Collision Theory

General Definition of a Collision Cross-Section

Cross-sections can be defined for a large number of “events due

to collision: simple scattering, excitation to some energy level,
ionization, etc.
To understand the definition of a cross-section, consider first
n1 
2 a simple situation where a population of “field particles  2
g n2 are effectively at rest, and are subjected to a shower of “test
particles 1 (a particle beam with a flux Γ12 = n1g ). The
collisions between the two populations produce a certain “event
at a rate R12 (per unit time, per unit volume), which, of course,
is proportional to n2 ,
R12 = n2 ν12
The rate is also proportional to the incoming flux n1g , and the cross-section for that event
is defined as,
# of events per particle 
2 per second
σ12 = (1)
Incident flux of particles 
1
Dimensionally:
t−1
[σ12 ] ≡ ≡ L2 (an area)
(L−3 )(Lt−1 )

Experimentally, detectors in the lab frame would count the events ν12 and the flux Γ12 . For
some “events the rate ν12 will be affected by the fact that the particles 2 are, in general,
also moving, and the cross-section definition must then specify the frame of reference used.
It will turn out that the most useful definition for all rate calculations is when the rela-
tive frame is used, i.e., the frame in which a particular particle 
2 is taken to be at rest.
Laboratory measurements must then be corrected to that frame. We will return to this later.

The Differential Scattering Cross-Section

For simple scattering (elastic), an “event is

defined as the deflection of particle 1 into a
dΩ range dΩ of solid angles about some obser-
g   Using Polar coordinates,
vation direction Ω.

g dΩ = sinχdχdφ (2)

1
b χ φ or if there is symmetry, for all φ,

2
dΩ = 2πsinχdχ (3)
The differential scattering cross-section is
then defined as,
# of particles 
1 scattered per second into dΩ
σ12 (χ) = (4)
n1 g

1
Notice that, in general, for a particular interaction potential V (r) between the particles, the
scattering angle χ depends on relative velocity g and “impact parameter b (miss distance).
If g is fixed, the number of particles scattered into the solid angle (ring) 2πsinχdχ is the
same as that arriving within the ring 2πbdb, provided χ = χ(b):

σ12 (χ)sinχdχ = bdb (5)

or,  
b  db 
σ12 (χ) = (6)
sinχ  dχ 
where the absolute value is used because the same argument applies whether χ increases or
decreases with b.

Total Scattering Cross-Section

Considering all possible impact parameters that lead to an interaction (notice that a cutoff
distance might be invoked), the total scattering cross-section is,
 bmax
Qtot
12 (g) = 2πbdb = π(b2max ) (7)
0

or,  π
Qtot
12 (g) = 2π σ12 (g, χ)sinχdχ (8)
0

Momentum-Transfer Cross-Section

As noted, the definition of Qtot 12 is generally divergent, unless a clear cutoff bmax can be
identified. A more useful total cross section results from consideration of the momentum
transferred during the collision. A more precise argument would require transformation of
the equations to the relative frame of  2 (which we will do later); for now, we notice that

the forward momentum of 1 before collision is m1 g, while after collision (accepting that in
an elastic collision the magnitude of the velocity does not change, i.e., g  = g), it is m1 gcosχ.
The momentum loss by  1 (or gain by  2 ) is then,

Δp1 = m1 g (1 − cosχ) (9)

The more complete argument (see later) would yield,

m 1 m2
Δp1 = μ12 g (1 − cosχ) with μ12 = (10)
m 1 + m2
We now multiply both sides of Eq. (4) times Δp1 and integrate for all deflections χ (including
the solid angle element dΩ = 2πsinχdχ),
 π
rate of momentum loss by all  1 due to one 2
2π σ12 (g, χ)Δp1 sinχdχ =
0 n1 g

2
and if we choose to represent the momentum loss rate as the momentum flux μ12 n1 g 2 times
a cross-section, we must define that cross-section as,
 π
∗
Q12 (g) = 2π σ12 (g, χ)sinχ(1 − cosχ)dχ (11)
0

or, alternatively,  ∞
Q∗12 (g) = 2π b [1 − cosχ(b)] db (12)
0

where we now can extend the range of b to ∞, since the factor 1 − cosχ(b), which becomes
very small for large b (small χ) ensures convergence in general (although not always!).
As a general rule, Q∗ and Qtot are comparable for nearly isotropic types of scattering (e.g.,
electron-neutrals at low energy, or neutral-neutral), but Q∗ is clearly lower than Qtot (by up
to 50%) for high-energy collisions, which tend to be more forward-biased.

