Sie sind auf Seite 1von 13

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Theory of the ordinary and extraordinary mode coupling in fluctuating plasmas

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2014 Plasma Phys. Control. Fusion 56 125011
(http://iopscience.iop.org/0741-3335/56/12/125011)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 111.68.101.153
This content was downloaded on 17/04/2015 at 06:25

Please note that terms and conditions apply.

Plasma Physics and Controlled Fusion


Plasma Phys. Control. Fusion 56 (2014) 125011 (12pp)

doi:10.1088/0741-3335/56/12/125011

Theory of the ordinary and extraordinary


mode coupling in fluctuating plasmas
AGShalashov and EDGospodchikov
Institute of Applied Physics of Russian Academy of Sciences, Ulyanova str. 46, 603950 Nizhny Novgorod,
Russia
E-mail: ags@appl.sci-nnov.ru
Received 31 July 2014, revised 16 September 2014
Accepted for publication 24 September 2014
Published 6 November 2014
Abstract

The theory of the linear coupling between the ordinary and extraordinary electron cyclotron
waves in the vicinity of the plasma cut-off has been generalized for configurations with
harmonic perturbations of plasma density, which is relevant, in particular, to the magnetic flux
fluctuations in a toroidal fusion experiment. It is shown that for a large aspect ratio, tokamak
fluctuations result in the amplitude and phase distortion of the quasi-optical beam on its way
towards the coupling region located near the cut-off layer, but have no essential influence on
the efficiency of the mode interaction inside the coupling region. However, this may not be
valid in a spherical tokamak, where the perturbations may be significant inside the coupling
region. We find that density fluctuations cannot be fully responsible for a low efficiency of the
OXB heating of the overdense plasma observed in the present-day experiments.
Keywords: mode conversion, cyclotron waves, fluctuations in toroidal plasma
(Some figures may appear in colour only in the online journal)

1.Introduction

the applicability of this theory to plasma with non-onedimensional fluctuations is not guaranteed and needs further
investigation.
In the present paper we propose a more rigorous approach
based on the 2D and 3D theory of the mode conversion that
can adequately describe the fluctuations in a toroidal plasma
without any additional qualitative speculations. The problem is
treated by means of a set of reference wave equationsobtained
from the Maxwell equationsin simplified geometry that allow
exact and physically clear solutions. Based on these solutions,
a new insight into the wave coupling in the presence of a harmonic perturbation near the plasma resonance (cut-off) has
been derived. As a result, the old, well-established view on
the role of fluctuations in the OX coupling in a fusion experiment has been significantly revised.

The linear conversion of the ordinary (O) wave to the extraordinary (X) wave plays an important role in the excitation of
the electron Bernstein waves, which in turn provide an effective way for the high-frequency (electron cyclotron) heating
and diagnostics of overdense plasma in spherical tokamaks
and optimized stellarators, see for example [1] and the references therein. In spite of a rather clear theoretical background
behind this process, existing theory tends to overestimate the
efficiency of the OX coupling in all the available experiments. Following Laqua, the significant discrepancy between
the predicted and observed efficiencies is usually explained
by the turbulent fluctuations of the cut-off layer where the
mode conversion occurs [2]. It should be noted, however, that
this explanation gives only a qualitative view based essentially on geometrical considerations. As a result, the poloidal
fluctuations of the cut-off surface are taken into account by
the effective broadening of the incident wave spectrum over
the poloidal wavenumbers within the 1D theory of mode coupling proposed by Mjolhus and Zharov for the plasma uniform in the poloidal and toroidal directions [3, 4]. Obviously,

0741-3335/14/125011+12$33.00

2. Reference wave equations


Following [5] we will use the Stix representation for the electric field components in the Cartesian coordinate system with
the z-axis directed along the external magnetic field:

2014 IOP Publishing Ltd Printed in the UK

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Ex iE y =

2 F exp(ik z it ), Ez = Fz exp(ik z it ).

N = (1 + / ce )1/2 ,

Here F denotes the slow field amplitudes that are smoothly


varying compared to the carrier wave length = 2 / k .
Assuming the plasma is weakly inhomogeneous in the x and y
directions and constant in the z direction, the Maxwell equations in the region of the OX wave coupling (close to the
plasma cut-off) can be reduced to [59]:


N (i/x /y ) F =
2 k 0 + N 2 F+,

N (i/x + /y ) F+ = 2 k 0 F .

L =

where = N 2(1 + N 2 )/(1 N 2 ) if the conservation of the


toroidal number N = const is taken into account and = N 2
if one assumes N = const. The coupling length is a slow function of all the parameters, so in most experiments it lies in the
0.51cm range.
In the presence of fluctuations along the z-axis, equation (1) must be updated to take into the account slow field
variations in this direction. This can be formally performed
by substitution N N + ik 01/z in (1) and by neglecting
all the second derivatives in the resulted equations. Finally,
we obtain the system which without perturbations has been
already studied in [8]:

(1)

Here k 0 = / c is the vacuum wavenumber, N = k / k 0 is the


refractive index parallel to the magnetic field, and +, are the
dielectric tensor components in the Stix frame:
2
= 1 pe
/ 2,

N (i/x /y ) F =
2 k 0 + N 2 2iN k 01/z F+,


(2)
N (i/x + /y ) F+ = 2 k 0 F .

2
+ = 1 pe
/[( + ce )].

ce and pe are, correspondingly, the electron cyclotron and


plasma frequencies. There are three spatial scales in the above
wave equations: the density variation L n = n e / n e , the
magnetic field variation LB = B / B , and the wave coupling
length that may be defined as
L =

N 2k 02 (+ N 2 )

1/4

L n / k 0 (ce / 2)1/4 (1 + L n / LB )1/4 ,

A study of the wave equationscan be done after the transformation to new field variables:
F+ =

g+ (A+ + A ),

F=

g (A A+ ),

with g+1 = (+ N 2 ) L and g 1 = L which may


be considered as being constant inside the coupling region.
Here the field variables are introduced in such a way that A+
and A are the amplitudes of the waves that propagate, respectively, in the positive and negative directions along the x-axis
in the WKB limit; see e.g. [6, 7]. In the new field variables,
the wave equation(2) take the following dimensionless form
convenient for further analysis:

Equation (1) have been obtained for a weakly inhomogeneous plasma, k 0L n, k 0LB >>1. Note also that the coupling
length L ~ L n / k 0 . Therefore, in tokamak conditions the
coupling is highly localized in space:
L < <L n < LB .

