Quantitative investigation of gamma radiation in accelerator produced ISO neutron fields

In standard monoenergetic ISO neutron fields, the neutron yield of neutron-producing reactions was measured in combination with the prompt photon yield, including photon energies up to 10 MeV, for the purpose of comparing the two yields. Separating the photons produced by the target (direct photons) from those generated by secondary neutron reactions was achieved using the time-of-flight method. Photon and neutron ambient dose equivalent values were calculated from measured spectral energy distributions. Quasi monoenergetic neutron fields are needed to systematically test the response of measuring instruments to neutron radiation. For this reason, ISO has defined a number of reference neutron radiation fields covering a wide energy range up to 19 MeV. Because neutron detectors may also be affected by photon radiation, the photon fluence in the ISO neutron fields has to be known. This work focuses on quasi monoenergetic accelerator-produced neutron fields in the energy range of 24 keV to 19 MeV.


Introduction
Precise radiation measurements are only possible if the detectors employed are well characterized.To achieve this, standardized radiation fields are needed so that comparable testing procedures can be performed at competent metrological institutes or other laboratories.Methods of producing reference neutron radiations are described in ISO 8529-1 [1].This standard lists reference neutron fields produced by reactors, by radioactive sources, and by accelerators.International key comparisons of neutron fluence measurements have been performed using most of these standardized fields (e.g.[2]) because they are an important metrological basis for neutron radiation measurements and radiation protection.Parts 2 and 3 of the standard are dedicated to the calibration of area and personal dosemeters [3] and to their response functions with respect to energy and angle of incidence [4].While sources can be used for calibration and routine testing (the latter for verifying instrument stability), compliance or type testing requires the application of monoenergetic neutron beams to determine the angular and energy response of an instrument.At the PTB ion accelerator facility (PIAF), monoenergetic neutron fields from 24 keV to 19 MeV are produced by nuclear reactions.The installed Tandetron can accelerate protons and deuterons up to 4 MeV, while the energy region up to 19 MeV (protons) and 13 MeV (deuterons) is covered by an isochronous cyclotron.Beams of both accelerators can be directed to the same experimental measuring stations and targets.
The accelerator ion beam also produces photons in the target.In addition to the neutron-producing nuclear reactions in the target, other parasitic reactions also produce photons, as do the neutrons that were created.The directly produced photons emitted from the target are of particular interest because they may also exert an influence on neutron detectors but are not subtracted by shadow cone measurements.Instruments that are calibrated in neutron fields often respond to photons as well.For such devices, it is necessary for both the neutron and photon components of the calibration field to be known.This work is dedicated to the measurement of the photon component relative to the neutron fluence, to the ambient dose equivalent, H * (10), and to the personal dose equivalent, H p (10) [5].Previous papers on this topic covered only a small number of neutron fields or had to cope with high uncertainties, e.g.[6][7][8].The total uncertainty of the data in this paper is less than 7%.In the following, the term measured spectrum denotes a measured pulse height distribution (whose x-axis is proportional to the deposited energy), while the term (unfolded) spectrum denotes an energy distribution of the photon fluence, i.e. the spectral fluence.
Photon spectra are recorded by means of scintillation detectors equipped with CeBr 3 crystals.Their pulse height resolution is much higher than that of the liquid scintillation detectors used for similar investigations in the past [9].In contrast to these detectors, the response of CeBr 3 crystals to neutrons is quite low.The neutron radiation is therefore quantified separately using long counters (or other neutron detectors).The photon spectra of the CeBr 3 detectors show a number of photon emission lines, which helps to interpret the spectra and to verify the measurements.The different processes that produce photons are separated, on the one hand by performing time-of-flight (TOF) measurements and on the other by systematically performing shadow cone measurements.The photon spectra are evaluated by applying different algorithms [10,11], in particular the GRAVEL unfolding algorithm, to obtain fluence values.
The methodology consists of the following steps: • The use of a CeBr 3 detector whose response has been validated over the entire energy range of the photons produced.The photon response matrix of the CeBr 3 detector has first been calculated and experimentally validated in photon reference fields.• The determination of corrections of these measurements for all photons other than those produced directly in the target (using in particular the TOF method).• Determination of the energy distributions of the photon fluence, after the previous corrections, and of the associated integral values after deconvolution.An analysis of the various photon peaks is found in the appendix.• Calculation of dosimetric quantities by applying conversion coefficients from fluence.
• A study of the relative contribution of prompt photons to total fluence and dose equivalents, including for irradiation of instruments on a calibration phantom with several angles of incidence.

Accelerator facility and neutron reference fields
The ion accelerator facility of PTB (PIAF) comprises a Tandetron (max.proton energy: 4 MeV), which is based on a doubled electrostatic Dynamitron, and a cyclotron (max.proton energy: 19 MeV, max.deuteron energy: 13 MeV, max.alpha energy: 26 MeV).For TOF applications, the Tandetron accelerator beam can be pulsed with a frequency of up to 2.5 MHz and a pulse width (FWHM) of 2 ns.The cyclotron can produce pulses with widths of a few nanoseconds and repetition rates of about 1 MHz.More detailed information on projectile energies and targets is provided in table 1.The deuteron beam for the 8 MeV fields was produced by the cyclotron.For all other neutron fields, ion beams from the Tandetron were used.The target at the end of the beam line is positioned in the middle of a cubic experimental hall of dimensions 24 m × 30 m × 14 m in order to minimize the contribution of neutrons scattered by the walls.In addition, as few objects as possible are located in the area surrounding the target.Detectors are positioned at a distance of between 1 m and just under 4 m from the target.Such distances are sufficient to separate the TOF signals of prompt photons and neutrons in the photon detector.The typical time evolution of the detection of the accelerator beam pulses is shown in figure 1.The solid targets are cooled with air and rotated around the accelerator beam axis such that the spot heated by the accelerator beam moves in a circle on the reactive surface of the target.The titanium solid-state targets, in which tritium is implanted, can be described by the chemical sum formula TiT x , whereby here x ≈ 1.5 can be assumed for fresh targets.The fact that this number is not exactly known is of no matter for this work because the main goal here is to derive the photon to neutron ratio for fluence or dose; the absolute yield is not important.

