Abstract
The accelerator-driven thermal neutron source (TNS) at SARAF is under the advanced design stage, to provide an accelerator-based substitution to existing neutron imaging and diffraction utilities of the IRR1 aging research reactor. A challenging task of the system design is to provide a safe and maintainable coupling of 40 MeV deuteron beam and a 200 kW power liquid metal target with a compact light-water moderator and reflector assembly. The verification process of the mechanical design is performed through Monte-Carlo simulations over the detailed CAD models. We outline the method of verification of neutron production and safety requirements over the detailed system design and highlight the advantages of this approach.
Introduction
SARAF (Soreq Applied Research Accelerator Facility) is a 40 MeV and 5 mA d/p LINAC under construction in Israel [7]. The thermal neutron source (TNS) is a moderated target station planned at SARAF as a replacement for neutron-imaging and diffraction systems of the IRR1 aging research reactor. The source is driven by a continuous 200 kW beam impinging a liquid metal jet of gallium-indium eutectic yielding
The system is entering a detailed design phase now, and the project is working on realistic 3D mechanical models that are geared towards production. The mechanical design of the system, shown in Fig. 1, includes fine details and material specification for the many different system components.
Since this is a high radiation target station, issues of target handling, maintenance and radiation protection must be considered at an early stage. As a result, it is required to verify that, firstly, the proposed mechanical design could provide the required neutron fluxes to all instruments and secondly that the radiation protection design is valid.
Artist view of SARAF phase II experimental Hall showing the location of the TNS system (left) horizontal cross section of the current mechanical design of the TNS system (right).
The validation tools are Monte Carlo (MC) simulations using CERN FLUKA version 4.2 [1] and MCNP version 4B [2]. Typically, geometry and materials for MC simulation codes are inserted manually using simple logic of geometrical shapes. However, the complexity of the mechanical design requires an automatic and direct process to port a complete CAD model to MC geometry. Our tools of choice are ANSYS SpaceClaim and DesignModeler [3] that are used to simplify and remove overlaps in the geometry. Coreform cubit [4] that is used for assignment of materials and the export to mesh format DAGMC (Direct accelerated geometry Monte-Carlo) [5], an open-source package that serves as a bridge between the CAD model and MC codes. Finally, we make an extensive use of bash/python scripting to link all the components into a unified workflow.
Cross section cuts of the simulation geometry showing the different components of (starting from bottom right, clockwise) the biologic shielding (the 6 m diameter “igloo”) and the n-imaging Hall, zoom in to the inner part of the igloo, the target-reflector-moderator assembly, side view of the target chamber and water tank of the reflector, side view of the entire igloo including service ports at its ceiling.
The resulted geometry is shown from different angles in Fig. 2 and contains an accurate interpretation of the original CAD model including most, if not all the relevant volume cells. DAGMC provides coupling to several different MC codes. Our tool of choice is CERN FLUKA, which has an excellent prompt and delayed radiation protection design tools. The calculations with FLUKA are fully analog simulations, performed using high-power computing cluster, and generally takes days to a week to complete. Another tool we are using is a direct conversion of the CAD geometry to an MCNP input, provided by ANSYS DesignModeler. This tool is limited to relatively simple geometries, and we use it only to export the water/beryllium part of the moderator-reflector assembly. Having an MCNP input allows us to test the sensitivity of the overall design to parametric changes of geometry and materials. For this purpose, we make an extensive use of variance reduction techniques available in MCNP, in particularly the F5 tally (next event estimator) that provides the thermal neutron brightness at the end position of each instrument in a short run time of no more than a few hours, on a single CPU.
Version 4B of MCNP that we are using does not yet have the meshtal option built in, therefore neutron flux maps were produced through smart usage of SD card and definition of separate universe at the tallied cross section area. With FLUKA, maps are produced with USRBIN cards and custom subroutine that scores over a specific energy range of choice. Thermal neutron maps provide a useful knowledge about the system performance, since the flux at the vicinity of the entrance irradiation port directly determines the available flux at the instrument, by correcting with a simply-calculated geometrical attenuation term [8].
We use these tools for verification of the design and to provide important information for the mechanical designers about different aspects of neutronic properties and radiation protection. In the following section we demonstrate the usefulness of this approach, highlighting problems we managed to address using these methods.
Robust operation is important for high-radiation system that is difficult to maintain and upgrade, that can be addressed using parametric sensitivity tests. The tests are preformed using MCNP and generally done by changing one parameter at a time. Figure 3 shows an example of a sensitivity test, where we scanned a range of parametric changes to the beryllium reflector vertical shape above and below the moderator location. The effect of the geometrical change is monitored through thermal neutron maps and neutron flux tally placed at the exit position of each instrument.
Sensitivity test using MCNP with parametric scanning of the insertion of beryllium extensions under and above the moderator location (upper left plot) the resulting thermal neutron maps (bottom right) and the relative gain in thermal flux with F5 tally (left) is showing a slight increase of the thermal flux for beryllium extension of up to 10 cm. More than 10 cm will result in a significant flux drop due to the better properties of light water as moderation substrate.
Figure 4 shows the thermal neutron flux maps that are produced at the center of the system, as calculated by MCNP and FLUKA. Notice that the MCNP model contains partial geometry with only the main system components are modeled, compared to the fully detailed geometry in the case of FLUKA model. Yet, the location and spread of the thermal neutron hot-spot, at the center of the system, is noticeably similar, except the distribution of thermal neutrons in FLUKA calculation is 15% more intense. The reason for that might be the additional reflection of fast neutrons from the surrounding that is modeled in the FLUKA case and missing with the simplified model used by MCNP.
