Arkady Serikov, Daniil Koliadko, Roman Afanasenko, Yuefeng Qiu, Dieter Leichtle Karlsruhe Institute of Technology (KIT), Institute for Neutron Physics and Reactor Technology, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany 5 th Fusion HPC Workshop, VC November 21-22, 2024 Computational neutronics analyses of deuteron interactions with lithium target in IFMIF-DONES for fusion applications
Part I: Basic processes defined for (d-Li) atomic and nuclear interactions. Illustration on a simple model of cylindrical solid Li inside the aluminum capsule. Part II: Application of the d-Li accelerator-based intense neutron source of IFMIF-DONES for fusion applications. Conclusions Backup slides: McDeLicious code parallelization on Marconi-Fusion HPC Use of On-The-Fly (OTF) Monte Carlo variance reduction technique
MCNP is a code for radiation transport calculations in 3D geometry. The abbreviation is translated as M onte C arlo N - P article. Neutron, photon, electron, or coupled neutron/photon/electron transport can be performed by MCNP. The MCNP code was developed by X-5 Monte Carlo Team in Los Alamos National Lab. (LANL), USA. Reference : [1] Avneet Sood, 2017. The Monte Carlo Method and MCNP-A Brief Review of Our 40 Year History, Presentation to the International Topical Meeting on Industrial Radiation and Radioisotope Measurement Applications Conference. History of the MCNP code development [1] Radiation transport with the MCNP code Contributors to MCNP6.2
McDeLicious code development history • 1999: McDeLi ( P. Wilson, Report FZKA 6218, 1999): - An enhancement to MCNP-4a to sample the generation of d-Li source neutrons based on embedded analytical formulas representing direct deuteron striping ( Serber model) and compound reactions. • 2001: McDeLicious ( S.P.Simakov et al. J.Nucl.Mat.307-311(2002)1710, FZKA 6743) - An enhancement to MCNP-4b,c to sample the d-Li source neutrons on the basis of tabulated double-differential d + 6,7Li cross-sections for deuteron energies up to 50 MeV (evaluated by A. Konobeyev et al., NSE 139 (2001)1). • 2005: McDeLicious-05 – compilation with MCNP-5 and use tabulated double-differential cross-sections from updated d + 6,7Li evaluation (made by P. Pereslavtsev et al., J.Nucl.Mat.367-370(2007)1531). • 2011: McDeLicious-11 - a new approach is implemented to enable direct sampling from the tabulated deuteron beam distribution data without using fitting functions. In this approach, the beam entry position is sampled from tabulated data representing the intensity distribution of the impinging deuteron beam – ( S. P. Simakov et al., “Status of the McDeLicious approach for the D-Li neutron source term modeling in IFMIF neutronics calculations,” Fusion Sci. Technol., 62 (2012), pp. 233-239 ) • 2017: McDeLicious-17 – the actual version of McDeLicious upgraded to MCNP version 6.1.0, an extension of the MCNP Monte Carlo code with the capability to simulate the deuterium-lithium neutron source on the basis of evaluated d + 6,7Li cross-section data. This code has been tested and confirmed to generate identical source particle data as the previous version McDeLicious-11 – (Y. Qiu et al., “IFMIF-DONES HFTM neutronics modeling and nuclear response analyses,” Nuclear Materials and Energy, 15 (2018), pp. 185-189) McDeLicious is an extension to the MCNP Monte Carlo code with the ability to simulate the generation of source neutrons based on deuteron - lithium ( D-Li) interaction processes
Part I Basic processes defined for (d-Li) atomic and nuclear interactions. Illustration on a simple model of cylindrical solid Li inside the aluminum capsule
Ref: S.P. Simakov, McDelicious Workshop, FZK/IRS, Institut für Reaktorsicherheit, Forschungszentrum Karlsruhe, 13-14 March 2008
MCNP model Simple Model: cylindrical solid Li inside Aluminum capsule Disk D = 3 mm D+ source Ed = 40 MeV Id = 1 microA X V1: Isotropic u niformly distributed disk D+ source Disk D = 3 mm D+ source Ed = 40 MeV Id = 1 microA X-Z cut N flux N flux
