Proceedings of the Ninth EGS4 Users' Meeting in Japan, KEK Proceedings 2001-22, p.

1-5

THE PHYSICS OF
ELECTRON/POSITRON TRANSPORT IN PENELOPE
F. Salvat , J.M. Fernandez-Varea , E. Acosta and J. Sempau
1

12

Facultat de F sica (ECM), Univ. de Barcelona. Diagonal 647, E-08028 Barcelona, Spain.
FaMAF. Univ. Nacional de Cordoba. M. Allende y H. de la Torre, 5000 Cordoba, Argentina.
3
ITE, Univ. Politecnica de Catalunya. Diagonal 647, E-08028 Barcelona, Spain.
1

Abstract

PENELOPE is a general-purpose Monte Carlo code system for simulation of coupled electronphoton transport in arbitrary materials and in the energy range from a few hundred eV to 1
GeV. In this communication we sketch the physical interaction models implemented in the code
as well as recent improvements of the electron/positron transport mechanics (i.e. the numerical
algorithm for generating random electron/positron tracks). The energy-dependence correction
adopted in the new transport mechanics leads to improved stability under variations of the
simulation parameters selected by the user.

1 Introduction
PENELOPE (an acronym for \PENetration and Energy LOss of Positrons and Electrons") is a
general-purpose code system for Monte Carlo simulation of coupled electron-photon transport in the
energy range from a few hundred eV up to 1 GeV 1,2]. The kernel of the code system is a library
of FORTRAN subroutines that perform simulation of electron-photon showers in homogeneous
materials of arbitrary compositions. Geometry operations and scoring of relevant quantities are
performed from the main program, which is to be provided by the user. The complete code system,
including the interaction database, auxiliary subroutine libraries and examples of main programs
are publicly available through the NEA Data Bank1 and the RSICC2.
The simulation of photons is performed by means of the conventional, detailed (interaction
by interaction) method and will not be considered here. The simulation of electrons/positrons is
much more di cult than that of photons, because of the large number of interactions su ered by
high-energy charged particles in the course of their slowing down this is a direct consequence of
the fact that inelastic interactions of charged particles involve preferentially small energy losses
(the average energy loss per inelastic collision is of the order of 50 eV). The traditional approach
to cope with this problem has been using class I (condensed) simulation methods in which the
global e ect of multiple interactions along a given path length is described by using approximate
multiple scattering theories 3]. This solution is not fully satisfactory due to the approximations
OECD Nuclear Energy Agency Data Bank. Le Seine Saint-Germain, 12 Boulevard des Iles. 92130 Issy-lesMoulineaux, France. e-mail: nea@nea.fr http://www.nea.fr
2
Radiation Safety Information Computational Center. PO Box 2008, Oak Ridge, TN 37831-6362, USA. e-mail:
pdc@ornl.gov http://www-rsicc.ornl.gov
1

underlying the multiple scattering theories and/or to the fact that these theories provide only a
partial description of the transport (see e.g. Salvat et al., 1996 1]). A di erentiating feature of
PENELOPE is the use of full class II (mixed) simulation of electrons/positrons, i.e. interactions
are classi ed into hard and soft, and the hard ones are simulated individually. The code also
implements a sophisticated transport mechanics algorithm to generate space displacements and
energy loss from soft events (the so-called random hinge method), which accounts for the energy
dependence of the interaction probabilities in an accurate way.
The models considered here correspond to the version 2001 of PENELOPE, which will be
released next November. Some of these models di er substantially from the ones in previous versions
of the code. In particular, the random hinge algorithm has been reformulated to account for most
of the energy dependence of the interaction probabilities. This largely improves the robustness
of the algorithm and minimizes the in uence of user-selected parameters on the accuracy of the
simulation results.

2 Electron/Positron Interaction Cross Sections

In this section we describe brie y the physical basis of the models used for the simulation of
electron/positron interactions. A much more detailed description of these models can be found in
the references and in the PENELOPE report 4]. We shall not consider positron annihilation, which
is simulated according to the Heitler DCS for two-photon annihilation with free electrons at rest
(see Salvat et al., 1996 1]).