Classical Elastic Collision Theory

Since collisions occur at atomic distance, their rigorous analysis requires Quantum Mechan-
√
ics. Specifically, this is so whenever the distance of closest approach, (of the order of Q∗ ) is
comparable
 to or√less than the Broglie wavelength for the relative momentum /p. Putting
p ∼ μ kT /μ = μkT , quantum effects dominate when,
 2 
∗ 
Q < (13)
μkT

For n-n collisions, μ > mH = 1.7 × 10−27 kg, and at T = 3000K this requires Q∗ <
10−22 m2  actual Q∗ . So for this type of collision, classical dynamics can be used. For
e-n collisions, μ ∼ me ∼ 10−30 kg, and taking T ∼ 10eV ∼ 105 K , the condition is
Q∗en < 8 × 10−21 m2 . Typical Q∗en values tend to be ∼ 10−19 m2 , so even in this case
we have some grounds for using classical dynamics. But in detail, many features of e-n
collision behavior are traceable to Quantum effects (such as the Ramsauer deep minimum
in Q∗ at energies where electrons resonate with the atoms potential well).
In what follows, we use Classical Mechanics for estimating some cross-sections, and then
also for calculating overall collisional effects using these cross-sections. In practical use, the
cross-sections are themselves obtained by laboratory measurements (or sometimes by precise
quantum computations), but since momentum and energy conservation are common to both
theories, the use of Classical Mechanics given the cross-section is on firm grounds.

Reduction to Relative Coordinates

Define,
w
1 ≡ velocity of particle 
1 in lab frame, before collision
w
2 ≡ velocity of particle 
2 in lab frame, before collision
 1
w ≡ velocity of particle 
1 in lab frame, after collision
 2
w ≡ velocity of particle 
2 in lab frame, after collision

3
Instead of the pair (w
 1, w
 2 ) the collision will be analyzed using the pair,

 = m1 w
G
 1 + m2 w
2
m 1 + m2 (14)
1 − w
g = w 2
 1 and w
Solving for w  2,
+ m2
w
1 = G g
m 1 + m2
m1 (15)
w −
2 = G g
m 1 + m2
Let F21 (r) be the force exerted by 
2 on 
1 , which is assumed to be a function of r = |r1 − r2 |,
and to be along the r1 − r2 vector. Then,
dw1
m1 = F21
dt (16)
dw2
m2 = −F21
dt
From (16),

dG
= 0 (17)
dt
 is not changed by the interaction. Also, from (16),
So the c.m. velocity G
 
dg 1 1 m 1 + m2 
= + F21 = F21
dt m 1 m2 m1 m2
or,
dg
μ12 = F21 (18)
dt
where μ12 is the Reduced Mass,
m 1 m2
μ12 = (19)
m 1 + m2
(notice if m1 << m2 , μ12 ≈ m1 , while if m1 = m2 , μ12 = m1 /2).
Comparing (18) to (16) we see that the relative motion of particle  1 (as seen from the
accelerated frame of 2 ), under their mutual force, is as if  2 were at rest, except that
the mass m1 is to be replaced by the smaller mass μ12 . All dynamical properties known
for motion about a fixed center of force can be applied now. In particular, the angular
momentum,
L = μ12r × g (20)
is a constant vector, which shows the motion is planar and that within this plane (using
polar coordinates),
L = μ12 r2 θ̇ ≡ constant (21)
The total kinetic energy in the lab frame is (with m = m1 + m2 ),
1 1
K = m1 w12 + m2 w22
2 2
or,
m 2 μ12 2
K = G + g (22)
2 2
4
whereas the overall momentum is,

p = m1 w
 1 + m2 w 
 2 = mG (23)

We already know that G is constant. If, in addition, the collision is elastic, then K is constant
as well, and then Eq. (22) shows that,

|g | = g ≡ constant (24)

So, the relative velocity vector g is only rotated by the interaction.

