© S. Benck et al., Published by EDP Sciences 2025

1 Introduction

The near-Earth space radiation environment is composed mainly of electrons and protons trapped within the Earth’s magnetosphere. The highly energetic particles that constitute the Van Allen radiation belts are primarily electrons of energy 100 keV–10 MeV and protons of energy 100 keV–400 MeV (Hess, 1968; Li W & Hudson, 2019; Koskinen & Kilpua, 2022). Furthermore, galactic cosmic rays, including heavy ion species, should be added to this radiation environment as a continuous quasi-steady background (Smart & Shea 1985; Usoskin et al., 2017; Sedrati & Bouchachi 2025). In contrast, solar energetic particles, originating from solar flares and/or coronal mass ejections, can lead to sporadic and intense increases in radiation levels (Mewaldt et al., 2005; Vainio et al., 2009; Anastasiadis et al., 2019).

During quiet periods the radiation belts consist principally of two regions: (a) a relatively stable, inner belt mainly composed of electrons with energy < 1 MeV (Li X et al., 2015; Li YX et al., 2023) and highly energetic protons (up to 400 MeV) originating from cosmic ray albedo neutron decay processes (Selesnick et al., 2013; Li X et al., 2020), and (b) a more variable outer belt filled primarily by electrons. This region is normally exempt from energetic protons with the variability linked to solar activity (Turner et al., 2019; Kanekal & Miyoshi, 2021). Finally, a relatively low flux slot region exists between both belts (Lyons & Thorne, 1973).

During active solar conditions, significant particle fluxes (mostly electrons with energies ranging from a few keV to about ten MeV) migrate from the outer belt to the slot region and can even penetrate the inner belt (Reeves et al., 2016; Kavanagh et al., 2018). They interact in complex ways with low-energy particles (ionosphere, plasmasphere, and plasmatrough), and with electromagnetic waves of different frequencies (Agapitov et al., 2015; Shklyar 2017; Baker, 2021; Koskinen & Kilpua 2022). When the space weather conditions return to normal, the created overpopulation of particles decays steadily in several hours to days, depending on the energy of the particles (Benck et al., 2010; Claudepierre et al., 2020a, 2020b).

All these particles form a hazardous medium for spacecraft (S/C) (Vampola, 2000; Horne, 2007; Akioka, 2009) and their damage effects depend strongly on particle species and energy e.g., high energetic electrons can induce internal charging of S/C payloads, while the cosmic ray background, as well as the hard long-lasting proton spectrum, can induce Displacement Damages and Single Event Effects in embedded components.

Detailed, wide-dynamic range monitoring of the various space particle populations at spacecraft level is therefore valuable for (i) the design and the operation of S/C (anomaly diagnosis, protective in-flight actions, etc), (ii) the assessment of radiation effects on S/C materials and electronic devices (including the human body), (iii) space weather services for now- and forecasting activities, (iv) the improvement of space radiation models and v) space physics research.

To characterize and predict the environment in the inner magnetosphere, two complementary approaches can be adopted: (i) fly radiation monitors on as many spacecraft as possible and (ii) use advanced physics models of the space weather processes at work, in an “assimilation” architecture where the boundary conditions are given by measurements. Examples of such models are, e.g., the VERB model (Drozdov et al., 2023) or Salammbô transport code (Bourdarie et al., 2005). One of the key inputs to those models is directionally resolved energy spectra, i.e., pitch angle distributions (PAD), where pitch angle refers to the angle between the velocity vector of the incident particle (or viewing direction of the detector) and the local magnetic field vector. In the lower energy range, accurate measurements of the seed electron population (100 keV range) help the monitoring of the particle injection and acceleration mechanisms (Horne, 2007; Tang et al., 2017) and ultimately lead to a better understanding of the dynamics of the high energy component of the radiation belt population. And, since the magnetic field traps charged particles, observing at a given point fluxes from different directions with respect to this magnetic field can be used to infer the distribution at other locations. In addition, those electron pitch angle distributions exhibit a strong energy dependency with respect to position in space and geomagnetic conditions (Gannon et al., 2007; Greeley et al., 2021; Smirnov et al., 2022).

The 3D Energetic Electron Spectrometer is a compact instrument that has been designed to characterize the above-mentioned electron population, and the following major requirements were targeted:

• Cover the electron energy range 0.1–7 MeV with 16–32 quasi-logarithmic bins

• Detection rate should be compatible with electron peak flux in GTO for E ≥ 100 keV (10⁹ cm⁻² s⁻¹)

• Ability to characterize the pitch angle distributions of electrons without having to resort to a spinning satellite

• Purity of electron channels ≥ 90%, also during intense SEP events

3DEES presently has no in-flight heritage and is operated as an in-orbit demonstrator onboard PROBA-3 (Project for On-Board Autonomy).

PROBA-3 is a mission dedicated to the in-orbit demonstration of precise formation flying techniques and technologies.¹ The PROBA-3 mission places two small satellites in a highly elliptical orbit around the Earth. During its apogee pass (_~6 h), the two satellites will fly in a precise formation, producing a very long baseline solar coronagraph called ASPIICS (Association of Spacecraft for Polarimetric and Imaging Investigation of the Corona of the Sun) (Shestov et al., 2021; Galy et al., 2023). One spacecraft carries the optical telescope (CSC – Coronagraph Spacecraft), and the second spacecraft carries the external occulter of the coronagraph (OSC – Occulter Spacecraft). In addition, PROBA-3 hosts two other science payloads: The Davos Absolute Radiometer (DARA), which is an absolute radiometer for measuring the total Solar Irradiance and is located on board the OSC (Suter, 2015), and the 3D Energetic Electron Spectrometer (3DEES), which is on board as a technology demonstration and is integrated on the CSC.

The satellite was launched on 5th December 2024 into a highly elliptical orbit: 60,530 km apogee, 600 km perigee, 59° inclination, 19.5 h orbital period. With these orbital parameters, the satellite will cover parts of the inner belt, outer belt, and mostly the border of the magnetosphere. Figure 1 shows a picture of the radiation belts represented by the integral flux of electrons > 1 MeV (z-axis) on an invariant coordinate map B/B₀ versus L; where B is the magnetic field, B₀ the field in the equatorial magnetic plane, and L the McIlwain parameter (McIlwain, 1961). The violet transparent area shows the part of the radiation belts covered by PROBA-3. Note, however, that 3DEES will be off during the apogee path when formation flying is done and ASPIICS measures. So, it is expected that 3DEES will deliver about 10 h of data per orbit.

Figure 1 Positional coverage of the radiation belts in (L, B/B0) coordinates for various missions: Here the radiation belts are represented by the electron fluxes with E > 1 MeV (background picture generated with the SPENVIS tool3). The red dashed outline shows the location of the proton belt for E > 10 MeV. The inset gives information on the type of orbit and the orbital period for each satellite. Within 3 orbits, PROBA-3 covers a large area of the radiation belts (only orbits 1 and 3 are shown).

