A prototype AlInP electron spectrometer

The development and characterization of a prototype temperature tolerant (capable of operation up to at least 100 (cid:1) C) particle counting electron spectrometer with an AlInP detector is reported. This is the ﬁ rst time that the response of an AlInP detector to electrons ( β (cid:3) particles) has been reported. The detector was a custom made circular mesa (200 μ m diameter) Al 0.52 In 0.48 P p þ -i-n þ (2 μ m i layer) photodiode; this was coupled to a custom made low-noise charge-sensitive preampli ﬁ er and otherwise standard readout electronics. The detector was electrically characterized and the spectrometer was investigated for its response to illumination from a 63 Ni radioisotope β (cid:3) particle source over the temperature range 100 (cid:1) C – 20 (cid:1) C. The absorbed electron energy within the active region (i layer) of the AlInP detector and the expected to be detected spectra were calculated using Monte Carlo simulations. Comparisons between the simulated and measured spectra indicated that the response of the spectrometer was in agreement with the Monte Carlo model. Future generations of electron spectrometers of this type are expected to be useful for space science missions where the instrumentation would be subject to high temperatures and intense radiation (e.g. to study the radiolytic processes in comets close to perihelion). In order to inform development of future generations of AlInP electron detectors for such applications, the response of the prototype instrument to illumination with solar wind electrons was modelled within the experimentally veri ﬁ ed energy range of the detector; avenues of future development to improve AlInP detector performance, identi ﬁ ed from the presently reported results, was investigated and discussed


Introduction
Electron spectroscopy can be used to study the surfaces, atmospheres, and magnetospheres of solar system bodies (e.g. comets, asteroids, planets, and moons), and their associated interactions with the solar wind and other sources of ionizing radiation (Livi et al., 2003). Consequently, electron spectrometers have formed a key part of the instrument package on numerous spacecraft, including: Galileo (to study the size, shape, and dynamics of the Jovian magnetosphere (Williams et al., 1992)); MESSENGER (to determine the structure and nature of Mercury's magnetic field and sparse atmosphere (Andrews et al., 2007)); and New Horizons (to study Pluto's interactions with the Solar wind (McNutt et al., 2008)).
Most semiconductor electron spectrometers in use today employ Si detectors. As an electron detecting material the performance of Si is, in many cases, excellent. However in high temperature (>20 C) and intense radiation environments other detector materials may be preferable. Indeed, the required radiation shielding and cooling systems for spectroscopic Si detectors to be used within space science instrumentation generally impose financial costs and technical restrictions to the mission. This means that harsh space environments can potentially render certain space environments inaccessible to or limit the scientific output of Si based instruments.
Such harsh environments include missions to study Mercury (Bepi-Colombo Mercury Planetary Orbiter external temperature % 300 C at 0.31 au (Huovelin et al., 2020)), the Sun's Corona (Parker Solar Probe external temperature % 1400 C at 0.05 au (Case et al., 2020)), the Solar wind (Solar Orbiter external temperature % 600 C at 0.22 au (Poncy et al., 2009)), and cometary radiolytic processes close to perihelion (temperatures up to 87 C were measured at P1/Halley at 0.8 au (Emerich et al., 1988)). Since understanding the origin and interactions of electrons within these environments is important for Solar System formation models, the development of electron spectrometers capable of operating at high temperatures is critical.
Over the past decade, it has been shown that photodiodes made from certain, carefully engineered, wide bandgap semiconductors are capable of photon and particle counting spectroscopy at high temperatures (Barnett et al., 2010) (Bertuccio and Casiraghi, 2003) (Bertuccio et al., 2011) (Bodie et al., 2021) (Lioliou et al., 2017) (Butera et al., 2018a)  . Among the various wide bandgap (E g ) III-V semiconductor compounds, Al 0.52 In 0.48 P (E g ¼ 2.31 eV at 20 C (Cheong et al., 2014)) has emerged as one of the most promising candidates. Photodiodes made from this material have been reported to operate in spectroscopic X-ray photon counting mode with energy resolutions (Full Width at Half Maximum) of 1.57 keV (Butera et al., 2016) and 1.31 keV  at 5.9 keV with the photodiodes and their custom low-noise charge-sensitive preamplifier both operating uncooled at 100 C. Al 0.52 In 0.48 P detectors are also expected to be more radiation tolerant than Si detectors of the same design. In fact, photovoltaic cell research has indicated reduced defect introduction rates within InP related materials cf. GaAs and Si, attributed to lower migration energies of In and P related defects (Yamaguchi et al., 2003) (Lee et al., 2007).
Here, we report for the first time the development and characterization of an AlInP direct detection electron spectrometer is reported. The spectrometer was capable of particle counting electron (β À particle) spectroscopy at temperatures up to and including 100 C. The proof-ofconcept spectrometer was comprised of a prototype p þ -i-n þ Al 0.52 In 0.48 P photodiode (200 μm diameter, 2 μm thick i layer) coupled to a custom made low-noise charge-sensitive preamplifier and standard onwards electronics. In order to investigate the instrument's response to energetic electrons, the detector was illuminated with β À particles from a 182 MBq (134 MBq, taking into account self-absorption) 63 Ni radioisotope β À particle source consisting of a 3 μm thick, 7 mm Â 7 mm, active 63 Ni layer, electroplated onto an inactive % 50 μm thick Ni foil substrate, and covered with a protective 1 μm thick inactive Ni overlayer.
The results demonstrate that spectroscopic particle counting direct electron detection can be achieved with Al 0.52 In 0.48 P detectors in environments of high temperature (20 C-100 C), and set the agenda for the development of a new generation of temperature tolerant (and likely radiation hard) electron spectrometers for future space science missions to extreme environments.

