You are on page 1of 10

Nuclear Instruments and Methods in Physics Research A295 (1990) 231-240 North-Holland

231

Low-energy photon scattering simulations with the Monte Carlo code ACCEPT
Fabrizio Cleri

ENEA, Dipartimento Reattori Velocl, CRE Casaccia CP 2400, 1-00100 Roma AD, Italy

Received 12 January 1990 and in revised form 20 April 1990 A version of the 3-D coupled photon-electron Monte Carlo transport code ACCEPT of the ITS system has been developed, focusing the attention on its use in the simulation of low-energy photon transmission/reflection experiments . As improvements, electron binding corrections to the Compton scattering and coherent Rayleigh scattering have been added. Biased photoelectric capture has been included to increase the calculation efficiency at low incident photon energies.
l. Introduction

Low-energy photons are used in both basic and applied science in a wide variety of applications. When X-rays are used either in computerized tomography for health physics or industrial applications, or in transmission/reflection experiments for nondestructive sample analysis, a computer simulation of the photon interaction and transport processes will always be a fundamental tool for the correct planning and interpretation of the experiment. Furthermore, coupled photon-electron simulations are of great importance when, for example, applied to the calibration of an X-ray detector crystal or to the study of radiation biasing in solid state devices . Several standard computer codes are available to solve the coupled photon-electron transport problem . The most widely used -are undoubtedly the 2-D cylindrical and the 3-D combinatorial geometry Monte Carlo [1,21 codes, owing to the uniquely accurate treatment of the physics and geometry of the transport process . In the present work we report on the improvements made to the general purpose photon-electron Monte Carlo transport code ACCEPT, of the ITS Series [11, to update it for low-energy photon scattering applications . The ACCEPT code is commonly recognized to be quite efficient and accurate in the treatment of many funda ....en tai photon-electron nroct-sse_C r1l_ I_n its basic ula.aa~wi rraavw ............. r- ...______ -~_ version, it follows atomic shell relaxation either via transition X-ray or Auger electron emission with an approximate model (it assigns all the cross-sections to the highest-Z element in a given material, based on the fact that the cross-sections are roughly proportional to Z4 or Z5; however, a more sophisticated ACCEPT-P version also exists, which fully accounts for the complex K-LM shell relaxation cascade) ; it samples the photoelectron emission angle from a Fischer distribution [41,

whereas other standard codes normally attribute the angle of the primary photon to the ejected electron ; the bremsstrahlung associated with electron-nucleus interaction includes the Elwert correction [51, which leads to a consistent enhancement of the low-energy bremsstrahlung yield; other numerical artefacts in the electron history termination, such as the inclusion of a "detour" factor accounting for the practical range of electrons whose energy falls below the fixed cut-off, insure a generally more accurate low-energy treatment than other commonly used codes. Despite this, some typical low-energy features like binding corrections to the Compton scattering and the coherent photon Rayleigh scattering are neglected . This is bound to lead to rather important discrepancies between experiment and calculations when low-energy source photons are treated, especially in diagnostic X-ray simulations or in photon backscattering from high-Z materials, as required for example in the determination of the Rayleigh/ Compton ratio for metallic alloy nondestructive analysis [6] . Moreover, because the coherent scattering and electron binding effects are confined to the very low-energy region (E :S 100-300 keV), searching for these effects can be difficult because of the overwhelming photoelectric cross-section, so for obtaining good statistics at large angles it may be necessary to use hours of CPU time. An option for the treatment of photoelectric capture in an implicit manner should thus be implemented ; for reasons to be discussed later, this implicit capture scheme with weight readjustment should work differently, according to whether the secondary electron production is considered or not . In the following, we will present the way the above effects and needs have been included in the three-dimensional module ACCEPT and in the cross-section

0168-9002/90/$0'3 .50 0 1990 - Elsevier Science Publishers B.V. (North-Holland)

232

F Cleri / Low-energy photon scattering simulations .

