Metropolis Monte Carlo simulation of the Ising
Model
Krishna Shrinivas (CH10B026)
Swaroop Ramaswamy (CH10B068)
May 10, 2013
Modelling and Simulation of Particulate Processes (CH5012)
Introduction
The Ising model, developed by Dr. Ernst Ising is used for
modelling ferromagnetic and anti-ferromagnetic materials
The model represents a lattice occupied by atoms which can
each have dipole moments or spins
The model predicts a second order phase transition occuring
at the Curie temperature for dimensions higher than 1
Phase transition is identified from ensemble properties and
compared with the theoretical model which has been solved
exactly for zero external field
Ferromagnetism
One of the fundamental properties of an electron is that it has
a dipole moment
This dipole moment comes from the more fundamental
property of the electron that it has quantum mechanical spin
The quantum mechanical nature of this spin causes the
electron to only be able to be in two states, with the magnetic
field either pointing "up" or "down"
When these tiny magnetic dipoles are aligned in the same
direction, their individual magnetic fields add together to
create a measurable macroscopic field
Ferromagnetism
Ferromagnetic materials are strongly ordered and have net
magnetization per site as 1/1 under temperatures T < Tc
They are able to maintain spontaneous magnetization even
under the absence of external fields
At temperatures above Tc , the tendency to stay ordered is
disrupted due to competing effects from thermal motion
The ferromagnetic substance behaves like a paramagnetic
substance at T > TC , showing no spontaneous magnetization
Ferromagnetism
The reason for this strong alignment/bonding arises from
exchange interactions between electrons
These exchange interactions are 1000 times more stronger
than dipole interactions, characteristic of paramagnetic and
diamagnetic substances
In antiferromagnetic materials, these exchange interactions
tend to favor the alignment of neighbouring atoms with
opposite spins
As far as the Ising Model goes, coupling parameter J > 0 for
ferromagnetic substances and J < 0 for antiferromagnetic
substances
Hamiltonian
I
Each atom can adopt two states, corresponding to
s = {1, 1}, where s represents the spin and the spin
interactions are dependent on the coupling parameter Jij
The lattice model has periodic boundary conditions and
extends infinitely
This model is defined in the Canonical Ensemble(N, V , T )
and the Hamiltonian is defined as below
H = Jij
X
<i j>
si sj h
si
where,
Jij = Coupling parameter between the adjacent atoms
h = External Field Strength
si,j = Spin of particle
Partition Function
The Partition function corresponding to the Hamiltonian for the
above model is defined as:
Qpartition =
e H
states
where =
1
kb T ,
kb Boltzmann Constant
Ising 2-D Square Model
For an isotropic case, where the coupling along the rows and
columns are equal, the critical temperature has been found to be
2
kB Tc
2.269
=
J
ln(1 + 2)
The 2-D square model in the absence of an external field has been
solved (Lars Onsager, 1944) in the absence of an external field
(h = 0)
Internal Energy:
P
U =< H >=
He H
states
Qpartition
Isothermal Susceptibility:
T = (
1
dM
)h =
(< Mv2 > < Mv >2 )
dh
kB T
Ising 2-D Square Model
Specific heat at constant volume:
(< H 2 > < H >2 )
dU
)h =
dT
kB T 2
Net magnetization per particle:
Cv = (
k=
)sinh( k2J
)
sinh( k2J
BT
BT
Spontaneous Magnetization per site, for T < Tc (isotropic):
M = [1 sinh4 (
2J 1
)] 8
kB T
Energy per site :
2
2J
2J
U = Jcoth(
)[1+ (2tanh2 (
)1)
kB T
kB T
1
q
4k
sin2 ()
(1+k)2
Applicability of Monte Carlo simulations
The rationale behind Monte Carlo sampling techniques is
inherently based on the sampling of time steps from an
exponentially distributed function and making stochastic
decisions
Since the Canonical ensemble arises from an exponential
distribution of states, a similar rationale can be used to
sample across the different states of the system until
equilibrium is reached
Setting up the Problem
An optimal value of 900 atoms was chosen to model this
system with periodic boundary conditions
The lattice was represented by a 31 31 random matrix, with
each element being randomly assigned with the values -1 or 1