It is of some interest to investigate the possible use of a different set of velocities for analysis.
We retain G  as one of them, but take as the other the velocity of  1 relative to the center of
mass,
 1G = w
w 1 − G (25)
From (14),
m1 w 1 + m2 w
2 m2 m2
1 −
 1G = w
w = 1 − w
(w  2) = g (26)
m 1 + m2 m 1 + m2 m
(and it follows that w 2G = w2 − G  = −m1g /m). We see from (26) that the velocity with
regard to the center of mass is just a scaled version of that with regard to particle  2 . It fol-
G
 1 , and therefore that the differential scattering
lows that the rotation χ of g is also that of w
cross-section could be calculated in either frame.

Energy and Momentum Transfer in Elastic Collisions

The momentum increase of 

1 (decrease for 
2 ) in the collision is,

 1 − m1 w
ΔP1 = m1 w 1

or,
m1 m 2 
ΔP1 = (g − g ) (27)
m
which justifies a result we advanced in a previous section.
Similarly, the increase in energy of 
1 (decrease for  2 ) is,
1 1
ΔE1 = m1 w12 − m1 w12
2 2
Hence, since g  = g,
ΔE1 = μ12 (g  − g ) · G
 = ΔP1 · G
 (28)

5
Important observation:

Even though the collision is elastic, and no total energy is lost, there is an exchange of energy
between the particles, unless their c.m. is at rest.

In some cases, particle  2 can be regarded as effectively at rest,

 2 = 0 (for example, if 
w 1 is an electron and 
2 is a heavy particle).
In that case we have,
g 
g  − g  = m1 g = m1 w
G 1
m m
χ
and since,  
g g

(g − g ) · = −g (1 − cosχ)
g
then we have,
g
ΔP1 · = −μ12 g (1 − cosχ) = −μ12 w1 (1 − cosχ)
g
 is along g /g, so,
 2 = 0), G
Also (for w
m1
ΔE1 = −μ12 (1 − cosχ) w12
m
ΔE1 2μ12
= − (1 − cosχ) (29)
E1 m
This is maximum for χ = π (a head-on collision), yielding,
 
ΔE1 4μ12 m 1 m2 m2 /m1 m1 /m2
= − = −4 = −4 2 = −4
E1 max m (m1 + m2 ) 2
[1 + (m2 /m1 )] [1 + (m1 /m2 )]2
 
ΔE1
which is largest if m1 = m2 ( E1
= 1 in that
max
  case). But for m1 /m2  1 (electron-heavy collision),
 ΔE1 

 E1 
  
 ΔE1  m1
1
max
  ≈ 4 1
 E1  m2
max

which shows a very poor energy transfer efficiency be-

m2 tween light and heavy particles, but a good one for
1 m1 like-mass particles. This is why heavy particles eas-
ily thermalize among themselves, but electrons can
end up decoupled thermally from the rest of the
gas.

6
The Classical Calculation of Elastic Scattering Cross-Sections

Let V (r) be the interaction potential energy, such

that,
F = ∇V
g  (30)
Then, by conservation of total energy,
M  
g rm 1 1
μ12 ṙ + r θ̇ + V (r) = μ12 g 2
2 2 2
(31)
r θm 2 2
b θ χ
and by conservation of angular momentum,

r2 θ̇ = gb (32)

Eliminating θ˙ and writing μ = μ12 ,

g 2 b2 2V
ṙ2 + r2 4
+ = g2
r μ
or,
b2 2V (r)
ṙ = ±g 1− − (33)
r2 μg 2
Time can be eliminated by dividing (33) by θ˙ = gb/r2 ,