generating module XGEN of the ITS Series, together with some minor operational improvements, resulting in a very powerful tool for the simulation of low-energy photon experiments. 2. Low-energy photon scattering Any theoretical approach to the photon scattering from isolated atoms considers the interaction as being the difference between an initial state, with an atom in state E; and an impinging photon of energy hca;, and a final state, where the atom is left in state Ef and a photon with energy hwf appears and/or other particles (electrons or nuclear fragments) with appropriate energy emerge from the atom itself. In the photon energy range considered here (E < 300 keV) the following interactions can take place: 1) incoherent Compton scattering, in which some fraction of the incident photon energy and momentum are transferred to one atomic electron ejected from the atom, with a consequent reduction of the photon energy (very rarely more than one photon can be found in the final state) ; 2) coherent Rayleigh scattering, in which only the photon momentum is changed and the final photon energy is left unchanged (no free electrons appear in the final state) ; 3) photoelectric absorption, in which electron ejection from the inner atomic shells occurs and other electrons undergo some rearrangement, resulting in a subsequent relaxation process that leaves characteristic X-rays and Auger electrons in the final state. The total photon-atom interaction cross-section can thus be written as Qtot -QC +Q R +Qph ,

glected, as long as for low-Z elements the K-shell binding is much smaller than the photon energies, and for high-Z elements the K-shell electrons - more tightly bound - are only a small fraction of the total. The standard photon-atom scattering description is thus given in terms of the Klein-Nishina cross-section formula [7], which assumes unpolarized photons to interact with free electrons at rest: daKN( 8 ) d ) (

U) a2 2(1 2 + ktt) -2 2(1 - + + ( 1 + kit )] ' 1

k2 2

(2)

where re is the classical electron radius, k is the incident photon energy in units of mc2, and p=1-cos (with 0 the photon deflection angle) is the increase in the photon wavelength X =1/k between the scattered and incident photon in Compton units. Rewriting the "Compton shift" in terms of the initial and final photon energies results in the familiar expression k'= k[1 + k(1 - cos 0)] -1 . However, it can be of interest in some cases to have a more refined description of the Compton scattering by considering the fact that atomic electrons are actually in motion and bound to the nucleus. The overall result of such a correction will be a reduction in the forward scattering, compensated by some amplification in the backward scattering. If binding corrections are treated in the impulse approximation, it can be shown that the corrected expression for the differential Compton cross-section is ( dac(8) S(q,Z) ( daKN(0) dQ d )

where ac, (FR and Qph are the partial cross-sections for the three processes listed above, respectively . As mentioned in the introduction, the ACCEPT code is a Monte Carlo program that, given some target in an arbitrary 3-D geometry and a photon source, is able to treat all these concurrent processes, computing the resulting distributions for escaping electrons and photons in space, angle and energy, the energy absorption and charge deuosition within the target, etc. However, this treatment is approximate in some respect, especially for low-energy incident photons. In the following subsections we will discuss the improvements needed for low-energy photon Monte Carlo simulations, with respect to the standard treatment. 2.1 . Compton incoherent scattering In the common photon transport theory, electron binding effects in Compton scattering are usually ne-

where aKN (8) is the free electron Klein-Nishina crosssection, and S(q, Z) is the incoherent scattering function depending on the atomic number Z and on the momentum transfer q in units of me that, at the small values at which the binding effects are important, can be approximated as q=2k sin (2' . S(q, Z) represents the probability that an atom be raised to an excited or ionized state as a result of a collision imparting a recoil momentum q to some atomic electron (not only in the K-shell but among all the shells, as a result of the impulse approximation) [8]. Upon inclusion of the target electron motion, the Compton line is also Doppler broadened [9]. The Compton shift relationship (3) now contains a term linearly dependent on the z-component of the electron's ground state momentum QZ that, under the same assumption of

