Fusion Simulation Program (FSP) Scientific and ...Fusion Simulation Project (FSP) Mission and Vision...

Fusion Simulation Program (FSP)Scientific and Computational Challenges

Arnold H. Kritz

Lehigh University, Bethlehem, PA USA

JET, July 6, 2010

Outline

• Fusion Simulation Project (FSP) Mission and Vision

• Background and Preparation for FSP during the last decade

• Role of High Performance Computing and ASCR

• FSP Planning effort

• Validation, Verification, Relation to Experimental Programs

• FSP Contributions to the Fusion Program

2Kritz JET ─ July 6, 2010

Goal of the Fusion Simulation Program (FSP) is to enable scientific discovery of important new plasma phenomena with associated understanding that emerges only upon integration

Requires developing a predictive integrated simulation capability for magnetically-confined fusion plasmas that are properly validated against experiments in regimes relevant for producing practical fusion energy

The Fusion Simulation Program (FSP) will develop advanced HPC-enabled tools to accelerate understanding of magnetized toroidalplasmas via efficient integration of multiple, coupled physical processes

This task will engage theory, experiment, and advanced HPC resources to deliver unprecedented capability for harvesting information from experiments and designing new devices with improved performance

FSP Mission

Kritz JET ─ July 6, 2010

• FSP will provide the capability to confidently predict toroidal magnetic confinement fusion device behavior

−With comprehensive and targeted science-based simulations of nonlinearly-coupled phenomena

in the core plasma, edge plasma, and wall region −On time and space scales required for fusion energy production

• Integrate the knowledge from key multi-scale physical processes −Continually improve fidelity for extending whole-device modeling

capabilities beyond current applicability domains• Produce a framework in which physics component-codes interact

efficiently• Enable unprecedented capabilities to compute experimental observables,

interpret experimental data, and explore the consequences of theoretical models

• Incorporate modern software engineering and software quality assurance

• To ensure reliability, robustness, and ease-of-use of tools that are developed

• Predictive integrated simulation capability for magnetically-confined fusion plasmas that are properly validated against experiments in regimes relevant for producing practical fusion energy

FSP Vision


FSP Background and Historical Perspectives • Initially computational modeling and simulation efforts not supported on

their own, but often as a component of an experimental program and sometimes a theoretical program−Support for computational efforts fluctuated and was not secure

• 1990: Significant change in approach regarding computational projects• Initiation of SciDAC projects in 1990 – three year grants, $500K to $1M

−Community wide projects involving physicists, computer scientists and applied mathematicians −MHD, RF, Atomic, Magnetic Reconnection

−Additional single topic projects added in subsequent years−Turbulence, Energetic Particles, many earlier projects renewed

• Plans for FPS also began in 1990−DoE Director of Science Call to Fusion Energy Advisory Committee−Letter to FESAC (Fusion Energy Scientific Advisor Committee) by

Jim Decker, Acting Head of DOE Office of Science, calling for −Establishment of Committee to study Feasibility of FSP


• What is the current status of integrated computational modeling and simulation?

• What should be the vision for integrated simulation of toroidal confinement fusion systems?

• What new theory and applied mathematics are required for simulation and optimization of fusion systems?

• How should integrated simulation codes be validated, and how can they best be used to enable new scientific insights?

• What computer science is required for simulation and optimization of fusion systems?

• What are the computational infrastructure needs for integrated simulation of fusion systems?

Questions Posed in Charge Letter From J. Decker

• Dahlburg Committee (14 members); Community inputReport: http://www.isofs.info/FSP_Final_Report.pdf

http://www.isofs.info/FSP_Appendix.pdfJournal of Fusion Energy Vol. 20, pp 135-196, 2001


• Committee appoint by Office of Fusion Energy in 2003 to further explore initiating the Fusion Simulation Project

• Recommended that Focused Fusion Initiatives be started Report of the Fusion Simulation Project Steering CommitteeD. Post, D. Batchelor, R. Bramley, J. Cary, R. Cohen P. Collella, S. JardinJournal of Fusion Energy, Vol. 23, No. 1, 2004

• Result: Funding of three prototype fusion simulation projects− Simulation of Wave Interactions with Magnetohydrodynamics (SWIM)

http://cswim.org − Center for Edge Plasma Simulation (CPES)

http://www.cims.nyu.edu/cpes− Framework Application for Core-Edge Transport Simulations (FACETS)

https://ice.txcorp.com/trac/facets • Funding Partners: Office of Science — Office of Advanced Scientific

Computing Research and Office of Fusion Energy

Focused Fusion Initiatives


https://ice.txcorp.com/trac/facets

DOE FSP Workshop (May 2007)• Planning of workshop involved 45 applied mathematicians,

computer scientists and physicists from 22 institutions• Considered the needs of the fusion program and emerging

opportunities−Scientific modeling issues for burning plasmas

Identified five critical scientific issues:1. Disruption effects, including avoidance and mitigation2. Pedestal formation and transient divertor heat loads3. Performance optimization and scenario modeling4. Tritium migration and impurity transport5. Plasma feedback control

