Compact formulas for guiding-center orbits in axisymmetric tokamak

geometry
Alain J. Brizard
Department of Chemistry and Physics, Saint Michaels College Colchester, Vermont 05439, USA
Received 12 November 2010; accepted 21 January 2011; published online 17 February 2011
Compact formulas for trapped-particle and passing-particle guiding-center orbits in axisymmetric
tokamak geometry are given in terms of the Jacobi elliptic functions and complete elliptic integrals.
These formulas can nd applications in bounce-center kinetic theory as well as guiding-center
FokkerPlanck kinetic theory. 2011 American Institute of Physics. doi:10.1063/1.3554696
I. INTRODUCTION
The presence of trapped-particle and passing-particle
guiding-center orbits in axisymmetric tokamak geometry has
had a signicant impact on magnetic fusion research see
Refs. 1 and 2 for a brief historical survey. Because of their
fundamental importance in understanding neoclassical and
anomalous transport in tokamak plasmas,
3
it is desirable to
nd compact analytic representations for the trapped-particle
and passing-particle guiding-center orbits in axisymmetric
tokamak geometry. In particular, compact expressions might
allow explicit calculations of bounce-angle averages in
bounce-center kinetic theory
46
beyond the deeply trapped
approximation.
Previous analytic representations have used a simplied
tokamak geometry
7
in which the magnetic surfaces are un-
shifted concentric circular toroidal surfaces centered on the
magnetic axis located at a major radius R for the axis of
symmetry. In this simplied tokamak geometry, trapped-
particle and passing-particle guiding-center orbits
6
are repre-
sented in terms of complete and incomplete elliptic
integrals.
8
The purpose of this work is to present compact
analytic formulas for these trapped and passing guiding-
center orbits expressed in terms of the Jacobi elliptic
functions.
9,10
These compact formulas are readily amenable
to explicit bounce-angle-averaging calculations that do not
require using simplifying limits such as the deeply trapped
limit in which case the Jacobi elliptic functions become
simple trigonometric functions. In particular, we show in
Appendix how the use of the deeply trapped limit in evalu-
ating bounce-angle averages can lead to expressions that are
both quantitatively and qualitatively incorrect.
The remainder of the paper is organized as follows. In
Sec. II, we introduce the simple axisymmetric tokamak ge-
ometry used here and in previous works.
7
In Sec. III, we
introduce the bounce and transit actions J

=J
b
, J
t
associ-
ated with the trapped-particle =b and passing-particle
=t guiding-center orbits projected onto the poloidal
plane at constant toroidal angle and their associated or-
bital frequencies

/ E
1
, which are given in terms
of complete elliptic integrals.
6
In Sec. IV, we present com-
pact expressions for the poloidal angle in terms of Jacobi
elliptic functions depending on the bounce angle
b
for
trapped-particle orbits and the transit angle
t
for passing-
particle orbits. In Sec. V, we introduce the canonical parallel
coordinates s

, J

and p

, J

for trapped-particle and

passing-particle guiding-center orbits, which satisfy a ca-
nonical condition s, p

1. This canonical condition forms

the basis of the bounce-center phase-space transformation
11
in general magnetic geometry. The fact that the parallel co-
ordinates explicitly satisfy the canonical conditions is a ma-
jor result of the present work. In Sec. VI, the slow toroidal
drift guiding-center dynamics for , is described in terms
of compact expressions involving the Jacobi elliptic func-
tions, where the poloidal magnetic ux is Taylor-expanded
about the guiding-center magnetic-ux invariant

. Lastly,
our work is summarized in Sec. VII and, in Appendix, we
demonstrate the power of compact Jacobian-elliptic expres-
sions by investigating periodic integrals associated with
bounce-averaging operations.
II. SIMPLE TOKAMAK GEOMETRY
In circular, large-aspect-ratio tokamak geometry,
7
the
magnetic-eld strength is approximated as
B, B
0
1
r
R
cos = B
e
+ 2B
0
sin
2

