High Performance Computing for Non-linear MHD Simulations for Fusion Plasmas - Shimpei Futatani (Universitat de Politècnica de Catalunya) - PRACE ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
High Performance Computing for Non-linear MHD Simulations for Fusion Plasmas Shimpei Futatani (Universitat de Politècnica de Catalunya) Acknowledgements: JOREK Team and Y. Suzuki (NIFS) Page 1 S. Futatani, PRACEdays19, May 13-17, 2019
What makes stars shine? “The starry night”, V. Willen van Gogh Page 2 S. Futatani, PRACEdays19, May 13-17, 2019
What makes stars shine? “The starry night”, V. Willen van Gogh Can we use the energy of shining stars as an alternative energy source? Page 3 S. Futatani, PRACEdays19, May 13-17, 2019
What makes stars shine? – Nuclear fusion Nuclear Fission U235 Nuclear Fusion He ++ + T + D n E = Dmc2 [nasa.gouv] Page 4 S. Futatani, PRACEdays19, May 13-17, 2019
Outline Introduction - Plasma / Fusion reactor / ITER Project Physics background - Tokamak - What is ELM? - JOREK - Nonlinear MHD code - www.jorek.eu - The JOREK team and themes - Numerical details and physics model - Mechanism of pellet triggered ELM - JOREK modelling of pellet triggered ELM Physics background – Stellarator - MIPS code - MIPS simulation result Conclusions and Perspectives Page 5 S. Futatani, PRACEdays19, May 13-17, 2019
What is plasma? For achivement the fusion à Plasma state is needed - Plasma is the 4th state of matter, obtained at high temperature (>105 degrees) - Plasma is an ionized gas which consists of ions and electrons. [www.nasa.gov] Page 6 S. Futatani, PRACEdays19, May 13-17, 2019
Galaxy fusion reactor in the universe Key parameters for the fusion = high-temperature and high-density è How can we confine the high-temperature plasma? - In stars: plasma particles are confined mainly by gravity. [www.setterfird.org] [wikipedia.org] Page 7 S. Futatani, PRACEdays19, May 13-17, 2019
Fusion reactor on Earth Key parameters for the fusion = high-temperature and high-density è How can we confine the high-temperature plasma? - On Earth: plasmas can be confined in Magnetic field lines = Magnetic Confinement • Charged particles spiral around magnetic field lines. • Toroidal (Donut shaped) system avoids plasma hitting the end of the container è Tokamak Page 8 S. Futatani, PRACEdays19, May 13-17, 2019
Fusion plasma reactors in the world Tokamaks (not listed all here) in the world JET (EU) and MAST (UK) ASDEX Upgrade (Germany) ITER (international project) JT60U (Japan) Page 9 S. Futatani, PRACEdays19, May 13-17, 2019
What is ITER? [www.iter.org] ITER is a major international collaboration in fusion energy research involving China, the EU (plus Switzerland), India, Japan, the Russian Federation, South Korea and the United States [www.iter.org] ITER ITER site (January 2017) Page 10 S. Futatani, PRACEdays19, May 13-17, 2019
Overview of the ITER Tokamak Pit Drone view of the tokamak pit and bioshield construction [Figure from Pinches, ITER Organization] Page 11 S. Futatani, PRACEdays19, May 13-17, 2019
A big challenge for control of fusion plasma Eruption of high temperature plasma = solar flares (for sun) = ELMs (for Magnetically confined plasma) è Big challenge for the control of the plasma Page 12 S. Futatani, PRACEdays19, May 13-17, 2019
Edge Localized Modes (ELMs) • Fusion plasma has a strong pressure gradient at the plasma boundary. • Edge pressure gradient is limited by an MHD instability (ballooning mode) • A crash of the edge profile occurs • Release of hot plasmas onto a plasma facing components • ELM removes up to 10% of the plasma energy in ~200 microseconds ELM • Periodic and bursty behaviors [Figure from Pitts, ITER Organization] Page 13 S. Futatani, PRACEdays19, May 13-17, 2019
