Jiaxin Wang and Robin P. Gardner
Oct 6th 2011, CEAR at NC State University, Raleigh, NC
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
2
Overview-PGNAA
Pb
S
Pb
S
Hg
C
C
Pb
C
S
O
Mg
Ca
S
Excited level
O
Mg
Neutron Source
Ca
Hg
Ca
Ground level
S
Hg
O
Bulk sample
3
3
Overview-PGNAA
 Advantages:
Nondestructive
Simultaneous
In Situ
Quantitative
sensitive to the entire
periodic table.
 shape of the sample are
relatively unimportant.





 Disadvantages
 Inherently large background






Interference from the neutron
excitation source.
Natural background
Structure materials
Detector activation (NaI)
Summing and pulse pile-up
effect
Hydrogen Peak
4
Overview-CPGNAA
 Solution: introduce gamma – gamma coincidence
technique
 Advantages
 Increase the signal – to – noise ratio
 Reduce the interference of background
 Eliminate the hydrogen prompt gamma-ray peak
 Disdvantages
 The coincidence response is about 2 order of magnitude lower
than single response
 Long measurement time
5
Overview-CPGNAA
 Source  Bulk Sample  Detector
 Reach maximum prompt gamma ray/coincidence
prompt gamma ray counting rates under certain
neutron source (cf-252) strength
6
Overview-MC Simulation
 MCNP5
The general purpose Monte Carlo simulation
parameter study, distribution maps
 CEARCPG
Specified code for prompt gamma and coincidence prompt gamma
Pulse height spectra, elemental library spectra
 Computation power
CEAR ‘Spectral’ cluster with 41 running nodes, each with a Quad-core
CPU.
7
Overview-MC Simulation
 Detector response function-> More scintillators, more
shapes, more size, etc.
-> New DRF generation code – CEARDRFs
 CEARCPG was written for serial computation only
-> Parallel feature implement of CEARCPG
 MCLLS Quantification
-> Differential Operator implement in CEARCPG
8
Overview-Quantification
 Peak analysis
 Matrix effect?
 Detector resolution (NaI, BGO, or HPGE)?
 Monte Carlo Library Least Square
9
Overview-Quantification
MCLLS - procedure
1.
Compositions of a unknown
sample are assumed and the
PGNAA measurement is simulated
2.
Elemental library spectra are
generated with the simulation
3.
Least-squares fit for the
experimentally measured sample
spectrum to obtain compositions
of it.
4.
Compare calculated values with
the originally assumed ones, if not
close enough, repeat the process
from step 1.
10
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Mc(do)lls Quantitative Analysis
 6. Conclusions and Future work
11
DRF-MC simulation
 Because the same detector has been repeatedly used
under different situations, the particle-transport inside
the detector (DRF) could be pre-calculated through
MC simulation to improve future simulation speed
and accuracy.
 MCNP5
General purpose for neutron, photon and electron transport
 G03
Specific for Cylindrical NaI detector
 *CEARDRFs
For more shapes, more scintillation detector: BGO, plastic, etc.
12
DRF-Advantages
 (1) on the order of one-half of the calculations per
history can be omitted by the use of a DRF
 (2) use of the DRF has a natural smoothing effect
which reduces the number of histories necessary for
the desired accuracy by a factor of about 100
