Physica A 387 (2008) 41484156

www.elsevier.com/locate/physa
Dimers on the kagome lattice I: Finite lattices
F.Y. Wu
a,
, Fa Wang
b,c
a
Department of Physics, Northeastern University, Boston, MA 02115, United States
b
Department of Physics, University of California, Berkeley, CA 94720, United States
c
Material Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, United States
Received 5 February 2008
Available online 25 February 2008
Abstract
We report exact results on the enumeration of close-packed dimers on a nite kagome lattice with general asymmetric dimer
weights under periodic and cylindrical boundary conditions. For symmetric dimer weights, the resulting dimer generating functions
reduce to very simple expressions, and we show how the simple expressions can be obtained from the consideration of a spin-
variable mapping.
c 2008 Elsevier B.V. All rights reserved.
PACS: 05.50.+q; 04.20.Jb; 02.10.Ox
Keywords: Finite kagome lattice; Close-packed dimers; Spin-variable mapping
1. Introduction
Acentral problemin lattice statistics is the enumeration of close-packed dimers on lattices and graphs. The origin of
the problem has a long history dating back to a 1937 paper by Fowler and Rushbrooke [1] in an attempt of enumerating
the absorption of diatomic molecules on a surface. A breakthrough in dimer statistics has been the exact solution of
the generating function for a nite square lattice of size M N, where M and N are arbitrary, obtained by Kasteleyn
[2] and by Temperley and Fisher [3] in 1961.
In view of the role of nite-size solutions in the conformal eld theory discovered by Bl ote et al. [4] in 1986, it
has been of increasing importance to consider solutions of lattice models for various nite two-dimensional lattices.
Thus, the dimer solution has been extended to cylindrical [5] and nonorientable [6] lattices. However, these lattices
are variants of the square lattice which may not necessarily exhibit special lattice-dependence features.
In a recent paper [7] we have reported enumeration results of close-packed dimers on an innite kagome lattice with
symmetric dimer weights (activities). The solution turned out to assume a very simple expression. In this paper we
extend the solution to nite lattices with general asymmetric weights and under two different boundary conditions. We
nd the solutions given by entirely different expressions. For symmetric weights, however, the solutions again reduce
to simple expressions. We show how the simple expressions can be deduced quite directly from the consideration of a
spin-variable mapping.

Corresponding author.
E-mail address: fywu@neu.edu (F.Y. Wu).
0378-4371/$ - see front matter c 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.physa.2008.02.054
F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156 4149
Fig. 1. (a) An M N lattice of 6MN sites with asymmetric dimer weights. (b) A Kasteleyn edge orientation of the kagome lattice. A unit cell is
the region bounded by broken lines.
2. Finite lattices
Consider a kagome lattice with asymmetric dimer weights x, y, z around up-pointing triangles and x

