Characterization of the Fast Neutron Generators for Calibration of Fusion Neutron Diagnostics

—Modern magnetic confinement fusion devices increasingly rely on extensive neutron diagnostic configurations to measure a plethora of key plasma parameters. Measurement accuracy for these diagnostics depends heavily on in situ calibration. In order to enable said calibration, we set out to propose a reliable and powerful fast neutron source, to define a characterization plan for this source in terms of yield, flux and energy distributions, propose an optimized set of tools suitable for online monitoring of neutron source performance and its metrological characteristics. In the framework of this research activity with the ultimate aim of ITER tokamak neutron diagnostics calibration we rely on industrial-grade fast neutron generator NG-24 (D-D neutron yield ~10 9 n/s, D-T neutron yield ~10 11 n/s) with sealed tube and stuffed, titanium target developed by FSUE VNIIA. The well-known analytical expressions for thick target NGs and our measurements using neutron spectrometers utilizing threshold reactions – diamond detector and LaCl 3 scintillator for D-T and D-D neutron generators respectively – were found to be significantly coherent. These data are supported through our multi-step forward modelling including ion stopping in target, fusion reaction kinematics modelling and calculating detector response based on modelled neutron spectra. We discuss methods of uncertainty mitigation of neutron spectrometer measurements. Application of both neutron activation analysis and gas-filled neutron flux monitors during source characterization and operation allows for lowing statistical uncertainty of neutron flux measurements to 1% level over 1 minute of time resolution. Over the course of several measurement campaigns the optimal set of measurement tools have been determined including detector dimensions, required acquisition time, calibration methods.


I. INTRODUCTION
HE task of fusion neutron diagnostic calibration consists of a continuous set of challenges including factory calibration of the diagnostic equipment, on-site testing and commissioning activities, in situ calibration with fast neutron source and crosscalibration during discharges.The objectives of current work revolve around characterization of a chosen source with proven reliability and suitable for the in situ neutron calibration activity.
Several aspects limit the choice of the fast neutron source for this task -available space, project schedule, cost, safety considerations and, of course, source neutron spectra.Majority of calibration campaigns of the operating magnetic confinement fusion devices relied on isotope sources, including JET [1], Globus-M2 [2] and TFTR [3].Only recently the calibration campaign at JET [4] had made use of compact sealed tube neutron generators (NGs) [5] to significant success.The key advantages of the modern sealed tube NGs are:  neutron spectrum consistent with fusion neutrons for both deuterium and deuterium-tritium plasma scenarios;  togglable operation without unnecessary irradiation going on during source transportation, storage, handling and between expositions;  point-like source (Ø2 cm) with yield of D-D and D-T neutrons comfortably reaching the yield of 10 9 and 10 11 n/s respectively;  modest weight (<200 kg) and dimensions (<Ø50×115cm) that correspond well to the Type-A package required for transportation and handling of 252 Cf source with comparable yield (10 9 n/s).The main disadvantages of this source type include -neutron yield and energy anisotropy and (for high-yield models) the need for active target cooling.Said disadvantages are an additional source of measurement error for the diagnostics undergoing the calibration, and the goal of characterization of sealed tube NG is elimination of this error.
The NG yield determination for both D-D and D-T cases requires relevant measurement techniques that allow for both time-resolved monitoring and absolute measurement of the number of neutrons produced by the source.The robust tools suitable for monitoring of fast neutron flux include threshold fission chambers [6,7] and gas counters based on boron [8] and helium-3 [9].Neutron activation analysis (NAA) technique is deemed the fusion industry standard for absolute measurements of neutron yield during discharge [10,11] and can be confidently applied for the task of characterization of a fast neutron source.
Given the nature of fusion reaction kinematics [12] the change in average neutron energy with respect to beam direction will present a significant source of uncertainty.Hence, to maximize the confidence and precision of the calibration we propose to apply for both tasks -source characterization and subsequent monitoring -a set of neutron spectrometers with the goal of determination of neutron energy distribution.For the D-D NG the detector of choice is LaCl3 scintillator [13,14] with fast neutron detection via the 35 Cl(n,p) 35 Sg.s.reaction, for the D-T NG -diamond detector with its respective 12 C(n,α) 9 Be [15,16] reaction.Both of these instruments allow for pulseheight analysis and feature a classical gaussian-shape response to fast neutrons.This approach allowed us to study the neutron energy distribution at various angle to the ion beam and thus verify the assumptions embedded in the forward modeling of the experiment.

