Calibration of the MEGANE gamma-ray and neutron spectrometer for JAXA’s Mars Moon eXplorer (MMX) mission to Phobos

JAXA’s Mars Moon eXploration (MMX) spacecraft carries a NASA-provided γ-ray and neutron spectrometer called MEGANE. MEGANE’s measurements will be used to characterize the elemental composition of Mars’ largest and innermost moon, Phobos, with the goal of providing data that directly tests existing hypotheses for the origin and evolution of the martian moons. We report results from the ground calibration of the MEGANE sensors. This information is required to accurately interpret the measurements collected by MEGANE during in-flight operations at Phobos.

The Mars-moon Exploration with GAmma Rays and NEutrons (MEGANE) instrument is a NASA contribution to JAXA’s Mars Moon eXploration (MMX) mission. The MMX mission’s primary objective is to elucidate key aspects of early solar system evolution as recorded on the martian moons Phobos and Deimos (Kuramoto et al. 2022). This will be achieved through both sample return from Phobos, as well as landed and orbital observations that will be made during three years of planned observations in the Mars system. MEGANE—the Japanese word for eyeglasses—contributes to MMX science goals by providing the project the ability to "see" γ-rays and neutrons, two forms of radiation that contain information about the elemental composition of near-surface materials on Phobos, to a depth of a few tens of cm. MEGANE data will be used to distinguish between the competing hypotheses for Phobos’ formation (Lawrence et al. 2019). These theories include the capture of a primitive asteroid and in situ formation from the debris generated by a giant impact on Mars. The two theories yield different elemental composition of Phobos’ surface, and the MEGANE investigation was designed to differentiate between these possible compositions. MEGANE will also attempt to resolve compositional differences between the two known spectral units on Phobos (Chabot et al. 2021). This information may be used as part of MMX sample collection site selection.

MEGANE’s γ-ray and neutron spectrometer subsystems have heritage from the MESSENGER and Psyche gamma-ray spectrometers (GRS; Goldsten et al. 2007; Lawrence et al., 2025) and the Lunar Prospector (LP) Neutron Spectrometer (NS; Feldman et al. 2004). The NS subsystem of MEGANE consists of two helium-3 (³He) gas-filled proportional counter (GPC) detectors with a single front-end electronics module containing a preamplifier for each sensor. The GRS subsystem consists of a high-purity germanium (HPGe) γ-ray detector, surrounded by a cup-shaped anti-coincidence shield (ACS) composed of a borated plastic scintillator with a photomultiplier tube (PMT) readout. GRS sensor front-end electronics include a sensor control unit module (SCUM) and preamplifier, high-voltage filter box (HVFB), and PMT preamplifier. A single data processing unit (DPU) provides high- and low-voltage power, cryocooler power for the GRS, and instrument commanding and signal pulse processing for both subsystems. Figure 1 provides an overview of the sensors, highlighting the active detector volumes.

As of late 2025, the MMX spacecraft is expected to launch aboard JAXA’s H3 rocket in October 2026. Interplanetary cruise will last approximately 11 months, and the spacecraft will enter orbit around Mars in August 2027. MMX will stay in Mars’ orbit for about 3 years, during which time it will investigate Phobos from a series of quasi-stationary orbits (QSO) with varying altitudes above the moon’s surface. The lowest altitude orbit, called "LC", places the spacecraft close enough to Phobos to perform the MEGANE investigation, as well as to perform high-spatial-resolution surveys of candidate landing sites using the suite of onboard imaging instruments.

As of December 2024, the LC phase of the mission is currently split into two phases. A short (~ 3-week-long) QSO-LC phase occurs in May 2028 and will provide ~ 5 days of high-quality MEGANE measurements prior to sample collection. A second QSO-LC phase, beginning in December 2029 and lasting ~ 2.5 months, will provide an additional ~ 30 days of data. During QSO-LC orbits, the MMX project will prioritize "nadir pointing" (i.e. spacecraft -Z axis pointing towards Phobos) to optimize the MEGANE measurements. Additionally, MEGANE will collect four, 15-day-long measurements of spacecraft-originating backgrounds during periods when the MMX spacecraft is sufficiently far from Phobos (altitude > 3.5 ×ばつ Phobos’ mean radius) to ensure that there is no Phobos-originating contribution to the measurements. These background measurements are important for proper analysis of the low-altitude data collected during the QSO-LC periods. See Nakamura et al. (2021) for details on the design of MMX orbital phases; however, note that the detailed planning has changed substantially relative to the plan in that paper due to a launch-date slip from 2024 to 2026. As of late 2025, the MMX operations plan remains in a state of revision, and thus, the final details of the QSO-LC orbits may differ from those outlined here.

There are several additional potential opportunities for MEGANE science. In 2029, the spacecraft will perform a series of rehearsal and actual landing attempts to collect samples for return to Earth. These short-duration activities (< 10 h each) occur at very low altitudes compared to QSO-LC, which would yield substantially improved signal-to-background relative to the QSO-LC measurements. MEGANE science data collection during these periods is under discussion with the MMX project. Additionally, a series of Deimos flybys in 2030 will provide an opportunity to compare the two moons. MMX will not be close enough to Deimos to facilitate γ-ray measurements of this moon; however, the NS will collect high-time-cadence data in the hopes of detecting neutrons originating from Deimos.

The in-flight particle and γ-ray environments experienced by the MEGANE instrument are unique to space and cannot be reproduced in the laboratory. The ground calibration campaign therefore relies on analogous measurements that provide validation of radiation transport simulations that are used to characterize in-flight performance. These same simulation tools will also be used to interpret the measurements made at Phobos. The calibration measurements are made using radioactive sources that emit γ-rays in the energy range of interest for MEGANE, as well as neutron sources to stimulate signals in the ACS and NS. Each of the radiation sources used for calibration is listed in Table 1. The γ-ray sources (⁶⁰Co, ¹³⁷Cs, ¹⁵²Eu, ¹⁵⁴Eu, and ²²⁸Th) are Type-D sources, with NIST-traceable activities with an uncertainty of < 3%. Gamma-ray energies are retrieved from Gilmore (2008). Cosmic-ray secondary muons are used to study the response of the sensors to charged particles similar to cosmic-ray protons. This calibration effort follows similar efforts to calibrate the NEAR (Peplowski et al. 2015), MESSENGER (Evans et al. 2017), and Psyche γ-ray spectrometers (Peplowski et al., 2025), as well as the characterization of Psyche and MEGANE neutron sensors at the NIST Center for Neutron Research (Peplowski et al. 2020) and the Neutron Spectrometer Subsystem (NSS) for NASA’s VIPER lunar rover (Peplowski et al. 2023).

During calibration, both of the MEGANE sensor units and the DPU were mounted on a deck mockup plate that replicated the relative positions of MEGANE components and harness routing when integrated on the MMX spacecraft. During calibration, a thin Delrin sheet with punch-outs for positioning γ-ray calibration sources was mounted above the deck mockup and GRS sensor. The central position of the punch-outs was directly above the centre of the HPGe crystal. The HPGe-centre-to-calibration-plate (bottom) distance is 55 cm. This configuration is shown in Fig. 2. The calibration plate has 121 source positions, which span the full range of incidence angles expected during MEGANE operations in the QSO-LC orbit. An outline of the Phobos field of view is shown relative to the calibration plate in Fig. 2. Note the hardware that mounts the calibration plate to the deck mockup interfered with calibration measurements at a small number of positions. This is detailed in Sect. 5.5.

3.1 Overview and spectral performance

The MEGANE NS sensors are neutron-sensitive gas proportional counters (GPCs) filled with ³He gas. Their dimensions are shown in Fig. 3. The tubes have a 1.4-mm-thick aluminium housing and are filled with 1013 kPa (10 atm) of neutron-sensitive ³He gas. The total fill pressure is 1018 kPa, with the difference (5 kPa) being a proprietary mixture of charge and/or UV quenching gas additives that improve the performance of the sensors. The two GPC sensors are differentiated only by different external material wraps, which act as neutron energy filters. The "bare" GPC has no external wrap and is thus sensitive to all neutrons (kinetic energies of up to ∼1 keV). The "Cd" GPC is surrounded by a 0.5-mm-thick, conformally wrapped sheet of cadmium (Cd) metal. Cd efficiently absorbs low-kinetic energy (≲0.35eV) neutrons, making the Cd-wrapped GPC only sensitive to neutrons with energies of approximately 0.35 eV to 1 keV. The difference between the bare and Cd sensor measurements yields the thermal (< 0.35 eV) neutron count rate at the NS.

Helium-3 gas is a common medium for neutron detection due to the high thermal-neutron-absorption cross Sect. (5330 b) of the neutron capture reaction ³He(n,p)t. The kinetic energies of the two reaction products, protons (573 keV) and tritons (191 keV), sum to the Q-value of the neutron capture reaction (764 keV). During an ideal neutron detection event, the proton and triton lose all of their kinetic energy within the ³He gas and a characteristic neutron capture peak appears in the energy deposition spectrum at the reaction Q-value (as noted in Fig. 4). Spectral fitting shows that the capture peaks are located in the MEGANE NS measurements at channel ~ 151 (of 256), yielding an approximate energy calibration of 5.06 keV/ch (from 764 keV/channel 151). The capture peak has a full width at half maximum of 6.5(± 0.5)%, a significant improvement over the heritage Lunar Prospector NS (FWHM ≈20%). Non-ideal events include reactions wherein the proton or triton reach the wall of the GPC before depositing all their energy in the gas. When this occurs, there is incomplete charge collection in the gas and the result is a "wall effect" feature in the spectrum (see Fig. 4).