2.1 Elastic scattering

Elastic interactions are described by means of the modi ed Wentzel (MW) model, in which
the di erential cross section (DCS) is expressed as a mixture of a Wentzel (screened Rutherford)
distribution and either a triangle or a delta distribution 4]. The coe cients in the MW model are
determined in such a way that the mean free path between elastic collisions and the rst and second
moments of the angular de ection = (1 ; cos )=2 in each collision coincide with the results from
accurate partial-wave calculations. The analytical form of the DCS facilitates the simulation of
random events and the formulation of the mixed algorithm. When transported particles undergo
multiple elastic scattering, the results obtained with the MW model do not di er signi cantly from
those simulated from the more accurate partial-wave numerical DCSs.

2.2 Inelastic scattering

Inelastic scattering is described by means of the plane-wave ( rst) Born approximation. The
fundamental feature of this approximation is that the DCS (di erential in the energy loss and
recoil energy) factorizes into the product of a purely kinematical factor and a structure factor (the
so-called generalized oscillator strength or GOS) that is independent of the energy of the projectile (see e.g. Inokuti, 1971 5]). Elementary excitations/ionizations are described by means of the
Sternheimer-Liljequist oscillator model, i.e. the GOS is represented as a family of delta-oscillators,
with resonance energies adjusted to to give the \correct" (ICRU) stopping power at high energies 1]. For relatively large recoil energies (i.e. large momentum transfers, each oscillator behaves
as a set of free electrons and its response is described by means of the M ller and Bhabha DCSs
for electrons and positrons, respectively. This GOS model provides values of the stopping power
that are essentially equal to those recommended by the ICRU and also yields realistic estimates of
the mean free path between inelastic collisions and the energy straggling parameter for the whole
energy range from 100 eV to 1 GeV. The details of the GOS model were selected so as to ensure
2

accuracy and, at the same time, allow the random sampling of the energy loss and recoil energy by
means of purely analytical methods.

2.3 Bremsstrahlung emission

The energy of bremsstrahlung photons is sampled from the scaled energy-loss DCSs tabulated
by Seltzer and Berger 6], which represent the state of the art in theoretical bremsstrahlung data.
For a given material, the scaled DCS depends only on the kinetic energy of the projectile E
and on the photon energy W , and it is easy to sample the photon energy by using appropriate
interpolation/sampling techniques.
The intrinsic angular distribution of bremsstrahlung photons (relative to the direction of the
projectile) is sampled from an analytical form that approximates (very closely) the partial wave
data tabulated by Kissel, Quarles and Pratt 7]. We have found that the actual angular distribution
reduces to a simple analytical form under a Lorentz boost (see e.g. Jackson, 1975 8]) with an
appropriate velocity, di erent from the speed of the projectile). This fact enables us to express
the angular distribution by means of an analytical formula (with parameters that vary smoothly
with the energies of the projectile and the emitted photon) from which the direction of the emitted
photon can be sampled analytically, i.e. in an easy and exact manner (see Salvat et al., 2001 4]).

3 Electron Transport Mechanics

As mentioned above, PENELOPE does class II simulation of electrons/positrons. To clarify the
meaning of this, let's assume a projectile electron with kinetic energy E moving in a medium of given
composition. The total DCS (i.e. the sum of elastic, inelastic and bremsstrahlung contributions)
is a function of the energy loss W and the angular de ection = (1 ; cos )=2. Interactions
are classi ed as \soft" and \hard" by introducing suitable cuto s Wc and c for the energy loss
and angular de ection, respectively. It is always possible to select the cuto s in such a way that
the average number of hard collisions along an electron track is small, say of the order of 10.
Hard collisions can then be simulated in a detailed way (individually). The soft collisions that
occur between a pair of consecutive hard collisions produce gentle energy losses and de ections,
which can be accurately described by means of a multiple scattering approach (i.e. using condensed
simulation). The strategy used in PENELOPE consists of moving the transported electrons by free
steps (where direction and energy remain constant) and simulating interaction events (i.e. energy
losses and de ections) at the end of each free step. The lengths of the steps and the kind of events
that occur are determined according to the \random hinge method" (Salvat et al., 2001).
Let d (h) =(dW d ) denote the total DCS for hard interactions (which vanishes for W < Wc and
for < c ). The total \hard" cross section is
Z 1
Z E
d (h)
(h)
(E ) = d
(1)
Wc dW d
c
and the associated mean free path is (h) = N (h) (E ), where N denotes the density of atoms
(or molecules) per unit volume. Notice that the mean free path is a function of the energy E
of the projectile, which varies between consecutive hard interactions due to the e ect of the soft
interactions. As a consequence, 1= (h) (E ) is the probability of interaction per unit path length, but
(h)
does not represent the average distance between hard events. Similarly, the \soft" interactions
are described by the cross section d (s) =(dW d ), which vanishes for W > Wc and for > c .
PENELOPE simulates multiple soft interactions by means of \arti cial" distributions that are
de ned by the rst moments of the soft DCS, which again vary with the energy of the projectile.
In previous versions of the PENELOPE algorithm, the energy dependence of these parameters
was disregarded. This required limiting the cuto values so as to have suitably small energy losses
3


