Principles of the spectrometer

As shown in Fig. 1, DINAS consists of three components: a miniaturized neutron generator for neutron production, detector modules for gamma-ray detection, and an electronics system. Neutrons from the neutron generator irradiate the deep-sea seabed, where polymetallic nodules or crustal minerals enriched with Mn are activated, releasing characteristic gamma-rays. The gamma-ray detector identifies these rays and converts them into electrical signals, which the electronics module then analyzes to determine the energy spectrum. The generated spectrum is used to determine the presence of polymetallic nodules or crusts on the deep seabed. Beyond processing gamma-ray signals, the electronics module also manages the neutron generator operation, provides comprehensive control of the entire detector, and implements data storage and communication.

Fig. 1Full-screen View

(Color online) Diagram of the principles

A miniaturized neutron generator is an essential component of the spectrometer based on neutron activation and primarily includes deuterium-deuterium (D-D) and D-T deuterium-tritium (D-T) neutron generators [13]. Although there are specific differences in terms of the neutron energy, neutron yield, and neutron emission angular distribution, both are suitable for the application scenarios explored in this study. Mn has a larger capture cross-section for thermal neutrons with energies less than 0.1 eV. However, D-D generators typically produce neutrons with an energy of approximately 2.45 MeV, whereas D-T generators generate neutrons with an energy of 14 MeV [14]. Therefore, it is necessary to moderate these neutrons to achieve a higher detection efficiency. Placing the spectrometer a few centimeters above the deep seabed allows fast neutrons to be moderated by seawater, enhancing the probability of Mn activation.

The neutron generator emits neutrons in all directions from the particle-targeting location. Therefore, when the target end of the neutron generator is placed near the seafloor, the seabed exhibits the most significant thermal neutron flux. To detect the activated gamma-rays of Mn efficiently, gamma detectors should be placed around the neutron generator close to the seafloor.

The operating environment of the deep seabed necessitates the encasing of the spectrometer in a titanium alloy shell that can withstand the water pressure of the deep ocean. In addition, a thermal neutron-shielding material with B₄C is installed on the outside of the alloy shell to absorb the thermal neutrons moderated and returned by the seawater [15]. Similar to the larger capture cross-section of Mn for thermal neutrons, the components inside the spectrometer may be affected if they are not shielded from thermal neutrons [16].

Gamma-ray detector

The design of a gamma detector, which detects and identifies activated gamma-rays, must consider the performance of the detector and the operating environment of the deep seabed. The deep-sea working scenario requires the gamma detector to have good energy resolution and detection efficiency so that the spectrometer can accurately and quickly determine the presence of target minerals. A high-purity germanium (HPGe) detector has the best energy resolution capability [17, 18]. However, its large size and the need for low temperatures reduce its suitability for applications in deep-sea in-situ environments. Inorganic scintillators coupled with photoelectric conversion devices are the most suitable choices for this purpose.

Owing to their high optical yield and energy resolution (<3.5% @662 keV), LaBr₃ scintillator crystals have become the preferred choice for gamma detectors [19-21]. The natural radiation background of LaBr₃ is typically considered a drawback; however, other studies have demonstrated that these backgrounds can be used to calibrate detectors in-situ on the seafloor [22]. In addition, despite the high atomic number and density of LaBr₃, the growth process limits its application to large crystals. This results in a limited full-energy peak detection efficiency of smaller-sized LaBr₃ for allosteric peak detection of high-energy gamma rays above 2 MeV. Therefore, we added larger, high-detection-efficiency Bi₄Ge₃O₁₂ (BGO) scintillation crystal detection channels to improve the overall detection efficiency of the spectrometer. The simulation results of the full-energy peak detection efficiencies of the LaBr₃ and BGO scintillators used in this study are shown in Fig. 3(a), which shows that ϕ2 × 3 inch BGO scintillation had a considerably higher detection efficiency for the 1810 keV and 2113 keV gamma-rays of Mn than ϕ1.5 × 1.5 inch LaBr₃ scintillation.

