0815 - 0900 Continental breakfast
NIMROD TEAM MEETING August 6-7, 2003 Sherwood Hills Resort Wellsville, Utah MINUTES _______________________________________________________ Wednesday August 6, Aspen Room , Sherwood Hills Resort, Wellsville
0900 - 1200 CODE DEVELOPMENT ISSUES/STATUS
Progress on fast direct solvers - Carl
SuperLU sparse direct linear solver (from ACTS) has been implemented in NIMROD. We are using the distributed memory parallel version. Implemented as a pre-conditioner to CG. Usually doesn't require CG iterations, but maybe a few for highly ill-conditioned systems. Performance can be 100 times faster than our previous pre-conditioned CG alone. Takes more memory. ("nybl is better than nxbl" - Carl). Factors of 100 improvement for linear problems, factors of 5 improvement for nonlinear problems (because of need o recompute the LU factors as the solution evolves).
Next steps: fortran utility for interfacing non-symmetric systems to SuperLU.
Hall SI operator - Carl
With Hao Tian. NIMROD has had 2-fluid for years. The SI operator is second order and not very accurate for large time steps. (It acts like an increased electron inertia.)
Hao is working on fourth order operator based on Harned-Mikic. Predictor-corrector is required. This is being debugged using AZTEC iterative solves.....almost there. Still needs boundary modifications. Direct non-symmetric solves will be ready soon.
Other code issues
By changing a parameter in NIMLIB, one can create executables that use "spectral elements" where the basis functions are Gauss-Lobatto-Legendre polynomials instead of Lagrange polynomials.
Resistive wall - Scott
Vacuum edge: H-mode equilibria are not very well constrained at the edge (a feature of efit). We do not yet have a successful strategy for dealing with these cases.
NOTE FROM DYLAN: The Li beam diagnostic is just coming online at DIII-D and has within the past few months begun to be included in EFITs. This data should help to solve this problem by accurately constraining the current at the edge. The kinetic data is well represented at the edge, the problem has been the current.
Implementation of the resistive wall has begun.
Energetic particles - Charlson
Particles getting low efficiency on seaborg (6/16**2). Progress is being made on M3D comparison working with G. Fu. Charlson spent a week with Guoyong at PPPL( Scott P. paid the travel from his SCIDAC funds). Initial results look pretty good, nice eigenfunctions and eigenvalues. Results depend a lot on resistivity and viscosity. Charlson and Fu must agree in detail on the equilibrium and parameters for a series of cases, and produce a plot of growth rate and real frequency as functions of beta_hot, compare with each other and with NOVA-K.
Kinetic heat flux - Eric
Paper showing reduction of parallel heat flux due to particle trapping has been accepted for publication. This is ready for utilization in the Big Splash case. Modifications have been worked out for field lines that intersect the wall.
The formalism may go over to calculating the parallel viscous stress.
Parallel nimfl is almost ready for use. This will be a large advance for post-processing.
1330 - 1430 TWO-FLUID ISSUES
Ordered fluid equations - Dalton
A very orderly presentation (except for the parts that were in error!)
CDXU/M3D comparison - Dalton
THE CDXU CASES FOR Q0=0.95 (The NOW agreed upon value) show resistive ballooning for n>9. They also show a robust n=1, as expected. One possibility is that the high n modes are secondary instabilities that are driven as the 1/1 helically distorts the plasma. We will check the linear stability of these modes.
During the Thursday meeting (thanks to SLU), we ran the linear CDXU q0=0.95 case and found instability.
1630 - ??? Golf, the 19th hole, and dinner
The Leader Board: Schnack 54 Held 57 Hegna 72 Sovinec 74
0900 - 1100 APPLICATIONS II
Alternates - Carl
LDX simulations have begun in collaboration with Jay Kessner.
PPCD in RFP: Jim Reynolds is working on simulation studies of plasma dynamics during PPCD operation.
Carl has provided a Corsica equilibrium interface for linear spheromak computations at LLNL.
Big Splash - Scott (maybe Dylan by phone)
It has been proven difficult to generate an NTM from ITER equilibria. Runs require lost of resolution, especially in phi. Need ways to stabilize high-n modes; presently using low-pass filter. (Carl is skeptical of this.) On-going runs look promising. We are under pressure to get results by the end of September.
High-beta disruption - Scott/Dalton
Calculated heat loads at DIII-D divertor are in good agreement with experiment. Collaboration with Dennis Whyte has begun. Expect paper publication in the fall, after some diagnostic development.
NSTX - Charlson
Charlson reviewed his recent visit to PPPL/NSTX. He presented a "wish list" from Jon Menard and Steve Sabagh for NIMROD applications. We will meet with the NSTX group at APS.
1100 - 1200 SPACE PHYSICS
Discussion with Jan Sojka
A very interesting talk about ionospheric plasma physics.
1200 - 1330 Lunch - White Owl
1330 - 1430 OTHER COMPUTATIONAL ISSUES
Useful scripts - Scott
Scott gave a presentation on scripts he has developed for running and maintaining NIMROD. These are all available on the web.
Data/viz issues using Vorpal - "The Pros from Boulder"
John and Sveta gave very interesting presentations on their networking and visualization projects. There was agreement that we should collaborate. Scott and John Cary are going to get started.
Carl reported on graphics work by our summer undergrad research assistant, Dan Swanson, to produce element-based plots of derivatives of solution fields.
Pre/post-processing - ??
Web - ??
1430 - 1530 PLANNING
What are the next development steps?
Are there opportunities for applications?
Follow through with SI Hall implementation.
Complete energy equation for drift-like model.
Implement drift like model.
Implement "full 2-fluid", including GV tensor.
Complete energetic particles. Need 2-fluid benchmarks.