Detector systems
A scintillation detector with a CeBr 3 crystal (in both length and diameter 1.5 ′′ ) is used to record the TOF photon spectra.The elastic scattering of neutrons in the massive CeBr 3 crystal only produces signals of very low energy that do not disturb the measurements.The advantages of this detector type outweigh the small disadvantage that the relative energy resolution is slightly lower than that of a LaBr 3 detector (about 4% compared to 3% at 662 keV).The light signal is amplified by a Hamamatsu 13 089 type photomultiplier.This very fast photomultiplier is well suited for TOF measurements.
In case the target produces intense electron and positron radiation, the detector crystal is shielded by a cylindrical polyethylene cap with a wall thickness of 20 mm.The absorption of photons by this cap is negligible in the energy range considered (>30 keV).The amplified detector signals are read out using the MPA-3 multi-parameter data acquisition system by FAST, which is especially useful if a TOF synchronization with a pulsed accelerator is necessary in the nanosecond range.
A De Pangher long counter is used to measure neutron fluences.Inside a moderator, which is shielded against scattered neutrons coming from the side and back, a BF 3 tube is included to detect thermal neutrons.The long counter has a rather flat neutron energy response and so allows a reliable quantification of the fluence.The neutron detector and the photon detector are positioned one after the other at the angles listed in table 1 (mostly at 0 • straight in front of the target on the axis of the beam line).The total deposited beam charge in the target is used to normalize the fluences and doses.Furthermore, another permanent neutron monitor exists, namely a long counter with a polyethylene moderator and a cylindrical 3 He proportional counter inside.It is positioned at an angle of 18 • .An energy compensated GM counter (ZP1313 Centronics) is installed backward at an angle of 118 • and serves as a photon monitor.Because the energy compensation  of this device is insufficient at photon energies above 1.5 MeV (the relative kerma response rises from 1 at 1.250 MeV to 2.3 at 10 MeV), field specific correction factors have to be determined and later applied in routine measurements [8].

Measurement procedure 2.3.1. General aspects
The CeBr 3 detector and the long counter are irradiated consecutively following the list of energies and angles given in table 1. Shadow cone measurements are performed to obtain background spectra, which include photons generated by secondary reactions of neutrons in the air and in the surroundings as well as by the natural background radiation.The photon spectra of the CeBr 3 detector are recorded continuously, triggered solely by the signal of the anode of the photomultiplier attached to the CeBr 3 crystal.The pulse height of the signal of the last dynode is amplified and fed into one analogue-to-digital converter (ADC) of the MPA-3 data acquisition system.Another ADC samples the time difference between the anode signal of the CeBr 3 detector and a stop signal generated by each beam pulse in a pick-up coil close to the target.This TOF difference depends on the velocity of the particles and is thus related to the kinetic energy (figure 1).The timing information is observed as the count rate of the anode of the CeBr 3 detector as a function of the time elapsed after each accelerator pulse.
In the case of a solid-state target, the count rate produced by the rotating target is monitored by the long counter at 18 • to check the homogeneity of the target with the irradiation time.In addition, the counts of the monitors and other parameters are recorded by digital scalers.The TOF conditions are implemented by the software of the data acquisition system.A coincidence condition is set by creating a window around the photon peak displayed in figure 1.In this way, only the spectrum of the prompt photons of the target is obtained.When a gas target is used, measurements are performed with a filled target cell and with an evacuated gas cell ('gas in' and 'gas out'), also performing a shadow cone measurement in both cases.Furthermore, 'blank target measurements' are performed in some cases to better understand the peak structure of the spectra.'Blank targets' are identical to the corresponding nominal targets except they do not include the designated neutron-producing isotope deuterium.
The time dependency of the total count rate, synchronized with the frequency of the pulsed accelerator beam, is depicted in figure 1.Only those events of the CeBr 3 detector that coincide with a window around the peak of the prompt photons (green) are collected in the foreground pulse height spectrum.After placing a shadow cone between target and detector, a background spectrum is recorded using the same coincidence condition.In figure 1, the photon peak produced by direct neutrons i.e. neutrons coming directly from the target and arriving at the detector about 100 ns later (depending on the neutron energy and length of the flight path) are visible to the right of the photon peak.A non-ideal behaviour of the pulse may produce 'shadow pulses' .These are, however, tiny in the figure shown here.Random photons are collected during the red time window (details below).
In general, the recording of one foreground or background photon spectrum lasts 3000 s to 6000 s.In order to better control possible interfering effects, e.g. a temperature drift of the detector or unexpected behaviour of the accelerator beam, all recorded spectra are stored in several runs of 600 s to 1000 s.Before adding the spectra of each run, they are checked for energy drift and normalized if necessary.