A particularity of 3DEES on-board PROBA-3 is that it will observe particles from 6 incoming directions simultaneously on an axis-stabilized, i.e., non-spinning satellite. The location and orientation of 3DEES onboard the CSC are shown in Figure 2. The Coronagraph Geometric Fixed Frame is also shown on this figure, as well as the direction to the Sun. The relevant point here is that the +X-axis is nominally pointing towards the anti-Sun direction (specific conditions prevail during formation flying when 3DEES is off) and that 3DEES measures in the ZY-plane, hence all apertures point perpendicular to the Sun direction. There is no restriction on the value of the roll angle around the X-axis: This roll angle may take any value between 0° and 360°. Simulations have shown that although there is always one aperture targeting at pitch angle 80–90°, the minimum targeted pitch angle is highly variable and occasionally all apertures may aim within pitch angles of 70–90°. Those moments can be used to cross-calibrate the apertures among themselves. A common limiting factor in PAD measurement is the dynamic range of the instrument that has been optimized for high fluxes and that may occasionally affect statistics of the data close to 0° or 180° (depending on the type of PAD). However, we expect that the various types of possible pitch angle distributions, e.g., the pancake, flat-top, butterfly, and cap PADs (Horne et al., 2003; Smirnov et al., 2022; Killey et al., 2025), can be identified by 3DEES during different geomagnetic conditions.

Figure 2 Picture of the PROBA-3 coronagraph spacecraft with indication of its geometric fixed frame (CGFF) and highlight of the 3DEES with its six viewing directions.

Last note: Although the nominal time resolution for a measurement is supposed to be 1 min, the integration time of the measurements can be adapted along the orbit, as this parameter is part of the configuration file that can be uploaded to the instrument through time-tagged commands. Depending on the operation possibilities for 3DEES, it can then be foreseen to measure with a larger integration time (up to 4 min) at the outer border of the radiation belts and to decrease the integration time during the perigee pass (down to 20 s), as long as statistics allow it.

The purpose of this paper is to provide a description of the 3DEES design challenges (Sect. 2), its performance evaluation through Geant4 simulations (Geometry and Tracking, version 4²: Agostinelli et al., 2003; Allison et al., 2006), including in-orbit predictions of channel count rates at several positions in space (Sect. 3). And finally, its calibration (proton beam and ⁹⁰Sr/⁹⁰Y source), against Geant4 simulations verifying the functionality of the instrument, is presented in Section 4. The paper concludes with a summary of the key findings (Sect. 5).

2 The 3DEES instrument

The 3DEES onboard PROBA-3 is a modular detector that is composed of a management unit (also called Docking Module (DM)) related to one sensor head named the Panoramic Spectrometer Module (PSM), which is composed of three OSMs (Orthogonal Sensor Modules). The 3 OSMs present a volume of about 11.5 cm × 10.5 cm × 13.5 (height) cm, basis with feet not included. The DM includes the Data Processing Unit (DPU), Electrical Interface Board, and Power Conditioning Unit of the whole instrument. Its size is about 23 cm × 14 cm × 7.5 cm (height), feet not included. Note that in its original design, 3DEES is composed of 2 PSMs fixed on top of the DM, but mass constraints on the mounting panel of PROBA-3 necessitated the separation of the sensor heads from their base, i.e., the DM, and only one PSM was kept with a specifically designed new base (Fig. 3).

Figure 3 Picture of the 3DEES showing the management unit and the sensor head with its six looking directions.

A CAD layout of the three OSM boxes is shown in Figure 4. From one OSM to the other, only the mechanical layout of the box changes and the routing of the cables inside (not shown); the sensor system (i.e., collimator heads, sensor stack, front sensors, shielding on the aperture sides) stays the same and is only rotated by 30°.

Figure 4 CAD view of the inside of the three OSM boxes, with tags for various elements inside.

The particles of interest enter the detector module via two apertures, defined by two identical collimator heads. A collimator head consists of an aluminum structure holding five 2 mm thick tungsten rings, as shown in Figure 5, followed by a cavity. The collimator heads should prevent electrons outside the field of view (FOV) from entering the detector directly. The upper diameter of a ring hole is 5 mm, and the lower diameter is 9 mm. Opposite to the aperture, the collimator head is closed by an aluminum foil of 40 μm thickness. This thin aperture window aims to protect the sensors from stray light and to block electrons of energy < 100 keV and protons of energy < 2 MeV. The angle between the axes of the two apertures is 90°. Behind the collimator head, as close as possible to the entrance foil, the particles encounter a front sensor, either labeled S1_F or S1_S (two silicon sensors of thickness 150 μm and diameter 4 mm, from Micron Semiconductors UK, model MSD004, depletion voltage < 30 V) that serves as a trigger for particle detection. The front sensors are mounted on small Printed Circuit Boards (PCB) that also include the respective Front-End Electronics (FEE) i.e., charge sensitive amplifier. The geometrical FOV defined by a collimator aperture and a front sensor is about 15°.

Figure 5 Section view of OSM2 with the field of view defined by the first collimator ring and the S1 sensor.

Behind the front sensors, obliquely to each of them (45°), there is a common sensor stack. All together they form a sensor module (Fig. 6a) that is identical for all the OSMs. The sensor stack (Fig. 6b) is composed of four 1.5 mm thick rectangular silicon detectors (active area 15 × 40 mm²) separated by a distance of 1.65 mm (back to front). They are obtained from Micron Semiconductors UK (model MSX060), and their depletion voltage is about 200 V. The silicon detectors are mounted on PCBs that include the FEE for signal read-out. The detector layers are followed by a tungsten plate and two electronic boards, where one comprises the Analog to Digital Converter (ADC) and the other the ADC-drivers for all six sensors of the module. The seven layers are molded within a compact structure, allowing for its robustness (manufactured by 3D Plus).

Figure 6 a) CAD view of a sensor module that comprises a sensor stack and two front sensors on their respective PCB that also holds the corresponding charge sensitive amplifier (CSA). b) Cross-section view through a sensor stack showing from right to left, the four sensor layers where each holds a silicon sensor (S2, D1, D2, or D3) and its adjacent CSA circuit, the tungsten plate W (violet), and the two layers that comprise the driver circuit and the ADC circuit.

The entire instrument is shielded by aluminum (Al) of thickness ranging from 12 mm, around the collimator heads, to 6 mm, on the sides at the back (Figure 5). This front shield stops electrons with energies less than ~10 MeV and protons with energies less than ~60 MeV from reaching the detector stack.

Table 1 summarizes the performances of the 3DEES instrument on board PROBA-3.

3 Instrument simulation

Measuring energetic electrons can be quite a challenging task (Vampola, 1998). In fact, in a mixed radiation environment, it requires a good understanding of how those electrons and the other particle species interact with the instrument. In the space radiation environment, the background contamination to electrons mainly comes from energetic protons and bremsstrahlung radiation originating from highly energetic electrons interacting with the surrounding medium (Claudepierre et al., 2020a). In addition, electronic limitations such as noise, saturation, pile-up, and deadtime (Knoll, 2000; Usman & Patil 2018) add complexity to the particle identification, especially when the flux intensity gets high. For this reason, it is important to take these problems into account right from the beginning of the development. One of the best measures that can be taken to avoid unexpected problems or to minimize their impact is to simulate the detector’s performance ahead of time. This was done using the Geant4 toolkit that simulates the passage of particles through matter (Agostinelli et al., 2003; Allison et al., 2006). In this development that started in 2012, the simulations were performed using Geant4.9.6-patch2 source code from 17 May 2013. The default constructor G4EmStandardPhysics is used for the standard electromagnetic physics models. And, the QGSP_BERT physics list, which was the default at the time, is used to model hadron physics in proton–nucleus interactions. Given the employed physics list and the involved processes, simulations using recent versions of GEANT 4 are not expected to yield statistically significant differences in the results.

