2 Analgorithmfor Monte Carlosimulationofbremsstrahlungemissionbyelectrons SCI
2 Analgorithmfor Monte Carlosimulationofbremsstrahlungemissionbyelectrons SCI
2 Analgorithmfor Monte Carlosimulationofbremsstrahlungemissionbyelectrons SCI
net/publication/315922577
CITATIONS READS
2 239
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Peng He on 10 April 2019.
Abstract An algorithm for Monte Carlo simulation of Keywords Bremsstrahlung Monte Carlo SuperMC
bremsstrahlung emission by electrons based on the frame- Differential cross section Angular distribution
work of SuperMC is presented in this paper with efficient and
accurate methods to sample the angular distribution and
energy of bremsstrahlung photons. The photon energy is 1 Introduction
sampled according to scaled energy-loss differential cross
sections tabulated by Seltzer and Berger. A novel hybrid As a result of the acceleration or especially the deceler-
model for photon angular distribution by low- and high-en- ation of charged particles caused by the electrostatic field of
ergy incident electrons is developed. The model uses Tsai’s atoms, electrons (or positrons) emit bremsstrahlung.
full form of angular distribution function with atomic form Bremsstrahlung emission is of basic interest in many fields.
factors for high-energy incident electrons. For electrons of Examples of applications where the bremsstrahlung inter-
\500 keV, a simple efficient and accurate analytical dis- action plays a vital role are: X-ray generators, radiation
tribution function is developed, using adjustable parameters shielding, medical physics and electron probe microanalysis
determined from the fitting of numerical values of the shape [1]. In medical physics, the bremsstrahlung phenomenon is
functions tabulated by Kissel et al. The efficiency of sam- very important in both diagnostics radiology and radio-
pling photon energy is 80%. Our angular sampling algorithm therapy because the fundamental characteristics of X-ray
for high-energy electron bremsstrahlung based on Tsai dis- machine are described by the spectral distribution of photon
tribution function is very efficient (sampling efficiency beam originating from the target. Bremsstrahlung photons
*70%) in the useful photon energy range. have much higher penetration than other charged particles
and thus have large contribution in dose in deep regions of
This work was supported by the National Natural Science Foundation the irradiated samples; therefore, an accurate Monte Carlo
of China (No. 81101132 and 11305203), the Strategic Priority simulation of bremsstrahlung is required in a general pur-
Research Program of Chinese Academy of Sciences (No. pose Monte Carlo code such as SuperMC [2].
XDA03040000) and the Natural Science Foundation of Anhui SuperMC is a general purpose, easy-to-use Monte Carlo
Province of China (No. 1508085QH180).
simulation program, developed by FDS Team [2–5], for
& Peng He modeling and simulating nuclear facilities. The released
[email protected] version of SuperMC can perform the simulation of neu-
1
trons, photons and coupled neutron and photon transport,
Key Laboratory of Neutronics and Radiation Safety, Institute
of Nuclear Energy Safety Technology, Chinese Academy of
while Monte Carlo simulation of electron transport has
Sciences, Hefei 230031, China been under development and will be released soon, this
2 research work is part of the development of electron
Collaborative Innovation Center of Radiation Medicine of
Jiangsu Higher Education Institutions, Suzhou 215006, China transport. The main technical features of SuperMC include
3 hybrid MC-deterministic methods and the adoption of
School of Nuclear Science and Technology, University of
Science and Technology of China, Hefei 230027, China advanced information technologies, while the main
123
71 Page 2 of 9 M. A. Wasaye et al.
usability features include automatic modeling of geometry electrons. We consider the process in which electron of
and physics, visualization and virtual simulation and cloud energy E (kinetic energy T) is accelerated or decelerated by
computing services. SuperMC has been applied to fusion the screened coulomb field of the target atom and emit
reactor FDS I [6], FDS II [7], FDS III [8], ITER [9, 10] and bremsstrahlung photons of energy K in the interval [0, E] in
other reactor studies [11] and is also designed for MC- the direction of polar angle h. Angular deflections of the
based dose calculation engine of accurate/advanced incident electron due to bremsstrahlung are already
radiotherapy system (ARTS) [12]. accounted for by electric scattering in SuperMC and
Bremsstrahlung process is described by the differential assumed that the direction of electron is not altered by the
cross section (DCS) in energy and direction of emitted emission of bremsstrahlung photon.
