zagorodny2009

NonMarkovian Effects in Turbulent Diffusion in Magnetized Plasmas
Anatoly Zagorodny and Jan Weiland
Citation: AIP Conf. Proc. 1177, 72 (2009); doi: 10.1063/1.3253966

View online: http://dx.doi.org/10.1063/1.3253966
View Table of Contents: http://proceedings.aip.org/dbt/dbt.jsp?KEY=APCPCS&Volume=1177&Issue=1
Published by the American Institute of Physics.
Related Articles
Residual turbulence from velocity shear stabilized interchange instabilities
Phys. Plasmas 20, 012301 (2013)
Freely decaying turbulence in two-dimensional electrostatic gyrokinetics
Phys. Plasmas 19, 122305 (2012)
Magnetic turbulence suppression by a helical mode in a cylindrical geometry
Phys. Plasmas 19, 122304 (2012)
Effect of secondary convective cells on turbulence intensity profiles, flow generation, and transport
Phys. Plasmas 19, 112506 (2012)
Asymmetric chiral alignment in magnetized plasma turbulence
Phys. Plasmas 19, 112301 (2012)
Additional information on AIP Conf. Proc.

Journal Homepage: http://proceedings.aip.org/
Journal Information: http://proceedings.aip.org/about/about_the_proceedings
Top downloads: http://proceedings.aip.org/dbt/most_downloaded.jsp?KEY=APCPCS
Information for Authors: http://proceedings.aip.org/authors/information_for_authors
Downloaded 13 Jan 2013 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
Non-Markovian Effects in Turbulent Diffusion
in Magnetized Plasmas
Anatoly Zagorodnya and Jan Weilandb
a
Bogolyubov Institute for Theoretical Physics, Metrolohichna Str.14-b, 03680 Kiev, Ukraine
b
Chalmers Technical University and Euratom-VR Association, S-41296 Göteborg, Sweden
Abstract. The derivation of the kinetic equations for inhomogeneous plasma in an external
magnetic field is presented. The Fokker-Planck-type equations with the non-Markovian kinetic
coefficients are proposed. In the time-local limit (small correlation times with respect to the
distribution function relaxation time) the relations obtained recover the results known from the
appropriate quasilinear theory and the Dupree-Weinstock theory of plasma turbulence. The
equations proposed are used to describe zonal flow generation and to estimate the diffusion
coefficient for saturated turbulence.
Keywords: plasma turbulence, non-Markovian diffusion, kinetic coefficients
PACS: 52.25 Dg, 52.25 Fi, 52.35 Ra
1. INTRODUCTION
In the present paper we make an attempt to derive consistent kinetic equations for
inhomogeneous plasma in an external magnetic field. We start from the Klimontovich
equations for the microscopic phase density. Statistical averaging of these equations
gives the kinetic equations with the collision integrals expressed in terms of the
correlation functions of microscopic fluctuations. In the general case such collision
terms are time-nonlocal (non-Markovian). Usually non-Markovian effects are
disregarded assuming that the correlation times for stochastic fields producing random
particle motion are much shorter than the times of changes of distribution function
(relaxation times). Since in the case of turbulent plasmas this condition could be
violated it is reasonable to introduce time nonlocality into kinetic theory of plasmas.
The appropriate treatment shows that it can be important for the description of various
phenomena such as partial diffusion in the real and velocity space, diffusion of the
magnetic field lines, turbulent saturation, etc. (see, for example, [1-10] and references
cited therein). Keeping time-nonlocality in the kinetic equations we are led to the
Fokker-Planck-type equations with time-nonlocal kinetic coefficients. These equations
are used to generalize various versions of kinetic theory of turbulence, such as
quasilinear theory and the Dupree-Weinstock renormalized theory to the case of non-
ATTACHMENT
Markovian description of inhomogeneous plasma in an external magnetic field. In
CREDIT LINE (BELOW) TO BE INSERTED ON THE FIRST PAGE OF EACH PAPER
particular a time-nonlocal renormalization of gyro-kinetic description is proposed.
In course of the derivation of kinetic equations we essentially use the Green
function method, similar to that proposed in Ref. [11] and applied to the non-
Markovian description of turbulent plasmas in the absence of external magnetic field
CP1177, From Leonardo to ITER: Nonlinear and Coherence Aspects, Workshop Proceedings
edited by Jan Weiland
© 2009 American Institute of Physics 978-0-7354-0716-9/09/$25.00
72
[12] as well as for plasma described in drift-kinetic approximation [13, 14]. Within the
physical context of the equation under consideration the Green function can be treated
as a probability of the particle transition in the phase space. Beside that this quantity is
needed for the derivation of the collision terms, it determines the dielectric response
function and thus makes it possible to describe electrodynamical properties of plasma
with regard to the influence of turbulence on particle motion provided that the
transition probability, taking into account such influence, is found.
Another important point of the theory of turbulent plasma concerns nonlinear
coherent wave interaction. Obviously, such effects cannot be described by the
collision terms corresponding to incoherent wave interactions. The obtained kinetic
equations give the possibility to calculate the electric fields with the accuracy up to the
second order and thus to obtain low frequency field generation due to the resonant
interaction. In particular, this makes possible to generalize the description of zonal
flows with regard to the influence of turbulent diffusion and turbulent collisions. One
more consequence of the resonant interaction is the effective decrease of the instability
growth-rate due to the velocity shear (the Waltz rule [15, 16]).
2. MICROSCOPIC EQUATION FOR A PLASMA IN AN