Investigation of the random background
The random photon background is obtained by defining a coincidence condition based on a window in the TOF histogram before the photon pulse (red window in figure 1).Besides the radiation of natural radionuclides in the surroundings, this background includes photon radiation caused by the activation of the target either by the incident beam or by neutrons (especially above 5 MeV), of the detector and of the elements in the accelerator hall, as well as the photons created by neutron interaction (mainly inelastic scattering) in the hall.Because the spectra are recorded in list mode, the coincidence conditions can be altered in any manner during a subsequent analysis.
The net random background is the difference of the foreground random spectrum and the background random spectrum, which is recorded behind the shadow cone.A net spectrum of the random background must be caused by photon emissions produced by one or two of the following mechanisms.Either long-lived activation products in the target (which are generated by the bombardment with the accelerator beam but decay slowly) lead to photon emissions (not synchronized with the accelerator beam), or nuclear states in the detector material are excited and decay slowly relative to the pulse repetition time of the accelerator beam, while the respective decay photons are not visible in shadow cone measurements during the small time window of the prompt emissions.It follows that these photo emissions must be induced by an activation of target materials or detector materials, with the latter only visible in the random net spectrum if the activated nuclei have decay times of fractions of seconds to minutes.A distinction between the two mechanisms is only possible if emission peaks in the random spectrum can be clearly identified.Photons originating from target activation are an integral part of photon spectra that exist in ISO neutron fields.
As an example, a random spectrum is shown in figure 2 (top) that is obtained when a 400 keV deuterium beam is shot at a titanium target with implanted tritium (backing: aluminium) producing 15 MeV neutrons at 0 • (inducing neutron activation of the target and surrounding materials).In this spectrum, much fewer peaks are visible than in the spectrum of the prompt photons (bottom) because the only long-lived excited states that are visible are those of the detector material ( 79 Br, 81 Br, 139 Ce) and of the alloys in the target region, perhaps of the target itself.The peaks in figure 2 (top) are for the most part dominated by the activation of detector materials.The 207 keV state of 79 Br has a half-life of 4.9 s and is excited by 79 Br(n, n ′ ) reactions.The 81 Br(n, n ′ ) reaction excites the 536 keV state of 81 Br (half-life 35 µs), which emits a 260 keV photon.The 754 keV state of 139 Ce, excited by the reaction 140 Ce(n, 2n) 139 Ce, decays with a half-life of 55 s by emitting photons of the same energy.Traces of tin, either in the aluminium backing of the target, in the detector housing, or in the steel of the beamline near the target, are activated by the reactions 120 Sn(d, p) 121 Sn, 122 Sn(n, 2n) 121 Sn and 124 Sn(d, p) 125 Sn.The latter decays with a half-life of 9.5 min, mainly by emitting 332 keV photons. 121Sn emits 176 keV photons via the 2835 keV level (half-life of 0.17 µs).
In comparison with the accelerator pulse repetition time, all described nuclear levels or isotopes decay slowly and a permanent background of the corresponding photons is observed.If the net random background is subtracted, however, excitations of the detector materials are no longer visible (figure 21, top).In addition to these photons, the spectrum of the prompt photons (including the random background) depicted in figure 2 (bottom) shows a number of 27 Al and 56 Fe peaks, which are all produced by 27 Al(n, n') Table 2. Relative random to net ratio for the prompt photons in terms of fluence, air kerma and ambient dose equivalent.The ratios are expressed in percent.and 56 Fe(n, n') reactions, i.e. by inelastic scattering.The excited levels of both isotopes have half-lives of between a femtosecond and some ten nanoseconds.They are produced by high-energy neutrons (almost) at the moment when an accelerated deuterium bunch hits the target.Because the deuteron beam has only an energy of 0.4 MeV, excited states above that energy must have their origin in neutron reactions.The levels, which have half-lives of some 10 ns, and the flight time of the neutrons cause the tailing of the peaks displayed in figure 1.