2
, 1
where the small inverse aspect ratio
r/R 1 2
is dened in terms of the minor radius r of a circular
magnetic surface labeled by , and the poloidal angle is
measured from the outside equatorial plane. In Eq. 1, B
0
denotes the magnetic-eld strength on the magnetic axis at
r=0 and B
e
B
0
1 denotes the magnetic-eld strength
on the outside equatorial plane. Lastly, we note that the
magnitude of the toroidal magnetic is B
tor
B while the mag-
nitude of the poloidal magnetic eld at the minor radius r is
B
pol
B
0
/ q, where q is the safety factor.
7
In all
expressions presented here, terms of order
2
are systemati-
cally omitted the symbol is used whenever this approxi-
mation is rst introduced.
On a single magnetic surface at xed , the magnitude
of the parallel momentum p

2mEB of a
guiding-center of mass m at xed total energy E, magnetic
moment , and magnetic ux is
PHYSICS OF PLASMAS 18, 022508 2011
1070-664X/2011/182/022508/8/$30.00 2011 American Institute of Physics 18, 022508-1
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
p

2mE B
e
1

2B
0
E B
e

sin
2

2
p
e
1
1
sin
2

2
mR

. 3
Here, the equatorial parallel momentum on the low-eld
side at =0 of the guiding-center orbit is dened as
p
e
=

2mE B
e
= 2

mR

, 4
where the bounce-transit parameter
4
is dened as
E, ,
E B
e
2B
0
=

e
2
1
21
e
2

, 5
and
e
p
e
/ p denotes the equatorial pitch-angle co-
ordinate. In Eq. 4, we also dened the characteristic paral-
lel frequency

1
R

B
0
m
, 6
and the connection length R

qR. Using these deni-

tions and Eq. 4, the equatorial poloidal angular velocity is
dened as

e
2

. 7
Note that the parallel frequency Eq. 6 is small compared to
the gyrofrequency
0
evaluated at the magnetic axis:

0
=
1
R

B
0
m
0
2

0
R

1, 8
where
0

2B
0
/ m
0
2
denotes the gyroradius of a
guiding-center with magnetic moment evaluated at the
magnetic axis.
The condition 1 in Eq. 3 implies that a trapped
particle bounces back and forth between the bounce poloidal
angles
b
, where

b
2 arcsin

, 9
while a passing particle is in transit for 1 in Eq. 3, with
the minimum parallel momentum p
min
=p
e

1
1
p
e
at-
tained at =. Using the denition 5, the bounce-transit
boundary =1 yields the trapping condition

e
2
2/1 + 2 10
in the simple tokamak magnetic eld Eq. 1.
III. BOUNCE AND TRANSIT ACTIONS
At the lowest order on the bounce-transit time scale, the
magnetic ux is frozen see Sec. VI and thus the minor
radius r=R and the connection length R

are constant.
Hence, the parallel momentum Eq. 3 depends only on the
poloidal angle as a guiding-center moves along a
magnetic-eld line s projected onto the poloidal plane at
xed toroidal angle .
A. Bounce action and bounce frequency for trapped
particles
By using the innitesimal parallel-length element
dsR

d, we easily obtain the following expression for the

bounce action for trapped particles 1:
J
b

1
2

ds p
e
R

b
1
1
sin
2

2
d

. 11
After making the substitution sin / 2=

sin in Eq. 11,

we obtain
6
J
b
= p
e
R

0
/2
cos
2
d

1 sin
2

mR

E 1 K

, 12
where we wrote cos
2
=
1
1 sin
2
1 while
K and E denote the complete elliptic integrals of the
rst and second kind, respectively.
9,10
By using standard properties of these complete elliptic
integrals,
9
the bounce frequency for trapped particles

b
J
b
/ E
1
is expressed as

b
=

2K
, 13
where we used / E=/ EB
e
and the relation
9
d
d
E 1 K =
1
2
K. 14
In the deeply trapped approximation, where the equatorial
pitch-angle coordinate is
e
2
2 so that 1, we
use K / 2 and E 1K / 4, and Eqs.
12 and 13 become J
b
E
e
2
/
0
and
b

