betaFeedback

WorksPhysics and Astronomy

Search for dark matter Particles via Invisible Decays in 46{}^{46}Sc Nuclear γγ Cascades with a CsI(Tl) Detector

Abstract

Dark matter remains one of the most compelling open problems in modern physics, motivating experimental searches for new light, weakly coupled particles beyond the Standard Model. Despite extensive efforts employing diverse detection strategies, large regions of parameter space remain unexplored. We report a high-statistics laboratory search for invisible decay modes in nuclear γγ-ray cascades using approximately 100 kg100~\mathrm{kg} of CsI(Tl) scintillators operated at Texas A&M University. The experiment employs a high-activity 46{}^{46}Sc radioactive source and a ``missing-γγ'' technique, in which the absence of a photon from a well-identified cascade serves as a signature of new physics. Unlike appearance-disappearance experiments, this approach requires only a single photon conversion into a dark-sector particle, enabling sensitivity to significantly weaker couplings. The setup provides simultaneous sensitivity to a broad class of light dark-sector candidates, including axions and axion-like particles, dark scalars, and dark photons in the 0.1−1 MeV0.1 - 1 \text{ MeV} mass region. Through careful control of detector containment, energy resolution, and environmental backgrounds, we exclude certain regions on the previously explored parameter space. With foreseeable improvements in detector volume and systematic uncertainty control, this technique has the potential to probe currently unexplored parameter space for axion-like particles and light dark scalars.

Introduction

Various astrophysical observations based on the behavior of galaxies in the galactic clusters [1, 2], rotational velocity curves of stars in galaxies [3], gravitational lensing [4]. have confirmed the existence of non-luminous matter in the universe or the so-called dark matter. About 83% of the matter content of the universe is in the form of dark matter. This experiment aims to search for several dark matter candidates, including axions and axion-like particles (ALPs), dark scalars, dark photons, and milli-charged particles.

Light, weakly coupled particles below the MeV scale arise naturally in many extensions of the Standard Model and are compelling dark-sector candidates, either as dark matter itself or as mediators of its interactions. They also appear in solutions to the strong CP problem [5, 6, 7], the gauge hierarchy problem [8], and the cosmological constant problem [9], and have been invoked in astrophysical contexts such as stellar energy transport and supernova shock revival [10]. While astrophysical and cosmological observations place strong bounds on these particles, the limits are highly model-dependent and can be evaded with modest theoretical modifications. These considerations highlight the importance of laboratory-based searches with controlled systematics to provide robust and complementary probes of light, weakly coupled particles.

In this work, we present a high-statistic, laboratory-based search for rare invisible decay channels in well-identified nuclear γ\gamma cascades. The experiment employs a 46^{46}Sc radioactive source placed at the center of a ∼100\sim100 kg CsI(Tl) scintillator array. By exploiting coincidence tagging of the 1120 keV γ\gamma line and searching for the associated 889 keV photon, the setup is designed to identify missing γ\gamma events that may signal the conversion of a photon into a hidden-sector particle such as an axion-like particle, dark photon, or scalar mediator. Through dedicated simulations, careful control of detector containment, and extensive background characterization using passive shielding and active vetoes, this work demonstrates the feasibility of probing rare processes in a controlled environment and sets the stage for next-generation ton-scale experiments with improved containment and sensitivity.

This paper is organized as follows. Section 2 presents the theoretical framework, encompassing the production of axions and axion-like particles (ALPs) from radioactive nuclear transitions, the resulting missing-γ\gamma probability in the detector, and the overall sensitivity of the search. Section 3 describes the experimental setup, including the choice of radioactive source, detector assembly, passive and active shielding, and detector stability and characterization. Section 4 discusses the background sources relevant to the missing-γ\gamma signal and quantifies their contributions. Section 5 presents the preliminary analysis performed with a 22^{22}Na source, followed by the missing-γ\gamma signal extraction with the 46^{46}Sc source and the associated statistical and systematic uncertainties. Section 6 summarizes the exclusion limits obtained for various dark-sector candidates, including axions and axion-like particles, dark scalars, dark photons, and milli-charged particles, and outlines potential upgrades and future improvements to extend the sensitivity to previously unexplored regions of parameter space.

Theoretical Framework

Nuclear γ\gamma cascades provide a clean and well-controlled environment to search for exotic decay channels. Radioactive isotopes such as 60Co{}^{60}\mathrm{Co} and 46Sc{}^{46}\mathrm{Sc} undergo beta decay followed by sequential γ\gamma emissions as the daughter nucleus de-excites to its ground state. The resulting γ\gamma spectra exhibit well defined and characteristic lines associated with each transition. Both isotopes feature two γ\gamma lines, allowing one γ\gamma to be used as a tag while the second is examined for anomalies. In practice, the large energy separation between the two γ\gamma lines in 46Sc{}^{46}\mathrm{Sc} compared to 60Co{}^{60}\mathrm{Co} makes 46Sc{}^{46}\mathrm{Sc} the preferred choice.

There are two generic mechanisms by which radioactive decays can produce dark sector particles. In the first, an intermediate excited nuclear state decays invisibly by emitting a dark sector particle instead of a γ\gamma photon. In this case, the standard γ\gamma cascade appears incomplete. In the second mechanism, a γ\gamma emitted in the cascade subsequently converts into a hidden sector particle through suppressed but non zero interactions with surrounding nuclei or electromagnetic fields. Examples include axions or axion-like particles produced via Primakoff type processes or related conversion channels. In both scenarios, the final state consists of one observed γ\gamma and one beyond Standard Model particle, leading to an apparent missing γ\gamma in an otherwise well understood two γ\gamma cascade. This missing γ\gamma signature provides a direct and experimentally robust handle to search for new physics.

This work focuses on a set of theoretically well-motivated models together with their effective interaction operators [12]. Table 1 summarizes the expressions for the missing branching fractions corresponding to different dark matter candidates. The parameter gpg_p denotes the effective scalar coupling to nucleons [11], while ϕ\phi represents a scalar mediator of mass mϕm_\phi. The symbol ϵ\epsilon refers to the kinetic mixing parameter between the Standard Model photon and the dark photon fields, where the dark photon has mass mA′m_{A'}. The parameter QQ corresponds to the fractional electric charge of a milli-charged particle χ\chi with

ModelLint\mathcal{L}_{\mathrm{int}}BRmiss\mathrm{BR}_{\mathrm{miss}}
Scalar (nucleon coupling)gpϕNˉNg_p\phi\bar{N}Ngp22e2(1−mϕ2ω2)5/2\frac{g_p^2}{2e^2}\left(1-\frac{m_\phi^2}{\omega^2}\right)^{5/2}
Dark PhotonϵFμνFμν′\epsilon F^{\mu\nu}F'_{\mu\nu}ϵ2(1−mA′2ω2)5/2\epsilon^2\left(1-\frac{m_{A'}^2}{\omega^2}\right)^{5/2}
Milli-charged Particle−QχˉγμAμχ-Q\bar{\chi}\gamma^\mu A_\mu\chiQ2925απκ(ω,mQ)ω5\frac{Q^2}{9}\frac{25\alpha}{\pi}\frac{\kappa(\omega,m_Q)}{\omega^5}
ALP (photon coupling)14faγaFμνF~μν\frac{1}{4f_{a\gamma}}aF_{\mu\nu}\tilde{F}^{\mu\nu}σPσP+σSM(1−e−ℓ/λ)\frac{\sigma_P}{\sigma_P+\sigma_{SM}}\left(1-e^{-\ell/\lambda}\right)

Table 1. Missing branching fractions of the dark matter candidates.

mass mQm_Q. Finally, aa denotes an axion-like particle (ALP) whose coupling to photons is characterized by the decay constant faγf_{a\gamma}. The coupling constants gpg_p, ϵ\epsilon, QQ, and faγf_{a\gamma} thus define the primary experimental parameters governing the sensitivity of the search to new physics beyond the Standard Model.

This approach offers an advantage over other laboratory-based axion searches, such as those using reactors or beam-dump experiments or light shining through the wall experiments. In our case, we focus solely on the production of rare particles at the source, without relying on their subsequent conversion back to photons or electrons at the detection site. As a result, the detection strategy requires only a single interaction during the nuclear decay itself thereby increasing the probability of observing such rare events.

Experimental set-up

Completing the experiment requires both the construction and detailed characterization of the detector. The proposed setup consists of a ∼100\sim100 kg CsI(Tl) scintillator array with a radioactive source placed at its center. The source is selected such that it undergo decays that produce at least two particles: one acting as a trigger or tag, and the other providing sensitivity to potential rare or invisible decay modes.

The choice of radioactive isotope is therefore crucial. Common candidates such as 24Na{}^{24}\mathrm{Na}, 60Co{}^{60}\mathrm{Co}, and 46Sc{}^{46}\mathrm{Sc} emit two primary γ\gamma rays suitable for coincidence measurements. Among these, 46Sc{}^{46}\mathrm{Sc} is particularly advantageous, as it emits well-separated γ\gamma lines at 889 keV and 1120 keV that can be clearly resolved by the detector. Its half-life of 84 days ensures a sufficiently long data-taking period, while the source activity can be tuned to provide high decay statistics without introducing excessive pile-up. These features make 46Sc{}^{46}\mathrm{Sc} a well-motivated choice for use with the CsI(Tl) detector. However, a key limitation of CsI(Tl) is its slow decay time (3μs3\mu\mathrm{s}), which restricts the choice of source activity to <37< 37 kBq (<1 μ< 1\ \muCi) to prevent pile-ups in the data acquisition system.

Detector Assembly

CsI(Tl) crystals were selected for their high density (4.51 g/cm3^3) and large scintillation light yield (∼5.4×104\sim5.4 \times10^4 photons/MeV), with a peak emission wavelength of 550 nm. It also provide efficient containment of MeV-scale γ\gamma rays and excellent energy resolution [13]. The detector array consists of CsI(Tl) crystals originally commissioned for the CLEO experiment [14], with more than 10 tons of material available for future detector expansions and ton-scale implementations of the missing-gamma technique.