−Verification and validation− Integration and management of code components−Computational and mathematical enabling technologies−Project Management and Structure

Journal of Fusion Energy 28, 1 (2009), by A. H. Kritz and D. Keyeshttp://www.science.doe.gov/ofes/programdocuments/reports/FSPWorkshopReport.pdf


Elements of an FSP Integrated Model

Magnetically confined plasmasin a tokamak are complex and demand integrated analysis


•Overlap in scales often means strong (simplified) orderingis not possible

•Needed to effectively harvest insights from ITER and to plan for DEMO

•Effective simulations at the petascale (1015 floating point operations per second) and beyond are required to address grand challenges in plasma science

Integrated predictive models must span huge range of spatial and temporal scales -- major challenges for theory and simulation

• Full predictive modeling of fusion plasmas requires cross coupling of a variety of physical processes and solution over many space/time scales


Advanced Modeling Codes --- “a measure of the state of understanding of natural and engineered systems”

T. Dunning, 1st SciDAC Director

Theory(Mathematical Model)

AppliedMathematics

(Basic Algorithms)

ComputationalPhysics

(Scientific Codes)

ComputerScience

(System Software)

Problem with Mathematical Model?

Computational Predictions

Agree* w/ Experiments?No Yes Speed/Efficiency?

Inadequate

AdequateUse the New Tool for Scientific Discovery(Repeat cycle as new

phenomena encountered )

*Comparisons: empirical trends; sensitivity studies; detailed structure (spectra, correlation functions, …)

“V&V” Loop

“Performance” Loop


Magnetically Confined Burning Plasmas: Unique opportunities for scientific discoveries

• BP/ITER physics elements raise mission-critical questions– Unprecedented size– Self-heating– Large energetic particle population– Multiple instabilities with unknown

consequences for fast ion confinement

Predicting fast ion confinement:

Critical for sustaininga burning plasma

•What is nonlinear interactionbetween energetic particlesand “sea of Alfvén modes?”•How is transport affected by presence of multiple instabilities?•How can predictive numerical tools be properly validated?•What scale of computational resources will be needed to answer BP/ITER mission-relevant questions?


Fusion Application on the Exascale (March 2009)• Fusion Energy Science and the Role of Computing at the Extreme Scale:

Workshop led by W. Tang and D. Keyes−Burning Plasma/ITER Science Challenges, N.Sauthoff, N. Gorelenkov−Advanced Physics Integration Challenges, J. Cary, A. Kritz−Plasma-Material Interaction Science Challenges, S. Zinkle, J. N. Brooks −Laser-Plasma Interactions and High-Energy Density Laboratory Physics,

R. Betti, W. Mori, −Basic Plasma Science/Magnetic Reconnection Physics,

A.Bhattacharjee, W. Daughton−Algorithms for Fusion Energy Sciences at Extreme Scale,

D. Keyes−Data Analysis, Management, and Visualization in

Fusion Energy Science, A Shoshani−Mathematical Formulations, P. Colella, R. Samtaney−Programming Models, Frameworks, and Tools, E. Lusk

• Forefront questions in fusion energy science and potential role of high-performance computing at the extreme scale to help resolve scientific issues

http://extremecomputing.labworks.org/fusion/PNNL_Fusion_final19404.pdf



Focus at Exascale Workshop (March 2009)• Over the next decade, significant advances in computing technology −Allow increases in computing performance by orders of magnitude−Major impact on ability to solve critical scientific and technological problems

• Workshop focused on 1. Identify forefront scientific problems in fusion energy science that could

be aided by computing at the extreme scale over the next decade 2. Establish specifics of how and why new high-performance computing

capabilities will address issues at the frontiers of fusion energy science 3. Provide fusion energy scientists with the opportunity to influence the

development of high- performance computing 4. Provide the FES community with plans for development of future high-

performance computing capability in collaboration with ASCR• Forum for the exchange of ideas from multiple scientific disciplines

−Fusion Energy, Computer Science, and Applied Mathematics − Identify and address challenges in fusion energy science as well as in

high-performance computing with an emphasis on the use of extreme-scale computing to accelerate progress in fusion energy http://extremecomputing.labworks.org/fusion/PNNL_Fusion_final19404.pdf



Advances in Computational Science and Next-Generation Computing Essential for

Leadership in Science and Technology• Future requires certification of complex

engineered systems and analysis of climate mitigation alternatives with quantified levels of uncertainty− New fuels and reactors− Stewardship without nuclear tests− Carbon sequestration alternatives− Regional climate impacts

• Broader application of next-generation computing can provide tremendous advantages for fundamental science and industrial competitiveness− Renewable energy and energy storage− Prediction and control of materials in

extreme environments− Clean and efficient combustion in

advanced engines

International Competition in HPCChart shows most capable

system for each year


Recent LCF-Enabled Simulations Provide New Insights into Nature of Plasma Turbulence

•Teraflops-to- petaflops computing power have accelerated progress in understanding plasma turbulence heat losses−Multi-scale simulations accounting for fully global

3D geometric complexity have been carried out on DOE-SC Leadership Computing Facilities