The Geant4 simulation study was done in two steps: First, the incident particles were simulated at the aperture level of a representative OSM module (Figure 7) and secondly, the particles were launched from different surfaces, completely enveloping the OSM but excluding the aperture. In both cases, the isotropic flux assumption is applied. The first study gives the instrument responses to FOV-entering particles and also allows for defining the effective FOV, which may be different from the purely geometrically defined FOV. The second study allows us to have an estimate of the detector’s responses to particles from outside the FOV passing through the shielding box. Electrons and protons were successively simulated to determine their response functions, but also their possible cross-contamination level.

Figure 7 Cut through the Geant4 geometry model of the representative OSM.

3.1 Geant4 simulations for events impinging on the aperture

Separately, 4 × 10⁹ electrons and 4 × 10⁹ protons are launched from the aperture S1_F from a disk with a diameter of 8 mm (outer-most diameter of the aperture of the collimator structure), and their corresponding energy deposition in each detector is registered. The incident particles are simulated uniformly distributed in the logarithm of the energy: 0.01 – 30 MeV range for electrons and 1.8–400 MeV range for protons. In this case, the number of particles per bin in log(E) is a constant: 2.9 × 10⁷ electrons and 4.3 × 10⁷ protons per bin of size Δlog(E) = 0.025 (40 bins per decade). At this stage, a uniform distribution in logarithms was preferred to increase the simulation speed.

3.2 The channel definition

For a given aperture, the particle discrimination and classification are based on the energy deposition information in all the sensors that have been hit. So, in order to identify and classify the incident particles in their correct physical channels (PC), the 3DEES needs as input a configuration file including a set of energy limits for each sensor stack and its apertures (S1 detectors).

An event is considered valid – meaning it meets the criteria for further processing – , if either the front sensor S1_F or S1_S is triggered individually, or if there is a continuous sequence of sensor activations without interruption. In the latter case, a hit pattern is defined, and the corresponding energy limits are applied to identify the particle type (the particle selection limits are pattern dependent). The particle identification is performed by using all available energy information from all the sensors hit.

Once the particle type is identified (electron or proton), it is registered in two kinds of physical channel spectra: The so-called High-resolution Channels (HC) and Coarse-resolution Channels (CC). The high-resolution PCs count the particles of a given type that deposit an amount of energy that is within a given energy range in the last sensor hit. The total number of channels defined within each last sensor hit is: 16 if the particle stops in S2, 8 if it stops in D1, 6 if it stops in D2, and 2 if it stops in D3. The total amount of HC channels is therefore 32. To this, one channel, named LC for “Low energy Channel”, must be added, corresponding to the particles that hit only the front sensor (either S1_F or S1_S) and no other sensor. Here, in order to separate electrons from protons, a pair of energy limits corresponding to each particle type is also applied. Please see Annex A for detailed information on particle identification and classification.

From those 32 high-resolution PCs, some are fully dedicated to electrons and others are optimized for protons. The resulting physical channel distribution is indicated in Table 2:

It is important to remember that within the 3DEES DPU, the channel allocation is not done based on energy criteria but on pulse height criteria (as measured by the 12-bit ADC). The energy limits in the configuration file are expressed in ADC-units (ADCu); and the energy calibration of the sensors permits knowing the sensitivity (S) of the sensor expressed in ADCu/MeV, which allows the conversion from MeV (for the deposited energy in the sensor) to ADCu (see Sect. 4).

The coarse-resolution physical channels count the particles of a given type that deposit a total amount of energy within a specified energy range. For analysis of results from the Geant4 simulation, this task is relatively simple, as the total deposited energy of a particle is the sum of the energies deposited by this particle in each sensor. However, within the DPU, the total energy is obtained by calculating:

$$E_{\mathrm{TOT}}\mathrm{ }= E_{\mathrm{S}1}·C_{\mathrm{S}1} + E_{\mathrm{S}2} + \sum E_{\mathrm{D}\left[\mathrm{x}\right]}·C_{\mathrm{D}\left[\mathrm{x}\right]} (\mathrm{Energy} \mathrm{expressed} \mathrm{in} \mathrm{ADCu}),$$(1)

with x = 1–3, S1 is either S1_F or S1_S, and with C_S1 = S_S2/S_S1 and C_D[x] = S_S2/S_D[x]. The sum is only calculated for valid events that have at least triggered S2.

The parameters C_S1 and C_D[x] are part of the configuration file. Ideally, all the sensitivities of the thick sensors should be equal, but slight deviations have been observed during the calibration campaign, and so 0.9 < C_D[x] < 1.1.

Due to the fact that E_TOT is based on the sensitivity values from all the sensors, or rather their ratio, the uncertainty on its value is larger than on the individual energy depositions in one sensor; therefore, the nomination “coarse-resolution”. For each particle type, 16 coarse-resolution channels can be added to the corresponding HCs.

3.3 The Energy response functions

For each physical channel, the particle detection efficiency as a function of incident energy is calculated and can be represented as a histogram. As the histogram extends roughly over two orders of magnitude in energy, its x-axis will be expressed in log10(E_inc), where E_inc (in MeV) is the incident particle energy. The efficiency for a given channel is then calculated as

$$\mathrm{Efficiency} \mathrm{per} \mathrm{bin} \left(\log 10\left(E_{\mathrm{inc}}\right)\right)=\frac{N_{\det } \left(\log 10\right(E_{\mathrm{inc}}\left)\right)}{N_{\mathrm{inc}} \left(\log 10\left(E_{\mathrm{inc}}\right)\right)}.$$(2)

Where N_det(log10(E_inc)) is the number of valid particle events originating from particles with incident energy log10(E_inc) ± Δlog(E_inc), log10(E_inc) is the centroid of the histogram bin, and Δlog(E_inc) = 0.0125 is the half bin-size (if 40 bins per decade). These histograms can be translated into final response functions by multiplying each bin height by the gathering power Γ of the surface from which the particles are launched (Sullivan, 1971): Γ =π∙A, with A the surface area from which the particles are isotropically generated within one hemisphere (i.e., directed towards the instrument). For the particles launched from the instrument aperture with radius R = 4 mm, Γ = 1.58 cm² sr.