EXTERNAL MAGNETIC FIELD

Let us consider a plasma in an external magnetic field B0 ( r ) . We assume that the
inhomogeneity scale LB ≫ ρ th (ρ th = vth / Ω) and kLB ≫ 1 , where k is the typical
wavenumber of perturbations under consideration. In such approximation the
Klimontovich equation, where
N
N (X , t )= ∑ δ (X − X (t )),

X ≡ (r , v ), ɶ = eB0 (r )
Ω (1)
i =1 mc
can be rewritten in the following integral form
e
t
' ' ∂N (X ' , t )
N (X , t ) = − ∫ −
' ' (0) ' '
dX dt W (X , X ; t t ) E (r , t ) . (2)
m −∞ ∂v '
Here W (0) (X , X ' ; t − t ' ) is the unperturbed transition probability satisfying the
equation
 ∂ ∂ ɶ ∂  (0)
 + v + Ω ( v × e ) W ( X , X ′; t − t ′) = 0 (3)
 ∂t ∂r ∂v 
with the initial condition
W (0) ( X , X ′;0 ) = δ ( X − X ′) .
(4)
73
In the approximation under consideration (LB ≫ ρ th ) W (0) ( X , X ′;τ ) is given by
 1
 Ω

( 
W (0) ( X , X ′;τ ) = δ  r⊥ − r⊥′ − vɶ⊥′ sin Ωτ + vɶ⊥′ × e (1 − cos Ωτ )  − vDτ  ×

)
(5)
{ ⊥ ⊥ ( ⊥ ) D }
× δ v − vɶ′ cos Ωτ − v′ × e sin Ωτ − v × δ ( z − z′ − v τ ) δ ( v − v′ ) ,
ɶ

where

vɶ⊥′ = v⊥′ − vD , vD ≡ vD ( v⊥′ v ' ), vD = v∇B + v ,
2
v2
ɶ ( r′ ), v = v⊥  e ⋅ ∇ ln B  , v = ( e × ) 

Ω=Ω , (6)
2Ω 
⊥ ∇B 0
Ω

= ( e∇ ) e .

The solution (5) can be also presented in the form
W (0) ( X , X ′;τ ) = W (0) ( X ′, X ; −τ ) =

 vɶ
= δ  r⊥′ − r⊥ + sin Ωτ −
⊥

(
vɶ⊥ × e )

(1 − cos Ωτ ) + vDτ  × (7)
Ω Ω
 
{ ( ) }
× δ v⊥′ − vɶ⊥ cos Ωτ + vɶ⊥ × e sin Ωτ − vD × δ ( z′ − z + vτ ) δ ( v′ − v ) .