F. Cleri / Low-energy photon scattering simulations

233

small momentum transfer, allows the final photon energy to be expressed as + 2Q, sin(0/2)1 . k'= k [1 [1+k(1-cos 0)1 Analysis of this broadened "Compton profile" in highresolution experiments can provide detailed information about the electron ground-state momentum distributions. 2.2. Rayleigh coherent scattering Like Compton scattering binding corrections, the coherent or Rayleigh scattering is generally also not considered in standard photon transport calculations, because it does not lead to photon energy degradation and is forward directed, resulting in a final state almost equal to the initial state. The definition of "coherent" for this process arises from the fact that forward scattering is greatly enhanced by the coherent addition of the scattering amplitudes from bound atomic electrons, without ionization or excitation of the atom. Because the photon energy is unchanged as a consequence of this scattering process, its effect will be particularly evident in backscattering experiments where the socalled "Rayleigh peak" with the same energy of the incident photon clearly appears besides the Compton backscattering peak . Moreover, in the low-energy region the coherent photon scattering cross-section is generally much larger than the Compton-corrected one, so that neglecting this effect - as it is done in the standard ACCEPT version can lead to wrong results in some cases, because the forward scattering probability is unphysically too small. Indeed, the commonly accepted argument that there could be a sort of compensation between the forward scattering reduction due to Compton binding corrections and the amplification due to the coherent Rayleigh scattering, has recently been shown to be of very limited validity [101. Besides this, a large number of successive coherent scatterings in the low-energy region could increase the pathlength travelled by the photons appreciably, increasing the chance for absorption [11]. It should also be noted that the angular spread associated with the Rayleigh scattering distribution is not negligible, the l_ b...at the limiting angle n `.V111C1111111g Q,VVY~ 75% of the scattered VC alb=t radiation being [121 l 0.0133Z' / 3 ), 0, = 2 arcsin' E with E the incident photon energy in MeV. Eq. (7) gives, for example, 0,; = 45" for 100 keV photons on iron, implying that the distribution is not so strictly forward dominated as assumed in the above approximations.

Taking the Rayleigh scattering into account can be done by correcting the low-energy limiting form of the Klein-Nishina cross-section (the so-called Thomson cross-section) by some form factor F, as C daR (0) = 2 d1 ) [F(q, Z)]2 i e ( 1 +cos20) .

F(q, Z,) represents the probability that the Z electrons in the target atom take up and share the recoil momentum q without absorbing energy [131. 2.3. Photoelectric capture In the atomic photoelectric effect, the initial state photon disappears and an electron appears in the final state with an energy equal to the incident photon energy minus the electron binding energy . Clearly, if the photon energy drops below the binding energy of a given shell, an electron from that shell cannot be ejected; hence, a plot of the photoelectric cross-section versus the photon energy will exhibit the characteristic absorption edges. At any photon energy the lower-shell photoelectric emission probability is always dominating (the experimental ratio of total cross-section versus K-shell contribution ranges from 1.0 to about 1.24 for Z from 1 to 92), and some approximate treatments consider Kshell emission only . A more accurate treatment should include, besides higher-shell emission, the relative probability of releasing the excess binding energy either in the form of fluorescent X-rays or as Auger electrons (fluorescent yield) . Theoretical calculations of the photoelectric scattering cross-sections and of the relative fluorescence yields, based on relativistic quantum mechanics, have been gathered for all the elements in compilations, the most widely used in the scientific literature probably being the one of Storm and Israel [141 . In those problems where low-energy events are searched for (e.g. X-ray diagnostic simulation or low-energy photon backscattering), the overwhelming value of the photoelectric cross-section can represent a serious problem for a Monte Carlo simulation, in which the interactions are modelled by stochastic sampling from the appropriate theoretical distribution functions: thousands of source photons can be lost to transmission or reflection, ending with high probability in a photo11 . W taaro Telayerthvle~~ nn implicit way of representing photon capture in a Monte Carlo simulation exists : at every collision the photoelectric absorption is skipped, the photon weight is somewhat readjusted while the photon continues to undergo scattering, and an appropriate fraction of the initial weight makes scoring for any quantity of interest to account for this "masked" capture effect . Provided too large photon weight fluctuations are prevented, such technique can improve the efficiency of a Monte Carlo calculation enormously.

234

F. Cleri / Low-energy photon scattering simulations

Fig. 1. The behavior of the incoherent scattering function S(q, Z)/Z (right) and of the atomic form factor F(q, Z)/Z (left), as a function of the momentum transfer q (in A-1 ) for two limiting values of Z = 6 and 82.

