You are on page 1of 31

Radial. Phys. Chem. Vol. 41, No. 4/5, pp.

673703, 1993
Printed in Great Britain. All rights reserved

0146-5724/93 $6.00 + 0.00


Copyright 1993 Pergainon Press Ltd

AN OVERVIEW OF THE TRANSPORT THEORY OF


CHARGED PARTICLES
2
Luo ZHENG-MING and ANrnsts Bit~mse
Center for Radiation Physics, Institute of Nuclear Science and Technology, Sichuan University,
Chengdu, 610064, P.R. China and 2Department of Radiation Physics, Karolinska Institute and
Stockholm University, P.O. Box 60211, S-104 01 Stockholm, Sweden
(Received 8 October 1992; accepted 22 October 1992)

AbstractAn unified description of the transport theory for charged particles including electrons, ions
and atoms has been formulated and is presented. The main content of the paper includes: the Boltzmann
equation and its different approximations, the theory of electron slowing down spectra, multiple
scattering theory, the theory of deep penetration and the theory ofsecondary particle emission. The scaling
properties of the transport of charged particles are also presented. Although some discussion of
stochastic (Monte Carlo) methods is included, the main focus of this review is on analytical approaches.
Finally, some important unsolved problems in the transport theory for charged particles are discussed.

1. INTRODUCTION
Since the

Bohrs article on

charged particle

penetration was published, numerous brief reviews


on charged particle transport theory have appeared.
Most of them, such as the summary written by
Bethe and Ashkin (1953), mainly summarized the
fundamental interactions of charged particles with
atoms. The study of charged particle penetration was
included, but only as a secondary subject. There are
a few reviews that mainly discussed the development
of transport theories of charged particles. For
example, Birkhoff (1958) and Scott (1963) have
reviewed and summarized the development of electron transport theories, and Winterbon (1975),
Brice (1975) and Ziegler et a!. (1985) formulated the
progress in ion transport theory. However, to the
authors knowledge, there is no unified summary of
the transport theory of charged particles, covering
both electrons and ions,
In the present paper we will review the historical
developments and present status of the transport
theory of charged particles including electrons and
ions. Because the relevant papers published are so
numerous, it is impossible to include all the main
developments of the transport theory for charged
particles in a brief review such as this one. Furthermore, the criteria for selecting material are largely
influenced by the authors knowledge and areas of
interest. Writing such a review may still be worthwhile, since the transport theory of charged particles
has been widely applied to many fields of science,
technology and medicine. In Table 1 we summarize
briefly the areas of applications of transport theory
of electrons and ions. Though incomplete as it is,
the table demonstrates clearly the importance of the
study of charged particle transport. Unfortunately,

the transport theory researchers in different fields are


often isolated from each other. The achievements
won by different researchers are therefore
seldom made use of by others. For this reason we

hope that a unified description of the transport


theory of charged particles can be useful. In the
present paper we attempt to formulate the transport
theory of electron, ion and atom beams in a
unified form. This main reason is that, the collisions
between charged particles and atoms are Coulomb
interactions, and therefore the transport of energetic
electron, ion and atom beams must be similar,
and differ substantially from that of uncharged
beams such as neutrons or y-rays. At the same
time, the authors are confident that the transport
theory of charged particles is a useful tool for the
scientists who are interested in the science and
technology of charged particle beams and radiation
in general. Consequently, a unified introduction to
the research results scattered in various fields will
hopefully be valuable. Our narration will start from
the well-known Boltzmann equation. In accordance
with the nature of the problems, we will treat the
existing theoretical results from the simple to the
complicated in a deductive way. Such a sequence
of formulation agrees roughly with the course of
historical development of the charged particles
transport theory. The deduction for some of the
results will differ from those in the literature to
help maintain the integrity of the unified formulation.
In addition, we also stress the differences between
the transport of charged particles and that of
neutrons or y rays. It appears to the authors that
a successful transport theory of charged particles
inevitably involves a deep understanding and careful
treatment of the influence of the Coulomb interactions upon the transport process, and transport

673

674

Luo ZHENO-MING and Ar~inuasBiwasa


Table I . Areas of application ofcharged particle transport theory

Field of Application
Atomic &
Nuclear Physics
Electron &
Ion Beam Fusion
Radiation Therapy

Bohr (1948)
Scott (1963)

Radiation Protection

~~and

I~~3Electrons~lons1

Energy Range/eV

Reissence

_____________
_______________

Brahme end
Nilison

.~.

~.

Keller (1967)

Electron & Ion


Beam Microscopy
Surface
Surface Analysis

...

.S.

_______________

RadiationDosiinetry
Space Radiation
Shielding
Radiation Processing

_________________

,.

~.

;...

~.

___________________________________

WUIOn end

Dean (1977)

_______________

_____________

Seller (1983)

_______________

i.s.glry (1~s4

________________________________

oec~w(19s4

Ion Implantation and


Modification
Surface Deposition

__________

_____________________________

.
..

.~

Lithography
theory neglect of such influences can usually only
achieve rough results.
2 THE BOLTZMANN EQUATION OF CHARGED
PARTICLE TRANSPORT

Before setting up the Boltzmann equation of


charged particle transport, let us first account for the
simplified interactions between charged particles and
atoms. The elastic scattering cross section of charged
particles in the Coulomb field of atoms is represented
by o~(E,T, cos 0). The inelastic scattering cross section is represented by O~e(E,T, cos 0). Other interaction cross sections, such as the bremsstrahlung
cross section, have been either ignored or considered
as inelastic collision through radiation stopping
power. This simplification is reasonable under many
circumstances such as for heavy charged particles and
low energy electrons. Nevertheless, in some specific
conditions such as in the transport of high energy
electrons, bremsstrahlung becomes increasingly
important, and needs careful consideration. Besides
this, we make two basic approximations as follows:
(1) as far as ions and atoms are concerned, as their
mass is far larger than that of electrons, we therefore
ignore their directional changes in inelastic collisions,
(2) While the inelastic collisions occurs between mcident electrons and atoms, the angular deflection effect
for the incident electrons is included in the elastic
scattering cross section by the correction factor

(1 + l/Z)(1 + c). Here c is the Fano factor (Fano,


1954). In the present paper we regard the transport
of cascade atoms as a spontaneous extension of the
transport of primary charged particles. The so-called
cascade process refers to the fact that, while a
charged particle collides with an atom, it possibly
transfers a part of its kinetic energy to the target atom
or its electrons. The atoms or electrons which have
obtained considerable energy may proceed to collide,
generating the cascade atoms or secondary electrons
of new generations. Such a process repeats itself till
all energy of primary charged particles and cascade
atoms of all generations are exhausted and the cascade process is ended. The cascade process makes it
impossible to deposit all the energy transferred by
charged particles to atoms or electrons at the point
where they interact but release the energy into a wider
region along the motion direction of cascade atoms.
As a consequence of this, the distribution of energy
transfer by primary charged particles is different from

the actual energy deposition distribution. Under certam conditions, for example, if the kinetic energy of
secondary cascade atoms (or electrons) is so low that
penetration of these secondary particles into a region
far away from their emergence point is impossible,
the distribution of energy transfer is very close to that
of energy deposition. The instances where the mass of
incident ions M1 is far larger than the mass of target
atoms M2 or conversely M, 4 M2, both belong to
the above-mentioned situation. The cascade collision

Transport theory of charged partides

occurring at a surface or an interface is particularly


interesting. When it occurs at the surface, the cascade
collisions lead to the sputtering of surface atoms and
the emission of secondary electrons. When it occurs
at the interface, the cascade collision results in the
atomic mixing at the interface, or, may give rise to a
special distribution of the electron energy deposition
in the transition area.

675

Equation
(1) is M
also applicable to electron transport.
1
For electrons,
Considering
the 1=rn,
identityy,~4rn/M2~0,
of electronelectron
but Y2=collision, the electrons with larger energy after each
collisions is considered as incident electrons, then
equation (I) can be rewritten as
U

2.1. The Boltzrnann equation

Vf + (~,,(E)+ E
1(E)Jf(r, a, E)

=N

duf(r, u, E).r1(E, u~u)

J4,i

According to the principle of statistical equilibrium, the Boltzmann equation of primary charged
particles can be written as

+ N,,

rz2

dEf(r, a, E)u1(E, E

E)

+ S(r,u,E)
,LI
=

(6)

dE
JE

duf(r, a, E)

x er1(E, EE, uu)

The transport equations for cascade atoms (or


electrons) are similar to those of the primary charged

ftL2

+ N,

dEf(r, u, E)o~,(E,E E)
Jr

+ S(r, a, E)

(1)

where f(r, a, E) is the differential fluence of charged


particles; f(r, u, E) dii dE represents the number of
charged particles at point r, whose energy is in the
interval from E to E + dE and direction is within
the element of solid angle from a to a + dii, passing
through a unit area perpendicular to a; N is the
number of target atoms per unit volume; N, is the
number of electrons per unit volume; S(r, a, E) dE dii
is the number of charged particles generated per unit
volume around r with energies from E to E + dE and
directions from a to u + dii, and

particles. Only for such a case, the source terms


comprise the following two: (I) displacement atoms
produced directly by incident charged particles;
(2) cascade atoms of a new generation produced after
the collisions between the cascade atoms of last
generation and the atoms of solid. Assuming that the
average displacement energy of atoms is Ed, the
Boltzmannequation for cascade atoms thus becomes
v~+ [I,(E) + ~e(E] f,(r,u, E)
~.

N ~E

E/(l
E2{Eo

V2)

E>E0(l71)
E<E0(ly1)

(2)

E>E0(ly2)
E <E0(l ~2).

(3)

E0

1.

E~

dE .142 duf,(r, a, E)

x a(E, E
+ N,

E/(lyi)
EI{Eo

(7)

E2 {Eo E ~ E0/2
2E E.<E0/2.

)4~

E, a a)

(1 +4m/M2)E

dEf,~r,a, E)u?(E, E

(8)

+ S~(r,u,E)+S,(r,u,E)

Here,

2, V2 = 4M, m/(M
2.
4M, M2/(M1 + M2)
1 + rn)
M,, M
2,rn are the muses of charged particles, target
atoms and electrons, respectively, and E0 is the initial

S,,(r, a, E)

duf(r

dE

U, E)

x a?(E,E+Ed,uu)
I~EoLa

energy of the incident charged particles.

S~(r,u, E) = NJ

E)

(9)

dE

duf~(r,a, E)

~ + La

E + Ed, uu)

(10)

ft
=

dT
.10

dua~,(E,T, ~ 0)

where S, is the source of secondary cascade atoms


generated by primary charged particles and S,, is the
source of cascade atoms generated by the cascade

J4~

t~~
L
=

I
Jo

N,

dTK,(E, T)
ft72E

(4)

dTa1,(E, ~

dTK,,(E, j~,

atoms with higher energy. The superscript (a in a~


and aa in a~represent the elastic collisions between
incident ions and the target atoms, and the collisions
between cascade atoms and target atoms respectively.
Both initial and boundary conditions must be added
to determine the solution of the I3oltzmann equation.

(5)

Generally, the
initial
condition
The boundary
condition
at the
surface
or is
thezero.
interface
is very

676

Luo ZHENG-MING and A~inmsBiwon

important. A general formalism for free boundary


conditions is given here (Davison, 1958)
f(f, a, E) =0 if a a ~ 0

(11)

where, F represents the boundary, a is the outer


normal unit vector to the surface. Under different
conditions the boundary conditions can be of barrier
type, injection type or free surface type.
2.2 The CSDA approximation and the scaling

transformation

and inelastic collision, only a small portion of the


energy of incident particles is transferred to the target
atoms or the electrons of target atoms. The stopping
process for charged particles seems to be the one in
which a pellet is gradually losing its kinetic energy
in the medium and finally stopped. In this case the
average energy loss per unit distance, i.e. the stopping
power, may be applied to describing the energy loss
process for charged particles. Although the physical
picture of CSDA is quite clear and simple, it is not
easy to deduce the SCDA Boltzmann equation from
the original Boltzmann equation in real space. On the
contrary, to make CSDA by applying Fourier transformation in its Fourier image space is straightforward and efficient. We use C1 to represent the
collision term, i.e.
=

dTa,(E,T,uu)

(16)

where p,, is the nuclear stopping power of charged


particles. For inelastic collisions, we can make similar
deductions. Since the angular deflection of incident
particles caused by inelastic collisions is not taken
into account, we have

IEf(l y~)

N,

dEf(r, u, E)ti,(E, E

E)

Now we shall study a very important special case


for equations (1) and (6), i.e. when M1 4 M2 or
M1 ~ M2. Under such conditions, after each elastic

C1(r, a, E)

Jo

dE

JE

duf(r,

(17)
Thus
equation
under the
the Boltzmann
CSDA is given
by for charged particles
fp
uVf~

NjdUCn(E~UU)

x [f(r, a, E) f(r, a, E)] + S(r, a, E)

(18)

where p = p, + p, is the total stopping power.


The Boltzmann equation under CSDA is sometimes
called the Lewis equation (Lewis, 1950). It was first
presented by Lewis to describe the electron transport.
Nowadays this equation is also applied to the transport of light ions in solids comprised of medium or
heavy atoms, or to the transport of heavy ions in low
atomic number media (Luo and Wang, 1987).
To allow a comparative study of the transport
behaviour of charged particles, it is desirable to use
a scaling transformation to equation (18). As a matter
of fact the scaling transformation corresponds to a
normalization of the transport equation of charged
particles with the average path length of charged

E, ~ii)

particles
as the
distance.
doing so,ofit charged
is more
convenient
to unit
compare
the By
transports
particles under different conditions. Let us first

(12)

introduce a single value function of energy: the

Applying Fourier transformation to equation (12)


and exchanging the sequence of the integral

relative residual path length t(E) =


R0 = R(E0)here R(E) is the path length of charged
particles with energy E, then the scaling transform-

E)c,(E, E

~,

J4i~

E,(E)f(r, U, E).

C7 (r, u, E) = N

I e~dE
Jco

du

ation is
ls

J4~i

7IF

Jo

Jt(E) = JoI dE/p(E).R0

dT[f(r, a, E)c,(E, T, uu) e_~T

f(r, U, E)a~,,(E,T, a u)J.

(13)

Taking a Taylor expansion of the exponential term in


equation (13) to 1st order, we have
1N~ dua,(E,u-u)
C7(r,u,E)~i I e~dE

J4*
_____

x (f(r, a, E) f(r, u, E)) +

(14)

then

y=r/R0.

