Notes from the NIMROD Team Working Meeting
November 9, 2013, Denver, Colorado
To avoid a scheduling conflict with the US-Japan JIFT Workshop on "Recent studies of extended MHD and MHD simulations," the NIMROD Team held a half-day working session in the morning. The group discussed four topics: interfacing the GRIN Green's function solver, implementing dynamic neutrals, applying Eric Held's continuum kinetic model, and using the new projection-based stabilization. An assigned discussion leader for each topic provided background and some details.
Discussion of the GRIN interface was led by Jake King, who has been developing multi-wall capability as part of an ITER grant. GRIN is linked as a separate library. While it has its own thin-wall computations, King recommends computing Bnorm evolution for the first wall within NIMROD. When setting-up the interface, GRIN vertices are specified to align with NIMROD's finite-element nodes for a given application. However, GRIN uses piecewise-constant bases for its magnetic potential, so some projection or differencing will be needed to evaluate Btang on the vacuum side of the wall. The Green's function information can be obtained from GRIN output files (hdf5 format) or by calling GRIN library routines. The group had differing opinions on which implementation method would be best for NIMROD. For the immediate future, the hdf5 file information is converted to a text file format to conform with the existing NIMROD implementation. This precludes the GRIN modeling with external coils, which will require linking the library, however it should expedite testing of the single wall model. King and Andi Becerra (formerly Montgomery) agreed to meet during APS week to cover details of the implementation.
Peter Norgaard updated the group on his development of dynamic neutrals with electron impact ionization and recombination. The reactions are represented as sources and sinks in the ionized-particle and neutral continuity equations. The group discussed implicit vs. explicit implementation of the reaction terms. High-temperature conditions are mathematically stiff, but few neutrals survive in these conditions. With linearization about the start of a time-step, making linear terms implicit may be sufficient for magnetic-confinement applications while avoiding complications of solving nonlinear algebraic systems. Norgaard has verified his initial implementation of the source and sink terms with analytical results for 0D configurations. Sovinec will help Norgaard learn how to implement 3D linearly implicit terms.
Discussion of the continuum kinetic modeling capability was led by Scott Kruger. He first provided an update on Eric Held's benchmark on neoclassical bootstrap current. Results obtained since the SciDAC review in October consider a high-temperature highly shaped NSTX equilibrium. In these conditions, the Sauter model is valid, and there is good agreement between Sauter's analytics, Emily Belli's 1 spatial-D NEO-code result, and Held's 2 spatial-D NIMROD result. Held made the NIMROD computation practical by completing the development of the Coulomb collision operator directly in the spectral-element v-space representation used for other terms of the drift kinetic equation. Kruger noted that essentially all of the continuum coding is in the nimdevel repository trunk. However, input includes a number of outmoded options that would confuse users. Thus, code cleanup is a critical step from this point. Dalton Schnack will use the development to investigate kinetics in giant sawtooth applications.
Carl Sovinec summarized the development of flow-stabilization terms to simultaneously provide stable-side convergence on localized MHD interchange and node-scale smoothing for nonlinear computations. King has already ported a clean version of the development to nimdevel, and only a couple of numerical parameters are needed to fully specify the added features. Sovinec showed encouraging linear and nonlinear results for MHD interchange and peeling-ballooning (ELMs). The group discussed how the terms affect linear convergence properties and advantages when carrying-out nonlinear convergence studies. The development is expected to facilitate a number of applications, but it alone is insufficient for stabilizing numerical modes that appear in some two-fluid applications.
NIMROD-related presentations in the half-day Center for Extended MHD Modeling meeting on Nov. 10 are linked below:
Carl Sovinec, "Spectral Projections in NIMROD"
Jacob King and Scott Kruger, "Update on EHO Modeling with NIMROD"