums were of order $400$0–$400$1. The reported signal-to-noise ratio for the $400$2 photoelectron separation is approximately $400$3–$400$4 across the four quadrants (Martinenghi et al., 29 Jul 2025).

In scintillation runs, a $400$5 $400$6 source was placed about $400$7–$400$8 from the detector inside the cryostat cooled to approximately $400$9. The trigger required each quadrant to exceed approximately $6\ \mathrm{keVee}$00, with a four-fold coincidence within $6\ \mathrm{keVee}$01. A $6\ \mathrm{keVee}$02 acquisition window and a $6\ \mathrm{keVee}$03 integration window were used to accommodate the longer NaI(Tl) decay time near $6\ \mathrm{keVee}$04, reported as about $6\ \mathrm{keVee}$05 (Martinenghi et al., 28 Oct 2025).

The published light-yield figures distinguish between gross and crosstalk-corrected values. The $6\ \mathrm{keVee}$06 line from $6\ \mathrm{keVee}$07 produced a peak at approximately $6\ \mathrm{keVee}$08 measured photoelectrons, corresponding to a gross yield of $6\ \mathrm{keVee}$09. Using the Vinogradov method, the prototype measured an optical crosstalk probability $6\ \mathrm{keVee}$10, and the adopted correction

$6\ \mathrm{keVee}$11

yields a net light yield of $6\ \mathrm{keVee}$12 (Martinenghi et al., 28 Oct 2025).

The same prototype literature reports a $6\ \mathrm{keVee}$13 trigger threshold, derived from a four-photoelectron trigger condition and the measured yield, together with $6\ \mathrm{keVee}$14 FWHM energy resolution at $6\ \mathrm{keVee}$15. A statistics-only estimate based on

$6\ \mathrm{keVee}$16

would predict $6\ \mathrm{keVee}$17 for $6\ \mathrm{keVee}$18 at $6\ \mathrm{keVee}$19. The reported excess broadening is attributed to correlated noise, non-uniform light collection, and electronics or operational effects at cryogenic temperature (Martinenghi et al., 28 Oct 2025).

A preliminary report on the same proof-of-concept phase emphasized the gross yield of $6\ \mathrm{keVee}$20, single-photon resolution, and an effective trigger threshold of approximately $6\ \mathrm{keVee}$21–$6\ \mathrm{keVee}$22. It also described a “black run” in which only sporadic triggers remained, attributable to laboratory grounding interference rather than SiPM dark noise. This was used to support the claim that cryogenic SiPM dark noise was not dominating the low-energy trigger population (Martinenghi et al., 29 Jul 2025).

6. Technical constraints, scaling trajectory, and nomenclature

Several ASTAROTH limitations are explicit in the literature. The first is optical inefficiency in the fused-silica/neon prototype encapsulation, where only about $6\ \mathrm{keVee}$23 of scintillation photons were estimated to reach the SiPMs. The second is correlated SiPM noise, with measured optical crosstalk probability $6\ \mathrm{keVee}$24, which degrades both light-yield interpretation and energy resolution. The third is that the prototype was operated above ground; the publications therefore treat the present detector as a technology demonstrator rather than a background-limited dark-matter instrument (Martinenghi et al., 28 Oct 2025).

Radiopurity considerations also change character in the SiPM architecture. Because the photosensors are essentially silicon, the dominant sensor-side background concern shifts from PMT bodies to front-end electronics, PCB substrates, components, and connectors. ASTAROTH’s mitigation strategy is to consolidate SiPMs and ASICs onto a single low-activity PCB, ideally without connectors, using low-radioactivity substrates such as Arlon or Pyralux (D'Angelo et al., 2022).

The scaling trajectory is correspondingly clear. Future modules are planned around cubic NaI(Tl) crystals read out on all faces by SiPM matrices in the $6\ \mathrm{keVee}$25–$6\ \mathrm{keVee}$26 range, with improved epoxy-based optical coupling, tiled SiPM PCBs, and eventually a liquid-argon bath instrumented as an active veto. The present chamber can cool and instrument up to two crystals, while future arrays are intended to use $6\ \mathrm{keVee}$27–$6\ \mathrm{keVee}$28 cubes to increase target mass. Quantitative exposure projections are not given, but the stated design purpose is to lower thresholds below $6\ \mathrm{keVee}$29 and reduce backgrounds in the $6\ \mathrm{keVee}$30–$6\ \mathrm{keVee}$31 interval that defines the DAMA region of interest (D'Angelo et al., 2022).

In arXiv usage, the name “Astaroth” also appears in an unrelated computational context: a GPU-accelerated code and library for high-order three-dimensional stencil computations used in isothermal resistive MHD, pseudodisk simulations, and iterative Poisson solvers for self-gravity (Väisälä et al., 2020, Väisälä et al., 2023, Krasnopolsky et al., 5 May 2026). That software lineage is distinct from the dark-matter detector program. Within astroparticle physics, however, ASTAROTH denotes the cryogenic-SiPM NaI(Tl) effort aimed at a decisive same-target test of DAMA and at extending NaI(Tl) searches into the sub-$6\ \mathrm{keVee}$32 regime.