Plasma Physics and Controlled Fusion

Related content
- Transport induced by ion-impurity friction
Toroidal rotation and momentum transport in strongly rotating, collisional tokamak
plasma
C T Hsu and D J Sigmar
To cite this article: J W Connor et al 1987 Plasma Phys. Control. Fusion 29 919
- Gyroviscous momentum transfer and
toroidal rotation in Tokamaks with
unbalanced NBI (authors' reply)
J W Connor, S C Cowley and R J Hastie

View the article online for updates and enhancements. - Gyroviscous momentum transfer and
toroidal rotation in Tokamaks with
unbalanced NBI
W M Stacey Jr

Recent citations
- Necessary Extensions and Modification of
Fluid Transport Theory for the Tokamak
Plasma Edge
W. M. Stacey

- Poloidal tilting symmetry of high order

tokamak flux surface shaping in
gyrokinetics
Justin Ball et al

- Investigations of turbulent transport and

intrinsic torque of toroidal momentum at
the edge of J-TEXT tokamak with
electrode biasing
Yue Sun et al

This content was downloaded from IP address 132.239.225.140 on 24/04/2018 at 21:48

Plasma Phyiicc and Controlled Fusion. Vol. 29. No. 7 . pp. 919 to 931, 1987 0741-3335 8753 00+ .OO
Printed in Great Britain IOP Publishing Ltd. and Pergamon Journals Ltd.

TOROIDAL ROTATION AND MOMENTUM TRANSPORT

J. W. COKKOR,S. C. COWLEY,
R. J. HASTIEand L. R. PAN*
Culham Laboratory, Abingdon, Oxon OX14 3DB, U.K.
(Euratom:UKAEA Fusion Association)
(Receii,ed 1 December 1986; unci it7 reci.red.form 13 Februroj, 1987)

Abstract--In this paper we re-examine the transport of momentum in a collisional plasma rotating at
sonic speeds. The slowing down rate due to collisional transport is found to be classical (T,,;' ~ ' , ~c?)p , ~
and the presence of impurities does not change the order of magnitude of this theoretical result. Thus
-
collisional transport theory cannot explain the anomalously high loss rates observed experimentally.
Some existing theories of collisional momentum loss are discussed in detail.

I. I N T R O D U C T I O N
IT IS well documented that many Tokamaks exhibit anomalously high rates of
toroidal momentum loss. Electron, and possibly ion, energy are also lost at rates
greater than predicted by neo-classical theory and it is common to seek an explana-
tion for this in terms of turbulent transport from some micro-instability. It is there-
fore natural to suppose a similar mechanism is responsible for the loss of toroidal
momentum and one such, ion temperature gradient turbulence, has been suggested,
(MATTORet ul., 1986).
However, STACEY and SIGMAR(1984 and 1985) have advanced the hypothesis
that, in a torus, classical gyroviscosity allied to inertial effects due to sonic flows, can
account for the experimental observations. This contrasts with an earlier discussion
by HOGAN(1984) of the same problem based on the full Braginskii stress tensor
which led to a Pfirsch-Schliiter enhancement of classical perpendicular viscosity.
We have therefore reconsidered this problem and show in Section I1 that, for an
up-down symmetric Tokamak. classical perpendicular viscosity (without a Pfirsch-
Schliiter enhancement) is appropriate. This result is consistent with kinetic treat-
ments of this problem for low flow velocities by ROSENBLUTHer al. (1971) and
HAZELTINE (1976) and for sonic flows, by HINTONand WONG (1985), (see also
COWLEY and BISHOP,1986).
Because of this confusion in the literature an important contribution in this paper
is the discussion, in Section 111, of the physical interpretation of our calculation and
the origin of the different results obtained by Stacey and Sigmar and by Hogan.

