Beyond Helium-3: Instruments for Cosmic-Ray Neutron Sensing based on Boron-10 Neutron Detectors

1. Introduction

Accurate knowledge about soil moisture as a key variable hydrological cycles, is most relevant in understanding weather patterns and agricultural productivity [1,2,3]. Cosmic-ray neutron sensing (CRNS) [4,5,6,7] has become in recent years an established methodology for non-invasive measurements of soil moisture [8]. With its sensitive volume of the top few decimeters of several hectares [9] it bridges an important gap between point measurements and remote sensing products [10,11,12]. The method relies on the inverse relationship between the near-surface cosmogenic neutron flux and the hydrogen content of the surrounding environment, with soil water providing the dominant contribution [13]. Low maintenance requirements and suitability for continuous long-term observations [14] lead to its integration in larger networks [15,16,17]. Instrumentwise the current development focuses on the transition from helium-3-based to helium-3-free instruments, which makes the method more cost-effective and better scalable. This allows the method to advance into fields of application not only in hydrology, but also in irrigation management, the validation of remote sensing data and climate-resilient monitoring strategies.

1.1. Neutron Detectors for CRNS

CRNS instruments are usually mounted 1–2 m above the ground and can be equipped with two types of sensors: one bare detector for the thermal neutron flux and one counter enclosed by ∼25 mm of polyethylene moderator, making the system most sensitive to epithermal-to-fast neutrons [18,19], which show the strongest scaling for soil moisture. Fast and thermal regimes exhibit a much weaker dependence on environmental hydrogen than the epithermal-to-fast range, but might provide useful information on vegetation and biomass dynamics [20,21,22]. The energy sensitivity (response function) from the usual epithermal detector extends, however, into the thermal and fast domains, which leads to artifacts such as the ’road effect’ [23]. Thermal neutron contamination can constitute up to 20 % of the signal [24] and vice versa the bare detector receives a comparable contribution from epithermal/fast neutrons [25]. Technically, systems are well understood with CRNS detectors tested in neutron reference fields are reported to be in good agreement to Monte Carlo simulations [26].

The common neutron converter types used in CRNS are: helium-3 or boron-trifluoride (${}^{10}\mathrm{BF}_3$) and boron-10-lined or lithium-6 gaseous counters as well as lithium-6-loaded scintillators with optical readout. The scarcity of helium-3 motivates adoption of alternative technologies [27,28,29]. The signal response to soil moisture depends on the moderator design, not on the converter itself. The moderator can be modified to partially suppress neutrons from undesired directions [30,31] or thermal contributions [32,33], though at the cost of counting statistics.

1.2. Conventional Detector Designs

Helium-3 acts simultaneously as converter and counting gas. A converted neutron releases 764 keV, shared between a proton (573 keV) and a triton (191 keV) emitted back-to-back. Partial energy loss of one product hitting the tube wall leads to the typical plateau features in the spectrum. Most of the first COSMOS instruments [6] and roving systems [34,35,36,37,38] are helium-3 based. ${}^{10}\mathrm{BF}_3$ also serves as a neutron converter but is also a quencher, requiring an admixture of a counting gas. This limits the absorption efficiency [39,40] and in combination with their toxicity and it impedes a wider deployment.

1.3. Alternative Solutions for Gaseous Detectors

A micrometer-thick coating of ${}^{10}\mathrm{B}$ or ${}^{10}\mathrm{B}_4$C on the inner tube wall is used with conventional counting gases such as Ar:CO₂ [41]. Because only part of each reaction-product track lies in the gas, the resulting continuous energy spectrum necessitates a low-energy cut-off that increases sensitivity to gain fluctuations or threshold drifts [19], imposing stricter requirements on frontend electronics. The comparably low conversion efficiency per layer can be compensated by multi-tube arrangements [42]. With a very similar design Lithium-6 metal can be used as a converter. An increasing number of CRNS sites now use lithium-based instruments [43,44,45].

