Photo-Neutron Source by High Energy Electrons on Target: Comparison between Monte Carlo Predicitons and Experimental Measurements

Transcription

1 Photo-Neutron Source by High Energy Electrons on Target: Comparison between Monte Carlo Predicitons and Experimental Measurements L. Quintieri, R. Bedogni, B. Buonomo, M. De Giorgi, A. Esposito, G. Mazzitelli, P. Valente and J. M.Gomez-Ros Abstract A photo-neutron source has been realized at the DaΦne Beam Test Facility: neutrons are produced by sending high energy electrons (510 MeV) to impinge on an optimized target. Neutron flux and spectra have been measured along well designed extraction lines. The experimental data have been compared with Monte Carlo predictions performed with FLUKA and MCNPX, respectively. The comparison between measured neutron rates, energy spectra and the correspective Monte Carlo expectations is presented, as well as the simulation results from the two different Monte Carlo codes are compared and analyzed. I. NEUTRONS AT THE DAΦNE BEAM TEST FACILITY: THE N@BTF PROJECT The DAΦNE Beam Test Facility (BTF)[1] is an electron/positron transfer line by which the beam accelerated from the Linac is modulated in intensity and energy and is transported in an experimental hall (100 m 2 area), where beam and experimental tests can be carried out. The facility can provide e-/(e+) with energy ranging from from 25 MeV up to 750(550) MeV, in a wide range of intensity: from single particle per bunch up to particles per pulse. The main applications of the facility are: high energy detector calibration, low energy calorimetry, low energy electromagnetic interaction studies, detector efficiency measurements, tests of beam diagnostic devices, and so on. The main idea of the project for neutron production at the BTF (n@btf) consists in sending high energy electrons to impinge on a suitable high Z target located in the BTF experimental hall, at the end of the transfer line. The interacting electrons produce, inside the target, a photon cascade shower that has energy spectrum end point equals to the maximum electron beam energy. The produced photons can be absorbed by the nuclei of the target, that in this way are excited. These nuclei decay into the fundamental state by boiling off neutrons, mainly according the well known mechanism of the Giant Resonance, even if also neutrons of higher energy are expected due to the Quasi Deuteron Resonance mechanism[2]. Manuscript received November 20, L. Quintieri, R. Bedogni, B. Buonomo, M. De Giorgi, A. Esposito, G. Mazzitelli are with the National Institute of Nuclear Physics, Laboratori Nazionali di Frascati, Frascati Italy (corresponding author lina.quintieri@lnf.infn.it). P. Valente is with La Sapienza University, P.le Aldo Moro, Roma, Italty. J.M.Gomez-Ros is with CIEMAT, Av. Complutense, 22 E-28040, Madrid Spain. II. DESIGN OF A PHOTONEUTRON SOURCE BY MONTE CARLO CODES The photonuclear physic is well known in accelerator field, mainly because related to shielding issues for neutrons, that are produced as a consequence of electron and photon interaction with matter (mainly in dumps and shields). The most effective way to shield electromagnetic radiation is to use high atomic number (Z) materials (Lead, Tungsten and so on). On the other side, high Z nuclei exhibit higher cross sections for photonuclear reactions, that means a related production of neutrons and protons that have to be properly taken into account for dose calculation and so far in the shielding design. At the Beam Test Facility of the DaΦe collider, we have designed and realized a photoneutron source, by exposing an heavy nuclei target to a high energy electron beam. In this way, what is normally an issue from the radioprotection point of view, has been successfully exploited as main source for our task. More than 80% of electrons loose energy by bremsstrahlung: the ensuing photon shower interacts with the nuclei exciting them (essentially on the base of Giant Diant Dipole Resonance mechanism for photon energy lower than 30 MeV and on the base of Quasi Deuteron Mechanism for higher photon energy). In going back to the fundamental state, the nuclei boil-off a nucleon, typically a neutron (in fact, the emission of protons, that is a possible channel, is strongly repressed in heavy nuclei due to the Colombian barrier). In the design of the experimental set-up for the photoneutron source, we used Monte Carlo codes in order to: Choose accurately the material and the geometry that could maximize the exiting neutron yield with respect to the characteristics of the electron beam (mainly energy spectrum, spot size, spatial distribution) we can dispose of. Know the spatial distribution of the emitted neutrons. In fact, even if, we could use semi-empirical correlation available in literature 1, to have a fast but rough evaluation of the neutron yield versus the thickness of the target, we must use Monte Carlo codes, in order to properly estimate the neutron spatial distribution as well as the energy spectrum of the leaving neutrons (mainly in relation to the fact that the final apparatus has a more complex geometry 1 It is worthwhile to mention the huge work done by Swanson[2], in this field /10/$ IEEE 915

