The use of diamond as substrate makes it possible to produce pure (almost free from defects) single crystal (sCVD) while polycrystalline (pCVD) has a large number of defects. As a consequence, pCVD exhibits a poor energy resolution and a low charge collection capability that makes it less attractive than sCVD for detector applications. Its main advantages rely on lower prices and higher achievable sizes up to - or greater than - 20 × 20 mm² (7 × 7 mm² for sCVD) [8]. An alternative is heteroepitaxial diamond grown on iridium substrate (DOI diamond). Indeed, DOI can be produced with large areas such as pCVD [8] but it has been recently proven [9, 10] that the transport of the holes is efficient whereas electron transport is hampered.

Compared to silicon detectors, diamond features several intrinsic advantages. High radiation hardness results from a high threshold displacement energy measured as 43 eV (silicon: 25 eV) [11], i.e., per incident ion fewer displacements per atoms (dpa) will occur. The high mobility of electrons 1,714 cm² V⁻¹ s⁻¹ (silicon 1,520 cm² V⁻¹ s⁻¹) and holes 2,064 cm² V⁻¹ s⁻¹ (silicon 600 cm² V⁻¹ s⁻¹) [12] grants it a fast time response and a good time resolution (lower than few tens of picoseconds [13]). The 0.3% RMS measured energy resolution with 5 MeV particles is comparable to that of Si detectors [14, 15], while the efficiency in diamond detectors can achieve 100% according to [16]. Finally, the 5.45 eV band gap results in a lower noise level and an almost negligible leakage current.

In the present study, we used sCVD from Element 6 [17], and pCVD diamonds from Element 6 [17], DDK (United States Applied Diamond Inc.) [18] and II-VI [19] and heteroepitaxial DOI diamonds from Augsburg University [20] (see Table 1).

The LOHENGRIN separator at ILL was used to provide mass- and energy-separated FF. Two 7 × 0.5 cm^{2 235}U and ²³³U targets respectively are mounted on a 0.4 mm thick Ti backing and covered by a 0.25 µm thick Ni foil [29]. The ²³³U target was supplemented by a spot of evaporated ⁶LiF. As illustrated in Figure 5, the targets are placed at an in-pile position and exposed to a thermal neutron flux of ≈5 × 10¹⁴ cm⁻² s⁻¹. Thermal neutron-induced fission of uranium isotopes produces a light FF with the mass yield distribution peaking around A ≈ 95 and a complementary heavy FF with the yields peaking around A ≈ 140. Light FF have a kinetic energy distribution peaking close to 100 MeV while heavy FF have kinetic energy distributions peaking at 50–80 MeV depending on their mass. The intrinsic FF energies are reduced by energy loss in the target and the Ni cover foil respectively, thus effectively covering a wider range of energies that can be selected by the spectrometer.

The recoiling FF that are typically ionized to charge states Q ≈ 20 are then separated according to A/Q and E/Q by the parabola spectrometer, where A, Q, E are the mass, ionic charge state, and kinetic energy of the FF (see Figure 5). Measurements at LOHENGRIN were carried out with various settings with mass numbers ranging from 84 to 144 and kinetic energies from 35 to 110 MeV in order to determine the characteristics of the detectors over a wide range of energy and mass/atomic numbers. Each scan comprised several runs with varying energy in steps of 5 MeV while keeping mass number and charge state constant. Mass-separated FF from LOHENGRIN are composed of several isobars. For illustration, the average nuclear charge from the JEFF3.3 [30] fission yield database is given for each mass. Tritons and α particles as light ions with lower stopping power were also used for characterization to verify the consistency with the off-line characterization. For this purpose ⁶Li (n,α)t reactions on the deposited LiF provide 2.7 MeV tritons and 2.05 MeV α particles. The ⁵⁹Ni(n,α) reaction provides 4.75 MeV α particles where the ⁵⁹Ni is bred by ⁵⁸Ni(n,γ)⁵⁹Ni reactions in the Ni target cover foil of the targets. Moreover, 1.4 MeV α particles are available from the ¹⁰B (n,αγ) reaction on boron collimators present in the LOHENGRIN beam tube. Given the thickness of these target materials some somewhat lower particle energies can be selected too with the LOHENGRIN spectrometer. The intrinsic FF energies are reduced by energy loss in the target and the Ni cover foil respectively, thus effectively covering a wider range of energies that can be selected by the spectrometer, 7 cm wide fission targets were used which leads to an energy acceptance of 1.0% of the spectrometer [4]. The energy calibration of the spectrometer is based on (n,α) and (n,p) reactions producing ions with well-known energies. The linearity of the electric and magnetic fields of the spectrometer are controlled by reference resistors and an NMR (Nuclear Magnetic Resonance) probe respectively. Thus the mean energy of ions selected by the spectrometer is known to be better than 0.2%.

