pattern formation via blag
DESCRIPTION
Pattern Formation via BLAG. Mike Parks & Saad Khairallah. Outline. Simulate laboratory experiments If successfully simulated, proceed to new computer experiments. Phase 1: Deposition. Gold particles incoming onto the surface from a heat source. The particles will not move much at T=20K. - PowerPoint PPT PresentationTRANSCRIPT
Pattern Formation via BLAG
Mike Parks & Saad Khairallah
Outline Simulate laboratory experiments
If successfully simulated, proceed to new computer experiments.
Phase 1: Deposition
Substrate
Xenon
T=20K
Gold particles incoming onto the surface from a heat sourceThe
particles will not move much at T=20K
Phase 2: Desorbtion
Substrate
Xenon particles desorbing
T>20K
Thin xenon film acts as timer
Gold particles walk randomly
With a sticking probability of one they form clusters when colliding
Final State: Clusters
SubstrateT>>20K
Final Equilibrium State: clusters on substrate(abrupt interface)
Control Parameters Parameters for Cluster Creation:
The thickness of the xenon layer acts as a timer
Sticking probability coefficient ~1 (DLCA) Surface coverage External potential (???)
No need to satisfy thermodynamics constraints: surface free energy and the three growth
modes
Results to simulate… Weighted cluster size grows as
S~t2 Density decays as N~t-2. Fractal dimension according to
DLCA size ~ (average radius)^Dimension.
…our contribution: Charge the particles Apply electric field perturbation
+-
Uniform E
-- -+
++
Simulation Start with uncharged particles interacting on a
square lattice with Lennard-Jones potentials. When two atoms become adjacent, they bond to
form a cluster. Update simulation time as
t = (# Atoms Moved)/(# Atoms), i.e. diffusion does not depend on time.
Simple metropolis algorithmNo KMC: 1. We are not describing the dynamics on the surface. 2. Pattern formation via BLAG does not depend on
time explicitly.
Implementation Issues: Need to efficiently
determine when to merge clusters
Use bounding boxes on clusters and check for adjacent atoms only when boxes overlap
Linked-cell method implemented for L-J potentials
The SIMULATIONS Performed
1. Uncharged particles: mimic experiment
2. Charged particles: uniformly distributed
3. Charged particles with uniform electric field: weak and strong
Results (Uncharged)
Initial Configuration
Final Configuration
Power Law Dependence(uncharged)
Experiment:
1.9 +/- 0.3
Simulation:
2.00 +/- 0.03
Agreement!
Fractal Dimension(uncharged)
Agreement!
Modification : Add Charge Add a positive or negative charge of
magnitude 1.6e-19 Coulombs to all atoms, such that the net charge is zero.
Distribute the charged particles uniformly over the lattice.
Clusters that form as to have no net charge interact only with L-J potential.
Results (Charged Particles)
Final Configuration
Fractal Dimension(charged plus charged with e-field)
Fractal:
New results. We see same dimension as with no charging.
Power law : Size~t2
coverage
Exp. No charge
Charge Charge with
Efield
21%
1.9 0.3
1.99 0.03 1.98 0.02 1.97 0.00
19%2.01 0.02 1.76 0.01 1.91 0.00
11%1.97 0.03 1.50 0.01 1.66 0.00
Interpretation…
The effect of charging subsides according to coverage:
1. Fast decay if high coverage: particles neutralize each other quickly
2. Slow decay if low coverage: particles neutralize each other slowly
Interpretation…
•When charging effect subsides fast, L-J takes over giving close results to exp.
•When charging effect subsides slow, Coulomb potential acts longer altering results from exp..
•So what does the electric field do?
Electric Field Effect… The electric field accelerates the
process of particles neutralizing each other making the charge effect decay fast.
We expect L-J to dominate on the long run
Hence results closer to experiment
Future work… The model, DLCA based on sticking
probability coefficient ~1: so change that number allowing for non-sticking collisions.
Have a metallic substrate to alter the potential with an image potential
Apply varying electric field More complicated: 3D clusters.