Within the geometrical optics (WKB) limit,


/x ikx, /y ik y, equation (1) result in a familiar dispersion relation for the left-polarized waves in the vicinity of
the cut-off region [10]:

(i/x + s ) A+ = (/y + a ) A + i (A+ + A )/ z ,


(i/x s ) A = (/y a ) A+ i (A+ + A )/ z .
(3)

Nx2 + Ny2 = 2 (+ N 2 )/ N 2 .

These are the basic equationsused in the subsequent theory.


Here = 2g / g+ ~ 1, and terms s,a are unambiguously
defined from equalities:

According to the usual notation, this solution describes the


O-mode wave for >0 and +>N 2, and the X-mode wave
for <0 and + < N 2. The propagation regions for the O
and X waves are separated by the evanescent region defined
by condition (+ N 2 ) < 0. Outside the geometrical optics
approximation, a wave field may tunnel through the evanescent region, which is exactly the OX mode coupling considered in the present paper. Thus, effective coupling is possible
when the right-hand sides of both equation(1) go to zero. The
latter condition defines the optimal value of the parallel wavenumber N = + at = 0. In this paper only a close vicinity
of this optimum will be considered. In a toroidally symmetric
geometry N N / r, where N is the conserving number
that is liable for a toroidal angle, r is the major radius, and
an exact equality corresponds to neglecting the poloidal component of a magnetic field. This means that N varies exactly
as the toroidal component of a magnetic field, which should
be taken into account when calculating (+ N 2 ); unfortunately this peculiarity was not mentioned in our previous
works. Explicitly one can obtain:

g = s + a ,

g+(+ N 2 ) = s a .

(4)

Here and below all the coordinates marked by prime and


corresponding wave vectors are normalized over the characteristic coupling length L (i.e. x = x / L etc). The terms
proportional to i are responsible for the toroidal inhomogeneity and should be omitted when the perturbations along the
magnetic field are ignored1. Note that the plasma dielectric
response is decomposed into two coupling terms, s and a,
which we call the symmetric and asymmetric parts, correspondingly. In the vicinity of the coupling region both the
terms are small (approaching zero for an ideal case), but their
particular dependence on spatial coordinates defines the entire
physics of the mode interaction. As shown below, the symmetric and asymmetric parts describe the coupling between
1

More precisely, if the plasma is homogeneous in the z direction,


then i/z may be taken into account by the formal substitution:
a,s a,s kz.
2

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

so we can formally consider limit . Approximation

(5) is valid until L < L n, which results in the more stringent


requirement n e < k 0n e / k 0L n . Besides, the shift in (5) due
to the fluctuations must also be limited, s, a < ; in dimensional units this is equivalent to n e < n e / k 0L n . All these
criteria may be fulfilled relatively easily for a parameter range
typical of fusion experiments.
Below we analyze equations (3) and (5) with harmonic
density perturbations. In the next three sectionswe consider
the formal theory for normal modes propagating along the x
direction. For the sake of clarity we study separately three
particular cases in which the wave vector of the perturbations is aligned along different coordinate axes (x, y and z). As
boundary conditions we set an amplitude of the incident wave
A+ at x and zero amplitude for the wave coming from
the other direction, A = 0 at x +. So we study the low
field side launch and the consequent wave propagation towards
the plasma density increase, this case being relevant to most
fusion experiments. With the described boundary conditions
we calculate the reflected and transmitted waves far outside
the coupling region, i.e. A at x and A+ at x +,
correspondingly. The final result of these calculations is in
most cases the reflection and transmission coefficients:

co- and counter-propagating waves, thus resulting in very different effects. In our previous works [59] we considered slow
variations of the coupling parameters in space, while here we
use a similar approach to the small periodic variations typical
of fluctuating plasma.
To study the effects of the fluctuations we consider the simplest nontrivial case in which:
n e =(1 + x / L n )ncr + n e(x, y, z ),

B =(1 + x / LB )B0z 0,

i.e. only the density is perturbed, and without this perturbation


1
the problem is reduced to a 1D case. Here ncr = / n e
is the plasma cut-off density at which = 0. Note that in a
toroidal geometry the non-perturbed density is constant on a
flux surface, thus x and y represent here the radial and poloidal
directions, correspondingly; the variation in the magnetic field
in the poloidal and toroidal directions is ignored. Bearing in
mind that the OX coupling occurs on a scale L shorter then
all the plasma inhomogeneity scales, we retain only the linear
dependences over x in the magnetic field and the non-perturbed density. Then, the coupling parameters (4) are defined
in a very simple way:
1 (g + g++)
n e,
2
n e
1 (g g++)
a =
n e ,
2
n e
s = x +

or, explicitly:
s = x s,

a = a,

n e L n
a ,
ncr L
n L
1
.
a = e n
ncr 2L 1 + LB / L n

Ln =

n e + n e
k 0L n
.

n e + n e
1 + L nn e / n e

A ()

A+ ()

T=

A+ (+)

A+ ()

,
.

The sum is over all modes constituting the spectrum of the


wave beams. Below these coefficients will be always averaged
over a random phase of harmonic perturbations. Note that T
is exactly the OX coupling efficiency mentioned as a fraction of the power flux penetrating through the cut-off layer.
In the non-dissipative media R + T = 1, so it is enough to calculate one of these coefficients. In section6 we compare our
results to the standard model mentioned in the Introduction.
Finally, in the last sectionwe apply our theory to studying the
effects of fluctuations for realistic conditions close to a fusion
experiment.
Before going into a detailed analysis, let us consider some
qualitative insights into the estimated physics. Apart from
terms responsible for the toroidal inhomogeneity (proportional to i ), the right-hand sides of our basic equations (3)
describe the coupling between the counter-propagating waves,
while the left-hand sides describe the coupling between the
co-propagating waves. Therefore, we can assume that the
coupling between the counter-propagating waves (reflection)
is affected only by the asymmetric part of the fluctuations.
Opposite to this, the symmetric part does not disturb the wave
coupling itself; however, it can modify the propagation of
separate waves outside the coupling region. So, even from the
structure of wave equationswe can argue that the symmetric
and asymmetric contributions of the density fluctuations
act essentially differently. Physically it may be understood
as s describing the shift of the evanescent region, while a
describes the fluctuation of its width, as shown in figure1. The
simplest way to detect this difference is to consider a static
limit when s,a = const. Then s does not modify the coupling

s =

(5)