Here vɶ⊥ = v⊥ − v D , Ω = Ω
ɶ (r) .
3. LINEAR RESPONSE
The first iteration of Eq. 2 with respect to the self-consistent perturbation
δ N ( X , t ) = N ( X , t ) − f 0 ( X ) results in the following dielectric response function for
fluctuation field
4π e 2 ∂f 0 ( X ′)
ε (k , ω ) = 1 − i ∫ dv ∫ dv ′W ω ( v , v ′) k

(0)
. (8)
mk 2 k
∂v ′
This result recovers the appropriate calculations on the basis of Vlasov equation.
Using Eq. 5 one obtains

Wk(0)
ω
( v′) = ∫ dvWk(0)
ω
(v , v′) =
∞

 ik⊥ ɶ  (9)
= ∫ dτ exp 
 (  )
v′⊥ sin Ωτ + vɶ⊥′ × e (1 − cos Ωτ )  + i(ω − kv − k⊥ vD )τ 
0 Ω 
74
or
ei ( n − m )ϕ
Wk(0) (v′) = i ∑ J n ( a ) J m ( a ) ,
ω
n ,m ω − kv − k⊥vD − nΩ + i0

where is the angle between vɶ⊥′ and k⊥ , ϕ v and ϕ k are the azimuth angles of the

vectors k and v , respectively. In what follows we assume that the distribution

function f(X) depends on the integral of motion R⊥ = r⊥ + ( v⊥ × e ) /¬Ω . Thus, for the

local Maxwellian distribution
3/2
 m   mv 2 
f0 ( X ) = n ( R⊥ )   exp  − 
 2π T ( R⊥ )   2T ( R ) 
one obtains
∂f ∂f m
( )

k 0 = k − k⊥ v⊥ − k⊥ v* f f 0 , (10)
∂v′ ∂v′ T
where
T
v* f =  e∇ ln f  .
mΩ 
Substituting Eqs. 9 and 10 into Eq. 9 we have
4π e 2  2  k ⊥ v⊥  ω − ω* f 
ε (k , ω ) = 1 + 1 − ∫ dv ∑ J    f 0 ( X ),
k2  n n  Ω  ω − nΩ − kv − k ⊥ vD + i 0 

(11)
  mv 2 3  T L
ω* f = ω* 1 + η  −  , ω* = k ⊥ [e ⋅ ∇ ln n ], η = n ,
  2T 2  mΩ LT
where Ln and LT are the density and temperature inhomogeneity lengths

respectively.
4. COLLISION TERM
Statistical averaging of Eq. 1 gives the equation for the distribution function
f ( X , t ) ≡ 〈 N ( X , t )〉
75
∂ ∂ e ∂ ɶ ( v × e ) ∂  f ( X , t ) = I,
 + v + 〈E〉 + Ω (12)
 ∂t ∂r m ∂v ∂v 

where
e ∂
I=− 〈δ E ( r , t )δ N ( X , t )〉. (13)
m ∂v
The equation for the fluctuation evolution is
∂ ∂ e ∂ ɶ ∂ 
 + v + 〈E〉 + Ω ( v × e ) ∂v  δ N ( X , t ) +
 ∂t ∂r m ∂v  (14)
e ∂f e ∂δ N ( X , T ) e ∂
+ δ E + δ E (r , t ) = 〈δ Eδ N 〉.
m ∂v m ∂v m ∂v
The collision term should be a functional of f(X,t) (in the opposite case kinetic
description cannot be introduced) and thus is a slowly varying quantity. This could be
a reason to disregard it in the equation for fluctuation evolution (but not in Eq. 12).

We also assume that 〈 E 〉 = 0 . In this case
∂f ( X ′, t ′)
t
e
δ N ( X , t ) = δ N (0) ( X , t ) − ∫ dt ′∫ d X ′W ( X , X ′; t , t ′)δ E ( r ′, t ′) , (15)
m0 ∂v ′
where W ( X , X ′; t , t ′) is the “microscopic” transition probability with regard to

fluctuation field influence. It satisfies the equation
∂ ∂ e ∂ ɶ ∂ 
 + v + δ E (r , t ) + Ω ( v × e ) ∂v W ( X , X ′; t, t′) = 0 (16)
 ∂t ∂r m ∂v 
with the initial condition
W ( X , X ′; t, t ′) = δ ( X − X ′). (17)