2
dr r b2 2V (r)
= ± 1− 2 − (34)
dθ b r μg 2

Here the (+) sign applies past the point M of closest approach, while the (−) applies before
M. At M, the distance rm follows from dr/dθ = 0, or,

b2 2V (rm )
1− 2
− = 0 (35)
rm μg 2

Turning (34) upside down, and integrating from (r = ∞, θ = 0), with the (−) sign, to
(r = rm , θ = θm ), we obtain,
 rm
(b/r2 )dr
θm = −

1 − rb2 − 2Vμg(r)
2
∞
2

or using ξ = b/r, we can write (35) as,

2 2V (b/ξm )
1 − ξm − = 0 (36)
μg 2
and therefore,  ξm
dξ
θm =
(37)
2V (b/ξ)
0 1 − ξ2 − μg 2

7
From the geometry,
χ = π − 2θm (38)
hence,
1 − cosχ = 2cos2 θm
sinχ = 2sinθm cosθm
and one can then complete the differential scattering cross-section using (6) and the total
and momentum transfer cross-sections using (8) and (11), or (12).

The hard sphere model

If each molecule behaves as a hard sphere (R1 for 1 , R2 for 2 ), the interaction can be
replaced by that of a point mass 
1 with a field particle 
2 of effective radius R0 = R1 + R2 ,
as in the figure.

This can be seen as the limit of a smooth potential (as we

will see below), or it can be dealt with directly. From the
g  geometry,
b b2
g χ sinθm = and cos2 θm = 1 −
R0 R02

1
b θm and from (38),
R0 sinχ = sin2θm = 2sinθm cosθm

2
 
2 b2
1 − cosχ = 1 + cos2θm = 2cos θm = 2 1 − 2
R0
We can now use Eq. (12) for the momentum transfer cross-section,
 R0    
b2 1 1
∗
Q12 = 2π 2b 1 − 2 db = 4πR0 2
− = πR02 = π (R1 + R2 )2 (39)
0 R0 2 4

We can also calculate the simple total cross-section from (7),

 R0
tot
Q12 = 2π bdb = πR02 = π (R1 + R2 )2 (40)
0

which, in this case, turns out to be the same as Q∗12 . An important observation is that
neither of them depends on g. All other interaction potentials yield cross-sections that do
depend on g.

Power law potentials

Consider interaction forces of the general type F (r) = ±α/rs , leading to,
 ∞
±α
V (r) = F (r)dr =
r (s − 1)rs−1

8
In terms of ξ = b/r, we have,
αξ s−1
V (ξ) = ± (41)
(s − 1)bs−1
Attractive potentials carry the (−) sign, repulsive ones the (+) sign. The relative velocity
influences cross-sections (Eq. 37) through the group,

2V 2 αξ s−1
= ±
μg 2 μg 2 (s − 1)bs−1

and defining a characteristic impact parameter,

  s−1
1
α
b0 = (42)
μg 2

2V 2 ξ s−1
= ± (43)
μg 2 s − 1 (b/b0 )s−1
Define now y = b/b0 , and substitute in (37) and then in (12),
⎡ ⎤
 ∞ 
2⎢
⎢ ξm dξ ⎥
∗
Q12 = 4πb0 2
ydycos ⎣  ⎥ (44)
 s−1 ⎦
0 0 2
1− ξ2 ∓ s−1
ξ
y

where ξm satisfies,  s−1

2 2 ξm
1− ξm ∓ = 0 (45)
s−1 y
The integrals in (42) need generally to be numerically completed. A conventional way to
write the result, and some numerical results, are as follows,

Q∗12 = πb20 2A1 (s, ±) (46)

s (+) or (−) A1
2 ± ∞
3 + 0.783
5 + 0.422
7 + 0.385
∞ + 0.5

The s → ∞ case is the hard-sphere limit, with b0 = R0 = R1 + R2 . The s = 2 case corre-

sponds to Coulombic interaction, and it is obvious that the (1 − cosχ) factor in the integrand
for Q∗ is not sufficient to produce a finite result. This case will be examined in more detail
below. One final comment is the absence of attractive potentials in the table above. The
integrations must be carried out very carefully in that case, because there are ranges of (b1 g)
which lead to capture in some cases.