Here the linear over the x term represents a regular variation


of n e, B and N B; the minus sign corresponds to / n e<0;
s and a represent the symmetric and asymmetric contributions of the density fluctuations. The latter parameters contain
a product of small n e / n e 1 and large L n / L ~ k 0L n 1
terms. As a result, the symmetric contribution s may be of an
order of unity, even though the relative density perturbations
are always small in a fusion experiment. It is important to
note that the asymmetric part appears only due to the regular
variation of a magnetic field and toroidal symmetry, and contains the additional factor L n / LB ~ 1 / A, where A is the ratio
between the major and minor radii of a plasma torus (aspect
ratio). For that reason condition a s always holds for large
aspect ratio tokamaks, while for spherical tokamaks a<
s.
In the present paper we consider only the harmonic density fluctuations n e ~ cos(r + ). Our basic equations (3)
are obtained under the assumption of weakly inhomogeneous
plasma, namely:

k 0L n 1,

R=

(6)

Therefore, in addition to the standard requirement k 0L n 1


we must accomplish n e k 0n e. It is important to note that
the fluctuations may be in principle much shorter then the electromagnetic wavelength, provided that its amplitude is small,
3

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 1. Schematic of the symmetric (left) and asymmetric (right) contributions of the density fluctuations. The dashing shows the region
where studied electromagnetic modes are evanescent in the geometro-optical approximation.

efficiency at all, while a results in the additional attenuation


factor exp(a2 ).

This can be interpreted as the coupling being much more sensitive to the fluctuations in the size of the evanescent region
than to fluctuations in its location.
In figure2 we compare the analytical results from the data
obtained by the direct numerical integration of equations(7).
Note that in a fusion experiment optimized for the wave
transmission, the poloidal modulation of a wave should be
minimized, k y 1. However, to enhance the role of the symmetric contribution we choose quite a large k y corresponding
to an unperturbed transmission efficiency of about 80%, but
even in this case the contribution of the symmetric perturbations is small. One can see that equation(8) fits perfectly the
numerical data in the whole range of the correlation lengths.
The observed oscillations over 2 are the result of the constant
gradients of regular inhomogeneity that may smooth out for
more realistic density and magnetic field profiles (however,
the envelope of these oscillations is robust).
The Born approximation becomes invalid for a sufficiently
large reflection. In this case equations (7) may be analyzed
based on geometrical optics. In particular, the coupling efficiency for an incident plane wave is well approximated by the
Budden formula [11]:

3. Radial fluctuations
First let us consider a trivial case of a harmonic fluctuation
in the radial direction given by axis x. This is essentially a
1D case, /y ik y, /z 0, and the corresponding wave
equations(3) and (5) take the following form:

i/x x a cos(x + ) A+
)
s
(

= (ik y aa cos(x + ) ) A ,

(7)
(i/x + x + as cos(x + ) ) A

= (ik y + aa cos(x + ) ) A+ .

For the sake of simplicity we assume here the initial wave field
that has an optimal phase modulation along the magnetic field,
i.e. E exp(ikzz ) with the same kz = k as used in the derivation
of equation(1). For non-optimized waves with kz = k + kz all
transmission coefficients considered below should be just multiplied by the same attenuation factor exp( 2 kz2 ).
The above equations may be solved using the standard
Born approximation of a scattering theory. Assuming a weak
reflection, A / A+ ~ r 1, one finds a solution as a series:
A+ = A0+ + r 2A2+ + ,

T = exp 2

A = rA1 + r 3A3 + .

kx2 = (x + as cos(x + ) )2 k y2 aa2 cos2(x + ).

Although it is formulated in geometric optics terms,


the above formula typically gives very accurate results in
smoothly inhomogeneous plasma. Hence, we may anticipate
that it works at least for 1. In this way one can obtain the
following estimate for the phase-averaged efficiency:

A () 2 d
=
A+ () 2 2

k y2 + 1 aa2
2

2k y2(aa2

4as2k y2 /


)sin ( / 4)+ .
(8)
2

Imkx dx ,

where kx is a wave vector corresponding to the WKB limit of


equations(7):

Then the solution may be obtained step-by-step. First a passing


wave A0+ is calculated by putting A 0; this wave pumps a
reflected wave A1, which in turn results in a second-order correction A2+ to the passing wave etc. Finally one can calculate
the reflection coefficient averaged over a random phase :
R=1T=

T exp k y2 1 aa2 + 2k y2as2J12(k y )


+

1 2 4
aa (1
32

k y2 2 )2 + .

(9)

This formula is obtained as a series expansion over fluctuation


amplitudes. The first two terms in the exponent correspond
exactly to those in equation(8), and all the other terms are small.
Note that in this approximation the main contribution in the
coupling coefficient comes again from the asymmetric perturbations. This contribution in the first non-vanishing order is purely
a static response with factor due to the phase averaging2.

Remember that both wave vectors are normalized over L1,


and the fluctuation amplitudes as,a are defined according equation(4). The first term is a familiar result of the non-perturbed
1D theory for a plane wave with k y 1; the other terms are due
to the fluctuations. Note that the dependence over a fluctuation
wavenumber appears only in O(r 4 )-order, and the symmetric
perturbations result in a strictly zero contribution up to O(r 3)
terms. Therefore, for comparable amplitudes the asymmetric
contribution always dominates in the coupling coefficient.

Strictly speaking, the phase averaging of the static coupling coefficient


results in T =<exp(aa2 cos2)>= I0( 1 aa2 ) exp( 1 aa2 ), but the pre2
2
exponent term may be ignored when aa<
0.5.
4

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 2.Distortion T ( ) of the OX coupling efficiency T = exp(k y2 ) (1 + T ) versus the radial fluctuation wavenumber for the

symmetric (left panel, as = 0.3, aa = 0, k y = 0.2, kz = 0) and the asymmetric (right panel, aa = 0.3, as = 0, k y = 0.2, kz = 0)
perturbations. A comparison of the analytical results (dashed lines) and numerical results (solid lines).