Each CsI(Tl) crystal has dimensions of 30.5×5.1×5.1 cm330.5 \times5.1 \times5.1\ \mathrm{cm}^{3} and a mass of approximately 4 kg. Crystals were selected after inspection for radiation damage, hygroscopic degradation, and mechanical defects [15]. Surface restoration and optical treatment were performed using progressively finer grit sandpapers followed by polishing liquids to improve surface smoothness, optical transparency, and scintillation light collection uniformity. Each crystal was subsequently wrapped with a high-reflectance PTFE layer providing >92%> 92\% reflectivity for CsI(Tl) scintillation photons [16], followed by an outer aluminized Mylar layer for optical isolation and electromagnetic shielding [17, 13]. These preparation procedures ensure efficient collection and optical isolation of the scintillation photons produced within each crystal.

The wrapping fully covers the crystal surfaces except for one end, which is optically coupled to a photomultiplier tube (PMT) using EJ-550 silicone optical grease [18]. The optical coupling compound provides near-unity transmission efficiency at the CsI(Tl) scintillation emission wavelength, maximizing photon transport from the crystal into the PMT photocathode. Scintillation photons incident on the photocathode of the PMT are converted into photoelectrons and subsequently amplified through a multi-stage dynode chain present in the PMT, producing an electrical pulse proportional to the deposited energy. The detector employs 5 cm diameter Electron Tube 9954B series PMTs, whose photocathode area closely matches the crystal cross section. The PMTs provide a quantum efficiency of approximately 10% at the 550 nm emission wavelength of CsI(Tl). Combined with the high scintillation light yield of the crystals, this results in sufficient photoelectron statistics to achieve excellent energy resolution, as demonstrated in § 5 [19]. The rear end of each PMT is connected to a voltage-divider base, which distributes the supplied high voltage across the photocathode and dynode stages through SHV (Safe High Voltage) coaxial cables. The amplified PMT output signal is extracted from the base and transmitted to the digitizer using coaxial BNC cables for waveform acquisition and analysis.

The detector assembly consists of 26 CsI(Tl) crystals arranged in a 5×55 \times5 geometry. The central crystal is divided into fixed and removable sections to accommodate the radioactive source at the detector center, with the fixed section made slightly larger to support the surrounding crystals when the removable section is extracted. The crystal stack is held in a custom acrylic support structure that ensures uniform alignment and stable optical coupling. Adjacent crystals are oriented with alternating PMT directions to minimize crowding and facilitate readout integration. Cylindrical openings in the holder accommodate the PMT assemblies. The complete detector assembly is shown in Figure 2.

Photograph of the approximately 100 kg CsI(Tl) detector assembly

Figure 2. Photograph of the ∼\sim100 kg CsI(Tl) detector assembly, serving as a first-generation proof-of-concept prototype for the proposed ton-scale detector design as shown in Figure 1. The detector consists of a stacked crystal array coupled to photomultiplier tubes (PMTs), with alternating PMT orientations adopted to optimize packing geometry and reduce readout crowding. The central detector module is divided into fixed and movable sections, allowing radioactive sources to be inserted or replaced without disassembling the detector array.

Schematic diagram of the proposed ton-scale stacked CsI detector array

Figure 1. Schematic diagram of the proposed ton-scale stacked CsI detector array designed to efficiently contain γ\gamma rays emitted from a radioactive source located at the center of the detector [11]. The emitted γ\gamma rays interact within the CsI crystal assembly, producing scintillation photons that are collected by photomultiplier tubes (PMTs) coupled to the crystals.

DAQ system

High-voltage biasing for the PMTs is provided by a CAEN SY5527LC mainframe equipped with two A7236 HV modules [20]. Each A7236 module provides 24 independent high-voltage channels with output capability up to 3.5 kV and 1.5 mA per channel [21], matching the operating requirements of the Electron Tubes 9954B PMTs, which typically operate between 1600–2200 V with current draw below 1 mA. Detector high voltage, channel currents, and system monitoring are controlled through the GECO2020 interface [22].

Signal acquisition is performed using a CAEN V1740D waveform digitizer installed in a compact CAEN VME mini-crate [23]. The digitizer provides 64 input channels with LEMO connectors and operates at a sampling rate of 62.5 MHz, corresponding to a 16 ns time resolution. This sampling rate is sufficient to accurately resolve the ∼3 μs\sim3\ \mu\mathrm{s} scintillation decay profile of CsI(Tl) crystals [24]. Data acquisition and online monitoring are handled using the CoMPASS software package [25], which provides real-time waveform processing and energy spectrum visualization.

Owing to the long CsI(Tl) scintillation decay time, a waveform acquisition window of 10 μs\mu\mathrm{s} was used, with the trigger position set at 40% of the acquisition window. Event energies were reconstructed by integrating the pulse waveform over a 3 μs3\ \mu\mathrm{s} interval beginning at the trigger time. CoMPASS performs online pulse integration and energy reconstruction during acquisition, eliminating the need to store individual waveforms and substantially reducing the data volume during detector commissioning and calibration runs. For offline analysis, reconstructed event information, including channel number, trigger timestamp, and deposited energy was recorded in Comma-Separated Values (CSV) format.

Passive Shielding and Active Veto

The detector assembly incorporates dedicated passive shielding and an active veto system to suppress background radiation that could mimic a missing-γ\gamma signature. The design and performance of these systems are described below. A comprehensive characterization of all the background sources relevant to the present search, including their estimated contributions to the missing-γ\gamma signal region, is presented in § 5.

Background radiation include intrinsic radioactive contaminants in the CsI(Tl) crystals and external sources such as ambient γ\gamma rays, β\beta particles, cosmic-ray–induced particles, and neutrons. γ\gamma rays with energies near the tagging line (∼1120\sim1120 keV) are particularly relevant, as they can add false count to the missing gamma signal. To suppress external backgrounds, the detector assembly is enclosed within a lead shield constructed from 20.4×10.2×5.1 cm320.4 \times10.2 \times5.1\ \mathrm{cm}^{3} lead bricks, providing a minimum shielding thickness of 10.210.2 cm on all six sides. Lead shielding of this thickness attenuates ∼1\sim1 MeV γ\gamma rays by approximately three orders of magnitude [26].

In addition, 11 plastic scintillator panels, each measuring 35.6×10.2×5.1 cm335.6 \times10.2 \times5.1\ \mathrm{cm}^{3}, are deployed as an active veto layer surrounding the detector inside the lead shielding. The scintillators are fabricated from BC-408 plastic scintillator material previously used in the CDMS-II experiment [27, 28]. The veto system is designed to reject residual backgrounds not sufficiently attenuated by the passive lead shielding, particularly cosmic-ray–induced events. Each scintillator panel was machined, polished, and wrapped following procedures similar to those used for the CsI(Tl) crystals. PMTs coupled to the scintillators provide veto signals through the readout channels. Events were selected by requiring activity in the CsI(Tl) detector array with no coincident signal in the veto scintillators within the predefined coincidence window [29].

A dedicated background measurement was performed using this configuration to characterize the residual environmental and cosmic-ray backgrounds. The corresponding analysis is presented in § 4, and the resulting background energy spectrum is shown in Figure 7.

Energy spectra recorded with the CsI(Tl) detector array

Figure 7. Energy spectra recorded with the CsI(Tl) detector array in the absence of an external radioactive source, showing contributions from environmental radioactivity and intrinsic crystal backgrounds. The spectrum is shown over the energy range 0–10 MeV, where peaks associated with 137Cs{}^{137}\mathrm{Cs} (660 keV) and 40K{}^{40}\mathrm{K} (1460 keV) are visible. The blue curve corresponds to the full background spectrum, while the orange curve shows the spectrum after application of the active veto.

Stability and Thermalization

Prior to extended data collection with the 46Sc{}^{46}\mathrm{Sc} source, we conducted detector stability studies by monitoring the temporal variation of the calibration peak positions. During the initial stages of operation, we identified a gradual downward drift of the 1275 keV calibration peak, which we attributed to time-dependent gain variations in the detector response. Further investigation revealed that sustained operation induced heating of the PMT bases, whose temperature-dependent behavior modifies the signal current and consequently alters the ADC response of the detector channels [30, 19]. To characterize and mitigate this effect, we implemented an active cooling scheme while continuously recording the PMT-base temperature via thermocouples. A thermal enclosure was constructed around the detector assembly to suppress heat exchange with the ambient environment, through which temperature-regulated air was circulated by means of an external cooling system. The temperature of the injected air was stabilized near 21°C.

With external cooling applied, the PMT-base temperature stabilized near 31°C, compared to temperatures exceeding 38°C in the absence of cooling, as shown in Figure 3. The improved thermal stability substantially reduced calibration drift during extended operation, as illustrated in Figure 4. Consequently looking into both the figures, all physics data runs were initiated only after a 40-minute stabilization period following high-voltage turn-on to ensure stable detector response. The combination of active cooling and controlled operating conditions effectively mitigated PMT-base overheating and enabled stable long-term detector performance.

PMT-base temperature measured during detector operation before and after the application of external cooling

Figure 3. PMT-base temperature measured during detector operation before and after the application of external cooling. Without cooling, the temperature rises above 38°C, while active cooling stabilizes the operating temperature near 31°C, significantly improving detector stability during long-duration operation.

Temporal evolution of the 1275 keV calibration peak position for base-19

Figure 4. Temporal evolution of the 1275 keV calibration peak position for base-19. Blue points correspond to measurements taken before cooling, showing a continuous downward drift from approximately 1274.5 keV to 1173 keV over ∼106\sim106 minutes. Red points correspond to measurements taken after cooling, showing significantly improved stability following an initial transient response.

In addition, dedicated analysis procedures were implemented to preserve detector stability throughout the full data-taking period. To further compensate for the time-dependent ADC gain variations, each detector channel was independently re-calibrated in successive 15-minute intervals. The choice of this calibration window was determined quantitatively by studying the temporal stability and uniformity of the calibration peak positions shown in Figure 3 and Figure 4. This procedure preserves the intrinsic detector resolution and maintains stable energy reconstruction during long-duration detector operation.

Detector Characterization

Several detector-characterization studies were performed to validate the performance of the experimental setup. These studies include measurements of energy linearity, energy resolution, and the effective trigger threshold of the data acquisition system.

