MONTE CARLO SIMULATION OF SIGNAL –TO –NOISE (S/N) RATIO IN NDE OF MILD

STEEL (MS) CORROSION BY GAMMA RAY SCATTERING

Shivaramu1, N. Chitra1, L. Thilagam2 and K.V. Subbaiah2

1
Radiological Safety Division, Safety Group, Indira Gandhi Centre for Atomic Research
2
Safety Research Institute of Atomic Energy Regulatory Board, Kalpakkam
e-mail: shiv@igcar.gov.in
Abstract
Monte Carlo (MC) simulations have been done by using the Monte Carlo N- Particle (MCNP) radiation
transport code in calculating the 661.6 keV gamma scattered intensity from the normal and corroded MS
flat plates with corrosion varying from 6.6 to 60%. The dimensions of the MS plates and their distances
from the collimator edge of the 137Cs radioactive source and HPG detector set up with the surrounding
lead shielding, the size of the collimators for both source and detector and the angle of scattering are
incorporated in to the modeling. A simple and straightforward reconstruction algorithm developed to
correct for attenuation of the incident and scattered photons in the material surrounding each scatter site.
Using this algorithm the densities of the normal and corroded regions of the MS flat plates are
reconstructed directly from the scattered intensities. Two cases of these simulations, one for a normal MS
plate and another one for a MS plate with 6.6% corrosion are compared with the experimental ones. A
good agreement in the shape of the Pulse Height Spectrum (PHS) is obtained between the experimental
spectra and those of MC simulated ones. The quantitative estimates of the values of S/N ratio derived
from these simulated scattered spectra i.e., the ratio of count rate of normal MS plate to corroded one vary
from 1.054 to 2.229 for 6.6 to 60% corrosion respectively.

1. INTRODUCTION
The gamma scattering method is a viable tool for inspecting material since it is an interaction which is
strongly dependent on the electron density of the scattering medium, and in turn, its mass density.
Therefore the information obtained by this technique is strongly related to the material density, thus
allowing changes in the material uniformity to be monitored. The material density is obtained from point
to point, and it is not necessarily complex mathematical models which determine the density of the test
object. By scanning a well-defined volume element through a plane of interest in an object, using a raster
motion, it is possible to obtain density distribution in this plane. In this process each recorded signal
comes from a different depth, and in this way, it is possible to get information about the density profile.
This approach can be widely applied, e.g., for nondestructive material examination in industry and
aviation, detection of landmines and explosives, investigating the constitution of archeological artifacts,
etc.. Hence, gamma scattering enables the detection of local defects and the discrimination between
materials of different density and composition.

Corrosion detection is a subset of the larger fields of Non-Destructive Evaluation (NDE) and Non-
Destructive Testing (NDT). Many of the technologies of NDE/ NDT lend themselves to the detection,
characterization and quantification of corrosion damage and generally the method for corrosion detection
depend on the reduced thickness of the component following corrosive attack. This is normally the
underlying concept for corrosion detection by NDT methods such as X-radiography, ultrasonics and eddy
currents. No single means of corrosion detection is either ideal or suitable for all forms of corrosion.
There are major advantages and disadvantages of the primary corrosion detection and characterization
technologies, as well as the corrosion mechanism it is used to detect. The need for advanced techniques
for detection and evaluation of embedded corrosion and a class of sub-surface defects that require access
only to the one side of any material or structure relatively thick to be inspected has drawn attention to X-
ray or gamma backscatter as a desirable choice [1-4]. The concept is based on the detection of scattered
radiation produced from a collimated beam aimed at the object. Corrosion results in chemical (oxidation)
and density changes, both of which affect X-ray or gamma scattering and absorption.