Eq. 16 differs from Eq. 3 by the term with the fluctuation field δ E ( r , t ) .
The quantity δ N (0) ( X , t ) is the microscopic density fluctuation in the system with
no interaction through the self-consistent microscopic field.
Substituting Eq. 15 into Eq. 13 one obtains
I = I pol + Iind , (18)
where
76
e ∂
I pol = −   〈δ E ( r , t )δ N (0) ( X , t )〉,
 m  ∂v
∞
(19)
e ∂ ∂f ( X ′, t ′)
=   ∫−∞ dt′∫ d X ′〈W ( X , X ′; t, t ′)δ Ei (r , t )δ E j (r′, t′)〉 ∂v′j .
ind
I
 m  ∂vi
The first part of the collision term I pol is determined by the correlations which are

not related to the collective interactions directly. As is seen, only the part of δ E
correlated with fluctuation of microscopic phase density with no collective
interactions contributes to the polarization add in the collision term. Assuming that in
the case of turbulent plasma the collective part of the fluctuation field is dominant, it is
possible to use the approximation
I ≃ I ind .
The next step is to calculate the triple correlation function in Eq. 28. Usually the
following approximation is used

〈W ( X , X ′; t , t ′)δ Ei ( r , t )δ E j ( r′, t ′)〉 ≅ 〈W ( X , X ′; t , t ′)〉〈δ Ei ( r , t )δ E j ( r′, t ′)〉 (20)
and thus,
e ∂ ∂f ( X ′, t ′)
2 t

I≅ 
 m  ∂vi
∫ dt′∫ d X ′〈W ( X , X ′; t, t′)〉〈δ E (r , t )δ E (r ′, t ′)〉
−∞
i j
∂v ′j
. (21)
The arguments concerning the substantiation of the approximation Eq. 20 are given
in [11]. Averaging Eq. 16 it is possible to show that in the approximations under
consideration the averaged transition probability 〈W ( X , X ′; t , t ′)〉 satisfies the equation
of type Eq. 12, namely
∂ ∂ e ∂ ɶ ∂ 
 + v + 〈E〉 + Ω ( v × e ) ∂v  〈W ( X , X ′; t, t′)〉 =
 ∂t ∂r m ∂v 
(22)
e ∂ ∂〈W ( X ′′, X ′; t ′′, t ′)〉
2 t

=  ∫ dt ′′∫ d X ′′〈W ( X , X ′′; t , t ′′)〉〈δ Ei ( r , t )δ E j ( r′′, t ′′)〉 × .
 m  ∂vi t′ ∂v ′′j
Obviously, 〈W ( X , X ′; t , t ′)〉 can be treated as the Green’s function of Eq. 12 and

thus could be used for calculation of the renormalized dielectric response function.
5. QUASILINEAR APPROXIMATION
Let us start our calculations of the collision term from the simplest case of particle
motion along the unperturbed orbit, i.e.
77
〈W ( X , X ′; t , t ′)〉 = W (0) ( X , X ′; t − t ′)
within the time-local approximation. In this case

e ∂ dk d ω  ∂f
2 t
I=  ∫ 3∫ ∫ dτ eiωτ −iΦ (τ ) ki 〈δΦ 2 〉 − kω  k ⊥ cos Ωτ −
 m  ∂vi (2π ) 2π 0  ∂v ⊥
(23)
 ∂f  ∂f 1 ∂f 1  ∂f  
−  k ⊥  sin Ωτ + k + k ⊥ sin Ωτ − k ⊥  (1 − cos Ωτ )  ,
 ∂v ⊥  z ∂v Ω ∂r⊥ Ω  ∂r⊥  z 
where
{ } ( )
1 ɶ
Φ (τ ) = − k ∆r (v , −τ ) = k⊥ v⊥ sin Ωτ − k ⊥ [v⊥ ·e ](1 − cos Ωτ ) + k⊥vD + kv τ .
Ω
In the case of homogeneous plasma Eq. 23 can be easily reduced to the collision
integral in the Fokker-Planck form obtained in Ref. [17].
In what follows we are interested in low-frequency processes and thus it is natural
to assume that the distribution function depends on coordinates through the integral of
motion r⊥ + ( v⊥ × e ) , i.e. f ( X ) = f ( r⊥ + (v⊥ × e ) / Ω, v , t ) and thus it is possible to
1
Ω
∂ ∂ eijz ∂
put = −
∂vi ∂vi Ω ∂ri
that gives