Transverse dimensions of detectors A and B are different but due to their housing (Figure 1), the two apparent metallized detection areas are similar and equal to ∼2.4 mm². The main difference arises from the fact that detector A (and detectors C, D, E and F as well) enables reversible bias and signal readout from both sides whereas detector B can be read only on one side. The signal readout of all detectors (diamond and SiC) was performed using preamplifiers.

A low-noise broadband RF amplifier: 2 GHz, 40 dB from CIVIDEC Instrumentation Company [31] (namely CIVIDEC C2) was used in the context of timing measurement and waveform analysis. Concerning data acquisition, it was performed with a 500 MHz bandwidth, 3.2 GS s⁻¹ digital sampling “Wavecatcher” [32] system. The advantage of using this system lies in the fact that it could be configured in a continuous acquisition mode, recording all waveforms, and thus enabling large statistics for offline analysis.

The spectroscopic performances of the detectors were studied with a dedicated low noise and fast response charge sensitive preamplifier developed by INFN. The preamplifier has two channels with charge sensitivity 0.93 (1.77) V/pC, with rise time 1 (2.5) ns, a decay time 100 μs for channel 1 (2), and a maximum output voltage of 4 V. The signal from electron/hole collection side of the detector was connected to each preamplifier channel and then sent to a spectroscopy amplifier [33] with a 0.5 µsGaussian shaping time. The readout acquisition was operated on a dedicated subset version [34] of a VME general purpose DAQ based on a 32-channel 12-bit peak sensing ADC (Analog to Digital Converter) V879 from CAEN [35].

A Keithley 6,487 voltage supply, which allows a programmable inversion of the bias, was used to perform voltage inversion cycles. Such a procedure, named “cycling”, was done in order to minimize the significant polarization [12] effect generated over time mainly in the pCVD and DOI diamond detectors. Operating inversion voltages and the transition ramps were controlled by a LabVIEW program developed at laboratory [23] using the following procedure: measurement during 3 min biased at + 450 V then the bias was inverted at −450 V during 1 min with a ramp of ±   50 V s⁻¹.

Series of measurements with a light FF of mass 98 (labeled later on FF98 considered as the reference fission fragment) were carried out prior to any detector performance measurement. It was aimed to find out an optimal combination of “face exposed to FF” versus “applied bias voltage”. Figure 6 illustrates the variation of the signal amplitude as a function of the applied bias voltage for detector A when exposed to FF98 at the reference kinetic energy of 90 MeV. Indeed, due to the CVD diamond growth process, the two diamond sides (“growth” for “top” versus “seed” for “bottom”) may exhibit some asymmetries. Furthermore, depending on positive or negative bias voltage applied on diamond material, the nature of the charge carrier, electron versus hole, collected on the electrode differs. That might be a cause of asymmetries as well due to a possible difference in charge trapping in the diamond bulk for electron and holes. In the present paper, the two diamond sides will be later on arbitrarily called 0° and 180° for reasons of clarity. This choice was motivated to avoid any misunderstanding in the conclusion of our analysis given that the two faces were not previously identified in relation to the growth conditions. In Figure 6, the maximum voltage that could be applied to diamond detector A was found to be 450 V. For higher voltages, discharge spikes were observed that might damage the diamond metallization. These observations led us to consider, for future experiments, a new model of detector holder in which the electrical connection is operated more robustly by wire bonding.

