Next: Quiz
Up: Lattice simulations. Random Walk.
Previous: Random Walk on Lattice
Subsections
This is a pet model to study phase transitions. It has been introduced
before
when we discussed phase transitions. It is the lattice of spins,
interacting via the nearest neighbor interaction -Js1 s2. We
consider here cubic lattice and the simplest model with .This model is fairly simple to simulate, and at the same time provides
good illustration of how does the MC method work for multi-particle
systems.
- Phase space is discrete, - set of the spin-states for
the lattice sites.
- The total energy is:
Average magnetization is zero at T>Tc, and has
some non-zero value m(T) at T<Tc, where Tc is a Curie
temperature, or an upper critical point, if we map the fluid mixture
on the Ising model.
Two following figures show the results for a 3D Ising model from the MC simulations.
Notice, that critical temperature is determined by the location
of the susceptibility maximum. When determined this way it also reflects
the system size effects on the criticality.
A method to correct for the system size effects is described in works
by K. Binder and coworkers. Very good description can be found in the
book by K. Binder and D. W. Heermann ``Monte Carlo Simulations in
Statistical Physics'' It involves calculations of the high-order
moments of the order parameter distributions, that require very long
simulations and use tricks to speed up the calculations. This is
beyond the scope of this course.
- 1.
- Choose the initial configuration of the system, calculate energy
- 2.
- Loop through the lattice:
- Pick the lattice site i to swap, calculate the energy change
if swapped (si=-si): .
- Decide whether to accept the move, based on the Metropolis criterium:
- if , calculate
- draw a random number r from 0 to 1.
- if w<r or , swap si=-si, update
the total magnetization and the energy,
- go back to the beginning of the lattice loop
- 3.
- update current values of the running averages.
- 4.
- go back to step 2
- Fortran code Example 4 :
- Here is an example of the Fortran
code for the 2D Ising model on the lattice.
Notice that this model simulates a system of finite size.
Code.
The previous example (3) deals with the finite system with
boundaries. As it will be seen from the homework this has a
huge size effect, due to the fact that sites on the border have
no neighbors. This is exactly surface tension contribution to the free
energy. How can we go around this, if we want to simulate bulk system
where the surface effects are negligible? The size of the system that
we can simulate is restricted by the computer resources, so we have to
somehow simulate the effect of the surrounding material on the
simulated region.
- Stochastic Boundaries
- We can surround the simulated region by
the layers of spins, that are taken randomly from inside the system,
e.g. if we have 2D lattice of Nx Ny spins, we make up two
additional columns at x=0,Nx +1, and two more rows at
y=0,Ny+1, and set the spin states in this columns (rows), from
the configuration of the randomly chosen inner column (row), at one
of the previous system configurations, that is also chosen at
random. of our system
- Periodic Boundaries
- This is the most popular way, used in
almost all the simulations in the bulk systems, not only for lattice
simulations. Here we again add two additional columns
(rows),
but the extra column at x=0 has all its spins at the same states
as the column at x=Nx, x=Nx+1 is the same as x=1 (Fig.
1). Similar for the border in the y-direction (see
the first Quiz problem).
Figure 1:
Periodic boundaries for the 2D Ising lattice.
|
Next: Quiz
Up: Lattice simulations. Random Walk.
Previous: Random Walk on Lattice
© 1997
Boris Veytsman
and Michael Kotelyanskii
Thu Nov 6 22:32:26 EST 1997