2 than a simple target and different materials are actually involved in the shielding around the extraction lines). Concerning the last item, we have to point out that the Monte Carlo predictions of the spatial distribution has determined important constraints in the design of the experimental set-up of the photoneutron source. In fact, while neutrons are emitted quite well isotropically around the target, photons are mainly peaked along the primary electron direction. As reported in the table I, that summarizes the Monte Carlo simulation results (by FLUKA code), going from a polar angle of 0, with respect to the electron impinging direction, to π/2, the photon flux is reduced by two orders of magnitude, whereas the neutron flux is almost unchanged. Moreover, the Monte Carlo simulations show that the harder component of photon flux is collimated in forward direction, along the electron beam. Polar Angle Photon Flux Neutron Flux [deg] [ph/cm 2 /pr] [n/cm 2 /pr] 0 o E-02 +/ % 5.782E-06 +/ % -30 o E-04 +/ % 7.325E-06 +/ % 30 o E-03 +/ % 6.987E-06 +/ % -45 o E-04 +/ % 6.731E-06 +/ % 45 o E-04 +/ % 6.373E-06 +/ % -60 o E-04 +/ % 5.841E-06 +/ % 60 o E-04 +/ % 5.353E-06 +/ % 90 o E-05 +/ % 4.379E-06 +/ % TABLE I NEUTRON AND PHOTON FLUX AROUND THE TARGET. POLAR ANGLE IS RESPECT TO THE ELECTRON IMPINGING DIRECTION Fig. 1. Experimental set-up (for feasibility test) Because the photoneutron source is located in the BTF experimental hall, at the end of the transfer line, we cannot allow the produced neutrons be scattered everywhere, but we have to shield the whole solid angle around the target, with the exception of well defined extraction lines, along which we should be able to properly collect neutrons. The shield is a multilayered structure made of lead and polyethylene (Boron Carbide sheets are also foreseen to be used in next future). The complete details of the shield design are reported in [3]. In figure 1 the experimental set-up, installed in May 2010, is shown, while in figure 2 the target inside the shield is visible. Thanks to the detailed map of particle distribution around the target, obtained with the Monte Carlo simulations, we defined two extraction lines (see figure 3) in a plane perpendicular to the electron beam direction. In this way we defined the best configuration of the lines along which collecting the signal, in order to maximize neutron to photon ratio. Fig. 2. Target located inside the shield apparatus III. PHOTONUCLEAR PHYSICS IN FLUKA AND MCNPX To estimate the neutron rates and energy spectra, along the extraction lines and all around the shield, we used both FLUKA (20083d release)[4] and MCNPX[5] (MCNPX2.5.0) codes. Simulations with GEANT4[6] are in progress and will be soon available for comparison. In fact, one of the main goals of n@btf is also to compare the simulation results of photonuclear physics from the different cited codes. Fig. 3. Extraction lines (with lead caps) are designed in the plane perpendicular respect to the impinging electron direction 916

3 We have chosen to use, as starting point, the FLUKA code because, in addition to the fact that it implements the photonuclear physics, since long time, on the whole energy range, it also offers a rich data base for the total photoneutron cross sections of 190 nuclides[7]. A more complete description of the implementation of photonuclear physics in FLUKA code can be found in [3]. Concerning MCNPX, the calculations relied on the latest ENDF/B-VII photonuclear data library, available, at the date in which the work has been done, on the website The photonuclear physics implementation in MCNPX had an important upgrade in 2005, when new physics packages for photons with energies from 5 MeV to about 2 GeV have been added. These packages include the models of Giant Dipole Resonance and Quasi Deuteron Resonance. These two models compliment the existing photonuclear data, on the base of the mix and match capability, according to which table data are used when available, otherwise physics models are applied. The mixing feature is accomplished with the MX card[8]. The threshold for photonuclear reactions in MCNPX is 1 MeV. A. Computational details for MCNPX simulation The final experimental set-up has been accurately reproduced in the MCNPX model, as well in the FLUKA simulations. Computational details about the simulations performed by FLUKA can be found in [3]. In our MCNPX model, the S(α,β) data have been used for the treatment of the thermal scattering in polyethylene. The photonuclear libraries for Tungsten are loaded by using the option pnlib=70u in the material definition card. As biasing technique