CRNS instruments need to be optimized for low count rates. In this respect it is important to note that alpha-emitter contamination in tube materials can reach ${10}^{-4}$ cm⁻² s⁻¹ and, with standard alloys, may account for up to one third of all recorded events, since the resulting continuous energy deposition spectrum overlaps with the signal region [46,47]. High-purity materials or inner coatings of such effectively suppress this background.

1.4. Scintillator Technologies

Neutron scintillation detectors require a converter embedded in a scintillator matrix, a lightguide and a photosensitive element [48]. Two strategies are common: a bulk LiF-loaded scintillator of relatively low converter concentration, or a thin ZnS(Ag) coating highly loaded with converter [49,50]. Scintillators respond to muons, gammas and electrons in addition to neutrons. Signal identification can be realized based on pulse shape analysis [51]. Unlike gaseous detectors, plastic scintillators contain hydrogen and therefore act as moderators themselves.

1.5. Measurement Accuracy and Representativeness

The measurement accuracy of Cosmic-Ray Neutron Sensing is primarily subject to the statistical nature of the low-intensity neutron flux. The precision follows counting statistics where the uncertainty is proportional to the inverse square root of the total number of detected neutrons, leading to the general rule that doubling the precision requires a fourfold increase in either measurement duration or count rate. The absolute uncertainty is also dependent on the measured soil moisture value, see Figure 1 (right). Intervals exceeding 12 hours are generally avoided as otherwise other quantities can be limiting factors including the hydrological dynamics itself. Environmental factors such as elevation significantly influence the count rate, which doubles approximately every 900 m.

The spatial representativeness of the signal is characterized by an exponential decrease in sensitivity in both horizontal and vertical directions, which translates to topsoil and objects in close proximity having a disproportionately stronger influence than deeper layers or distant topologies, see Figure 1 (left).

2. Materials and Methods

Microcontrollers have become a practical basis for the development of neutron detector electronics, particularly when used as digitizer and slow-control units in combination with dedicated analog front-end boards, for further examples see also [52,53,54,55].

Table 1. Summary of selected microcontrollers and typical development boards or systems in which they are used.

MCU	Type / Core	Clock	I/O	Voltage	Used also in
ATmega328P	8-bit AVR	16 MHz	14 digital + 8 analog	5 V	Arduino Nano
ATSAM3X8E	32-bit ARM Cortex-M3	84 MHz	54 digital + 12 analog	3.3 V	Arduino Due
			+ 2 DAC
ATmega16U2	8-bit AVR USB MCU	16 MHz	22 GPIO	2.7–5.5 V	Arduino USB port
STM32L412KB	32-bit ARM Cortex-M4F	32 MHz	26 digital + 10 analog	1.71–3.6 V	Nucleo-L412KB
STM32U385RG	32-bit ARM Cortex-M33	32 MHz	51 digital + 17 analog	1.71–3.6 V	Nucleo-U385RG
			+ 2 DAC

Table 1. Summary of selected microcontrollers and typical development boards or systems in which they are used.

MCU	Type / Core	Clock	I/O	Voltage	Used also in
ATmega328P	8-bit AVR	16 MHz	14 digital + 8 analog	5 V	Arduino Nano
ATSAM3X8E	32-bit ARM Cortex-M3	84 MHz	54 digital + 12 analog	3.3 V	Arduino Due
			+ 2 DAC
ATmega16U2	8-bit AVR USB MCU	16 MHz	22 GPIO	2.7–5.5 V	Arduino USB port
STM32L412KB	32-bit ARM Cortex-M4F	32 MHz	26 digital + 10 analog	1.71–3.6 V	Nucleo-L412KB
STM32U385RG	32-bit ARM Cortex-M33	32 MHz	51 digital + 17 analog	1.71–3.6 V	Nucleo-U385RG
			+ 2 DAC

In our proportional-counter readout the architecture was adapted to fully make use of the technical capabilities of such microcontrollers, which allow for example to measure pulse height and pulse length at very low power consumption at low to medium count rates. Other functional elements like high-voltage supply and the analog signal path require separate components and logic.

The first nCatcher generation was implemented on an ATmega328P-based Arduino Nano platform, see also application references in [17,19,47,56,57,58,59,60,61,62]. In the current system generation, the logger relies on the ATSAM3X8E architecture of the Arduino Due, whereas newer detector-side controller concepts are be implemented on STM32 microcontrollers, see for example [63,64].

