30May 1994

PHYSICS LETTERS A

ELSEVIER PhysicsLetters A 188 (1994) 376-383

A note on the toroidal plasma dispersion function

Marco Brambilla
Max-Planck-Institut f~r Plasmaphysik, Euratom Association, Garching near Munich, Germany

Received 24 February 1994;accepted for publication 17 March 1994

Communicatedby M. Porkolab

Abstract

We have found a simple yet accurate approximation for the toroidal plasma dispersion function, which described the effects
of parallel space dispersion on the propagation and absorption of waves near ion cyclotron resonances in tokamak plasmas. It
can be easily implemented in numerical codes solving the finite Larmor radius wave equations in toroidal geometry, allowing a
quantitative discussion of the effects of toroidicity on ion cyclotron damping and on linear mode conversion to ion Bernstein
waves.

1. Introduction

Due to its importance for auxiliary plasma heating, considerable effort is being devoted to the theoretical and
numerical modelling of ion cyclotron (IC) heating, and in particular to the solution of the wave equations

V×VXE= -~ E+ JP (1)

in toroidal geometry (for a review, cf. Ref. [ 1 ] ). In a tokamak, the h.f. plasma current JP is found to depend on
integrals of the form [2-4]

Lr(E)=
;
--oo N/r~ Utha
[ }
dvll exp[--(--vlt/v--2th~)2] -ia~
.-oo
(i
dt'exp i
U
(o9-12g) dr r-
t'
(vnlb) dz
)] , (2)

where for each species the t'-integration is along magnetic field lines, with b = Bo/Bo, and vth = x/~T/m. The non-
local nature of the relation between JP and E described by ~ (E) is due to the fact that a hot plasma is a spatially
dispersive medium. This, combined with the effects of toroidicity on the ion motion, gives rise to qualitatively
new features of ion cyclotron resonances, which do not have a counterpart in the homogeneous plasma theory.
In this note we show how the most important of such effects, namely the toroidal broadening of the cyclotron
damping layer over its thermal Doppler width, can be taken into account accurately with negligible numerical
effort, and we discuss its influence on the efficiency of ion cyclotron damping and of linear mode conversion of
the externally excited magnetosonic waves into ion Bernstein waves near ion-ion resonances.

0375-9601/94/$07.00 © 1994 Elsevier ScienceB.V. All rights reserved

SSDI0375-9601 (94)00224-D
 M. Brambilla / Physics Letters A 188 (1994) 376-383 377

2. The toroidal plasma dispersion function

Eq. ( 1 ) can be solved assuming a solution of the form [5]

E~=exp(in~0) ~ E'~(n~; 9,) exp(imO) (3)

m~ --oo

( ~, labels magnetic surfaces, 0 and ~oare the poloidal and toroidal angles, respectively). Because of axisymmetry
there is no coupling between different n~ components; in the following the argument n~ will be omitted. Poloidal
modes on the other hand are strongly coupled by toroidicity. Defining

n~
k~= m
r sin O+ ~- cos O (4)

with tan O=Bpo~/Btor (for simplicity we have used notations appropriate to circular plasmas, but the extension
to the general case is trivial) and

o~- nt2c~ (5)

x,.,~ =x,,,,~(m, ~, O) - k'~vth,~

the ansatz (4) transforms the integrals (2) into the algebraic equation

~(ve)= ~ [-Xo~Z(x.~)].Em, (6)

m~ ~oo

with

-Xo~2(x,,,~)= du e x p ( - u 2 ) -ioJ exp i k'~Pth(X'~a--U" ) dt" dt' . (7)