0
, where

0
=R

E/ m.
B. Transit action and transit frequency for passing
particles
For a passing particle 1, the transit action is
J
t
p
e
R

1
1
sin
2

2
d
2
= mR

E
1
. 15
The corresponding transit frequency
t
J
t
/ E
1
is ex-
pressed as

t
=

K
1

K
, 16
where we used the relation
9
d
d

E
1
=
K
1

1
2
K.
In the large- energetic-passing limit, we recover
J
t
p
e
R

and
t
p
e
/ mR

from Eqs. 15 and 16, which

both scale as

for 1.
022508-2 Alain J. Brizard Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
C. Unied representation for action integrals
We note that the bounce action Eq. 12 and the transit
action Eq. 15 can both be combined into a single expres-
sion for the normalized action J

J/ mR

for arbitrary
bounce-transit parameter :
J

0
K
cn
2
udu, 17
where =11 or =1/ 21. Note that the fac-
tor =1/ 2 yields a consistent relation between the poloidal
angle and the transit angle
t
for passing particles see
Eq. 25.
For 1, we use the denition
9
for the elliptic integral
E

0
K
dn
2
udu 18
and the identities
cn
2
u + sn
2
u 1 dn
2
u + sn
2
u, 19
to obtain
J

b

8

0
K
1 + dn
2
udu
=
8

E 1 K, 20
so that J

b
=1 =8/ . For 1, we use the con-
version relation
8,9
cnu dn

u
1
associated with
1
1
1. By using the change of variable v=

u,
Eq. 17 becomes
J

t

8

0
K
1

dn
2
v
1
dv =
4

E
1
, 21
so that J

t
=1 =4/ .
Figures 1 and 2 show the normalized action
J

J/ mR

, dened by Eq. 17, and the normalized fre-

quency /

versus the bounce-transit parameter . Fig-

ure 1 shows the discontinuity of the bounce and transit ac-
tions at =1. This discontinuity is related to the fact the
transit period for a marginally passing particle is associated
with the motion from = to +, while the bounce period
for a marginally trapped particle is associated with the mo-
tion from = to + and then back to = i.e., dou-
bling the path taken to complete a bounce period. The
bounce frequency Eq. 13 and the transit frequency Eq. 16
both vanish i.e., the corresponding periods become innite
at the bounce-transit boundary =1 since Kx as
x1.
IV. BOUNCE AND TRANSIT ANGLES
The angles canonically conjugate to the bounce action J
b
and the transit action J
t
are the bounce angle 0
b
2 and
the transit angle
t
, respectively. These angles have
been expressed previously in terms of the incomplete elliptic
integrals, which often prevented immediate analytical appli-
cations. In the present Section, we use the Jacobi elliptic
functions to give compact expressions of the poloidal angle
in terms of the bounce or transit angles.
A. Bounce angle for trapped particles
The bounce angle
b
for trapped particles 1 is de-
ned as

b
+

b
2

1
1
sin
2
/2
= +

2K

/2
d

1 sin
2

3
2
+

2K
sn
1
sin , 22
where sin sin/ 2 /

and we dened
b
at
=
b
i.e., at =/ 2. Equation 22 can be inverted to
give
1 2 3 4
1
2
3
4
FIG. 1. Normalized action J

J/ mR

, given by Eq. 17, versus the

bounce-transit parameter . The normalized bounce action Eq. 20 is shown
for 1 while the transit action Eq. 21 is shown for 1.
1 2 3 4
0
1
2
3
4
FIG. 2. Normalized frequency /

versus the bounce-transit param-

eter . The bounce frequency Eq. 13 is shown for 1 while the transit
frequency Eq. 16 is shown for 1.
022508-3 Compact formulas for guiding-center orbits in axisymmetric tokamak geometry Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
sin