Typically, it is assumed that a photomultiplier tube (PMT) coupled with scintillator crystals performs best. However, in deep-sea in-situ environments, silicon photomultipliers (SiPMs) are more suitable for photomultiplier devices because of their small size, low operating voltage, and high gain. LaBr₃ and BGO scintillator crystals coupled to SiPMs can also achieve good performance [23-25].

Structure of the spectrometer

Similar to passive gamma-detection instruments operating in-situ on the deep seabed [26-28], the mechanical structure of the spectrometer consists of a watertight outer shell resistant to deep-sea pressure and an inner core with detectors, electronics, and a neutron generator. This design allows for greater flexibility in the inner-core design, making it a relatively self-contained unit that can be laboratory-operated and tested independently of the shell.

The design of the inner-core mechanical structure following the analysis in the previous sections is shown in Fig. 2. The inner-core mechanical structure comprises three parts: an inverted T-shaped D-D neutron generator, ring-shaped electronic modules, and gamma detectors consisting of multiple scintillation crystals. DINAS utilizes a compact D-D neutron generator provided by the China Academy of Engineering Physics with a neutron yield of approximately 10⁶ neutrons/s and a maximum operational power consumption of 50 W. The neutron generator integrates a high-voltage power supply at the larger-diameter top of the inverted T-shape, with the deuterium-targeting position at the bottom placed near the base of the spectrometer to provide greater neutron flux to the seabed. The size of the neutron generator determines the overall dimensions of the inner core, which has an overall diameter of 250 mm and a height of 330 mm. LaBr₃ and BGO scintillation crystal gamma detectors are positioned around the neutron generator at the base of the spectrometer to achieve the maximum geometric efficiency of gamma-ray detection. The ring-shaped electronic modules are situated above the gamma detectors, with the SiPM array directly mounted on a printed circuit board (PCB) via board-to-board connectors for convenient and reliable analog signal transmission from the detectors. Additionally, the modular ring-stacked electronics structure maximizes the PCB area, reducing the layout and wiring complexity.

Fig. 2Full-screen View

(Color online) Structure of DINAS: (a) View of the inner core, (b) exploded view, and (c) view of neutron generator separation

The spectrometer is located on the deep-sea floor, necessitating a pressure-resistant outer shell to shield the inner core from the immense underwater pressure. The design of this shell is influenced by two factors: the inherent strength of the material, which determines the required thickness to withstand water pressure, and the impact of the material on gamma-ray detection efficiency. Titanium alloys, which are commonly used for deep-sea instruments, offer high strength and low density. They do not significantly obstruct gamma-rays, thus facilitating the penetration of characteristically activated gamma-rays through the casing to reach the detector. In this study, the spectrometer’s casing was designed as a cylindrical shell enveloping the inner core.

A finite element analysis (FEA) simulation was used to preliminarily simulate the thickness requirements of the spectrometer shell under varying depths of seawater pressure, ensuring that the maximum equivalent Von-Mises stress on the shell did not exceed the permissible stress of the titanium-alloy material. For titanium-alloy materials (with a yield strength of approximately 850 MPa) calculated with a safety factor of 1.3, the permissible stress was approximately 650 MPa. The simulation results indicated that, for exploring crusts buried at a depth of 1800 m, the safe titanium alloy thickness was 20 mm, whereas for Mn nodules buried at a depth of 4000 m, the safe thickness was 35 mm.

Simulations were conducted in Geant4 to analyze the impact of these two thicknesses of titanium-alloy shells on the gamma rays. The simulation assessed the percentage of gamma-rays that could pass through the titanium-alloy shell without any energy loss, and the results are shown in Fig. 3(b). Simulation results indicated that 847-keV gamma-rays had probabilities of approximately 55% and 36% of passing through 20-mm and 35-mm-thick titanium-alloy shells, respectively, without energy loss. Despite the shielding provided by the titanium-alloy shell, a considerable number of gamma-rays could still be detected.