Acomparison with the numerical data shows that equation(9)


is accurate for <1 while for shorter wavelengths it sufficiently
overestimates the role of the symmetric contributions (see
figure2, the red line in the online version). This also remains
true in the case of a strong reflection.
The results may be summarized as follows. Harmonic density modulation results in conditions for the Bragg resonance.
However, this resonance is strongly depressed for the co-
propagating modes, thus degrading the role of the symmetric
perturbations. On the other hand, the Bragg back-scattering
may degrade the coupling efficiency of the counter-propagating
waves. However, this effect is defined by the asymmetric perturbations that are small in tokamak conditions.

4. Poloidal fluctuations
In this section we study the harmonic modulation in the
poloidal direction represented by axis y. For waves with the
optimal toroidal wavenumber, equations(3) and (5) take the
following form:

i/x x a cos(y + ) A+
)
s
(

= (/y aa cos(y + ) ) A ,

(10)
(i/x + x + as cos(y + ) ) A

= (/y + aa cos(y + ) ) A+ .

/2

A =

/2

dk y

n =

(i/x x ) A + 1 a (A + + A + )
s n1
n
n+1

= i(k y + n )An aa(An 1 + An+ 1),

(11)
1

(i/x + x ) An + as(An 1 + An+ 1)


2

= i(k y + n )An+ + 1 aa(An+ 1 + An++ 1).

These equations can be effectively solved both analytically


and numerically using the techniques proposed in our previous papers [8, 12].
The main idea introduced in [8] is that the coupling on the
left- and right-hand side of this kind of wave equation acts
on essentially different scales. Namely, the coupling defined
by the left-hand side occurs over a much wider distance. The
reason is that the phase synchronism required for the coupling is very different for the co- and counter-propagating
modes. Indeed, in the WKB limit one obtains (neglecting
perturbations):


kx2 = x 2 (k y + n )2 .

(12)

Therefore, for x k y + n the wavevector kx x is for


all the modes contributing to A+, and kx +x is for all the
modes contributing to A (the inverse sign here reflects the
opposite direction of a group velocity with respect to kx for
the studied waves [5]). In this region the only pronounced
coupling mechanism is between the co-traveling modes,
while the reflection is strongly suppressed. The region with
x < k y + n is much smaller then the total distance of the
co-traveling coupling, so its presence is of no importance
for this type of mode interaction. However, only inside this
region is an effective phase synchronism between the counterpropagating waves possible, which results in the reflection
A+ A. Obviously, this process occurs on a short scale, and
therefore the interaction of the co-traveling waves inside the
reflection zone may be ignored. These considerations show
us that the symmetric (as) and asymmetric (aa) perturbations
in equations(11) may be considered separately. This circumstance significantly simplifies the analysis.

Although the non-perturbed equationsare homogeneous over


y , we will consider the beams with finite distribution in the
poloidal direction, i.e. inhomogeneous boundary conditions.
Bearing this in mind, we retain here a phase of harmonic
perturbation.
The standard technique for equations (10) is based on
Blochs theorem. Using this theorem, wave fields may be
found as a convolution of independent modes corresponding
to a particular quasi-wavevector k y:

In this case, equations(10) may be reduced to a set of ordinary differential equations:

An (x , k y ) exp i(k y + n )y .

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 3. The transmission coefficient of the optimal plane wave with k y = 0 and kz = 0 in the presence of the symmetric poloidal
perturbations versus distance L from the plasma cut-off. The analytical (dashed line) and numerical (solid line) results are compared for
as = 0.3, = 0.25 (left panel) and = 2 (right panel).

The symmetric contribution of the density perturbations


results in the coupling of the co-propagating harmonics, so
one can neglect the reflection, i.e. take A... = 0 in the first of
equations(11). In this case an exact analytic solution may be
obtained:

An+ (x) = Am+ (L) Jn m

L as dx exp i m 2 n
x

over the poloidal wavenumbers is depressed when the wavevector (12) becomes evanescent while n increases. In numerical modeling scattering to the nth mode is only efficient in a
conical region defined by condition kx2(x , n )>0. This stabilizes the process for a long enough trace.
Our goal here is mainly to show that the described type of
interaction does not influence the processes inside the reflection region, rather than giving a detailed description. It should
be noted that the spectrum modification is efficient far outside the cut-off region where our basic approximation may be
not valid. Another important consideration is that the random
phase of fluctuations may essentially slow the spectrum modification due to the transition from the convective to the diffusive regime. Note that the diffusion in k-space of a beam
induced by the fluctuations has been discussed previously; see
e.g. [13] and the references therein.
The asymmetric perturbation couples the counter-
propagating modes in the vicinity of the plasma cut-off with
a typical length L. Since this scale is much less than all the
plasma inhomogeneity scales, L L n, LB, the spectrum modification of the co-traveling modes may be ignored, as = 0 in
equations(11). Then, these equationsmay be reformulated as:

(13)

This solution describes the slow modification of the poloidal


spectrum of an incident beam as it approaches the plasma cutoff. It is important to note once more that this effect is collected
along a much longer path than the localized reflection region
where the incident wave couples with a counter-propagating
wave. Inside the reflection zone the symmetric contribution
itself is of minor importance for all meaningful parameters,
which is proven by direct numerical calculations. However
a modification of the poloidal structure of the incident beam
before the cut-off layer may result, of course, in a modification
of the net coupling efficiency. As a hortative example, let us
consider the coupling efficiency for the beam spectrum given
by equation (13) neglecting all the perturbations (including
asymmetric ones) inside the reflection zone:
T = Am+ (L)

m, n

Jn2

L as dx
0


(i/x x ) An+ = Rnm
Am ,

+ +
(i/x + x ) An = RnmAn ,

exp ( (k y +(m n ) )2 ) Am+ (L)


m

where Rnm
is just a matrix notation of the right-hand sides of
equations (11). Now one of the counter-propagating waves
may be excluded:

(14)

(i/x + x ) (i/x x ) An+ = nm Am+

This formula is obtained after averaging over a random phase.