Correspondingly, the distribution function and the


source of charged particles then become
f(y, u, t) =f(r, u, E)pR0
S(y,u,t)=S(r,u,E)pR
0.
(20)
Thus, the transport equation of charged particles
under the CSDA approximation reduces to (Spencer,
1955)

Cf=N

R0N

dii~,(t,uu)
J4~

I dua,,(E,u~u)

[f(y, U, t)
+ S(y, u, t).
X

x [f(r, a, E) f(r, u, E)] +

(19)

(15)

f(y, U, t)]
(21)

Transport theory of charged particles


2.3. The FokkerPlanck approximation to the
Boltzmann equation
Applying the small angle approximation to the
collision term in equation (21), we can obtain
the FokkerPlanck approximation to the CSDA
Boltzmann equation. This results was obtained by
Bethe et a!. (1939).
Taking a as the polar axis. We can write
a = a + dii. We define 9 as the angle between unit
vectory a and u, and ~ is the azimuth of a.
I daI =2 sin 9/2 4 1. dii ~ ~a, + ~,u,. u~,u~are the
two unit vectors laid on the azimuth plane, which
are orthogonal each other. We can make a Taylor
expansion of the distribution function f(r, iii, ~)in
powers of du, we have
&
f (r, u, t) f(r, u, t) ~ ~

8!
+~ ~

____

3x

Thus, the collision term to second order becomes as

Assuming there is a source of particles at a


given point in a infinite homogeneous solid. When
the energy is exhausted, its range distribution is
F(r, a, E), where r is the independent position variable, a and E are the initial direction and energy
of charged particles. When a charged particle starts
moving from its incident point at a given instant,
it will move an infinitely small distance AS a in its
initial direction u. After the infinitely small displacement, it is apparent that the final range distribution
is identical with the original range distribution. Viewing this from another angle, after the infinitely small
displacement, the distribution of charged particles
must consist of two parts, one of which is the range
distribution resulted by charged particles that arrive
at the position AS a without being collided. The
contribution made by them to the final range
distribution is
(1

(22)

AJ

(23)

where

677

EnASEeAS)F(ruAS;u,E).

(26)

The contribution of charged particles that have


undergone a single collision (the dual or higher order
collisions are negligibly small and neglected) to the
range distribution is
r ftyE
ft
ASI NJ dTJ duo~(E,T,uu)F(r;u,ET)

T(E) = R

29.

(24)

0N
duo,(E, cos 9)sin
+ N~J dTa,(E, T)F(r; a, E fl]. (27)
T is the scattering power which is a function of
0
particle energy. The analogy of this basic interaction
quantity with the ordinary stopping power concept The total contribution made by the two parts form
is quite clear. Now, we obtain the FokkerPlanck the
we range
have distribution of charged particles. Therefore,
approximation to the CSDA Boltzmann equation as
follows

a VF = NJ

dT

~+uVf=~A
8:

9f+S(r,u,t).

(25)

The FokkerPlanck equation (25) is very suitable for


description of the transport of heavy ions in low
atomic number media, because the incident ions are
scattered to a narrow cone in a single collision, no
large angle scattering event occurs. This equation has
also often been used to describe the transport of high
energy electrons.
2.4. The backward Boltzrnann equation
Lindhard et a!. (1963) have introduced another

kind of transport equations for the study of range


distributions and radiation damage of charged partides. Such equations are not built on the basis of
statistic equilibrium, but are based on a translational
invariance principle, particularly suitable for calculation of the range distribution of ions in amorphous
solids. We obtain here the backward Boltzmann
transport equation by using the deduction proposed
by Sanders (1968).

dua,,(E, T, u u)

[F(r;u, E

T) F(r; u, E)]

+ N,

dTu,(E, T)
Jo
x [F(r;a, E T) (r; u, E)J.

(28)

The backward Boltzmann equation is particularly


useful to calculate the range distribution of charged
particles in an infinite homogeneous media. It played
an important role during l960s for the development
of the investigation of the ion implantation and the
radiation effects in solids. At that time, the ordinary
transport theory or Monte Carlo method were not
refined enough to solve this problem. The major
restriction of the backward Boltzmann equation lies
in its difficulty to treat the transport problems with
surface and interface. Such problem, however, play a
crucial role in the practical application of transport
theory of charged particles.

678

Luo ZHENO-MING and Aimnas Biwum

3. THE SLOWING-DOWN SPECTRUM OF ELECTRONS


IN AN INFINITE HOMOGENEOUS MEDIUM

From this section onward, we shall start to


formulate the solutions to Boltzmann equation under
different conditions. In the present section, we will
first of all study the simplest situation: a monoenergetic electron source uniformly distributed in an
infinite homogeneous medium. As the distribution
function is independent of the spatial co-ordinates,
the transport equation is degenerated to an integral
equation. The result is the SpencerFano theory of
electron slowing-down spectrum (Spencer and Fano,
1954). Integrating equation (6) over 4n directions,
we have

Thus, we have
04(E)p(E)

(EE

0)

(35)

(/)(E) = l/p(E)
(36)
where p(E) is the electron stopping power. Now lets
consider the contribution of secondary electrons to
slowing-down spectrum. When E <E0/2, the electrons in the energy range 2E ~ E ~ E~are able to
produce secondary electrons with energy E, the
equation (30) can thus be written as

~(E)

Ff2

K,(E, 7) dT

Jo

25

4i(E) =

f f(u, E) du

(29)

dE41(E)Ke(E, E

J4~

and
S(E) =

ftE/2

4i (E)

N,r,(E, T) dT

Jo

!E2

N,

4(E)c,(E, E

E) dE

5 (E

(30)

E0).

In equation (30), the accumulation of secondary


electrons in the energy spectrum is not considered.
We expand the definition domain of equation (30)
and make an analytical extension as follows
4(E)

~b(E) Oi~E~E0
.c(9
E <0 or E > E0

(31)

then equation (30) can be rewritten as

~(E)

r1 +

dEb(E)K,(E, E).

.<

dW

JE

~(E~)K,(E~,
W) dE]/P(E).

(39)

When the Mller formula is applied to K,


equation (39) can be written as

~(E)k(E, E) dE]/P(E)

(40)

J2E

and
2E

JR

q~(E)K,(E,EE)dE

k(E

+(EE

0).

2 [~E_T+2(E+l)2
1
E2T
fl 2mr~NZIl
p

(32)
2E+l

By applying Fourier transformation to equation (32),


we obtain the same basic results, though the method
is different from that of Spencer and Fano

(38)

25

ftE0/2

cb(E) =11 +

E) + S(E) (37)

Note that, while calculating 4(E), the slowing-down


spectrum of electrons in the energy range [2E,E0] is
known already. As a result, by similar calculations
the total slowing-down spectrum of electrons can be
obtained

K,(E,T)dT

JR

dE4(E) eiPE

+ E(E +

1)2 ln

(ET)
T

(41)

If ignoring the interfering terms of the Mller formula, equation (41) becomes
(E, T) = 27tr~NZ
E~
(42)

-~

IE/2

dTK,(E, T)(l

e~T)= e(~L0 (33)

JO

where c,r,(E, T) is the inelastic collision cross section


of electrons. o~,(E,T) may be expressed by the Mller
(1932) cross section.
When
the energy
transfer the
is
2. In
equation
(33), therefore,
fairly contribution
main
small, ;~ l/Toriginates from the first two terms
in the Taylor-expansion of e_~~T.Therefore we have
ii~(

dE4 (E)p(E) eq)E = e~.

(34)

The results of equations (36) and (42) are the


simplified forms of SpencerFano slowing-down
spectrum, which have been used in SpencerAttix
cavity theory (Spencer and Attix, 1955). To test
SpencerFanos
slowing-down
spectrum
theory,
McConnell
et al. (1965)
have measured
the electron
energy spectrum in a cavity. Their results indicate
that the SpencerFano theory and experimental
results coincide well, though the experimental spectrum
is higher than the theoretical spectrum at the
very low
energy range, particularly in high atomic

Transport theory of charged particles


number materials. Figure 1 shows a comparison
between the experimental results (McConnell et a!.
1965) and the SpencerFano theory.
4. THE TRANSPORT OF CHARGED PARTICLES
THROUGH THIN LAYERS

The study of the energy spectrum and angular


distribution of charged particles passing through thin
layers witnessed important developments in 1940s
and 1950s. The study was of great significance for
analysing and explaining the scattering experiments
of fast charged particles and to investigate atomic and
nuclear structure. Although these theoretical studies
had only interest for pure research at that time, they
are still playing an important role even at present,
since they successfully describe the behaviour of the
transport of charged particles in thin layers. Particularly, when the Monte Carlo method is applied to
examine the deep penetration phenomena of charged
particles, the transport rules in thin layers that substitutes for the single collision between charged particles
and atoms is regarded as basic sampling processes,
so that the Monte Carlo method becomes a practical method with modem computers. The valuable
achievements won by Monte Carlo simulation, particularly Monte Carlo simulation for electron transport, therefore, contain the great contributions made
by the transport theory of charged particles in thin

679

tides, R0, and about 100 times larger than the mean
free path for single interactions of the charged partides. In this way, it guarantees the establishment of
a steady state distribution of the charged particles
that have suffered sufficiently many collisions with
the atoms in solids, and ensures that the average
energy of charged particles having penetrated
through this layer changes only slightly, and at the
same time the angular spread is still small. By doing
so~it is possible to simplify the Boltzmann equation
tremendously and to obtain the analytical solutions.
Whats more, the techniques that have been developed in the study of the transport ofcharged particles
in thin layers can also serve as a starting point and
as reference for the deep penetration of charged
particles.
4.1. The L~d~theory
Landau (1944) developed the theory for the energy
spectrum of charged particles passing through thin
layer. By definition, the energy spectrum of charged
particles is given by
#(x, E)

duf(x, p E)

(43)

As charged particles penetrate through thin layers,


their directional change may be disregarded. From
equation (1), therefore, we have

layers.

A thin layer refers to a thickness that is much


smaller than the mean path length of charged par-

a ~

CF

iD

xj 0 K,(E,T)dT+(ERo)(x). (44)
In Landaus paper, K,(E, E E) did not include

10~

the influence of identical particle collision. Extending


the definition domain of ~(x, E) as follows

-I

4(x, E) = f~(x,E) 0 ~ E ~ E
(0
E<OorE>E0.
0

4~~(x,p)jCoo dE*fr(x, E)e~

.~

(45)

(44), we have

io-~

dE~(E)K,(E,E E) b(E)

io~

Applying Fourier transformation to the equation


1
4i(x,E)=
JdP#*(x,P)e~P5.
(46)
~

14

10s

,A

0
0)

5
io~
10Electron
102energy
1o~
io
in eV i~
above bottom
III~

I lid

I 111111

liii

106~
of

III

While applying the Fourier transformation to equation (44), we replace k(E, T) with k(E0, T) and use
the 1~~utliet~1orc.l
cioss section
K (E
0 ~

band
Fig. 1. Flux of electronsconduction
in copper containing
uniformly
distribution MCu. (~) SpencerAttix theory and
SpencerFano theory. The former is broken down into
primary and secondary contributions as shown
The
step in the theoretical primary flux is calculated from
experimental Auger emission data (McConnell eta!., 1965).

Thus, we have

2ivNZr~12
T

(47)

(.. )

().

d~*

(to

_4~*~~~0)j
CITK,(EO, 7)(e~T_ 1)
+ (x) e~.

(48)

680

Luo ZHENO-MING and Aiws Ba*mmem

The integral in equation (48) can be evaluated as


follows

i~

JO dTKe(Eo,T)(lC~T)

;E

800

iPB+IPIJrtolpl
dx
~(1cosx)
0
rtoIpI dx

+iPj

With the Landau assumption pE0

~.

U 700
.~600

,~5OO
U

dTK~(E~,
T)(l

Most probable
total energy loss

1, we have
~.400

Jo

(49)

1(sinxx).

Landau-Synion theory
Experimental points

e~~T)

Mean energy loss

U
0.

oo300

=p{41_~(lnIPIEo_l+cI+IPI~}

(50)

So

Q 200

Maximum

energy
loss
100 _in
a single
collision

~ *(X, p)

eXP{iPE0

ippx

x [i

On

I~ 1E0

1 + C)]_ IPIPx-~}. (51)

By inverse Fourier transformation, we get


B 1 (~
t~(xE)=_._j
exp(xu/2)

8
12
16
20
24
28
Pulse height (arbitrary units)

Fig. 2. Frequency distribution of energy losses of 37MeV


protons
traversing
10 cm(Gooding
argon proportional
counter
pressure
of 1.2 aatoms
and Eisberg,
1957).at a
where

px ir

2 dT.

(to

r E~

px L

ln e

x cos[u(A + ln u)] du

(52)

px (ln px +

(53)

~ c)]

Be

~
12
(54)
2,nv2
Here, B is the stopping number of charged particles,
p is the stopping power, C is the Euler constant,
which is 0.5772. 1 is the mean excitation energy.
Figure 2 shows the energy spectrum of 37 MeV
protons after penetrating through thin layers
(Gooding and Eisberg, 1957). It can be seen that the
experiment confirms the asymmetry characterised by
the Landaus theoretical spectrum. Landaus results
can be interpreted such that the low energy part of the
energy spectrum is constituted by single collisions
with large energy losses and the high energy part is
constituted by multiple coffisions with small energy
losses. As the high energy part of the, spectrum
consists of multiple small energy loss events, according to the central limit theorem of probabilitytheory,
their distribution will tend to be Gaussian, which is
just the results of the Landau theory. Actually, if
e~nin the equation (48) is expanded to 2 orders, the
solution under the FokkerPlanck approximation is
obtained, the energy spectrum of charged particles
under this approximation is a Gauss function
= ~

4o(x, E)

~/2C~xe~~S/~~x

(55)

E0 = E0 px, (I

K(EO, T)T

In a single collision, if the largest energy loss of a


charged particle is much smaller than their energy,
the FokkerPlanck approximation is a good approximarion to the energy spectrum of charged particles.
However, when the largest energy loss approachesthe
energy of the incoming particles, Landaus theory
describes the energy spectrum much better. There are
several developments to the Landau theory (Blunck
and Leisegang, 1950; Blunck and Westphal, 1951)
among which an important one is the work done by
Vavilov (1957). Vavilov has extended the Landau
theory to the condition of arbitrary largest energy
transfer in a single collision, i.e. Vavilovs solution
is therefore suitable to describe the transition from
the FokkerPlanck approximation to the Landau
solution.
4.2. The multiple scattering theory

To understand the scattering behaviour of electrons after penetrating through thin layers, various
multiple scattering theories were developed during
1920s1950s. Wentzel (1922), Williams (1939),
Goudsmit and Saunderson (1940), Moliere (1948),
Snyder and Scott (1949) and Want and Guth (1951)
had developed the multiple scattering theories in
different ways. Among these theories Molieres and
Goudsmit-Saundersons results have been most widely
used. Lewis (1950) and Bethe (1953) theoretically
analysed the inherent relation among these theories.

Transport theory of charged particles


4.2.1. The GoudsmitSawsderson theory. Let us
consider a monoenergetic electron beam penetrating
a thin layer, behind which the energy variation is
disregarded and only its angular distribution is
considered. Then we have
to

f(x,p)=

f(x,p,E)dE.

Jo
Integrating equation (18) over energy, we have
-~=N
3X

J4R

681

Therefore, the final result of Goudsmit-Saunderson


theory is
=
V+i
f(x, p) ~
P,(p)
1-0

xex

(56)

P{_2nr~NxZ(Z+ l)(1 +c)

(E+ 1)2
~ E2(E + 1)2 1(1 + 1)

dua
0(Eo,uu)(f(x,p)f(x,p)J
(57)

+ S(x)5(1 p)/2ir.