2
=

sn
2K

b
3K

cd2K
b
/, 23
where we used the standard notation pqpn/ qn with p and
q either c, s, or d and the identities
9
snu3K snu+K
cdu, with u2K
b
/ and we used the 4K-periodicity
of sn.
Equation 23 yields the poloidal angular velocity

1 sd2K
b
/

e
1
1
sin
2

2
, 24
where

e
is dened in Eq. 7 and we used the relations Eq.
19. Figure 3 shows the normalized poloidal angle / ver-
sus the normalized bounce angle
b
/ for various values of
1. In the standard deeply trapped approximation 1
Ref. 6
b
cos
b
is shown by the solid curve.
B. Transit angle for passing particles
Next, the transit angle
t
for passing particles 1 is
dened as

t


t
2

1
1
sin
2
/2
=

K
1

0
/2
d

1
1
sin
2

K
1

sn
1
sin

1
, 25
which can be easily inverted to give
sin

2
= snK
1

1


snK

, 26
with = when
t
= for all values of 1 and we used
sn

u
1

snu and K
1

K. We note
that the denition Eq. 25 yields the correct poloidal angular
velocity

e
cnK
t
/

e
1
1
sin
2

2
, 27
which is consistent with the denition of the transit action
Eq. 15. Figure 4 shows the normalized poloidal angle /
versus normalized transit angle
t
/ for various values of

1
1. Note that in the energetic-passing limit
1
1, we
nd
6

t
shown by a solid line in Fig. 4.
V. CANONICAL PARALLEL COORDINATES
In this section, we derive suitable expressions for the
canonical parallel coordinates s, p

, which are required to

satisfy with dsR

d the canonical condition

s, p

J

s
J
p

=
E B
e

p
e

E


E
p

= 1 28
for both trapped and transit particles, where 2
K/ for trapped-particle =1 and passing-particle
=1/ 2 orbits. In Eq. 28, we also made use of the identity
R

EB
e
/ p
e

, which follows from the denitions

Eqs. 46.
A. Parallel coordinates for trapped particles
For a trapped particle, using Eq. 23, the parallel mo-
mentum Eq. 3 is dened as
p

= p
e

1
1
sin
2
/2 p
e

1 sd
b
. 29
Figure 5 shows the normalized parallel momentum
p

/ mR

versus the normalized bounce angle

b
/ for
various values of 1, where the standard deeply trapped
approximation p

b
p
e
sin
b
is shown as a solid curve.
Next, we verify that the parallel coordinates E,
b
and
p

E,
b
, dened by Eqs. 23 and 29, satisfy the canonical
condition Eq. 28. First, using Eq. 23, we obtain
1.0 0.5 0.5 1.0
1.0
0.5
0.5
1.0
FIG. 3. Normalized poloidal angle / versus normalized bounce angle

b
/ for =0.1 solid curve, 0.8 dashed curve, and 0.999 dotted curve.
In the deeply trapped limit 1, we nd
b
cos
b
and 0 at

b
=/ 2 for all values of 1.
0.2 0.4 0.6 0.8 1.0
0.2
0.4
0.6
0.8
1.0
FIG. 4. Normalized poloidal angle / versus normalized transit angle

t
/ for =5 solid curve, 1.05 dashed curve, and 1.0001 dotted curve.
In the energetic-passing limit
1
1, we nd
t
solid curve.
022508-4 Alain J. Brizard Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions

b
= 2

1 sd, 30
and

E
=

/1
E B
e

cn + dn2
cd

, 31
where the expression for cd/ is not needed in what
follows.
12
Next, using Eq. 29, we obtain
p

b
= p
e

1 cd nd, 32
and
p

E
=
p
e
2E B
e

1 2

1
sd + 2

1
sd

. 33
By combining Eqs. 3033 into Eq. 28, we obtain
s, p

b
=

1 sd
2
+ cd
2
= 1, 34
where we used the identity 1sd
2
+cd
2
1 obtained from
Eq. 19.
B. Parallel coordinates for passing particles
For a passing particle, using Eq. 26, the parallel mo-
mentum Eq. 3 is
p

