The Ising Model in One Dimension

Ben Sauerwine
Department of Physics, Carnegie Mellon University
Pittsburgh, PA 15213
(Dated: January 25, 2010)

I. INTRODUCTION

Here I consider a one-dimensional chain of spins, either spin-up or spin-down, interacting with their nearest neighbors
and an external field. I take the generalized Hamiltonian for the linear Ising chain of N spins to be:
N
X −1 N
X
ĤN = − Ji,i+1 σi σi+1 − mHi σi (1)
i=1 i=1

Above, Ji,i+1 indicates the strength of the interaction between the i and i + 1-th spins, σi ∈ {−1, 1}. m indicates
the magnetization of the particles and Hi indicates the strength of the external field at the i-th spin.

II. THE PARTITION FUNCTION FROM TRANSFER MATRICES

I begin by writing the partition function in terms of transfer matrices in the most general case, which I will later
apply to specific special cases. In the canonical ensemble, the partition function is given by:
X
Z= e−βEs (2)
s

In equation 2, the sum is over each state s and utilizes the energy of the state Es . In that spirit, I may then
write Ẑ2 , the canonical partition function for N = 2 spins. Referring to σ = 1 as an up or + spin and σ = −1 as a
down or − spin, I may then write for s ∈ {(++), (+−), (−+), (−−)}, where the first and second elements in the pairs
respectively indicate the spins of particle 1 and 2:

Ẑ2 = e−βE++ + e−βE+− + e−βE−+ + e−βE−− (3)

= eβ(J12 +mH2 ) eβmH1 + eβ(−J12 −mH2 ) eβmH1 + eβ(−J12 +mH2 ) e−βmH1 + eβ(J12 −mH2 ) e−βmH1 (4)

Notice in equation 4 that the terms split neatly into those involving only particle 1 and those involving particle 2
and the interactions with particle 1. I now introduce the transfer matrix M (i, j) which includes the effects of particle
i on the states of particle j. In particular,

 
(++) (−+)
M (2, 1) =   (5)
(+−) (−−)
 
eβ(J12 +mH2 ) eβ(−J12 +mH2 )
=   (6)
eβ(−J12 −mH2 ) eβ(J12 −mH2 )
 2

Clearly, I may then write

   
eβmH1 e−βE++ + e−βE−+
M (2, 1)  =  (7)
e−βmH1 e−βE+− + e−βE−−

Finally, I notice that the sum of the results from equation 7 is just the parition function from equation 3. In order
to retrieve these, I may take:

 †  
βmH1
1 e
Ẑ2 =   M (2, 1)   (8)
1 e−βmH1

This suggests that with a generalized transfer matrix from equation , I may systematically extend my partition
function to any number of particles by means of repeated application of transfer matrices. For instance, I may verify
this for Ẑ3 :

 †  
1 eβmH1
Ẑ3 =   M (3, 2)M (2, 1)  
1 e−βmH1

= e−β(−J12 −J23 −mH1 −mH2 −mH3 ) + e−β(J12 −J23 +mH1 −mH2 −mH3 )

+ e−β(−J12 +J23 +mH1 +mH2 −mH3 ) + e−β(J12 +J23 −mH1 +mH2 −mH3 )

+ e−β(−J12 −J23 +mH1 +mH2 +mH3 ) + e−β(J12 −J23 −mH1 +mH2 +mH3 )

+ e−β(−J12 +J23 −mH1 −mH2 +mH3 ) + e−β(J12 +J23 +mH1 −mH2 +mH3 )

In the case where all J and H are taken to be equal, each transfer matrix M (i, j) becomes the same, namely

 
eβ(J+mH) eβ(−J+mH)
M =  (9)
eβ(−J−mH) eβ(J−mH)

Diagonalizing the matrix M with an eye on applying the operator repeatedly, I get:

 
− 21 e2βJ + 12 e2β(J+Hm) − eβHm 1 + e4βJ Sinh[βHm]2 − 21 e2βJ + 21 e2β(J+Hm) + eβHm
p p
1 + e4βJ Sinh[βHm]2
M=  
1 1
 p 
eβJ Cosh[βHm] − e−2βJ + e2βJ Sinh[βHm]2 0
× p 
0 eβJ Cosh[βHm] + e−2βJ + e2βJ Sinh[βHm]2
 √ 
e−βHm e2βJ Sinh[βHm]+ 1+e4βJ Sinh[βHm] 2
− √ √
2 1+e4βJ Sinh[βHm]2 2 1+e4βJ√
Sinh[βHm]2
×
 
−e2βJ Sinh[βHm]+ 1+e4βJ Sinh[βHm]2

√ e−βHm √
2 1+e4βJ Sinh[βHm]2 2 1+e4βJ Sinh[Bβm]2
 3

In the Jordan normal diagonalization above, the columns of the matrix on the first row are the eigenvectors
corresponding to the eigenvalues along the diagonal of the matrix of the second row. Respectively, I will refer to these
as Ψmin , Ψmax , λmin and λmax .
Now I may easily calculate ẐN using the property of diagonal matrices where M N = (SDS −1 )N = SDN S −1 .

 †  
1 eβmH
ẐN =   M N −1   (10)
1 e−βmH

Taking H = 0, this gives simply:

 †   N −1   
1 −1 1 2Sinh[βJ] 0 − 12 1
2 1
ẐN,H=0 =         = 2(2Cosh[βJ])N −1 (11)
1 1
1 1 1 0 2Cosh[βJ] 2 2 1

To determine free energy per spin scales in the long chain, no external field limit, we may then take:

1 ln2 ln[Cosh[βJ]]
F = lim − ẐN,H=0 = − − (12)
N →∞ βN β β