11. C A L C U L A T I O K O F C O L L I S I O N A L T O R O I D A L M O M E N T U M L O S S
In this section we derive a momentum diffusivity from Braginskii's stress tensor
(BRAGINSKII, 1965) given in Appendix A, so that our result is valid for a collisional
plasma in which imfp, the mean free path for all species, is small compared with the
macroscopic scale size. For simplicity we first consider the case with no temperature
*Institute of Mechanics. Chinese Academy of Sciences, Beijing, P.R.C. Academic Visitor, Imperial College,
London SW7 2A2, U.K.
@ Crown copyright.
919
920 J. W. COKKOR
et al.

gradient or impurities; the modifications due to temperature gradient are discussed

in Appendix B and the role of impurities is discussed in Appendix C.
In this situation the plasma is described by the following set of equations for each
species x :

= -Vp, + Z,en, -Vq + E, + ~

and

2B
-==VxE=-VxEI
2t

where the electric field E = - V q +

E, with E, the inductive component. v , is~ the
collision frequency between species a and /3, while S, is the particle source and P,
the momentum source for species a and the other symbols have their usual meaning.
The viscosity tensor n, given in Appendix A, has a decomposition

where no, is the dominant parallel component characterised by a coefficient of

- -
parallel viscosity qo n,T,/v,,, n33and n4aare the gyroviscous terms with viscosity
coefficients q 3 , 4 (v,,~oc,)qo and n l z and nzz are the perpendicular components
-
with q 1 , 2 (v,,/coc,)zqo. Here we have introduced the gyro-frequency w,, =
Z,eB'm, and it is also convenient to define the thermal velocity V[h3 = (2T,/m,)'
and gyroradius p a = Vlhz,oca.
We solve equations (1)-(5) by an expansion in p ~ wherea a is the minor radius,
treating 2mf,,/u as a subsidiary expansion parameter, and seeking an evolution of
toroidal momentum on a timescale

as will be justified a posteriori. Therefore with E labelling order in pila,

%(V) = no,(V) + E(%(V) + K4,(V)) + Ez(nl,(v)+ n,,(V)). (8)

 Toroidal rotation and momentum transport 921

Momentum sources do not change the order of magnitude of this answer if iV.n,l %
lPil and, since this is always satisfied, we consider the case Pi = 0 for simplicity.

(a) Low,est order flows

In an axisymmetric torus V-B = 0 implies that the magnetic field can be expressed as

B = V$ x eT
-
R
+ Z-R
eT
(9)

where R is the major radius of the torus, is the poloidal flux function and eT is a
unit vector in the toroidal symmetry direction.
When the equilibrium flow velocity is of order the sound speed the lowest order
ion momentum equation yields the familiar result

and the electrons obey an identical equation. The magnetic field is therefore frozen
into the zeroth order flow. By taking the scalar product of equation (10) with B we
deduce that cpo = cpo($). Since V q o is perpendicular to the flux surface Vpis in the
surface. We resolve Vp into two components, one parallel to B and one in the
toroidal direction (see Fig. 1). It follows from equation (10) that

where o = c 2 q o / 2 $ and ii is arbitrary. For stationarity on this fast timescale the

continuity equation [equation (l)] constrains this flow to be of the form

FIG.1.-The lowest order flow.

922 J. W. CONNORet al.

This form of the flow is discussed from a physical point of view in Section 111. We
will refer to the toroidal part of the zeroth order flow as the rigid rotator part. The
parallel flow contains a poloidal component which is strongly damped by parallel
viscosity [also known as magnetic pumping, (HASSAM and KULSRUD,19781. On the
long timescale of equation (7) we therefore expect Ki to be small. This is demon-
strated mathematically by considering the ion momentum equation in the next order
(
of the vii/wciexpansion. The ZienO E ~ + ~ term is annihilated by the opera-
dl
tion ImB., where dl is an element of length in the poloidal direction and B, is the
poloidal magnetic field. Physically, we are computing the flux surface average of the
work done by the parallel force on the parallel flow. Then, writing