THE KAHN'S ALGORITHM

qeax = -ax . table lenght ( 8 . A -1 )

a = inverse Compton wavelenght / ,, 2 - 29 .1445 A-1


A =

initial photon energy ( in mc Z )

A' = final photon energy ( in mcZ ) 0 = photon scattering angle

Fig. 2. Flow diagram of Kahn's method for Compton distribution sampling and of the rejection method used to include electron binding corrections in Kahn's algorithm .

F. Cleri / Low-energy photon scattering simulations


3 . Improvement of low-energy Monte Carlo simulations

235

3.1 . Binding corrections and coherent scattering Inclusion of the Compton binding corrections and Rayleigh scattering in the ACCEPT module (we will call this new version ACCEPT-E) has been performed by adapting some parts of the CQLIDP subprogram from the neutron-photon Monte Carlo code MCNP [15] as follows. The incoherent scattering functions S(q, Z) [16] and the coherent scattering squared form factors F 2 (q, Z) [17], are tabulated in a dedicated file, on two grids for q and q 2, with 21 points between 0.0 and 8.0 -1 and 55 points between 0.0 and 36.0 -2 respectively, for all elements with Z =1 to 92. In the simulations, the two grids are indeed kept much sharper and the values of the S and F 2 functions are correspondingly log-log interpolated by the cross-section generating module XGEN. Their typical behavior is shown in fig . 1 for two limiting Z-values.

The capture survival probability includes both coherent and incoherent scattering probability, and the relative probability of the two scattering processes is tabulated up to 10.0 MeV (above this energy the cross-section for coherent scattering becomes negligible with respect to Compton scattering even for Z = 100). Data relative to the coherent scattering cross-sections are taken from the Storm-Israti compilation [14]. Sampling the Compton scattering distribution in the standard ACCEPT module is performed with the celebrated Kahn algorithm [18]; the inclusion of binding corrections has been realized by slightly modifying this rejection technique, so that only one further random number call is required as shown in the flow diagram of fig . 2. Rayleigh scattering distribution sampling is performed in a completely similar way from the Thomson cross-section rather than from the Klein-Nishina one (see fig. 3). In the calculation of the total Rayleigh scattering crosssection for a material made up of many elements, the atoms of the various species are supposed to act independently and their form factors are weight-summed ; this approximation can have some impact for particular

2 = max . table lenght gnax

( 36 . A-

=2a2 a = inverse Compton wavelenght k = photon energy ( in mc 2 ) 0 = photon scattering angle (

/ V = 29 .1445t-1

1(4) - S

l0

14

2
F(4' , Z)dq~~

-1 is readily obtained 4(x) being a tabulated function, .V by inverse interpolation : GIVEN COMPUTE THEN 3(xi-1)<y<_ .4(x,) P= 9 .4(x i) J(xi-I) - .0(x 1 )

x= .-lly)=xt + P(xt-1 - xt)

Fig. 3. Flow diagram of the rejection method for Rayleigh scattering distribution sampling.

236

F Cleri / Low-energy photon scattering simulations .

0.8

Na

without binding corrections . Source photons are the 24'Am source 59.6 keV gamma-rays from a collimated normally impinging on the target plane. The effect of the electron binding (continuous line) versus free elec1 .20

AI
0.

1 .10 1 .00 0 .90 0 .80


_e -- ._________________ .__ .___________ ____ .__ .__________ . .___________

-0 .5

cos(8)

0.

0.5

1 .0r

1 .20 2
,

Pb

-6 .5

0. cos(6)

1 .ka0

--- .'.__ . . . . . . .__ . . . . . .__ ._ .___ . . . .__ .__ . ._______ . -----

- ar

________ . ._____ . . ._ .

Fig. 4. Angular distributions of Compton-scattered 59.6 keV photons from a plane target of sodium (upper figure) and uranium (lower figure) . Continuous and dashed curves are obtained using bound electron and free electron Compton cross-section, respectively . The target is thin enough to insure single-scattering only to occur. aggregates, like water, in which the molecular correlation is not negligible (see for example ref . [19]). The target electron momentum in Compton-bound scattering should be in principle extracted from a suitable Fermi-Dirac distribution . However, as far as the final result should be of a Gaussian broadening of the Comp:on peak, it has been chosen to derive the electron momentum modulus from the distribution
l02 l 1 ; ti P(Q) dQ = exp' -E F dQ, Ks7.~2m