During operation, a high-voltage bias is applied to the anode wire of the GPCs (shown in Fig. 3), and a uniform, radial electric field permeates the gas in a volume known as the "active region" of the sensor. Charged-particle interactions in the GPC produce ionization electrons that are drawn to the anode wire by the resulting electrostatic potential. The total number of ionization electrons produced by a neutron capture event scales with the bias voltage that is applied to the sensor, resulting in the gas gain vs. bias relationship shown in Fig. 5. We chose a nominal operating voltage of 1360 V to maximize gas gain without compromising peak shape, and an electronic gain that places the neutron capture peak near the centre of the energy deposition spectrum at this bias level. At both ends of the sensor are regions where the anode wire is surrounded by a field-control tube and therefore the corresponding gas volume has a reduced electric field. This "dead region" of ³He gas does not contribute events to the neutron capture peak, but instead produces the "no gain" peak observed in some neutron spectra (e.g. Figure 4). Neutron events that occur near the transition between the active and dead volumes have intermediate gain values, resulting in a low-energy tail on the neutron capture peak. This phenomenon is discussed in detail in Peplowski et al. (2020).

The electric charge produced in the gas is detected (and amplified) by the preamplifier module and then sent to an analogue-to-digital converter (ADC) that tracks the event amplitude (charge) versus time. The signal passes through a bipolar shaper, and the total amplitude is derived from the positive peak to negative peak amplitude of the bipolar-shaped pulse (Fig. 6). The MEGANE FPGA records these peak-to-peak values on an event-by-event basis, where they are passed to the instrument flight software (iFSW), which sums the events over a commandable integration time to produce a histogram of pulse-height values, hereafter called a spectrum (Fig. 4).

Temperature-dependent variations in the performance of the electronics components that populate the preamplifier and processor boards can induce changes in the amplitude of the pulse-height signal that manifest as gain changes in the NS pulse-height spectra. We quantified these gain variations during instrument thermal-vacuum (iTVAC) testing. iTVAC included six temperatures cycles that included dwells at cold and hot soak plateaus, and the exact temperatures of these plateaus varied depending on the specific instrument component in question but are nominally −20°C and 50°C, respectively, for the NS sensor. During iTVAC, a moderated AmBe neutron source was placed outside of the TVAC chamber, and 60-s-integration-period neutron spectra were collected over the course of two full temperature cycles. The centroid position of the neutron capture peak was recorded for each measurement, and the resulting positions (centroids) were plotted as a function of both the preamp and processor board temperatures (Fig. 7). The results show a ~ 7 channels (4%) variation in the centroid position of the neutron capture peak of over the full temperature range of the TVAC cycles sampled, with observable hysteresis as the temperature increased versus decreased. The hysteresis may be due to differences between the locations of the temperature sensors and the temperature-sensitive components in the preamplifier and processor board. Note that the expected temperature range in flight (~ −25 to ~ 40° C) is smaller than the range during iTVAC, and over this expected range the temperature-induced gain shifts are significantly smaller (~ 2–3 channels). Any temperature-dependent gain shifts observed in flight for the NS will be corrected in the calibrated (L1) PDS data products.

3.2 Neutron detection efficiency

It was not possible to characterize the MEGANE NS efficiency in a laboratory environment with flight-like neutron energy and geometric properties. As a consequence, our knowledge of the response of the NS sensors relies on simulations using version 11.1.2 of the GEANT4 radiation transport toolkit (Agostinelli et al. 2003; Allison et al. 2006, 2016). A GEANT4 model of the NS was developed using the geometry shown in Fig. 3, with the addition of the Cd-wrap for one sensor and the proper relative spacing for the two sensors in their flight configuration. The precision and accuracy of the GEANT4 simulations were benchmarked using measurements with a monoenergetic neutron beam (Peplowski et al. 2020) and mixed-energy neutron sources (²⁴¹AmBe and ²⁵²Cf).

Monoenergetic neutron measurements were performed at the NIST Center for Neutron Research (NCNR). NCNR provided a monoenergetic, tightly collimated neutron beam with a kinetic energy of 0.05 eV. That beam systematically scanned the length of six separate engineering-model GPC sensors, and the ratio of the measured-to-incident neutron flux was used to calculate the neutron detection efficiency as a function of position at 0.05 eV. Those measurements are detailed in Peplowski et al. (2020), and they informed the development of the GEANT4 simulation used throughout MEGANE NS calibration activities.

We further validated that simulation with two measurements made with the completed flight-model MEGANE NS. For those measurements, a moderated ²⁵²Cf neutron source was located first in the -Z direction, which will be directed at the spacecraft sub-nadir point on Phobos during QSO-LC observations. Then, a second set of measurements were performed with the source ~ 45° off nadir (towards the -Y direction). Following a process used for calibration of the Neutron Spectrometer Subsystem for the VIPER lunar rover (Peplowski et al. 2023), direct comparison of measured and modelled spectra yields a quantification of the accuracy of the simulations. That comparison is shown in Fig. 8, and the result is that the simulations reproduce the measured rates to within < 10% for both sensors and both orientations.

With the validation of the GEANT4 simulations complete, we simulated the efficiency as a function of neutron energy for a plane wave of neutrons incident on the NS from the -Z (i.e. "nadir") direction. This is the direction to Phobos, as viewed by NS, during the MEGANE-dedicated observations in QSO-LC. A plane-wave source is not a perfect analogy for our measurements at Phobos; however, it provides valuable quantification of the average response of the NS. See Fig. 9. All final analyses of MEGANE data will take into account the shape of Phobos and the actual distribution and incidence angles of neutrons from the surface as detailed in Chabot et al. (2021).

3.3 Background suppression

The NS data processing algorithms calculate the time between the positive and negative lobes of each bipolar-shaped event. This quantity is called the delta time or "dt". Figure 6 illustrates dt and peak-to-peak (p2p) values for a typical neutron pulse. The dt values provide a valuable tool for characterizing the source of a given measured pulse. Figure 10 plots dt versus the p2p values for various radiation sources. The features of interest are:

• The neutron capture peak (signal of interest), which has a p2p value of ∼18,000 ADC units and a narrow range of dt values centred at dt ≈ 270 (in units of 0.1 μs).

• The neutron capture artefacts resulting from the sensor geometry, which includes the wall effect events and the no-gain peak events from interactions in the sensor dead volume, have a range of p2p values (∼4000 – ∼18,000 ADC units) but a narrow range of dt values centred at dt ≈ 270 (0.1μs).

• Gamma-ray interactions in the sensor, which are low p2p value (< 5000 ADC units) and have a wide range of dt values.

• Fast-neutron capture events (capture peak + some neutron kinetic energy). These events have a p2p value of > 20,000 ADC units and a narrow range of dt values centred at dt ≈ 270 (0.1 μs).

Additionally, we know from Psyche NS data that cosmic-ray protons produce a wide distribution of dt values (Peplowski et al., 2025). The exact range of values for MEGANE will not be known until post-launch instrument commissioning. Regardless, these data indicate that a pulse-processing cut that retains only those events with dt values inclusive of neutron events (wall effect and capture peak; i.e. 26.3 μs < dt < 28.3 μs) will retain all of the neutron events, while removing low-energy γ-ray backgrounds and cosmic-ray protons. To that end, the MEGANE pulse-processing architecture provides a means of restricting the events that are added to the pulse-height spectra, based on their dt values.

There are three dt-cut pulse-processing options available. Option #1 produces neutron spectra with no dt cut (e.g. black histogram in Fig. 11). Option #2 removes events outside of the valid dt range (e.g. red histogram in Fig. 11). Option #3 repurposes upper portion of the spectrum to plot the otherwise-rejected early dt events (not plotted here). The lower-limit channel for the repurposed portion that displays the rejected events is also programmable.

3.4 Conclusion

Neutron spectrometer data are reported as pulse-height spectra produced using programmable pulse-processing parameters. Nominally, the raw NS spectra will be recorded in dt cut mode 2, which removes the low-energy γ-ray events and a significant portion of cosmic-ray-proton events, while preserving the neutron signal of interest (Fig. 11). These raw (L0) data products will be temperature-corrected during processing in the MEGANE Science Operations Center (SOC) to remove the gain shifts, as guided by the iTVAC data (Fig. 7). Spectra are also filtered for data quality at this stage, removing measurements that might be compromised by environmental (e.g. solar particle events) or instrument housekeeping (e.g. unstable HV) conditions. The resulting calibrated (L1) data facilitate spectral summing that preserves energy resolution while maximizing statistical precision.

Radiation transport simulations are used to predict the neutron signal at MEGANE for varying elemental composition scenarios at Phobos’ surface. These simulations incorporate the calibration-benchmarked (Fig. 8) efficiency versus energy (Fig. 9) and angle information detailed here. The best-fit simulations to the data will be used to report the surface composition of Phobos, which for the NS is focused on deriving the mean hydrogen content of Phobos’ surface, to a depth of a few tens of centimetres.

4.1 Overview and measurement modes

The MEGANE anti-coincidence shield (ACS) is a cup-shaped piece of borated polyvinyl toluene (plastic) scintillator that surrounds the γ-ray sensor. The ACS provides several functions:

• Charged-particle detection for the anti-coincidence veto trigger of the γ-ray spectrum,

• Low-energy (< 0.1 MeV) neutron detection,

• High-energy (Fast; > 0.1 MeV) neutron detection, and

• Galactic-cosmic-ray (GCR) monitoring via direct measurements of minimum-ionizing particles (MIPs).

The choice of the sensor material and shape of the AC shield was informed by these functions. Low-density, low-Z plastic scintillator provides charged particle detection while minimizing direct γ-ray interactions, increasing the efficiency of the veto trigger without unduly shielding the γ-ray signals of interest from interacting in the HPGe γ-ray sensor. The cup-like shape of the AC Shield (Fig. 12) allows the HPGe sensor to be placed within the AC Shield, maximizing the geometric probability that cosmic-ray protons will interact with both sensors and thus maximizing the efficiency of the veto trigger. A cross section of the ACS is shown in Fig. 12. The total mass of scintillator is 1.63 kg.

