Solutions to Chen’s Plasma Physics

Kalman Knizhnik

1-1. Compute the density (in units of m−3 ) of an ideal gas under the following condi-
tions:
a) At 0o C and 760 Torr pressure (1 Torr = 1mm Hg). This is called the Loschmidt
number.
b) In a vacuum at 10−3 Torr at room temperature (20o C). This number is a useful one
for the experimentalist to know by heart (10−3 Torr = 1 micron).
a) Avogadro’s number is NA = 6.022 × 1023 . One mole of gas at STP occupies 22.4 liters. 1 liter is
1 × 10−3 cubic meters. Thus, the number per cubic meter is NA /n = 6.022 × 1023 /(22.4 × 10−3 ) =
2.66 × 1025 m−3 . Thus, the Loschmidt number is 2.66 × 1025 2
b) Using PV=NkT, we obtain (with R= 1.4 × 10−23 J K −1 and 1 Torr = 133 Pa):
N P 10−3 × 133
n= = = = 3.3 × 1019 m−3 2 (1)
V kT 1.4 × 10−23 × (20 + 273)
1-2. Derive the constant A for a normalized one-dimensional Maxwellian distribution
2
fˆ(u) = Ae−mu /2kT (2)

such that Z ∞
fˆ(u)du = 1 (3)
−∞
This one is straightforward. Just integrate:
s
Z ∞ r
−mu2 /2kT 2πkT m
1= Ae du = A ⇒ A= 2 (4)
−∞ m 2πkT

1-4. Compute the pressure, in atmospheres and in tons/f t2 , exerted by a thermonuclear

plasma on its container. Assume kTe = kTi = 20keV , n = 1021 m−3 and p = nkT , where
T = Ti + Te .
This is just unit conversion, albeit with units that nobody really ever remembers. For reference,
1 keV = 1.6 × 10−19 J, so

p = 1021 × (20keV + 20keV ) = 4 × 1022 m−3 keV = 4 × 103 m−3 J = 4 × 103 N/m2 (5)

But 1 atm = 105 N/m2 = 1 ton/f t2 , so (Note: I think there is a mistake in Chen’s solutions here.
If I am mistaken, please let me know).

p = 0.04 atm = 0.04 ton/f t2 2 (6)

1-5. In a strictly steady state situation, both the ions and the electrons will follow the
Boltzmann relation
nj = n0 e−qj φ/kTj (7)
For the case of an infinite, transparent grid charged to a potential φ, show that the
shielding distance is given approximately by
ne2 1 1
λ−2
D = ( + ) (8)
0 kTe kTi
Show that λD is determined by the temperature of the colder species.
We’ll use Poisson’s equation
ene − eni en0 eφ/kTe en0 eφ −eφ e2 n φ φ
∇2 φ = = (e − e−eφ/kTi ) ≈ (1 + −1− )= ( + ) (9)
0 0 0 kTe kTi 0 kTe kTi
Now we’ll suppose the φ goes like a decreasing exponential: φ = φ0 exp(−x/λD ). Thus, the Lapla-
cian acting on this is

1 ne2 φ φ 1 ne2 1 1
∇2 φ = 2 φ = ( + ) ⇒ 2 = ( + ) 2 (10)
λD 0 kTe kTi λD 0 kTe kTi

To show that λD is determined by the colder species, we suppose first that the electrons are the
colder species: Te  Ti . Then,
s
1 ne2 kTe 0
= ⇒ λD = 2 (11)
λ2D kTe 0 ne2

Alternatively, if the ions are colder, Ti  Te , then a similar analysis yields:

s
1 ne2 kTi 0
= ⇒ λD = 2 (12)
λ2D kTi 0 ne2

1-6. An alternative derivation of λD will give further insight to its meaning. Consider
two infinite, parallel plates at x = ±d, set at potential φ = 0. The space between them
is uniformly filled by a gas of density n of particles of charge q.
a) Using Poisson’s equation, show that the potential distribution between the plates
is
nq 2
φ= (d − x2 ) (13)
20
b) Show that for d > λD , the energy needed to transport a particle from a plate to the
mid plane is greater than the average kinetic energy of the particles.