0.80

1 .30

1 .20

centered at the Fermi energy E F and smeared over a FWHM of the order of KT Having sampled the electron momentum modulus Q by direct inversion of eq. (9), its z-component is readily obtained by spherical polar projection, sampling a random angle ~ between 0 and rr so that QZ = Q cos ~. In fig . 4, the ACCEPT-E calculated angular distributions of Compton-scattered photons corn 'L4 a and I' plane targets after a first collision are shown with and

0.90 0--

Fig . 5. Ratios of the average volume photon flux including coherent and bound electron incoherent scattering (continuous line), and including binding corrections only (dashed line), with respect to the reference calculation including only free
Al (upper), Pb (mid)

depth (MFP)

cc.trvn Ctropon scattcr t,. TgCts a[E ` MF' hick slabs of

and water (lower) .

F. Cleri / Low-energy photon scattering simulations

23 7

tron scattering (dotted line) is evident, and larger for the high-Z element, as expected. Examples of the relative effect of coherent and incoherent scattering are shown in fig . 5, where photon transmission through 5 MFP thick slabs of aluminum, lead and water is simulated. The figures represent the ratio of the average volume photon flux (CM-2S-1) between runs in which coherent scattering and binding effects (continuous line) or binding effects only (dashed line) were included, and runs with free electron Compton scattering only. The source is the same as in the previous example. The effect of coherent scattering tends to shorten the penetration lenght of the photon beam more extremely for high-Z materials as expected. The opposite effect due to binding corrections to Compton scattering is also evident, but it does not wholly compensate the increased flux attenuation, as is supposed by simplified treatments. The behavior of water is singular, as the binding correction looks to overwhelm the effect of coherent scattering with a resulting effect of enhanced penetration . A similar result had been arrived at in ref. [10], where curves of the buildup factor, instead of the photon flux, were presented. 3.2. Biased photoelectric capture The standard ACCEPT module allows to switch off the secondary electron production to save computing time in problems where electron transport can be neglected with respect to other effects. The implicit photon capture scheme has to be kept different according to whether secondary electrons are to be followed or not because, if secondary electrons are followed, the complete exclusion of photoelectric events would lead to a parallel exclusion of bremsstrahlung and shell-relaxation radiation . This biased photoelectric capture is based on the concept of particle weight, common to any Monte Carlo simulation . The weight represents the contribution to an observable quantity from a physical particle track : the two coincide if the weight is unity. For various computational purposes, the weight of a particle can be altered during the calculation ; a true particle with current weight w can be split into n pseudoparticles with weights such that their sum equals w. We define a weight switch parameter R as R _ clot - ~ph
Qtot

fractions of energy EW(1 - R), and flux W(1 - R), are tallied, and the noncaptured weight W' continues to undergo other events . If secondary electrons are followed, the photoelectric cross-section Qph is exchanged with its complementary Qtot - oph and the game is continued as usual ; the weight is changed to W
1-R

WR

(12)

if the biased photoelectric event takes place, or to W ,. == W(1 -R) R


(13)

if the particle undergoes any other kind of event. The exchange between the total and photoelectric cross-se( tions is a sort of "uniform" exponential transform, by which the too large values of photoelectric cross-sections (R is near 0 for any material, below 50-150 keV) can be avoided. However, this process can induce considerable changes in the running particle weight, which will range from very small to very large values . To avoid unconditional weight increase or decrease, that could force the calculation in a long dummy loop, Russian roulette is performed at each nonmultiplicative event as follows : a weight test factor WT is obtained, via the current weight W and a bound factor WB of the order of 10-3, as WT = WBW if W tends to increase, or as

WT

WB

if W tends to decrease, and is used as Russian roulette parameter (i.e., a random number between 0 and I is compared with WT and, if smaller, the particle is "killed") . If the particle survives, its weight is now readjusted as W W'= 1- WT .
(14)