Figures 8 and 9 show the response functions for electrons and protons, respectively. The response functions are expressed as the gathering power of the instrument, in units of cm² sr, as a function of incident particle energy. It can be noted in Figure 8 that each channel is generally well-resolved, implying that the applied energy selection limits are sufficient at classifying the particles, and the simulation statistics also look adequate. The first high-resolution channels, up to about HC10, are quasi-differential, showing a well-pronounced peak that also allows us to define the incident threshold energy above which the channel becomes active. The tail following the peak is due to electron scattering in the system (collimator, entrance foil, sensors). The upper channels, defined by electrons stopping in one of the detectors D1, D2, and D3, show a more integral characteristic. Their threshold energies may occasionally be identical to those of channels defined in a preceding sensor. This is also due to the strong scattering of the electrons when slowing down in the medium. However, the sharply rising efficiency curves, each restricting the incident energy domain to which the detector is sensitive, will allow straightforward unfolding of the measured spectra. Note that for the first measurements, in order to test the energy sum of equation (1), the limits that define CC01 for electrons and protons were set in such a way to have zero efficiency.

Figure 8 Energy response functions for incident electrons: The top figure represents the coarse resolution channels and the bottom figure shows the high-resolution channels (due to overlapping response functions, channels based on energy deposition in D2 and D3 i.e., HC25 to HC32, are not shown). Some channels are highlighted and can be identified by their indicated label. The incident particles were simulated uniformly distributed in logarithm of the energy: 0.01–30 MeV range.

Figure 9 Energy response functions for incident protons: The top figure represents the coarse resolution channels, and the bottom figure shows the high-resolution channels. Some channels are highlighted and can be identified by their indicated label. The incident particles were simulated uniformly distributed in logarithm of the energy: 1.8–400 MeV range.

The response functions of electrons (Fig. 8) peak at energies between 0.14 and 6 MeV. However, the channel with the lowest threshold, i.e., LC in Figure 8b, shows a very broad peak, and in the future, special attention will be paid to this channel in order to resolve the incident energy bin that will cover the 140–280 keV energy range. For the channels that show a rather pronounced peak in their response function, by defining the nominal energy resolution as the full width at half maximum of the response function for each channel (definition from Khoo et al., 2022), a nominal energy resolution (ΔE/E) of 20–25% can be estimated. As 3DEES was designed to operate in the heart of the outer radiation belt, where electron flux levels can be very high, up to 10⁹ cm⁻² s⁻¹, the response function’s peak value generally lies between 10⁻⁴ and 10⁻³ cm² sr.

When launching protons from the aperture within a diameter of 8 mm, some of those protons will traverse partly the saw-like collimator rings, whose largest diameter is only 5 mm. Hence, some high-energy protons will present themselves at the triggering S1 sensor as a proton of lower energy. This causes the appearance of the tails in the proton high-resolution channels of Figure 9b. However, the situation for the last channel, HC32, is somewhat different. Here, the energy limits were adjusted to mainly register protons that cross the sensors: The small peak at lower energy is due to protons that stop in D3, and the large plateau comes from the protons that traverse the sensor. Adjusting the energy limits for this last channel can increase or decrease the plateau region that extends from 50 to 400 MeV. This last channel will be a good monitor for the presence of high-energy protons >50 MeV and indicate when proton spectra will need to be treated with care (see also Sect. 3.4). For the proton coarse resolution channels (see Fig. 9a), which are defined by the total deposited energy according to equation (1), the first peak at lower energy corresponds to protons that enter the geometric FOV of the aperture without striking the collimator and stop in one of the thicker sensors. The subsequent tail arises from protons that interact with the collimator, while the peak at higher energy is due to protons that traverse all the sensors without stopping.

3.4 Channel impurity estimations

Cross-contamination, as well as outside-FOV particles, are two of the most deleterious causes of loss of the flux data quality. In order to get a first-order approximation of the particles that are incident on the instrument out of its aperture but generate a valid pattern, the following simulation is done: Particles are generated from 7 surfaces of a half cylinder prolonged by a rectangular box, as shown in Figure 10. On each surface, 4 × 10⁷ incident particles are isotopically generated. To limit simulation time, protons are generated in an energy range of 41–400 MeV and electrons in an energy range of 3.2–30 MeV (same spectral shape as before), after testing that the particles of lower energy are generally not able to traverse the shielding walls. In addition, the side of the OSM facing the S/C is not simulated as no particles are coming from that side. This is a worst-case scenario as either an OSM is shielded by another OSM or by two of them, as is the case for the OSM in the middle.

Figure 10 CAD view of an OSM without cover. The Aperture of S1_F is shown (AP). The coloured outlined volume includes the seven surfaces from which the particles are generated to estimate the amount of additional channel counts coming from particles that are incident on the shielding box but outside the aperture. The bottom and top of the OSM were covered by a representative 8 mm thick Al cover for the simulation. No particles were simulated from the rectangular surface behind the sensor stack.

For each defined virtual channel, its gathering power as a function of incident energy coming from all the simulated surfaces is analyzed. Figure 11 shows the result for a selected high-resolution electron channel, i.e., HC18. This channel is defined for electrons that have stopped in D1. It can be observed that a high contamination of outside-aperture electrons can be expected if the presence of >7 MeV electrons is high compared to electrons in the 2–7 MeV range. Additional counts in that channel could come from protons in the presence of >200 MeV protons. The situation is roughly the same for all the electron channels: Electrons of E > 7 MeV and protons of E>200 MeV may contaminate electron spectra if those populations show relatively high fluxes.

Figure 11 Response function for electron channel HC18 as defined when the particles are launched from the aperture (black squares) together with the response functions from outside-aperture particles classified by color (for the launch surfaces as defined in Fig. 10) and incident particle type (electron data is represented by squares and proton data is presented by triangles).

This may be especially problematic for the electron channels defined within D2 and D3, as those are the detectors that are sensitive to in-aperture electrons of energy >2 MeV or >3 MeV, respectively. The channel defined in D3 for electrons is also the most sensitive to >200 MeV protons.

For the LC channel, the situation is different. It is based on the condition that only S1 is hit and a given amount of energy is deposited within S1. For electrons, this channel LCe may contain some outside-of-aperture electrons of E > 4 MeV and some outside-of-aperture protons >50 MeV, if those are present in the space environment. Protons of about 2.5 MeV from the aperture, that just traverse the entrance foil and stop in S1 while depositing an energy equivalent to the LC electrons, may add some contamination to the LCe channel.

The situation for protons is quite different. Electron contamination in the proton channels is quasi non-existent unless they have an energy >10 MeV; however, high-energy out-of-aperture protons >50 MeV highly contaminate all the channels. This makes it impossible to resolve incident energy spectra when only considering the aperture response functions. Detailed simulation will be needed to calculate representative response functions and deduce proton spectra in the presence of a strong population of E > 50 MeV protons. The presence of this population will be indicated by counts in HC32 (see Fig. 9b).

In order to evaluate the effect of these limitations on the 3DEES mission on-board PROBA-3, in-flight count rates were simulated by folding the response functions with energy spectra derived from radiation belt model predictions – specifically AE8 for electrons (Vette, 1991) and AP8 for protons (Sawyer & Vette, 1976) – along the PROBA-3 orbit. Solar cycle maximum is assumed for the model predictions. An integration time of 1 min is taken as a baseline. Here, only two positions will be presented:

• Position 1: L = 4.43, B/B₀ = 3.1

• Position 2: L = 2.15, B/B₀ = 4.6

where L is the McIlwain parameter (McIlwain, 1961), B is the magnetic field, and B₀ is the magnetic field at the magnetic equator for a given L.