e
2
dk d ω  ∂ 1  ∂  
I=  ∫ (2π )3 ∫ 2π 〈δΦ 2
〉 k
kω  −  k⊥   ×
m  ∂v Ω  ∂r⊥  z 
(24)
∞
iωτ −iΦ (τ )
 ∂  ∂  1  ∂  ∂ 
×∫ dτ e  k⊥ cos Ωτ − k ⊥  sin Ωτ − k⊥  + k  f ( X , t ).
0  ∂v⊥  ∂v⊥  z Ω  ∂r⊥  ∂v 
As is seen, collision operator 24 includes all combinations of the spatial and

velocity derivatives. Obviously, the part containing the velocity derivatives only
generalized the result obtained for the collision term for homogeneous plasma and
generates the diffusion coefficient in the velocity space which is in agreement with
that derived by M. Rosenbluth and S. Ichimaru [17] on the basis of direct calculations
of the particle velocity displacements. The diffusion in the velocity space at large
evolution times in turn can produce the diffusion in the real space (see, for example,
[18]), however, such spatialdiffusion is definitely of the lower order in FLR than that

associated with the direct Ê × B - random motion. This conclusion follows from the
comparison of the diffusion coefficient in the real space induced by the diffusion in
the velocity space ( D r ≃ D v / Ω 2 , where D v is the diffusion coefficient in the velocity
78
space) and that calculated in the drift approximation (see, [19] and references cited
therein).
Taking into account the reasons mentioned above it is possible to assume that
the main contribution in the case under consideration to the collision operator 38 is
given by the pure spatial part for the perturbations propagating across magnetic field
and by the velocity part associated with the parallel motion, i.e.
∞
dk d ω ∂ 2  ⊥ k⊥i k ⊥ j  2
2
e 1
I ≃ +  2 ∫ 3∫
δ
 ij − k
 ⊥ 〈 δ Φ 2
〉 kω ∫
dτ eiωτ −iΦ (τ ) f ( X , t ) +
 m  Ω (2π ) 2π ∂r⊥i ∂r⊥ j  k⊥2  0
(25)
∞
dk d ω 2 ∂ iωτ −iΦ (τ ) ∂f ( X , t )
2
e
+  ∫ 3∫
 m  (2π ) 2π ∂v 0
k ∫ dτ e ∂v
,
or
∂2 f ∂  ∂f 
I ≃ Dij +  d , (26)
∂r⊥i ∂r⊥ j ∂v  ∂v 
where
∞
dk d ω ∂ 2
2
e  ⊥ k⊥i k⊥ j  k⊥2 2
Dij =   ∫ (2π )3 ∫ 2π ∂r⊥i∂r⊥ j δ
 ij −  2 ⊥k 〈 δΦ 2
〉 kω ∫
dτ eiωτ −iΦ (τ ) f ( X , t ), (27)
m  k 2
⊥  Ω 0
∞
dk d ω 2
2
e iωτ −iΦ (τ )
d =  
m
∫ (2π )3 ∫ 2π ∫0 dτ e
k . (28)
Averaging (41) over gyro-rotation gives

dk d ω k⊥2  ⊥ k⊥i k ⊥ j  i 〈δΦ 〉 kω 2
2 2
e
Dij =  
m
∫ (2π )3 ∫ 2π Ω2  ij k⊥2  ∑n ωɶ − nΩ J n (a ).
δ − (29)
That is in agreement with the result of calculations on the basis of the drift

description of random transverse motion [19]. Here, ωɶ = ω − kv − k⊥vD .
Averaged value of (42) is given by

dk d ω 2
2
e i
d =   ∫ 3∫
k ∑ J n2 (a ) . (30)
 
m (2 π ) 2π n ω
ɶ − nΩ + i 0
That is the parallel part of the diffusion coefficient in the velocity space obtained in
[18].
79
6. RENORMALIZED TRANSITION PROBABILITY
The quasilinear time-local description given above disregards the turbulent field
influence on particle trajectories. In order to take into account such influence it is
necessary to use Eq. 22 for the averaged transition probability. With regard to explicit
form of the collision term and assuming that its velocity part is most important for the
description of parallel motion we are led to the following set of equations for the
transition probability and time-nonlocal kinetic coefficients
∂ ∂ ɶ ∂  ∂  ∂〈W ( X , X ′; t ′′, t ′)〉 