Each CsI(Tl) crystal is characterized using radioactive sources such as 22Na{}^{22}\mathrm{Na}, 60Co{}^{60}\mathrm{Co}, 57Co{}^{57}\mathrm{Co}, 137Cs{}^{137}\mathrm{Cs}, and 133Ba{}^{133}\mathrm{Ba}. γ\gamma-ray lines spanning the energy range 122–1460 keV were measured with the sources placed directly along the long axis of the CsI(Tl) crystal at its midpoint, with scintillation photons collected at one end of the long axis by a coupled PMT. The measured ADC response as a function of deposited energy is well described by a linear calibration of the form [31]

EADC=m⋅Eγ+b,(1)E_{\mathrm{ADC}} = m \cdot E_{\gamma} + b, \tag*{(1)}

where m=(0.117±0.006)m = (0.117 \pm0.006) ADC/keV is the gain and b=(7.64±5.7)b = (7.64 \pm5.7) ADC is the offset, the fit yields an excellent value of R2R^2 of 0.9897. The detector demonstrates stable and reliable performance of the crystal–PMT–base assembly for precision γ\gamma-ray spectroscopy, as shown in Figure 5a.

Figure 5. Characterization of a representative CsI(Tl) detector. Left: energy linearity measured using multiple radioactive calibration sources over the energy range 100–1600 keV. Right: relative energy resolution, σ/μ\sigma/\mu, as a function of deposited energy, showing the expected scintillator scaling behavior.

The detector’s energy resolution was also checked; it achieves an σ/μ\sigma/\mu of approximately 7% around 1 MeV (Figure 5b), which is among the best values obtained for PMT-based CsI(Tl) systems. The measured energy resolution follows the expected scintillation-detector behavior,

σμ=aE+c,\frac{\sigma}{\mu} = \frac{a}{\sqrt{E}} + c,

where EE is the energy at which σ\sigma and μ\mu were calculated [32]. Here a=(1.5362±0.147)a = (1.5362 \pm0.147) and b=(0.018±0.007)b = (0.018 \pm0.007), the fit yields an excellent value of R2R^2 of 0.9647. The observed behavior is well described by this relation over the full measured energy range, as illustrated in Figure 5b.

The threshold efficiency of each detector channel was also evaluated to ensure reliable signal acquisition. For every channel, the digitizer threshold was set manually within the CoMPASS software [25] interfaced with the CAEN V1740D digitizer [23]. Hardware threshold values are expressed in units of least significant bits (LSB), which translate to an equivalent voltage in millivolts determined by the dynamic range and the selected energy resolution of the respective channel. The threshold for each channel was optimized to maximize the collection efficiency of signal γ\gamma-rays while simultaneously suppressing the dark-current pulses inherent to PMT operation; only pulses whose amplitude exceeds this threshold are accepted by the digitizer for further analysis.

To quantify the threshold efficiency, two energy spectra were acquired independently for each channel: one recorded at a threshold of 0 LSB, which imposes no amplitude discrimination, and a second recorded at the operationally selected threshold value. The threshold efficiency as a function of energy was then obtained by computing the bin-by-bin ratio of the threshold spectrum to the zero-threshold reference spectrum. The resulting efficiency curves for all channels are presented in Figure 6.

Effective hardware trigger threshold efficiency curves for all detector channels

Figure 6. Effective hardware trigger thresholds for all detector channels, exhibiting the expected error-function-like turn-on behavior [33]. The majority of channels reach full collection efficiency near 200 keV, well below the 889 keV and 1120 keV signal energies of the 46^{46}Sc cascade, confirming that the detector operates with high efficiency across the entire energy region of interest.

The effective hardware trigger thresholds exhibit the expected error-function-like turn-on behavior [33], with the majority of channels reaching full efficiency in the vicinity of 200 keV. A small subset of channels display elevated thresholds near 500 keV, while a few channels attain full efficiency at lower energies near 100 keV, as illustrated in Figure 6. To ensure complete collection of signal γ\gamma-rays across all channels, a common software threshold of 100 keV was applied, thereby retaining all events above the lowest hardware threshold observed in the array.

Since the characteristic γ\gamma-ray energies from the 46^{46}Sc decay (889 keV and 1120 keV) lie well above the threshold region, the detector operates with high efficiency for the signals of interest. These characterization studies provide confidence in the suitability of the detector system for the radioactive-source-based missing-γ\gamma search and associated dark-sector investigations.

Data Analysis

A total of more than 300 hours of data were collected using a 46^{46}Sc radioactive source with the CsI(Tl) detector array and analyzed through the data analysis pipeline.

The analysis pipeline processes the CSV output produced by the DAQ system (§ 3.2) and reconstructs individual decay events by clustering all triggers occurring within a 0.5 μ\mus coincidence window relative to the first triggered channel, provided that the recorded energy exceeds the software threshold of 100 keV (the determination of the coincidence window is discussed in § 8.2). A single decay event may therefore contain energy deposits recorded in multiple detector crystals.

From this event-building procedure, two key observables are constructed. The first is the event multiplicity, defined as the number of detector channels registering a trigger within a given decay event. Physically, this quantity reflects the number of correlated energy depositions arising from particles emitted in a single radioactive decay. The second is the co-added energy spectrum, defined as the sum of the energies deposited in all triggered crystals associated with the same decay event. These observables provide complementary information on the event topology and energy deposition pattern and form the basis of the missing-γ\gamma search described in the following sections.

Source-free Background Data Analysis

Data were first collected without any radioactive source to characterize the environmental background. The resulting energy spectrum, shown in Figure 7, serves as a reference for estimating the environmental background contribution to the rare-event search. Application of the active veto reduces the background event rate by more than 15%, demonstrating the effectiveness of the veto system in suppressing cosmic-ray-induced and other residual background events.

The event multiplicity distribution for the background dataset is shown in Figure 8. As expected, the distribution is dominated by multiplicity-one events, reflecting the largely uncorrelated nature of environmental background interactions. Events with higher multiplicities arise primarily from Compton scattering of higher-energy γ\gamma rays across multiple detector crystals. This distribution provides an important reference for comparison with source-induced events, which frequently produce correlated energy depositions in multiple detectors.

Event multiplicity distribution recorded without an external radioactive source

Figure 8. Event multiplicity distribution recorded without an external radioactive source. The spectrum is dominated by multiplicity-one events, consistent with uncorrelated environmental background interactions.

Preliminary data analysis with 22Na{}^{22}\mathrm{Na}

The 22Na{}^{22}\mathrm{Na} source was used to verify detector functionality, containment efficiency and analysis pipeline. In each 22Na{}^{22}\mathrm{Na} decay, a positron is emitted in coincidence with a 1275 keV γ\gamma ray. The positron subsequently annihilates with an electron, producing two additional 511 keV γ\gamma rays, such that a single decay can yield up to three γ\gammas. Although all three γ\gamma rays are emitted nearly simultaneously on the nanosecond timescale [34], the corresponding detector triggers are spread over longer times due to the relatively long decay time of the CsI(Tl) scintillator [24] and the trigger efficiency of the digitizer. This makes the choice of an appropriate time coincidence window a critical aspect of the analysis (discussed in § 8.2).

Event multiplicity is shown in the Figure 9. This distribution peaks at a multiplicity of three, consistent with the emission of three γ\gamma rays in the 22Na{}^{22}\mathrm{Na} decay cascade. In some cases, more than three signals are also observed when one or more γ\gamma rays undergo Compton scattering within the detector. Conversely, fewer signals may be recorded if γ\gamma rays escape without interaction or deposit energy below the trigger threshold. Events with fewer than the expected number of signals constitute potential signal candidates and are discussed in detail in later sections.

Event multiplicity distribution for the detector array with a ${}^{22}\mathrm{Na}$ source

Figure 9. Event multiplicity distribution for the detector array with a 22Na{}^{22}\mathrm{Na} source. The peak at multiplicity 3 corresponds to the three-particle decay signature of 22Na{}^{22}\mathrm{Na}, consisting of one 1275 keV γ\gamma ray and two 511 keV annihilation γ\gamma rays. Events with higher multiplicities arise primarily from Compton scattering of γ\gamma rays across multiple detector crystals, while lower multiplicities result from incomplete energy containment, detector inefficiencies, accidental background events, or rare decay topologies.

In addition, the co-added energy spectrum as shown in the Figure 10 shows good agreement with the expected spectrum, providing further confidence to proceed with measurements using the 46Sc{}^{46}\mathrm{Sc} radioactive source. The peak at 2297 keV corresponds to full energy absorption of all three γ\gamma rays (1275 + 511 + 511 keV), while the remaining peaks arise from partial containment. The 511 keV peak represents events in which only one annihilation γ\gamma is detected; the 1022 keV peak corresponds to detection of both 511 keV γ\gammas alone; the 1275 keV peak arises either from escape of both annihilation γ\gammas or from the alternate 22Na{}^{22}\mathrm{Na} branch emitting a single 1275 keV γ\gamma; and the 1786 keV peak corresponds to escape of one 511 keV γ\gamma. As seen in Figure 10, as the effective detector volume increases from the inner 3×33 \times3 to the full 5×55 \times5 array, escape peaks are suppressed and the total absorption peak becomes more prominent, consistent with improved γ\gamma-ray containment. Additional details of the 22Na^{22}\mathrm{Na} studies are provided in the Appendix 8.3.

Co-added energy spectrum of $^{22}\mathrm{Na}$ for the inner $3 \times 3$ detector channels and the full $5 \times 5$ detector array

Figure 10. Co-added energy spectrum of 22Na^{22}\mathrm{Na} for the inner 3×33 \times3 detector channels (top) and the full 5×55 \times5 detector array (bottom). The orange curve represents the experimental data with the veto rejection cut applied, while the blue curve shows the spectrum without the veto cut. The 5×55 \times5 configuration exhibits a more pronounced full-energy absorption peak at 2297 keV, whereas the 3×33 \times3 configuration shows additional structures associated with partial energy containment and escape events.

DAQ Validation for 46Sc^{46}\mathrm{Sc}

Using the defined coincidence window of 0.5 μs0.5\ \mu\mathrm{s} and the previously defined energy threshold of 100 keV for all detector channels, the event multiplicity distribution was constructed and is shown in Figure 11. The distribution peaks at a multiplicity of two, consistent with the expected emission of two γ\gamma rays in the 46Sc^{46}\mathrm{Sc} decay cascade. This agreement confirms again the proper functioning of both the detector system and the analysis framework. Events with multiplicity greater than two arise primarily from Compton scattering, in which a γ\gamma ray deposits energy in multiple detector channels. Events with multiplicity equal to one define the signal region and include contributions from detector containment losses, environmental backgrounds, and potential exotic decay signatures. The composition of this region is quantified in the following sections.