Detector structure
A custom made circular (200 μm diameter) Al 0.52 In 0.48 P p þ -i-n þ (2 μm i layer) photodiode of the structure shown in Table 1 was produced.
Al 0.52 In 0.48 P p þ -i-n þ epilayers were grown lattice matched on a commercial (100) n þ GaAs substrate (dopant Si) by metalorganic vapour phase epitaxy (MOVPE). The thicknesses of the Al 0.52 In 0.48 P p þ , i, and n þ layers were 0.2 μm, 2 μm, and 0.1 μm, respectively; the doping concentrations of the Al 0.52 In 0.48 P p þ and n þ layers were 5 Â 10 17 cm À3 (dopant Zn) and 2 Â 10 18 cm À3 (dopant Si), respectively. Due to the prototype nature of AlInP radiation detectors, a relatively thin (2 μm) i layer was grown in order to maximise material quality and help enable proof of concept of AlInP as a material for spectroscopic electron counting detectors. From the epiwafer, mesa devices were chemically etched using a solution of de-ionized water, phosphoric acid, and hydrogen peroxide, followed by a 1:8:80 solution of de-ionized water, sulphuric acid, and hydrogen peroxide for 10 s. Ti/Au (20 nm/200 nm) and InGe/Au (20 nm/200 nm) metal contacts were evaporated onto the top of the Al 0.52 In 0.48 P structure (covering 45 % of its surface) and onto the rear of the GaAs substrate, respectively. The photodiode was packaged in a TO-5 can and gold ball-wedge wirebonded. A schematic (not to scale) of the Al 0.52 In 0.48 P photodiode structure and top contact area is shown in Fig. 1.

Detector electrical characterization
Since the achievable energy resolution of the spectrometer depends, in part, upon the leakage current and capacitance of the detector (Lioliou and Barnett, 2015), preliminary leakage current and capacitance measurements of the photodiode were performed over the temperature range 100 C-20 C.
The detector leakage current and capacitance were measured as functions of reverse applied bias (from 0 V to 30 V, in 1 V increments) using a Keithley 6487 picoammeter/voltage source and a HP 4275A Multi Frequency LCR meter, respectively. A TAS Micro MT climatic cabinet was used to control the detector temperature (from 100 C to 20 C, in 20 C increments). The climatic cabinet, in which the detector was installed, was continuously purged with dry N 2 (<5 % relative humidity) to eliminate any humidity related effects .
Dark current measurements of the detector are shown in Fig. 2a. The dark current was found to decrease with decreasing temperature over the investigated range (100 C-20 C). At a reverse applied bias of 15 V (the optimal for the detector -see Section 4.3), the dark current decreased from 1.9 pA AE 0.4 pA at 100 C to 0.1 pA AE 0.4 pA at 20 C (i.e. below the measurement uncertainty of AE 0.4 pA). Since the detector was measured when it was packaged, the properties of the TO-5 package itself were also measured separately so that they could be subtracted. The TO-5 package's leakage current was subtracted from the data presented in Fig. 2a, resulting in the dark current contribution of the AlInP photodiode itself (i.e. with no current contribution from the TO-5 package) as shown in Fig. 2b. At a reverse bias of 15 V and a temperature of 100 C, the photodiode contributed a leakage current of 0.9 pA AE 0.4 pA to the total.
Dark capacitance as a function of reverse applied bias and temperature over the same ranges was also measured for the detector. A sinusoidal 50 mV rms magnitude test signal, at 1 MHz frequency, was used. At 15 V reverse applied bias, the capacitance of the packaged photodiode remained constant (within the calculated uncertainties) with temperature, measuring 2.47 pF AE 0.03 pF at 100 C and 2.45 pF AE 0.03 pF at 20 C. Had the uncertainties associated with the measurements been smaller, it may have been appropriate to conclude that there was a dependence of capacitance upon temperature. The capacitance of the detector itself, shown in Fig. 3b, was calculated by subtracting the capacitance of the TO-5 package from the capacitance of the packaged photodiode. At 15 V reverse applied bias, capacitances of 1.65 pF AE 0.04 pF at 100 C and 1.63 pF AE 0.04 pF at 20 C were calculated for the detector. The corresponding depletion width (calculated assuming a parallel plate capacitance (Sze, 2006)) was 1.9 μm AE 0.2 μm for both temperatures. The calculated capacitance uncertainty included the uncertainty associated with the measurement apparatus (AE 0.004 pF) and the repeatability accuracy of the measurement (AE 0.03 pF). The Debye length (0.04 μm at 100 C and 0.03 μm at 20 C (Sze, 2006)) was taken into account for the estimation of the uncertainty associated with the depletion width (see Fig. 4).

Introduction
The response of the presently reported AlInP detector, when coupled to a custom made low-noise charge-sensitive preamplifier and otherwise standard readout electronics, was investigated under the illumination of a 63 Ni β À particle source (endpoint energy 66 keV), as a function of temperature (100 C-20 C). Monte Carlo simulations were conducted using the computer modelling package CASINO    to predict the electron quantum detection efficiency of the detector, the spectrum incident on the detector, and the spectrum expected to be detected. CASINO was selected due to its wide availability, simplicity of use, and the successful prediction (good agreement between experimentally measured spectra and that predicted to be detected) of similar situations in InGaP ( 2018) electron detectors. More elaborate models, using more complex software, (e.g. Geant4 (Agostinelli et al., 2003) and FLUKA (B€ ohlen et al., 2014) (Ferrari et al., 2005)) may prove beneficial for future simulations.