Expanding f(x, p) into a series of Legendre ~1,.


nomials, we have
f(x,

~ 21 + 1 P(p)A()

(58)

/...0

X [_~in~
+ l)_E~]}.
(65)
4.2.2. The Moliere theory of small angle multiple
scattering. From the deduction of Goudsmit and
Saunderson (1940), their theory is valid for arbitrary
large scattering angles. The difference between the
Moliere theory and the GoudsmitSaunderson
theory is that the former chiefly investigates the small
angle scattering of charged particles. In the smallangle approximation, sin 8 ~ 0, cos 8
I 82/2 and
simultaneously the upper limit of the angular variable
changes from it to c~.Consequently, equation (57)
becomes
~

Using the addition theorem for the Legendre polynomials, equation (57) becomes
S
1(Eo)A1(x)+ 5(x)

(59)

-~

2nN

8 d0a~(E0,0)[f(x, 8) f(x, 0)]

Jo

and

(66)
According to Bethes derivation of the Moliere
theory, using FourierBessel transformation to the
distribution function and equation (66), one obtains
g(x, v) =
~ dOJ0(vO)f(x, 0)
Jo
+ 5(x)S(02/2)/2n.

S,(E0) = 2nN

I dp[l

oo

21

As a result
f(x,p)= ~

+
4ir

P1(p)Ja,(E0, p). (60)

1
P,(jz)e.

(61)

10

In the work of Goudsmit and Saunderson (1940),


the screening Rutherford formula is used, i.e.
2iur,,(E, p) = 2irr~Z(Z+ 1)(1 + e)
2
I
X E2(E+2)2(1+2~p)2~
(62)
(E+l)

113

After careful calculations, Goudsmit and Saunderson


(1940) obtained
2
(E+l)
S,= 2itr~NZ(Z+ 1)(1 + c)E2(E
+ 2)21(1 + 1~

0(E0,8)[110(vO)]+S(x)

(68)

and

g(x, v)2itN

8d0a

hence
~

\2

I Z\2(E+l)2j
1 / Z
~ _121.25)[l.l33.7~UY)E(E+2)J
1
E(E + 2)~
(63)

(67)

jo

If using the Moliere screening factor o~,and o 4 1,


we have

v dvJ0(vO)g(x, v)

f(x 0) =

g(x,v)=ex

2itNx

0d0a~(E
0,0)

x [I J0(vO)J
).
(69)
From FourierBessel inverse transformation, we
have

vdvl0(vO)

f(x,0)=Jx exp(_2nNx

0 d0a,,(E
0, 0)

~
xE_~on~)_
kI

Jo

(64)

x [I Jo(vO)]).

(70)

Luo ZuE1~o-auNoand Annaas BaAnaem

682

The central problem is now how to estimate the


integral in the exponential term. Let
Q=2nNxj

OdOtr0[1J0(v0)].

Usually, inclusion of the first two or three terms of


the series is sufficient, and the angular distribution of
scattered charged particles is given by

(71)

Introducing Moliere :ymbol


2
= 4nNxr
2
2
(E+1)
0Z(Z + D EE 2\2~

I
Thus

0)0 dO = ~ d~[fb0) +

Here

t~0].

t01(C) = 2 e~2

(72)

(80)
(81)

fw(C)=2e~(Cz~
l)[E,(C2)lnC9

Cl = 2~$ 0 dO(l

2 + 4~)2 (73)
0(vO)]/(0
Divide the integral interval into two parts: [0,k] and
[k, cc], let k satisfy 2,.J~4k 4 ~ then, in the range
[0,k], .1
2v2/4, but
in the range
0(Ov) can
[k, cc],
be approximated
q can be ignored,
as 1 thus
0 we may
have

~ ~(x~v)2[In(x~v)
+ ~+ In 2
/
y2\
= ~~2I
b In -i-)

CJ
(74)

and

dO

~
2

~*

+ ~ Ink

multiple scattering theories were tested in detail


by some experiments (Kulchitsky and Latyshev,
(75~ 1942; Hanson et a!., 1951). The comparisons showed
that they are in good agreement with experimental
2C, C = 0.5772. results. A typical comparison between the Moliere
theory and the experimental results is shown in
Fig. 3.
J

where x~v= y, b =ln(.Zc/Xa)2 + I


Let O/x. = A, then

AdA

y dyJ
0(Ay)

2\l
xexpJ~y2fb+ln)J.
(76)
I
I
y
L \
Defining a new variable u = B~~y,
,~= B22, then B
should satisfy a transcendental equation

where E,(x) is the integral exponential function.


Gauss distribution, representing the
effect of multiple small-angle scattering, fw(C)
represents the large angle Rutherford scattering.
The higher order terms express the correction of
the angular distribution resulting from multiple
large angle scattering. The Moliere theory has
rather skilfully merged multiple small angle scattering and single large angle scattering into a single
theory, which is superior to other small angle

f~(C)
is the

_1 ln 4

dO =

(82)

improved by Nigam et a!. (1959). The small-angle

LJo (0 + 4i)

f(x, 8)0

2(1 2 e~2)

multiple scattering theories. The Moliere theory was

r
ln L~ limi I
=

In B

10_i

_______

18.66mg/cm2Au

~ io~..M~
37.28

~.~

mg/cm2 Au

then

Single scattering

~d~$

f(x, 0)0 dO =

u duJo(uC)
u2

x exp(_

u2

u2

7+~.j In7).

(77)

B is a larger number of the order of 10, so the

10

._~-.S..

Gaussian
10~ -

function in the integral can be expanded into a series


of the reciprocal of B, thus we have
f(x, 8)0 dO

~d~(~ f(N)/B0)

~j1

(.0

,~
0.1

(78)

1tJ-~
0

Oi~U~1

2\Iu2 /u2\l~
x expi/ fl
u ln{ ) I
\ 4/L4
.

\4JJ

(79)

I
5

I
10

I
15

I
20

I
25

I
30

Scattering angle in degrees


Fig. 3. Angular distribution of electrons from thick and
thin
foils from
00 to 30.
()
Represents
the
theorygold
ofMoliere
extrapolated
through
the region
where his
small
angle
and
large
angle
approximations
give
different
values. At small angles (--.) represent the continuation of
the gaussians. At large angles,
represent the single
scattering contribution (Hanson et a!., 1951).
(---)

Transport

theory of charged particles

5. THE PROBLEM OF DEEP PENETRATION

From a practical point of view, the problem of the


deep penetration of charged particles is of common
interest for all areas of science and technology of
charged particle beams. Compared with thin layer
transport, the problem of deep penetration of
charged particles is far more difficult. The unsolved
transport problems of charged particles are mainly
relevant to the deep penetration of charged particles,
which therefore is a developing area of charged
particle transport theory.
The deep penetration refers to the depths that are
comparable with the range of charged particles,
Under this condition, due to elastic and inelastic
collisions of charged particles with target atoms, their
energy and direction at various depth have then
changed considerably. Hence, the basic assumptions
leading to the transport theory in thin layer are no
longer valid. On the other hand, though the energy
spectra of charged particles, particularly the angular
distribution of charged particles at the certain depth
of penetration appear the diffusion behaviour, the
global transport behaviour, as pointed out by
Spencer, is quite different from diffusion behaviour.
Owing to the fact that small angle scattering and
small energy transfer collisions occupy a dominant
role in Coulomb collisions, the successful methods in
the neutron transport and y ray transport, such as
spherical harmonic approximation, discrete ordinate
method or S~method are not very suitable for
calculating the transport of charged particles. This
shows that new theoretical tools able to more
accurately reflect the features of multiple Coulomb
interactions must be found.
5.1. The moment method

The moment method developed by Spencer (1955)


may be regarded as the first successful theory treating
the deep penetration problem of charged particles.
Before that, the moment method was successfully
used by Spencer and Fano (1951) for studying the
neutron transport in hydrogen-containing media as
well as the deep penetration theory of X- and y rays.
Generally speaking, it is not a very difficult job to
determine the spatial moments of the distribution
function of charged particles by means of the
Boltzmann equation. However, it is rather difficult to
recover the transport quantities, such as electron
energy deposition, by using the first several spatial
moments. If the general characteristics of the
spatial distribution of those transport quantities are
unknown, especially their asymptotic behaviour at
large depth, it would usually be impossible to construct reliably and precisely the spatial distribution
of those transport quantities from the first few
moments. It is easy to obtain the asymptotic
behaviour of the spatial distribution of neutron and
y rays in the deep penetration region. It is, however, far less evident for the spatial distribution of

683

charged particles, particuarly for the asymptotic


behaviour at large depth. The success of Spencers
theory lies just in his successful application ofWicks
method to determine the asymptotic trend of the
distribution function of electrons at large depths
and finding the form of basic functions which can
meet the requirements of asymptotic behaviour, so
as to create necessary conditions for constructing
the energy deposition distribution of electrons from
the first few moments.
5.1.1. The energy deposition of electrons. Assume
there is a monoenergetic and monodirectional electron source uniformly incident into an infinite plane
of a homogeneous medium. The direction of the
source electrons is perpendicular to the plane. Let the
plane of this source be the plane of x =0, from
equation (21), we have

of
~

Of

R.~NI d~a0(t,u-u)

+~

.140

x [f(x, p, t) f(x, p, t)]


