Speaker
Elizaveta Kaveeva
Description
See the full Abstract at http://ocs.ciemat.es/EPS2018ABS/pdf/P5.1015.pdf
Speed-up of SOLPS-ITER code for tokamak edge modeling
E. Kaveeva1, V. Rozhansky1, I. Senichenkov1, I. Veselova1, S. Voskoboynikov1,
X. Bonnin2, D. Coster3
1
Peter the Great St.Petersburg Polytechnic University, St.Petersburg, Russia
2
ITER Organization, St Paul Lez Durance Cedex, France
3
Max-Planck Institut für Plasmaphysik, EURATOM Association, Garching, Germany
Understanding of edge plasma performance and divertor exhaust is crucial for operation of
ITER and other tokamaks. Traditionally this is done by transport codes like SOLPS and
others, based on Braginskii model for parallel transport, experimentally based description of
anomalous transport and Monte-Carlo model for neutral transport. In the early versions of
SOLPS self-consistent electric fields, drifts and currents were ignored. These effects were
introduced into the version which is known as SOLPS5.0 [1]. The physics of the edge plasma
with drifts is treated much better by the new version, however one has to pay price by slower
convergence of the code. Later modifications - SOLPS5.2 [2] and its upgrade by the ITER
Organization to form a new package, SOLPS-ITER [3] still exhibit slow convergence.
Account of drifts and currents dramatically decreases the accessible time step for the
integration of time dependent equations of the code. Running the code with sophisticated
EIRENE Monte-Carlo model for neutrals and large number of fluid equations for multiple
ion species makes the computation time unacceptably long.
In the present paper the mechanisms leading to the time step limitations in
SOLPS-ITER are analyzed as well as the ways to relax these limitations. The numerical
instability driven by drifts is associated with poloidal redistribution of particles inside the
separatrix by ExB drift in combination with modification of the radial electric field by
diamagnetic currents. It can be overcome by implementation of one of two algorithms. The
first method uses artificial slowing down of poloidal density and temperatures redistribution.
In the second method equations are modified to get faster convergence to a solution close to
the true one, which then is used as an initial approximation for convergence to the true
solution. Application of these schemes decreases the time of convergence for a steady state
solution by more than an order of magnitude. Ways to improve convergence by introducing
artificial particle sources and artificial rise of time derivatives are also suggested.
[1] Rozhansky V. et al., Nucl. Fusion 41 (2001) 387
[2] Rozhansky et al., Nucl. Fusion 49 (2009) 025007
[3] S. Wiesen et al., J. Nucl. Mater.463 (2015) 480