Note that the effect is accumulated along the trace; here we
have to introduce the finite length L at which the incident
beam is launched. In figure3 we compare the analytical and
numerical results for the transmission efficiency of a plane
wave with k y = 0 and kz = 0 (optimal in the non-perturbed
case) in the presence of the symmetric poloidal perturbations.
Note that equation(14) does not depend on when 1. This
is not justified by numerical calculations, so our approximation becomes invalid for large fluctuation wavenumbers (see
figure 3, right). A physical reason is that the redistribution

R + . This matrix is Hermitian, and therefore it


with nm = Rnk
km
may be diagonalized. In other words, there is a unitary linear
+
transformation, An = OnmAm+, that converts equations(11) to a
set of independent parabolic cylinder equations[14]:

+
+
(i/x + x ) (i/x x ) An = nAn .

(15)

Each of these equationsdescribes a 1D coupling with the efficiency T = exp(n ) for the nth mode. The typical spectrum
n(k y ) is shown in figure4. The coupling efficiency of a wave
6

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 4. Example of the spectrum of eigenvalues n(k y ) defining

Figure 5. The transmission coefficient of the optimal plane wave with


k y = 0 and kz = 0 in the presence of the asymmetric perturbations
with different poloidal wavenumbers . Analytical (dashed line) and
numerical (solid line) results are compared for aa = 0.3.

the coupling efficiency in the presence of the asymmetric poloidal


perturbations with aa = 0.3, = 2.

beam with a finite spectrum may be calculated as a convolution over all modes:

which is valid when exp(as ) 1, and in particular when


. On the other hand, for the high-order poloidal modes
with aa2, the potential may be approximated by its
average. This results in a continuous spectrum = k y2 + 1 as2
2
and the familiar static expression for the coupling efficiency
averaged over a random phase:

1
T = OnmAm+ () 2 exp(n ) An+ () 2 . (16)
n

n, m

The calculation of this sum for any finite beam can be easily
performed numerically.
Decomposition over Blochs functions is a convenient
technique for numerical calculations. However, the expected
results may be more naturally understood directly from equations(10). First of all, let us notice that these equationswith
as = 0 allow solutions with a strictly zero reflection3:

A 0

A+ 0.

+
Azero
= exp 1 ix2 +(aa / )sin y ,

T exp[k y2 1 aa2].
2

This approximation is valid when a sufficient fraction of the


beam power is carried by the high-order poloidal modes,
which in practice is guaranteed by the condition 1.
As an important example, we may find the expected degradation of the transmission coefficient for the ideal mode of
unperturbed geometry, namely a plane wave with k y = 0 and
an optimal kz. Combining the two techniques proposed above,
the transmission coefficient may be estimated as:

and

Azero
= exp + 1 ix2 (aa / )sin y ,

This means that the lowest eigenvalue in (15) is always zero,


0 = 0 at k y = 0. Next, equations(10) with as = 0 allow a separation of the variables, after which the poloidal part of the
traveling wave amplitudes A is described by a Schrdingertype equation:
2
2
2
2
( /y + aa cos y aa sin y ) Y = Y .

(17)

+
A+ * dy
Azero

( A

+
2
zero

dy

A+

dy

(18)

As shown in figure5, a comparison with the numerical results


proves that this formula works well over the whole range of
perturbation wavenumbers. One can see that the influence
of the poloidal fluctuations decreases for shorter fluctuation
wavelengths. Note that this differs from the radial perturbations, where the dependence on appears only in a higher
order over the fluctuation amplitude (see figure2).

The radial part is the same as equation(15), therefore = n.


Equation(17) allows qualitative speculations. For low modes
with aa2, the potential in (17) may be approximated by
a quadratic well with the known spectrum n = asn. For a
strong enough scattering, the gap between 0 and 1 becomes
so large that we can assume that all the modes except the
lowest are totally reflected. Therefore, the perfect structure

Azero
acts as a spatial filteronly a fraction of the initial beam
that is projected to this structure is transmitted. This consideration results in a simple approximate formula for the coupling coefficient:
2

T exp[ 1 aa2(1 + 2 )1 ].

5. Toroidal fluctuations
In this section we study the harmonic modulation in the
toroidal direction represented by axis z. Wave equations (3)
and (5) take the following form:

3
It is interesting to note that with 0 these solutions split into a chain of
2D problems studied in [5-7].

i/x i /z x a cos(z + ) A+
)
s
(

= (/y + i /z aa cos(z + ) ) A ,

(19)
(i/x + i /z + x + as cos(z + ) ) A

= (/y i /z + aa cos(z + ) ) A+ .

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 6. The dispersion curves of Bloch modes in the presence of toroidal perturbations. All the possible coupling events caused by a
particular incident mode are indicated as circles.

An important feature of these equations is that the symmetric and asymmetric perturbations cannot be as clearly separated as in previous cases. Moreover, by a formal substitution
both perturbations may be concentrated in either the right- or
left-hand side of these equations. In particular, introducing:


A = B(x , z ) exp s sin(z + )+ ik yy


i

In spite of their numerical toughness, equations(22) allow


a rather clear qualitative analysis in terms of the geometrical
optics. First, note that when the coupling is ignored, each
mode propagates along x with wave vector k n = x n,
where the plus and minus signs correspond to the forward and
backward propagation. These wavevectors form a grid combined of parallel straight lines on the kx, x-plane. The intersection of these lines, k n+ = k m, is indeed a condition for the
effective scattering of mode n into the counter-propagating
mode m. So, in the vicinity of these points the dispersion
curves are split, as shown in figure6. One can see that in contrast to the poloidal case, where all modes are reflected at the
same point, here the interaction regions for the consequent
modes are separated in space with step / 2. This has two
consequences. First, the effects of the spectrum modification
due to the interaction of the co-traveling waves (important for
poloidal perturbations) is not pronounced. Second, as a good
approximation all reflection points may be treated independently, which should work at least when 1. The isolated
coupling efficiency Tn, m between modes n and m can be easily
calculated from equations(22) with only two particular modes
being considered. In this way one obtains the parabolic cylinder equationssimilar to equation(15) with standard transmission and reflection coefficients:

(20)

we arrive at:


(i/x i /z x) B = (ik y + i /z a cos(z+ )) B ,

(i/x + i /z + x) B = (ik y i /z + a cos(z+ )) B ,


(21)
where a = as + aa is a sum of the symmetric and asymmetric
contributions. Using Blochs theorem, solutions to these equations may be found as a convolution of independent modes
corresponding to a quasi-wavevector kz:
/2

B =

/2

dkz

n =

Bn(x , kz ) exp [i(kz + n )z ] .

