Vous êtes sur la page 1sur 6

A Brief Study of Ferromagnetism and Temperature

The Square Lattice Ising Model By: Nathan Parzuchowski The Ising Model Problem The Ising Model problem is a common demonstration of one of the most fundamental elements of statistical physics: a lattice of magnetic spins. The spins are dipole moments oriented either up or down. We dene a spin with a value of 1 or -1, depending on its orientation. Spins can cancel if anti-aligned, or sum if aligned. In order to give a more manageable quality to the macroscopic quantities of the square lattice, we dene a quantity m, known as the magnetization of the lattice. This is equivalent to the expectation value of the dipole moment over the entire lattice. This value is easy to calculate in the absence of an external magnetic eld: m= 2s N (1)

where 2s is the spin excess dened as the difference 2s = ( N+ N ), the number of spins up and down respectively. N is the total number of spins in the lattice. In the Ising Model problem, we are tasked with determining the behavior of the magnetization of the lattice in response to varying temperature. The problem was created in 1920 by physicist Wilhelm Lenz and solved in 1925 by his student Ernst Ising, for which it is named. The problem in general involves a lattice of spins placed in an external magnetic eld. The behavior of the lattice can be predicted using its energy hamiltonian: H = J

<ij>

Si S j HB

i =1,N

Si

(2)

Here J is the ferromagnetic exchange constant, which measures the electrostatic exchange energy between spins. J is difcult to calculate and varies based on the material represented in the lattice. Si and S j are spins, and further more are specied as nearest neighbor spins by the notation

< i, j >. Nearest Neighbor refers to the spins to the immediate right, left, front, or back of the
spin in question. As is evident from the hamiltonian; J is the proportion by which spin effects contribute to the energy of the lattice. The magnetic effects on energy are determined by the magnetic eld HB and the atomic dipole moment of the surroundings: .2 1

Developing a Simulation For simplicity, we have set the external magnetic eld to zero. The hamiltonian simplies to: H = J

<ij>

Si S j

(3)

This result shows us that when spins are aligned, their energy is lower than when anti-aligned. This effect is mostly contributed to the Pauli Exclusion Principle, which states that two electrons cannot occupy the same quantum state. If two electrons from different atoms have the same spin, they then occupy the same orbital state, and thus cannot be brought close together. Electrons in anti-aligned states can be brought together, because there is no violation of the Pauli Exclusion Principle in this case. The interaction between the atoms at larger distances will produce lower energies due to the electrostatic nature of the exchange energy. Since nature tends toward lower energy, it is energetically favorable for spins to be aligned.2 We are interested in the Curie Temperature Tc . for T > Tc , a ferromagnet will lose its magnetism. Tc is related to J, and like J, is tricky to calculate, and varies by material. We dene T0 = J/K B . Where T0 is the temperature where magnetization begins to decrease. In 1941 H.A. Kramers and G.H. Wannier used a duality argument to show that:1 Tc = 2.269 T0 (4)

We nd that when temperature is brought near Tc , a ferromagnet does not necessarily behave as one could predict from energetic favorability. A useful Monte Carlo simulation can be created to model the behavior of the lattice when its temperature is changed. The general premise of the Monte Carlo method is random sampling. A random spin shall be selected from the many possible, and then it will be subjected to tests, to determine if it is energetically favorable to alter its current state. There are many different algorithms which could be considered in running a Monte Carlo simulation, the most popular of which being the Metropolis Algorithm. This is the algorithm which will be used in simulating the Ising Model. The Metropolis Algorithm takes into account the probability of nding certain energy states in

response to temperature. This probability is proportional to the Boltzmann factor. P( E) exp( E) Where = 1/(K B T ). With some manipulation of this proportion, we nd: P( E f ) = exp( E) P( Ei ) (6) (5)

From here the Metropolis Algorithm is developed. First we select a spin at random. The change in energy, E, is calculated for changing the orientation of that spin. If E is negative, we keep the change in accodance with energetic favorability. This is also evident in (5), as the P( E f ) term is shown to dominate P( Ei ) in the ratio. If E is positive, the probability of the spin changing is determined by (5). We keep the change based on whether or not a random number is found that is less than the ratio calculated in (5). If the random number is less, than we keep the change. This process should be repeated many times to nd an accurate sample of the systems behavior.3 In our simulation, we arrange the lattice in a two dimensional array, where Si j denotes a single spin in the lattice. We seek to nd the change in energy associated with changing the orientation of a single spin, as a result of nearest neighbor interactions. Through manipulation of (3), we nd: E = 2( J (Si j (Si + 1 j + Si 1 j + Si j + 1 + Si j 1 ))) (7)