Ratio in % concerning
If the time-dependent structure of the spectra is plotted in a two-dimensional diagram (figure 3), further details can be made visible.In this figure, the time axis is similar to that of figure 1; but here both the sum of the photon counts as well as the spectral distribution are coded in false colours, as shown in the legend of this figure.Consequently, the long-lived states leading to quasi-permanent peaks in figure 2 cause structures in figure 3 that can be perceived as horizontal lines.In a completely background-free environment, only the peaks around 51 ns and 117 ns (time axis) would be present, while the rest of the diagram should be empty.The percentage of net random fluences and doses with respect to net fluences and doses of the spectra of the prompt photons ('net' in each case meaning the difference between a foreground and background spectrum) is listed in table 2.
The total fluence and dose rate values of the net background spectra could not be evaluated by unfolding (section 2.4) due to high statistical fluctuations.In many spectrum bins, differences of values of nearly identical size are found, and many bins even have negative entries (in some regions, the spectra oscillate around zero).Instead, the conversion method [10,11] was employed, which uses a pulse-height weighting function to modify the energy dependence of the photon detection efficiency ε(E) in such a way that the weighted integral of the pulse-height distribution becomes proportional to integral quantities such as the photon fluence Φ or ambient dose equivalent H * (10).This procedure does not allow the calculation of the energy distribution of random photons, but it does provide an indication of the total relative contribution of random photons to the dominating prompt component.Concerning the random background, the mean value of each net value and gross value (the latter derived from spectra without shadow cone measurements) will probably be close to the actual random background level as the net values can be regarded as lower limits and the gross values as upper limits of the true values.It should also be noted that the net fluence values are affected by the flaws of the shadow cone method, e.g., partial shielding of the areas behind the photon source and the detector.Moreover, the simple subtraction of measurements with shadow cone in place from those without shadow cone does not account for the time dependence of the fluence of photons resulting from target activation.
When a triangular probability function is assumed, it is possible to calculate uncertainties of the obtained mean values table 3).The relative uncertainties in this table show that the precision of the fluence and dose ratios is limited.However, the values can be used to judge the quantitative influence of the random photons relative to the prompt photons.As a major source of uncertainty, the data in tables 2 and 3 are strongly influenced by the size of the TOF windows (prompt and random region) because the peak of the prompt photons in figure 3 has a Gaussian shape with a tail, meaning that the included amount of prompt photons decreases exponentially with a growing TOF window size while the amount of random photons depends proportionally on the TOF random window size.This table therefore serves to show the magnitude of the random background.
An evaluation of the foreground random spectra will yield upper limits of fluences and doses of the total random background because these spectra include not only detector events caused by photons but also neutron-induced events.However, the net random spectra no longer include the diffuse radiation of the detector environment (after applying the shadow cone method, only photons with a direct path from the target region are detected).They may serve as an estimate of the lower limits of the fluences and doses of the random radiation.Even though the influence of the random photons on the results is small, the net random background (i.e. the random background of the foreground spectrum minus the random background of the background spectrum) was in the following subtracted from all the net spectra of prompt photons in order to obtain background-corrected results.
The conclusion drawn from a quantitative evaluation of gross and net random background spectra (table 3) is that either the percentage of random photons is low (especially at low neutron energies) or the amount of photons in comparison with neutrons is small (at high neutron energies), meaning that the random background does not play an important role in either case.The random background was nevertheless systematically taken into account when creating the catalogue of photon spectra (see annex).The result of unfolding (after seven iterations) is displayed in red.The reconstructed spectrum using the response matrix (used by GRAVEL to adapt and check the unfolded spectrum [10]) is plotted in light blue.
The lack of knowledge concerning most of the time scales for the build-up of activations and excited states, and for their de-excitation, both of which depend on the irradiation history, prevents the exact quantitative measurement of the random background fluence.

Long-lived F-19 nuclear excitation state
When a target containing fluorine, typically LiF, is used, the following special effect has to be taken into account.The 197 keV level of 19 F emits a line that dominates the measured photon spectra with a very pronounced peak at the same energy (visible, for instance, in figure 4 in the form of the leftmost peak-see also table 10 in the Annex).This energy level has a relatively long half-life of 89.3 ns [12], meaning that the time window around the TOF photon emission peak of some 10 ns is too short to enable the detection of a major part of the photons of the decay.As such, the measured 197 keV peak in the photon spectrum is far too small.The height of this peak therefore has to be corrected so that there is no dependency of the peak height on the length of the TOF window of the direct photons.The number of detected 197 keV photons of one spectrum was measured as a function of the size of the TOF window of the prompt photons (figure 5) by replaying a list-mode file.The half-life used for the fitted curve is 89.3 ns.The relative amount of detected 197 keV photons as a function of TOF window size is corrected using this curve when a LiF target is applied.For a correct unfolding of the whole spectrum, the peak height in the photon spectrum likewise has to be altered.
Because the 197 keV peak is more pronounced than every other peak, it needs to be given special consideration.Other long-lived nuclear states observed in the photon spectra of the neutron ISO fields produce only small peaks, which are of no importance when quantitative evaluations of total fluences and doses are made.Peaks produced by long-lived states of Ce and Br, of which the detector crystal is made, are removed by subtracting the random background.

Calculation of fluences and doses
The TOF net spectra of the prompt photons are obtained by subtracting the shadow cone measurements from the foreground measurements.The measured TOF spectra are normalized to the accumulated accelerator beam charge on target.For each neutron beam quality and run, the neutron fluence rates and dose rates in terms of Ḣ * (10) and Ḣp (10) per charge are derived from the long counter measurements.The yield (particles per steradian) can be directly calculated from the fluence.From the recorded photon spectra, the quantities of fluence (and yield), air kerma, H * (10) and H p (10) are derived by unfolding the spectra.The iterative GRAVEL algorithm (based on the SAND II algorithm but including uncertainties as input parameters) is used.It takes a first estimate of the fluence spectrum and in a few iterations adapts it to produce the undisturbed fluence spectrum at the detector position.The GRAVEL formulas alter the estimated fluence spectrum in such a way that a test spectrum (the unfolded spectrum times the response matrix) is adapted to the measured spectrum in iterative loops [11].
The fact that GRAVEL cannot handle bins with negative or zero entries is a relevant drawback, especially when processing difference spectra.Simply excluding such bins may lead to flawed results.Another problem is the abort criterion of the iterations.Ideally, the iteration will be stopped when a good agreement is reached.The uncertainties in the measured spectrum and the response matrix make the abort criterion more complicated.However, the total fluence (or derived dose) is only weakly dependent on the exact termination of the iteration process.After a few loops, the integral under the unfolded spectrum remains almost constant.The measured spectrum can readily be used as a first estimate of the unfolded spectrum, especially if it shows clear photo peaks.If the likely peak structure of the unfolded fluence spectrum is anticipated in advance, an artificial input spectrum may even lead to a more realistic result.In either case, a good result is generally obtained after about 10 iteration loops.Example spectra are displayed in figure 4. The conversion of fluence spectra to air kerma is done using a fit function of the ICRU coefficients published in [13].A conversion to H * (10) or H p (10) is carried out by utilizing fits of the extended tables of ICRU coefficients [14].Fit functions are much easier to implement in computer codes and they avoid local interpolation problems.