Ensuring an understanding of the detector through modelling is essential for space science applications, where the spectrum incident upon the detector must be reconstructed from that detected by the instrument in   The percentage of electron energy absorbed within the AlInP detector's active region (i layer) was calculated by dividing the total electron energy deposited within the i layer of the detector with the total electron energy incident upon the detector's face. CASINO simulations, across the incident electron energy range 1 keV-66 keV ( 63 Ni endpoint energy) in 1 keV steps, informed the values used in this calculation. In order to account for the metal contact coverage of the detector's top surface (45 % of the face of the detector) two sets of simulations were conducted, each consisting of 4 k electrons at each incident electron energy (i.e. 264 k electrons in total). The first set simulated electrons as incident upon a portion of the photodiodes face covered by the top Ti/Au metal contact. The second set simulated electrons as incident upon a portion of the photodiodes face not covered by the top Ti/Au metal contact (the so-called optical window). The simulation results were then combined in the appropriate proportions according to the Ti/Au metal contact surface coverage (45 % of the detector's face). The GaAs p þ and n þ layers, the Al 0.52 In 0.48 P p þ and n þ layers, and the GaAs substrate of the Al 0.52 In 0.48 P p þ -i-n þ mesa structure (see Table 1), were considered to be inactive for each set of simulations; i.e. only charge created by electrons within the active region (i layer) was assumed to be usefully absorbed. For each material included within the simulations, the associated density at 20 C was used.
CASINO was configured as per refs (Butera et al., 2019) (Whitaker et al., 2020)     which simulated similar situations in InGaP, AlGaAs, GaAs, and SiC electron detectors, respectively. For convenience of computation, the electron beam width was set to 1 nm, with incident electrons simulated as being 90 to the detector's face at the start of their track. Due to the large area (7 mm diameter) of the 63 Ni β À particle source cf. the relatively small diameter of the detector (200 μm), and the relatively short length of the electron tracks cf. the relatively large diameter of the detector in comparison, angular effects of electrons impinging the detector's surface were considered to be negligible. The simulations did not include non-localised effects of secondary electron generation. One computer, with 32 GB of random access memory and an Intel i7-6700 processor (3.40 GHz, 8 threads), was used to run the simulations. The percentage of electron energy absorbed within the active region of the detector as a function of incident electron energy, as predicted by the simulations, is presented in Fig. 5.
The percentage of electron energy absorbed within the i layer of the detector increased with incident electron energy up to: % 19 keV (71 %) for electrons incident upon the optical window; % 28 keV (34 %) for electrons incident upon the top metal contact; and % 23 keV (47 %) assuming uniform illumination across the face of the AlInP photodiode i.e. the weighted case. The increase in electron energy absorbed within the i layer of the detector for incident electrons of energy 23 keV given the weighted case, suggested that low energy incident electron absorption was limited by inactive top layer (metal contacts and p þ layers) absorption of electrons. For incident electrons of energy > 23 keV, in the weighted case, the percentage of electron energy absorbed within the i layer decreased with increasing incident electron energy (reaching 13 % at 66 keV). Although losses within the inactive top layers (metal contacts and p þ layers) of the detector played a part, the results suggested that the relative thinness of the Al 0.52 In 0.48 P p þ -i-n þ mesa structure i layer (2 μm) limited absorption of high energy incident electrons. A thicker active region would therefore be required for optimal absorption of incident electrons of energies > 23 keV. Through CASINO simulations, it was found that Al 0.52 In 0.48 P layer thicknesses of 2.8 μm AE 0.1 μm and 16.5 μm AE 0.2 μm would be required to fully absorb 95 % of incident electrons (excluding backscattered electrons) with energies up to 23 keV and 66 keV, respectively.
4.2.2. 63 Ni β À particle (electron) spectrum expected to be detected The electron (β À particle) spectrum expected to be detected by the spectrometer, as a result of illumination by the 63 Ni β À particle source, was estimated by first simulating the expected to be incident spectrum upon the detector's face. The effects of self-absorption mechanisms within the active 63 Ni layer itself, as well as the attenuation of emitted electrons through the 63 Ni source's inactive Ni overlayer (1 μm thick) and the dry N 2 atmosphere (3 mm thick) separating the 63 Ni source and detector (see Section 4.3) were included in these simulations. For each material included in the simulations, its density at 20 C was used. A total of 66 CASINO simulations, across the incident electron energy range 1 keV-66 keV ( 63 Ni endpoint energy) in 1 keV steps, were conducted. The 63 Ni relative emission probability, accounting for self-absorption, informed the number of electrons simulated at each energy. A total of 18.5 M electrons were simulated using 12 computers (each with an Intel i7-6700, 8 thread, 3.40 GHz processor, and 32 GB random access memory), which performed the simulations in parallel. The number of simulated electrons was selected to ensure acceptable statistics for data interpretation rather than to reflect the number of electrons emitted from the 63 Ni source during the experimentally measured spectra as reported in Section 4.3.
The simulated electron paths were tracked and used to calculate the remaining energy of each electron after passing through the active 63 Ni  layer, the inactive Ni overlayer, and the dry N 2 atmosphere separating the 63 Ni source and AlInP detector. The remaining energies of the 18.5 M simulated electrons were then binned into channels of 1 keV width, producing the spectrum expected to be incident upon the detector's surface, as shown in Fig. 6. Simulated electrons that lost all their energy prior to reaching the detector's face were excluded from the spectrum. The discontinuity at % 60 keV remaining electron energy was a consequence of the simulated electron step size (1 keV) and typical energy loss (!6 keV) from electrons at energies ! 61 keV in passing through the inactive Ni overlayer and the dry N 2 atmosphere before reaching the detector's surface.