+

-~--

~~~it)S(l

(83)

p).

2ir
Expanding f(x, p, t) into a Legendre polynomial
series
21 + 1 P1~)N,(x,t).
(84)
f(x, p. t) = ~
.0

Applying the addition theorem and the orthogonal theorem of L.egendre polynomials. From the
equation (83), we have
~

ON, +21+I l[~+ l)~Lt~~


+13N~~1
Ox
+ S,(t)N,(x, t) = S(x)S(l

t)

(85)

where S,(t) are the scaled scattering coefficients,


whose difference from equation (60) is that there is
one more factor R0, i.e.
S,(t) = 2itNR0

~0I

p)[l

P,(p)] dp.

(86)

~,-

Here o~(t,p) is the elastic scattering cross section


for electrons. In Spencers calculation, the effect of
relativity and the screening of electrons outside nuclei
upon the nuclear electric field have already been
taken into consideration in the elastic scattering cross
section. Spencer (1955) has shown that S,(t) can be
approximately represented as
______

S,(t) =

t(t + o~

(87)

This approximate formula has proved its usefulness


within the energy range from 10 keY to 10MeV.

Luo ZHENG-MINO and ANDeRS Ba*nass

684

Ir(t) is an integral of the spatial moment 17(z).


By the definition
17(t) =

= .10
I

to

dxN,(x, t)x,

J-1

their distribution function approaches (l x z).


Therefore the spectrum of the Laplace image for
their distribution function must be very wide. Now,
lets
turn back
equation (85). For the sake of
simplification, we introduce a linear transformation

dt(\t+(LJ
17(t).
/ ~

(88)

After multiplying equation (85) by x[t/(t + a)], then


integrating the equation, we have the recurrence
relation
1-1-1

= 1
{z=t

(93)

~.

Thus equation (85) becomes


ON, ON,
1
+ l)~~+i~~]
Ot~Oy2I+I

.0

2l+l,~0d,+p+i+l
x [(1+

1)J7;lI.~+1+1+

+S,(t)N,(y,t)(1e)(lyt). (94)
The domain for the variable y and t are (0, cc)
and (0, 1) respectively. Making the Laplace transformation for N, and equation (94), we have

l17:~~]

i+l

+oOfla.;d,+P+!+lO+2)P+,+I.

N~(t,p)=fdye_P7N,(y,t)

(89)

Jo

The spatial distribution of energy deposition is


D(x) =

dtp(t)N0(x, t)

(90)

N,(y,

the stopping power can be approximately written as

Thus, we have

1/2

+ A2

\3/2

+ A31
/

~.

dp 8 N~(t,p). (95)

ONf

11+1 N*
N7+21~1N~*_1]
~T~L2l+1 1+I

\1/2

+A1(~~---)
p(t)=Ao(__L)
t+a

t) =

(91)

+S,N7=(l t). (96)


Taking I as a continuous variable, we have

\5/2

Thus, we have

ON*(I,p,t)
Ot

p
~N*(1,p, t)
2
+S,N*(l,p,t)=5(l_t)
~N(l, p, t) = 2N* + ~I ON~
0

_________

dxD(x)x
= A
+2A, j~/2 + A 312+ A
512 (92)
0I~~
21~
317,~
17.1 can be obtained through equation (89). The
problem is now changed to how to get D(x) from
D
0. This problem is relevant to the study of the
asymptotic distribution of energy deposition.
The asymptotic trend in deep penetration for
electron
provides
important
knowledge
about thetransport
electron energy
deposition
which
is a key
factor for the recovery of the electron energy deposition from the first several spatial moments. Before
formulating Spencers result, lets pay attention to the
fact that the deeply penetrating electrons which can
almost reach the end of electron range must have only
suffered the successive collisions with extremely small
angular deflections, therefore their tracks are almost
straight lines and their angular distribution must be
concentratedin a very small cone around the incident
direction. This means that the orders of the spherical
Jl

harmonic moments of distribution function must be


very
high,
the number
of orders
considered
as ahence
continuous
variable.
On the can
otherbehand,
due to their tracks being almost straight lines, the
relation x + t I ~ 0 is valid for them. Other electrons which do not satisfy the relation, do not belong

-~

-~--.

(97)
(98)

Under the small angle approximation, 5, ~ ad!


x) 12. We then use Yangs method (1951) to
solve equation (97). Introducing the following eigen
value problem
2~, 1 O~i1
ad 121/,
p~IO + 7
+ ~ + ~)
t(t +

~j

= A(p, t)~fr
(l,p, t). (99)
Equation (99) is the Schrodinger equation for a two
dimensional oscillator, and its eigenvalues are as
follows (Spencer and Coyne, 1962)

(~+ I) [ 2adp
t(t +

]1/2

n =0,

1,2,...

(100)

8ad )l/4 e~2L 2) (101)


0(s
+ a)
2l2. 4(s2) is the Laguerre poly~2
= [2adft(i
a)p]p. t) can be expanded into a series
nomial.
Thus +N*(1,
of the eigenfunction 1/(I, p, 1)
*~(l

0p~
t)

(t

.0

N*(l,p, t) = ~ a
5(p, t)1/i0(l,p, t).

to the deeply penetrating electrons. That is to say,

(102)

Transport theory of charged particles


Using the orthogonality relation for the Laguerre
polynomials, we have
Oct0

.0

~ M,,,,a~(p,t)+2~a0=A0(p)(103)
m-0

and

685

In this way, the largest it is chosen for calculating A.


From the calculated 1.). and D0 -2 parameter A can

be determined from the following formula


1
ln(D0_2/D0)
2.Jn-i-l/4-i-A/l2,.,/n7/4+A/12
~

(111)
Spencer assumes that D(x) consists of an odd
function Dodd and an even function D~, which can
be written respectively as

1/,,,1~-~!!
dl

M,.,,,, =
Jo
00

0Co)=

(104)

hfr(I,p,l)dI.

As a zeroth order approximation, Spencer neglected the cross term involving the coefficients Ma,,

D00(x) = ~ a,F(x, b,)

/1 +(l+a)~

(106)

_~)exP(_~~~)(113)
Dod(x) = ~ ;F(x, fl
1)
a,(~

~.

x exp[_ph12(8ad)h12in~~+~)].

Ax

in equation (103). Thus, we have


a~(p,t) = Yn(P)~~P[Ph/22(2it+ l)(2ad)~
2\]
/1 + (1 + a)~
x ln(
+~
(105)
As mentioned above,p 1, therefore a~( a~),
Thus
N*(I,p, t) ~ v~(p)fr~(I,p,
t)

(112)

x~ I

Ax \

~,)
ex~(~:~).

(114)

we

have

Thus, for the even J~and for odd J0,


respectively
Ea,b7=D0/~, it =0,2,4,...

(115)

From the inverse Laplace transformation, we have


1 IC+i00
dpey0(p)1/0(0,p,t)
2xi Jc1.0
2)] (107)
x ex~[ 2(8a~2ln~~
/1 + +
(1 + a)~

Ea,fl7=D0/co~, n=l,3,5,...

N0(y,t)=
~

When p

~.

1, y

0(p) and *~(0,p,

t)

vary

slowly with ,,,,

(116)

discontinuous
condition
at x = 0 and
the discontinuApart
from that,
satisfy
the
ous condition
of Dwe
(x)also
at xrequire
=0. ByD(x)
doingtoso,
we have
two additional conditions, i.e.

therefore, we have
/ A~
N0Q,t)~y01/i0(0,z)
exp(
~--)
2~/~\
12 \2
A = 8ad(ln ~1 +(l
+ ~).
(108)
+a)

~aibi2=Ad~~t_D dt

_____

The energy deposition at large depths is

J(x) ~

N~(x,t)p(t) dt

Jo

(1

~a,flT =p(t = 1).


(118)
To guarantee the correct asymptotic tendency, we set
=

x)312

1. Here, the definition of w~is


dx(l

Jo

[_Ao1.

(117)

x)~x0cx

A
(119)

(109)

A, D, o~having been calculated, a,, b,, a, and ~, can


be determined by the equations (115), (116), (117),

This is the asymptoticdistribution for electron energy


deposition. Practically, parameter A was regarded as
a unknown constant. To determine the parameter A
in equation (109)., we study the high order spatial
moments, because they reflect the influence of the
asymptotic distribution

(118), so as to determine the distribution of energy


deposition D(x).
Spencers theory has been proved successful after
many experimental tests (Hoffman et al., 1957; GrOn,
1957; Clark et aI., 1955). Generally, in the energy
range from 10 keY to 1 MeV, Spencers theory is
convincing for infinitely homogeneous medium. In
the higher energy range, because the bremsstrahlung
and the energy loss straggling have a large influence
on the electron transport, the Spencer theory can not
be expected to generate accurate results. In the energy
range below 10 keY, the stopping process of electrons
is cons~derablydifferent from the Bethe formula, and
the deviation of the elastic scattering cross section

x exp

X]

1I

dxD (x)x

i-1

1~

~ JoI k(l

exP(i~)xh

dx

1/2
=

k(~) exP{A/2_[44(n + + i)]

RPC 41-4/3--H

1/2

}.(l 10)

686

Luo ZHENO-MING and ANDass BRAHME


3

112

Bipartition model
Moment method

Gums experimental data

~. 2-

~.

oI,

Fig. 4.

-~~~1

\~.l.

Depth in CSDA range


by Spencers (1959) moment method, the bipartition model
(Luo, 1985a,b) and GrOns (1957) experimental data for 32 keV electrons in air.

A comparison among results obtained

from the Rutherford formula is also rather great.


Therefore Spencers theory needs modification so
that it can be applied to the energy range below
10 keY. Figures 4 and 5 show the comparisons
between the calculational results and the experimental data for the energy deposition distribution of
32 keY electrons in air and 100 keV electrons in Al.
It can be clearly seen that the results of Spencers
moment method are in good agreement with the
experimental data given by Grfin (1957) and by
Hoffman et a!. (1957).
5.1.2. The range distribution of ions in solids,
The range distribution of charged particles behind
the surface of a semiconductor or metal is a crucial
problem in the research of surface science and technology, particularly for semiconductor technology.
The calculation of ion range distributions by the
moment method has a fairly long history of development: from the 1960s when Lindhardet al. (1963) first
proposed using the moment method for calculation

of ion ranges, then throughout the developments


made by Sanders (1968), Winterbon et al. (1970),
Brice (1970) and finally the newer calculation of
Littmark and Zeigler (1981). In 1960s, either the
analytical transport theory based on the original
Boltzmann equation or the Monte Carlo calculation
of charged particles transport had not developed to
the stage where the range distribution of ions could
be effectively calculated, so the moment method
for range distribution advanced by Lindhard et al.
(1963) played an important role for the growing ion
implantation technology, Recently, Littmark and
Ziegler (1981) have accomplished extensive calculations of ion range parameters. In the following we
shall present their results.
They assume that an infinitely wide ion beam is
incident into an amorphous solid at the reference
plane x =0. As the largest energy transfer in an
inelastic collision between ion and atom, compared
with ion energy, is very small, the inelastic collision

40

ill

Bipartition model
Moment method

4/

Huffmans data

Depth in CSDA range


Fig. 5. A comparison among results obtained by Spencers (1955) moment method, the bipartition model
(Luo, 1985a,b) and Huffman et a!. (1957) experimental data for 0.1 MeV electrons in aluminium.

Transport theory of charged particles

687

term can be expanded to the first order of energy

equation (127) can be calculated by an iteration

transfer, thus we obtain from the equation (28)


= N ~
OE + ~ X
Jo dT J4x
I duo~,,(E,T, u us)
x [F(x; p, E T) F(x; ,a, E)] (120)
where

procedure. The spatial moments having been calculated, the spatial distribution of ion range can be
similar to that
constructed
fromof spatial
Spencers
moments.
theory. The
But,problem
under the
is
present condition, there is a factor that is specially
advantageous for the construction of range distribution, which is known to people beforehand, the
range distribution of incident ions is roughly

u0(E, T, u us) = a(E, T)~5(uu


cos q

cos q,)/2x (121)

T/E)~+ ~(1 .412/M,)


2
x (T/E)(l TIE)

(1

Gaussian. Therefore, the Edgeworth or Pearson

(122)

the fitting (Baroody,


function is 1965;
oftenHofker
used toetdetermine
expansion
a!., 1975) the
as

0 is the electronic stopping power. The


equations (121) and (122) represent the conservation

spatial distribution of ion range. From the spatial


moment of the distribution function F7(E), the
average projected range R,,, the standard deviation
AR,,, and the kurtos v and skiness fi can all be
calculated. Their definitions are as follows

where p

of energy and momentum in the process of elastic


collision. a(E, T) is the differential cross section of
the elastic collision. In the Littmark and Ziegler
(1981) calculation,
Winterbon
al.s
form (1970) for
elastic
collision cross
sections et
was
used.
Nna2
At_l_m
t
Na
0(E, ~~2[l
.1~(2AtI~m)9h/oT (123)
where ~ is the reduced energy, e = [M
2)E; a is the screening 2f(M,+
radius, M2)]
a=
x (cc/Z,Z2 e 3+ Z~3)2, a
0.8853a
0(Z~y=4M
0 is the
2/yE.
2. ABohr
=3.35,radius,
m=
t=Te
1M2/(M1+M2)
0.233 and q = 0.445.
Expanding F(x, z, E) into a
Legendre polynomial series
21+ 1
F(x; p, E) = ~
P,(p)F,(x; E). (124)
Defining the spatial moment of F(x;
P7(E)=

100

~,

E) as F7(E)

dxF,(x,E)x0.

(125)

AR,=[f
U

1~

U
=

2F(x;E, l)dx]

21+ 1

(x R,)4F(x; E, 1) dx]/(AR,)4.

[F~(E)p0J+ N

jo

(126)

(132)
(133)

(LtR,)(fl + 3)y
(134)

A 2-6)

Expanding F7(E T) into a Taylor series of the


energy transfer T, we have
21

_~R,,) A

x (F7(E)F7(ET)P,(coso)].

(131)

ciP
(xa)
dxbo+bix+b
b0

dTa0(E, T)

(130)

As an example, the ion rangedistribution constructed


by using Pearson IV with 4 moments satisfies the
following differential equation

1
=

(129)

~
00

2x2~)
2(4$ 3y2)

~IF7:I + ~l+ 1)F7;I]

1/2

R,,)3F(x; E, 1) dx]/(AR,)3

J-oo

From equation (120), we can have

(128)

(x R,)

-.0

(x

xF(x; E, 1) dx

.0

00

______________

.0

00

2=

(2~3y
A

b1

218.

(135)
(136)
(137)

A=10$12y

~ [iF,:
1+
+NF7

~yE

(1+ l)F7;,] =

[F7(E)p0J

dTa0(E,T)(1P,(cosq,)]

~ d~nr(_l~IE
dra,,(E, T)
~
~
L-rr
x TP,(cos ]

(127)

This is an integral equation system, which can be used


to calculate spatial moments of the order (it +1)
from the order it. The differential terms in the

Having
solved equation
(132),
the range for
distribution
can
be obtained.
When the
ion reflection
a certain
iontarget combination is weak, the moment method
can give good range distributions. Figure 6 shows a
comparison between the calculated range parameters
R, and AR, given by the moment method and the
experimental results (Ziegler et a!., 1985). Generally,
the moment method can give reasonable range distributions for ions incident into amorphous solids.
However, forthe case of ions incident into crystalline
solids, the channelling effect will strongly influence
the ion transport process in solids. Under this condition the moment method will not be valid. The

Luo ZHENG-MING and

688

Total path length


Mean range
Straggling

ANDmts Ba~ijan

Ion
Target
ZBL

14N7
-

Ni28
Universal
stopping

Symb

Ref

Year

1
2
3
4

848
1183
796
220

5
6
7
8
9

Zi

Ml

Z2

1976
1979
1975
1960

714
7 14
7 14
7 14

28
28
28
28

1351
1373
104
366

1980
1980
1954
1970

7
7
7
7

14
14
15
15

28
28
28
28

235

1965

15

- -

95 S

- -

28./8

s ~S

S_
5-

fi

0_i

001

-,.-

II

IIIII~

101

113111

Ion

102
energy

(keV)

111111

10~

111111

io

Fig. 6. The range of ions in solids is shown both experimentally and theoretically (Ziegler et a!., 1985).
The letter S indicates straggling data which is defined differently in experiments and theory. The theoretical
straggling curve is about 30% too low.

moment method was also used to evaluate the radiation damage profile and ionization profile resulting
from an ion beam. However, the reliableconstruction
of these profiles is still difficult due to lack of good
fitting functions. The Brice method (1970) seems to be
more reliable,

obtained by the moment method. The limitations of


the moment method justifies the need to develop
a highly precise and flexible transport theory of
charged particles. The present paragraph will therefore describe the bipartition model of transport of
charged particles which satisfies most of the requirements mentioned above (Luo, 1976, l9SSa, b; Luo

5.2. The bipartition model of charged particle

and Wang, 1987; Luo and Brahme, 1993).


Two aspects contradictory to each other must be
considered in solving the deep penetration problem of
charged particles: the large angle scattering events of
small probability and the small angle scattering
events of large probability both exert important
influence on the transport of charged particles. The
existing transport theories for neutrons and y rays can
be used to describe the transport process of charged
particles suffering only large angle scattering events.
However, as those theories can not estimate correctly
the influences of small angle multiple scattering, they
can hardly generate precise results for charged partides. On the other hand, as mentioned initially, the
theory of small angle multiple scattering itselfcan not
directly be applied to the deeppenetration calculation
of charged particles, because the influences of large
angle scattering events upon the transport are important too. Moreover, many prerequisites for the small
angle approximation are not fulfilled. By introducing
a bipartition condition, the bipartition model coordinates the two different tendencies, and produces
accurate results. The bipartition model decomposes