1-9. A distant galaxy contains a cloud of protons and antiprotons, each with density
n = 106 m−3 and temperature T = 100o K. What is the Debye length?
The Debye length is given by
X 0 kTj
λD = (14)
j
nj e2j

Plugging in the numbers:

8.85 × 10−12 × 1.4 × 10−23 × 100

λD = = 0.48 m 2 (15)
106 × (1.6 × 10−19 )2

As a check, use the SI unit form for the Debye length given in Chen. If T is in Kelvin, and n is in
cubic meters, then: s s
T 102
λD = 69 m = 69 × m = 69 × 10−2 m 2 (16)
n 106
This is the same order of magnitude so we are ok.

1-10. A spherical conductor of radius a is immersed in a plasma and charged to a

potential φ0 . The electrons remain Maxwellian and move to form a Debye shield, but
the ions are stationary during the time frame of the experiment. Assuming φ0  kTe /e,
derive an expression for the potential as a function of r in terms of a, φ0 , and λD . (Hint:
Assume a solution of the form e−br /r.)
Let’s assume a solution of this form: φ = Ae−br /r. Then,

1 ∂ 2 ∂φ e
∇2 φ = 2
(r ) = b2 φ = (ne − ni ) (17)
r ∂r ∂r 0
Since the electrons are Maxwellian, they obey ne = n0 eeφ/kTe ≈ n0 (1 + eφ/kTe ). The ions, however,
are stationary, so ni = n0 . Thus we have:

e eφ n0 e 2 φ φ 1
b2 φ = (n0 + n0 − n0 ) = ≡ 2 ⇒ b= (18)
0 kTe 0 kTe λD λD

Thus, so far we have:

e−r/λD
φ=A (19)
r
But we also need to match the boundary condition that φ(a) = φ0 . So,

e−a/λD
φ0 = A ⇒ A = aφ0 ea/λD (20)
a
So, finally we have our answer:
e−r/λD
2
φ(r) = φ0 ea/λD a (21)
r
You know what they say: if it satisfies Poisson’s equation and the boundary conditions then it
must be the answer.

2-3. An ion engine (see Fig. 106) has a 1-T magnetic field, and a hydrogen plasma
is to be shot out at an E × B velocity of 1000 km/s. How much internal electric field
must be present in the plasma?
The E × B velocity is given by
E×B
v= (22)
B2
Plugging in the numbers:

|E|
106 m/s = ⇒ |E| = 1000 V /m 2 (23)
1T

2-4. Show that vE is the same for two ions of equal mass and charge but different
energies, by using the following physical picture (see Fig. 2-2). Approximate the right
half of the orbit by a semicircle corresponding to the ion energy after acceleration
by the E field, and the left half by a semicircle corresponding to the energy after
deceleration. You may assume that E is weak, so that the fractional change in v⊥ is
small.
If the energy of the right part of the orbit is E1 and the energy of the left part of the orbit is E2 ,
then we have
E1 = E0 + eEr1 , E2 = E0 − eEr2 (24)
p
where E0 is the initial energy and E is the electric field. The velocity is determined by v = 2E/m,
so s s
2E0 + 2eEr1 2E0 − eEr2
v1 = v2 = (25)
m m
The Larmor radius is determined via r = mv⊥ /qB, so
s s √ s
m 2E0 eEr1,2 2mE0 eEr1,2 2E0 1 Er1,2
r1,2 = 1± = (1 ± )= ± √ (26)
qB m E0 qB 2E0 m ωc 2 mE0 ωc

Thus, s s s
m E 1 2E0 eE 2E0
r1,2 (1 ∓ )= (1 ± ) (27)
2E0 qB ωc m 2E0 ωc m
The guiding center moves a distance r1 − r2 :
s s
eE 2E0 1 2E0 2eE
r1 − r2 = = (28)
E0 ωc m ωc m mωc2
The velocity of the guiding center is
r1 − r2 ωc 4eE 2eE 2E E
vgc = 2 = 2 (r1 − r2 ) = = = ≈ 2 (29)
T 2π 2πmωc πmωc πB B
since ωc = eB/m. This is a pretty good approximation.