II. MEASUREMENT SETUP
The measurement setup required for characterization of the fusion neutron source is formed by the source itself -the neutron generator, the neutron spectrometers typically located within 50 cm from the target due to their low sensitivity, neutron monitors -gas-filled ionization chambers -located at a higher distance to avoid pulse pileup.The neutron activation system is constituted of gamma-ray spectrometer and a pneumatic system delivering activation samples to and from the irradiation zone.Coverage of several neutron emission angles with spectrometers is achieved using an automated positioning system operated, together with NG and measurement system I&C from a secure control room.

A. Neutron generator
Considering the dimensions of ITER machine with major radius exceeding 6 m and minor -exceeding 2 m [17], we provision the use of NG-24 neutron generator manufactured by FSUE VNIIA [18], the main technical specifications of this model are listed in Table I.The sketch of the NG irradiation unit is illustrated in Several details that constitute the know-how of the manufacturer were omitted.The key element of the NG is the sealed tube and the detailed design around the radiative spot -NG target -has been thoroughly modelled for further research to include the copper cooling circuit, stuffed, titanium target, and sealed tube housing.The supporting structure for the scintillator-based neutron spectrometer to be located in the frontal sphere of the NG has also been designed.The duration of expositions necessary for the successful calibration of the key ITER neutron diagnostics lie in the order of hours [19], hence, a 1-minute time resolution for measurements of NG yield and energy we presume sufficient.
The tests of the NG-24 model for yield stability measured using laboratory helium-3 neutron counter have yielded good results of only ±2,5% yield change over 1 shift (7 hr 30 min) of continuous operation.Test results further underline the lack of necessity to further minimize the time resolution.

B. Neutron monitors
The task of determination of NG yield starts with determination of neutron flux at detector location.This section comprises a brief overview of achievable results with regard to determination of neutron flux value at several suitable locations around NG irradiation unit.
For this, in D-D NG case operating with the yield of ~10 9 n/s, the boron gas-filled counter is deemed most suitable.Given the Q = 2792 keV the 10 B(n,α) 7 Li reaction with the cross-section at En = 2.5 MeV reaching 1 barn, this monitor is highly sensitive to both thermal and D-D neutrons.With a fairly unsophisticated I&C system this sensor constitutes a robust solution achieving >10 4 counts per minute at locations near the front on the NG.The sensor mass being as low as 12 g allows designating its location near the back side of the front NG flange on the outer wall of the NG housing.Nevertheless, its sensitivity to thermal neutrons requires mitigation using shielding materials to prevent counts from backscattered fast neutrons.The low abundance of helium-3 [20] and the need increase bias voltage for detector operation (1200 V versus 500 V) rule out the use of said counters for the task of D-D NG characterization and monitoring during operation.
The D-T NG with the yield being 2 orders of magnitude higher (~10 11 n/s) over the D-D modification, allows for use of the threshold fission chambers (FC) with radiators covered by uranium-238 [21].The FC with 0.5 g of fissile material operating at 300 V of bias voltage is a perfect candidate for NG monitoring purposes due to reaction threshold (~1 MeV) and due to the cross-section increasing with energy up to 2 barn for En = 14 MeV.FC location directly behind the back flange of the NG allows reaching count-rates above 10 4 per minute, lowering the statistical uncertainty of flux reconstruction down to 1%.

C. Neutron activation analysis
The more precise way to provide absolute measurements of the neutron yield integrated over time is the use of neutron activation technique also applicable to magnetic confinement fusion devices [10,11,22].For this, we have utilized several activation samples listed in the Table II with the key characteristics derived based on ENDF-B/V.III [23].The D-D neutron energy range ~ 2.5 MeV is covered with only a few threshold reactions occurring in common NAA samples.The two select isotopes feature a gamma-ray energy of the same order of magnitude, but vastly different thresholds, with 115 In cross-section more closely matching the required range [24].In laboratory conditions under the sealed tube D-D neutron generator operating at accelerating voltage of 150 kV, 150 μA of ion current amounting to 10 7 n/s yield the irradiation time required exceeded 10 hours for a 0.2 g sample located at 1.5 cm away from NG target.Increasing the yield to NG-24 operation parameters, increasing the distance to ~50 cm and increasing the weight of the sample tenfold would amount to the same irradiation time needed to achieve the same 3% of statistical uncertainty in neutron fluence determination.This leaves little room for application of NAA technique for detailed multi-angle characterization of the source, but the technique still remains crucial for absolute scaling of measurements by neutron flux monitors and spectrometers provisioned for characterization and monitoring of the D-D NG.
Precise D-T NG yield determination requires less irradiation time and allows the use of samples more fit for purpose of NAA.The pure aluminum sample allows measurements using two gamma-ray energies, further lowering statistical uncertainty.The Teflon sample while having lower crosssection has a more favorable threshold reaching slightly above 10 MeV.The use of these samples with characteristic weight of 1 g under irradiation using NG-24 D-T model operating at 200 kV of accelerating voltage, 2 mA of ion current and producing ~5×10 10 n/s allowed determination of the local D-T neutron flux with the accuracy within 2% after 1 hour of irradiation.
Activation of each sample has been studied with the laboratory Ø2×2-inch LaBr3(Ce) scintillator calibrated using a set of standard 60 Co, 137 Cs and 22 Na isotope sources.
The NAA technique highlighted in this section provides an independent and robust measurement of the absolute neutron flux value necessary to minimize the final uncertainty of NG parameters after characterization.