Then, equation (21) are reduced to a set of ordinary differential equations:


(i/x x + ) B + =(ik )B 1 a(B + B ),
n
n
y
n n

n1
n+1
2

(i/x + x n ) Bn =(ik y + n )Bn+ + 1 a(Bn+ 1 + Bn++ 1),

2
(22)

Tn, n = tn = exp(k y2 2 n2 ),
Tn, n 1 = t =

exp( 1 a2 ),
4

Rn, n = 1 tn,
Rn, n 1 = 1 t .

(23)

The modes with other indices do not couple within this model.
Our approximation is valid for >1, therefore there is a strong
reflection Tn, n 1, for all modes except there may be resonances with k y 1, n 1. Now we are ready to construct
the net transmission coefficient Tn for a mode with given n.
As shown in figure6 as circles and arrows, there are only six

where n = kz + n. Unfortunately, this system cannot be


diagonalized for the left- and right-hand sides simultaneously
(this is equal to the inability of the separation of the variables
in equations(20) and (21)), but it can be effectively integrated
numerically using the impedance technique proposed in [12].
8

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 7. The transmission coefficient of the lowest mode with n = 0, k y = 0, kz = 0 (left panel) and kz = 1 (right panel) in the

presence of perturbations with different toroidal wavenumbers . Analytical estimates are indicated as dashed lines: (1) T = exp( 1 aa2 )
2
[limit 1], (2) T = t , t = exp( 1 (as + aa )2 ) [limit 1, kz 1], (3) T = 2t (1 t ) [limit 1, kz ~ 1], and (4)
4
T0 = exp( kz2 ) [no fluctuations]. The fluctuation amplitudes are as = 0.3, aa = 0.3.

possible coupling processes for each incident mode. The net


transmission coefficient is a sum over all channels:

figure7 (right). The latter is not important for optimized wave


beams, so the simple equation(25) for >
1 and the static formula for 1 may be recommended for fusion applications.
It is interesting to note that opposite to the poloidal case, the
influence of the toroidal fluctuations tends to increase for
shorter wavelengths.

Tn = Tn, n 1Tn, nTn, n + 1 + Tn, n 1Rn, nRn, n 1


+ Rn, n 1Rn 1, n 1Tn 1, n + Rn, n 1Tn 1, n 1Rn 1, n 2 .

The terms here describe n n, n n 1, n n 1 and


n n 2 transitions, correspondingly. Note that there are
two channels resulting in n n 1 transmission, so in principle their interference is possible. Here for the sake of simplicity we neglect (average) this interference assuming the
random phase of the initial perturbations. The above formula
may be simplified using (23):


Tn = t 2tn + t (1 t )(2 tn tn + 1)+ (1 t )2 tn + 1 .

6. Comparison to geometrical optics


As mentioned in the Introduction, a standard approach to
account for the density fluctuations in the OX coupling process is based on geometrical optics. Following [2], let us consider the fluctuations with poloidally directed wavevectors,
which introduce an effective poloidal beam divergence much
higher than the intrinsic one due to a finite aperture. The probability density function of the poloidal component k y (similar
to a poloidal beam divergence) takes the following form in
our notations4:

(24)

Assuming that tn 1 one obtains Tn 2t (1 t ). Note that


without perturbations this channel is closed, while the scattering
on the perturbations results in cutting the highly reflecting point
from behind (see figure6). However this effect is pronounced
for the non-optimized beams with a small energy content in
the n = 0 mode, so is of minor importance in a fusion experiment. On the other hand, the maximum transmission efficiency
is expected for the lowest mode corresponding to n = 0. In this
case t1 t0 ~ 1 and equation(24) results in:


T0 [1 (2t 1)(1 t0 )] t .

p ( k y )=

1
2
2k y
2
exp
.
(as + aa )k 0L
2 (as + aa )k 0L

In many senses this statistical description is equivalent to


averaging over a random face of harmonic perturbation in our
theory. Then the OX conversion efficiency of the optimal
plane wave may be calculated as:

(25)

In particular, T0 t = exp[ 1 (as + aa )2 ] is for the optimal


4
mode with k y = kz = 0. It should be stressed that this is potentially the largest effect of density perturbation since here the
symmetric contribution acts in the same order as the asymmetric one. At the opposite limit of a static perturbation,
0, equations(19) result in the phase-averaged transmission coefficient T = exp( 1 aa2 ), which depends on the asym2
metric perturbations only. As expected, formula (24), which
gives T t 2 for a small , is inconsistent with the static limit.
Our qualitative theory is in reasonably good agreement
with the results of the numerical integration of equations(19)
in cases of good transmission, see figure7 (left), and slightly
overestimates the transmission in highly reflecting cases, see

T=

p(k y)exp(k y2) dk y

= 1 / 1 + (as + aa )2 2(k 0L)2 / 2 .

Of course this formula is obtained assuming the fluctuation


scale length being large compared to the wavelength of the
cyclotron waves, k 0L.
One can compare the above result to equation(18) and find
it inconsistent with our theory. In particular:
We use the equationfrom page 3468 [2] with Ny = ky/L 1, y = 2L/,
x =(as + aa )L.

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

(a) Geometrical optics gives a large factor (k 0L)2 k 0L n 1,


which considerably enhances the impact of the
fluctuations.
(b) Geometrical optics cannot distinguish the asymmetrical
contribution of the fluctuations, which is the only important contribution for the poloidal fluctuations.
(c) Geometrical optics does not reproduce the clear static
limit with 0.
(d) Geometrical optics predicts the increasing role of short
wave fluctuations, T 0 with or until the approximation is valid. While the wave approach predicts the
opposite: very short waves do not affect the coupling at all.
(e) Geometrical optics predicts a similarity low corresponding to n e / fl = const, where fl is a wavelength of
fluctuations; however, this is not in agreement with the
numerical simulations.