(A)=
CI.
-A

we obtain the equation (temporarily retaining the effect of temperature variations)

Since the induced electric field E, is of order y J and the resistivity y -

mivie/ne2
the last two terms in equation (14) are comparable and give no contribution in zero
and first orders, while the time derivative also appears only in second order. In zero
order we insert Vp into equation (14) noting that the inertia term vanishes identically.
Using the form of zoi given in Appendix A we find that zOi(wRe,) = 0 so that,
ignoring temperature gradients, equation (14) reduces to

where b = B/B. The parallel viscosity therefore constrains the parallel flow to
be zero at this order. The relaxation of e
to zero clearly occurs on a timescale
Vthi/u2vii. We discuss the physical mechanism responsible for this relaxation in
Section 111.
To calculate the slowing down of the toroidal rotation o,($,t),we will still require
Kx$), the longitudinal part of the first order flow. KK$) is obtained from the ion
momentum equation in next order of the v/wcexpansion, but we delay the calculation
until certain equilibrium properties have been established, in particular the nature of
the poloidal variation of ni($,0.
 Toroidal rotation and momentum transport 923

(b) Equilibrium
Adding the momentum equations (equation 2) for the ions and electrons we obtain
the zeroth order pressure balance equation

+
wherep = ni(Ti Te)and J o = Znie(Vf - Vf), and to this order -4nJo = cV x B.
Taking the scalar product of equation (16) with B we obtain,

+
where cs = (2(Ti Te)/mi)' is the constant sound speed. The density variation in
the surface provides a parallel pressure gradient which balances the component of
the centrifugal force in the surface. The toroidal component of equation (16) yields
Jo.V$ = 0, which with equation (7)implies I o = Io($).The V $ component ofequation
(16) yields the equilibrium equation

where p = E(Ti + Te) and A* is the usual Grad-Shafranov operator,

A*$
a 12$
R--- + -.8%)
2RR2R aZ2

When temperature gradients are included in the problem (see Appendix B)

B.VTo = 0 and equation (18) is unchanged except that cs = e,($). The equilibrium
specified by equation (18) is determined by the three free flux functions p, o/c, and I
(MASCHKE and PERRIN, 1980), which evolve through the transport equations.

(c) Euolution of toroidul ungulur momentum

We derive an equation governing the evolution of toroidal angular momentum by
annihilating the leading order part of the momentum equation [equation (16)]. This
annihilator is JgReT., and, applying this to a plasma with a single ion species, we
obtain

Physically the terms in equation (20) represent, from left to right: the rate of change
924 J. W. CONNOR
et ul.

of angular momentum, the convection of angular momentum, the toroidal J x B

force and the viscous drag between surfaces.
First we show that the toroidal J x B and convection terms can be neglected. It
follows from Gauss's law and charge conservation that the toroidal J x B force is
also equal to the rate of change of charge inside the surface. Thus

(21)

where we have used equations (9j and (12). Comparing the J x B force to the rate
of change of angular momentum we find

where V , is the Alfven velocity V , = B/v/4nmini. Since V,/C < 1 in most experi-
ments, the toroidal J x B force will be ignored.
We may express the convection term (which involves the second order radial flow)
in terms of the lower order flows by taking the scalar product of the ion momentum
equation with wR3e, and flux surface averaging. T o second order we obtain

e
-(niwR2Vf.V$) = - mi(oR2V.(V!wR2ni))- (oR2V*n.eTR)
c
+ nzi(wR2vi,ni(Vi - V!).eT). (23)

The first order flow Vi can be calculated from the first order ion momentum equation

VR2
- minPw2- -Vql x ~

2 C

together with equation (17). We first obtain the potential 9' from the parallel
component of equations (24) and (17)

ecp' 1 miw2R2
-- - -
T, 2(T, +
Ti)'