9
The Case of Coulomb Collisions

This is a particular case of a power-law potential with,

z1 z2 e2
α = and s=2 (47)
4π0

(z1 , z2 are the charge numbers of the particles; z = −1 for electrons, +n for an nth -charged
positive ion). From this, the characteristic impact parameter is,

z1 z2 e2
b0 = (48)
4π0 μg 2

which can be positive (repulsion) or negative (attraction). Notice that for the case of electron-
electron collisions, μee = me /2, whereas for electron-ion, μei ≈ me . Hence,

b0ee = 2 |b0ei | (49)

Since b > 0, the quantity y = b/b0 can be positive or negative, so both cases are represented
by (from (37)),
 ξm
dξ
θm =  (50)
0 1 − ξ 2 − 2ξ/y
2
where ξm + 2ξm /y − 1 = 0, i.e.,

1 1
ξm = − ± +1 (51)
y y2

In order to have ξm = b/rm > 0, the (+) side must be adopted for either attraction or
repulsion in (51).
Eq. (50) can be integrated explicitly to,
 ξm    
ξ + 1/y 1 1
θm = sin−1  = sin−1 (1) − sin−1  = cos−1  (52)
1 + 1/y 2 0
1 + y2 1 + y2
or,
χ 1 χ
cos2 θm = sin2 = leading to y = cot (53)
2 1 + y2 2
or,
χ
b = b0 cot (54)
2
For attraction, both b0 and χ are negative, so b is still positive.
We can now calculate the differential scattering cross-section from (6) and,

db b0 1 χ χ
= − 2 and sinχ = 2sin cos
dχ 2 sin χ/2 2 2

b20
σ = (55)
4sin4 (χ/2)

10
which was first derived by Rutherford. This is clearly very forward-biased (strong decrease
of σ(χ) as χ increases from zero).

In terms of b, using,
χ 1
sin2 =
2 1 + y2
 2
b20 b2
σ = 1+ 2 (56)
4 b0
The momentum transfer cross-section is then,
 ∞  ∞
∗ 2 2 ydy  ∞
Q = 4π b0 cos θm ydy = 4π 2
= 2πb20 ln 1 + y 2 0 → ∞ (57)
0 0 1+y

Here, even the factor 2cos2 θm = (1 − cosχ) is not enough to eliminate the divergence that
happens at large b (small χ). Clearly, that is because of the strong singularity of σ(χ) at
χ = 0. The divergence is weak (logarithmic) and should not arise for any potential that is
less spread-out than the Coulomb potential.
Physically, we know that the plasma has a strong tendency to shield away any region of
concentrated charge. We make a small detour here to show that this modifies the Coulomb
potential of an isolated charge (ze) in an essential way, and we will then use this result to
complete the calculation above.

Consider a plasma where electrons have (as a fluid) negligible inertia, so that only pressure

gradients and electric fields matter, ∇Pe ≈ −ene E.
 = −∇φ,
With constant Te , using Pe = ne kTe and E
∇ne e∇φ eφ
= → ne = ne0 e kTe (58)
ne kTe
where ne0 is the electron density where φ = 0. Consider now an isolated ion of charge (ze)
and assume the ion density in its neighborhood is undisturbed (equal to ne0 ) due to their
large inertia, while the electron density may locally increase (in a statistical sense) due to
the ions attraction. The net charge density is then −e(ne − ne0 ), and from Poissons equation
in spherical coordinates,
 
1 d 2 dφ ene0  kT
eφ

r = e e − 1 (59)
r2 dr dr 0

Not very near the ion, φ << kTe /e, so expand the exponential to 1st order,
 
1 d 2 dφ ene0 e
2
r ≈ φ
r dr dr 0 kTe
e2 n
The group 0 kTe0e is 1/λ2D (λD = Debye’s length). Also, put φ = ψ/r. Substituting and
simplifying,
d2 ψ ψ
2
− 2 = 0
dr λD

