Imaging with Neutrons

Hassina Bilheux
VENUS Instrument Scientist

Yuxuan Zhang
MARS Instrument Scientist

Jean Bilheux
Computational Instrument Scientist

August 7, 2024

ORNL is managed by UT-Battelle, LLC for the US Department of Energy

 The Neutron Imaging Team for the General User

Yuxuan Zhang, James Torres, Jean Bilheux, Shimin Tang, Artificial

HFIR MARS Scientist HFIR MARS Scientist Computational Intelligence, Machine
Instrument Scientist Learning, Hyperspectral
Imaging

Chen Zhang
Imaging Software
Developer Roger Hobbs, Hassina Bilheux,
Kevin Yahne, Harley Skorpenske,
Imaging Scientific Imaging SA SNS Group Leader SNS VENUS Scientist
Associate (SA)
2
 Outline
• Imaging at the High Flux Isotope Reactor MARS beamline:
– Principle of neutron radiography and computed tomography at a
continuous source

Yuxuan
– The CG-1D imaging beamline
– Examples

• Imaging at the Spallation Neutron Source:

Hassina
– Principle of neutron radiography at a pulsed source
– Examples
– The VENUS !!!construction project

Jean
• Scientific programming

3
 Outline
• Imaging at the High Flux Isotope Reactor MARS beamline:
– Principle of neutron radiography and computed tomography at a
continuous source

Yuxuan
– The CG-1D imaging beamline
– Examples

• Imaging at the Spallation Neutron Source:

Hassina
– Principle of neutron radiography at a pulsed source
– Examples
– The VENUS !!!construction project

Jean
• Scientific programming

4
 Neutron imaging measures the transmitted neutrons
though an object

Source Object Detector

(Figure source: aven.amritalearning.com,. (2013). Shadows and Pin Hole Camera.)

5
 Transmitted neutrons recorded as image/images

Continuous neutron beam

θ
I0 I

Incident Transmitted y) θ)
I(x, y,

Volume
reconstruction

6
 Neutrons interact uniquely with matter

(Zhang Y. The University of Utah; 2016)

• Non-destructive
• High penetration through metals
• Sensitive to light elements (H, Li, etc.)
• Isotopic contrast
A. Tengattini, et al. Geomechanics for Energy and the Environment 27 (2021)

7
 Neutrons interact uniquely with matter (cont’d)
X-ray

Neutron

Gupta, D., Zhang, Y., Nie, Z., & Koenig, G. M. (2024).

IPTS-33592, Y. Zhang, 2024 Journal of Industrial and Engineering Chemistry
8
 Neutrons interact uniquely with matter (cont’d)
Neutron
6Li

H D2O H2O

D
7Li

IPTS-26032, Y. Zhang, 2022

https://neuit.ornl.gov/

9
 The MARS instrument at HFIR

Beam stop Detectors

Flight tube (He filled)

Sample table

10
 Example: visualize live root system

Switchgrass Maize

Al chamber

12 (J. Warren (PI), H, Bilheux, M. Kang, S. Voisin, C. Cheng, J. Horita, E. Perfect, Plant Soil, 2013)
 Resolve the composition gradient in graded superalloy
René41

IN718

Transmission image

Huang S., Shen C., An K., Zhang Y., Spinelli I., Brennan M., Yu D., Frontiers in
Metals and Alloys, 1, 1070562 (2022)
Reconstructed 3D volume Cross-sectional view after
(virtual slicing location in orange) virtually slicing
13
 Measures molten salt densities at MSR operating
temperatures

Moon J., Andrews H.B., Agca C., Bilheux J., Braatz A.D., McAlister A., McFarlane J., McMurray J.W., Robb K.R., Zhang Y., Industrial &
14 Engineering Chemistry Research, 61, 17665-17673 (2022)
 Resolutions and FOVs
Type of Highest spatial Typical acquisition time
Field-of-view (FOV) Pixel size (μm) Maximum speed @16 bit
detector resolution (μm) of 1 radiograph

High-res (1x) 36 x 24 mm2 3.8 10-15 900 s 1 image/second

High-res (.5x) 50 x 48 mm2 7.63 20-25 300 s 1 image/second

High-speed 88 x 88 mm2 43 ~100 — 74 image/second

Balanced 88 x 88 mm2 16 ~50 30-90 s 1 image/second

High-speed High-res (.5x) High-speed High-res (.5x)

Gap between
lines = 20 μm

Gap between
lines = 100 μm
(Sample credit: George Williams)
15 IPTS-27734, Y. Zhang, 2022 IPTS-27734, Y. Zhang, 2022
 Resolutions and FOVs
Type of Highest spatial Typical acquisition time
Field-of-view (FOV) Pixel size (μm) Maximum speed @16 bit
detector resolution (μm) of 1 radiograph

High-res (1x) 36 x 24 mm2 3.8 10-15 900 s 1 image/second

Transmission mode
High-res (.5x) 50 x 48 mm2 7.63 20-25 300 s 1 image/second

resolves micro features (10-100 μm)

High-speed 88 x 88 mm2 43 ~100 — 74 image/second

Balanced in cm sized samples ✓

88 x 88 mm2 16 ~50 30-90 s 1 image/second

High-speed High-res (.5x) High-speed High-res (.5x)

Go down to smaller length scale ?

Gap between
lines = 20 μm

Gap between
lines = 100 μm
(Sample credit: George Williams)
16 IPTS-27734, Y. Zhang, 2022 IPTS-27734, Y. Zhang, 2022
 What is neutron grating interferometry (nGI)?
A neutron imaging instrumentation that enable
the utilization the wave properties of neutron, to
spatially resolve sub μm internal features

Interference
Sample pattern
(cm size)
T. Reimann et al., J. Applied
Crystallography. (2016)

Ls

Source Phase grating Analyzer Raw data resulted from

a step-wise scan Transmission Image (TI)
grating p = 10-50 μm grating
Y. Kim et al., Applied Sciences (2022).
(Sample courtesy of Dr. Chris Fancher)

17
 What is neutron grating interferometry (nGI)?
A neutron imaging instrumentation that enable
the utilization the wave properties of neutron, to
spatially resolve sub μm internal features

Interference
Sample pattern
(cm size) T. Reimann et al., J. Applied
Crystallography. (2016)

Ls

Source Phase grating Analyzer Raw data resulted from

a step-wise scan Dark Field Image (DFI)
grating p = 10-50 μm grating
Y. Kim et al., Applied Sciences (2022).
(Sample courtesy of Dr. Chris Fancher)

18
 More about this sample along with the nGI results

Photo EBSD Transmission (TI) Dark Field (DFI)

Inconel

Two different
cooling rate regions
in an Inconel
sample
(Sample courtesy of Dr. Chris Fancher)
Color relates to the
SANS from features
EBSD image is from Plotkowski et al., Additive Manufacturing. 46, 102092, (2021).
in certain sizes
19
 More about this sample along with the nGI results

Photo EBSD Transmission (TI) Dark Field (DFI)

Inconel

How to interpretate the DFI?

Two different
cooling rate regions
in an Inconel
sample
(Sample courtesy of Dr. Chris Fancher)
Color relates to the
SANS from features
EBSD image is from Plotkowski et al., Additive Manufacturing. 46, 102092, (2021).
in certain sizes
20
 Scan through various correlation lengths to understand
the length scale
Interference
Sample pattern
(cm size)

Neutron
p Autocorrelation length
Wavelength

𝜉 = 𝜆𝐿𝑠 /𝑝
Ls

Source Phase grating Analyzer

grating p = 10-50 μm grating
Y. Kim et al., Applied Sciences (2022).

(Sample courtesy of Dr. Chris Fancher)

21
 Scan through various correlation lengths to understand
the length scale Neutron Correlation length
Wavelength
Interference
Sample
(cm size)
pattern 𝜉 = 𝜆𝐿𝑠 /𝑝

ξ = 223 nm ξ = 313 nm ξ = 401 nm ξ = 491 nm

Ls

Source Phase grating Analyzer

grating p = 10-50 μm grating
Y. Kim et al., Applied Sciences (2022).

(Sample courtesy of Dr. Chris Fancher) ξ = 580 nm ξ = 670 nm ξ = 759 nm ξ = 848 nm

22
 Plotting the dark field intensity vs. ξ in region-of-interest
(ROI)
1

0.8

Dark field intensity

0.6

0.4

0.2
ROI#1 ROI#2 ROI#1 ROI#2
0
0 1000 2000 3000 4000
Autocorrelation length (nm)

DFI • Indicating different microstructures between

(Sample courtesy of Dr. Chris Fancher) selected ROIs
23
 Measurement of suspended PMMA spheres in solution as
a calibration standard
1

0.8

Dark field intensity

0.6

0.4

0.2
ROI#1 ROI#2 ROI#2
0
0 150 300 450 600 750 900
Autocorrelation length (nm)

• The curves indicate the size of the PMMA spheres,which

match well with the spec. (~100 nm in the top cell, ~150
nm in the bottom cell)

TI • Some settling observed in the bottom cell, resulted

DFI
difference in number density of PMMA can be observed
(Sample courtesy of Dr. Fankang Li)

24
 nGI capability at MARS
Two setups to cover ACL ranges from ~40 nm to ~3400 nm

Transmission Dark field

(Sample courtesy of R. Dehoff)

25
 Outline
• Imaging at the High Flux Isotope Reactor MARS beamline:
– Principle of neutron radiography and computed tomography at a
continuous source

Yuxuan
– The CG-1D imaging beamline
– Examples

• Imaging at the Spallation Neutron Source:

Hassina
– Principle of neutron radiography at a pulsed source
– Examples
– The VENUS !!!construction project

Jean
• Scientific programming

26
 Imaging at a pulsed source (SNS)
Energy-dependent
Detector radiographs
Neutron pulse
Hg
Target

Proton
pulse I0 I
Moderator
Incident Transmitted
T0 T I(x, y, t)
L
Courtesy of Y. Zhang, ORNL

I(x, y, 𝜆)
−  ( ) x
N A or
I ( ) = I 0 ( )e  ( ) =  t ( )
M I(x, y, E)

27
 Higher energy neutrons can also be used for imaging (neutrons of energies
higher than 1 eV): Resonance Imaging

Zhang Y., Myhre K.G., Bilheux H.Z., Tremsin A.S., Johnson J.A., Bilheux J., Miskowiec
A., Hunt R.D., Santodonato L., Molaison J.J., "Neutron Resonance Radiography
and Application to Nuclear Fuel Materials", Transactions of the American Nuclear
28 Society, (2018).
 Resonance imaging (> 1eV or < 0.286 Å): preparing
your experiment
• Soil surveys, contaminants in soil, etc.: neutrons Overall
thickness
through beam
– transmission through 0.01 mm thickness of natCo (between 1 and 5 Å) = 99.5 %
– transmission through 1 mm thickness of natZn (between 1 and 5 Å) = 96.4 %

Simulated resonance for elements of interest(*)

100 eV = 0.0286 Å

29 (*)Soil surveys in UK (1978-1983, 2000, 2011-2012)

 Resonance imaging: preparing your experiment (cont’d)

• Hg contamination in soil
– Assumptions: 0.1 mm Hg (13.6 g/cm3) + 12.5 mm SiC (with 1.5 g/cm3)
• Transmission (1 and 5 Å) = 66.4 %

Resonance peaks plotted

with:

iNEUIT (“I knew it”)

30
 Using epithermal neutrons (energy > 1 eV), resonance
imaging can map the isotopic content in advanced
nuclear fuel materials in 3D

➢ Distribution of elements drive the performance of the novel advanced

nuclear fuel materials Gd U

Gd

➢ Quantitative analysis is being developed using in-house open-source

Python package (ResoFit)
Myhre K.G., Zhang Y., Bilheux H.Z., Johnson J.A., Bilheux J., Miskowiec A., Hunt R.D.,
"Nondestructive Tomographic Mapping of Uranium and Gadolinium Using Energy-Resolved
31 Neutron Imaging", Transactions of the American Nuclear Society, (2018).
 Bragg edge imaging: how does it work?
Different Bragg edges I ( ) = I 0 ( )e −  (  ) x
measure different phases N A
and lattice spacings, dhkl.  ( ) =  t ( )
M

The height of Bragg edge

Transmission

provides the amount of a

Neutron

specific phase.

Barton J.P., Bilheux H.Z., Bossi R.,

Herwig K.W., Santodonato L.,
Taylor M., "Chapter 12: Neutron
Radiography for Nondestructive
Testing", Nondestructive Testing
Handbook, Fourth Edition: Volume
3, Radiographic Testing
(RT) (2019).

λhkl
Neutron Wavelength
3D printed
Inconel 718 The position of the Bragg
Radiograph
at HFIR edge, λhkl = 2 dhkl, is a measure
32
1 cm of the strain in the sample
 Principle of Bragg edge Transmission
✓ Utilizes thermal and cold neutrons (approximately between 1 and 10 Å)
✓ Obeys Bragg’s Law λhkl = 2dhkl sin θhkl simplifies: λhkl = 2dhkl
March-Dollase Sabine’s primary
model extinction model 𝑉0: volume of unit cell
𝜆2 2𝑑ℎ𝑘𝑙 >𝜆 Fhkl: Structure factor including Debye-
𝜎𝐵𝑟𝑎𝑔𝑔 𝜆 = ෍ 𝐹ℎ𝑘𝑙 2 𝑑ℎ𝑘𝑙 𝑃 𝛼ℎ (𝜆൯ 𝐸ℎ𝑘𝑙 𝜆, 𝐹ℎ𝑘𝑙 Waller factor
2𝑉0 ℎ𝑘𝑙
Crystallite orientation
Total cross section for d111 Crystallite size effect
Inconel 718 (Ehkl) effect (𝑷 𝜶𝒉 (𝝀൯ : r and β)

Total Cross Section () r= =

Coherent Elastic 100
100 Incoherent Elastic 1 m 300
No texture
Coherent Inelastic r = 1 for all (hkl)
Incoherent Inelastic 5 m
80 Absoption
80 10 m

tot (Barns)
tot (Barns)
() r= =
Barns

60 200
() r= =
222 311 220 200 111
40 60 () r= =
100
20

0 40 222 311 220 200 111 222 311 220 200 111
0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1.5 2.0 2.5 3.0 3.5 4.0 4.5
Wavelength (Å) Wavelength (Å) Wavelength (Å)
G. Song et al., Journal of Imaging, 3, 65 (2017).
33
 The perfect case study: powders

(200)

Song G., Lin J.Y., Bilheux J., Xie Q., Santodonato L., Molaison J.J., Skorpenske H.D., dos Santos A.M., Tulk C.A., An K., Stoica A.D., Kirka M.M., Dehoff
R.R., Tremsin A.S., Bunn J.R., Sochalski-Kolbus L.M., Bilheux H.Z., "Characterization of Crystallographic Structures Using Bragg-Edge Neutron
34 Imaging at the Spallation Neutron Source", Journal of Imaging, 3, 4, 65 (2017).
 Materials Behavior: Monitoring residual strain relaxation and preferred grain orientation of
additively manufactured Inconel 625 by in-situ neutron imaging

Microstrain
35
 Engineered Materials: Monitoring residual strain relaxation and
preferred grain orientation of additively manufactured Inconel
625 by in-situ neutron imaging (10 min measurements)

Microstrain
Microstrain
Modeled and experimental results.

AM Inconel 625 strain evolution as a function of

temperature

Tremsin et al, Additive Manufacturing, 2021.

36 Tremsin et al, Nuc. Instr. Methods in Phys. Res. A, 2021.
 Autonomous Hyperspectral Light scan and
preselection of
Neutron CT Experiment Sample Detector
projection angles
at ORNL
ue Goniometers
tin
on

C
Stop Up to factor 5 improvement in time
Optimization of the scan based on the unique
sample geometry
Ability to provide real-time reconstructed data
Autonomous using advanced iterative reconstruction
Decision methods

4.2
1.5
AI Quality Sample Adaptive
Evaluation Scanning Angles
37
(active learning)
 <311> Bragg edge reconstruction at ~ 2.17 Å ± 0.2 Å
Conventional reconstruction methods Our advanced algorithms/methods

Final result
Ni

Cu

FBP – 15 FBP – 27 svMBIR– 15 sVMBIR– 27

projections projections projections projections

Thresholds
MSE1 5.689x 10-3
SSIM2 0.634

1Zhou Wang; Bovik, A.C.; ,”Mean squared

error: Love it or leave it? A new look at
Signal Fidelity Measures,” Signal
Processing Magazine, IEEE, vol. 26, no. 1,
pp. 98-117, Jan. 2009.
1Z. Wang, A. C. Bovik, H. R. Sheikh and E.

P. Simoncelli, “Image quality assessment:

From error visibility to structural similarity,”
IEEE Transactions on Image Processing,
38 vol. 13, no. 4, pp. 600-612, Apr. 2004.
 Neutron Imaging Capabilities at VENUS

• Bragg edge imaging

• Resonance imaging
• Epithermal imaging
• Largest field-of-view thermal/cold imaging
• Neutron grating interferometry (to be implemented)
• Polarized imaging (to be implemented)

39
https://neutrons.ornl.gov/venus
 Bragg edge imaging:
20 x 20 cm2, spatial resolution ~ 100
Cave µm, time resolution is 5 µs.
shielding
Resonance imaging:
4 x 4 cm2, spatial resolution ~ 150 µm,
time resolution is 150 ns.

Beam stop
Front-end optics
buried in shielding VENUS wall of signatures

Cave door

Control hutch

Radiological
Materials Area
(RMA)
40
 Largest field of view: 20 x 20 cm2
20 cm

Imaging detector at VENUS

Spatial resolution mask First 20x20 cm2 large field-of-view radiograph

measured at VENUS (July 24, 2024)!!!
41
 Resonance radiography demonstrated with Tantalum foil and
the microchannel plate (MCP) Timepix (TPX) detector

42
 Bragg edge radiography demonstrated with Nickel powder
and the microchannel plate (MCP) Timepix (TPX) detector

43
 We hope to see you at VENUS!
The VENUS control hutch VENUS

44
 Outline
• Imaging at the High Flux Isotope Reactor MARS beamline:
– Principle of neutron radiography and computed tomography at a
continuous source

Yuxuan
– The CG-1D imaging beamline
– Examples

• Imaging at the Spallation Neutron Source:

Hassina
– Principle of neutron radiography at a pulsed source
– Examples
– The VENUS !!!construction project

Jean
• Scientific programming

45
 Scientific programming

5 things you must know

Jean Bilheux
Neutron Imaging Computer Instrument Scientist

ORNL is managed by UT-Battelle, LLC for the US Department of Energy

 The 5 things that will save your life

47
 The 5 things that will save your life

1 Pick the right language

48
 Which language ?

Python

Matlab

49 https://www.tiobe.com/tiobe-index/
 Which language ?

50 https://www.tiobe.com/tiobe-index/
 Python

• Huge community (help, libraries, …)

• Easy to learn (no compiler needed)
• Easy to build GUI (standalone application, Web interface)
• Run on any platform ( …)
• Notebooks

51
 Notebooks

https://imaginglectures.github.io/Quantitative-Big-Imaging-2024/
52
 Notebooks

53
 Notebooks

54
 The 5 things that will save your life

1 Pick the right language

2 Stay green

55
 Stay green We need to preserve our environment!

Anaconda.com

3.8
3.12

MyProgram

Python
environments

3.12

3.13
3.9.0

3.9.2

56
 The 5 things that will save your life

1 Pick the right language

2 Stay green

3 Write good code

57
 Write good code

There is a good chance that

later, you will be the one trying
to understand your own code.

58
 Write good code

• Name of variable is
what they represent
• Name of method
indicates what they do
• Explain strange choices
• Add examples at top of
methods/classes

59 https://www.youtube.com/results?search_query=clean+code+uncle+bob
 Write good code
Advantages
Workflow • The unit tests are often used as documentation to
• decide the code to write learn how the software works
• Write the test • You write better code (simpler)
• It should fail • People will trust your code
• Write the code to pass the test • You can check the unit test coverage
• It should pass • Later on, when you make changes, you will quickly
• Move on to the next one find out if the software still works
• If someone report a bug, first write a unit test to
reproduce the bug, then fix it!

Disadvantages
• It seems slower to code (but overall, it’s not)

60
 Write good code
DOI

Code
Unit tests
Documentation
JOSS
License file
Future work file
… Software X

Getting a publication out of your software ! F1000Research

61
 The 5 things that will save your life

1 Pick the right language

2 Stay green

3 Write good code

4 How to keep your job!

62
 iBeatles

• Program that perform

automatic strain
mapping calculation

• 1 million lines of code

• 5 years development

• Kept top secret until

today (only copy is on
this machine!)

63
 iBeatles

Demo time !

64
65
66
67
 Repositories

• Backup of your project

• Provide a full history (easy to reverse changes, …)

• Ideal for collaboration (parallel work, …)

• Documentation

• Necessary tool for publication of the code

• Easy to share code (web interface, …)

• GitAction (automatic test, deployment, build documentation ….)

68
 The 5 things that will save your life

1 Pick the right language

2 Stay green

3 Write good code

4 How to keep your job!

5 Use the best debugging tool

69
 Each of you will
Best debugging tool leave today with
that tool!

• Any OS

• Any computing language

• It takes no time to learn how to use it

• It never needs any software update

• It has a very small carbon footprint

70
 Best debugging tool

• Tell your new friend what

your program does, and
you will find what is wrong
with it !

The Yellow Duck

71
 The 5 things that will save your life

1 Python

2 use environments

3 Good naming & unit tests

4 Repository

5 Talk to your yellow duck

72
 Acknowledgements

Research sponsored by the Laboratory Directed Research and Development Program of Oak
Ridge National Laboratory, managed by UT-Battelle LLC, for DOE.

Resources at the High Flux Isotope Reactor (HFIR) and Spallation Neutron source (SNS), U.S. DOE
Office of Science User Facilities operated by ORNL, were used in the research.

73
 https://forms.office.com/g/gtdnJGM87r

Thank you

Questions

ORNL is managed by UT-Battelle, LLC for the US Department of Energy

neutronimaging.ornl.gov