The electronic noise contributions of the spectrometer (white parallel noise; white series noise (including induced gate drain current noise); 1/f series noise; and dielectric noise) were calculated as per refs (Bertuccio et al., 1996) (Gatti et al., 1990) (Barnett et al., 2012). The white parallel noise included the leakage current of the AlInP photodiode, its packaging, and the input JFET of the preamplifier used to obtain the spectra (see Section 4.3). The white series noise included the capacitance of the photodiode, its packaging, and the input JFET. The dielectric noise included the dielectric contributions of the photodiode, its packaging, and input JFET. The expected Fano limited energy resolution (Barnett, 2011) of the detector was also calculated as a function of energy and temperature (the Fano factor (Fano, 1947) quantifies the deviation between the observed and the predicted variance of the generated electron-hole pairs (Bertuccio et al., 1997)). For this purpose, we used the known electron-hole pair creation energy of Al 0.52 In 0.48 P (5.34 eV AE 0.07 eV at 20 C; 5.04 eV AE 0.07 eV at 100 C (Butera et al., 2018b)), and the Fano Factor of GaAs, 0.12 (Bertuccio et al., 1997), (the Fano factor for Al 0.52 In 0.48 P has not yet been reported). The calculated noise contributions as a function of energy at 20 C and 100 C, at the optimal shaping time (10 μs at 20 C; 1 μs at 100 C) and detector reverse bias (15 V) (see Section 4.3) are presented in Fig. 7.
The 63 Ni β À particle spectrum incident upon the detector's surface ( Fig. 6) was combined with the weighted deposited electron energy histogram used to predict the percentage of electron energy absorbed within the i layer (active region) of the detector (see Fig. 5). The result was the spectrum predicted to be detected by the spectrometer, excluding the effects of Fano noise, spectrometer electronic noise, and incomplete charge collection, shown in Fig. 8.
The effects of the calculated Fano noise and spectrometer electronic noise upon the spectrum expected to be detected were then computed by distributing, stochastically, each simulated detected count, at each bin energy, across the total calculated Gaussian noise distribution at that bin energy. The energy resolutions (Full Width at Half Maximum) at 6 keV and 60 keV, as determined by the calculated noise contributions, were 0.93 keV AE 0.04 keV and 1.03 keV AE 0.04 keV, respectively, at 20 C and 1.07 keV AE 0.04 keV and 1.15 keV AE 0.04 keV, respectively, at 100 C. The predicted 63 Ni β À particle spectrum, accounting for the calculated spectrometer noise at 20 C and 100 C, can be seen in Fig. 8. The effects of any incomplete charge collection were not included. Since the Full Width at Half Maximum (FWHM) were comparable to the bin-width used for the simulations, the effect of including the spectrometer's noise in the simulations was negligible.
The difference between the 63 Ni β À particle spectrum incident upon the AlInP detector (see Fig. 6) and that predicted to be detected (see Fig. 8) was a consequence of the detector's top inactive layers (metal contacts and p þ layers) and the relatively thin active region (2 μm i layer). Although the inactive top layers significantly attenuated low energy incident electrons (e.g. given 5 keV electrons incident upon the detector's surface, 2 % of electron energy was absorbed within the active region (see Fig. 5)), thus reducing the detected counts at low energy (<5 keV), the inactive layers also attenuated higher energy incident electrons before they entered the active region. The result was a decrease in energy of the detected counts relative to the incident electrons impinging the detector's surface. In addition, the active region of the detector was not sufficiently thick to completely absorb the energy of incident electrons at energies >15 keV, resulting in only a portion of their energy being deposited in the active region.

Experimental 63 Ni β À particle spectra
In order to confirm the validity of the presently reported Monte Carlo model, and to investigate the prototype electron spectrometer's response to β À particle radiation, the AlInP detector (2 μm i layer) was connected to a custom low-noise charge-sensitive preamplifier and placed within a TAS Micro MT climatic cabinet for temperature control. The climatic cabinet contained a dry nitrogen atmosphere to eliminate any humidity related effects . The output of the preamplifier was connected to an ORTEC 572A shaping amplifier and an ORTEC EASY-MCA 8k multi-channel analyser (MCA). The 182 MBq (134 MBq, taking into account self-absorption) 63 Ni β À particle source (endpoint energy 66 keV) was placed % 3 mm above the p þ side contact of the detector inside the chamber. A block diagram of the experimental setup is shown in Fig. 9.
Spectra were collected over the temperature range 100 C-20 C (in 20 C decrements) with the detector reverse biased at 15 V. This operating reverse bias was chosen in order to minimise the spectrometer's noise; the sum, in quadrature, of the series white noise (proportional to the total capacitance at the input of the preamplifier) and the parallel white noise (proportional to the combined leakage current of the detector and the preamplifier's input JFET) was lower at 15 V than any other voltage studied. Each spectrum had a live time of 1800 s. To ensure the best energy resolution at each temperature studied, shaping times of 1 μs and 10 μs were chosen across the temperature ranges: 100 C-60 C; and 40 C, respectively, in accordance with ref. (Butera et al., 2016). The acquired 63 Ni β À particle spectra at 100 C and 20 C are shown in Fig. 10. Energy calibration of the accumulated β À particle spectra was achieved using the apparent endpoint energy as determined from the simulations, accounting for the relative probability of detection and the position of the zero-energy noise peaks (not shown in the figures). The change in density of the 63 Ni source, the N 2 atmosphere, and the detector as a function of temperature, was considered negligible within the investigated temperature range (100 C-20 C); as such, the β À particle spectra predicted to be detected at all temperatures were the same, excluding changes in the performance of the spectrometer itself. A linear variation in spectrometer output as a function of detected energy between the zero-energy noise peak and endpoint energy was assumed.