11
the solution that does not explode as r → ∞ is,
C −r/λD
ψ = Ce−r/λD or φ = e (60)
r
Near the ion (r  λD ) this behaves as φ ∼ C/r, so C must be ez/4π0 ,
ez −r/λD
φ = e (61)
4π0 r
This shows that the ions potential is Coulombic only inside its Debye sphere, while it decays
much faster (exponentially) outside. This screened Coulomb potential should be used in Eq.
(37), but this would require numerical integrations. Instead, a simple device is to exclude in
the impact parameter integration (57) all values of b larger than one Debye length. Given
the weakness of the integral’s divergence, this should be adequate. We therefore return to
(57) and set the upper limit of the integral equal to λD /b0 ,
 
∗ 2
 
2 λD /b0 2 λ2D
Q ≈ 2πb0 ln 1 + y 0 = 2πb0 ln 1 + 2 (62)
b0

The quantity,
λ2D
Λ = 1+ (63)
b20
is called the “Coulomb logarithm, and our result can be written,

Q∗ = 4πb20 lnΛ (64)

According to (48),
z1 z2 e2
b0 =
4π0 μg 2
which shows Q∗ ∼ 1/g 4 , which implies that Coulomb collisions are most important at low
energies. Inside the logarithm, it is sufficient to use the average value of μg 2 , from,
 
1 2 3  
μg = kTe → μ g 2 = 3kTe
2 2

so that, 
λD 0 kTe 12π0 kTe 12π (0 kTe )3/2
= = (65)
b0 e2 ne z1 z2 e2 z1 z2 e3 n1/2
e
This ratio is related to the average number of electrons inside the Debye sphere,
 3/2
4 4π 0 kTe
ND = πλ3D ne = ne (66)
3 3 e2 ne

Comparing (65) and (66) we see that,

λD
= 9ND (67)
b0

12
For validity of our statistical treatment of the ion’s neighborhood, we should have ND  1
(and hence Λ  1). For verification, assume Te = 1eV = 11600K, ne = 1018 m−3 , z1 = z2 =
1 . We calculate λD = 7.4 × 10−6 m and b0 = 4.8 × 10−10 m,
λD
≈ 1.5 × 104  1
b0
and so, to a good approximation, (63) becomes simply,
λD
Λ ≈ (68)
b0
For our example, lnΛ = ln(1.5 × 104 ) ≈ 9.6. For most plasmas of interest, lnΛ ranges only
from about 5 to about 20.
The parameter b0 has in this case a simple interpretation.

g  g
b0
g or
(−)
b0 (+) g 

From (54), b = b0 cot(χ/2), so that b = b0 implies χ = ±90◦ . b0 is called the Landau impact
parameter, or “90◦ deflection parameter”.

Relationship to Lab-Frame Quantities

Data leading to the experimental determination of cross-sections are invariably taken by

instruments rooted in the laboratory frame, and then converted to the relative (or center of
mass) frame. The general reduction is fairly tedious, but it is useful to consider the simpler
case where particles of kind  2 are much slower than those of kind  1 (say, heavy particles
vs. electrons), in which case we can take w  2 = 0 . From Eqs. (14) then G = m1 w
 1 /m and
g = w 1 . From (15), expressed after collision,

 1 = G
w  + m2 g  = m1 w1 +
m2 
g
m m m
These relationships are best viewed graphically in the fig-
 1
w m2 
g ure.
m The lab-frame deflection of  1 is the rotation χL of w
 1 as
it becomes w 1 . We have,
χL χ
m2
m 1
w sinχ
 = m1 w
m2 tanχL = m2
G 1 w
1 m 1
w cosχ + mm1 w1
m m
or,
w
1
sinχ
tanχL = (69)
cosχ + m 1
m2

13
In the simple case m1 /m2 = 1, this gives χL = χ/2, and in general χL < χ, due to the recoil
of particle 
2 (notice we only assumed w  2 = 0).
 2 = 0, but not w
Suppose now a detector measures the number of particles scattered into unit solid angle
2πsinχL dχL in the lab frame, and this is used to determine the differential scattering cross-
section σL (χL ). As long as χ and χL are interrelated through (69), we can also express the
number in terms of relative frame quantities, so that,