2-5. Suppose electrons obey the Boltzmann relation of Problem 1-5 in a cylindri-
cally symmetric plasma column in which n(r) varies with a scale length λ; that is
∂n/∂r = −n/λ.
a) Using E = −∇φ, find the radial electric field for a given λ.
b) For electrons, show that the finite Larmor radius effects are large if vE is as large
as vth . Specifically, show that rL = 2λ if vE = vth .
c) Is (b) also true for ions?
Hint: Do not use Poisson’s equation.
a) We simply solve for φ from the Boltzmann relation for electrons.
kTe n
n = n0 eeφ/kTe ⇒ φ= ln( ) (30)
e n0
Therefore,
∂φ kTe n0 1 ∂n kTe
E = −∇φ = − r̂ = − r̂ = r̂ 2 (31)
∂r e n n0 ∂r eλ
b) We start with the definitions of vE , vth , and rL :
s
E 2kTe mv⊥
vE = , vth = , rL = (32)
B m eB
So, calculating the magnitude of vE :

kTe 2
mvth 1 rL vth
vE = = = (33)
eλB 2 eλB 2λ
where in the last step I have assumed that the perpendicular velocity is the thermal velocity. Now,
setting vE = vth , it is easy to see that
rL = 2λ 2 (34)
c) Sure, why not?

2-6. Suppose that a so-called Q-machine has a uniform field of 0.2 T and a cylindrical
plasma with kTe = kTi = 0.2 eV . The density profile is found experimentally to be of
the form
n = n0 exp[exp(−r2 /a2 ) − 1] (35)
Assume the density obeys the electron Boltzmann relation n = n0 exp(eφ/kTe ).
a) Calculate the maximum vE if a = 1 cm.
b) Compare this with vE due to the earth’s gravitational field.
c) To what value can B be lowered before the ions of potassium (A = 39, Z = 1) have
a Larmor radius equal to a?
We solve for φ:
kTe −r2 /a2
n0 exp[exp(−r2 /a2 ) − 1] = n0 exp(eφ/kTe ) ⇒ φ= (e − 1) (36)
e
Thus, the electric field is
∂φ kTe 2r −r2 /a2
E=− r̂ = e r̂ (37)
∂r e a2
and so vE (and it’s maximum) is
E 2rkTe −r2 /a2
vE = = e (38)
B ea2 B
s
∂vE 2kTe 2 2 4r2 kTe −r2 /a2 a2
= 2 e−r /a − e =0 ⇒ r= 2 (39)
∂r ea B ea4 B 2
So, with a = 1 cm,
s
2kTe a2 −1/2
vE,max = 2 e ≈ 8.5 km/sec 2 (40)
ea B 2 a=1 cm,B=0.2 T,kTe =0.2 keV

b) If we assume these are potassium ions, we have mg = 39 × 1.6 × 10−27 × 9.8 = 6.4 × 10−25 N .
Meanwhile, if we plug in the numbers above into the expression for the electric field (equation 37),
we’ll get that E = 17 V /m. Thus, the force due to the electric field is eE = 1.6 × 10−19 × 17 =
2.8 × 10−18 N . Thus the gravitational drift is
Fg 6.4 × 10−25
= ≈ 1.5 × 10−7 (41)
FE 2.8 × 10−18
times smaller. 2
c) The Larmor radius is rL = mvth /qB, so, in terms of the constants of vth , we have (setting
rL = a): s s
m 2kTe 2mkTe
rL = =a ⇒ B= (42)
qB m q 2 a2
Plugging in the numbers:
s
2 × 39 × 1.6 × 10−27 × 0.2 × 1.6 × 10−19
B= = 4 × 10−2 T 2 (43)
(17 × 1.6 × 10−19 )2 × (0.1 × 10−2 )2

