TRANSP and PTRANSP: Status and Plans

1
TRANSP and PTRANSP Status and Plans

• Presented at JET/MAST TRANSP Users Meeting,
• Sept. 19, 2007

2
TRANSP Vision Statement
Provide a comprehensive end-to-end modeling
capability for magnetic confinement fusion energy
experiments of today and tomorrow.
3
Traditional TRANSP Overview
Experiments (Asdex-U, C-Mod, DIII-D, ITER, JET,
KSTAR, MAST, NSTX)
MDS
20-50 signals f(t), f(x,t) Plasma position,
Shape, Temperatures, Densities Field, Current, RF
and Beam Injected Powers.
Preliminary data Analysis and Preparation (largely
automated)
Diagnostic Hardware
Pre- and Post-processing at the experimental site
MDS
TRANSP Analysis Current diffusion, MHD
equilibrium, fast ions, heating, current drive
power, particle and momentum balance.
Experiment simulation Output Database 1000-2000
signals f(t), f(x,t)
Visualization
Load Relational Databases
Detailed (3d) time-slice physics simulations
GS2, ORBIT, M3D
FusionGrid TRANSP on PPPL servers
4
TRANSP Developers and Users

• PPPL TRANSP team
• Robert Andre
• Eliot Feibush
• Kumar Indireshkumar
• Jae-Min Kwon
• Long-Poe Ku
• Christiane Ludescher
• Doug McCune
• Lew Randerson
• Paid mostly by NSTX Collaboration projects.
• Also SciDAC, PTRANSP.

• User Sites
• Culham (MAST)
• GA (DIII-D)
• HL2A (China)
• IPP (Asdex-U)
• JET
• MIT (C-Mod)
• PPPL (NSTX)
• PPPL (Collaborations)

KSTAR/NFRC visitor (returns to Korea Oct. 1)
5
PPPL TRANSP Run Production
Fusion Grid TRANSP (SciDAC Collaboratory)
6
Major New TRANSP Features

• Monte Carlo RF Operator (Jae-Min Kwon)
• TORIC wave field solutions coupled to NUBEAM
• Two passes after first pass orbits are
  recalculated with E renormalized to get power
  absorption right.
• MPI-parallel TRANSP Server
• Serial clients share server for 8- or
  16-processor NUBEAM calculations.

7
Beam ion RF-power absorption match between
NUBEAM and TORIC5
total RF-power from NUBEAM
total RF-power from TORIC
8
RF-field renormalization constant for power match
9
Beam ion RF-power absorption profiles
10
Beam ion RF-power absorption profiles by NUBEAM
and TORIC5
TORIC5
NUBEAM
11
NUBEAM RF Operator Issues

• MPI runs needed for better statistics.
• Wave code needs non-Maxwellian target
  distribution function.
• NonMax version of TORIC exists.
• Need to install and test in TRANSP.
• Need to learn how to fit noisy MC-binned
  distribution function data in TORIC.
• Extensive validation will be needed.

12
MPI-Parallel Module Server
Serial TRANSP Run (Client 1)
Input File Package, e.g. XPLASMA NetCDF state.
Server Queue

• MPI-Parallel TRANSP
• Module Server(s)
• NUBEAM monte carlo
• TORIC5 full wave
• GenRAY ray tracing
• CQL3D fokker planck
• GCNM transp. solver
• ...

Serial TRANSP Run (Client 2)
Output File Package, e.g. XPLASMA NetCDF state.
Serial TRANSP Run (Client 3)

viability of method depends on keeping files
small.
Serial TRANSP Run (Client N)
NTCC container module for equilibrium,
profiles, distribution functions, etc.
(http//w3.pppl.gov/NTCC) to be used for Fusion
Simulation Project prototype and tested in TRANSP
deployment.
network
13
NUBEAM Parallel Server

• In operation for early volunteer users.
• Reliability and performance evaluation in
  progress.
• Client server file communications overhead is
  significant
• Only large NPTCLS runs will benefit
• We plan to evaluate a more traditional (no
  client-server) deployment for midrange runs.