The program ising model. f 90 was written to execute the Metropolis Algorithm described above. The program prompts the user to input the simulations temperature in kelvin, the size of the lattice, and the amount of Monte Carlo steps that the algorithm will be run over. The initial lattice can be selected as a random lattice, or a ferromagnet with all spins oriented up. This program also gives the option of running the entire Monte Carlo simulation repeatedly for a specied amount of trials. The net magnetization is stored for each Monte Carlo step. At the end of the simulation, the magnetization at each step is averaged over the amount of trials, and written to le. The magnetization is also averaged into a single value so that data concerning the temperature can be collected.

Simulations and Results For the simulations, we chose an arbitrary J and used temperatures in a range near T0 . We set J = K B such that T0 =1. The Boltzmann factor thus simplies to exp(E/T ) where E is now (6) with J dependence divided out.

FIG. 1: Random 40x40 Lattice, 100 Trials Averaged

Our rst simulation involves the time dependence of a random lattice subjected to varying temperatures near T = 1.0 K. The simulation is run over 10000 Monte Carlo steps (our time dependence), and it is averaged over 100 trials. The results are displayed in Fig. 1. The magnetization is relatively unchanged throughout the simulation. At all four temperatures, the magnetization remains very close to zero. The lattice has no signicant trend towards ferromagnetism at any point during the simulation. In theory, when a disordered lattice of spins is cooled below Tc , it is able to return to ferromagnetism. We do see a minor upward trend at T=0.5 K. This trend was investigated by running a single simulation over one million steps. The trend proved to continue and complete ferromagnetism was re-established around nine hundred thousand steps. The second experiment involves the time dependence of a ferromagnet subject to varied temperature. This simulation was run over 10000 Monte Carlo steps, and was averaged over 1000 trials. Fig. 2 shows the results of the simulation run in a single trial. A random ferromagnet behaves as such. The general behavior is shown in Fig. 3. We see from these graphs clear evidence of a decaying magnetization as temperature is increased. The lattice becomes rapidly demagnetized at temperatures as high as T=4.9 K. At T=1.7 K we see barely any change at all. The magnet loses 4

FIG. 2: Ferromagnetic 20x20 Lattice, 1 Trial

FIG. 3: Ferromagnetic 20x20 Lattice, 1000 Trials averaged

some of its magnetization, but stabilizes. We are seeing stabilization at T=1.7 K, some decay at T=2.7 K, and near complete decay at T=3.8 K and T=4.9 K. This agrees with the theoretical decay trend near Tc . Finally we must examine the overall temperature dependence of the system. The simulation is run repeatedly in steps of 0.25 K between T=0.5 K and T=5.0 K. Fig. 4 shows the response of magnetization to changing temperature. It is noted that the magnetization decreases slowly until T=1.5 K, when it begins to fall rapidly. Magnetization levels out around T=4.5 K at a much lower

FIG. 4: 20x20 Lattice Magnetization Response to Temperature

value m < 0.1. This is the relation predicted in Fig. 3. The trend begins to fall at T0 . From this we observe that: Tc 2.269 K. There is some error here, as the graph does not give completely conclusive evidence about Tc . From the calculated results, we have found that for J=K B , for either initial condition, temperatures below 2.0 K will lead to a ferromagnetic state in the lattice, while temperatures above T=4.5 will produce a completely demagnetized state. In the temperatures between we see uctuation in the magnetization, with an exponential decay in response to temperature increase. Our

simulation does have some room for error, as there are likely small energy contributions that are not accounted for in the theory. However, this trend is very close to what is predicted by theory, and for our purposes, these results are an accurate sampling of the magnetized square lattice.

1 2 3

Chapter 8: Statistical Mechanics, Phase Transitions and the Ising Model. Richard Fitzpatrick. The Ising Model. University of Texas, 2006. Peter N. Saeta. The Metropolis Algorithm. Harvey Mudd College, 2011.

Vous aimerez peut-être aussi