Verification of spectrum analysis methods
The detector response matrix has been calculated using the MCNP6.2Monte Carlo code, which transports both photons and electrons.The photon and electron cross-section data were taken from the libraries mcnplib84 (library 84p) and El03 (library 03e) electrons).Both photons and electrons were tracked down to 1 keV (according to the MCNP standard settings).Doppler broadening was switched on.Bremsstrahlung photons were generated with a thick-target bremsstrahlung approximation (according to MCNP standard settings).The geometry was implemented as meticulously as possible in terms of geometries and materials, even including, for example, the mu-metal-shield around the photomultiplier tube and the detector plastic holder.The correct implementation and use of the Monte Carlo code MCNP was verified using test measurements in the photon fields of radioactive sources and the accelerator produced R-F photon field [10].Measured spectra on the one hand were compared with simulated spectra on the other.The simulated MCNP spectra were folded with the detector resolution (obtained from measurements in a variety of photon fields covering a wide energy range) in order to compare simulated with measured spectra.The measured spectra were normalized on the basis of the known activity of the source (uncertainty 1%) and the live time of the measurement.As a result, a very good agreement between simulation and measurement was found.
The GRAVEL code was verified by unfolding measured spectra of radioactive sources and by a subsequent analysis of the unfolded spectra to obtain total fluences and doses, which can then be compared with fluence and dose values calculated from the known activities of the sources (by taking into account the geometry of the experiment).The sum of all events in the unfolded spectrum yields the total fluence of the Table 4. Data of photon fields (produced in neutron fields).Calculated fluence rate Φ , air kerma rate Ka, ambient dose equivalent rate Ḣ * (10) and personal dose equivalent rate Ḣp (10), the latter at an angle of 0 • .All values are normalized to the charge on the target Q (using the factor 2 • 10 −6 ) and related to a distance of 1 m between target and CeBr3 detector.The total uncertainty of each value is below 5%, as the data are based on unfolding.

En
Target measured spectrum.Moreover, a total dose can be calculated by multiplying the unfolded spectrum with fluence-to-dose conversion coefficients [14].In all examples studied, the difference between the calculated and measured data was only on the order of a few percent [10].From this investigation and similar measurements in other photon fields, one can conclude that the applied procedures lead to reasonable quantitative results.
Examples demonstrated that even effects that are not explicitly included in the response matrix may nevertheless be visible in the resulting unfolded spectrum (example: broad electron peak in the spectrum of an R-F field [10]).However, structures in unfolded spectra below 1% of the most pronounced photon peaks can generally be neglected since they depend strongly on input parameters.Unfolded spectra are therefore typically plotted with a linear ordinate axis in order to avoid over-interpretation.

Evaluation of photon spectra recorded in the ISO neutron fields
The ISO reference neutron fields listed in table 1 were investigated with respect to the photon and neutron spectra emitted by the targets.The reference fields up to 565 keV are produced by bombarding either a LiF target or a pure Li target with protons.Due to their stability, LiF targets are often used, even though fluorine emits a significant amount of high-energy photons and makes the photon contamination of the neutron field considerable.This disadvantage can be avoided by using a Li target.However, such a target is difficult to handle due to its high chemical reactivity in air.For this reason, the characteristic data for both targets were measured.

Relative photon content of neutron fields
The photon spectra recorded using the CeBr 3 detector are evaluated below.First, the parameters of the photon fields are listed in table 4. Then in table 5, the normalized yield and dose rate of each neutron field are compared with those of the produced prompt photons.The beam charge is chosen as a normalization factor, so the measuring time of the spectra plays no role.The respective photon spectra are depicted in the Annex.Data on the measurands of fluence rate, absolute air kerma rate, absolute ambient dose equivalent rate, and yield are available for the photon field.The neutron field is characterized in terms of fluence rate and yield as well as by the absolute ambient dose equivalent rate.This allows the calculation of the relative amount of photons in each ISO reference neutron field.

Discussion and results
The fluences and dose data shown in table 4 depend on the beam current of the accelerator (typically between 0.6 µA and 1.4 µA) and the target properties, in particular the area density of the target material.The beam current is proportional to the frequency of the beam pulser.The listed values are normalized to a standard distance between target and detector of 1 m and to the measured target charge and they are typical under the measurement conditions at PIAF.The related target area densities are listed in table 1. Rough Table 5. Yield Y per charge, fluence rate and ambient dose equivalent rate Ḣ * (10) of neutrons and photons in comparison.The ratios in the last columns are beam charge corrected.All data are related to a distance of 1 m between target and CeBr3 detector.The total uncertainty of the listed photon or neutron data is 5% and of the ratios 7% (see text).

