Date post: | 05-Jan-2016 |
Category: |
Documents |
Upload: | erik-dixon |
View: | 215 times |
Download: | 0 times |
Tech-X Corporation1
Boundary Conditions In the NIMROD Code
S.E. Kruger
Tech-X Corporation
D.D. Schnack
U.W. - Madison
Tech-X Corporation2
Experimental Evidence Demonstrates Important Effects of Boundary Conditions on Long-Wavelength Instabilities
• Resonant Magnetic Perturbation (RMP) Experiments– Apply perturbation to plasma using coils. Has demonstrated
suppression of ELMs
• Locked modes– Rotating islands slow down due to interaction with field errors,
grow, and cause problems
• Resistive Wall Modes– Ideal modes that are stable to a perfectly conducting wall, are
allowed to grow due to the finite resistivity of the walls
– Near ideal boundary, intrinsic fields become larger (Field Error Amplfiication) and interact with tearing modes
Tech-X Corporation3
Physics of Stability Depend Upon Time Scale of Perturbation
• Let:RMP= time scale of external perturbation
rec= reconection time scale (depends on diffusivities in resistive MHD model)
• Different orderings:ARMPrectransport Forced ReconectionArecRMPtransport Field-Error Penetration ArectransportRMP Static Field Error
Tech-X Corporation4
Different Experimental Issues in Different Regimes
• Resistive wall modes– Experimentally static
error fields known to be important
• RMP Experiments– Field-error penetration
problem • NTM Seeding
– Forced reconnection problems
Tech-X Corporation5
Taylor Problem Canonical Problem For Studying Effects of Boundary Perturbations
• Start with stable slab plasma
• Perturb boundary by adding sinusoidal displacement
• Island forms
• Well studied in literature. A sampling:
Hahm and Kulsrud, Phys. Fluids 28, 2412 1985.• Analytic studies of problem
Wang and Bhattacharjee, Phys. Fluids B 4, 1795 1992.• Analytic studies of problem with flow
Ma, Wang, and Bhattacharjee, Phys. Plasmas 3, 2427 1996. • Numerical studies of problem with flow
Fitzpatrick, Phys. Plasmas 11, 937 2004.• Analytic and numerical studies with Hall MHD
Fitzpatrick, Phys. Plasmas 10, 1782 2003.• Extended problem to study field-error penetration problem
Tech-X Corporation6
Need to specify boundary conditions consistent with primitive variable, finite-element formulation with NIMROD
• Faraday’s law:
• Finite-element formulation:
• Need to specify electric field on boundary.What is electric field consistent with displacement?
• At boundary and using Fitzpatrick’s notation for displacement:
€
∂B
∂t= −∇× E ,
€
dBq
dtαp(x)αq(x)d3x
V∫q
∑ = − αp(x)∇× Ed3xV
∫ ,
= ∇αp(x)× Ed3xV
∫ − αp(x) ˆ n × EdSS
∫ ,
€
E = −V × B +ηJ
€
E z = −VxBy = −˙ Ξ By
Tech-X Corporation7
Tangential electric field is insufficient numerically for boundary condition
• Recalling finite-element formulation:
• Normal component has no boundary condition; For Taylor problem, Bx does not have boundary condition directly
• Numerical problem that NIMROD really solves is:
• Div-B correction term changes order of equation.
• Need to specify Dirichlet boundary condition for magnetic field
€
dBq
dtαp(x)αq(x)d3x
V∫q
∑ = − αp(x)∇× Ed3xV
∫ ,
= ∇αp(x)× Ed3xV
∫ − αp(x) ˆ n × EdSS
∫ ,
€
∂B
∂t= −∇× E +κ∇∇⋅B .
Tech-X Corporation8
Summary of Boundary Conditions Needed for Consistency Within NIMROD
• Velocity Field Eq:
• Magnetic Field Eq.:
• Density Equation:
• Temperature Equation:
€
Vx = ˙ Ξ
€
E z = −VxBy = −˙ Ξ By
€
Bx =∂Ξ
∂yBy
€
n = −Ξ∂n
∂x
€
T = −Ξ∂T
∂x
Tech-X Corporation9
Code Implemented Using New Data Structures
vert: DO iv=0,rb(ibe)%mxe
tang=rb(ibe)%tang%fs(:,iv)
btangeq=SUM(rb(ibe)%beq_edge%fs(1:2,iv)*tang)
btang=SUM(rb(ibe)%b_edge%fs(1:2,iv,imode)*tang)
vnorm=dbydt*rb(ibe)%displacement%fs(1,iv,imode)
! E_tor=-V_n X B_t
rb(ibe)%e_surface%fs(3,iv,imode)=-vnorm*(btangeq)
!rb(ibe)%e_surface%fs(3,iv,imode)=-vnorm*(btangeq+btang)
! Interior nodes
IF(seam(ibe)%vtype(iv)=='v') THEN
ENDIF
ENDDO vert
Tech-X Corporation10
… but doesn’t seem to work
• Div(B) errors persist• Bug? Formulation problem?
Tech-X Corporation11
Conclusions
• Since APS various problems with new edge finite element reformulation have been found
• Loop voltage benchmark with nimuw worked well– No obvious bugs
• Resistive wall still needs the pre-processor and matrix-vector multiply work, but waiting on this benchmark to pass
Tech-X Corporation12
Development
• Regression tests have been completely redone– Nate’s stuff is out. David Wade-Stein’s stuff is in
• Much more convenient for developer’s and responsive
• Want to have in place and working for upcoming merges
• Merging is worrying me. Changes from:– Carl
– Eric
– Scott
– Val
• Unlikely to be as bad as last year, but …
• Targeting July visit so that I can do it in person