, y

, z

around
down-pointing triangles as shown in Fig. 1(a). We consider a lattice of MN unit cells having a total of 6MN lattice
sites; the case of M = 1, N = 3 is shown in Fig. 1(a). A unit cell of the lattice contains 6 sites numbered 1, 2, . . . , 6
as indicated in Fig. 1(b). The Kasteleyn edge orientations adopted in Ref. [7] are also shown in Fig. 1(b).
The method of Pfafans [2] is concerned with the evaluation of a 6MN 6MN antisymmetric Kasteleyn matrix
A written down according to edge weights and orientations (under specic boundary conditions) which can be read
off from Fig. 1(a) and (b), and by adopting the prescription
A
i j
=
_
+w
i j
orientation from i to j
w
i j
orientation from j to i
(1)
where w
i j
is the weight of edge i j . The dimer generating function is then given by the square root of the determinant
of the matrix A. We consider the kagome lattice under two different boundary conditions.
2.1. The periodic boundary condition
First we consider the periodic (toroidal) boundary condition (PBC) for which the lattice is periodic in both the
horizontal and vertical directions. Kasteleyn [2] has shown that under the PBC the dimer generating function Z
PBC
is
a linear combination of four Pfafans Pf| A
i
|, i = 1, 2, 3, 4,
Z
PBC
=
1
2
[Pf| A
1
| +Pf| A
2
| +Pf| A
3
| +Pf| A
4
|]. (2)
Up to signs yet to be determined, the Pfafans are the square root of the determinants specied by the Kasteleyn
orientation of lattice edges with, or without, the reversal of arrows on edges connecting two opposite boundaries. A
perusal of Fig. 1(b) and the use of the prescription (1) lead to the four 6MN 6MN Kasteleyn matrices,
A
1
= a
0,0
I
M
I
N
+a
1,0
I
M
T
N
a
T
1,0
I
M
T
T
N
+a
0,1
T
M
I
N
a
T
0,1
T
M
I
T
N
+a
1,1
T
M
T
N
a
T
1,1
T
T
M
T
T
N
A
2
= a
0,0
I
M
I
N
+a
1,0
I
M
H
N
a
T
1,0
I
M
H
T
N
+a
0,1
T
M
I
N
a
T
0,1
T
M
I
T
N
+a
1,1
T
M
H
N
a
T
1,1
T
T
M
H
T
N
A
3
= a
0,0
I
M
I
N
+a
1,0
I
M
T
N
a
T
1,0
I
M
T
T
N
+a
0,1
H
M
I
N
a
T
0,1
H
M
I
T
N
+a
1,1
H
M
T
N
a
T
1,1
H
T
M
T
T
N
A
4
= a
0,0
I
M
I
N
+a
1,0
I
M
H
N
a
T
1,0
I
M
H
T
N
+a
0,1
H
M
I
N
a
T
0,1
H
M
I
T
N
+a
1,1
H
M
H
N
a
T
1,1
H
T
M
H
T
N
.
(3)
4150 F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156
Here, the superscripts T denote transpose, is the direct product, I
M
is the M M identity matrix, and H
N
, T
N
are
the N N matrices
H
N
=
_
_
_
_
_
_
0 1 0 0
0 0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 1
1 0 0 0
_
_
_
_
_
_
, T
N
=
_
_
_
_
_
_
0 1 0 0
0 0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 1
1 0 0 0
_
_
_
_
_
_
, F
N
=
_
_
_
_
_
_
0 1 0 0
0 0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 1
0 0 0 0
_
_
_
_
_
_
, (4)
where F
N
is to be used later in (15), and
a
0,0
=
_
_
_
_
_
_
_
0 z

0 0 0
z

0 x

0 0 0
y x

0 y 0 0
0 0 y 0 z

0 0 0 z

0 x

0 0 0 y

0
_
_
_
_
_
_
_
, a
1,0
=
_
_
_
_
_
_
_
0 0 0 0 0 0
0 0 x z 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 x
0 0 0 0 0 0
_
_
_
_
_
_
_
,
a
0,1
=
_
_
_
_
_
_
_
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
y 0 0 0 0 0
_
_
_
_
_
_
_
, a
1,1
=
_
_
_
_
_
_
_
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
z 0 0 0 0 0
0 0 0 0 0 0
_
_
_
_
_
_
_
, (5)
a
1,0
= a
T
1,0
, a
0,1
= a
T
0,1
, a
1,1
= a
T
1,1
.
The determinant of a matrix is equal to the product of its eigenvalues. To determine eigenvalues of A
1,2,3,4
,
we rst block-diagonalize the 4 matrices by appropriate Fourier transforms. Since T
N
and T
T
N
commute, they can
be simultaneously diagonalized and replaced by respective eigenvalues e
i
n
and e
i
n
, where
n
= 2n/N, n =
0, . . . , N 1.
Similarly, T
M
and T
T
M
can be simultaneously diagonalized and replaced by eigenvalues e
i
m
and e
i
m
, where

m
= 2m/M, m = 0, . . . , M 1.
Likewise, H
N
and H
T
N
commute and they can be simultaneously diagonalized and replaced by eigenvalues e
i
n
and
e
i
n
, where
n
= (2n +1)/N, n = 0, . . . , N 1; H
M
and H
T
M
can be simultaneously diagonalized and replaced
by eigenvalues e
i
m
and e
i
m
, where
m
= (2m +1)/M, m = 0, . . . , M 1. Then we nd
det | A
1
| =
N1

n=0
M1

m=0
det | A(
n
,
m
)|, det | A
2
| =
N1

n=0
M1

m=0
det | A(
n
,
m
)|,
det | A
3
| =
N1

n=0
M1

m=0
det | A(
n
,
m
)|, det | A
4
| =
N1

n=0
M1

m=0
det | A(
n
,
m
)|,
(6)
where the 6 6 matrix A(, ) is anti-Hermitian and is given by
A(, ) = a
0,0
+a
1,0
e
i
+a
1,0
e
i
+a
0,1
e
i
+a
0,1
e
i
+a
1,1
e
i(+)
+a
1,1
e
i(+)
=
_
_
_
_
_
_
_
_
0 z

