April 4-6

Log in

Abstract Details

files Add files

status:file name:submitted:by:
approveddimits_shwd16_abstract_v3.pdf2016-02-15 16:24:58Andris Dimits


Author: Andris Dimits
Requested Type: Poster
Submitted: 2016-02-15 16:24:29

Co-authors: I. Joseph, L. L. LoDestro, M. E. Rensink, T. D. Rognlien, B. Sjogreen, D.P. Stotler, M. V. Umansky

Contact Info:
Lawrence Livermore National Laboratory
PO Box 808/7000 East Avenue
Livermore, CA   94550

Abstract Text:
An accurate coupled treatment of fluid-plasma and kinetic-neutrals models of the tokamak edge and scrape-off layer is necessary for the interpretation and prediction of divertor performance [1]. We investigate various approaches to efficient implicit coupling of such models.
While a highly efficient implicitly coupling of fluid-neutrals and plasma models exists within the UEDGE code [2], a kinetic [e.g. Monte-Carlo (MC)] neutrals treatment is necessary in some important regimes. Existing couplings of these with plasma-species models are explicit [3], often making them very slow computationally due to numerical "stiffness" associated with the large disparity between the fast (e.g., charge-exchange and ionization) and slow (transport/global equilibration) time scales.
The approaches investigated are based upon Newton's method, and include quasi-Newton (e.g., generalized Broyden's methods and Anderson mixing) and Jacobian-free Newton-Krylov methods (e.g., GMRES).
In our case the constituent codes are the plasma component of UEDGE and the DEGAS-2 MC-neutrals code [4]. For algorithm testing, we use a Cartesian box geometry and plasma conditions based on previous studies [3]. Additional insight is gained through comparison with and analysis of simplified model systems consisting of either two coupled fluid fields diffusing in one dimension or of one fluid field coupled to a MC species, with controllable sources and coupling terms.
Key issues addressed include the development of adequate robust "preconditioners," necessary for the Newton-like methods to converge rapidly, and optimization of the convergence of the calculation and of the accuracy of the resulting solution with respect to the MC "noise."
[1] A.S. Kukushkin et al., Fus. Eng. & Design 86 (2011) 2865–2873.
[2] T.D. Rognlien, et al., Contr. Plasma Phys. 34 (1994) 362.
[3] M. E. Rensink et al., Contr. Plasma Phys. 38, 325 (1998).
[4] D. Stotler and C. Karney, Contrib. Plasma Phys., 34, 392 (1994).