Event multiplicity distribution for the detector array with $^{46}\mathrm{Sc}$ source

Figure 11. Event multiplicity distribution for the detector array with 46Sc^{46}\mathrm{Sc} source. The peak at multiplicity 2 corresponds to the two-particle decay of the radioactive source, while higher multiplicities arise from Compton scattering across multiple crystals.

Calculation of Missing signal with 46Sc^{46}\mathrm{Sc}

To identify potential missing-γ\gamma events, we examine the co-added energy spectrum, defined as the total energy deposited across all detector crystals within a single decay. The spectrum, shown in Figure 12, exhibits a clear total-absorption peak at approximately 2 MeV (889 keV + 1120 keV) and single-γ\gamma absorption peaks at 889 keV and 1120 keV, corresponding to events in which one γ\gamma ray is detected while the other escapes the detector volume. Good agreement is observed between data and Monte Carlo simulation, with the detector modelling and simulation validation described in Sec. 5.1.2. The single-γ\gamma peaks provide the baseline for the missing-γ\gamma search, as discussed in the following sections.

Co-added energy spectrum of $^{46}\mathrm{Sc}$ for all detector channels

Figure 12. Co-added energy spectrum of 46Sc^{46}\mathrm{Sc} for all detector channels. The orange curve represents experimental data, while the blue curve shows the simulated spectrum. The peak near ∼2\sim2 MeV corresponds to the total decay energy, whereas lower-energy peaks indicate incomplete containment of the emitted radiation within the detector array. Note: The spectrum shown here is presented without background subtraction, where the measured ambient background spectrum is shown separately in Fig. 7.

For the present analysis, the 1120 keV γ\gamma ray is used as the tagging photon, and the search is performed for the coincident 889 keV γ\gamma ray. This choice is motivated by two considerations. First, the 889 keV peak is susceptible to contamination from Compton-scattered 1120 keV events; using the 889 keV line as the tagging photon could therefore lead to misidentification of Compton-scattered 1120 keV γ\gamma rays as true cascade events. Second, the detector exhibits higher containment efficiency at 889 keV than at 1120 keV. Taken together, these considerations make the 1120 keV line the more reliable tagging channel for this analysis.

While Figure 12 shows good overall agreement between data and simulation, discrepancies are observed in several regions, most notably at the 1120 keV single-γ\gamma absorption peak, which corresponds to events with a missing 889 keV γ\gamma. The number of missing-γ\gamma events is estimated by counting events within the 1120 keV±1.5σ1120\ \mathrm{keV} \pm1.5\sigma region of the co-added energy spectrum across all channels, where the choice of 1.5σ1.5\sigma is motivated by the separation between the 889 keV and 1120 keV peaks. For a total exposure of 7.54×1097.54 \times10^{9} 46Sc^{46}\mathrm{Sc} decays, the data yield $690.19 \pm 25.6×10625.6 \times10^{6} (9.15%) missing-γ\gamma events, while the simulation predicts 565.27±9.08×106565.27 \pm9.08 \times10^{6} (7.49%) events from standard processes, leaving an unaccounted excess of 125.08±27.2×106125.08 \pm27.2 \times10^{6} (1.65%) events beyond detector containment alone. The background contributions to this excess are evaluated quantitatively in Sec. 5.

Statistical and Systematic Uncertainty

The statistical uncertainty on the missing-γ\gamma count is given by σstat=NB\sigma_{\mathrm{stat}} = \sqrt{N_B} [35], where NBN_B is the number of events in the signal region. For the present dataset, this yields a statistical uncertainty of 0.00035%, which is negligible owing to the large event statistics accumulated over the full data-taking period. Systematic uncertainties were evaluated using a cut-variation approach [36], in which each analysis variable is varied around its nominal value and the resulting spread in the missing-γ\gamma count is taken as the systematic uncertainty for that variable. The individual contributions are tested to be mutually independent and are therefore summed in quadrature to obtain the total systematic uncertainty.

The total systematic uncertainty receives four significant contributions. (1) Coincidence window: varying the window from 0.40 to 0.60 μ\mus yields an uncertainty of 0.058%, (2) Energy threshold: varying the software threshold from 60 to 140 keV gives an uncertainty of 0.267%. (3) Calibration shifts: automated peak-finding can misidentify the true photo-peak position by 1–2 ADC bins per channel; shifting each channel independently and repeating the full analysis yields an uncertainty of 0.21%. (4) Detector resolution: the energy-resolution function σ/μ\sigma/\mu was sampled from Gaussian distributions defined by the fit covariance matrix and propagated through the simulation, yielding an uncertainty of 0.12%. Summing these contributions in quadrature gives a total systematic uncertainty of 0.373%. Table 2 summarizes the missing-γ\gamma fraction, the relevant background contributions, and the associated uncertainties. Further details of the uncertainty evaluation are provided in Appendix 8.4.

Background Estimation

Despite the inherent advantages of the missing-γ\gamma search strategy, several background processes can produce signatures indistinguishable from a genuine dark-sector decay. These arise from four primary sources: environmental and intrinsic radioactive backgrounds, which are partially suppressed by the passive shielding and active veto system described in Sec. 3.3; finite detector containment, which causes a fraction of signal γ\gamma-rays to escape the active volume undetected; DAQ-induced pile-up, in which a secondary pulse falling within the post-trigger window is misidentified as a missing-γ\gamma event; and bound-state internal conversion via the BrICC process, which mimics an invisible nuclear transition. Each of these contributions and their combined effect on the missing-γ\gamma signal region is summarized and quantified in the following Sec. 5.

Detector Containment

Accurate evaluation of the detector containment efficiency is critical for estimating the fraction of missing γ\gamma-ray events prior to performing the experiment. As a first-order approximation, the missing fraction was estimated numerically using the detector geometry, the attenuation properties of CsI(Tl), and isotropically emitted 889 keV γ\gamma rays propagating through the crystal array.

Analytical Calculation of Containment

Missing gamma signal is calculated with 1132 keV γ\gamma ray being captured and 889 keV γ\gamma ray being missing. The probability that an 889 keV γ\gamma ray from 46Sc{}^{46}\mathrm{Sc} escapes the CsI(Tl) crystal array without depositing detectable energy was estimated using the Beer–Lambert attenuation law [37],

I=I0e−μx,I = I_0 e^{-\mu x},

where the total mass attenuation coefficient at 889 keV is μ/ρ=6.132×10−2 cm2/g\mu/\rho= 6.132 \times10^{-2}\ \mathrm{cm}^{2}/\mathrm{g}, taken from the NIST XCOM database [38]. Using the CsI(Tl) density ρ=4.51 g/cm3\rho= 4.51\ \mathrm{g}/\mathrm{cm}^{3}, this corresponds to a linear attenuation coefficient of μ=0.2766 cm−1\mu= 0.2766\ \mathrm{cm}^{-1}, or an attenuation length of λ=3.62\lambda= 3.62 cm. A Monte Carlo integration over isotropically distributed 4π4\pi emission directions through the rectangular crystal geometry shown in Fig. 13 yields a geometric missing fraction of (the calculation is shown in further detail in Sec. 8.1)

fmiss=2.773%.f_{\mathrm{miss}} = 2.773\%.

This estimate accounts only for events in which the 889 keV γ\gamma escapes the detector without interacting. Applying the same calculation to the 1120 keV γ\gamma yields an absorption probability of 95.7%. The combined probability of fully absorbing the 1120 keV γ\gamma while the 889 keV γ\gamma escapes is therefore 0.957×0.0277=2.65%0.957 \times0.0277 = 2.65\%. This is an order-of-magnitude estimate of the missing-γ\gamma fraction due to incomplete containment; the actual value may differ substantially because of effects not captured here. In particular, partial energy deposition via Compton scattering of one or both cascade γ\gammas, together with the finite detector resolution and energy threshold, can shift this result. A quantitative treatment would require a full detector-response simulation as explained in the following section.

Monte-Carlo Estimation of the Containment

The detector response is modeled using the GEANT4 simulation framework with an appropriate PhysicsList package for low-energy electromagnetic interactions. Figure 14 illustrates the simulated experimental setup. The simulation geometry is constructed to closely reproduce the physical detector configuration used in the experiment. The detector assembly consists of 26 CsI(Tl) crystals, including two central crystals, arranged in a 5 × 5 square matrix. All detector elements are implemented in GEANT4 using their measured dimensions, positions, and material properties. The 46Sc{}^{46}\mathrm{Sc} radioactive source is placed as given in the detector, matching the experimental configuration.

For this first-order study, the simulation does not include external background sources. Consequently, passive lead shielding and the external plastic veto system are omitted from the geometry. The purpose of this simplified implementation is to estimate the intrinsic fraction of missing γ\gamma-ray events arising solely from detector containment effects for the radioactive source. Detector response effects, including finite energy resolution and threshold efficiency, are incorporated during the analysis stage by convolving the simulated energy spectra with experimentally determined detector response functions.

CategoryNumber (×106\times10^{6}) (%)Systematic (%)Statistical (%)
Total Decay7542.8 (100)––
Missing γ\gamma events – data (Tag 1120 keV)690.19 (9.15)0.340.00035
Containment Inefficiency (Simulation)565.27 (7.49)0.120.00031
Pile-up (in the ROI at 889 keV)70.53 (0.94)0.010.0001
Radioactive background4.16 (0.055)0.0010.00003
BRICC (Conversion electron from a decay)1.21 (0.016)theoretical0.00001
Remaining excess49.04 (0.65)0.373< 0.001

Table 2. Summary of missing γ\gamma contributions and associated uncertainties. Numbers are given relative to the total 7.54×1097.54 \times10^{9} decays, with percentages in parentheses.

Cross-sectional views of the CsI(Tl) detector geometry

Figure 13. Cross-sectional views of the CsI(Tl) detector geometry with the 46^{46}Sc source located at the center. Left: YZ cross-section (view along the X-axis) showing the stacked crystal arrangement along the vertical direction. Right: XY cross-section (view along the Z-axis) showing the 5 × 5 array of 2″ × 2″ CsI(Tl) crystals surrounding the source position. The central region is segmented symmetrically about the source to optimize containment and event reconstruction.

Simulation of the experimental setup

Figure 14. Simulation of the experimental setup. For visualization purposes, a reduced 3 × 3 detector geometry is shown instead of the full 5 × 5 configuration used in the analysis. The 46^{46}Sc radioactive source is placed at the center of the detector array, emitting γ\gamma rays (green) and beta particles (pink). The green trajectories illustrate that a fraction of the emitted γ\gamma rays escape the detector volume without full energy deposition, while others are absorbed within the CsI(Tl) crystals. In contrast, beta particles are rapidly absorbed near the source position because of their comparatively short range in the detector material.

To validate the simulation framework, simulated energy spectra were compared with experimentally measured spectra for individual detector channels. Figure 15 shows the comparison for Channel 16. Figure demonstrates good agreement between simulation and experimental data after including detector response effects. In particular, the two characteristic γ\gamma-ray peaks from 46^{46}Sc at 889 keV and 1120 keV are well reproduced in both position and relative intensity. Similar characteristics in all the other channels also provided enough validation check for similarity between experiment and simulation. This agreement provides confidence in the accuracy of the simulation framework and supports its use for estimating detector containment efficiency and evaluating missing- γ\gamma backgrounds.

The number of missing- γ\gamma events is estimated by counting events within the 1120 keV ±1.5σ\pm1.5\sigma region of the co-added energy spectrum across all channels. For a total exposure corresponding to 7.54×1097.54 \times10^{9} 46^{46}Sc decays, the simulation predicts 565.27±9.08×106565.27 \pm9.08 \times10^{6} events (7.49%) arising from standard detector containment losses, the cascade events in which the 889 keV γ\gamma ray escapes the active volume and goes undetected (Figure 12). This defines the baseline containment inefficiency of the detector and serves as the primary background against which the experimental missing-γ\gamma rate is compared.

Energy spectrum for detector Channel 16

Figure 15. Energy spectrum for detector Channel 16 acquired with the 46^{46}Sc source. The blue histogram represents the experimentally measured spectrum, while the orange histogram shows the simulated spectrum obtained from GEANT4 after convolving with the measured detector energy resolution and threshold efficiency functions. The good agreement between data and simulation validates the detector response model built in GEANT4.

Environmental Background

Environmental radioactivity is discussed in detail in Sec. 3.3. Using the background spectrum shown in Fig. 7, the environmental background contribution to the missing-γ\gamma event sample was estimated to be approximately 0.055% by counting events within the ROI. This contribution is negligible compared to the background from incomplete γ\gamma-ray containment.

Pileup as a Significant Background

A significant background contribution arises from pile-up in the data acquisition system. As described earlier in § 3.2, each recorded waveform spans a total of 10 μ\mus, of which only a central 3 μ\mus window is used for energy integration. If a second radioactive decay occurs within the post-pulse region of an already-triggered channel, the γ\gamma ray from that decay may fall partially or entirely outside the integration window and thus go unrecorded shown in Figure 16. Such events are then misidentified as missing-γ\gamma signatures in the subsequent trigger.

Digitized waveform showing a pile-up event

Figure 16. Digitized waveform showing a pile-up event. The primary pulse is captured within the 3 μ\mus integration window (blue); the pile-up pulse arrives in the post-window region (red) and its energy is not reconstructed, producing a false missing-γ\gamma signature.

The number of pile-up events NpileupN_{\mathrm{pileup}} is estimated by identifying pulses for which the ratio of the 3 μ\mus integral to the full 10 μ\mus integral falls below a characteristic threshold (Figure 17),

Distribution of the ratio between pulse integrals calculated using 3 $\mu$s and 10 $\mu$s integration windows

Figure 17. Distribution of the ratio between pulse integrals calculated using 3 μ\mus and 10 μ\mus integration windows. Two Gaussian-like populations are observed. The dominant second peak corresponds to non-pileup pulses, while the smaller first peak originates from pileup events and dark-current-induced triggers, as confirmed by inspection of representative waveforms from the respective regions of the distribution.

F=∫03 μsV(t) dt∫010 μsV(t) dt(2)\mathcal{F} = \frac{\int_{0}^{3\,\mu\mathrm{s}} V(t)\,\mathrm{d}t}{\int_{0}^{10\,\mu\mathrm{s}} V(t)\,\mathrm{d}t} \tag*{(2)}

where V(t)V(t) is the digitized pulse amplitude as a function of time. Low values of F\mathcal{F} indicate the presence of additional pulse energy outside the standard integration window, consistent with pile-up events. To study the impact of pile-up, a dedicated dataset was acquired in which full waveforms were recorded. This was necessary because in the high-rate data taken with high-intensity radioactive sources, storing complete waveforms for all events in the main data set was not feasible due to severe memory and data throughput limitations. The results obtained from the dedicated waveform dataset were subsequently used to model and extrapolate the pile-up contribution in the main experimental data.

In this dedicated dataset, the full waveform associated with each triggered event was recorded in addition to the reconstructed energy, detector channel, and trigger timing information. A representative subset of datasets containing full pulse information was successfully collected and analyzed to quantify the contribution of pile-up events. To identify such events, a pile-up filter was defined as the ratio of the pulse integral (PIR) calculated over a 3 μ\mus window to that calculated over a 10 μ\mus window, as illustrated in Figure 17 for Channel 24.

The resulting distribution exhibits two distinct gaussian populations. Events with high PIR values correspond to clean, well-contained scintillation pulses, while events with low PIR values arise primarily from either low-energy electronic noise (dark currents) or pile-up events, as confirmed through waveform inspection. Using the low filter value and an energy threshold of 100 keV, pile-up pulses were collected. To estimate the contribution of pile-up to the missing-γ\gamma signal region, the number of events within the energy window 889 keV ±\pm 1.5σ\sigma was evaluated. Finally contribution for all the channels added together to estimate the exact pileup value contribution for this run.

To estimate the total number of pile-up events over the full data-taking period of the main data, the time dependence of the radioactive source activity had to be taken into account. Since the activity of the 46Sc{}^{46}\mathrm{Sc} source decreases with time because of radioactive decay, the pile-up rate correspondingly decreases during the experiment. To characterize this dependence, dedicated datasets were acquired using source configurations with different activity levels. From these measurements, the pile-up event rate per unit source activity was determined, as shown in Figure 18.

Plot showing the experimentally measured pileup rate for different source activities

Figure 18. The plot shows the experimentally measured pileup rate for different source activities. These measurements are used to estimate the total number of pileup events accumulated over the full data-taking period.

Assuming that pile-up events occur randomly and scale linearly with source activity, this measured rate was integrated over the activity profile of the source throughout the full data-taking period to estimate the total pile-up contribution in the experimental dataset. Using this procedure, the total number of pile-up events was estimated to be 0.95% of the total events which is

(70.53±0.75)×106.(70.53 \pm0.75) \times10^{6}.

After subtracting the estimated pile-up contribution, an excess of approximately

49.04×10649.04 \times10^{6}

events remained, corresponding to roughly 0.65% of the total event population. Further statistical analysis showed that this residual excess is not statistically significant in the context of the expected rare-event signal sensitivity.

Internal Conversion: The BRICC Process

In addition, an intrinsic source of missing-γ\gamma events is the bound-state internal conversion process, characterized by the BRICC (Band-Raman Internal Conversion Coefficients) [39]. In this mechanism, instead of emitting a γ\gamma ray, the excited nucleus transfers its de-excitation energy to an orbital electron via a virtual photon. The electron is then ejected from the atom with kinetic energy equal to the transition energy minus its binding energy. Because the emitted electron has a very short range in matter and is readily absorbed by the surrounding crystal and structural materials, it produces no detectable signal in the CsI(Tl) array. Consequently, whenever the 889 keV transition in 46Sc{}^{46}\mathrm{Sc} undergoes internal conversion rather than γ\gamma emission, the event appears as a genuine missing-γ\gamma signature. Based on the BRICC tabulation [39], the internal conversion coefficient for this transition is approximately 0.016% of additional missing-γ\gamma events to the total background budget.

Summary of the Missing Signal and the Estimated Background

For a total exposure of 7.54×1097.54 \times10^{9} 46Sc{}^{46}\mathrm{Sc} decays, the data yield 690.19×106690.19 \times10^{6} (9.15%) missing-γ\gamma events, of which 565.27×106565.27 \times10^{6} (7.49%) are attributed to detector containment inefficiency, 70.53×10670.53 \times10^{6} (0.94%) to DAQ-induced pile-up, 4.16×1064.16 \times10^{6} (0.055%) to environmental and radioactive backgrounds, and 1.21×1061.21 \times10^{6} (0.016%) to bound-state internal conversion via the BRICC process. After subtracting all identified background contributions, a remaining excess of 49.04×10649.04 \times10^{6} (0.65%) events is observed. The total systematic uncertainty is 0.373%, dominated by the software energy threshold (0.267%) and calibration shifts (0.21%), while the statistical uncertainty is negligible at < 0.001%. The corresponding test statistic and sensitivity plot are drawn in the following section.

Conclusion And Experimental Reach

Exclusion Plot

The statistical significance of the observed excess was evaluated using the test statistic [40]

Z=Sσtotal,(3)Z = \frac{S}{\sigma_{\mathrm{total}}}, \tag*{(3)}

where the total uncertainty is defined as

σtotal=σstat2+σsys2.(4)\sigma_{\mathrm{total}} = \sqrt{\sigma_{\mathrm{stat}}^{2} + \sigma_{\mathrm{sys}}^{2}}. \tag*{(4)}

Using the number of decays, the estimated backgrounds, and the systematic uncertainties discussed above, the resulting significance is

Z=1.73.Z = 1.73.

This value falls well below the conventional thresholds typically used in rare-event searches: Z>5Z > 5 corresponds to a discovery claim, while Z>3Z > 3 is considered evidence for a possible signal. Thus, the present measurement does not indicate a statistically significant deviation attributable to new physics.

Sensitivity plot is drawn using CLs_{\mathrm{s}} (Confidence Level) method. These are obtained by computing the exotic branching fraction for each dark-sector model using the corresponding theoretical expressions (Tab. 1). The fundamental condition for sensitivity is that the expected signal yield equals the experimental threshold,

BRDM×ndecay=θ,(5)\mathrm{BR}_{\mathrm{DM}} \times n_{\mathrm{decay}} = \theta, \tag*{(5)}

where θ\theta depends on the statistical treatment of backgrounds. In the zero-background limit, the Poisson process gives θ=2.3\theta= 2.3 at 90% C.L. In the presence of background and systematics, the Gaussian approximation yields θ=1.28σtotal\theta= 1.28\sigma_{\mathrm{total}}, corresponding to a 90% C.L. exclusion with mass and gpg_{p}, ϵ\epsilon, faγf_{a\gamma} as degrees of freedom [41].