the charged particle beam into a forward directed

transport

Although the moment method has been successful


in solving the deep penetration problem of charged
particles, its limitations are still obvious. First, it can
not be applied to the transport problem of charged
particles in solids with surfaces or interfaces, nor can
it be used to calculate the transport of the charged
particles obliquely incident into a solid. Secondly, to
construct the spatial distribution of a transport quantity of charged particles by the first several spatial
moments, a preliminary understanding of the spatial
behaviour of the transport quantities in question is
required. If such an understanding is missing, the
construction of the distribution by moments is not
likely to be successful. Finally, a severe limitation of
the moment method is that many transport quantities
which people are interested in, such as energy spectrum, angular distributions, reflection coefficients,
transmission coefficients, etc. are actually those which
are difficult to get by calculating spatial moments. To
our knowledge, there are no significantly successful
instances in the existing works. This situation shows
that only a part of the transport information can be

Transport

theory of charged particles

charged particle group and a diffusion charged partide group. The angular distribution of the diffusion
group is rather isotropic. When elastic scattering
events occur, their angular distribution becomes even
more isotropic. Therefore scattering does not alter
the diffusion nature of this group. As for theforward
directed charged particle group, when these particles
collide elastically with the atoms, a fraction of them
will be scattered into the large angle direction. Hence
the small angle features of this part of the forward
directed charged particle group will be lost. The
natural solution is thus to treat those rather isotropically scattered charged particles as a source term and
bring them into the diffusion group. The rest of
forward directed charged particles maintain their
small angle features, and remain in the forward
directed charged particle group. The essential step is
to set up the bipartition condition for the collision
term of the Boltzmann equation. In the following, we
shall use the transport of electrons as an example to
illustrate the solution of the bipartition model of
charged particle transport. We start from equation
(83). Following the bipartition model, the distribution function f(x, /4, t) comprises the distribution
function of forward directed electrons f~(x,p, t) and
the distribution function of diffusion electrons
fd(x, p, t), i.e.

(138)

f(x,/4,t)=fs(x,P,t)+fd(x,/4,t).

689

directions are exactly equal to the number of electrons scattered to m + 1 chosen large angle directions
after elastic collisions. Expandingf,,f~
and Sdinto the
Legendre polynomial series, we have
00

f(x, p, t) = ~ (2/ + I )P,(p)A,(x, t)/4n (143)


1-0
.0

fd(X, p, t)

~ (2/ + 1)P,(,p)N,(x, t)/4n (144)

Sd(x, p, t) = 10
E (2/ + l)P,(p)S,(x, t)/4ir. (145)
In this way, the bipartition condition can be
rewritten as
.0

S,(x, t) =

S,(t)A,(x, t)

D,~
S~A~
(x, t),

fm+1

1 =0, 1,.

..

m (146)

where D~is the bipartition coefficient (Luo, l985a,


b). By inserting equation (146) into the Boltzmann
equation of forward directed electrons, since the large
angle scattered electrons have been excluded, the
small angle approximation is tenable. Replacing the
actual directional cosine p by the average directional
cosine p,~,we thus have
,

M,
~

00
~

D,

1.S,.(t)A,.(x,t)

tm+I

Correspondingly, the Boltzmann equation is divided


into
~

+(x)(lt)

l~m

(147)

+(x)(lt)

l>m

(148)

R,~N dua~(t,vu)

~3j~

J4e

x [f3(x,p,t)f~(x,p,t)]

and

Sd(X, p, t)

+-~(x)(lt)(lp) (139)
2x

ff

Is(x1P~t)thth!1/f

jo.Ji
=

t9t

R0N

du~(t,u~u)

J(x, p, t)dx dp.

(149)

+Pa:c

x [fd(x,p, t)

fd(x,

It can be proved that, for A,(l ~ m), we have

i~t)]

+Sd(x,1u,t).
(140)
Assuming the collision term for the forward directed
electrons C~to be
C1,(x, p, t) = R0N

.0

A,(x, t) =

J4e

t) f,(x,

p, t)]

tm+l

D,rA,.(x, 1).

(150)

For A,(l > m), a Fourier transformation of equation


(148) is performed to obtain the following results

f duo~(t,vu)

x [J(x, p,

(141)

then the bipartition condition is


Cf(x,p,,t)=Sd(x,p,,t) i=0, l,2~~m (142)
where p, are the cosines of the chosen large angle
directions, numbered from 1 to m + 1. It is not
advisable to take m too big, 3, 4, 5 are generally
enough. Equation (142) shows that the intensities of
the diffusion electron source in the chosen large angle

.4,(x, t) =
p,(x)

51r x

Pa(t)

L
I S,=

Jo

Pa

dt]e_~4 (151)

S,(t)dt.

(152)

ji

At last, we have
A,(x, t) =

o[x
L

~ ~a(t)dt]cti(x)

(153)

690

and Arinsas BRABME

Li,io Ziuno-MING

ct,(x) =

~ D,,e~
r - m +~

1~m

[e~)

(154)
I > m.

At the other (n + 1)12 outwardly-moving directions,


writing the Boltzmann equation along the p, direction
in a discretized form, we have
0fd(O, p,,

Having obtained A,(x, t), S,(x, t) can be calculated


by equation (146). For the diffusion electrons, since
their angular distribution appears rather isotropic,
P~approximation should be used. Under the p
approximation, from equation (140) we get
ON,
+
Ot

t) =

n, it + 1.

0, p,> 0.

(156)

(157)

0.3

.~

AX

At 1
N,(Ax, t)p, -~j

(160)

here N,(0, t + At) is the calculated value of the last


step spherical harmonic moment at the boundary;.
N,(Ax, t) is the calculated spherical harmonic
moment at the first discrete point under the surface.
Thus, from the equations (159) and (160), the value
of the spherical harmonic moments of diffusion electrons at the boundary surface can be determined
by the calculated boundary spherical harmonic
moments in the last step and the spherical harmonic

Au, Pb

~-c-~
a
I~

~\

~.

Al

o.i~*.\.

5 i03 2

5 i0~2

(159)

~ (2! + 1)P,(J1,)N,(0, t) = ~ (2/ + 1)P,(p,)


- r
/
x N,(o, t + At)( 1 +
AtS,(t)

~:~~

Cu

1022

(158)

The equation (158) is the additional boundary


condition, which can be written as

~I,\

~
~02

S,(t)N,(0, t)P,(ji,), p, <0.

-.0

0.5

0.4

2! + 1

~ (21 + 1)N,(0, t)P,(ji,) = 0, p,>O.


(155)

At the surface, the distribution function of the


diffusion electrons at (it + 1)/2 inwardly-moving
directions should be zero due to the absence of
the incidence of diffusion electrons, i.e.
f(0, p,,

Equation (157) is the Mark condition, which can be


written as

1=0,l,...,n.

1,2,.

t)

ox

To use a numerical method to solve this hnear system


of hyperbolic equations, the initial and the boundary
conditions must be given. The boundary conditions
for a surface or an interface problem are very important, because they determine whether the transport
problem with the boundary can be solved with high
accuracy. It has been proved (Chandrasekhar, 1950)
that, the P~ approximation is equivalent to the
motion of electrons in n + I special directions, whose
directional cosine satisfies
,(p,) =0,

+
X

1 [
ON,, ON,_ ~
I (1+1)+1
I
21+lL
Ox
Ox j

+S,(t)N,(x,t)=S,(x,t),

Ofd(O, p,,

t)

5 i0~2

106 2

s io~2

io

cV
Fig. 7. The dependence ofreflection coefficients on electron energy and the atomic numbers of media. The
solid lines 1, 2, 3, 4 and 5 are the calculated results for C, Al, Cu, Tin and Pb respectively. The dashed
lines 3, 4 and 5 are the interpolation results for Fe, Ag and Au (Luo, l985a,b).

Transport theory of charged particles

bipartition model, which agrees very accurately

1.0

with the experimental results. Figure 9 shows a

nU
S

4.
4)

\\
\\

Pb\ ~

o00

II-

0.5

1.0

Thickness (CSDA range)


Fig. 8. Transmission coefficients as function ofthe thicknes5
of slabs. () The results from the bipartition model (Luo,
l985a,b). (s). (U), (~)Are Seligers (1955) experimental
data for Al, Ag and Pb, respectively
moments calculated at the first discrete point behind
the surface in this step. The initial condition and the
boundary condition at the infinitely far distance are
zero. By so doing we obtain the spherical harmonic
moments of diffusion electrons and determine further
the distribution of their transport quantities insPace.
By adding the contributions of the forward directed
electrons, the distribution of the total transport
quantities is obtained,
The bipartition model of charged particles transport has been proved efficient and useful, which has
shown the following merits:
1. Wide range of application: the bipartition
model can be applied not only to the transport
of electrons, but to the transport of ions and
atom beams; it can describe not only the transport of primary charged particles, but also the
transport of secondary atoms and secondary
electrons,
2. High precision: Figs 4 and 5 also give the energy
deposition distribution of 32 keV electrons in air
and 100 keV electrons in Al achieved by the

i0~

on
U
00

691

:/

io3

102

/
P11111

100

I 1111111

101

I 1111111

102

I 1111111

Hydrogen energy (key)


Fig. 9. Mean projected range as a function of ion energy.
Results given by the bipartition model. (~)Represent ~
data ofDemond eta!.;(U) Marcinkowski eta!.; (0) Ligeon
and Guivarch; (~)Chu et aL (from Luo and Wang, 1987).

results. The reflection coefficients (see Fig. 7)


hydrogen
comparison
ions
of the
in silicon
calculated
withprojected
the
experimental
of
and
the transmission
coefficients
(see
Fig.range
8), the
reflection spectra and angular distributions calculated by the bipartition model conform very
the experimental
results.
3. well
urgewith
flexibility:
the bipartition
model can be
used for normal and oblique incidence. It can
also calculate the transport of charged particles
under the conditions of infinite homogeneous,
semi-infinite homogeneous and multilayered
medium. Recently, an extension of the bipartition model to the high energy electrons
(550 MeV) has been completed considering
bremsstrahlung, energy loss straggling and secondary
electron
transport
and toBrahme,
1993). The
bipartition
model(Luo
is likely
obtain
almost all statistical information of transport,
such as the distribution of energy deposition,
range
bution,distribution,
the reflectionionization,
coefficients,damage
angular distridistributions and energy spectra of the reflected
particles.
4. Very high efficiency: a systematic comparison
shows the computational efficiency for bipartition model is as l0~times higher than that for
Monte Carlo for electron transport (50,000
samplings) or as l0~l0~
times higher than that
for Monte Carlo for ion transport (TRIMS6)
(50,000 samplings).
5.3. The pencil beam solution
The first analytical pencil beam solutions of the
Boltzmann equation were derived by Fermi (1941) in
thelate thirties. One of the early aims was to describe
the development of cosmic ray showers in the atmosphere. Towards the end of the forties Eyges (1948)
managed to account also for energy losses within the
CSDA approximation and Yang (1951) set up the
equation for the path length distribution. During the
last 2 decades the pencil beam solution has received
renewed interest as it can describe the spreading out
of clinical electron beams in a patient during radiation therapy of malignant tumours (Brahme, 1975;
Hogstrom et a!., 1981; Jette, 1983). Figure 10 shows
an energy deposition distribution given by the pencil
beam
enough both
the analytical algorithm.
solution forInterestingly
the more superficial
transport
and
the deep penetration follow similar Gaussian distribution profiles in the first approximation. The deep
diffusion type solution governed by Fermis age
equation by coincidence can therefore be fitted to the
more superficially valid pencil beam equations. This
is the main reason why pencil beam models have been
very useful to describe the transport of high energy
charged particles over about 2/3 of their range. Near
the surface it is assumed that the electrons are

692

Luo ZI4ENG-anNo and ANDERS Biwnia

10

1.0

~0025

005

010

0.25

25

001

10

10(cm)

Fig. 10. A beautiful energy deposition diagram of a practically point monodirectional 10 MeV electron
beam incident on a uniform water phantom. The primary electrons are largely contained inside a sphere
with a radius equal to the continuous slowing down range and centred at a point about half the range
below the point of incidence. The contours outside this range are due to bremsstrahlung generated in the
first few centimetres of their ranges (Brahme, 1985).
scattered mainly over small angles, whereas the deep
penetration rests on quasi-isotropic diffusion.

The general solution of this equation for an incident


Gaussian beam is given by
2(x)y2
2r0 (x)yO~+ r2(x)0~
r 0
exP[_
02(x)r2(x) (rO(x))2
~(x,y, 0~)=~(0,0)
(165)

1~(x)~(x)
(r(x))2J2

L 02(0)r2(0)
Generally, equation (25) can be used to describe
the transport for a pencil beam. Assuming that the
beam is incident along the positive x-axis and y, z are
the other two spatial ordinates. 0~and 0. are the
deflection angles projected on the xy plane and the
xz plane respectively. Thus equation (25) can be
written as follows

~
t3y

(rO(O))2

where 02(x) is the mean square angular spread of the


beam at depth x
02(x) =

02(0) +

T(u) du

(166)

the covariance rO(x) of r and 0 at depth x is


given by

(161)

rt9(x)=rO(0)+02(0)x+ IX(x_u)T(u)du (167)

Assuming Tis a function of the single variable x, and


introducing an integral fluence ~(x, y, z, 0,, Os),

and r2(x) finally, is the mean square radius of the


beam as a function of depth:

~(x, y, z, 0,, 0,)

~
4\~02+0,

j f(x, y, z, 0,, 0,,

t) dt.

(162)

Due to the axial symmetry, the solution to equation


(161) can be separated as follows
(x, y, z, 0~,0,)

~(x,

~,

O,)~(x, 0,) (163)


~,

Jo

r2(x)

,.2(O) +

2r0(0)x + 02(0)x2

Px
(xu)2T(u)du. (168)
Jo
If equation (165) is integrated over all angles the

purely spatial variation of the fluence is obtained:


0) expF_2-l
L p2(x)]

and o~(x,y, 0,) is governed by thefollowing equation


(

Tv32~

(164)

(x,y)

~(o,

rr2(x)]2

(169)

Transport theory of charged particles

The quantity ~(0, 0),.J~(0)is a measure of the


number of incoming particles per unit length in the
direction perpendicular to the y- and x-axes. Under
the above assumptions, the beam will thus remain
Gaussian with a spatial spread given by equation
(169) at all depths. As equation (169) is derived in
the small angle approximation it describes the penetration of the beam quite accurately from the plane
x =0 to about half the electron range (Brahme, 1971)
when the net deflection angles are small. To treat the
deep penetration of the electrons more accurately,
a quite different assumption has to be made, namely
that the electrons have suffered a large number
of collisions and are almost random in direction.
Under this assumption of quasi-diffusion, the fluence
differential in angle is well approximated by:
f(r, u, E) =

-.~

47r

[f

0(r, E) + 3u11(r, E)1

(170)

where f1 (r, E) is the vectorial electron fluence, a


vectorfield which indicates the predominant direction
of motion of the electrons. This equation therefore
implies the assumption of an essentially cosine
distributed electron flow,
By using this approximation, the FokkerPlanck
equation can, after integration over all angles, u, be
rewritten as (Bethe et a!., 1938)

2
i~f0, s=lt.

693

A comparison of this solution which only holds


for the deep penetration of the electrons, with the
small angle approximation equation (169) which is
valid at least over the first half of the electron range,
shows that both solutions result in Gaussian spatial
distributions. This somewhat surprising result is
one of the fundamental reasons for the success of
Gaussian beam models used in most current electron
beam dose planning algorithms. It can also be concluded that it should be possible to generalize the
solution given by equation (169) so it is valid also for
the deep penetration of the beam. Similarly, it should
be possible to generalize equation (173) to be valid at
small depths where the angular spread is far from
isotropic. These facts have important consequences
for the development of generalized beam models for
electron beam dose planning (Lax and Brahme,
1985).
5.4. Discretization of the Boltzmann equation
Discretization of one or several variables in the
Boltzmann equation to simplify the equation into a