Furthermore, in Figure 6, it can be noticed that the signal amplitude is increasing with bias voltage but, even if the slope of the curves decreases at higher voltage, no plateau is reached. All the measurements carried out with the different detectors confirm this observation. The significant space charge along the incident particle trace generates an electric field in the opposite direction to that applied, increasing local recombination and thus preventing an optimum charge collection. This is in relation with the PHD effect described in Pulse Height Defect.

The detection efficiency here is defined by number of ions detected by candidate detectors divided by the number of incident ions measured with a Si detector which has 100% detection efficiency in the sensitive area. Collimators with sizes corresponding to the diamond detector aperture: 2.4 mm (measured on the detector holder for both sCVD A and pCVD D) and 7 mm (same but for pCVD E) were installed in front of the Si detector. The readout electronic chain was the one used for spectroscopic measurement (Detector front-end electronic and data acquisition). The detection threshold was set above the noise level ∼1 MeV α particles with energies 1.4, 4.5 and 4.7 MeV and FFs of mass 98 (136) with different energy 100, 90 and 80 MeV (77, 70 and 63 MeV) respectively, were used to determine the efficiency.

The loss in the detection efficiency results from incomplete charge collection efficiency of pCVD caused by grain boundaries [36]. The average charge collection efficiency of CVD diamonds vary from 100% (Figure 3) to 5% depending on the diamond detector quality [23]. The loss in charge collection efficiency is maximal for highly ionizing FF since the stopping range of FF is very short (6–8 μm for light FF) while α particles have a longer range (e.g., ∼11 μm for 4.7 MeV) [26, 37].

The sCVD detector A showed ∼102% efficiency compared to the Si detector (see Table 2) where the systematically larger value is probably due to the uncertainty in the collimator geometry. This indicates good charge collection efficiency in all regions of the detector surface.

The pCVD detectors showed lower efficiency compared to the sCVD detector. The small detector D showed ∼83% efficiency. The decrease in efficiency was mainly due to the bias recycling procedure, where the detector was at the nominal bias during 70% of the measurement time (see Detector polarization effect handling), indicating the good charge collection efficiency during the normal bias up to ∼100%.

On the other hand, the 7 mm pCVD detector E showed low efficiency ∼31%, which cannot be explained by bias cycling. This could be due to the low charge collection efficiency of this diamond detector, causing the collected charge falls below the detection threshold (see Table 2).

As previously mentioned, due to the pCVD detector structure (grain distribution), the charge collection depends on hit position on the detector, where efficient charge collection is carried out in region with large grain size, while region between the large grains in the surface shows dead region. This surface effect is enhanced for low energy fission fragment which stops at the surface of the detector ultimately causing the decrease in the efficiency. As the pCVD detector E is an optical grade and not an electronic grade detector, the results are worse than for the detector D.

Detection of ionizing particles with high energy loss in solid state detectors may lead to a pronounced pulse height defect (PHD) [38, 43–47], which is defined as the energy difference ΔE between the kinetic energy E_k deposited by an incident ion impinging the detector and the apparent energy E_DD derived from the measured electric signal [38]: Δ E   =   Ek   -   EDD (4) where E_DD = N_q E_eh (N_q is the number of collected charge carriers and here E_eh=13.6 eV is as defined before the energy to produce electron-hole pairs in diamond [6, 7]). Up to now, detailed studies of PHD were performed mainly for Si detectors, less for diamond detectors [38].