= p
e
dnK
1

1
p
e
cnK

, 35
where p
min
p
e

1
1
when
t
=. The modulation in par-
allel momentum
p

= p
e
1

1
1
= 2mR

1 36
decreases with increasing values of . Figure 6 shows the
normalized parallel momentum p

/ mR

versus the nor-

malized transit angle
t
/ for various values of 1.
We now verify that the parallel coordinates E,
t
and
p

E,
t
, dened by Eqs. 26 and 35, satisfy the canonical
condition Eq. 28. First, using Eq. 26, we obtain

t
= 2

cn, 37
and

E
=

E B
e

sd + nd2
sn

. 38
Next, using Eq. 35, we obtain
p

t
= p
e
dn sn, 39
and
p

E
=
p
e
2E B
e

cn + 2
cn

. 40
By combining Eqs. 3740 into Eq. 28, we obtain
s, p

t
=

sn
2
+ cn
2
= 1, 41
which follows from Eq. 19.
C. Phase portrait
We have demonstrated in Eqs. 34 and 41 that the
parallel coordinates s, p

, expressed in terms of Jacobi el-

liptic functions, are valid canonical coordinates for the
trapped-particle and passing-particle guiding-center orbits.
Figure 7 shows the phase portrait of the guiding-center
trapped-particle and passing-particle orbits in the
,

-plane, which combines expressions for the poloidal

angle and the normalized poloidal angular velocity

e
for trapped and passing particles. Here, we easily recognize
the well-known separatrix structure of the standard pendu-
lum problem, which separates trapped-particle orbits inside
from passing-particle orbits outside. In Sec. VI, we will
1.0 0.5 0.5 1.0
1.0
0.5
0.5
1.0
FIG. 5. Normalized parallel momentum p

/ mR

versus the normalized

bounce angle
b
/ for =0.1 solid curve, 0.8 dashed curve, and 0.999
dotted curve. In the deeply trapped limit 1, p

p
e
sin
b
solid
curve.
0.5 1.0 1.5 2.0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
FIG. 6. Normalized parallel momentum p

/ mR

versus the normalized

transit angle
t
/ for =1.5 solid curve, 1.05 dashed curve, and 1.0001
dotted curve. The modulation of the parallel momentum of a passing par-
ticle decreases as increases.
022508-5 Compact formulas for guiding-center orbits in axisymmetric tokamak geometry Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
project the trapped-particle and passing-particle guiding-
center orbits onto the poloidal plane at a constant toroidal
angle .
VI. TOROIDAL GUIDING-CENTER DYNAMICS
In this Section, we discuss the guiding-center dynamics
in the toroidal direction i.e., the direction of axisymmetry
and its implications on the canonically conjugate drift action.
So far, the toroidal angle was considered xed, which
enabled a study of the trapped and passing guiding-center
orbits projected onto the poloidal plane, and the magnetic
ux was assumed to be frozen on the bounce-transit time
scale.
The axisymmetry of the tokamak magnetic geometry im-
plies that the toroidal canonical momentum or drift action
P

=
e
c
+ p

B

e
c

42
is conserved, where B

B
0
R is approximately constant. We
note that, to lowest order in orbit width, the results presented
in Secs. IIIV are based on the identity

in the zero-
orbit-width limit
13,14
. The results presented in this section,
therefore, investigate the expansion
13,14

, p

, p

, + , 43
in which p

, appearing in the rst-order deviation

, p

, are given in Sec. V for =

.
A. Trapped-particle guiding-center orbits
For a trapped-particle guiding-center orbit, this conser-
vation law is expressed as

