particle simulation, gyrokinetics, turbulence and beyond w ... · particle simulation,...
TRANSCRIPT
Particle Simulation, Gyrokinetics, Turbulence and Beyond
W. W. Lee
Theory Department SeminarPPPL
April 2015
AcknowledgementS. Ethier and R. Ganesh (IPR, India)
1000
Other pioneers: Oscar Buneman
Ned Birdsall
103
104
105
106
1536 3072 6144 12288 24576 49152 98304
com
pute
pow
er (m
illion
s nu
mbe
r of p
artic
les
per s
econ
d pe
r ste
p)
number of nodes
A200$
B200$
C200$
D200$
Sequoia*
BG-Q Performance: Weak Scaling Results
Mira$
$$$$$Mira$
Mira$
*NNSA’s Sequoia (16.3 PF)
x 16 → number of cores
2014
80,000 particles/core
Fluctuation-Dissipation Theorem for Weakly Damped Normal Modes [Klimontovich ‘67]
Particle Noise in a Simulation Plasma
|eΦ(k,ws)/Te|th =1√N
|eΦ(k,wpe)/Te|th =1√
NkλDe
ε ≡ 1 + |Sk|2[1 + ξeZ(ξe) + τ + τξiZ(ξi)]/(kλDe)2 = 0
k2λ2
D� 1 k
2a2� 1
The need of a quasineutral simulation model
P. J. Catto
Nonlinear
Vlasov equation
Gyrokinetic
Density response due to
+
+
B
W. W. Lee
F = F0 + f
δf simulation for
F (x,v, t) =
NX
j=1
[x xj(t)][v vj(t)]
f(x,v, t) =
NX
j=1
wj(t)[x xj(t)][v vj(t)]
w = f/F
Computational Science & Discovery 1 (2008) 015010 W W Lee et al
Figure 5. Time evolution of (a) the ion thermal flux, (b) the particle weights, (c) the field energy,and (d) the radial modes, as well as (e) the zonal flow structure for a/ρ = 500 including both thenonlinearly generated zonal flows and the velocity space nonlinearity (VNL).
than the quasilinear value
collisions
5vti < vk < 5vti µ < 12.5v2ti/⌦i
⌫ii = 0.001vti/Ln
dRdt
= vkb + vd − @φ
@R⇥ b,
Applications to tokamak transport physics δ
[Lee, Jenkins, and Ethier, CPC 2011; Ganesh, Ethier and Lee, ICPP, 2014]
vd ⇡✓
v2k +
v2?2
◆b ⇥ @
@RlnB
dvk
dt= b ·
✓v2?2
@
@RlnB
◆ b⇤ · @φ
@R
b⇤⇡ b + vkb ⇥ @
@RlnB
↵1 + ↵2 = 1
⇢ = eX
j
[↵1pj + ↵2wj ]δ(x xj)]
dp
dt= (p w)(vd + vE⇥B) ·
dw
dt= (p w)
(vd + vE⇥B) · +
Te
Ti(vkb + vd) · E
�vd
d(p w)
dt= (p w)
Te
Ti(vkb+ vd) ·E
µB ⌘ v2?2B
⇡ const.,
↵1 =|wj |2
|pj |2 + |wj |2↵2 =
|pj |2
|pj |2 + |wj |2
and Krommes (PoP ’94)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0 200 400 600 800 1000 1200 1400
sum
w^2
/N (
red)
, sum
p^2
/N-1
(gr
een)
, sum
(p-
w)^
2/N
-1 (
blue
)
time [a/Cs]
Entropies : comparision, p-w system. p-weight not included in dynamics,dt=0.1, micell=100, arho=125
a
⇢s= 125
particle interaction
NL stage.
losing energy in the saturated state
scattering are similar physical processes !
@F↵
@t+ v · @F↵
@x+
q
m
E+
1
cv ⇥ (B0 + δB)
�· @F↵
@v= 0
E = −rφ− (1/c)@A/@t B = r⇥A
L =1
2mv2 q+
q
cv ·A
@F↵
@t+ v · @F↵
@x+
q
m
−r(φ− 1
cv ·A) +
1
cv ⇥B0
�· @F↵
@(v + q↵A/m↵c)= 0
β
v ! v +q↵m↵c
A?
@F↵
@t+ v · @F↵
@x+
q
m
−r(φ− 1
cv? ·A?)−
1
c
@Ak
@t+
1
cv ⇥ (B0 + δB?)