--oo v/r~ -oo tt

The function 2 thus defined might be called the toroidal plasma dispersion function (PDF). It reduces to the
Fried-Conte PDF [ 6 ]
1 e xup-(x-. u2) du+ix/-~ exp ( - x ~ ) (8)
Z(x.) = ~,~ -~

in the limit of a homogeneous plasma. Note that in the toroidal case the argument x,~ depends explicitly on g/
and O, and is different for each poloidal mode.
The integrals (7) have formally to be evaluated according to the Landau prescription, Ira(co)~0 +. If this
limit is taken literally, however, they show resonances not only at the harmonics of the cyclotron frequency, but
also at the harmonics of the toroidal bounce frequencies of passing and trapped particles [ 2-4 ]. This is not only
analytically and numerically untractable, but also physically incorrect. As discussed e.g. in Ref. [ 4 ], collisional
phase diffusion efficiently destroys the phase memory of the particles, thereby eliminating the bounce frequency
dependence of the plasma response. Hamiltonian stochasticity induced by repeated resonant wave-particle in-
teractions can lead to the same effect [ 7 ].
Because of phase diffusion, only the contribution from the last stationary point in the z integral, i.e. from the
last transit through a cyclotron resonance, has to be taken into account in Eq. (7). Models for the resulting
toroidal PDF have been discussed in [ 8-12 ]. The most satisfactory treatment is due to Lamalle [ 12 ], who has
also indicated how to evaluate the integrals (7) for more general distribution functions. The expressions ob-
tained in [ 12 ] for 2 are also consistent with the quasilinear diffusion coefficients which should be used in the
kinetic equation for the steady-state distribution function in toroidal geometry [ 13,14 ]; they therefore would
378 M. Brambilla / Physics Letters A 188 (1994) 376-383

allow a truly self-consistent treatment of ion cyclotron heating. A different method, based on the solution of the
Vlasov equation in action-angle variables, has been proposed in Ref. [ 15 ].

3. Approximating the toroidal plasma dispersion function

Unfortunately, a wave code taking into account all details of the particle motion along magnetic field lines
would be extremely slow, since different classes of ions (passing, trapped on banana orbits which cross reso-
nance, deeply trapped on banana orbits which do no more cross resonance) must be treated differently. This
makes the evaluation of Z a very complicated task, which becomes prohibitive when it must be performed at
each point of a huge two-dimensional mesh in the poloidal plane, and separately for each poloidal wavenumber
m. Simple considerations, on the other hand, suggest that such a complicated approach is not really necessary.
In the first place [ 16 ], as long as the wavelengths are large compared with the thermal ion Larmor radius, the
coefficients of the wave equations depend only on the distribution of parallel velocities (assumed to be Max-
wellian in Eq. ( 1 ) ) and on low order momenta of the distribution of perpendicular velocity, namely (simplify-
ing slightly) the density and the perpendicular pressure. In the second place, ion cyclotron resonances are always
optically thin or at most marginally thick: this means that by far the largest part of the absorption is due to
particles in the thermal domain, whose steady-state distribution function is relatively insensitive to toroidal
trapping. Thus when solving the wave equations neglecting toroidal trapping should be a fair approximation.
In the simplest approximation, retaining only quadratic terms in the expansion of the phase around the sta-
tionary point, the velocity integral can then be performed analytically to give

2~(x,) -~ Z(x,, y) = - i ~ exp [ix, r - ¼z2(1 +yZ) 2 ] exp( - ~vefrz3) dz, (9)
0

with

Y" 2k~-~vth [slnOsinO[ (I0)

evaluated at the cyclotron resonance. This approximation is equivalent to assuming that ions cross the reso-
nance layer with constant velocity, and strictly speaking should not be used in the limit 0 = 0 (magnetic surface
tangent to the cyclotron resonance layer). Excluding this case, for typical tokamak parameters y is of order unity,
and can be much larger than unity for small values of kjl. Phase diffusion is explicitly modelled by the last
exponential factor in the integrand, with vefr--- Uco11/Iku Ivth [ 10] (u~fr~ 1/co11~2/Ik~ [V3haccording to Ref. [ 13 ] ).
For realistic values of the parameters, however, this factor has a negligible influence on the behaviour of Z. For
large argument, Ix, I >> 1, the function Z(x,, y) has the same asymptotic behaviour as the Fried-Conte PDF
(8), to which it also reduces when y-,0; ~(x,, y) differs appreciably from Z(x,) only in the resonance layer
Ix, I = O ( 1 ) where dispersion and damping are important.
In spite of the simpler form of Eq. (9), the numerical evaluation of the z-integral is still rather lengthy, and
would largely dominate the execution time of a wave code if implemented without further approximations. In
addition, it is far from obvious that (9) contains the correct ion-cyclotron resonance physics, since, as shown in
Fig. 1, the imaginary part of Z(x,, y) is not negative everywhere. As already noticed in Ref. [ 9 ], this gives rise
to difficulties in interpreting the local power balance, since it makes the local power deposition by ion cyclotron
damping negative in some regions even in a Maxwellian plasma. By a more careful treatment of the parallel
particle motion, Lamalle [ 12 ] has shown that the local power deposition in a Maxwellian plasma is positive
defined everywhere, as it should [ 17 ].
The other important (and real) effect of toroidicity expressed through the parameter y is an appreciable
broadening of the resonance region. Toroidal broadening, as already predicted in [ 15 ], decreases the efficiency
 M. Brambilla / Physics Letters A 188 (1994) 376-383 379