  
  
 

(

OKFP


OKFP


(


  
  
 

(

(

(

(
(

$O

$X

( ( ( ( ( ( (

( ( ( ( ( ( (

(H9

(H9

Figure 1: Inverse mean free path (interaction probability per unit path length) for hard interactions
of electrons in Al and Au for various values of the simulation parameters (which correspond to
di erent cuto de ections c ). In al cases, the cuto energy loss was Wc = 100 eV.
along each step and, therefore, it set a limit on the simulation speed. In the 2001 version of the
code, this energy dependence is accounted for in a quite accurate manner, which is to the advantage
of code accuracy and stability. The adopted solutions are described in detail in the PENELOPE
manual (Salvat et al., 2001). Essentially, they make use of two facts. First, our energy straggling
model allows us to set an e ective upper limit for the energy loss along a step of given length. And,
second, the inverse hard mean free path (as a function of E ) does not have maxima3 (see g. 1).
This implies that the energy dependence of the hard mean free path can be accounted for exactly
and very e ciently. The energy variation of the moments of the soft interaction DCS are described
by means of a linear approximation, which is accurate enough to permit step lengths of the order
of 10 per cent of the Bethe range, which is conveneient for e cient class II simulation.
The transport mechanics adopted in PENELOPE is very accurate and stable under variations
of the simulation parameters, due mostly to the e ectiveness of the energy-dependence corrections.
This is illustrated in g. 2, which displays results from simulations of 511 keV electrons in aluminium (in nite medium) with radiative events switched o . Electrons started o from the origin
of coordinates moving in the direction of the z axis. The simulation of each track was discontinued
when the electron had travelled a path length s equal to 200 m, and the distributions of the
nal electron energy and position coordinates were tallied no secondary radiations were followed.
Crosses represent results from detailed (interaction by interaction) simulations, continuous curves
are results from class II simulations with relatively large cuto s, which are 55 times faster than detailed simulation. Both calculations agree almost exactly, which proves that our class II simulation
is indeed very accurate.

3
This is so because our hard interaction mechanism includes elastic scattering. The inverse mean free path for
hard (restricted) inelastic scattering and bremsstrahlung emission has a maximum at intermediate energies.

H
 NH9
V FPLQ$O

(

S(H9


S]FP









H
 NH9
V FPLQ$O

(

(





]FP

(



(

(

(H9

(

Figure 2: Results from the simulations of 511 keV electrons in aluminium described in the text.
Crosses, detailed simulation continuous curves, mixed simulation. p(z ) and p(E ) are the distributions of the z -coordinate and the energy E of the electron, after traveling the prescribed 200
m.

Acknowledgements
F.S. wishes to express his gratitude to the Ministry of Education, Culture, Sports, Science and
Technology of Japan for a Foreign Visiting Scientist Fellowship and to the sta of the Radiation
Center of KEK for their warm hospitality.

References
1) F. Salvat, J. M. Fernandez-Varea, J. Baro and J. Sempau, Informes Tecnicos CIEMAT n. 799
(CIEMAT, Madrid)(1996).
2) J. Sempau, E. Acosta, J. Baro, J. M. Fernandez-Varea and F. Salvat, Nucl. Instrum. and Meth.
B132(1997)377-390.
3) M. J. Berger, Methods in Computational Physics , vol. 1, eds. B. Alder, S. Fernbach and M.
Rotenberg (Academic Press, New York) p. 135 (1963). (C)
4) F. Salvat, J. M. Fernandez-Varea, E. Acosta and J. Sempau, To be published (2001).
5) M. Inokuti, Rev. Mod. Phys. 43(1971)297-347.
6) S. M. Seltzer and M. J. Berger, Nucl. Instr. and Meth. B12(1985)95-134.
7) L. Kissel, C. A. Quarles and R.H. Pratt, At. Data and Nucl. Data Tables 28(1983) 381.
8) J. D. Jackson, Classical Electrodynamics (John Wiley and Sons, New York)(1975).
5