We selected EJ-254, an inorganic scintillator with boron added (5% by mass) to facilitate neutron detection via the ¹⁰B(n,α)⁷Li reaction (Drake et al. 1986). EJ-254 is Eljen Technology’s commercial equivalent of the Bicron BC-454 scintillator that was used for the Lunar Prospector ACS (Feldman et al. 2004), MESSENGER NS (Goldsten et al. 2007), Dawn GRAND (Prettyman et al. 2012), and the Psyche GRNS ACS (Lawrence et al., 2025). EJ-254 is inherently sensitive to thermal and epithermal neutrons. Pulse-processing algorithms performed by the MEGANE flight software further facilitate identification of fast (> 0.1 MeV) neutrons (Feldman et al. 1991) as detailed in Sect. 4.3.

When any charged particle interacts in the ACS, it loses kinetic energy via a variety of processes, primarily ionization, that lead to the production of scintillation photons. Our charged particles of interest include galactic-cosmic-ray (GCR) protons and the ¹⁰B + n reaction daughters α and ⁷Li. Internal reflection guides these photons towards the photomultiplier tube (PMT) located at the base of the scintillator. ~ 50% of the photons reach the PMT contact surface and are converted to electrons by the PMT cathode. That electron signal undergoes amplification (electron multiplication) in the PMT. The amplification is proportional to the PMT bias voltage. The PMT output signal next passes to the ACS preamplifier and then to an analogue-to-digital converter (ADC) located in the DPU processor board.

The magnitude of the PMT output signal is proportional to the number of scintillation photons entering the PMT, not the energy deposited in the ACS by the parent particle. The number of photons produced in the ACS depends on a number of factors, including the particle species and energy, as well as detector properties such as internal photon quenching and geometry-dependent light collection efficiency. These factors complicate the relationship between ACS measurements (e.g. signal pulse height) and the interaction that produced the signal. For instance, the ¹⁰B(n,α)⁷Li neutron capture reaction products have a summed kinetic energy of 2.79 MeV (the reaction Q-value) with the daughter nucleus in its ground state (⁷Li^g.s.), or 2.31 MeV with the daughter nucleus in its 1st excited state (⁷Li^*). Yet the measured signal for this type of event does not correspond to 2.31 MeV or 2.79 MeV, but instead an electron-equivalent energy deposition of ~ 80 keV_ee (electron-equivalent keV). See Sect. 4.2 for further discussion of the neutron signal in the ACS.

The ACS is operated in two different gain regimes defined by the PMT voltage setting. The first is high-gain mode, which is used for neutron detection and the HPGe veto. The HPGe veto function is discussed in Sect. 5.6. During the calibration activities detailed in this report, the bias voltage of the ACS photomultiplier tube was 542 V and the energy scaling was ~ 1.2 keV/channel (Fig. 13) for the high-gain mode measurements. Additional details for high-gain mode are discussed in Sect. 4.2. The second ACS gain regime is the low-gain mode, which is used to make direct measurements of galactic-cosmic-ray (GCR) protons. During calibration, the bias voltage of the ACS photomultiplier tube in low-gain mode was 260 V and the energy scaling was ~ 68 keV/channel. Low-gain mode neutron measurements are discussed in Sect. 4.3.

Note that the ACS gain modes are exclusive. When MEGANE is commanded into one of the two modes, high-voltage, gain, and threshold settings are changed as needed. Thus, the instrument can either produce the high-gain spectrum and fast-neutron (FN) spectra or the low-gain GCR spectrum. The primary function of vetoing HPGe-coincidence signals works identically in either mode. The expected operational strategy is to operate in high-gain mode for the majority of the time in orbit around Phobos to provide neutron science with the ACS. The low-gain GCR mode will be used in cruise, when far from Phobos, and for a small fraction of time in orbit around Phobos, as needed to sample the local GCR environment. Ideally, low-gain measurements made in QSO-LC will be timed to occur during non-optimal Phobos measurement periods, such as data downlink periods when MMX is not providing Phobos nadir pointing for MEGANE.

4.2 High-gain (neutron) spectrum

The ACS high-gain mode is tuned for making neutron measurements with the ACS via observation of the ~ 80 keV_ee peak resulting from the ¹⁰B(n, α)⁷Li neutron capture reaction. The cross section for this reaction is inversely proportional to the neutron energy, making this mode sensitive to thermal and epithermal neutrons. Sensitivity to > 100 keV neutrons is achieved via detection of double pulses as detailed in Sect. 4.3.

A high-gain ACS spectrum is shown in Fig. 14. The moderated AmBe measurements (blue and red spectra) show a clear spectral peak at ∼80 keV_ee that is not present when a moderated neutron source is absent (black spectrum). This spectral feature is due to the detection of the ¹⁰B + n reaction products. Its presence is indicative of neutron detection, and its amplitude provides the neutron rate at the AC Shield.

We provide an approximate conversion from channel to energy, as derived via the use of several signals observed in the ACS, including the neutron capture peak (~ 80 keV_ee), the minimum-ionizing particle (MIP) peak produced via muon interactions in the ACS (see Sect. 4.4), and Compton edges from γ-ray sources. While there is a near-zero probability of fully stopping a > 100 keV γ-ray in the scintillator (i.e. producing a photopeak), there is a small but non-zero probability that a γ-ray will have a single scattering interaction with the plastic to produce an event that populates the Compton edge.

Compton edge energies are calculated from photopeak energy values as:

where E_γ is the full energy of the γ-ray and m_ec² is the rest mass of an electron (511 keV). For instance, the Compton edge of the 662-keV γ-ray from ¹³⁷Cs is located at 477 keV. The presence of the ¹³⁷Cs Compton edge in Fig. 14 (bottom panel) shows that the high-gain mode ACS spectrum extends to ~ 500 keV. Note that this range facilitates detection of neutrons via the ¹⁰B(n,α)⁷Li* reaction, where 478-keV γ-rays are produced from the decay of the ⁷Li* excited state. These γ-rays are occasionally detected in the ACS via Compton scattering.

4.3 Fast-neutron detection

Fast-neutron detection is achieved with the ACS by using it as a proton recoil spectrometer via the use of a double-pulse detection algorithm first developed by Feldman et al. (1991). The scenario is as follows:

• A fast (≥ 100 keV) neutron interacts with the ACS via proton scattering on the H in the scintillator,

• The recoiling proton loses energy via ionization in the ACS, resulting in the production of scintillation photons (called the "prompt" event),

• The now lower-energy neutrons are in an energy range where the ¹⁰B(n,α) reactions are more likely to occur. Any neutron capture event that does occur is detected and called the "capture" event.

EJ-254 was designed specifically for this process, as it contains H (14% by mass) and 5% (natural) boron loading. At this boron loading, the characteristic time between fast-neutron moderation and capture is 2 μs.

Figure 15 (top panel) details the double-pulse detection process, along with the implementation as carried out by the MEGANE DPU. When a signal is identified in the ACS, its amplitude is checked against the software-specified prompt threshold value. If the event amplitude exceeds the prompt threshold, then it is classified as a prompt event and a 25-μs-long coincidence window is opened. If a second event whose amplitude exceeds the capture threshold value occurs within this coincidence window, then the second event is labelled a capture event and the two events are classified as a candidate fast-neutron (FN) event. The time between the prompt and capture events is recorded as the time-to-second-pulse (TTSP). The prompt and capture thresholds are commandable values that are checked against the unipolar shaped and filtered ADC input. As currently configured (late 2024), the prompt and capture thresholds are identical (ADC value = 28), and these values correspond to 42 and 2.6 keV_ee in the Capture and Prompt spectra, respectively (Fig. 16).

MEGANE’s FN event-mode diagnostic data product illustrates FN signal processing. It provides event-by-event information for each event meeting the coincidence timing and threshold requirements, including the Prompt and Capture pulse peak amplitudes, and the TTSP values. A histogram of measured TTSP values as recorded during irradiation with a fast-neutron source (²⁵²Cf) is shown in Fig. 15 (bottom panel). The real FN rate at the ACS is determined by first removing accidental coincidences that mimic the double-pulse signature of the fast-neutron recoil plus capture. A constant-value function is fit to the counts vs. TTSP data for events whose TTSP values are much longer (> 20 μs) than the characteristic capture time (~ 2 μs) to determine the accidental coincidence rate. See Fig. 15 (bottom panel) for an example. This rate is subtracted from the TTSP data to yield the background-subtracted, fast-neutron-only population of events (Fig. 15).

The FN event-mode data product is data volume intensive and is only produced during diagnostic periods. For Phobos observations, fast-neutron science is done using the ACS FN spectral product. This product yields four separate 256-channel-long spectra:

• Prompt events with TTSP < 5 μs, called "Prompt Early",

• Prompt events with TTSP of 20 μs to 25 μs, called "Prompt Late",

• Capture events with TTSP < 5 μs, called "Capture Early", and

• Capture events with TTSP of 20 μs to 25 μs, called "Capture Late".

Here "early" and "late" refer to the time that the second (i.e. Capture) event arrives. The pulse height values reported in these spectra include software gain of the ADC signal, meaning that the channel values reported here do not correspond to the software threshold values. To derive the fast-neutron rate in a given spectrum, the "Capture Late" spectrum should be subtracted from the "Capture Early" spectrum. For example, FN spectrum data, collected during a MEGANE performance checkout performed soon after installation on the MMX spacecraft, are shown in Fig. 16.