photons and the direction of outgoing electrons. Atomic In conventional MC simulations, bremsstrahlung emis-
DCSs of bremsstrahlung emission by electrons are derived sion is modeled using analytical DCSs obtained from
from conventional quantum electrodynamics methods, and simple approximations and this was done to minimize the
comprehensive review on the DCSs is described by Koch computer memory requirements. With the advancement in
and Motz [13] and Tsai [14]. Efforts have been made to use computer technology, however, it is now convenient to use
accurate methods for sampling the photon energy and, the combination of numerical DCSs and analytical for-
especially, the angular distribution of bremsstrahlung mulae, which are more accurate and reliable at present. In
photons, but still, there is need to use simple, efficient and this study, we used the models and numerical data which
accurate methods to sample energy and angular distribu- are considered as most reliable at present and their form
tion. The 2BS formula of Koch and Motz was implemented suits for Monte Carlo simulation.
by Bielajew et al. [15] in EGS4 [16] for the angular dis-
tribution of bremsstrahlung photons. Acosta et al. [17] 2.1 Sampling photon energy
developed a parameterization with adjustable parameters
determined from the fitting of shape functions provided by In this section, we describe the algorithm to sample
Kissel et al. [18] based on the calculations of partial wave photon energy, K, from scaled energy-loss total DCSs
methods by Tseng et al. [19]. Geant4 [20] uses a parame- (electron–electron, electron–nucleus), v(Z, T, w = K/T),
terization of Tsai [14] DDCS for sampling photon angular tabulated by Seltzer and Berger [23] which are based on the
distribution, while Rodrigues et al. [21] implemented 2BN combination of numerical partial wave results and high
formula of Koch and Motz, in Geant4 as more accurate energy theory.
option for low-energy region. EGSnrc [22] incorporated
b2 dr
modified version of 2BS formula which converges to 2BN k ¼ vðZ; T; w ¼ K=T Þ; ð1Þ
Z 2 dK
for the treatment of low-energy region.
In this article, an algorithm is described for Monte Carlo where w is reduced photon energy,b is incident electron
simulation of electron bremsstrahlung with kinetic energy velocity in units of the velocity of light, and T is the kinetic
range of 1 keV–10 GeV and for Z = 1–100. We also energy of incident electron. The set includes results for 57
describe the efficient and accurate methods by which the values of electron kinetic energies from 1 keV to 10 GeV
angular distribution and energy of bremsstrahlung photons versus 30 values of reduced photon energy points span on
are sampled. Photon energy is sampled according to most the interval [0, 1], and for atomic numbers Z = 1–100. The
reliable Seltzer and Berger [23] scaled DCSs, and for values of electron kinetic energies are suitably spaced
angular distribution of bremsstrahlung photons, a novel (spacing is nearly logarithmic) to allow accurate linear
hybrid model is developed. In Sect. 2, DCSs used for interpolation in lnT. The continuous scaled DCSs are,
sampling energy and angular distribution of emitted pho- therefore, obtained by cubic spline interpolation in w = K/
tons, and corresponding random sampling algorithms are T and linear interpolation in lnT.
presented. Results of the newly developed MC simulation For sampling photon energy, a simple and efficient
algorithm are presented in Sect. 3. composite sampling procedure is implemented. The dis-
tribution function p(w) for the reduced photon energy,
w = K/T, and for given Z and T over the interval [Kc/T, 1]
is given as,
2 Monte Carlo simulation of electron
bremsstrahlung 1
pðwÞ ¼ vðZ; T; wÞ; ð2Þ
w
Bremsstrahlung process is described by the atomic dif- where Kc is threshold photon energy. As shown in Fig. 1,
ferential cross section, in the direction h and energy K of v(Z, T, w) is relatively flat over the interval [Kc/T, 1] and
the emitted photons and the direction of the outgoing can be used for rejection sampling, while direct (inverse
123
An algorithm for Monte Carlo simulation of bremsstrahlung emission by electrons Page 3 of 9 71
transform) sampling method is used to sample w from the with X-rays of about a few keV to a few hundred keV.