In the present work MC simulations have been done by using the Monte Carlo N- Particle (MCNP) code
in calculating the gamma scattered intensity from the MS flat plates and compared with the experimental
ones. These simulations were performed in support of the experimental measurements helped provide an
understanding of the physics of gamma scattering NDE in addition leading to optimization of design
parameters such as source and detector collimation angles, source and detector offsets, and scan step size,
as well as sample conditions and corrosion region size, shape, and density and for deriving the
quantitative estimates of S/ N ratio for different rates of MS corrosion. A simple and straightforward
reconstruction algorithm developed to correct for attenuation of the incident and scattered photons in the
material surrounding each scatter site is described. It provides accurate reconstruction of the Compton
scattering data through an adequate correction of the absorption phenomena. Using this algorithm the
densities of the normal and corroded regions of the MS flat plates are reconstructed directly from the
scattered intensities. A detailed procedure to reconstruct the densities and an unambiguous interpretation
of the signal I(P) as a function of material density at any point of the object being inspected is given. In
this method the ρ of the target need not be known and only the knowledge of the target material’s mass
attenuation coefficients (composition) for the incident and scattered energies is enough to reconstruct the
ρ of the each voxel of the specimen being studied.

2. MONTE CARLO CALCULATIONS

The MC simulations have been done which takes into account the detailed characteristics of the source,
detector and the scatterer, in calculating the PHS. The MCNP4C [5] radiation transport code is applied to
perform the calculations in this work and it is a general purpose, three-dimensional general geometry,
time-dependent code, which is used to calculate coupled neutron-photon-electron transport in bulk media.
For photons, the code accounts for incoherent and coherent scattering, the possibility of fluorescent
emission after photoelectric absorption, absorption in pair production with local emission of annihilation
radiation, and bremsstrahlung. MCNP provides a nearly predictive capability of how radiation interacts
with matter. In MCNP simulations, each particle (photon) is tracked from creation until termination with
all interactions based on physics models and cross-sections, and all decisions (location of interaction,
scattering angle, etc.) are based on pseudo-random numbers. Usually, the results of a simulation are
normalised per starting source photon. New source photons are randomly created until a preset number of
histories are tracked and the simulation is ended.

The desired result in the simulation component of this study is a PHS since it produces the distribution of
the energy deposited in a “cell”, i.e. the gamma ray energy spectrum in a physical model of a detector.
The PHS simulations are implemented in MCNP in the so-called “f8 tally”. This estimator is based on the
following fundamental approaches: i) the energy bins in gamma ray spectra simulations correspond to the
energy deposited in the detector cell summed over all the tracks of a history; ii) no integral is evaluated,
but the deposited energy in a cell is calculated based on the detailed microscopic radiation tracking; iii) If
the photon does not undergo interactions in a specified cell, no energy is deposited and the photon does
not contribute to the PHS. Hence for each history, only one count is added to the spectrum. The pulse
height tally was used to obtain the energy distribution of pulses created in the volume of the germanium
crystal. The f8 energy bins correspond to the total energy deposited in a detector in the specified channels
by each physical particle. Thus, the detector geometry is modeled with the MCNP code, which simulates
the detection process to obtain the spectral shape. To obtain a good representation of the experimental
spectra, a standard MCNP Gaussian broadening of the pulse-height response is used to define the Full
Width at Half Maximum (FWHM) of the crystal and is given by:

FWHM (MeV) = a + b ( E + c E 2 ) .