0 ze
i(+)
ye
i
z

0 x

+ xe
i
ze
i
0 0
y x

xe
i
0 y 0 0
0 ze
i
y 0 z

ze
i(+)
0 0 z

0 x

+ xe
i
ye
i
0 0 y

xe
i
0
_
_
_
_
_
_
_
_
. (7)
F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156 4151
This yields
det A(, ) = 2A +2D cos() +2E cos(2 +) +4
2
sin
2

with
A = (xy

z + x

yz

)
2
+(x

z + xyz

)
2
D = (xy

z x

yz

)
2
E = (x

z xyz

)
2

2
= (x

yz xy

)
2
.
(8)
The desired generating function is now obtained by substituting either Pf| A
i
| = +

det | A
i
| or

det | A
i
| into (2).
In the present case the signs in front of the square roots can be determined by considering the case of M = N = 1.
By explicit enumeration we have
Z
PBC
= 2(x + x

)(yz

+ y

z), M = N = 1. (9)
It is readily veried that the expression (9) is reproduced by (2) if all 4 terms in (2) are positive. Thus, we are led to
the nal expression
Z
PBC
=
1
2
_
_
det | A
1
| +
_
det | A
2
| +
_
det | A
3
| +
_
det | A
4
|
_
, (10)
where det | A
i
|, i = 1, . . . , 4, are given by (6).
For symmetric dimer weights x

= x, y

= y, z

= z, we have D = E =
2
= 0, det | A
1,2,3,4
| = (4xyz)
MN
, and
the simple result
Z
PBC
= 2 (4xyz)
MN
, x

= x, y

= y, z

= z. (11)
We shall see in Section 3 that this simple result can be understood and deduced directly using a spin-variable mapping.
In the case of an innite lattice, (2) leads to the per-dimer free energy
f = lim
M,N
1
3MN
ln Z
PBC
=
1
24
2
_
2
0
d
_
2
0
d ln
_
2A +2D cos( ) +2E cos( +) +4
2
sin
2

_
. (12)
This free energy is independent of the boundary condition. The free energy (12) can also be deduced using the vertex-
model approach introduced in Ref. [7], the details of which are straightforward and will not be given.
For symmetric dimer weights x

= x, y

= y, z

= z, (12) reduces further to

f =
1
3
ln(4xyz). (13)
This result for an innite lattice was rst reported in Ref. [8] with the full derivation given in Ref. [7]. The exact
per-dimer entropy s =
2
3
ln 2 obtained from (13) at x = y = z = 1 has been cited earlier by Phares and Wunderlich
from Ref. [9] and by Elser [10] from different considerations.
2.2. The cylindrical boundary condition
Consider next the cylindrical boundary condition (CBC) for which the lattice of M N unit cells is periodic in
the horizontal direction. The Kasteleyn orientation is achieved by reversing the orientations of the 4M 1 edges
connecting unit cells in the Nth column to those in the rst column. This gives the dimer generating function as a
single Pfafan
Z
CBC
=
_
det | A
CBC
|, (14)
4152 F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156
where A
CBC
is the 6MN 6MN matrix
A
CBC
= a
0,0
I
M
I
N
+a
1,0
I
M
H
N
a
T
1,0
I
M
H
T
N
+a
0,1
F
M
I
N
a
T
0,1
F
M
I
T
N
+a
1,1
F
M
H
N
a
T
1,1
F
T
M
H
T
N
. (15)
Here, matrices a are those in (5) and F
N
has been given in (4). Again, the matrix A
CBC
is block-diagonalized by
replacing H
N
and H
T
N
by their respective eigenvalues. This leads to
det | A
CBC
| =
N1

n=0
det |B
M
(
n
)|,
n
= (2n +1)/N (16)
where B
M
() is the 6M 6M Kasteleyn matrix
B
M
() = B I
M
+ B
+
F
M
+ B

F
T
M
=
_
_
_
_
_
_
_
_
_
_
B B
+
0 0 0 0
B

B B
+
0 0 0
0 B

B 0 0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 B B
+
0
0 0 0 B

B B
+
0 0 0 0 B

B
_
_
_
_
_
_
_
_
_
_
, (17)
B = a
0,0
+e
i
a
1,0
e
i
a
T
1,0
, B
+
= a
0,1
+e
i
a
1,1
, and B