−Excellent Scalability of Global PIC Codes enabled by strong ASCR-FES collaboration in SciDACproto-FSP CPES projects

•Exascale-level production runs are needed to enable running codes with even higher physics fidelity and more comprehensive and realistic integrated dynamics−Current petascale-level production runs

on ORNL’s Jaguar LCF require 24M CPU hours (100,000 cores × 240 hours)

• FSP Applications Must be Prepared to Take Advantage of Coming Hybrid Hardware

Mission Importance:Fusion reactor size and costare determined by balance

between loss processes and self-heating rates


60% of ASCR LCF Resources Awarded through INCITE Program

Fusion applications are some of the most “ready” to effectively exploit

LCF resources to further their science goals

05

10152025303540

2006 2007 2008 2009

Num

ber o

f IN

CIT

E

Proj

ects

Year

Industry

Fusion

Engineering

Earth Science

Computer Science

Chemistry

Biology

Physics

Materials


• Plasma physics received the largest fraction in 2010INCITE (Innovative and Novel Computational Impact on Experiment)

DoE-SC Office Advanced Scientific Computing Research (OASCR) Program and the Fusion Simulation Project • ASCR has three basic thrust areas:

– Research• Applied Mathematics, Computer Science, Integrated Network Environments

– Facilities• NERSC (capacity), LCF’s (capability), ESnet

– Cross-cutting projects• Scientific Computing Through Advanced Computing (SciDAC), Innovative

and Novel Computational Impact on Experiment (INCITE), Multi-scale Mathematics Initiative

• ASCR Strategic Directions for 2010 and beyond:W. Brinkman, Director, DoE-SC, ASCAC, Aug., 2009:− Advance the state of the art in computational capability− Develop tools and methods for harnessing that capability − Bring this capability to bear on scientific questions with national need:

e.g., Fusion – “advances in computing power move fusion simulations closer to real world conditions and help engineers improve efficiency and design control systems”

FSP will directly benefit from ASCR activities & vice versa– a true partnership is essential


Co-Design Teams: A Better Way to Successfully Build Applications

FSP targeting a co-design team structure & will work with ASCR and FES to communicate lessons learned and best practices to other co-design science teams

Code Project Focus

Science Driver

Model DevelopmentPhysics Component 1Physics Component 2Physics Component NCoupled Multi-physicsValidation

Code Architecture

Integration FrameworkBuild and Testing EnvironProgramming ModelResiliencePerformance/AccelerationScalabilityDynamic ParallelismTask/Data DistributionSQADocumentationMath LibrariesI/O

Enabling Methods

DiscretizationSolution AlgorithmsVerificationSolution AdaptivitySensitivityUQOptimizationUpscaling MethodsTime Acceleration

System Architectecture

Apps MicrokernelsPerformance ModelsKernel SimulatorsNode ArchitectureNetwork ArchitectureI/O ArchitectureRuntime ArchitecturePower Efficiency

Science Delivery

RequirementsChallenge ProblemsAssessmentData AnalyticsVisualizationData ManagementSteering & Workflow


ASCR is pushing to formalize this concept

FSP Will Make Unique Contributions to the ASCR Program• The FES community is well-positioned to be a major applications area

for demonstrating the benefits of exascale computing. (ref. D. Kothe’spresentation):“Computational fusion science projects on the ASCR LCF platforms are demonstrably leading the overall computational science community”

• FES advances in models, algorithms, and software have and will continue to demonstrate “applications readiness” of benefit to the ASCR mission “to develop the algorithms, computer programs and hardware that advance scientific research”

• FES can “show the way” for other applications domains by its prominent role in cross-cutting ASCR-led HPC programs SciDAC (Scientific Discovery through Advanced Computing) and INCITE (Innovative and Novel Computational Impact on Experiment)

• Positive impact of availability of impressive suite of well-diagnosed FES experimental facilities & associated large foundational data sets aligned with FSPhelp drive Uncertainty Quantification and R&D involving sensitivity analysis


FSP -- Strategic Opportunity to Accelerate Scientific Progress in Fusion Energy Sciences

• Need reliable predictive simulation capability for BP/ITER (especially in the US)

• Powerful (“Leadership Class”) computational facilities moving rapidly toward petascale and beyond

• Interdisciplinary collaborative experience, knowledge, and software assembled over the course of nearly a decade under SciDAC plus OFES and OASCR base research programs in the US


FSP Project Planning• Requirements for a two year (Aug 2009 to July 2011) detailed “planning

study” was specified in a DoE Request for Proposal• FSP team funded ($4-6M) to carry out the study consists of:

• 6 national labs – PPPL, ORNL, LANL, LBNL, LLNL, ANL9 universities – Chicago, Columbia, Lehigh, MIT, New York, Princeton,

Purdue, Texas, UCSD 2 companies – GA, Tech-X

• Current “project definition” phase managed as a project– Includes specification of FSP program scope, planning and deliverables

Work Breakdown Structure (WBS)Targeted goals, schedules, milestones, responsible groups

– Build on “lessons learned” from other major scientific software development projects

For example, Advanced Simulation Computation Projects advanced at National Laboratories associated with the stockpile Stewardship Program

Performance, reliability, and safety—particularly in the absence of full-scale testing


US advances in 21st Century science, technology, and engineering requires leadership incomputational science and high performance computing ” exascale” applications & technology.

