NUCLEAR INSTRUMENTS AND METHODS IO4 (I972) I63-I68; © N O R T H - H O L L A N D PUBLISHING CO.

A M O N T E CARLO P R O G R A M F O R E V A L U A T I N G T H E R E S P O N S E OF
A S C I N T I L L A T I O N C O U N T E R T O M O N O E N E R G E T I C G A M M A RAYS

R. J. D. B E A T T I E and J. B Y R N E

School of Mathematical and Physical Sciences, University of Sussex, Falmer, Brighton, Sussex, England

Received 15 M a y 1972

A simple technique is described for calculating the response o f a N a l scintillation counter to monoenergetic 7-rays. The m e t h o d
employs empirically determined counter parameters a n d ?,-ray cross-sections in a M o n t o Carlo routine a n d has been successfully
applied in the analysis o f internal b r e m s s t r a h l u n g spectra.

1. Introduction cross section for pair production in the crystal was

One of the main problems encountered when a scintil-
lation spectrometer is used to measure g a m m a ray
yields is to find a suitable means of evaluating the
response of the counter to a monoenergetic g a m m a ray.
The observed spectrum consists of the photopeak, the
continuous Compton distribution, the backscatter
peak, and possibly also annihilation g a m m a ray
escape peaks. The problem becomes even more acute
when the g a m m a ray spectrum under investigation is
continuous rather than being made up from discrete
lines. In this case each energy cell in the observed
spectrum contains contributions from all g a m m a rays
of energy higher than that of the cell, and some contri-
bution from those of lower energy.
The method described below was evolved to correct
for detector response in the measurement of internal
bremsstrahlung energy spectral). Once the response of
the counter to a monoenergetic g a m m a ray can be
predicted it is an easy matter to extend the treatment
to a continuous spectrum. This paper is concerned
mainly with the response of a given counter to an
incident g a m m a ray of known energy.
It is not practicable to use an analytic expression for
the response function, since in general the extremely
complicated analytic expressions which exist are
approximations not easily related to practical counter
configurations. The empirical method described below
has the advantage that it is tailored to give results for
an arbitrary system. This means for example that if
some special shielding is being used, the backscatter
contribution from this shielding can be estimated with
fair accuracy.
2,. Physical assumptions
In the work for which this method was evolved, the
g a m m a ray energy was small ( < 1.5 MeV) and so the
163
always very much less than that for other competing
processes. The results given later in this paper do not
include treatment of pair production, but an indication
is given in the flow diagram of the program where such
effects could be included. The treatment of the escape
of annihilation quanta is relatively complicated, and
adds considerably to the run time of the program. It is
therefore probably not worth including until the pair
production cross section reaches about 5% of the total
at least.
When a gamma ray enters a N a I crystal the secondary
electrons produced in the various interactions are the
particles which are directly detected. The photopeak
corresponds to the detection of photoelectrons which
absorb the full energy of the incident g a m m a ray, and
are detected with that energy decremented by the
binding energy of the electron which itself is usually
absorbed in the crystal following the emission of one
or more low energy X-rays and Auger electrons. The
Compton background observed corresponds to the
electron spectrum which results when the incoming
gamma ray is inelastically scattered by electrons in the
absorber (usually considered as being free electrons).
The kinematics of the problem show that this Compton
electron energy spectrum extends from zero energy up
to a maximum of 2E2/(2Eo+l), where E o is the
primary g a m m a ray energy in units of m 0 c 2.
Compton scattering and photoelectric absorption
are the only two interaction processes which are
considered in this treatment. However, multiple
scattering processes in the crystal, whether leading to
total absorption or escape of the gamma ray, are
included. Also some allowance is made for backscat-
tering from materials in the vicinity of the crystal.
Another assumption which is made is that a secondary
electron once produced is always detected i.e. no
164 R. J. D. B E A T T I E A N D J. B Y R N E

account is taken of escape of secondary electrons. This TABLE 1

is likely to occur only in the outermost regions of the Details of calibration sources used to determine the response
crystal and to allow for it would be very difficult and o f the Nal crystal to monoenergetic electrons released within
time consuming. the crystal.
Having made these assumptions it is necessary to
know the response of the system to a monoenergetic Source Energy (mo c e) Resolution
electron. This can be measured experimentally since the
photoelectrons produced are essentially monoenergetic
137Cs [ .296 0. 124
and give rise to photopeaks which can be shown to be
SSy 1.768 0.115
Gaussian to a high degree of accuracy. This is to be 6°Co 3.62 0.079
expected from the statistical nature of the production 6°Co 2.29 0.101
of photoelectrons in the photocathode of the photo- 2.60 0.089
22Na 1.00 0.138
multiplier tube. Knowing that the response is Gaussian,
2.485 0.089
we require a relationship between the energy of the ~°3Hg 0.133 0.324
electron and the full width at half maximum of the 0.546 0.180
response peak. 9~Nb 1.507 0.126
Several monoenergetic gamma ray sources can be (~aZn 2.172 0.096
S4Mn 1.635 0.106
used, with energies in the range 0.068 MeV-1.85 MeV.
In each case the resolution of the photopeak, defined as