0.3cm V1: Isotropic u niformly distributed disk D+ source D+ flux Total deuteron flux, d/cm2/s
Total deuteron flux, d/cm2/s D+ flux V2: Monodirectional along X-axis disk D+ source, Ed = 40 MeV, Id = 1 microA 0.3cm X
Part II Application of the d-Li accelerator-based intense neutron source of IFMIF-DONES for fusion applications
Radiation Isolation Room RIR-1 Target Interface Room TIR Test Cell (TC) D+ beam The International Fusion Materials Irradiation Facility—DEMO Oriented NEutron Source (IFMIF-DONES) aims to evaluate and validate the structural and functional materials for developing DEMO-type reactors. To achieve this ambitious goal, several projects have been promoted in recent years, which together form the DONES Programme. DONES building CAD model
The CAD model of IFMIF-DONES building is prepared (simplified and decomposed) for the CAD-to-MCNP conversion using the codes: McCad (INR-KIT developed) or SuperMC (developed by FDS-team, China) McDeLicious-17 code package developed at INR-KIT – an MCNP6 code modification for deuteron-lithium (d-Li) nuclear reactions in Li of IFMIF-DONES Test Cell. The beam of deuteron ions accelerated up to 40 MeV with current of 125 mA impinges the liquid Li target delivering 5 MW power. The Li target volume is 5×20×2.5 cm 3 DONES building model horizontal cut at the beam level. DONES building model vertical cut at the target center Total n-flux mapped at the horizontal cut of the IFMIF-DONES Test Cell MCNP model MCNP model 13 IFMIF-DONES Test Cell X-Y scale in cm IFMIF-DONES Test Cell IFMIF-DONES Test Cell D+ beam
Neutron flux spectra at several locations of the Test Cell
Neutronics geometry of the accelerator systems The test cell (TC), which houses the key components of the Target Assembly (TA) and the High Flux Test Module (HFTM) Neutron flux (n/cm 2 /s) at Target Assembly (TA) Photon flux ( ph /cm 2 /s) at Target Assembly (TA)
D+ ion beam stops in the lithium jet delivering a total power of 5 MW on a volume of 5×20×2.5 cm 3, with d-Li footprint area of 5x20 cm 2. Deuterons lose their energy in Li by interactions with Li electron clouds and nuclei – all the processes have been taken into account in the MCNP6 energy deposition calculations with the TMESH card. For calculation of deuteron beam energy deposition in Li at the d-Li footprint area, transport of neutrons, photons, deuterons, and protons – 4 particles have been transported with the MCNP6 mode n p d h Deuteron beam energy deposition in the Li jet at the TA d-Li footprint area Li Steel HFTM DONES High-Flux Test Module (HFTM) Energy deposition, W/cc MCNP6 horizontal cut of the D+ beam energy deposition at the d-Li footprint area with heat peak of 110 kW/cc D+ beam MCNP6 TMESH result for 0.5x1x1 mm 3 (xyz) mesh Horizontal cut of the MCNP6 geometry at d-Li footprint 16 Bragg peak of heat is 1.86 cm deep in Lithium. As thickness of the Li-jet is 2.5 cm, only a 0.64 cm distance separates the Li heat peak and the back plate made of steel (Eurofer-97)
Ref: S.P. Simakov, McDelicious Workshop, FZK/IRS, Institut für Reaktorsicherheit, Forschungszentrum Karlsruhe, 13-14 March 2008
Deuteron track depth (longitudinal, Ld ) dependence on the D+ energy (Ed) Ld =19.8 mm for Ed=40.0 MeV Ld =20.3 mm for Ed=40.5 MeV
Curved surface with Bragg peak at the d-Li footprint area D+ ion beam energy deposition in Li target with Li(d, xn ) neutron source in IFMIF-DONES
DONES Target Assembly (TA) components 20 MCNP modeling of the d-Li source Target Assembly (TA) in DONES MCNP model vertical cut of the DONES TA covered with mesh-tally MCNP mesh-tally
21 Nuclear heat density (W/cc) in the TA materials of the MCNP model TA materials: Steel SS316L material density 7.93 g/cc EUROFER steel with density 7.87 g/cc Lithium (Li) with impurities, its density is 0.512 g/cc. Heating in Li jet at the area of deuteron footprint requires inclusion of the heat contributions of charged particles.
