university of colorado at boulder santa fe campus center for integrated plasma studies implicit...
DESCRIPTION
University of Colorado at Boulder Santa Fe Campus Center for Integrated Plasma Studies IMP Algorithm Fluid equations –Quasineutral, no displacement current –Electrons are massless fluid (extensions possible)TRANSCRIPT
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Implicit Particle ClosureIMP
D. C. BarnesNIMROD Meeting
April 21, 2007
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Outline• The IMP algorithm
– Implicit fluid equations– Closure moments from particles– f with evolving background– Constraint moments
• Symmetry – Energy conservation theorem– Conservation for discrete system– absolutely bounded, no growing weight problem!
• Present restricted implementation• G-mode tests• Future directions
f 2
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
IMP Algorithm
• Fluid equations– Quasineutral, no displacement current– Electrons are massless fluid (extensions possible)
Pen e BJBuE
0
0
uu
u
eeeeee PPt
P
nDtDn
eneJuu
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
IMP Algorithm• Fluid equations
– Ions are massive, collisionless, kinetic species– Use ion (actually total) fluid equations w. kinetic
closure
P�
iePMnDtDMn BJgu
BJEB
0
1;t
0
vBvF
xv f
Mef
tf
uvwwww ;P fdMi�
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
IMP Algorithm• f particle closure algorithm
– Background is fixed T with n, u evolving– Particle advance uses particular velocity w (very
important)
ev
nffff vw
T
T22 2/
32
;~
Axx
uwBww
uwx
~12
nn
vMe
dtddtd
T
wwwAA fdnMn
nTkMn
Bi ~1~;~P�
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
IMP Algorithm
~
Aw
xu:wwu ~1~
~11
2vdtd
T
ff /~~
• Note: (perturbation) E does not enter closure directly
• There is some kind of symmetry between advance and A~
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Constraint Moments
• With infinite precision and particles, should have …
vef
SWfd
SWfd
T
vw
T
pppppT
ppppT
T
2 2/3
2/
2
~0~
~0~
22
xxwww
xxw
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Constraint Moments
• Satisfy constraints by shaping particle in both x and w
corrections
sfdsfd
ssW
w
wTwT
ppwppp
www
wwxxwx
0
,~,~
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Constraint Moments
• Using Hermite polynomials, find
vfs
T
ppTpw 2
1, wwwww
• Projection of weight equation is then
xu:wwu
Awxu:wwuwww
v
vsd
dtd
T p
Tpw
p
p
2
2
1
~1,~
~11
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Constraint Moments
• …and, closure moment has symmetric form
ppppTn
ppppppn
ppwTppn
sWvCsWC
sfdsCn
xxxxww
wwwwwxxA
~~
,~ ~
2
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Symmetry Leads to Energy Integral
Aux
x
~
log122
2
0
22
ndM
nnvMPBMnuMnddtd
Te
e
Usual fluid w. isoT ions Interchange w. closure
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Symmetry Leads to Energy Integral
ppp
Tn
ppppTn WvCM
dtd
dtd
WvCM
ndM
~2
~~1log
~
...
22
2
Aux
Usual fluid w. isoT ions Closure energy
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Symmetry Leads to Energy Integral
.~~1log
log122
2
2
0
22
constWvCM
nnvMPBMnuMnd
ppppTn
Te
e
E
x
EE
nnvMPBMnuMnd-
WvCMWvCM
Te
e
ppp
Tn
ppppTn
log122
~2
~~1log
2
0
22
22
2
x
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Symmetry Leads to Energy Integral
• r.m.s. of particle weights absolutely bounded
• Stability comparison theorem– Kinetic system more stable than isoT ion fluid
system– But only for marginal mode at zero frequency
• This is absolutely the most important point!