we used the BBREM card to make more efficient the bremsstrahlung photon production inside the target. In fact, as documented in the MCNPX manual, the bremsstrahlung process generates many low-energy photons, but the higher-energy photons are often of more interest. One way to generate more high-energy photon tracks is to bias each sampling of a bremsstrahlung photon toward a larger fraction of the available electron energy. Moreover we used techniques of variance reduction based on assigning relative importance to different cells. Finally, concerning the post-processing, we used both F4 tally (average flux over a cell, i.e the Bonner sphere volume) and F5 tally (flux at a point, i.e. at the center of the Bonner sphere) to estimate the neutron flux and the energy spectrum at the point of interest. B. Optimization of the target In table II we compared the semi-empirical neutron rates provided by Swanson[9] with the correspective neutron rates calculated by Monte Carlo code (FLUKA), that could be obtained with 500 MeV electrons on thick targets of different materials: Tungsten, Tantalum and Lead, respectively. As we can see, Tungsten and Tantalum offer a much higher neutron yield with respect to Lead. For all the examined materials the spectrum is pretty well a Maxwellian, peaked around 1 MeV (with little difference, according to which Material Swanson FLUKA nyield [n/s] [n/kw/s] E+12 [n/kw/s] BTF Tantalum E+10 Lead E+10 Tungsten E+11 TABLE II COMPARISON BETWEEN PREDICTED AND EMPIRICAL NEUTRON YIELD FOR DIFFERENT HEAVY MATERIALS the maximum for Lead is a little bit shifted toward higher energies). So that the final choice has fallen onto Tungsten, that, in addition, offers a better thermal diffusivity respect to Tantalum. We have chosen to use a cylindrical shape for the target, whose dimensions have been optimized by several Monte Carlo calculations for fine tuning. In all the simulations, the primary beam has been supposed to be along the cylinder axis. In this way, we have also validated the FLUKA predictions of the source term (neutron yield per primary) with respect to the well known and widely used semi-empirical correlations of Swanson: as table II shows, there is always an agreement better than 10% between predictions and experimental values. This makes us confident in using properly FLUKA code to estimate the photoneutron source term. Finally, the dimension of the optimized target, made of Tungsten, have been determined: 35 mm of Radius and 60 mm of Length. The estimated neutron spectrum obtainable from this optimized target is reported in figure 4. Up to 100 MeV the spectrum is described as a Maxwellian distribution, with average around 0.7 MeV and FWHM of about 0.9 MeV. Approaching the higher energies the Quasi-Deuteron effect adds a tail of high-energy neutrons to the Giant resonance spectrum. The slope becomes stepper as the incident electron energy is approached. IV. FINAL EXPERIMENTAL SET-UP: THE FEASIBILITY TEST The experimental set-up shown in figure 1 has been accurately simulated by FLUKA and MCNPX code. The values estimated by FLUKA code of the neutron flux in different positions of the final experimental set-up are summarized in figure 5: in the shown table, values normalized per primary are reported in the first column, while the expected fluxes, corresponding to the maximum electron beam, that at present time can be delivered in the experimental BTF hall (about 5E+11 e/s), are reported in the second column. Estimations of the neutron rates, as well as of the neutron energy spectra at different positions around the target, have been compared with the experimental data. Measurements have been done by using Bonner Spheres [10] neutron detectors. The complete set of Bonner Spheres (LNF-ERBSS) used for the measurements is shown in figure 6. It includes: 8 polyethylene spheres (density 0.95 g/cm3) 3 polyethylene spheres (density 0.95 g/cm3) loaded with copper and lead a 4x4 6 LiI(Eu) active scintillators 917

4 2010 IEEE Nuclear Science Symposium Conference Record N31-6 EdF/dE neutron spectrum Double Log Spectrum n/pr/cm2 1e-05 Fig. 7. ERBSS Response Function 0.5 FLUKA Dy-BSS 0.4 1e-09 MCNPX 1e Energy (GeV) E df/de Fig Letargic Energy Spectrum of neutron exiting the target n@btf Flux[n/cm2/pr] (on all solid angle and spectrum) FLUX n@btf n/cm2/s (all spectrum) exiting the target=ϕ1 ( A ) 1.80E E+08 entering the shield=ϕ2 ( B ) 4.10E E+08 leaving the shield=ϕ3 ( C ) 4.90E E+07 at 1m from shield=ϕ4 ( D ) 5.99E E e-09 1e-05 1e-04 1e-03 1e-02 1e e+01 1e+02 Neutron Energy [MeV] Fig. 8. Experimental and MC computed neutron spectra, 150 cm apart from the target at 90o wr to the impinging electron beam Fig. 5. MC estimated neutron rates. First column: values per primary; second column: values obtainable in BTF when 5E+11 e/s are sent onto the target Monday, 10 May 2010 Each sphere hosts in the center a detector for thermal neutrons, that can be active or