14
More TRANSP Improvements

• Incremental improvements to equilibrium
• TEQ somewhat more reliable for STs.
• Some cases still fail.
• Equal Arc poloidal angle option for LEVGEO8
  (scrunch2) runs.
• NUBEAM deposition distribution function data
  set OUTTIM() in namelist use get_fbm on
  ltrunidgt.DATAn output files.
• ElVis RPLOT runs in web browser

15
RPLOT and TRDAT in ElVis
RPLOT session shown
Single window for plots.
Command line i/o
16
Future Directions

• More MPI services TORIC, GENRAY,
• Better MHD equilibrium reconstruction
• Continued improvement to Monte Carlo RF operator
  and wave code coupling.
• More code development collaboration for more
  rapid progress.
• PTRANSP

17
TRANSP is Big Software

• Source builds 219 executable programs
• Sources for 199 subroutine libraries
• 418 (219199) directories containing F90, C, and
  C source code.
• 26 directories of scripts and documents (not
  compiled).
• 178 directories with changes in CY-2007.
• 1.66 Mlines of source (w/comments) 1.13 Mlines
  of source (comments not counted).

18
Code Development is Feasible

• Unified source code control (cvs server at PPPL).
• Unified build system with makefile generator.
• Precise control of contents of debug
  executables for code development.
• Lehigh University physicist Glenn Bateman and
  student Federico Halpern acquired TRANSP code
  developer skills after one week visit to PPPL.

19
PTRANSP Phase 1 (2006 APS)

• Stiff solver upgrades completed
• Free Boundary (TSC) L. P. Ku, JP1.00123
• Prescribed Boundary G. Bateman, JP1.00126
• PTRANSP Client-Server Configuration
• TSC free boundary predictive code client
• Compute evolution of equilibrium and profiles
• TRANSP server
• Compute heating and current drive sources
• Standard analysis of predictive code results.
• See JP1.00123.

20
The PTRANSP Coupled Te,Ti Temperature Solver
Without linearization
With linearization
GLF-23
TSC solver method was ported into TRANSP by
Lehigh U. Group Extensive use in TRANSP for ITER
Simulations by R. Budny Lehigh U. team.
21
Phase 2 PTRANSP is TRANSP

• A project to upgrade TRANSP predictive
  capability.
• Retained from TRANSP
• Code base
• Production system
• Connection to experimental data
• Connection to post-processors
• Connection to user community.

22
PTRANSP Phase 2

• non-renewable 3 Year Grant, 650k/year
• Funding approved late in FY-2007
• General Atomics (25)
• Lehigh University (25)
• LLNL (25)
• PPPL (25).
• Work scope of grant clearly focused on predictive
  upgrades to TRANSP itself.

23
PTRANSP Phase 2 GA Role

• Add GCNM-P Solver to TRANSP
• Allow flexible applications
• Prescribe electron density (T/F)
• Prescribe ion densities and impurity levels (T/F)
• Include depletion by fast species.
• Separately computed MHD equilibrium and q(r,t).
• Import TGLF predictive Transport Model into
  TRANSP via GCNM-P.
• Likely to use SWIM Plasma State software.
• Support other uses of TGLF as needed.

24
PTRANSP Phase 2 Lehigh U.

• Program of direct improvements to TRANSP internal
  solvers (with PPPL).
• Predictive Sawtooth and Pedestal models.
• Intensive use of PTRANSP for research
  applications.

25
PTRANSP Phase 2 -- LLNL

• Provide Free Boundary TEQ model to PPPL.
• Enhance TEQ to enable concurrent prediction of
  poloidal field diffusion and MHD equilibrium.
• Additional TEQ enhancements (e.g.
  hyper-resistivity).

26
PTRANSP Phase 2 -- PPPL

• Provide TRANSP system and development support to
  all participants.
• Install the TEQ model upgrades provided by LLNL.
• Place PTRANSP capabilities in production and
  trouble-shoot applications.
• Provide additional TRANSP/PTRANSP upgrades as may
  be needed.

27
PTRANSP Focus on Plasma Core

• True whole device predictive modeling requires
  validation with close coupling to
• Scrape-off Layer (Edge) Plasma Model.
• Wall Model.
• Many other things SOL Atomic Physics, etc.
• Current PTRANSP plans are short term.
• Not high performance super-computing.
• No true whole device predictive model.
• Such capabilities require a much larger effort.
• Fusion Simulation Project 24M/year??

28
Collaboration Opportunities

• Improve TRANSP RF capabilities.
• Improve PTRANSP capabilities
• Performance options (e.g. fast source models).
• TGLF installation in TRANSP-native solver (when
  TGLF is available).
• Significant commitment required for success
  0.5 person-year / collaboration.

29
TRANSP Users Group at APS

• Annual TRANSP Users Group Meeting.
• Review/discuss (P)TRANSP status plans.
• Monday evening satellite meeting at APS.
• This year
• Orlando, Florida APS-DPP conference.
• Monday, Nov. 12, 8pm.
• Apologies in advance for the inevitable schedule
  conflicts