Fig. 3Full-screen View

Geant4 simulation results: (a) Full-energy peak detection efficiency of different scintillators. (b) Influence of different thicknesses of titanium alloys on gamma-ray detection. (c) Simulation results of the count of neutrons moderated by seawater. (d) Influence of different thicknesses of seawater on gamma-ray detection. (e) Simulation spectrum of the LaBr₃ detector for in-situ exploration. (f) Simulation spectrum of the BGO detector for in-situ exploration

Monte Carlo simulation

Monte Carlo methods are commonly used for analysis and simulation [29, 30]. In this study, Geant4 11.2 [31, 32] was used to perform a model simulation of the deep-sea in-situ neutron activation spectrometer.

A simulation model of DINAS was established in Geant4 with an overall height of 400 mm and an overall diameter of 290 mm (including the spectrometer shell). Four BGO and four LaBr₃ scintillation crystals were alternately arranged around the neutron generator, as shown in Fig. 2. The thermal neutron shielding material was a 3-mm-thick flexible material containing 70% B₄C. The neutron generator was abstracted as a point source 3 cm from the bottom of the spectrometer with an energy of 2.45 MeV, emitting neutrons randomly over a 4π solid angle. The physics list used in the simulation is QGSP_BERT_HP.

The distance between the spectrometer and seabed influences the moderation of fast neutrons produced by the neutron generator, as well as the detection of gamma-rays from activated Mn within the seabed. With increasing seawater thickness, there was an initial increase followed by a decrease in the thermal neutron flux at the seabed, whereas the gamma-rays found it increasingly difficult to penetrate the seawater for detection, as shown in Figs. 3(c) and (d). Based on analysis and simulations, the spectrometer was positioned 6 cm above the seabed to maximize the gamma count rate. Polymetallic nodules on the seabed generally appear spherical with diameters of several centimeters, with seawater filling the voids between the spheres. For modeling purposes, the seabed minerals were simplified to a mixture of polymetallic nodules and seawater with a thickness of 5 cm, with seawater comprising 50% of the mass. The chemical compositions of the polymetallic nodules were derived from a study by Pak [10].

After being emitted from the particle source, neutrons interact with the elements in the model, resulting in the release of promptly captured gamma-rays and their potential inelastic scattering, which produce a significant number of counts in a short period. Therefore, in the simulation, only gamma events occurring 1 ms after particle emission were recorded along with the corresponding reaction processes and nuclide types. The results were normalized per neutron to calculate the count per neutron. Such preliminary outcomes necessitate further processing to align them more closely with the actual measurement conditions. The count per neutron, denoted as N, as provided by Geant4, must also consider the influence of the nuclide half-life τ_h and duration of neutron irradiation T, as indicated in Eq. (1). $n(t)=N\text{Φ}\left(1-e^{-\lambda T}\right)e^{-\lambda t}\mathrm{,}$ (1) where $\lambda =\frac{\ln 2}{{\tau }_{\text{h}}}$, and Φ = 10⁶ n/s. For instance, according to the simulation results, the value of N was 7.1d-6 /neutron for the 847-keV gamma-rays of Mn measured by the LaBr₃ detector. For a neutron generator irradiation time of 60 min and measurement time of 30 min, the total peak count rate of the different scintillator channels during the measurement time was calculated using Eq. (1). According to calculations, the count rates for the three primary activation gamma-rays of Mn detected by the BGO detector channels were 9.3 cps at 847 keV, 3.0 cps at 1811 keV, and 1.7 cps at 2113 keV. In contrast, the count rates for the LaBr₃ detector channels were 1.6 cps, 0.4 cps, and 0.2 cps, respectively.

In addition, the energy deposition spectra obtained from the Geant4 simulation must be Gaussian broadened according to the energy resolution of the gamma detectors: 3.5% at 662 keV for the LaBr3 detector and 12% at 662 keV for the BGO detector. For a scintillating gamma detector, σ of the Gaussian peak in the spectrum and energy E approximately satisfy the function $\sigma \propto \sqrt{E}$. Therefore, by using an energy resolution of 662 keV as a reference, σ at all other energy points can be approximated. For each gamma event within the energy deposition spectrum, the Gaussian-distributed random energy was regenerated based on the respective energy and its corresponding σ at that energy. Subsequently, the energy spectrum was recalculated to facilitate the Gaussian broadening.