2.1. Frontend Electronics: nCatcher

The nCatcher readout board, see Figure 2(left) or [47] for the previous version, integrates high-voltage bias generation, up to five-channels with analog amplifiers, pulse discrimination, peak detection and slow control. The design is intended for low-count-rate neutron detectors which require a maximum of 2000 V. Due to its adaptability, signal pulses of helium, boron or lithium-based detectors can be recorded with a pulse height resolution of 1024 channels. The low-voltage supply features digital 5 V for the controller, digital-to-analog converters (DACs), comparator logic and communication interface.

2.1.1. High-Voltage Supply and Bias Distribution

The bias voltage is generated on-board by a dedicated high-voltage module. Its adjustable output voltage is highly stable against temperature drifts and the actual monitor values of voltage and current are provided.

The high-voltage supply board ’nAccelerator’ converts a 5 V low-voltage input into the detector operating voltage. This module was developed due to the challenging demands of CRNS instruments with respect to high temperature stability, low ripple and noise as well as low power consumption.

On the nCatcher board, see Figure 2(left), the high-voltage output is then routed through a passive filtering network and is distributed to the detector channels through individual 10 M$\mathsf{\Omega }$ resistors with a 270 pF decoupling capacitor.

2.1.2. Analog Signal Stage

Each channel has an inverting charge-sensitive preamplifier based on a low-bias-current operational amplifier with clamp diode protection. The signal is integrated by 4.4 M$\mathsf{\Omega }$ with a 1 pF capacitor in parallel leading to a characteristic pulse time of around 4.4µs. The signal then passes a band pass with cutoff frequencies of 24 kHz and 12 kHz and non-inverting amplifier with a gain of 22. One branch is forwarded to the comparator that triggers an output if the analog signal passes a threshold. The time difference between the falling and rising edge is measured as the pulse length. The other branch is fed into an active peak-detector circuit based on a dual operational amplifier, a Schottky diode network and a hold capacitor. The peak value is in that way buffered and read by the microcontroller through an analog input. After conversion, the peak detector is reset by discharging the hold capacitor through an analog switch.

2.1.3. Controller and Interfaces

An STM32 microcontroller serves as the main control unit and interface of the board. It reads the high-voltage monitor outputs, enables or disables the HV module, counts the signal length and reads out the peak-detector via its ADC. The on-board temperature and humidity are recorded in parallel.

For communication with an external system, the board provides a differential I²C interface, which allows for operation with longer cable lengths than foreseen for single-ended I²C and in addition an RS-485 connection. Alternatively, an SDI-12 interface is available for operation with standardized scientific data loggers.

2.2. Pulse Height and Pulse Length Selection

The discrimination of neutron-induced events from gamma-ray background and electronic noise is achieved through a dual-parameter selection of pulse height and pulse length. As boron-10-lined proportional counters have a continuous energy deposition spectrum, one cannot rely on energy thresholds alone as usually the case for helium-3 detectors. The nCatcher electronics analyzes the specific correlation between the peak signal amplitude and the time-over-threshold [19]. This two-dimensional approach relies on a parametrized ’banana-shaped’ signal region characteristic for neutron events and is shown in Figure 3. All events which lie outside the dashed lines are rejected. It has to be noted that due to the color coding the noise region appears to have more events than actually there are - in the inlet one can for example not even see the radionuclide line which stretches out to the right. This procedure allows the instrument to maintain stable performance even under noisy conditions.

2.3. Data Logger

The integration of autonomous sensors into cloud-based networks follows the broader trend of Internet of Things (IoT) applications in environmental monitoring, specially adressing real-time decision support for Smart Farming [65,66]. The logger, see Figure 2(right), serves as the central control, communication and power-management unit of the instrument. Its architecture is not limited to simple data acquisition, but is designed as a complete field controller for autonomous CRNS instruments.