The established Equation 2, derived from α and triton spectroscopic measurement on detector B gives the value of the expected ADC channel as a function of the kinetic energy. It was evaluated on particles that are not sensitive to the PHD, as explained in α particles and tritons. Using Equation 2 for FF, one would expect, if all the charges deposited by the incoming FF are collected, that the FF98 peak at the energy of 100 MeV measured at channel 2755 in Figure 8B, would be measured in fact at channel 5594 (out of range of the 12 bit ADC set-up). The difference is 2839 channels. That indubitably means that about 50% of the deposited kinetic energy is not detected. This observable attests to the PHD which means that the energy derived from the measured electric signal is different from the kinetic energy deposited by the incident ion. For detector A this difference is equal to 48%.

The electric bias for detector A was at the level of 0.9 V μm⁻¹ against 4 V μm⁻¹ for detector B. It is well known for semi-conductor detectors that space charge effects are reduced when higher electric fields are applied. This is illustrated by Figure 6, but one should notice in addition that with FF there is no saturation of the signal amplitude when the bias increases as it can be observed in the literature with α particles [23]. It means that the PHD is never cancelled. Furthermore in the present experiment, the thickness of detector A is 10 times larger than detector B, which favors charge trapping on the path to their collection by the electrodes deposited on both sides of the diamond. Consequently, one might expect a better result for detector B. However, the measured PHD is identical on the two diamond detectors. This shows that the electric field, for the two tested detectors, does not play a major role on the recombination during the charge drift toward the electrodes.

Ions deposit a maximum of their energy at the end of their path, in the Bragg peak. In the present case, given the energy transported by the ions, this will lead, very locally, to a very large amount of charges that will be generated. Taking into account the differences between detectors A and B, as detailed above, which did not lead to differences in the experimental observables for the PHD, we can therefore reasonably assume that the proportion of missing charge comes from in-situ recombination in the region of the Bragg peak.

On the other hand, what is paradoxical in our results is that, in addition, we observe a linear variation of the collected charge compared to the energy of the incident FF at given FF (Figure 9) over an energy range of 70–100 MeV for FF98. When the energy of the FF increases, its depth of penetration, and therefore the position of the Bragg peak in the material, increases also. However, the energy loss rate per unit of FF path length is quite smaller than it is in the region of the Bragg peak where the FF stops. The results of Figure 9 would therefore imply that the Bragg peak occurs at the same FF degraded energy, regardless its initial value, and energy loss at the Bragg peak depends on FF mass. Experiments with a heavier FF (FF144) in a lower energy range, from 40 up to 65 MeV, lead to the same observations. Other conclusions could have been drawn if it had been possible to go below 70 MeV for the light FF, but this was not possible in this case.

For a better understanding, another experiment was done with diamond detector B in the same conditions. One should notice that the difference is that the gain is set to a factor 1.7 lower on DAQ than for previous measurements. A set of FF were selected by LOHENGRIN: 84, 93, 98, and 102 for the light fragments with a range in energy from 70 MeV up to 105 MeV by step of 5 MeV. As well, FF 132 and 144 were selected with energy ranging respectively from 59 MeV up to 78 and 46 MeV up to 65 MeV by steps of 5 MeV. Linear fits of ADC channels versus kinematic energies were done for each FF. The results are plotted in Figure 11A.

In Figure 11B,C, the offset and slope fit parameters found for each FF are plotted as a function of the FF mass. An interdependence has been demonstrated. It seems difficult to connect with a smooth curves the measurement points for light FF and heavy FF mainly for Figure 11B (offset parameter). Nevertheless, it can be concluded that, the mass A, and respectively the nuclear charge Z of the incident FF, influences the PHD (note here that FF are selected over their Q parameter, and not Z).

This dependency seems to be complex even if we consider separately the light FF (four data points) from the heavy FF (two data points).