t
 + v + Ω ( v × e )  〈W ( X , X ′; t , t ′)〉 = ∫ dt ′′ d (t , t ′′) −
 ∂t ∂r ∂v  t′
∂v  ∂v 
(31)
t
∂  ∂〈W ( X , X ′; t ′′, t ′)〉 
+ ∫ dt ′′  Dij (t , t ′′) ,
t′
∂ri  ∂rj 
d (t , t′)   dω
 d ω − iω ( t −t′) 
=∫ e  , (32)
 2π 
Dij ( t, t′)   Dijω

dk dω′
2
e iJ n2 (a)
dω =  
 m
∫ (2π )3 ∫ 2π 〈δΦ2
〉 k ×
2
k ω′ ∑n ω − ω′ − k v − k v − nΩ + ik k D + iν ,
(33)
⊥ D i j ijω−ω′ ω−ω′

dk d ω′ k⊥2  ⊥ k⊥i k⊥ j 
2
e
Dijω =   ∫ (2π )3 ∫ 2π Ω2  δ ij − k ⊥2  〈δΦ 〉 kω′ ×
2
m
(34)
iJ 2 (a )
×∑ n ,
n ω − ω ′ − kv − k ⊥ vD − nΩ + iki k j Dijω −ω′ + iν ω −ω′
1/3
 k 2d ω 
νω =   . (35)
 3 
Deriving Eqs. 31 – 35 we used the following approximation to calculate the

integrals of the type
∞
 k2d ω 3  ∞
∫0 dτ exp  − i Φ (τ ) − τ + iωτ  ≃ ∫ dτ exp [ −iΦ (τ ) − ν ωτ + iωτ ] . (36)
 3  0
Eqs.47 – 50 form the set of integral equations for the kinetic coefficients. This set
of equations is not yet solved even for the case of prescribed statistical properties of
turbulent fields. This makes serious problem for the application of the obtained
relations. However, even in such general form they can be used to obtain some useful
80
estimates, in particular, as is shown below, to estimate the diffusion coefficient for
saturated turbulence.
Now, let us use Eq. 12 to describe wave propagation in the system with turbulent
collisions. The equation for perturbation has the form
∂ ∂ e ∂ ɶ ∂ 
 + v + E0 + Ω ( v × e ) ¬  δ f ( X , t ) +
 ∂t ∂r m ∂v ∂v 
(37)
e ∂f 0 e ∂δ f ( X , t )
+ E + E = I { f 0 + δ f } − I { f 0 }.
m ∂v m ∂v
Here, we assume that

〈 E ( r , t )〉 = E0 + E ( r , t ) ,
f ( X , t ) = f 0 ( X ) + δ f ( X , t ),

where E0 is the external electric field, if present, E ( r , t ) is the field of regular
perturbation.
The solution of the linearized Eq. 37 can be rewritten as
∂f
t
e
δ f ( X , t) = − ∫ dt ′∫ d X ′〈W ( X , X ′; t − t ′)〉 E ( r′, t′) 0 , (38)
m −∞ ∂v′
where 〈W ( X , X ′; t − t ′)〉 satisfies Eq. (31).

Eq. 38 generates renormalized response function 8 which describes nonlinearities
associated with random-turbulent field, but disregards nonlinearities produced by the
e ∂δ f
term E in Eq. 51, i.e. by regular perturbations E and δ f .
m ∂v
7. ZONAL FLOWS
Let us estimate the influence of the nonlinear term in Eq. 37 within the perturbation
theory. We thus obtain

dk ′, d ω′
2
e
δ f (v ) = −   ∫ dv′∫ dv′′∫ 〈W (v , v′)〉 ×
(2π )3 ∫ (2π ) k ω
( 2)

kω
m (39)
∂ ∂f ( X '' )
× k ′ 〈Wk′′ω′′ ( v′, v′′)〉 k ′′ 0 Φ k′ω′Φ k′′ω ′′ ,
∂v′ ∂v′′

where k ′′ = k − k ′, ω ′′ = ω − ω ′.
81

Integration of Eq. 39 over v leads to the generalization of the result presented in
Ref. [20] to the case of plasma with diffusion generated by stochastic turbulent fields.