n field γ field
Ratios in % γ data/n data concerning  estimates of the fluence rates (in cm −2 ) and dose rates (in µGy h −1 or µSv h −1 , respectively) actually prevailing 1 m in front of the target are obtained by multiplying the normalized values by a factor of 10.The uncertainties to be expected under the frame conditions of this work were investigated in detail in [11].Therein, the effect of variations in Monte Carlo simulations and variations of input parameters was used as a method to quantify the uncertainty of the results because there is no analytical access to the uncertainties of the results.The uncertainty budget was calculated according to the guide to uncertainty in measurement [15].As a result, the total uncertainty of integrated fluence (rate) or dose (rate) data was always below 5%, when the analysis was based on unfolding and below 3%, when the conversion method was applied.The total uncertainty of the neutron fluence measurements (using a long counter) was generally below 5%, so that the combined total uncertainty of photon and neutron fluence ratios is always below 7%.A more detailed analysis of single values is not possible because a step-by-step uncertainty analysis of the evaluation of the photon spectra is not possible [11].
The Ḣ * (10) and Ḣp (10,0 • ) values of one particular photon field are very close because these quantities differ only slightly, especially if integrated data of wide spectra are considered.The neutron and photon yields normalized to the accelerator current are found in tables 5 and 6, allowing the data of different neutron fields to be compared.In addition, the Ḣ * (10) and Ḣp (10) values produced by the neutrons during the measurements are displayed normalized to the target charge to allow comparison with corresponding values of the same radiation quality.Still, these normalized values depend on the target properties (including impurities and ageing).The last columns of table 5 and table 6 display the ratio of the normalized gamma dose and neutron dose related to Ḣ * (10) and Ḣp (10), respectively.As noted above, these ratios do not depend very much on either quantity.Relatively high photon doses are produced by LiF targets (figure 6).
Under standard conditions at PIAF, the fields of 8 MeV and 5 MeV neutrons exhibit the highest neutron yields.In contrast, the 5 MeV and 19 MeV neutron fields show the highest photon yields.The most important result is the relative gamma to neutron yield (last two columns of table 5).The neutron fields with a neutron energy of 0.024 MeV (LiF target), 0.250 MeV (LiF target) and 19 MeV have considerable ratios of photon to neutrons yields of between 100% and 200%.The 0.072 MeV, 0.144 MeV, 0.250 MeV and 0.560 MeV neutron fields, all produced by a LiF target, also have a relatively high photon yield >30%.In all other cases, the relative photon yield amounts to just a few percent.With respect to the relative dose rate, the influence of photons is less considerable.The highest value of about 16% is associated with the 0.024 MeV neutron field by a LiF target.All other relative dose rate values of the photons are below 10%.They are even as low as a few per mille in several cases, especially when a pure Li target is applied.From this study, it is recommended to use pure Li targets, though their handling is technically demanding.
The relative photon dose in the 0.144 MeV, 0.250 MeV and 0.560 MeV neutron fields of a LiF target was obtained through simulations by Tanimura et al, which additionally took account of considerable scattering in the not very extensive irradiation room [6].Generally speaking, their values agree with the respective H * (10) ratios in table 5, but differences in detail are seen (they found 6.9 instead of 10.5 for the 250 keV field).Roberts et al measured the relative H * (10) and H p (10) ratios in the 70 keV, 144 keV, 565 keV and 1.2 MeV neutron fields using a LiF target (first three energies) and a Ti target on tantalum [7].The results of the different dosemeters used in the same field differed in part by a factor of 2 or more because the dosemeters were operated in non-linear energy ranges.For the 1.2 MeV field, the authors only published a limit because the dosemeters used were not sensitive enough to obtain quantitative data.The order of magnitude and general tendency were similar to those shown in this paper, but a more detailed comparison Table 7. Angular photon conversion coefficients for ISO neutron fields concerning the quantity Ḣp(10, α).

En
Target Ḣp(10, α) / Ḣp(10, 0 is not possible because of the relative disagreements of the data in [7].In [8], photon/neutron fluence ratios only for the 2.5 MeV, 8 MeV 15 MeV and 19 MeV ISO neutron fields are found.These data should be identical with the respective yield ratios in table 5 (yield and fluence ratios are identical), but the ratios in this publication are smaller by a factor of 10 or even 20, though very small uncertainties are reported.The results in [8] are outdated because they were flawed by different problems.The calculation of total fluences and doses is related to the energy range covered by the measured data.The lower limits of the measurements are slightly different.It is 80 keV for the neutron fields below 144 keV when using a Li target, 90 keV for the 8 MeV neutron field and 120 keV for the other examined neutron fields.The photon spectra measured using the Li target are dominated by a single, very pronounced peak, i.e. the 478 keV peak of the first exited nuclear state of 7 Li (figures 9ff).Other peaks are hardly visible.

Results regarding the angular dependence of the personal dose equivalent
The H p (10) values in table 6 correspond to H p (10,0 • ) values.This quantity simulates the irradiation of a dosemeter (mounted on a slab phantom) under 0 • .To obtain data for a different angle of incidence, α, the quantity Ḣp (10, α) has to be taken into account.After choosing an angle α, the Ḣp (10) values in table 6 need to be multiplied with a coefficient to account for the different Ḣp (10, α) caused by the different geometry.
The first step is to obtain angular conversion coefficients for the quasi-monoenergetic neutron fields and for the related photon spectra.The conversion of the units is done using the photon and neutron conversion coefficients published in the ICRU report 57 [14], which are, with respect to the data used here, identical with the conversion coefficients published in the ICRP publication 74 a few years earlier.Here, the measured and unfolded photon spectra were converted using fits of the conversion coefficients.For the purpose of calibrating personal dosemeters, these need to be placed on the ISO recommended phantom [4], whereas the neutron and photon fluence spectra used for the calculation of the conversion coefficients have to be determined free in air without the presence of the phantom.The angles stated in the tables below refer to the orientation of the calibration phantom with respect to the incoming neutron irradiation.The neutron radiation is regarded as mono-energetic.For the purpose of a conversion of neutron data at the energies needed, the neutron conversion coefficients were also fitted.All fits were done with smooth curves, so the resulting data are physically as exact as possible (avoiding the imprecision of a local interpolation).The obtained angular conversion coefficients, which allow the calculation of the personal dose equivalent (rate) under angles different from 0 • , are listed in table 7 (for the neutron fields) and in table 8 (for the associated photon fields).These coefficients have to be multiplied by the personal dose equivalent rate values in column 4 of table 6 (neutron radiation) and column 6 of table 6 (photon radiation) to obtain dose data valid for the angles of incidence defined in ICRU report 57.
In table 9, coefficients are listed which translate the ratios of the photon to neutron personal dose equivalent rate (column 8 in table 6) under 0 • to ratios under different angles.This table reveals that the ratio of the photon-induced dose to the neutron-induced dose rises sharply when the angle of incidence is increased, especially if the neutron energy is low.Because in some ISO fields the absolute photon dose even Table 8.Angular conversion coefficients for the associated photon spectra of ISO neutron fields concerning of the quantity Ḣp(10, α).