FSP Planning and Management Teams• FSP Planning Team:W. Tang (PPPL/PU) D. Batchelor (ORNL) H. Berk (IFS)J. Brooks (Purdue U.) J. Cary (Tech-X/U. Colorado) V. Chan (GA)C.S. Chang (NYU) P. Colella (LBNL) L. Diachin (LLNL)P. Diamond (UCSD) M. Greenwald (MIT) D. Keyes (Columbia U.)D. Kothe (ORNL) A. Kritz (Lehigh U.) W. Nevins (LLNL)A. Siegel (ANL/U.Chicago) X. Tang (LANL) G. Tynan (UCSD)

• FSP Managemen Team:Director (W. Tang, PPPL)Deputy Director for Plasma Science (M. Greenwald, MIT)Deputy Director for Advanced Scientific Computing (D. Kothe, ORNL)Science Drivers (Lead, A. Kritz, Lehigh U.) with Supporting TeamFrameworks/Physics Integration (Lead, J. Cary, Tech-X) with Supporting TeamExperimental Validation (Lead, M. Greenwald, MIT, Co-Lead, V. Chan, GA) with

Supporting TeamAdvanced Physics Modules (Lead, X. Tang, LANL Verification Lead, L. Diachin,

LLNL) with Supporting TeamProject Managers (R. Strykowsky, PPPL, Doug Hudson, ORNL)


FSP Planning Effort• The FSP planning effort has an active outreach to the theory, modeling

and experimental communities−Project planning must involve input from broad range of scientists

associated with US and international communities in FES −Applied math and computer science communities in ASCR to help define

scientific priorities and establish mechanisms for productive collaborations • Seminars and discussion at fusion energy science research centers

−GA, MIT, Maryland, LLNL, ANL, LANL, ORNL, …..JET• The FSP planning team has posted on its national web-site a FAQ

(Frequently Asked Questions) section http://www.pppl.gov/fsp/

• Planning team and management teams welcome input, comments and suggestions from the FES and ASCR communities−ASCR (Advanced Scientific Computing Research)

• A DOE-Office of Science review will be held at the end of the 2-year planning study (shortly after July 2011)


Organization of Planning Effort• The project team is organized into five planning areas 1) Science applications (A. Kritz, M. Greenwald)2) Software architecture, frameworks and integration (J. Cary)3) Advanced modules and code verification, applied math and

advanced algorithms (X. Tang, L. Diachin)4) Experimental validation (M. Greenwald, V. Chan)

5) Management (W. Tang, D. Kothe, M. Greenwald)

• Deliverables associated with Planning 1) Identification of an initial set of science applications/drivers best suited for

development. 2) The definition of the software architecture and infrastructure required and a

plan for its development. 3) The identification of the initial set of required physics components and a plan

for their development and integration4) The definition of the requirements and a plan for verification and validation 5) The definition of the management structure and a program plan for the FSP


• Science drivers are a set of compelling scientific problems chosen to focus FSP development

• Science drivers serve to– Motivate & prioritize production of useful tools for broader fusion community– Define and exercise the needed wide range of capabilities

• Key objective of FSP– Accelerate the predictive understanding and control of magnetic fusion

experiments through the integrated modeling of multi-scale physics • By exploiting the computational power of emerging peta-scale, and

eventually exa-scale, computers – Stakeholders — major facilities, small experiments, ITER, advanced design

projects, analytic theory, modelers, advanced simulation • Goal of the FSP science drivers team

– Articulate key issues that must be addressed through integration of multi-scale physical phenomena in fusion experiments

• Science objectives can be achieved only through close interaction between the plasma theory and experimental communities

– Collaboration will ensure the fidelity of each physics module and the validity and relevance of the simulation results

FSP Science Drivers Objectives and Goals


• Ultimately, the Fusion Simulation Project is expected to produce a kind of “numerical tokamak” or “fusion system simulation’’ to compute:

– All relevant physical phenomena, e.g., particle distribution functions, turbulence spectra, large-scale instabilities, …

• From the magnetic axis to the plasma-facing wall• Bridging the gap from microsecond to whole discharge time scales

• It is expected that different researchers will use FSP in different ways:– A few heroic simulations will be carried out for whole-device modeling at the

highest level of physics fidelity– Validated reduced models will be used for more rapid whole-device scenario

simulations including start-up, shut-down and performance optimization– Some users will focus on particular parts of the plasma, such as the

scrape-off layer, or particular physical phenomena, such as turbulence• FSP will enable focused research to be carried out in the context of

self-consistent whole-device simulations– Some researchers will use FSP simulations to analyze and understand

existing tokamak discharges and to improve design of new fusion experiments and facilities