e2n0 1 dk ′ d ω′ ( k × k ′′)
δn
Ti mi Ω ∫ (2π ) 3 ∫ 2π ∫
(2)
=i dv ×
kω
ωɶ
 k ′ v  ω ′′ − ωɶ ′′ 2  k ⊥′′v⊥ 
× J 02  ⊥ ⊥  * J0   f0 (v )Φ k ′ω′Φ k′′ω′′ , (40)
 Ω  ωɶ ′′  Ω 

ωɶ = ω − kv − k ⊥vD + iki k j Dijω + iν ω .
At Dij =0 , ν = 0 Eq.40 reduces to the result obtained in Refs. [20, 21].

The particle flux associated with the perturbation δ nk(2)

ω
can be estimated as
Γikω = eijz nU jkω ,
where

′ d ω ′  ω − ωɶ  2
2
i  e  m dk
dvf 0 ( v ) k ′j 1 −
n  m  T Ωn (2π )3 ∫ 2π ∫
∫
U jkω =   J (a ) ×
 ωɶ  0 (41)
ω*′ − ω′ 2
× J 0 (a ′)Φ k′ω′Φ k′′ω ′′ .
ω′
In the case of narrow spectrum of δΦ kω ′ the stationary value of the velocity is given
by

dk d ω ω − ωɶ * 2
2
ie m
n  m  T Ω ∫ (2π )3 ∫ 2π
Ui =   f 0 ( v ) ki J 0 ( a ) | Φ kω |2 . (42)
ωɶ
∂δ f
Thus, the nonlinear term E in Eq. 37 generates, in particular, additional
∂v
background velocity Û which can be treated as a zonal flow velocity. Restricting
ourselves to the description of this effect it is possible to use the following
approximation
e ∂δ f ∂
E ≃ eijz U jδ f ( X , t )  . (43)
m ∂v ∂ri 
This leads to the further modification of the linear response function given by
∂
( )

ω → ω − k × U e + ieijz Uj. (44)
∂ri

Introducing (43) and (44) we assume that the velocity Û is dependent on the
coordinate through the density and temperature profiles.
82
With regard to (61) the renormalized dielectric response function is

kD2 
εɶ ( k , ω ) = 1 + 1 − ∫ dv ∑ J ( a ) f0 ( v )  ×
2
k2
 n n 
 ∂U j 
( )

 ω − ω* − k ⊥ × U e − ik⊥i k ⊥ j Dijω − iν ω − ieijz ∂r 
(45)
×  i 
.
∂U j
( )

ω − kv − k⊥vD − nΩ − k ⊥ × U e + ik ⊥i k⊥ j Dijω + iν ω + ieijz − nΩ
∂ri
In turn, this leads to the following condition of the instability saturation
∂U j
γ k = ν ω + k⊥i k⊥ j Dijω + eijz . (46)
∂ri

k k
Together with Eq. 48 this condition gives the following estimate for the diffusion
coefficient for slow perturbations
3
 ∂U j 
 γ k − eijz 
 ∂ri 
k⊥i k ⊥ j Dij 0 = 3
. (47)
∂U j 
( ) )
2 
(

ωk − k⊥vD − k⊥ × U e +  γ k − eijz 
 ∂ri 
In order to get Eq. 64 we took zero-th order cyclotron harmonic, neglected the
parallel motion and estimated other quantities in Eq. 48 using assumption that in the
case of small ωk
∂U j
k⊥i k ⊥ j Dij 0 = γ k − eijz . (48)
∂ri
8. SUMMARY AND CONCLUSIONS