= a
T
0,1
e
i
a
T
1,1
are the 6 6 matrices
B =
_
_
_
_
_
_
_
_
0 z

0 0 0
z

0 x

+ xe
i
ze
i
0 0
y

xe
i
0 y 0 0
0 ze
i
y 0 z

0 0 0 z

0 x

+ xe
i
0 0 0 y

xe
i
0
_
_
_
_
_
_
_
_
, (18)
B
+
=
_
_
_
_
_
_
_
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
ze
i
0 0 0 0 0
y 0 0 0 0 0
_
_
_
_
_
_
_
, B

=
_
_
_
_
_
_
_
0 0 0 0 ze
i
y
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
_
_
_
_
_
_
_
.
The matrix B
M
() is of a form of that occurring in the evaluation of an Ising partition function under the cylindrical
boundary condition [5,6], and the determinant det |B
M
()| can be evaluated as follows:
Let B
i, j
denote the 6 6 matrix B with row i and column j removed. By Laplacian expansion and the use of a
lemma established in Ref. [6], the determinant of B
M
can be expanded as
det |B
M
| = det |B| det |B
M1
| + z
2
det |B
5,5
| det |B
1,1
M1
|
+y
2
det |B
6,6
| det |B
1,1
M1
| + yze
i
det |B
5,6
| det |B
1,1
M1
| + yze
i
det |B
6,5
| det |B
1,1
M1
|. (19)
Similarly, the determinant of the matrix B
1,1
M
can be expanded as
det |B
1,1
M
| = det |B
1,1
| det |B
M1
| + z
2
det |B
1,1;5,5
| det |B
1,1
M1
| + y
2
det |B
1,1;6,6
| det |B
1,1
M1
|
+yze
i
det |B
1,1;5,6
| det |B
1,1
M1
| + yze
i
det |B
1,1;6,5
| det |B
1,1
M1
|, (20)
where B
i, j ;k,
is the 6 6 matrix B with rows i and k, and columns j and deleted.
F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156 4153
Write B
M
det |B
M
| and C
M
det |B
1,1
M
|. Expansions (19) and (20) are recursion relations of B and C,
B
M
= a B
M1
+b C
M1
C
M
= c B
M1
+d C
M1
(21)
where
a = det |B|
b = z
2
det |B
5,5
| + y
2
det |B
6,6
| + yze
i
det |B
5,6
| + yze
i
det |B
6,5
|
c = det |B
1,1
|
d = z
2
det |B
1,1;5,5
| + y
2
det |B
1,1;6,6
| + yze
i
det |B
1,1;5,6
| + yze
i
det |B
1,1;6,5
|,
(22)
subject to the initial condition B
0
= 1, C
0
= 0. Explicitly using (18) and (22) gives
a = (x
2
+ x

2
)(y

2
z
2
+ y
2
z

2
) +4x
2
y

2
z

2
2(x

z + xyz

)(xy

z + x

yz

) cos
+4xy

(x

yz xy

) cos
2

b = 2i(x

yz + xy

)(y

2
z
2
+ y
2
z

2
) sin 4iyy

zz

(x

yz xy

) cos sin
c = 2i
_
(x
2
+ x

2
)(x

yz + xy

) 2xx

(x

yz xy

)
_
cos sin
d = (x
2
+ x

2
)(y

2
z
2
+ y
2
z

2
) +4x

2
y
2
z
2
+2(x

z + xyz

)(xy

z + x

yz

) cos
4x

yz(x

yz xy

) cos
2
.
(23)
The recursion relation (21) can be solved by introducing generating functions
B(t ) =

M=0
B
M
t
M
, C(t ) =

M=0
C
M
t
M
. (24)
The recursion relation (21) gives
B(t ) = 1 +t [a B(t ) +b C(t )]
C(t ) = t [c B(t ) +d C(t )], (25)
where we have made use of the initial condition B
0
= 1, C
0
= 0.
Solving (25) for B(t ) and C(t ), we obtain
B(t ) =
1 d t
1 (a +d)t +(a d b c)t
2
=
1 d t
(1
+
t )(1

t )
, (26)
where

= (a +d)/2
_
(a d)
2
/4 +bc
= A +2
2
sin
2

_
(A +2
2
sin
2
)
2
D
2
E
2
2DE cos
2
. (27)
Here, A, D, E,
2
have been given in (8). Partial fraction and expand the right-hand side of (26), and compare the
resulting expansion with (24), one obtains
B
M
() =

