ISSUES WITH THE HIGH ENERGY RADIOGRAPHY SIMULATIONS FeyziInane Iowa State University Center for Nondestructive Evaluation Ames, IA 50010, USA ABSTRACT. Although most of the industrial radiography relies on x-ray tubes, in some cases, gamma rays or linear accelerators form the interrogating radiation beam. Due to high energy levels of photons in such cases, new physics start to accompany radiation absorption, scattering events. Most of these new physics are interactions of charged particles produced by high-energy photons. Charged particles with sufficient energy can produce photons. High-energy radiography simulations need to account for both photon and charged particle transport for proper computation of photon fluxes. This work will outline various physical mechanisms that need to be taken into consideration and challenges that come with designing deterministic algorithms that can handle the mathematics related to these new physical mechanisms. INTRODUCTION Although industrial radiography and medical diagnostic radiography employ x-ray tubes as the interrogating beam sources, there are other cases where the radiation source is not a x-ray tube but either a radioisotope emitting high-energy photons or a linear accelerator tube. In such cases, the photon beam will be composed of either multiple high-energy photon lines or a spectrum that can go from a few keV to many MeV levels. In the x-ray tube based implementations, the photon energy levels dictate that only photon probable interaction mechanisms are photoelectric absorption, coherent scattering and incoherent scattering. Since the primary photons are already low energy photons, any secondary photon resulting from such interactions will not be instrumental in the outcome of the radiography because of their very low energy levels and relatively small quantities. With the increase in the primary photon energy levels, another photon interaction mechanism, pair production, becomes possible and photoelectric absorption loses its dominance over the other interaction mechanisms. One byproduct of this trend is increasing amount of energy transferred to charged particles through scattering, pair production and photoelectric absorption interactions. Since amount of energy transferred to the charged particles in terms of kinetic energy is instrumental in production of secondary photons, secondary photons start to increase both in quantity and importance. The major difference between the primary photons and secondary photons is that secondary photons are less energetic compared to the primary photons. Since the photon detector usually are more sensitive to lower energy photons than the higher energy ones, estimation of secondary photon fluxes incident upon the detectors becomes an important issue in the high-energy radiography. In this article, we will provide some insight into the issues that are important in developing a deterministic algorithm that can be used in simulating high-energy radiography procedures. CP657, Review of Quantitative Nondestructive Evaluation Vol. 22, ed. by D. O. Thompson and D. E. Chimenti © 2003 American Institute of Physics 0-7354-0117-9/03/$20.00 561 LITERATURE REVIEW Early Monte Carlo charged particle transport computations started 1960's [1]. Monte Carlo computations were used for photon transport even before that date [2]. After those initial attempts, Monte Carlo based transport calculations became widely available with the advent of computers. Today, there are many general purpose Monte Carlo codes that can be used for various forms of transport calculations. Some of the well-known general-purpose Monte Carlo codes are MCNP [3,4] and EGS4 [5]. These codes or their derivatives are routinely used in various transport calculations and dosimetry. In contrast with the Monte Carlo methods, research for development of deterministic transport tools for charged particle transport started much later and gained speed in the recent years. In 1967, Zerby et. al. provided an extensive review of the electron transport theory and some experimental and computational results [6]. Most of the deterministic methods use discrete ordinates method. The first one has been reported by Bartine et.al. in early 1970's [7,8] and they implemented an extended transport correction for handling forward peaking scattering cross section. This has been followed up by a series of articles reporting implementation of Fokker-Planck operators [9-15]. Following these initial reports, there have been many reports studying various aspects of the Boltzmann-Fokker-Planck equations used in the charged-particle transport. Some of the later ones have been implemented to solve the pencil beam or collimated beam problems that are typical to medical radiotherapy implementations [16-18]. MOTIVATION Detection efficiency is quite important in radiography for improving spatial resolution. Regardless of the detector type, detection efficiency is a strong function of the photon energy. Since higher energy levels result in lower interaction rates, detectors will detect the incoming photons with lower efficiency levels. As a result, photon detection efficiency is much better for lower energy levels because of higher interaction coefficients. Very energetic photons transfer significant amounts of energy to charged particles emitted as a result of photon object interactions. These charged particles, in turn, generate secondary photons. These photons are likely to have energy levels low compared to the primary photons that caused emission of charged