Renormalization Group Theory
The goal of this section is to introduce several concepts of Renormalization
Group Theory and to illustrate such concepts with the 1-dimensional
Ising model.
Consider the task of computing the canonical partition function
of the one-dimensional Ising model in the absence of an external magnetic
field. According to Eq. (286),
 |
(317) |
where coupling parameter
and N is the total number of spins. Note that according to Eq. (317),
 |
(318) |
The renormalization group strategy for the 1-dimensional Ising model
can be described as follows.
Step (1). Sum over the even numbered spins in Eq. (317). Note
that summing, e.g., over
we obtain
![$\displaystyle Z(K, N) = \sum_{S_1,S_3,S_4,S_5,S_6,S_7...} [e^{K (S_1+ S_3)}+e^{-K (S_1+ S_3)} ] e^{K S_3 S_4} e^{K S_4 S_5} e^{K S_5 S_6} e^{K S_6 S_7} ...,$](img681.png) |
(319) |
summing over
and
we obtain,
![$\displaystyle Z(K, N) = \sum_{S_1,S_3,S_5,S_6,S_7...} [e^{K(S_1+ S_3)}+e^{-K(S_1+ S_3)}] [e^{K(S_3+ S_5)} + e^{-K(S_3+ S_5)}] e^{K S_5 S_6} e^{K S_6 S_7} ...,$](img683.png) |
(320) |
and summing over all even numbered spins we obtain
![$\displaystyle Z(K, N) = \sum_{S_1,S_3,S_5,S_7...} [e^{K(S_1+ S_3)}+e^{-K(S_1+ S...... [e^{K(S_3+ S_5)} + e^{-K(S_3+ S_5)}] [e^{K(S_5+ S_7)} + e^{-K(S_7+ S_9)}]... .$](img684.png) |
(321) |
Step (2). Rewrite the remaining sum (i.e., the sum over odd numbered
spins introduced by Eq. (321)) by implementing the Kandanoff transformation
 |
(322) |
where both
and
are functions of
.
Substituting Eq. (322) into Eq. (321) we obtain
 |
(323) |
Note that such transformation allow us to rewrite the partition function
in terms of a renormalized partition function
(i.e., a partition function with new parameters that describes an Ising
model with half the number of spins and a different coupling parameter
).
In order to determine the renormalization group equations (i.e.,
and
as a function of K) and show that
,
we note that when
,
Eq. (322) gives
 |
(324) |
and when
,
Eq. (322) gives
 |
(325) |
Therefore, solving for
in Eq. (325) and substituting into Eq. (324) we obtain
and substituting Eq. (326) into Eq. (325) we obtain
cosh |
(327) |
Eqs. (326) and (327) are called renormalization group equations
since they provide the renormalization scheme.
Step (3). Go to (1), replacing
by
.
Step (3) is repeated each time on the subsequent (renormalized) partition
function (i.e.,
,
,
,
,
... etc.) until the renormalized parameters become approximately constant
(i.e., until the renormalized parameters reach a fixed point and
become invariant under the Kadanoff transformation). Note that, according
to Eq. (326),
,
etc., so after a few iterations the coupling parameter becomes negligibly
small and the partition function can be approximated by using Eq. (318)
as follows:
![\begin{displaymath}\begin{array}{ll}\text{ln} Z(K, N) &\approx \frac{N}{2} \tex......cosh}^{1/2} (2K^{V}) ] + \frac{N}{2^6} \text{ln} 2.\end{array}\end{displaymath}](img706.png) |
|
|
(328) |
The renormalization group strategy thus involve computing the total
sum, introduced by Eq. (317), step by step. The success of the approach
relies on the fact that after a few iterations the sum converges to an
expression that can be easily computed.
Sometimes the partition function is known for a specific value of the
coupling parameter (e.g., for
in the 1-dimensional Ising model). The renormalization group theory can
then be implemented to compute the partition function of the system for
a different value
of the coupling constant. This is accomplished by inverting Eq. (326) as
follows:
and computing
from
according to Eq. (323).
One could also define the function
as follows
ln |
(330) |
and substituting Eq. (329) into Eq. (323) we obtain
Therefore, given the partition function
for a system with coupling constant
,
one can compute
and K according to Eqs. (330) and (329), respectively. The partition function
exp
is then obtained by substituting the values of
and K in Eq. (330). Note that according to this procedure,
and the subsequent iterations give larger and larger values of
.
This indicates that the flow of
has only two fixed points at K
(e.g., at infinite temperature) and K
(e.g., at 0 K).
Systems with phase transitions, however, have nontrivial fixed points
at intermediate values of
.
For instance, following a similar procedure, as the one described in this
section, it is possible to show that the 2-dimensional Ising model has
an additional fixed point
and that the heat capacity
divergers at
.
Thus,
determines the critical temperature where the system undergoes a phase
transition and spontaneosly magnetizes.