Sairam Tangirala

PHYS 8602, set-2

Molecular Dynamics Simulation of Argon

Using Lennard-Jones Potential
Requirement ε
Reduced Time t * = ⋅t [2]
mσ 2
To implement molecular dynamics simulation of
Argon system using Lennard-Jones interaction L
Reduced Length L* =
potential. σ

Introduction In my simulation the reduced quantities were:

The Lennard Jones potential is characterized by two ρ * = 0.636

independent parameters, σ and ε . The potential is E* = 101.79
written as: σ * = 1.0 [3]
ε * = 1.0
σ
12
σ
6 M* = 1.0
V (r ) = 4ε ⋅ − [1]
r r
Molecular Dynamics Approach
The attractive term is the Van der Walls interaction
σ
6
In this formulation, the central assumption is that
and its functional form comes from the quantum thermodynamic every state is equally likely to be
r
12
occupied. This is a reasonable assumption for a
σ system at thermodynamic equilibrium only.
mechanical calculations. The term is the
r
repulsion term. We study the dynamical evolution of the coordinates
and momentum of a classical system of ‘N’ particles
Reduced Units and solve Newton’s (or Hamilton’s) equations of
motion. In general these differential equations of
To minimize the effects of round off errors and for motion have no analytic solution. However they can
computational advantages, reduced units are used in be solved numerically on a computer. The correct
computations. They are derived from the Lennard- formulation of these equations of motion and their
Jones potential parameters: σ , ε , the mass M of a solution constitute a “molecular dynamics”
single atom and the Boltzmann’s constant. The simulation.
reduced parameters are related to the real parameters
using: In an actual experiment, what we are really do is
taking a series of measurements of the physical
quantity over some time interval, and then averaging
Reduced Density ρ * = ρσ 3
over this interval. Thus, an MD simulation more
k BT closely represents both the dynamical evolution of a
Reduced Temperature T * =
ε system of particles in time, as well as the
E measurement procedure.
Reduced Energy E * =
ε
 System Initialization Calculating v(t − ∆ )

When an MD simulation is started for the very first Using initial positions r (t − ∆ ) , calculate the Potential
time, one needs to specify the initial arrangement of
Energy (P.E) of the system using Lennard-Jones
the atoms in space, as well as their initial velocities.
interaction.
Below I mention the initial conditions that were used
in my simulation.
Kinetic energy of the system can be calculated using
conservation of energy.
•Atoms were arranged on a 7x7x7 lattice. Only 256
lattice sites were used to place atoms and rest were N
1
left unoccupied. K.E (t − ∆ ) = = E* - P.E (t − ∆ )
2
mvi
2 i =1

•The potential and the kinetic energy is distributed [6]

among the particles such that the conservation of
energy is satisfied.
•The system satisfies periodic boundary conditions
Using Verlet Algorithm
Verlet integration is used for calculating the
trajectories of particles in molecular dynamics
simulations. It offers more accurate solutions than the
Euler’s integration method, besides having time-
reversibility. This algorithm calculates the positions
and velocities using iterative method as shown below:
N
Using K.E, I determined the total velocity vi of
i =1
all the atoms. This total velocity was distributed
among N atoms equally in magnitude. The directions
of velocities for N atoms are chosen randomly.
Then the time-progression of positions and velocities
is implemented using Verlet algorithm as described in
above section.
The simulation is started after choosing a time-step ∆ .
Data are calculated and stored in files for analysis.

r (t + ∆ ) = 2 r (t ) + r (t − ∆ ) + a(t ) ∆2 Results

r (t + ∆ ) − r (t − ∆ )
v(t ) = [4]
2∆

It is seen that for calculating r (t + ∆ ) we need r (t )

and r (t − ∆ ) . So we need a different method to get
r (t ) using r (t − ∆ ) , for this I chose the Taylor’s
method for calculating the positions at the first time-
step using the formula:

r (t ) = r (t − ∆ ) + ∆ v(t − ∆ ) [5]

r (t − ∆ ) is calculated using initialization of Argon

atoms on 7x7x7 lattice.
Plot(1): Time-propagation of error in E*
Plot (1) shows the propagation of error in calculation Energies fluctuate at initial times and at later times
of total energy as a function of time. Error become more stable about their mean values. This
propagation for varying time-step ( ∆ ) has been reduction in the fluctuations of energies is a sign of
shown. In general the error increases as the time the system reaching equilibrium.
progresses; this is due to the truncation-errors
introduced by the limited terms in the series expansion
as implemented in the Varlet’s algorithm

The rate of the increase in the error depends on the

value of time-step ∆ . For a larger ∆ the error
increases at a larger rate as compared to that of a
smaller time-step ∆ .

For a given simulation the optimum value of time-

step ∆ to be used depends on the system and the
computer resources. It is possible to get very small
errors in the calculations by choosing small ∆ , but the
drawback is that it takes a larger amount of time to do
simulations and to reach equilibrium.
Plot(3): Time-propagation of potential energy
In my case I choose the value of ∆ =1.0 x 10-4 and
perform the simulations for a total of 50x103 time-
steps. During equilibrium, I calculated the average kinetic
energy of the system from time 10.0x103 to 50.0x104.
The plot (2) and (3) shows the time
dependence of kinetic energy and potential energy Average reduced temperature T* is calculated from
respectively. this average kinetic energy using the equi-partition
theorem.
2 K.E
T* = ⋅ [7]
3 N

Answer : T* = 2.58(3)

Statistical error bars are small because I have very

large production run (=40.0x104 time-steps). Also I
suspect that the reported average T* may be larger
than the expected value because the total energy is no
more to initial value (E*= 101.79) at larger times.
These errors are due to Verlet’s method for smaller
time-step ∆ =1.0 x 10-4 that I used for simulation.

References
Plot(2): Time-propagation of kinetic energy [1]. “Computer Simulation of Liquids”
by Allen and Tildesley
Both kinetic and potential energy evolve with time n a
complementary way such that the total energy remains [2]. Centre for Molecular Simulation
a constant. http://www.swinburne.edu.au