This potential provides an electron density variation which ensures charge neutrality
when the ions are expelled by the centrifugal force. The perpendicular component of
equation (24) and the condition of stationary density in the continuity equation at
this order yield
 Toroidal rotation and momentum transport 925

where prime denotes differentiation with respect to $. Note that the diamagnetic
toroidal velocity has a contribution due to velocity shear.
One may verify that substituting Vy and V! into equation (23) yields, to second
order,

(nioR2Vi.V$) = 0 (27)

-
where we neglect the friction term because vie/vii ( w I J w I ~ ) " ~ Thus
. we conclude
that the toroidal J x B and convection terms in equation (20) may be neglected.
The slowing down is therefore determined by the viscous term in equation (20);
evaluation of this term is aided by two identities:

for any velocity, and

V$.z,i(wReT).e, = V**7C,,(wReT).eT = 0. (29)

Thus the viscous term in equation (20) is, to second order,

so that the gyroviscosity contributes the same order as the perpendicular viscosity.
The slowing down rate can therefore be calculated once K! is determined as we do
not require Vz.
K! is determined from the parallel work done at first order using equation (14). In
this order the friction term and time derivative in equation (14) are negligible, the
inertia term again vanishes identically, and we obtain,

At this point we exploit the fact that the collisionality parameter, imfplfu,is small
in the collisional regime. The contribution to equation (28) from the diamagnetic
parts of Vr in z,, and xqiis smaller than that from the perpendicular viscosity terms,
z I i and 7 c 2 i , by a factor i m f p / u and can be ignored. However, K ! , driven by gyro-
viscosity in equation (31), is given by

w'
K! = -
' 3

where q o and g , are given in Appendix A. Thus

926 J. W. CONNOR
et al.

so that K! gives a contribution to equation (30) through the gyroviscous term

comparable to that from the rigid rotation in the perpendicular viscosity.
Clearly, when the magnetic field is up-down symmetric ( B = B(R, I))),Kf is zero.
Thus, in an up-down symmetric Tokamak, the viscous damping of toroidal flow
comes entirely from the rigid rotator flow in the perpendicular viscosity and equation
(20) takes the form

where B, = JVI)(/R,B, = I / R and q1 = q2/2 = 3/10 nT(vii/w:i)(BRAGINSKII,1965).

The density evolution is given by

and is therefore slower than the evolution of w by a factor (vie/vii).

We estimate the damping of toroidal rotation from equation (34) as

and observe that there is no Pfirsch-Schluter enhancement. One may verify that in
an up-down asymmetric Tokamak the term resulting from K! in gyroviscosity gives
a momentum diffusivity of order q2.viip’/a2 times the square of the degree of
asymmetry.
111. D I S C U S S I O N
In this section we will discuss the physical processes that govern the form of the
flow and the slowing down and compare our results with others in the literature.
The lowest order consequences of the momentum equation in a gyroradius
+
expansion is the familiar E Vi x Bjc = 0, equation (10). The magnetic field is
therefore frozen into the lowest order flow. We may deduce the properties of the
flow from the constraints that the magnetic field and density be stationary. Thus
there is no flow perpendicular to the flux surface since it would distort the surface. It
is convenient to split the flow in the surface into two components: a toroidal flow
which, because of axisymmetry, does not perturb the density and a flow parallel to
the magnetic field which therefore does not perturb the field. The angular velocity of
the toroidal flow, w, must be a flux function otherwise differential rotation within the
surface would “wind up” the field. T o discuss the parallel flow V l l iwe consider a flux
tube through which the fluid flows. The continuity of the particle flux within the
tube requires that V l i n i 6 Abe constant (where 6 A is the cross-section of the tube).
Since V.B = 0, JAB is also a constant along the tube so that = Ki($)B/ni.The
 Toroidal rotation and momentum transport 921