For light FF, in Figure 11B,C, the FF98 corresponding point stands out of the two distributions. It seems to exclude a purely linear dependency as reported in [5] for silicon detector operated at LOHENGIN in very similar experimental conditions. The main difference may arise from the fact that in [5], the silicon detector is a ΔE-E telescope detector device. The combined measurement of energy loss in the thin layer (ΔE) that the FF crosses and in the thicker layer (E) where it stops, improves strongly the FF identification.

The large difference in Figure 11B between light and heavy FF could suggest that, since the energy range in which we have taken these measurements for heavy FF is lower than for the light FF, the offset parameter is more sensitive than the slope parameter, to the incident FF energy. Indeed, there are some evidence of a “dead detector thickness” in Figure 8C (detector A) and Figure 8D (detector B) at the entrance under the metallization layer which creates an offset in the charge collection.

To conclude, as in [38], it appears from the present study that the PHD is to be explained in the framework of recombination models. For heavy ions, such as FF, the main source appears to be the recombination of electron-hole pairs in the plasma bulk produced by the FF when it stops. This theory was first proposed by [48]. As demonstrated in the present analysis, this process is not easy to characterize. It may depend on such factors as energy and intrinsic parameters of the incident FF (A, Z and Q), the applied bias voltage (the maximum electric field investigated here was 4 V µm⁻¹) as well as the distribution of recombination and trapping centers in the detector. There is an indication, with the tilt experiment, of some localized “dead zone” close to the surface, which would not be in favor of an optimal charge collection.

Diamond detectors are expected to be fast responding detectors. Consequently, a procedure was established to determine the time resolution of the FF detectors, based on off-line analysis of signal waveforms recorded with the Wavecatcher test bench. Actually, no coincidence between two detectors is possible here. Therefore, the timing resolution can be inferred only by comparing the timing difference from the signals of the two detector sides, whenever possible [24].

A numerical Constant Fraction Discrimination (CFD) was performed by averaging the background on the waveforms, determining the maximum of the pulses, and interpolating the 50% rise time value. The time difference measured between side 0° and side 180° on diamond detector A biased at −450 V is equal to 10.2 ps RMS (Figure 12A) which is the best value ever obtained on diamond to our knowledge. A comparable measurement done with detector C (pCVD) lead to a time resolution of 23.8 ps RMS against 34.1 ps RMS for detector D (pCVD) and 16.8 ps RMS for detector F (DOI) which is nearly as good as the sCVD results despite the observed inhomogeneity in charge collection (see Characterization using focused particle beams and [10]). For the three diamonds mentioned last, a cycling operation was implemented on the supply voltage as described in Detector Polarization Effect Handling to improve the charge collection. No timing resolution was evaluated with detector E due to its too poor charge collection efficiency which makes it not a good candidate for our applications. No measurement were done as well with detector B or the SiC detector but for technical reason linked to detector housing.

The FIPPS experiment can be divided schematically (Figure 13) into blocks. On the periphery of the instrument, the detector block is equipped with High Purity Germanium (HPGe) detectors (in blue in Figure 13) that allow the measurement of the γ rays emitted during the reactions taking place within the target (in red in Figure 13) chamber. These high-resolution detectors are arranged around a sphere to maximize geometric detection efficiency. Then, the collimation unit part in front of the target cell that delivers a shaped beam of thermal neutrons leaving the guide towards the target chamber. The beam stop block (called beam stop hereafter): this part makes it possible to stop the neutrons passing through the target without interactions. It is located at the rear of the instrument. Finally, the sample environment block, that is located in the center of the instrument, in the middle of the detection system. The sample environment is the site of interactions between neutrons and the target. Depending on the physics cases of interest, the targets can be stable or radioactive.

In the context of the present study, Figure 13 illustrates three possible configurations for a target (red) and FF diamond or SiC detector (grey) arrangement: (a) single-sided FF detection with a passive stopper (black) on the opposite side, (b) double-sided FF detection, (c) single-sided adjacent FF detector plus a distant FF detector wall on the opposite side. These configurations will be discussed now with regard to FIPPS expected performances versus diamond and SiC FF detectors measured performances.