 (3) use of the DRF yields better accuracy in spectral
simulations because they can be more accurate than
calculations of particle transport with existing physics
inside the detector.
13
DRF-Simulation VS Exp
ENG-MeV
Probability
0.6617
0.8998
Cs-137 by 2x2 BGO
EXP
MCNP5
G03
MCNP5-unspread
0
0.038197
0.0318
0.020703
0.0364
0.0139
Half life
30.2 years
* Cs-137 --> Ba-137m
Normalized counts
0.0322
10
-1
10
-2
10
 Compton Edge
 Flat continuum
 X-ray escape peaks
EXP
EXP
MCNP5
CEARDRFs
MCNP5
EXP
EXP
CEARDRFs-Add
CEARDRF
MCNP5
-3
10
0
0.1
0.2
0.3
0.6
0.5
0.4
Energy (MeV)
0.7
0.8
0.9
from BGO
14
1
DRF-Simulation VS Exp
Co-60 by 2x2 BGO
Na-24 by 2x2 BGO
EXP
MCNP5
CEARDRFs
0
0
Normalized counts
10
Normalized counts
10
EXP
MCNP5
CEARDRFs
-1
10
-1
10
Co-60 by Box NaI 2x4x16in
0
0.2
0.4
0.6
0.8
1
1.2
Energy (MeV)
1.4
0
10
Na-24 by Box NaI 2x4x16in
-2
10
1.6
1.8
2
EXP
MCNP5
CEARDRFs
0
0.5
EXP
MCNP5
CEARDRFs
1
1.5
Energy (MeV)
2
2.5
3
0
Normalized counts
Normalized counts
10
-1
10
-1
10
-2
10
0
0.2
0.4
0.6
0.8
1
1.2
Energy (MeV)
1.4
1.6
1.8
2
0
0.5
1
1.5
2
Energy (MeV)
15
2.5
3
3.5
DRF-Accuracy and Speed
 The DRFs generated by CEARDRFs have much better
agreement with experiments than commonly used
MCNP5
 The speed of CEARDRFs is very fast. It costs 69
seconds for 2.754 MeV energy and 29 seconds for 0.662
MeV, which almost hundreds of times faster than
original MCNP5. Thus, a complete set of DRF could be
simulated in a reasonable time.
16
DRF-Usage
 1. A complete set of DRF needs to be generated by MC simulation, i.e.
CEARDRFs. For example, an energy range from 0 to 11 MeV in 1024 channels.
 2. Build up the model of surrounding geometry of detector, run the MC
simulation to record the photon energy flux reaching the detector surface and
its path length.
 3. Adjust the photon weight according to the path length and convolute the
recorded energy flux with DRF to get the final simulated spectra.
MC simulation
outside detector
(CEARCPG)
DRF generation
(CEARDRF)
Convolute the
incident gamma
flux with DRF
Simulated pulse
height spectra
Elemental
analysis
Experimental
spectra
17
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
18
CEARCPG-Overview
 CEARCPG (Han, 2005) was developed as the first
specific code that can be used to simulate both the
single and coincidence spectrum of coincidence
PGNAA, including relatively complicated neutron and
photon transportation.
 The most important contribution of CEARCPG is a
new algorithm is developed to sample the neutronproduced coincidence gamma-rays following nuclear
structure.
19
CEARCPG-Parallel implement
File I/O path preparation
Random seeds generated and
distributed to slave nodes
Master node collects recorded
data from each slave nodes
20
CEARCPG-DO
 The Differential Operator method is very powerful tool for
measurement sensitivity study and system optimization.
The basic idea of the differential operator technique is, if
the magnitude of perturbation is very small, the ratio of
changed response can be found by using Taylor series
expansion.
m
 (r ; x j )
j 1
x j
 (r ; x0  x)   (r ; x) x 0  
1 m m  2 (r ; x)
 
2 j11 j 21 x j1x j 2
x j
xj0
x j1x j 2  o(x 3 )
x j 10 , x j 20
21
CEARCPG-comparison
Nuclear data
Neutron interaction
CEARCPG
MCNP5
ENDF/B-VII
ENDF/B-VII
ENSDF
ENSDF with NJOY format
EPDL
EPDL
Neutron capture reaction
Same
Neutron elastic scattering reaction (Free
gas thermal Treatment)
Same
Neutron inelastic scattering reaction (n, n’ )
All inelastic scattering reaction, such as
(n.n’) (n,2n) etc.
The number is function of neutron
weight, photon limit weigh, photon
production cross section, etc.
Generation of neutroninduced photons
Sampling from isotope scheme
Photon interaction
Simple Physics Treatment
Variance reduction
technique
Stratified sampling
general
Parallel Computation
Simple Scripts
Full MPI/OPENMP
Perturbation
Differential operator with variance reduction
technique considered. Applicable to pulse
height spectra
Only to flux
Expanded photon production method
&30X20 photon production method
Simple Physics Treatment &
Detailed Physics Treatment
22
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
23
Prompt Gamma-ray Modeling
 General optimization
o Moderator
o Neutron distribution
o Prompt gamma-ray distribution
 Detector
o Cross-section
Source
Moderator
Bulk sample
o Detector Efficiency
o Neutron response
o Plastic detector setup
 Geometry arrangement
Detector 1
Detector 2
o Lab sample
o Large sample
24
Modeling-Neutron Maps
 Radioactive capture
reaction happens all
through the coal sample
with highest production in
the center area, if the
source is placed under the
sample
Neutron capture rate
Thermal neutron
Fast neutron(1-10MeV)
 Thus, it is better to place
the source under the large
size bulk sample. No
moderator is needed as
the self moderation of
sample is enough for 252Cf
neutron source
25
Modeling-Photon Maps
 In coincidence detection, it is better to place the detectors facing the top and
bottom surfaces separately if possible. Otherwise, placing the two detectors
together on the opposite side of neutron source is also a good arrangement.
 Photon flux spatial distribution maps around the large rectangular shape coal sample
and the conveyor belt shape coal sample.
26
Modeling-Geometry
 Lab size sample
(55cm x 9.7cm x
6.7 cm)
 6”x6” NaI
Cylindrical
detector
 2”x4”x16” Slab
NaI detector
27
Modeling-Geometry
Comparsion for single spectra of Sulfur
-7
10
-8
-6
10
10
-7
Normalized counts
-10
10
-11
10
Ref
Thin Paraffin
6x6 sides
Slabs Sides
-12
10
-13
0
1
2
10
-8
10
-9
10
Ref
Thin Paraffin
6x6 sides
Slabs Sides
-10
10
-11
3
4
5
6
Energy (MeV)
7
8
9
10
10
0
1
Ratio of increase for different lab size sample setup
2
3
4
5
6
Energy (MeV)
7
8
9
Ratio of increase for different lab size sample setup
40
50
Thin Paraffin VS Ref
Side arrangement VS bottom
Slab VS cylinder
35
30
Thin Paraffin VS Ref
Side arrangement VS bottom
Slab VS cylinder
40
25
30
Ratio
Ratio
Normalized counts
-9
10
10
Comparsion for single spectra of Mercury
-5
10
20
20
15
10
10
5
0
0
1
2
3
4
5
Energy (MeV)
6
7
8
9
0
0
1
2
3
4
5
Energy (MeV)
6
28
7
8
10
Modeling-Geometry
Comparsion for total coincidence spectra
10
10
Ratio of increase in total coincidence for lab size sample setups
50
8
10
Thin Paraffin VS Ref
Side arrangement VS bottom
Slab VS cylinder
40
6
10
Ratio
Counts
30
20
4
10
10
2
Ref
Thin Paraffin
6x6 sides
Slabs Sides
10
0
10
0
50
100
150
0
200
250
300
Channel
350
400
450
0
50
100
150
200
Channel
250
500
29
300
350
Modeling-Geometry
 Thinner paraffin (7.3cm)will increase the overall detector response about
a factor of 4.3 and 3.4 for single and coincidence, respectively.
 Changing the 6”x6” detectors position from bottom to left-right sides can
further increase the overall detector response another factor around 1.6
and 3.8.
 Two slab detectors replacing the 6”x6” cylindrical NaI detectors can gain
another increase of a factor around 9.5 and 17.2.
 In sum, the slab detector left-right arrangement can detect around 65 and
223 times more gamma-ray events than the reference setup.
 The ratio of increase (ROI) for different setup as a function of energy:
Higher efficiency for higher energy
30
Modeling-Geometry
ROI* in Single
ROI in Coincidence
response
response
Thin wax
4.3
3.4
0.79
Side Cylinder
6.9
13.0
1.88
Side Slab
66.5
223.7
3.36
Relative ROI
* All ROI values are calculated based on reference setup
31
Modeling-Geometry
 Large size sample (25cm
x 100cm x 100 cm)
 6”x6” NaI Cylindrical
detector
 2”x4”x16” Slab NaI
detector
 70cm x 50cm x 10cm
plastic detector
32
Modeling-Geometry
Comparison for single spectra of Sulfur
-6
10
-7
-5
10
10
-8
-6
Normalized counts
10
-9
10
-10
10
-11
10
-7
10
-8
10
-9
10
10
Two 6"x6" NaI
Two slab NaI
Plastic with NaI
-12
10
1
2
3
Two 6"x6" NaI
Two slab NaI
Plastic with NaI
-10
4
5
6
Energy (MeV)
7
8
9
10
10
1
2
Ratio of increase for different lab size sample setup
5
6
Energy (MeV)
7
8
9
10
Slab vs cylinder
Plastic arrangment vs cylinder
6
5
5
Ratio
4
3
2
4
3
2
1
0
4
7
Slab vs cylinder
Plastic arrangment vs cylinder
6
3
Ratio of increase for different lab size sample setup
7
Ratio
Normalized counts
Comparison for single spectra of Mercury
-4
10
1
1
2
3
4
5
Energy (MeV)
6
7
8
9
0
1
2
3
4
5
Energy (MeV)
6
337
8
Modeling-Geometry
Comparison for total coincidence spectra
9
10
Ratio of increase in total coincidence for lab size sample setups
8
20
10
Slab vs cylinder
Plastic arrangment vs cylinder
15
7
Ratio
Counts
10
10
6
10
5
5
10
4
10
0
Two 6"x6" NaI
Two slab NaI
Plastic with NaI
50
100
150
200
Channel
250
300
350
0
50
100
150
200
Channel
250
300
400
34
350
Modeling-Geometry
 Replacing the two 6”x6” cylindrical detectors with two 2”x4”x16” slab
NaI detectors could gain the ROI of 1.6 and 6.2 in single response and
coincidence response, respectively
 The plastic/NaI special setup could gain the ROI of 2.5 and 1.7.
 NaI detector in the special setup has a better efficiency to high energy
gamma-rays in single response while the slab detectors setup has better
efficiency to high energy gamma-rays in coincidence response
ROI in Single
response
ROI in
Coincidence
response
Relative
ROI
ROI in
Coincidence
Events
Side Slab
1.6
6.2
3.8
9.7
Plastic with NaI
arrangement
2.5
1.7
0.68
8.0
35
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
36
Results
 Through CEARCPG, the 2D coincidence spectrum of
these setups has been simulated with a coal sample
(H-2.892%, C-5.28%, N-%1.4, O-5.487%, Na-1.121%, Al2.38%, Si-1.943%, S-5.6%, Cl-1.729, Hg-2.168%).
 Three setups: slab detectors for lab and large sample,
the special setup with plastic detector.
37
Results
38
Results-Plastic Projection
Peak
Energy(MeV)
Source
1
0.511
Pair Production
2
3
4
0.841
2.379
2.931
Sulfur
Sulfur
Sulfur
5
3.22
Sulfur
6
4.4308
Sulfur
7
4.869
Sulfur
8
5.4205
Sulfur
9
7.31
Hg
10
7.8
Sulfur
11
0.367
Hg
12
6.457
Hg
39
Results-Interference
 Fission gamma and prompt gamma-rays from
structure materials still contribute to true coincidence.
40
Results-Interference
 Everything source of gamma-rays
could be included in the
coincidence response through
chance coincidence.
 When R1  R2  104 / s, The chance
Rc  R1  R2  
coincidence counting rate is only
2% of the true coincidence rate.
 However, when the single
detector counting rate increases
to 105/s the chance coincidence
counting rate is 20% of the true
coincidence rate
41
Results-Dose Rate
 MCNP5 F4 mesh tally
and FM card (flux-todose conversion factor
for human)
 For neutron and photon
separately.
 If a 10 microgram (μg)
source is used, it is
allowed to stay close the
device behind the
shielding material for
2000 hours annually,
even under the public
limits
42
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
43
Simulated total coincidence library spectra for sample 2
10
10
MC(DO)LLS
Normalized counts
 Two set of libraries
8
10
6
10
4
10
2
10
0
10
0
100
7
x 10
200
300
Channel
400
500
600
Q-value projection library spectra for sample 2
Al
Cl
Hg
N
S
Si
fission
total
4.5
4
3.5
Normalized counts
 Two coal samples
Al
O
C
Cl
N
Na
S
Si
Hg
fission
extra
total
3
2.5
2
1.5
1
0.5
0
50
100
150
200
Channel
250
44
300
350
400
MC(DO)LLS-DO results
-4
16
f
x j
x 10
Hg
14
Relative Counts
12
x0 j
2 f
2
x j
10
8
6
2 f
x j1 x j2
4
2

x0 j

 j D exp[  0  x0 j j D]
1  exp[  0  x0 j  j D]
 ( j D) 2 exp[  0  x0 j j D]
x0 j1 x0 j2
1  exp[  0  x0 j j D]
  j1 j2 D 2 exp[   0  x0 j1 j1  x0 j2  j 2 D]

1  exp[   0  x0 j1 j1  x0 j2  j 2 D]



0
-2
0
2
4
6
8
10
12
Energy(MeV)
-6
0.5
x 10
0
Relative Counts
-0.5
-1
-1.5
-2
-2.5
-3
Si
-3.5
0
2
4
6
Energy(MeV)
8
10
12
45

MC(DO)LLS-Fitting Results
 Sample 2, the results of both sulfur and mercury are improved through Q-value projection.

 Sample 1, the result of sulfur is improved while the result of mercury has degradation.
 This result is reasonable since there is little interference in the high-energy window. The
reason of mercury result in sample 1 is that the 8-9 MeV windows is too close to Mercury
Q-value to include the whole peaks.
 When the concentration of Mercury is low as in sample 1, the benefited of less interference
might be canceled out by the drop of signal due to energy window projection.
Number
Element
Total coincidence
Fitting results
Sample 1 Sample 2
1
2
Sulfur
Mercury
1.2479
0.9876
1.0483
0.9992
Q-value projection
Fitting results
Sample 1
Sample 2
1.1066
0.9095
1.0278
1.0005
46
Agenda
 1. Overview
 2. Detector Response Function
 3. Code CEARCPG
 4. Prompt Gamma-ray Modeling
 5. Simulation Results
 6. Mc(do)lls Quantitative Analysis
 7. Conclusions and Future work
47
Conclusions
 1. A new code named CEARDRFs has been developed to generate pretty
accurate detector response function at a very fast speed to improve
accuracy and efficiency of CEARCPG.
 2. Parallel computation feature has been implemented in CEARCPG by
a simple script approach, which dramatically simplified the job while
keeping all the original features and could nearly reach the ideally
linear speed-up feature.
 3. With derivatives to second order Taylor expansion, the DO has also
been implemented into CEARCPG and validated, including the
consideration of collision kernel, transportation kernel and variance
reduction kernel.
48
Conclusions
 4. For lab size sample, replacing the detectors with two 2”x4”x16” slab
NaI detectors could gain the ROI of 66.5 and 223.7 for single and
coincidence response, with higher efficiency for higher energy gammarays.
 5. For large size sample, two 2”x4”x16” slab NaI detectors setup could
gain the ROI of 1.6 and 6.2 in single response and coincidence
response, respectively and the special setup of plastic VS NaI could
gain the ROI of 2.5 and 1.7. The NaI detector in the special setup has a
better efficiency to high energy gamma-rays in single response while
the slab detectors setup has better efficiency to high energy gammarays in coincidence response
49
Conclusions
 6. The simulated 2D coincidence spectra show the feasibility of using
the plastic detector as a trigger to another detector that has better
energy resolution.
 7. Among all the interference, in the total coincidence spectra, the
fission gamma remains the major factor while the interference from
structure material still contributes.
 8. Q-value projection on the 2D spectra could further suppress the
interference. The MCLLS analysis on the Q-value projected spectra
shows better accuracy than using the total coincidence spectra.
 9. With proper shielding, the dose rate around the analyzer is pretty
low.
50
Future Works
 1. Validate the results with benchmark experiments,
 2. New elemental analysis method is also need to be developed
with elemental libraries, eg. Restraind LLS, true 2D LLS.
 3. Other neutron sources like D-T generator are worth a look.
 4. Looking for more complete nuclear structure data, especially
angular correlations between prompt gamma-rays
 5. The light transport in large size detector is also an interesting
area to look.
51
Thank you!
Questions and comments?
The authors are also grateful for the financial support of CEAR through
the Associates Program for Nuclear Techniques in Oil Well Logging
presently supported by Baker Hughes, Weatherford, EXXON Mobil,
Halliburton, Pathfinder, and Los Alamos National Laboratory
52