Figure 8. The transmission coefficient as a function of the aspect


ratio LB / L n for k 0L n = 10, 20, ..., 60 (greater k 0L values correspond
to a greater transmission coefficient). The fluctuation amplitudes
are ne / n e = 10% (red dashed lines) and n e / n e = 20% (navy
solid lines). The plot is drawn for / ce = 1, which corresponds to
N = 1 / 2 and = 3 / 2.

All these issues systematically and considerably reduce the


OX conversion efficiency as compared to our results obtained
within a first-principles wave model. So we assume that the
geometro-optical approach may not be adequate, thus the role
of the plasma density fluctuations in the OXB experiment
needs to be revised.

conceivable experimental parameters. However, in a


spherical tokamak the asymmetric part of the density perturbations may be significant inside the coupling region
resulting in a degradation of the transmission efficiency.
In figure8 we illustrate the transition from low to large
aspect ratios for the realistic (red lines) and very high
(navy lines) fluctuation levels. One can see, for example,
that for k 0L = 60 fluctuations below 20% result in the
degradation of the transmission efficiency less than 10%,
even in a compact tori, and fluctuations below 10% can
always be neglected.

7. Discussion, application to the fusion experiment


Summarizing the results of the formal theory, we may state
the following basic conclusions for the transmission of cyclotron waves through the cut-off layer in the low reflection conditions close to the optimal OX mode conversion.

(c) The above approach underestimates the effects in the case


of plasma density modulated in a toroidal direction. In
this case the symmetric response is significant for not
very long fluctuations ( 1), so there is no aspect ratio
factor in the coupling efficiency:

(a) Apart from a specific case of perturbations along the


magnetic field, the plasma density fluctuations modify
the transmission efficiency inside the mode-conversion
region only in the presence of the regular inhomogeneity
of magnetic field strength or toroidal curvature due to
the asymmetric response in the coupling equations.
This results in reducing the effective amplitude of the
fluctuations by a factor L n / LB ~ 1 / A, which is small in
conventional large aspect ratio devices.

T = exp 1 (as + aa )2

= exp
4

(b) We do not find a strong dependence over a fluctuation


wavenumber. The effects of the fluctuations remain
limited in both limiting cases of the long and short scale
perturbations. A physically intuitive static approach
would give a reasonable estimate for the degradation of
the transmission efficiency averaged over a random phase
of fluctuations. In dimensional variables it is:

T = exp 1 aa2
2



= exp
4

n e2
k L (L n / LB )2 (1 + L n / LB )3/2 ,
2 0 n
2ce ncr
(26)

where is a factor of the order of unity defined in section2. For large aspect ratio devices such as conventional
tokamaks and stellarators, all the fluctuation effects are
strongly depressed by a small factor (L n / LB )2 for any

2 ne2
k L (1 + L n / LB )1/2 .
2 0 n
ce ncr

However, we are not sure that such density perturbations with wavevectors directed along the magnetic
field and amplitudes comparable to the poloidal or
radial turbulence level may be present in an experiment.
It should be stressed here that the radial and poloidal
density modulations are actually due to magnetic flux
tube oscillations, and they are present even if the density is homogeneous over a flux surface. Opposite to
this, a rigorous analysis of our model shows that the
z direction should be treated as a direction along the
magnetic field [8], so only the pure density perturbations on a flux surface should be accounted for in n e in
case of the toroidal modulation.

(d) The poloidal fluctuations can additionally modify the


amplitude and phase distributions in the incident quasioptical beam far outside the coupling region, which in

10

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

Figure 10. The transmission coefficient for a Gaussian beam

A exp(y 2 / 2w 2 ) as a function of a beam waist. The different


curves correspond to (1) the theoretical transmission efficiency
T = w / L2 + w 2 without fluctuations and curvature effects
being taken into account (black solid line), (2) the spread of the
transmission efficiency due to poloidal fluctuations and its phaseaveraged value (blue dashed line), and (3) the degradation of the
transmission efficiency due to poloidal curvature in the case of a flat
phase-front of the incident beam (gray solid line). The plot is drawn
for n e / n e = 15%, = 1 cm -1, k 0L n = 60, L n = 10 cm, LB = 20 cm,
/ ce = 1.

Figure 9. The length at which the transmission efficiency degrades

by factor two due to the optimal beam distortion in the presence of


poloidal fluctuations as a function of fluctuation level. The plot is
drawn for k 0L n = 10, 20, ..., 60, / ce = 1 and LB / L n = 6.

turn affects the consequent mode-conversion process


simply because the beam loses its optimal form during
its transit to the vicinity of the cut-off layer. Such a
beam distortion may be pronounced even in large aspect
ratio devices because it is sensitive to the symmetric
perturbations. Moreover, the effect is collected over a
trace much longer than the OX coupling region itself.
Aqualitative feeling of the scale of this effect can be got
from figure 3, however the degradation of the coupling
efficiency is overestimated because the statistical phase
jumps resulting in spectrum diffusion are neglected.
Alternatively, the length at which the optimal beam is
contaminated may be estimated noting that the diffusion
coefficient in k y-space is approximately D ~ s2 / fl
[13]. Assuming that the diffusion results in the Gaussian
spectrum with k y ~ D x , one obtains that the coupling
efficiency T = (1 + k y2 )1/2 drops from 1 to 1/2 at length
x ~ 1 / D (see figure9). As already mentioned, there are
separate studies on this process, which is itself completely
independent of the physics of the OX coupling at the
plasma cut-off, so we do not need to go into further detail
in this paper.