A. Chlorine-based neutron spectrometer
Recent developments in the field of fast neutron spectrometry have led to a series of scintillating materials being utilized due to availability of 35 Cl(n,p) 35 Sg.s.(Q=615 keV, σ=0.15 barn for En = 2.5 MeV) reaction for measurement of fast neutrons.For the purposes of neutron generator characterization and consequent monitoring we propose the LaCl3(Ce) scintillator, having in mind the previous multi-step modelling and analysis of experiment with a low-power NG yielding promising results [25].With the light yield being higher and with response being faster than that of CLYC crystals, this detector looks most promising for operation in the harsh environment -within 50 cm and in direct view of the D-D NG target.The pulse-shape discrimination capability of LaCl3(Ce) is lower than that of CLYC (Figure-of-Merit = 1.0 versus >2.0) and requires digitization of the signal with sampling rate of at least 500 MSps at 14-bit resolution.The detector energy resolution in our previous work was estimated at 8% [26].
We coupled the Ø2×0.5-inchcrystal to the Hamamatsu R6231-100 PMT and conducted a series of expositions at a fixed distance of 43.5 cm from the target center of D-D modification of the NG-24 operating at 220 kV of accelerating voltage and 1 mA of ion current which corresponds to 5×10 8 n/s yield.The only parameter varying in between the expositions was the angle between the detector line of sight and the NG axis.Following the digital pulse-shape discrimination procedure (PSD) [27], we conducted the pulse-height analysis of the detector pulses, associated with slowing down of protons born in the (n,p)-reaction for each of the expositions.The results are illustrated on Fig. 2. The figure-of-merit of current crystal sample was found to be lower and averaged at 0.45 across the expositions, this is primarily due to increased 227 Ac contamination [28] of the crystal compared to previously used samples.At the same time, background radiation arising from this contamination in the form of α-decay events was recorded in between each irradiation for the duration of ~5 minutes.No visible deviation has been found in the shape of the pulse-height distribution or the position of the α-decay event peaks.This fortunate circumstance allowed us to self-calibrate the detector using known locations of the peaks, and continue to monitor the energy calibration stability between each exposition.Using the p/β-ratio derived from the previous measurements (~0.75) and (n,p)-reaction kinematics allowed us to infer average neutron energies for each of the irradiation angles.The results of the measurement are summarized in the Fig. 3 and Table III.The agreement between measurement and the data for thick targets by J. Csikai [29] is within 2%.The absolute measurements of neutron flux is undermined by discrepancies in the cross-section data for the 35 Cl(n,p) reactions in the literature [23,30,31].Experiment data yield incoherent results in terms of absolute yield reconstruction and requires further investigation using a metrological fast neutron source.
Detector performance in terms of relative measurements and average neutron energy determination is on substantial level, allowing it to be used as a primary tool for both characterizing the NG and monitoring its performance during operation.Provisioned placement of LaCl3(Ce)-based detector in the frontal sphere of the NG within 50 cm of its target allows reaching count rates above 10 4 events per minute in the (n,p)-associated peak without significant pulse pileup and PSD degradation, which is effectively aligns with results of the experimental campaign.