(10)

the partial cross-sections being defined in eq. (1) . If secondaries are not included, the photoelectric capture is completely overriden ; at each collision the photon weight W is readjusted as

It is seen that, as the weight becomes increasingly smaller or larger, the parameter WT brings about more and more hazardous Russian roulette games (note that, owing to the restrictions on the random number range, the running weight can tend to infinity from the right, but is pre-vented to assume negative values) . This expedient has been shown to be very useful in maintaining a not too large weight range, thus reducing also variance fluctuations. An example of the use of this option is given in table 241 Am source 1 for the following problem: the same photons impinge on a small iron disk with a depth of 0.5 MFP, subdivided into five zones of equal depth, within which the energy deposition is to be cstimated . In table I we report, for the four different calculations

238

F. Cleri / Low-energy photon scattering simulations

Table 1 Summary of the iron disk energy deposition experiment simulation for the four different options. Zone Energy deposition (MeV/g) Photon + electron Analog 4 .765 1 a) 4 .387 1 4 .111 2 3.674 3 3.169 4 4.73E - 2 b) 5.65E - 3 1.04E -2 1.15E -4 7.27E - 3 5.42E -3 2.2 4 2370 0.10 Photon only Implicit 4.776 4.299 4.122 3.749 3.318 1 1 1 2 4 Analog 4.833 4.375 4.171 3.750 3.383 1 1 2 2 3 Implicit 4.699 4.419 4.092 3.791 3.333 1 1 1 2 2

1 2 3 4 5

Average energy/event PE CO RA BR RX AU Statistics (a>


Amax T a) b)

4.70E - 2 5.54E - 3 1 .19E - 2 1 .10E -4 7.15E - 3 5 .85E -3 1.8 4 2110 0.18

4.86E - 2 5.67E-3 9.52E -3 0.0 6.47E - 3 6.47E - 3 1 .8 3 91 4.1

5.89E - 2 5.81E-3 1.14E -2 0.0 0.0 0.0 1.4 2 53 11.9

Means 4.765 f 1% (MeV/g/source photon) deposited in this zone. Means 4.73 x 10-2 (MeV/particle) average loss due to this event.

(photon + electron and photon only with analog or im-

plicit capture), the values of energy deposition by zone and the average values of energy,/ event (PE _ photoelectric, CO = Compton, RA = Rayleigh, BR = bremsstrahlung, RX = X-ray, AU = Auger electron) together. with statistics data such as the average standard deviation (a), the maximum standard deviation Amax in

percent of the result, the CPU time -r in seconds on the IBM-3090/120S computer, and the figure of merit Q = ( o2 T M)-I , M being the computer scalar speed (equal to 8.4 MIPS for the 120S). Note that the Q-value allows us to compare the calculation efficiency no matter whether the game was biased or not, and quite independently on the machine the problem was run on.

1E-~01

1E-03

1E+00

-Cl

aD

1E-01

E
C

F 1E-02~

E-04~

1 1 1

1 1 1

0 (degrees) energy (KeV) Fig. 6. Energy (right) and angular (left) distributions of forward transmitted 59.6 keV photons through a 0.5 MFP thick iron target, calculated with the unbiased capture (dashed) and with the implicit capture method (continuous line) .

20

30

40

50

80

90

~0

35

40

45

J0

55

60

F. Cleri / Low-energy photon scattering simulations It is seen that the overall energy deposition results are consistent . In the photon + electron case, the implicit capture technique largely increases the number of Compton and Rayleigh events allowing for better statistics. The running time is obviously smaller in the photon only runs, with a considerably better Q in the implicit capture option. In this latter case, the energy balance among different events is somewhat modified, the whole population of relaxation radiation being cumulated onto the photoelectric capture scoring as described in eq. (10) and below . This seems to have no large impact on both angular and energy distributions as can be seen from fig . 6. Also note that the implicit method gives Q-values of about 3.5 and 50 times larger, respectively, than the unbiased capture method .