Position 1 at L = 4.43 is a region where electron flux is very intense and no high-energy protons are present. The differential directional spectra deduced from the integral spectra of AP8/AE8 are shown in Figure 12 (pink lines).

Figure 12 Differential flux-energy spectra as deduced from AE8 for electrons (a) and AP8 for protons (b) for position 1: L = 4.43, B/B0 = 3.1 (black squares with pink line) and for position 2: L = 2.15, B/B0 = 4.6 (black squares with blue line).

For that position, the predicted counts in the high- and coarse-resolution electron channels are shown in Figure 13 and Figure 14. The number of counts coming from the different launch surfaces can be identified by their respective colors. With respect to Figure 10, some of the surfaces are grouped: SF, SBF, and SBS become “side”, BF and BB become “bottom”, TF and TB become “top”. If a particle type from a given surface does not add any counts, it is not listed on the figure; in addition, sometimes the number of added counts is so small that they are not visible. For the HC channels, the labels on the x-axis are color-coded: The shades of blue, green, red, and violet link the channel to the last sensor that was hit, respectively S2 (1–16), D1 (17–24), D2 (25–30), and D3 (31, 32). The dark and light tone of a given color provides information about the particle type for which the respective channel is optimized (see also Table 2).

Figure 13 Counts in the high-resolution electron channels as observed at position 1 (L = 4.43, B/B0 = 3.1). The contribution from the different launch surfaces is indicated and can be identified by their color code.

Figure 14 Counts in the coarse-resolution electron channels as observed at position 1 (L = 4.43, B/B0 = 3.1). The contribution from the different launch surfaces is indicated and can be identified by their color code.

It can be observed that the contribution of outside-aperture particles to the counts in the various channels will be negligible. Also, statistics in the spectra will be sufficient with a 1-min integration time.

For position 2, the differential directional spectra deduced from the integral spectra of AP8/AE8 are shown in Figure 12 (blue line). This position, with L = 2.15, is close to the slot region, and the high-energy electron flux is very low, stopping at 3 MeV. The proton flux extends up to 200 MeV but with very low intensity.

For that position, the predicted counts in the high-resolution channels are shown in Figure 15 for electrons and in Figure 16 for protons. The statistics in the spectra are very low; however, outside-aperture contamination is still insignificant for electrons, while for protons, the situation may still be satisfactory. In the lowest proton channel, LCp, the contamination is 10%, while the HCp32 channel cannot be used, but indicates that the >50 MeV proton population starts to be relevant.

Figure 15 Counts in the high-resolution electron channels as observed at position 2 (L = 2.15, B/B0 = 4.6). The contribution from the different launch surfaces is indicated and can be identified by their color code.

Figure 16 Counts in the high-resolution proton channels as observed at position 2 (L = 2.15, B/B0 = 4.6). The contribution from the different launch surfaces is indicated and can be identified by their color code.

Among others, in that region of the orbit, it is not advisable to increase the integration time as the satellite is at its perigee, and its speed is the highest, and so the space resolution is already degraded with respect to measurements above ~5,000 km altitude. Additionally, the flux gradients are higher in that region, making it unsuitable for a long integration time.

In the heart of the inner belt, when L < 2, high-energy protons >200 MeV become an issue, especially for the electron channels defined in D2 and D3, and so reliable count rates are only available for a more limited number of electron channels. The simulation along the orbit also confirmed that in that region, proton count rates in all the channels mainly originate from outside-aperture particles. During its mission on board PROBA-3, the proton spectra from 3DEES will not be deduced for L < 2, and electron spectra will be treated with care.

4 Instrument calibration

Incorrect calibration of radiation detectors may induce systematic errors in trigger threshold and selection limit settings and consequently in count measurements. On the other hand, in-flight calibration of instruments may be helpful whenever channels are suspected of over- or under-estimating counts. Hence, it is important to both well calibrate the instrument on-ground and to survey this calibration in-flight.

4.1 Calibration in the proton beam

The 3DEES calibration process includes the definition of the sensitivity, that is, the relationship between the energy deposited in sensors (expressed in MeV) and the corresponding digital information from the ADC that is transmitted as input to the Data Processing Unit (named ADCu, cf. Sect. 3.2). This is done by registering the detector pulse height of the individual particle events originating from energy deposition of protons of various incident energies at the Light Ion Facility (LIF) in Louvain-la-Neuve (Belgium). Figure 17 shows the 3DEES on its Ground Support Equipment and its installation in the proton beam.

Figure 17 a) The 3DEES on its ground support equipment (GSE). The red caps on the 3DEES apertures are protection covers during storage and transport. b) 3DEES in the proton beam facility at Louvain-la-Neuve, Belgium.

The maximum proton energy at the LIF facility is 62 MeV, and degraders can be used to decrease this incident energy down to ~10 MeV. The protons of 62 MeV have sufficient energy to traverse all the sensors. The nominal intensity of the beam was about 10⁴ proton s⁻¹ cm⁻². With this, individual pulses could be registered without pile-up, and they were analyzed offline. While all the sensor stacks were calibrated twice, i.e., before and after integration into the final OSM boxes, the S1 sensors were only characterized during the last calibration campaign. The following proton energies were used: 62.0, 57.2, 49.7, and 46.8 MeV (hit all the sensors up to D3), 40.8 MeV (stop in D2), 33.9 and 30.1 MeV (stop in D1), and 20.5 MeV (stop in S2). So, degrading the incident energy was mainly beneficial to the calibration of the front sensors S1_F and S1_S and to the sensor stack’s first sensors S2 and D1, as for them a larger set of calibration spectra could be measured.

The calibration process then consists of normalizing the generated measured pulse height spectra for each sensor to the Geant4-simulated energy deposition spectra. For the simulation, the Geant4 geometry model of Figure 7 was used together with a simplified model of the proton beam facility. More precisely, for all simulated protons, their energy deposition in a given sensor is multiplied by an assumed sensitivity S for that sensor, and a simulated histogram in ADCu is built taking the same bin size as for the measured spectra. The bin size of the measured spectra is selected manually based on the statistics accumulated for each incident proton energy. The search for the sensitivity factor starts with S_nominal × 0.5 and ends at S_nominal × 1.5, with an increment of 1 ADCu/MeV. S_nominalis a first guess for S, derived from the position of the peaks in the measured and simulated spectra. And for each assumed S, a normalization factor (N) range is browsed to search for the best agreement between measured spectra and the simulated one, the criteria being based on a chi-square evaluation.

An example of outputs of the fit procedure is shown in Figure 18.

Figure 18 Deposited energy spectra in ADCu (primary X-axis) and in MeV (secondary X-axis) for the different sensors for incident protons of 62 MeV. Black dots with error bars and grey histograms represent the measured data, and the red histograms are derived from the simulated data. The resulting values for the sensitivity S are indicated.

The final sensitivity value (energy deposition to channel conversion factor) for each sensor is an average of all available values from the two in-beam calibration campaigns. The estimated uncertainty on the final sensitivity values, based on the standard deviation of the mean, is of the order of 3%.