The sensitivity condition then becomes

Pγ→a(ma,faγ)=θndecay,(6)P_{\gamma\rightarrow a}(m_{a}, f_{a\gamma}) = \frac{\theta}{n_{\mathrm{decay}}}, \tag*{(6)}

where the right-hand side is determined entirely from experimental quantities. Solving Eq. (6) over the (ma,faγ)(m_{a}, f_{a\gamma}) parameter space exclusion plots is drawn for ALPs, Dark Scalar and Dark Photons.

CandidatesMass (MeV)Coupling ConstantPresent LimitFuture Limit
Dark Scalar0.1 - 1gpg_{p}>8e−2> 8e - 2>1e−8> 1e - 8
Dark Photon0.1 - 1ϵ\epsilon>1e−1> 1e - 1>1e−8> 1e - 8
ALP0.1 - 1faγ(GeV−1)f_{a\gamma}(GeV^{-1})>6e−1> 6e - 1>1e−7> 1e - 7

Table 3. Exploration for parameter space for different Dark-Matter candidates such as Dark Scalar, Dark Photon and ALPs.

The exclusion limits obtained in this study are shown in Fig. 19, probing regions of parameter space for axion-like particles (ALPs), scalar dark matter, and dark photons. The parameter-space ranges investigated in this work are summarized in Tab. 3.

ties.

Future improvements in detector mass, containment efficiency, background suppression, and systematic uncertainty control are expected to substantially enhance the experimental sensitivity, extending the reach into previously unexplored regions of parameter space and potentially probing areas that are currently constrained only by astrophysical observations.

ALPs

(a)

Scalar dark matter

(b)

Dark Photon

(c)

Figure 19. Exclusion limits for different dark-sector models: (a) ALPs, (b) scalar dark matter, and (c) dark photons. The solid region represents the currently excluded parameter space. The dash–dot line shows the projected sensitivity with improved systematic uncertainties, while the dashed line indicates the expected reach of a future ton-scale detector with enhanced systematics and containment efficiency. The proposed future experiments aim to probe the remaining unexplored parameter space.

Using the first-generation experiment with approximately 100 kg of CsI(Tl) scintillators, the excluded region for ALPs is relatively limited and largely overlaps with parameter space previously explored by beam-dump experiments. In the cases of scalar dark matter and dark photons, the excluded regions overlap with existing constraints from the NA64 experiment [42]. Nevertheless, these results validate the missing-gamma technique and demonstrate the feasibility of the experimental approach. The current sensitivity is primarily limited by detector containment, residual backgrounds, and systematic uncertainties.

Future Outlook

The primary limitations of the present experiment arise from limited γ\gamma-ray containment and relatively large systematic uncertainties, which motivate the design of the next-generation upgrade. An upgraded, ton-scale experiment is currently under development and is expected to be one of the leading lab-based experiment to extend the sensitivity into previously unexplored regions of parameter space and broad, untested regions for scalar dark matter. This study highlights the importance of reducing systematic uncertainties, which are dominated by the energy threshold and the choice of coincidence window. The energy threshold is strongly influenced by the use of photomultiplier tubes (PMTs), whose gain instabilities can lead to threshold fluctuations; replacing PMTs with silicon photomultipliers (SiPMs) is expected to provide improved gain stability and enable lower, more uniform thresholds. In addition, systematic uncertainties associated with the coincidence window can be further reduced through optimized analysis strategies and faster data processing using C++-based algorithms.

Additional upgrades under consideration include the use of high-purity CsI crystals with faster decay time (∼20\sim20 ns), which could increase statistics by up to two orders of magnitude, and the deployment of auxiliary β\beta detectors to improve tagging efficiency and coincidence rejection. Together, these developments will lay the foundation for the next-generation program aimed at exploring rare processes in nuclear γ\gamma cascades with unprecedented sensitivity. Furthermore, alternative radioactive sources such as 90^{90}Nb, 60^{60}Co, and 24^{24}Na are being evaluated as substitutes for 46^{46}Sc to probe different dark matter mass ranges and other dark sector candidates [12].

The detector is also being relocated to Los Alamos National Laboratory (LANL) to facilitate axion searches in a beam dump environment. A similar detector setup is planned for deployment at Oak Ridge National Laboratory (ORNL) to operate in the HFIR 85 MW reactor. This versatile detector system is designed to be compatible with a wide range of dark matter search environments, including reactor-based, beam-dump, and radioactive source experiments.

Acknowledgements

The work presented in this paper has been supported by LANL - TRIAD and DOE Grant Nos DE-SC0018981. We also acknowledge Mitchell Institute and the Nuclear Science and Engineering Center for providing institutional support with lab space and other required facilities to carry out experimentation.

Appendices

Missing Fraction Calculation

The missing fraction fmissf_{\mathrm{miss}} for the 889 keV γ\gamma ray from 46Sc{}^{46}\mathrm{Sc} was computed as follows. The total mass attenuation coefficient of CsI(Tl) at 889 keV, excluding coherent (Rayleigh) scattering which deposits negligible energy, is μ/ρ=6.132×10−2 cm2/g\mu/\rho= 6.132 \times10^{-2}\ \mathrm{cm}^{2}/\mathrm{g} [38]. Combined with the density ρ=4.51 g/cm3\rho= 4.51\ \mathrm{g}/\mathrm{cm}^{3}, this gives a linear attenuation coefficient μ=0.2766 cm−1\mu= 0.2766\ \mathrm{cm}^{-1} and mean free path λ=3.62 cm\lambda= 3.62\ \mathrm{cm}. The 46Sc{}^{46}\mathrm{Sc} point source is located at the origin of a rectangular CsI(Tl) array with face distances of ±10.16 cm\pm10.16\ \mathrm{cm} in XX and YY, +18.29 cm+18.29\ \mathrm{cm} in +Z+Z, and −12.19 cm-12.19\ \mathrm{cm} in −Z-Z. For an isotropically emitted γ\gamma in direction (θ,ϕ)(\theta,\phi), the path length to the nearest crystal face is

dmin⁡(θ,ϕ)=min⁡(10.16∣sin⁡θcos⁡ϕ∣,10.16∣sin⁡θsin⁡ϕ∣,zface∣cos⁡θ∣),(7)d_{\min}(\theta,\phi)=\min\left(\frac{10.16}{\lvert\sin\theta\cos\phi\rvert},\frac{10.16}{\lvert\sin\theta\sin\phi\rvert},\frac{z_{\mathrm{face}}}{\lvert\cos\theta\rvert}\right), \tag*{(7)}

where zface=18.29 cmz_{\mathrm{face}} = 18.29\ \mathrm{cm} for cos⁡θ>0\cos\theta> 0 and 12.19 cm12.19\ \mathrm{cm} otherwise. The escape probability for that direction is e−μdmin⁡e^{-\mu d_{\min}}, and the total missing fraction is the 4π4\pi-averaged escape probability,

fmiss=14π∫02πdϕ∫0πsin⁡θ e−μdmin⁡(θ,ϕ) dθ.(8)f_{\mathrm{miss}}=\frac{1}{4\pi}\int_{0}^{2\pi}d\phi\int_{0}^{\pi}\sin\theta\,e^{-\mu d_{\min}(\theta,\phi)}\,d\theta. \tag*{(8)}

This integral was evaluated numerically by Monte Carlo sampling of 2×1072 \times10^{7} uniformly distributed directions on the unit sphere, yielding fmiss=(2.773±0.004)%f_{\mathrm{miss}}=(2.773 \pm0.004)\%. This represents a geometric lower bound; the (−Z-Z, 12.19 cm12.19\ \mathrm{cm}) face corresponds to ∼3.4\sim3.4 mean free paths and the lateral faces (±X\pm X, ±Y\pm Y, 10.16 cm10.16\ \mathrm{cm}) to ∼2.8\sim2.8 mean free paths, so exponential attenuation strongly suppresses escape in all directions.

Estimation of Time Coincidence Window

To associate all signals originating from a single physical decay, a coincidence window is defined in the analysis. Events occurring within this window are grouped together and treated as a single decay event, ensuring that correlated multi-hit signals are not mistakenly identified as independent events.

The distribution of time differences between successive hits exhibits a pronounced excess at short time separations, indicating correlated signals originating from the same decay, while larger separations are consistent with uncorrelated background events.

Based on the observed event clustering below 0.5 μ\mus (Figure 20), the 16 ns digitizer time resolution, the ∼3 μ\sim3\ \mus CsI(Tl) decay time constant, and the measured pulse rise times, a coincidence window of 0.5 μ\mus was adopted for event grouping. Any loss of correlated events outside this window was treated as a systematic uncertainty and propagated through the final analysis.

Histogram of the time differences between consecutive events

Figure 20. Histogram of the time differences between consecutive events. A pronounced exponential component is observed for time intervals below 0.5 μ\mus, corresponding to correlated radioactive decay events, while the approximately flat distribution at larger time differences is consistent with uncorrelated background events. Panel (a) shows the event rate on a semilogarithmic scale for time differences between 0 and 5 μ\mus. Panel (b) presents the same distribution on a linear scale for 0–0.5 μ\mus (top) and 0.5–5 μ\mus (bottom).

Preliminary Analysis with 22Na{}^{22}\mathrm{Na}

While a comprehensive analysis of the 22Na{}^{22}\mathrm{Na} data is presented in Sec. 4.2, an additional comparison is included here to further demonstrate detector performance and validate the Monte Carlo simulation.

A detailed examination was performed for events within 2-detector multiplicity category, defined as cases in which exactly two crystals triggered within the coincidence window. Figure 21 shows a two-dimensional histogram of the energy deposited in the first triggered detector versus that in the second triggered detector. Distinct clusters are observed at the expected energy combinations (511, 511), (511, 1275), (511, 1786), (1275, 511), and (1786, 511) keV, corresponding to the possible γ\gamma-ray combinations produced in 22Na{}^{22}\mathrm{Na} decays.

Figure 21. Two-detector shared-energy spectra for coincident events in 22Na{}^{22}\mathrm{Na} data (a) and Monte Carlo simulation (b). Distinct clusters corresponding to the expected γ\gamma-ray energy combinations and diagonal bands arising from energy sharing between detectors are reproduced by the simulation, demonstrating good agreement with the measured data and validating the detector response model.