The difference in apparent endpoint energy between the experimentally obtained 63 Ni β À particle spectra (% 48 keV, see Fig. 10) and that expected to be detected through simulation (% 60 keV, see Fig. 8) was a Fig. 6. Simulated 63 Ni β À particle spectrum as emitted from the active 63 Ni layer including self-absorption (triangles), and incident upon the Al 0.52 In 0.48 P detector's surface (squares) including attenuation in the inactive Ni overlayer and N 2 atmosphere, at 20 C, as determined by Monte Carlo modelling. The associated uncertainties were smaller than the symbols. consequence of the experimental spectra live time (1800 s), which was not long enough to accumulate high energy (!50 keV) electrons; given the simulated expected to be detected spectrum, a live time of % 182 h would be required to accumulate 1 count at 60 keV. The difference in multi-channel analyser channel width (in units of energy) between the two spectra was a consequence of the change in conversion factor of the spectrometer as a function of temperature and a change in the shaping amplifier's gain as a function of shaping time. In order to compare the simulated and experimentally measured spectra, and thus determine the validity of the Monte Carlo model of the spectrometer, each spectrum (i.e. all accumulated, not just those at 20 C and 100 C) was independently normalised to the mean number of counts detected per channel within the broadly flat region (13 keV-15 keV) associated with each spectrum. The results for the spectra accumulated at 20 C and 100 C are shown in Fig. 11, comparable results were obtained for all investigated temperatures.
At electron energies !8 keV, the experimentally measured 63 Ni β À particle spectra were found to be in good agreement with that predicted to be detected by the Monte Carlo model (within the measured uncertainties), thus indicating that the exclusion of the non-localised effects of secondary electron generation within the simulations was not significant for the present purposes. This was consistent with preliminary simulations conducted during the pre-computational planning of the reported simulations. Those preliminary simulations investigated the effects of including the non-localised effects of secondary electron generation for selected parts of the simulations. The results indicated that the time cost of including execution of the relevant non-localised secondary electron codes for the whole simulation outweighed the benefits of including them. At electron energies <8 keV, the differences between the experimentally measured spectra and those expected to be detected were attributed to the right hand side of the zero-energy noise peak tail, which was not entirely eliminated by the low energy threshold (2 keV) set for the experimental accumulations. The results demonstrate agreement between the Monte Carlo model and the experimentally obtained 63 Ni β À particle spectra with the prototype spectrometer when operated within the temperature range 100 C-20 C, and when illuminated with electrons of energy 66 keV.

Introduction
A prototype Al 0.52 In 0.48 P p þ -i-n þ photodiode has been demonstrated, for the first time, as suitable for particle counting electron (β À particle)   8. 63 Ni β À particle spectrum expected to be detected by the electron spectrometer, excluding (black diamonds) and including calculated Fano noise and spectrometer electronics noise at 20 C (green circles) and 100 C (red squares), as predicted using Monte Carlo simulations. The effects of incomplete charge collection were not included. The relative counts of each spectrum, normalised to the mean number of counts detected per channel within the broadly flat region (13 keV-15 keV), is also presented. Fig. 9. Block diagram of the experimental 63 Ni β À particle spectra accumulation setup. spectroscopy at temperatures up to 100 C when coupled to appropriate readout electronics. Since Al 0.52 In 0.48 P is also expected to be more tolerant to intense radiation environments cf. Si (Lee et al., 2007), the material may serve as a suitable replacement to Si for electron spectroscopy applications in environments of intense radiation as well as high temperature; such environments are to be encountered during many space science missions likely to be flown in the 21st Century.
Therefore, the development of radiation hard and temperature tolerant instrumentation, including electron spectrometers, is of pressing concern. Conventional Si based instrumentation used within space science missions often requires radiation shielding of greater mass than the instrument itself, placing significant constraints upon spacecraft mass and volume budgets. For example, in the case of the Jupiter Energetic Particle Detector Instrument (JEDI) on board the Juno spacecraft (Mauk et al., 2017), the instrument has a mass of 1.4 kg with an additional 5 kg of radiation shielding (Mauk et al., 2017). Had the instrument used detectors and electronics which were more radiation hard, the science package could have been more extensive, or the mission cost could have been reduced by virtue of eliminating some, or all, of the radiation shielding. For example, a reduction in shielding mass of JEDI by 64 % (i.e. a reduction to 1.8 kg) would have enabled two instruments of the same mass to be flown within the same mass budget. High temperature environments also pose challenges to conventional Si instrumentation: in high temperature environments such as those experienced at Mercury (e.g. 400 C (Benkhoff et al., 2010)), or comets close to perihelion (e.g. 87 C (Emerich et al., 1988)), Si detectors require substantial cooling systems to operate. The Si detectors used within the Solar Intensity X-ray and particle Spectrometer (SIXS), on board the BepiColombo Mercury Planetary Orbiter (MPO), require an optimal temperature range of À10 C to À25 C (Huovelin et al., 2020) whilst the external temperature of the instrument is expected to reach 300 C (Huovelin et al., 2020); a heat shield and radiator was therefore installed. Such cooling systems, as is the case with radiation shielding, place constraints upon spacecraft mass and volume budgets, in turn limiting the spacecraft science instrument suite, and/or increasing space mission costs. Instrumentation that could operate at higher temperatures, even if not as hot as 300 C, would substantially simplify the thermal design of spacecraft for such environments.