particles. Introduction of these secondary photons into the overall photon spectrum is likely to cause distortions in the lower portions of the overall photon spectrum. A typical example of such a case is provided by Takeuchi and et. al. [19]. One of the problems they worked on is made up of a 10 cm. thick lead plate with 6.2 MeV photons incident on one side. They provide the photon flux spectrum on the other side of the lead plate with and without bremsstrahlung photons. According to their finding, two cases provide similar results above 3 MeV. Around 1 MeV, case with bremsstrahlung photons provides a spectrum that is about 60% larger than the case without bremsstrahlung. Those numbers are about 40, 30, 15 and 10 % for 1.5, 2, 2.5 and 3 MeV respectively. In another problem, they study the photon flux inside the lead shield. In this case they analyze the flux at 4 and 10 mean free path (mfp) distances from the boundary with a 8 MeV photon source. The differences at 4 mfp are 500, 350, 220, 100 and 50 % at 1, 1.5, 2, 2.5 and 3 MeV respectively. There is not much difference above 5 MeV. For 10 mfp case, these numbers are about 160, 100, 50, 25 and 10 % for the same energy levels. As seen from these numbers, introduction of charged particle transport is likely to introduce significant amounts of secondary photons at energy levels where the detectors are more sensitive. For lower atomic number materials, levels of the secondary photon fluxes are likely to be much lower than the ones given for lead. 562 GOVERNING EQUATIONS AND ISSUES One of the major differences between x-ray tube and high energy radiography mathematical formulation is that high-energy radiography requires a set of simultaneous transport equations representing photon transport and charged particle transport rather than only photon transport. If we use the integral transport equation notation, we can write these equations for a high-energy photon source as given below. R" -l^(r-R"Q,E)clR" RT 0 Ik(r,E,G)=\qk(r-KQ,E,Q)e KT -JZ* (r-R"Q,E)dR" dR+Ik(rr,E,O)e ° k = 1,2,3 (1) where Ik(r,E,Q) is photon, electron and positron flux respectively for k=l,2,3. Since the interrogating radiation in radiography is usually an external photon source, the second term in the right hand side of equation (1) will be nonzero only for k=l. The source term in equation (1) is given in equation (2). (2) The first term in the RHS of equation (2) is Boltzmann operator, the second is continues angular deflection term and the third is continues slowing down operator. The fourth term is the source term induced by the other types of particles. The second and third terms are usually known as Fokker-Planck operators used for cases where the scattering is strongly forward peaked and energy loss per interaction is very small. Photon transport equation in equation (1) does not have Fokker-Planck operators. The fourth term in the RHS of equation set (1) provides the coupling among three equations. Photons contribute to electron flux through photoelectric absorption, incoherent scattering and pair production. In the photoelectric absorption, photoelectron carries out the photon energy minus binding energy. In the incoherent scattering, recoiled electron receives the energy that scattered photon looses. In the pair production interaction, electron and positron receive their shares of energy that is left over after the pair production energy is taken off from the incident photon. This is the only mechanism that causes a positron flux. Electrons contribute to photon flux through bremsstrahlung and fluorescence radiation following an ionization event. Positrons contribute to photon flux though bremsstrahlung, fluorescence and annihilation processes. All these mechanisms form the source term in the RHS of equation (2). While photons and charged particles interact with the matter and contribute the fluxes of other types, some of the interaction mechanisms that were not listed here contribute to the flux of its own particle or photon type. Photon Transport Issues The major issue with the high-energy photon transport is severity of the anisotropy seen in the incoherent scattering. With the level of anisotropy shown in figure 1, the sampling of the scattering kernel becomes quite challenging. The conventional Legendre polynomials approximation used in most of the deterministic methods face convergence difficulties for scattering kernels that are highly anisotropic for such cases. 563 Scattering Angle FIGURE 1. Behavior of incoherent scattering kernel with incident photon energy levels. Issues With Charged Particles One of the issues that makes charged particle transport different than the photon transport is amount of energy lost through charged particle interactions. While photons can slow down to photoelectric absorption dominant energy levels through a few scattering events with relatively large energy losses, charged particles slow down MeV levels to a few keV energy levels only after thousands of interactions. Since energy groups can not be chosen that small with the current computational power levels, conventional multigroup approach is insufficient to represent this type of energy loss. The continuous slowing down operator in equation (1) is meant to handle this energy loss process. Through that operator, charged particle is assumed to loose energy continuously. Continuous slowing down term requires scattering kernels to be decomposed into two components where one of the components represent the cases where the energy loss is relatively large enough to cause