distribution w-1 in the interval [Kc/T, 1]. Therefore, an accurate treatment of the angular distribu-
The algorithm is then tion of the bremsstrahlung photons needs to be done for
low-energy incident electrons. Thus, a new, accurate and
1. Calculate vmax (find maximum value of v in K/T grid
fairly simple analytical distribution function has been
for given T and Z).
developed for photon angular sampling with
2. Generate two random numbers r1 and r2 over the
adjustable parameters determined from the fitting of the
interval [0, 1].
numerical values of shape functions tabulated by Kissel
3. Sample w from the distribution w-1 as w = exp[ln(Kc-
et al. [18].
r1/T)]
4. If r2 B v(Z, T, w)/vmax, accept w; else, go to Step 2.
Efficiency and accuracy of this sampling algorithm are 2.2.1 High-energy bremsstrahlung angular distribution
discussed in Sect. 3.1. A number of comparisons between
measured and calculated scaled differential cross section in In this study, for high-energy electrons, the angular
photon energy have been done [23]. These comparisons distribution of emitted photons has been sampled accu-
show that Seltzer and Berger’ scaled differential cross rately, according to the ‘‘exact form’’ of Tsai’s double
sections are the most reliable representation of brems- differential cross section (DDCS) which includes atomic
strahlung at present. form factors. Thus, the sampling is fully consistent with
theory. It should be noted that the parameterization of the
2.2 Hybrid model for photon angular distribution angular distribution instead of ‘‘exact form’’ may produce
uncertainties and the accuracy of the parameterization
A new hybrid model for sampling the direction of needs to check against, many, possible electron and photon
bremsstrahlung photons has been developed which char- energies and materials. The double differential cross sec-
acterizes the explicit treatment of photon angular distri- tion, energy angle distribution of electron bremsstrahlung,
bution for low- and high-energy incident electrons. High- described by Tsai [14] is
(" #
energy analytical formula of DDCS (differential in energy dr 2a2 E2 2y 2 12u2 ð1 yÞ 2
and angle of bremsstrahlung photon) described by Tsai ¼ 4 2
þ 4
Z þZ
dXdK pme K 2
ð1 þ u Þ 2
ð1 þ u Þ
[14] based on high energy theory of Davies, Bethe, " # )
Maximon and Olsen (DBMO) [24, 25] with emphasis on 2 2y þ y2 4u2 ð1 yÞ 2
2
þ X 2Z fc ðaZ Þ ;
form factor screening corrections is used to sample the ð 1 þ u2 Þ 2 ð1 þ u2 Þ4
direction of emitted photons. However, at low energies, ð3Þ
2BS formula of Koch and Motz and Tsai’s DDCS gives
rise to deviations which could reach 50° on the most where y = K/E and u = hE/me (me = 0.511 MeV), a = 1/
probable angle [21]. This is of particular importance when 137, Z is atomic number, and fc is coulomb correction. For
simulating bremsstrahlung emission in very thin targets the calculation of the function X, Tsai [14] provided simple
and addressing topics like medical imaging which works expressions:
" 2
#
2 a2 m2e ð1 þ u2 Þ
X ¼ Xel þ Xinel ¼ Z ln 1
a2 tmin þ 1
" 0 2
# ð4Þ
a 2 m2e ð1 þ u2 Þ
þ Z ln 1 ;
a0 2 tmin þ 1
123
71 Page 4 of 9 M. A. Wasaye et al.
" #
2y 2 12u2 ð1 yÞ 2 Since true maximum are 3% (at the most) higher than
Rð u Þ ¼ 2
þ 3
Z þZ the estimated maximum, this small error can be corrected
ð1 þ u Þ 2
ð1 þ u Þ
" # by introducing an additional multiplicative factor to the
2 2y þ y2 4u2 ð1 yÞ 2
2
estimated maximum, which is 1.03 (3%) for entire energy
þ 3
X 2Z f ð aZ Þ ;
ð 1 þ u2 Þ ð 1 þ u2 Þ range.