Full width at half maximum

R= , (l) Since the photopeak response is Gaussian, the
Energy at maximum following relationship holds:

is measured. Table 1 gives a list of suitable calibration 2.354a

R - - - , (2)
sources, with their photopeak energies. Fig. 1 shows a E
graph of log R plotted against log E 0. The straight line
shown is a least squares fit to the experimentally where a is the standard deviation of the Gaussian
observed points. response at energy E. R is known as a function of E
from an experimental graph such as that shown in
fig. 1, and so an empirical relation between E and
Log R may be established.
When an electron of any energy is produced in the
-1.4
Nal crystal, the response of the detector to that
\,,
\
electron can be calculated. If the energy distribution
\
- 1.'6
~ 203Hg
under investigation is split up into elementary energy
,~, cells, and it is established in which cell Eo the energy
\
\ of the secondary electron falls, the smearing effect of
-1.@
\, the detector response can be calculated from the
22
Na equation
-2.0

-2.2
"/~f,
137 " > , ~
Cs
85Nb

\\ (if/
88, Ye - dE
O-o</2 7c
exp ~ - ( E - E o ) Z ~ ,

20-2o J
(3)

,.',
,,,,Y ",,f 60Co
where dE is the width of the energy cell and }rE is the
,e \ Co relative count at energy E.
-2.L 65 / . ~ ,
Each time the detection of a gamma ray produces an
electron in the NaI crystal, the electron energy is
-2.5 I I I I ~,
calculated, and the smearing process described above
1.5 1.o o .5 0.0 0.5 1.0 1.5
is carried out. The subroutine called P H O T O performs
Log E [moc2) this operation and determines how many cells on either
Fig. 1. Empirically determined photo-peak resolutions observed side of the central one have non-negligible contri-
for the various calibration sources listed in table 1. butions.
 RESPONSE OF A SCINTILLATION COUNTER
3. Description of the program
A description of the Monte Carlo method may be
found in refs. 2 and 3. Information on photo-electric
and Compton cross sections for g a m m a rays in Nal
and glass is given in ref. 4. Since the cross sections are
given in tabular form it is necessary to set up inter-
polation routines to give the cross sections at any
required energy. Since the cross sections vary rapidly
with energy, the logarithms of the energies and cross
sections are used as input values for the cubic spline
interpolation routine employed. In this way the inter-
polation can be made more accurate.
A diagram of the source-detector geometry is given
in fig. 2. The source is taken to be isotropic, and so a
direct application of the Monte Carlo technique would
be very wasteful of computer time since only a small
percentage of the simulated gamma rays created at the
source would actually hit the crystal. To avoid this
difficulty, the direction cosines of the trajectory of the
g a m m a ray are chosen so that the trajectory lies within
a cone whose apex is at the source and whose base is
the front face of the N a I crystal. The initial gamma
rays are confined to the cone by placing a geometrical
constraint on the expression choosing the direction
cosine with respect to the z-axis.
The transmission of the NaI crystal can be quite
high i.e. a large proportion of g a m m a rays incident
upon the crystal passes through it without interaction.
This is undesirable in the present program, and so the
technique is used of forcing the first collision in the
crystal. This involves no loss of generality since the
program builds up a qualitative response function
which can later be quantified by suitable normalisation.
These two procedures noted above are necessary in
order to keep the computing time of the program down
to a reasonable minimum.
Account is taken of the transmission of the crystal
by weighting each g a m m a ray appropriately. The
weighting factor is proportional to the probability p
that the g a m m a ray will be absorbed in its passage
through the crystal. This probability is given by
p = l - e x p ( - IJ2), (4)
No.I crystol Gloss
rf h hg
so~
~:/Y ,,~.
Fig. 2. G e o m e t r y o f source a n d c o u n t e r c o m b i n a t i o n .
TO MONOENERGETIC GAMMA RAYS 165
where lc and 2 are the gamma-ray path length and
mean free path respectively in the crystal.
When the first collision within the crystal has been
forced, the (x,y,z) coordinates of the collision point
are calculated. Also the decision is made as to whether
the interaction is by photoelectric absorption or
Compton scattering. This is done by comparing the
cross-section ratio O'p.E./O'tota 1 with the next random
number r in the series. If r is greater than this ratio, the
event is a Compton scattering; if r is less, the photon
has been absorbed in a photoelectric event.
If the event is photo-electric absorption, a count is
registered in the energy cell which includes energy E o.
The history of this particular g a m m a ray is now
complete, and a j u m p is made back to the start of the
routine to follow another simulated g a m m a ray. If the
event is a Compton scattering, the Monte Carlo
principle has to be used to find the scattered electron
energy i.e. the equation
r = e(eo) (5)
is solved for Ec, where P(Ee) is the probability distri-
bution function for electron energy E,. This is a very
complicated expression (given by the Klein-Nishina
formula) and would be very time consuming in a
calculation of the present type. Fortunately, Cashwell
and Everett 2) give an approximation to the inverse
function which is valid for g a m m a ray energies
< 4 mo c2.
Eo
Ee = E0 (6)
1 + Sr + ( 2 E o - S ) r 3'
where S = Eo/(1.05625 E0). This makes the determi-
nation of the Compton electron energy very much
simpler and quicker.
Once the energy of the scattered electron has been
found, the angle of scatter can be calculated, and hence
the direction cosines of the trajectory of the scattered
g a m m a ray may be determined. The point at which the
next collision within the crystal medium occurs is
calculated, and checked to see whether or not it lies
within the physical dimensions of the N a I crystal. I f
it does not, the gamma ray is considered as having
escaped from the crystal and a count is registered in the
appropriate cell for an energy equal to the incident
g a m m a ray energy minus the energy of the escape
gamma ray.
If the scattered gamma ray does not escape, exactly
the same procedure as outlined above is followed to
determine whether or not this second collision is
photo-electric absorption or Compton scattering. At
 166 R. J. D. B E A T T I E A N D J. B Y R N E
this stage the gamma ray being followed does not have
the primary g a m m a ray energy, but that retained by the
gamma ray after the original Compton scattering. If at
any stage the gamma ray is completely absorbed in the
crystal, the total energy transfer to the crystal is the
same as if the photon had been photo-electrically
absorbed at the first collision.
A check is kept in the program to see that the
number of multiple scatterings does not become
improbably high. If the energy of a scattered gamma
ray falls below 10 keV it is assumed that the gamma
ray is totally absorbed. This is a very small source of
error in the program. If the scattered gamma energy
falls below the energy of the KX-ray in iodine, another
set of absorption coefficients must be used because of
the discontinuity in the total absorption coefficient
occurring at that energy.
The response spectrum is built up by following many
simulated gamma rays until they are totally absorbed
or lost from the crystal. This gives a spectrum as would
be observed if there were no 'smearing' of the distri-
bution by the counter. The observed distribution is
calculated from this spectrum by taking the number of
counts in each cell and applying the Gaussian formula
(3) for the appropriate values of E o and a0-
An important factor which has not been discussed
above is backscattering into the N a I crystal from
surrounding materials. A gamma ray from the source
can Compton scatter in some material other than the
crystal, and then enter the crystal with a degraded
energy, thus distorting the spectrum of gamma rays
from the source. G a m m a rays scattered through angles
of over 90 ° have a fairly narrow energy spread as can
be seen from a graph of scattered gamma ray energy
against angle of scatter. This means that truly back-
scattered gamma rays which enter the crystal are
almost monoenergetic and tend to give a peak in the
crystal response around the energy Eo/(l+2Eo) (in
units of mo c2).
In principle it is relatively easy to allow for back-
scattering, but a truly representative treatment would
be prohibitive as regards computer time. In the present
scheme scattering material is treated as an absorber of
gamma rays exactly as is the N a I crystal, but the
interest lies in the scattered g a m m a rays rather than the
electrons from which they scatter. In this program the
main source of scattered g a m m a rays was taken to be
the glass windows in the container of the crystal and
on the photomultiplier tube. These were treated
together as a slab of glass, of the same diameter as the
crystal and of depth as shown in fig. 2. Two classes of
gamma ray may enter the glass; scattered gamma rays
direct from the source, or those which have already
been Compton scattered in the Nal crystal. When one
of these gamma rays interacts in the scattering medium,
the decision is made as to whether the event is a
Compton scattering or a photoelectric absorption. If it
is the latter, it is of no further interest. If the former the
angle of scatter is then calculated, and if the scattered
g a m m a ray would enter the NaI crystal, the point at
which it suffers its next interaction is calculated. If this
interaction is within the Na[ crystal, there is a back-
scatter contribution to be added to the spectrum. Such
a gamma ray entering the crystal is followed as
described above until it finishes in one of the terminal
categories.
One factor which complicates such a treatment is the
different absorption coefficient of each material,
necessitating the splitting up of the trajectory of a
g a m m a ray into different zones, in each of which the
mean free path of the gamma ray is different. Thus if a
g a m m a ray is backscattered in the glass windows, its
path length to its next collision must be calculated
allowing for the fact that there is a boundary between
media at the crystal-glass interface. Another compli-
cation is that absorption coefficients must be known,
and it is next to impossible to treat a composite
physical object such as, for example, another photo-
multiplier tube because
a) its exact chemical constitution is unknown and
so absorption coefficients are not available,
b) it cannot be zoned into simple geometrical regions.
The flow diagram for the main part of the programme
is illustrated in figs. 3a and 3b; the backscatter section
1+++++°i
absorption eoeffs.
and energies
I Call
+
interpolation
routines
Read geometrical
eonsts., primary I
enel-gy nun~hel"of
itcrat ons.
+
I Initialise I
randorn number
generator
+
Fig. 3a. Initiation procedures for main program.
RESPONSE OF A S C I N T I L L A T I O N COUNTER TO M O N O E N E R G E T I C GAMMA RAYS 167

?
Calculate ]
initial d.c. 's

Calculate ]
trajectory length
in Xtal + absorber
i
Calculate" path
length to 1st [ [11pdat . . . . . ning ]
~ - ~ t o t a l of fraction ]
,ollJsion(using ] It........ itted ]
we ighting

Cartesian
~'stal
; .
coordinates .
of flrst
collision

)ton Yes
channel P
( ~ Tally in Ep
channel

Yes
Get appropriate
(~----~ total absorption
~" ~ coefficient

c(mrrtinates of ~,o-----~length to next

pt. of collision. ] [collision

~ Tally in E - E [
channel P 3'

Calculate E~2 ]
scattered energy ]

~ .~ Tally in Ep No Yes Tally no. of

JcYes
Transpose
variables
e ° * eT1
eT1 ey2
xI ) x2
YI "* Y2
z 1 ~ z2
u 1 ~ II2
v 1 ~ v2
Ep Primary T Ray Energy Wl , '~2
E° Energy of 3~ Ray before a scattering |

~'T = Energy of T Ray after a scattering

Fig. 3b. F l o w diagram for main program.

168 R. J. D. BEATTIE AND J. BYRNE

scattered back into it as it may be estimated that such

events make a negligible contribution. When all
iterations have been completed subroutine P H O T O is
called for each energy channel and the events in that
channel 'smeared out' in accordance with the Gaussian
response. When all channels have been processed in
this way the final result is the calculated spectrum for
(EY
B )ray
y bnerg3'. the particular experimental arrangement which is being
simulated.

4. Conclusions
I' Calculate I ~e
No x V z - Get appropriate [ A Monte Carlo program has been described which
~ ? r y ~ a l ~ eoo3dl~nat3s of n e x t ' t o t a l absorption
~ I ~'°fc°l'isi°n" it oo~ff..... t. I computes the response of a NaI crystal to mono-
I I energetic 7-rays. The method makes use of empirically
Transpgse
variables determined photo-peak distributions and tabulated
e° -~ eTB
absorption coefficients and cross sections. It may be
x1 ~ x3
applied for arbitrary counter sizes and experimental
Yl ~ Y3
z1 ~ z3 arrangements. The accuracy of the procedure has been
u 1 -* u 3
checked by comparing experimental and predicted
v I -~ v 3 spectra for the 667 keV 7-ray from a source of 137Cs.
w1 -* w 3 The agreement is very close and indicates that the
method is a useful one for application in nuclear
physics experiments.

Fig. 4. Flow diagram for computing the contribution of back- References

scattering.
is shown in fig. 4. It will be noticed that only scattering
o f a primary y-ray into the N a I crystal is taken into
consideration. N o account is taken of C o m p t o n
scattered y-rays which leave the crystal and are then
1) R. J. D. Beattie and J. Byrne, Nucl. Phys. A160 (1971) 225.
2) E. D. Cashwell and C. J. Everett, .4 practical manual on the
Monte Carlo method .for random walk problems (Pergamon,
London, 1959).
3) j.M. Hammersley and D.C. Handscomb, Monte Carlo
methods (Metbuen, London, 1964).
4) K. Siegbahn, Alpha-, beta-, and gamma-ray spectroscopy, vol. l
(North-Holland Publ. Co., Amsterdam, 1965) p. 827.