A flight-like neutron environment that covers the full range of energies and incidence angles expected during QSO-LC measurements at Phobos could not be produced for the MEGANE calibration campaign. Instead, we adopted the Psyche FN efficiency calibration (Peplowski et al., 2025). The Psyche effort involved the use of a 2.5-MeV neutron source, produced by a deuterium-deuterium neutron generator, to benchmark GEANT4 simulations of fast-neutron detection efficiency. The simulations are detailed in Appendix A. For MEGANE, we simply adjusted the prompt and capture threshold values, which differ between Psyche and MEGANE, and re-ran the simulations to produce the efficiency information for MEGANE. The MEGANE and Psyche thresholds differ, in part, because the ACS PMTs for the two systems are operated at different voltages and thus the systems have different gains.

Using the MEGANE prompt and capture thresholds, we simulated the response of the ACS to a plane wave of monoenergetic neutrons originating from the nadir (-Z) direction, covering the entire energy range of interest for the MEGANE investigation. Events were tagged as singles if they resulted in a capture peak but no prompt event exceeding the prompt threshold, and double pulses if they met the fast-neutron detection requirements detailed in Appendix A. The resulting simulations were used to determine the neutron detection efficiency as a function of energy (Fig. 17). We define the lower-energy threshold for neutron detection using the half-maximum value of the rising edge of the efficiency versus energy (Fig. 17). This half-maximum value occurs at 445 keV. The maximum neutron detection efficiency (23%; 48 cm² active area) occurs at 700 keV. Note that singles efficiency increases at > 10 MeV due to resonances in the ¹⁰B(n,α)⁷Li neutron capture reaction that increase the direct (no-moderation) probability for the reaction.

4.4 Low-gain (GCR) spectrum

The majority of γ-ray and neutron radiation observed by MEGANE is produced as a result of interactions between galactic-cosmic-ray (GCR) particles and the surface of Phobos. As a consequence, it is vital to know the spectral shape and integrated particle flux of GCRs as a function of time throughout MEGANE observations in order to properly interpret the MEGANE data. To that end, MEGANE makes direct measurements of GCR protons with the ACS shield via the data product called the GCR spectrum, which is produced with a lower-gain version of the ACS data.

The ACS low-gain spectrum (called GCR spectrum) has a full-energy range (~ 45 MeV) that was selected to be inclusive of the minimum-ionizing particles (MIPs) spectral features caused by galactic-cosmic-ray proton interactions in the ACS. The MIP peaks provide unambiguous measurements of MIPs during flight, and tracking the MIP peak rate during MEGANE operations provides an in situ measurement of the local GCR flux, a crucial parameter for converting GRS-measured γ-ray peak rates to elemental abundances for Phobos’ surface. This low-gain GCR mode was first developed for the Psyche GRNS (see Peplowski et al., 2025).

Calibration of the ACS GCR spectral product was complicated by the fact that GCR protons do not reach Earth’s surface due to our thick atmosphere. The pre-launch performance of the proton measurements can only be studied via radiation transport simulations. The GEANT4 model of GRS is detailed in Appendix A. First, we simulated the GCR measurements by irradiating the GRS with a uniform (4π steradian) distribution of protons with a GCR-like energy spectrum and isotropic angular emissions. That result is shown in Fig. 18C, where three broad peaks are observed. These peaks result from the irregular geometry of the ACS. Additional GEANT4 models, wherein the ACS was irradiated from discrete positions, show that the peaks at ~ 4.5, ~ 7.5, and ~ 18.5 MeV correspond, generally, to protons interacting in (1) the thin walls of the annular section (i.e. incident along the x–z plane), (2) the thin length of the base (i.e. along the y axis), and (3) the thick portion of the base (again incident along the x–z plane). These results are shown in Fig. 19. These simulations are consistent with the in-flight performance of MEGANE’s sister instrument, the Psyche GRNS (Peplowski et al., 2025), and will be confirmed for MEGANE in post-launch instrument commissioning activities.

For GCR spectrum calibration activities, we use GCR-secondary muons as a stand-in for GCR protons. Muons are produced by cosmic-ray proton interactions in the upper atmosphere, and unlike the protons they persist to sea level. Muons and protons both have minimum dE/dx values near their mean energies, and the minimum dE/dx values in plastic scintillator are similar. As a consequence, muons make a convenient and readily available proxy for GCR protons during ground testing of space-based particle detection systems (e.g. Moiseev et al. 2007). There are two differences between GCR-secondary muons and GCR-primary protons that are relevant for this calibration activity:

• At sea level, the muon flux has a ∝ cos²θ dependence that is peaked for incidence angles directly overhead (θ = 0°),

• The muon flux is a factor of ∼300 lower than the primary CR proton flux.

Because of the low muon flux, long integration times are required to provide GCR-comparable, statistically significant measurements, and the signal to background of these measurements is considerably worse than the in-flight measurements will be.

A 6.7-h-long room-background measurement was made in the GCR configuration for the purpose of identifying the MIP spectral features. With the instrument mounted on the deck mockup (Fig. 2), the muons preferentially arrive at downward angles along the spacecraft -Z axis that separately sample the thin walls and thick base of the ACS, due to the ∝ cos²θ incidence angle dependence. In order to interpret the muon measurements, we performed a GEANT4 simulation of muons incident on the GRS sensor. The simulations used a realistic energy spectrum for muons at sea level. We performed separate simulations for positive and negative muons (μ⁻,μ⁺), and in two geometries—plane-wave incident from above the sensor, and a 2π steradian source geometry. The simulations were weighted and summed to give a single result that mimics the mixed field (μ⁻,μ⁺) with the cos²θ incidence angle dependence. The result is shown in Fig. 18A and has high statistical precision and lacks the non-muon signals (backgrounds) present in measurements. The simulations highlight the presence of MIP peaks resulting from muons interacting with the complex geometry of the ACS. These muon MIP peaks are similar to those in the modelled GCR proton measurements (Fig. 18C). The muon simulations (Fig. 18A) can be directly compared to the measurements (Fig. 18B), highlighting that the measurements have a significant background continuum and lower signal to noise. Regardless of these differences, this comparison confirms that the MIP peaks can be measured by the ACS. This validation provides confidence in GEANT4 simulations of GCR proton measurements (Fig. 18C) and, by extension, that the ACS low-gain GCR spectral measurements will provide valuable measurements of GCR protons. The comparison also yielded the channel-to-energy conversion shown in the second x-axis of Fig. 18 (panel B). Final confirmation of the performance of this measurement will be made during post-launch instrument checkout; however, in-flight performance of the nearly identical ACS on Psyche has shown that the low-gain GCR mode does yield three MIP peaks resulting from GCR proton interactions (Peplowski et al., 2025).

4.5 Conclusion

In addition to its primary function of vetoing HPGe-coincident cosmic-ray background events (see Sect. 5.6), the ACS provides three scientifically valuable measurements detailed here; (1) low-energy neutron rates, (2) fast (> 400 keV) neutron rates, and (3) a diagnostic measurement of galactic-cosmic-ray protons. The low-energy neutron rate is derived from the area of the neutron capture peak in the high-gain ACS spectrum. Because the spectrum is not a high-energy-resolution system, temperature corrections are not required. Spectral analysis to determine the event rate in the ~ 80 keV_ee peak (Fig. 14), with background subtracted, will be performed to produce measurements that can be compared to models of the expected low-energy neutron environment from Phobos for a variety of surface compositions in order to constrain Phobos’ elemental composition. Those simulations will leverage the modelled efficiency shown in Fig. 17 (top panel). These measurements are particularly sensitive to the hydrogen content of the surface and compliment NS measurements.

The fast-neutron (FN) rate is derived from the area of the neutron capture peak in the FN spectral product. This product reports events within the "early" and "late" time windows derived from the TTSP analysis (Fig. 15). The FN peak can be derived from the difference between the early and late events as shown in Fig. 16 (bottom panel). Because the spectrum is not a high-energy-resolution system, temperature corrections are not required. Spectral analysis to determine the event rate for the ~ 80 keV_ee peak (Fig. 16) with background subtracted will be performed, and the resulting measurements will be compared to models of the expected fast-neutron environment from Phobos for a variety of surface compositions. Those simulations will leverage the modelled efficiency shown in Fig. 17 (bottom panel). These measurements are particularly sensitive to the average atomic mass of the surface, e.g. Gasnault et al. (2001).

Finally, relative variations in the local galactic-cosmic-ray proton flux can be determined by tracking the counting rate of the MIP peaks in the GCR spectrum as a function of time. This provides in situ information of cosmic-ray variability, analogous to the triple coincidence counter on the MESSENGER NS (e.g. Lawrence et al. 2016) but without the complicated geometric response. In situ measurements are an improvement over other investigations, which used terrestrial and Earth-orbiting space weather assets to infer the environment at other locations in the solar system. Comparisons to simulations will provide absolute measurements of the GCR proton flux. Because the GCR mode measurements are mutually exclusive with the high-gain measurements, the GCR environment will be sampled intermittently during the mission, continuously when MMX is far from Phobos, and possibly once per day for ~ 4 h during the QSO-LC periods.

5.1 Overview

The MEGANE gamma-ray spectrometer (GRS) subsystem uses a reverse-electrode (n-type) high-purity germanium (HPGe) sensor for γ-ray detection. N-type germanium was chosen due to its inherent tolerance to radiation damage compared to the p-type germanium used in most commercial applications (Pehl et al. 1972). The geometry of the MEGANE HPGe detector is shown in Fig. 12 and draws heritage from the size and geometry used in the MESSENGER gamma-ray spectrometer (Goldsten et al. 2007). The mass of the HPGe crystal was measured to be 525 g prior to system assembly.

HPGe is a semiconductor that requires cryogenic cooling in order to lower thermal noise such that the noise is well below the magnitude of the signals from γ rays. MEGANE includes a Ricor 508N cryocooler to achieve the temperatures needed to safely operate the detector (< 105 K). We operated at temperatures between 90 and 95 K during calibration activities.