passive. In our feasibility test we used a Disprosium activation foil, that appears to be more suitable for working in a high photon background. The response functions of the ERBSS were calculated with MCNPX code and are shown in figure 7. The response matrix of the ERBSS was validated in reference neutron fields and its overall uncertainty was estimated to be +/-3%. A special unfolding program, FRUIT[11], has been used in order to reconstruct the neutron final spectra from the raw data of each sphere. The measurement campaign took about 15 days and ended by 7th May V. F INAL TEST: C OMPARISON BETWEEN M ONTE C ARLO P REDICTIONS AND E XPERIMENTAL DATA Fig. 6. Complete set of Bonner Spheres used for measurements In figure 8 the lethargic (EdΦ/dE) spectra, normalized to the total neutron flux, estimated by Monte Carlo codes (continuous line for FLUKA and dashed for MCNPX) are shown together with the normalized lethargic spectrum (red dots), obtained with Bonner Spheres. As we can see, there is a good agreement between measurements and simulation results concerning the shape of the 918

5 neutron spectrum: statistical tests have been performed in order to quantify properly the accordance of Monte Carlo predictions with experimental values. In particular we used the χ 2 Goodness of Fit test in order to asses quantitatively the Monte Carlo accuracy around the Giant Dipole Resonance. The p-values that we found, respectively for MCNPX and FLUKA, are and This allows to conclude that MCNPX and FLUKA provide an accurate reconstruction of the experimental resonance, both in energy position and amplitude. As predicted for, the majority of produced neutrons belongs to the energy range from 10 KeV to 20 MeV: the calculated neutron spectrum has a Maxwellian shape with average around 0.7 MeV. The experimental and calculated neutron fluxes, collected over a sphere of 10 cm, whose center is 1.5 m far away from the target along one of the two extraction lines, are reported in Table III, confirming again the good agreement between experimental and predicted values. REFERENCES [1] B. Buonomo, G. Mazzitelli, F. Murtas, L. Quintieri A wide range electrons, photons, neutrons beam facility, Proceedings EPAC-2008,June , Genoa,Italy [2] W.P. Swanson Calculation of Neutron Yields Released by Electrons near the Photoneutron Threshold(*), SLAC-PUB-2211, October [3] L.Quintieri et al., Feasibility Study of a Neutron Source at the DaΦne Beam Test Facility Using Monte Carlo Codes, Nuclear Science Symposium Record, 2009 IEEE, Orlando, Oct. 2009; Pages: [4] A. Fasso et al. A multi-particle transport code, CERN (2005), INFN/ TC-05/11. [5] [6] [7] A. Fassò, A. Ferrari, P.R. Sala, Photonuclear Reactions in FLUKA: Cross Sections and Interaction Models, Presented at the International Conference on Nuclear Data for Science and Technology, Santa Fe, NM, USA, 26 Sep - 1 Oct 2004 In: AIP Conf. Proc. 769 (2005) pp [8] J. S. Hendricks et al, MCNPX ExtensionVersion 2.5.0, LA-UR [9] W.P. Swanson Calculation of Neutron Yields Released by Electrons Incident on Selected Materials, SLAC-PUB-2042, November 1977 [10] A. Esposito and M. Nandy, Measurement and unfolding of neutron spectra using Bonner Spheres, Radiation Protection Dosimetry, (1-4): (doi: /rpd/nch385) [11] A. Esposito et al, Instruments and Methods in Physics Research A 580 (2007) Measurements [BSS] MCNPX Fluka 1/[cm2/pr] 1/[cm2/pr] 1/[cm2/pr] 8.04E-7±3%. 8.06E-7± 4% 8.12E-7±5% TABLE III TOTAL NEUTRON FLUX PER PRIMARY: COMPARISON BETWEEN EXPERIMENTAL MEASUREMENTS AND PREDICTIONS We can conclude that at present, because we are authorized to deposit only a small fraction of the total Linac power (40-50 W), we are able to have, at 1.5 m from the target, a maximum neutron rate of 4E+5 n/cm2/s, that mainly ( 80% ) is made of neutrons with energy around 1 MeV, as correctly predicted by Monte Carlo simulations. VI. CONCLUSION AND OUTLOOK We have successfully accomplished the first main step of the n@btf project: the feasibility of a neutron source at the DaΦne Beam test Facility has been demonstrated. According the preliminary analysis, neutrons are produced in good accordance with the expectations. As predicted by Monte Carlo simulations, neutrons are measured over 10 decades in energy (from ev up to about 160 MeV), even if more than 80% is found around the Giant resonance. A very good agreement between measurements and simulations has been found also for the shape of normalized spectrum, as confirmed by statistical tests. The measurement of the total neutron flux per primary coincides with the predicted values, within the simulation uncertainties. Calculations with Geant4 are also in progress and results should be soon available for benchmarking. Differences between all the cited Monte Carlo codes are going to be deeply and accurately investigated and quantified with respect to the experimental values, since this is one of main objectives of n@btf project. 919