En
Target Ḣp(10, α) / Ḣp(10, 0 Coefficient regarding Ḣp(10, α) dominates at high angles (60 • , even more at 75 • ), problems may arise in practice if a neutron dosemeter under test whose response to neutron radiation is to be investigated is also sensitive to photons.In a 24 keV neutron field produced using a LiF target, the photon to neutron fluence ratio is as much as 1.6 at an angle of 75 • , with the consequence that the absolute photon-induced personal dose is 20% higher than that of the neutrons.

Conclusion and summary
This study examined the photon and neutron fluences of ISO neutron reference fields with neutron energies from 24 keV to 19 MeV.A quantitative investigation of gamma radiation emitted by the targets used for neutron production revealed that the photon yield can indeed be considerable depending on the composition of the target.The TOF method was used to distinguish the prompt photons coming from the target from those produced in secondary processes, such as neutron-induced reactions in or near the detector.The results show that the photon yield of an ISO neutron field can be even greater than the neutron yield.This is the case when 24 keV neutrons are produced in a LiF target and 19 MeV neutrons are produced in a solid-state tritium target, which was realized by a TiT x layer on an Ag backing.The highest energy of most recorded photon spectra is about 8.1 MeV, but higher energies are possible (albeit with a low emission probability), especially in the 15 MeV neutron field.The results prove that Li targets emit much less photon radiation than LiF targets, so that the use of Li targets is preferable to produce low energy neutron fields.
The new data are relevant for future measurements performed in standard ISO neutron fields, not only at PTB.When instruments are tested or calibrated in these fields, the additional photon radiation may have an unwanted influence on neutron detectors or other investigations.The photon spectra, fluences and doses are now known with a high precision and may support the judgement of their influence on any measurements in ISO neutron fields in the covered energy range.
The CeBr 3 spectrometers, operated in neutron fields as a photon detector, demonstrated good performance.This type of scintillation detector has an energy resolution far superior to that of a conventional NaI detector, thus enabling the identification of a number of photo peaks (in particular gamma peaks) generated by the activation of nuclear states.This helps in understanding the nuclear processes taking place in the target.Moreover, this detector is fast and features high detection efficiency even in the MeV range.After exposing it to the direct neutron beam, no disturbing effects caused by long-lasting activation are visible.The isotopes that make up the detector material have sufficiently small cross-sections for nuclear reactions with fast neutrons.The isotopes produced by neutron-induced activation, including 82 Br, have such a long lifetime that they do not disturb the TOF measurements.Their low background is even subtracted by both shadow-cone measurements and random background subtraction of the net spectra.
The photon spectra were evaluated by employing the GRAVEL unfolding method to calculate fluence as well as area and personal dose values from the spectra.In addition, the spectra reveal important information about physical processes in the target and the surroundings since gamma lines at certain energies can be attributed to specific reactions of particular isotopes.Because the measured spectra already show peak structures, the unfolding is simplified since the measured spectrum can be used as an initial estimate when starting the iterative calculation of the fluence spectrum.There is no need to have advance knowledge of the peak structure, as is otherwise often the case.In additon, predefining expected peaks may flaw the results if the expectation is not correct.Besides an independent verification of the results, there is no method to uncover such flaws in unfolded spectra.