2.3.1. Electronics Design

Its digital core is based on an ATSAM3X8E microcontroller, i.e. a 32-bit ARM Cortex-M3 running with 3.3 V logic in combination with a dedicated ATmega16U2 for USB communication. The controller interfaces the detector frontend electronics, manages auxiliary sensors and communication to peripherals as well as subsequently responsible for data acquisition and transfer.

The logger board features the following components: A DS3231(SN) temperature-stabilized real-time clock with a backup battery. Non-volatile SPI memory is provided by an FRAM and for removable mass storage SD cards can be used. The board furthermore includes a Bosch BME280 sensor for local measurements of pressure, temperature and relative humidity. Communication interfaces are implemented for both local sensors and telemetry. The main peripheral interface is realized by a half-duplex RS-485 transceiver, optionally an I²C connection is provided and for external environmental sensors the logger has an SDI-12 interface.

An onboard Simcom SIM7600 (4G) modem provides cellular communication and GNSS functionality. Alternatively, a Quectel BC95-G (NB-IoT) in combination with an UBlox M[6-9]N GPS module can be configured.

For autonomous field operation, the logger additionally integrates a solar charger for photovoltaic input with battery-backup output.

2.3.2. Logger Functionality

The logger serves as the central control, acquisition, and communication unit of the measurement system. During each acquisition cycle, the logger collects neutron and slow-control data from the nCatcher frontend, supplements the data stream then with local or external sensor feeds, timestamps the data and stores or transmits the resulting records.

In addition to raw data from the neutron detector and environmental sensor information, the logger acquires diagnostic parameters such as high voltage, threshold settings, temperature, humidity, and general status information. A number of external SDI-12 sensors can be configured, these may include meteorological, such as ATMOS14/41 (METER Group), or atmospheric like LI-710 (LI-COR Environmental) and in-situ soil sensors such as SMT-100 (Truebner GmbH) or HygroVUE5, SoilVUE10, TEROS-12 and CS616 (Campbell Scientific) as well as PR2 profile probe (Delta T Devices) and Drill & Drop (Sentek). For sensor references and calibration see [67,68,69,70,71,72,73]. Remote communication can be realised by modems with either UDP, FTP, HTTP POST, MQTT, or optionally LoRa, depending on hardware and firmware configuration.

Figure 4. Schematic of the current data logger components and input/output logic.

Figure 4. Schematic of the current data logger components and input/output logic.

2.3.3. Telemetry and Data Integration

The telemetry framework implemented into the logger is designed for data transmission from the sensor to remote storage systems, either through cellular or local LoRa networks, supporting various protocols such as FTP, MQTT, or UDP for maximizing compatibility with different network infrastructures. Once transmitted, the records are stored into a time-series database, such as InfluxDB, where the data is further processed. This allows for seamless integration with data visualization and analysis tools to facilitate accessing the information through dedicated (API) interfaces or dashboards like Grafana for real-time visualization of time-series and statistical interpretation of soil moisture dynamics, see also Figure 5.

2.4. Instrument Pool

The stationary Stx instrument family comprises several modular CRNS systems that address different requirements in terms of detector size and temporal resolution, see also Figure 6.

The S1 is a slim detector configuration intended primarily for long-term agricultural monitoring at a resolution of around 10 h. The SP presents itself as a larger installation with improved counting statistics and thus better suitability for hourly averages, challenging environments [17,19]. The intermediate S2 and S2+ systems provide a compromise between installation size, time resolution and cost. All these systems are based on helium-3-free boron-converter proportional counters with Ar:CO₂ counting gas and combine one or several detector tube, the same nCatcher readout electronics, logger and autonomous PV/battery supply in a modular outdoor platform [19,64]. Depending on the measurement context different external SDI-12 sensors are adopted. These systems have been installed both in application-oriented agricultural projects such as ADAPTER and in larger scientific monitoring networks [17,64]. In addition, the S2L system with a lithium-foil multi-wire proportional chamber [43] achieves around 3800 cph. All count rates are stated for sea level, 10% soil moisture and average German latitude. In comparison, under the same conditions the Hydroinnova CRS1000 [6] yields around 900 cph.