Vision for FSP Science Drivers


Key Scientific Challenges for Burning Plasmas

• Disruptions: Large-scale macroscopic events leading to rapid termination of plasma discharges– Avoid or mitigate because ITER can sustain

only a limited number of full-current disruptions– Pedestals: Formation of steep spatial gradients leading

to transient heat loads in plasma periphery(divertor region)

– Predict onset and growth because pedestal heightis observed to control confinement

– Predict frequency and size of edge localized mode (ELM) crashes to mitigate erosion of divertor and plasma-facing components

• Tritium migration/retention and impurity transport• Performance optimization and scenario modeling• Plasma feedback control

– Burning plasma regime is fundamentally new withstronger self-coupling and weaker external control

ELMs in MAST

Plasma disruption in DIII-D


• Examine ReNeW, Exoscale, FESAC (Priorities, Gaps and Opportunities) and past FSP workshop reports and generate a list of science drivers

• Establish and apply criteria to prioritize and sequence science drivers– Evaluate component and framework plans in order to determine criteria and

prioritization for both near term and longer term science drivers• Get feedback from the community and get buy-in

– Presentations and fact finding at fusion facilities and other research groups– Interact and iterate with other FSP activities (e.g., framework and

components) to get feed back• Identify plans to acquire, develop and complete required physics modules

• Science drivers team will need to develop a plan and mechanism to monitor FSP progress in advancing science

• Document and disseminate science driver planning• Establish independent review process for components and framework• Science Drivers Team Activities

– Minutes of conference calls available on Science Drivers Wiki

http://fspscidri.web.lehigh.edu/index.php/Main_Page

Tasks for Science Drivers Team


Example Questions to be Addressed by FSP-ReNeW• How rapidly will large reactor plasmas rotate,

and what effect does rotation have on performance limits?

• How will fully steady-state reactor-relevant wall affect the core plasma performance, and what approaches can be used to control the erosion, transport, and redeposition of wall material?

• How do we adequately fuel the plasma and remove fusion reaction byproducts and how do we manage and control the tritium inventory in a steady-state reactor?

• How do the energetic particles and/or the background plasma turbulent transport, MHD equilibrium and RF driven currents respond to energetic particle modes and/or turbulence?

• Can we successfully and accurately predict the approach to a disruption event and initiate control techniques to avoid the disruption or, if a disruption becomes unavoidable, can we mitigate the consequences?


Example Questions to be Addressed by FSP-ReNeW

• Can we accurately identify the physics trigger(s) for L-H transition, onset of edge and internal transport barriers, and the control of ELM crashes?

• Can we predict the coupling of radiofrequency, beam and pellet sources across the pedestal?

• How do alpha particles affect radiofrequency wave propagation and absorption and what possible transport effects from radiofrequency waves on alphas may be expected?

• How can modeling bridge disparate spatio-temporal scales that link different aspects of high-performance, steady-state burning plasma physics?

• Can the reliability of each component, as well as the framework connecting components, be improved to the point where integrated models can be used routinely to plan and analyze burning plasma experiments in ITER?


Examples of Gaps

• Structure, size and stability of H-mode pedestal– Need to fill gap in the spectrum between fine scaled turbulence and large

scale instabilities• Core turbulence and transport on transport time scales

– Bridge turbulence and transport time scales– Turbulence in the presence of magnetic islands– Full electromagnetic treatment of turbulence and transport

• Integrated whole-device modeling – Tight coupling among heterogeneous components with high physics fidelity

• From magnetic axis to first wall, from startup to shutdown

• Integrated scrape-off layer – divertor – plasma-wall interaction – Lack of reduced models for cross field transport– Coupling of atomic physics, plasma-wall interactions together with turbulence

• Onset, consequences and mitigation of disruptions– Full non-linear treatment of disruptions and their effects on plasma profiles

• Wave particle resonances – Self consistent coupling of alpha particle power and

plasma transport in a burning plasma– Interaction of flow shear and energetic particle induced perturbations on the

radial electric field due to the action of fast ion driven instabilities


• Criteria for selecting FSP scientific problems would include: – Clear need for multi-scale, multi-physics integration:

• Topic should require the capabilities envisioned for the FSP –beyond scope of the current modeling programs

• Solving the scientific problem should demonstrate that FSP “is more than the sum of its parts”

– Importance and urgency:• Importance measured against the OFES mission to create knowledge base

needed for an economically and environmentally attractive fusion energy source• Should answer questions or address scientific issues integral to knowledge base• Urgency should take into consideration schedules, dependencies and

critical paths for program elements that the FSP would support– Readiness and Tractability:

• Underlying physics along with the required applied math, computer science and computing platforms should be sufficient to initiate work at outset of FSP

• Need for the FSP to impact ongoing research at an early date should also be considered

• Some reasonably clear path toward solving the research problem, posed by the topic, should be envisioned

– Opportunity to open up new lines of research:• Attacking the problem should offer possibility of new insights or potential

breakthroughs, particularly those not accessible by other means

FSP Science Drivers Selection Criteria


Science Drivers

• Integrated Scrape-off Layer, Divertor, Plasma Wall Interactions– Turbulent perpendicular transport, classical parallel transport physics,

atomic physics and plasma wall interactions– Issues include:

• Power and particle loads on divertor plates and first wall impurity generation, erosion, redeposition and transport;

• Fuel deposition/tritium retention; fueling and density limit; • Choice of material; temperature for Plasma Facing Components; • Effects of off-normal events; • First wall conditioning dust generation; transport and effects; • Recycling and pumping, flow boundary conditions; • Boundary conditions for temperatures, density and flow at last closed flux surface

including turbulence spreading between sol and pedestal

• Structure, Size and Stability of H-mode Pedestal– Nonlinear MHD, turbulence, ELMs and other relaxation mechanisms– Issues include:

• Long-time solutions to turbulence problem• Scale separation between nonlinear MHD and nonlinear electromagnetic turbulence• Finite Larmor radius effects in narrow pedestal region


Science Drivers

• Nonlinear Turbulence and MHD on Transport Time Scales– Electron and ion channels, particle and momentum transport using

gyro-kinetic transport physics extended to transport time scales– Issues include:

• Evolution of equilibrium profiles of temperature, density and rotation frequency • Self-generated and driven rotation • Interaction between turbulence and magnetic islands or MHD modes• Local versus non-local turbulence and transport

• Integrated Whole-Device Modeling– Multi-physics, multi-scale integration to compute turbulence spectra,

onset and consequences of large-scale instabilities, and particle distributions as a function of space and time

– Issues include:• Strong interactions between the different physical processes

that occur on a wide range of different time and space scales• Used for scenario modeling (including start-up and shut down), optimization of

discharge parameters (such as maximizing fusion power production) and the development of discharge control techniques

• Self-consistent simulations are also used for testing theoretical predictions and validation of models by comparison with experimental data


Provisional Science Drivers

• Detection Avoidance and Mitigation of Disruptions– Mitigation requires integration of MHD instability modeling, atomic and

radiation physics, plasma wall interactions, and relativistic electron physics– Issues include:

• Reliable predictions of triggering conditions leading to onset of disruption• Reliable predictions of nonlinear evolution of disruptions including rapid evolution

of temperature, current density, fast particle and density profiles• Interaction with plasma facing wall

• Wave-Particle Resonances– Wave coupling and propagation, wave-particle interactions, nonlinear plasma

responses, generation of waves from non-Maxwellian distributions– Issues include:

• Antenna or launcher coupling; wave propagation and damping; RF sheaths• Plasma response — heating, current drive; flow drive;

loss or redistribution of fast particles;• Nonlinear stability of energetic particle modes• Computation of saturated mode amplitudes• Anomalous velocity space diffusion; impact on alpha heating• Effects of background turbulence


Science Driver Questions

• What is need for integration or what is the integration aspect?

• What is the physics challenge?

• Why is the science driver important and urgent?

• To what extent is science driver ready to be addressed and tractable?

• Will the science driver provide an opportunity to open up new lines of research?

• What are examples of issues to be addressed?

• Why can existing SciDAC projects not address the questions associated with the science driver?

• Why will it be possible to address these questions can by FSP?


Integrated SOL-Divertor-PWI• Scope

– Simulations span collisional plasma at bottom of pedestal, scrape-off-layer (SOL), the divertor, and the front surface of the first wall

• At material boundary, include only the first few microns of the first wall• Modeling includes phenomena such as recycling, sputtering, and fuel retention

• Importance and urgency– Fusion reactors represent a tremendous extrapolation

in power loading and pulse length – Simultaneous matching of SOL and core dimensionless parameters

requires integrated edge modeling– Improved understanding and predictability will affect operating limits,

choice of materials, geometry, and operating temperature for reactors• Power and particle loads on divertor plates and first wall • Impurity generation, erosion, redeposition and impurity transport • Fuel deposition/tritium retention • Effects of off-normal events • Dust generation • Fueling and density limit • Recycling and pumping • Production and effects of RF sheaths


Integrated SOL-Divertor-PWI

• Physics Challenges – Anomalous cross-field transport has a dominant influence– Lack of clear spatial scale separation; large amplitude of fluctuations

high degree of intermittency; coupling to atomic and material physics– Missing physics includes

• Turbulent transport in scrape off layer – More complete fluid (short mean-free-path) codes — especially with

correct vorticity equations– Matching fluid and gyrokinetic results at appropriate limits (especially vorticity) – Proper calculation of self-consistent E field and self-generated flows

(requires keeping higher order terms) – Proper treatment of open field lines (no flux surface averaging)

• Issues with ordering: scale separation and large fluctuations• Fully integrated divertor physics including private flux region and X-point • Sheath physics including transmission factor for heat loads, also RF sheaths • Plasma-wall interactions • Multi-scale material physics including chemistry and morphology • Self-consistent solution in realistic 3D geometry


Integrated SOL-Divertor-PWI• Tractability

– Current research is yielding better sets of equations, which begin to address the shortcomings

– Experimental measurements are quite good in the region of the plasma (and could be significantly improved without enormous expense)

– In situ diagnostics for surface analysis will be coming on line in a few years• More parameters can be measured and spatial coverage is better than in core