0
in terms of the magnetic
ux
0
at the bounce points, so that Eq. 42 becomes

0
, 44
where

/ m. By substituting the Taylor approxima-

tion
0
+r
0
B

/ q
0
, the normalized radial deviation

0
from the magnetic surface
0
is expressed
as

b

q
0

R
= 2

1 sd2K
b
/, 45
where, using Eq. 8, we have dened the trapped-orbit
parameter

q
0
2

0
=
q
0

2
0

0
R
. 46
The maximum radial deviation Eq. 45 is
b

max
=2

2.
A trapped-particle guiding-center orbit can be projected
onto the normalized poloidal plane x x/ R, z z/ R and
represented as

x
b
, = 1 +
0
+
b
cos
b
,
z
b
, =
0
+
b
sin
b
,

. 47
Figure 8 shows the trapped banana guiding-center orbits
for
0
, =0.5, 0.1 for various values of 1, with the
magnetic surface =
0
shown as a dashed circle. We note
that, in the simple orbit topology considered here, a trapped-
particle orbit does not enclose the magnetic axis.
15
B. Passing-particle guiding-center orbits
For a passing-particle guiding-center orbit, the conserva-
tion law Eq. 44 may be expressed in terms of the magnetic
ux

t
at the inside equatorial midplane at =. By
using the Taylor approximation
t
+r
t
B

/ q
t
, the nor-
malized radial deviation
t

t
from the magnetic sur-
face
t
is expressed as

t
2

dnK
1

1


1

, 48
where orbit parameter Eq. 46 is now evaluated at
t
note
that
t
vanishes on the inside equatorial plane where
p

=p
min
=p
e

1
1
. The maximum radial deviation Eq.
48 is
t

max
=2

1, i.e., the largest radial devia-

3 2 1 1 2 3
1.5
1.0
0.5
0.5
1.0
1.5
FIG. 7. Phase portrait in ,

-plane of guiding-center trapped and transit

orbits. The separatrix dashed curves separates trapped-particle orbits in-
side from the passing-particle orbits outside.
0.0 0.5 1.0 1.5 2.0
0.5
0.5
FIG. 8. Trapped banana guiding-center orbits in the normalized poloidal
x , z -plane for =0.1, 0.5, 0.99, with the invariant

=
0
chosen as the
magnetic ux
0
of the bounce points shown as a dashed circle.
022508-6 Alain J. Brizard Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
tion for a passing-particle orbit occurs for marginally passing
particles 11.
A passing-particle guiding-center orbit can be projected
onto the normalized poloidal plane and represented as

x
t
, = 1 +
t
+
t
cos
t
,
z
t
, =
t
+
t
sin
t
,

. 49
Figure 9 shows the passing guiding-center orbits for
t
,
=0.5, 0.1 for various values of 1, with the magnetic
surface =
t
shown as a dashed circle.
C. Bounce-averaged toroidal drift frequency
The bounce-averaged toroidal drift frequency
7

c
e

B

50
is approximated ignoring magnetic shear as

d

2B
0
m
0
q
R
2

E
K

1
2
, 51
where we used B/ q/ R
2
cos . We note that the
bounce-averaged toroidal drift frequency Eq. 51 exhibits a
drift reversal for
r
0.826, where K
r
2E
r
see
Fig. 10; a more complete expression for the bounce-
averaged toroidal drift frequency can be found elsewhere.
1
Lastly, the toroidal drift frequency Eq. 51 can also be
expressed in terms of a radial derivative of the bounce action
as follows. First, we write
J
b

4m

E 1 K +
1
2
K

= 2
B
0

E
K

1
2
, 52
where we used Eq. 14 and / =1/ 2 / . The toroidal
drift frequency Eq. 51 can thus be approximately expressed
as

d

q
b

J
b
mR
2
. 53
Hence, the toroidal drift reversal occurs when the radial de-
rivative of the bounce action Eq. 12 changes sign.
VII. SUMMARY
The analytical formulas presented in this paper provide
compact expressions in terms of the Jacobi elliptic functions
cn, sn, and dn that describe trapped-particle 1 and
passing-particle 1 guiding-center orbits in simple axi-
symmetric tokamak geometry for arbitrary bounce-transit pa-
rameter . The parallel coordinates s, p