In addition to these clusters, diagonal bands are visible at total energies of 1275 keV, 1786 keV, and 2297 keV, indicating events in which a γ\gamma ray shares its energy between two detectors, most likely through Compton scattering or partial energy deposition. A slight asymmetry is also observed in the distribution, with brighter clusters associated with the first triggered detector. This effect arises because higher-energy γ\gamma rays, such as those at 1275 keV or 1786 keV, tend to produce faster-rising pulses that cross the digitizer threshold earlier, making them more likely to be recorded as the first trigger in the detector pair.

Errors and Uncertainty

A summary of the statistical and systematic uncertainties is provided in § 4.5, while a detailed discussion of their evaluation is presented in this section. In general, the uncertainties associated with the experiment can be broadly classified into two categories:

  • Statistical uncertainty: This quantifies the uncertainty arising from finite data samples and random fluctuations, and it decreases as the amount of collected data increases.

  • Systematic uncertainty: This arises from potential biases in the experimental setup and analysis procedures.

Calculation of Statistical Uncertainty

The statistical uncertainty is straightforward to estimate. For the large data sets considered in this analysis, the statistical fluctuations are well described by a Gaussian distribution and given by the square root of the number of events in each case [35].

σstat=∑iσstat,Bi2\sigma_{\mathrm{stat}} = \sqrt{\sum_i \sigma_{\mathrm{stat},B_i}^{2}}
σstat,Bi=NBi\sigma_{\mathrm{stat},B_i} = \sqrt{N_{B_i}}

Where:

  • σstat,Bi\sigma_{\mathrm{stat},B_i}: Statistical uncertainty from the ithi^{\mathrm{th}} background source.

  • NBiN_{B_i}: Number of missing γ\gamma counts from the background source ithi^{\mathrm{th}}.

CategoryNumber (×106\times10^{6})Statistical Uncertainty (%)
Total Decays7542.8
Data690.193.5×10−43.5 \times10^{-4}
Simulation565.273.1×10−43.1 \times10^{-4}
Pile-up Contribution70.531×10−41 \times10^{-4}
Background Contribution4.163×10−53 \times10^{-5}
BRICC Process1.211×10−51 \times10^{-5}
Excess (Unaccounted)49.04<0.001%< 0.001\%

Table 4. Estimated statistical uncertainties in the missing γ\gamma contributions.

Calculation of Systematic Uncertainty

Several sources of systematic uncertainty were evaluated for this experiment. The dominant contributions are outlined below. Unless otherwise stated, the systematic uncertainties were estimated using the cut-variation method described in § 4.5.

Coincidence window selection:

The choice of the coincidence window is one of the most critical factors influencing the overall systematic uncertainty of the analysis. A coincidence window of 0.5 μs0.5\ \mu\mathrm{s} was selected to group events originating from a single decay, with all subsequent triggers within this interval treated as part of the same decay sequence. An inaccurate choice of this window can significantly affect the extracted missing-γ\gamma fraction.

To estimate the associated systematic uncertainty, the coincidence window was varied to 0.4 μs0.4\ \mu\mathrm{s}, 0.45 μs0.45\ \mu\mathrm{s}, 0.5 μs0.5\ \mu\mathrm{s}, 0.55 μs0.55\ \mu\mathrm{s}, and 0.6 μs0.6\ \mu\mathrm{s}, and the full analysis chain was repeated for each case. Owing to computational constraints, this study was carried out using a representative 15-minute subset of the data. The resulting spread in the number of missing-γ\gamma events was quantified, and the standard deviation was taken as the systematic uncertainty. This corresponds to a standard deviation of 3217.15 events, or 0.058%, relative to a total of 5,517,417 decay events in the 15-minute data set.

Time Window (μs\mu\mathrm{s})Missing γ\gamma Counts
0.40525088
0.45522753
0.50520570
0.55518348
0.60515918

Table 5. Event counts as a function of coincidence time window.

Energy threshold:

A similar procedure was followed to evaluate the systematic uncertainty associated with the software energy threshold. The analysis was repeated with threshold values of 60 keV, 80 keV, 100 keV, 120 keV, and 140 keV. The resulting variation in the missing-γ\gamma counts at the end of the analysis pipeline was used to quantify the associated uncertainty. The standard deviation of the missing-γ\gamma counts is 14,749.84 events, corresponding to 0.267% of the total 5,517,417 decay events in the 15-minute data set.

Calibration shifts:

Energy Threshold (keV)Counts
60550392
80550080
100520570
120520073
140519757

Table 6. Event counts as a function of energy threshold.

Energy calibration for all detector channels was performed using an automated procedure that smooths the energy spectrum and identifies the first peak in the 46Sc^{46}\mathrm{Sc} spectrum. In some cases, the identified peak position deviates by 1–2 ADC bins from the true peak, which can affect the extracted missing-γ\gamma yield.

To estimate the associated systematic uncertainty, the calibration peak for each channel was manually shifted by ±(1\pm(1–2)2) ADC bins relative to the automatically determined position, and the full analysis pipeline was repeated for each variation. The resulting spread in the missing-γ\gamma counts was used to determine the systematic uncertainty due to calibration shifts. The individual channel contributions were then combined using a weighted quadrature sum,

Total uncertainty=∑ipi⋅σi2,(9)\text{Total uncertainty} = \sqrt{\sum_i p_i \cdot\sigma_i^2}, \tag*{(9)}

where pip_i is the relative probability of each channel contributing to the detected events and σi\sigma_i is the corresponding channel-wise systematic uncertainty. The total systematic uncertainty from calibration shifts was found to be 0.21%.

Variation of the detector resolution:

Detector energy resolution is an important input to the simulation and can influence the estimated contribution of missing-γ\gamma events. The resolution was characterized by evaluating σ/μ\sigma/\mu at multiple energies using calibrated radioactive sources and fitting these values with a resolution model. Variations in the fitted parameters modify the simulated energy spectra for individual channels and can therefore affect their contribution to the missing-γ\gamma signal.

To quantify this effect, the resolution-function parameters were varied by sampling Gaussian random distributions defined by the fit covariance matrix (Figure 22). This procedure induces small shifts in the simulated energy response, as observed in the co-added energy spectrum for individual channels, and was repeated independently for all channels. The resulting variations in the missing-γ\gamma counts were used to estimate the corresponding systematic uncertainty.

Variation of $\sigma/\mu$ as a function of energy for a representative channel

Figure 22. Variation of σ/μ\sigma/\mu as a function of energy for a representative channel (Channel 17). Variation 0 corresponds to the nominal fit to the measured σ/μ\sigma/\mu data points, while the remaining variations are generated by randomly sampling the fit parameters within their covariance-defined uncertainties, yielding a set of perturbed σ/μ\sigma/\mu curves used to propagate the resolution uncertainty into the missing-γ\gamma yield.

As in the previous cases, the channel-wise uncertainties were combined using a weighted quadrature sum,

Total uncertainty=∑ipi⋅σi2\text{Total uncertainty} = \sqrt{\sum_i p_i \cdot\sigma_i^2}

where pip_i is the relative contribution of each channel to the total number of detected events and σi\sigma_i is the systematic uncertainty in the missing-γ\gamma yield for that channel. The total systematic uncertainty arising from detector resolution variations was found to be 0.12%.

The total systematic uncertainty was obtained by summing the individual contributions in quadrature, resulting in an overall systematic uncertainty of (0.373%).

Effect of $\sigma/\mu$ variation on the total co-added energy spectrum

Figure 23. Effect of σ/μ\sigma/\mu variation on the total co-added energy spectrum. Each curve corresponds to a distinct variation of the resolution function, illustrating the resulting shift in the total absorption peak and the consequent change in the reconstructed missing-γ\gamma count. The spread across variations is used to quantify the systematic uncertainty attributed to detector energy resolution, as described in the text.