2.5. Field Deployment

The handling of the Stx instruments is designed for straightforward and autonomous field deployment. For the installation typically a mechanical insertion of a ground screw into the soil is chosen, which for usual soils can be drilled in by hand. Once the base is positioned, the main detector unit is mounted and aligned, see Figure 7. Then, the subsystems are connected to the foldout logger unit: photovoltaic supply, the neutron detector and environmental sensors. Data acquisition parameters and other parameters are managed via a configuration file on the SD card. Upon powering the system, the integrated display allows for directly checking the proper instrument operation, such as current count rates, the high-voltage status and telemetry signal quality.

2.6. Applications and Data Management

The integration of CRNS technology into environmental monitoring networks has significantly advanced hydrological monitoring with the current installations COSMOS [6], COSMOS-UK [16], CosmOz Australia [15], ADAPTER [17] and smaller-scaled networks in India [76], Germany [64] and Ireland [77]. Due to the comparably large footprint, CRNS data can serve as a validation tool of remote sensing products [78,79,80]. Soil moisture data is especially valuable for drought monitoring [81] and early warning systems [82], for the calculation of specialized indices such as the Standardized Precipitation Evapotranspiration Index (SPEI) or the Palmer Drought Severity Index (PDSI) [83]. In agricultural management, accurate information on soil water content allows for the optimization of irrigation practices and water use efficiency [74,84,85]. Furthermore, long-term monitoring [86] supports hydrological research into water flow dynamics, infiltration rates and groundwater recharge, as well as the assessment of climate change impacts on ecosystem resilience [8]. The system integration involves automated data collection by instruments at predefined intervals and subsequent transmission via cellular, LoRa or satellite networks. Data storage, quality control and analysis are performed on cloud platforms.

3. Results

3.1. Thermal Stability of Signal Discrimination

The stability of the nCatcher electronics was evaluated for the S1 system over a period of several years, see Section 3.3, with temperature range from −10 °C to +55 °C.

In a linear multivariate regression analysis, the pulse length exhibited a drift coefficient of −0.02 units/°C. Given the typical signal range of 100 to 350 units, a 30 °C temperature shift results in a variation of less than $0.7$ units ($<0.3\%$). Similarly, the pulse height drift was quantified at 0.046 units/°C. These measured drifts are an order of magnitude smaller than the characteristic width of the neutron ’banana’ region in the PL-PH space. Hence, the analysis of the pulse length and pulse height confirms good thermal stability.

3.2. Power Consumption and Energy Efficiency

Energy efficiency of the system is a critical factor for long-term autarcic operation, not only for dimensioning PV panel and battery, but it also determines whether an instrument can be deployed in challenging environments like forests. The power consumption varies between operational states and could be reduced within each generation. The newer generation (S1p) runs in combination with the optimized nCatcher 4.0 electronics with the nAccelerator high voltage module, which consumes only 2 mA at 12 V. In contrast, the previous S1 system, with the nCatcher 3.x and a CAEN A7508P high-voltage module, requires approximately 8 mA. The total power consumption, see Table 2, refers to a complete setup without external SDI-12 sensors. An SD card is included in the budget, current draws can vary slightly between models though.

For more complex configurations, such as the S2+ (triple analog stage) or the S2L (lithium-based detector), the power requirements increase accordingly. The S2L has a CAEN A7502P module, which adds approximately 4 mA of quiescent current and requires higher output power for its multi-wire proportional chamber, resulting in a total increase of roughly 6 mA compared to the base configuration. Hence, even the more complex Stx system configurations are well within a power budget of less than 0.5 W and the most recent S1p achieves a significant reduction in consumption with the instrument reaching nearly 50 mW.

3.3. Long-Term Operation

We show an exemplary timeseries of a CRNS instrument deployed within an urban environment. Around 40 % of the surface is covered by infrastructure. Located at a former military base most soils had been excavated and back- or refilled. This anthropogenic sediment therefore provides a base with laterally rather homogeneous conditions and average densities of around 1.48 g/cm³, yet with a very high hydraulic conductivity. The challenge in such an environment for in that case the smallest sensor is to resolve the rather fast hydrological response with a significant signal damping. The result, see Figure 8, under these circumstances is a time series which is consistent with the expected behaviour of a stationary CRNS deployed in a heterogeneous urban footprint.