• New lines of research– Success will provide greatly improved models for the plasma

at the boundary between open and closed field lines• Temperature, density and flow profiles • Viscosity and turbulence spreading • Fueling and density limit

– Necessary for understanding the behavior of the core plasma and for whole device modeling

• Integration– Multi-physics phenomena including collisional and turbulent transport,

neutral transport, atomic physics and material physics– Generation of impurities via sputtering through RF-produced sheaths

broadens the integration still farther– Wide range of spatial and temporal scales are involved


Integrated Whole-Device Modeling

• Need for integration:– Simulation of plasma evolution requires self-consistent combinations of

all relevant physical phenomena in tokamak discharges

• Physics and computational challenge– Compute turbulence spectra, onset and consequences of large-scale

instabilities, and particle distributions as a function of space and time using self-consistent combinations of all of the relevant physical phenomena

– Strong interactions occur between the different physical processes that occur on a wide range of different time and space scales

• All relevant modes of turbulence and channels of transport must be included along with relevant large-scale instabilities for self-consistent whole-device simulations

– Transport (driven by turbulence, charged particle drifts and collisions) affects plasma profile evolution, which, in turn, impacts large-scale instabilities, which then alter plasma profiles, which then alter transport

• All interacting physical phenomena must be included in order to make stand-alone predictions that go beyond limited extrapolations from existing experimental data


Integrated Whole-Device Modeling• Importance and urgency

– Clear and compelling need for modern whole-device integrated modeling code • Span entire plasma from magnetic axis to first wall • Bridge gap between the microsecond time scale of basic physics phenomena and

hundreds of seconds time scale for duration of plasma discharges

• Readiness and tractability– Recent FSP prototype SciDAC projects focus on combining pairs of

regions or physical phenomena in high-performance computations– Given progress in FSP prototype projects, as well as the mature capability

developed for comprehensive integrated modeling using reduced models• Integration of all of the physics components is tractable in a comprehensive

framework for high performance, whole-device modeling

• Opportunities for new lines of research– Provides opportunity to investigate the physics of transport,

computed self-consistently with turbulence, sources and sinks • In all regions of the plasma including the scrape-off layer

– Investigate the interactions between turbulence, MHD instabilities and fast particles driven by radio-frequency and neutral beam injection

– Scenario modeling and detailed comparisons with experimental data using comprehensive whole-device simulations with high-fidelity physics

• Significant improvements to our basic understanding of physical phenomena


Integrated Whole-Device Modeling• Examples of issues to be addressed

– Reliable, validated, self-consistent predictions of evolution of plasma profiles from magnetic axis to the plasma-facing wall in variety of tokamak discharges

• Include predictions of height and width of internal and edge transport barriers• Include turbulence spectra and the distributions of all particle species

– Complete, reliable scenario modeling from discharge start up to shut down• Include free-boundary evolution of plasma shape, evolution and coil currents

– Prediction, control and mitigation of the onset and consequences of large-scale instabilities (such as sawteeth, ELMs, NTMs, and RWMs)

• Include interactions with turbulence, transport, sources, sinks and fast particle distributions

• Why these questions cannot be addressed by existing SciDAC projects– Required capabilities involve strong interaction among different physical

phenomena and different parts of the plasma • Consequently, these issues cannot be addressed by focusing on

only one or few physical phenomena or one region of the plasma.

• Why these questions can be addressed by FSP– Capabilities require integration of high-performance components central to

FSP in order to investigate the strong interaction among different physical phenomena and different parts of the plasma with high physics fidelity


FSP Web Sites and Wikis• Work associated with the US Fusion Simulation Project is described on

the project home Web page and on wikis associated with the project

• Wikis is a website that allows the easy creation and editing of any number of interlinked web pages

• Wikis are simple web pages that groups of researchers can edit together

• Home Web page: http://www.pppl.gov/fsp

• Science Driver Wiki: http://fspscidri.web.lehigh.edu/index.php/Main_Page

• Framework Wiki: https://ice.txcorp.com/trac/fspfrmwrkplan/wiki

• Components Wiki: http://fspcomp.web.lehigh.edu/index.php/Main_Page

• Validation Wiki: http://www.psfc.mit.edu/FSP-Validation/index.php/Main_Page


FSP Planning Workshop (March 15-18, 2010)• Nearly 50 participants from broad range of institutions and disciplines:

−Boundary – T. Rognlien (LLNL), CS Chang (NYU), M. Greenwald (MIT), S. Krasheninnikov (UCSD), D. Knoll (LANL), D. Stotler (PPPL), L. Berry (ORNL)…

−Pedestal – P. Snyder (GA), X. Xu (LLNL), S. Parker (U.Colorado), J Cummings (CalTech), S. Klasky (ORNL), J. Hittinger (LLNL), ….

−Core Profiles – W. Nevins (LLNL), P. Diamond (UCSD), J. Candy (GA), C. Holland (UCSD/GA), S. Parker (U. Colorado), …..

