Molecular Dynamics

Molecular dynamics
Content: The basic technique

The Verlet algorithm Average quantities Temperature and pressure scaling Potential energy surfaces Length- and time-scale problems
CompPhys / Molecular Dynamics
Gran Wahnstrm, Department of Applied Physics, Chalmers
Molecular dynamics simulation
A numerical technique to compute the equilibrium and transport properties of classical many-particle systems by solving Newton's equation of motion.
Molecular dynamics simulation

The aim: - to compute the macroscopic behaviour from microscopic interactions
The basic problems:
- obtain a sufficiently accurate description of the microscopic interaction, the interaction between the particles
- sample sufficiently efficiently the relevant part of the configuration space in order to determine the macroscopic behaviour
The molecular-dynamics simulation technique

is used for fundamental studies of interacting classical many-particle systems is used to increase our understanding of disordered systems, liquids, melting, diffusion, anharmonic effects in solids, fracture, defects, phase transitions, plastic deformation, glassy dynamics, ... - on the microscopic level
With increasing realism for the description of the interaction between particles the technique is used to interpret and understand experiments on the microscopic level to study systems under extreme conditions, which may be inaccessible or very expensive to study experimentally to predict properties of new materials, e.g. in nanotechnology to test properties of materials without need to synthesize them, e.g. in drug design in the pharmaceutical industry
Starting point:
The inter-particle interaction potential
Classical mechanics is
Input:
Initial positions and momenta for all particles
used to determine the timedependence of the trajectory in phase-space, given the initial condition
Statistical mechanics is Output:

The time-dependent trajectory in 6N-dimensional phase-space
then used to evaluate various physical quantities, to bridge the gap between the microscopic and macroscopic descriptions.
Example: The Lennard-Jones system
CompPhys / Molecular Dynamics 5/33
Numerical problem
Newton's equation of motion
a set of 3N coupled 2nd-order ODEs
or using
Hamilton's formulation
a set of 6N coupled 1st-order ODEs
with
The numerical algorithm
Molecular dynamics is a statistical mechanics method

The aim is not to predict precisely what will happen to a system prepared in a precisely known initial condition, but to predict the average behaviour of the system prepared in an initial state about which we know something.
should conserve energy well. Moderate short-time energy change (energy fluctuations) is not so critical, systematic changes over long times (energy drift) is more serious. should be time-reversible and area preserving (symplectic). This often implies little energy drift. should duplicate the classical trajectory as closely as possible. The algorithm should be accurate on the timescale for decay of typical time-correlation functions. should be fast to evaluate and simple to implement. The calculations of the forces are the most time-consuming part. The Verlet algorithm is the simpliest that fulfills these requirements.
In fact, it is impossible to predict the true trajectory over long times due to the extreme sensitivity on the initial conditions (chaotic systems).
The original Verlet algorithm

Make a Taylor expansion of the position coordinates both forward and backward in time
Add these two equations
and determine the accelerations from
The algorithm is not self-starting. Initially, the positions at two subsequent time steps are needed. The velocities can be determined from
The velocity-Verlet algorithm

The velocity-Verlet algorithm is exactly equivalent (in exact arithmetic) to the ordinary Verlet algorithm. However, it handles the velocities in a more satisfactory manner and minimizes roundoff errors. It takes the form
The algorithm is self-starting. It can be implemented in an efficient way in two stages, with a force (acceleration) evaluation in between.
Instability of trajectories
Lyaponov instability
Lennard-Jones system (Argon), N=1000
A typical program
Initialization
temperature scaling pressure scaling
Equilibration
Production
integrating the equation of motion force evaluation minimum image convention neighboring lists
Analysis
Average quantities
Time average - the MD approach
An average taken along the trajectory in phase-space for a single system. It corresponds to that N, V, and E are conserved.
Ensemble average - the MC approach

An average is taken over a set of identical systems (an ensemble) at a fixed time. Microcanonical ensemble (fixed N, V, and E):
Ergodic hypothesis
Believed to be valid for most systems, but difficult to prove rigoursly.
Example of average quantities

Simple averages
temperature pressure
Structural quantities
pair-distribution function static structure factor
Fluctuations
heat capacity
Dynamic quantities
mean squared displacement
velocity correlation function

diffusion coefficient
dynamic structure factor
Simple averages
Simple averages
Kinetic, potential and total energy
Temperaure and pressure
Pair distribution function
distance ()
CompPhys / Molecular Dynamics Gran Wahnstrm, Department of Applied Physics, Chalmers
Dynamic quantities
Dynamic quantities
Mean squared displacement
Dynamic quantities
Melting
T=18K, a solid T=183K, a fluid
Temperature scaling
Pressure scaling
Technical details
Minimum image convention Neighbour list
To speed up the calculations (for Larue systems) it is useful to introduce a list of neighbours, within the chosen cut-off distance. For a liquid that list has to be updated regularly due to diffusion of atoms. The list then also has to contain atoms at distances slightly larger than the cut-off distance.
Cell list method

Another, similar technique, is the so called cell list method.
The potential energy surface (PES)

The electrons in the system are treated explicitly
ab-initio molecular-dynamics (the Car-Parrinello method)
The electrons in the system are not treated explicitly

molecular-dynamics based on classical model potentials
very time consuming

not so much used
not so time consuming

extensively used Different models are used for different kinds of systems: van der Waals systems ionic systems metallic systems covalent systems biomolecular systems
van der Waals system

van der Waals bonding - fluctuating electric dipole moments _ _ Model for rare gas solids _
__ _ + ++ + + + _ ++ _
Ex. Lennard-Jones potential (the generic model potential in MD)
_
_ _ _
_ + + _ + + + + ++ _ _
Ionic system
Ionic bonding - transfer of valence electrons
_ _ with
_ + + + _ _
+ + + + _ + _
_ _ + + + _ _ _ ++ ++ + _
+ + + + + _ _
_ _ +++ _ _
The long-range interaction has to be treated with special techniques.
Covalent system
Covalent bonding - sharing of valence electrons _
_ + ++ + + in general, problem with transferability Ex. Stillinger-Weber potential for Si _ _ _
_
_
__
_ _
+ + + + +
+ + + + +
_
_
- the angle formed by two bonds
the model favours configurations in the diamond-like tetrahedral structures
Metallic systems
Metallic bonding - delocalization of valence electrons Physical fact: Pairwise interaction fails. The bonds become weaker when the local environment becomes more crowded.
Pair-functional model
_
_
_
+ ++ + + _ _ _
_
_ __ _ _ _ _ + + + + +
++ ++ +
+ + + + + _ with _ _ + + _ + + + _ _ _ _
Biomolecular systems
Polymers and biomolecules
Models (force-fields) based on geometrical quantities Dihedral angle
Intra-molecular interactions
Bond stretch
Inter-molecular interactions
Dispersion and electrostatic interaction
Bond angle
Length-scale?
Typically: 100 x 100 x 100 atoms
106 = 1023 ? Use periodic boundary conditions
Time-scale?
Time-step:
t 1 fs
Number of time-steps ~ 106 - 109
i.e. tmax ~ 10-9 - 10-6 s
Techniques to reach long times rare event simulations Problems long-range interaction long-wavelength fluctuations different phases kinetic modelling
Length-scale problem - Ewald sums

The Ewald method computes the long-range contributions to the potential energy in a system with periodic boundary conditions.
Time-scale problem rare events

Transition state theory approximation
Activated processes cannot be studied directly by conventional MD simulations due to the long time scales. The reason is not that the relevant dynamics is slow, but rather that it involves a step that involves a rare event. However, if this rare event does take place, it usually happens quite quickly, i.e. on the time scale that can be followed by MD simulations.
Formally exact expression for the rate constant
References:
Molecular dynamics simulation: - Frenkel and Smit, Understanding Molecular Simulation - Allen and Tildesley, Computer Simulation of Liquids - Thijssen, Computational Physics
Molecular dynamics simulation in C: - Rapaport, The Art of Molecular Dynamics Simulation

Molecular Dynamics

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Molecular Dynamics

Caricato da

Copyright:

Formati disponibili

Molecular dynamics

Content: The basic technique

CompPhys / Molecular Dynamics

Gran Wahnstrm, Department of Applied Physics, Chalmers