system of algebraic equations is a good method to
obtain the numerical solution of the distribution
function directly. Such a method is a direct developnient from the numerical solution to the transport
equation ofneutrons or y rays. Among the algorithms
of this kind there are the discrete ordinate method, S~

(171)
6.0

This equation can be transformed to the form of the


ordinary heat conduction or diffusion equation by
introducing:

5~5

(172)

Perpendicular incident emission

By separating the variables again and restricting


the treatment to the (x, r) plane, equation (171)
reduces to:
2f~
(173)

Isotropic emission
Cosine distributed emission

5.0

2 C ds

of0

4.0

5.

35
3 0

5
-

Is

an equation which also was first used by Fermi


to describe the diffusion of neutrons in a medium.
It is seen that r corresponds to the time in the
usual
diffusion
conduction and is
therefore
often equation
called the for
ageheat
parameter.
For an isotropic or cosine distributed electron
line source representing the electron distribution at a
given depth in the medium, this equation has the
elementary solution:
x21
f
0(x, ~)

4(0,0)

[ti

2.5

~.

2.0

r(o)I
where it is assumed that the source is located at x = 0
and that the value of r depends on the depth x under
consideration.

1.0

o.~

1.5

rL..

.~

1..~

-~

~.

L~
I

(174)

-i
-

L.~
-~~

Depth

I
8

10

(cm)
Fig. 11. Calculated depthdose curves in water at 18 MeV
for difl~erentangular distributions in plane beams. The
smooth curves were obtained by the discrete ordinate
method (Prillinger, 1977) and the histograms by the Monte
Carlo method. The ordinate is the energy imparted AE in a
layer of thickness pAx.
in water

Luo ZHENG-MING and ANDsas BRAHME

694

method, the discretized Boltzmann equation and the


phase space time evolution method.
The earliest development of discrete ordinate
method was proposed by Wick and Chandrasekhar,
which performed discrete treatment to the directional
cosine variance in the neutron transport equation
with plane symmetry only (Davison, 1959). Later, the
discretization was extended to the position coordinates and the energy. When making the discretization
of the directional cosine as suggested by Carlson, the
5,, method is obtained (Carlson, 1955). The S,,
method has been rather successful in the transport of
neutrons and y rays and has been widely used.
Successful works have been reported on the application of the discrete ordinate method to successfully
calculate the energy deposition of a plane isotropic
electron source and monodirectional plane electron
source in a medium, with the excellent results shown
in Fig. 11. Hoffmann et a!. (1978) applied the S,,
program of neutrons and y rays to calculating the
sputtering induced by charged particles bombarding
the surface of solid, and obtained good agreement
with some experimental data (see Fig. 12). Christel
et al. (1980) adopted a discretization of the Boltzmann equation directly to obtain the range distribution of ions and the radiation damage distribution,
Recently, Huizenga and Storchi (1989) used the phase
space time evolution method to calculate the energy
deposition of high energy electrons in water, obtaining results that agree with the Monte Carlo method
(see Fig. 13). The electron energy loss straggling, the
transport of secondary electrons and the influence of
bremsstrahlung have all been put into consideration
in their calculation,
Generally, a fine discretization method of the
Boltzmann equation is likely to achieve precise
results, but its price is to handle meticulously the
small angle part in the collision terms, which leads
to a fine division of the angular variance. As a

S Experimental data
A

Marlowe results

0 ANISN results

ci

Os
~,

1J.

100

200

500

~
~

>.

~
~
~

10

12

Depth

in water (cm)
Depthdose curves of 5, 10, 20 MeV

Fig. 13.
electrons in
water. (---)Monte Carlo data given by Andreo(1985). ()
Numerical data obtained by Huizenga and Storchi (1989).
result, the spatial variance and energy variance must
be carefully divided. All these increase the calculation
time considerably. In addition, as other analytical
theories, the capacity to solve the transport problem
of charged particles in non-homogeneous media
remains a crucial test to the efficiency of the method.
.

5.5. The low order spherical harmonic approximation


Among the analytical theories of this kind are
chiefly the P1 approximation, the B1 approximation
and the age diffusion theory for charged particles. In
the early period of investigating the deep transport
phenomenon of charged particles, these studies were
common and numerous. Even today, they are quite
commonly used. The major advantages of such
methods are their simplicity, ease of grasping and
application, and at the same time they are able to give
a preliminary description of the transport behaviour
of charged particles. By means of them, it is possible
for us to calculate energy deposition, radiation damage, range distribution, reflection and sputtering.
The defect is that the results obtained by such
theones are generally of hmited accuracy, sometimes
.

with large errors. With the ever increasing performance of computers, the importance of these theories
may therefore be reduced.
CAUSED BY CASCADE
COLLISIONS

50

6. THE SURFACE EMISSION

)-
3

1000 2000

5000

10,000

Ion energy (CV)


Fig. 12. Sputtering yield for argon incident on copper.
ANISN is a program compiled by using discrete ordinate
method (Hofilnan et a!., 1978).

The surface emission phenomena caused by


charged particle bombardment mainly include the
secondary electron emission phenomena induced
by electrons, the emission of the surface atoms caused
by charged particles, i.e. the sputtering, and the
emission phenomena of the secondary electrons from
a surface caused by heavy charged particles. In a
broad sense, the light and X-ray emission caused by
charged particles, and the atomic mixing phenomenon caused by charged particles passing through an

Transport theory of charged particles


interface as well as the transport of secondary electrons passing through an interface also belong to
these problems. The sputtering and the emission of
secondary electrons may be among such problems
that have been studied most extensively. As there
exists an important field of application, such as the
particlesurface-interaction (PSI) problem in fusion
technology, a quantitative explanation of the main
phenomena is an important area of development of
contemporary transport theories of charged particles.
6.1. A semi-empirical theory of secondary electron
emission

To date, the theoretical research on the secondary


electron emission remains at stage of a senu-quantitative phenomenological theory, which is related to
the complicated nature of the interaction between low
energy electrons and surfaces of solids. Apart from
a few Monte Carlo calculations, the research of
secondary electron emission as a transport problem
has just begun. Therefore, we shall mainly focus on
some of the most fundamental results of the existing
research on secondary electron emission,
The most influential of the conceptional theoretical
studies of secondary electron emission is the so-called
three-step model (Salow, 1940; Bruining, 1954). The
three-step model is so named because the emission
process of secondary electrons can be divided into
three steps. As the first step, primary electrons and
atoms in the solid interact with each other, exciting
and ionizing the atoms in the solid. In the second
step, secondary electrons produced in the ionization
are transported to surface of the solid. In the third
step, secondary electrons sufficiently energetic to
overcome the surface barrier are emitted into the
vacuum or surrounding gas. The empirical theory of
secondary electron emission is based on this threestep model for the calculation of the yield of
secondary electrons. The yield is defined as the number of secondary electrons emitted from a surface due
to the impingement of an incident electron.
Assuming that n (x) represents the number density
of secondary electrons produced by an incident electrons at a position x under the surface, a simple
estimate is
n(x) = p[E(x)]/c
(175)

695

where R0 is the range of the incident electrons. At the


lower energy, Young (1956) found
~ EO/RO.
(178)
Thus
(1

e~0/A).

From2E~35
the
/D, we
energy
have range
x 10
BD.~

relation

(179)
R0 = 1.15

35.lo2(l
_e_1.15I&4iw) (180)

~ =.

1~E~

The main result of this formula is the existence of


a maximum m for the yield. The experimental observations have also indicated the existence of the
maximum yield of secondary electrons at a certain
incident energy.
6.2. Sigmunds sputtering theory

Sigmund (1969) developed a sputtering theory in


the late sixties, which serves as the foundation of the
modern sputtering theories. Differing from the earlier
researchers, Sigmund based his sputtering theory on
a model of binary collisions of cascade atoms, and
completed the cascade transport theory within the
frame of LSS transport theory. As it is difficult for the
LSS theory to solve the problem of surface transport,
we will mainly discuss Sigmunds physical sputtering
model and a simplified formula for sputtering yield
based on this model. By definition, the sputtering
yield is

the number of sputtered atoms


the number of incident ions

(181)

Let N(E) be the mean number of atoms with energy


greater than E in a cascade initiated by a primary ion
or recoil of initial energy E
0. From the theory of
cascade collision (Sigmund, 1981), we have
N(E)

= ~m

(182)

1. T(E~)is the energy which is


used for creation of recoil atoms. Assuming that
F(E) dE expresses the mean number of atoms recoiling into a given energy interval (E, E + dE) in a
cascade. Then
where p is the stopping power of the primary elec2 E4E
trons with the energy E in the solid, and e is the
F(E) fmT(Eo)/E
0.
(183)
average energy needed to produce a secondary elec- If F~(r;u0, E0) represents the density of the energy
tron. Letf(x) to be the probability of the secondary transferred from an elastic collision of an incident
electrons at the depth x, to reach the surface and be ions of energy E0 and direction u0 in the solid, then
emitted into the vacuum, then the following form is we have
assumed for f(x)
f(x)Be~
(176)
JdVF.i(r;uo~Eo)=T(Eo).
(184)
where B is a constant less than 1. Thus, the total yield
of secondary electrons from the surface is
For low energy cascade atoms, as their ranges are
short, their density of energy transferred from elastic
=
n(x)f(x) th
(177)
collision is actually proportional to the energy
o
deposited by the nuclear collision at the point. Hence

f,,, is a parameter

T(E0)/E E 4 E0.

696

Luo ZEENO-MING and Aimaas BRARME

we have the density distribution of recoil atoms in a


cascade as follows
F(r, ii, E)dE clii dV ~ fmFa(r; E

0, u0)

dEdu
24ir
xdV
E

(185)
where F(r, u, E) dE dii dV is mean number of atoms
recoiling into the energy interval (E, E + dE), with
directions (u, a + dii) in a volume element d v at r in
a cascade. Equation (185) implies that the angular
distribution of recoil atoms is isotropic. Generally
this assumption is reasonable. Of course, the assumption only is a rough approximation for the region
near a surface. Let S be the ion source with energy

11

5.0

3.0

2.0

0.7

r3 1.0

~A~
~

Andersen and Bay (1971)


Sigmund theory

jS

0.5

0.3

//
I

0.1
0

10

I
20

Measuring uncertainty
I
I ~l
I
I
30 40 50
60 70 80
I

I
90

Z
1

E0 and G(E) dE be the mean number of atoms with


energy interval (E, E + dE) for a steady state, After
a time interval ~t, atoms with energy E will lose a
share of their energy AE. Obviously,

AE =p(E)~x=p(E)v(E)z~t.

Fig. 14. Sputtering yields of polycrystalline copper for


various 45 keV ions at perpendicular incidence, normalizer
to Cu~-.Cu selfsputtering. Experimental results from
Andersen and Bay (1981) theoretical curve from Sigmund
(1972).

(186)

Thus, after the time interval &, the atoms with


energy (E, E + i.tE) will not belong to the atoms with
energy higher than E due to energy loss. For a steady
state, these atoms leaving will be compensated by
newly created recoil atoms, thus

For low energy recoil atoms, the electronic stopping


power can be neglected. The total stopping power can
then be taken as the nuclear stopping power
N
mE~.
(192)
(E) ~ p,,(E) =

C
I-rn

G(E)p(E)o(E)At

SN(E)ttt

Thus, the sputtering yield is

or
SEF(E)
VP

r
y=Fd(O,~,EO)R(;m

(187)

Let n(r, a,E) be the steady state atom density.


n(r, a, E) dE du dV is the mean number of atoms
with energy (E, E + dE), in the directions (a, a + dii),
and in the volume element dV at r. From equations
(185) and (187), we have
n(r, a, E)

SFmFd(r a0, E0)

(188)

4irEvp

On the surface, the outward current of recoil atoms


through the surface plane x =0, is
I

j u dE

and Cm are the parameters in the nuclear stopping


power formula. The formula (193) is the main result
of Sigmunds theory. It demonstrates that the sputtering is directly proportional to the energy loss in
nuclear collisions at the surface. If m = 0, the yield is
inversely proportional to the surface barrier. These
results qualitatively coincide with observed sputtering
data. Figure 14 shows a comparison of the sputtering
yields predicted by Sigmund theory with those
from experimental measurements, the agreement is
excellent.

(189)

model
6.3. The sputtering theory based on the bipartition

According to the definition of sputtering yield,


we have

Sigmund has proposed a reasonable model for the


sputtering process. This model has been completely
confirmed by Monte Carlo simulation. Sigmunds
sputtering theory builds itself on the LSS transport

V
A

$2,,

AFd(0; u0, E0)

~2

~
~ F

on(O a E)Ijz dii.

(193)
~m)NCmU~
where N is the number of atoms per unit volume. m

dp

P(E,

~).

(190)

-i

P(E, ii) represents the probability of cascade atoms


leaving the surface. The simplest model for surface
binding ofa solid is based on a planar surface barrier
U. In this case
20 > U
P(E, ~)= P(E, cos 0) = I E
cos2G
< u.
(191)
E cos

theory, which has a slight disadvantage for solving


the sputtering problem, since as Sigmund pointed
out: There is no sputtering without a target surface. However, the LSS theory is not appropriate for
handling the transport problem with surfaces and
interfaces.
Viewing
sputtering
phenomena
from the
point
of
view One
of the
transport
theory,
two calculate
problems
must
be
solved.
is how
accurately
the
distribution function
of to
primary
incident
particles
at the

Transport theory of charged particles


surface, which determines the source intensity of
cascade atoms. The other is how to determine the
boundary conditions for a barrier surface. The bipartition model can solve these problems because it can
furnish the distribution function of the primary jp,~j..
dent particles at the surface. Also, the recent development of the bipartition model has solved the problem
of boundary condition for a barrier surface (Luo,
1988). An outline concerning the application of the
bipartition theory to solve the problem of sputtering
was recently reported at an IAEA symposium in
Vienna (Luo, 1991). Here, we shall given a brief
account of the main results as follows. From the
bipartition model in Section 5.2, we have already
obtained the distribution function of incident charged
particles everywhere in the solid. The cascade atoms
of the first generation are produced by the collision
between charged particles and atoms. The direction
of motion of recoil atoms is generally perpendicular
to the direction of incident particles. Moreover, as
their energy is low, the frequency of collisions is high
and their angular distribution is nearly isotropic.
Thus, it is reasonable to use the F,, approximation.
Expanding the distribution function of cascade atoms
into a series of Legendre polynomials we have:
f(x,

~ 21+ 1 P

j~,E)

1(p)N7(x, E)

(194)

In this way, equation (8), (9) and (10) become


1
+21 + ~

apN~

+NJ
x

+N

/ M, + M2 IM, E + Ed\
2M1

Sj

M2

~5d- Ed dEN~(x,E)o~(E,E

+ Ed)

~ + + Ed)
~ p(\/E

(195)

a,,

NSE dTa~,(E,r)T

f~(0,~ E) ~

pj,

E)

0< p, < ~

(197)

When
has other
values, the
boundary
condition
is
still thep,same
as equations
(159),
(160). Now
we may
