Scribd
Upload
179 views2 pages

TransferMatrix 1D-Ising

1) The document discusses using the transfer matrix method to solve the 1D Ising model in the presence of an external magnetic field. 2) It defines the transfer matrix P as a 2x2 matrix with elements given by the Boltzmann factors. 3) It shows that the partition function can be written as the trace of the transfer matrix raised to the power N, which depends only on the eigenvalues of P. 4) The eigenvalues are calculated, allowing the free energy and magnetization to be expressed in terms of them, demonstrating that the 1D Ising model does not exhibit a phase transition.

Uploaded by

Vigneshwaran Kannan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
179 views2 pages

TransferMatrix 1D-Ising

1) The document discusses using the transfer matrix method to solve the 1D Ising model in the presence of an external magnetic field. 2) It defines the transfer matrix P as a 2x2 matrix with elements given by the Boltzmann factors. 3) It shows that the partition function can be written as the trace of the transfer matrix raised to the power N, which depends only on the eigenvalues of P. 4) The eigenvalues are calculated, allowing the free energy and magnetization to be expressed in terms of them, demonstrating that the 1D Ising model does not exhibit a phase transition.

Uploaded by

Vigneshwaran Kannan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Physics 335

Spring (12) 2011

Transfer matrix solution to the 1D Ising model


The most popular approach to solving the 2D Ising model is via the so called transfer matrix
method. We can get some idea of how this method works by using it to solve the 1D model. In
particular we can use this technique to solve the 1D Ising model in the presence of an external
magnetic field, something not possible with our previous direct summation approach. This solution
will allow us to rigorously demonstrate that the 1D model does not exhibit a phase transition (i.e.,
no spontaneous magnetization).
Consider an N-site 1D Ising model with nearest neighbor ferromagnetic coupling J and periodic
boundary conditions (i.e., i+N=i) in an external magnetic field B. The Hamiltonian for this model
can be written as follows:
N

H = J si si+1 B si = {Jsi si+1 + B(si + si+1 ) /2} .


i=1

i=1

i=1

The partition function is:

Z(N,T) = e H
s1

sN

= exp {Jsi si+1 + B(si + si+1 ) /2}


i=1

s1
sN
N

=
s1

sN

exp({Js s

i i+1

+ B(si + si+1 ) /2})

i=1
N

=
s1

sN

si si +1

i=1

where the four possible Boltzmann factors Pmn are simply given by:

P11 = e (J +B ) ,

P11 = P11 = e J , and P11 = e (J B ) .

P11 P11
Treating the Pmn as matrix elements, we define the "transfer matrix" P =
.
P11 P11

Before going on, convince yourself that the partition function really can be written in terms of these
"matrix elements" as shown above.

Now recall from linear algebra that the square of a matrix P can be defined in terms of matrix
elements as follows: (P2)ij = k Pik Pkj. Similarly, the cube of P is defined by matrix elements:
(P3)ij = k l Pik Pkl Plj. With this in mind, we rewrite the above partition function as

Z(N,T) = Ps1 s2 Ps2 s3 Ps3 s4 PsN 1 sN PsN s1 = (P N ) s1 s1 = Tr(P N )


s1

sN

s1

where Tr( ) denotes the trace. Notice that PN is just a 22 matrix. Okay, make sure you agree with
the above result since it's the key to our solution.
From here we will let two theorems of linear algebra do all the work for us. The first theorem is
that the trace of a matrix is independent of the representation of the matrix and is equal to the sum
of the matrix eigenvalues (which, of course, give the diagonal representation of the matrix). The
second theorem we need, which may not be quite as familiar, is that if 1 and 2 are the eigenvalues
of the matrix P, then 1N and 2N are the eigenvalues of the matrix PN. The implications of these
two theorems are that all we need to do is find the eigenvalues of the transfer matrix P and we're
done(!) since the partition function is now simply:
Z(N,T) = 1N + N2 .
The eigenvalues are given, as usual, by the vanishing of the secular determinant:

|P I|= 0.

Leaving the algebra as an exercise for the reader we arrive at the result:
1/ 2

= e J cosh(B) (sinh 2 (B) + e 4 J )

We can achieve further simplification by considering the free energy, which is given by

F = ln Z(N,T)
= ln( N+ + N )
= N ln( + ) ln[1+ ( / + ) N ].

Since <+ , the second term in the above expression becomes negligible with respect to the first
for large N. Thus we are completely justified in writing the partition function simply in terms of

the larger of the two eigenvalues as:

1/ 2 N

Z(N,T) = N+ = e NJ cosh(B) + (sinh 2 (B) + e 4 J )

You should verify for yourself that in the zero-field case (B=0) this expression reduces to our
previously derived result. Well,
that's about it. As I said above, this result can be used to prove that
the 1D Ising model does NOT undergo a phase transition. To demonstrate this fact we must
compute the magnetization given by m = (1/N) i si. Referring back to our first expressions for the
Hamiltonian and partition function, notice that m can cleverly be written as a derivative of ln Z(N,T)
with respect to B:
1 ln Z(N,T)
m=
.
N (B)
In order to demonstrate the absence of a phase transition we must show that m0 for B=0, T>0 is
not possible ... but I think I'll leave this as another exercise for the reader (or perhaps as a final exam

question).

You might also like