derived in Sec. V,
which satisfy the canonical condition Eq. 34 for trapped
particles and Eq. 41 for passing particles, guarantee that the
bounce-center transformation
5
in axisymmetric tokamak ge-
ometry can be explicitly carried out beyond the deeply
trapped and energetic-passing limits.
6
Applications of the compact Jacobi elliptic representa-
tions of trapped-particle and passing-particle guiding-center
orbits include the derivation of bounce-center-averaged uc-
tuating potentials in bounce-kinetic theory
5,6
and the deriva-
tion of a bounce-center Fokker-Planck collision operator
13,14
beyond the zero-banana-width approximation.
ACKNOWLEDGMENTS
The author is grateful to Joan Decker for his insightful
comments concerning applications of the present work.
APPENDIX: PERIODIC INTEGRALS OF ELLIPTIC
FUNCTIONS
In this Appendix, we investigate how the averaging op-
eration

1
4Km

0
4Km
d A1
depends on the modulus m. For this purpose, we consider the
function Fx+ cn, y+ sn evaluated parametrically on a
0.0 0.5 1.0 1.5 2.0
0.5
0.5
FIG. 9. Passing guiding-center orbits in the normalized poloidal x, z-plane,
with the invariant

=
t
chosen as the magnetic ux
t
on the inside equa-
torial plane shown as a dashed circle. The dotted and solid circles are
passing guiding-center orbits with =1.05 and 1.5, respectively.
0.2 0.4 0.6 0.8 1.0
0.4
0.2
0.2
0.4
FIG. 10. Normalized drift frequency E / K 1/ 2 as a function
of . Note that the toroidal drift frequency Eq. 51 changes sign for

r
0.826.
022508-7 Compact formulas for guiding-center orbits in axisymmetric tokamak geometry Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions
circle of radius centered at x, y. First, we dene the
function
Fx, y;, m Fx + cn, y + sn, A2
which smoothly depends on the coordinates x, y as well as
the parameters , m. Next, we Taylor-expand Eq. A2 in
powers of to obtain up to second order
Fx, y;, m = Fx, y +

2
2

2
F
x
2
cn
2
m
+

2
2

2
F
y
2
sn
2
m + , A3
where we used the identities cn m 0sn m. The
coefcients of the second-order terms involve the averages
sn
2
m
1
4K

0
4K
sn
2
md =
1
m
1
Em
Km
,
A4
cn
2
m 1 sn
2
m =
1
m

Em
Km
1 m,
A5
where Km and Em are the complete elliptic integrals of
the rst and second kinds, with
Em
Km
= 1
m
2

m
2
16
+ A6
when m1. In the limit m0, where we nd sn
2
0
=1/ 2=cn
2
0, we obtain the Taylor series
Fx, y;, 0 = Fx, y +

2
4

2
F
x
2
+

2
F
y
2
+ , A7
which is reminiscent to the standard nite-Larmor-radius ex-
pansion found in gyrokinetic theory. We note, here, that the
spatial derivatives in x and y are symmetric in Eq. A7.
When m0, however, Eqs. A3A5 yield the
expansion
Fx, y;, m = Fx, y +

2
4

2
F
x
2
1
m
4
+
+

2
F
y
2
1 +
m
4
+
+ , A8
which now introduces an asymmetry between the spatial de-
rivatives in x and y. In the limit m1, where Eqs. A4 and
A5 yield sn
2
m 1 and cn
2
m 0, we nd
Fx, y;, 1 = Fx, y +

2
2

2
F
y
2
+ , A9
where
2
Fx, y / x
2
has now completely disappeared
from the Taylor expansion Eq. A3. This result is easily
understood as follows. If we dene the physical angle
, m =tan
1
sn m / cn m, so that Fx+ cn, y
+ sn Fx+ cos , y+ sin , we nd that, during a
period 04Km, the angle stays near / 2 for a
longer portion of that period as m1, so that sin
2
1
and cos
2
0. Hence, the average Eq. A2 would have
been dramatically wrong both quantitatively and qualita-
tively if we had insisted in using the deeply trapped result
Eq. A7.
We conclude this Appendix by noting the simplicity of
the averaging process A1 associated with the Jacobi elliptic
functions, which yields the explicit compact results Eqs.
A3A5. Based on these explicit results, we now have a
clear understanding of the implications of performing
bounce-angle averages beyond the deeply trapped approxi-
mation. For example, using the parallel momentum Eq. 29
for trapped particles, we nd p