References

  1. [1]F. Zwicky, Die Rotverschiebung von extragalaktischen Nebeln, Helv. Phys. Acta 6 (1933) 110–127. doi:10.1007/s10714-008-0707-4.
  2. [2]F. Zwicky, 107. On the Masses of Nebulae and of Clusters of Nebulae, Harvard University Press, Cambridge, MA and London, England, 1979, pp. 729–737 [cited 2024-05-27]. doi:doi:10.4159/harvard.9780674366688.c115. URL https://doi.org/10.4159/harvard.9780674366688.c115
  3. [3]V. C. Rubin, J. W. K. Ford, Rotation of the andromeda nebula from a spectroscopic survey of emission regions, Astrophys. J. 159 (1970) 379–403.DOI
  4. [4]C. Alcock, C. W. Akerlof, R. Allsman, T. Axelrod, D. Bennett, S. Chan, K. Cook, K. C. Freeman, K. Griest, S. L. Marshall, et al., Possible gravitational microlensing of a star in the large magellanic cloud, nature 365 (6447) (1993) 621–623.DOI
  5. [5]R. D. Peccei, H. R. Quinn, Constraints imposed by CP conservation in the presence of pseudoparticles, Phys. Rev. D 16 (1977) 1791–1797. doi:10.1103/PhysRevD.16.1791. URL https://link.aps.org/doi/10.1103/PhysRevD.16.1791DOI
  6. [6]S. Weinberg, A new light boson?, Phys. Rev. Lett. 40 (1978) 223–226. doi:10.1103/PhysRevLett.40.223. URL https://link.aps.org/doi/10.1103/PhysRevLett.40.223DOI
  7. [7]F. Wilczek, Problem of strong p and t invariance in the presence of instantons, Phys. Rev. Lett. 40 (1978) 279–282. doi:10.1103/PhysRevLett.40.279. URL https://link.aps.org/doi/10.1103/PhysRevLett.40.279DOI
  8. [8]P. W. Graham, D. E. Kaplan, S. Rajendran, Cosmological relaxation of the electroweak scale, Physical Review Letters 115 (22) (Nov. 2015). doi:10.1103/physrevlett.115.221801. URL http://dx.doi.org/10.1103/PhysRevLett.115.221801
  9. [9]P. W. Graham, D. E. Kaplan, S. Rajendran, Relaxation of the cosmological constant, Physical Review D 100 (1) (Jul. 2019). doi:10.1103/physrevd.100.015048. URL http://dx.doi.org/10.1103/PhysRevD.100.015048
  10. [10]K. Mori, T. Takiwaki, K. Kotake, S. Horiuchi, Shock revival in core-collapse supernovae assisted by heavy axionlike particles, Phys. Rev. D 105 (6) (2022) 063009. arXiv:2112.03613, doi:10.1103/PhysRevD.105.063009.DOI
  11. [11]J. B. Dent, B. Dutta, A. Jastram, D. Kim, A. Kubik, R. Mahapatra, S. Rajendran, H. Ramani, A. Thompson, S. Verma, Pathfinder for a high statistics search for missing energy in gamma cascades, Physical Review D 105 (1) (Jan. 2022). doi:10.1103/PhysRevD.105.015030. URL http://dx.doi.org/10.1103/PhysRevD.105.015030
  12. [12]G. Benato, A. Drobizhev, S. Rajendran, H. Ramani, Invisible decay modes in nuclear gamma cascades, Phys. Rev. D 99 (2019) 035025. doi:10.1103/PhysRevD.99.035025. URL https://link.aps.org/doi/10.1103/PhysRevD.99.035025DOI
  13. [13]K. N. Vincent, S. Mahapatra, I. Poddar, S. Verma, Improving the resolution and light yield in csi(tl) scintillators, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1056 (2023) 168624. doi: https://doi.org/10.1016/j.nima.2023.168624. URL https://www.sciencedirect.com/science/article/pii/S0168900223006149
  14. [14]E. Blucher, B. Gittelman, B. K. Heltsley, J. Kandaswamy, R. V. Kowalewski, Y. Kubota, N. B. Mistry, S. Stone, A. Bean, Tests of Cesium Iodide Crystals for an Electromagnetic Calorimeter, Nucl. Instrum. Meth. A 249 (1986) 201. doi:10.1016/0168-9002(86)90669-8.
  15. [15]S. D. Verma, Fabrication of cdms dark matter detectors using bi-layer lift-off technique and detection of axions/axion like particles (alps) using csi(tl) scintillator detectors, Ph.D. thesis, Texas A-M (2022).
  16. [16]Thor Labs, High-Reflectance PTFE Sheets, available at: https://www.thorlabs.com/high-reflectance-ptfe-sheets?tabName=Overview (2020).
  17. [17]3M™, 3M™ EMI Aluminum Foil Shielding Tape 1170, available at: https://www.3m.com/3M/en_US/p/d/b00041317/ (2021).
  18. [18]Eljen Technology, EJ-550 and EJ-552 Optical Coupling Compounds: Technical Datasheet, Technical report, Eljen Technology, accessed: 2025-02-18 (2020). URL https://eljentechnology.com/products/accessories/ej-550-ej-552
  19. [19]ET Enterprises, PMT 9954B Response Curve, available at: https://et-enterprises.com/images/data_sheets/9954B.pdf (Accessed: 2025-02-18) (2025).
  20. [20]CAEN HV SY5527LC, SY5527LC Universal Multi-channel Power Supply System (Low Cost), available at: https://www.caen.it/products/sy5527lc/ (2026).
  21. [21]CAEN HV A7236, 12/24/32 Channel 3.5 kV, 1.5/0.15 mA (4W) Common Floating Return Dual Range Boards, available at: https://www.caen.it/products/a7236/ (2026).
  22. [22]CAEN GECO2020, GEneral COntrol Software for CAEN HV Power Supplies, available at: https://www.caen.it/products/geco2020/ (2026).
  23. [23]CAEN Digitizer V1740D, 64 Channel 12 bit 62.5 MS/s Digitizer supporting DPP-QDC firmware, available at: https://www.caen.it/products/v1740d/ (2026).
  24. [24]GammaData, CsI(tl) and na scintillation material data sheet, https://gammadata.se/wp-content/uploads/2024/01/CsITl-and-Na-data-sheet-v2.pdf, technical datasheet (2024).
  25. [25]CAEN COMPASS, Multiparametric DAQ Software for Physics Applications, available at: https://www.caen.it/products/compass/ (2026).
  26. [26]FIRGELLI Automations, Shielding Thickness Gamma Interactive Calculator, available at: https://www.firgelliauto.com/blogs/calculators/shielding-thickness-gamma-calculator?srsltid=AfmBOorGhw_vWYU4ajgIg9Xp7D1czA3J3YV-BmGr9zisl9dDpdUpQSF (2026).
  27. [27]Luxium Solutions, BC-400, BC-404, BC-408, BC-412, BC-416, available at: https://luxiumsolutions.com/radiation-detection-scintillators/plastic-scintillators/bc400-bc404-bc408-bc412-bc416 (2026).
  28. [28]Z. e. Ahmed, Results from a low-energy analysis of the cdms ii germanium data, Phys. Rev. Lett. 106 (2011) 131302. doi:10.1103/PhysRevLett.106.131302 URL https://link.aps.org/doi/10.1103/PhysRevLett.106.131302DOI
  29. [29]S. Sahoo, S. Verma, M. Mirzakhani, N. Mishra, A. Thompson, S. Maludze, R. Mahapatra, M. Platt, Reactor-based search for axion-like particles using csi(tl) detector (2026). arXiv:2407.14704. URL https://arxiv.org/abs/2407.14704
  30. [30]G. LI, Y. WANG, X. WANG, Adaptive compensation method for photomultiplier tube counting temperature drift, Journal of Measurement Science and Instrumentation 15 (2) (2024) 244–252. doi:10.62756/jmsi.1674-8042.2024025. URL https://www.scipopen.com/article/10.62756/jmsi.1674-8042.2024025DOI
  31. [31]N. Yavuzkanat, M. Şenyiğit, M. Kaplan, Investigation of the gamma-ray efficiency for various scintillation detector systems, Revista Mexicana de Física 68 (06 2022). doi: 10.31349/RevMexFis.68.041201.DOI
  32. [32]E. Aprile, A. et al., Energy resolution and linearity of xenon1t in the mev energy range, The European Physical Journal C 80 (8) (Aug. 2020). doi:10.1140/epjc/s10052-020-8284-0. URL http://dx.doi.org/10.1140/epjc/s10052-020-8284-0
  33. [33]G. Angloher, S. Banik, G. Benato, A. Bento, A. Bertolini, R. Breier, C. Bucci, J. Burkhart, L. Canonica, A. D’Addabbo, S. Conforti Di Lorenzo, L. Einfalt, A. Erb, F. Feilitzsch, N. Iachellini, S. Fichtinger, D. Fuchs, A. Fuss, A. Garai, V. Zema, Results on sub-gev dark matter from a 10 ev threshold cressst-iii silicon detector, Physical Review D 107 (06 2023). doi:10.1103/PhysRevD.107.122003.
  34. [34]S. R. Cherry, J. A. Sorenson, M. E. Phelps, Physics in Nuclear Medicine, 4th Edition, Elsevier, 2012.
  35. [35]F. C. Porter, Confidence intervals for the poisson distribution (2026). arXiv:2509.02852. URL https://arxiv.org/abs/2509.02852
  36. [36]P. K. Sinervo, C. M, Definition and treatment of systematic uncertainties in high energy physics and astrophysics (2025). arXiv:2510.24313. URL https://arxiv.org/abs/2510.24313
  37. [37]M. Baur, N. Uhlmann, T. Pösch el, M. Schröter, Correction of beam hardening in x-ray radiograms, Review of Scientific Instruments 90 (2) (Feb. 2019). doi:10.1063/1.5080540. URL http://dx.doi.org/10.1063/1.5080540
  38. [38]M. Berger, J. Hubbell, S. Seltzer, J. Chang, J. Coursey, R. Sukumar, D. Zucker, K. Olsen, XCOM: Photon cross sections database, NIST Standard Reference Database 8 (XGAM), National Institute of Standards and Technology, available at https://physics.nist.gov/PhysRefData/Xcom/html/xcom1.html (2010).
  39. [39]Australian National University, Bricc conversion coefficient database, available at: https://bricc.anu.edu.au/ (Accessed: October 2025) (2025).
  40. [40]D. Mondal, W. Baker, M. Chaudhuri, J. B. Dent, R. Dey, B. Dutta, V. Iyer, A. Jastram, V. K. S. Kashyap, A. Kubik, K. Lang, R. Mahapatra, S. Maludze, N. Mirabolfathi, M. Mirzakhani, B. Mohanty, H. Neog, J. L. Newstead, M. Platt, S. Sahoo, J. Sander, L. E. Strigari, J. Walker, Cevns search with cryogenic sapphire detectors at miner: Results from the triga reactor data and future sensitivity at hfir (2025). arXiv:2510.15999. URL https://arxiv.org/abs/2510.15999
  41. [41]X. Qian, A. Tan, J. Ling, Y. Nakajima, C. Zhang, The gaussian cl method for searches of new physics, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 827 (2016) 63–78. doi:10.1016/j.nima.2016.04.089. URL http://dx.doi.org/10.1016/j.nima.2016.04.089
  42. [42]Y. M. Andreev, A. Antonov, M. A. A. Torres, D. Banerjee, B. B. Oberhauser, V. Bautin, J. Bernhard, P. Bisio, A. Celentano, N. Charitonidis, P. Crivelli, A. V. Dermenev, S. V. Donskov, R. R. Dusaev, V. N. Frolov, S. V. Gerts​​enberger, S. Girod, S. N. Gninenko, A. V. Ivanov, Y. Kambar, A. E. Karyne​​u, G. Kekelidze, B. Ketzer, D. V. Kipchnikov, M. M. Kirsanov, V. A. Kramarenko, N. V. Krasnikov, S. V. Kuleshov, V. E. Lyubovitskij, A. Marini, L. Marsicano, V. A. Matveev, R. M. Redes, R. M. Yanssen, L. M. Bueno, M. Mongillo, D. V. P​​eskhonov, V. A. Polyakov, B. Radics, K. Salamatin, V. D. Samoylenko, H. Sieber, D. Shchukin, O. Soto, V. O. Tikhomirov, I. T​​lisova, A. N. Toropin, M. Tuzi, P. V. Volkov, I. V. Voronchikhin, J. Zamora-Saá, A. S. Z​​hevlakov, Searching for light dark matter and dark sectors with the na64 experiment at the cern sps (2025). arXiv:2505.14291. URL https://arxiv.org/abs/2505.14291

Paper details

vlnc:W_xqfbgmxh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents