You are on page 1of 95

1

Lecture notes for Atomic Physics in Fusion


Henric Bergsker, November 29
th
, 2011


Introduction

The following lecture notes define the structure and content of the course in Atomic Physics in
Fusion, which since 2009 is included in the the nuclear fusion masters program at KTH. When I
was first asked to set up this course, it was intended to substitute a more general and basic atomic
physics course with something that is more directly relevant to fusion plasma physics. I quickly
realized that there is no textbook or set of a few textbooks available that covers the subject of the
course. The first year the course was given, I handed out a fairly large number of original papers
and excerpts from literature, largely because I was worried that the course content wouldnt
otherwise be covered thoroughly enough or presented well enough. Experience and feedback
from the students have shown that large volumes of text were not very helpful and that the rather
condensed and short lecture notes on the other hand are nearly sufficient. In the 2011 version of
the course only a couple of complementary texts will be recommended. There is very little
original material in these notes, they are mostly based on and condensed from the listed literature
and if anyone is interested in seeing any particular topic discussed in more depth or in different
style I will be happy hand out more comprehensive literature excerpts (within copyright limits ).

It was clear from the beginning that the students in the masters program have variable
background, both in micro physics and in plasma physics. The intention therefore is that the
course should not strictly require any significant prior training in either quantum mechanics or
plasma physics. A solid background in engineering mathematics, classical mechanics and basic
electromagnetic theory is necessary, however. On the other hand, the course will cover enough
ground to go through the important basic physics and deal with some fusion relevant applications.
Consequently the first lectures cover wave mechanics pretty much from scratch, the structure of
the hydrogen atom and, more superficially, the theoretical framework for many electron atoms.
The following lectures move quickly over to atomic collisions, of prime importance in fusion
plasma physics but frequently left out in a first course in atomic physics. Having gone through
the basic concepts of atomic collisions, applications in various directions that are of interest in
fusion plasma physics are explored, with examples related to recent research activities. Basic
plasma physics concepts, such as plasma resistivity and collisional transport are also covered,
lightly but from scratch. This way I hope that the course is accessible, but still interesting and
challenging to follow, also for someone who already has a good background in either basic
atomic physics or plasma physics. For those who are already familiar with basic quantum
mechanics and/or plasma physics, the introductory sections on these subjects may be used for
repetition.

Atomic physics in fusion applications is a huge field, especially if the wide range of plasma
diagnostics is taken into account. The content selected for this course emphasizes the aspects of
atomic collisions that are essential for the performance of a fusion plasma device. This obviously
includes processes in the bulk plasma, at the plasma edge and at material plasma facing surfaces,
since plasma-surface interactions are now largely and increasingly in focus in the world wide
2
efforts to achieve controlled fusion as a viable energy source. Examples of important areas that
receive less attention in this course are plasma diagnostics techniques involving different kinds of
spectroscopy, diagnostics instrumentation and surface physics. Molecular processes are
completely left out, as well as laser physics and related diagnostics.

The intended learning outcomes of the course are that having completed the course the student is
expected to be able to

1) Recall the sizes of key atomic quantities and the definitions of basic concepts in atomic
physics.
2) Apply methods of wave mechanics, classical mechanics and electrodynamics to problems
in atomic physics, in particular to atomic collisions.
3) Explain the significance of different approximations that can be made when estimating
atomic quantities.
4) Predict the general behavior of atomic physics quantities that are relevant in fusion
plasma physics, based on dominating physical mechanisms.
5) Construct simple numerical and analytical models of fusion plasma phenomena with input
from atomic physics databases.
6) Present and communicate atomic physics related material in a clear and efficient way,
both as written reports and orally.
They can be described as skills and can only be acquired by actively doing these things. The
theory presented in the lectures and in the lecture notes should provide the support to stand on,
but the most important to make progress is to work on problems and exercise the listed skills.
Most importantly the students in this course are expected to work on a number of assignment
tasks, individually and in groups, to present their results in writing and orally and to take active
part in discussing the problems and solutions. Each lecture section in the notes also includes a
number of exercise problems, which are intended to illustrate the theory.
3


Contents

Lecture 1 Review of wave mechanics 4
Lecture 2 Spherical coordinates and the hydrogen atom 11
Lecture 3 Perturbation methods amd many-particle methods 18
Lecture 4 Kinematics in atomic collisions. Cross sections
and rate coefficients 23
Lecture 5 Elastic collisions and the Rutherford cross section 31
Lecture 6 The partial wave expansion for the scattering problem 38
Lecture 7 Bremsstrahlung and effective Z 43
Lecture 8 Plasma resistivity 49
Lecture 9 Collisions and particle transport 54
Lecture 10 Interatomic potentials and the Thomas-Fermi model
of the atom 60
Lecture 11 Ion implantation and backscattering 66
Lecture 12 Sputtering
Lecture 13 Collisional ionization 74
Lecture 14 Excitations, inelastic scattering 81
Lecture 15 Equilibria and power balance 84
Lecture 16 Charge exchange and neutral transport 90

References 94

4

Lecture notes for Atomic Physics in Fusion



Lesson 1. Review of wave mechanics.

Both waves and particles are familiar objects from everyday experience. Waves are characterized
by properties like wavelength, frequency, amplitude and phase velocity, in many cases also
polarization. Phenomena such as interference and diffraction are observed where waves with
different phase are added. In spatially restricted regions standing waves are formed and give rise
to discrete frequency spectra. Particles are characterized by properties like position, mass,
velocity and momentum.

Vast experience has shown that microscopic objects exhibit both wave- and particle properties (in
the everyday sense), depending on experimental conditions. Thus, a particle with momentum
p mv = and energy E can be associated with a wave of wavelength / h p = and frequency
/ E h v = , where
34
6.6261 10 h

= Js is Plancks constant. Conversely it is useful to be able to
discuss a prima facie wave, such as an electromagnetic wave or an elastic wave as consisting of
particles (quanta: photons and phonons respectively) with momentum / p h = and energy
E hv = . Alternatively these relations are expressed in terms of angular frequency 2 e tv = , wave
number 2 / k t = and =h/2. For reference:
p k = (1.1)

E e = (1.2)

More specifically it has turned out that a particle can be conveniently described by means of a
complex wave function ( , ) r t , such that the probability of finding the particle in any volume
element
3
d r is given by


3 2 3
( ) P particle in d r d r =, , , (1.3)

provided that the wave function is normalized,
2 3
| | 1 d r =
}
over the relevant region.
For a given particle state expectation values of coordinates or functions of coordinates are
calculated as usual with probability distributions. For instance, if the wave function depends only
on x, the expectation value for the particle position is


2
| | x x dx x dx

-

< >= =
} }
(1.4)


The evolution of the wave function is determined by the time dependent Schrdinger equation:
5
i H
t


c
=
c
(1.5)
where H is an operator representing the total energy of the particle. H is called the Hamiltonian
of the system. The correct expression for the Hamiltonian for a given problem can often be
arrived at from classical, macroscopic physics. If H is independent of time, separation of
variables by looking for solutions of the form ( , ) ( ) ( )
E
r t r t = then gives ( )
i t
t Ne
e


= and
the time independent Schrdinger equation:

E E
H E = (1.6)
with constant energy E. Solutions
E
to (1.6) represent stationary states, with specific energy.
General solutions can be written as linear combinations of such solutions.
For a particle with mass m, moving in a potential ( ) V r , the Hamiltonian operator is


2
2
( )
2
H V r
m
= V + . (1.7)

For a free particle, i.e. with V=0, the plane wave

( )
( , )
i k r t
r t Ae
e


= (1.8)
is a solution of (1.7), provided that =
2
k
2
/2m. This expression has the disadvantage that it can
not be normalized to fit (1.3). If normalization is required, the easiest trick is to introduce
periodic boundary conditions, ( , , , ) ( , , , ) x y z t x L y L z L t = + + + , and let L . Since the
differential equation (1.5) is linear, any linear combination of plane waves is also a solution to the
equation. For example, if we consider only plane wave solutions moving in the x direction (the
wave function is independent of y and z), any superposition


( )
1
( , ) ( )
2
i k t ikx
x t k dk e e
e

t

=
}
(1.9)
where (k) = k
2
/2m is a possible solution to (1.7) with V=0. As we can see more clearly if we
consider the situation at t=0, the amplitude function ( ) k specifies the wavelength- (or
momentum- ) distribution in a propagating wave pulse:


1
( , 0) ( )
2
ikx
x k e dk
t

=
}
(1.10)
in the sense of a Fourier transform. The momentum distribution can be recovered from
( , 0) x by the inverse transform [1]:

1
( ) ( , 0)
2
ikx
k x dx e
t

=
}
(1.11)
Either of the functions ( , 0) x or ( ) k contains all information about the particle. They are
alternative representations of the same system, a coordinate representation and a momentum
representation, respectively.

6
For the interpretation (1.3) to make sense, a decreasing probability of finding a particle in one
volume must be exactly balanced by an increase in the probability of finding it somewhere else.
Indeed, the continuity equation

*
( ) s
t

c
= V
c
(1.12)
can be derived from (1.5), with the so called probability current

* * *
1
( ) Re( )
2
i
s p
m m
= V V = (1.13)
being a measure of the flow in the direction of p .

Equation (1.6) is an eigenvalue equation. Depending on the boundary conditions the equation
may either have solutions for a continuum of different energies E, or only for discrete values of
E. All measurable quantities can be represented by operators and the corresponding eigenvalue
equation gives the possible results in a measurement of the quantity. Moreover, if Q is the
operator representing some observable quantity, then for any two normalizable funcions
1
and
2



3 * 3
1 2 1 2
( ) Q d r Q d r
-
=
} }
, (1.14)

where integration is over the appropriate region, with appropriate boundary conditions. This
important property is called hermiticity (the operator Q is said to be hermitian, or self adjoint
over the region). Hermiticity entails a number of useful properties, such that the eigenvalue
equation

Q q = (1.15)

only has real eigenvalues q and that eigenfunctions
1
and
2
corresponding to different
eigenvalues
1
q and
2
q are orthogonal, in the sense that

3
1 2
0 d r
-
=
}
.
(1.16)


If there are two or more eigenfunctions corresponding to the same eigenvalue (degeneracy) it is
always possible to write them as linear combinations of a set of orthogonal eigenfunctions. The
wave functions of physically distinguishable states are orthogonal.

Given a wave function , the expectation value of any observable quantity that is represented by
an operator Q is calculated as:


3
Q Q d r
-
< >=
}
(1.17)

7
By operating first with B and then with A the combined operator AB is formed. It is often
convenient to work with an operator algebra, without necessarily writing out the functions they
operate on. One particularly useful shorthand notation is the commutator of operators A and B ,
defined as
, A B AB BA
(
=

(1.18)
The operators are said to commute if , 0 A B
(
=

.
If a set of hermitean operators , , .... A B C commute with each other, admissible wave functions
can be expanded in a set of functions that are eigenfunctions of all the operators , , .... A B C . If
the operators commute, the corresponding physical quantities can in principle be measured and
known simultaneously. The time development of an expectation value can be expressed as
[ , ]
d i Q
Q H Q
dt t
c
( ) = ( + )
c
(1.19)
Thus, if an operator commutes with H and does not depend explicitly on time, the expectation
value is constant, such that the corresponding physical quantity is a constant of the motion.

Very important quantum mechanical operators in atomic physics besides H are those associated
with linear and angular momentum. In view of (1.1) the expectation value
x
p for a particle
moving in the x-direction can be calculated simply as

2
*
( ) ( ) ( )
x
p k k dk k k k dk


= =
} }
, (1.20)
such that the momentum operator in the momentum representation is simply k . It follows from
(1.10) and (1.11) that in the coordinate representation the momentum operator in cartesian
coordinates is

x
p i
x
c
=
c
(1.21)
and similar for y and z: p i = V. Analogously the operator corresponding to position x in the
momentum representation is simply

x
x i
k
c
=
c
(1.22)
The operators x and
x
p do not commute:
[ , ]
x
x p i = . (1.23)
Hence admissible wave functions can not be eigenfunctions of both x and
x
p , which reflects the
fundamental physical fact that the exact position and momentum of a particle can not both be
determined simultaneously in any experiment (an example of Heisenbergs indeterminacy
relations). The solutions (1.8) for example are eigenfunctions of
x
p , with eigenvalues
x
k , but
the position is completely unknown.

8
The operator corresponding to the z-component of the angular momentum is z
y x
L x p y p = .
The other components of L follow by cyclic permutation. If the potential V is symmetric with
respect to rotations, i.e. a central field of force ( ) ( ) V r V r = , both
2
L and any component z L
commute with H , and
2
L commutes with any component z L , but different components of
L dont commute with each other:[ , ] x y z L L i L = , [ , ] y z x L L i L = , [ , ] z x y L L i L = . Thus, if V is
spherically symmetric the wave function can be expanded in common eigenfunctions of
2
L , z L and H . Experimentally the total energy, the magnitude of the angular momentum and one
component of the angular momentum are in principle possible to measure simultaneously.

A system with many particles is represented by a wave function
1 2
( , ,..., ) r r t with the
coordinates of each particle as arguments and the Hamiltonian likewise is a function of the
coordinates of every particle. Solutions to a given problem are usually built up from product
functions
1 2 1 1 2 2
( , ,...) ( ) ( )..... r r r r = . For particles with spin, factors representing the spin of
each particle also need to be included. The full wave function of a system of identical fermions
(particles with spin or odd multiples thereof) has to be antisymmetric with respect to
interchange of particle labels. A consequence of this is that no two particles can be in the same
state, including spin. This is called the Pauli principle.

Basics of wave mechanics are given e.g. in refs [2-5], where chapter 9 in [2] is most elementary,
[3,4] more complete and higher level, [5] at still slightly higher level, although all the texts start
from the beginning. Quantum mechanics can also be introduced in a more abstract way by
leaning more heavily on the theory of linear vector spaces, c.f. [6] or appendix A of [7].


Problems

1.1 Show that in the case of V=0, the wave (1.8) satisfies the time dependent Schrdinger
equation (1.5) and that the spatial part satisfies (1.6). What is the phase velocity and the
direction of propagation of the wave in (1.8)? What are the energy and linear momentum
and how do they correspond to classical mechanics?
1.2 What is the probability current associated with (the spatial part of ) (1.8)?
1.3 Show that Hermiticity (1.14) of an operator Q implies that the eigenvalues q are real.
Show that eigenfunctions belonging to different eigenvalues
1 2
q q = are orthogonal.
1.4 Show that the operators H (with a real potential V) and p are hermitean, with
appropriate boundary conditions for the integrations.
1.5 Derive (1.19) for a hermitean operator Q using (1.5) and (1.17). The potential V is
assumed to be a real function.
1.6 Suppose that a moving particle is described by a wave packet that is shifting position with
time. Derive Ehrenfests first theorem

x
d
m x p
dt
< >=< >
9
and Ehrenfests second theorem
x
d V
p
dt x
c
< >=
c
.
The proofs are easiest using operator algebra for commutators.
The two theorems establish a firm link between wave mechanics and classical mechanics
and can often justify the use of classical approximation.

1.7 Suppose that a flux of electrons with energy E are propagating in the positive
x-direction and is represented by the wave function
1
in
ik x
Ae = ,
1
2 / k mE = .
At x=0 the particles encounter a potential step of height V
0
:

0
0, 0
( )
, 0
{
x
V x
V x
<
=
>

Consider the case when E>V
0
. The wave function should be continuous at x=0. What
other boundary condition should there be (try taking the integral of (1.6) over a short
interval [ , ] c c and let 0 c )? Determine R and T, where R is the ratio of the reflected
probability current to the incoming probability current and T is the ratio of the transmitted
probability current to the incoming probability current. Show that R+T=1. Do R and T
change if V
0
is negative?

1.8 Solve the problem in 1.7 for the case E<V
0
.
1.9 Consider the case of a particle trapped in an infinitely deep potential well: V=0 in the
interval 0 < x < a, V for x<0 and x>a. What are the appropriate boundary conditions
at x=0 and x=a? What are the energies of the possible particle states and what do the wave
functions with lowest energy look like?
1.10 Consider the limiting case of an infinitely deep and narrow potential well,
( ) ( ) V x x |o = , where the Dirac delta function by definition has the property
( ) ( ) ( ) f x x a dx f a o

=
}
for any continuous function f(x). What should the boundary
conditions on ( ) x be in this case (compare with exercise 1.7)? Solve the Schrdinger
equation for bound states (E<0). How many bound states are there? Determine the
normalized wave functions and binding energies for the bound state(s).
1.11 For the same potential as in exercise 1.9, solve the problem for E>0, i.e. the scattering
problem. Determine the reflection and transmission coefficients R and T, defined
analogously as in 1.6. Do R and T change if the potential changes sign?
1.12 Determine the transmission of a beam of particles incident on a potential barrier of limited
width: V(x) = 0 for x<-a, V(x)=V
0
for a < x < a, V(x)=0 for x>a, in the case where V
0
>
E. In this situation, clearly there would be no transmission at all predicted by classical
mechanics.

1.13 Already in 1914 the Swedish physicist A. kesson, when working on his PhD thesis in
Heidelberg, found that the scattering cross section for electrons penetrating into some
gases decreased at very low electron energies, in a way that appeared inexplicable by
classical collision models [8]. The experiments were criticized however, and not generally
accepted. The issue was revived when in 1920 Carl Ramsauer studied the elastic
10
scattering of electrons impacting noble gas atoms. The measured reflection probabilities
are shown in the figure [9].

The scattering probability drops to nearly zero at
impact energies just below 1 eV. That energy was
known to be far below any possible excitation energy
of the atoms. It is hard to imagine how this may come
about within the framework of classical collisions
between hard spheres or similar. If the electrons are
associated with waves however, zeroes in the reflection
coefficient can be understood qualitatively (how?).
As a first model of this scattering problem, note that
an electron penetrating a noble gas atom would be
exposed to electrostatic attraction from the partly
unscreened nuclear charge. This could be modeled in
one dimension by a potential well of depth V
0
:

0
0,
( ) ,
0,
{
x a
V x V a x a
a x
<
= < <
<

E>0 and there are incident particles only from one
direction. Solve the one dimensional Schrdinger equation for this situation, by ansatz in the
three different regions and fitting of coefficients at the two boundary positions x a = . Calculate
the reflection and transmission coefficients and plot them as functions of energy. Discuss how the
calculated R compares with the Ramsauer scattering probabilities.

References

[1] Anders Vretblad, Fourier Analysis and its Applications, Springer Verlag
New York 2003, chapter 7, or similar literature.
[2] Hermann Haken and Hans Christian Wolf, Atomic and Quantum Physics,
Springer Verlag, New York 1984.
[3] David Park, Introduction to Quantum Theory, 3
rd
ed. McGraw-Hill 1992.
[4] Stephen Gasiorowicz, Quantum Physics, 3rd ed. John Wiley Sons 2003.
[5] Eugene Merzbacher, Quantum Mechanics, John Wiley & Sons 1970.
[6] P.A.M. Dirac, The Principles of Quantum Mechanics, 4
th
ed. Clarendon Press 1951,
latest printing 2009.
[7] Charles J. Joachain, Quantum Collision Theory, Elsevier Science,
3rd ed. Amsterdam 1983.
[8] Max Jammer, The conceptual development of quantum mechanics, McGraw-Hill
1966, and references therein.
[9] Carl Ramsauer, Annalen der Physik 66(1921)545. Reproduced in R. B. Brode,
Reviews of Modern Physics 5(1933)257. The Ramsauer effect or Ramsauer-
Townsend effect was independently discovered also by J.S. Townsend.


11
Lesson 2. Spherical coordinates and the hydrogen atom.

The hydrogen atom consists of a positively charged nucleus and a negatively charged electron.
More generally we consider any ion with nuclear charge +Ze and a single electron with charge
e. In the first approximation, ignoring spin, relativistic effects and taking the electron mass =m
e

to be negligible compared to the nuclear mass, the time independent Schrdinger equation for the
electron moving in the electrostatic field of an immobile nucleus is

2 2
( )
2
H V r E

V
= + = (2.1)

with the Coulomb potential

2
0
( )
4
Ze
V r
r tc
= (2.2)
As will be discussed in lesson 4, if the electron mass is not neglected compared to the nuclear
mass, the reduced mass should be used in equation (2.1).
Since the potential is spherically symmetric it can be shown that all solutions to (2.1) can be
written as a superposition of simultaneous eigenfunctions of H ,
2
L and z L . Classically this is
not surprising: the angular momentum should be a constant of the motion. Classically
( ) ( ) ( ) x y z
z y x z y x
L r p yp zp e zp xp e xp yp e = = + + .
Likewise in quantum mechanics z
y x
L x p y p = and so on, while
2 2 2 2
x y z L L L L = + + .
In spherical coordinates
z L i

c
=
c
(2.3)
and

2
2
2 2
2 2
1 1
sin
sin sin
L u
u u u u
c c c
=
c c c
. (2.4)
The eigenfunctions to z L and
2
L are the so called spherical harmonics


2 1 ( )!
( , ) (cos )
4 ( )!
m
m
im
m
Y P
m
e

u u
t
+
=
+
(2.5)
where ( )
m
P z are polynomials called the associated Legendre functions. The spherical harmonics
( , )
m
Y u are also eigenfunctions to the Laplace operator
2
V in spherical coordinates, so if the
potential V(r) is independent of angles ( , )
m
Y u are also eigenfunctions of H , as required. The
eigenvalues are given by

( , ) ( , ), , 1....., z
m m
L Y m Y m u u = = + (2.6)
and

2
2
( , ) ( 1) ( , ), 0,1, 2.....
m m
L Y Y u u = + = (2.7)

12
The eigenfunctions of H and hence the stationary states of a particle moving in a spherically
symmetric potential V(r) are obtained by separation of variables in the Schrdinger equation:
( , , ) ( ) ( , )
n m n m
r R r Y u u = , (2.8)
where 1, 2, 3.... n = and 0,1, 2... 1 n = . The differential equation for the radial function ( )
n
R r is

2 2
2
2 2
1 ( 1)
( ) ( ) ( ) ( )
2 2
n n
d d
r V r R r ER r
r dr dr r
( +
+ + =
(

(2.9)
In the case of the hydrogen atom, with V(r) given by (2.2), there are analytical solutions to this
equation. With the boundary conditions ( ) 0
nl
rR r , both for 0 r and r , solutions exist
only for eigenvalues
n
E E = ,

2 4 2
2 2 2 2
0
1
2 (4 )
n
Z e Z
E Ry
n n

tc
= = (2.10)
where n is an integer n=1,2,3 and 0,1, 2... 1 n = . The numerical value of the energy unit Ry
(Rydberg) is about 13.6058 eV. Clearly the ground state of the hydrogen atom (Z=1), the state
with lowest energy, has a binding energy of 1 Ry, the ionization energy of hydrogen. The integer
n is called the principal quantum number. There are infinitely many excited bound states, with
n>1. Transitions from higher bound states
2
n to lower states
1
n give rise to the emission of
electromagnetic radiation with energy

2
2 2
1 2
1 1
E Z Ry
n n
| |
A =
|
\ .
(2.11)
Spectral lines with
1
n =1 and
2
n = 2,3,4 are called the Lyman series. Spectral lines with
1
n =2
and
2
n = 3,4,5 are called the Balmer series
1
, which was investigated experimentally already in
the 19
th
century. The line corresponding to Z=1,
1
n =2 and
2
n = 3 is often called the H
o
line and
is frequently used for diagnostics in fusion plasma experiments. The wavelength of the H
o
line
is 6563 . The integer is called the angular momentum quantum number. By convention letters
are often used instead of numbers, states with 0,1, 2, 3... = are called s,p,d,f.. states. The
integer m is called the magnetic quantum number.


In this approximation, the binding energies of the hydrogen atom depend only on the principal
quantum number. If the effects of electron spin and relativistic mass are taken into account, there
is a slight shift in binding energies depending on and the relative orientation of the electron
spin and the angular momentum, the so called fine structure. In the presence of external electric
or magnetic fields, the energies also depend on the magnetic quantum number m.


1
The wavelengths of the first four hydrogen spectral lines were measured by Anders ngstrm before 1868. Johann
Jakob Balmer was a school teacher in Basel. He published the Balmer formula (expressed in wavelengths) in 1885.
Reportedly Balmer was a devotee of numerology, interested in such things as the number of the beast or the number
of the steps of the pyramids. One day, chatting with a friend who happened to be a physicist or chemist, Balmer
complained that he had run out of things to do. Whereupon the friend replied: well, you are interested in numbers,
why dont you see what you can make of this set of numbers that come from the spectrum of hydrogen? [8] Janne
Rydberg claimed in 1890 that he had been using the Balmer formula long before its publication and later generalized
it [8].
13
The solutions to the radial equation may be expressed in terms of so called associated Laguerre
polynomials, multiplied by an exponentially dropping function. Qualitatively the radial functions
for higher n peak further out from the nucleus, so that the electrostatic binding energy decreases
with n. For atoms other than the one-electron atoms, the radial equation has to be solved
numerically.

The ground state radial function is easy to find by guessing, since 1 0 n = = the second term
in eq. (2.9) is zero and the wave function must be spherically symmetric. Trying a solution of the
form
0
10
/
( )
r a
R r Ne

= in (2.9) gives

2
0
0 2
4
a
Ze
tc

= (2.12)
The characteristic size
0
a of the hydrogen atom (Z=1) in the ground state is called the Bohr
radius and is found to be
( )
2 2
0
4 / e tc =0.529177 . We also find
( )
2 2
0
/ 2 E a = . What is the
appropriate normalization constant N ?

We have seen how by separating variables in Cartesian coordinates it is easy to find the plane
wave solutions to the Schrdinger equation in case V=0. Obviously the equation can be solved
also in spherical coordinates, with the solutions in terms of simultaneous eigenfunctions of H ,
2
L and z L . Just like the plane wave solutions these eigenfunctions constitute a complete basis in
terms of which all solutions to the Schrdinger equations can be expressed. In particular the plain
wave functions can be expanded in spherical eigenfunctions. If we consider a plane wave
propagating in the positive z-direction, the wave function can be expressed as

cos 0
0
(2 1) ( ) (cos )
ikz ikr
e e i j kr P
u
u

=
= = +

(2.13)
where
0
P P = are also called simply Legendre functions and

1/ 2
( ) ( )
2
j kr J kr
kr
t
+
= (2.14)
are spherical Bessel functions of the first kind. ( ) J z
v
are ordinary Bessel functions of the first
kind. The first j function is

0
sin
( )
z
j z
z
= (2.15)
and subsequent ( ) j z can be found by recursion formulae like [10]:

1 1
2 1
( ) ( ) ( ) j z j z j z
z
+
+
= (2.16)
or

1 sin
( ) ( 1)
d z
j z z
z dz z
| |
=
|
\ .
(2.17)

In literature the associated Legendre functions and spherical harmonics sometimes come with
different sign, phase and normalization conventions, the conventions used here correspond to refs
[1-3]. If you want to play around with these functions numerically, the associated Legendre
14
functions for 0 m> are obtained from the MATLAB function legendre. For negative m, use
m
P
in equation (2.5) and change the sign in front to
( ) / 2
( 1)
m m
. The ordinary Bessel functions of the
first kind ( ) J z
v
are obtained with the MATLAB function besselj. Figures 2.1-2.4 give an
impression of how the partial wave expansion (2.12) works out. The physics of the hydrogen
atom is covered in chapter 10 of [2] , sections 7.3 and 14 of [3] and in chapters 2-3 of [11].





Figure 2.1. A plane wave propagating
in the positive z-direction is
represented by the wave function
ikz
e .
This wave function is easily obtained
by separation of variables in the
Schrdinger equation with Cartesian
coordinates. The plane wave is an
eigenfunction of the linear momentum
operator p , as of the free particle H .
The plot shows Re( )
ikz
e .

Figure 2.2. The solutions to the free particle
Schrdinger equation can also be expressed as linear
combinations of eigenfunctions to H ,
2
L and z L in
spherical coordinates. The figure shows the real part
of the first term in the expansion (2.13). The s terms
( 0) = are spherically symmetric.














15


Figure 2.3. The real part of the sum of the
first three terms in (2.13) is shown. Only
even terms contribute to the real part.










F igure 2.4. The real part of the sum of the first 16
terms in (2.13) is shown.
Only even terms contribute to the real part. The sum
starts to resemble the plane wave in figure 2.1.








Problems

2.1 Where in space is it most likely to find the electron in a hydrogen atom in the ground state?
What is the most probable distance between the electron and the proton? What is the
expectation value r for the radial position of the electron in a hydrogen atom in the ground
state?
2.2 What is the expectation value of the potential energy in the hydrogen atom in the ground
state? In classical mechanics it follows from the virial theorem that for central motion in a
potential ( )
n
V r r the kinetic energy T and the potential energy V are related by / 2 T nV = .
How does that compare with the quantum mechanical description of the hydrogen atom in the
ground state?
2.3 What is the characteristic velocity of the electron in a one electron atom in the ground state?
2.4 We have found the ground state
100
of the hydrogen atom. Since 1 n = in the ground state
0 = and 0 m= and the wave function is spherically symmetric. Look for another
spherically symmetric state whose wave function is of the form (1 )
r
cr e

and use equation
16
(2.9) to determine c , and the energy of the state. Normalize the solution. Is it appropriate to
call this state
200
?
2.5 Investigate the radial functions in the hydrogen problem numerically. Firstly,write the radial
wave function as ( ) ( ) /
nl nl
R r P r r = . Measure the energy in units of
2 2
0
/(2 ) Ry a = and
distance in units of
0
a . Show that equation (2.9) then cleans up to

2
2 2
( 1)
( ) ( ) ( )
n n
d
V r P r EP r
dr r
( +
+ + =
(

(2.18)
The appropriate boundary conditions for
n
P are (0) 0
n
P = and lim ( ) 0
n
r
P r

= . It can also be
convenient to introduce the effective potential

2
( 1)
( ) ( )
eff
V r V r
r
+
= +
so that the radial differential equation becomes

( ) ( ) ( )
n eff n
P r V r E P r '' ( =


Now integrate this equation numerically, by treating it as an initial value problem, starting
near 0 r = . To get started properly, verify analytically that for small , r > 0
1
( )
n
P r r
+
~
(disregarding normalization). Use this to start the integration from a small r, say 0.01 r =
with a suitable slope. Use some MATLAB routine like ODE45 or ODE113, or any other
suitable numerical scheme to step outwards, but do not proceed beyond about twice the
classically allowed region, since the solution will sooner or later blow up due to numerical
instabilities. The classically allowed region is defined by ( ) 0
eff
V r E < . Try first with some
allowed values of
2
1/ E n = and 0,1... 1 n = for small integers n. Do the solutions behave
correctly at large r ? Then try to change E slightly (e.g. 0.23 E = instead of 0.25 E = )
and see what happens. By trying out a few small n and , try and figure out how the number
of zeroes (nodes) and maxima (antinodes) of
2
( )
n
P r depend on n and . In a many-
electron atom the electrons are screening each other from the nuclear charge but can be
thought of as moving in a resulting modified potential. In that situation there is no exact
solution to the radial equation like in the hydrogen atom, but numerical solutions can still be
found and the n can be identified by counting the nodes and antinodes.
2.6 When 0 V = the wave equation is
2 2
( ) 0 k V + = , which is called the (homogenous)
Helmholtz equation. In spherical coordinates, after separation of variables the radial equation
is

2 2
2 2
1 ( 1)
( ) 0
d d
r k R r
r dr dr r
+ ( | |
+ =
| (
\ .
(2.19)
The spherical Bessel functions ( ) j z in eqs. (2.14) (2.17) provide solutions
( ) ( ) R r j kr = that satisfy the boundary condition ( ) 0 krj kr = . Using eq. (2.16) or (2.17), find
the explicit expressions for ( ) j z for the first few . Show that these examples satisfy the
generally valid asymptotic behaviour:
17

sin( )
2
( ) ,
kr
j kr r
kr
t

(2.20)


References

[10] Milton Abramovitz and Irene Stegun eds. Handbook of Mathematical Functions, 9
th

printing 1972, Dover Publications.
[11] Robert D. Cowan, The theory of atomic structure and spectra, University of California
Press 1981






























18
Lesson 3. Perturbation methods and many-particle systems.

Exact analytic solutions in quantum mechanics are usually not possible for realistic
problems. A common approximate approach is to use a known analytic solution to a simplified
problem, which is close to the actual one and regard the difference between the actual problem
and the simplified one as a small perturbation. Say that the solutions
(0)
n
and eigenvalues
(0)
n
E
are known for the problem

(0)
(0) (0) (0)
n n n
H E = . (3.1)
Suppose that the Hamiltonian for the actual problem is

(0) (1)
H H H = + (3.2)
where the perturbation
(1)
H is small in some sense. It then feels plausible that the solutions
n
and eigenvalues
n
E of the actual problem

n n n
H E = (3.3)
can be approximated with
(0) (1)
n n n
= + and
(0) (1)
n n n
E E E = + , where
(1)
n
and
(1)
n
E are relatively
small corrections. Indeed, in a first order perturbation calculation for non degenerate states
(0)
a
the energy shift due to the perturbation
(1)
H is

(1) (1)
(1) (0)* (0) 3
a a a
E a H a H d r =
}
(3.4)
i.e. the expectation value of the perturbation, evaluated in the unperturbed state. Here we are
using the common, shorthand bracket notation for the integral.
It can be shown that the correction
(1)
a
can be chosen orthogonal to
(0)
a
. Then, the perturbation
can be expanded in the unperturbed states

(1) (0)
a n n
n a
c
=
=

(3.5)
where

(1)
(0) (0) m
m a
m H a
c m a
E E
= =

(3.6)
A second correction to the energy eigenvalue can be calculated from

2
(1)
(2)
(0) (0) a
m a
m a
m H a
E
E E
=
=

(3.7)
so that
(0) (1) (2)
a a a a
E E E E ~ + + .

Similarly, in a non stationary situation, where as a result of a perturbation a transition can take
place between any of the stationary states of the system, the transition rate can be estimated by a
perturbation technique. It can be shown that if
a
and
b
are eigenstates belonging to the
unperturbed problem and the perturbing term in the Hamiltonian is
(1)
H (possibly time dependent)
and the system is initially in the state
a
, the time derivative of the probability of finding the
system shortly afterwards in state
b
is
19

2
(1)
2
( )
b b a
P b H a E E
t
o = (3.8)
where we are once again using the bracket shorthand
(1) (1)
3
b a
b H a H d r
-
=
}
. Frequently a
transition to a single state can not be detected. Rather what is measured is transitions into a
bundle of states around
b
. One can measure for example the number of particles scattered into
the solid angle dO subtended by a detector. Let us by b denote such a bundle of states. Let us
define the density of states ( )
b
E such that the number of final states in the energy range
| | ,
b b b
E E dE + is ( )
b b
dN E dE = . The rate of transitions is then given by

2
(1)
2
( )
a b
P b H a E
t
= (3.9)
This formula is called Fermis golden rule, or just the golden rule.

For systems with more than one particle involved, such as for many-electron atoms, the
state of a system is completely characterized by a wave function
1 2
( , ,... , )
N
r r r t +
involving the position coordinates of every particle. The wave function is determined by
the Schrdinger equation, with a Hamiltonian H that depends on the position coordinates
of all particles and on derivatives of these coordinates. E.g. in a multi-electron atom, if
the potential energy is assumed to be given by the electrostatic interaction between the
electrons and the nucleus and between the electrons (the electron spin is disregarded), the
Hamiltonian can be written:

2 2 2 2
1 1
0
0
2 4
4
N N
i
i i i j
i
i j
Ze e
H
m r
r r
tc
tc = = >
V
= +


(3.10)
Here
2
i
V is the Laplace operator with respect to the position coordinates of particle i and the first
term represents the kinetic energy. The final electron-electron interactions term is a sum over all
electron pairs. If it werent for the electron-electron interaction and the Pauli exclusion principle
the problem could easily be separated so that the total wave function could be given as a product


1 2 1 1 2 2
( , ,... ) ( ) ( )....... ( )
N N N
r r r r r r + = (3.11)

or a linear combination of such product functions. For numerical calculations the electron-
electron interactions may be regarded as a perturbation of the stronger interaction with the
nucleus and hence the solution to the many particle problem may be expanded in product
functions like (3.11), where angular parts of the single particle wave functions are taken as those
of the known single particle wave functions of the hydrogen atom. At this point, let us also
introduce the electron spin, albeit without going into any detail. Suffice it to say that electrons
have an inherent magnetic moment, related to a spin. The spin is described by a spin quantum
number s=1/2 (analogous to the orbital angular momentum quantum number ) and that the
orientation of the spin is specified by a spin magnetic quantum number
s
m (analogous to the
orbital magnetic quantum number m), which can assume any of the two numbers 1/ 2 .
Mathematically the spin of an electron can be represented as a spin factor ( )
s s z
m z m s
s o o , i.e.
1/ 2
o
20
is zero unless the spin is oriented in the z-direction ( 1/ 2
z
s = + ), in which case
1/ 2
o =1 and
likewise
1/ 2
o

is zero unless 1/ 2
z
s = . For brevity then, the single particle functions in (3.11)
can be specified as ( ) ( ) ( , ) ( )
i i i i s z
i
i i n i m i i m i
r R r Y s u o = . These ( )
n
R r are generally different from
the radial functions of the hydrogen atom and have to be calculated numerically. This step
involves solving the Schrdinger equation loosely speaking of each electron moving in the
average field from the nucleus and the other electrons. There are a number of different iterative
and variational schemes to do this, such as the Hartree-Fock method.

In fact we know that the total wave function must be antisymmetric with respect to interchange of
particles, no two electrons can be in exactly the same state, including spin. A convenient way to
take this into account is to define antisymmetrized combinations of product functions, called
Slater determinants

1 2 3
1 2 3
1
( 1) ( ) ( ) ( )..... ( )
!
N
p
j j j N j
P
r r r r
N
+ =

(3.12)
where the summation is meant to be over all possible permutations P of the N particle labels
1 2
....
N
j j j and terms with odd permutations have negative sign. With this definition 0 + = for
any combination where two single particle states are identical. Further linear combinations of this
kind of functions may be built up, which meet specific requirements on total orbital angular
momentum, total electron spin, total angular momentum of the atom etc, but still with the single
electron state product functions (3.11) as building blocks. The result is in any event a set of
many-electron basis functions
b
+ , which are orthogonal to each other, normalized and constitute
a complete basis for the possible many electron wave functions. Now the generally applicable
numerical method to calculate the possible energies and the corresponding many-electron wave
functions of the atom consists in approximating the solution to the Schrdinger equation


k k k
H E + = + (3.13)
with a truncated series

1
M
k k
b b
b
y
=
+ = +

(3.14)
Since the basis functions are orthonormal, multiplying with the conjugate of one basis function
*
' b
+ and integrating over all 3N spatial coordinates and summing over both possible spin
directions for each of the N spins one gets the matrix equation

' '
1
M
k k k
b b b b
b
H y E y
=
=

(3.15)

where the matrix elements
*
' ' b b b b
H H dt = + +
}
. The matrix equation can be diagonalized
numerically to find the eigenvalues
k
E and the corresponding approximate eigenfunctions.

If the total many electron wave function involves
1
w single electron wave functions with
quantum numbers
1
n and
1
,
2
w single electron wave functions with quantum numbers
2
n and
2
etc, it is said to be in the (electron) configuration
21

1 2
1 1 2 2
( ) ( ) .....
w w
n n
where the usual letters are used for . For example the configuration
2 3
1 2 2 s s p involves two 1s
electrons, one 2s electron and three 2p electrons.

For many atoms and ions the orbital angular momenta of the single electron functions can be
thought of as adding up vectorially to a total orbital angular momentum L of the atom, while the
spins add up to a total spin S . Finally L and S can add vectorially to make a total angular
momentumJ . This is called LS-coupling. In spectroscopic notation the atomic levels are then
called
2 1 S
J
L
+
. Thus
2
1/ 2
S is called a doublet S one half level, and so on. Perturbation techniques
are discussed in refs. [3-5] and to some extent in [2]. Many-electron atoms and the Hartree-Fock
method are thoroughly covered in ref. [11].

Problems

3.1 Derive the first order energy shift in equation (3.4). Argue along the following lines:
Suppose that
(0)
(0) (0) (0)
a a a
H E = has been solved, so that the energy eigenvalues
(0)
a
E and the
corresponding eigenvalues are known. Say that the problem to be solved is
( )
(0)
(1)
a a a
H H E + = , where is an arbitrary real smallness parameter and
(1)
H is the
perturbation. Moreover, assume that the exact energies and wave functions can be expanded
in powers of :
( )
0
( )
0
n n
a a
n
n n
a a
n
E E

=
=
=



Insert the expansions (or at least the first few terms) in the problem equation and sort the
terms in orders of . This gives an equation for the coefficient of each
n
. The equation for
0 n = should give just the unperturbed, solved equation again. The equation for 1 n = ,
multiplied in appropriate way with
(0)
a

-
and integrated gives
(1)
a
E . Dont forget the
hermiticity property of
(0)
H .

3.2 Considering a perturbation
0
( ) / V x V x a = to the the infinitely deep potential well
in problem 1.8. Calculate the first order energy shifts for the energies of the eigenstates.

3.3 Calculate the density of states ( ) E for free particles with kinetic energy E. Consider
first particles which are confined in a (large) cubic box with volume
3
V L = . Impose
appropriate boundary conditions for the free particle wave functions: either periodic
boundary conditions or that the wave functions vanish at the walls. Imagine the particles
filling up a space ( , , )
x y z
k k k and determine the density of allowed states in this k space, i.e.
the number of allowed states in a volume
3
d k in k space. Finally convert this result to the
22
energy density of states, i.e. the number ( ) E of allowed states in the energy interval
[ , ] E E dE + .

3.4 Following up on problem 3.1, show that once the first order energy shift
(1)
a
E has
been determined, the first order correction to the wave function can be arrived at in principle
by solving the differential equation:

( ) ( )
(0) (1)
(0) (1) (1) (0)
a a a a
H E H E = (3.16)

3.5 Provided that
(1)
a
can be chosen as perpendicular to all the unperturbed eigenstates

(0)
n
, show that if the perturbation of the wave function is expanded in the
unperturbed states:

(1) (0)
a n n
n a
c
=
=

(3.17)
the coefficients are

(1)
(0) (0) n
n a
n H a
c
E E
=

(3.18)
3.6 Apply (3.16) or (3.18) to the perturbation in problem 3.2 to see how the wave
functions are modified due to the perturbation. Make the necessary calculations numerically
or analytically.

3.7 Consider two electrons with equal spin direction, trapped in the infinitely deep
potential well of problem 1.8. If the Coulomb interaction between the electrons is
neglected, what is the ground state of the two particle system?

3.8 The ground state of the Helium atom can be analyzed with the perturbation
technique, if the electron-electron interaction is introduced as a perturbation of the situation
where the two electrons move independently of each other in the field from the nucleus, like
in two superimposed hydrogen atoms. What are
(0)
H ,
(0)
E and
(0)
in this case? The
perturbing term in the Hamiltonian should be

2
(1)
0 1 2
4
e
H
r r tc
=


Write down the integral expression for calculating
(1)
E as in equation (3.4). For the
evaluation of the integral, separate the problem into first calculating the electrostatic potential
1
( ) V r at radius
1
r due to the charge distribution of electron number 2, then integrate over
1
r
get the energy shift due to the overlap of the two charge distributions.






23
Lesson 4. Kinematics in atomic collisions. Cross sections and rate coefficients.

4.1. Types of collisions.

Atomic collisions can be classified as elastic, inelastic or reactions. In an elastic collision
A B A B + +
the collision partners A and B remain unaltered by the collision, they only exchange momentum
and kinetic energy. In an inelastic collision

*
A B A B + = +
immediately after the collision at least one of the collision partners, say A, has changed its
internal state. Typically the atom A was in its ground state before the collision and emerges in an
excited state
*
A after the collision. In this case
*
A may later relax to the ground state, e.g. by
emission of radiation. A reaction is a collision
A B C D + +
or
.... A B C D E + + +
where reaction products, which are different from the original collision partners emerge after the
collision. The outcome of a collision is largely determined by the conservation of momentum and
energy, the kinematics, regardless of the detailed dynamics of the interaction.

4.2. The centre of mass coordinate system.

Consider two particles with masses
1
m and
2
m , located at
1
r and
2
r , which interact. In classical
mechanics the equations of motion for the two particles are


1 12 1 1
F F mr + = (4.1)


2 21 2 2
F F m r + = (4.2)
where
1
F and
2
F are any external forces acting on particle 1 and 2 respectively, while
12 21
F F =
are the forces with which the particles act on each other. The mass centre of the two particles is
defined as

1 1 2 2
1 2
mr m r
R
m m
+
=
+
(4.3)
It follows immediately that
1 2
F F MR + = , where
1 2
M m m = + . In other words, the mass centre
moves as a result of external forces in the same way as a single particle with mass M would.
With
2 1
r r r = it follows from (4.1) and (4.2) that


1 2 2 1 1 2 21 1 2
( ) mF m F m m F mm r + + = (4.4)

Now, if either
1 2
0 F F = = or
1
F and
2
F are parallel and
1 2 1 2
/ / F F m m = (as in a homogenous
gravitational field) it follows that
24

1 2
21
1 2
mm
F r r
m m
= =
+
(4.5)
where is called the reduced mass. It is also straightforward to show that the kinetic energy

2 2 2 2
1 1 2 2
1 1 1 1
2 2 2 2
E mr m r MR r = + = + (4.6)
In conclusion, the motion of two interacting particles can be separated into the motion of the
centre of mass and the relative motion, both obeying Newtons laws for a single particle.

In quantum mechanics the two particle system would be described by a wave function
1 2
( , ) r r such that the probability of finding particle 1 in a volume
3
1
d r centered at
1
r and particle
2 in a volume
3
2
d r at
2
r is
2
3 3
1 2 1 2
( , ) r r d r d r . This wave function should be a solution to
H E = . If there are no external forces acting on the particles, only an interaction between
them in the form of a potential ( ) V r , then

2 2 2 2
1 2
1 2
( )
2 2
H V r
m m
V V
= + (4.7)
where as before
2 1
r r r = and
2
1
V and
2
2
V denote differentiation with respects to the position
coordinates of particle 1 and 2. Through a somewhat tedious but straightforward series of partial
differentiations, e.g. in Cartesian coordinates beginning with


1
1 1 1
( , ) m r R x X
x x x X x x M X
c c c c c c c
= + = +
c c c c c c c


it is possible to transform the kinetic energy terms in (4.7) so that the Schrdinger equation can
be written

2 2 2 2
( ) ( , ) 0
2 2
r R
V r E r R
M

| | V V
+ =
|
\ .
(4.8)
where the Laplace operators operate on r and R instead of
1
r and
2
r . This new equation can
easily be separated after writing ( , ) ( ) ( ) r R r R = + into


2
2
1
( ) ( ) ( )
2
r
V r r E r

| |
V + =
|
\ .



2
2
2
( ) ( )
2
R
R E R
M
V + = +
with
1 2
E E E = + . Thus, in quantum mechanics as well as in classical mechanics, the two particle
problem can be reduced to a single particle moving in a potential and the system as a whole in
uniform motion.

25
Since the dynamics in a binary collision is much simplified in this way, any detailed calculations
are always performed using a coordinate system that is moving as the centre of mass, the centre
of mass system, CM-system for short.

Let us first have a look at how to transform scattering- and recoil angles between the lab
coordinate system and the CM system. Suppose that the CM system is moving in the z-direction
with a constant velocity
CM CM z
R V V e = = . It is usually most convenient to express directions in
the usual spherical coordinates, with
L
u as the polar angle (angle to the z-axis) in the lab system
and u in the CM-system. Figure 4.1 shows how to convert a velocity v with respect to the CM-
system to
L
v in the lab system.


Vector addition
L CM
v V v = + gives:
sin sin
L L
v v u u =
cos cos
L L CM
v V v u u = +
whence

sin
tan
cos
L
u
u
t u
=
+
(4.9)
or

2
cos
cos
2 cos 1
L
t u
u
t t u
+
=
+ +
(4.10)
where /
CM
V v t = . Note how the relation between
L
u and u changes qualitatively depending on
the value of the parametert . As expected
L
u u when 0 t . For 1 t < ,
L
u increases
continuously from 0 to t as u increases continuously from 0 to t . For 1 t = , / 2
L
u u = and is
never larger than / 2 t . For 1 t > ,
L
u reaches a maximum of arcsin(1/ ) t and for smaller
L
u the
same
L
u appears for two different u .

Let us now specialize to the case of an elastic collision between a particle with mass
1
m and
initial velocity
1
v and a stationary particle with mass
2
m . We also assume that the forces between
the colliding particles are only along the line connecting the particles, which entails among other
26
things that the trajectories of both particles

will remain in a single plane. The collision is depicted in terms of classical trajectories in the lab
system in figure 4.2 and in the CM-system in figure 4.3. In the lab system the purely kinematic
relations between the speeds and directions after the collision,
1
v' ,
2
v' ,
L
u and
L
| can all be
calculated from conservation of energy and momentum. Already at this stage however, the
description is much simpler in the CM-system. In the CM-system particle 1 is perceived as
moving initially with speed
1 CM
v V , while particle 2 is moving at the speed
CM
V . The total
momentum is always zero in the CM-system, so
1 1 2
( )
CM CM
m v V mV =

1
1 1
1 2 2
CM
m
V v v
m m m

= =
+
(4.11)
The initial speed of particle 1 in the CM-system is

2
1 1 1
1 2 1
CM
m
v V v v
m m m

= =
+
(4.12)

As there are no external forces acting on the system,
CM
V is constant throughout the whole
collision event. The velocities of the two particles change during the collision, in the region
where they are so close that the potential energy is not negligible. After the collision however,
when the particles are far away from each other again and the potential energy is zero, since the
collision was elastic, the total kinetic energy is once
27
again the same as before the collision and by equation (4.6) not only
CM
V but also the relative
speed of the two particles, in our case
1
v , remains the same. In conclusion, in the CM-system
both particles travel away after the collision at the same speeds at which they approached to
begin with.


























Figure 4.1 and equations (4.9) and (4.10) apply directly to the scattering angle of particle 1 if we
insert
1 CM
v v V = . From (4.11) and (4.12) then
1 2
/ m m t = . For the recoil angle 1 t = and it
follows from figure 4.1 or equation (4.10) that / 2
L
| | = . From figure 4.3 we also see that
| t u = , so alternatively the lab recoil angle can be calculated as ( ) / 2
L
| t u = . In conclusion,
for elastic scattering the conversion of scattering and recoil angles between the lab system and the
CM-system is given by

1 2
sin
tan
/ cos
L
m m
u
u
u
=
+
(4.13)
and
/ 2 ( ) / 2
L
| | t u = = (4.14)

We are going to be considering scattering where the interaction between the collision partners
can be described simply with an interaction potential V(r), which depends only on the distance
28
between the colliding particles. As we have seen, this scattering problem is equivalent to a single
particle with reduced mass being scattered by the potential V(r) in the CM-system.

4.3. Cross sections.

The central concept in the dynamics of atomic collisions is the cross section. Suppose that
something can happen to an atom as a result of particle bombardment. It might be for instance
elastic scattering, excitation from one bound electronic state to another, ionization or a nuclear
reaction. The bombarding particles could be e.g. electrons, ions, neutral atoms or photons. The
cross section for the event is defined as the ratio of the average number of events per unit time
and target atom to the incident flux density of particles. Thus if I is the number of incident
particles per unit area and unit time and
A
N is the number of target atoms exposed to the flux,
then the number of events per unit time is

A
dN
IN
dt
o = , (4.15)
where o is the cross section. We are frequently interested in scattering events where the incident
particle is deflected as a result of a collision. If o in equation (4.15) is referring to scattering into
any solid angle sin d d d u u O= in the CM-system, it is usually called the integral cross section.
If we are interested in the scattering probability as a function of u we have to define a differential
cross section, usually denoted either / d d o O or ( ) o u .The integral cross section is then
sin 2 sin
d d
d d d
d d
o o
o u u t u u = =
O O
} }
(4.16)
where / d d o O has been taken as independent of the azimuthal angle .

To see how the differential cross section is transformed between the CM-system and the lab
system, let the solid angle 2 sin d d t u u O= in the CM-system correspond to
2 sin
L L L
d d t u u O = in the lab system. The solid angles could for instance represent a physical
detector. Clearly the number of scattering events into the detector must be independent of the
coordinate system that is used, so the differential cross section has to be transformed so that

L
L
d d
d d
d d
o o | |
O = O
|
O O
\ .
(4.17)

By differentiating equation (4.10) we get


2 3/ 2
cos ( 2 cos 1)
cos 1 cos
L L
d d d d
d d d d
o u o t t u o
u t u
+ + | |
= =
|
O O + O
\ .
. (4.18)
Recall that for an elastic collision with
1
m hitting a stationary
2
m ,
1 2
/ m m t = .

4.4. Rate coefficients.

If a target is bombarded with particles with a distribution of velocities, the cross sections have to
be averaged over the distribution of relative velocities. If the targets are once again stationary and
have a density
2
n , the incident flux density of particles 1 is
1 1
n v and the number of events per
29
unit volume and time is
1 2 1
n n v o , where o may be a function of the velocity
1
v . If the velocity
distribution is
1 1
( ) f v , i.e. the probability of a particle having a velocity within
1 1 1 x y z
dv dv dv at
1 1 1 1
( , , )
x y z
v v v v = is
1 1 1 1 1
( )
x y z
f v dv dv dv , then the number of events per unit time and unit volume is

1 2 1 2 1
( )
x y z
n n v n n v f v dv dv dv o o =
}}}
, (4.19)
where v o is called the rate coefficient or rate constant for the reaction. For instance, if the
distribution is Maxwellian:

1
2
3/ 2
1
1
2
( )
2
mv
kT
m
f v
kT
e
t

| |
=
|
\ .
(4.20)
and switching to spherical coordinates
2
4
x y z
dv dv dv v dv t = we get

2
1
3/ 2
3 1 2
4 ( )
2
mv
kT
m
v v v dv
kT
e o t o
t

| |
=
|
\ .
}
(4.21)
More commonly the cross section is given as a function of energy, in which case the substitution
2
1
/ 2 E mv = gives

3/ 2
1
2
1 0
8
( )
2
E
kT
m
v E E dE
m kT
e
t
o o
t


| |
=
|
\ .
}
(4.22)
The condition with an (almost) stationary target is appropriate e.g. in the case of electrons
colliding with atoms or ions, in view of the disproportionate masses. However, in case of
collisions between particles with comparable masses the averaging should actually be over
relative velocities
1 2
v v :


3 3
1 2 1 2 1 1 2 2 1 2
( ( ) ( ) v v v v v f v f v d v d v o o = )
}}
(4.23)

In case both
1
f and
2
f are Maxwellian with the same temperature T, some manipulations lead to

2
3/ 2
3
2
4 ( )
2
v
kT
v v v dv
kT
e

o t o
t

| |
=
|
\ .
}
(4.24)
where not too surprisingly (?) the reduced mass has taken the place of
1
m in (4.21).
If the cross section is given as a function of energy for particles 1 bombarding stationary particles
2 then

1
3/ 2
2
1 0
8
( )
2
E
mkT
v E E dE
m kT
e

t
o o
t


| |
=
|
\ .
}
(4.25)

4.5 Collision time and mean free path

In approximate discussions it is often very useful with concepts like the typical time interval
between two successive collisions a particle experiences, or the average distance a particle travels
between collisions. Depending on precisely how the averaging is defined, these quantities come
out slightly differently. We will use the collision time defined as
30

1
n v
t
o
= (4.26)
where n is the density of possible collision partners and the rate coefficient is averaged over
relative velocities. The mean free path for particles of type 1 we take as

1
v t = (4.27)
where
1
v is the average speed of particles 1.

The classical scattering in a spherically symmetric potential is well and clearly covered by
chapter 3 of ref. [12]. Kinematics and cross sections are treated in chapter 2 of ref. [13], in
sections 2.1 2.2 of ref [14], in chapter 9 of ref. [3] and in section 2.1 of ref [7].

Problems

4.1 How much is the wavelengths of the D
o
and T
o
lines from deuterium and tritium
shifted with respect to the H
o
line?

4.2 Since the ionization energy of neutral hydrogen is 1 Ry = 13.6 eV, clearly there is an
absolute threshold energy for ionization. Discuss how this threshold depends on the
projectile mass in the case of cold deuterium atoms being ionized by electron impact or by
deuteron or triton impact.

4.3 Derive classically the differential and total cross section for a collision between hard
spheres, in the case of a sphere with radius a impinging on a much heavier stationary sphere
with radius b.

4.4 Derive classically the total and differential cross sections in the lab system for a collision
between hard spheres, in the case of a sphere with radius a and mass
a
m impinging on a
stationary sphere with radius b and mass
b
m .Compare the results for mass ratios between 0.5
and 2.
4.5 For a reaction cross section of the form
2
R R R
P b o t = , where the cutoff impact
parameter
R
b is given by /
x
R
b C E = , show that if both the reacting species have
a Maxwellian velocity distribution the temperature dependence for the rate
coefficient is
( 2 1/ 2) x
R
k T o
+
.



References

[12] Herbert Goldstein, Classical Mechanics. Addison-Wesley, 2
nd
ed. 1980.
[13] R.E. Johnson, Introduction to Atomic and Molecular Collisions, Plenum Press 1982.
[14] J.N. Murrell and S.D. Bosanac, Introduction to the Theory of Atomic and Molecular
Collisions, John Wiley & Sons 1989.
31

Lesson 5. Elastic collisions and the Rutherford cross section

Let us now analyze in more detail the case of elastic scattering and determine the differential
scattering cross section for Coulomb collisions, both classically and quantum mechanically.
Figure 5.1 illustrates an experimental setup with a particle arriving from the left being scattered
by a potential that is spherically symmetric about the origin. In classical mechanics particles
follow precise trajectories. The impact parameter b is defined as the distance between the
scattering centre and a straight line extrapolated from the trajectory of the incident particle at
large distance, i.e. b would be the closest approach if the particle were not deflected at all.



As discussed in the previous lesson, this scattering problem is also representative of two particles
colliding, provided that the reduced mass is used. Since the potential is spherically symmetric, all
particles arriving with impact parameter b will be scattered into the same polar angle u and we
need not be concerned about the azimuthal angle. If particles with impact parameters in the range
| | , b b db + are scattered into the deflection angles | | , d u u u + then by the definition of the
differential cross section the number of particles per unit time scattered into the solid angle
sin d d u u O= is

2 2 sin
d
Ib db I d
d
o
t t u u =
O
,

where I is the incident flux density, assumed to be uniform. Depending on the shape of the
scattering potential, the deflection angle may either increase or decrease with increasing impact
parameter, but the absolute values in the above equation make sure that the number of scattered
particles is always positive. It follows that
32

sin
d b db
d d
o
u u
=
O
(5.1)

Let us now derive from scratch, in a straight forward pedestrian manner, the relation between
impact parameter and deflection angle by classical mechanics for the case of scattering in the
Coulomb potential. We can write the Newtonian equation of motion in polar coordinates as

1 2
2
0
4
r
q q
r e
r

tc
= (5.2)
where here is the reduced mass. With, say, the direction of incidence in the positive z-direction
and the scattering taking place in the plane y=0 we can write

(cos sin )
r z x
r re r e e u u = = +
and

sin cos
z x
e e e
u
u u = +
whence
r
r r e r e
u
u = + and
2
( ) ( 2 )
r
r r r e r r e
u
u u u = + + . Using the abbreviation
1 2 0
/(4 ) K q q tc the vector equation (5.2) can then be written:

2 2
/ r r K r u = (5.3)

2 0 r r u u + = (5.4)

By equation (5.4) it follows that

2
( ) ( 2 ) 0
d
r r r r
dt
u u u = + =
This means that
2
r u is a constant of the motion. It can easily be identified with the angular
momentum:
2
y
L r p r r r e u = = . If the initial relative speed of the two collision partners
is
0
v , in fact
0
L v b = and
2
0
r v b u = . At this point, the really helpful trick is the substitution
( ( )) 1/ u t r u = , which gives
2
/ / r u u Lu u ' ' = = and consequently
2 2 2
/ / r Lu L u u u '' '' = = . Inserting this in equation (5.3) the equation of motion becomes:

2 2
/ u u K L '' + = (5.5)

The general solution is
2 2
cos sin / u A B K L u u = + , where the coefficients A and B can be
determined from the initial conditions. Firstly, the incident particle is initially infinitely far away,
such that
2 2
( ) 0 / u A K L t = = . Secondly the particle is initially approaching with speed
0
v ,
so that
2 2
0 0
( ) ( ) / 1/ r v L u v B b t t ' = = = .

2
2
1
( ) sin (1 cos )
K
u
b L

u u u = + (5.6)
The scattering direction must be given by the conditions 0 u = and u t = , which gives

2
2
sin
(1 cos )
K
L b
u
u + =
33

Rewriting once again and reinserting the value of K we get


1 2
2
0 0
sin
2 1 cos 4
tan
q q
v b
u u
u tc
= =
+
(5.7)

Equation (5.7) defines the relation between the impact parameter b and the scattering angle u . To
get the differential cross section for classical Coulomb scattering we insert (5.7) in (5.1) and
differentiate b with respect to u to get


2
1 2
2 2
0
4
2
( )
(16 ) sin
q q d
d
E
u
o
tc
=
O
, (5.8)
where
2
0
/ 2 E v = is the kinetic energy in the CM-system. Equation (5.8) is the famous
Rutherford cross section formula, which Ernest Rutherford used in 1911 [15] to explain the alpha
particle scattering experiments by Geiger and Marsden from two years earlier
2
[16], surprisingly
showing that most of the mass of an atom is concentrated in a small, positively charged nucleus.

For the general case of scattering in a spherically symmetric potential ( ) V r , the relation between
b and u can also be derived in the form of an integral expression in terms of the potential. The
total energy in the CM-system is constant,
2
0
/ 2 E v = . When the particle separation is small and
the potential energy is not negligible, the energy in polar coordinates is:

2 2 2 2
0
( ) ( )
2 2
E r r V r v

u = + + = (5.9)
After inserting
2
0
r v b u = in (5.9) we can extract

2
This experiment and its background and importance are discussed in G.L. Trigg, Crucial experiments in modern
physics, Crane, Russac & Co, New York 1971, chapter 5. The most likely looking theory about atomic structure in
the first years of the 20
th
century was the one developed by lord Kelvin and J.J. Thomson. The electron had been
discovered by Thomson in 1898 and its properties had been studied. It was widely believed then that the neutral atom
consisted in electrons immersed in a uniformly distributed positively charged medium, which also carried most of the
atomic mass. The model has been called a plum pudding model, with the electrons scattered around as raisins in a
pudding. Hans Geiger was a post doc working in Rutherfords lab in Manchester, he had been looking at small angle
scattering of alpha particles while penetrating thin foils. Geiger and Rutherford thought it might be a good exercise
for the young PhD student Ernest Marsden to look for particles scattered also through large angles,, although nobody
believed that there would be any. Surprisingly if was soon discovered that some projectiles were scattered also
through large angles, even backwards, when a beam of alpha particles was directed at a thin gold foil. Rutherford and
his co-workers realized that the large angle scattering could only be understood as the result of interaction with one
single, massive target nucleus, since a scenario with many successive collisions with light positive charge carriers
would be far too improbable. Rutherford derived the formula (5.8) and in improved experiments in 1913 both the
angular and energy distributions could be confirmed. The picture of a massive, positively charged nucleus, with
electrons orbiting around it somehow, was thus firmly established experimentally, but it could not be understood
classically how such an atom could be stable. Another young post doc arrived to Rutherfords group and started
immediately in 1911 to work on the penetration of charged particles through matter. His name was Niels Bohr and he
promptly went on to developing the first quantum model of the atom, the Bohr model.
34

2
0
( )
1
V r b
r v
E r
| |
=
|
\ .
(5.10)
and using once again
2
0
/ v b r u = we can get a differential equation for the trajectory:

2
2
( )
1
d b
dr r
V r b
r
E r
u u
= =
| |

|
\ .
(5.11)
The plus sign should be used for the approach phase, when r and u are both decreasing, while the
minus sign corresponds to the later part of the trajectory, when r is increasing.
Formally integrating along the trajectory and with the initial value
i
u t = one gets the scattering
angle, since the two integrals, from infinity to the point of closest approach and back out to
infinity are identical

min
2
2
2
2
( )
1
r
b dr
V r b
r
E r
u t


=

}
(5.12)
The evaluation requires calculating the distance of closest approach
min
r , as defined in figure 4.3.
The distance
min
r can be determined by putting 0 r = in equation (5.10).

Just like in classical mechanics, an exact solution to the Coulomb scattering problem can be
found also in wave mechanics. However, we will concentrate on an approximate solution, in the
sense of first order perturbation theory, the technique of which is applicable also in inelastic
scattering, mutatis mutandi. The central scattering integral can be arrived at along different
routes, let us first discuss the scattering problem for a spherically symmetric potential ( ) V r in the
light of diffraction of any kind of wave. We consider once again a particle, or a beam of particles,
moving in the positive z-direction and being deflected by a scatterer at the origin. The incident
wave vector is
a a z
k k e = . Let us say that the scattered particle is represented by the wave vector
b
k , which is in a direction u from
a
k , as shown in figure 5.3. Since the collision is elastic, the
wave number far away from the scattering centre is the same for the scattered particle as for the
incident particle:
b a
k k k = = . The total energy is obviously
2 2
/(2 ) E k = and the time
independent Schrdinger equation

2
2
( )
2
V r E

| |
V + =
|
\ .
(5.13)
can be rewritten as

( )
2 2
2
2 ( )
( ) ( ) ( )
V r
k r r S r

V + = = (5.14)
where we have introduced the notation ( ) S r for the right hand side of the equation and are
temporarily going to treat it as if it were a given source distribution. Equation (5.14) is then called
the inhomogenous Helmholtz equation. For a given S, the complete solution to (5.14) should be
any solution to the homogenous equation plus a particular solution to the inhomogenous
equation,
h s
= + . The solutions to the homogenous Helmholtz equation in Cartesian
35
coordinates are the familiar plane waves
ik r
e

and we choose
a
ik z
h
e = ,which is just the incident
beam. Turning to the particular solution
s
, we can use spherical coordinates instead of
Cartesian, since we know that the scattered wave is going to be propagating from the origin.
( ) S r is not spherically symmetric, but we can think of the source as being built up by an
aggregation of point sources ( ) ( ) S r r r o ' . We then get the full solution by integrating the
solutions to the point source problem over the whole region where ( ) S r is not zero (or not
negligible). Clearly that region is small when we are analyzing an atomic collision and want
the solution at
Figure 5.3

macroscopic distance from the target. Looking at equation (5.14) in this light, the problem is
familiar, it is identical to calculating the electric potential at a point in space due to a time
harmonic charge distribution (cf e.g. ref [17] pp. 333-334,339). The contribution from a source
element at a point r' is spherically symmetric about that point and is easily calculated in the
spherical coordinate R r r' = after the substitution ( ) ( ) / R u R R = . The contribution from the
whole source is

3
1 ( )
( )
4
ik r r
s
S r e
r d r
r r

t
'
'
' =
'
}
(5.15)
At large r, r and r' are practically parallel (cf figure 5.2), so
r
r r r e r ' ' ~ . Inserting this and
the value of S in (5.15) and adding the homogenous solution we get

2
3
2
4
( ) ( ) ( )
r
ikr
ike r ikz
r V r r d r
r
e
e e

t

'
' ' ' =
}
(5.16)
This expression as it stands is not a solution to the problem, since appears in the integrand.
However, (5.16) forms a basis for approximation. In the first place in the integrand can be
replaced with the unperturbed incident beam
a
ik r ikz
e e

= . This is called the first Born


approximation. If we write the scattered wave vector as
b r
k ke = , the Born approximation can be
written as

3
2
( ) ( )
2
a a b
ikr
ik r ik r ik r
r V r d r
r
e
e e e

t
' '
' ' ~
}
(5.17)

Conceivably this approximate solution can be inserted again in equation (5.16) for iterative
improvement. Generally, the scattering amplitude ( ) f u is defined by
( ) , ( )
a
ikr
ik r
r r
r
e
e f u

~ + (5.18)
so in the first Born approximation
36

) 3
2
( '
( ) ( )
2
a b
i k k r
f V r d r e

u
t

' ' ~
}
(5.19)
We see that the scattering amplitude in the Born approximation is essentially a 3D Fourier
transform of the potential. The incident wave function represents a probability current density
/ I k = . By eq. (1.13) the second term in (5.18) at large distance carries a radial flux density
2
2
( )
s
f
k
I
r
u

= plus terms of order


3
1/ r etc. By the definition of the differential cross section
2
s
I d I r d o = O, so

2
( )
d
f
d
o
u =
O
(5.20)

The integral in (5.19) is best evaluated in spherical coordinates with respect to
a b
k k . From
figure 5.3 we can see that
2
2 sin
a b
k k k k
u
A = = . We get:

( )
2
2 2
0 0 0
cos
( ) ( ) sin ( )
i kr i kr i kr
f r V r d rV r dr
i k
e e e
t
0

u 0 0

A A A
= =
A

} } }
(5.21)

For a general potential V(r) the integral is more conveniently written

2
0
2
2
2
( ) sin(2 sin ) ( )
2 sin
f kr rV r dr
k
u
u
u

=
}
(5.22)
The integral diverges if the Coulomb potential is inserted directly, but the integral is easily
evaluated for a screened Coulomb potential (Yukawa potential, Debye Hckel potential):

1 2
0
4
( )
r
q q
r
V r e
o
tc

= (5.23)
In this case we get

( )
( ) ( ) 1 2 1 2
2 2 2 2
0 0 0
2
( )
4 4 ( ( ) )
i k r i k r
q q q q
f e e dr
i k k
o o

u
tc tc o

A A +
= =
A + A
}
(5.24)
In the limit of 0 o we finally get


2
2
1 2
2 2
0
4
2
( )
( )
(16 ) sin
q q d
f
d
E
u
o
u
tc
= =
O
(5.25)
i.e. exactly the same as the classical Rutherford differential cross section. If the quantum
mechanical calculation is carried out exactly it is found that ( ) f u gets a different phase compared
to (5.24), but the absolute value is the same. Consequently, for the Coulomb potential classical
mechanics, exact quantum mechanics and the Born approximation all give the same differential
cross section. This is a remarkable coincidence, usually the agreement is not that good.

37
The simplest exact quantum mechanical solution to the Coulomb scattering problem is found
through separation of the Schrdinger equation in parabolic coordinates ( r z = + , r z q = ,
arctan( / ) y x = ), cf. Merzbacher [5] section 11.8, Joachain [7] section 6.1.

Problems

5.1 Derive the Rutherford cross section using equation (5.12) for the Coulomb potential.

5.2 Derive the Born approximation formula from time dependent perturbation theory, i.e.
Fermis Golden rule (3.9).

5.3 Show by direct insertion in the equation that


'
4 '
( ' )
ik r r
r r
G r r
e
t

= (5.26)
is a solution to
( )
2 2
( ' ) ( ') k G r r r r o V + = , satisfying the boundary condition of
being an outgoing spherical wave at r .

5.4 A variation of the divergence theorem in vector analysis is Greens second identity:


( ) ( )
2 2
V S
u v v u dV u v v u dS V V = V V
}}} }}
(5.27)
for any two scalar functions ( ) u r and ( ) v r . Use (5.27) to show that if

( )
2 2
( ' ) ( ') k G r r r r o V + = and
( )
2 2
( ) ( ) k r f r V + = , where ( ) f r drops
sufficiently quickly as r , then

3
( ) ( ' ) ( ') r G r r f r d r ' =
}}}
(5.28)

5.5 Consider a particle with kinetic energy
0
E , which is scattered in an attractive inverse-cube
force potential:
2
/ V A r = . Show that by classical mechanics the cross section for capture
is
0
/ A E t . The problem is taken from [3].


References

[15] E. Rutherford, Phil. Mag. 21(1911)212.
[16] H. Geiger and E. Marsden, Proc. Roy. Soc. 82(1909)495.
[17] David K. Cheng, Field and Wave Electromagnetics, 2
nd
ed. Addison Wesley 1989






38
6. The partial wave expansion for the scattering problem

As we have seen in the discussion of bound states in hydrogen, the angular part of the Laplace
operator in spherical coordinates fits with the angular momentum operator
2
L , so that for a
spherically symmetric potential ( ) V r the Hamiltonian in spherical coordinates can be written

2
2
2
2 2
1 1
( )
2 2
L
H r V r
r r r r
c c
= + +
c c
(6.1)
It is clear from this expression that the Hamiltonian commutes with
2
L . z L also commutes with
2
L and H , so eigenfunctions to H can be chosen as simultaneously eigenfunctions of
2
L and
z L . The latter eigenfunctions are the spherical harmonics (2.5) and the corresponding
eigenvalues are given by (2.6-2.7):


2
2
( , ) ( 1) ( , ), 0,1, 2.....
m m
L Y Y u u = + = (6.2)

( , ) ( , ), , 1....., z
m m
L Y m Y m u u = = + (6.3)

Inserting the product ansatz functions ( ) ( , )
m
R r Y u = in the time independent Schrdinger
equation (1.6) with the Hamiltonian (6.1) gives, just as for the hydrogen atom a set of ordinary
differential equations for the radial functions:

2 2
2
2 2
1 ( 1)
( ) ( ) ( )
2 2
d d
r V r R r ER r
r dr dr r
| | +
+ + =
|
\ .
(6.4)
The boundary conditions in the scattering problem are different from the hydrogen atom, we no
longer look for solutions that vanish at infinity (bound states). This means that there are no
restrictions on the energy E , the total energy is just the kinetic energy of the incident particle (in
the CM-system), which can take on any positive value. The functions R will depend on E
however (or more conveniently they will depend on 2 / k E = ). The boundary condition is
that
( ) , ( )
a
ikr
ik r
r r
r
e
e f u

~ + (6.5)
so that we have a plane wave representing incident particles with a well defined k and a term
representing a spherical scattered wave coming out from the scattering region. We can also
require that the solution be bounded at 0 r , so that ( ) 0, 0 rR r r . Due to conservation of
angular momentum the particle classical trajectory lies in a plane. Experimentally, with a
spherical potential and no external field we have no way of specifying the azimuthal angle , so
we need not be concerned with 0 m= , we can write the appropriate eigenfunctions
( ) (cos ) R r P u = . If we define a normalized potential
2
( ) 2 ( ) / U r V r = and substitute
( ) ( ) / R r u r r = , equation (6.4) simplifies to

2
2
2 2
( 1)
( ) ( ) 0
d
k U r u r
dr r
| | +
+ =
|
\ .
(6.6)
39
It is also sometimes convenient to define an effective potential

2
( 1)
( ) ( ) U r U r
r
+
= + (6.7)
where the first potential term represents the centrifugal force. We then get

2
2
2
( ) ( ) 0
d
k U r u r
dr
| |
+ =
|
\ .
(6.8)

If we can assume that the potential ( ) 0 U r sufficiently quickly when r , at large radii
equation (6.8) becomes
2
0 u k u
''
+ = with asymptotic solutions ( ) sin( ) u r a kr q + for large
r . For 0 U = we know that equation (6.8) has solutions
, 0
( ) ( )
U
u r kr j kr
=
,where ( ) j z are the
spherical Bessel functions of eqs. (2.13)-(2.17). The asymptotic behaviour of these functions is
( ) sin( ) ,
2
kr j kr kr r
t
[8]. It is therefore convenient to write the asymptotic phase
shift o so that ( ) sin( )
2
u r a kr
t
o + . In the absence of a scattering potential, 0 o = . We
are now in a position to introduce the boundary condition (6.5) in (6.8), if we expand

0
1
( ) ( ) (cos ) r u r P
r
u

=
=

(6.9)
and

0
( ) (cos ) f f P u u

=
=

(6.10)

With the incident k vector in the z-direction
a z
k ke = and the plane wave expressed in spherical
coordinates, eq. (2.13)

cos 0
0
(2 1) ( ) (cos )
ikz ikr
e e i j kr P
u
u

=
= = +

(6.11)
the asymptotic boundary condition (6.5), separated in becomes

1
( ) (2 1) ( )
ikr
f
u r i j kr
r r
e + + (6.12)

and consequently

/ 2
) (2 1) )
2 2
sin( sin(
ikr
l l
ka kf kr e kr e
t
t t
o + + + (6.13)
It follows that

( )
( )
( / 2) ( / 2)
( / 2) ( / 2)
/ 2
(2 1)
2
i kr i kr
i i ikr i kr i kr
i
e e
ikf e ka e e e e
e
t t
o o t t
t

+
=
(6.14)
The coefficient for inward going waves is zero, which gives
(2 1) /
i
a i k e
o
= + (6.15)
This can now be inserted in the equation for the coefficient of outward going waves

40

/ 2 2
2 [ (2 1) (2 1) ]
ikr ikr i i
ikf e i i e e e
t o
= + + (6.16)
so that

( )
2
1
2 1 ( 1)
2
i
f e
ik
o
= + (6.17)
and finally the scattering amplitude can be written

( )
( )
( )
0
0
2
1
( ) 2 1 1 (cos )
2
1
2 1 sin (cos )
i
i
f P
ik
P
k
e
e
o
o
u u
o u

=
= + =
= +

(6.18)
This formula shows how the scattered wave can be conceived as being composed of separate
contributions, partial waves, from different angular momenta / impact parameters. Each
trajectory is associated with a phase shift o . This is very different from the classical picture, as
can be most clearly seen when we form the differential cross section

( )( )
( )
'
2
2
0 ' 0
' '
1
( ) 2 1 2 ' 1
sin (cos ) (cos ) sin
i d
f
d k
P P
e
o o o
u
o u u o

= =

= = + +
O



(6.19)
We can see that there are interference terms connecting different partial waves ' = . Figure 6.1
shows an example of this effect.


Figure 6.1 Comparison of the differential
cross section calculated classically and fully
quantum mechanically. The example is from
ref [12] for an Ar-Ar potential. This potential
has an attractive part and a repulsive part,
the well depth is 12 meV and the incident
energy is 24 meV.





These interference terms are averaged out
when the integral cross section is calculated, thanks to the orthogonality property of the Legendre
functions [8]:

'
'
0
2
(cos ) (cos )sin
2 1
P P d
t
o
u u u u =
+
}
(6.20)
Inserting (6.20) in (6.19) gives the integral cross section
( )
2
2
0
4
2 1 sin
k
t
o o

=
= +

(6.21)
Incidentally, this expansion diverges for the Coulomb potential and the above discussion is not
completely valid unless the potential vanishes faster than 1/ r at infinity.

41
By eq. (6.18) the phase shifts are only determined up to a multiple of t . However, o is taken as
zero in the absence of a potential and is then allowed to vary continuously if a nonzero potential
is established by smoothly changing some parameter, thus the phase shifts can take on values
outside the t interval. Figure 6.2 shows the phase shifts for the Ar-Ar potential used in figure
6.1 [12].

Figure 6.2. Phase shifts o for the Ar-Ar
potential. The incident energy is twice the
depth of the potential well.
The negative phase shifts correspond to
trajectories entering the repulsive part of the
potential, while the positive phase shifts are
associated with the attractive part of the
potential. Large imply a large centrifugal
term in ( ) U r and correspond classically to
large impact parameters, where the potential
is weak, hence 0 o .

We have seen how the effect of the scattering
potential can be expressed through the phase shifts of the partial waves. The phase shifts still
need to be calculated, which generally involves solving equation 6.6 for each , numerically if
necessary. In numerical solutions the radial functions inside the range of the potential have to be
matched smoothly to the oscillating solutions far away from the target. There are also a number
of simplifying approximations possible in calculatingo . In practice the partial wave expansion is
most useful in case of low energy and short range
of the potential. In that situation only a few terms
in the partial wave equation may have to be
included. This is illustrated semi-classically in
figure 6.3. A particle with mass m is approaching
the scatterer at impact parameter a. The angular
momentum in the direction perpendicular to the
particle trajectory is a k = , so that / a k = .
The value 0 = may be associated with trajectories
inside the circle with radius
1
1/ a k = , the angular
momentum 1 = with trajectories with impact
parameters between
1
1/ a k = and
2
2/ a k =
etcetera. If for example the potential is essentially zero for r d > and 1 kd , then we expect that
only the spherically symmetric s-wave ( 0) = will contribute to the scattering amplitude. More
generally, the partial wave equation can be truncated at
max
if
max
kd .

One route to calculating approximate phase shifts is the semi classical method using the WKB
approximation. This is a generally applicable method for one dimensional problems in quantum
mechanics. The wave function is approximated with
42

( )
( )
1
( ) ( )
( )
i k r dr
iS r
r F r
k r
e e

=
}
= (6.22)
where

2 ( ( ))
( )
E V r
k r

= (6.23)
This approximation can arguably be good both in the classically allowed region ( 0 E V > , k
real) and in the classically forbidden region ( 0 E V < , k imaginary), but clearly not around
E=V. This can be mended by assuming that the potential has a simple form near the classical
turning points, for which the Schrdinger equation is solvable, and use that to join the WKB
solutions for the different regions. The WKB approximation is discussed in Park [3] section 4.4
and more thoroughly in Merzbacher [5] chapter 7. The semi classical method in scattering is
treated in chapter 3 of ref. [11] and in chapter 2 of ref. [12].


Problems

6.1 The issue of how fast ( ) U r should vanish at infinity in order to warrant the
discussion of the asymptotic phase shift following eq. (6.8) requires further investigation.
Show that U needs to drop at least as fast as s>1
s
r

, . Put ( ) ( )
ikr
u r F r e

= and assume
that ''/ F F can be neglected beside '/ kF F and try to show that the phase of F becomes
independent of r for large r .

6.2 Show that the integral cross section and the scattering amplitude in the forward
direction are related by

4
Im (0) f
k
t
o = (6.24)
This is called the optical theorem.

6.3 Calculate the s wave phase shift
0
o and scattering amplitude
0
f for a hard sphere of
radius
0
r .

6.4 Consider the case of a spherical potential well,
0 0
( ) , V r V r r = s , and zero
outside. Find the s wave phase shift
0
o . What is the condition for higher partial
waves not contributing appreciably to the scattering amplitude? Calculate an approximate
value for the integral cross section in this case. Under what conditions can the integral cross
section be zero (cf Ramsauer effect)?







43
Lesson 7. Bremsstrahlung and effective Z

Classically we know that time varying currents and charge distributions entail the emission of
electromagnetic radiation. When charged particles collide they are accelerated. An electric charge
being accelerated is equivalent to a non stationary current and consequently is associated with the
emission of radiation. This type of radiation obviously slows the colliding particles down and is
called bremsstrahlung. From a practical point of view it is the total radiated power density and the
frequency distribution of the emitted radiation that are most interesting. The polarization and
angular distribution of the bremsstrahlung radiation may also be of interest, but are usually
isotropic in plasma applications, due to averaging over random collisions.

In elementary electromagnetic field theory the fields due to moving charges are calculated from
the retarded scalar and vector potentials [17]:


3
0
( ' / )
1
( , ) '
4 '
t r r c
V r t d r
r r

tc

=

}
(7.1)
and

3 0
( ' / )
( , ) '
4 '
J t r r c
A r t d r
r r

t

=

}
(7.2)
where is the charge distribution, J is the current density and both should be evaluated at the
retarded time ' ' / t t r r c = , i.e. at the time of the emission event at the source. The
integration is over the volume of the source. We want to specialize this to the case of a single
charge q moving along some trajectory
0
( ) r t . However, some care should be taken about the
source volume element
3
' d r and it is easiest to view the point charge as the limiting case of a
charge density filling up a volume V A , which can be made arbitrarily small while keeping
q V = A constant. If the source is small and located near or around the origin, ' '
r
r r r r e ~ .
Thus ' / ' /
r
t t r c r e c ~ + . If we first consider the case of v c (i.e. non relativistic conditions),
the retarded time is approximately independent of the variable of integration in eqs. (7.1) and
(7.2). It is then easy to just insert
0
( ' ) J qv r r o = in equation (7.2) to get

0
4
A qv
r

t
= (7.3)
where
0
v r = is the particle velocity. We are interested in the radiation field far away from the
source, where we know that it is similar to plane wave waves
( ) i k r t
e
e
, / k c e = . In that region
we have

0 0 0
1 1
k k
ik
H A e A e A
c
= V = = (7.4)
Inserting (7.3) in (7.4) gives
sin
4 4
k H
e
q q
H v e a
rc rc
u
t t
= = (7.5)
where we have introduced the acceleration a and the angle u between the acceleration vector
and the direction
k
e in which the plane wave is observed. The direction
H
e of the magnetic field
44
vector is perpendicular both to the direction of propagation and to the acceleration vector of the
particle. We also know that for plane waves the electric field is
0 k
E H e q = , where
0 0 0
/ q c = . Bringing together these results, we get in the non relativistic case the radiating
power density from the Poynting vector

2 2
2
2 3 2
0
sin
16
k
q a
S E H e
c r
u
t c
= = (7.6)
Integrated over all directions the total radiated power is

2 2 2 2 2
3
2 3 3
0 0 0 0
sin
16 6
rad
q a q a
P d d
c c
t t
u u
t c tc
= =
} }
(7.7)
This is called the Larmor formula for the radiated power and was first derived by J.J. Larmor in
1897. The power loss is proportional to the acceleration squared. Obviously, in collisions
between electrons and ions, due to their much smaller mass and consequently larger acceleration
it is the motion of the electrons that gives rise to most of the bremsstrahlung radiation.

The discussion above considers the emission from a single particle with acceleration a r = . In
case of collisions between identical particles, e.g. electron-electron collisions, qv in equation
(7.3) should be replaced with
1 2
( ) q v v + . However, since by Newtons third law
1 2
v v = , when
taking the time derivative the total dipole emission is zero. Consequently the bremsstrahlung due
to collisions between like particles is negligible and we are left with electron-ion collisions as the
important source of bremsstrahlung.

At particle velocities not negligible compared to the speed of light, equations (7.1) and (7.2) have
to be handled with more care. To state the result first, the relativistically correct potentials are the
Linard-Wiechert potentials [18-22]:

0 0
1
4
ret
q
V
r r tc k
(
=
(


(7.8)
and


0
0
4
ret
q v
A
r r

t k
(
=
(


(7.9)
where there is a relativistic correction ( )
0 0
1 /
v
r r r r
c
k = and everything within the
brackets is to be evaluated at the retarded time ' ' / t t r r c = . One way to show (7.8) and (7.9)
is to rewrite e.g. (7.1) in a way that handles formally the retarded time condition [20]:

( ', ')
( , ) ( ' ' / ) ' '
'
r t
V r t t t r r c dt dV
r r

o = +

}}
(7.10)
where
0
( ', ') ( ' ( ')) r t q r r t o = . The spatial integration then gives

0
0
( , ) ( ( ) / )
( )
d
V r t q t r r c
r r
t
o t t
t
= +

}


45
and using the formula ( ( )) ( ') / '( ') F t F t o t o t = , where ( ') 0 F t = , gives (7.8). Similar for (7.9).
A more intuitive way [22] of understanding the factor k in the denominator of (7.8) and (7.9)
may be outlined as in figure 7.1.

Figure 7.1. A rod of length L with charge density is moving in the direction v . The radiation is
observed in the direction r .

If we apply eq. (7.1) to a rod shaped particle with cross section area S and its axis in the
direction of v , the potential

0
( , )
4
ret
V r t V
r

tc
~ A (7.11)

where
ret
V A is the set of points ' r that lie within the particle at the retarded time ' ' / t t r r c = .
As ' t is different at different points ' r within the particle,
ret
V A doesnt represent a
simultaneous position of the rod. The position of the rod is initially A1-B1 and the time t A later it
is A2-B2. Events in the whole interval A1-B2 contribute to the field at r at time t. From figure
7.1 ( )cos c t L v t u A = + A , such that

cos
cos
L
t
c v
u
u
A =


and the volume element in the integral is

0
( )
cos
1 cos
ret
V SLc
V S L v t
v
c v
c
u
u
A
A = + A = =



where
0
V A is the static, simultaneous volume of the particle. Inserting this and
0
q V = A in
(7.11) gives

0
1
( , )
4
(1 cos )
q
V r t
v
r
c
tc
u
(
(
~
(
(

(7.12)
46
where the quantities in the bracket should be evaluated at the retarded time. A particle of any
shape could be divided up in rod shaped pieces and when we let the size of the particle go to zero
the equation becomes precisely (7.8). Equation (7.9) follows by the same argument when handled
one component at the time.
When the potentials (7.8) and (7.9) are used for calculating H and E by differentiation, care
needs to be taken about the retarded time:

' 1 t
t k
c
=
c

etc. As a result, the radiated far field becomes

3
0
1
4
r r
ret
q v a
E
c r c c
e e
tc k
( | |
=
| (
\ .
(7.13)
and the radiated power density

2
2
2 2 6
0 0
1 1
16
r r r r
ret
q v a
S E e e
cr c c
e e
t c k
(
| |
= = (
|
\ .
(

(7.14)
which reduces to eq. (7.6) when v c .

The radiated power
2
/ d W d de O per unit solid angle and unit angular frequency is obtained from
Fourier transforms of the above results [14]:

0
2
2 2 2
( ' )
3
0
'
16
r
i t e r
r r
d W q v
e e e dt
d d c c
e
e
e t c

=
O
}
(7.15)
In the frequency range 0.1 / 10
e
kT e s s , even with quantum mechanical corrections the
temperature dependence of the radiated power frequency distribution is
/
/
e
e
kT
kT e
e
[18].
Thus, by measuring the slope of the bremsstrahlung spectrum the electron temperature can be
determined. In principle it is sufficient to measure the relative intensities in two selected energy
ranges to determine the temperature. However, when such a method is used it is very important to
ascertain that no other radiation is important in the selected energy ranges, such as impurity line
radiation.

A first estimate of the bremsstrahlung power loss in a plasma can be made as follows [23]. The
acceleration of an electron at a distance b from an ion with charge state Z is

2
2
0
4
e
Ze
a
m b tc
= (7.16)
The duration of a collision is approximately 2 / b v t ~ and the energy loss per such collision
consequently 2 /
rad
E P b v o = . Inserting (7.16) and (7.7) gives

2 6
2 3 3
0
48 ( )
e
Z e
E
m c vb
o
tc
= (7.17)
In the same way as when calculating the differential cross section in scattering, we get the total
energy loss rate by summing contributions from different impact parameters b, keeping in mind
that the flux density of electrons is
e
n v . With a density of ions
Z
n the radiated power per unit
volume becomes
47

2 6
2 2 3 3 2
0
2
24
e Z
br e Z
e
n n Z e db
P n n Ev b db
m c b
o t
t c
= =
} }
(7.18)
At this point it is necessary to enter some reasonable atomic distance as lower limit for the
integral. Using the uncertainty relation /
e
p r b m v A A > > and approximating the velocity
with its typical thermal value at electron temperature
e
T , so that
2
/ 2 3 / 2
e e
m v kT ~ , we get

6
2
3/ 2 2 3 3
0
3
24
br e Z e
e
e
P n n Z kT
m c t c
~ (7.19)

which is a little less than a factor 2 higher than a calculation with more realistic assumptions
about the electron trajectory, velocity averaging and quantum mechanical corrections [14,19].

When there are ions present with many different charge states, collisions with each species
contribute as in equation (7.19). The total bremsstrahlung power then becomes proportional to
2
e Z
Z
n n Z

. Here
e Z
Z
n n Z =

and it is customary to define an effective Z:



2
Z
Z
eff
Z
Z
n Z
Z
n Z
=

(7.20)
This way the bremsstrahlung power is proportional to
2 2
e eff
n Z .


Bremsstrahlung is discussed in sections 5.1 and 5.3 of ref. [18]. The emission radiation from an
accelerated point charge is thoroughly treated in refs. [19-21].

Problems

7.1 At what electron energy can the relativistic effects proportional to v/c no longer
be neglected? In what electron temperature range would you say that relativistic
effects can be neglected?

7.2 Compare the bremsstrahlung cooling rate with the fusion heating power at different
electron temperatures, in a DD plasma and in a DT plasma with 50% of each isotope.
Take the electron temperature to be equal to the ion temperature. For simplicity ignore DT
fusion in the DD case and DD fusion in the DT case. The heating power is the amount of
energy released to charged particles per fusion reaction, times the number of fusion
reacions. In the DT case 3.5 MeV per fusion reaction is imparted to alpha particles. In the
DD case there are two almost equally probable reaction channels, giving either protons
and tritons with in total 4.04 MeV energy, or
3
He ions with 0.82 MeV energy for heating.
For simplicity, lump the two channels together and take the average 2.43 MeV heating
energy per DD fusion event. The fusion cross sections can be approximated with the
expression
/
( ) ( ) /
G
S e
c c
o c c c

= , wherec is the energy in the centre of mass system and


( ) S c is a relatively slowly varying function of energy. A first approximation at low
energy is to use a constant (0) S . For the sum of the two DD reactions, use
48
28 2
(0) 110 10 keV m S

= and 986 keV
G
c = . For the DT reaction use
24 2
(0) 1.2 10 keV m S

= and 1182 keV
G
c = . For DT this approximation is not very
good above about 50 keV. In calculating the rate coefficients, either directly numerically
or with some analytic approximation, it is wise to note that the integrand is quite peaked
in a relatively narrow energy range.

References

[18] Ian H. Hutchinson, Principles of plasma diagnostics, Cambridge University Press
1987.
[19] J.D. Jackson, Classical Electrodynamics, John Wiley & Sons, 2
nd
ed 1974.
[20] L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields, Pergamon Press
1975.
[21] D.B. Melrose and R.C. McPhedran, Electromagnetic Processes in Dispersive Media,
Cambridge University Press 1991.
[22] Anders Ramgard, Relativitetsteori, KTH 1977.
[23] John Wesson, Tokamaks, Oxford Science Publications, 2
nd
ed. 1997.



















49
Lesson 8. Plasma resistivity

A plasma contains ions and electrons, which are free to move in the direction parallel to the
magnetic field, if there is one. The parallel resistivity is due to collisions. If the collisions are
predominantly between charged particles and neutrals, we call the plasma weakly ionized, while
if only collisions between charged plasma constituents are important we call the plasma fully
ionized.

If we first consider the case of a homogenous, weakly ionized plasma with no magnetic field (or
we consider the particle motion parallel to the magnetic field), the charge carriers will be
accelerated in the presence of an electric field and there will consequently be a systematic particle
velocity in the direction of the electric field. The average velocity is limited only by collisions
with neutral atoms or molecules. If on average a plasma particle loses all its acquired forward
momentum every time it makes a collision, then a stationary average drift velocity is established,
so that the momentum loss of any plasma constituent particle equals the momentum gain in
between collisions:

i
i
i
q
u E
m
t
=
where t is the collision time, the average time between two successive collisions of a particle
(4.26). The coefficient /
i i i
q m t = is called the mobility of species i . The current density
i i i i i i
i i
J n q u n q E
| |
= =
|
\ .

. Since the mobility
e
of the electrons is likely to be much larger
than that of any ion, in view of the small electron mass, the ion conductivity can be neglected.
Thus, if the resistivity q is defined by E J q = , then for the weakly ionized plasma and collisions
with a single neutral species with density
0
n :

0
2 2
e e
e e
m n v m
n e n e
o
q
t
= (8.1)

where equation (4.26) has been used for the last step.

In case of a fully ionized plasma it is the collisions between ions and electrons that determine the
resistivity. If the ions are fully stripped, the basic interaction is the elastic scattering in the
Coulomb potential. The integral cross section of the bare Coulomb potential is infinite, and while
the electrostatic electric field from a point charge drops as
2
r

, the volume element for


integration increases as
2
r , so if there were no spatial correlation between the charges in a
homogenous plasma, a particle would effectively be interacting with other particles at all
distances. What actually happens is obviously that electrons will tend to approach positively
charged ions and ions will repel each other, so that the bare point charge field is screened.

If we first consider a one dimensional case, if in a plasma volume the electron density and the
density of singly charged ions are both equal to n and the electrons are then displaced or removed
in a slab
D
x s , a space charge region with charge density ne = is created. By Poissons
equation the potential in the space charge follows from
0
( ) / V x c '' = . The solution is
50
( )
2 2
0
( )
2
D
ne
V x x
c
= and if we equate the potential energy of an electron at 0 x = with the typical
one dimensional kinetic energy of the free electron / 2
e
kT we get the Debye length:

0
2
e
D
e
kT
n e
c
= (8.2)

Secondly, consider a positive charge e at the origin, surrounded by spherically symmetric clouds
of singly charged ions and electrons [23]. In spherical coordinates the Poisson equation becomes:

2
2
0
1
( )
i e
n n d d
r V r e
r dr dr c

= (8.3)
If the densities of ions and electrons are Boltzmann distributed with the same temperature T

/ /
0 0
,
eV kT eV kT
i e
n n e n n e

= =
where
0
n is the density far away from the test charge, assumed equal for ions and electrons.
Moreover we tentatively assume that eV kT and get
0
2 /
i e
n n n eV kT ~ and

2
2 0
2
0
2 1
( ) ( )
n e d d
r V r V r
r dr dr kT c
= (8.4)
As boundary conditions we want ( ) V r to approach the bare Coulomb potential at small r and it
should vanish faster than the Coulomb potential at infinity. Thus, if we introduce the Debye
screening function ( )
D
r u :

0
( ) ( )
4
D
e
V r r
r tc
= u (8.5)
and we require that ( ) 1, 0
D
r r u and ( ) 0,
D
r r u . Equation (8.4) simplifies to

2
0
0
2
( ) ( ) 0
n e
r r
kT c
'' u u = (8.6)
and the solution that satisfies the boundary conditions is

2 /
( )
D
D
r
r e

u = (8.7)
and the screened potential

0
2 /
( )
4
D
r
e
V r
r
e

tc

= (8.8)
has the form of a Yukawa potential. The assumption eV kT is fulfilled if

2
0
4
e
r
kT tc

If we take r as the typical distance between plasma particles,
1/ 3
r n

~ and insert once again the


value of
D
, then this condition is equivalent to ( )
3/ 2
3
4
D
n t

. That this is so is clearly
necessary for the continuous description, indeed it is part of the definition of what a plasma is
[24]. We can define the number of electrons in a sphere with radius
D
:

3
4
3
D e D
N n
t
= (8.9)
51
The requirement is then essentially that 1
D
N .

Now we are prepared to go ahead and discuss Coulomb collisions in a fully ionized plasma. It is
convenient to divide them up in hard collisions and distant collisions. A hard collision can
conveniently be taken as one where the electron is deflected at least / 2 t . We use the classically
calculated deflection function from lecture 5. By putting / 2 u t = in equation (5.7) we get the
impact parameter corresponding to / 2 t scattering:

2
0 / 2 2
0
4
e
Ze
r b
m v
t
tc
= = (8.10)
where v is the electron speed and we have neglected the reduced mass correction. If the impact
parameter is smaller, the deflection is larger, so clearly the cross section for hard collisions is

2 4
2
0 2 2 4
0
16
hard
e
Z e
r
m v
o t
tc
= = (8.11)
The collision frequency 1/
hard hard Z hard
vn v t o = = , for electrons with velocity v . In order to
estimate a resistivity analogous to (8.1) it is of course necessary to average over the electron
velocity distribution.

The distant collisions are by our definition the collisions with impact parameter
0
b r > . Due to the
long range of the Coulomb potential and the Debye screening one can think of any electron
effectively colliding simultaneously with about
D
N ions, the number of particles within a distance
of one Debye length
D
. However, it is possible to estimate an effective collision time
dist
t , an
effective collision cross section
dist
o etc. for distant collisions, which are comparable to 8.11. One
way to do this is to consider the loss rate of parallel momentum due to distant collisions. If an
electron with momentum p is deflected an angleu in an elastic collision, the loss of momentum
in the direction parallel to the initial velocity is

2
2
2
(1 cos )
1 cot
p
p p
u
o u = =
+
(8.12)
But by eq. (5.7) and eq. (8.10)
0
2
cot
b
r
u
= , so we can write

0
2
2
1
e
b
r
m v
p o =
| |
+
|
\ .
(8.13)
Now, if we consider the ions as stationary, the total rate of change of momentum for the electron
is given by multiplying with nv (the apparent flux density of ions an observer travelling with the
electron would see) and integrating over all impact parameters:
52

max
0
2
2
0
2
max
0 2 2 2
0
0
2 4
1
1
2 ln 4 ln
2
b
e
r
D
e e hard
dp b db
nv p b db nm v
dt
b
r
b
r
nm v r nm v
r
o t t

t o

= =
| |
+
|
\ .
(
| |
( +
|
(
( \ .
= ~
(
(

(
(

} }
(8.14)
where the upper limit of integration has been taken as
D
(at which distance we know that the
screened interaction potential vanishes) and we assume that
0 D
r . The ratio of upper to lower
limits of integration
0
/
D
r is conventionally written Aand ln Ais called the Coulomb logarithm.
In fact, combining (8.2) and (8.10) and inserting
2
e e
m v kT ~ for Z=1 we get 3
D
N A ~ , i.e. it is
roughly the number of ions in the Debye sphere. For fusion plasmas ln Ais in the range 10 to 20.
If we compare the effects of hard collisions and distant collisions in terms of collision frequencies
or effective cross sections we get for hard collisions the braking force


hard Z hard
dp
p n v p
dt
v o = = (8.15)
while for distant collisions
4ln
Z hard Z dist
dp
n v p n v p
dt
o o = A = (8.16)
where 4ln
dist hard
o o = A is about 40 times larger than
hard
o . Thus we can conclude that although
the electron loses little speed in each distant collision, since there are so many collision partners
the braking force due to distant collisions is much larger than that due to hard collisions. Another
route to practically the same result is to consider the small changes in perpendicular velocity that
occur due to collisions with small scattering angle u . Using once again eq. (5.7) one gets for a
single collision
2
0
2
4
e
Ze
v v
m vb
o u
tc

~ ~ . Since the v o

in different collisions are uncorrelated the
expectation value for the total change in
2
v

per unit time the sum of the individual


2
v o

so that

max
min
2
2
2
0
2
4
b
Z
e b
d v
Ze
vn b db
dt m vb
t
tc

| |
=
|
\ .
}
(8.17)
This time, by defining the effective collision time as the time to produce a / 2 t deflection
through many distant collisions, the effective cross section becomes 8ln
hard
o A . The conclusion
is the same, the many distant collisions are much more important for the average deflection rate
than the few hard collisions.

If the plasma resistivity q is defined by E J q = , then equating the electrostatic force on the
electrons with the rate of momentum loss due to collisions at the stationary electron drift velocity
e
v :
53
/
e e
eE m v t =
and with
e e
J en v = we get

2
e
e
m
n e
q
t
= (8.18)
Using
1
Z dist
vn o
t
= and
2
3
e e
m v kT ~ gives
( )
1/ 2 2 2
3/ 2
2
0
ln
12 3
Z e
e
e
n m Z e
kT
n
q
tc
A
=

(8.19)
A more precise result, taking into account also electron-electron collisions is the Spitzer
resistivity

( )
( )
1/ 2 2
3/ 2
9
3/ 2
2
0
ln
0.51 1.65 10 ln
3 2
e
s e
e
m e
kT m
kT
q
c t

A
= = A O (8.20)
with
e
kT in keV [23].

Problems

8.1 At what electron temperature would the effective cross section for distant Coulomb
collisions in a fully ionized hydrogen plasma be comparable to the scattering cross section
of neutral hydrogen atoms?

References

[24] Francis F. Chen, Introduction to Plasma Physics and Controlled Fusion, Plenum Press 2
nd

ed. 1984.
[25] Dwight R. Nicholson, Introduction to Plasma Theory, John Wiley & Sons 1983.
[26] Carl-Gunne Flthammar, Inledning till plasmafysiken, KTH 1977.

54
9. Collisions and particle transport

In the previous section we considered the special case of electron collisions with neutrals
or ions and the resulting response to an electric field, but paid no attention to the role of a
magnetic field. The resistivity could then be calculated from a velocity dependent
collision time. The effects of distant collisions could be handled by limiting the impact
parameter integration range to the Debye length, leading to the Coulomb logarithm.

Most of that discussion applies as well to other collision phenomena in plasmas. In completely
uniform and stationary conditions, collisions will eventually lead to all plasma constituents
(electrons, ions of different masses and charge states, even neutrals) being Maxwell-Boltzmann
distributed in energy, with a common temperature. In practice this is usually not the case, but it is
a common approximation in plasma physics to assume that each species can be described by a
drifting Maxwellian velocity distribution, with a determined fluid drift velocity at every point in
space. Besides equations for particle and energy conservation, there follow in such a model fluid
equations of motion like [23,24]:

( )
( )
j j
j j j j j j
j j j j j
dv v
n m n m v v
dt t
n q E v B p R
c | |
= + V =
|
c
\ .
= + V +
(9.1)
The total time derivative to the left is for a given fluid element that is following the fluid motion,
it is conveniently broken up in the part that is an explicit time derivative of the velocity in a fixed
volume element and the part that is due to the fluid element drifting into regions with different
fluid velocity. The scalar pressure
j j j
p n kT = . Viscosity is omitted in (9.1). Source terms that
could arise from ionization, recombination or charge exchange are also omitted. The final term
j
R is meant to include all effects of collisions between different species (collisions between
identical particles do not change the momentum of the fluid element). The most obvious effect of
collisions is a frictional drag, which can be written

( )
j j
j i
i j
ji
fj
n m
R v v
t
=
=

(9.2)
if the effective collision time
ij
t is defined as corresponding to the particle losing all its forward
momentum in a collision with a stationary target. In addition to the frictional force there is also a
thermal force, which is proportional to the temperature gradients of the other species. The reason
for this is the velocity dependence of the collision cross section. If we consider for example the
thermal force on an ion due to electron-ion collisions, by (8.11) the collision time increases
as
3
e
v t . This means that the electrons coming from the colder region will collide more
frequently with the ion than the ones arriving from the hot region. The net effect of this is a
collision force on the ion in the direction of the electron temperature gradient (thermophoresis).
Thus the collision force is a sum of friction and thermal forces,
j Tj fj
R R R = + .

Depending on the problem at hand, different collision times or relaxation times are appropriate
and can be defined, either as a function of relative velocity of the colliding particles, or as
averaged relaxation times over a Maxwellian velocity distribution [23,26,27]. The numerical
factors in the averaged quantities often differ slightly in the literature, depending on the precise
55
premises. In many cases this is less important than keeping track of the dependences on the key
parameters, such as particle mass, energy (temperature) and charge state. For a test particle
moving at velocityv with respect to the target particles with density n , the relation between
collision frequency, collision time and cross section is 1/ ( ) nv v v t o = = . Here are some
examples of relaxation times [27]:

Effective collision time for electron-ion momentum transfer:


1/ 2 2 4
6 2 3/ 2
2 1/ 2 3/ 2
0
4(2 ) ln 1
3.63 10 ln /
(4 ) 3 ( )
i
i e
ei e e
Z e n
Z n T
m kT
t
t tc

A
= = A (9.3)

with temperature in Kelvin. Effective collision time for electron-electron momentum transfer:


1/ 2 4
6 3/ 2
2 1/ 2 3/ 2
0
4 ln 1
2.57 10 ln /
(4 ) 3 ( )
e
e e
ee e e
e n
n T
m kT
t
t tc

A
= = A (9.4)

and the collision time for like ion-ion momentum transfer:


1/ 2 4 4
8 4 3/ 2 1/ 2
2 1/ 2 3/ 2
0
4 ln 1
6.01 10 ln /( )
(4 ) 3 ( )
i
i e i
ii i i
Z e n
Z n T M
m kT
t
t tc

A
= = A (9.5)

where
i
M is the ion mass number. Another important relaxation time is the temperature
equilibration time, that is the characteristic time
eq
t for species i to assume the temperature of
species j through collisions

i j
i
eq
T T
dT
dt t

= (9.6)
where

3/ 2
2
0
2 2 4
(4 ) 3
8 2 ln
i j j
i
eq
i j j i j
mm kT
kT
m m n Z Z e
tc
t
t
| |
= +
|
|
A
\ .
(9.7)

In a hydrogen plasma with approximately equal electron- and ion temperatures

22 950
eq ii ee
t t t ~ ~ (9.8)

It follows that the electron and ion temperatures can easily be different in an experiment.

As a first example of the implications of eq. (9.1), let us consider the case of weakly ionized
plasma, either with 0 B = or we consider only motion parallel to the magnetic field. Moreover,
suppose that the fluid motion is stationary, that the collisions are frequent enough that particles
wont drift into regions of higher fluid velocity within a collision time (i.e. ( ) 0 v v V ~ ) and that
56
the flow is isothermal. Finally we assume that the neutrals with which the plasma particles collide
are immobile. We then get the stationary solution

( )
0
0 0
, ,
j
j j j j j
j j
j j j
j j j j j
v q n E kT n
m n
q kT n
E j e i
m m n
t
v v
= V
V
= =
(9.9)

expressed in collision frequency for collisions with neutrals. We have considered only electrons
and a single ion species. The mobility
j
and the diffusion coefficient
j
D are

| |
j j
j j
j j j j
q kT
D
m m

v v
= = (9.10)
The mobility and the diffusion coefficient are thus related by

q D
kT
= (9.11)
called the Einstein relation. Let us now take the ions to be singly charged and
e i
n n n = = . If the
collision cross section is approximately independent of impact velocity, then the collision
frequency nv v o = is proportional to
1/ 2
m

and
e i
,
e i
D D . However, in a stationary
situation the transport has to be ambipolar:


i i e e
i e
i e
nE D n nE D n
D D n
E
n


V = V
V
=
+

An ambipolar electric field is created, so that the flux of ions and electrons becomes equal. The
flux is
a
D n I = V , with the ambipolar diffusion coefficient

i e e i
a
i e
D D
D


+
=
+
(9.12)
and the equation of continuity is

2
a
n
D n
t
c
= V
c
(9.13)
It becomes more clear what the diffusion coefficient in (9.4) means if we recall that the mean free
path (4.27) between collisions is / v v t v . With
2
kT mv ~ we get

2
/ D v t ~ ~ (9.14)
i.e. it represents a random walk with step length equal to the mean free path for collisions with
neutral atoms.

Next, let us consider transport perpendicular to the magnetic field in a weakly ionized plasma.
Once again the left hand side in eq. (9.1) is assumed to be zero and the temperature uniform. We
consider singly charged ions. If the magnetic field is in the z-direction the equations of motion for
the perpendicular velocity components for either electrons or ions are
57

x x y
y y x
n
mn v enE kT env B
x
n
mn v enE kT env B
y
v
v
c
=
c
c
=
c

which has the solution [24]:

2 2
1
E D
c
v v n
v E D
n

e t

+ V
= +
+
(9.15)
where

2 2 2 2
1 1
c c
D
D

e t e t

= =
+ +
(9.16)
the cyclotron frequency /
c
eB m e = and
E
v and
D
v are the E B drift and the diamagnetic drift:

2 2 E D
E B p B
v v
B enB
V
= =

In this case, if the plasma is strongly magnetized
2 2
1
c
e t and (9.10) and (9.16) give

2
2
2 2
/
L
mv m
D r
m eB
v t
v

| |
~ =
|
\ .
(9.17)
where
L
mv
r
eB
= is the Larmor radius for a gyrating particle with perpendicular velocity v . So in
this case we have a random walk with step length equal to the Larmor radius.

Let us finally consider particle diffusion in a fully ionized plasma. In ref. [24] the single fluid
MHD equations are derived from the equivalent of (9.1) after putting:


( )
( )
i e
i i e e
i e
i e
e i
n m m
mv m v
v
m m
J ne v v
p p p
= +
+
=
+
=
= +
(9.18)
and keeping the already defined resistivity q :

( )
1
e
v
J B p
t
E v B J J B p J
en

q q
c
= V
c
+ = + V ~
(9.19)
In stationary conditions one finds by taking the cross product of the second equation with B that

2 2
E B
v p
B B
q


= V (9.20)
where the second term represents a diffusive perpendicular transport D n

I = V if

2
( )
i e
n kT kT
D
B
q

+
= (9.21)
58
This behaviour is called classical diffusion. Once again, with
2
kT mv ~ and recalling from (8.18)
that
2
/( ) m ne q t = we get
2
/
L
D r t

= , a random walk process with step length equal to the


Larmor radius. Figures 9.1 and 9.2 show roughly what happens when particles collide in a fully
ionized plasma.


Figure 9.1. Two identical particles making a Figure 9.2. Oppositely charged particles
/ 2 t collision. There is no net transport making a t collision.The gyro centres of
perpendicular to the magnetic field [24]. both orbits move in the same direction[24].

Collisions between like particles dont lead to any transport perpendicular to the magnetic field.
but collisions between oppositely charged particles do.

The main issue in magnetically confined fusion is to use fact that transport is reduced by a
magnetic field, in the direction perpendicular to the magnetic field. Classical diffusion is typically
not observed in magnetic confinement experimental devices, like tokamaks, stellarators, reversed
field pinches and magnetic mirror machines. The theory can be modified for toroidal geometry to
include random walk diffusion with a larger step length than the Larmor radius, so called
neoclassical transport [23,24]. Frequently anomalous transport is observed in the experiments and
can be related to turbulence. The particle diffusion can often be close to

1
16
e
B
kT
D
eB
= (9.22)
(Bohm diffusion).

In order to have well defined plasma-surface interaction in the experimental fusion devices, the
plasma edge is usually defined either by limiters or by a divertor, as shown in figures 9.3 and 9.4.

59


Figure 9.3 . A limiter tokamak cross section. Figure 9.4. A poloidal cross section
of a tokamak with single null divertor.

In both cases the central plasma region has closed magnetic field lines and magnetic surfaces. At
the edge there is a so called scrape-off layer, where the field lines terminate on material surfaces.
The plasma consequently diffuses perpendicularly to the confining magnetic field as far as the
last closed flux surface, further out there is a much faster parallel transport along the magnetic
field lines to the surfaces in the scrape-off layer. In the divertor case the last closed flux surface is
determined by the topology of the magnetic field, rather than by a material object, so that the
areas with intense plasma-surface interaction can be somewhat separated from the main plasma
volume. In the scrape-off layer it is mainly the parallel transport that is of interest.

References

[27] Andr Anders, A Formulary for Plasma Physics, Akademie-Verlag Berlin 1990

60
10. Interatomic potentials and the Thomas-Fermi model of the atom

There is an interesting approximation method for calculating either the electron radial charge
distribution in any atom, or the interaction energy when the electron clouds of two atoms overlap
during an atomic collision, which is based on the properties of the free electron gas. This is called
the local density approximation, the semi-free electron model or the Thomas-Fermi method.

In the free electron gas, the electrons can be represented as plane waves ( )
ik r
k e

. If we
imagine the electron gas as being confined to a cubic box of size
3
L O= , the wave vectors can
only assume discrete values. The boundary conditions can be taken either as requiring that the
wave functions vanish at the edges of the box, in which case we get standing wave solutions, or
periodic boundary conditions are used. We go for the latter alternative and say that
( , , ) ( , , ) x y z x L y z = + and similar for y and z . This gives
2
, .... 1, 0,1, 2....
x x x
ik x ik x ik L
x x x
e e e k n n
L
t
= = = (10.1)
and similar for
y
k and
z
k . Allowing for two possible spin directions, the number of single
electron states in an element
3
d k of k-space is consequently

( )
3
3
3
2
2
L
dN d k
t
= (10.2)
Since by the Pauli principle no two electrons can have the same wave vector and spin, in the
ground state all the states are occupied, up to a wave number
F
k , such that the total electron
density is

( )
3
3
4 2
3
2
F
k N t
t
=
O
(10.3)
If we introduce as the number density of electrons, the wave number of the most energetic
electrons in the ground state (the Fermi wave number) becomes

( )
1/ 3
2
3
F
k t = (10.4)
The total kinetic energy is obtained by summing the energies
2
/ 2
k
E k m = up to the Fermi wave
number. If we express the kinetic energy as an energy density:

2 2
2
3
0
2/ 3
2
5/ 3 5/ 3
1
4
4 2
3 3
5 2
F
k
kin
kin
dE k
k dk
d m
m
t
t
t
k
t
2
=
O
| |
= =
|
\ .
}
(10.5)
where the numerical value of the coefficient is
2
21.88 eV
kin
k = . The electrostatic potential
energy density due to the electron-electron repulsion can be calculated as usual as / 2 e V ,
where V is the electrostatic potential. If the many-electron wave function is taken as a product of
single-electron wave functions (as appropriate for completely non-interacting electrons), the first
order estimate of the potential energy involves (cf eq. 3.4) summing electron-electron interaction
contributions
61

2
2
(1) 3 3
0
( ) ( ) '
4 '
pot i j
i j
e
E r r d r d r
r r

tc
2
>
' =

}
(10.6)
Obviously this can not be the whole truth: firstly it does not take into account the Pauli principle,
secondly one would expect some anti-correlation between the positions of different electrons,
since they repel each other. The first step to improve the potential energy estimate for the electron
gas is to use the anti-symmetrized determinatal wave functions (eq. 3.12) rather than the simple
product functions. Doing this results in mixed terms when taking the expectation value of the
electron-electron interaction energy for electrons with the same spin direction:

2
* * 3 3
0
) ') ) ') '
4 '
i j
ex s s i i j j
i j
e
E r r r r d r d r
r r
o
tc
>
= ( ( ( (

}
(10.7)
This is carried out for the plane wave single-electron wave functions in chapter 7 of ref. [11] and
somewhat differently in chapter 17 of ref. [28]. The result is a so called exchange energy density


1/ 3
2
4/ 3 4/ 3
0
3 3
16
ex
ex
dE e
d
k
tc t
| |
= =
|
O
\ .
(10.8)
where the coefficient 10.635 eV
ex
k = . The fact that antisymmetric wave functions are used
reduces the potential energy a bit, since electrons with the same spin directions tend not to be
close to each other. This does still not include all the correlation there should be for an electron
gas with interacting electrons, eqs. 10.5 and 10.8 are just the first two terms in a high density
expansion of the ground state energy of the electron gas [28]. It is a rather good approximation
however, compared to the further approximations we are about to discuss.

In lesson 3 we discussed very briefly how the atomic structure of a many-electron atom can be
handled numerically. The energy levels and electronic charge state distributions of an isolated
complex atom can be determined reasonably accurately, albeit at the price of considerable effort
[11]. In the case of two complex atoms colliding, a calculation at the highest level of achievable
accuracy would require approximate solution of the Schrdinger equation for the overlapping
system at any separation distance of the two atoms. Even if one is only interested in elastic
scattering, each projectile-target combination presents a tremendous task. For the purpose of
calculating things like ion penetration in matter, the backscattering probability of an ion
impinging at the surface or the ejection probability of atoms from a solid target as a result of ion
impact (sputtering), one is interested in much simpler recipes for treating atomic collisions. In
particular it is valuable with methods that can immediately be applied to arbitrary projectile-
target combinations. Such an approximation scheme is described in detail in [29].

Firstly, we assume that the electronic charge distributions have been calculated or approximated
somehow for a set of isolated atoms. A feasible approximation for the interatomic potential in a
collision can then be to calculate the total energy of the merged system as a function of the
distance of separation between the two atoms, assuming that neither of the two electronic charge
distributions changes during the collision. The interaction energy must have a positive
contribution from the repulsion between the two nuclei, a negative contribution from the
attractive interaction between the electrons on atom 1 and the nucleus in atom 2 and vice versa, a
positive contribution due to the electrostatic repulsion between the two electron clouds, a positive
contribution due to the increasing kinetic energy in the region where the electron clouds overlap
62
and finally a negative contribution from the exchange energy in the overlapping region. In each
case we need the difference between the situation where the atoms are close to each other and the
situation when the atoms are distant and dont interact. For the interatomic potential we need the
interaction energy as a function of interatomic distance. Naturally we assume that the charge
distributions are spherically symmetric, or can be averaged to spherically symmetric
distributions, and the interatomic potential too will be spherically symmetric. We can call the
separation between the atomic nuclei R .

( )
nn en ee kin ex
E R E E E E E A = + + +A +A (10.9)
Obviously
2
1 2 0
/(4 )
nn
E Z Z e R tc = .
en
E should be calculated as sum of the potential energy of the
first nucleus due to the electron cloud on the second atom and vice versa. The first term is then

2 2 2 2 2
1 1
0 0 0
( ) ( ) 1
4 4
4 4
R
en
R
r r
E Z e r dr r dr
r R

t t
tc tc

(
= +
(

} }
(10.10)
and analogously for
2 ne
E . Defining

0
0
2
0
0
2
0
( ) ( )4
( )4
( )
r
i i
i
i
r
Q r r r dr
r r
r dr
r
t
t


=
=
}
}
(10.11)
one gets
| | | |
2 2
1 2
2 2 1 1
0 0
( ) ( ) / ( ) ( ) /
4 4
en
Z e Z e
E R Q R R R Q R R
tc tc
= + + (10.12)
To get the electron-electron potential energy one needs to integrate over all charge elements of
one electron cloud the potential energy due to the other cloud:
| |
2
3
2 2 2 2 2 1 1
0
( ) ( ) / ( )
4
ee
e
E r Q r r r d r
tc
1
= +
}
(10.13)
The change in kinetic energy is

5/3 5/3 5/3 3
1 2 1 2
( ) ( )
kin kin
E d r k ( A = + +
}
(10.14)
and the exchange energy is

4/3 4/3 4/3 3
1 2 1 2
( ) ( )
ex ex
E d r k ( A = + +
}
(10.15)

where the integrations are over the overlap region. In ref [29] benchmark, more accurate electron
charge state distributions are used to calculate the interaction potentials for a selected number of
projectile-target atom combinations, using the recipe (10.9)-(10.15). These results are then used
to construct an approximate universal interatomic potential function for use in problems
involving atomic collisions in solids.
A widely used and historically important universal atomic potential function is the Thomas-Fermi
model of the atom
3
. In this model it is assumed that the electron charge distribution in a multi-

3
Historical note: the method was introduced alredy soon after the birth of quantum mechanics, L.H. Thomas, Proc.
Cambridge Phil.Soc. 23(1927)542 and E. Fermi, Atti Accad. Lincei 6(1927)602. It was to prove immensely useful
and continues to be of interest and a subject for discussion, cf. E.H. Lieb, Rev. Modern Phys 53(1981)603-641, F.M.
Fernandez, Phys. Lett. A372(2008)5258.
63
electron atom can be represented with a spherically symmetric number density of electrons ( ) r ,
such that

0
3
0
( )
r
r d r N =
}
(10.16)
where N is the number of electrons. The cut off radius
0
r is some distance beyond which the
charge density can be taken as zero,
0
r can also be taken to infinity. The total energy of the
configuration is calculated as the sum of potential and kinetic energy, with the kinetic energy
given by the semi-free electron value (10.5). The electrostatic potential at radius
1
r is given by

0 1
1
2
1 2 2 2 2 2 2
0 1 0 1 0
1
( ) ( ) 4 ( ) 4
4 4
r r
r
Ze e
V r r r dr r r dr
r r
t t
tc tc
(
= + (
(

} }
(10.17)
The total energy is


0 1
1
2 2
2/3 2 2
1 2 2 2 2 2 2 1 1 1
0 1 0 1 0
1 1
( ) ( ) 4 ( ) 4 ( )4
4 2 4
r r
kin
r
E
Ze e
r r r dr r r dr r r dr
r r
k t t t
tc tc
=
( | |
( + + |
|
(
\ .
} } }
(10.18)

At this point we require that the charge distribution must be such that the energy is minimized,
subject to the condition (10.16). This implies that there should be a Lagrange multiplier e
0
V , such
that

0
2
0 1 1 1
0
( )4 0
r
E eV r r dr o t
| |
+ =
|
|
\ .
}
(10.19)
for small variations
1
( ) r o of the charge distribution. Applying (10.19) to (10.18) gives the
relation

2/ 3
0
5
( ) 0
3
kin
eV r eV k + = (10.20)
With
0
( ) 0 r = we get

0 0
0 0
( )
( )
4
e Z N
V V r
r tc

= = (10.21)

For a neutral atom
0
0 V = . Anticipating that the potential will be a screened Coulomb potential
and defining

0
( ) ( / )
4
TF TF
Ze
V r r a
r tc
= u (10.22)
we get

2
2 2
0
1
( ) ( / )
4
TF
TF
d Ze
rV r r a
dr a tc
'' = u (10.23)
At the same time (10.17) gives

64

0
0
( ) ( )4
4
r
r
d e
rV r r r dr
dr
t
tc
=
}
(10.24)
and consequently

2
2
0
( ) ( )
d e
rV r r r
dr

c
= (10.25)
Substituting from (10.20) and equating (10.23) and (10.25) gives

3/ 2
( )
( )
x
x
x
u
'' u = (10.26)
with /
TF
x r a = and

1/ 3 1/ 3
2 5 2
0 0 0
2 1/ 3 1/ 3
0.8853 36 9
2 128
TF
a a
a
e m Z Z Z
c t t | | | |
= = ~
| |
\ . \ .
(10.27)
using the Bohr radius
0
0.529 a ~ . The characteristic length
TF
a is the Thomas-Fermi screening
length. Several analytic approximations to the Thomas-Fermi screening function have been used
[29], such as

3
3/
3/
/ 3
( ) 1 1
144
TF
x x
x

(
(
| |
| |
( u ~ + +
(
| |
\ . ( ( \ .


~ (10.28)
with 0.8034 = and
2/ 3
12 o = . There is also the Moliere approximation to Thomas-Fermi:

0.3 1.2 6
( ) 0.35 0.55 0.1
x x x
Moliere
x e e e

u = + + (10.29)

the Lenz-Jensen potential:


1.038 0.3876 0.206
( ) 0.7466 0.2433 0.01018
x x x
L J
x e e e

u = + + (10.30)

or the simply exponential screening:

( )
Bohr
x
x e

u = (10.31)

These classical atomic potentials are plotted in figure 10.1. A number of classical expressions for
interatomic potentials use the simple atomic potentials and just change the screening length. Two
suggestions have been [29]:

( ) ( )
0 0
1/ 2 2/ 3
2/ 3 2/ 3 1/ 2 1/ 2
1 2 1 2
0.8853
,
Bohr Firsov
a a
a a
Z Z Z Z

= =
+ +
(10.32)

The choice in ref. [29] is often used today and was obtained by fitting to the above mentioned
benchmark calculations. It is called the universal screening, or ZBL-screening and uses

3.2 0.9423 0.4029 0.2016
( ) 0.1818 0.5099 0.2802 0.02817
x x x x
U
x e e e e

u = + + + (10.33)
with /
U
x r a = ,

0
0.23 0.23
1 2
0.8854
U
a
a
Z Z

=
+
(10.34)
65


Figure 10.1. Classical atomic screening functions.

The universal potential is a considerable improvement compared to the classical potentials, when
compared with accurate numerical benchmark calculations. It has been widely used for
calculations of range distributions, backscattering and sputtering.

Problems

10.1 Calculate and plot the Thomas-Fermi electron charge distribution using any two of the
approximate screening functions (10.28) - (10.31). Plot the Thomas-Fermi charge
distributions for carbon, iron and tungsten atoms with the radial coordinate in ngstroms.



References

[28] N.W. Ashcroft and N.D. Mermin, Solid State Physics, Holt, Rinehart and Winston
1976.
[29] J.F. Ziegler, J.P. Biersack and U. Littmark, The Stopping and Range of Ions in
Solids, Pergamon Press 1985.
66
11. Ion implantation and backscattering

The kinematics of elastic collisions were discussed in section 4. The energy transferred to the
recoil particle in Figure 4.2 is
2
2 2
1
( ')
2
T m v = . Since the speed of the recoil particle in the CM-
system is
CM
V , the transformation of the recoil angle between the lab system and
the CM-system, following Figure 4.1 is given by Figure
11.1. Obviously the CM recoil angle is twice the lab
recoil angle (eq. 4.14). By the law of cosines
( )
2
2
2
' 2 (1 cos( ))
CM
v V t | = . Again by (4.14) t | u = ,
the scattering angle of the projectile in the CM-system,
so we get
2
2
(1 cos )
CM
T mV u = . Passing to the half angle
and using eq. (4.11) we get

( )
2 2 1 2
1 1 2
1 2
2
4 1
sin
2
mm
T mv
m m
u
=
+
(11.1)
If we introduce the incident energy
0
E and write the transferred energy in a head on collision as
max 0
T E = we get

( )
2 1 2
0 2
1 2
,
2
4
sin
mm
T E
m m
u
= =
+
(11.2)
The classical scattering angle in the CM system is given by the classical deflection function of eq.
(5.12). It may be convenient to transform (5.12) to reduced units by the substitutions / x r a = ,
/ b a | = ,
( )
2
1 2 0
/ / 4
CM
E Z Z e a c tc = , where a is an interatomic screening length as discussed in
lesson 10. In terms of the lab system energy the reduced energy is

0 1
2
1 2 1 2
4
L
a m
E
m m Z Z e
tc
c

=
+
(11.3)
Experimental data for atomic collisions in solids are often plotted as functions of this reduced
energy, where the screening length may be
U
a a = or
Firsov
a a = of lesson 10, or

0
2/ 3 2/ 3
1 2
0.8853a
a
Z Z

=
+
(11.4)
Using an interatomic screening function as in lesson 10, (5.12) becomes:

0
2
2
( )
1
2
x x
x x
dx
x
|
c
|
u t

u

=
| |
|
\ .
}
(11.5)
Equations (11.2) and (11.3 or 5.12) provide the tools for calculating the slowing down of a
particle through elastic collisions. The rate of energy transfer is equal to the energy loss rate of
the projectile particle and can be written

1
1 2 1
( )
n
dE
v n S E
dt
= (11.6)
where
67

max max
2
0 0
2
( ) 2 2 sin
b b
n
b db S E T b db E
u
t t = =
} }
(11.7)
is called the nuclear stopping cross section. From (11.6)

1
2 1
( )
n
dE
n S E
dx
= (11.8)
for a particle of type 1 moving in the x-direction through a matrix of particles of type 2 with
density
2
n . A reduced nuclear stopping power ( )
n
s c can be defined by [29]:

2
( ) ( )
n n
U
s S E
a E
c
c
t
= (11.9)
and calculated stopping powers by (11.7) using the ZBL universal potential can be well fitted to
the analytic expression [29]:

( )
0.21226 1/ 2
1.1383
0.01321 0.19593
ln(1 )
( ) , 30
2
n
s
c
c c
c c c
+
= <
+ +
(11.10)
and 2 ( ) ln /
n
s c c c = for 30 c > . This can be taken as an approximate universal nuclear stopping
power. Now, eq. (11.8) can be used for calculating the slowing down of a fast particle in matter
due to elastic collisions, for example a first estimate of the implantation range of an ion
impinging at a surface may be made as

2 0
'
( )
( ')
E
n
dE
R E
n S E
=
}
(11.11)
where
2
n is the density of solid target atoms. However, eq. (11.8) does not address the statistical
aspects of few collisions and it does not say anything about the accumulated deflection due to
collisions. The elastic scattering is the dominating stopping mechanism at low energies, however
at energies above a few keV for light ions, the inelastic stopping due to interaction with the target
electrons becomes more important [30,31]. The total stopping power is then the sum of nuclear
stopping and electronic stopping. We will return to the inelastic losses in a later lesson.

A transport equation for the fate of energetic particles penetrating a solid and being elastically
scattered may be written [13,29]:

| |
2
( , , )
cos ( , , ) ( , ', )
I E e x
n I E e x I E T e x d
x
u o
c
=
c
}
(11.12)
where e is the direction of motion, cos
x
e e u = and ( , , , ') d E T e e o is the cross section for
scattering from energy E and direction e to energy E T and direction ' e . The transport
equation can be solved numerically or with simple cross section assumptions analytically.
Another strategy, which has been very widely applied to fusion relevant problems, is Monte-
Carlo calculations. Conceptually the latter technique is simple, particles are assumed to make
binary collisions with one single target atom at the time and between the collisions an inelastic
continuous stopping due to interactions with target electrons can be included. By following the
particle trajectories, both the implantation range distribution and the distributions of
backscattered particles can be calculated. The target atom positions can be pre-stored at their
original lattice sites in a crystal, or they can be allowed to have random thermal motion around
their equilibrium positions. The best known code that includes crystal structure in this way is
MARLOWE [32]. More common is however to simulate randomly distributed target atoms by an
68
appropriate random distribution of the distance between collisions. The best known code that
does this is TRIM, with different variations [29,33]. TRIM versions updated 2008 can be
downloaded from the website at [33]. Essential for speedy execution is that the collision
dynamics are handled efficiently. TRIM uses among other things an efficient magic analytic
formula for evaluating the deflection angle from (11.5). With the Monte Carlo methods it is
relatively easy to include complications such as peculiar surface texture or a changing target
composition as a result of high fluence ion implantation.


Figure 11.2. Comparison of calculated and measured implantation ranges for boron in silicon
[33].

Figure 11.2 shows an example of how TRIM calculations compare with experiment for
implantation ranges in the 10-1000 keV range. Figure 11.3 shows a set of experimental data for
light ion reflection probability and Figure 11.4 shows corresponding calculated values. Figure
11.5 shows an example of an analytic fit for reflection coefficients, but only for hydrogen and
helium ions impinging at iron and nickel [34]. The fitted formula for the particle reflection
coefficient is

( ) ( )
2/ 3
3/ 2 3/ 2
0.34334 3/ 2
1 3.2116 1.388
N
R c c

(
= + +
(

(11.13)
and for the energy reflection coefficient

( ) ( )
2/ 3
3/ 2 3/ 2
0.3520 3/ 2
1 7.1172
E
R c c

(
= + + 5.2757
(

(11.14)
69

Figure 11.3. Experimental values for the kinematic backscattering probability [34].

Figure 11.4. Calculated values for the kinematic backscattering probability [34].

The theoretical approaches described so far all employ the binary collision approximation
and the potentials that have been used were spherically symmetric and only repulsive.
70

Figure 11.5. Experimental and fitted values for the kinematic backscattering probability [34].

It is intuitively clear that the binary collision approximation must become invalid at very low
energies, i.e. energies comparable to the binding energy between the target atoms. But it is not
clear precisely when that kind of failure will be noticeable. This can be investigated by relaxing
the binary collision approximation and going over to a molecular dynamics approach. In this case
the equations of motions are solved simultaneously for a number of target atoms, which interact
not only with the incident projectile, but also with each other. A comparison between TRIM and
molecular dynamics was made in [35]. The selected problem was carbon backscattering at a
graphite surface. TRIM calculations show the backscattering probability going to zero at low
energy, for normal incidence. This is bound to happen with the binary collision approximation,
since it is kinematically impossible to get backscattering in a single collision, and the projectile
will lose a substantial fraction of its energy in any hard collision (11.2). Molecular dynamics on
the other hand shows significant backscattering at around 20 eV incident energy, even for normal
incidence, as shown in Figure 11.6.
71


Figure 11.6. Comparison of calculated
carbon backscattering at a graphite surface,
using TRIM and molecular dynamics [35].



Figure 11.7 may give more concrete feeling for what is happening. It shows the projectile at
normal incidence making a head on collision with one of the target atoms, atom number 91. At 58
eV the collision is almost binary, in the sense that practically all forward momentum is
transferred to the target atom, however the projectile is still reflected, with a very small backward
velocity. At 60 eV the projectile becomes trapped at the surface. At lower energies the collision is
not at all binary, the nearest neighbors in the top target plane effectively participate in the
collision. A free molecular dynamics program package can be downloaded at [35a].




Figure 11.7. A more detailed plot of what happens
around 50-60 eV. Atom 1 is making a head on collision
at normal incidence with atom 91 at the surface. At 58
eV the projectile is barely reflrected, while at 60 eV it
barely sticks to the surface. In this situation the
collisions is almost binary, in the sense that practically
the whole projectile energy and momentum is
transferred to the target atom [35].