Whilst the presently reported electron spectrometer is only a proof-ofconcept prototype, consideration of eventual use-cases can inform future development. An important area of study that would benefit from such a temperature tolerant, potentially radiation hard, electron spectrometer, is the investigation of solar wind electron interactions with cometary ices. Such radiolytic interactions result in the destruction and synthesis of molecules and are consequently of astrobiological significance and important for the proper understanding of the Solar System's formation (Luspay-Kuti et al., 2018). As the solar wind density increases, so too does the number of radiolytic interactions, thus the in-situ study of comets close to perihelion is desirable. Space science missions exploring such environments would benefit from instrumentation capable of operating at high temperatures (e.g. temperatures as high as 87 C were observed at comet Halley at 0.8 au (Emerich et al., 1988)), therefore AlInP detectors may be preferable compared to Si detectors in order to provide spectrometry of the incident solar wind electron flux upon the comet, in addition to potentially measuring photoelectrons generated by the radiolytic interactions perhaps on a cometary lander (Burch et al., 2007).

The solar wind electron environment at 0.99 au
As the predicted and detected 63 Ni β À particle spectra obtained with the Al 0.52 In 0.48 P detector spectrometer were in agreement (see Section 4.3), the response of the prototype electron spectrometer to solar wind electrons during a typical solar electron event (Krucker et al., 2009) at a distance of 0.99 au from the Sun was investigated as an indicative space Fig. 10. Accumulated electron spectra using the presently reported spectrometer, under the illumination of a 63 Ni β À particle source, with 15 V reverse bias applied to the detector, at (a) 20 C and (b) 100 C. Energy calibration was achieved using the apparent endpoint energy as determined from the simulations, taking into account the position of the zero-energy noise peaks (not shown in the figures) and the relative probability of detection. Spectra were normalised to the mean number of counts detected per channel within the broadly flat region (13 keV-15 keV). The same 63 Ni β À particle spectrum shape and apparent endpoint energy was observed at all investigated temperatures. Fig. 11. Experimentally measured 63 Ni β À particle spectra (grey line) at (a) 20 C and (b) 100 C compared to that predicted to be detected from the simulations (squares), including the calculated Fano noise and spectrometer electronics noise. Each spectrum was normalised to the mean number of counts detected per channel within the broadly flat region (13 keV-15 keV) of the spectrum. Comparable results were obtained for the other investigated temperatures. electron environment. Omnidirectional electron flux measurements, using data taken by the 3DP instrument on board the WIND spacecraft on April 4, 2000 were used (Krucker et al., 2009), and can be seen in Fig. 12. To remain consistent with the Monte Carlo model of the presently reported electron spectrometer (see Section 4.2), only the low energy electron (1 keV-66 keV) component of the environment was considered. It should be noted that the solar wind environment contains electrons of much greater energies (up to MeV) that would also contribute to the detected spectrum in any such mission.
The omnidirectional electron flux shown in Fig. 12 was halved and multiplied by the area of the Al 0.52 In 0.48 P detector (3.14 Â 10 À4 cm 2 ), thereby producing the expected to be incident spectrum upon the detector of the prototype electron spectrometer. The predicted to be detected electron spectrum was then calculated by combining the expected to be incident spectrum and the weighted histogram of deposited electron energies within the active region (i layer) of the detector as a function of incident electron energy, as produced by the Monte Carlo simulations (see Section 4.2). Fig. 13 shows the expected to be incident electron spectrum upon the prototype electron spectrometer compared with that predicted to be detected, accounting for calculated Fano noise and spectrometer electronic noise as per Section 4.2.2, assuming detector temperatures of 20 C and 100 C.
As was the case for the 63 Ni β À particle spectra, electron energy losses within the inactive top layers of the detector (top metal contact and p þ layers) and the relative thinness of the active region (2 μm i layer) explain the difference between solar wind electron spectra expected to be incident upon the instrument and predicted to be detected by the instrument. The inactive top layers of the detector attenuated the incident electrons before they entered the active region of the detector, resulting in: low energy (<5 keV) incident electrons not being detected, and a reduction in the energy of incident electrons with energies ! 5 keV prior to them entering the active region, thereby reducing the energy of the detected events. The active region of the detector (2 μm i layer) was not sufficiently thick to completely absorb the energy of incident electrons at energies > 15 keV, resulting in only a portion of their energy being deposited in the active region.

Expected spatial and temporal resolution at 0.99 au
Consideration of the required spatial and temporal resolutions for such a spectrometer is also useful to inform future development. For a spectrometer on board a moving spacecraft, the distance travelled per unit time required to obtain an acceptable (statistically significant) number of counts over the energy range of interest can be a limiting factor. In the case of the IES instrument on board the Rosetta spacecraft (Burch et al., 2007), which provided in situ electron ( 22 keV) population data at comet 67P/Churyumov-Gerasimenko, a temporal resolution of 128 s was required (Burch et al., 2007). This resulted in a solar wind electron flux spatial resolution of 4390 km per spectrum at 67P/Churyumov-Gerasimenko perihelion (1.24 au). Given the omnidirectional solar wind electron flux (4.3 Â 10 6 keV À1 cm À2 s À1 ) at 0.99 au on April 4, 2000, and in order to match the spatial resolution (4390 km) of the Rosetta spacecraft IES instrument, a temporal resolution of 147 s would be required for the presently reported spectrometer, assuming a circular solar orbit. The total number of counts predicted to be detected by the presently reported electron spectrometer over this accumulation time (147 s) would be 5 Â 10 3 AE 1 Â 10 2 within the temperature range 20 C-100 C, assuming the presently reported solar wind omnidirectional electron flux obeys the law of Poisson statistics (Jenkins et al., 1995). Temperatures experienced by an electron spectrometer within such an orbit (0.99 au), although largely dependent upon spacecraft geometry and material choices, could be as high as 150 C; use of an AlInP detector could therefore reduce spectrometer cooling system mass and volume requirements.