Annex
The following pages show the measured net spectra as well as the unfolded photon spectra (figures [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22].This means that spectra recorded behind a shadow cone were subtracted from the respective foreground spectra.Afterwards, the net random spectra were also subtracted.The random background was recorded by defining a coincidence condition between detector events and recording time using a time window, which is positioned before a bunch of photons arrives at the detector.Some expected emission lines are not directly visible because they are superimposed by other lines, a pattern of Compton scattered photons, i.e. the Compton continuum, or other effects.The nuclei with low atomic numbers Z may give rise to high-energy peaks in the photon spectrum, mostly below 9 MeV.However, when the target region consists of elements with a higher Z, the number of nuclear excitation states in the MeV region is so high that single peaks are no longer visible.Isotopes that emit gamma rays because of inelastic neutron scattering are marked in the figures below.Labels are primarily provided for detected photon peaks, but some positions of expected peaks are also marked.The latter are, however, covered by other structures caused by Compton scattering or by other emissions lines of higher or similar intensity.In the figures below, the plotted counts may show values below 1 because the spectra are differences of charge normalized spectra (foreground minus background spectra minus the net random background spectra).
The measured photon spectra that were detected in ISO reference neutron fields show a number of peaks that can be identified (using [12]).The peaks originate from photons emitted by excited nuclear states of the target material or its proximate surroundings, such as the air in front of a gas target.Peaks whose identification may be questionable because of overlapping structures or unclear detection are indicated by a question mark.Because the detection of neutron-induced reactions in the detector is suppressed by the TOF method (for short-lived nuclear states) and by the subtraction of the random background (for longer-lived nuclear states), excited states of detector materials are not visible, nor are the gamma particles emitted due to detector or target activation.Some peaks are emitted by nuclear states of nuclei, small traces of which must have been present in the target or its surroundings.In the case of pair production in the MeV range, electrons can reach any detector in front of the target even at a distance of a few meters if no shielding exists (any such shield should preferably be made of PE).However, a plastic shielding has the disadvantage that it catches a considerable amount of positrons, which annihilate inside the plastic by creating an intense 511 keV photon radiation.The photon fluence thus created is almost proportional to the area of the shielding if this is positioned close to the detector.As a compromise, a cylindrical PE cap with a wall thickness of 25 mm was placed on top of the detector when pair production was expected.
Relevant or identified gamma lines that are present in standard ISO neutron fields are listed in table 10, grouped by target type and neutron field.Only the emissions of nuclear states below 100 ns are listed since the photons emitted by the longer living states are not visible due to the subtraction of the random background.The nuclear reactions that produce gamma emissions relevant in this context are as follows: • Inelastic proton scattering: A X(p, p ′ ), or deuteron scattering: A X(d, d ′ ), where X is the element symbol and A the atomic number; a nuclear state is excited by a neutron and emits a photon when it goes back to the ground level.• Inelastic neutron scattering: A X(n, n ′ ), where X is the element symbol and A the atomic number; a nuclear state is excited by a neutron and emits a photon when it goes back to the ground level.Because this is a secondary process with a comparably low cross-section, it is an unlikely process.• Neutron capture or deuteron breakup: A X(n, γ) A+1 X or A X(d, p) A+1 X.After a neutron is caught by a nucleus, the new isotope that is produced de-excites by emitting one or several photons.• Proton capture: The neutron fields are generated by the proton capturing reactions A X(p, n) A Y or A X(d, n) A Y, both increasing the atomic number of the target nucleus by 1.Here, the nuclides 7 Li, D or T are used as a target (table 1).In the tables below a proton induced reaction only happens in the α-particle emitting reaction 19 F(p, α) 16 O, which occurs when a LiF target is used.• Internal pair production: 16 O * → 16 O + e − e + .Because the reaction mentioned before produces 16 O in an excited state with the considerable energy of about 6 MeV, and because a gamma transition to the ground state is forbidden because of identical spin and parity of both states, the de-excitation to the ground happens by electron/ positron pair emission.The energy distribution of the electrons peaks at 2.2 MeV, but rises to 4 MeV, when 4 MeV protons are shot onto the target [17].

Figure 1 .
Figure 1.Time structure of the pulsed accelerator proton beam.The normalized count rate of the CeBr3 detector is depicted as a function of time.In this example, the distance between the target and the detector is 150 cm.The curves representing the integrated number of counts of one run are normalized to the total beam charge in 500 µC.The actual count number was higher by about a factor of two.Accelerator pulse frequency: 1.25 MHz.

Figure 2 .
Figure 2. Random background (top) and spectrum of the prompt photons of the 15 MeV neutron field without random background subtraction (bottom).Long-lived excited states of the CeBr3 crystal are shown here.Both diagrams are based on the same measurements with a total live time of 6000 s.

Figure 3 .
Figure 3. Two-dimensional diagram of the spectral photon distributions over time using a false-colour plot to depict the bin content.This plot was recorded in a 15 MeV neutron field.

Figure 4 .
Figure 4. Measured photon spectrum in a 250 keV neutron field by using a CeBr3 detector (crystal size: 1.5 ′′ × 1.5 ′′ ) in dark blue.The result of unfolding (after seven iterations) is displayed in red.The reconstructed spectrum using the response matrix (used by GRAVEL to adapt and check the unfolded spectrum[10]) is plotted in light blue.

Figure 5 .
Figure 5. Number of detected 197 keV photons with width of the TOF window obtained through the evaluation of one example spectrum.

Figure 6 .
Figure 6.Comparisons of the neutron dose and the photon dose measured in standard ISO neutron fields.The diagram displays Ḣ * (10) data.All values are normalized to a distance of 1 m between target and CeBr3 detector and to the beam charge.

Figure 12 .
Figure 12.Measured (top and middle) and unfolded photon spectrum (bottom).Here and in the following, high-energy photon peaks and their referring escape peaks (E1 & E2) are labelled in the same colour.

Table 1 .
List of neutron beam qualities and information on parameters that govern the beam production.
a Thickness of all backing plates: 0.5 mm.b Target only usable once (because of chemical instability).

Table 3 .
Mean random photon fluences and dose rates divided by net fluences and dose rates of the spectra of the prompt photons.The ratios are expressed in percent.The uncertainties are stated relative to these values.

Table 6 .
(10)nce rate and personal dose equivalent rate Ḣp(10)in comparison.The ratios in the last columns are beam charge corrected.All data are related to a distance of 1 m between target and CeBr3 detector.The values related to the yield are identical with those in the table above.They are repeated for a better comparison with the Ḣp(10) data.The total uncertainties are given in the same way as in table5.

Table 9 .
Angular conversion coefficients to be applied to the personal dose equivalent ratios in table 6.