It is chosen as a 31 31 matrix so as to ensure that all the edges
are periodic in nature. The values at the first and last column, first
and last row are made the same
Setting up the Problem
All quantities are manipulated in normalized units
Temperature is normalised and kB 1 unit. The coupling
strength is taken as J = 1 unit
The whole procedure was done for different temperatures
ranging from 0.5 3 and the number of iterations for
equilibration was taken as n 108
Initial Lattice Structure
Figure: Intial lattice, blue squares represent s = 1 and red squares
represent s = 1
Algorithm
Initialise the system randomly with spins, at a given
Temperature
Set the value of the external field, in most cases h = 0
Make a random flip in the spin of some atom
Compute the Energy change arising from this, due to only the
neighbouring atoms
Ensure that the periodic boundary conditions are in place to
take care of edge effects
If 4E < 0, accept this configuration and continue this process
Algorithm
If 4E > 0, accept this configuration with a probability of
), else retain the old configuration
p = exp( k4E
BT
Once every m, iterations , sample the system for important
ensemble properties
This sampling has to be done after discarding the edges
because they only represent the periodic boundary conditions
Now allow the system to equilibriate (typically takes
n3 iterations)
Estimate the average properties, variance terms(Susceptibility
and Cv )
Repeat this procedure at different temperatures
Visualization of MMC
30
25
20
15
10
10
15
20
Figure: T = 0.5
25
30
Visualization of MMC
30
25
20
15
10
10
15
Figure: T = 1
20
25
30
Visualization of MMC
30
25
20
15
10
10
15
20
Figure: T = 1.5
25
30
Visualization of MMC
30
25
20
15
10
10
15
Figure: T = 2
20
25
30
Visualization of MMC
30
25
20
15
10
10
15
20
Figure: T = Tc 2.269
25
30
Visualization of MMC
30
25
20
15
10
10
15
20
Figure: T = 2.5
25
30
Visualization of MMC
30
25
20
15
10
10
15
Figure: T = 3
20
25
30
Visualization of MMC
30
25
20
15
10
10
15
20
Figure: T = 3.5
25
30
Visualization of MMC
30
25
20
15
10
10
15
Figure: T = 4
20
25
30
Results
Energy per atom versus Temperature
1
1.5
Energy per atom
2.5
3.5
4
0.5
1.5
2
Temperature T
Figure: Energy per site versus Temperature
2.5
Results
Heat Capacity versus Temperature
5000
4500
4000
3500
Heat Capacity
3000
2500
2000
1500
1000
500
0
0.5
1.5
2
Temperature T
Figure: Heat Capacity versus Temperature
2.5
Results
Magnetization versus Temperature
1
0.9
0.8
0.7
Magnetization
0.6
0.5
0.4
0.3
0.2
0.1
0
0.5
1.5
2.5
Temperature T
Figure: Magnetization per site versus Temperature
Results
Magnetic Susceptibility versus Temperature
0.18
0.16
Magnetic Susceptibility per atom
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0.5
1.5
2.5
Temperature T
Figure: Magnetic Susceptibility per site versus Temperature
Ehrenfst Classification of Phase Transitions
For an nth order phase transition, at the transition point all
the nth order derivatives of the ensemble property diverge
In the Ising model, both specific heat (Cv ) and magnetic
susceptibility (T ) have sharp discontinuites at the Curie
temperature
Since, Cv and T are second order derivatives of ensemble
properties, this is classified as a second order phase transition
Physical understanding of the Phase Transition
As the temperature increases, the tendency to stay ordered
reduces because of thermal fluctuations
The net magnetization, which is a function of net order in the
system starts dropping
Beyond the curie temperature Tc , there is no more tendency
to stay ordered, and due to complete disorderness, the net
magnetization per site drops to zero
Hysterisis
The equilibriation is acheived with some value of h = l using
the aforementioned algorithm
Now h is slowly changed to h = l in discrete steps
During each of these steps, the previous equilibriated
configuration is given as input to the system to undergo
equilibriation again
Average and variance quantities are calculated and plotted
Results
1
Reverse/Decreasing Field Strength
Hysteresis increasing field strength
0.8
0.6
Magnetization per site
0.4
0.2
0.2
0.4
0.6
0.8
1
4
0
External field strength
Figure: Hysteresis Loop
Critical Exponents