High-voltage bias must also be applied to the cold HPGe crystal to enable γ-ray detection. The HPGe inner (positive) high-voltage contact is chemically diffused lithium, and the outer (negative) HV contact is ion-implanted boron. While under high-voltage bias, electron–hole pairs produced from γ-ray interactions within the bulk crystal drift towards the respective opposite-polarity electrodes, and the resulting charge signal yields the magnitude of energy deposition within the HPGe. The two high-voltage contacts are mechanically separated by a groove in the crystal. Contamination within the groove can cause leakage currents (see Sect. 5.2). The surface within the groove is protected by a passivation layer that prevents external contamination from causing leakage currents that can degrade the quality of the γ-ray measurements.

5.2 Intrinsic characteristics

The depletion voltage is an important characteristic of an HPGe detector. The depletion voltage is the minimum voltage necessary to fully deplete the active (intrinsic) volume of free charge carriers. HPGe detectors are typically operated at ~ 500 V above the depletion voltage to ensure an electric field sufficient for the drifting electrons and holes to reach saturation velocity. Thus, knowledge of the depletion voltage is critical for proper operation of the system. To determine the depletion voltage, we characterize the peak position (peak amplitude as reported as a pulse-height value) and energy resolution of the system as a function of detector bias voltage. For these tests, we used a pulser that is injected into the HPGe sensor by the DPU, via the sensor control unit module (SCUM) and high-voltage filter box (HVFB). This SCUM pulser has a programmable rate that facilitates rapid, high-statistics measurements (see Sect. 5.5) and, because it is injected into the HPGe crystal via the high-voltage filter box, it samples the intrinsic properties of the HPGe crystal. For the depletion voltage tests, the high-voltage bias of the system is slowly lowered to zero volts from a starting value of − 2500 V. Housekeeping and pulse-height spectra were collected once per second while the high-voltage was lowered to zero volts. The spectra were analysed, and the position (centroid) and width (FWHM) of the pulser peak are plotted as a function of the bias voltage during that spectral integration period. The results are shown in Fig. 20.

The position (peak centroid) and width (full width at half maximum; FWHM) of the pulser peak varies strongly with bias voltage from 500 to ~ 1800 V. For voltages > 1800 V, the variability in performance versus bias voltage was significantly reduced. These data suggest that the depletion voltage is 1800 V (± 100 V), suggesting a minimum operative voltage of 2400 V. Initially, this depletion voltage depends on the intrinsic level of impurities in the HPGe crystal. In the interplanetary environment, radiation-induced breakdown of the atomic lattice structure of the germanium crystal causes the depletion voltage to decrease in proportion to the total amount of radiation damage (Peplowski et al. 2019). Full annealing of the detector can restore the pre-damage depletion voltage value.

Leakage current is the electrical current across a detector in the absence of charge-generating signals. This leakage current is a source of noise that can degrade the performance of a HPGe-based sensor. For a fully depleted HPGe detector, leakage current can arise from several mechanisms that include charge injection at electrical contacts, charge flow along the surface of the detector, and thermal generation in the HPGe. Leakage currents for pristine HPGe detectors are typically < 10 picoamps (pA), and the value is constant as a function of bias voltage.

The MEGANE HPGe detector has exhibited a wide range of leakage current values throughout ground testing. Following long, high-temperature bakeouts, the system typically has leakage currents of an order a few tens of pA at 2500 V. See Fig. 21, which reports a leakage current measurement obtained during calibration. However, when the system has gone months without a bakeout, we have observed leakage currents in the > 1000 pA range at -2500 V bias, which has a measurable impact on performance. We believe that inadequate bakeout durations, a small pumping aperture (through the microvalve), and a persistent source of internal contamination are likely contributing causes of the observed leakage current issues. In response, we are planning to perform extended, 14-day-long high-temperature bakeouts prior to future ground testing and in-flight operations to accelerate the outgassing of the contamination into space. Leakage current should be carefully monitored in flight. However, as long as the value is < 1 nA, we expect to meet our required pre-launch energy resolution requirement of < 3.5 keV at 1332 keV.

5.3 Energy calibration, integral, and differential nonlinearity

The MEGANE HPGe data products are summed histograms of pulse-height amplitudes whose values are directly proportional to the energy loss (dE) in the HPGe, which are in turn proportional to the total charge produced within the HPGe on an event-by-event basis. The HPGe event data are histogrammed into a 16,348-ch-long spectrum. For ideal events, dE = E_γ, a single peak (photopeak) is observed in the pulse-height spectra at the element-diagnostic γ-ray energy. However, the majority (~ 50 + %) of detector-incident γ-rays are not dE = E_γ events, but instead are dE < E_γ events that populate a continuum of pulse-height values (the Compton edge) that does not contain element-diagnostic information.

A conversion from pulse-height spectrum channel to energy (dE in the HPGe) is required to associate measurements with element-diagnostic signatures. We determined this conversion using a combination of γ-ray calibration sources (⁶⁰Co, ¹³⁷Cs, ¹⁵²Eu, ¹⁵⁴Eu, and ²²⁸Th; see Table 1), as well as with a pile of intermixed salts (KCl) and polyethylene assembled around a ²⁵²Cf neutron source (see Fig. 22 and Table 2). The calibration sources provided gamma rays with energies up to 2615 keV, which is just one third of the full-energy range of the HPGe pulse-height spectrum. The neutron source + salt measurements provide higher energy calibration peaks via neutron capture reactions. In this scenario, the ²⁵²Cf releases high-energy (~ MeV) neutrons, which are downscattered to lower energies by the hydrogen-rich polyethylene. The low-energy neutrons are then captured by the Cl atoms in the salt, producing a wide array of > 5 MeV γ-rays that are used to determine the channel-to-energy calibration for the upper end of the pulse-height spectrum.

An energy calibration to convert pulse height (channel) to energy was made based on a linear fit from the data listed in Table 2. The result, shown in Fig. 23, is:

The maximum energy of the HPGe spectrum is 8038 keV. That value was chosen to include the highest-energy γ-rays of interest for the MEGANE science investigation, the Fe capture γ-rays at ~ 7650 keV, and enough of the surrounding continuum to enable accurate peak fitting.

The quality of energy calibration fit (Eq. 2) is characterized via the integral nonlinearity (INL) of the system, which shows the deviation between the predicted (Eq. 2) and actual γ-ray energies for the γ-ray peaks listed in Table 2. The INL of the system is ± 1 channel (± 0.5 keV) across the entire energy range reported by the HPGe measurements. Individual INL values are listed in Table 2. INL quantifies the accuracy of the γ-ray energy measurement scale.

The energy calibration (Eq. 2) is temperature dependent. During instrument thermal vacuum (iTVAC) testing, the temperature of the DPU and sensors were varied to qualify the instrument for flight. iTVAC included six temperatures cycles that included dwells at cold and hot soak plateaus, and the exact temperatures of these plateaus varied depending on the specific instrument component in question but are nominally −35°C and 30°C, respectively, for the GRS. 60-s-integration-period gamma-ray and pulser peak data were collected during iTVAC, and the relationship between the pulser and γ-ray peak positions were recorded as a function of the temperatures of the preamplifier board and the processor board, which are the two major electronics assemblies in the HPGe signal processing chain. Pulser position and width data are shown in Fig. 24, which reveals that peak position can vary by ± 1 channel (± 0.5 keV) across the range of temperatures experienced during iTVAC. The in-flight temperature range will be smaller.

The iTVAC data are colour-coded to indicate when the iTVAC heaters were on (red) or off (black), as heaters were a source of interference that caused degradation of performance (broader peaks). See Fig. 24 for details. Because the heaters were typically on when temperatures were warmer, it is challenging to disentangle heater-induced performance changes from thermal-induced variations. The iTVAC data suggest a maximum temperature-induced variation of 1.2 channels over a temperature variation of ΔT = 35°, which is a significantly larger ΔT than the expected variation in flight. Careful examination of in-flight data will be required to refine this correction in the interference-free environment in flight. We will perform this evaluation in flight, and our raw (L0) PDS data products will be temperature-corrected during processing in the MEGANE Science Operations Center (SOC) to remove the gain shifts as needed.

The HPGe detector analogue output is readout by an ADC. In principle, the output of an ADC is a linear relationship between input voltage and ADC channel. This value should be uniform for all codes, however deviations from the average value can exist and are quantified by a measurement called the differential nonlinearity (DNL) of the ADC. DNL is a measure of how evenly spaced the channel transitions occur. For a HPGe system, poor DNL can degrade energy resolution in a non-uniform way.

The DNL of the HPGe channels was measured by injecting artificial pulses, spread uniformly over a portion of the spectrum, into the DPU processor board. The pulses were produced using an external Berkeley Nucleonics BH-1 tail-pulse generator and a BNC LG-1 ramp generator. The output from these modules, validated using a calibrated oscilloscope, was a ~ 9.9 kHz pulse whose amplitude varied such that it spanned the full input range of the HPGe channels. This output was fed directly into the HPGe channel input in the DPU (Ch 0). Two hundred and nine individual spectra (1200-s-long accumulations) were summed to produce a final spectrum having a total of 250,800 s (69.7 h) of data collection at a Ch 0 input rate of 9909 cps.

A linear fit to the summed spectrum removed the mean value, and the residual (adjusted spectrum) was then used to calculate variations from its mean value. The total DNL is calculated by taking the square of the difference between the counts in each channel in the adjusted spectrum and the overall average counts per channel. A linear fit of that data yields a width of 315 counts (from a mean of 106,191 counts/channel), which is a 3-σ width (DNL) of 0.89%. This value is sufficiently small that it does not compromise the performance of the HPGe measurements, and accordingly, no DNL correction will be required for the calibrated (L1) HPGe data products.