ionization and the other component represents the cases where the energy loss is small. The slowing down operator is constructed on the basis of the component where the energy losses are assumed to be small. The other component of the scattering kernel is kept with the Boltzmann integral operator. The stopping power term, Sk(E), in equation (2) is given below [20]. £, S(E) = fa (3) where ass(E—>E',jio) represents scattering interactions with small energy losses. The decomposition of charged particle scattering kernels into two components is not a unique process and it is an issue that needs to be dealt carefully. Momentum transfer, a^E), in continues angular deflection term in equation (2) is given by equation (4) [20]. cKE) = (4) 564 Computational Issues The most important issue in the computational aspect of the photon-charged particle transport is how to solve three integral equations simultaneously. Solving these in a simultaneous manner with the current computational resources poses a major difficulty. Luckily, it is possible to be able separate these three equations from each other and implement an iterative algorithm that would provide the answer sought in the computations. Usually, the nature of the problem dictates the character of the iterative algorithm. In this specific case, an interrogating photon beam induces charged particle fluxes. Those charged particle fluxes, in return, end up contributing to the photon flux. This secondary photon flux contributed by the charged particle interactions is quite small compared to the primary photon flux. Therefore, any charged-particle flux resulting from the interactions of the secondary flux will be negligibly small. Therefore, all we need to compute is the primary photon flux, primary charged particle fluxes and then secondary photon flux. Therefore, the iteration will start with solving the photon transport equation by using the external source. Once the photon fluxes are computed, electron and positron source distributions through the problem domain will be calculated. Electron and positron charged particle transport equations will be solved by using these source distributions. Once the charged particle fluxes are known, they will be used for calculating bremsstrahlung, annihilation and fluorescence photon source distributions. Secondary photon flux will be done based on these secondary photon source distributions. The resulting photon fluxes will be obtained by summing the primary and secondary photon fluxes. Another issue with the photon problem is sampling of the incoherent scattering kernel. As it is seen in figure 1, mean scattering angle at high energy levels is very small. To avoid any potential sampling problems, we plan to use a dynamic approach to the approximation of scattering integral by quadratures. The scattering angle will be divided into intervals and a different quadrature will be used in each interval. Partitioning of the scattering angle domain will vary with the energy levels. Since a direct integration approach will be used for solving the integral transport equations, this approach should prove to be efficient enough to sample scattering kernel appropriately. Another issue that needs is attention is presence of slowing down terms in the RHS of the charged particle transport equations. Photons do not gain energy through scattering. Therefore, the usual approach in the photon transport is to start with the highest energy group equation and then work down to the lowest energy group equation. Because of the slowing down terms in the charged particle transport, this approach is not feasible. In the multigroup approximation, the slowing down term and the flux is to be expressed in terms of the slowing down and flux terms in the higher and lower energy groups. This generates a strong coupling between adjacent group equations. Therefore, instead of a top down approach, either a simultaneous solution technique or an iterative technique should be adapted for solving the energy group equations. Another issue that is not very visible with the photon transport is the convergence of the scattering source iterations. The scattering source iteration is a very popular approach in deterministic transport based algorithms. This approach is very efficient in solving the photon transport equation. Because high-energy photons can escape the object boundaries or slow down to photoelectric absorption dominant energy levels very rapidly through large energy loss incoherent scattering, scattering source iteration converges very rapidly. In contrast with the photon transport, energy loss in the charged particle transport is usually very small. In addition, charged particles do not travel far and escape from the boundaries is minimal. Therefore, the convergence of scattering source surfaces as a major issue in the computations. This issue has been studied for discrete ordinates methods exhaustively and various algorithms developed for enhancing the convergence rate of 565 scattering source iterations are reviewed in an article [21]. Although that article addresses the issue for discrete ordinates method, facing similar difficulties with integral transport based methods should not be surprising. Data Issues Gathering and arranging data for high-energy radiography computations is an important issue. Although data for all types of interactions are available in the literature, they need to be organized into proper forms for utilization in the computations. In the photon transport, the most crucial data are scattering and