B. Diamond neutron spectrometer
The semiconductor detector based on a CVD-diamond mono-crystal is a widely adopted [32][33][34][35] method for measurement of D-T neutron energy and flux.Having a high critical fluence of ~10 14 n/cm 2 and impeccable energy resolution ~1% this detector highly benefits form predominantly (except for impurities) carbon composition and a threshold reaction that produces a classical gaussian-like response -12 C(n,α) 9 Be, Q=-5.7 MeV.
Utilizing the 4×4×0.5 mm 3 crystal coupled to a Canberra 2004 preamplifier, Ortec 673 Amplifier and Ortec 926 analyzer we have conducted a set of measurements at 10 cm from the target of the the NG-24 (D-T modification) operating at 200 kV of accelerating voltage and 0.5 mA of ion current, corresponding to the yield of ~2×10 10 n/s.Detector energy calibration was performed using isotope source of α-particles 226 Ra and its analytically estimated sensitivity is ~6×10 -5 cm 2 .
This set of measurements allowed us to study the change in the pulse-height distribution with the angle between detector LOS and NG axis.The PHA of detector for various irradiation angles is illustrated on Fig. 4 with each exposition time being equal to 180 s.Verification of measurement results was done for two critical location -0 and 90 degrees.For this, we utilized the multi-step modelling approach briefly described in [36].The neutron generator beam ion composition was presumed 90% molecular ions (45% DT + , 22.5% D2 + , 22.5% T2 + and 10% single (5% D + and 5% T + ).The ranges for stopping of said ions were calculated using SRIM software for the titanium bulk material, each ion had its velocity vector recorded, with that, beam-target reaction kinematics was calculated and the resulting neutron energy and angle distribution was modelled.Beam componentwise energy distribution of the neutrons born in the direction along the NG axis is illustrated on Fig. 5.
The resulting energy distribution convoluted with (n,α)reaction cross-section was plotted against the results obtained in experiment and shows significant coherence in shape and absolute number of events as illustrated on Fig. 6.Diamond detector performance for the purposes of characterizing the D-T NG demonstrates stellar results.With that, its location within 30 cm from target of D-T NG operating at 10 11 n/s leads to a count rate of the order 5×10 3 per second and its increase would likely constitute a pulse pileup due to significant number of pulses occurring at lower amplitudes simultaneously.The statistical uncertainty in determination of local neutron flux achievable over 1 minute of spectra integration time falls well below 1% in this case nevertheless.

IV. CONCLUSIONS AND OUTLOOK
In present work we outline the set of key tools necessary for conversion of the industrial-grade sealed tube neutron generator into metrological source fit for calibration and testing of the fusion neutron diagnostics.
Based on extensive experience of operation of sealed tube NGs our recommendation is to use multi-vector approach with simultaneous measurements done by means of neutron activation samples, conventional neutron monitors and unique neutron spectrometers, thus allowing to cover both the absolute neutron fluence measurements, consistent neutron flux monitoring and detailed energy versus angle characterization of the source.
Tests in laboratory conditions allow us to pinpoint the optimal sensor locations for each type of the source provisioned to be used in the ITER calibration campaigns -the sealed tube NG-24 neutron generator.The increase of the number of sensors in each case is a subject to a detailed reliability study.
For D-D source (yield 10 9 n/s) the boron counters are to be located directly behind the front flange of the NG, on the outer surface of the NG body, with careful consideration of the shielding conditions.The neutron spectrometer suitable for both characterization and subsequent monitoring is the lanthanum chloride scintillator, shown able to consistently (<2%) determine the average energy and therefore monitor the stability of NG accelerating voltage.The location in the frontal sphere within 40-50 cm range will allow to drive the statistical uncertainty of neutron flux determination based on this set of two tools well below 1% over each minute of the ongoing exposition.
For D-T source (yield 10 11 n/s) the uranium-238 fission chambers are a robust tool that may be located behind the back flange of the NG.This location will also require careful design of shielding to prevent backscatter counts.The diamond neutron spectrometer seems a perfect candidate for monitoring of both neutron flux and neutron spectrum from a location in the frontal sphere of the NG.The low dimensions of diamond detector (2-3 cm 3 ) constitute very little obstruction to the diagnostics undergoing calibration.The cumulative statistical uncertainty of neutron flux determination of this set of tools also reach values well below 1% over 1 minute of irradiation.
The multi-step modelling approach developed and tested over the span of these experiment campaigns provided us with important insight into NG performance and verified a set of measurements of both diamond and chlorine-based neutron spectrometers.The sum total of modelling the ion stopping in the target and the beam-target reaction kinematics allows for analytical definition of the NG source that is both usable in future modelling and is consistent with the measurements.

Fig. 2 .
Fig.2.Pulse-height analysis of proton-associated events in the LaCl3(Ce) response after PSD plotted for several angles between detector line-of-sight and NG axis, calibrated in the gamma-ray equivalent energy.

Fig. 3 .
Fig. 3. Neutron energy measured using LaCl3(Ce) versus the angle plotted against analytical approximation of thick target NG data.

Fig. 4 .
Fig. 4. Diamond detector PHA under D-T NG irradiation plotted for various angles between detector LOS and NG axis, exposition time = 180 s.

Fig. 5 .
Fig. 5. Diamond detector PHA under D-T NG irradiation plotted for various components of the ion beam for the detector LOS along NG axis, exposition time = 180 s.

TABLE III NG
NEUTRON ENERGY MEASUREMENT VERSUS THICK-TARGET DATA