Figures 3(e) and (f) show the neutron activation gamma-ray spectra of the LaBr₃ and BGO detectors derived from Geant4 simulations and subsequent Gaussian broadening. The spectra were smoothed via a seven-point moving average to reduce statistical fluctuations and enhance clarity. The simulation results showed a finer full-energy peak resolution for LaBr₃ and a higher detection efficiency of BGO for gamma-rays with energies above 2 MeV. It is essential to note that in the simulated energy spectrum of the LaBr₃ detector, not only were the gamma-rays produced by the activation of seabed elements from neutron tubes considered, but also the radioactive background present in the scintillation crystal itself. This background is primarily attributed to the decay of the naturally occurring radioactive isotope ¹³⁸La. For a 1.5-inch scintillation crystal, the calculated activity of ¹³⁸La was approximately 60 ∼ 70 Bq. In addition to ¹³⁸La, the crystals may contain trace amounts of impurity elements, such as ²²⁷Ac. The decay of ²²⁷Ac and its decay products emit α particles, contributing to counts in the regions above 2 MeV in the LaBr₃ detector background energy spectrum. The design of the LaBr₃ detector in this study focused solely on the energy region below 2 MeV, thus eliminating the need to consider the contribution from ²²⁷Ac. The effect of the radioactivity of ¹³⁸La on the energy spectrum was simulated using Geant4, which served as a background for the activated gamma energy spectrum.

Front-end electronics module

In this study, the LaBr₃ scintillator was coupled to SiPM arrays of 16 OnSemi Micro-J60035 SiPMs using EJ-550 silicone optical grease, and the BGO scintillator was coupled to SiPM arrays of 64 OnSemi Micro-J60035 SiPMs. The area of the SiPM array was matched to the scintillator area for better light transmission and collection performance. The assembled scintillation crystal and SiPM detectors were mounted to the FEM via board-to-board connectors, and the FEM realized the signal readout and amplification.

Data process module

3.2.2

FPGA firmware design

The core of the firmware running within the FPGA in DPM processes digitized waveforms to obtain information about the charge corresponding to the nuclear pulse waveform, which translates into the energy of the incident gamma-ray. The logical structure of the waveform processing is shown in Fig. 5.

Fig. 5Full-screen View

Spectra of the radiation sources tested using the LaBr₃ (a) and BGO detectors (b)

In this study, the digital integration method was used to obtain energy information from digitized waveforms. The raw waveforms were processed using an infinite impulse response (IIR) digital low-pass filter to eliminate high-frequency noise, producing filtered waveforms that were then passed to the next-stage modules. Additionally, a digital constant-fraction timing algorithm generated trigger signals for waveforms that exceed a certain amplitude threshold, which served as the starting point for the digital integration window. Compared to simple threshold triggering, this method mitigates the impact of starting point drift due to waveform amplitude variations on energy resolution.

The time characteristics of the raw waveforms from the LaBr₃ and BGO channels differed significantly. The LaBr₃ channel waveform exhibited faster rise and fall times than the BGO channel waveform, which is indicative of higher-frequency components. Different filtering parameters were required for each channel to achieve optimal results.

The filtered waveforms, denoted as W_f, were first delayed by a configurable period to produce W_d, which was implemented based on FIFO in the FPGA. Upon detecting a trigger signal, the W_f signal was utilized to compute the waveform area using digital integration, whereas the W_d signal was used to calculate the pre-waveform baseline. The difference between these two calculations represented the actual energy information of the digitized waveform.

Once the charge information was obtained in real time from the eight waveform channels, the energy spectrum was dynamically compiled using SRAM external to the FPGA. For each waveform, the charge values normalized to 4096 channels served as addresses in the SRAM. The data value at each corresponding address was incremented, facilitating continuous accumulation of the energy spectrum.