σsinχdχ = σL sinχL dχL (70)

We can then use (69) to calculate dχL /dχ and χ(χL ). The result, after some algebra, is,
2
(1 + r2 + 2rcosχ) m1
σL = σ with r= (71)
1 + rcosχ m2
or, perhaps more directly useful,
 
1 − r2 sin2 χL + rcosχL 1 − r2 sin2 χL
σ = σL    3/2 (72)
r2 − 1 + 2 1 − r2 sin2 χL 1 − r2 sin2 χL + rcosχL

Because of (70), the total cross-section Qtot will be the same when computed in either frame;
but since the factors 1 − cosχ and 1 − cosχL are not accounted for, the momentum transfer
cross-section will be different. But Q∗ is only useful in the relative (or c.m.) frame. In fact,
the momentum transfer from  1 to  2 is not μ12 w1 (1 − cosχL ), because w1 = w1 .

So the process is:

1. Measure σL (χL )

2. Calculate σ(χ)
π
3. Integrate to Q∗ = 2π 0
σ(χ)(1 − cosχ)sinχdχ

14
Application: Thompson’s calculation of ionization cross-section by
electron impact

Ionization occurs when a free e imparts more than E∞ − En to a bound electron in the nth
state. Assume this e is at rest and free.
m1 2 me me 1
ΔE2 = −ΔE1 = μ(1 − cosχ) w1 = (1 − cosχ)  w12 = (1 − cosχ)me w12
m 2 2me 4

For this to be more than ΔE = E∞ − En (the ionization energy),

 
4(E∞ − En ) −1 4(E∞ − En )
1 − cosχ > or χ > cos 1 −
me w12 me w12
π
The cross section for ionization is then Qi = 2π χmin σ(χ)sinχdχ, where, for the Coulomb
interaction,
b20 /4 e2 e2
σ(χ) = → b0 = =
sin4 χ2 4π0 μg 2 2π0 me w12
 π  1  
b20 /4
χ χ 2 ds 2 1
Qi = 2
π 4χ 2
sin cos dχ = 2πb0 = πb0 −1
χmin sin 2 2  2! " sin χmin
2
s3 sin2 χmin
2
2d(sin χ
2
)

and,
χmin 1 − cosχmin 2ΔE
sin2 = =
2 2 me w12
therefore,  
me w12
Qi = πb20 −1
2ΔE
Now define,
1
m w2
2 e 1 e4
u ≡ → πb20 = 
π
ΔE 16π
2 20 (ΔE)2 u2
In terms of the Bohr radius,
e2 me v 2 4π0 2
= and me va0 =  → a0 =
4π0 a
20 a0
 me e2

and the H atom ionization energy (from n = 1 to n = ∞),

 
e2  H e2
EiH = recall that in bound orbits Ei  = |V (a0 )| = 1
8π0 a0 2 2 4π0 a0
then writing,
e4 e4 4πa20  H 2 4πa20
πb02 = = = Ei
16π20 (ΔE)2 u2 64π 2 ε20 a20 (ΔE)2 u2 (ΔE)2 u2
Therefore,
 2
EiH 1
πb20 = 4πa20
ΔE u2

15
and the ionization cross section is,
 2
EiH u−1 Ee
Qi = 4πa20 2
with u =
Ei u Ei

where, for instance, Ee is the energy of an incident electron and Ei is the ionization energy
of the target particle in the nth state. We observe that,
 
u−1 1
2
= when u = 2
u max 4

Finally,
 2
EiH
Qmax
i = πa20
Ei

Qi (Ee )
Qmax
i

∝ 1/E
e
∝E

Ee
Ei 2Ei

For H this would predict,

˚2 ) = 3.52 u − 1
Qi (in A with u=
Ee (in eV )
u2 13.6
This gives Qmax
i = 0.88A ˚2 at Ee = 27.2eV (actual is 0.7A
˚2 at 70eV ). It also gives
˚2 , actual is 0.44A
Qi (200eV ) = 0.22A ˚2 .

For helium, it would predict Qmax

i = 0.88(13.6/24.6)2 = 0.27A
˚2 at 49.2eV . Actual is 0.36A
˚2
at 90eV .