70.3

Re(Z) ,
I
I

35.1 I I I

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0.0

t
-,.35.1 \
\,

-70.3 Z 1 1 I
-0.32 -0.19 -0.06 0.06 0.19 X 0.32

10. I
I

--,----,.--,,-.
I I I
! J I
-18. t I I
I I I
! f !
t I I

-45.
I !
~,,~
I !
tu,,
i I
I I I
It!
li I
-73. Ill
II I
II I
II!

Ira(z) I
I
I
-I00. I 1 r ! :
-0.32 -0.19 -0.06 0.06 0.19 X 0.32

Fig. 1. Approximation to the toroidal plasma dispersion function. ASDEX Upgrade parameters (R = 1.65 m, f = 30 MHz, np = I 0, Ti = 2
keV). Thick line: Eq. ( I 1 ); thin line: Z(x, 7), Eq. (9); dotted line: uniform plasma Z(x), Eq. (8).

of mode conversion to ion Bernstein waves compared to the predictions made in plane-stratified geometry. In
minority heating scenarios, moreover, it shifts the transition between the mode-conversion dominated regime
and the regime dominated by ion cyclotron damping towards higher minority concentrations or lower toroidal
wavenumbers. A very good approximation to the toroidal PDF which takes this effect into account consists in
replacing 2 with the ordinary PDF Z with argument evaluated using an appropriate effective parallel wavenumber

-Xo2(X~, y) ___-XoZ(X~),
w-nI2¢
~n - ( k, )efrv,h ' (I 1 )

with (kll) errgiven by

380 M. Brambilla / Physics Letters A 188 (1994) 3 76-383

(k,)eff~-k, x/1 + 4 y - 1 (12)

2y
This equation can be understood by noting that Lr~s ~- (2RVth/nO sin 0 ) 1/2 is the resonance length of a single
particle; this length adds to the thermal Doppler broadening (k~jvth/co)R of the cyclotron resonance layer. In
particular (k,)err remains finite even in the limit ktl--,0:

(k,,)~ff_.+(O)2__~Vthsin 0 ) 1/2 for k,h--,0. (13)

As shown in Fig. 1, Eqs. ( 11 ), (12) offer an excellent approximation to the toroidal PDF (uniformly good for
all values of k, ), which reproduces accurately the toroidal broadening of the cyclotron resonance layer, while
suppressing the unphysical oscillatory behaviour of the imaginary part of Z which would follow from using (9).
At the same time Eq. ( 11 ) allows to take advantage of the efficient algorithms available to evaluate the function
Z [ 18 ]. Not taken into account by ( 11 ) are only the effects of toroidal trapping, as described e.g. in Ref. [ 12 ].
These effects are certainly important for ions in the tail of the distribution function, and will therefore influence
the behaviour of 2~ in the wings of the resonance region. From the arguments given above, however, we feel
entitled to claim that in practice ( 11 ) is always an adequate approximation.

