Discussion of developments surrounding the transformation of the scanning transmission electron microscope from an imaging platform into a manipulation platform.
Atomic level manipulation of matter using Scanning Transmission Electron Microscopy
1. Atomic level manipulation of matter using
Scanning Transmission Electron Microscopy
Ondrej Dyck
Center for Nanophase Materials Science
Oak Ridge National Laboratory
Oak Ridge, TN
2. E v o l u t i o n o f i m a g i n g : d e s c r i p t i o n , u n d e r s t a n d i n g , c o n t r o l
Need to find out :
• Why do atoms do it ?
• How we direct them
to do what we want ?
X-ray and
neutron
scattering:
where the atoms
are on average
Electron and
probe
microscopy:
where exactly
are the atoms
Dynamic
microscopies:
what atoms do Nanotechnology
Beyond Moore
Molecular Machines
Present
Time
3. S c a n n e d p r o b e t e c h n i q u e s
D. Eigler, 1990
M. Simmons, 2012
Single Atom Transistor
4. E m e r g i n g f i e l d o f e l e c t r o n - b e a m m a n i p u l a t i o n
5. W h y n o w ?
Interest in machine
learning through time
2007
2018
• Aberration correction
• 2D materials
• End of Moore’s law
• Quantum race
• Artificial intelligence
6. A t o m i c F a b r i c a t i o n i n S T E M : P i e c e s o f t h e p u z z l e
7. E - b e a m I n d u c e d P h e n o m e n a
Electron-irradiation “damage”
Elastic
(electron-nucleus)
scattering
Inelastic
(electron-electron)
scattering
Deposition, e.g
hydrocarbon
contamination
Mass loss
Structural damage
Specimen heating
E-beam sputtering
Atomic displacement Electrostatic charging
Direct nuclear recoil
Dependent on electron energy,
beam fluence, and atomic
number
Direct nuclear recoil and
ejection
Electron-electron energy
transfer
Dependent on beam current
and material; usually negligible
Absorbed electrons
Sample conductivity
Backscattered electrons
Secondary emission (Auger)
Radiolysis
Electronic excitations
Non-reversible decay
Altered chemical bonds
Depends on current density,
sample chemistry, and temp.
Polymerization of hydrocarbons
Adapted from Egerton and Malac, Micron (2004)
8. E - b e a m - i n d u c e d P h e n o m e n a
Crystallization of amorphous material
Elastic-plastic transition
Ferroelectric domain switching
Phase transitions
Vacancy formation and dynamics
Creation of molecular bonds
Atomic motion
Sculpting (erosion)
Liquid electrochemistry
S. Jesse et. al. (2015) Small
J. Kotakoski et. al. (2014) Nat. Com.H. Komsa et. al. (2012) Phys. Rev. Lett.
O. Dyck et. al. submitted T. Susi et. al. (2017) arXiv
9. E - b e a m I n d u c e d P h e n o m e n a
Electron beam induced
fragmentation
Catalytic growth of C nano-onions
Oku et. al., J. Mat. Chem. (1998)Gonzales-Martinez et. al., Nanoscale (2016)
Fullerene formation
Chuvilin et. al., Nat. Chem. (2010)
Nanotube welding
Terrones et. al., Phys. Rev. Lett. (2002)
10. E - b e a m I n d u c e d P h e n o m e n a
Nanotube growth
Gonzalez-Martinez et. al., Nano Letters (2014)
Formation of 2D Fe and ZnO
Zhao et. al., Science (2014)
Quang et. al., ACS Nano (2015)
Catalytic etching via Fe
nanoparticle
Wang et. al., Sci. Rep. (2012)
11. E - b e a m I n d u c e d P h e n o m e n a
Single atom catalytic activity
Ta et. al., Nano Res. (2017)
Ni M2,3 EELS
Ramasse et. al., ACS Nano (2012)
Catalyst-free formation
of graphene from a-C
Börrnert et. al., Adv. Mat. (2012)
Nanowire formation via
e-beam sculpting
Lin et. al., Nat. Nano. (2014)
12. E - b e a m I n d u c e d P h e n o m e n a
Nanoparticle nucleation and growth
(liquid cell)
Liao et. al., Science (2014)
Creation of point defects (WS2)
Zhou et. al., Nano Lett. (2013)
Crystallization in Strontium titanate
Jesse et. al., Small (2015)
13. E - b e a m I n d u c e d P h e n o m e n a
Dyck et. al., J. Vac. Sci. Tech. B (2017)
Patterning of C Dopant movement through 3D crystal
Jesse et. al., arXiv (2017)
Movement and assembly of
single atoms in 2D materials
Dyck et. al., arXiv (2017)
14. E - b e a m I n d u c e d P h e n o m e n a
• There are a wealth of e-beam induced phenomena to explore
• Specialized holders offer additional parameters to explore
(heating, cooling, electrical biasing, gas/liquid cell,
nanomanipulators etc.)