form of the flow given by equation (12) is therefore deducible from relatively simple
physical considerations.
The value of K, is determined by parallel viscosity. The damping of the parallel
flow by parallel viscosity can be understood in terms of the work done against P ,
and P , , the components of pressure perpendicular and parallel to the magnetic field.
A fluid element flowing along the field line from the outside to the inside of the torus
is elongated along the field and compressed across the field. The rate of doing work
against parallel viscosity in a large aspect ratio Tokamak is of order (PL- P I I ) ~ V l l i , ’ a
(where E is the inverse aspect ratio). The difference P- - P , larises from competition
between collisionless particle motion which increases the anisotropy and collisions
-
which decrease it and is given by P , - P I , &PViliJuvii. Balancing the loss of kinetic
energy of the parallel flow with the work done against parallel viscosity leads to a
damping rate for poloidal flow of order
It is interesting to estimate the bulk parallel ion flow resulting from neutral beams
by balancing this parallel ion viscosity with the friction between the beams and the
ions. Since the frictional force is of order minbvii(V ~ h i / V b )V,f,
~ wherefis the fraction
of beam momentum going into parallel motions, Lf,li/Vt,,i can be estimated as

where ljbl’lji = n,Tb,’niTiwith n, and T, the beam density and “temperature”. In

typical Tokamak experiments Vi’i/Vthi estimated from equation (37) is less than a
per cent.
Our result equation (32) differs from those of STACEYand SIGMAR (1985) and
HOGAN(1 984). Hogan has argued that momentum diffusion is analogous to thermal
diffusion and that a Pfirsch-Schliiter enhancement of momentum diffusivity occurs.
To understand this we review the calculation of thermal diffusion. In a non-rotating
plasma (see Appendix B)

? T, K
n,,
ot
2
B
KA
B.V-~B.VT, - v.-B
B
x VT, + V-K-VT,

-
where the parallel thermal conductivity K i , niTi/miviiand K,/K
vii/wCi.Expanding equation (38) in powers of vii/wCiyields to lowest order
KJK,,- -v

and therefore TP = To($). T! is determined in the next order by the equation

This is readily solved for T! which is u p d o w n asymmetric. The evolution equation

for To($) is given by the flux surface average of equation (38) in the second order,
928 J. W. COKVORet al.

x VT: + V.K,VTO

where the contribution from T: yields the Pfirsch-Schluter enhancement.

While there is some similarity between the thermal conduction and viscosity
problems (parallel viscosity is similar to parallel conduction etc.) there is a crucial
difference between the role of viscosity in the momentum equation and the role of
conduction in the temperature equation. Parallel thermal conduction dominates
equation (38) whereas parallel viscosity does not dominate the momentum equation
(2). In the small gyro-radius ordering employed in this paper the momentum
equation is in fact dominated by the Lorentz force (cV.n,/ VBne piimfp/a2).
5 In order
to determine the constraints imposed by parallel viscosity it is then necessary to
annihilate the Lorentz force in the momentum equation by taking the scalar product
with B/n, and flux surface averaging. Thus K' is determined from equation (31) by
computing the work done over the whole surface, whereas T,! is computed locally
on the flux surface. Hogan's calculation neglects the Lorentz force and obtains a local
equation for the toroidal velocity V' driven by 7c3,,(wRe,),analogous to equation
(39) for T!. The contribution to the toroidal momentum loss from 7c3,,(V') then
yields a Pfirsch-Schliiter factor.
In a series of papers STACEY and SIGMAR(1984, 1985) have advanced the hypo-
thesis that classical gyroviscosity can account for the observed anomalous toroidal
viscosity. In order to achieve their effect they postulate O(E),where E is the inverse
aspect ratio of the Tokamak, variations of density n and V,,'R within a magnetic
surface, corresponding to non-rigid rotation. In particular these variations must be
up-down asymmetric. Gyroviscosity is then able to produce a radial transport of
toroidal momentum with a viscosity coefficient p, - E'D, where D, is the Bohm
diffusion coefficient. It is interesting to note that a similar result for thermal con-
ductivity would arise if one considered the effect of the finite gyro-radius heat flux in
the presence of an O ( E temperature
) variation in a surface! To justify these variations
in V,/R and n Stacey and Sigmar use results obtained from momentum equations in
which the Braginskii viscous stress tensor is simply replaced by a drag-njmjvDjVj
(STACEY and SIGMAR, 1984). The arbitrary drag coefficients vDj are chosen to simulate
the anomalously large toroidal viscosity observed in experiments, and thus greatly
exceeds the equivalent value arising from Braginskii's equations. However, it grossly
underestimates the parallel triscous drag of the classical stress tensor. This is the
essential reason for the result obtained by Stacey and Sigmar. It allows much greater
parallel flows, i.e. KO($) # 0, than are permitted by the Braginskii anistropic stress
tensor. In particular non-rigid rotator flows occur, resulting in u p d o w n asymmetric
density perturbations. The gyro-viscous stress then provides the anomalously large
toroidal viscosity. The authors can then envisage a bootstrapped situation in which
this anomalous viscosity is responsible for the initially assumed anomalous drag. As
we have seen the correct form of Braginskii's stress tensor does not permit these
effects-one would have to remove the dominant parallel viscosity effects by invoking
a greatly enhanced ion collision frequency. These essential points are not significantly
affected by complications due to impurities, temperature gradients etc.
 Toroidal rotation and momentum transport 929