Edge Localized Modes (ELMs) • Fusion plasma has a strong pressure gradient at the plasma boundary. • Edge pressure gradient is limited by an MHD instability (ballooning mode) • A crash of the edge profile occurs • Release of hot plasmas onto a plasma facing components • ELM removes up to 10% of the plasma energy in ~200 microseconds ELM • Periodic and bursty behaviors Fast camera image of ELM (MAST) [A. Kirk et al.] Page 14 S. Futatani, PRACEdays19, May 13-17, 2019
Edge Localized Modes (ELMs) [Linke 2007] ELMs lead to a large erosion of and a limited lifetime of the plasma facing components. èRequires physics understanding of ELMs and ELM control • Techniques to control ELM: • stabilisation by external magnetic perturbations • triggered by pellet injection (pellet : deuterium solid ice cube) • Etc… Page 15 S. Futatani, PRACEdays19, May 13-17, 2019
Demonstration of ELM pacing by pellets without fueling Divertor [Baylor APS 2010] - Pellets can control the ELM frequency - Heat flux of the pellet triggered ELMs on the fusion reactor wall becomes small. Theoretical and Numerical Modelling studies are needed. Page 16 S. Futatani, PRACEdays19, May 13-17, 2019
Non-linear MHD code JOREK • JOREK has been developed with the specific aim to simulate ELMs, developed by Dr. G. Huijsmans (CEA/Univ. Eindhoven). • G.T.A. Huysmans, Plasma Phys. Control. Fusion 47, B165 (2005) • G.T.A. Huysmans and O. Czarny, Nuclear Fusion 47, 659 (2007) • O. Czarny and G. Huysmans, J. Comp. Phys. 16, 7423 (2008) • See [https://www.jorek.eu/] Page 17 S. Futatani, PRACEdays19, May 13-17, 2019
Non-linear MHD code JOREK • European Enabling Research Project (PI: M. Hoelzl) • JOREK collaborations (>30 members, >10 international institutions): • JOREK main development and application • Involved institutes: CEA, IPP Garching, ITER, CCFE, Eindoven, UPC etc. • Pellet triggering of ELMs / Shattered pellets for disruption • S. Futatani , D. Hu etc. • ELM mitigation/suppression by external fields • F. Orain, M. Becoulet, K. Wittawat, M. Hoelzl, etc • Full orbit particle tracer • D. Van vugt, A. Dvornova, C. Somariva, etc. • Disruptions (MGI, REs, etc) • E. Nardon, C. Sommariva, V. Bandaru, M. Hoelzl, D. Meshcheriakov, F. Wieschollek, etc • Simulation of ELMs • G. Huijsmans, S. Pamela, M. Becoulet, F. Orain, M. Hoelzl, A. Cathey etc. • Solvers (PaStiX, HIPS, Interface MURGE) • P. Ramet, P. Henon, X. Lacoste, Univ. Bordeaux/INRIA • Full MHD model/numerical methods • B. N’Konga, G. Huijsmans, H. Guillard, S. Pamela • Resistive Wall/Free boundary version (VDEs, RWMs, vertical kicks, ...) • M. Hoelzl, J. Artola-Such, etc • Numerical methods • B. N’Konga, E. Sonnendruecker, H. Guillard, E. Franck etc Page 18 S. Futatani, PRACEdays19, May 13-17, 2019
Non-linear MHD code JOREK • Numerical features: • Discretisation: Xpoint geometry • Cubic finite elements flux-aligned poloidal grid • Fourier series in toroidal angle • Time stepping: • fully implicit Crank-Nicholson • Solver sparse matrices (PastiX library) • GMRES iterative solver with physical preconditioner [Provided by M. Hoelzl] • Parallelisation using MPI/OPENMP • ~30000 poloidal elements • Typically 1880-3800 cores • MareNostrum IV (BSC), Marconi-Fusion (CINECA), CURIE (France), hydra (Germany), etc Acknowledgement to PRACE project. [Provided by S. Pamela] Page 19 S. Futatani, PRACEdays19, May 13-17, 2019
Non-linear MHD code JOREK • Reduced MHD model (JOREK also has the full MHD model). • Braginskii parallel conductivity • Spitzer resistivity • Mach-1 boundary condition, free flow on divertor target • Magnetic field and the velocity • Mass density Coupled with the pellet ablation model • Poloidal momentum (vorticity) • Parallel momentum • Temperature • Poloidal flux Page 20 S. Futatani, PRACEdays19, May 13-17, 2019