have
N1(0, E),..., at the boundary, in
whichN~(0,E),
the sputtering
yield is
lEo
=

Jo

N~(0,E) dE

(198)

The other relevant sputtering quantities, such as


angular distribution and energy spectrum for
sputtered particles, can also be obtained from the
distribution function for the cascade atoms.
7.

THE MONTE CARLO METHOD

The Monte Carlo method has become one of the


most successful methods in the study of charged
particle transport during the past 30 years with the
rapid development of computer performance. The
further development of computers will make it possible for the Molecular Dynamics (MD) Simulation
technique to become a powerful and applicable
means to study the multibody collision processes of
charged particles.
The advantages of the Monte Carlo method in the
calculation of charged particle transport are due to

can obtain almost all the transport quantities one


is interested in, such as the distribution of energy
deposition, radiation damage, range distribution,

dE[N,(x, E) + A,(x, E)l

(E, E + Ed)P, ~,,

quently invalid. Thus, for the case of a planar barrier,


we have the boundary condition from 0 < <p,, as
follows

several factors: first, it is flexible and can deal with the


transport problems in complicated situations which
many
other
analytical
to the
address.
Modern
Monte
Carlo theories
methodsfind
treathard
either
electron transport, or the ion and atom transport. They

+ ~ 3N~+1+ 13AT~_t
~x j
k 1
S7(E)N~(x,E) + i~2 ~ (co,N7)
~

697

(196)

The influence of the surface barrier on therecoil atom


transport can be expressed by using a critical angle
0~:cos 9,, =
= (U
2 which is dependent on the
0/E)atoms. When 0< u, < it,,, the
energy E of cascade
Mark condition (160) will change, because, in this
angular interval, the cascade atoms in the angular
interval it,, < p, <0 have been reflected back by the
surface barrier, and the Mark condition is conse-

reflection
transmission
coefficients.
They
also
handle theand
transport
of charged
particles
incan
infinite,
semi-infinite
and multilayered
materials.
Principally,
media,
on such
calculations
Monte though
charged
Carlo
particles
ispublications
capable
in irregular
of treating
and the
inhomogeneous
transportare
of
still rare. Secondly, the Monte Carlo method has
a high precision, as long as the speed of computer
is high enough, the interaction between charged
particles and atoms can be carefully considered
principally in the Monte Carlo calculation, and the
sampled trajectories are sufficiently numerous to
minimize errors. Thirdly, the Monte Carlo method is
easy to grasp. As long as the interaction processes
between charged particles and atoms or the solid are
known, the difficulties of preparing Monte Carlo
programs will be reduced to effectively realizing
statistical sampling of relevant interaction processes.
The development of analytical theories requires that
the researcher has a profound knowledge oftransport
theory. Therefore, the Monte Carlo technique is

698

Luo ZHENG-MING and A~maasBRARME

conceptually simpler for application to experimental


research. The Monte Carlo programs simulating the
transport of charged particles are basically of two
kinds. The first is the ordinary statistical sampling
of the elastic scattering of bindary collisions as its
basic process, which is suitable for the transport
calculation when the number of elastic collisions in a
penetration history from start to rest are below about
10. Monte Carlo programs of this kind are the
Monte Carlo codes of ion transport, such as TRIM
(Biersack and Haggmark, 1980), MARLOWE
(Robinson and Torrens, 1974), ACAT (Takeuch and
Yamamura, 1983), TCIS (Cui and Li, 1985) programs, etc. The Monte Carlo transport program of
electrons below 10keV belongs also to this kind,
which are often called Detailed history technique
in literature. Some others are generally called condensed history techniques and are suitable to such
transport problems where the elastic collision events
are about i0~i0~
collisions before an incoming partide come to rest. If the microscopic technique is
applied in this case, the calculation efficiency is too
low. To overcome such difficulties, the path length of
charged particles is divided into segments. At the end
of each segment, the small-angle multiple scattering
theory is used to predict its angular distribution, and
the Vavilov theory or Landau theory is applied to
calculating its energy spectrum. For every segment
penetrated, the energy and angle are sampled. All
segments are sampled one after the other in the same
way until a certain cut off energy is reached. The
penetration history of an incident particle is thus
ended. Such a technique is usually applied to the
transport of high energy electrons. ETRAN (Berger
and Seltzer, 1968), EGS (Nelson et al., 1985), ITS
(Halbleib and Vandevender, 1974), etc. belong to this
kind. Figures 1518 show some successful calculated
results using Monte Carlo method.
3.0 MCV

~j

e4

V
U
V

~\

polystyrene

2~fr

I
0.4

y
~
2
o
~

Water

Comparison

~,Air

~luminium%~

o
.~
1
~,

.n

0
.0

.~0

10

12

14

Depth (cm)
Fig. 16. Depth dose distributions from a 20 MeV beam of
electrons incident on a water phantom with air and Al
cylinder at 2mm depth. (0) Represents homogeneous
phantom data; (X) the case with the Al cylinder; (A) the
case
the air cylinder. The histograms are EGS4 calculated with
results
for a point source of 20 MeV electrons. The
experimental data are all normalized to the calculations via
one point on the homogeneous curve (Rogers and Bielajew,
1988).
8. SCALING PROPERTIES

A recent significant but somewhat unexpected


result of the transport theory of charged particles is
the discovery of extensive scaling properties possessed by the transport behaviour ofcharged particles
(Luo, 1990). The term scaling in this paper also
means normalization. At a first glance, charged partides move in a 6 dimensional phase space. Incident
particles and solid target atoms are very different.
The regularities of the transport behaviour ofcharged
particles are hard to see. However, some simple
compact and inherent laws do actually exist in the
transport behaviour of charged particles. The socalled scaling property of charged particle transport
defined here refers to the existence of such parameters, called scaling parameters. Once they are

determined, the transport behaviour of charged par-

ylon

tides
of
target
charged
~Satoms.
mainly
particles,
determined,
this can
their
way,
energies
independent
the
and the
ofbyof
the
species
species
of
particles
can
beIn
scaled,
and
the transport
transport
behaviour
of
charged
particles
be
determined
acharged
simple
parameter. We have found that the scaled transport
cross section is the most fundamental scaling parameter. According to its definition, the scaled transport cross section is

0.8

1.2

1.6

2)

15.

mm

z (g/cm
Fig.

20 MeV. 2

:-

Mylar

Polystyrene

of calculated

and

measured

depth-dose curves from 3MeV electrons in polystyrene. The


histograms were calculated with the ETRAN code. The
curves are from a moment method calculation of Spencer
done in the continuous slowing down approximation. The
points represent measurements with radiochromic dye ~
dosimeters by McLaughlin and Hussmann (Berger, 1988).

where

2nNRo$,diwn(Eo~!z)E1~J4]

(199)

0 is the average path length of charged


particles. o~,has a simple physical meaning: under the
constant cross section approximation, ~ represents
the effective collision number experienced during the

period from a charged particle initial state up to its


exhaustion of all energy. An effective collision refers

Transport theory of charged particles


1.0

0.8

699

+,~r

0
0
0

0.6

,~t
0.4
0

0.2

0
.0

I
0

I
0.4

0.2

I
0.6

0.8

1.0

1.2

z/R~
Fig. 17. Central axis relative depth absorbed distributions obtained Andreo and Brahme (1984),
and Berger and Seltzer MC data for 20 MeV electrons, and experimental results of Harder and
Schulz for 21,2MeV in Lucite (+;
and Brahme and Svensson (1979) for 22.5 MeV in
water (0;
)

).

to a change of the average direction of a charged


particle from its initial direction to the perpendicular
direction after certain collisions. It is evident that the
larger a~is, the more effective are the collisions of the
charged particle, and the faster it reaches a diffusion
state.
The study of scaling properties starts first from
empirical formulae of transport quantities of charged

Ion I Caic. Mess,


H
+

DIXIS
4Hc I I
Ne (
0
Ar I i 0
Xe

10

Ni

~ 10

~ io~

particles. Empirical
mentioned
penetration
the
phenomena.
remarkable
formulae
Luoscaling
(1967)
of of
pointed
thethe reflection
electron
out the

e
~

13

iii

10

coefficient
tides
are often
and the
needed.
range In
parameters
the studyofoncharged
the scaling
parproperties for electron transport, Spencer (1959)

Ar
Ne

1
102

I
10~

iii

iii

uI
~

I ii

ii

10~

E0, incident energy (eV)


Fig. 18. Comparison of experimental and calculated values
of the sputtering yield Yversus incident energy E~for H, D,
He, Ne, Ar and Xe on Ni at normal incidence. Lines connect
the calculated points and are drawn to guide the eye
(Biersack and Eckstein, 1984).

possibility of using Spencers d


1 parameter
for fitting
empirical formulae. The parameter
d1 namely
is the
scaled transport cross section. Later, Harder (1970)
made a systematic parameter analysis for the electron
transport equation. He suggested2/E
the atomic number
Z and another parameter Zm0c 0 as the scaling
parameters for middle energy and for high energy
respectively, and obtained empirical formulae for the
reflection coefficient, the transmission coefficient and
the energy deposition. Lax and Brahine
investigated the scaling property of the FermiEyges solution and found the scaling property for the energy
(1985)

deposition by a pencil beam incident onto a multilayer inhomogeneous medium. Their result is shown

Luo Zmwo-wNo and Ar.waas BaARsm

700

in Fig. 19. For the scaling of ion transport, Weissman


and Sigmund (1973) initiated the use of the reduced
energy as a variance to fit experimental data, because
when the ion energy is low, if electronic stopping
power is ignored, the reflection coefficient will only
rely on the reduced energy a. Even at the low energy
the reflection coefficient is closely relevant to the
electronic stopping power. It has been found that for
the light ions of H, D, T and He for the same a there
is a 34 times (Eckstein, 1984, quoted from Langley,
1984) difference among the particle and the energy
reflection coefficients which is not a satisfactory
result. Therefore, is not an ideal scaling parameter.
Figure 20 shows that the particle reflection as a
function of the scaled transport cross section at
possesses good one to one correspondence. The
scaled transport cross section reflects the similarity
of the transport behaviour of charged particles, and
it is a measure of the transport behaviour of charged
particles. We may predict that, by means of the
basic fact that the distributional function is deterA

s mm

aluminium
Water

8 mm aluminium
100
50 mm
Depth of _~
measurement: 0.73 r0

47 mm

4.~jU

0
C)

n0

01

EU
U

UT
H
He

6
0.01

I
1.0

I I I I

I I

I
10

Scaled transport cross section (ok)


Fig. 20. The dependence of the particle reflection coefficient
on the scaled transport cross section. The Monte Carlo data
are represented by
(~)and (A) (Luo, 1990).

(s),

9. SOME UNSOLVED TRANSPORT PROBLEMS

As the sciences and technologies associated with


electron, ion and atom beanis have presented the ever
more complex transport problems which usually bear
many characteristics of interrelated disciplines it has
caused the transport problem to be a cross disipline
problem, providing impetus to enrich the existing
transport theories of charged particles. Some of
unsolved problems are:
Highly accurate transport calculation for high
energy electron beanis. Modern medical radiation physics has shown that a dose accuracy
better than 5% is often required to minimize
radiation damage. Therefore modern radiation
therapy requires highly accurate transport calculation of electron and X-ray beams (Brahme,

.g

0
-

0
V
.0
V

:~

mined by u~,it is possible to obtain more scaling


formulae.

20 MeV

Water

1o~

B
13.3 MeV

10~

1.0

1985).
io~

Transport calculation of high energy electron


beams in homogeneous media. It is a serious
challenge to the capacity of analytical electron
transport theories to solve this problem. Considering the geometric complexity of human organs
and their great difference in stopping and scatter-

3
Relative radius (r/rj,e)

Fig. 19.
Corresponding profiles for the equivalent
energy 13.3 MeV. The historgram was obtained by interpolation of Monte Carlo data for 10 MeV and 15MeV to the
energy 13.3 MeV in water at the scaled depth 0.7 to
0.75 csda ranges (Lax and Brahme, 1985).
()

ing properties which may be unprecedented in


the area of transport problems, and the requirement for a highly precise calculation of electron
beam dose distributions, it is a very challenging
research subject.
The transport theory of charged particles in
crystals. Some experimental data show that the
range distribution of charged particles in crystals
is quite different from that in amorphous
materials (Davies et a!., 1968). What can be

Transport theory of charged particles


imagined is that the radiation damage and
ionization distribution of charged particles in
monocrystalline media will also differ obviously
from the corresponding distributions in amorphous materials. It seems that existing channelling theories can not easily explain such
problems [110].
The analytical sputtering theories, such as the
Sigmund theory, though winning considerable
successes as a theoretical model, still leave
some problems unsolved in the transport calculation, such as how to correctly set up the
barrier boundary conditions; how to describe
the transport of the cascade atoms; what is
the influence of oblique incidence. Before these
problems are finally solved, the quantitative
analytical sputtering theory can not be
considered as fully established.
There is a history of several decades of the
study on the secondary electron emission from
surfaces by charged particles (including electrons
and ions). The existing theories fundamentally
belong to the phenomenological and semiquantitative type. A consistent quantitative analytical
theory based on least physical hypotheses and
strict transport calculation is still needed.
The transport of low energy electrons in solids,
The transport calculationof electrons within the
energy range from several eV to several keY are
still rarely seen, with the exception of a few
Monte Carlo calculations. The main problem is
that at such low energies, the interaction cross
section data are very uncertain.
The problem of electric charge state of ions.
The distinctions between a charged ion and
an uncharged atom with the same incident
energy in the transport behaviour in the solid
have not been considered by present transport
theories. Nevertheless, as judged from physical
intuition, there must be some considerable
differences between their transport behaviour.
What acutally happens to the electric charge
state of a charged particle after it enters a solid
is a complex problem and it is little known at
present.
Binary collisions and multibody simultaneous
collisions. For this problem there is only MD
(molecular dynamics) simulation and a reliable
analytical theory is needed.
The interaction between extremely low energy
ions and surface. Experimental evidence shows
that, under this condition, neither the Monte
Carlo method nor existing analytical theories
can given the correct description of transport
behaviour (Winters et a!., 1990; Coufal a a!.,
1991). In fact, the range of ions of several tens
of eV is only a few
Therefore the surface
electric field and local crystal structure must
have a strong influence. As in secondary particle
emission, this is a cross discipline problem
A.

RPC 41.4/5I

701

between the surface physics and charged particle


transport theory.
Both analytical theories and Monte Carlo
calculations require more precise data about the
interaction between charged particles and atoms.
An incomplete preliminary list shows that we