IV. C O N C L U S I O N
We have examined the diffusion of momentum in a collisional toroidal plasma
with a sonic toroidal flow. We have demonstrated that there is no Pfirsch--Schliiter
enhancement of momentum diffusivity in an up-down symmetric Tokamak, i.e. the
momentum confinement time, as given in equation (34), is classical, (T;'
With an u p d o w n asymmetry of the magnetic geometry, we obtain a Pfirsch--
viip?/a2). -
Schluter-like confinement time .I.;' - +
(viip'/a2)(1 iJ2q2), where 1. is a constant of
order unity, and 6 is a measure of the asymmetry. Since experimental slowing down
rates are more than fifty times these rates (ISLERet al., 1986; BRAUet al., 1983), the
momentum loss, like the energy loss, is anomalously fast. An explanation in terms of
turbulent transport, such as that due to temperature gradient turbulence (MATTOR
et al., 1986), could afford an explanation of this anomaly.

REFERENCES
BRAGINSKII S. I. (1965) Reviens of Plasma Physics, Vol. 1 . Consultants Bureau, New York, p. 205.
BRAUK. et al. (1983) Nucl. Fusion 23, 1643.
COWLEY S. C. and BISHOPC. M. (1986) Culham Lab. report.
HASSAM A. B. and KULSRGD R. M. (1978) Pllysics Fluids 21. 2271.
HAZELTINE R. D. (1976) Phy.sic.7 F1uid.r 19, 747.
HIUTONF. L. and WONGS. K. (1985) Physics Fluids 28, 3082.
HOGANJ. T. (1984) Phj.ric.7 Fluids 27, 2308.
ISLER R. C. et ul. (1986) h'ucl. Fusion 26, 391.
MASCHKE E. K. and PERRIN H. (1980) Plusnzu Physics 22, 579.
MATTOR N., DIAMOND P. H. and LEEG. S. (1986) Sherwood Meeting, New York, paper 2C-11.
MlKHAlLoVsKll A. B. and TSYPINV. S. (1971) Plasnza Plzysics 13, 785.
ROSENBLUTHM. N., RUTHERFORDP. H., TAYLOR J. B., FRIEMAN L. M. (IAEA,
E. A. and KOVRIZHNIKH
Vienna 1971), Plusnza Physics und Controlled Nuclear Fusion, Vol. 1, p. 495.
STACEY W. M. JR. and SIGMAR D. J. (1984) Physics Fluid.r 27, 2076.
STACEY W. M. JK. and SIGMAK D. J. (1985) Physics Fluids 28, 2800.
TswIN V. S. (1985) Soviet J . Plasma Phq's 11, 661.