Thermal neutron flux maps at the center of the system showing the hot-spot of thermal neutrons at the center of the light-water moderator and calculated using MCNP (left) and FLUKA (right).
Deuteron-induced neutron sources are complicated to model and in the case of our system, to our knowledge, the double-differential neutron yield of 40 MeV deuterons impinging a gallium-indium target has not yet measured. For our calculations, we typically make use of an evaluated neutron spectrum that was generated based on partial cross section measurements, predicting a total yield of
Thermal (top) and fast (bottom) neutron maps with FLUKA starting from a deuteron source (left) and a fast neutron source definition. The thermal distribution match in both cases to better than 5%, while the fast distribution is evidently very different.
The new version of CERN FLUKA (4.2) has, for the first time, the ability to use deuteron reaction directly in simulations, based on models of deuteron stripping and break-up reactions [6], allowing cross-checks of our evaluated neutron source with FLUKA prediction. Figure 5 shows the thermal and fast neutron flux maps at the center of the system, starting from the source definition of deuterons (left hand side plots) or fast neutrons (right hand side plots). While the thermal neutron flux maps show a high match of better than 5%, the fast neutron maps differ by more than 200% at the moderator center location. The reason for the difference in fast neutron distributions is that the definition of both sources is fundamentally different, the neutron source is approximated as isotropic emission, while the neutron source generated by FLUKA has a forward tendency that is a result of the high momentum transfer from the 40 MeV beam. Nevertheless, this is a nice evidence for the robustness of the physical system design to the definition of the fast neutron source itself, as the thermal neutron distribution is a property that is mostly determined from the moderator material characteristics due to the many thousands of scattering events that are required to bring down the energy from fast to thermal energies.
According to radiation safety design, the dose rate level at working areas outside of the biological shielding is required to be lower than 2 uSv/hr to allow free access to the infrastructures that surrounds the system. The intense neutron source is therefore surrounded by 240 cm thick magnetite-concrete walls that attenuates the neutron and gamma dose rates to this level. Figure 6 shows an example of such calculation of the prompt neutron dose rate field using the highly detailed model and FLUKA simulations. High doses are recorded at the n-imaging hall, where the neutron beam emerging through the designated collimator and scattered at the beam-dump at the end of that line. This calculation is a nice example of the usefulness of this unified workflow, as it allowed us to reveal design flaws that could not be detected earlier. It revealed a flaw in the design of the collimator that allowed a direct line-of-sight to the source, effectively increasing the dose inside the room by an order of magnitude. Another obvious flaw of the beamstop design being too narrow, is evident from the scattered neutrons circumventing the beamstop and hitting the back wall, thus resulting with high doses outside. Following this work, those flaws were fixed by changing geometry and adding shielding at the appropriate places.
Neutron dose field surrounding the system and at the n-imaging Hall (right), the calculated distribution of fast neutrons at the position of n-imaging instrument showing a hot-spot of neutrons resulting from a design flaw that allowed a direct line-of-sight to the fast neutron source (bottom left) as simulated using FLUKA. The design flaw was fixed by geometry changes to the collimation system shown as a side view of the system in the upper left plot.
The same method is used for neutron activation studies that is of high importance to a high radiation system that has to be maintained remotely and be properly planned in advance. Figure 7 shows the evolution of the activation dose (in units of uSv/hr for 5 mA current) at the center of system following consequent cooling times. Shortly after the end of bombardment, aluminum (6061 alloy) components are highlighted, as short-lived isotopes of aluminum activation products are most significant. After 4 hours, and as the short half-life isotopes have mostly decayed, the high iron content magnetite concrete walls are highlighted to show the highest contribution to the surrounding dose. One week later, most of the remaining activation is located at the vicinity of the neutron target chamber, made from stainless-steel (316LN alloy), where longer half-life isotopes are generally produced from neutron reactions. This information is useful for the maintenance design of different parts that require timely attention or replacement.
Evolution of the gamma dose rate distribution at the center of the system, after 500 hours of full power operation and different cooling times using FLUKA calculations.
Another useful feature of the CAD based simulation approach is the ability to share the calculation grid with other mechanical design tools. We use this method for the calculation of heating power in the system that is resulted by neutron capture and gamma scattering reactions that deposit significant energy throughout the system. The heat power distribution shown in Fig. 8 is calculated using FLUKA using the full and detailed model and includes both neutron and gamma driven power. The energy deposition is not uniform and most of the power is deposited in the region of moderating water in vicinity to the neutron producing target. The total power that has to be evacuated from the reflector-moderator assembly is ∼600 Watt when running in full-power. The resulted power distribution could now serve as input for further calculations of heat removal strategies using tools such as ANSYS that operate on the same exact CAD grid.
Cross cut of deposited energy distribution by neutrons and gammas across the system, highlighting areas where additional cooling might be required.
We discussed the benefit of using advanced tools for verification of mechanical design using Monte-Carlo simulation and the particular workflow that we established for the design process of the continuous thermal neutron source of SARAF. The unified approach allows better precision planning of the neutron/gamma flux balance, radiation shielding, heat removal and maintenance that is vital for a high radiation system at this advanced stage of design. It has already proved very useful in many cases, in the current system design, and has leveled up our ability to detect design flaws early on. This method is used extensively and exclusively now, also for other systems at SARAF, as we move forward in the design and construction of the neutron research complex in the next coming years.