72

References

[30] H.H. Andersen and J.F. Ziegler, Hydrogen Stopping Powers and Ranges in All
Elements, Pergamon Press 1977.

[31] J.F. Ziegler, Helium Stopping Powers and Ranges in All Elements. Pergamon
Press 1977.

[32] M.T. Robinson and I.M. Torrens, Phys. Rev. B9 (1974) 5008.

[33] http://www.srim.org/ (J.F. Ziegler)

[34] W. Eckstein and H. Verbeek, Nuclear Fusion special issue 1984.

[35] H. Bergsker, F. Lama, R. Smith and R. Web, Vacuum 44(1993)341-344.

[35a] http://sites.google.com/site/Kalypsosimulation/Home ( Marcus Karolewski )

73
12. Sputtering

To be completed.

Problems

References

[36] P. Sigmund in Sputtering by particle bombardment
[37] H. Bay ibid.
[38] J. Bohdansky, Nucl. Fus. Spec. Iss. 1984
[39] I. Gudowska et al
74
13. Collisional ionization

The rate of ionization of neutral atoms and subsequently of low charge state ions to higher charge
states is clearly an important process in plasma physics. Due to the small mass and consequently
high thermal velocity of the electrons, the rate coefficient for electron impact ionization is
normally much higher than for ion impact ionization, so it is the former that is primarily of
interest, e.g. for estimating the actual charge state distribution in a plasma.

A classical estimate of the collisional ionization cross section is most easily arrived at if the
process can be viewed as a binary collision between the projectile and a bound electron. This is
called the binary encounter approximation (BEA). It is most likely to be valid in small impact
parameter collisions, while in the case of more distant collisions, e.g. between a free electron and
a many-electron atom, one would expect a more complicated collective response, involving many
target electrons simultaneously. A small impact parameter is equivalent to a relatively large
energy transfer.

A good starting point for exploring ionization in the classical BEA approximation is the energy
transfer in an elastic collision, eq. (11.2):


( )
2 1
2
1
,
2
4
sin
e
e
mm
T E
m m
u
= =
+
(13.1)
where T is the energy transferred to the (bound) target electron, E is the projectile energy in the
lab system,
1
m is the projectile mass and u is the projectile scattering angle in the CM-system.
For electron impact 1 = , while for ion impact 4 /
e i
m m ~ . It is sometimes convenient to define
an energy-transfer cross section (ref. [13] chapter 2), related to the differential scattering cross
section:

d d d cos
= 2
dT
d d
d dT d dT
o o o u
t
O
=
O O
(13.2)
As
2
2
cos 1 2sin
u
u = we get for a fixed
0
E :

4 d d
dT E d
o t o

=
O
(13.3)
Since we are considering close collisions anyway, it is natural to use in the first place the bare
Coulomb differential scattering cross section (eq. 5.8 or 5.25):

( )
2
2
2 2
0 0
4
2

(16 ) sin
Ze
d
d
E
u
o
tc
=
O
(13.4)
where we allow for an ion with charge
1
q Ze = as projectile. Inserting (13.4) and (13.1) in (13.3)
gives:

( ) ( )
2 2
2 2
2 2 2
2 2
0 0
0 0
4
2
4
4
=
(16 )
(16 ) sin
Ze Ze E
d
dT E E T
E
u
t
o t
tc
tc
= (13.5)
75
Next, we assume that if the energy transferred to the bound electron (which is taken as stationary)
exceeds the ionization energy
i
E , ionization will take place. We get the total cross section by
integrating over all energy transfers between
i
E and a maximum energy transfer, which we can
take as E . This integration is equivalent to the integration over impact parameters that we have
used already a couple of times.

( ) ( )
0
2 2
2 2
2 2 2 2 2
0 0 0
4 4
1
(16 ) (16 )
i
E
ion
i E
Ze E Ze
dT E
E T E E

t t

o
tc tc
| |
= =
|
\ .
}
(13.6)
If we specialize to the case of electron impact 1 Z = , 1 = and note that the cross section was
given in terms of the CM energy
0
/ 2 E E = , where E is the projectile energy in the lab system,
we get the classical cross section for electron impact ionization related to one electron with
binding energy
i
E :

( )
2
2 2
2
0 2 2
0
4
1 4 1
(8 )
i
i i
Ze
E Ry E
a
E E E E
t

o t
tc
| | | |
| |
= =
| | |
\ .
\ . \ .
(13.7)
where we have used the Bohr radius (2.12)
10
0
0.529 10 m a

= and the Rydberg energy
(2.10) 13.6eV Ry = . For a many-electron atom with
i
N electrons in shell i with binding energy
i
E the total cross section for electron impact ionization would be

2
2
0
4 1
iontot i
shells
i
Ry E
a N
E E
o t
| |
| |
=
| |
\ .
\ .

(13.8)
where of course
i
shells
N N =

is the total number of electrons of the target atom.


This cross section estimate was first derived by J.J. Thomson in 1912 [40]. Equation (13.8) can
easily be improved a bit [18] in the energy range near the threshold by arguing that an electron
with binding energy
i
E is in orbit with a potential energy 2
i
E . Presumably the projectile
electron would be accelerated by the same energy before making the binary collision with the
bound electron. Also, if the projectile energy after the collision is less than
i
E , it will itself be
bound and there will be no net ionization. Including this correction (13.8) is modified to

2
2
0
1
4 1
2
iontot i
shells
i i
Ry E
a N
E E E E
o t
| |
=
|
+
\ .

(13.9)
Other possible corrections to the Thomson formula are discussed in [13]. When the transferred
energy (13.1) corresponds to an impact parameter that is comparable to the uncertainty of the
particle positions due to the Heisenberg relation, the higher limit of the integration (13.6) should
be modified accordingly. Likewise, at the lower limit of integration, where the impact parameter
is large, the limit of integration could be modified to a a suitable screening length. The
corresponding classical rate coefficient (for a single electron) is evaluated analytically in [18]:

( )
1/ 2
1/ 2
/ 2
0
/
8
4
1
i
i
E kT
e i
E E kT
i i
Ry Ry Ry
v a e
m E kT
E E E E
e Ei
kT kT
o t
t
+

+
+ +
| | | |
| |
=
| | |
\ .
\ . \ .
+ + ( | | | |

| | (
\ . \ .
(13.10)
76
where
i
E is the ionization energy and E
+
can be the correction 2
i
E or (apparently better)
4
i
E E
+
= . The final square bracket with the exponential integral (the last factor in (13.10)) can
be approximated with [18]:

| |
( ) /
1
i
E E kT
i
kT
e
E E
+
+
+
(
=

+
(13.11)

The same kind of classical calculation as for the ionization cross section can be used to estimate
the electronic stopping power in the high energy limit [13]. In the same way as the high energy
nuclear stopping power, the energy loss rate of a charged particle due to electronic excitations
can be written as

( )
max
min
2
2
0 max
2 2 2
0 0 min
4
ln
(16 )
T
T
Ze
dE T d
n N T dT n N
dx dT E T
t
o
tc
| |
~ =
|
\ .
}
(13.12)
but with
min
T now equal to a minimum electron excitation energy of the atoms. In (13.12)
2
n is the
density of target atoms and N is the number of electrons on each atom. We see that the
classically predicted energy dependence for the stopping power at high energy is ln / E E.

Quantum mechanically, in the Born approximation a scattering amplitude can be defined
similarly to the case of elastic scattering (5.19), also for inelastic collision processes, including
ionization:

* 3 3
0 0 2
( , ) ( , ) ( , )
2
f f
f R r V R r R r d R d r

~ + +
}
(13.13)
Here
f
represents the final state,
0
the initial state and V is the total interaction potential,
which should be regarded as small in the sense of the time dependent perturbation theory.
Neglecting the need for anti-symmetrization in case of identical particles, the many-particle wave
functions can be written as plane wave projectile wave functions before and after scattering,
multiplied with the target atom wave functions before and after the collision. The calculations in
this case are similar as for elastic scattering, except that the magnitude of the final k-vector is
smaller than the initial k.

For practical use in plasma modeling, many semi-empirical formulae have been constructed for
the electron impact ionization cross section and rate coefficients [27, 41].
Many of them are available online [42]. The present author has mostly used the recommended
values of Lennon et al. [43]. In the Lennon report the ionization cross sections are fitted to the
formula:

0
0
1
0
1
( ) ln 1
ion
j
N
ion
j
j
ion
E E
E A B
E E E E
o
=
| |
| |
| |
= + |
| |
|
\ .
\ .
\ .

(13.14)
and the rate coefficients to
77

/ 10
0
0
log , /10 10
1
ln
/
i
i
m
E kT
n i i
n
i
n
m
n i
n
i i i
n
kT
E
kT
v e a E kT E
E
kT kT
v kT
E E kT E
o
o o |

=
( | |
= s s
( |
\ .
(
| | | |
( = + , 10E <
| |
(
\ . \ .

(13.15)
The fitting parameters are tabulated in [43] for all ground state ions up to Ni. A selection of
fitting parameters is given in tables 13.1-13.3.

Table 13.1. Fitting parameters for eq. (13.14). The ionization energy is in eV. The A and B
parameters are in units of 10
13
eV
2
cm
2
[43].

Ion/atom E
i
A B
1
B
2
B
3
B
4
B
5

H I 13.6 0.185 -0.019 0.123 -0.19 0.953
Be I 9.3 0.924 -0.770 0.362
Be II 18.2 0.269 0.389 -1.836 3.939 -2.275
Be III 153.9 0.796 -0.5 0.884
Be IV 217.7 0.4
C I 11.3 2.114 -1.965 -0.608
C II 24.4 1.082 -0.161 -0.856 0.906
C III 41.4 0.691 -0.508 0.699 0.014 -0.433
C IV 64.5 0.450 -0.318 1.026 -2.859 1.995
C V 392.1 0.796 -0.500 0.884
C VI 490.0 0.4
Fe I 7.9 1.142 -0.920 1.782 -1.694
Fe II 16.2 2.124 -0.530 -9.617 29.17 -40.81 19.29
Fe III 30.6 4.42 -4.81 5.692 -12.977 6.698

Table 13.2. Coefficients for eq. (13.15) in units of cm
3
s
-1
[43].

Ion/atom a
0
a
1
a
2
a
3
a
4
a
5

H I 2.3743e-8 -3.6867e-9 -1.0366e-8 -3.801e-9 3.415e-9 1.6834e-9
Be I 7.4206e-8 -1.552e-8 -3.9403e-8 7.2155e-9 1.1098e-8 -2.5501e-9
Be II 3.2777e-8 -1.3769e-8 -2.1724e-8 -1.1818e-9 1.2575e-8 4.4398e-10
Be III 1.6057e-9 -6.4406e-10 -7.7888e-10 3.3563e-10 2.1913e-10 -1.0612e-10
Be IV 4.9686e-10 -3.6944e-10 -3.9362e-11 8.9106e-11 8.0167e-12 -1.7149e-11
C I 5.9849e-8 1.1903e-8 -3.0141e-8 -1.3693e-8 8.3749e-9 4.015e-9
C II 2.8395e-8 -1.6698e-8 -2.3557e-9 3.2161e-10 9.6017e-10 5.2713e-10
C III 9.0159e-9 -6.293e-9 -1.3198e-9 1.7365e-9 3.2537e-10 -3.8135e-10
C IV 2.0527e-9 -1.2430e-9 -9.0089e-12 2.8863e-10 -2.1088e-10 8.1898e-11
C V 3.9483e-10 -1.5837e-10 -1.9152e-10 8.253e-11 5.3883e-11 -2.6094e-11
C VI 1.4716e-10 -1.0942e-10 -1.1658e-11 2.6391e-11 2.3744e-12 -5.079e-12
Fe I 1.4438e-7 -8.0018e-8 -6.1752e-8 5.6502e-8 1.2350e-8 -1.7668e-8
Fe II 4.622e-8 -2.4962e-8 2.8052e-9 -2.8845e-9 4.2137e-9 -2.1823e-9
Fe III 3.209e-8 -3.7148e-9 -1.5464e-8 6.2761e-9 1.0119e-9 -1.4006e-9





78
Table 13.3. Coefficients for eq. (3.15) in units of cm
3
s
-1
[43].

Ion / atom

0

1

2

H I 2.4617e-8 9.5987e-8 -9.2464e-7 3.9974e-6
Be I 2.1732e-7 -2.1648e-7 2.811e-7 5.307e-7
Be II 6.4891e-8 -1.1198e-7 6.4469e-7 -1.9617
Be III 2.7903e-9 -2.435e-10 -9.8088e-9 5.2155e-8
Be IV 8.3329e-10 -4.6056e-10 3.7364e-10 2.5630e-9
C I 3.7442e-7 -6.5826e-7 2.052e-6 -4.4688e-6
C II 6.015e-8 -4.0215e-8 -2.7928e-8 5.551e-7
C III 1.4438e-8 -5.3357e-9 -1.0432e-8 1.0585e-7
C IV 5.8147e-9 -5.5184e-9 -4.9521e-9 8.9187e-8
C V 6.8613e-10 -5.9877e-11 -2.412e-9 1.2825e-8
C VI 2.468e-10 -1.3641e-10 1.1066e-10 7.5911e-10
Fe I 3.4615e-7 -4.3391e-7 1.7618e-6 -5.3387e-6
Fe II 2.1839e-7 -3.829e-7 1.0867e-6 -1.8547e-6
Fe III 1.744e-7 -3.0837e-7 9.849e-7 -2.1688e-6

Figure 13.1 shows the electron impact ionization cross section for hydrogen in the ground state,
calculated in different ways. The curves labeled Classic are calculated according to equations
(13.8) and (13.9) respectively. The curve lasbeled Lennon is the semi-empirical fit from [43], eq.
(13.14) with coefficients from table 13.1. The curve labeled Janev is taken from the ALADDIN
database [42].



Figure 13.1. Cross section for electron impact ionization of hydrogen. The solid line is the semi-
empirical fit of ref. [43]. The dashed line is the classical Thomson cross section.


79
Problems

13.1 Plot and compare the classical cross sections (13.8) and (13.9) with semi-empirical,
experimental or calculated cross sections for some many-electron atom. Use data from
[42], [43] or other data available online. It is necessary to estimate the contribution to the
cross section from electrons with different binding energy.
To do this, either make some reasonable assumption about the screening of the nuclear
charge due to inner shell electrons, or find approximate shell binding energies by some
other method. As an example, the Los Alamos online program that is linked to at [42]
allows among other things calculation of average binding energies for atomic subshells.
Running the code for neutral Be, C and Fe gives binding energies as follows:

Atom subshell N
i
E
i
(eV)
Be I 1s
2
2 129.8
Be I 2s
2
2 9.17

Atom subshell N
i
E
i
(eV)
C I 1s
2
2 309.8
C I 2s
2
2 20.3
C I 2p
2
2 12.0

13.2 Consider a fusion device with a strongly magnetized plasma. The edge magnetic field is
parallel to the wall. Assume that the majority ion in the plasma is hydrogen and that
impurities entering the plasma from the wall are ionized at a distance from the wall.
The plasma dimensions are assumed to be large compared to . Following ionization,
impurities are assumed to diffuse in the confined plasma with a perpendicular
diffusivity D. How does the equilibrium impurity concentration at the centre of the
plasma depend on D and ?
13.3 In the device discussed in problem 13.2, compare the cases when the wall is made of
beryllium, graphite or stainless steel and the impurities are released from the wall by
physical sputtering, or by thermal evaporation. Estimate the differences in impurity
penetration for the different wall materials, assuming that the edge electron temperature is
about 30 eV and the edge electron density
18 3
10
e
n m

~ . As estimates of the surface


binding energies of Be, C and Fe, use the cohesive energies 3.32
Be
H A = eV, 7.37
C
H A =
eV and 4.28
Fe
H A = eV [40].
13.4 Develop the example of problems 13.2 and 13.3 further by taking into account the role of
the magnetic field. Discuss the impurity penetration based on a simple model, which
includes release of impurity atoms at the wall by physical sputtering, electron impact
ionization and gyro motion of ions due to the magnetic field. Assume that the neutrals
dont make any collisions before being ionized and that if they are ionized within one
gyro radius from the wall they will hit the wall again and stick there. Discuss how the
impurity penetration depends on edge electron density and temperature and on the
magnetic field. Under what conditions can the plasma be effectively screened from
sputtered impurities?


Atom subshell N
i
E
i
(eV)
Fe I 1s
2
2 7181
Fe I 2s
2
2 885
Fe I 2p
6
6 751
Fe I 3s
2
2 117
Fe I 3p
6
6 76.6
Fe I 3d
6
6 17.2
Fe I 4s
2
2 7.93
80
References

[40] J.J. Thomson, Phil. Mag. 23(1912)419.
[41] M.F.A. Harrison in D.E. Post and R. Behrisch eds. Physics of Plasma-Surface
Interactions in Controlled Fusion, NATO ASI series, Plenum Press 1984.
[42] E.g. ALADDIN, GENIE, OPEN-ADAS, cf links at http://www-amdis.iaea.org/
[43] M.A. Lennon, K.L. Bell, H.B. Gilbody et al. Culham Laboratory CLM-R.270.
[44] C. Kittel, Introduction to Solid State Physics, 5
th
ed. John Wiley & Sons 1976.

81
14. Excitations, inelastic scattering

Besides ionization, inelastic atomic collisions also cause transitions between bound states of
atoms and ions. In fusion plasma physics these excitations are important in the first place as the
primary cause of radiation power loss, and for diagnostic methods using spectroscopy. Inelastic
processes also contribute to the slowing down of fast charged particles penetrating into solids,
electronic stopping.

One way to analyse inelastic processes, which is appropriate for distant collisions, is to regard the
bound electrons as classical harmonic oscillators (Johnson [13], appendix C). A more realistic
semi-classical strategy to calculate approximate excitation cross sections is to assume a classical
trajectory for the projectile and calculate the transition probability between different states of the
target atom by perturbation techniques as a function of impact parameter. Once the transition
probability
0
( )
f
P b

between the given initial state 0 and the final state f is calculated, the
integrated cross section can be derived as usual by integration over the impact parameter:

0 0
( ) 2
f f
P b b db o t

=
}
(14.1)
A more purely wave mechanical approach is to treat the collision in analogy with elastic
scattering and calculate the differential cross section in the Born approximation.

Elaborating on the discussion of perturbation techniques in lecture 3, we can write the
Hamiltonian for the target as (3.2):

(0) (1)
H H H = + (14.2)
where
(1)
( ) H V t = is a time dependent perturbation due to a projectile passing by. Just as for
ionization the rate coefficients are likely going to be larger for electron impact than for ion
impact, simply because of the smaller electron mass and consequently higher electron speed, so
as far as the fate of the target atoms is concerned it is primarily electron impact that is of interest.
The stationary states
(0)
n
+ are taken as known, along with the corresponding energy eigenvalues
(0)
n
E , such that

(0) (0) (0) (0)

n n n
H E + = + (14.3)
and the total unperturbed target wave functions are the periodically oscillating

(0) (0) (0)
,
n
n n n n
i t
e E
e
e

u = + = (14.4)
The perturbed solution is expanded in the unperturbed eigenfunctions

(0)
( ) ( )
n
n n
n
i t
t c t e
e
u = +

(14.5)
and inserted in the time dependent Schrdinger equation (1.5):
i H
t
cu
= u
c
(14.6)
After multiplication of (14.6) with any
(0)
n
u and integration over the spatial coordinates one
obtains

( )
( )
mn m
n mn
n
i t
dc t
i c t V
dt
e
e
=

(14.7)
where
82

(0)* (0)
mn m n
V V = + +
}
(14.8)
and
(0) (0)
mn m n
E E e = . If we next assume that the target is initially for sure in state 0 then

0
( ) 1 , ( ) 0 for m 0
m
c c = = = (14.9)
As V is taken to be small, the coefficients in (14.9) are assumed to change very little, so that

0
0
( )
m m
m
i t
dc t
i V
dt
e
e
= (14.10)
Since we are considering a transient perturbation due to a passing projectile, (14.10) can be
integrated with the result

0
0
( )
f
i t
f f
i
c V e dt
e

=
}
(14.11)
The transition probability is the probability of finding the target in state f after the collision:

2
0
( )
f f
P c

= (14.12)
The perturbation V naturally depends on the separation ( ) R t between the projectile and the target
and the approximate calculation of the transition probability thus involves integrating (14.11)
over the classical projectile trajectory. For example we may wish to consider excitations due to
distant collisions, where the projectile deflection is small and the trajectory may be approximated
with a straight line
2 2 2
( ) R t b v t = + . Substituting z vt = in (14.11) gives

0
/
0
( ) ( ( ))
f
i z v
f f
i
c V R t e dz
v
e

=
}
(14.13)
Generally speaking, for a given projectile velocity the integral will be large for small
0 f
e , i.e. for
transitions between states which are closely spaced in energy. For small velocities the integral
will vanish due to the exponential factor, while for large velocities ( )
f
c goes to zero because of
the factor 1/ v , so the transition probability and hence the cross section will have a maximum at
some intermediate velocity. The evaluation of (14.11) is particularly simplified if the duration of
the collision, the time interval t A when V is significant is so short that
0
1
f
t e A , in which case
the exponential factor can be ignored. For Coulomb interaction the actual form of the
perturbation is

2
2
1
1
0
1
4
Z
j
j
Z e
V
r R tc
=
=

(14.14)
for a projectile
1
Z , summing over
2
Z target electrons. For large R it is convenient to expand

( )
2
2
2
0
1 1
1 2 cos
1 1
1 cos 3cos 1 ....
2
1
(cos )
j
j j
j
j j
j j
j
j
r R
r r
R
R R
r r
R R R
r
P
R R
u
u u
u

=
=

| |
+
|
\ .
(
| |
= + + + (
|
( \ .

| |
=
|
\ .

(14.15)
83
Since the unperturbed wave functions in (14.8) are orthogonal, the first (monopole-) term in
14.15 is zero. Unless for symmetry reasons the second (dipole-) term in (14.15) is zero, the series
can be truncated after the second term.

For a general projectile scattering trajectory, if only the scattering potential is spherically
symmetric, the angular momentum is conserved, hence
2
R bv u = is a constant of the motion.
Thus (14.13) can be written

( )
0
( )
2
1
0 0
0
( ) sin cos
4
f
f
b
i t
f f f
Z e i
c x z e d
vb
u
e
t
u u u
tc
= +
}
(14.16)
where

2
(0)* (0) 3
0 0
1
Z
f jf j j j
j
x x d r
=
=

}
(14.17)
and similar for
0 f
z are components of the dipole moment.