After correction for incoming neutron intensity, atmospheric pressure and air humidity, the neutron signal shows a comparatively smooth temporal evolution, with emphasized links to precipitation-driven wetting and subsequent drying phases. The comparatively low noise level of the corrected neutron data, despite count rates of roughly 700–1000 cph, also indicates that the expected precision for a stationary CRNS, after temporal aggregation, can be achieved. The CRNS soil moisture evaluation reproduces the main seasonal and precipitation-wise soil-moisture dynamics seen by the weighted SMT reference, see also [9,87], but with reduced amplitude. After applying the urban areal correction, see Appendix A, the CRNS-derived soil moisture shows a stronger wetting and drying response and agrees more closely with the weighted in situ reference, which is consistent with the theoretical expectation that sealed surfaces contribute a comparatively invariant neutron background and therefore damp the apparent soil-moisture dynamics of the hydrologically active fraction. As the in-situ SMT-100 sensors [88] do not cover the full support volume an exact overlap is not to be expected. Particularly distinct during the winter period is the appearance of snow: here, the derived CRNS moisture signal rises strongly as a distinct water layer forms ontop of the soil. For the interpretation of snow water equivalent by CRNS we refer to [89,90,91,92,93]. The timeseries demonstrates that the corrected CRNS-derived soil moisture enables the precise measurement of the field-scale dynamics with good temporal resolution.

4. Outlook

4.1. Roving

By mounting a detector on a vehicle the measured area can be extended from several hectares to the square-kilometer scale.

With stationary CRNS being optimized for monitoring within a fixed footprint, mobile applications are carried out to resolve spatial heterogeneities or along regional transects. This technique, commonly referred to as roving, uses the same detection principles as stationary probes but with much larger detectors to maximize the count rate and hence minimize acquisition time. Yet, data typically needs to be corrected for the ’road effect’ [23] or changing soil texture along the track. Roving provides a scalable solution for characterizing soil moisture from the field scale up to mesoscale assessments of entire catchments and basins. This approach has been applied in a variety of environmental contexts [35,36,37,38,94,95,96]. An illustrative example of the spatial coverage achieved during a representative roving campaign of 2 h is presented in Figure 9.

4.2. Monolithic Instrument Architecture: The S1p System

The latest evolution of the S1 series, the new S1p, introduces a major overhaul by transitioning from a modular assembly to a fully monolithic instrument architecture, see Figure 10. Conventional CRNS setups typically rely on a distributed configuration, consisting of a separate logger and battery enclosure connected via external cabling to a detector on a pole. The S1p integrates all main components into a single cylindrical enclosure. The main development of this design is the S1p ’penthouse’ unit. This integrated head houses the new logger, the telemetry modules and the battery as an extension of the detector housing. The solar panel and the environmental sensors (pressure, temperature and humidity) are mounted directly on top of the instrument cap. This design shift removes the need for external cables or mounting solutions. Consequently, the S1p can be deployed as a single unit on a ground screw, which reduces installation time and effort and as a more holistic solution presents itself as a more sleek and cost-effective solution for long-term agricultural and hydrological monitoring.

5. Conclusions

We presented the Stx instrument family as a scalable, cost-effective and helium-3-free solution for Cosmic-Ray Neutron Sensing. The modular architecture relies on boron-10-lined proportional counters with dedicated frontend electronics, data logging and telemetry. The pulse shape analysis provides reliable discrimination of neutron events from gamma-ray background and electronic noise, which is a necessity because of the continuous energy spectrum characteristic of boron-10-lined detectors. Main results are the complete overhaul in electronics and system design. We achieved a significant gain in efficiency with close to 50 mW total system power with the latest S1p generation. This represents a roughly fivefold improvement over the previous S1 generation. With the transition to the monolithic S1p architecture and complementary roving applications, the Stx platform is well positioned to support the continued expansion of CRNS monitoring networks in hydrology, precision agriculture and climate-resilient land management.