239

4. Concluding remarks Benchmarking a computer code against experimental results is a hard but necessary step for the assessment of its successful operation. Comparison between calculation and experiment in the low-energy range is often complicated by experimental difficulties (like the onset of Laue-Bragg scattering for uncollimated beams, insufficient knowledge of the exact sample thickness, sensitivity to the orientation of the sample surface with respect to crystallographic directions, and to the rolling texture for rolled foils). Moreover, in the energy range where the cross-sections for the photoelectric effect, Rayleigh scattering and Compton scattering become comparable, rather poor agreement is known to exist

140 120-' 100-

Ag-K (-V-) A9-Kp Aq-Ka 189 Key

Cmnptm pede 18.5 Key elmtc " 59.6 Key

25.1 Key

6040~ 20 0 i20 30 40 50 energy (KeV)

between theory and experiment [20]. This is only partly due to the neglect of thermal-diffuse scattering [21], and mainly due to incorrect assessment of the Comp' on component of the attenuation coefficient : in ref. [20] it was shown that, despite the inclusion of electron binding corrections, a residual discrepancy between theoretical and experimental Compton cross-sections persists, roughly above 150 keV (examples were given for Z = 29, 50 and 82), also when very special precautions are kept on the experimental side. A typical application example of the ACCEPT-E code is reported in fig. 7, where we show the results of the simulation of a 59.6 keV photon backscattering experiment on an Ag (26.896)-Cu (73 .296) alloy target (the continuous line represents the calculation ; the dotted line represents experimental data from ref . [6]). The experiment registered the backscattered radiation under an angle of 155* t 12"; at these angles, the energy of Compton scattering photons ranges between 48.3 and 49.2 keV about 2% coming from Doppler broadening. The two curves have been normalized so as to give the same area for the Compton peaks. The Ag-K line and the Rayleigh line have been statistically smeared over a larger energy interval by a Gaussian weighting function (from the calculation they arose as very high, infinitely thin lines at a well defined energy) preserving the peak areas. The calculated Rayleigh/ Compton ratio (i.e., the ratio of the two integrated peak counts) is 0.09 5%, which is in quite good agreement with the experimental value of 0.075 15`x. Although less sophisticated codes exist for the calculation of gamma-ray buildup factors for shield design, it can be useful to assess the impact of coherent scattering and binding corrections to incoherent scattering on the buildup factors. We have thus performed sample calculations of gamma-ray buildup factors (defined as the ratio of the complete dose calculation versus the pointattenuation dose calculation) for some cases where a comparison can be made, with and without the physics refinement described above. In table 2 we report such calculations for a parallel beam on a plane target, for G.06 MeV photons on lead and for 0.1 MeV photons on iron. For comparison, similar results from ref. [10] (Monte Carlo calculations with EGS4) and ref. [I1] (discrete ordinates calculations with ASFIT) are reelectron binding corrections, normally neglected in this kind of calculations, can play a consistent role for very low- and very high-Z materials. The results are globally consistent, also if some discrepancy appears at very deep penetration where the variance is also worse . It must be rioted that the ACCEPT code is quite inefficient. in such problems, due to the lack of specific variance reduction techniques for photon transport . An effort to implement a very efficient technique of coupled spatial splitting and Russian roulette algorithms is
-~nrtrti Tt is r~nnfirrnPA that nnharPnt cr-attprino anti

60

70

Fig. 7. Gamma ray spectrum of the 155 scattered radiation of 59.6 keV rays by an Ag-Cu (26 .8-73.2) alloy . Elastic and inelastic peaks are shown, together with experimental Ka and K fluorescence peaks (from ref. [6]). The calculated Ag-K peak is displaced due to the "highest-Z" approximation (see text) .