Pellet model and implementation in JOREK • Realistic pellet ablation model (NGS model [Gal, NF(2008)]) is implemented in JOREK : • Pellet moves at fixed speed and direction • Pellet is modelled as an adiabatic localized time-varying density source rp: pellet radius [m] ne: plasma electron density [m-3] Sp: Density source Te: plasma electron temperature [eV] JOREK solves a very complex, multi-physics which requires HPC Interactions Non-linear MHD physics Pellet ablation physics Page 21 S. Futatani, PRACEdays19, May 13-17, 2019
Mechanism of pellet triggered ELM Page 22 S. Futatani, PRACEdays19, May 13-17, 2019
Plasma dynamics after the pellet injection 1.1mm pellet 1.7mm pellet Page 23 S. Futatani, PRACEdays19, May 13-17, 2019
Heat flux on the divertor target Heat flux profile on the divertor target ELM [Wenninger et al. (2010)]. (a) spontaneous ELM (b) pellet triggered ELM Spontaneous ELM Pellet density Pellet triggered source creates ELM (1.7mm) another channel to the divertor target. è Distribution of the heat flux in a wide area. Page 24 S. Futatani, PRACEdays19, May 13-17, 2019
Simulations of stellarator plasma • Stellarator is an alternative type of fusion reactor. It exploits strangely-shaped magnets that is hard to build but potentially easier to operate. • The global MHD dynamics of the Large Helical Device (LHD) in Japan has been analyzed with the MIPS-code. Page 25 S. Futatani, PRACEdays19, May 13-17, 2019
Fusion plasma reactors in the world Stellarators (not listed all here) in the world Wendelstein 7-X (Germany) NCSX (US) TJ-II (Spain) LHD (Japan) Page 26 S. Futatani, PRACEdays19, May 13-17, 2019
MIPS code • MIPS code solves the full MHD equations starting from the initial equilibrium. [Y. Todo et al., PFR (2010), K. Ichiguchi et al., NF (2015)] • Those equations are solved in the rectangular grids of the cylindrical coordinates (R, f, Z). • Any plasma geometry can be implemented by implementation of magnetic field. • Good scaling of parallelization up to 4096 cpu cores. Page 27 S. Futatani, PRACEdays19, May 13-17, 2019
Pellet cloud profile • Realistic pellet ablation model (NGS model [Gal, NF(2008)]) is implemented in MIPS. • The pellet cloud profile shows the m/n=1/1 shape. The density (pressure) source is localized in the toroidal and the poloidal direction Pellet injection Page 28 S. Futatani, PRACEdays19, May 13-17, 2019
Pellet cloud profile • Realistic pellet ablation model (NGS model [Gal, NF(2008)]) is implemented in MIPS. • The pellet cloud profile shows the m/n=1/1 shape. Page 29 S. Futatani, PRACEdays19, May 13-17, 2019
MHD excitation by pellet • The three-dimensionally localized pressure perturbation excites the MHD modes. • Magnetic energies evolve. • m/n=2/1 (solid lines), m/n=2/2 (dashed lines), m/n=2/3 (dotted lines) • The plasma is stable in the absence of the pellet. • After the full pellet ablation, the energies suddenly go relax state (t=400- 500µs). Page 30 S. Futatani, PRACEdays19, May 13-17, 2019
Conclusions and perspectives JOREK simulation by G. Huijsmans (CEA/ITER) Conclusions • JOREK has been performed to study the non-linear MHD physics. • JOREK allows us to compute the ELM physics and calculate the heat flux onto the plasma facing components. • Qualitative agreements with the experiment results of JET are observed via simulation. • MIPS has been performed to study the pellet effect in stellarator plasma. • The excitation of MHD modes caused by pellet injection is observed. Future works • JOREK pellet simulation will be performed including more physics effects to allow for a more quantitative comparison. • Improvement of physics model in MIPS will be carried out. Page 31 S. Futatani, PRACEdays19, May 13-17, 2019
You can also read