Energy resolution experiment

Multiple radioactive sources were used to evaluate the energy resolution of the gamma-ray detectors. The detector was calibrated using these radioactive sources according to the quadratic function of the energy-address relationship, as shown in Figs. 6(a) and (b). Figure 6(c) shows the energy spectra of ¹³⁷Cs (665 keV), ²²Na (511 keV and 1275 keV), ⁶⁰Co (1173 keV and 1332 keV), and ¹⁷⁶Lu (202 keV and 307 keV) measured using the LaBr₃ detector. By performing a Gaussian fit on the full-energy peak at 662 keV, the full width at half maximum (FWHM) was obtained, resulting in an energy resolution of 3.2% at 662 keV. This result is comparable with the energy resolution outcomes reported in other studies. Figure 6(d) shows the energy spectra of ¹³⁷Cs and ⁶⁰Co measured using the BGO detector with an energy resolution of 5% at 2.5 MeV (superimposed peak of two gamma-rays of ⁶⁰Co).

Fig. 6Full-screen View

Energy calibration of the LaBr₃ (a) and BGO detectors (b) and the spectra of the radiation sources tested using the LaBr₃ (c) and BGO detectors (d)

Test of the gamma detector at CSNS

To evaluate the capability of the DINAS gamma detectors and electronics system to measure the gamma spectrum of activated Mn, thermal neutron beams from the China Spallation Neutron Source (CSNS) were employed to activate Mn as a preliminary experiment for spectrometer validation. Thermal neutron beamline No. 20 at the CSNS can provide thermal neutron beams with wavelengths ranging from 1 to 5 Å, efficiently activating Mn.

Figure 7(a) shows a photograph of the experimental setup, with the Mn metal located on the thermal neutron beamline, the detectors and electronics positioned beside the Mn, and lead bricks shielding them from the accompanying gamma-rays from the beamline. During the experiment, the irradiation time of the thermal neutron beam was approximately 10 min and the measurement time was 5 min.

Fig. 7Full-screen View

(Color online) Experimental results of the prototype: (a) Photograph of the experiment setup at the CSNS. (b) LaBr₃ detector spectrum measured at the CSNS. (c) BGO detector spectrum measured at the CSNS. (d) Photograph of the laboratory model experiment setup. (e) LaBr₃ detector spectrum of the model experiment. (f) BGO detector spectrum of the model experiment

The resulting activated gamma spectra are presented in Figs. 7(b) and (c), revealing significant gamma-ray peaks from Mn activation. The limited detection efficiency of LaBr₃ detectors above 2 MeV, combined with their inherent radiation background, makes the detection of the 2113-keV peak of Mn challenging when using the D-D neutron generator within the spectrometer as a neutron source, which can provide a considerably lower thermal neutron flux. This underscores the importance of the BGO detector.

Spectrometer model experiment

A model experiment was conducted in the laboratory to assess the functionality of the spectrometer prototype. Simulations of the model experiment were conducted using the Monte Carlo method outlined in Section 2.4, and the results were compared with experimental data to validate the reliability of the proposed simulation model. As shown in Fig. 7(d), a 6-cm-thick polyethylene board was used to simulate the presence of seawater between the spectrometer and the seabed, with a total mass of 10 kg of Mn metal placed on the surface of the SiO₂ sand. The spectrometer prototype was powered by a DC power supply and controlled by an upper computer for the data readout. The neutron generator used in the experiment was the D-D neutron generator mentioned previously, with a neutron yield of approximately 10⁶ neutrons/s. After the neutron generator was activated, irradiation continued for approximately 1 h. Subsequently, the neutron generator was deactivated, and a 1-h measurement period commenced. Turning off the neutron generator before commencing the measurements served a dual purpose. It reduced the impact of promptly captured gamma-rays and inelastically scattered gamma-rays on the detection of Mn activation gamma-rays. Moreover, it prevented an increase in dark counts and a decline in the performance of SiPMs caused by neutron irradiation.