2
Here a, b and c are constants and E is the energy of the photon in MeV.

Two 137Cs radioactive source capsules each of strength 76.96 GBq (total 153.92 GBq) deposited inside
stainless steel capsule and the HPGe detector is modeled with the surrounding lead shielding and a
0.007m beam collimator for both source and detector. The HPGe detector consists of a crystal of size
0.066 m × 0.066 m surrounded and sealed by an aluminum layer 0.002 m thick in front and 0.001 m on
sides. The dimensions of the MS flat plates and their distances from the collimator edge of the source and
detector set up, the size of the collimators and the angle of scattering are incorporated in to the modeling.
The MC PHS contains 93 bins, each with a width of 3 keV and the photon energies ranged from 25 to 304
keV. The measured PHS was rebinned to the same energy grid as the MCNP calculation for the purpose
of comparison. For the pulse-height (energy-deposition) estimates used in this study, the source biasing
represents the only feasible method to improve computational efficiency. Herewith, the directional
(source) biasing function is used to make the source emitting particles to move as a fine beam towards the
MS flat plates. Two different MCNP simulations are carried out to study the normal and corroded MS flat
plates. Each simulation is run with 2.1 billion source particles. This number of source particles was
chosen because the absolute efficiency value approaches a steady value and the relative variance
decreases as the number of source particle increases. The calculated spectrum is normalized to the
experimental ones absolutely by multiplying the number of photons emitted by 137Cs source during a time
interval for which the experimental PHS is recorded. To do this, the values of the simulated PHS are
multiplied by a factor of 153.92 GBq (source activity) × t s (acquisition time of experimental spectrum),
which allowed direct comparison with the experimental spectrum corrected for background or the
experimental data are divided by the same factors to compare with the MCNP simulated PHS.

3. EXPERIMENTAL PROCEDURE
The scanning system consists of source and detector units and a four axis job positioning system. The
137
Cs radioactive source of strength 153.92 GBq with a lead shielding and a collimated and shielded 50%
efficiency coaxial HPGe detector providing high resolution energy dispersive analysis of the scattered
spectrum are mounted separately on the source and detector sub assemblies of 6- axis system respectively.
The voxel to be analyzed is geometrically established by the intersection of the incident and scattered
beams, as defined by the input and output collimators of source and detector respectively and depends on
size of the collimators employed and on source- structure, detector- structure distances and can be easily
chosen by proper adjustment of X, Y, Z and θ positions of 6- axis and 4-axis job positioning systems. The
scattered intensity from the specified voxel of the MS flat plates is detected using HPGe detector and the
PHS was accumulated and displayed using an 8K-channel analyzer which was interfaced with a PC for
data storage and analysis. Two MS flat plates, one normal and another corroded of dimensions 0.235 m x
0.155 m x 0.01 m are chosen for corrosion detection and quantification. They are placed at a distance of
0.772 m from the collimator edge of the source set up and the detector’s distance from the plates is
1.448m with the angle of scattering 115.5º. The incident photon energy emitted by 137Cs is 661.6 keV and
the energy of the scattered photon is 229.6 keV. The diameter of both the source and detector collimators
is 0.007 m and the resulting voxel is 27.103 x 10- 6 m3. The MS flat plates are mounted on the 4-axis job
positioning system and the scattered spectra from these plates are recorded with the number of channels
nearly equal to 2000 and each channel width is 0.374 keV.

4. GAMMA SCATTERING TECHNIQUE AND A ALGORITHM FOR RECONSTRUCTION

OF THE OBJECT’S DENSITY

A simple and straightforward reconstruction algorithm developed to correct for attenuation of the incident
and scattered photons in the material surrounding each scatter site is described. It provides accurate
reconstruction of the Compton scattering data through an adequate correction of the absorption
phenomena. Using this algorithm the densities of the normal and corroded regions of the MS flat plates

3
are reconstructed directly from the scattered intensities. As per the present experimental arrangement we
will consider backscatter geometry (Fig.1.) for reconstruction of the object’s density. For simplicity, we

Fig. 1. Backscattering geometry for reconstruction of the object’s density

will assume that the photons from the source undergo only one scattering. We will also ignore the small
attenuation by the air between the object and the backscatter system. The number of detected photons
depends on the number of scattered photons, as well as on the attenuation all along the path within the
material. The path of photons from the source to the detector can be divided into three stages which
determine their contribution to the measured signal. The first stage is the photons travel from the source to
the scattering point P along path α. In a homogeneous material gamma rays are attenuated according to
the Beer-Boucher law:

I1 = Io exp – [{µ(E)/ ρ} ρ x], (1)

where I1 and Io are the transmitted and incident flux, respectively, µ(E)/ρ is the mass attenuation
coefficient of the material for photons of energy E, ρ is the density of the material and x is the length of
path α. The second stage of development for our signal is the scattering towards the detector which
occurs at point P. The scattered flux, I2, is determined by:

I2 = I1 dσCKN(E,θ)/ dΩ S (E,θ, Z) ΩD ρe(P) ∆L, (2)

where dσCKN(E,θ)/ dΩ is the differential scatter cross section as governed by the Klein-Nishina formula
(a function of the incident gamma energy, E and the scatter angle θ), S is the incoherent scattering
function (a function of the incident gamma energy E, the scatter angle, θ and atomic number of the
element, Z ), ΩD is the solid angle subtended by the detector and its collimator, ∆L is an element of path α
in the vicinity of point P (the voxel thickness), and ρe(P) is the electron density at point P. The electron
density at P is the material property we are attempting to measure. It is proportional to the physical
density ρ according to the formula:

ρe = ρ N Z / A, (3)

where N is Avogadro’s number, Z is atomic number and A is atomic weight. The detected spectrum is
complicated by two effects. The first one is due to the preferential removal of photons from a specific
spectral region due to the absorption in the sample and the second one is due to the generation of
secondary radiation due to multiple interactions. The third stage of development is the transport of the

4
scattered photons back through the materials towards the detector. The signal is further attenuated, so
that:
I3 = I2 exp – [{µ’(E’,θ)/ ρ} ρ x’], (4)

where I3 represents the flux intensity reaching the detector, µ’(E’,θ)/ρ is the mass attenuation coefficient
for scattered photons of energy E’ , now a function of θ by the virtue of the Compton energy shift at P,
and x’ is the length of path β. Combining the expressions for the three stages, the signal intensity
corresponding to point P can be written as:

I(P) =Io exp– {(µ/ ρ} ρ x} dσCKN(E,θ)/dΩ S(E,θ,Z) ΩD ∆L ρe(P) exp– {(µ’/ρ) ρ x’}. (5)

The ratio NZ/A in eqn.(3) is nearly constant among elements. This and other parameters in eqn.(5) which
are constant in a given inspection set- up can be combined into a system parameter. Combining the
exponentials in the same set-up, we can rewrite eqn. (5) in terms of density ρ, and a system parameter k:

I(P) = k ρ(P) exp -{(µ/ ρ) ρ x + (µ’/ ρ) ρ x’}, (6)

Where k = Io {dσCKN(E,θ)/ dΩ} S(E,θ, Z) ΩD ∆L N Z / A.

According to (6) we can expect an unambiguous interpretation of the signal I(P) as a function of material
density at point P. Non- homogeneous materials (or those with irregular surfaces) present a different
problem. In these cases µ becomes a function of position, which requires that the attenuation terms be
integrated over the path lengths involved. Equation (6) becomes:

I(P) = k ρ(P) exp -{∫α (µ/ ρ) ρ x dx + ∫β (µ’/ ρ) ρ x’ dx’ },

= k ρ(P) AF, (7)
where the attenuation factor, AF = exp -{∫α (µ/ ρ) ρ x dx + ∫β (µ’/ ρ) ρ x’ dx’ }. Here x and x’ are positions
along paths α and β, respectively.

The geometry for calculating the AF for the voxel of thickness T, is shown in Fig.2. Let C be the point in
the centre of the voxel and t' be the corresponding thickness of the MS flat plate at which scattering takes
place. Then for this case the incident and scattered paths are given by x = t' / cosθ1 and x' = t' / cosθ2
respectively.

Fig. 2. The geometry for calculating the attenuation factor

5
The AF is given by:
T
AF = (1/ T) ∫ {exp- (µ/ ρ) ρ (t' / cosθ1) dt' + (µ’/ ρ) ρ (t' / cosθ2) dt'}. (8)
0
= (1/T) [1- exp - {(µ/ ρ) (t'/ cosθ1) + (µ’/ ρ) (t'/ cosθ2)} ρ] /
{(µ / ρ) (t'/ cosθ1) + (µ’/ ρ) (t'/ cosθ2)} ρ. (9)

5. RESULTS AND DISCUSSION

The photo peak intensity of PHS plotted as a function of x-axis position of normal and corroded MS flat
plate is shown in Fig. 3. It can be seen clearly from this figure that the intensity decreases as one move
from normal region to the corroded region. The measured gamma scattered photo peak intensity from
corroded MS flat plate decreases up to 7.9% for the same size voxel compared to a normal MS flat plate
and this corresponds to an average density loss of 0.524 x 103 kg / m3 (6.66%) due to corrosion in a MS
flat plate of size 0.235m x 0.155m x 0.01m. This can be explained based on the following reasoning. As
the normal MS flat plate intersects the sensitive volume, there is an increase in the total electron density
of the material comprising of the sensitive volume, hence an increase in detector response. As the
sensitive volume intersects the corroded MS flat plate, there is a reduction in the total electron density
and, therefore, a decrease in the detector response. The ρ values of the normal and corroded MS flat plates
regions are reconstructed independently (blind trial) from the corresponding experimental scattered
intensities using the reconstruction algorithm described in section 3. In this method the ρ of the specimen

Fig.3. The photo peak intensity plotted for normal and for corroded MS flat plates

need not be known and only the knowledge of the target material’s (composition) µ/ρ and µ’/ρ and voxel
position and its size are enough to reconstruct the ρ of the each voxel of the specimen being studied.
These mass attenuation coefficients are independent of the physical state of the matter (independent of
density) and the knowledge of composition of the material is sufficient to generate these data. The voxel
position and its size are calculated from the size of the collimators and source-sample and sample-detector
distances. The µ/ρ and µ’/ρ are computed using the XCOM program which can generate photon cross
sections for scattering, photoelectric absorption and pair production, as well as total attenuation
coefficients, for any element, compound or mixture (Z 100), at energies from 1 keV to 100 GeV [6].
The result of MCNP PHS simulations and their comparison of the spectral shape with the experimental
ones are shown in Fig. 4 for the normal MS flat plate and in Fig. 5 for the corroded one. The
reconstructed densities of the normal and corroded MS flat plate’s voxels from the experiment are used in
these simulations. A good agreement in the shape of the PHS is seen between the experimental spectra
and those of MC simulated ones. Thus measurements support the MC simulations done using the
experimental reconstructed densities

6
Fig.4. The scattered PHS and its comparison Fig.5. The scattered PHS and its comparison
with MCNP simulated one for normal MS flat plate with MCNP simulated one for corroded plate

In field conditions one will be looking for structures with density losses varying from few percent
onwards. The ratio between experimental scattered intensities of normal and corroded plate, Signal-to-
Noise (S/ N) is calculated from MC simulation of MS flat plate corrosion varying from the present case of
6.6% extending up to 60%. The photo peak parts of the simulated scattered spectra are shown in Fig. 6

Fig.6. The photo peak parts of the MC Fig. 7. The quantitative estimates of the values
simulated scattered spectra for corroded MS flat plate of S/ N plotted as a function of the % of corrosion

and the quantitative estimates of the values of S/ N from these spectra i.e., the ratio of count rate of
normal MS flat plate to corroded one is plotted as a function of the percentage of corrosion and is shown
in Fig.7. The S/ N values estimated from simulated MC spectra vary from 1.054 (6.6% corrosion) to
2.229 (60% corrosion). The experimental reconstructed densities of the normal and corroded MS flat
voxels of same sizes are compared with the average density values obtained by weighing the MS flat plate
in a weighing balance and estimating its volume by fluid displacement method. A good agreement is seen
between the experimental values and average density value within estimated errors although the
reconstructed corroded density values are slightly on the lower side compared to the average density
value. This can be explained based on the following. It is seen that the mass loss due to corrosion is not
uniform and because of this the average density value obtained for the whole MS flat plate by weight and
volume method will not be exactly matching with the experimental values obtained in individual voxel of
size, 27.103 x 10-6 m3. The experimental results show that the scattering method is highly sensitive to
changes in electronic and physical densities of the volume element under study and the magnitude of
embedded corrosion can be clearly identified and quantified by monitoring single scattered events. The
multiple scattering events which affects sensitivity are highly minimized by selection of very narrow
energy range of 20 keV of the photo peak in the scattered spectrum recorded with very high resolution
HPGe detector (soft collimation) in addition to choosing 0.007 m collimators (hard collimation) for both
source and detector.

7
The effectiveness of this inspection technique can be defined by the spatial resolution and the density
contrast achievable. Good resolution requires a small sensitive volume, while high contrast demands large
sensitivity to changes in composition. The size of the inspection volume defines the spatial resolution.
Reducing the collimator aperture to improve the spatial resolution leads, however, to a decrease in the
count rate. This can be compensated for by increasing the source strength and the counting period. A
practical compromise is therefore necessary to achieve a reasonable resolution within an appropriate
counting period and without exposure to a high dose of radiation. In order to increase the contrast, the
contribution to the detector from the material contained within the sensitive volume should be enhanced
while that of the surrounding media should be reduced. This can be achieved by reducing the attenuation
of the radiation as it travels to and from the sensitive volume, and/ or by increasing the probability of
scattering within the volume. The attenuation and scattering probability, however, depend on the radiation
energy and the angle of scattering. The source energy and the scattering angle are, however, the two most
important design parameters as they directly affect the detector response. The main advantage of this
method as compared to other techniques is the ability to detect and quantify corrosion in any material or
structures with out the need for all- round access to the structures under investigation. It is also possible to
inspect a chosen volume and the method is relatively simple and results are of direct nature. The photon
scattering preferable to other forms of NDE in certain cases: high contrast even for thick samples, good
spatial resolution, 3-dimensional capability, and the ability to make direct real-time density measurements
of only the volume of interest within the sample.

6. CONCLUSIONS

The MC simulations have been done by using the Monte Carlo N- Particle (MCNP) code in calculating
the gamma scattered intensity from the MS flat plates and compared with the experimental ones. A simple
and straightforward reconstruction algorithm developed to correct for attenuation of the incident and
scattered photons in the material surrounding each scatter site which can be used in practice is described.
The quantitative estimates of the values of S/ N are calculated from MC simulation of MS flat plate
corrosion varying from 6.6% extending up to 60% is given to give an indication of the magnitude of S/N
one can get in this type of investigations. The measurements support the MC simulations done using the
experimental reconstructed densities. The experimental and MC simulation results show that the
scattering method is highly sensitive to changes in electronic and physical densities of the volume
element under study and the magnitude of embedded corrosion can be clearly identified and quantified by
monitoring single scattered events.

REFERENCES
[1]Banerjee K, Dunn W L, On X-ray back-scattering to detect hidden
cracks in multi-layer structures Appl. Radiat. Isot. 2007, 65, 176-182
[2] Lawson L R, Compton X-ray backscatter depth profilometry for aircraft corrosion inspection Mater.
Eval. 2002, 60, 1295- 1316
[3]Harding G, Inelastic photon scattering: Effects and applications in biomedical science
and industry Radiat. Phys. Chem. 1997, 50, 91-111
[4]Hussein E M A Whynot T M, A compton scattering method for inspecting concrete
structures Nucl. Instr. Method 1989, A283, 100-106
[5] Briesmeister J F (Ed.), MCNP™, LA-13709-M Manual, 2000
[6] Berger M J, Hubbell J H, N B S I R 87-3597, MD 20899. 1987