Validation of calculational determination of 18 O(p,n) secondary neutron field

. Cyclotrons used to produce medical isotopes are relatively widespread. Nowadays, it is popular to place small and compact accelerators directly in hospitals. This approach simplifies handling of the produced radiopharmaceuticals, but it imposes strict radiation safety measures during production. For optimal utilization of isotope production cyclotrons, the exact knowledge of leakage neutron field is essential due to the deep penetration ability of the high energy neutrons and the accompanied secondary radiation production. Our paper presents measurement of the neutron leakage spectra for various angles from an open target assembly at the cyclotron U-120M at NPI of CAS. These spectra are compared with data obtained from a compact medical cyclotron IBA Cyclone 18 / 9 in UJV Rez and also with activation measurements of reactions with di ff erent threshold energies. All data were also compared with calculations made with di ff erent Monte Carlo codes using both models and data libraries. The preliminary results show significant disagreement between experiments and theoretical predictions. These findings could have implications not only to the nuclear data community but also to the production accelerators operators at the licensing stage.


Introduction
Positron emission tomography is widely expanding and becoming more widespread. The most common radionuclide used in PET is 18 F, which is usually produced in compact accelerators by protons bombarding water targets enriched with 18 O. It can be further expected that new compact cyclotrons will be installed in hospital buildings, because it makes the manipulation with radiopharmaceuticals easier and enables utilization of short-lived isotopes. The importance of radiation safety during operation of cyclotron, which is source of fast neutrons, rises with this phenomenon. Radiation protection issues are gaining importance as production increases on current cyclotrons which lead to higher radiation loads than originally designed.
We have already measured the secondary neutron field around the target at the cyclotron IBA Cyclone 18/9 in UJV Rez [1,2], which allows stilbene spectra measurement only at specific angles because of the steel shielding around the cyclotron. But spectra in other angles are also needed, for example to determine the activation of the cyclotron's structural parts during decommissioning. For this measurement, an open target of the cyclotron U-120M at the Nuclear Physics Institute of the Czech Academy of Sciences, which is also used to produce 18 F, was chosen. * e-mail: marek.zmeskal@cvrez.cz * * e-mail: michal.kostal@cvrez.cz * * * e-mail: matej.zdenek@mail.muni.cz

Experimental setup and methods
Spectra of secondary neutrons produced by the reaction of 18 MeV protons in a water target enriched in 18 O were measured at different angles at the cyclotron U-120M at NPI of CAS. The obtained data were compared with previously measured data at the IBA Cyclone 18/9 production cyclotron. A comparison of the two target systems modeled in Geant4 is shown in figures 1 and 2. The neutron spectra were measured using an organic scintillator (dimensions 1 × 1 cm) coupled to a fast two-parameter spectrometric system NGA-01 [3] equipped with an active voltage divider. Pulse Shape Discrimination unit is than used to distinguish the type of the detected particle by analyzing the pulse shape, whereas particle energy is evaluated from the integral of the whole response. Acquired proton recoil spectra are then subjected to the deconvolution by the Maximum Likelihood Estimation technique. In addition to the spectra measurements, measurements of the reaction rates in activation foils of aluminium and nickel were performed. These were irradiated at the same angles at which the spectra were measured, at distances of 100 mm to determine the spectral indices and 269 mm to determine the absolute values of the reaction rates. The activation foils were then transferred to a well-characterized HPGe detector whose efficiency is determined by a computational model.
All data were also simulated using multipurpose Monte Carlo codes Geant4 [4] and MCNP6.2 [5]. In Geant4, two approaches were used to simulate the proton reaction. Both the default model with the Bertini cascade through the Shielding reference physics list and the TENDL-2019 data library [6] implemented through the QGSP_BIC_AllHP physics list were used. In MCNP6.2, the default model CEM03.03 was used.

Neutron spectra
In figure 3 a comparison of neutron spectra taken at 1m flight path for different target configurations is shown. As can be seen, they all exhibit a similar shape, which roughly corresponds to the evaporation spectrum. Compared to the spectra from the IBA Cyclone 18/9 facility, the open-geometry spectrum at U-120M has a smaller fraction of low-energy neutrons because the target is smaller and there is no shielding, where neutrons can be slowed down.
In figure 4, the spectra at 0 • and 90 • measured on the U-120M cyclotron are compared with spectra from the EXFOR database obtained in an experiment at the Japanese TIARA facility [7]. The spectrum at 0 • is compared with the spectrum from 15 • . The neutron flux at 0 • is significantly lower than at 15 • , because the target in the second experiment was much less massive and the neutrons are heavily scattered by the water cooling located in 0 • direction. At 90 • there is already a fairly good agreement between these two spectra. The measured spectrum is also compared with the calculated spectra, where a strong underestimation of the spectrum can be observed for the models in Geant4. A similar pattern for spectra calculated with MCNP6.2 and TENDL-2019 can be seen, which overestimate the spectrum at lower energies and underestimate it at higher energies.    Table 1 contains the absolute values of the measured reaction rates and their comparison with the calculated ones. A reasonably good agreement of the calculated reaction rates using the measured neutron spectra can be seen for reactions with different energy thresholds. The Geant4 calculation using the physic models strongly underestimates all reaction rates as was seen in the comparison of the spectra. The calculation using TENDL-2019 predicts the reaction rates of the production of 58 Co reasonably well, but underestimates the reaction rates with higher energy thresholds. The calculation in MCNP6.2 then even overestimates reaction rates with lower energy thresholds and underestimates those with higher ones. Spectral indices of reaction rates with different energy thresholds can be suitably used to compare the shape of the spectra. The ratio of the reaction rates of formation of 58 Co and 57 Co in nickel is well suited for this purpose. In table 2, one can see how this ratio increases from 0 • to 90 • due to the loss of fast neutrons originating from the reaction kinematics. It can be seen that the spectral index calculated using data from the EXFOR database is quite strongly underestimated, indicating a lower yield of high energy neutrons. With the measured spectra, the spectral indices are in agreement within 10%, which is a very good result. As seen with the spectra comparison in figure 4, in addition to the overall underestimation of the spectra, Geant4 also underestimates the yields of the highest energies, as indicated by the overestimation of the spectral indices. As for MCNP6.2, it shows a strong anisotropy of the high energy neutron production, while for 0 • it overestimates the spectral index by only 40%, for 90 • it is already over 1200%.

Conclusion
Characterization of the secondary neutron spectrum produced during the production of the radionuclide 18 F at two different angles (0 • and 90 • ) has been performed using a stilbene spectrometer with accompanying measurements of the reaction rates. The possibility of measuring such spectra with the NGA-01 spectrometer set has been demonstrated. Furthermore, the inconsistency of the EX-FOR data with the measured reaction rates and their overestimation of high energy neutrons were shown. Discrepancies were also shown for the data calculated by simulation programs, where none of the models provide satisfactory results for all neutron energies and angles studied in this work. Among other things, a strong underestimation of the neutron yield for the physics models in Geant4 was shown.