�· @F↵
@(v + q↵A?/m↵c)= 0
to obtain
F ⌘ F (x,v, t)
F ⌘ F (x, vk, µ/B, t)
r2A− 1
v2A
@A?@t2
= −4⇡
c
X
↵
q↵
ZvF↵dvkdµ
v ⇡ vkb+c
B0E⇥ b
E = −r(φ− v? ·A?/c)− (1/c)@Ak/@t
b = b0 + B?/B0 b0 = B0/B0 B? = r⇥Ak
@F↵
@t+
vkb− c
B0r(φ− 1
cv? ·A?)⇥ b0
�· @F↵
@x− q
m
"r(φ− 1
cv? ·A?) · b+
1
c
@Ak
@t
#@F↵
@vk= 0
µ = v2?/2 v
Tp = (mc/eB2)(@2
A?/@2t)
d
dt
⌧Z(1
2v2k + µ)(meFe +miFi)dvkdµ+
!2ci
⌦2i
|r?Φ|2
8⇡+
|rAk|2
8⇡
�
x
= 0
⌘ − v? ·A?/c
v
Lp = −(mc2/eB2)(@r?φ/@t)
k2?⇢2i ⌧ 1r2φ+
!2pi
⌦2i
r2?φ = −4⇡
X
↵
q↵
ZF↵dvkdµ
!2 ⌧ k2?v2A
b
⇤ ⌘ b+vk
⌦↵0b0 ⇥ (b0 ·r)b0 b = b0 +
r⇥ A
B0
F↵ =
N↵X
j=1
(RR↵j)(µ µ↵j)(vk vk↵j)
@F↵
@t+
dR
dt· @F↵
@R+
dvk
dt
@F↵
@vk= 0
⌦↵0 ⌘ q↵B0/m↵c
⌘ − v? ·A?/c
dR
dt= vkb
⇤ +v2?
2⌦↵0b0 ⇥rlnB0 −
c
B0rΦ⇥ b0
dvk
dt= −v2?
2b
⇤ ·rlnB0 −q↵m↵
✓b
⇤ ·r+1
c
@Ak
@t
◆
v? ·A? = 1
2⇡
eB0
mc
Z 2⇡
0
Z ⇢
0
δBkrdrd✓
(m,n) tearing modes [APS 2004, Sherwood 2005] using GTS [Wang et al., PoP 2003].
Gyrokinetic Current Densities
Jgc(x) = Jkgc(x) + J
M?gc(x) + J
d?gc(x)
=X
↵
q↵hZ
F↵gc(R)(vk + v? + vd)δ(R− x+ ⇢)dRdvkdµi'
[Qin, Tang, Rewoldt and Lee, PoP 7, 991 (2000); Lee and Qin, PoP 10, 3196 (2003).]
p↵? = m↵
Z(v2?/2)F↵gc(x)dvkdµ
p↵k = m↵
Zv2kF↵gc(x)dvkdµ
J?gc = J
M?gc + J
d?gc
J?gc =c
B
X
↵
b⇥rp↵
=c
B
X
↵
hb⇥rp↵? + (p↵k − p↵?)(r⇥ b)?
i
J
d?gc =
c
B
X
↵
hp↵k(r⇥ b)? + p↵?b⇥ (rlnB)
i
J
M?gc(x) = −
X
↵
r? ⇥ cb
Bp↵?
vd =v2k
⌦↵b⇥ (b · @
@R)b+
v2?2⌦↵
b⇥ @
@RlnB
p↵ = p↵k = p↵?
ρv - ion
Rx
b - out of the board
k2?⇢2i ⌧ 1 F ! F ! Ak ! Ak v? ·A? ! 0
r2?Ak = −4⇡
cJk
J? =c
B
X
↵
b⇥rp↵
b ⌘ B
BB = r⇥A
B = B0 + B
d
dtr2
?φ− 4⇡v2Ac2
r · (Jk + J?) = 0d
dt⌘ @
@t− c
Brφ⇥ b ·r
dp↵dt
= 0
Ek ⌘ −1
c
@Ak
@t− b ·rφ = ⌘Jk ! 0
! = ±kkvA
r2?A? − 1
v2A
@2A?@t2
= −4⇡
cJ? !2 ⌧ k2?v
2A
J? =c
B
X
↵
b⇥rp↵
r · (Jk + J?) = 0
d
dtr2
?φ+v2Ac(b ·r)r2
?Ak − 4⇡v2Ac2
r · J? = 0
Ek ⌘ −1
c
@Ak
@t− b ·rφ = 0
d
dt⌘ @
@t− c
Brφ⇥ b ·r
@Ak
@t! 0 ! 0
n(x) = n +12⇢2
t
1Tr2
?nT
0 500 1000 1500 2000 2500 3000 3500 4000 45000
0.5
1
1.5
2
2.5
3
t(a/cs)
i (cs
s2 /a)
0 500 1000 1500 2000 2500 3000 3500 4000 4500!18
!16
!14
!12
!10
!8
!6
t(a/cs)
ln(r
adia
l m
ode)
(vE× B
/cs)
ITG simulation using GTC (a/rho =125)
J? =c
Bb ⇥rp? + qnvE⇥B + qn
⇢2t
2
hr2
?vE⇥B +vE⇥B
nTr2
?nTi
p? = m
ZµFgc(x)dvkdµ
J =c
Bb ⇥r(p?e + p?i) + eni
⇢2i
2
r2
?vE⇥B +vE⇥B
niTir2
?niTi
�
VE⇥B
cs=
1
2pi⇢s
Ti
Teb⇥ x
pi⇢s ⌘ −⇢srpipi
J =c
Bb⇥r(p?e + p?i −
1
2⇢2ir2
?p?i)
@⇢m@t
+r · ⇢mV = 0
⇢m
✓@V
@t+V ·rV
◆=
1
cJ⇥B−rp
E+1
cV ⇥B = ⌘J
r⇥E = −1
c
@B
@t
r⇥B =4⇡
cJ
! = ±kvA ! = ±kkvA
⇢/L ⇠ !/⌦ ⇠ o(✏)
Discussions
parameter exchanges.