The Fokker-Plank-type kinetic equations with time-nonlocal kinetic coefficients are
formulated for the case of inhomogeneous magnetized plasma. These equations
generalize the results known from the traditional quasilinear theory and the Dupree-
Weinstock theory with renormalized time-local diffusion coefficients.
Relations for the diffusion coefficients in the velocity and real spaces are
found. It is shown that the diffusion in the real space associated with the drift motion
induced by random electric fields is better pronounced then that generated by the
diffusion in the velocity space.
The possibility of the renormalized kinetic description of low frequency field
generation due to coherent wave interaction is shown.
83
REFERENCES
1. O. Ishihara, C. Grabowski and A. Hirose. Phys. Fluids B2, 270 (1990).
2. O. Ishihara, H. Xia and A. Hirose, Phys. Fluids B4, 399 (1992).
3. H. Xia, O. Ishihara and A. Hirose, Phys. Fluids B5, 2892 (1993).
4. R. Balescu, Phys. Rev. E51, 4807 (1995).
5. R. Balescu, Phys. Rev. E55, 2465 (1997).
6. A. Hirose and O. Ishihara., Ca. J. Phys. / Rev. Can. Phys. 77 (10), 829 (1999).
7. R. Balescu, Eur. J. Phys. 21, 279 (2000).
8. A. Hirose, K. Elder, E. Bolster and O. Ishihara, Plasma Sci. IEEE Trans. 31, 3 (2003).
9. R. Balescu, M. Vlad, F. Spieand and J. Misguich. Intern. J. Quant. Chem. 98, 125 (2003).
10. R. Balescu, Aspects of Anomalous Transport in Plasmas. Institute of Physics, Bristol, 2005.
11. S.A. Orszag and R.H. Kraichnan, Phys.Fluids 10, 1720 (1967).
12. A. Zagorodny and J. Weiland, Phys. Plasmas 6, 2359 (1999).
13. A. Zagorodny and J. Weiland, Cond. Mat. Phys. 4, 603 (2001).
14 J. Weiland, A. Zagorodny and V. Zasenko, „On Advanced Fluid Modeling of drift Wave
Turbulence” in Recent Research Developments of Plasma Physics, Transworld Research Network,
Trivandrum, Kerala 2007,
pp. 209-234.
15. R.E. Waltz and G.O. Kerbel, J. Milovich, Phys. Plasma 1, 2229 (1994).
16. R.E. Waltz, G.O. Kerbel, J. Milovich, and G.W. Hammet, Phys. Plasma 2, 2408 (1995).
17. M. Rosenbluth and S. Ichimaru, Phys. Fluids 12, 2778 (1970).
18. I. Holod and A. Zagorodny, J. Weiland, Phys.Rev. E71, 046401 (2005).
19. S. Ichimaru, Basic Principles of Plasma Physics. Benjamin, N.Y., 1973.
20. J. Weiland, Phys.Scripta 29, 234 (1984).
21. J. Weiland, A. Eriksson, H.Y. Nordman and A. Zagorodny, Plasma Phys. Control. Fusion 49, A45
(2007).
84

zagorodny2009

Uploaded by

Copyright:

Available Formats

zagorodny2009

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

zagorodny2009

Uploaded by

Copyright:

Available Formats

NonMarkovian Effects in Turbulent Diffusion in Magnetized Plasmas

Anatoly Zagorodny and Jan Weiland

Citation: AIP Conf. Proc. 1177, 72 (2009); doi: 10.1063/1.3253966

Additional information on AIP Conf. Proc.

2. MICROSCOPIC EQUATION FOR A PLASMA IN AN

can be rewritten in the following integral form

with the initial condition

The solution (5) can be also presented in the form

W (0) ( X , X ′;τ ) = W (0) ( X ′, X ; −τ ) =

local Maxwellian distribution

Substituting Eqs. 9 and 10 into Eq. 9 we have

where Ln and LT are the density and temperature inhomogeneity lengths

The equation for the fluctuation evolution is

where W ( X , X ′; t , t ′) is the “microscopic” transition probability with regard to

with the initial condition

Obviously, 〈W ( X , X ′; t , t ′)〉 can be treated as the Green’s function of Eq. 12 and

within the time-local approximation. In this case

As is seen, collision operator 24 includes all combinations of the spatial and

Averaging (41) over gyro-rotation gives

Averaged value of (42) is given by

∂ ∂ ɶ ∂  ∂  ∂〈W ( X , X ′; t ′′, t ′)〉 

Deriving Eqs. 31 – 35 we used the following approximation to calculate the

Here, we assume that

The solution of the linearized Eq. 37 can be rewritten as

where 〈W ( X , X ′; t − t ′)〉 satisfies Eq. (31).

At Dij =0 , ν = 0 Eq.40 reduces to the result obtained in Refs. [20, 21].

In turn, this leads to the following condition of the instability saturation

8. SUMMARY AND CONCLUSIONS

You might also like