ð6Þ The algorithm is then
123
An algorithm for Monte Carlo simulation of bremsstrahlung emission by electrons Page 5 of 9 71
Fig. 3 Location of the maximum of the rejection function as function of reduced photon energy for a aluminum and b silver and for several
incident electron kinetic energies. It can be seen that the z-dependence of umax is very week
Fig. 4 Comparisons of angular distribution of bremsstrahlung pho- incident on silver and gold, obtained with Tsai DDCS, Kissel et al.
tons for 50 and 100 keV electrons with reduced photon energy K/ data and analytical distribution function. All plots are normalized to
T = 0.6, respectively, and for 500 keV electrons with K/T = 0.8 unity
method) was applied to sample h. The normalization con- retrieves, for particular T, Z and K/T. The polar angle is
stant NA is constant chosen to normalize the integral of then sampled as
q(h) in the overall range of h to unity.
r B
Bþp h ¼ B C ln 1 ln þe 1eC
; ð11Þ
e C eC
B
NA A C
NA ¼ A C exp e e ; ð10Þ
where r is random number uniformly distributed over [0,
Since the calculation of NA involves five exponential 1]. This algorithm requires one log(log) calculation and
calculations, in order to make algorithm efficient we cal- two exponential calculations in Eq. (11). As it uses direct
culated NA for 144 benchmarked values and stored in sampling method, the algorithm is more efficient than the
tables along with adjustable parameters for their rapid sampling method proposed by Acosta et al. [17] which uses
123
71 Page 6 of 9 M. A. Wasaye et al.
four random numbers, on average, to sample each value of the photon energy sampling algorithm works as expected.
polar angle. All plots in Fig. 5 are normalized to unity.
2.2.3 Comparison between Kissel shape functions, Tsai 3.1.2 Sampling efficiency
DDCS and fitting function
The sampling efficiency, P, of the photon energy sam-
Figure 4 shows comparisons of the polar angle distribu- pling algorithm is given in Table 1. The sampling effi-
tion of bremsstrahlung photons using Tsai DDCS, Kissel ciency is defined as the ratio of the accepted sampled
et al. data and the analytical distribution function, Eq. (9), values to the total number of trial. We have calculated the
with reduced photon energy, K/T = 0.6, for 50 keV, efficiency in terms of incident electron energy. Although
100 keV and K/T = 0.8 for 500 keV electrons incident on the sampling efficiency severely depends on the threshold
Silver and Gold. For low energy about 50 keV and 100 keV, photon energy and also less on Z, the qualitative behavior
there is a significant deviation in the most probable angle remains the same for other values of Kc. We are therefore
between Tsai’s DDCS and Kissel’s data/Analytical function more interested in calculating P for different electron
(Eq. 9). With an increase in the electron energy, all three energies while keeping Kc = 1 keV and Z = 43 (Ag).
approaches tend to overlap and present a good agreement. From Table 1, the sampling efficiency in energy region
Since Kissel et al. data is very accurate for electron energies of E \ 100 keV and E [ 1 MeV is higher than that in
below 500 keV, we recommended to use analytical distri- 100 keV–1 MeV, which is obvious from the shape of the
bution function given by Eq. (9) in this energy range. For energy spectrum, as shown in Fig. 1. Even for higher
energies of [500 keV, Tsai DDCS is a good choice as it incident electron energies, the energy spectrum is flat
retains the high sampling efficiency and physics accuracy. except at end points of K/T.
123
An algorithm for Monte Carlo simulation of bremsstrahlung emission by electrons Page 7 of 9 71
Fig. 7 Angular distribution of photons a with reduced photon energy K/T = 0.6 for 50 keV electrons incident on aluminum. b For 100 keV
electrons incident on silver with reduced photon energy K/T = 0.95
123
71 Page 8 of 9 M. A. Wasaye et al.
The sampling efficiency of our proposed angular sam- 1. F. Salvat, J.M. Fernández-Varea, J. Sempau et al., Monte Carlo
simulation of bremsstrahlung emission by electrons. Radiat. Phys.
pling algorithm based on Tsai’s DDCS and angular sam-
Chem. 75, 1201–1219 (2006). doi:10.1016/j.radphyschem.2005.
pling algorithm proposed by Bielajew et al. [15] 05.008
implemented in EGS5 [26] and Geant4 [20] is plotted in 2. Y. Wu, J. Song, H. Zheng et al., CAD-based Monte Carlo pro-
Fig. 8. The sampling efficiency is plotted for incident gram for integrated simulation of nuclear system SuperMC. Ann.