M+1
+

M+1

M
+

M

d. (28)
Finally, by combining (14) and (16), we obtain the desired generating function
Z
CBC
=

_
N1

n=0
B
M
(
n
),
n
= (2n +1)/N.
4154 F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156
Fig. 2. (a) An M = 3 kagome lattice. Sublattice 1 sites are denoted by open circles. (b) The spin conguration deduced from a typical dimer
conguration. The lattice is decomposed into 4 strips or loops depending on the boundary condition (see the text).
For symmetric weights x

= x, y

= y, z

= z we have
a = 8x
2
y
2
z
2
(1 cos ), b = 8ixy
3
z
3
sin
c = 8ix
3
yz sin , d = 8x
2
y
2
z
2
(1 +cos )

+
= 16x
2
y
2
z
2
,

= 0.
We obtain B
M
() = (16x
2
y
2
z
2
)
M
sin
2
(/2) and hence
Z
CBC
= (4xyz)
MN
N1

n=0
sin(
n
/2) = 2
1N
(4xyz)
MN
. (29)
Again, we shall see in Section 3 that this simple expression can be understood using a spin-variable mapping.
In the case of an innite lattice and since
+
>

, we have from (28) B

M
()
M
+
. This gives the per-dimer
free energy
f = lim
M,N
1
3MN
ln Z
CBC
= lim
N
1
6N
N1

n=0
ln
+
(
n
)
=
1
12
_
2
0
ln
+
() d (30)
where
+
() is given by (27). It is readily veried that (30) is identical to the free energy (12) after carrying out the
integration over in (12).
3. A spin-variable mapping
The exact enumeration results (11) and (29) for a lattice of M N unit cells with symmetric dimer weights
are strikingly simple, suggesting the possibility of a simple derivation. Indeed, Zeng and Elser [11] and Misguich
et al. [12] have introduced a pseudo-spin consideration of enumerating quantum states which can be transcribed to
the present x = y = z = 1 case [13]. However, the pseudo-spin consideration was presented in contexts of spin 1/2
antiferromagnets and quantum dimer models, and the application to the classical dimer problem with general weights
x, y, z is not immediately obvious. A simpler formulation is very much needed.
We elucidate the matter by describing an alternate spin-variable mapping valid for general x, y, z. First we note
that the numbers of x, y, and z dimers are always xed for nite lattices. This is due to the fact that the three principal
axes do not intersect at common points. The kagome lattice has three sublattices as numbered in Fig. 2(a). Denote the
number of sites on sublattice i by N
i
, i = 1, 2, 3, and the number of x dimers by N
x
, etc. Then as a consequence of
the fact the principal axes do not intersect at common points, we have the relations
N
y
+ N
z
= N
1
, N
z
+ N
x
= N
2
, N
x
+ N
y
= N
3
.
F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156 4155
This gives rise to N
x
= (N
2
+ N
3
N
1
)/2, etc., which are xed numbers. The dimer generating function is therefore
a single monomial of the form
Z = x
N
x
y
N
y
z
N
z
,
so we need only to compute the constant . In the case of N
1
= N
2
= N
3
= N we are considering, this leads to the
expression
Z = (xyz)
N/2
, (31)
where we have N = 2MN for both the PBC and CBC boundary conditions.
We next map dimer congurations on the lattice to spin congurations on one sublattice, say, 1. Consider the
example of the lattice shown in Fig. 2(a). Let the lattice consist of M rows of sublattice 1 sites and M+ 1 rows of
equal 2 and 3 sites with open boundaries in the vertical direction. The boundary condition in the horizontal direction
can be either open or periodic. Denote the number of sublattice 1 sites in the mth row by n
m
, which must satisfy the
sum rule
n
1
+n
2
+ +n
M
= N
1
= even, (32)
since the lattice must admit dimer coverings.
Assign spin variable

mn
= 1, m = 1, . . . , M, n = 1, . . . , n
m
to sublattice 1 sites, where m is the row number counting beginning from the top. Adopt the convention that
= +1(1) if the dimer covering the site also covers a site above (below) the row. For +1 (1) sites we remove
the two edges below (above) the site as well as the horizontal edge directly above (below) it as shown in Fig. 2(b).
This procedure decomposes the lattice into strips (loops) for open (periodic) boundary conditions in the horizontal
direction.
Now every strip (or loop) must have an even number of sites to accommodate one (or 2) dimer covering(s).
This condition imposes constraints on spin congurations that can be realized by this mapping. To best describe
the constraints it is convenient to dene a row variable

m
=
n
m

n=1

mn
, m = 1, 2, . . . , M. (33)
It is then readily veried that we must have

1
= (1)
n
1
,

m1

m
= (1)
n
m
, m = 2, 3, . . . , M,

M
= 1. (34)
The constraint
M
= 1 is automatically satised due to (32) and the fact that