2-8. Suppose the Earth’s magnetic field is 3 × 10−5 T at the equator and falls off as
1/r3 , as for a perfect dipole. Let there be an isotropic population of 1 − eV protons and
30 − keV electrons, each with density n = 107 m−3 , at r = 5 earth radii in the equatorial
plane.
a) Compute the ion and electron ∇B drift velocities.
b) Does an electron drift eastward or westward?
c) How long does it take an electron to encircle the earth?
d) Compute the ring current density in A/m2 .
Note: The curvature drift is not negligible and will affect the numerical answer, but
neglect it anyway.
a) The grad-B drift is given by
1 B × ∇B 1 ∇B
v∇B = v⊥ rL | 2
| = v⊥ rL | | (44)
2 B 2 B
We can calculate v⊥ from the energy, and rL from the magnetic field:
s s
2E mv⊥ m 2E
v⊥ = , rL = = (45)
m eB eB m
Thus, s s
1 2E m 2E ∇B E ∇B
v∇B = | |= | | (46)
2 m eB m B eB B
Since B ∼ r−3 , we obtain
∂B 3 B ∇B 3
∇B = r̂ = − 4 = −3 ⇒ | |= (47)
∂r r r B r
So
3E 3Er3 3Er2 3E(eV )r2
v∇B = = = = (48)
eBr eB0 Re3 r eB0 Re3 B0 Re3
where I have used the fact that B = B0 Re3 /r3 . Now we can plug in the numbers:
3 × 30 × 103 × (5 × 6 × 106 )2
v∇B,e = = 1.3 × 104 m/s 2 (49)
3 × 10−5 × (6 × 106 )3
3 × 1 × (5 × 6 × 106 )2
v∇B,i = = 0.42 m/s 2 (50)
3 × 10−5 × (6 × 106 )3
b) The magnetic field is azimuthal, from north to south, i.e. the −θ̂ direction. The gradient of the
magnetic field is clearly in the radial direction, so we have B × ∇B ∼ r̂ × θ̂ = φ̂, which is eastward.
This is for the electrons. The ions, which come without the minus sign to cancel the minus sign in
equation 47, will go in -φ̂, which is westward. 2
c) Well, it has to travel a distance L = 2π(5Re ), with the velocity v∇B,e . So,
L 2π × 5 × 6.4 × 106
T = = ≈ 4.5 hours 2 (51)
v∇B,e 1.3 × 104
d) The current density is given by J = nev, so, using the grad-B velocity in this expression, we get
J = nev∇B,e = 107 × 1.6 × 10−19 × 1.3 × 104 = 2 × 10−8 A/m2 2 (52)
2-10. A 20 − keV deuteron in a large mirror fusion device has a pitch angle θ = 45o at
the mid plane, where B = 0.7 T . Compute it’s Larmor radius.
The Larmor radius is given by
mv⊥
rL = (53)
qB
In natural units, a deuteron has m = 2 and q = +1. Furthermore, v⊥ = vsin(θ). To find v, we
convert the energy to velocity:
s s
m 2E 2 × 1.6 × 10−27 2 × 20 × 103 × 1.6 × 10−19
rL = sin(θ) = sin(45o ) = 0.3 m 2 (54)
eB m 1.6 × 10−19 × 0.7 2 × 1.6 × 10−27
2-12. A cosmic ray proton is trapped between two moving magnetic mirrors with
Rm = 5 and initially has W = 1 keV and v⊥ = v|| at the mid plane. Each mirror moves
toward the mid plane with a velocity vm = 10 km/sec (Fig. 2-10).
a) Using the loss cone formula and the invariance of µ, find the energy to which the
proton will be accelerated before it escapes.
b) How long will take to reach that energy?
1. Treat the mirrors as flat pistons and show that the velocity gained at each bounce
is 2vm .
2. Compute the number of bounces necessary.
3. Compute the time T it takes to traverse L = 1010 km that many times. Factor of
two accuracy will suffice.

a) The loss cone formula is sin2 (θm ) = 1/Rm , where θm is the angle of the magnetic mirror and
Rm is the mirror ratio Bmax /B0 . We can also write down a formula for sin(θm ) in terms of the
parallel and perpendicular velocities:
1 v⊥,f 1
sin(θm ) = √ =q =q v||,f 2 (55)
Rm 2
v⊥,f 2
+ v||,f 1 + ( v⊥,f )
Squaring both sides, and noting that, since µ is invariant, v⊥,f = v⊥,i , we have

1 1 1
= v||,f 2 = ⇒ v||,f = 2v⊥,i (56)
Rm 1 + ( v⊥,i ) 5

Thus, we can get the final energy:

1 2 2 1 2 2 5 2
Wf = m(v⊥,f + v||,f ) = m(v⊥,i + 4v⊥,i ) = mv⊥,i (57)
2 2 2
But we can’t evaluate this without knowing what the original v⊥,i is. Fortunately, we know that
initially v⊥ = v|| , so

1 2 2 1 2 2
Wi = m(v⊥,i + v||,i ) = m(2v⊥,i ) = mv⊥,i = 1 keV (58)
2 2
So, finally,
5
Wf = Wi = 2.5 keV 2 (59)
2
b) In the frame of the piston, when the particle bounces off, it’s velocity doesn’t change. In the
piston’s frame, the velocity of the particle as it is coming in is vi − vm , where vm is the velocity of
the piston (it is negative). It’s final velocity is the same but negative. Thus,

vi0 = vi − vm , vf0 = vm − vi (60)

where the prime denotes the velocity in the piston’s reference frame. But, in the lab frame,
vf = vf0 + vm , so we have
vf = 2vm − vi (61)
Thus, the change in velocity on each bounce is 2vm = 20 km/sec. The initial proton velocity is
s s
2Wi 2 × 1 × 103 × 1.6 × 10−19
vi = = = 447 km/s (62)
m 1.6 × 10−27

The proton final energy is 2.5 keV . This corresponds to a velocity of

s s
2Wf 2 × 2.5 × 103 × 1.6 × 10−19
vf = = = 707 km/s (63)
m 1.6 × 10−27

Thus, the total change in velocity needed is ∆vtot = 707 − 447 km/s = 260 km/s. This corresponds
to
∆vtot 260
Nbounces = = = 13 bounces 2 (64)
∆v1bounce 20
We can neglect the distance the mirrors move in the time the particle travels the distance, since
vm  vproton . Thus, the time it takes to travel a distance Nbounces L is, using v̄proton = (vf − vi )/2,

Nbounces L 2 × 13 × 1010
T = = = 109 s ≈ 32 years 2 (65)
v̄proton 707 − 447

2-13. Derive
R
the result of Problem 2-12(b) directly by using the invariance of J.
a) Let v|| ds ≈ v|| L and differentiate with respect to time.
b) From this, get an expression for T in terms of dL/dt. Set dL/dt = −2vm to obtain
the answer.
a) The quantity J = ab v|| ds is invariant. Thus, is approximate it as v|| L, then it’s time derivative
R

must be 0:
d
(v L) = Lv̇|| + v|| L̇ = 0 (66)
dt ||
b) We can solve this expression:

Lv̇|| = −v|| L̇ ⇒ dv|| L = 2v|| vm dt (67)

2-14. In plasma heating by adiabatic compression, the invariance of µ requires that

kT⊥ increases as B increases. The magnetic field, however, cannot accelerate particles
because the Lorentz force qv × B is always perpendicular to the velocity. How do the
particles gain energy?
Maxwell tells us that an electric field will be induced by a changing magnetic field. The induced
electric field is what accelerates the particles. 2

4-1. The oscillating density n1 and potential φ1 in a “drift wave” are related by
n1 eφ1 ω∗ + ia
= (68)
n0 kTe ω + ia
where it is only necessary to know that all the other symbols (except i) stand for
positive constants.
a) Find an expression for the phase δ of φ1 relative to n1 . (For simplicity, assume that
n1 is real.)
b) If ω < ω∗ , does φ1 lead or lag n1 ?
a) Solving for φ1 leads to

ω + ia n1 kTe n1 kTe (ω + ia)(ω∗ − ia) n1 kTe ωω∗ + a2 + i(aω∗ − aω)

φ1 = = = (69)
ω∗ + ia n0 e n0 e ω∗2 + a2 n0 e ω∗2 + a2

Now, in a drift wave we can suppose that φ1 ∼ exp(iδ), which in turns tells us that tan(δ) =
Im(φ1 )/Re(φ1 ). We have

n1 kTe ωω∗ + a2 n1 kTe a ω∗ − ω

Re(φ1 ) = 2 2
; Im(φ1 ) = (70)
n 0 e ω∗ + a n0 e ω∗2 + a2

Thus,
ω∗ − ω
δ = tan−1 (a ) 2 (71)
ωω∗ + a2
b) For ω < ω∗ , δ > 0. We can set the phase of n1 to be 0, since all that matters is a phase difference.
Thus, φ1 lags n1 . 2