Nucl. Energy 82, 161–168 (2015). doi:10.1016/j.anucene.2014.
electron energies from 500 keV to 1000 MeV for silver and
08.058
gold, and K/T = {0, 0.5, 1}. One can realize that the sam- 3. Y. Wu, CAD-based interface programs for fusion neutron trans-
pling efficiency based on Tsai’s DDCS is generally very port simulation. Fusion Eng. Des. 84, 1987–1992 (2009). doi:10.
good for both Z = 47 and Z = 79 and about 70% of average 1016/j.fusengdes.2008.12.041
4. Y. Li, L. Lu, A. Ding et al., Benchmarking of MCAM 4.0 with
for K/T * 0, 0.5 and reduces by few percent when K/T = 1.
the ITER 3D model. Fusion Eng. Des. 82, 2861–2866 (2007).
The sampling efficiency for Z = 79 decreases, at most, to doi:10.1016/j.fusengdes.2008.12.041
55% at 1 MeV, only when K/T = 1 it retains 62% up to 5. Y. Wu, Z. Xie, U. Fischer, A discrete ordinates nodal method for
1000 MeV. So, the sampling technique employed here is one-dimensional neutron transport calculation in curvilinear
geometries. Nucl. Sci. Eng. 133(3), 350–357 (1999)
more efficient in the useful photon energy range (lower
6. Y. Wu, S. Zheng, X. Zhu et al., Conceptual design of the fusion-
reduced photon energies) than the sampling technique used driven subcritical system FDS-I. Fusion Eng. Des. 81(Part B),
to sample the angular distribution of photons, reported by 1305–1311 (2006). doi:10.1016/j.fusengdes.2005.10.015
Bielajew et al. [15] using 2BS formula of Koch and Motz 7. Y. Wu, F.D.S. Team, Conceptual design of the China fusion
power plant FDS-II. Fusion Eng. Des. 83(10–12), 1683–1689
[13] which is implemented in EGS5 [26] and Geant4 [20].
(2008). doi:10.1016/j.fusengdes.2008.06.048
8. Y. Wu, F.D.S. Team, Fusion-based hydrogen production reactor
and its material selection. J. Nucl. Mater. 386–388, 122–126
4 Conclusion (2009). doi:10.1016/j.jnucmat.2008.12.075
9. Y. Wu, F.D.S. Team, Conceptual design and testing strategy of a
It can be concluded that the proposed algorithm provides dual functional Lithium-Lead test blanket module in ITER and
accurate and efficient methods for sampling energy and EAST. Nucl. Fusion 47(11), 1533–1539 (2007)
10. Y. Wu, F.D.S. Team, Design analysis of the China dual-functional
direction of bremsstrahlung photons. The efficiency of lithium lead (DFLL) test blanket module in ITER. Fusion Eng. Des.
sampling photon energy is more than 80% with this new 82, 1893–1903 (2007). doi:10.1016/j.fusengdes.2007.08.012
algorithm. Hybrid model provides very accurate treatment 11. Y. Wu, Progress in fusion-driven hybrid system studies in China.
of photon angular distribution by low as well as high-en- Fusion Eng. Des. 63–64, 73–80 (2002). doi:10.1016/S0920-
3796(02)00239-9
ergy incident electrons. Angular sampling which is based 12. Y. Wu, G. Song, R.F. Cao et al., Development of accurate/ad-
on Tsai DDCS is very efficient for useful photon energy vanced radiotherapy treatment planning and quality assurance
range, compared to other available angular sampling system (ARTS). Chin. Phys. C 32(Suppl.II), 177–182 (2008)
techniques in high-energy region (Bielajew et al. [15]). 13. H.W. Koch, J.W. Motz, Bremsstrahlung cross section formulas
and related data. Rev. Mod. Phys. 31, 920–955 (1959). doi:10.