• Aberration correction and 2D materials have allowed the
addressing of single atoms in STEM
How do we develop fabrication techniques at the atomic scale?
15. D e t e c t i o n
Imaging
Spectroscopy
Ptychography
Compressed sensing
3D imaging
Super-resolution
Read-out speed
Signal to noise ratio
Data generation from
beam-sample
interactions.
16. I n t e r p r e t a t i o n
Machine Intelligence
DFT Modelling
Molecular Dynamics
Image Simulation
Offline analysis
Library building
Physical insight
Real-time analysis
Decision making
What are we looking at?
17. R e a c t i o n D e c i s i o n
Prior Experience
Physical Laws
Libraries
Machine Intelligence
DFT Modelling
Molecular Dynamics
Image Simulation
What do we want to
make happen? How can
we alter parameters to
achieve this?
18. R e a c t i o n
Beam Control
Feedback
Stability
Perform some action, monitor
and correct errors and state of
the sample.
electron beam
Specimen Advanced
DAQ
Fast Direct
Electron
Detection
To scan coils, or
in situ holder
ADF/ABF
19. P u t t i n g t h e p u z z l e t o g e t h e r : c r y s t a l l i z a t i o n
electron beam
Specimen Advanced
DAQ
To scan coils, or
in situ holder
Start inside crystal,
Fast advance
Amorphous/crystalline
interface
Growth of new crystalline atomic layer
Beam advances to next atomic layer
Crystalline detection threshold
FFT reveals real-time frequency distributions
Stephen Jesse
20. P u t t i n g t h e p u z z l e t o g e t h e r : c r y s t a l l i z a t i o n
After
2 nm
Before
2 nm
Beam induced dopant motion
Paul Snijders
Andrew Lupini
Beth Hudak
21. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g
Stephen Jesse
22. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I
Series of fast spiral scans (~50 ms/scan) at a chosen
location. The rapid change in brightness indicates the
formation of a hole.
X
Hole formation
Time
23. P u t t i n g t h e p u z z l e t o g e t h e r : b e a m - i n d u c e d h e a l i n g
• Start with pristine,
cleaned graphene lattice
• Material at ~1200 C
• Electron beam at 100 kV
• Drill hole using spiral scan
• The formed hole is
metastable
• If we continuously
scan at 100 keV, the
hole will grow as edge
atoms are easily
removed
• If we turn the beam
off, or scan elsewhere,
the hole will heal in
less than a minute
24. P u t t i n g t h e p u z z l e t o g e t h e r : b e a m - i n d u c e d h e a l i n g
60 kV with light room-temperature contamination
E-beam deposited
multi-layer graphene
25. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I I
Graphene Lattice Viewed Edge-on
26. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I I
Graphene Lattice Viewed Edge-on
27. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I I
Agitate
source
material
Create
hole/defect
Position
beam
Dyck et. al., Appl. Phys. Lett. (2017)
Scan over large area
to mobilize Si atoms
Drill hole100 kV Si dopant positioning
28. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I I I
Susi et. al., Ultramicroscopy (2017)
Positioning single dopant atoms
within a graphene lattice
12
Dyck et. al., Appl. Phys. Lett. (2017)
Toma Susi
Jannik Meyer
Jani Kotakoski
29. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g I V
Dyck et. al., arXiv (2017)
Assembly of primitive, few atom structures from single dopant atoms
Elisa Jimenez-Izal Anastasia N.
Alexandrova
30. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
• Stationary beam prevents real-time sample monitoring
• Drift causes frequent “misses”
• Need to detect when a structural change happens
• Need to automatically (and intelligently) position beam
• We need help from the machine
31. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Custom software/hardware for beam control and feedback
Stephen Jesse
Overview
scan pattern
Dwell time
Raw overview
Filtered
overview
Sum of spiral scans
Real-time image of
manipulation mode
32. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Automated drift compensation
Drift compensation off Drift compensation on
33. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
34. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Next step: Automated beam positioning for atomic movement
Rule: chose the closest
neighboring lattice position
to the target
35. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Spiral Scan + Compressive Sensing
Conventional scan pathStop and wait for
scan coil hysteresis Significant oversampling
Can we scan continuously?
Can we sample “just enough”?
36. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Lissajous scan paths
Is there a scan path that
optimizes data collection?
Do we need an image?
37. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Next step: Automated atom position recognition
Rapid-scanned image
(simulated) with 5 times
more noise than signal
FFT of the image revealing
periodicity
Ring filtered image can be
compared to an expected
lattice generated from the
FFT to find atom positions.
We help the computer by
giving it an assumption: it
is looking at a lattice of
some kind and this is what
is important.
38. P u t t i n g t h e p u z z l e t o g e t h e r : d e f e c t p o s i t i o n i n g V
Next step: Automated atom position recognition
Raw data from microscope Ring filtered image
Atom positions found
• Rapid scan fast “peak” at the sample
• Lower electron dose less chance of
unintentional sample modification
• Computer knows where atoms are atomic
movement can be automated
39. I n t e r p r e t a t i o n : b u i l d i n g l i b r a r i e s
One C atom displaced Two C atoms displaced
Anastasia N.
Alexandrova
Elisa Jimenez-Izal
• What atomic configurations can be stable?
• Are they useful/have interesting physics?
• How would they look in STEM?
40. I n t e r p r e t a t i o n : A u t o m a t e d D e f e c t C l a s s i f i c a t i o n
Convolutional Neural Network
Pixel-wise localization and classification
Maxim Ziatdinov Artem Maksov
M. Ziatdinov et al., ACS Nano 11, 12742 (2017)
M. Ziatdinov et al., arXiv:1801.05133 (2018)
A. Maksov et al., arXiv:1803.05381 (2018)
Experiment cNN output
Time
C lattice atoms Si dopants Vacancies
41. I n t e r p r e t a t i o n : A u t o m a t e d D e f e c t C l a s s i f i c a t i o n
Analysis of Atomic Defect Kinetics During E-beam
Induced Transformations in WS2
ConvNet + Gaussian mixture model
42. N a n o - M a c h i n e s
Sensors
(multimodal)
Signal processing
and control
Locomotion
Power unit
Nature provides us with examples
incredibly well adapted machines
- Autonomous
- Adaptable
- Self-growing
- Self-healing
- Self-replicating
- Capable of (limited)
network formation
25-200 nm30 nm
Viruses are
the smallest
organisms
43. N a n o - M a c h i n e s
• Not even nature gives us examples of “high level” functionality on the
nanometer scale
• They don’t do anything “on purpose”, no decisions being made
• General problem: We are trying to integrate to many dissimilar functionalities in
an extremely small package
– Thinking (signal processing, decision making, feedback etc)
– Motility
– Energy sources (internal, chemistry, control fields?)
44. N a n o - M a c h i n e s
What if we use a single atom or small atom assembly as an
functional element of a moving nanomachine, and defer
control and power functions to external entities?
Remove thinking and energy source from the machine itself
45. Scanning over this bright
nanoparticle caused it to start
eating the carbon raft it was
stuck to.
As the
nanoparticle eats
the carbon it is
pulled along the
edge of the bilayer
46.
47. N a n o - M a c h i n e s
Dyck et. al., J. Vac. Sci. Tech. B (2017)
Patterning of C
Single layer
Double layer
Triple layer
Maybe heat and custom beam control can
direct the deposition rate and produce
patterning of graphene on graphene
Nanoparticle
Pathway defined by in situ
grown graphene
48. N a n o - M a c h i n e s
600 oC
600 oC
Trigger on intensity
rather than frequency
Graphitic nanowires
grown at various
temperatures