need improved accuracy of: the elastic scattering
cross sections of electrons with atoms, the inelastic collision cross section, the stopping power
and the energy loss straggling data of low
energy electrons, the differential cross section of
bremsstrahlung and the energy spectnun and
angular distribution of recoil electrons induced
by emitted bremsstrahlung, the inelastic collision
cross section, stopping power and the energy loss
straggling of energetic ions in the solid, and
elastic collision cross section of ions with atoms.
10. CONCLUDING REMARKS

When comparing the developmental process of


the transport theory for charged particles with that of
neutrons, large difference are seen. For neutrons,
particularly since the fission reaction was discovered,
the transport theories have rapidly been developed.
Through 1940s and 1950s, the transport theory for
neutrons rapidly matured. The powerful driving force
to this rapid development was the use of nuclear
energy production. And, though the discovery of
charged particles was an event at the beginning of the
century, there was no such period of rapid developmeat as for the neutron transport, starting from
Wentzels research as the beginning of the study on
the transport of charged particles. Yet it is actually in
a state of continuing exploration and development.
The impetus to the development of modern transport
theory of charged particles comes mainly from
the following sciences and technologies: radiation
therapy, ion implantation and modification, and surface analysis. The future driving force for the development originates from an interdisciplinary study
between atomic physics, nuclear physics, solid state
physics, surface physics and medical radiation
physics.
It seems that a coordinative developments between
MC, MD and the analytical transport theory would
be desirable, since each of them has its own merits
and demerits. The development ofmodern computers
makes it possible for MC and MD to overcome the
disadvantages brought by their low calculation
effIciency. On the other hand, the limitations of the
existing analytical transport theories will also be
removed. Contemporary development of analytical
theories breeds the possibility of overcoming these
limitations completely. Some analytical transport
theories and calculationmethods of charged particles
which are very flexible, precise enough and highly
effective will hopefully be developed and effectively
applied to the physical and medical sciences and
to technology during the last decade of this century.

702

Luo ZHENG-MING and Ai~ceasBa.ussni


REFERENCES

Christel L. A., GibbQns J. F. and Mylroie S. (1980)


An. application of the Boltmmnn transport equation to

Andersen H. H. and Bay H. L. (1981) Sputtering by Particle


Bombardment, Vol. 1, p. 145. Springer, Berlin.
Andreo P. and Brahme A. (1984) Restricted energy loss

ion range and damage distribution in multilayered


targets. J. Appi. Phys. 51, 6176.
Clark R. K. Brat S. S. and Marinell L. D. (1955) Ionization
of air by beta rays from point sources. Radiology 64, 94.
Cosselett V. E. and Thomas R. N. (1964) Multiple scattering
of 530 keV electrons in evaporated metal films. I. Total
transmissions and angular distribution. fr. .1. APP!. Phys.

straggling and multiple scattering of electrons in mixed


Monte Carlo procedures. Radiat. Res.
16.
Baroody J. (1965) Influence of anisotropic scattering on

too,

stopping by elastic.collisions. Appl. Phys. ~, 3565.


Behrisch R. (1981) Sputtering by Particle Bombardment
(Edited by Behrisch R.), Vol. 47, p. 1. Springer, Berlin.

Berger M. 3. (1968) ETRAN Monte Carlo code system for


electron and photon transport through extended media,
CCCIO7, Oak Ridge Nat. Lab. Oak Ridge. Tenn.
Berger M. J. (1988) ETRANExperimental Benchmark,
Monte Carlo Transport of Electrons and photons (Edited
by Jenkins T. M., Nelson W. ft. and Rindi A.),
pp. 183219. Plenum Press, New York.
Bethe H. A. (1953) Molieres theory of multiple scattering,
Phys. Rev. 89, 1256.
Bethe H. and Ashkin J. (1953) Passage of radiation through
matter. Experimental Nuclear Physics Vol. 3 (Edited by
E. Segre). Wiley, New York.
Bethe H. A., Rose M. E. and Smith L. P. (1938) The
multiple scattering of electrons. Proc. Am. Phil. Soc. 78,
573.

Can. J. Phys. 46, 573.

Davies J. A., McIntyre 3. D., Cushing R. L. and Lounsbury


M. (1960) The range of alkali ions of kiloelectron volt
energies in Al. Can. J. Chem. 38, 1575.
Davison B. (1958) Neutron Transport Theory. Oxford
University Press, London.

Biersack J. P. and Eckstein W. (1984) Sputtering studies


with the Monte Carlo program TRIM.SP. Appi. Phys.
A34, 73.

Biersack J. P. and Haggmark L. G. (1980) A Monte Carlo


computer program for the transport of energetic ions in
amorphous targets. Nucl. Instrum. Meth. 174, 257.
Birkhoff ft. D. (1958) The passage of fast electrons through
matter. Handbuch der Physik (Edited by Flgge), Vol. 34,
p. 53. Springer, Berlin.
Blunck 0 and Leisegang S. (1950) Zur energieverlust
schneller elektronen in dOnnen schichten. Z. Phys.

~,

500.
Blunck 0. and Westphal K. (1951) Zur energieverlust
energiereicher elektronen in dnnen schichten. Z. Phys.
130, 641.
Bohr N. (1948) Penetration of atomic particles through
matter. Kg!. Dansk. Videnskab. Selskab Mat. Fys. Medd.
18, 8.
Bothe W. (1949) Einige einfache uberlegungen zur rckdiffusion schneller elektronen. Ann. Phys. 6, 6.
Brahme A. (1971) Multiple scatteringofrelativistic electrons

in air. TRITA-EPP-7l-22.
Brahme A. (1975) Simple relations for the penetration of

high energy electron beams in matter. National Institute


of Radiation Protection, Stockholm, report SSI (1975011).
Brahme A. (1985) Current algorithms for computed electron
beam dose planning. P.adio:her. Oncol. 3, 347.
Brabme A. and Nilsson B. (1984) Limitation of pencil
beam algorithms in electron beam dose planning.
Proc. 8th ICC.R Toronto, p. 157. IEEE Computer Society
Press, Silverspring.
Brhme A. and Svensson H. (1979) Radiation beam characteristics of a 22 microtron. Acta Radio!. Oncol. 18,
1244.

Brice D. K.

15, 883.

Coufal H., Winters H. F., Bay H. L. and Eckstein W. (1991)


Energy transfer from noble gas ions to surfaces: collisions
with carbon, silicon, copper, silver and gold in the range
1004000eV. Phys. Rev. 844, 4747.
Cui F. and Li H. (1985) Recoil implantation and interface
mixing: A computer simulation study. Nuci. Instrum.
Meth. 87/8, 650.
Davies J. A., Eriksson L. and Whitton J. L. (1968) Range
measurements in oriented tungsten single crystal III.
The influence of temperature on the maximum range.

(1970) Ranges of energetic ions in

matter.

App!. Phys. Len. 16, 1039.


Bruining H. (1954) Physics and Application of Secondary
Electron Emission. Pergamon Press, Oxford.
Cardans M. C. and Zucker M. S. (1971) A method for
solving time-dependent electron transport. Nuci. Sci.

Engng 45, 107.

Carlson B. 0. (1955) Solution ofthe Transport Equation by


Sn Approximation, LA-l89l, Los Alamos Scientifical
Laboratory.
Chandrasekhar S. (1950) Radiative Transfer. Oxford
University Press, London.

Dekker A. J. (1958) Secondary

electron emission.

Solid State Phys. 6, 251.


Byges L. (1948) Multiple scattering with energy loss.
Phys. Rev. 74, 534.
Fano U. (1954) Inelastic collisions and the Moliere theory
of multiple scattering. Phys. Rev. 93, 117.
Fermi E. (1941) (cited by Rossi B. and Greisen K.)
Cosmic ray theory. Rev. Med. Phys. 13, 265,
Frank H. (1959) Z. Naturf. 14*, 247.

Gooding T. J. and Eisberg ft. M. (1957) Statistical fluctuations in energy losses of37 MeV protons. Phys. Rev. 105,
357.

Goudsmit S. and Saunderson J. L. (1940) Multiple scattering of electrons. Phys. Rev. 57, 24.
GrOn A. E. (1957) Lumineszenz-photometrische messungen
der energieabsorption im strahlungsfeld von elektronenquellen eindimensionaler fall in lull. Z. Naturf. 12*, 89.
Halbleib J. A. and Vandevender W. H. (1974) Tiger: a one
dimensional multilayer electron/photon Monte Carlo
transport code, SLA-73-l026, Sandia Laboratories.
Hanson A. 0., Lanzl L. H., Lyman E. M. and Scott M. B.
(1951) Measurement of multiple scattering of 15.7 MeV
electrons. Phys. Rev. 84, 634.
Harder D. (1970) Some general results from the transport
theory of electron absorption. In 2nd Symp. on Microdosimetry, Stress, Italy (Edited by Ebert).
Hoffman T. 3. et at. (1978) Sputtering calculations with
discrete ordinate method. Nucl. Sd. Engng. 68, 204.
Holier W. K. et a!. (1975) Concentration profiles ofboron
implantations in amorphous polycrystalline silicon.
Radial. Eff. 24, 223.
Hogstrom K. R., Mills M. D. and Almond P. R. (1981)
Electron beam dose calculations. Phys. Med. Bin!. 26,445.
Huffman et a!. (1957) Spatial distribution of energy
absorbed from an electron beam penetrating aluminium.
Phys. Rev. 106, 435.
Huizenga H. and Storchi P. R. M. (1989) Numerical
calculation of energy deposition by broad high-energy
electron beams. Phys. Med. Biol. 34, 1371.
Jespersgard P. and Davies J. A. (1967) Ion implantation
depth distribution:

energy deposition into

atomic

processes and ion locations. Can. J. Phys. 43, 2983.


Jette D. (1983) The application of multiple scattering
theory to therapeutic electron dosimetry. Med. Phys. 10,
141.
Kessaris N. D. (1964) Calculated absorbed dose for
electrons. Radiat. Res. 23, 630.

Transport theory of charged particles

Kulchitsky L. A. and

Latyshev C. D. (1942) The multiple


scattering of fast electrons. Phys. Rev. 61, 254.
Landau L. (1944) On the energy losses of fast particles by
ionization. I. Phys. U.S.S.R. 8, 201.
Langley R. A. (1984) NucI. Fusion Special Issue, 9.
Lax I. and Brahme A. (1985) Electron beam dose planning
using Gausian beams. Ada Radio!. Oncol. 24, 75.
Lewis H. W. (1950) Multiple scattering in an infinite
medium. Phys. Rev. 78, 526.
Lindhard J., Scharif M. and Schitt H. E., (1963) Range
concepts and heavy ion ranges (note on atomic collisions
II). Kg!. Danske. Vidensk. Seisk. Mat. Fys. Medd. 33, 14.
Lindhard J., Nielsen V., Scharif M. and Thomson P. V.
(1963) Integral equations governing radiation effect. Note
on atomic collisions III. Kg!. Danske Videnskab. Selskab.
Mat. Fys. Medd. 33, 10.
Littmark U. and Ziegler J. F. (1981) Ranges of energetic
ions in matter. Phys. Rev. A23, 64.
Luo Z. M. (1976) The multiple scattering and losses of
electrons. Chin. J. Atom. Energy Sri. Technol. 1, 77.
Luo Z. M. (1985a) Improved bipartition model of electron
transport. I: a general formulation. Phys. Rev. B32, 812.
Luo Z. M. (1985b) Improved bipartition model of electron

transport. II. Application to inhomogeneousmedia. Phys.


Rev. B32, 824.
Luo Z. M. (1988) Inifuence of surface barrier of solids on
range profile. Nuci. Instru,n. Meth. B33, 102.
Luo Z. M. (1990) A new parameter for describing the
reflection coefficients of light ions: scaled transport cross
section. NucI. Instrwn. Meth. B48, 444.
Luo Z. M. (1991) An outline of sputtering theory based on
bipartition model for ion transport. The oral report at 1st
Coordinating Meeting of Plasma Induced Eroson of
Fusion Reactor Materials by IAEA in Vienna.
energy electron
transport. Phys. Rev. In press.
Luo Z. M. and Wang S. M. (1987) Bipartition model of light
ion transport: an outline of new range theory. Phys. Rev.
B36, 1885.
McConnell W. J. et a!. (1965) Electron slowing-down spec-

Luo Z. M. and Brahme A. (1993) High

trum in Cu of beta ray from Cu. Phys. Rev. 138, 1377.


Moliere 0. (1948) Theorie der streuung schneller geladner
teilchen II, Mehrfach und vielfach streuung. Z. Naturf
3a, 78.
Mller C. (1932) Zur theorie des durchgangs schneller
elektronen durch materie. Ann. Phys. 14, 117.

Nelson W. R. Hiyayama H. and Rogers D. W. 0. (1985)

Standard Linear Accelerator Center, Report SLAC-265.


Nigam B. P., Sundaresan M. K. and Wu T. Y. (1959)
Theory of Moliere scattering: second Born approximation
and correction to Molieres work. Phys. Rev. 115, 491.
Oechsner H. (1984) Thin Film and Depth Profile Analysis,

(Edited by Oechsner H.). Springer, Berlin.


Paretzke H. 0. (1975) Proc. 5th Symp. Microdosimetry
(Edited by Booz J., Ebert and Smith), Verbamia Pallanta,
Italy, pp. 4159.

(1977) Berechung von elektronen und


bremsstrahlungs- feldern in heterogenen gewebschichten,

Prillinger 0.

IKE 6-96, Institut ffir Kernenergetik, Stuttgart.


Robinson M. T. and Torrens I. M. (1974) Computer
simulation of atomic displacement cascades in solids in

the binary collision approximation. Phys. Rev. B9, 5008.


Rogers D. W. 0. and Bielajew A. F. (1988) Experimental
Benchmark of EGS, Monte Car!o Transport of Electrons
and Photons (Edited by Jenkins T. M., Nelson W. R. and
Rindi A.), pp. 307321. Plenum Press, New York.
Roos H., Drepper P. and Harder D. (1970) The transition

from multiple scattering to complete diffusion of high


energy electrons. In Proc. of the 5th InS. Betatron Symp.

Bukarest.

and Ottenberger W. (1979)


IPP-Report 9/26.
Salow H. (1949) Phys. Z. 41, 434.
Roth 3., Bohdansky 3.

703

Sanders J. B. (1968) Ranges of projectiles in amorphous


materials. Can. I. Phys. 46, 455.
Scott W. (1963) The theory of small angle multiple scattering of fast charged particles. Rev. Mod. Phys. 35, 231.
Seiler H. (1983) Secondary electron emission. Electron
Beam Interaction with Solids, SEM. Inc. AMF OHare
(Chicago), IL 60666, U.S.A., pp. 3342.

Seliger H. H. (1955) Transmission of positrons and deetrons. Phys. Rev. 100, 1029.
Sigmund P. (1969) Theory of sputtering. I sputtering yield
of amorphous and polycrystalline targets. Phys. Rev. 184,

383.
Sigmund P. (1972) Collision theory ofdisplacement damage.

Rev. Roum. Phys. Tome 17, 1079.


Sigmund P. (1981) Sputtering by Particle Bombardment,
Vol. 1 (Edited by Behrisch R.). Springer, Berlin.
Snyder H. S. and Scott W. T. (1949) Multiple scattering of
fast charged particles. Phys. Rev. 76, 220.
Spencer L. V. (1955) Theory of electron penetration. Phys.
Rev. 98, 1597.
Spencer L. V. (1959) Energy dissipation by fast electrons,

NBS Monograph I.
Spencer L. V. and Attix F. H. (1955) A theory of cavity
ionization. Radiat. Res. 3, 239.

Spencer L. V. and Fano U. (1951) J. Res. NBS 46, 446.


Spencer L. V. and Fano U. (1954) Energy spectrum resulting from electron slowing down. Phys. Rev. 93, 1172.
Symon K. R. (1948) Fluctuations in energy loss by high
energy charged particles in passing through matter. PhD
Thesis, Harvard University.

Takeuchi W. and Yamamura Y. (1983) Computer studies of


the energy spectra and reflection of light ions. Radiat. Eff.
71, 53.
Vavilov P. V. (1957) Ionization losses of high energy heavy
particles. JETP 5, 749.
Wang M. C. and Guth E. (1951) On the theory of multiple

scattering, particularly ofcharged particles. Phys. Rev. 84,


1092.

Weissmann R. and Sigmund P. (1973) Sputtering and

backscattering of keV light ions bombarding random


targets. Radiat. Eff. 19, 7.
Wentzel G. (1922) Ann. Phys. 69, 335.
Wick 0. C. (1949) On the space distribution of slow
neutrons. Phys. Rev. 75, 738.

Williams E. 3. (1939) Concerning the scattering of fast


electrons and of cosmic-ray particles. Proc. Roy. Soc.

Lond,, Ser. A 169, 531.


Wilson J. W. and Denn F. M. (1977) Method of shield
analysis for protection against electrons in space. Nuc!.
Tech. 35, 178.
Winterbon K. B. (1975) Ion Implantation Range and energy
Deposition Distribution, Vol. 2. Low Incident Ion Energy.

Plenum, New York.

Winterbon K. B. (1972) Heavy ion range profiles and


associated damage distributions. Radiat. Eff 13, 215.
Winterbon K. B., Sigmund P. and Sanders J. B. (1970) Kg!.
Danske. Videnskab. Seiskab. Mat. Fys. Medd., 37, No. 14.
Winters H. F., Coufal H., Rettner C. T. and Bethune D. S.
(1990) energy transfer from rare gases to surfaces: collision with gold and platinum in the range 14000eV.
Phys. Rev. Mi, 6240
Yamamura Y., Matsumami N. and Itoh N. (1983) Theoretical studies on an empirical formula for sputtering yield at

normal incidence.

Radiat. Eff

75, 65.

Yang C. N. (1951) Actual path length of electrons in foils.


Phys. Rev. 84, 599.
Young J. R. (1956) Penetration of electrons in A1
203 films.
Phys. Rev. 103, 299.
Zerby C. D. and Keller F. L. (1967) Electron transport,
calculations, and experiments. Nuci. Sci. Engng. 27, 231.
Ziegler J., Biersack 3. P. and Littmark U. (1985) The
Stopping and Range ofIons in Solids (Edited by Ziegler
J. F.). Pergamon Press, Oxford.

You might also like