The validation of the calibration is subsequently performed through the recording of beam particles in adequate physical channels, as shown in Figure 19. It can be noted that the incident 62 MeV protons are mainly recorded in the 3DEES channel HC32 devoted to the protons that traverse the sensor stack. The low-energy tail accompanying the dominant peak protons is also well reproduced. Experimental results are in good agreement with simulation predictions.

Figure 19 Proton high-resolution spectra HCp as observed with the OSM1_S1S aperture during different runs (see color code) with integration time 30 s. The simulated counts are represented by the green histogram bar. The spectra have been normalized to the integral counts of the run named output_3dees_42.

3DEES is equipped with an in-flight calibration survey channel. This is achieved through the registration of the running average of energy deposition (in terms of pulse height) for all the sensors when the last sensor has been hit by protons depositing between 12 and 25 MeV in the last sensor (adjustable calibration limits). This corresponds to incident protons in the energy range ~50–70 MeV. This procedure presumes that the calibration of the sensors has not degraded in such a way that triggering is not possible anymore, and that the D3 calibration has not significantly changed. Calibration values will be updated mainly when passing closer to the inner belt.

4.2 Validation of the calibration with a Strontium-90 source

Tests with a Strontium-90 source were done to both verify the calibration and instrument performances and validate the related Geant4 simulations, and also to get a better definition of the detection thresholds that should be used, as well as to confirm the effective electron FOV of the apertures. Sr-90 has a half-life of 28.8 years and undergoes beta decay into Y-90, with a maximum electron energy of 546 keV. Subsequently, Y-90 decays into the stable Zr-90 with a half-life of 64 h while producing electrons wit h a maximum energy of 2.28 MeV (Mougeot, 2015). Due to the energy limitation, the electrons from the strontium source can mainly reach only as far as D1 in the sensor stack, while very few possess sufficient energy to reach D2.

Figure 20a shows the experimental setup with the Ground Support Equipment (GSE) in the lab. The position of the source was manually adjusted to a distance of 15 mm between the source holder’s (Fig. 20b) aperture and the selected collimator aperture of 3DEES. The geometry of the source-holder and the OSM of 3DEES was implemented in the Geant4 simulation, and electrons were launched from a spherical volume (1 mm diameter) representing an active bead of Sr-90. The launch point within the sphere was selected randomly, as well as the launch direction. The simulated energy spectrum was that of a Sr-90/Y-90 radioactive source (Mougeot, 2015). The number of particles launched was increased until the sum of counts in the physical channels equals that of a measured spectrum registered with an integration time of 30 s. This corresponds to ~10⁹ simulated particles; divided by the integration time of 30 sec, this simulation represents a source of 33 MBq. This value corresponds to the nominal activity given by the source provider i.e., 37 MBq±30% (aging of the source not included: About −10% after 5 years).

Figure 20 a) 3DEES on its GSE with the Sr-90 source holder aligned in front of it. The distance between the holder’s collimator exit hole and the 3DEES aperture was set to ~15 mm. The GSE allows for precise rotation of the aperture under test by a given angle with respect to a rotation axis passing through the center of the selected aperture. b) Cut view through the source holder, highlighting the elements (absorbing materials) that affect the electron spectrum.

Figures 21 and 22 show the measured high- and coarse-resolution spectra for the aperture labeled OSM2_S1S and their comparison to results from Geant4 simulations. Each graph displays results from three measurements, represented as bar charts in varying shades of blue. The grey histograms are the results of the simulation. For the latter, squares are displayed at the center of each histogram bar, with error bars visible when they extend beyond the size of the squares. In general, the shape of the measured spectrum is well reproduced by the simulation. However, for all the apertures, it was systematically observed that there are counts in HCe11 and CCe12 that are not represented by the Geant4 simulation. This could neither be explained by the imprecision in the alignment procedure nor by uncertainty in the calibration values. However, those counts could be explained by adding pile-up effects to the simulation data (light red histograms in Figs. 21 and 22 with corresponding squares and error bars).

Figure 21 For the aperture labeled OSM2_S1S, measured high-resolution count spectra (blue bar-charts) as obtained with the Sr-90 electron source, compared to a normalized simulated count spectra (grey and light red histograms (pile-up included)) with corresponding squares at the center of each histogram bar, please see text for more information).

Figure 22 Coarse resolution electron spectra (blue bar charts for three measurements) as obtained with a Sr-90 electron source. The light red and grey histograms represent the simulated spectra, respectively, with and without pile-up effects included.

The simulation of the pile-up effect for particles coming from the electron source was done in the following way:

We start with the distribution function for time intervals “t” between adjacent random events that are emitted at a rate of “r” (Knoll, 2000). In our case, r = 33 × 10⁶ electrons per second (corresponding to the intensity of the source):

$$I_1\left(t\right) \mathrm{d}t =r e^{-\mathrm{rt}}\mathrm{d}t.$$(3)

This represents the probability of emitting a particle during the infinitesimal time interval “dt” after a time interval of length “t” following a previous particle, and during which there has been no other particle emitted. The corresponding cumulative distribution function can then be expressed as:

$$\mathrm{CDF}\left(t\right)={\int }_0^{t}r e^{-\mathrm{rt}}\mathrm{d}t=1-e^{-\mathrm{rt}}.$$(4)

To each electron launched within the active bead of the source, a time interval “t” is associated by generating a random number between 0 and 1 (0 ≤ CDF(t) ≤ 1) and by extracting “t” from equation (4). Then, prior to classifying an event that has hit at least S1, the time stamp of the following event that has also hit the sensors is looked up. If the two events are within a time interval Δt = 1 μs, then the energy depositions in the sensors corresponding to both events are added up before classification, and the procedure continues with the next event following those two. No triple coincidences are considered. The 1 μs corresponds roughly to the capacity of the electronic system and subsequent signal processing to resolve two consecutive events.

The results of this pile-up effect simulation are represented by the light red histograms in Figures 21 and 22, and the comparison with the measured spectra indicates that the counts registered in HCe11 and CCe12, as well as a few counts in CCe11, can find well their origin in random overlapping signals. When a pile-up occurs, the pulses corresponding to lower energy deposition of coinciding events are added up and generate a single high-energy pulse. Consequently, the system registers fewer counts for lower energy channels and more counts for higher energy channels. In our case, only a few counts could “migrate” due to this pile-up effect from lower energy channels to the higher ones, while the lower energy channels are not visibly affected.

3DEES implements a function called “test counts”. It simply consists of defining for each sensor an extra channel that increments when the sensor is triggered, independent of any coincidence pattern. This allows for identifying in-orbit the regions and/or environmental conditions that are favorable for pile-up, estimating its effect, and if needed, starting specific analysis to take this into account. However, comparing the count rate deduced from Figure 21 to that from Figure 13, which is an estimation for a region where PROBA-3 will encounter the highest fluxes, it can be expected that, in general, the pile-up effect will be relatively small or negligible.