−Wave-Particles (EP&RF) – R. Nazikian (PPPL), P.Bonoli (MIT), H. Berk (IFS), G. Fu (PPPL), D. Spong (ORNL), ….

−Disruptions – S. Kruger (Tech-X), C.Sovinec (U. Wisconsin), A. Reiman(PPPL), V. Chan (GA), S. Humphreys (GA), …..

−Whole Device Modeling – C. Kessel (PPPL), T. Casper (ITER), D. McCune (PPPL), D. Batchelor (ORNL), G. Bateman (Lehigh), …..

•AGENDA with Presentations: https://ice.txcorp.com/trac/fspfrmwrkplan/wiki/FspPlanningAgenda


https://ice.txcorp.com/trac/fspfrmwrkplan/wiki/FspPlanningAgenda

FSP Integrated Planning Teams• Each team focuses on a particular science driver

• Each team is led by a modeler and an experimentalist and includes at least one computer scientist or applied mathematician

• Charge to the integrated planning teams:−Preparing the overall logic, and the specifics to the extent possible, of the

implementation plans for each of the FSP science drivers

−Physics components that need to be integrated in order to achieve the goals associated with the science driver

−Plans for adapting older components and as well as plans for developing new components that are required

−Analysis of the requirements for composition of the physics components (including data exchanges and algorithms) and for the full workflow (including such items as visualization and job preparation)

−Plans for validation of critical physics associated with the science driver

− Needs for collaboration with other efforts within the FSP as well as requirements for work to be accomplished outside the FSP


Charge to FSP Integrated Planning Teams cont.• Charge to the integrated planning teams:

−A projected schedule of the work to be carried over a 15 year time period and a realistic estimate of resources required •

− Suggested high-level goals and milestones, at roughly the 2, 5, 10 and 15 year marks.

−The process of developing these plans should begin with the descriptions and roadmaps of the FSP science drivers as summarized in the March workshop report.

− Interim report, covering the technical aspects of each plan due at the end of July, 2010 and Final report, including schedules, resource requirements and milestones due at the end of September, 2010

• Charge to the Integrated Planning Teams details the expectations with regard to the plans developed by the teams, particularly in terms of components, frameworks and validation


Verification and Validation Challenges

• Establishing the physics fidelity of modern plasma science simulation tools demands proper Verification & Validation (V&V)

−Reliable codes demand solid theoretical foundations and careful experimental validation

• Verification assesses degree to which a code (both in the advanced direct numerical simulation and reduced models categories) correctly implements the chosen physical model− More than “essentially a mathematical problem”

For example, accuracy of numerical approximations, mesh/space and temporal discretization, statistical sampling errors, etc.

• Special emphasis should be placed on code verification via− Comparisons with theoretical predictions

Threshold/onset conditions for instabilities; weakly nonlinear evolution; nonlinear saturation estimates

− Cross-code benchmarking (codes based on different mathematical formulations/algorithms but targeting the same generic physics)

Finite difference, finite elements, spectral methods, implicit schemes, and/or models as Particle-in-Cell, Vlasov/Continuum, Hybrid PIC-Fluid

• Validation assesses degree to which a code (within its domain of applicability) “describes the real world,”


Collaboration With Experimental Facilities• Strong collaboration with experimental facilities essential for

validation of FSP codes − Regarded as significant value to the facilities for planning and

interpretation of experiments• Draft agreement with major facilities (DIII-D, C-MOD, NSTX) defining:

− General principles for collaboration and intellectual property sharing− Proposed mechanisms for short-term and long-term planning− Roles & Responsibilities for the FSP and for experimental teams in their

collaboration− Lessons learned from the major experimental facilities useful in planning

the FSP research program– e.g., open community research forums• The existing collaboration engagement agreements for and

approaches to research governance used by the facilities provide a proven model for the FSP.

• Elements for Success (from past experience) dictates:-- ongoing partnerships and mutual interactions institutionalized through formal agreements; and -- regular participation in planning and reporting activities with cross-membership in planning groups


The FSP Will Make Unique Contributions to Fusion Program• FSP will address multi-physics and multi-scale problems that are now

often treated in isolation − Scientific discovery of new phenomena which emerge only with integration

• FSP will carry out a rigorous and systematic validation program, in collaboration with experiments− Intent to put models on firmest and most realistic possible foundation

• FSP will develop predictive models which improve our capabilities for reliable scenario modeling− Especially for ITER, and for design of future machines, e.g., DEMO

• FSP incorporates powerful HPC capabilities to help accelerate scientific understanding and modern software engineering approaches

• Leverages ongoing activities (theory, experiment, modeling in FES and applied math, computer science in ASCR)− Develop unprecedented simulation capabilities.

• Embodies our state of knowledge in a suite of advanced codes under a unified framework and made widely available to the FES community

• Train the next generation of simulation/modeling-oriented CS, Applied Math and applications-oriented computational scientists and engineers


Fusion Simulation Program (FSP) Scientific and ...Fusion Simulation Project (FSP) Mission and Vision...

Documents

Transcript of Fusion Simulation Program (FSP) Scientific and ...Fusion Simulation Project (FSP) Mission and Vision...