5.4 Energy resolution

The defining performance characteristic of HPGe-based γ-ray spectrometers is their superior energy resolution as compared to other γ-ray detector technologies. Energy resolution, which quantifies how narrow a γ-ray peak appears in the spectrum, improves both signal-to-background and the ability to unambiguously distinguish individual peaks. The 1332-keV γ-ray from ⁶⁰Co is an industry-standard energy to characterize the performance of an HPGe γ-ray spectrometer (ANSI N42), and we adopted that standard for the results reported here. For this standard, energy resolution is quantified as the full width at half maximum (FWHM) at 1332 keV.

Laboratory-based, LN₂-cooled systems with the size of the MEGANE sensor routinely provide 1.7- to 1.9-keV energy resolution. Comparable spaceflight systems have exhibited notably worse energy resolution (> 3.5 keV), due to radiation damage and the inclusion of sub-optimal electronic components as dictated by the need to use flight-qualified components (e.g. Evans et al. 2006; 2017; Kobayashi et al. 2013). Energy resolution can also be affected by environmental conditions, as it depends on HPGe operating temperature, and is sensitive to microphonics and electromagnetic interference.

MEGANE HPGe signal processing is performed using a tripolar pulse shaping algorithm with an adjustable shaping time. Tripolar shaping was selected to minimize the impact of microphonics in the performance of the system. The energy resolution for the MEGANE SCUM pulser is shown as a function of shaping time in Fig. 25. The data reveal a classic shape, with a local minimum between 6 and 8 μs that highlights the trade between partial charge collection (short shaping times) and acceptance of noise into the pulse-processing algorithm (long shaping times). Our selected value of 7.1 μs shaping time is calculated to be ideal to reject interference from the cryocooler, which is present at the pulse-width modulation frequency of ~ 60 kHz.

There is no single energy resolution value for MEGANE. The resolution at any given time depends on the temperature of the system, pulse-processing settings (e.g. Figure 25), the cryocooler power level, and degradation from local electromagnetic interference. Additionally, like all HPGe-based systems, the resolution also varies with γ-ray peak energy. Figure 26 plots FWHM energy resolution versus energy during MEGANE ground calibration. For this measurement, the FWHM energy resolution at 1332 keV was 2.3 keV. This is a typical value for MEGANE; as-measured values during qualification testing and calibration ranged from 2.2 to 2.6 keV.

Microphonics from the cryocooler, which varies with cryocooler power level, introduces noise to the HPGe measurements that manifest as deviations from the Gaussian distribution typical of ideal γ-ray peaks. This phenomenon is highlighted in Fig. 27. These deviations are not included in FWHM calculations, so their presence and magnitude cannot be quantified using standard metrics. The fraction of events outside of the typical Gaussian distribution varies as a function of cryocooler power level, and this phenomenon must be properly accounted for in order to derive correct live times from the pulser (Sect. 5.7) and derive other γ-ray peak rates. In addition to this "continuous" degradation, there is additional microphonics noise which manifests at discrete cryocooler power levels; however, we suppress the impact of this phenomenon by dithering the cryocooler power level when the system is in regulation such that it does not spend any time lingering at a problematic (noise-inducing) power level.

5.5 Intrinsic gamma-ray detection efficiency

The number of events in a γ-ray peak can be used to determine the concentration of the corresponding element on Phobos’ surface. Once a γ-ray peak has been identified, the next step is to convert the peak area to the γ-ray flux at the detector. This requires knowledge of the full-energy γ-ray peak (photopeak) detection efficiency, which varies with γ-ray energy and detector incidence angle.

The wide variety of γ-ray energies and detector incidence angles expected during in-flight operations is impossible to accurately reproduce in a laboratory environment. Instead, we made a series of measurements, at different energies and angles, that were used to benchmark a GEANT4 simulation that provides the efficiency at arbitrary energies and angles. Measurements were made with radioactive sources that emit gamma rays with known energies and rates (Table 1). The uncertainties in the emission rates are < 3% and based on NIST-tracible standards. The details of the GEANT4 simulation are documented in Appendix A.

We adopt the -Z axis geometry as the performance reference for the HPGe detector. During QSO-LC operations, the MMX spacecraft will nominally align the spacecraft -Z axis with the spacecraft-to-Phobos-centre (nadir) vector. Figure 28 reports a comparison of simulated versus measured peak rates in the nadir geometry for measurements made with all of the γ-ray sources listed in Table 1, at position 61 (Fig. 2) on the source calibration plate. Position 61 corresponds to a source that is directly above the HPGe crystal, centred on the centre of the HPGe crystal. The plate sits 55 cm above the centre of the HPGe crystal. The nadir measurements are used as a stringent benchmark for our simulated detector response; therefore, long-duration measurements (1200 s) were made with each source to ensure that counting statistics in the γ-ray peaks do not dominate the final uncertainties of these measurements.

A comparison of the simulated-to-measured rates (Fig. 28) shows that the simulations generally reproduce the measurements to within ± 5% in the energy range required for MEGANE (400 keV–8 MeV). Below 400 keV, there are two outliers that likely result from differences in the simulated and actual instrument housings as the low-energy γ-rays are the most easily attenuated by the instrument housing. These values fall outside of the required measurement range for MEGANE observations. For the > 400 keV γ-rays, a normalization of 1.025 to the simulations would improve the simulation-to-model agreement. We adopt this value (2.5%) as the systematic uncertainty on the precision of the simulations.

Following the nadir-orientation calibration, additional measurements were made at 18 γ-ray energies and 97 different angles, for a total of 1,746 separate benchmarking measurements. The measurements were made with the calibration plate shown in Fig. 2, which was designed to cover the entire angular field of view of Phobos as observed by MEGANE. These measurements were 600 s long. Figure 29 highlights a subset of those measurements, along with a comparison to the equivalent simulations. A summary plot is shown in Fig. 30. Exempting regions where ground-support equipment is present in the measurements that is not included in the models (see white boxes, Fig. 29), the simulations were within 15% of the measurements, with increasing accuracy at higher energies and smaller angles (relative to nadir). Figure 30 shows that the vast majority of the simulations (> 90%) were within 10% of the measured value, with the simulation generally overestimating these measurements by ~ 5%.

The calibration measurements were made using point-like γ-ray sources, which is not a flight-like measurement geometry, and they are limited to energies of < 2.62 MeV. GEANT4 simulations are required to determine the efficiency in a flight-like orientation. To that end, a series of simulations were performed at a wide range of energies and a different geometry that more closely mimics that expected in flight. The geometry had a plane wave of γ-rays incident on the GRS from infinity and directed along the + Z axis. The number of detected events was divided by the number of HPGe-intersecting events to report the intrinsic detection efficiency, which is reported in Fig. 31. The data include the full-energy photopeak events (dE = E_γ), along with single- (dE = E_γ—511 keV) and double-escape (dE = E_γ—1022 keV) peaks. Escape peaks are a consequence of electron–positron pair production in the detector, followed by the escape (non-detection) of one (single escape) or both (double escape) of the resulting 511-keV annihilation γ-rays.

As seen in Fig. 31, the efficiency for γ-ray detection is higher for the escape peaks than the photopeak above 6000 keV. This fact, coupled with lower overall detection efficiency at these energies, makes the use of escape peaks important for improving the statistical precision of high-energy γ-ray measurements (like the 6.129 MeV gamma ray for oxygen). However, it is important to note that the precision of the escape peak simulations has not been evaluated using the calibration data, as none of the calibration sources (Table 2) produce sufficiently high-energy gamma rays to serve as a benchmark. Likewise, the polyethylene pile measurements could not be used to verify the relative efficiency, since the absolute γ-ray emission rates are not known. Thus, the escape peak efficiencies reported in Fig. 31 have an uncharacterized systematic uncertainty associated with the accuracy of the GEANT4 simulations.

The GEANT4 model yields a cross-sectional area for the HPGe detector, as viewed from nadir (-Z) of 25.5 cm², and the intrinsic γ-ray detection efficiency at 1332 keV is 10.4 ± 0.5%. 10.4% efficiency corresponds to an effective area of 2.64 cm² at 1332 keV. Efficiency and effective area values at a variety of other energies are reported in Appendix B. Note that for measurements in QSO-LC, the efficiency will vary somewhat from these values, as Phobos-originating γ-rays will not be a plane-wave source at this distance. Instead, a simulation of γ-ray emissions from Phobos, incorporating a proper shape model to account for Phobos’ irregular shape, will be combined with a radiation transport model of the MEGANE GRS (e.g. Appendix A) to account for all of the relevant geometric considerations. This process is detailed in Chabot et al. (2021).

5.6 Anti-coincidence spectra

In flight, the raw HPGe spectra include a significant rate of events originating from cosmic-ray (CR) protons. While the energy of these CR protons (E_p) is generally much higher than the maximum energy of the HPGe spectrum (8038 keV), a fraction of these events deposit small amounts of energy (i.e. dE < 8 MeV) in the HPGe and thus contribute to the spectral continuum that lowers the signal-to-background for all γ-ray peaks. MEGANE reduces this CR-generated background by vetoing events that are time-coincident in the ACS and HPGe, using a time-coincidence (veto) window with software-configurable upper and lower bounds. GEANT4 simulations of an isotropic, 4π steradian source of CR with a galactic-cosmic-ray-like energy spectrum reveal a total rejection efficiency of 97.4%, showing that the GRS geometry is well-suited for efficient rejection of HPGe + ACS coincident events. Because γ-rays have a low probability of interacting in the low-density, low-Z plastic ACS scintillator, all time-coincident events in the two GRS sensors are assumed to be CR protons and they are digitally removed (vetoed) from the HPGe anti-coincidence spectrum. The data product where these events have been vetoed is called the HPGe anti-coincidence (AC) spectrum.