4.3 The angular resolution of an aperture

3DEES was designed to characterize angular distributions in the radiation belts. Hence, it is important to verify its effective FOV. This was done with the Sr-90 source by turning the aperture under test by a selected angle with respect to a rotation axis passing through its center (Fig. 20a). In the geometry for the Geant4 simulation, the angle between the source axis and the aperture axis was changed in the same range and for each configuration, up to 5 × 10⁷ electrons were simulated at the source level for each angle. While the measurements were done at angles between −20° and 20°, the simulations were only done in one angular direction and mirrored around 0°. The normalization between measured and simulated number of counts is done with respect to the total number of counts (sum over the counts in all the channels) when the selected angle is 0°. The results are shown in Figure 23 for two selected channels, LCe and HCe7. It can be observed that the measured distribution closely matches the simulated one and can be well approximated by a Gaussian function. The geometrical FOV defined by the size of the hole of the upper collimator ring and the size of the S1 sensor is 14.5° (−7.25° to 7.25°). It has been calculated that about 90% of the electrons are within this FOV. While the electrons respect the FOV determined by the collimator relatively well, this is not necessarily the case for the protons. In fact, protons above 55 MeV can traverse the 12 mm aluminum front shielding that surrounds the collimator head, and with energy higher than 150 MeV, they can even traverse the 1 cm of tungsten of the collimator, along with some additional mm of aluminum coming from the mechanical structure that keeps the collimator in place. This is especially problematic for the first proton channel LCp.

Figure 23 For OSM1_S1S: Counts in the selected channels LCe and HCe7 (blue dots) as a function of rotation angle of 3DEES aperture with respect to electron source aperture. Comparison with Geant4 simulations (orange dots) after renormalization. The simulated data points (from Geant4) are mirrored at 0 degrees. The error bars for the measurements are within the size of the dots and are not shown.

5 Summary

This paper summarizes the development of an instrument that has been designed to measure angle-resolved electron spectra in the heart of the radiation belts, i.e., that measures pitch angle distributions without resorting to a spinning satellite. The importance of Geant4 simulations was revealed by its application at various stages throughout the development: design of the collimator and sensor module, definition of the detection channels, evaluation of out-of-aperture or inter-species particle contamination, energy calibration, and measurement interpretation (e.g., pile-up effect). Tests with a proton beam and an electron source allowed final calibration and validated the Geant4 simulations and the performance of 3DEES.

Presently, the development of the tools for flux extraction is ongoing. From the response functions shown in Section 3.3 the efficiency matrices can be determined, and the electron and proton spectra can be resolved either by applying a procedure like that used for the EPT data (Cyamukungu et al., 2014) or by applying a Single Value Decomposition (SVD) approach (Höcker & Kartvelishvili, 1996). Both are presently under investigation, and results will be shown in a future paper.

PROBA-3 with 3DEES on board has been launched on 5th December 2024 onto a highly elliptical orbit where 3DEES will monitor space weather effects in a large area of the radiation belts. This in-orbit demonstration unit should provide its first angle-resolved energy spectra (6 angles, electrons 0.14 to about 10 MeV) with absolute flux values by the end of 2025, following a several-month commissioning phase of the S/C and 3DEES instrument.

Acknowledgments

The authors would like to thank the people at Redwire (Belgium), UCLouvain (Belgium), and SRI (Czechia) for their sporadic contributions during the early stages of the 3DEES project, as well as the many others who later offered temporary contributions to the project. Carolien Goubert (formerly at Redwire) is especially thanked for her project management role during a later stage of the project. Here, special acknowledgments also go to ESA (European Space Agency) technical officer David Rodgers, who is now retired, as well as Petteri Nieminen from ESA, who always encouraged us in our projects.

The UCLouvain/CSR team especially acknowledges Werner Verschueren from BELSPO (Belgian Science Policy) for his engagement and support throughout the 3DEES development. They would also like to thank their former academic head, Véronique Dehant, for her support of the project during her mandate, as well as Viviane Pierrard and Eddy Neefs, who followed and supported the 3DEES project at BIRA-IASB. They also acknowledge the UCLouvain cyclotron team for their good and flexible service in delivering the proton beam. Data storage resources have been provided by the supercomputing facilities of the Université catholique de Louvain “Calcul Intensif et Stockage de Masse” (UCLouvain/CISM). Finally, we thank the reviewers for providing comprehensive and valuable comments that have improved this manuscript. The editor thanks Fan Lei and an anonymous reviewer for their assistance in evaluating this paper.

Funding

The UCLouvain and BIRA-IASB teams thank BELSPO for the provision of financial support in the framework of the PRODEX Program of ESA, under contract number 4000134131, for the project called “3DEES – Phase C2/D”. For the same period, the Redwire team thanks ESA for its financial support under the GSTP contract. The whole 3DEES consortium also thanks ESA for support in the 3DEES development (Phase A/B/C1) through the GSTP program(Contract numbers 4000105532 and 4000115696).

Data availability statement

The data that support the findings of this study are not publicly available, but may be available from the corresponding author upon reasonable request.

Appendix A

An example of particle identification and classification is given for the case where the hit pattern is “D1 hit” and the particle entered the detector through aperture S1_F. This means that sensors S1_F, S2, and D1 were activated in coincidence, with signal heights above threshold.

Figures A1–A4 present 2D histograms, with the number of counts indicated by the color-scaled Z-axis, plotted as a function of incident energy (X-axis) and energy deposited in various sensors (Y-axis). In Figure A4, the Y-axis represents the total deposited energy. The panels on the left show the results of the electron simulations, while the panels on the right correspond to simulations with incident protons. The red horizontal lines correspond to the energy limits to be identified as an electron, while the blue horizontal lines correspond to those energy limits to be identified as a proton. As can be observed, the energy deposition from all the triggered sensors, as well as the total deposited energy, is used to identify a particle. Combining all these conditions to identify a particle, electrons and protons can clearly be separated. Only protons above ~200 MeV can cross-contaminate electrons.

Figure A1 Energy deposited in S1_F as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton.

Figure A2 Energy deposited in S2 as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton.

Once the particle type is identified, the electron or proton is then registered in two kinds of physical channel spectra, the so-called High-resolution Channels (HC) and Coarse-resolution Channels (CC).

• The HCs result from energy limits defined for the last sensor hit, as can be seen in Figure A3 (black dashed lines). For D1 there are 9 limits for the 8 HCs, wherein 6 are optimized for electrons and 2 for protons. The number of limits defined in that way is different for each pattern, e.g., if S2 is the last sensor hit, 17 limits are used to define the 16 HC channels for that pattern.

• The CCs result from energy limits defined in the total deposited energy range. Here, the limits for electrons and protons are separated into two independent sets. However, those limits and herewith subsequently defined channels are common to all patterns. As can be seen in Figure A4a, the electrons that hit D1 do not populate the lowest CCs, as their total deposited energy is too high. Those channels are, however, populated by the electrons that stop in S2.

Figure A3 Energy deposited in D1 as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton. The black dashed lines show the 9 classification limits for this pattern (same for protons and electrons); they define the 8 high-resolution channels where 6 are optimized for electrons and 2 optimized for protons.

Figure A4 Total deposited energy as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton. The black dashed lines show the 17 classification limits (different for electrons and protons); they define the 16 coarse resolution channels for each particle type.