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
IMP2 Implementation
• 2D, Cartesian• TE polarization
– B normal to simulation plane– E in plane
• Linearized, 1D equilibrium• Uniform T
0|| k
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Time Centering
• Moments use Sovinec’s time-centered implicit leap-frog– Direct solve
• Particles use simple predictor-corrector– Present, use full Lorentz orbits w. orbit
averaging (Anticipating Harris Sheet or FRC calculations)
– Iteration required (3 – 5 typical count)
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Time Centering
t u u
n,T,B n,T,B
S.I. MHD step Implicit induction
equation step
Leap-frog particle substeps give orbit averaged A ~
Particles use average of u(depends on A, so need iterate)
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Space Differencing
• Use Yee mesh, w. velocity with B
* * *
* * *
* * * ux,Ey
uy, Ex
Bz ~ ~
n,Pe ~ ~ ~
~ ~
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
G-Mode Tests1
yx
00 0.1
Contours of ux for Roberts-Taylor G-mode
g
• Following Roberts, Taylor, Schnack, Ferraro, Jardin, …
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
G-Mode Tests
• Two series– Low Hall stabilized – with and w/o closure– High gyro-viscous stabilized (Hall turned off)
• Numerical parameters– Nx x Ny = 30 x 16– 9 – 25 particles/cell– Typically 100 particle steps/fluid step
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
0 0
.5 1
1.5
2 2
.5 3
3.5
4 0 0
.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
x
y
pertu
rbed
den
sity
min
= -5
.694
E+16
m
ax=
5.6
94E+
16
0 0
.5 1
1.5
2 2
.5 3
3.5
4 0 0
.5
1 1.5
2 2.5
3 3.5
4
x
y
pertu
rbed
den
sity
min
= -2
.864
E+18
m
ax=
2.8
64E
+18
Perturbed density
0
0.2
0.4
0.6
0.8
1
1.2
0 0.5 1 1.5 2 2.5 3 */ g
M
g/g M
G-Mode Tests
• Low – 0.02– B = 6.0 T– n = 2. x 1020 m-3
– g = 1. x 1012 m/s2
– Ln = 120 m– T = 8.94 keV
– = 2.28 mm
• Arrow marks k = 0.15
Fluid only
Closure
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
G-Mode Tests• High
– 1.0– B = 0.482 T– n = 5.78 x 1019 m-3
– g = 2.7 x 108 m/s2
– Ln = 10 m– T = 10 keV– g/i = 2.25 x 10-4
– = 2.99 cm• Arrow marks
k = 0.15 0
0.5
1 1
.5 2
2.5
3 3
.5 4
0 0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
x
y
pertu
rbed
den
sity
min
= -5
.694
E+16
m
ax=
5.6
94E+
16
0 0
.5 1
1.5
2 2
.5 3
3.5
4 0 0
.5
1 1.5
2 2.5
3 3.5
4
x
y
pertu
rbed
den
sity
min
= -2
.864
E+18
m
ax=
2.8
64E+
18
Case10.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
kv*/ G_M
/gM
g/g M
Perturbed density
0 0
.5 1
1.5
2 2
.5 3
3.5
4 0 0
.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
x
y
pertu
rbed
den
sity
min
= -5
.694
E+16
m
ax=
5.6
94E+
16
0 0
.5 1
1.5
2 2
.5 3
3.5
4 0 0
.5
1 1.5
2 2.5
3 3.5
4
x
y
pertu
rbed
den
sity
min
= -2
.864
E+18
m
ax=
2.8
64E+
18
Case10.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
kv*/ G_M
/gM
g/g M
Perturbed density
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
G-Mode Tests
• Hall and gyro-viscous stabilization of fundamental G-mode observed– Stability seems consistent with Roberts-
Taylor, modified by Schnack & Ferraro, Jardin
• New, higher kx mode observed at k > 0.2 or so– Drift wave?– Present in fluid (+ Braginskii) also?
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Temperature Variation?
• Vlasov equation is linear, so superposition allowed– Superimpose number of uniform T solutions– Slightly different equation
• Mild restriction on equilibrium T– Density must depend only on potential
nnefdf wM 2/2ˆ
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Future Directions
• Short-term– Understand high kx mode– Finish manuscript
• Medium-term– Add T variation– Add full polarization, check Landau damping
• Long-term– Gyrokinetic version– Parallel implementation
University of Colorado at BoulderSanta Fe CampusCenter for Integrated Plasma Studies
http://cips.colorado.edu/
Preprints