Distribution function used for photon angular sampling 1103/RevModPhys.31.920
using Tsai DDCS has contribution of atomic form factors 14. Y.S. Tsai, Pair production and bremsstrahlung of charged leptons.
and hence explains physics phenomena very well espe- Rev. Mod. Phys. 46, 815–851 (1974). doi:10.1103/RevModPhys.
cially in high-energy region. The parameterization of the 46.815
15. A.F. Bielajew, M. Rahde, S.C. Chen. Improved bremsstrahlung
shape functions given by Eq. (9) provides a very accurate photon angular sampling in EGS4 code system. NRC Ottawa,
fit to the Kissel’s benchmark data (Fig. 4); therefore, Canada, PIRS-0203R, (1989)
simulated angular distribution is fully consistent with the 16. W.R. Nelson, H. Hirayama, D.W.O. Rogers. The EGS4 code
most accurate and reliable theory at present and the sam- system. Stanford Linear Accelerator Centre, Stanford. CA.
Report SLAC-265, (1985)
pling algorithm is more efficient than the previously 17. E. Acosta, X. Llovet, F. Salvat, Monte Carlo simulation of
developed method by Acosta et al. [24], which uses four bremsstrahlung emission by electrons. Appl. Phys. Lett. 80,
random numbers, on average, to sample one photon energy. 3228–3230 (2002). doi:10.1063/1.1473684
It is worth mentioning that the approaches used for 18. L. Kissel, C.A. Quarles, R.H. Pratt, Shape functions for atomic
field bremsstrahlung from electrons of kinetic energy 1–500 keV
sampling energy and direction of bremsstrahlung photons on selected neutral atoms 1 B Z B 92. At. Data Nucl. Data
can be used for a wide range of incident electron energies Tables 28, 381–460 (1983). doi:10.1016/0092-640X(83)90001-3
and materials and hence suitable for general purpose Monte 19. H.K. Tseng, R.H. Pratt, C.M. Lee, Electron bremsstrahlung
Carlo simulations. angular distribution in the 1–500 keV energy range. Phys. Rev. A
19, 187–195 (1979). doi:10.1103/PhysRevA.19.187
20. S. Agostinelli, J. Allison, K. Amako et al., Geant4—a simulation
Acknowledgements The authors would like to show their great toolkit. Nucl. Instrum. Methods A 506, 250–303 (2003). doi:10.
appreciation to other members of FDS Team for their support and 1016/S0168-9002(03)01368-8
contribution to this research.
123
An algorithm for Monte Carlo simulation of bremsstrahlung emission by electrons Page 9 of 9 71
21. P. Rodrigues, R. Moura, C. Ortigao et al., Geant 4 applications 24. H. Davies, H.A. Bethe, L.C. Maximon, Theory of bremsstrahlung
and developments for medical physics experiments. IEEE Trans. and pair production. II. Integral cross section for pair production.
Nucl. Sci. 51, 1412–1419 (2004). doi:10.1109/TNS.2004.832314 Phys. Rev. 93, 788–795 (1954). doi:10.1103/PhysRev.93.788
22. I. Kawrakow, E. Mainegra-Hing, D.W.O. Rogers et al., The 25. H. Olsen, L.C. Maximon, Photon and electron polarization in
EGSnrc code system. Monte Carlo simulation of electron and high-energy bremsstrahlung and pair production with screening.
photon transport. NRCC Report PIRS-701 (2011) Phys. Rev. 114, 887–904 (1989). doi:10.1103/PhysRev.114.887
23. S.M. Seltzer, M.J. Berger, Bremsstrahlung energy spectra from 26. H. Hirayama, Y. Namito, A.F. Bielajew et al., The EGS5 code
electrons with kinetic energy 1 keV–10 GeV incident on system. Stanford Linear Accelerator Centre, Stanford. CA.
screened nuclei and orbital electrons of neutral atoms with Report SLAC-R-730, (2015)
Z = 1–100. At. Data Nucl. Data Tables 35, 345–418 (1986).
doi:10.1016/0092-640X(86)90014-8
123