An AlInP detector for improved performance
Given the above, two avenues of future development, identified from the presently reported results, may improve the performance of an AlInP electron spectrometer. Firstly, given that the relatively thin active region of the detector was found to limit the detection of incident electrons >15 keV, increasing the active region thickness would significantly improve the detection statistics of higher energy electrons. Secondly, given that the inactive top layers of the detector (metal contacts and p þ layers, see Section 2) were found to limit low energy incident electron absorption such that < 5 keV incident electrons could not be detected, reducing the thickness of the inactive top layers of the detector would improve low energy incident electron counting statistics (see Fig. 5). Whilst this may be possible through modifying the structure grown during the epitaxial growth, an alternative approach would be to etch away, i.e. "thin", a window region in the Al 0.52 In 0.48 P coincident with optical window of the contact. Alternatively, the opposite side of the detector (i.e. the n type layer side) could be "back-thinned", if illumination was to be from that side. Table 2 shows a proposed layer structure of an Al 0.52 In 0.48 P photodiode that has been shown in modelling to provide better performance.
It should be noted that the epitaxial structure proposed in Table 2 would change the electrical characteristics of the detector cf. those of the detector which was experimentally investigated. As a first  (Krucker et al., 2009), using data accumulated by the 3DP instrument on board the WIND spacecraft. Fig. 13. Expected to be incident (black line) and expected to be detected electron spectra (1 keV-66 keV) of the solar wind electron environment at Lagrangian point L 1 (0.99 au) on April 4, 2000 using the prototype electron spectrometer, accounting for calculated Fano noise and spectrometer electronic noise at 20 C (green triangles) and 100 C (red circles). approximation, assuming the capacitance of such a detector can be approximated as a parallel plate capacitance (Sze, 2006) and that the detector was a 200 μm diameter circular mesa device, the expected device capacitance would be 0.19 pF AE 0.01 pF at full (16.5 μm) depletion of the i layer. Assuming that the effective carrier concentration of the i layer was 1 Â 10 15 cm À3 , then the required reverse applied bias to reach a full depletion would be 220 V (electric field of 133 kV cm À1 ), given the equation for general non-uniform distributions (White, 1982). In consideration of the presently reported leakage current results (see Fig. 2), the expected leakage current of the proposed Al 0.52 In 0.48 P photodiode at full depletion was estimated to % 0.1 pA at 20 C and % 4 pA at 100 C. Given these properties, the energy resolution (FWHM) of a spectrometer employing such a detector can be estimated to improve to 0.81 keV AE 0.02 keV and 0.92 keV AE 0.02 keV at 6 keV and 60 keV at 20 C, and 0.89 keV AE 0.02 keV and 0.99 keV AE 0.02 keV at 6 keV and 60 keV at 100 C, by virtue of reduced spectrometer electronic noise contributions. At 20 C, the series white noise reduced to 6.9 e À rms AE 0.1 e À rms, the parallel white noise reduced to 11 e À rms AE 2 e À rms, and the dielectric noise of the detector reduced to 13 e À rms AE 1 e À rms, compared with 10.2 e À rms AE 0.1 e À rms, 11 e À rms AE 2 e À rms, and 39 e À rms AE 4 e À rms, for the present 2 μm thick detector. An introduction to the various noise contributors in semiconductor radiation spectrometers can be found in ref. (Lioliou and Barnett, 2015). The expected energy resolutions reported here are comparable to those recently measured using a In 0.5 Ga 0.5 P electron detector coupled to similar front-end electrons (0.85 keV and 1.23 keV at 20 C and 100 C respectively, at 5.9 keV ), but not as good when compared to high quality Si electron detectors coupled to ultra low-noise front-end electronics. A single-pixel Si Drift Detector, coupled to low-noise electronics designed for the Tritium Investigation on Sterile to Active Neutrino mixing (TRISTAN) project, reported energy resolutions of 190 eV and 265 eV at 5.9 keV and 20 keV, respectively, at room temperature (Gugiatti et al., 2020).

Monte Carlo simulations with the proposed detector
Considering this proposed detector structure, additional Monte Carlo simulations, similar to those of Section 4.2.1, were conducted. To enable direct comparisons between the effects of the change in epitaxial structure only, the top metal contact thickness (20 nm Ti/200 nm Au) and coverage (45 % of the detector's face) was kept the same. The results can be seen in Fig. 14.
The percentage of electron energy absorbed within the active region of the proposed detector increased to 80 % cf. 28 % in the 2 μm thick detector for 44 keV electrons incident upon the optical window. At electron energies >44 keV the percentage of energy absorbed within the active region remained constant at 80 % up to the maximum electron energy investigated. At the maximum energy, 66 keV, the percentage of energy usefully absorbed in the 2 μm thick detector was just 12 %. The results also demonstrated how reducing the thickness of the inactive layers at the front of the detector improves detection of low energy electrons; given electrons of 4 keV energy incident upon the optical window, the percentage of electron energy usefully absorbed within the detector was 11 % for the proposed detector cf. 0 % for the 2 μm i layer detector. As was the case in Fig. 5, the inactive top metal contacts attenuated incident electrons across the investigated range, limiting the performance of the proposed detector at lower energies.