Table A1 gives an example of the selection limits for the particle identification expressed in keV. In the configuration file, those are expressed in ADCu, taking into account the deposited energy to ADCu conversion for each sensor. The limits in the thick sensors are independent of the aperture through which the particles enter the detector.

Table A2 gives an example of the classification limits expressed in keV for one OSM. For the pattern “S1 hit”, where only one front sensor is hit, particle identification and classification are done within the same step.

References

All Tables

All Figures

Figure 1 Positional coverage of the radiation belts in (L, B/B0) coordinates for various missions: Here the radiation belts are represented by the electron fluxes with E > 1 MeV (background picture generated with the SPENVIS tool3). The red dashed outline shows the location of the proton belt for E > 10 MeV. The inset gives information on the type of orbit and the orbital period for each satellite. Within 3 orbits, PROBA-3 covers a large area of the radiation belts (only orbits 1 and 3 are shown).

In the text

	Figure 2 Picture of the PROBA-3 coronagraph spacecraft with indication of its geometric fixed frame (CGFF) and highlight of the 3DEES with its six viewing directions.
In the text

	Figure 3 Picture of the 3DEES showing the management unit and the sensor head with its six looking directions.
In the text

	Figure 4 CAD view of the inside of the three OSM boxes, with tags for various elements inside.
In the text

	Figure 5 Section view of OSM2 with the field of view defined by the first collimator ring and the S1 sensor.
In the text

Figure 6 a) CAD view of a sensor module that comprises a sensor stack and two front sensors on their respective PCB that also holds the corresponding charge sensitive amplifier (CSA). b) Cross-section view through a sensor stack showing from right to left, the four sensor layers where each holds a silicon sensor (S2, D1, D2, or D3) and its adjacent CSA circuit, the tungsten plate W (violet), and the two layers that comprise the driver circuit and the ADC circuit.

In the text

	Figure 7 Cut through the Geant4 geometry model of the representative OSM.
In the text

Figure 8 Energy response functions for incident electrons: The top figure represents the coarse resolution channels and the bottom figure shows the high-resolution channels (due to overlapping response functions, channels based on energy deposition in D2 and D3 i.e., HC25 to HC32, are not shown). Some channels are highlighted and can be identified by their indicated label. The incident particles were simulated uniformly distributed in logarithm of the energy: 0.01–30 MeV range.

In the text

	Figure 9 Energy response functions for incident protons: The top figure represents the coarse resolution channels, and the bottom figure shows the high-resolution channels. Some channels are highlighted and can be identified by their indicated label. The incident particles were simulated uniformly distributed in logarithm of the energy: 1.8–400 MeV range.
In the text

Figure 10 CAD view of an OSM without cover. The Aperture of S1_F is shown (AP). The coloured outlined volume includes the seven surfaces from which the particles are generated to estimate the amount of additional channel counts coming from particles that are incident on the shielding box but outside the aperture. The bottom and top of the OSM were covered by a representative 8 mm thick Al cover for the simulation. No particles were simulated from the rectangular surface behind the sensor stack.

In the text

	Figure 11 Response function for electron channel HC18 as defined when the particles are launched from the aperture (black squares) together with the response functions from outside-aperture particles classified by color (for the launch surfaces as defined in Fig. 10) and incident particle type (electron data is represented by squares and proton data is presented by triangles).
In the text

	Figure 12 Differential flux-energy spectra as deduced from AE8 for electrons (a) and AP8 for protons (b) for position 1: L = 4.43, B/B0 = 3.1 (black squares with pink line) and for position 2: L = 2.15, B/B0 = 4.6 (black squares with blue line).
In the text

	Figure 13 Counts in the high-resolution electron channels as observed at position 1 (L = 4.43, B/B0 = 3.1). The contribution from the different launch surfaces is indicated and can be identified by their color code.
In the text

	Figure 14 Counts in the coarse-resolution electron channels as observed at position 1 (L = 4.43, B/B0 = 3.1). The contribution from the different launch surfaces is indicated and can be identified by their color code.
In the text

	Figure 15 Counts in the high-resolution electron channels as observed at position 2 (L = 2.15, B/B0 = 4.6). The contribution from the different launch surfaces is indicated and can be identified by their color code.
In the text

	Figure 16 Counts in the high-resolution proton channels as observed at position 2 (L = 2.15, B/B0 = 4.6). The contribution from the different launch surfaces is indicated and can be identified by their color code.
In the text

	Figure 17 a) The 3DEES on its ground support equipment (GSE). The red caps on the 3DEES apertures are protection covers during storage and transport. b) 3DEES in the proton beam facility at Louvain-la-Neuve, Belgium.
In the text

	Figure 18 Deposited energy spectra in ADCu (primary X-axis) and in MeV (secondary X-axis) for the different sensors for incident protons of 62 MeV. Black dots with error bars and grey histograms represent the measured data, and the red histograms are derived from the simulated data. The resulting values for the sensitivity S are indicated.
In the text

	Figure 19 Proton high-resolution spectra HCp as observed with the OSM1_S1S aperture during different runs (see color code) with integration time 30 s. The simulated counts are represented by the green histogram bar. The spectra have been normalized to the integral counts of the run named output_3dees_42.
In the text

Figure 20 a) 3DEES on its GSE with the Sr-90 source holder aligned in front of it. The distance between the holder’s collimator exit hole and the 3DEES aperture was set to ~15 mm. The GSE allows for precise rotation of the aperture under test by a given angle with respect to a rotation axis passing through the center of the selected aperture. b) Cut view through the source holder, highlighting the elements (absorbing materials) that affect the electron spectrum.

In the text

	Figure 21 For the aperture labeled OSM2_S1S, measured high-resolution count spectra (blue bar-charts) as obtained with the Sr-90 electron source, compared to a normalized simulated count spectra (grey and light red histograms (pile-up included)) with corresponding squares at the center of each histogram bar, please see text for more information).
In the text

	Figure 22 Coarse resolution electron spectra (blue bar charts for three measurements) as obtained with a Sr-90 electron source. The light red and grey histograms represent the simulated spectra, respectively, with and without pile-up effects included.
In the text

	Figure 23 For OSM1_S1S: Counts in the selected channels LCe and HCe7 (blue dots) as a function of rotation angle of 3DEES aperture with respect to electron source aperture. Comparison with Geant4 simulations (orange dots) after renormalization. The simulated data points (from Geant4) are mirrored at 0 degrees. The error bars for the measurements are within the size of the dots and are not shown.
In the text

	Figure A1 Energy deposited in S1_F as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton.
In the text

	Figure A2 Energy deposited in S2 as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton.
In the text

Figure A3 Energy deposited in D1 as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton. The black dashed lines show the 9 classification limits for this pattern (same for protons and electrons); they define the 8 high-resolution channels where 6 are optimized for electrons and 2 optimized for protons.

In the text

Figure A4 Total deposited energy as a function of incident energy when the hit pattern is “D1 hit”: a) incident electrons, b) incident protons. The red lines show the particle selection limits to be identified as an electron, and the blue lines show the limits to be identified as a proton. The black dashed lines show the 17 classification limits (different for electrons and protons); they define the 16 coarse resolution channels for each particle type.

In the text