Note:The cross sections were for a stationary target, and as such contain relative speed g and
reduced mass μ. To account for the motion of the target, we need the velocity distribution
function f , and will treat that later.
For the Coulomb case, it turns out (1st order Chapman-Enskog theory) that the effective mo-
mentum transfer cross-section (to be associated with the mean thermal speed) in 6π 2 p20 lnΛ
instead of 4π 2 p20 lnΛ.

Discussion:

Notice that in a Coulomb orbit K = − 12 V and Etot = K + V = −K, i.e., the electron K is
numerically equal to the ionization potential from its orbit. Now, this is the least energy in

16
impairing electron can have if it is to produce ionization; therefore, the energy of the bound
electron is always less than that of the ionizing electron, and this partially justifies Thomp-
son’s assumption of neglecting the bound electron motion. The other assumption, namely,
that the electron is effectively free, has a similar justification, since |V | = −2K, so that |V |
is at most 2 times the energy of the impinging electron, and so the assumptions are expected
to be good at several times the ionization potential, and moderately good near the threshold.

More refined cross-section models (see Mitchner-Kruger, pp. 26-29)

 H
2
E1λ
Drawin: Q(k→λ) (E) = 2.66πa20 Ekλ
ξk β1 g(u)

u−1 E
g(u) = ln(1.25β2 u) , u =
u2 Ekλ

ξ = number of equivalent electrons in k th level

β1 ,β2  1 For β2 = 0.8, gmax = 0.2603 at u = um = 4.244

Gryzinski: QGR (E,Ekλ ;ΔE) = Cross section for transfer of energy > ΔE by electron at
E to electron in level k
 2
H
E1λ E Ek
QGR = 4πa02 ξk g(u, v) ; u = , V = λ
ΔE ΔE ΔE
 3/2   v+1
v #   $  1/2 %&
u−1 u 1 2v 1 u−1
g(u1 v) = 1− 1+ 1− ln e +
u2 u+v u 3 2u v

For ionization, ΔE = Ekλ (v = 1)

For a k → l transition, difference between Q s for ΔE = l + 1 and for ΔE = l

Estimate of 3-body recombination rate (Thompson)

Applies for Te  Ei , because at high Te it is hard to arrange that any e loses so much energy
as to be captured.

R ≡ Rate of e − i recombination (3 body, electron 3rd body) in a gas at Te .

R = R1 pee
R1 ≡ Number of times per second that an electron will pass within an ion’s “sphere of
influence” (SOI),
e2
r0 ∼
4π0 23 kTe

17
pee ≡ Probability of an e-e collision while the 1st electron is within the ion’s SOI

Strictly speaking, the e-e collision would have to be such that one of the electrons would
afterward have negative total energy and be captured (both have initially positive total en-
ergies before being accelerated into the ion’s field, and one of them should surrender enough
K to the other, so that its own K is now less than the magnitude of its (negative) potential
energy in the field of the ion). But since e-e collisions are effective for energy transfer, the
fraction of all possible e-e collisions satisfying this is of order 1 (for low Te ).

Now,
r0
R1  ne ni c̄e πr02 and Pee = 1 − e−r0 /λee  = r0 ne Qee with Qee ∼ πr02
λee

  5
8 kTe e2 1.04 × 10−20
R ne ni c̄e πr02 r0 ne πr02 R n2e ni π 2 r05 c̄e α=π 2
=
 ! " π me 6π0 kTe Te
9/2
α

The more rigorous value (Hinnov-Hirschberg) is,

1.09 × 10−20
R= 9/2
n2e ni
Te

Notice we did not need screening considerations here, since r0  λD (Λ  1) in cases of

interest. In other words, all of this happens within a Debye sphere.

Experimentally, Bates’ law gives

R 1.64 × 10−20
∼ 9/2
ne
ne ni Te

So, good within an order of magnitude, and has the right trends in it.

18
MIT OpenCourseWare
http://ocw.mit.edu

16.55 Ionized Gases

Fall 2014

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.