4. An example

We have incorporated Eq. ( 11 ) into the toroidal wave code TORIC (a thoroughly revised version of the code
described in Ref. [ 5 ] ). Previously, the ordinary PDF Z ( x n ) with k ngiven by Eq. (4) was used in this code. We
would like to stress that this is not a "local approximation" omitting toroidal effects, as sometimes stated in the
literature [ 12 ]. Comparison with Eq. ( 11 ) should make clear that using Z ( x ~ ) , whose argument is a function
of both rn and O, simply underestimates that width of cyclotron resonances by the amount Lres. On the other
hand it already takes into account the other most important toroidal effect, namely the dispersion of k, values
due to its dependence on the poloidal mode m. By smoothing out somewhat the 0 variation of the toroidal PDF,
toroidal broadening of the IC resonances actually decreases coupling between poloidal modes, thus reducing the
toroidal spread of k v Comparing results of TORIC with and without toroidal broadening of the IC resonances,
and of the slab code FELICE [ 19 ] in which k~l has a unique, constant value, allows one to estimate the relative
importance of the two effects.
Fig. 2 shows the electric field along the equatorial plane of ASDEX Upgrade (R = 165 cm, a = 50 cm, Bo = 2
T, n~=0.8× 1019 m -3, Ti= Te=2 keV on axis, 5% H + in D + , f = 30 MHz, n~= 10) evaluated with FELICE and
with TORIC in the two approximations; the fractions of power damped in the various channels are summarized
in Table 1. The trends are just as expected. The relatively large electron heating predicted by FELICE is due to
the moderate standing wave developing in the outer half of the plasma, between the antenna and the cutoff
associated with the ion-ion resonance (in slab geometry the chosen value of kll ~- 6.06 m - 1 is at about the tran-
sition between the low kll regime dominated by mode conversion, and the high k u regime dominated by ion
cyclotron damping). The spread of klLvalues due to toroidicity reduces the standing wave ratio, and thus favors
minority cyclotron damping compared with electron damping. Since lower values of k~ are present as well,
however, the efficiency of mode conversion into ion Bernstein waves is only moderately reduced. Toroidal
broadening of the cyclotron resonance further increases somewhat the efficiency of ion cyclotron damping. Its
most important effect, however, is to shift the transition to the mode-conversion regime towards lower values
of klL,thereby reducing drastically the excitation of ion Bernstein waves (the relatively large amplitude of these
waves in spite of the tiny amount of power transported is due to their very low group velocity). Not surprisingly,
the power deposition profile is also broadened, as shown in Fig. 3.
 0 .0 19

Ex (V/cm) i
i
(for 1 A in the antenna short) , A
i-
0.010 /
/
Ill II
,~ n
I II
'
#
t
/
0.000
...... ~':,::,, , t, ,_/ _._ _;_ _L
V i

-0.009

a) Slab geometry i
-0.018 r I 1 I
-50,0 ~29.0 --8.0 13.0 34.0 X 5s.o

0.47
T
E+ (V/m) II (for 1 A in the antenna short)

0.17

--0.14

-0.44

b) Without toroid£1 "

broadening. .
-0.75 I I E I
-50.0 --30.0 --10.0 10.0 30.0 ~ 50.0

0.19
it o
:"I:P:] ~ I |l It I ~
Ill I I I I l I l II i ~ P'%
. f X
IJ"
tflll I ii k l tt l l

I I I " I i It
-0.07

-0.32

-0.57 c) W i t h toroidal
E÷ (V/m) broadening.
(for 1 A in the antenna short)
-0.82

-50.0 -30.0 -10.0 10.0 30.0 X 50.0

Fig. 2. Electric field in the equatorial plane of ASDEX Upgrade ( parameters given in the text): (a) FELICE (slab geometry); (b) TORIC
without toroidal broadening of cyclotron resonances; (c) TORIC taking toroidal broadening into account.
382 M. Brambilla /Physics Letters A 188 (1994) 376-383

Table 1

Absorption channel FELICE (slab) TORIC with Eq. ( 8 ) TORIC with Eq. ( 11 )

electrons 23.99% 4.71% 3.67%

H÷ 54.91% 70.13% 79.00%
D+ 15.65% 21.58% 17.25%
Bernstein waves 5.51% 3.57% 0.07%

0.050

W/cm 3 (for 1 A in the antenna short)

0.038

H+
O. 025 :D+

0,013
a) Without toroidal
broadening.
o,ooo : I
o.oo 0.20 0.40 o.~o o.8o ~ loo

O.050
/ ~ W/cm 3
o. 022 / ~ (for 1 A in the antenna short)

i
0.015 iN+ /~
0.o08 ", k..x b) With toroidal
',.. ~ broadening.
e '..
0 ooo e . . . . . :
o.oo 0.20 0.40 0.60 0.80 1.g 1.0o

Fig. 3. Power deposition profiles in toroidal geometry: (a) without toroidal broadening of cyclotron resonances; (b) taking toroidal
broadening into account.