The MEGANE pulse-processing electronics process events from the HPGe and ACS on an event-by-event basis. The relative time between any two events (Δt) is therefore measurable. A "global coincidence" window of 50 ADC units (5 μs) is used to broadly define coincident events, i.e. any events with Δt ≤ ± 5 μs are flagged as coincident by the GRS electronics. A tighter veto window was established to then reduce the likelihood of accidental coincidence events being vetoed from the HPGe spectrum. Signals from a single pulser, offset by known amounts using a delay generator, were separately input into Ch0 (high-gain HPGe) and Ch1 (high-gain AC Shield) to calibrate the veto time window. The real Δt values of the input signals were confirmed via oscilloscope. An offset was identified for the positive time values for Ch0 + Ch1 coincidence, and veto window low/high values of −0.8 to 0.2 μs yield a symmetric veto time window of ± 0.5 μs.

Figure 32 shows the veto spectra measured with a γ-ray source and an overnight measurement. For these measurements, cosmic-ray secondary muons are a stand-in for the cosmic-ray primary protons expected in flight (see Sect. 4.4). The muons produce a large signal in the ACS, significantly higher than room-background neutrons and γ-rays. On a per-particle basis, cosmic-ray secondary muons are a good proxy for CR protons; however, the muon flux is a factor of ~ 300 lower than the primary CR proton flux. Thus, the background subtraction observed in these laboratory measurements underestimates the performance of the anti-coincidence pulse processing relative to in-flight performance. The in-flight performance of the AC veto will not be known until MEGANE measurements are made in the space environment.

Figure 32 highlights additional features of interest in the AC spectrum. First, we observe low-level Compton suppression via the ACS veto. Specifically, we see that a small fraction of the spectral "shoulders" at ~ 480 keV and lower-energy are vetoed (i.e. they are observed in the rejected event spectrum). These are Compton events from the 662-keV γ-ray peak, which are true-coincidence events that interacted in both sensors. Because these Compton events contribute to the background continuum, Compton suppression improves the signal-to-background for γ-ray peaks. Finally, accidental coincidence can remove events of interest from the anti-coincidence spectrum. This is observed in Fig. 32, where some rejected events (blue spectrum) include a small pulser peak and 662-keV peak. We found that ~ 0.5% of pulser events were included in the rejected spectrum, and that 3% of the 662-keV γ-ray events were rejected. The 662-keV rejection was due to "accidental" coincidence between a photopeak 662-keV event and a random signal in the ACS. This phenomenon is rate dependent, and is expected to be insignificant in flight due to low (< 200 s⁻¹) rates during measurements at Phobos.

5.7 Throughput

The MEGANE instrument flight software (iFSW) processes signals on an event-by-event basis. The time required to process an event is determined by pulse-processing parameters like the shaping time, analysis window, and flat top duration. When a signal from any sensor exceeds a preset threshold, the iFSW classifies it as an event and an analysis window is opened. If a second event begins during the analysis window, then the system cannot distinguish between the two events and the signal is compromised (e.g. pileup). The fractional time for which the GRS electronics are busy processing an event and cannot process a follow-on event is known as the dead time. The inverse, the time for which the system is not busy and is available to process an event, is known as the live time. Live time can be reported as an actual time, or as a fraction of time for which the system is responsive to new events. Knowledge of the live time is crucial for converting measured γ-ray count rates to γ-ray fluxes and subsequently to element concentrations on Phobos.

The live time of the GRS is measured in real time using a signal pulser. Pulsers are artificial signals with known input rates. The difference between the known (input) rate and the measured rate is the throughput of the system and can be used to calculate live time. The fractional live time (t_L) is calculated from the pulser as:

where C_p is the measured number of events in the pulser peak, R_P is the pulser repetition rate, and t_R the real measurement time.

MEGANE has two pulsers. The first is a hardware (SCUM) pulser with a fixed amplitude but selectable frequency (settings are 1 Hz, 10 Hz, 100 Hz, 1 kHz, and 10 kHz). Rate selection is a trade between dead time measurement precision (higher rates), and increased dead time due to processing of the pulser signal (lower rate = lower dead time). The SCUM pulser appears near ~ 110 keV in the calibrated HPGe spectral products. The SCUM pulser is injected into the signal processing chain at the high-voltage filter box, and thus travels through the HPGe crystal. As a consequence, the SCUM pulser can be used to diagnose the state of the HPGe sensor, for example the depletion voltage measurements shown in Fig. 20 and the energy resolution measurements shown in Figs. 25 and 27.

MEGANE also has a digital pulser, which injects artificial events into the signal processing chain within the pulse-processing FPGA and thus samples the effects of the pulse-processing chain past the analogue front-end electronics, and not the performance of the HPGe. The digital pulser has the advantage of having an arbitrary pulse height and repetition frequency. Either pulser provides a peak that can be used to derive the fractional live time. Because the pulsers are both driven by the processor board, they can interfere with each other, regardless of their respective rates, to produce real pileup that causes unphysical spectral artefacts. Thus, only one pulser should be used at any time.

Pulser measurements were performed to verify the functionality of the pulser and lower-level discriminator. Pulser live-time determination accuracy was confirmed via a set of tests using a ⁶⁰Co γ-ray source, which was in a fixed position during tests and therefore provides a fixed peak rate of 1173- and 1332-keV γ-rays. A ¹³⁷Cs source was introduced at varying distances from the detector to change the total detector rate, facilitating measurements of the ⁶⁰Co rate versus total detector input rate. ¹³⁷Cs was used for rate control as it has a single γ-ray peak at lower energies that does not interfere with the ⁶⁰Co peaks.

For each measurement, the detector input rate was derived from the event rate per-second counters, which reports the rate of events processed by the ADC. For the pulser peaks, least-squares fitting of Gaussian functions to the measurements is used to determine their properties (centroid, width, and area) and the uncertainty of those properties. From these measurements, and Eq. 4, the fractional live time is derived. From that, the live time is calculated as the fractional live time times the real time, and the dead time is as 1 – fractional live time.

To verify the performance of the pulser, the 1332-keV peak rate from the fixed position ⁶⁰Co source was calculated using the real time and the calculated live-time values. For these measurements, the dead time manifests as an apparent decrease in the measured ⁶⁰Co peak rate with increasing trigger rate, despite the fact that the ⁶⁰Co peak position (and thus real rate at the detector) is fixed. In contrast, the ⁶⁰Co peak rates as calculated using the live time values recover the real ⁶⁰Co γ-ray rate. See Fig. 33. That the real ⁶⁰Co peak rates are recovered for all input rates confirms that the pulser-derived live-time corrections are accurate. As seen in Fig. 33, the accuracy of the recovered peak areas is 3.5%, which we adopt as the systematic uncertainty of the recovered live time during analysis of HPGe data products.

5.8 Conclusion

The MEGANE HPGe data are used to measure the elemental composition of Phobos’ surface. The HPGe detector performance is optimized by operating the system at its depletion voltage, plus at least 500 V. In the absence of radiation damage, the depletion voltage was measured to be ~ 1850 V (Fig. 20) and we recommend operating at 2500 V. Similarly, we recommend operating the system with a shaping time between 6.6 and 7.3 μs (Fig. 25) to optimize energy resolution.

Accurate energy calibration (Fig. 23) and optimized energy resolution allows element-diagnostic features to be unambiguously identified. Count rate areas are derived from fitting of these peaks and applying proper live-time corrections (Fig. 33). The measured counting rates can be converted to γ-ray fluxes at the detector using the simulated efficiency (Fig. 31), which was benchmarked to a comprehensive set of calibration measurements. These measured γ-ray peak fluxes can then be compared to simulated γ-ray production rates to constrain the elemental composition of Phobos’ surface.

The Japanese-led MMX mission includes contributions from Germany, France, and the USA. The MEGANE instrument is one of two US-contributed hardware systems for the MMX payload. Together, this multinational effort seeks to resolve one of the biggest remaining questions in planetary science – what is the origin of Mars’ moon Phobos? As a side benefit, the project will strengthen the ties between these countries and inspire a new generation of scientists and engineers to tackle difficult problems.

MEGANE serves a key role in the MMX project by providing the only remote-sensing data on the elemental composition of the surface. MEGANE measurements will provide key constraints of the formation of Phobos prior to the more detailed sample analysis that will occur once the samples are returned to Earth. The MEGANE γ-ray and neutron sensors have been calibrated in order to provide high-quality, accurate knowledge of the performance and its systematic uncertainties. This information will be vital for interpretation of MEGANE measurements, which will provide valuable knowledge about Phobos that will guide MMX mission decisions (like landing site selection) as well as context for both remote-sensing and sample analysis efforts.

Calibration activities were performed prior to the completion of the MEGANE ground data system, and they do not meet PDS data archiving standards. The data used for calibration was produced by our ground-support equipment (GSE) software GSEOS. The format and names of the GSEOS-produced data products are different from the names of the formal NASA Planetary Data System (PDS) data products that will be used for science analysis once in orbit, and there are some data format differences. Additionally, these products lack the documentation needed to be effectively used by scientists outside of the MEGANE team. As a result, the calibration data will not be archived on the PDS, and thus, they will not be publicly available. The modelled response data for all four MEGANE sensors are included in tabular form in Appendix B. All MEGANE data produced during MMX flight operations will be made available via NASA’s Planetary Data System (PDS).

We thank the JAXA and the MMX team, whose hard work and dedication are enabling this exciting future exploration of Phobos and Deimos. E. Menyhart, T. Palmer, S. Cheng, M. Graziano, E. Hoffer, and E. Shea of JHU/APL provided additional support during calibration activities. Additionally, J. Stinchcomb, B. Massey, and K. Peplowski assisted with various aspects of the design and assembly of the test equipment used for calibration activities.

The MEGANE project is supported by the NASA Discovery Program under contract NNN06AA01C to the Johns Hopkins University Applied Physics Laboratory (JHU/APL).

