2 Collisions and cross sections

Elastic collisions are of paramount importance for Ohmic heating and en-
ergy dissipation into the plasma; inelastic collisions are responsible for all the
electronic excitation processes. For both processes, the deductive approach
is chosen. By electronic and ionic impact, electrons can be released from sur-
faces; this process is required for the carrier avalanche in DC discharges, but
has regained interest for stabilizing processes in production reactors whose
inner surface is subject to various (mostly unintended) coating reactions.

2.1 Introduction

Plasmas are chaotic gaseous systems which contain electrically charged carriers
and which maintain themselves by collisions of electric carriers with neutrals.
These impacts lead to further carriers, but other energy-consuming processes,
such as optical transitions, can be triggered as well. These collisions are sub-
sumed as inelastic. Nearly all of the gaseous ionizations are due to collisions
between slow heavy particles and rapidly moving electrons. If an electron falls
short of the required kinetic energy to excite a heavy particle (atom or molecule),
it can but exchange momentum with its counterpart, and we denote this process
an elastic collision.
This process dominates all other possible collisions since only a very small
fraction of electrons have kinetic energies which exceed the threshold of excita-
tion. However, due to the unfavorable mass ratio of individual neutrals/ions over
electrons, only a very small portion of energy is exchanged. The lack of efficiency
causes an athermal plasma (Chap. 3): Electrons and heavy particles (sometimes
even ions alone and neutrals alone) are in separate thermal equilibrium, and the
temperatures differ tremendously.
In the case of electron trapping by a neutral molecule or ion, respectively,
the electron density is reduced. This process is denoted electron attachment;
it will become effective beyond pressures of about some Torr. Below this limit,
diffusion processes dominate the loss of electrons: The particles exhibit random
walks, caused by stochastic collisions. When the reactor wall becomes a partner
of a collision process, the charge is terminated.

G. Franz, Low Pressure Plasmas and Microstructuring Technology, 5

DOI 10.1007/978-3-540-85849-2 2, 
c Springer-Verlag Berlin Heidelberg 2009
6 2 Collisions and cross sections

With elastic collisions, only a very small fraction of kinetic energy is ex-
changed between particles which significantly differ in mass, which leads to
only a small increase of kinetic energy of the translational energy of the gas
constituents, i. e. a small rise in gas temperature. The electron kinetic energy is
significantly reduced by inelastic collisions, which cause excitations of molecular
rotations, vibrations and eventually electronic transitions. It is this excitation
which generates ions, radicals and other dissociation products; furthermore,
metastable species can be formed. When these inelastic collisions are possible,
they dominate all other processes because of their large losses of energy. All
these species can again react among themselves, but they can also relax. An
electronic transition is often combined with radiation of light (UV/VIS range),
which is specific to the ambient conditions (cf. Chap. 9).
Our interest in these processes is twofold: Elastic scattering is mainly re-
sponsible for energy transfer and dissipation of energy; by inelastic scattering,
optical levels can be occupied or neutrals are excited to ions, and electrons are
generated. For electron generation, we distinguish between certain classes which
are denoted primary processes, i. e. thermal emission and field emission, and
secondary processes:

• Generation of secondary electrons at the cathode and the walls by particles

abundant in energy (γ-reaction): photons, electrons incident to the anode,
ions incident to the cathode.

• Ionization by impact by particles abundant in energy in the negative glow

and the dark spaces of either electrode (impact by electrons: α-reaction,
impact by ions: β-reaction).

Primary processes can be neglected in glow discharges. Ions are produced

only by gaseous reactions; otherwise, they can be generated by thermal emission,
however, with very low yield [13].

2.2 Elastic collisions

2.2.1 Cross section

In an approximation to first order, the cross section for elastic scattering is

adapted from kinetic gas theory, which regards molecules as hard spheres, and its
fourfold area is defined as the cross section; we begin with the relation between
mean free path λ and particle density n

1
λ≈ . (2.1)
nσ
The channel of a molecule with cross section σ as inner diameter is now ex-
posed to an electron beam with flux I0 , which is shot into an ensemble of
2.2 Elastic collisions 7

non-interaction particles. After having covered a certain distance x, the flux has
been attenuated to

I = I0 e−μx (2.2)
according to Beer and Lambert with μ a reciprocal length which can be
interpreted as mean free path λ following Eq. (2.1)
1
= nσ (2.3)
λ
with n the density of the molecules and σ the cross section of the specific in-
teraction. For a fundamental understanding of the cross section, the integral
attenuation of an electron beam, its angular and eventually its energetic depen-
dence are required.

2.2.1.1 Cross section and mean free path in kinetic gas theory. The
time τ between two collisions is of the order of τ ≈ λ/ < v > with < v >
the mean thermal velocity. The mean free path λ itself depends on particle
density n and cross section σ according to Eq. (2.3) because it roughly scales
with the target area of the molecule. In the simple approach of kinetic gas
theory, this cross section is the only one considered and is the cross section for
elastic scattering (see [14]). In this approximation, the cross section for elastic
scattering is independent of temperature (energy) and is tabulated in units of
πa20 —the cross section of the hydrogen atom with a0 Bohr’s radius (0.529 Å),
8.82 × 10−17 cm−2 [sometimes the averaged collision number Pc across 1 cm for
a gas at 1 Torr (133 Pa) and 0 ◦ C is found; the relation is σ = 0.283Pc in Å2 ].
A typical value for λ is 4.5 cm for nitrogen at 1 mTorr (0.2 Pa).
In fact, the cross section weakly depends on temperature (kinetic energy):
With decreasing temperature (reduced molecular speed), it rises slowly due to
prolongation of the interaction. For example, by reducing the temperature by
200 ◦ C from +100 ◦ C to −100 ◦ , σel rises by 30 % for nitrogen and oxygen.

2.2.1.2 Cross section and mean free path in plasmas. This temperature
dependence is caused by weak intermolecular forces (polarization), which exhibit
a weak r−6 dependence for the attractive part of the Lennard-Jones potential
ΦLJ . In plasmas, however, we have bare charges which lead to intense, far-
reaching interactions which can be described analytically only in some simple
cases. To begin with, we introduce some experimental details for atoms and
electrons:

• Atoms

– In a sputter system operated with argon at 50 mTorr (7 Pa), the

mean free path of the argon atoms is approximately 1.54 mm (25
meV Ar, σtot = 26 Å2 ).
8 2 Collisions and cross sections

• Ions

– The mean free path of ions is considerably smaller. λ for Ar+ ions of
medium velocity in argon at 30 mTorr (4 Pa) and 4 eV is about 0.87
mm and rises for 10 eV ions to 1 mm and for 100 eV atoms to 1.4
mm.
– In ion beam sources which are driven at significantly lower pressures,
they are larger by orders of magnitude: For 400 eV Ar+ ions and a
gas pressure of 10−4 Torr (13 mPa), λ equals 51.6 cm.

• Electrons

– λe of fast electrons in argon [16 eV, which is at the maximum of σ

(26 Å2 )] amounts to about 2.5 mm at a pressure of 50 mTorr (7 Pa),
corresponding to about 7.5 mm at 20 mTorr (3 Pa).
– Slower electrons (Ekin ≈ 4 eV) with a smaller σ (9.7 Å2 ) show a
longer λe (at 20 mTorr about 13 mm) which applies also to rapid
electrons (λe for 100 eV electrons is about 75 mm at 20 mTorr in
argon). The frequencies for elastic collisions between electrons and
neutrals yield, according to νm = nσv:
∗ for 1 Torr (133 Pa, 3.54 × 1016 cm−3 ) 3.8 × 109 s−1 ,
∗ for 100 mTorr (13.3 Pa) 382 × 106 s−1 and
∗ for 10 mTorr (1.3 Pa) 38 × 106 s−1 .

2.2.2 Definitions

First, we consider a beam of monoenergetic particles orientated in a parallel

direction with respect to the x-axis and exhibiting a flux of N particles per
cm2 sec (Fig. 2.1).

dW

Fig. 2.1. Definition of the dif-

q ferential cross section. Np pro-
Nt jectiles collide with Nt tar-
y
j gets and are deflected into dΩ,
the differential solid angle ele-
x ment.
Np
2.2 Elastic collisions 9

In the case of hitting an annular target with area 2πb db (b: inner diameter,
b + db: outer diameter), we define the differential scattering cross section for
solid angle dΩ = 2π sin ϑdϑ:

dσ(v, ϑ) b db
= , (2.4)
dΩ sin ϑ dϑ
and the total cross section (“c” for collision)
 π
dσ(ϑ)
σc (v) = 2π dΩ, (2.5)
0 dΩ
which, in turn, does not allow any statement of the angular dependence. Fur-
thermore, we define the frequency of collision according to

νc (v) = nσc (v)v, (2.6)

the mean free path again, but decorated with a subscript “c”
1
λ≈ (2.7)
nσc (v)
and the probability of scattering
1
Pm = . (2.8)
pλ
The change of momentum is in the center-of-mass system (Fig. 2.2):

Fig. 2.2. Collision between two

q
spheres in the center-of-mass sys-
b r tem. b: scattering parameter, θ,
CM
j ϕ: scattering angle, CM: center of
mass.

Δp = mv(1 − cos ϑ) : (2.9)

◦
Collisions in the forward direction are minimally weighted (cos 0 = 1), those in
the backward direction (cos 180◦ = −1), however, are considered most promi-
nent: The diffusion is slowed down most effectively by collisions which are ori-
entated backwards.
For the treatment of the transfer of energy and momentum, the most im-
portant property is the cross section for momentum transfer, which is σc (v), but
weighted by the factor 1 − cos ϑ:
10 2 Collisions and cross sections
 π
dσ(ϑ)
σm (v, ϑ) = 2π (1 − cos ϑ)dΩ, (2.10)
dΩ0
along with the frequency of momentum transfer:
v
νm = = νc 1 − cos ϑ. (2.11)
λ
The cross section of momentum transfer σm can be extracted from the total
collision cross section when its angular dependence is known. For a head-on
collision (ϑ = 180◦ = π), we calculate, using the laws of conservation for energy
and momentum, the maximum amount of transferred energy to (m the mass of
the pushing sphere, M the mass of the stroken sphere)
4mM
EM,kin = Em,kin (2.12)
(m + M )2
which is often rewritten using Langevin’s energy loss parameter
2m
L= (2.13)
m+M
eventually yielding [15]
2M
EM,kin = L Em,kin . (2.14)
m+M
It can easily be shown that the maximum of this function is given for m =
M : Collisions between ions with high kinetic energy and slowly moving parent
atoms are very effective (cf. Sect. 2.5.2). On the other hand, the energy transfer
between an electron and a heavy neutral is negligible (me /mAr ≈ 70 000).

2.3 Elastic collisions between electrons and neutrals

2.3.1 The total cross section

Measurement of the total cross section (sum of all possible scattering processes)
was first performed by Ramsauer using the instrument which is sketched in
Fig. 2.3. By exposing a plate of zinc sulfide to UV light, electrons are released
at point A by the photoelectric effect. Without collisions, they are forced into
orbital motions when a perpendicular magnetic field is applied. They pass all
the slots from S1 to S5 until they hit the collector C. With elastic collisions,
however, electrons are deflected and cannot pass; with inelastic collisions (with-
out deflection), kinetic energy is lost, the orbital diameter increases and these
electrons fail to reach the collector as well.
Under a certain pressure p, the electron current is measured in B (current i)
and C (current I). For known distance s between S4 and S5 , the cross section of
absorption can be evaluated from the currents i and I at two different pressures
p1 and p2 according to Beer’s law with α = nσ:
2.3 Elastic collisions between electrons and neutrals 11

S2 Fig. 2.3. Ramsauer’s apparatus to

S1 measure the total cross section [16].
S3
First, electrons are released from a
slab made of ZnS. By applying a mag-

hn
S4 netic field orientated normally to the
plane of the paper the electrons are
forced in orbital motions. Their diam-
A
eter is given by the distance between
S5
slots S2 and S5 . Scattering, however,
will lead to losses in the Faraday
cages B and C.
B C

i = Ie−αps (2.15)
 
I1 i 2
(p1 − p2 )αs = ln , (2.16)
I2 i 1
and for the lightest targets (hydrogen and helium), a hyperbolic dependence
as pictured in Fig. 2.4, is obtained. For zero energy, the curve saturates at a
certain constant level.

12

10

8 helium
cm ]
2

6
-16
s [10

Fig. 2.4. Experimental cross

4 sections for the elastic collision
of electrons through helium as
2
function of electron energy ac-
cording to [17].
0
0 10 20 30
electron energy [eV]

2.3.2 Differential cross section

We consider now the total cross section and the cross section of momentum
transfer, which can be evaluated after having measured the angular depen-
12 2 Collisions and cross sections

dence of the scattering amplitude according to Eq. (2.10). It was again Ram-
sauer (this time together with Kollath) who applied the measuring princi-
ple sketched in Fig. 2.5 [18]. A reactor contains a filament that emits electrons
which are accelerated across an electric field. These electrons are deflected by
gas molecules and are caught in Faraday cages and eventually measured.

electrometer with
Faraday cups

filament
grid system reactor collector

Fig. 2.5. Sketch of the apparatus developed by Ramsauer and Kollath to deter-
mine the angular dependence of the cross section for elastic scattering of electrons
after [18].

Some general guidelines are:

• At sufficiently high electron energies, we find an intense zero beam: Almost
no deflection and then a gradual decline for large scattering angles.
• Approaching low electron energies, this picture becomes more compli-
cated. Maxima and minima occur which resemble X-ray diffractograms
of fluids (Fig. 2.6). For the first time, this was observed by Bullard and
Massey in argon. They explained these patterns by diffraction of electron
waves at symmetrically scattering atoms, comparable with the pattern of
light waves which are scattered at spheres of similar dimension [19]. How-
ever, for the quantitative interpretation of this phenomenon, a quantum
mechanical treatise is required.
• The complexity of this pattern rises with increasing atomic number of the
scattering atom.
For elastic scattering of electrons of low energy, a successful collision is
assumed to have occurred for a deflecting angle of 90◦ . According to Eq. (2.11),
νm then equals νc . For the noble gases, both dependencies (on velocity and on
angle) have been measured; hence, σc (v) and σm (v, ϑ) can be compared. For low
2.3 Elastic collisions between electrons and neutrals 13

1.00
normalized scattering intensity

4.00 eV
0.75

0.50
2.80 eV

2.00 eV Fig. 2.6. Scattering of elec-

0.25 trons at argon atoms for very
1.15 eV low electron energies [20] (
c
J. Wiley & Sons, Inc.).
0.00
0 45 90 135 180
scattering angle [°]

energies, the agreement is satisfactory (argon) or almost perfect (helium, Figs.

2.7).

25
6 argon
helium 20 sc
sm
sc
s [10-16cm2]

s [10-16cm2]

4 sm 15

10
2
5

0 0
0 20 40 60 0 20 40 60
electron energy [eV] electron energy [eV]

Fig. 2.7. Comparison between total cross section σc and cross section for momentum
transfer σm for the lightest noble gas helium and the medium-light noble gas argon,
which behaves entirely differently for low electron energies [20]. For helium, σm exceeds
σc for low electron energies: The scattering in the backward direction becomes more
important than forward scattering.

2.3.3 Modeling

2.3.3.1 Cross section for the interaction between a point charge and
an induced dipole. For potentials which decrease with r−4 (i. e. for forces
which decline with r−5 ), there exists a simple correlation between the frequency
of collisions and number density, which was pointed out by Maxwell [21].
14 2 Collisions and cross sections

An electron is moving in a field of a potential Φ which declines with r−n (Fig.

2.2). Introducing polar coordinates yields
 2  2
∂r ∂φ
v 2 = v2 + v⊥
2
= + r2 , (2.17)
∂t ∂t
with v and v⊥ the two components of the velocity v which are orientated in parallel
or in perpendicular fashion with respect to the radius vector r, we find for energy
and angular momentum
 2  
m0 v 2 m0 ∂r ∂φ 2
E= +Φ= · +r 2
+ Φ = const, (2.18)
2 2 ∂t ∂t
∂φ
L = m0 r2 = const. (2.19)
∂t
With b the scattering parameter, the distance between the two particles under
consideration (for lacking interaction, the trajectory of the electron would be a straight
line) we get for the angular momentum

∂φ
m0 r2= m0 bv0 , (2.20)
∂t
with the boundary condition for the initial energy E0 (initial velocity v0 ) for vanishing
potential Φ = 0

m0 v02
E0 =. (2.21)
2
Since the time-dependent derivations of the coordinates are

∂φ ∂r ∂r ∂φ ∂r
= v0 br2 ∧ = · = · v0 br2 , (2.22)
∂t ∂t ∂φ ∂t ∂φ
we obtain

2
∂r
∂φ E0 b2
E0 b2
+
+ Φ = E0 . (2.23)
r4 r2
From Eq. (2.23), the angular dependence of r (∂r/∂φ) yields
 1/2
∂r r2 b2 Φ
=± 1− 2 − , (2.24)
∂φ b r E0
with the minus sign for approaching particles, the plus sign for disappearing ones.
This function exhibits a minimum at ∂r/∂φ = 0, when the argument of the square
root vanishes. With Φ = ar−n and b = cr, we find for the minimum distance rmin

n 2a
rmin = . (2.25)
m0 v02 (1 − c2 )
2 ,
Applying the hard sphere model, the minimum distance rmin equals to σscatt = πrmin
and we can write it down with combined constants:
2.3 Elastic collisions between electrons and neutrals 15

A
σscatt = . (2.26)
v 4/n
For n = 4, this yields the simple form

A
σscatt = , (2.27)
v
which is the potential which will form between a point charge and an induced dipole
with moment μind (with α the polarizability):

1 1
μind = αE ∧ Epot = − μ · E = − αE 2 . (2.28)
2 2
For a Coulomb potential, we calculate the potential energy to

1 e2
Epot = − α 04 . (2.29)
2 r

With this derivation, we have obtained the important result that for a force
with a declining behavior according to r−5 , e. g. a point charge in a neutral gas
which exerts polarization effects, the specific conduct of the EEDF (and hence
ve ) does not matter at all. In this very special case, the frequency of elastic
collision between electrons and neutrals νm (and also the mean free path λ) will
become a√single function of the number density [21].1 σ will then scale with
1/v ∝ 1/ Ekin (Fig. 2.8).

12

10

8
s [10-16 cm 2]

6 Fig. 2.8. Experimental and

calculated cross sections for
4 the elastic collision of electrons
through helium as a function
2 of electron kinetic energy ac-
cording to [17].
0
0 10 20 30
E [eV]

1
For the interaction between point charges Coulomb’s law holds: The potential drops
according to 1/rn=1 . In this case, σscatt = A/v 4 , which determines the generalized resistance
η in the case of high-density plasmas (cf. Sect. 14.7.2).
16 2 Collisions and cross sections

2.3.3.2 Ramsauer effect. For electron energies larger than some tens V, the
cross section decreases gradually; it scales with the inverted ionization potential
and is proportional to the polarizability (to a first order approximation: the
atomic number) since scattering takes place at the bound electrons of the atoms.
Very low values of the scattering cross section are caused by weak interaction
with the higher noble gas atoms (Ramsauer effect, Fig. 2.7.2 [22]). This effect
was explained for the first time using quantum mechanical methods by Allis
and Morse using the partial wave method.
According to this theory, the cross section of elastic scattering is composed of
partial cross sections, which are denoted s scattering for l = 0, p scattering
 for l = 1
etc. (ηl is the partial phase shift of the waves with angular momentum L = l(l + 1)h̄,
p = h̄k):
 4π
σ0 = σl ∧ σl = (2l + 1) sin2 ηl . (2.30.1)
l
k2
The affiliated potentials become deeper but simultaneously more short-ranged—
the atomic radius grows by about 150 % when going from helium (0.49 Å) to xenon
(1.24 Å) leading to a ratio of atomic volumes of about 16, whereas the number of
electrons increases by a factor of 27! For slow electrons (k → 0 ∧ λ → ∞), L becomes
very small, and for the noble gas atoms argon and heavier, only the component of
zero order, i. e. s scattering, remains of importance:
4π
σ0 = sin2 η0 . (2.30.2)
k2
η0 → π at k = 0 ⇒ σ0 = 0. Hence, a head-on collision (ϑ = π) between an electron
and an atom cannot be observed. For the heaviest noble gas atoms (Xe and Rn) this
effect is much more distinct than for the lighter ones (Ar and Kr) and vanishes for
He and Ne.

2.3.4 The frequency of elastic collisions between electrons and

neutrals

For the special cases of hydrogen and helium, the collision frequency simply
scales with discharge pressure (number density): νm = const × p0 , which follow
the equation with p0 = 273/T × p [p in Torr (in Pa), Figs. 2.9]:
• He: νm = 2.31 × 109 p0 (3.07 × 1011 p0 );
• H2 : νm = 5.93 × 109 p0 (7.89 × 1011 p0 ).
This does not hold true for the heavier noble gases (at least for neon, we see a
fairly good approximation with an asymptotic behavior) and a strange conduct
is observed for chlorine. Here, the maximum is foreshadowed by a minimum at
very low energies in Cl· and Cl2 (Fig. 2.10).
Following Massey, we estimate the mean free paths for electrons λe in argon
according to Eq. (2.7) [25]. For typical electron energies of several electronvolts,
we calculate a λe of several centimeters (Figs. 2.11).
2.3 Elastic collisions between electrons and neutrals 17

6 25
Xe
5 H2 Kr
20
n/p [10 9 sec Torr] -1

n/p [10 9 sec Torr] -1

4 Ar
15
3 He
10
2

1 5 Ne

0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Ekin [eV] Ekin [eV]

Fig. 2.9. Frequency of elastic collisions between electrons and neutrals for hydrogen
and helium, and the heavier noble gases [23].

108
nm [sec -1]

107 Fig. 2.10. Frequency of elas-

Cl2
tic collisions between electrons
and neutrals for Cl· and Cl2 af-
106
Cl ter [24]; discharge pressure 10
mTorr.
10-2 10-1 100 101 102
Ekin [eV]

2.3.5 Cross section and rate constant for argon

By Eqs. (2.6) + (2.11), the cross section for a specific scattering process is
related to the probability of its occurrence. This frequency ν depends on the
density of target molecules, and reducing ν to the unit density yields the rate
constant k,

k = < ve > σ (2.31.1)

in particular for the momentum transfer km :

km = < ve > σelast . (2.31.2)

This relation allies atomic properties which must be evaluated by scattering ex-
periments with macroscopic kinetic parameters, with the mean electron velocity
being the connecting link. The directed energy E from the scattering experi-
18 2 Collisions and cross sections

100 100 argon

mean free path [cm]

mean free path [cm]

argon
10
10 0.5 eV
1.0 eV
1 1 mTorr
1.6 eV
5 mTorr 1 2.4 eV
0.1 10 mTorr 4.3 eV
8 eV
50 mTorr 12 eV
100 mTorr
0.01 0.1
0 3 6 9 12 15 0 10 20 30 40 50 60
electron energy [eV] pressure [mTorr]

Fig. 2.11. Calculated mean free paths of low-energy electrons (below the ionization
threshold) in argon after [25].

ments has to be replaced by the thermal energy E for a discharge. Most fre-
quently, a Maxwellian distribution is assumed but especially for low plasma
densities (E = kB Te ), however, this assumption has turned out to be highly
questionable (Sect. 14.1). The connection between σ and k is pictured for ar-
gon in Figs. 2.12. The data for σelast have been compiled from Nakanishi and
Szmytkowski [26, 27].

30

2 100
10
20 k
k [10 -6 cm3sec-1]

-1
selast [10-16 cm2]

s 10
k [10 cm sec ]
selast [10-16 cm2]

-1
3

1 -2
1 10
-6

10 k
s
10-3
0.1
0 0 10-4
0 25 50 75 100 0.01 0.1 1 10 100
electron energy [eV] electron energy [eV]

Fig. 2.12. Argon: total cross section for elastic scattering σelast and the corresponding
rate constant k = σelast × < ve > after [25] − [27].

2.4 Elastic collisions between heavy particles

The scattering among heavy particles exhibits a remarkably distinct maximum

in the forward direction [28]. Measuring such an angular dependence is almost
hopeless. For energies of a few electronvolts, the requirements for angular resolu-
2.4 Elastic collisions between heavy particles 19

tion are difficult to meet, and these demands increase with rising momentum of
the projectiles. The accuracy necessary to measure the total elastic cross section
for argon is for thermal atoms 0.70◦ , those with 1 000 K temperature require an
accuracy of 0.30◦ , which is even sharper for atoms of 1 eV kinetic energy (11 600
K, 0.11◦ ). These difficulties are reinforced for the measurement and calculation
of the differential cross section which means the determination of the fraction
of molecules which hit target molecules and are subsequently scattered into the
solid angle dΩ = 2π sin θdθ. In Table 2.1, the results of calculations for the
differential cross section of helium against a beam of protons for two different
scattering potentials (Coulomb and Hartree, a simple pseudopotential) are
compiled [29].

Table 2.1. Differential cross sections dσ(ϑ)/dΩ per unit solid angle for protons of
kinetic energy of 110 eV in helium in units of a20 .

 dσ(ϑ)/dΩ
Hartree potential Coulomb potential
0 9 × 103 ∞
12 7.85 124.0
28 2.00 6.10
34 0.72 2.85
57 0.21 0.40
80 0.08 0.12
114 0.04 0.04
137 0.03
167 0.02

This forward direction is more distinct for larger kinetic energies of the pro-
jectiles (coming from low energies, only elastic scattering processes can occur).
Normally, the determination of the angular dependence is confined to the range
without the zero beam. One of the pioneering experiments was carried out by
Berry and Cramer in neon and argon [30] − [33] (Fig. 2.13). We note that
the cross section remains at several Å2 up to kinetic energies of about 500 eV.
Hence, an ion beam is almost entirely dissipated after having passed a distance
of only 1 cm at high pressures (cf. Sect. 6.8).
However, some advanced, indirect methods are available, e. g. measuring the
complex conductivity (Sect. 5.6), and the line width of the electron cyclotron
resonance, which is mainly determined by the collisions between electrons and
neutrals [34] (Sect. 7.5).
20 2 Collisions and cross sections

4 A
A - 250 eV
B - 500 eV
C - 750 eV
3 B D - 1000 eV
E - 1250 eV
s[10-16 cm2]

C
F - 1500 eV
D G - 1750 eV
2
E
F
G
1 Fig. 2.13. Cross section of
Ar+ ions in argon as a func-
0 tion of scattering angle [30].
0 45 90 135 180
scattering angle [°]

2.5 Inelastic collisions

2.5.1 Inelastic collisions between electrons and heavy particles

During inelastic collisions, internal grades of freedom are excited by transfor-

mation of kinetic energy of the colliding projectiles into internal energy levels
of the target. For electron impact, the most important reactions are with A∗ an
excited molecule:

• Excitation in upper molecular states (vibration levels or electronic states,

[Eq. (2.32)].

• Dissociation into radicals (which is very effective in discharges through

electronegative gases [35], [Eq. (2.33)].

• Dissociation into ions [Eq. (2.34)].

• α-ionization [Eq. (2.35)].

• Ionizing dissociation [Eq. (2.36)] and

• Electron attachment (dissociative or electron attachment, respectively),

for electronegative gases [Eq. (2.37)].

e− + A2 −→ A∗2 + e− (2.32)

e− + A2 −→ 2 A∗ + e− (2.33)

e− + A2 −→ A+ + A− + e− (2.34)
2.5 Inelastic collisions 21

e− + A2 −→ A+
2 + 2e
−
(2.35)

e− + A2 −→ A+ + A + 2 e− (2.36)

e− + A2 −→ A− + A∗ . (2.37)

2.5.1.1 Experimental methods. In atomic gases, dissociation reactions do

not take place; one of the “classical” reaction is the ionization of mercury vapor
by electron impact (Franck-Hertz experiment, Figs. 2.14). In a triode tube

anode
grid

filament

0.5 V

_ +
20 V

400
14.7 V
30
300 9.8 V
anode current [a. U.]

sion [10-16 cm2]

20
200 4.9 V

100 10

0
0 2 4 6 8 10 12 14 16 0
10 12 14 16
anode voltage [V] electron energy [eV]

Fig. 2.14. Franck-Hertz experiment. Top: experimental setup, bottom: the anodic
current (LHS) and the cross section of ionization (RHS) as functions of acceleration
voltage (electron energy) in mercury vapor [36]. At the ionization potential of 4.9
eV and its multiples, the characteristic exhibits distinct deviations from the expected
behavior, and the cross section does not exhibit a smooth behavior [37].

filled with mercury vapor, the electrons which are emitted by a glowing cathode
are accelerated through a uniform field to a grid with the anode plate placed
directly behind the grid. Those which have passed it will be slowly decelerated
by elastic collisions with gas atoms, but in general, we observe a slight increase
of the anodic current with growing anodic voltage. However, when the ioniza-
tion potential of mercury (4.9 eV) has been reached, the I(V ) characteristic
22 2 Collisions and cross sections

drops sharply. This phenomenon is caused by inelastic collisions of the electrons

with mercury atoms which are ionized by electron impact. A small retarding
field between grid and anode causes the anodic current to vanish at this value,
because the electrons have less than 4.9 eV kinetic energy and are not able to
move against this retarding potential:

Hgvapor + e− −→ Hg+ + 2 e− . (2.38)

This setup has been improved by Maier-Leibnitz [38] and further by

Schulz and Fox who employed a retarding potential difference method [39].
They concluded that the electron energy distribution function, EEDF, in the
original Frank-Hertz experiment was very broad, too broad to obtain an
accurate measurement of the excitation probability. For that purpose, the energy
of the bombarding electrons as well as the energy of those that have lost energy
in inelastic collisions must be known. Therefore, they constructed an energy
filter system which made use of also in the determination of the ion energy
distribution function IEDF (cf. Chap. 6). By means of this three-grid optics,
it is possible to extract a nearly monoenergetic beam of electrons (Figs. 2.15).

retarding potential F1 F2 F3
aperture 0
F1 F2 F3
F [a.u.]

collision
filament chamber

collector
x [a. u.]

Fig. 2.15. By employing the retarding potential difference method, Schulz and Fox
could significantly improve the accuracy of the determination of the inelastic scatter-
ing cross section [39].

The first grid is positively biased by approximately 3 V with respect to

the filament and serves to draw electrons. The second grid is slightly negative
with respect to the filament, the electrons must run uphill against a retarding
potential, and as a result, the EEDF is chopped off. It is only the high-energy
fraction of the electrons which is able to pass. The third grid shows the same
potential as the first grid. Since the EEDF can be sensitively influenced by
variation of Φ2 , Φ2 + ΔΦ2 , we obtain two values for the collector current which
can be solely referred to the electrons with energy ΔΦ2 . By variation of the
retarding potential Φ2 , grid 2 also varies the voltage to the collision chamber
which determines the collector current, i. e. the EEDF.
2.5 Inelastic collisions 23

2.5.1.2 Cross section. The cross section for an inelastic scattering process
exhibits a threshold which is followed by a steep rise (several orders of mag-
nitude) to reach a maximum. This is sharply peaked for optical transitions at
energies which are close to the threshold, and a little broader for ionizations at
about 100−150 eV, which is consistent with the de Broglie wavelength of the
electrons of approximately 1 Å, the typical diameter of the targets where the
electron waves are expected to be deflected most intensely. For higher energies,
we find a slow decrease since the time of interaction between the collision part-
ners is gradually shortened (Fig. 2.16 in linear scale, Figs. 2.17 in logarithmic
scale).

3
cross section [10-16cm2]

argon
hydrogen
Fig. 2.16. Typical energy de-
1
pendence of the cross sec-
tion for inelastic scattering be-
tween electrons and molecules
0 for argon and hydrogen.
10 100 1000
electron energy [eV]

The cross section of ionization has been comprehensively described by

Bethe and Salpeter [40, 41]. According to Born’s approximation, the total
cross section of a discrete excitement into state n with energy En of the excited
state can be approximated by
 
4πa20 z 2 C 1
σn =
A ln(4 B T ) + + O , (2.39)
T T T2
with a0 Bohr’s radius, z the charge and T the kinetic energy of the projectile,
mostly given in Rydberg (= 13.6 eV), irrespective of the nature of the projectile.
A is the squared matrix element of the specific excitation, and the B, C are
constants which will be generated during the integration [42], and the factor 4
in the logarithmic summand which is left for historical reasons. The third term
is a very small correction, and most frequently, the second term is neglected as
well, and we arrive at the following simplified approximation [43, 44], which is
known as Bethe’s asymptotic cross section [41]:

4πa20 z 2
σn =
[A ln(4 B T )] . (2.40)
T
Unfortunately, this theory holds best for large kinetic energies of the projectiles.
However, for ionization and excitation processes by electron impact, the energy
24 2 Collisions and cross sections

dependence just above threshold is of extreme importance. To describe this

regime properly, in particular the steep rise beyond threshold and the maximum,
Wannier developed a theory for single ionization which states that in the
energy range near threshold, the probability of ionization rises exponentially
with the energy in excess of the ionization energy (more correctly: raised to
the 1.127th power) [45, 46], which had been improved by Geltman [47]. Lotz
introduced an equation with three parameters (a, b and c), which approximates
the cross sections of most of the interesting gases within a deviation of less than
±10% to experimental data [44] (Fig. 2.17.2):
  
ln(T /En ) T
σn = a 1 − b exp −c −1 . (2.41)
T En

cross section [10-17 cm 2]

1 10
cross section [10-17 cm2]

21P

0.1 1
2S
1

argon
3
2P 0.1 hydrogen
0.01

0.001 0.01
10 100 1000 10 100 1000
electron energy [eV] electron energy [eV]

Fig. 2.17. Typical dependence of the inelastic cross section of molecules and electrons
on the electronic kinetic energy. Note the double logarithmic scale to make the steep
increase visible. LHS: energy dependence of the cross section for excitation into various
optical levels of helium, according to the formula of Mityureva and Smirnov [48].
RHS: Lotz’s theory for ionization of hydrogen and argon, experimental data for argon
are from one of the most cited papers by Rapp and Englander-Golden [49].

The main difference between ionization and excitation into optical levels
by electron impact is caused by the electrostatic interaction of three charges
after the collision process. For the first time, this problem has been addressed
by Wannier ([46], also cf. [50], Fig. 2.17.1). As a main result, the maximum
for excitation into optical levels is located very close to the threshold, whereas
for ionization, the difference in energy between maximum and threshold will be
no less than a factor of 4 in excess of the ionization energy [48]. The graphs of
Figs. 2.17.1 are plotted applying an approximation containing three adjustable
parameters σ0 , φ, and γ:

 −γ
T − En T
σn = σ0 −1+φ . (2.42)
T En
2.5 Inelastic collisions 25

In summary, we find the cross section for ionization by electronic impact

to rise sharply from zero at a specific threshold by several orders of magnitude
until it reaches a maximum at approximately 100 eV because the de Broglie
wavelength of about 1.2 Å fits the atomic dimensions. For further increases
of electron energies, the cross section gradually decreases. At 500 eV, it has
typically weakened by a factor of about 2 compared with its maximum value
(Fig. 2.18).

Hg

10 air
sion [10-16 cm2]

Ar

1 Ne
Fig. 2.18. Cross section for
H2
ionization by electron impact
He for various gases in double-log-
0.1 arithmic scale [51] (c Oxford
University Press).
10 100 1000 10000
electron energy [eV]

2.5.1.3 Rate constant for ionization. Employing Eqs. (2.31), we ally the
frequency of ionization with the rate constant for this process, pictured in Figs.
2.19 for ionization of argon, with a Maxwellian distribution assumed. In fact,
it is the rate constant we are interested in to calculate the carrier generation.
Especially at the threshold, the slope of the k(E) curve repeats the steep increase
of the cross section. In low temperature plasmas, it is but these electrons in the
high-energy tail that are responsible for the maintenance of the discharge.

2.5.1.4 Electron attachment. Electron attachment plays a significant role

in discharges through electronegative gases. One of the main processes is disso-
ciative electron attachment, e. g.:

O2 + e− −→ O− + O, (2.43)

SF6 + e− −→ SF−
5 + F, (2.44)
a two-body collision leading to the formation of a negatively charged oxygen ion
and and oxygen atom, or a very special case, the formation of a SF− 5 ion and a
fluorine atom.2
2
It is this ion which captures almost all the electrons at very low discharge pressures.
26 2 Collisions and cross sections

3
10

rate constant [10-11 cm3sec-1]

1000
cross section [10-16 cm 2]

cross section [10-16 cm 2]

k
2 Lotz s
Li 1
100

1 0.1
argon argon 10

0 0.01 1
0 250 500 750 1000 10 100 1000
electron energy [eV] electron energy [eV]

Fig. 2.19. Ionization by electron impact for argon. LHS: Two different formulae,
according to Lotz and Li, yield an almost indistinguishable result [44, 52]. RHS:
For cross section and rate constant, a Maxwellian distribution for the electrons
assumed. The calculation is based on the Lotz formula [44].

1.0 Fig. 2.20. Normalized cross

Schulz section of electron attachment
normalized cross section

Craggs et al.
0.8 and formation of O− ions. The
asterisks were evaluated by
0.6
Schulz, the circles are due to
measurements by Craggs et
0.4
al. The maximum (Schulz)
0.2 has been found at 1.25 × 10−18
cm2 and 6.7 eV [53] ( c J. Wi-
0.0 ley & Sons, Inc.).
4 6 8 10 12
electron energy [eV]

Atoms are not able to capture electrons in a two-body collision process

because the laws of conservation of momentum are violated. Therefore, the cross
sections are very low (in the order of 10−18 cm2 ), but exhibit sharp maxima at
electron energies of a few eV (Fig. 2.20). The density of negative ions can far
surpass that of the electrons which changes the characteristics of the discharge
completely.
As is evident from Fig. 2.22, the cross section for dissociative attachment
exhibits two maxima for oxygen. The first can be referred to reaction (2.43),
whereas the maximum at higher energies is due to the reaction

e− + O2 −→ O− + O+ + e− . (2.45)
2.5 Inelastic collisions 27

One of the most prominent cases is that of BCl3 as an additive in discharges

through chlorine. In principle, three reaction paths are possible [54]:

BCl3 + e− −→ BCl∗3 , (2.46.1)

BCl3 + e− −→ BCl2 + Cl− , (2.46.2)

BCl3 + e− −→ BCl + Cl−

2, (2.46.3)
with BCl∗3the metastable parent anion, which can be effectively generated by
extremely low energy electrons up to 0.1 eV [54]. Although the pyramidal BCl− 3
ion is expected to be long lived, (Petrovic et al. detected this anion in a mass
spectrometer, which indicates a lifetime in the microsecond range [55]), due to
its small electron affinity of 0.33 eV, it will be easily discharged by electron
impact. All of the discharges which are dealt with in the following chapters
exhibit higher mean electron energies. Therefore, BCl− 3 anions are most likely
to elude detection.3 Tav et al. have estimated the cross section for dissociative
electron attachment to BCl3 to be ≤ 5 × 10−18 cm2 [according to Eqs. (2.46.2)
and (2.46.3)] and for electron attachment to BCl3 according to Eq. (2.46.1) to
4.5 × 10−15 cm2 .

2.5.1.5 Total collision cross section. Since the numerous possibilites of ex-
citation exclude each other, their probabilities must be added up, and we can
define a total cross section for all possible excitations [57]:

σtot = Pi σi (v), (2.47)
i

with Pi the probability and σi the differential cross section for reaction i. For
example, the maximum of the total cross section for electron impact for argon
amounts to 26 × 10−16 cm2 = 26 Å2 and can be found just above the ionization
threshold of 15.76 eV. This means an electronic mean free path of 1.5 mm at
a discharge pressure of 50 mTorr (7 Pa, n = 2 × 1015 cm−3 ). For low electron
energies, the cross section for elastic scattering almost equals the total scattering
cross section since the velocities of the electrons do not suffice for atomic or
molecular excitation. Especially for the noble gases, numerous measurements
have been performed and modeled. A compilation of these measurements is
pictured in Fig. 2.21.
It was Myers who split up the total cross section σO2 ,tot into its separate
parts for oxygen (Fig. 2.22) [58]. For energies up to approximately 50 eV, the
3
However, Gottscho and Gaede succeeded in generating and observing this anion in
low-frequency discharges at 50 kHz where the electron density will completely relax during
each half cycle leaving ample time for the slow electrons to generate the BCl−
3 anion [56].
Using a non-invasive method (laser photodetachment), they could detect this anion before
the electrons were reheated in the next half cycle.
28 2 Collisions and cross sections

40 Fig. 2.21. Experimental scat-

Xe tering total cross sections of
Kr
the heavy noble gases for
30
low electron velocities. The
s[10-16 cm2]

almost complete transparency

20 Ar at very low velocites is strik-
ing; the distinct maximum cor-
10 responds well with the ioniza-
Ne tion energy Eion [25] (c Ox-
0 ford University Press). Com-
0 2 4 6 8 10 pare with Fig. 2.7.2.
electron velocity [V1/2]

cross section for elastic scattering is the most important contribution to the
cross section for ionization.

Fig. 2.22. The single contribu-

10-15 A tions of the total cross section
for electronic impact were split
G
C up by Myers for O2 [58]. The
B letters denote the contributions
10-17 B
as follows: A: elastic scattering,
s [cm2]

F
C B: rotatory excitation, C: oscil-
E latory excitation, D: excitation
10-19 E
to singlet-oxygen, E: dissocia-
D D
tive attachment (two maxima!),
C G F: excitation to upper electronic
10-21 states, G: ionization.
0.1 1 10
E [eV]

2.5.2 Inelastic collisions between heavy particles

To begin with, the literature covering the low-energy range, which is of paramount
importance in glow discharges, is widely scattered, in particular when we com-
pare this range with the countless number of papers dealing with energies ex-
ceeding 1 keV. This is mainly caused by the low cross sections for inelastic
processes. Hence, elastic scattering will still dominate the interaction between
the gas constituents. The main reactions between heavy particles are:
• Generation of electrons by ionic impact [so-called β-ionization, Fig. 2.23,
Eq. (2.48)].
2.5 Inelastic collisions 29

• Excitation to electronic levels [Eq. (2.49)].

• Electron stripping [Eq. (2.50)].

• Simple charge transfer [Figs. 2.25 − 2.28, Eq. (2.51)].

• Double charge transfer [Eq. (2.52)].

+
Ar in Ar

H2+ in H 2
10
Ne+ in Ne
sion [10-16 cm2]

Fig. 2.23. β-ionization: cross

He+ in He section of ionization by ions
1 of the parent gas after [59].
The values are not corrected
by ionizations caused by sec-
ondary electrons ( c Oxford
0.1 2 University Press).
10 103 104
ion energy [eV]

−
2 + B −→ A2 + B + e ,
A+ + +
(2.48)

A+,∗ + B −→ A+ + B∗ , (2.49)

A+ + B −→ A2+ + B + e− , (2.50)

B+ + A −→ A+ + B, (2.51)

A+ + B −→ A− + B2+ . (2.52)
As we can see from Fig. 2.23, the threshold for ionization by ionic impact is
relatively low but amounts to about twice the ionization potential. The classical
theory of J.J. Thomson yields cross sections equal in size for equal velocities
of ions and electrons. The same result is obtained with quantum mechanics
with the restriction that the impact of the colliding particles is large compared
with their pair potential (Born’s approximation). Choosing the abscissa scale
skilfully (in fact, it is the momentum of the colliding particle), we can find a
similar behavior between the ionic and the electronic cross section, however,
dilated by some orders of magnitude for the heavy particles (Fig. 2.24). The
30 2 Collisions and cross sections

electron energy [eV]

100 1000
sion [10-16 cm2]

Hooper et al., 1962

Smith, 1930
Fig. 2.24. Comparison of the
1 Tozer, Craggs; 1930 ionization cross sections of ar-
Bleakney, 1930 gon for protons (dashed) and
Lampe et al., 1957
electrons (solid) [60] (
c J. Wi-
ley & Sons, Inc.).
0.1 1
proton energy [MeV]

cross sections for ionization (β-ionization) do not reach hogh values until the
ions have been accelerated to velocities which are comparable to those of the
electrons for α-ionization. Hence, for energies up to several hundreds of eV, the
ionic cross section is smaller by about two orders of magnitude. For the same
kinetic energy, the ionic momentum is considerably greater, which leads to very
short de-Broglie wavelengths. Hence, even for inelastic collisions, there are
hardly any diffraction effects.

2.5.2.1 Charge transfer. For charge transfer, an almost complete exchange

of momentum is of importance. Consider a rapidly moving ion colliding with a
slow molecule (Fig. 2.25). This collision can result in a slowly moving ion and
a rapid neutral, and the components of velocity are conserved. In the case that
both colliding particles have the same mass, this process is very efficient, and
this collision is denoted (symmetric or resonant charge transfer, Fig. 2.26). We
expect the cross section for this process to be comparable with that for elastic
collision.
To begin with, we assume A = B, i. e. the ionization potentials should be
differ by the amount ΔEion :

Eion,B = Eion,A + ΔEion . (2.53)

Massey discussed this charge transfer by means of the correspondence principle
[62].
For a particle velocity v smaller than that of the bounded electrons, they will
adapt to this disturbance without an electron transfer (so-called adabatic approxi-
mation). We find the time of collision τ to scale inversely with the frequency ν of an
electron oscillation with τ ≈ b/v (b the so-called adiabatic parameter):
2.5 Inelastic collisions 31

15

12 H+, Kr
s [10-16 cm2]

9 H +, Xe

6
Fig. 2.25. Cross section for
3 the asymmetric charge trans-
fer of protons in noble gases
0 after [61].
0 5 10 15 20 25
V1/2

a 1 ΔEion aΔEion
= ∧ν =τ= =⇒ v = . (2.54)
v ν h h
When the velocity of the particles exceeds this threshold, no adiabatic collision will
take place but an ionization will occur. The larger the difference in ionization potential
ΔEion , the higher the energetic threshold to become a competitive process. On the
other hand, the time for interaction declines with rising energy which eventually
reduces the probability of ionization: A maximum has been evolved (Fig. 2.26).

100
Fig. 2.26. Principal energy
dependence of the two mech-
s [10-16 cm2]

anisms for charge transfer.

in double-logarithmic scale,
10 the resonant charge trans-
fer linearly scales with en-
ergy, whereas the asymmetric
symmetric charge transfer
asymmetric charge transfer
charge transfer exhibits an en-
1 ergetic maximum.
1 10 100 1000
E [eV]

2.5.2.2 Resonant charge transfer. For a resonant charge transfer with

ΔEion = 0, we expect a monotonous increase in the cross section for ionization
for lower energies in the adiabatic approximation. Albeit the correspondence
principle predicts this conduct qualitatively correctly, more extensive descrip-
tions of the height of the maximum (e. g. ΔEion /Ekin ) are definitely impossible.
32 2 Collisions and cross sections

The principal shape of the energetic dependence of the cross section is de-
picted in Fig. 2.26. It can be described by the formula

σ = (a − b ln v)2 (2.55)
with a, b empirical constants and v the relative velocity of the colliding particles
[63]. For a resonant reaction, there exists a linear relation between the cross
section and the logarithmic kinetic energy (Fig. 2.27).

Firsow formula
60 Gilbody, Hasted (1956)
Flaks, Solovev (1958)
Gustafsson, Lindholm (1960)
sCT [10-16 cm2]

40 Fig. 2.27. Resonant charge

transfer: Ar+fast + Arslow −→
Ar+slow + Arfast ; comparison of
20
experimental data with theory
(Firsow formula) [64] ( c J.
0 Wiley & Sons, Inc.).
0 20 40 60 80
v [eV1/2]

The cross section reaches a maximum for an ion in its parent gas at slow
velocities. In this energy range, it can deliver the main contribution to the total
cross section (Fig. 2.28). The electron transfer mostly happens without any
exchange of momentum: The generated ions exhibit negligible velocities, and
the scattered atoms alter their direction only in a very narrow scale; hence, the
scattering cone exhibits a very low opening angle [65].
The measuring of the charge transfer was made possible by an apparatus
which consists of a magnetic field which deflects the ion beam and focuses it on
to the sample. To prevent multiple ionizations, the cross section of the reactor
has to be made sufficiently thin. The ions produced are sucked off by an electric
field which is directed perpendicular to the reactor. The ions are measured by
a Faraday cage which is negatively biased to avoid distortions by secondary
electrons (so-called condensor method) [66].
Ternary collisions occur relatively rarely; the ratio of the probabilities of a
two-body collision over a three-body collision is about thousand [67], therefore
reactions

2A+ + B −→ A+
2 +B
+
(2.56)
(recombination) are not very likely to happen. Since the number of collisions
scales with the squared pressure p (at constant temperature: With the squared
number density n), the probability for recombination increases with rising pres-
sure (cf. Sect. 4.7).
2.5 Inelastic collisions 33

50 50
s (total) s (total)
s (resonant charge transfer) s (resonant charge transfer)
s (elastic collision) s (elastic collision)
s [10-16 cm2]

s [10-16 cm2]
helium neon
25 25

0 0
0 5 10 15 20 0 5 10 15 20
v [eV1/2] v [eV1/2]

75 s (total)
s (resonant charge transfer)
s (elastic collision)
s [10-16 cm2]

argon
50

25

0
0 5 10 15 20
v [eV1/2]

Fig. 2.28. Cross sections for the processes of elastic scattering and symmetric (reso-
nant) charge transfer of He+ in He, Ne+ in Ne, Ar+ in Ar [31] − [33]. At low energies,
the resonant charge transfer can deliver the main contribution to the total cross sec-
tion.

2.5.2.3 Penning ionization. Among the numerous reactions which can occur
between heavy particles, we should mention the Penning ionization, collisions
of second order. This mechanism becomes important in discharges of reactive
gases which are doped with noble gases (cf. Sects. 10.4.3, 10.6.4 and 11.8.6) and
consists of ionization by metastable species.
For ionization by collisional impact, the metastable level must be higher
than the ionization potential of the neutral molecule or atom. Since the return
to the ground level from a metastable level is forbidden by selection rules, these
levels exhibit considerable lifetimes. The metastable levels for neon (16.62 and
16.7 eV) and those of argon (11.55 and 11.72 eV) are higher than the ionization
potentials of most of the gaseous elements and gaseous compounds. By dop-
ing a discharge of argon with neon, metastable neon atoms can ionize argon
34 2 Collisions and cross sections

atoms. The state of metastability is denoted with an asterisk (cf. also Fig. 4.4,
enlargement of α).

Ne∗ + Ar −→ Ne + Ar+ + e− . (2.57)

The cross sections at thermal energies are somewhat larger than the gas
kinetic cross sections (≈ 10−15 cm2 ), therefore, its probability approaches unity,
and this process becomes dominant very quickly [68, 69] and will be manifested
in a rise of the discharge current [70].
Mierdel discussed consecutive staircase processes which cause an addi-
tional ionization after a collision of two metastables [71]; this effect would be-
come more probable with rising density of the metastable atoms.

Ar∗ + Ar∗ −→ Ar+ + Ar + e− . (2.58)

Furthermore, ionization of metastables can happen by electron impact:

Ar∗ + e− −→ Ar+ + 2 e− . (2.59)

Since the energy difference from the metastable level to ionization has shrunk
from 15.76 eV to only 4.21 eV, considerably more electrons are able to succeed
in an ionization. Although the density of the metastable species is very low,
Ingold has estimated this mechanism to be the most important ionization
source in discharges of Hg/Ar [72].

2.5.3 Collisions between photons and molecules

As final collision processes between individual particles, we must consider im-

pacts of photons (Fig. 2.29). Again, the threshold of ionization has to be ex-
ceeded. But, in contrast to the impacts of charged particles, we observe for
photons the occurrence of a steep absorption edge of several orders of mag-
nitude at E = Eion to decline precipitously beyond this sharp maximum. For
example, the maximum of the photoelectric excitation is found for argon at
15.5 eV and exhibits a value of about 0.36 × 10−16 cm2 .
Hence, the maxima for the yield of ionization for electrons, ions and photons
are located at very different values.

2.6 Generation of secondary electrons at surfaces

2.6.1 Electrons

The impact of electrons incident on a surface causes the following reactions:

• Elastic scattering, the energy does not change but the momentum, very
high contribution in the spectrum.
2.6 Generation of secondary electrons at surfaces 35

40
M2 M3

30 Ar
s [10-18 cm 2]

20 M1

Fig. 2.29. Cross section

10 Ne for photo absorptions of
L1 K L2 L
3 noble gases [73] (
c Sprin-
He ger-Verlag).
0
20 40 60 80
l [nm]

• Inelastic scattering, both energy and momentum are changed, small con-
tribution in the spectrum.
• Towards low energies, the second rise peaking at energies between 5 and
10 eV is caused by secondary electrons (Fig. 2.30).

1.00

0.75

Fig. 2.30. Energy distribu-

f(E)

0.50 tion of secondary electrons,

f (E)se , for 160 eV electrons
0.25
incident on a shiny gold sur-
face (normalized to the low-en-
ergy maximum) [74] ( c Ox-
0.00 ford University Press).
0 50 100 150
Ese [eV]

The yield of secondary electrons depends sensitively on

• The energy of the electrons incident on the surface, which causes a max-
imum at medium energies. At low energies, the electrons are rather ab-
sorbed, at higher energies, they are scattered elastically.
• The chemical composition of the topmost layer: It rises with increasing
density and work function with the well-known periodic dependencies.
The yield increases with growing fill-up of the electronic shells. When the
filling of a new shell has begun before an inner shell has been completely
filled, we observed discontinuities, in particular for half-occupied shells.
36 2 Collisions and cross sections

• The quality of the surface: the shinier, the higher the yield.

For many metals with a clean surface the yield for secondary electrons, δ, equals
unity (Fig. 2.31). For insulating layers, the values for δ can reach 15 (to measure
δ, momentum methods are applied, however, the errors are rather large).

2.0

1.5 Ag

W
Cu
1.0 Mo
d

Fig. 2.31. Coefficients for sec-

0.5 ondary emission, δ, for several
metals as a function of the en-
Be
ergy of the primary electrons
0.0 incident on the surface.
0 500 1000 1500
electron energy [eV]

In insulating materials, the conduction band lacks electrons almost entirely

whereas the valence band is nearly filled. Hence, the primary electrons can
generate secondary electrons only by interaction with valence band electrons.
But these electrons gain a higher amount of kinetic energy than comparable
metal electrons, which enhances their probability of reaching the surface and
eventually leaving the substrate. For the same reason, the yield increases with
flatter angles of incidence.
To differentiate between the electrons, the coefficient for backscattering η
refers to those electrons with energy between 50 eV and the energy of the
electrons incident on the surface, and the coefficient for secondary electrons δ
refers to those electrons having energy between 0 and 50 eV.

2.6.2 Ions

Ions incident on the surface can release electrons as well, and this mechanism
is of considerable importance in DC discharges and also plays a role in RF dis-
charges which are capacitively coupled. The energy distribution of the secondary
electrons responds relatively flatly to the kinetic energy of the incident ions and
exhibits a broad maximum between 5 and 10 eV [75] (Fig. 2.32). The distribu-
tions for Ei ≤ 200 eV are very similar; γi rises steeply for growing energies, and
a maximum becomes visible at very low electron energies. Eventually, for the
highest ion energies, the energy distribution of the secondaries peaks close to
zero.
2.6 Generation of secondary electrons at surfaces 37

30 1000
Fig. 2.32. Energy distribution
10 E(He+) [eV] of electrons which are released
25 40 10
600
40 from a surface being atomi-
100
200 100 cally cleaned by He+ ions with
20 200
600 various kinetic energies. The
n(Ei) [10-3]

1000 dashed lines without symbols

15
separate the Auger processes

Eion - 2 Wa
1000
10 from the other mechanisms
which release secondary elec-
5 He+ Mo 10 trons [75] (c The American
Physical Society).
0
0 5 10 15 20
electron energy [eV]

An electron can be released only when the sum of the kinetic energy of the
ion incident on the surface and the ionization potential exceeds twice the value
of the work function Wa : For every electron emitted from the surface, a second
one is required to neutralize the incident ion.

Ekin (e− ) ≥ Eion − 2 Wa . (2.60)

This condition is remarkably violated by neon which can be attributed to an

additional Auger process [76], and which is not observed when the electron
is released by Ne ions of 10 eV kinetic energy. This strange phenomenon is
explained by the discharging of ions of higher energy in close vicinity to the
surface. Afterwards, these atoms can generate rapidly moving electrons in a
process denoted Auger relaxation. The yield of this process, however, exceeds
that of the Auger neutralization.
Typical values for the yield for secondary electrons, γ, vary between 0.05
and 0.1 for the heavy noble gases incident on molybdenum and tungsten (Fig.
2.33). The distinct energetic minimum of the number of secondary electrons for
He+ is absolutely striking. This is repeated for the heavier noble gas, however,
to a weaker extent, which is due to a larger penetration depth with rising kinetic
energy, and the electrons which are now released are attenuated during their
course to the surface. This reduction in secondary electrons is overcompensated
by the growth of the yield at further growing kinetic energies. For neon, this
effect is covered by the above mentioned Auger relaxation.4 Although the
curves of yield are very similar, the electron yield is

• always higher for molybdenum than for tungsten,

4
This is valid for ion energies which are typical for glow discharges. Positive ions moving
very rapidly (α-particles with energies of 1 MeV) can release up to thirty electrons; their
energy amounts then to some keV.
38 2 Collisions and cross sections

0.30
Ne+

0.25 He+
gi [electron/ion]

0.20 W
Mo
Fig. 2.33. Comparison of the
0.15 yield for secondary electrons,
Ar+
γi , for the ions of the heavy
0.10 noble gases incident on pure
Kr+ molybdenum and tungsten
0.05
Xe
+ [75, 77] ( c The American
0.00 Physical Society).
0 250 500 750 1000
ion energy [eV]

• and depends on the ionization energy of the incident ion: The lower its
ionization potential the less the yield of secondary electrons.

This is due to the low work function of molybdenum—for constant energy

of the incident ions, the energy of the secondary electrons rises and enhances
the probability of being released. On the other hand, Fermi energy and density
of states at the Fermi level (this equals the number of conduction electrons
divided by the Fermi energy in the free-electron model) energy and the number
of released electrons: A falling Fermi energy causes a larger kinetic energy of
the emitted electrons, simultaneously, the density of state goes up.
Furthermore, Hagstrum investigated the influence of the surface quality
on the yield for secondary electrons [78]. Fig. 2.34 impressively illustrates the
difficulties obtaining reliable data for δ and γ (γ as anisotropic property addi-
tionally depends on the crystal orientation).

0.10
W
+
0.08 Ar

N2/W
Ar+
gi [electron/ion]

0.06 W Fig. 2.34. γi as a function of

+
Kr kinetic energy of the heavy no-
0.04 ble gas ions for pure tungsten
Kr+ N2/W and tungsten covered with a
0.02 W mono layer of nitrogen [78] (
c
Xe+
N2/W
The American Physical Soci-
0.00 ety).
0 250 500 750 1000
ion energy [eV]
2.6 Generation of secondary electrons at surfaces 39

But also the energy distribution is altered by the contaminated surface: The
fraction of rapid electrons is reduced. This is not only caused by a change in
the work function but also by fundamental alterations of electron generation
mechanisms with the exception of Auger processes.

2.6.3 Photons

The electron yield caused by photons, triggered by the Einsteinian photo effect,
oscillates between 100 and 1000 ppm in the near UV (NUV) (dependent on the
work function Wa ) rising to values of up to 0.5 in the vacuum UV (VUV) (Figs.
2.35 and 2.36).

E [eV]
0
2015 10 5 Fig. 2.35. Yield for photoelec-
10
trons γhν of gold and plat-
Pt inum as functions of wave-
length (lower abscissa) and en-
_ ergy (upper abscissa). γhν ad-
_

-3
ditionally depends on the po-
10 potassium
g hn

Au
on platinum larization of the incoming light
[79]. ⊥ and : E of linearly
_
_

polarized light is orientated

in perpendicular or in parallel
fashion, respectively, with re-
10-6 spect to the surface (
c Oxford
100 200 300 400 500 600
l [nm] University Press).

E [eV]
30 25 20 15
0,20
Fig. 2.36. The yield for
0,15
photoelectrons depends sen-
sitively on the pretreatment
of the surface [80]: squares:
0,10 untreated tungsten electrode;
ghn

triangles: 5 min heated at

T > 1 000 ◦ C and 10 Torr;
0,05 circles: heated up to repro-
ducibility at T > 1 000 ◦ C (
c
0,00
J. Wiley & Sons, Inc.).
40 60 80 100
l [nm]
40 2 Collisions and cross sections

Two main mechanisms have been identified which contribute to the steep
increase in yield: First, most of the radiation in the long-wavelength range is
reflected, and second, due to conservation of momentum, the excitation of free
electrons in the valence band by a two-electron process seems very unlikely
to happen. Going to short wavelengths, the reflectivities of the metals begin
to decline to very low values (with simultaneous rise in transparency), and it
becomes more likely to excite valence electrons.
All these observations add experimental proof to the suggestion that this
process consists of two steps which both belong to the activation type with a cer-
tain threshold, generating the electrons as well as their escaping from the solid.
Measuring the work function and the various coefficients γ, δ and η remains
a great challenge still, since the yield depends strongly on the purity (volume
property), but even more sensitively on the contamination (surface property).
Theoretically, the main problem is the exact calculation of the barrier which
has to be overcome (model potentials etc., cf. [81]).
Although the cross sections for photoionization are extremely small at low
energies, this does not mean that they are without significance in glow dis-
charges. In DC discharges, the charged carriers exhibit kinetic energies of sev-
eral thousands of eV, therefore, photons of high energy are generated by Auger
processes.
http://www.springer.com/978-3-540-85848-5