4.2 Calculate the plasma frequency with the ion motions included, thus justifying our
assumption that the ions are fixed. (Hint: include the term n1i in Poisson’s equation
and use the ion equations of motion and continuity.
We will use Gauss’s Law, Fourier transforming the field and charge perturbations into plane waves
of the form x = x0 + x1 , where x is any quantity, vector or scalar. The subscript 0 indicates the
equilibrium value, and the subscript 1 indicates the perturbation. We only keep terms of to first
order in small quantities.
ρ e(ni − ne )
∇·E= ⇒ ik · E1 = (72)
0 0
Similarly, the equation of motion for the electrons,
dve
me = −eE ⇒ iωme ve = eE1 (73)
dt
ions,
dvi
mi = eE ⇒ iωmi vi = −eE1 (74)
dt
and continuity equation for electrons,
∂ne
+ ∇ · (ne ve ) = 0 ⇒ −ωne1 + ne0 k · ve = 0 (75)
∂t
and ions
∂ni
+ ∇ · (ni vi ) = 0 ⇒ −ωni1 + ni0 k · vi = 0 (76)
∂t
We now have 9 equations and 9 unknowns. I will skip the boring algebra. Solving for ω yields

ω 2 = ωp2 + Ω2p 2 (77)

q
2
ni e
where Ωp = m i 0
is the ion plasma frequency. Clearly, omitting the ion plasma frequency from
the calculation is justified since mi  me .

4.4 By writing the linearized Poisson’s equation used in the derivation of simple plasma
oscillations in the form ∇ · (E) = 0 derive an expression for the dielectric constant 
applicable to high-frequency longitudinal motions.
Fourier transform Poisson’s equation:
ρ e
ik · E = = n1 (78)
0 0
We also know that
∂n1 k·v
+ n 0 ∇ · v1 = 0 ⇒ iωn1 − n0 ik · v = 0 ⇒ n1 = n0 (79)
∂t ω
and we know Newton’s Law:
∂v e
m = −eE ⇒ v=i E (80)
∂t mω
Thus, equation (78) gives us

ie e 2 n0
k·E=− n0 k · v = k·E (81)
0 ω 0 ω 2 m
We can do a trick here, and pull everything over to the left side. Writing ωp2 ≡ e2 n0 /m0

ωp2 ωp2
k · E(1 − )=0 ⇒ ∇ · {(1 − )E} = 0 (82)
ω2 ω2
and thus we obtain
ωp2
=1− 2 (83)
ω2
4.6
a) Compute the effect of collisional damping on the propagation of Langmuir waves
(plasma oscillations), by adding a term −mnνv to the electron equation of motion and
rederiving the dispersion relation for Te = 0.
b) Write an explicit expression for Im(ω) and show that its sign indicates that the
wave is damped in time.

a) The cold electron equations of motion are

∂v
mne = −ene E − mne νv ⇒ iωmv = −eE − mνv (84)
∂t
and
∂ne
+ ∇ · (ne v) = 0 ⇒ iωne − ine k · v = 0 ⇒ k·v=ω (85)
∂t
We also have Gauss’s law:
ne e ne e m
∇·E=− ⇒ k·E=i = i ωp2 (86)
0 0 e

We will dot equation (84) with k:

(iωm − mν)
k·E=k·v (87)
e

Plugging in equations (86) and (87), we obtain

m 2 (iω 2 m − mων)
i ω = ⇒ ωp2 = ω 2 + ων 2 (88)
e p e

So we see that if we include collisions, the oscillation frequency is different from the plasma fre-
quency.

b) Lets let ω = ωR + iωI . Then expression (88) becomes

ωp2 = ωR
2
− ωI2 + 2iωR ωI + iωI ν + ωR ν (89)

This means that

ν
2ωR ωI + ωI ν = 0 ⇒ ωI = −2 (90)
2
Now we suppose a plane wave solution for the field quantities, i.e.

E ∝ e−iωt (91)

we obtain
νt
E ∝ e−iωR t+ωI t ⇒ E ∝ e−iωR t e− 2 2 (92)
Thus, the wave is exponentially attenuated in time.