From the point of view of the Born approximation the unperturbed wave function for projectile
and target is the product of a plane wave describing the projectile and a wave function for the
target atom. We can write the initial wave function
0
(0) (0)
0 0
ik R
e

+ = and the final wave function


(0) (0) f
ik R
f f
e

+ = . Generalizing (5.19) the transition probability amplitude can be written



0
(0)* (0) 3
0 0 0
2 2
f
ik R
ik R
f f f
m m
f V e V R e d R
t t

= + + = ( )
} }
(14.18)
with the integral over the target coordinates as defined in (14.8). Equation (14.14) looks just like
the Born approximation for elastic scattering (5.19), the difference being that for inelastic
scattering obviously
0 f
k k = .

Derived or theoretical cross sections and rate coefficients for many fusion plasma relevant
transitions from the ground state to excited states are found in the databases that are available
online [42].

84
15. Equilibria and power balance


15.1 Thermal equilibrium

In thermal equilibrium all particles, radiation, atoms and other subsystems are in equilibrium,
with a common temperature. By the fundamental postulate of statistical mechanics, all accessible
microstates are equally probable [45]. For the atoms and other bound systems this means that if
there are
i
g degenerate states at an energy level with energy
i
E , the number of atoms at that
energy level is proportional to the Boltzmann factor
/
i
E kT
e

and the ratio of the numbers of atoms


at any two levels i and j is

/
,
ij i i
ij i j
j j
kT N g
E E
N g
e
e
e

= = (15.1)
For free particles, including photons, the density of states in k-space is given by (10.2), i.e. the
number of states with wave number in the interval | | , k k dk + is

( )
2
3
( ) 2 4
2
V
k dk k dk t
t
= (15.2)
In case of electrons the factor 2 accounts for two possible spin directions, while in the case of
photons it is related to two linearly independent directions of polarization. The radiation field is
conveniently seen from a thermodynamic point of view as a photon gas, with indeterminate
number of photons but well defined total energy. Photons have integer spin (1) and therefore
populate the available microstates at energy E h e v = = with the Bose-Einstein probability
distribution

/
1
( )
1
BE E kT
f E
e
=

(15.3)
The energy density u of the radiation field is equal to the density of states (15.2) times the
average number of photons per state (15.3) times the photon energy. If we write this energy
density in terms of e we need to insert / k c e = in (15.2) and the energy per unit volume in the
angular frequency interval | | , d e e e + becomes

( ) ( )
3
3 / 3
8
( )
1 2
kT
d
u d
e c
e
t e e
e e
t
=

(15.4)
or, perhaps more familiar, the energy per unit volume in the frequency interval | | , d v v v + is

( )
3
3
/
8
( )
1
h kT
h d
u d
c
e
v
t v v
v v =

(15.5)
and the formula is called Plancks radiation law
4
. Laboratory plasmas are normally not in
complete thermal equilibrium, mainly due to their limited size, in particular the radiation escapes
from the plasma volume and is not in equilibrium with the ions, atoms and free electrons.
However, the hypothetical condition of thermal equilibrium can be used to derive some relations

4
Historical note: this derivation of Plancks radiation law is due to Satayanadra Nath Bose, S.N. Bose, Z. Physik
26(1924)178-181. Bose sent the manuscript to Einstein, who translated it to german and arranged for its publication.
Einstein worked out other interesting consequences for a gas of particles obeying (15.3) [8].
85
pertaining to the interaction between radiation and bound systems. The probability of radiative
transition between atomic bound states can be calculated by the time dependent perturbation
technique. Using the correct perturbation Hamiltonian for the effect of a radiation field on an
atom, Fermis Golden rule (3.9) gives for the probability of stimulated (dipole) transition [18]:

3 2
2
* 3
2
0
8
3 (4 )
j
ij i j
i
g
e
B r d r
g h
t

tc
=
}
(15.6)
If i and j are two atomic levels, such that
i j ij
E E hv = , then the Einstein coefficients
ij
A ,
ij
B and
ji
B are defined so that
ij
A is the probability per unit time of spontaneous transition, while
( )
ji ij
B u v is the probability per unit time for absorption of radiation and ( )
ij ij
B u v is the probability
per unit time for stimulated emission. Since in equilibrium the transition rates in both directions
must be equal:

( )
ij ij i ji j
A B u N B uN + = (15.7)
From (15.1) and (15.7)
5


( ) ( )
/
/ /
ij
ij ij
h kT
j i ji ij j i ji ij
A A
u
N N B B g g e B B
v
= =

(15.8)

whence by (15.5)

3
3
8
,
ij
ij ij j ji i ij
h
A g B g B
c
t v
= B = (15.9)
Although the relations (15.9) were found with reference to thermal equilibrium, the quantities are
clearly atomic quantities, so the relations between them must be valid also when the atoms are
not in thermal equilibrium with the radiation field.

15.2 Local Thermal equilibrium

Even if the radiation field is not in thermal equilibrium and not thermally coupled to the bound
systems and the free particles, the atoms/ions and free electrons may be nearly in thermal
equilibrium. This may come about in a very high density plasma, where the collision frequencies
are much higher than the the radiative transition probability rates. This is called local thermal
equilibrium (LTE). For bound states the ratio of numbers of atoms in different states are given by
(15.1), but if we want to calculate the ratio of the equilibrium densities of different ion charge, it
is necessary to takew into account the density of states of the free electrons in an appropriate
way. The density of free electron states in k-space is given by (15.2) as usual. Rewritten into the
number of states with speed in the interval | | , v v dv + it is, using mv k = :

( )
3
2
3 3
( ) 2 4
2
e
m
V
L v dv v dv t
t
= = (15.10)

5
This derivation of Plancks law is due to Einstein in 1916-1917: A. Einstein, Phys. Zeitschr. 18(1917)121-128.
However, the coefficient /
ij ij
A B was still derived classically (Rayleigh and Jeans), with the energy quanta
hv introduced ad hoc [2,8].
86
By the Maxwellian velocity distribution the actual number of electrons in the same speed interval
is

2
3/ 2
2 / 2
2
4
e e
e
m v kT
m
K n V dv
kT
e v
t
t

| |
=
|
\ .
(15.11)
We need the statistical weight that comes from the different number of ways of arranging K
electrons among L states, compared to the number of ways to arrange K+1 electrons among L
states. The electrons have to be treated as indistinguishable, so different ordering among the
electrons should not be counted separately. Then, the number of ways of ordering K electrons is
( 1).....( 1) / !
K
G L L L K K = + (15.12)
while the number of ways to similarly arrange K+1 electrons is

1
( 1).....( 1)( ) /( 1)!
K
G L L l K L K K
+
= + + (15.13)
This gives the ratio

1
1
K
K
G L K L
G K K
+

= ~
+
(15.14)

If we compare the situations before and after an ionization, the change in energy going from the
lower to the higher charge state is
2
/ 2
i e
E E m v A = + , where v is the velocity of the free electron.
Including the number of states at both levels we consequently expect the ratio of numbers of
atoms in charge states i and 1 i + to be

2
1 1 1
( / 2)/
i e i i K
i K i
E m v kT
N g G
N G g
e
+ + +
+
= (15.15)
By (15.10), (15.11) and (15.14)

3/ 2
3
/ 1 1
3
2 2
i
E kT i e i
i e i e
N m g kT
e
N n h g m
t
+ +
| |
=
|
\ .
(15.16)
The relation (15.16) is called the Saha equation and states the ratio of ion charge states in a
situation when the bound systems (atoms) are in thermal equilibrium with the free electrons
(LTE).

15.3 Coronal equilibrium

At low density, the collision frequencies are much lower than the radiative transition rates. These
conditions are called coronal equilibrium, since they prevail approximately in the solar corona.
Coronal equilibrium can also often be a reasonable approximation for the core plasma in a fusion
device. In this limit, essentially all ions/atoms will be found in their ground states. The
distribution of charge states of atomic species in this case is determined by the balance between
the rates of ionization and recombination:

1 1
1
i i i
i i i
v N
N v
o
o
+ +
+
= (15.17)
where the rate coefficients are defined to include all ionization and recombination processes
respectively, and the recombination rate coefficient is the sum of recombination rates into all
possible states of the atom. The possible atomic processes are, for ionization (Z refers to the
nuclear charge, N is the number of electrons) :
87

*
collisional ionisation ( , ) ( , 1)
photoionisation ( , ) ( , 1)
autoionisation ( , ) ( , 1)
Z N e Z N e
Z N Z N e
Z N Z N e
e
+ + 2
+ +
+


and for recombination the reversed processes:
*
three-body recombination ( , 1) ( , )
radiative recombination ( , 1) ( , )
dielectronic recombination ( , 1) ( , )
Z N e Z N e
Z N e Z N
Z N e Z N
e
+ 2 +
+ +
+


Of these, often only collisional ionization, radiative- and dielectronic recombination need to be
taken into account. Collisional ionization we discussed in lesson 13. We will not discuss the
recombination processes any further, but for modeling of plasma behaviour the rate coefficients
for many common plasma ions are tabulated in the online databases [42]. Typically radiative
recombination dominates at the lowest electron temperatures and dielectronic recombination
(does not exist for hydrogen, obviously) is more important at higher temperatures. Figure 15.1
shows the distribution of charge states for oxygen as a function of electron temperature, in
coronal equilibrium [46]. Typically one charge state dominates in a particular temperature
interval. This means that for a fusion plasma device, which is heated in the centre and loses
energy and particles at the edge by transport, so that the plasma is hot in the centre and cool at the
edge, impurity charge states are found in fairly well defined shells, with low charge states at the
edge and high charge states in the centre of the plasma.

Figure 15.1. Charge state distribution for oxygen in coronal equilibrium [46].


15.4 Equilibria with transport

88
In a fusion device, which is necessarily of finite size, the impurity charge state distribution
obviously depends also on how ions are transported in the plasma. A 1D model in cylinder
geometry can be set up for instance like this, radial flux:
( ) ( ) ( )
i
i i i i
n
D r V r n r
r
c
I =
c
(15.18)

Charge state balance:

| |
1 1 1 1
( ) 1
( )
( )
i i
e i i i i i i i i
i
i
i
n r
n s n s n n n
dt r r
n
S r
r
o o
t
+ +
c c I
= + +
c
+
(15.19)
This model allows for radially dependent diffusivity and pinch velocity. The source term S would
be a neutral source at the edge and the final term is meant for parallel loss in the scrape-off layer
region.

15.5 Power loss.

For fusion applications, equilibria come in particularly for estimating power losses due to
impurity radiation. Reference [47] is a thorough investigation of radiative losses by different
mechanisms, based on the assumption of coronal equilibrium.


Figure 15.2. Fitted total cooling rate due to carbon impurity in a plasma in coronal equilibrium
[47].

89

Figure 15.2. Fitted total cooling rate due to tungsten impurity in a plasma in coronal equilibrium
[47].

The paper gives fitting parameters for the total radiative cooling rates of ions of all elements up to
Uranium and is useful for modeling the power balance in a fusion plasma. The limitations of the
coronal equilibrium assumption should be kept in mind though.


References

[45] Tony Gunault, Statistical Physics, Chapman & Hall 1995.
[46] Yuan Ding, Modelling of the radiative power loss from the plasma of the Tore
Supra tokamak, KTH masters thesis 2008.
[47] D.E. Post, R.V. Jensen et al. Atom. Dat. Nucl. Dat. Tabl. 20(1977)397-439.
90
16. Charge exchange and neutral transport


Collisions resulting in the transfer of an electron between two atoms/ions are important in plasma
physics. Electron capture into an excited state of an impurity ion

0 ( 1) * n n
A H A H
+ + +
+ + (16.1)
is an important process for diagnostics, if the subsequent radiative emission can be measured
spectroscopically. The source of neutral hydrogen could be for example a high energy neutral
beam, used for heating the plasma. Charge transfer between neutral hydrogen atoms and protons

0 0
H H H H
+ +
+ + (16.2)
is also important, even though it doesnt change the ion density. In particular, when cold neutral
atoms enter the plasma from the edge, charge exchange processes lead to the escape of the
previously charged particle, which after neutralization is no longer confined by the magnetic
field. Charge exchange determines the neutral density profile, in particular near the edge of the
plasma, it constitutes an energy loss mechanism and a source of energetic particles bombarding
the walls of a plasma experiment. The flux of energetic neutrals from the plasma core also offers
a possibility to determine the core ion temperature, if the energy distribution of the neutrals can
be measured.

As discussed in lecture 14, the time dependent perturbation method can be used to calculate
transition probabilities. If the system is initially for sure in a state
0
, the subsequent state can be
expanded in unperturbed eigenstates
k
and the coefficients (probability amplitudes) are
approximately given by the differential equations (14.10). In case of a transient perturbation
V the probability amplitude for finding the system in the new state
f
was given by (14.11):

0
0
( )
f
i t
f f
i
c V e dt
e

=
}
(16.3)

and the probability of transition is given by the coefficient squared (14.12):


2
0
( )
f f
P c

= (16.4)
The coupling matrix element

* 3
0 0 f f
V V d r =
}
(16.5)

determines transition from state
0
, to state
f
. For example, in the case of charge exchange
between a hydrogen atom and a proton, the initial state could in the first approximation be taken
as the single electron wave function describing an electron in the ground state, centred at one of
the two nuclei involved in the collision. The final state could be identified with the electron being
in the ground state, centred on the other nucleus. The normalized wave functions for hydrogen in
the ground state according to (2.12) are then

0 0
0
3/ 2 3/ 2
0 0
/ /
1 1
,
r R
f
a r a
a a
e e
t t

= = (16.6)
91
if we use a spherical coordinate system with the origin at the initially neutral atom and the
distance between the two nuclei is R . The characteristic size
0
0.529 a ~ is the Bohr radius. The
perturbing interaction Hamiltonian can be taken as

2
0
4
e
V
r R tc
=

(16.7)
while the collision lasts. The transition is resonant, in the sense that the energy difference
between the final and initial states is zero, so that
0
0
f
e = , and insertion of (16.7) and (16.6) in
(16.5) gives the probability for transition within the time interval t A if the interaction matrix
element is taken to be constant during the encounter. The spatial integral is hard to handle in
spherical- cylinder or Cartesian coordinates, since it involves the distances to two focal points.
The convenient thing to do is to make a substitution and use prolate spheroidal (or elliptical)
coordinates:

( )
3 3 2 2
,
1 , 1 1,
1
8
r R r r R r
R R
d r R d d d
q
q t
q q
+
= , =
s < s s 0 s < 2
=


The result for the transition probability assuming constant separation during a short interaction
time t A becomes:

0
2
2 2
0 2
0 0 0
/
1
4
f
R a
t e R
P
a a
e
tc


| | A
=
|
\ .
(16.8)
The distance between the nuclei can be taken as constant and equal to the impact parameter:
R b ~ . The time duration of the collision can be approximated with / t b v A ~ . These
approximations give

0
2
2 2
0 2 2
0 0 0 0
2 /
1
( ) 1
4
f
b a
e b b
P b
a v a a
e
tc


| | | |
=
| |
\ . \ .
(16.9)
This can not be straight forwardly integrated over impact parameters to give an integrated cross
section, since the probability can never be larger than 1, in fact it should be quite a bit smaller in
order to warrant the approximations underlying the perturbation calculation. The argument used
in Hutchinson [18] is to truncate (16.9) at the largest impact parameter
1/ 4
b where the transition
probability becomes equal to , and assume that transition occurs for all impact parameters
smaller than
1/ 4
b . Numerically one finds that
1/ 4 0
6 b a ~ for a 100 eV proton colliding with a
stationary atom. This gives a charge transfer cross section
19 2
(100 ) 3.3 10 m
CX
eV o

~ in good
agreement with the experimental values plotted in figure 16.1 [41].

92





Figure 16.1. Experimental charge
exchange cross section for the
reaction
(1 ) (1 ) H H s H s H
+ +
+ +

[41]. The fitted curve is based on the
expression
10 2
(7.07 1.83 log ) v o =



Figure 16.2. Rate coefficients for charge
exchange and collisional ionization. Figure 16.3. Experimental cross section
for
2
(1 ) (1 ) (1 ) He s H s He s H
+ +
+ +
[41].

Generally speaking the charge exchange cross section is high in conditions of resonance, and in
this symmetric case it is high for low impact velocity. At higher proton velocity one needs to take
into account the translational velocity, so the unerturbed wave functions become more
complicated and the cross section drops faster than indicated by (16.9). Figure 16.2 shows a
comparison of rate coefficients for charge exchange and collisional ionization for the hydrogen-
proton collision. Figure 16.3 shows an example of a charge exchange cross section that is not
resonant at zero impact energy. More charge exchange cross sections and rate coefficients can be
found in the online databases [42].

Neutral transport in the plasma, at the edge as well as in the core is a result mainly of charge
exchange. It is difficult to treat the problem analytically, but as for many other scattering
93
problems Monte Carlo codes are very useful, especially for specific geometry and realistic source
distributions. The MCNC code is a Monte Carlo program for neutral transport, worked out and
written by Marco Cecconello, both in Mathematica and MATLAB. It is primarily intended for
the interpretation of time of flight neutral particle spectrometry at the T2R reversed field pinch
experiment [48,49]. The algorithm is that of Hughes and Post [50]. The same algorithm has been
taken up again recently in [51].

Problems.

16.2 Study the documentation for MCNC and run the code for input data in approximately the
same range. How far into the plasma can one say that neutrals penetrate from the edge? Try to
find out how the neutral density at the centre of the plasma scales with electron temperature and
to what extent the energy distribution of escaping neutrals is representative of the central ion
temperature.

94

References

[1] Anders Vretblad, Fourier Analysis and its Applications, Springer Verlag
New York 2003, chapter 7, or similar literature.
[2] Hermann Haken and Hans Christian Wolf, Atomic and Quantum Physics,
Springer Verlag, New York 1984.
[3] David Park, Introduction to Quantum Theory, 3
rd
ed. McGraw-Hill 1992.
[4] Stephen Gasiorowicz, Quantum Physics, 3rd ed. John Wiley Sons 2003.
[5] Eugene Merzbacher, Quantum Mechanics, John Wiley & Sons 1970.
[6] P.A.M. Dirac, The Principles of Quantum Mechanics, 4
th
ed. Clarendon Press 1951,
latest printing 2009.
[7] Charles J. Joachain, Quantum Collision Theory, Elsevier Science,
3rd ed. Amsterdam 1983.
[8] Max Jammer, The conceptual development of quantum mechanics, McGraw-Hill
1966, and references therein.
[9] C. Ramsauer, Annalen der Physik 66(1921)545. Reproduced in R. B. Brode,
Reviews of Modern Physics 5(1933)257. The Ramsauer effect or Ramsauer-
Townsend effect was independently discovered also by J.S. Townsend.
[10] Milton Abramovitz and Irene Stegun eds. Handbook of Mathematical Functions, 9
th
printing 1972, Dover Publications.
[11] Robert D. Cowan, The theory of atomic structure and spectra, University of
California Press 1981
[12] Herbert Goldstein, Classical Mechanics. Addison-Wesley, 2
nd
ed. 1980.
[13] R.E. Johnson, Introduction to Atomic and Molecular Collisions, Plenum Press
1982.
[14] J.N. Murrell and S.D. Bosanac, Introduction to the Theory of Atomic and
Molecular Collisions, John Wiley & Sons 1989.
[15] E. Rutherford, Phil. Mag. 21(1911)212.
[16] H. Geiger and E. Marsden, Proc. Roy. Soc. 82(1909)495.
[17] David K. Cheng, Field and Wave Electromagnetics, 2
nd
ed. Addison Wesley 1989
[18] Ian H. Hutchinson, Principles of plasma diagnostics, Cambridge University Press
1987.
[19] J.D. Jackson, Classical Electrodynamics, John Wiley & Sons, 2
nd
ed 1974.
[20] L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields, Pergamon Press
1975.
[21] D.B. Melrose and R.C. McPhedran, Electromagnetic Processes in Dispersive
Media, Cambridge University Press 1991.
[22] Anders Ramgard, Relativitetsteori, KTH 1977.
[23] John Wesson, Tokamaks, Oxford Science Publications, 2
nd
ed. 1997.
[24] Francis F. Chen, Introduction to Plasma Physics and Controlled Fusion, Plenum
Press 2
nd
ed. 1984.
[25] Dwight R. Nicholson, Introduction to Plasma Theory, John Wiley & Sons 1983.
[26] Carl-Gunne Flthammar, Inledning till plasmafysiken, KTH 1977.
[27] Andr Anders, A Formulary for Plasma Physics, Akademie-Verlag Berlin 1990.
[28] N.W. Ashcroft and N.D. Mermin, Solid State Physics, Holt, Rinehart and
Winston 1976.
95
[29] J.F. Ziegler, J.P. Biersack and U. Littmark, The Stopping and Range of Ions in
Solids, Pergamon Press 1985.
[30] H.H. Andersen and J.F. Ziegler, Hydrogen Stopping Powers and Ranges in All
Elements, Pergamon Press 1977.
[31] J.F. Ziegler, Helium Stopping Powers and Ranges in All Elements. Pergamon
Press 1977.
[32] M.T. Robinson and I.M. Torrens, Phys. Rev. B9 (1974) 5008.
[33] http://www.srim.org/ (J.F. Ziegler)
[34] W. Eckstein and H. Verbeek, Nuclear Fusion special issue 1984.
[35] H. Bergsker, F. Lama, R. Smith and R. Web, Vacuum 44(1993)341-344.
[35a] http://sites.google.com/site/Kalypsosimulation/Home ( Marcus Karolewski )
[36] P. Sigmund in Sputtering by particle bombardment
[37] H. Bay ibid.
[38] J. Bohdansky, Nucl. Fus. Spec. Iss. 1984
[39] I. Gudowska et al
[40] J.J. Thomson, Phil. Mag. 23(1912)419.
[41] M.F.A. Harrison in D.E. Post and R. Behrisch eds. Physics of Plasma-Surface
Interactions in Controlled Fusion, NATO ASI series, Plenum Press 1984.
[42] E.g. ALADDIN, GENIE, OPEN-ADAS, cf links at http://www-amdis.iaea.org/
[43] M.A. Lennon, K.L. Bell, H.B. Gilbody et al. Culham Laboratory CLM-R.270.
[44] C. Kittel, Introduction to Solid State Physics, 5
th
ed. John Wiley & Sons 1976.
[45] Tony Gunault, Statistical Physics, Chapman & Hall 1995.
[46] Yuan Ding, Modelling of the radiative power loss from the plasma of the Tore
Supra tokamak, KTH masters thesis 2008.
[47] D.E. Post, R.V. Jensen et al. Atom. Dat. Nucl. Dat. Tabl. 20(1977)397-439.
[48] M. Cecconello, TRITA-ALF-2001-03, KTH 2001.
[49] M. Cecconello, TRITA , KTH 2001.
[50] M.H. Hughes and D.E. Post, J. Comput. Phys. 28(1978)43.
[51] D. Yu, L. Yan, G. Zhong et al. Plasm. Sci. Techn. 9(2007)2.

You might also like