pair production cross section data. Since coherent scattering and binding effects are negligible for such high energy levels, Klein-Nishina formula can provide incoherent scattering cross sections. Pair production and photoelectric absorption cross sections are readily available in the literature [22]. Most of the charged-particle interaction cross sections are described in a report by Lorence et al. [23]. We hope to make use of that document in generation of the cross section to be used in the computations. CONCLUSIONS As seen from the discussions in the previous sections, high-energy radiography is radically different than the x-ray tube based radiography. Therefore, it is quite difficult to adapt the x-ray tube based radiography simulation codes to high-energy radiography cases with small modifications. One reason for such a radical shift is that number of physical interactions in the high-energy regimes is much larger than the physical interactions seen in the lower energy levels. Therefore, simulation codes should be able to accommodate the physics introduced at higher energy levels. In addition, charged particles play a role in the higher energy levels. Although bremsstrahlung generation starts to dominate chargedparticle physics at higher levels, it may still have enough contribution to the overall photon flux to distort the flux in the lower end of the spectrum where the detectors are more sensitive to the incoming radiation. One other thing that needs to be kept in mind that there are three transport equation in high energy radiography simulations in contrast with the one transport equation in the x-ray tube radiography cases. This increases computational efforts significantly. Therefore, quick and easy solutions should not be expected for high-energy radiography simulations. Such simulations can be performed only with considerable computational resources that may include multiple processor platforms. Such computational resources may not be available to typical users who are planning to simulate high-energy radiography scenarios. ACKNOWLEDGMENTS This manuscript has been authored by Iowa State University of Science and Technology under Contract No. W-7405-ENG-82. REFERENCES 1. Berger M.J., "Monte-Carlo calculations of the penetrations and diffusion of fast charged particles:, in Methods in Computational Physics, Vol. 1, ed. by B. Alder, S. Fernbach and M. Rotenberg, Academic Press, New York, 1963. 2. Kahn H., "Applications of Monte Carlo", U.S. Atomic Energy Commission Report, AECU-3259, 1956. 3. Hendricks J. S., "MCNP4C2," LANL Memo X-5:RN (U)-JSH-Ol-Ol, 2001. 566 4. Briesmeister J. F., Ed., "MCNP - A General Monte Carlo N-Particle Transport Code, Version 4C," LA-13709-M, 2000. 5. Nelson W.R., Hirayama H., and Rogers D.W.O., "The EGS4 Code System", SLAC265, 1985. 6. Zerby C.D., Keller F.L., "Electron Transport Theory, Calculations and Experiments", Nucl. Sci. Eng., 27, pp.190-218, 1967. 7. Bartine D. E., Alsmiller R.G. Jr., Mynatt F.R., Eagle W.W. Jr., Barish J., "Low-Energy Electron Transport by the Method of Discrete Ordinates", ORNL-TM-3438, 1971. 8. Bartine D. E., Alsmiller R.G., JrMynatt, F.R., Eagle W.W., Jr., Barish J., "LowEnergy Electron Transport by the Method of Discrete Ordinates", Nucl. Sci. Eng., 48, pp.159-178, 1972. 9. Ligou J., "Discrete and Continous Interactions in Charged Particle Transport", Nucl. Sci.Eng.,71,pp.216-218, 1979. 10. Mehlhorn T.A., Duderstadt J.J., "A Discrete Ordinates Solution of the Fokker-Planck Equation Characterizing Charged Particle Transport", J. Comp. Phys., 38, pp.86-106, 1980. 11. Wienke B.R., "Charged Particle Differential Collision Terms in SN Applications", Nucl. Sci. Eng., 79, pp.430-432, 1981. 12. Morel J. E., "Fokker-Planck Calculations Using Standard Discrete Ordinates Transport Codes", Nucl. Sci. Eng., 79, pp. 340-356, 1981. 13. Tran T.M., Ligou J., "An Accurate Numerical Method to Solve the Linear FokkerPlanck Equation Characterizing Charged Particle Transport in Spherical Plasmas", Nucl. Sci. Eng., 79, pp.269-277,1981. 14. Przybylski K., Ligou J., "Numerical Analysis of the Boltzmann Equation Including Fokker-Planck Terms", Nucl. Sci. Eng., 81, pp.92-109, 1982. 15. Caro M., Ligou J., "Treatment of Scattering Anisotropy of Neutrons Through the Boltzmann-Fokker-Planck Equation", Nucl. Sci. Eng., 83, pp.242-252, 1983. 16. Borgers C., Larsen E.W., "On the Accuracy of the Fokker-Planck and Fermi Pencil Beam Equations for Charged Particle Transport", Med. Phys., 23, 1996. 17. Prinja A.K., Pomraning G.C., "A Generalized Fokker-Planck Model for Transport of Collimated Beams", Nucl. Sci. Eng., 137, pp.227-235, 2001. 18. Franke B.C., Larsen E.W., "Radial Moment Calculations of Coupled Electron-Photon Beams", Nucl. Sci. Eng., 140, pp. 1-22, 2002. 19. Takeuchi K., Tanaka S., Kinno M., "Transport Calculation of Gamma Rays Including Bremsstrahlung by the Discrete Ordinates Code PALLAS", Nucl. Sci. Eng., 78, p. 273283, 1981. 20. Morel J.E., "A Hybrid Multigroup/Continuous-Energy Monte Carlo Method for Solving the Boltzmann-Fokker-Planck Equation", Nucl. Sci. Eng., 124, pp.369-389, 1996. 21. Adams M.L., Larsen E.W., "Fast Iterative Methods for Discrete-Ordinates Particle Transport Calculations", Prog. Nucl. Energy, 40 (1), pp.3-159, 2002. 22. Berger M.J., Hubbell J.H., "XCOM, Photon cross sections on a personal computer", MBSIR 87-3597, 1987. 23. Lorence L.J., Morel J.E., Valdez G.D., "Physics Guide to CEPXS:A Multigroup Coupled Electron-Photon Cross-Section Generating Code", SAND89-1685, 1989. 567
© Copyright 2026 Paperzz