5.6. Solar wind electron environment spectra at 0.99 au In order to emphasise the importance of the improvements in design for the structure shown in Table 2, the response of the detector to solar wind electrons was modelled and is presented in Fig. 15; this can be compared with the spectrum which would be detected with the 2 μm thick detector, presented in Fig. 13. The effects of the various noise sources contributing to the energy resolution of the spectrometer were included given the expected properties of the device at 20 C and 100 C. However, given that the FWHM at 6 keV and 60 keV were 0.99 keV AE 0.02 keV, comparable to the bin width used for the simulations (1 keV), the noise was inconsequential. The number of electrons predicted to be detected by the proposed detector during an accumulation time of 147 s was 7 Â 10 3 AE 1 Â 10 2 cf. 5 Â 10 3 AE 1 Â 10 2 electrons within the temperature range 20 C-100 C. The 39 % increase in counts per unit time could be used to improve the temporal and spatial resolutions achievable with such an instrument employing a 16.5 μm thick detector cf. a 2 μm thick detector, or alternatively improve the spectral data by increasing the number of electrons in each spectrum if the accumulation time was kept constant between the two instruments. A spectrum of duration 106 s, accumulated using the 16.5 μm thick detector, would contain the same number of electrons as a 147 s duration spectrum accumulated with the 2 μm thick detector. This means that a spatial resolution of 3160 km per spectrum would be achievable with the former cf. 4390 km per spectrum with the latter, thus enabling spatial variations in electron population density to be explored in greater detail.
The size (face area) of the detector could also be increased to improve the detection statistics: the count rate would be expected to scale in proportion to the area of the detector. However, detectors of larger surface area would have greater device capacitances and leakage currents, and thus increased noise (i.e. degraded energy resolutions). A spectrometer employing a 16.5 μm thick Al 0.52 In 0.48 P detector of 1 mm 2 area would be expected to have an energy resolution of 1.3 keV AE 0.1 keV at 6 keV and 60 keV, and would provide count rates % 32 Â greater than the smaller detectors previously discussed.

Conclusions and further work
For the first time, an Al 0.52 In 0.48 P p þ -i-n þ photodiode (200 μm diameter; 2 μm i layer) has been investigated for its performance as a temperature tolerant (up to 100 C) particle counting spectroscopic electron detector. The detector was coupled to a custom made low-noise charge-sensitive preamplifier and standard onwards readout electronics. The instrument was investigated for its response to electrons (β À particles) emitted by a 182 MBq (134 MBq, taking into account selfabsorption) 63 Ni β À particle source (endpoint energy 66 keV) over the temperature range 100 C-20 C. An apparent endpoint energy of % 48 keV (instead of 66 keV, see Fig. 10) was measured at all temperatures.
This was attributed to the relative thinness of the i layer (2 μm) and the top inactive layers of the detector (top metal contact and p þ layers). The accumulated 63 Ni β À particle spectra were compared to that predicted to be detected from computer simulations (see Fig. 11). At electron energies !8 keV, the experimentally measured and predicted to be detected 63 Ni β À particle spectra were found to be in good agreement (within the measured uncertainties); at electron energies < 8 keV, the difference was attributed to the right hand side of the zero-energy noise peak tail, which was not entirely eliminated by the low energy threshold (2 keV) set for the experimental accumulations.
As the predicted to be detected and experimentally measured 63 Ni β À particle spectra were in agreement, and in order to inform future development, a potential use-case of the prototype electron spectrometer was investigated (see Section 5). Due to the temperature tolerant and potentially radiation hard (Lee et al., 2007) nature of the AlInP detector, the response of the spectrometer to incident solar wind electrons experienced by a comet close to perihelion was studied. Given the omnidirectional solar wind electron flux (4.3 Â 10 6 keV À1 cm À2 s À1 , within the energy range 1 keV-66 keV) during a typical solar electron event (Krucker et al., 2009) at Lagrangian point L 1 (0.99 au), the total number of counts predicted to be detected by the presently reported electron spectrometer over an accumulation time of 147 s (representing a solar wind spatial resolution of 4390 km assuming a circular solar orbit at 0.99 au; similar to the solar wind spatial resolution of the IES instrument on board the Rosetta spacecraft (Burch et al., 2007)) was 5 Â 10 3 AE 1 Â 10 2 counts.
The reported results set the agenda for future development of AlInP detectors capable of high temperature (up to at least 100 C) direct electron detection without the need for cooling, potentially replacing Si detectors as the gold-standard choice for space-based electron spectrometers. In future work, characterization of thicker i layer AlInP electron detectors will be reported in order to further improve performance. Detectors of larger active area will also be investigated, in addition to pixel array detector designs. It would be beneficial to characterise the response of the spectrometer with additional electron radiation sources in order to extend the energy range with which it has been investigated; radioisotope β À particle sources of potential utility include: 3 H (end point energy 18 keV); 14 C (end point energy 156 keV); and 90 Sr (end point energy 2.3 MeV).

Data availability
The data that supports the findings of this study are available within the article.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Fig. 15. Expected to be incident (black line) and expected to be detected electron spectra (1 keV-66 keV) of the solar wind electron environment at Lagrangian point L 1 (0.99 au) on April 4, 2000 using the proposed Al 0.52 In 0.48 P (16.5 μm i layer) structure, accounting for calculated Fano noise and spectrometer electronic noise at 20 C (green triangles) and 100 C (red circles). For clarity, the associated uncertainties have been omitted.