see the perfect match. For comparison, in figure10 we plot


the unperturbed 1D transmission efficiency without thefluctuations being taken into account. For narrow beams the
effects of the fluctuations are much more pronounced in the
spread than in the averaged value. Such a spread may appear,
for example, as anomalous noise in the spontaneous BXO
emission, which may be important in the peculiar plasma
diagnostics based on the BXO noise measurements [16].
However, the spread becomes negligibly small for wide
beams. The reason for this is that a variation of the phase
is formally equivalent to a shift of the whole perturbation in
the poloidal direction, see equations(10). Obviously such a
shift becomes unimportant when the beam is wider than the
fluctuation wavelength. This results in a very low effect of the
fluctuations for realistic beams with a waist greater than 3cm.
Note that the 2D theory results in an optimal beam waist that
lies in the same range [57].
Recently, an alternative mechanism explaining the degradation of the coupling efficiency in toroidal geometry
has been found in numerical simulations [17] and was later
explained theoretically [9]. In this paper we show that the
finite curvature of a cut-off surface may play a crucial role in
such a degradation. It is found that the poloidal and toroidal
curvatures may be interpreted as an additional phase variation across the incident beam aperture. This phase modulation results in the effective widening of the k-spectrum of a
wave beam, which leads to the significant degradation of the
OX coupling efficiency, even if the geometrical curvature
itself is weak as compared to the beam width. In figure10 we
illustrate this effect with line (3) representing the transmission coefficient for a beam focused exactly on the coupling
region such that the phase-front is flat there. One can see that
the potential degradation of the coupling efficiency due to

Therefore, tokamak fluctuations may in principle result in


the distortion of a wave beam on its way towards the mode
coupling region, but most likely they have no influence on the
efficiency of the mode interaction inside the coupling region.
The only possible exception is the case of small aspect ratio
devices with high turbulence. In figure 10 we illustrate
this case for the parameters close to the spherical tokamak
MAST [15]. Only the poloidal fluctuations are taken into
account. To make the situation more realistic, a Gaussian beam
with a finite poloidal waist and an optimal carrier wavevector
is considered. Opposite to our previous analysis related to the
transmission coefficients averaged over random fluctuations,
here we plot additionally the spread of the transmission coefficient corresponding to the different phases of the harmonic
perturbations. The spread is calculated numerically, while the
averaged transmission is given by equation(26), and one can
11

A G Shalashov and E D Gospodchikov

Plasma Phys. Control. Fusion 56 (2014) 125011

curvature effects is much greater than those due to the fluctuations even in an unfavorable case of a spherical tokamak.
The good news is that most of the curvature effects may be
avoided if the incident wave beam is focused in such a way
that it possesses a constant phase over the cut-off surface.
This tailoring of the optimal beam is technically possible, but
requires an accurate description of a magnetic configuration,
e.g. taking into account a triangularity, etc. However, this does
not require dealing with such poorly predictable phenomena
as plasma turbulence.
Finally, it should be noted that there are other effects that
may impact the OXB mode-conversion process in toroidal
plasmas such as parametric decay instabilities [18] and the
collisional damping of the electron Bernstein waves [19].
Both effects occur in the vicinity of the upper hybrid resonance after the OX conversion point is passed and the wave
is effectively trapped by dense plasma [20]. Therefore, in the
heating experiment these effects lead mainly to the redistribution of the rf power deposition, but do not affect the net
coupling between the incident radiation and plasma (except
a small amount of energy escaping as a low-frequency wave
emission in the case of parametric decay). On the other hand,
in the BXO emission experiment the nonlinear effects are
not pronounced, while the collisional damping at the plasma
periphery may shield the Bernstein waves coming from the
central parts.

OXB heating experiment both in spherical tokamaks and


large aspect ration traps.

Acknowledgments
The authors are grateful to Timur Khusainov for his valuable
and inspiring discussions. This research was supported by
RFBR (projects 12-02-00648, 14-02-31024) and the President
Council for Grants (project MD-1736.2014.2).
References
[1] LaquaH P 2007 Plasma Phys. Control. Fusion 49 R142
[2] LaquaH P et al 1997 Phys. Rev. Lett. 78 3467
[3] MjlhusE 1984 J. Plasma Phys. 31 7
[4] ZharovA A 1984 Sov. J. Plasma Phys. 10 642
[5] GospodchikovE D, ShalashovA G and SuvorovE V 2006
Plasma Phys. Control. Fusion 48 869
[6] ShalashovA G, GospodchikovE D and SuvorovE V 2006
Sov. Phys.JETP 103 480
[7] GospodchikovE D, ShalashovA G and SuvorovE V 2008
Fusion Sci. Technol. 53 261
[8] ShalashovA G and GospodchikovE D 2010 Plasma Phys.
Control. Fusion 52 115001
[9] GospodchikovE D, KhusainovT A and ShalashovA G 2012
Plasma Phys. Control. Fusion 54 045009
[10] TimofeevA V 2004 Phys.Usp. 47 55582
[11] BuddenK G 1952 Proc. R. Soc. Lond. A 215 21533
[12] ShalashovA G and GospodchikovE D 2010 Plasma Phys.
Control. Fusion 52 025007
[13] SysoevaE V, GusakovE Z and HeurauxS 2013 Plasma Phys.
Control. Fusion 55 115001
[14] WhittakerE T and WatsonG N 1952 A Course of Modern
Analysis vol 2 (Cambridge: Cambridge University Press)
[15] PreinhaelterJ, ShevchenkoV, IrzakM A, VahalaL and
VahalaG 2000 ECRH in spherical plasmas: OXEBW
mode conversion in MAST Report UKAEA FUS 444
Culham Science Centre, Culham, UK
[16] PopovA and IrzakM 2014 Plasma Phys. Control. Fusion
56250029
[17] KhnA et al 2008 Plasma Phys. Control. Fusion 50 085018
[18] GusakovE Z and SurkovA V 2007 Plasma Phys. Control.
Fusion 49 6319
[19] DiemS J et al 2009 Phys. Rev. Lett. 103 015002
[20] BalakinaM A, ShalashovA G, GospodchikovE D and
SmolyakovaO B 2006 Radiophys. Quantum Electron.
4961732

8.Conclusion
Based on the developed theory we revise the role of the
fluctuations in the OXB experiment. We find that plasma
fluctuations most likely cannot be responsible for the low efficiency of the OXB heating of overdense plasma observed
in present-day experiments. We guess that the main factor
impeding the tunneling of the electromagnetic waves through
the plasma cut-off in these experiments may be associated
with a curvature of the toroidal magnetic surfaces. It seems
that such a degradation may be quite essential for the typical
beams used in fusion installations; however unlike all turbulence effects, the curvature effect may be compensated by a
proper tailoring of the optimal beam. Accordingly, this should
be taken into account in the interpretation and planning of the

12

Das könnte Ihnen auch gefallen