Critical exponents describe the behaviour of physical
quantities near continuous phase transitions
It is widely believed, but not proved formally, that these
exponents are independent of the physical properties of the
system at consideration
These only depend on the following properties
I
I
I
Dimension of interaction
Range of Interaction
Spin Dimension
Critical Exponents
Near the phase transition temperature TC , we define the
reduced Temperature = (T Tc )/Tc
It is stipulated that properties of the system, fi vary in a
power law order, i.e., f , asymptotically as 0
Some fluctuations are only observed in one phase, i.e. ordered
or disordered
The difference in the phases is determined by the order
parameter , which is Magnetization for Ferromagnetic
materials
Critical Exponents
Some important critical exponents are defined for CV ,
( ) and T
These are estimated from the simulations performed through
Partial Least Squares Regression Technique(of the log values)
Analytical
0
1/8
7/4
Simulation (MMC)
0.013
0.1288
1.81
Table: Comparison of Analytical and Simulation based Critical
Exponents
Anti-Ferromagnetic Materials
For, anti-ferro magnetic materials, the coupling parameter
J < 0 between spins.
This ensures that all the spins are always oppositely aligned at
T < Tc,as this maximizes the energy(From the Hamiltonian)
of the system.
Since it is symmetric with all other respects(except
Magnetization), all the variation in ensemble properties
resemble those of the ferromagnetic system
At temperatures beyond T > TC , the thermal fluctuations
start to weigh in and the tendency to remain ordered is
removed
Anti-Ferromagnetic Materials
30
25
20
15
10
10
15
Figure: T = 1
20
25
30
Anti-Ferromagnetic Materials
30
25
20
15
10
10
15
Figure: T = 3
20
25
30
Anti-Ferromagnetic Materials
Antiferromagnetic materials always prefer neighbouring atoms
to have alternating spins. Hence presence of external field
does not impose any kind of magnetization on the material.
This is observed from the simulations that the Net
Magnetization Per Atom M, is invariant under external fields
Since there is no external way to magnetize the system and a
state having zero net magnetization is preferred, these kind of
materials do not show any cyclic properties or hysteresis loops.
Conclusions
Phase transition is observed and signified by the change in the
order parameter()/net magentization(M) with increasing
temperature in the absence of any external field
Phase transition is observed (ordered to disordered state) at
the curie Temperature of Tc 2.26, in agreement with the
analytical result.
This phase transition is characterized as a second order phase
transition based on the divergence/discontinuity characteristic
of the second derivative properties, for example Magnetic
Suscpetibility T and Constant Volume Heat Capacity Cv .
Conclusions
The efficacy of Metropolis Monte Carlo Algorithms is evinced
by the rapid convergence to equilibrium, taking only a few
minutes on a personal computer
The ability of the algorithm to predict equilibrium values with
great precision is very evident in the close overlap between the
theoretical and simulation based results of the values of the
critical exponents
An important aspect of ferromagnetic materials, memory of
past configurations while subjected to varying external
fields(hysteresis) is evinced by these simulations
In addition, antiferromagnetic systems were also studied and
characterized
Possible Extensions to this Work?
Working with a different force coupling parameterJij which is
distance dependent (Like Lennard-Jones potential)
Simulating a 3 D Ising model and deducing Phase transition
properties and critical exponents in 3 D
Extrapolating the Ising Model to a Lattice based Gas Model
and predicting phase transition based on revised order
c
parameter =
c
Try and establish validity of Ising model for alloys/mixtures of
ferromagnetic substances and predicting phase transition
phenomena
References
Onsager, Lars (1944), "Crystal statistics. I. A two-dimensional
model with an order-disorder transition", Phys. Rev. (2) 65
(34): 117149
Statistical Physics, 2nd edition, Landau&Lifshitz
Markov Processes, Gillespie, 4th edn
Class Notes!