A P P E N D I X A. T H E B R A G I N S K I I V I S C O S I T Y T E N S O R
In this Appendix we give the ion viscosity tensor, (BRAGINSKII, 1965). This involves:

parallel viscosity

1 1
no; -3tlo(bb - -I)(bb - -1):VV
3 3

where 'lo = 0.96 ~ , T , / vand

, ~ b is a unit vector along B.

gyroviscosity,

and

n4&= q4[b x Webb - bb*W x b]

where q4 = n,T, w C ,= 2q3, the rate of strain tensor

2
w = vv + (VV)' - -V.VI
3

and I , = ( I - bb)
930 J. W. Cowon et al.

perpendicular viscosity,

n Z i= qz[Zl*W*bb + bb*W*Z,]
where q z = = 4ql.
12 10 (niTiviL’w~i) Thus

q O : r l 3 ,q 4 : q l , q2 = I : V ~ ~ ~ W , ~ : ( L ’ ~ ~ / O ~ ,~)*
(A.7)
The full viscosity tensor is
4
xi = 1 xni.
n=0

A P P E N D I X B. T E M P E R A T U R E E V O L U T I O N
In this Appendix we discuss the modifications imposed by the presence of temperature gradients. The
ion temperature equation is (BKAGIMKII, 1965),

where Q, is the heat source (we ignore heat exchange with electrons) and q, is the ion heat flux,

qt = -K +
i,bb*VT, tiA,b x V T , - hllVT,. (B.2)
The thermal conductivities are,

In the lowest order equation ( B . l ) becomes,

where we have used equation (10). The flux surface average of this equation is equation (14) to the lowest
order. Dividing equation (B.4)by TP we obtain the condition,

Since both averages in equation (B.5) are positive definite we conclude that

KP = 0, B*VTP = 0. (B.6)
The order of the slowing down time can now be determined from equation (20). Using equations (28) and
(29) and Vp = tu($)Re,., we find that the temperature gradient does not effect the order of magnitude of
the slowing down. If we wished to evaluate the precise effect of temperature gradients on the slowing
down we would integrate equation (B.l) to determine B.VT: and then obtain K! from equation (14). To
calculate the correct coupling of the momentum loss to the temperature gradient we must include the
contributions to the viscous stress from heat flow (MIKHAILOVSKII and TSYPIN.1971; TSYPIN,1985)
which were not included by Braginskii. Since the order of magnitude of the result is unchanged by the
temperature gradient and the correct result is given in HINTONand WONG(1985), we will not pursue the
matter further here.
 Toroidal rotation and momentum transport 93 1

A P P E N D I X C. T H E I N F L U E N C E O F I M P U R I T I E S
In this Appendix we discuss the modifications to the theory of momentum transport introduced by the
addition of an impurity. We shall assume that the impurity density n1 is such that,

In this ordering impurity-ion collisions are comparable with ion-ion collisions. Equation (10) holds for
both species so that the zeroth order flow for both is given by equation (11). In calculating the parallel
flow from equation (14) we must include the friction term. Equation (14) for the ions becomes,

and for the impurities

so that KI = K, = 0. The zeroth order flow is the rigid rotator, just as in the case without impurities. The
viscous slowing down time is now determined from equation (20), and again arises from the rigid rotator
flow in perpendicular viscosity (and V’ in gyroviscosity in the asymmetric case). The order of magnitude
of the slowing down time is thus not changed by the presence of impurities [i.e. it is still given by
equation (31)] and so it is unnecessary to evaluate the details here.
Finally we note that the poloidal density variations can be obtained from the parallel component of the
zeroth order momentum equations;

ne = fi,($)exp
3
--

The poloidal variation of ‘p’ is determined from charge neutrality. If Z , 5 1, Zi= 1 and equation (C.l)
holds, then ‘p’ is given approximately by equation (25). In this limit the impurities play a role in the
collisional dynamics but not in charge neutrality.