The activation gamma spectra of Mn detected using the two types of gamma detectors are shown in Figs. 7(e) and (f), revealing significant characteristic peaks of the Mn activation gamma-rays. For the LaBr₃ detector, the characteristic peaks were superimposed on the radiation background of LaBr₃, similar to the results obtained at the CSNS. Calculations showed that the peak count rate at 847 keV measured by the LaBr₃ detector was 1.8 cps, whereas the BGO detector measured a peak count rate of 1.7 cps at 2113 keV. Figures 7(e) and (f) also present the activated gamma-ray spectra when the initially pure Mn sample was substituted with a ferromanganese mixture containing approximately 60 wt% Mn. No significant differences were observed in the activated gamma-ray spectra when the sample was replaced with the mixture. The characteristic gamma-rays of Mn were still distinctly detectable, with a count rate of approximately 0.8 cps at 847 keV detected by the LaBr₃ detector and 1.0 cps at 847 keV detected by the BGO detector.

The experimental results showed a high degree of consistency with the Monte Carlo simulation spectra of Geant4. For the LaBr₃ detector, the experimental background had a higher count than the simulation because of the imprecision of the LaBr₃ background simulation model, although the characteristic Mn peaks were similar. The peak near 1460 keV resulted from the decay of ⁴⁰K and electron capture (EC) decay of ¹³⁸La in the LaBr₃ scintillator. For the BGO detector, the experimental and simulation results aligned well at the Mn characteristic peaks, with the experimental spectrum featuring two additional background peaks corresponding to the 1460.8 Ke V gamma rays of ⁴⁰K and 2614.5 Ke V gamma rays of ²⁰⁸Tl.

Discussions

The model experiment confirmed the viability of the in-situ exploration technique proposed in this study, proving that the designed prototype is capable of effectively detecting metal resources on the deep-sea floor in-situ. Despite these advancements, precise quantitative measurement of the elemental content of deep-sea metal minerals continues to pose a significant challenge, necessitating additional research. The following section includes an initial discussion of this topic.

In the in-situ measurements conducted in deep sea, the full-energy peak count rate A of the characteristic gamma-rays for Mn, as observed in the activated gamma energy spectrum, is theoretically described by Eq. (5). $A=\frac{m{N}_{\text{A}}\theta }{M}{P}_{\gamma }ϵ\text{Φ}\sigma \left(1-e^{-\lambda t_{\text{r}}}\right)e^{-\lambda t_{\text{c}}}\mathrm{,}$ (5) where m is the represents the mass of the target element (Mn) to be measured, M is the molar mass of Mn, N_A is Avogadro’s constant,θ denotes the isotopic abundance of ⁵⁵Mn, Pγ is the branching ratio for the characteristic gamma-rays, $\sigma$ represents the capture cross-section of the nuclide, Φ is the thermal neutron flux, ϵ represents the gamma-ray detection efficiency of the spectrometer, λ is the decay constant of ⁵⁶Mn, and t_r and t_c are the irradiation and measuring time, respectively.

In a deep-sea in-situ work environment, accurately obtaining the neutron flux Φ is challenging because of factors such as the neutron yield stability of the neutron generator itself, structure of the spectrometer shell, and distance between the spectrometer and seabed. Similarly, the measurement of detection efficiency ϵ is impeded by the attenuation caused by seawater. Consequently, the mass of the target element cannot be calculated directly using Eq. (5). By employing a relative analysis method from the field of neutron activation analysis, a reference element with known mass can be placed directly on the seabed. The mass of Mn can be indirectly determined by measuring the characteristic peak counts of the reference and target elements (Mn). Indium-115 (¹¹⁵In) serves as an ideal reference element because of its substantial neutron capture cross-section (¹¹⁵In(n,γ)^116mIn, 157 barns for thermal neutrons) and the suitable half-life (54 min) of its activation product,^116mIn [11]. Furthermore, ^116mIn emits various gamma-rays, which addresses the discrepancies in detection efficiency caused by differences in gamma-ray energy. Considering the impact of seabed minerals on the neutron flux and the blocking of activated gamma-rays, empirical underwater experiments and simulations are essential to achieve more accurate results.