Accelerated Molecular Dynamics Methods
|
|
Tim
Germann - tcg@lanl.gov - Los Alamos
National Laboratory |
Outline of Lecture
- Background
- Rare event examples: epitaxial growth, radiation damage
- e.g. plutonium aging due to self-irradiation, ...
- Pitfalls of KMC: concerted multiatom mechanisms, unconventional
crystal structures or grain boundary diffusion, ...
- One possibility: Henkelman-Jonsson saddle-point enumeration
+ KMC Seek MD acceleration technique(s) w/o biasing the dynamics
- Hyperdynamics
- Basic idea: boost PES in wells w/o affecting saddle points
Mathematical justification (assuming TST) & definition
of an accelerated time
- Demonstration on model system(s) using simple boost functions
- More sophisticated boost functions
- Application to more complex systems
- Unsolved problems: extension to systems with low barriers
(or a range of barriers, e.g. glasses, liquids, proteins)
- Related approaches: temperature accelerated dynamics
- Basic idea: increase simulation temperature to find neighboring
saddles, but reject transitions until low-temperature path
can be accepted
- Mathematical justification (assuming harmonic TST)
- Demonstration on model system(s)
- Related approaches: parallel replica dynamics
- Basic idea: independent realization of system on each processor
to explore phase space more rapidly; first processor to find
escape pathway is accepted
- Mathematical justification (assuming infrequent
events)
- Demonstration on model system(s)
- Brownian Monte Carlo
- Combining II and IV -> parallel replica hyperdynamics
- Multiplicative boost achieved
- Example: epitaxial growth of Cu/Cu(100) on 1000 procs ->
0.3 s
- Outlook & Prospects
- Lab
- Will focus on *'d topics; vary boost function, system size
and complexity (e.g. single adatom diffusion vs. islands with
both low and high barriers) to demonstrate where hyperMD works
well and where it runs into difficulties.
|
|
|