Authors and Affiliations

1. Johns Hopkins Applied Physics Laboratory, 11,000 Johns Hopkins Road, Laurel, MD, 20723, USA

   Patrick N. Peplowski, Samuel Fix, John O. Goldsten, David J. Lawrence, Kathryn Marcotte, Brian Schratz, Zachary W. Yokley, Sarah Bucior, Erik Ramseth & Jack T. Wilson

2. Lawerence Livermore National Laboratory, 7000 East Avenue, Livermore, CA, 94550, USA

   Morgan T. Burks

Contributions

PNP designed and performed the experiments detailed here, performed all data analysis, and wrote the manuscript. SF, MTB, JOG, KM, BS, and ZWY led the technical development of the instrument. DJL provided administrative and scientific leadership. All authors, including JTW, assisted with data collection and the review of the manuscript.

Corresponding author

Correspondence to Patrick N. Peplowski.

Competing interests

The authors have no competing interests.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix A. GEANT4 Radiation Transport Simulations

1.1 A.1 Neutron spectrometer

Our GEANT4 simulations of the MEGANE Neutron Spectrometer subsystem were developed from the GEANT4 example code AnaEx03. AnaEx03 natively tracks energy deposition events in active detector volumes on an event-by-event basis. We made the following changes to AnaEx03:

• We edited DetectorConstruction.cc to replace the original detector geometry with two ³He gas proportional counters with the geometry shown in Fig. 3 and filled with 10 atm of pure ³He gas. One sensor included the 0.5-mm-thick Cd wrap. The relative placement of the detectors mimics that shown in Fig. 1.

• We edited AnaEx03.cc to replace the default physics list (FTFP_BERT) with G4PhysicsListFactory and G4VMoldularPhysicsList, allowing the physics list to be selected by the user prior to running the application, and without recompiling.

• We edited PrimaryGenerator.cc to replace G4ParticleGun with G4GeneralParticleSource.

Our simulations were performed with the QGSP_INCLXX_HP physics list. Note that "HP" means high-precision neutron cross sections, and for that the ENDF VIII.0 neutron cross section library is utilized. Additional details on the simulation approach can be found in Peplowski et al. (2020).

The GEANT4 NS application was used twice for this manuscript. The first application was a validation exercise wherein two calibration measurements were compared to outputs from the simulation to verify the accuracy of the simulations. For this activity, the table, deck mockup plate, lead brick, high-density polyethylene (HDPE) neutron moderator, and ²⁵²Cf source encapsulation were all included in the simulation geometry. Modelling of the HDPE neutron moderator and ²⁵²Cf neutron source capsule and neutron energy spectrum are detailed in Peplowski et al. (2023). Following Peplowski et al. (2020), the simulated spectra produced from the GEANT4-output event-by-event data were processed to produce a spectrum comparable to the measurements. For that processing, 85% of the events are subjected to a simple Gaussian smooth with an energy resolution matching the measured value. The remaining 15% of events are subjected to a smooth that uses an exponentially-modified Gaussian function. This step reproduces peak tailing observed in the data, and is attributed to neutron capture events occurring near the boundary of the dead region, which results in incomplete charge collection. This processed spectrum appears in Fig. 8.

The second application of our simulation was to predict the detector response in flight. Here, the two sensors were the only structures in the simulation – no spacecraft materials are included. A plane wave of monoenergetic neutrons was incident on the NS from the -Z (nadir) direction. 10⁸ neutron histories were simulated over an area of 1225 cm² (larger than the entire NS footprint), resulting in 9.9% of the histories intersecting each NS sensor (121.2 cm² active area). Efficiency was calculated as number of neutron events (include capture peak and wall effect events) detected, divided by the number intersecting each detector (9.89 ×ばつ 10⁶). Effective area was calculated as efficiency times the active area. These results are detailed in Fig. 9. Note that, in flight, the incident neutron geometry will not be a plane wave. Instead, the simulations will handle neutron emissions from the surface as detailed in Chabot et al. (2021).

1.2 A.2 Gamma-ray spectrometer

1.2.1 A.2.1 Overview

Our GEANT4 simulation of the MEGANE Gamma-Ray Spectrometer subsystem was also developed from the GEANT4 example code AnaEx03. AnaEx03 natively tracks energy deposition events in active detector volumes on an event-by-event basis. We made the following changes to AnaEx03:

• We edited DetectorConstruction.cc to replace the original detector geometry with the HPGe detector, ACS detector, and associated housing materials as shown in Fig. 12.

  1. o

     Note that the PMT, radiator, SCUM, and Preamp electronics boxes are not included in the model, as they are never between the sensors and Phobos and thus do not affect the science measurements.

• We edited AnaEx03.cc to replace the default physics list (FTFP_BERT) with G4PhysicsListFactory and G4VMoldularPhysicsList, allowing the physics list to be selected by the user prior to running the application, and without recompiling.

• We edited PrimaryGenerator.cc to replace G4ParticleGun with G4GeneralParticleSource. The density of the ACS plastic was increased from 1.026 g cm⁻³ (manufacturer value) to 1.053 g cm⁻³ to get a match between the measured ACS mass and that reported by GEANT4 for the ACS volume.

Our simulations were performed with the QGSP_INCLXX_HP physics list. Note that "HP" means high-precision neutron cross sections, and for that the ENDF VIII.0 neutron cross section library is utilized.

1.2.2 A.2.2 Gamma-ray detection

The GEANT4 GRS application was used twice for γ-ray analysis in this manuscript. The first application was a validation exercise wherein measurements made with the γ-ray source plate were compared to the model. For this set of simulations, the geometry of Fig. 2 was added to the GEANT4 simulation. This includes the Delrin plate with punch-outs for the sources, as well as the 1′′ diameter, 0.25′′ thick Type-D HDPE plastic disc sources (Table 1).

The simulated spectra produced from the GEANT4-output event-by-event data were processed to produce a spectrum comparable to the measurements. 10⁷ γ-ray histories (H) were simulated, with 4π isotropic angular emission from a point source, separately for each source position. 10⁷ H is converted to an equivalent physical time (t) as:

where A is the activity of the source and I_γ are γ-ray emission intensities for each γ-ray. Both quantities are listed in Table 1. Equation A.1. is performed for γ-ray energy that was simulated in GEANT4. We then derived the simulated count rates by summing the events in the simulated spectrum (C_s), divided by the time as calculated using Eqn. A.1. The simulated rates were compared to the measured rates to evaluate the accuracy of the GEANT4 simulation as a function of energy and angle (see Figs. 30, 31, and 32).

The second application of our simulation was to predict the detector response in flight. Here, a plane wave of monoenergetic gamma rays was incident on the GRS from the -Z (nadir) direction. 10⁷ γ-ray histories were simulated over area of 36 cm² (larger than the entire HPGe), resulting in 68.7% of the histories intersecting the HPGe sensor (~ 25 cm² active area). Efficiency was calculated as number of events in the relevant peak (photopeak, 1st, or 2nd escape peak), divided by the number of γ-ray histories that intersected the HPGe detector (10⁷ ×ばつ 68.7%). Effective area was calculated as efficiency times the active area. These results are detailed in Fig. 31. Note that, in flight, the incident γ-ray geometry will not be a plane wave, which will alter the efficiency relative to that reported in Fig. 31. The proper flight geometry will be simulated during orbital operations using actual spacecraft ephemeris and Phobos shape models, as described in Chabot et al. (2021).

1.2.3 A.2.3 Fast-neutron detection

Fast-neutron efficiency modelling was performed with the MEGANE GRS model detailed here. Two changes were made to the model:

• The ACS was divided into four separate subsections, such that the location of the detected event could be coarsely tracked, and

• Event processing was modified relative to the γ-ray simulations to approximate the pulse-processing performed by the MEGANE instrument flight software (iFSW).

The iFSW classifies events at candidate fast neutrons when: 1) An initial (prompt) pulse is detected with an amplitude that exceeds the prompt threshold, and 2) A delayed (capture) pulse is detected with an amplitude that exceeds the capture threshold and that occurs within 25 μs of the prompt pulse. The GEANT4 simulations take a different approach to FN event classification. The simulations do not incorporate a time component, as the output is produced on an event-by-event basis for each neutron history. Thus, there is never uncertainty in associated prompt and capture pulses, e.g. there is no possibility of accidental coincidences being confused for legitimate FN events. The simulations instead produce two spectra:

• Proton-recoil spectrum, containing all pulse-height values for events where the energy deposition originates from a proton with the ACS, and

• Capture spectrum, containing all pulse-height values for events where the energy deposition originates from ¹⁰B(n,α)⁷Li reaction daughter products (α or ⁷Li).

We make two manual modifications to the GEANT4-simulated energy deposition values (dE). The first correction is a position-dependent adjustment to the dE value, based on the event location in the ACS as derived from experimental measurements of light output made using a ¹⁰⁹Cd source. This is why the ACS was segmented into four volumes. See Peplowski et al. (2025) for details. The second correction is due to differences in light output based on parent particle. Note that for α and ⁷Li, the pulse height is equal to the energy deposition (dE) value, whereas for protons, light quenching of the scintillator signal is included by modifying the effective energy loss (dE’) using Birk’s law as:

The FN rate is then derived from the two simulated spectra by summing all events whose dE_prompt and dE_capture values both exceed the respective threshold values. Those values were derived from the GEANT4 benchmarking experiment detailed in Sect. 4.3, as well as the final efficiency results shown in Fig. 17.

Appendix B. Modelled response data

Tables

3,

4, and

5 contain the results from detector response simulations for the plane-wave radiation sources, incident the NS and GRS, from the -Z spacecraft axis. This geometry roughly mimics that expected during QSO-LC measurements at Phobos. The GEANT4 models used to produce these results will be used by the MEGANE team to calculate the response for all measurements at Phobos, taking into account measurement geometry and Phobos’ true shape, as detailed in Chabot et al. (2021).

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions