A Study of Computer-Based Simulations for Nano-Systems and their types

Tamal Sarkar1, Samir Chandra Das2, Ardhendu Mandal3

North Bengal University
P. O., N.B.U, Siliguri-734013, India
1
[University Science Instrumentation Centre (USIC) , [email protected]]
2
[USIC, [email protected]]
3
[Computer Science and Applications (CSA), [email protected]]

ABSTRACT
In most of the cases, the experimental study of Nanotechnology involves high cost for Laboratory set-up and the experimentation
processes were also slow. So, one cannot rely on experimental nanotechnology alone. As such, the Computer-Based molecular
simulations and modeling are one of the foundations of computational nanotechnology. The computer based modeling and
simulations were also referred as computational experimentations. In real experiments, the investigator doesnt have full control over
the experiment. But, in Computational experimentation the investigator have full control over the experiment. The accuracy of such
Computational nano-technology based experiment generally depends on the accuracy of the following things: Intermolecular
interaction, Numerical models and Simulation schemes used. Once the accuracy of the Computational Scheme is guaranteed one can
use that to investigate various nonlinear interactions whose results are completely unexpected and unforeseen. Apart from it,
numerical modeling and computer based simulations also help to understand the theoretical part of the nano-science involved in the
nano-system. They allow us to develop useful analytic and predictive models. In this paper, a brief study of Computer-BasedSimulation techniques as well as some Experimental result obtained using it were given.

INTRODUCTION
The computer based simulation techniques [1] are widely used for computational nano-technology. The
frequently used simulation approaches are Monte Carlo (MC) and Molecular Dynamics (MD) methods. The
other various simulation methods come from these two basic methods. Apart from it, the optimization
techniques were also needed in MC and MD. The computer- based simulation methods, developed for nanosystems, generally consist of a computational procedure performed of few atoms or molecules confined in a
small geometrical space. This geometrical space in which the simulation is performed is termed as cell. In the
subsequent section, a brief classification of simulation methods based on Accuracy, Computational Time etc
were given. After it, classifications of optimizations in molecular simulations were discussed. In next section
after it, we have discussed the result of computer -based simulation.
MONTE CARLO AND MOLECULAR DYNAMICS SIMULATION METHODS [2, 3, 6]
The two basic used simulation approaches are Monte Carlo (MC) and Molecular Dynamics (MD) methods. All
the other various simulation methods come from these two basic methods. A brief over-view with areas of
application of the both are discussed below. These concepts are essentially required to understand the
methodology of classification of Computer-Based Simulation methods based on accuracy and time-complexity.
MC method uses random numbers to perform calculations. There are many areas of application of MC
Methods including nano-material. Some important areas where we apply MC method are: - (i) Estimation of
large-dimensional integrals (ii) Generating thermodynamic ensembles in order to compute thermal averages of
physical equilibrium quantities of interest and simulation of non-equilibrium phenomena such as growth and (ii)
Computation of distribution functions out of equilibrium known as Kinetic Monte Carlo [5].
MD deals with predicting the trajectories of atoms subject to their mutual interactions and eventually an
external potential. Some important areas of application of MD are: - (i) Computation of transport properties
such as response functions, viscosity, elastic moduli and thermal conductivity (ii) Thermodynamic properties
such as total energy and heat capacity and (iii) Dynamical properties such as phonon spectra.
CLASSIFICATION
OF
SIMULATION
METHODS
BASED
ON
ACCURACY
AND
COMPUTATIONAL TIME
Computer based methods used for simulation of various properties of nano scale systems differ in their level of
accuracy and time-complexity to perform such calculations. Based on it, the required time scale for these
methods can be from tens of picoseconds (e.g. ab initio MD calculations) to few microseconds or more
nd

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067

(classical molecular dynamics simulation). There are also methods which require very long computational time
such as cluster growth and may require super computers to achieve fast results.
Based on these facts we may classify the methods into following groups (i) Methods with highest degree of
accuracy (ii) Methods with second highest degree of accuracy (iii) Semi-empirical method and (iv) Stochastic
method (shown in Table 1) [4, 5].
Methods
Methods
with highest
degree
of
accuracy

Purpose
Investigation of both electronic and
atomic ground state, optical and
magnetic properties of weakly
interacting and also strongly
interacting correlated systems

Methods
with second
highest
degree
of
accuracy

Accurate calculations of ground

state
structure
by
local
optimization;
Calculation
of
mechanical, magnetic and optical
properties of small clusters and
perfect
crystals
of
weakly
interacting
electron
systems,
estimation of reaction barrier and
paths
Search for ground state structure
by GA, Simulated Annealing (SA)
or local optimization if a good
guess for the structure is known;
simulation of growth or some
reaction mechanisms; calculation
of response functions

Semiempirical
method

Stochastic
method

Investigation of long timescale

non-equilibrium phenomena such
as transport, growth, diffusion,
annealing, reaction mechanisms
and also calculation of equilibrium
quantities and thermodynamic
properties

TABLE-1
Input
Atomic
species,
coordinate, systems
symmetry,
interaction
parameter

Output
Total energy, excitation
energy
and
spin
densities, force on atoms

Atomic species and

their
coordinates,
symmetry of the
structure, pseudopotential
or
Hamiltonian
parameters for the
species considered

Total energy, charge and

spin densities, forces on
atoms, electron energy
eigen values, capability
of doing Molecular
Dynamics, vibrationalmodes
and
phonon
spectrum

Atomic
species,
their
coordinates;
parameters of the
inter-particle
potential,
temperature
and
parameters of the
thermostat or other
thermo-dynamic
variables.
Parameters of the
inter-particle
potential,
temperature
and
parameters of the
thermostat or other
thermodynamic
variables

Output
of
TightBinding(TB):
Total
energy, charge and spin
densities,
force
on
atoms,
particle
trajectories,
phonon
calculation; mechanical
magnetic and optical
properties of clusters and
crystals
Statistics of several
quantities
such
as
energy, magnetization,
atomic displacements

Examples
Ab initio methods for
electronic
structure
calculations of correlated
systems, Quantum MC,
Quantum Chemistry and
Many body
Ab initio methods for
normal Fermi liquid systems
based on either Hartree
Fock or Density Functional
Theories

Semi-empirical methods of
large systems or long time
scale. TB or LCAO
molecular dynamics based
on classical potential or
force -field

MC
walk
towards
equilibrium,
kinetic
or
dynamical MC (growth and
other
non-equilibrium
phenomena)

CLASSIFICATION OF OPTIMIZATIONS IN MOLECULAR SIMULATIONS

Optimization [4, 7] in any numerical computation is to look for the best value of a multivariate function among
the many numbers generated during the computation. These are common in
molecular simulations where a system consisting of assembly of N particles is
described in a 6N-dimensional phase space. In such cases, a cost (here energy) is
associated with every configuration and objective is to seek sets of data points that
maximize (or minimize) the objective function. There are two kind of optimizations
required in every molecular simulation (i) Local optimization (ii) Global
optimization (Fig. 1)
Some examples in each type of optimizations are discussed below:Local Optimization Methods: (i) Steepest Descent Methods (ii) Damped Newtonian
Fig.1. Local and
Dynamics (iii) Conjugate Gradient Method (iv) Quasi- Newton Method
Global Maxima
Global Optimization Methods: (i) Simulated Annealing Method (ii) Genetic
Algorithms
1

0.5

-0.5
5

nd

-5

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067

-5

Local Optimization Methods

Steepest Descent Method (SDM)
This is the simplest optimization method. In this, one has to move along force applied on the particle. This
method certainly works because at each step the potential energy is lowered, but its convergence is often slow.
Even a situation may arrive where a particle will keep oscillating around the anisotropic valley and may take
may step for reaching the minimum.
The simple iterative formula to implement SDM is x(t + t ) = x (t ) + f [ x (t )] where f ( x ) = E ( x) the force
is. In its discretized form, it can be written as x(i + 1) = x(i ) + f [ x(i )] .
Damped Newtonian Dynamics Method
This is faster than SDM. Here we consider, the algorithms
x(i + 1) = [ 2 x(i ) + f ( x(i )) x(i 1)(1 )] /(1 + )
Conjugate Gradients Methods (CGM)
It is more intelligent method than SDM and Damped Newtonian Dynamics Method. CGM will find the
minimum of the anisotropic valley faster than the SDM. In principle, it finds the minimum of a quadratic
function in N dimensions in exactly N line
minimization steps. If one is near t he minimum, the
convergence to the minimum is almost as fast as the
Populationof
String
SDM.
Quasi- Newton Methods

Genetic
Operation

Parentsof
Strings

Newtons method: It is a method to find the zero of a

function f in one dimension. The expression for the
iterative scheme is as follows:
x(i + 1) = x(i ) f ( x((i )) / f ' ( x(i ))

GLOBAL OPTIMIZATION METHODS

Evaluationof
FitnessValues

Creationof
MatingPool

Fig.2.FlowchartforprocessofaGAglobaloptimization
ofamolecularsimulation
Start

Genetic Algorithm (GA)

It is a biological evolutionary process in intelligent search,
machine learning and optimization problems. This
algorithm is basically search algorithm based on the
machines of natural selection and nature genetics shown in
Fig 2. It is the balance between efficiency and efficacy
necessary for survival in many different environments. In
order for GA to surpass their more traditional in the quest
for robustness, GA must differ in some fundamental ways.
GA are different from more normal optimization and search
procedures in four ways:
i. GA work with a coding of the parameter set, not the
parameters themselves.

MolecularSimulationathigh
temperature
Rearrangementofsystemfromits
statusquo
Lowering of temperature to bring
the system into a stable state

No

Is system is in
equilibrium?

Yes
Stop

nd

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
Fig.3.Flowchartforprocessofasimulatedannealing
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067

globaloptimizationofamolecularsimulation

ii. GA search from a population of point not a single point.

iii. GA use pay off (objective function) not derivatives or other auxiliary knowledge.
iv. GA use probabilistic transition rules, not deterministic rules.
Simulated Annealing Method
This method was first proposed by Kirkpatrick et al. this method was chosen for large scale optimization
problem in which required global minima is hidden between many local minima. Simulated annealing
distinguishes between the different local optima in a same way analogous to the physical process of annealing.
The flowchart for simulated annealing to go through for global optimization is shown in Fig. 3.

COMPUTER BASED EXPERIMENT AND RESULTS OBTAINED USING GA

Present work declares that a soft computing tool, GA is used to get the optimized system parameters of GaAs
QW for a desired high frequency response characterized by a cutoff frequency (f3dB). By the present work, a
model for optimized system parameters of GaAs, QW is obtained for a high frequency under hot electron
condition. In GA, a fitness function is the main criteria for reproduction. The fitness values are used to favor
high fitness individuals over low fitness individuals to take part in the process of reproduction. In the present
application of GA, we find the f3dB for a semiconductor quantum structure for its different system parameters.
So we find the variation of f3dB (cut off frequency at which the mobility falls to 0.707 of its low frequency
value) and one particular parameter of the system where the other parameters are optimized by the GA. By
taking the other parameters in one form, we can be able to find the fitness values. These fitness values are
converted to binary form and then proceed for further GA operation. After reproduction, simple crossover may
proceed in the following steps. First, members of newly reproduced strings in the mating pool are mated at
random. Second, each pair of strings undergoes crossing over as follows: - An integer position t along the
strings is selected uniformly at random between 1 and the string length less one [1, P - 1]. Two new strings are
created by swapping all characters between positions (t + 1) and P inclusively. For example consider two strings
X and Y.

TABLE2:Comparisonofoptimizedsystemparameters
5
forfrequency=300GHzandF
0=0.7510 V/mofGaAs.
|
ac
n2D
TL
L z
Te
mobility
5
2
X = 10011|1011
(K)
(K)
(nm)
(10
m
)
ac(m2/V.s)
1.61
77
335
9.0
95
Y = 00110|0010
1.83
100
170
5.9
90
|
1.91
125
215
7.0
95
1.60
150
190
5.0
110
Suppose in choosing a random number between 1 and 8 (9 1.98
175
300
10.0
120
1), as P = 9 as indicated in the above by dotted line. Then the
1.78
200
275
8.0
120
result of cross over which produces two new strings indicated
1.65
225
285
8.0
95
1.56
250
305
8.0
90
by X1 and Y1 (say).
1.66
275
350
9.0
115
X1 = 100110010
1.37
300
345
7.0
110
[TL= Lattice temperature, Te= Electron temperature,
Y1= 001101011
n2D=Carrierconcentration,Lz=Channellength]

The mutation operator plays a secondary role in the simple

GA. We simply note that the frequency of mutation to obtain good results in empirical GA studies is on the
order of one mutation per thousand bit transfers.
NUMERICAL RESULTS

nd

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067

In this optimization technique, we consider a square quantum well of GaAs and electron temperature model is
employed for numerical calculations with the following system parameters : Electron effective mass m* =
0.3735 x 10-31kg, Static dielectric constant Ks = 13.88,optic dielectric constant K = 11.34, LO phonon
angular frequency 0 = 5.37x 1013 rad s-1, longitudinal elastic constant CL = 14.03 x 1010Nm-2, acoustic
deformation potential constant.E1= 17.6x 10-11 J, background ionized impurity concentration nbi = 6.0 x 1021m3
-1
, longitudinal acoustic velocity, u1 = 5.22 x 10ms .
The frequency of the applied is varied and at each frequency the optimum parameters are determined by
employing the Genetic algorithm for a particular dc biasing field. For a particular dc biasing field, it is possible
to predict the optimum values of the system parameters like electron temperature, channel width, carrier
concentration, for the realization of desired high frequency response and ac mobility[8-11]. The optimized
parameters will obviously save the search time for the technologists involved in the fabrication of the devices.
We have computed the optimized parameters using GA to get
TABLE 3: Optimized parameters for dc bias field
desired millimeter and sub-millimeter wave response
F0 = 1.0 x 105 v/m
n2D
TL ( k)
Lz ( nm )
characteristics. The application of GA in the device parameter Frequency
15
-2
(
GHz)
(
10
m
)
optimization for getting the best performance is a new area of
110
10.0
108
85
research. The optimized system parameters presented here
135
8.0
118
115
predict the better performance of GaAs QWs in the
160
9.0
115
182
microwave and millimeter wave regime and can be used to
280
180
10.0
122
285
210
9.0
103
analyse the experimental data when they appear in the
237
230
6.0
88
literature. Artificial neural network and genetic algorithm
85
290
250
6.0
based computations are performed with the parameter values
[n
=
Carrier
Concentration,
L
=Channel
Width,
TL( K)
2D
z
of In0.53Ga0.47As for the dc biasing field F0 of 0.75105 V/m
= Lattice Temperature]
and the frequency of the applied ac field of 300GHz.
Optimized system parameters for the frequency 300GHz of the applied small-signal electric field and the dc biasing field
F0 of 0.75105 V/m are given in Table 2, which reveal that for the desired cutoff frequency at a particular dc biasing field
it is possible to predict the optimum values of the system parameters like carrier concentration, channel length and
electron temperature for realizing a particular high frequency response characterized by a cutoff frequency. It is found that
at the lattice temperature TL of 200K, the optimized system parameters are: electron temperature Te=275K, 2D carrier
concentration n2D= 8105m2 and the channel length Lz= 120 nm.The corresponding parameters obtained by GA are
obtained placed in the Table 3. Comparison recalls that the ANN results are also equally suitable device GA based results.

Conclusion
This paper demonstrates the various processes through which computer based simulations and
optimizations are used in various field starting from classical to nano-levels. The same will also help the reader
to choose the best method suitable for a particular application of their interest. To give a better view we have
given one example of computational technique that we have used for nano device simulation from one of our
experimental work.
Reference:
[1]. Narsingh Deo, System Simulation with Digital Computer, PHI Pvt. Ltd. (2001)
[2]. Landau David P. et al, A Guide to Monte Carlo Simulations in Statistical physics, Cambridge University
Press, Second Edition (2005)
[3]. Field Martin J., A Practical Introduction to the Simulation of Molecular Systems, Cambridge University
Press, First Edition (2007)
[4]. Mansoori G. Ali, Principles of Nanotechnology, World Scientific, First Edition, (2005).
[5]. Lakhtakia A., A Handbook on Nanotechnology Nanometer Structures Theory, Modeling and
Simulation, PHI Pvt. Ltd (2007).
nd

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067

[6]. Chandra S., Computer Application in Physics, Narosa Publishing House (2003).
[7]. Jorge J. More' et al, Optimization Software Guide, SIAM Publications (1993).
[8]. Das S. C., Neuro-Genetic Approach for Optimizing Parameters for the Polar Semiconductor,
Proceedings of NATCOMNNAMTECH 2007.
[9]. Sarkar S. K. et al, Nanotechnology, 9 321 (1998).
[10]. Karmakar A. et al, CODEC-2006, December (2006).
[11]. Sarkar S. K. et al, Physica B, 325, 189 (2003).

nd

PublishedinProceedingsof2 NationalConferenceonNanomaterialsandNanotechnology(2123December,2009),UniversityofLucknow,U.P.
ISBN:9789380043616,PublishedbyExcelIndiaPublishers,61/28,DalpatSinghBuilding,PratikMarket,Munirka,NewDelhi110067