M
= (
1
)(
1

2
) (
M1

M
) = (1)
m
1
++ m
M
= (1)
N
1
= 1. (35)
We remark that the row variable (33) can also be used to analyze the dimer models in higher dimensions considered
in Ref. [14].
We can now compute the constant . Beginning with an overall spin state degeneracy 2
N
1
of sublattice 1 sites,
each constraint in (34) reduces the spin states by a factor of 2. Since there are M1 such constraints, we have
=
_
2
N
1
(M1)
, open boundaries
2
N
1
+2
, horizontal periodic boundary condition.
(36)
Note that there is an extra factor 2
M+1
for periodic boundary conditions in the horizontal direction since each loop
has 2 dimer coverings. Expression (36) is a very general result independent of specic values of n
m
.
4156 F.Y. Wu, F. Wang / Physica A 387 (2008) 41484156
For a lattice of M N unit cells with toroidal boundary conditions PBC considered in Section 2.1, we have
N
1
= N
2
= N
3
= N = 2MN, M = 2M. Hence (36) gives = 2
N
1
2
(M1)
2
M
, where as in (36) the second
factor is due to M 1 constraints with the Mth constraint automatically satised, and the third factor is due to the
2-fold dimer coverings of each loop. This leads to Z
PBC
= 2 (4xyz)
MN
in agreement with (11).
For a lattice of M N unit cells with cylindrical boundary condition CBC considered in Section 2.2, we again
have N
1
= N
2
= N
3
= N = 2MN, M = 2M. However, the Mth row of N sublattice 1 spins must be all +1 (Cf.
Fig. 2(b) with the bottom row of sites removed) reducing the counting by a factor of 2
N
and the number of rows by
1. Hence = 2
N
2
N
1
(M1)
2
M
and Z
CBC
= 2
1N
(4xyz)
MN
in agreement with (29).
We remark that our spin-variable mapping is akin to the one used recently by Dhar and Chandra [14]. However,
the DharChandra approach focuses on an innite lattice by ignoring what happens on the boundary. Here, we treat
the boundary effect rigorously and apply the mapping to nite lattices.
Acknowledgements
We are grateful to D. Dhar for sending us a copy of Ref. [14] and to G. Misguich for calling our attention to Refs.
[1113]. The work by FW is supported in part by grant LBNL DOE-504108.
References
[1] R.H. Fowler, G.S. Rushbrooke, Trans. Faraday Soc. 33 (1937) 1272.
[2] P.W. Kasteleyn, Physica (Amsterdam) 27 (1961) 1209.
[3] H.N.V. Temperley, M.E. Fisher, Philos. Mag. 6 (1961) 1061;
M.E. Fisher, Phys. Rev. 124 (1961) 1664.
[4] H.W.J. Bl ote, J. Cardy, M.P. Nightingale, Phys. Rev. Lett. 56 (1986) 742.
[5] B. McCoy, T.T. Wu, The Two Dimensional Ising Model, Harvard University Press, 1973.
[6] W.T. Lu, F.Y. Wu, Phys. Lett. A 258 (1998) 157;
W.T. Lu, F.Y. Wu, Phys. Lett. A 293 (2002) 235.
[7] F. Wang, F.Y. Wu, Phys. Rev. E 75 (2007) 040105(R).
[8] F.Y. Wu, Internat. J. Modern Phys. B 20 (2006) 5357.
[9] A.J. Phares, F.J. Wunderlich, Nuovo Cimento Soc. Ital. Fis. B 101 (1988) 653.
[10] V. Elser, Phys. Rev. Lett. 62 (1989) 2405.
[11] C. Zeng, V. Elser, Phys. Rev. B 51 (1995) 8318.
[12] G. Misguich, D. Serban, V. Pasquier, Phys. Rev. B 67 (2003) 214413.
[13] G. Misguich, Private communication.
[14] D. Dhar, S. Chandra. arXiv:0711.0971.