September 25, 2007 IEEE Nano-Net, Catania, Italy 1
Networking Behavior in Thin Film and
Nanostructure Growth Dynamics
Tansel Karabacak
U of Arkansas – Little Rock
Applied Science Department
Murat Yuksel
U of Nevada – Reno
Hasan Guclu
Los Alamos National Lab
Complex Systems Group
September 25, 2007 IEEE Nano-Net, Catania, Italy 2
Talk Outline
Motivation: Dynamic Effects in Growth
A Network Modeling Approach Initial Results Future Work
September 25, 2007 IEEE Nano-Net, Catania, Italy 3
Motivation: Deposition Techniques
definition
plasma
Sputter deposition
sin/cos~
Precursor gas
Chemical vapor deposition (CVD)/plasma enhanced CVD
cos~o70
Oblique angle deposition
Deposition angle (
Flux distributions:
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
flux
dis
trib
utio
n F
(
CVD Obliqueangle dep
Sputter dep
Thermal evap
Thermal evaporation(not including MBE)
0
definition
definition
plasma
Sputter deposition
sin/cos~
plasmaplasma
Sputter deposition
sin/cos~
Precursor gas
Chemical vapor deposition (CVD)/plasma enhanced CVD
cos~
Precursor gas
Chemical vapor deposition (CVD)/plasma enhanced CVD
cos~o70
Oblique angle deposition
Deposition angle (
Flux distributions:
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
flux
dis
trib
utio
n F
(
CVD Obliqueangle dep
Sputter dep
Thermal evap
Deposition angle (
Flux distributions:
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90
0.0
0.2
0.4
0.6
0.8
1.0
Nor
mal
ized
flux
dis
trib
utio
n F
(
CVD Obliqueangle dep
Sputter dep
Thermal evap
Thermal evaporation(not including MBE)
0
Thermal evaporation(not including MBE)Thermal evaporation(not including MBE)
0
September 25, 2007 IEEE Nano-Net, Catania, Italy 4
Motivation: Dynamic Effects in Growth
A typical thin film and nanostructure growth involves four main dynamic effects: shadowing surface diffusion reemission noise
Characterizing these effects is vital to understand the shapes of nanostructures
September 25, 2007 IEEE Nano-Net, Catania, Italy 5
Motivation: Dynamic Effects in Growth
theory
theory(shadowing)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
ObliqueCVDSputteringEvaporation
Gro
wth
exp
onen
t,
Deposition method
(smoothing)theory
theory(shadowing)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
ObliqueCVDSputteringEvaporation
Gro
wth
exp
onen
t,
Deposition method
(smoothing)
Theoretical models (e.g., dynamic
scaling) have not been able explain growth behavior
well..
root-mean-square roughness = time
September 25, 2007 IEEE Nano-Net, Catania, Italy 6
Motivation: Dynamic Effects in Growth
Monte Carlo simulations have been able to explain growth behavior well..
2
3
1
4
x
y
z
2
3
1
4
x
y
z
Fundamental Questions:
Can we explain the growth behavior in simple theoretical terms? Are there any universal behaviors in the growth
process?
September 25, 2007 IEEE Nano-Net, Catania, Italy 7
A Network Modeling Approach…
Key Idea: Use the simulations data to map the growth process to a corresponding network model.
For now, we focus on the reemission and shadowing effects, as they are the dominant ones.
AB
Non-sticking & “re-emission”
Sticking & deposition
Incident particles(atoms/molecules)
Particles being captured mostly by the hills due to the “shadowing” effect
AB
Non-sticking & “re-emission”
Sticking & deposition
Incident particles(atoms/molecules)
Particles being captured mostly by the hills due to the “shadowing” effect
A reemitting particle means that there is a “relationship” between the starting and ending points of the reemission.
September 25, 2007 IEEE Nano-Net, Catania, Italy 8
A Network Modeling Approach…
What is a “node”? grid-based
raw grid points
cluster-based mainly models shadowing effects defined based on the current surface morphology
granularity matters..
hill
valley
September 25, 2007 IEEE Nano-Net, Catania, Italy 9
A Network Modeling Approach…
What is a “link”? mainly models the reemission effect unidirectional vs. bidirectional
again, granularity matters.. possible to classify the nodes:
source (hill), sink (valley), routers
September 25, 2007 IEEE Nano-Net, Catania, Italy 10
A Network Modeling Approach…
Several more abstractions are possible link capacity: maximum # of particles that can
physically reemit from point A to point B on the surface
link propagation delay: physical distance between point A to point B
traffic: particles/time bits/time
The challenge is to illustrate physical meanings to these abstractions..
September 25, 2007 IEEE Nano-Net, Catania, Italy 11
Initial Results Looked at thin film growth simulations
Chemical vapor deposition (CVD) surface size: 512 x 512 lattice units two different sticking coefficients: s=0.1, s=0.9 simulated reemission, shadowing, and noise effects
Network model assumed each lattice unit on the 512 x 512 grid is
potentially a node took four snapshots at different surface thicknesses
during the simulation each snapshot is the network corresponding to the
trajectories of 10 x 512 x 512 particles (adatoms)
September 25, 2007 IEEE Nano-Net, Catania, Italy 12
Initial Results
Grown surfaces at different thicknesses and their corresponding network models
s=0.9
September 25, 2007 IEEE Nano-Net, Catania, Italy 13
Initial Results
Degree distributions: mainly Exponential, becomes
power-law-like as time goes by
Degree distribution converges to the same for both sticking coefficients!
Degree distribution converges to the same for both sticking coefficients!
September 25, 2007 IEEE Nano-Net, Catania, Italy 14
Initial Results
Non-intuitive insight: degree distribution is virtually the same for surfaces grown with different sticking coefficients!
September 25, 2007 IEEE Nano-Net, Catania, Italy 15
Initial Results
Distance distributions are
clearly power-law!
September 25, 2007 IEEE Nano-Net, Catania, Italy 16
Future Work How does the granularity affect? cluster-based network modeling diffusion effect