240

F. Cleri / Low-energy photon scattering simulations

Table 2 Comparison of buildup factors in iron and lead medium for plane parallel beam (EGS-4 from ref. [10], ASFIT from ref . [ll]). B = buildup for free electron Compton scatttering, Bc = buildup with binding corrections and coherent scattering included. Iron 0.1 MeV Lead 0 .06 MeV Data Depth (MFP) Be & B B 1.270.002 1 .260.004 1.02 0.001 1.03 O-001 0.5 EGS4 ) 1.260.005 1.260.005 1.0190 .001 1.0430.005 pr. work '~ 1.270.002 1.270.002 1.02 0.001 1.03 0.001 EGS4 b) b) 1.03 1.24 1.27 1.02 ASFIT 1.0390.005 1.260.005 1.260.005 1.0190.001 Pr. work b) 0.9320.003 1.830.020 1.900.010 1.04 0.010 EGS4') 5.0 0.9440.007 1.930.050 1.970.050 1.0480.020 Pr. work b) 0.9190.006 1 .830.020 1.720.020 1 .04 0.010 EGS4 1.73 0.903 1.82 1.04 ASFIT b) 0.9280.007 1.930.050 1.780.050 1.0480.020 Pr. work b) 0.7640.008 2.230.060 2.180.020 1.07 0.020 10.0 EGS4') 2.290.074 2.190.080 1.0730.045 0.6650.050 Pr. work 0.7330.021 2.230.060 1.740.020 1.07 0.020 EGS-4 b) 0.689 2.17 1.74 1.06 ASFIT b) 2.290.074 1.980.080 1.0730.045 0.6270 .050 Pr. Work b) ') Bound electron cross-section is used for Compton scattering. b) Free electron cross-section is used for Compton scattering. under way, which will greatly enhance the ACCEPT-E performances also in this respect. Acscnowledgements The author is indebted to Dr. Maurizio Angelone for useful suggestions. Critical reading of an early version of this manuscript by Dr. Kenneth W. Bum was much appreciated . References [1] J.A. Halbleib and T.A. Melhorn, Nucl. Sci. Eng. 92 (1986) 338. [2] W.R. Nelson, H. Hirayama and D.W.O. Rogers, Stanford Linear Accelerator Center report SLAC-265 (1985). [3] T.M. Jenkins, W.R. Nelson and A. Rindi (eds.), Monte Carlo Transport of Electrons and Photons (World Scientific, Singapore, 1989) chs. XI and XIV . [4] RR Roy and RD. Reed, Interactions of Photons and Leptons with Matter (Academic Press, New York, 1968). [5] G. Eiwert, Ann. Phys. 34 (1939) 178.
(61 R rr_carPn Meel inctr anti Moth 170 10R11 rAS [7]

O. Klein and Y. Nishina, Z. Phys. 52 (1929) 858 .

1892. [9] M.J. Cooper, Rep. Pmgr. Phys. 48 (1985) 415. [10] H. Hirayama and D.K. Trubey, Nucl. Sci. Eng. 99 (1988) 15. [ll] K.V. Subbajah, A. Natarajan and D.V. Copinath, Nucl. Sci. Eng. 101 (1984) 352. [12] RD. Evans, in: Handbuch der Physik, (Springer, Berlin/G5ttingen, 1958) vol . XXXIV, p. 218 . [13] D.T. Cromer and I.T. Waber, Acta Crystallogr. 18 (1964) 104. [14] E. Storm and H.I. Israel, Nucl. Data Tables A7 (1970) 565. [15] Los Alamos Monte Carlo Group (X-6), MCNP - A general Monte Carlo Code for Neutron and Photon Transport, RSIC code package CCC-200, Oak Ridge Nat . Lab. (1983). [16] J.H. Hubbell, H.A. Gimm and I. Overbo, J. Phys. Chem. Ref. Data 9 (1980) 1023. [17] D. Schaupp, M. Schumacher, F. Smend, P. Rullhusen and J.H. Hubbell, J. Phys. Chem. Ref. Data 12 (1983) 467. [18] H. Kahn, U.S. Atomic Energy Commission report AECU3259 (1956). [19] P.C. Johns and M.J. Yaffe, Mod. Phys. 10 (1983) 40. [20] D.C. Creagh, Nucl. Instr . and Meth. A255 (1987) 1 . [21] H. Sano, K. Ohtako and Y.H. Ohtsuki, J. Phys. Soc. Jpn 27 (1%9) 1254 .

[8] D.T. Cromer and J.B. Mann, J. Chem. Phys. 47 (1%7)

You might also like