5. Conclusions

We have f o u n d an a p p r o x i m a t i o n to the t o r o i d a l constitutive relation which takes accurately into account the
most i m p o r t a n t t o r o i d a l effect on ion cyclotron resonances, n a m e l y their b r o a d e n i n g due to the finite w i d t h o f
 M. Brambilla / Physics Letters A 188 (1994) 376-383 383

the resonance of single ions, and which can be incorporated in toroidal wave codes with a minimal numerical
effort. This approximation does not take into account toroidal trapping of energetic ions; we argue however that
toroidal trapping has only a minor influence on wave propagation and absorption. The comparison of results
from a toroidal and a slab wave code confirms the expected effects of toroidal broadening: an increase in the
efficiency of ion cyclotron damping and a reduction of the mode conversion efficiency near ion-ion resonances.
The results suggest however that the most important effect of toroidicity is the spread of the kjj spectrum due to
the poloidal inhomogeneity, the broadening of the ion cyclotron resonance interactions playing only a second-
ary, although nonnegligible, role. Since Lres scales as R t/2 while the thermal width of the resonance scales as R,
however, this role will decrease further with increasing size of the plasma.

References

[ 1 ] M. Brambilla, Plasm Phys. Contr. Fus. 35 (1993) A141.

[2] G. Cattanei and R. Croci, Nucl. Fus. 17 (1977) 239.
[ 3 ] T.D. Kaladze and A.B. Mikhailowskii, Sov. J. Plasma Phys. 1 ( 1975 ) 128.
[4] S.V. Kasilov, A.I. Pyatak and K.M. Stepanov, Nucl. Fus. 30 (1990) 2467.
[5] M. Brambilla and T. Krficken, Nucl. Fus. 28 (1988 ) 1813.
[ 6 ] B.D. Fried and S.D. Conte, The plasma dispersion function (Academic Press, New York, 1961 ).
[7] A. Becoulet, D.J. Gambier, J.-M. Rax and A. Samain, Proc. 8th Top. Conf. on Applications of RF power to plasmas, Irvine, CA,
USA (1989) p. 197.
[8] S.I. Itoh, A. Fukuyama, K. Itoh and K. Nishikawa, J. Phys. Soc. Japan 54 (1985) 1800.
[9] D.W. Faulkoner, Plasma Phys. Contr. Fus. 29 (1987) 433.
[ 10] D. Smithe, P. Colestock, T. Kammash and R. Kashuba, Phys. Rev. Lett. 60 (1988) 801.
[ 11 ] P.J. Catto and C.N. Lashmore-Davies, Europ. Topical Conf. on Radiofrequency heating and current drive of fusion devices, Brussels
(1992) p. 165.
[ 12 ] P.U. LamaUe, Phys. Left. A 175 ( 1993 ) 45.
[ 13 ] G.D. Kerbel and M.G. McCoy, Phys. Fluids 28 ( 1985 ) 3629.
[ 14 ] P.J. Catto and J.R. Myra, Phys. Fluids B 4 (1992) 187.
[ 15 ] D.J. Gambier and A. Samain, Nucl. Fus. 25 ( 1985 ) 283.
[ 16 ] M. Brambilla and C. Hoffman, 20th EPS Conf. on Controlled fusion and plasma physics, Vol. III, Lissabon (1993) p. 957.
[17] M. Brambilla and T. Krficken, Plasma Phys. Contr. Fus. 30 (1988) 1083.
[ 18] W. Gautschi, SIAM Rev. 9 (1967) 24.
[ 19] M. Brambilla, Plasma Phys. Contr. Fus. 31 (1989) 723.