=0 and
p

2
=
p
e
2
1
4K

0
4K
sd
2
d
=
p
e
2

E
K
1 , A10
which implies that p

2
p
e
2
/ 2 in the deeply trapped limit
0 but p

2
0 as we approach the bounce-transit
boundary 1 since the particle now spends an increas-
ingly longer period of time near the turning points where p

vanishes.
1
A. A. Galeev and R. Z. Sagdeev, in Reviews of Plasma Physics, edited by
M. A. Leontovich Consultants Bureau, New York, 1979, Vol. 7,
pp. 257343.
2
F. L. Hinton and R. D. Hazeltine, Rev. Mod. Phys. 48, 239 1976.
3
E. J. Doyle, W. A. Houlberg, Y. Kamada, V. Mukhovatov, T. H. Osborne,
A. Polevoi, G. Bateman, J. W. Connor, J. G. Cordey, T. Fujita, X. Garbet,
T. S. Hahm, L. D. Horton, A. E. Hubbard, F. Imbeaux, F. Jenko, J. E.
Kinsey, Y. Kishimoto, J. Li, T. C. Luce, Y. Martin, M. Ossipenko, V.
Parail, A. Peeters, T. L. Rhodes, J. E. Rice, C. M. Roach, V. Rozhansky, F.
Ryter, G. Saibene, R. Sartori, A. C. C. Sips, J. A. Snipes, M. Sugihara, E.
J. Synakowski, H. Takenaga, T. Takizuka, K. Thomsen, M. R. Wade, H. R.
Wilson, ITPA Transport Physics Topical Group, ITPA Connement Data-
base and Modeling Topical Group, and ITPA Pedestal and Edge Topical
Group, Nucl. Fusion 47, S18 2007.
4
F. Y. Gang and P. H. Diamond, Phys. Fluids B 2, 2976 1990.
5
A. J. Brizard, Phys. Plasmas 7, 3238 2000.
6
L. Wang and T. S. Hahm, Phys. Plasmas 16, 062309 2009.
7
R. B. White, The Theory of Toroidally Conned Plasmas World Scien-
tic, Imperial College Press, London, 2001.
8
L. M. Milne-Thomson, in Handbook of Mathematical Functions, edited by
M. Abramowitz and I. A. Stegun Dover, New York, 1965, Chap. 17.
9
L. M. Milne-Thomson, in Handbook of Mathematical Functions, edited by
M. Abramowitz and I. A. Stegun Dover, New York, 1965, Chap. 16.
10
A. J. Brizard, Eur. J. Phys. 30, 729 2009.
11
F.-X. Duthoit, A. J. Brizard, Y. Peysson, and J. Decker, Phys. Plasmas 17,
102903 2010.
12
The partial derivative of Jacobi elliptic functions with respect to the
modulus is expressed in terms of the indenite integral cn
2
d
1
1
+
1
E , where E denotes the incomplete elliptic
integral of the second kind Ref. 8.
13
A. J. Brizard, J. Decker, Y. Peysson, and F.-X. Duthoit, Phys. Plasmas 16,
102304 2009.
14
J. Decker, Y. Peysson, A. J. Brizard, and F.-X. Duthoit, Phys. Plasmas 17,
112513 2010.
15
L. G. Eriksson and F. Porcelli, Plasma Phys. Controlled Fusion 43, R145
2001.
022508-8 Alain J. Brizard Phys. Plasmas 18, 022508 2011
Downloaded 17 Feb 2011 to 192.80.64.240. Redistribution subject to AIP license or copyright; see http://pop.aip.org/about/rights_and_permissions