Power balance for the D+ beam energy released in Test Cell (TC) and internal components Type of heat power Heat, kW Input heat power of D+ beam delivered by the IFMIF-DONES one accelerator beam current of I=125 mA 5000 D+ heat released in Li Target 4858.8 Neutron + Photon heat released in the TC components (numbered and displayed in the next slide): 1) TC liner 15.2 2) Removable Biological Shielding Blocks 77.3 3) Bucket liner 0.2 4) Bucket 1.6 5) Piping and Cabling Plugs (PCP) 2.6 6) Lower Shielding Plug (LSP) 9.9 7) Upper Shielding Plug (USP) 0.01 The sum of 7 TC components: 107 Target Assembly (TA) structural parts 17.3 High Flux Test Module (HFTM) 16.9 Integral neutron and photon heat in all considered TC components ~141.2 Output: total integral heat released by D+, neutrons, and photons: 4858.8+141.2= 5000
Integral nuclear (neutron + photon) heating in the Test Cell (TC) components
The interactions of deuterons with lithium target for the energies relevant to fusion applications, particularly Ed=40 MeV in IFMIF-DONES facility, are most accurately described with the McDeLicious code in its actual version McDeLicious-17, as an extension of the MCNP6.1 Monte Carlo radiation transport code. The McDeLicious code has been validated & verified in experimental and computational benchmarks. Using the D+ beam settings, McDeLicious samples neutrons and photons using evaluated d+ 6,7 Li data. The simple model of D+ interactions with cylindrical solid Li inside Aluminum capsule allows to investigate the D+ flux attenuation, track length, Ed attenuation, and D+ energy deposition. This work presented simple model with two settings of the D+ sources: V1: Isotropic uniformly distributed disk V2: Monodirectional directed source defined at a disk The (d-Li) reactions defined in McDeLicious-17 have been studied for the IFMIF-DONES facility. The beam of deuteron ions accelerated up to 40 MeV with current of 125 mA impinges the liquid Li target delivering 5 MW power. The presented results include distributions of D+ energy deposition, neutron and photon fluxes and heating. The integral heating calculations in IFMIF-DONES Test Cell (TC) components reveals that D+ energy deposition in liquid Li at thin Bragg peak with a D+ beam footprint area of 20x5 cm 2 contributes 97% of total heating in the whole Test Cell volume. The 5 MW heat power of D+ beam delivered by the IFMIF-DONES is released by 97% in liquid lithium. 24
Backup slides: Additional information about McDeLicious code parallelization on Marconi-Fusion HPC Use of On-The-Fly (OTF) Monte Carlo variance reduction technique
Figure 1. The MCNP5 speed-up on IFERC-CSC Helios supercomputer. Figure 2. The speed-up of McDeLicious code on Marconi-Fusion HPC. MCNP5 tested on the F4E Broader Approach IFERC-CSC Helios: 2x8 Intel Sandy-bridge EP processors with 2.7 Hz and 64 GB RAM per node: Excellent scalability of MPI/ OpenMP parallel runs of MCNP5 code up to 1024 cores in analogue runs, no variance reduction. Speed-up equals ~450 on 512 cores, and ~850 of speed-up for 1024 cores. OpenMP /MPI hybrid, the satisfactory speed-up of more than 2500 on 4096 cores was achieved for not-biased MCNP5 calculations, as it is illustrated in Figure 1 McDeLicious tested on the EUROfusion HPC Marconi-Fusion with conventional partition (A3) based on INTEL Skylake with peak performance ~9.2 Pflops (2848 nodes). Each node is equipped with 2x24-cores Intel Xeon 8160 CPU ( Skylake ) at 2.10 GHz and 192 GB of RAM per node. Speed-up MPI-parallel performance has been measured and presented in Figure 2 for the McDeLicious code for IFMIF-DONES radiation deep-penetration shielding tasks with variance reduction. The optimal number of CPUs used in MCNP5/6 parallel calculations is dependent on complexity of the model. To improve the statistical errors of the MCNP5 results we are using the ADVANTG approach and the recently developed at KIT On-The-Fly (OTF) Monte Carlo variance reduction technique with dynamic Weight Window upper bounds, see Ref. [ Yu Zheng, Y. Qiu, “Improvements of the on-the-fly MC variance reduction technique with dynamic WW upper bounds,” Nuclear Fusion 62 (2022) 086036, https://doi.org/10.1088/1741-4326/ac75fc ] MCHIFI ( M onte C arlo Hi gh Fi delity ) project has been organized for massively parallel computations on the EUROfusion Marconi-Fusion HPC for the most urgent and computationally demanded fusion neutronics tasks. The MCHIFI project was founded in 2012 to use the IFERC-CSC Helios supercomputer in the framework of the F4E Broader Approach (BA) to serve the ITER neutronics tasks. MCHIFI project: fusion neutronics computations on HPCs of F4E BA and EUROfusion
OTF-GVR : On-The-Fly Global Variance Reduction Weight windows mesh (WWM) is a common method used for MC shielding calculation. Performs “on-the-fly” iterations to get a global flux map and a weight-window mesh. Using novel dynamic WW upper bound method to solve the neutron streaming and “long-history” particles Comparing with ADVANTG, OTF-GVR shows enhancement by a factor of 20 MCHIFI: Development of the On-The-Fly (OTF) MC variance reduction technique Analogue run ADVANTG WWM run OTF-GVR run Percentage of mesh cells and rel. error [ Yu Zheng et al 2022 Nucl. Fusion 62 086036, https://doi.org/10.1088/1741-4326/ac75fc ] Weight window On-the-fly Global weight window mesh generation OTF-GVR : Definition of “c” to avoid “long-history” by limiting the splitting in the OTF run in Ref. [ Yu Zheng, Yuefeng Qiu, et al., “An improved on-the-fly global variance reduction technique by automatically updating weight window values for Monte Carlo shielding calculation”, Fusion Eng. Des. 147 (2019) 111238, https://doi.org/10.1016/j.fusengdes.2019.06.011 ] Ref.:
