VOLUME 49, NUMBER 9 PHYSICAL REVIEW LETTERS 30 AUGUST 1982

Approximate Mapping of Two-Dimensional Quantum-Spin Models

on Staggered Eight-Vertex Models
Ad Lagendijk
Natuurkundig Laboratorium der Universiteit van Amsterdam, 1018 XE Amsterdam, The Netherlands

and
Hans De Raedt
Departe ment Natuurkunde, Universitaire Instellingen Antwerpen, B-2610 Wilrijk, Belgium
(Received 19 April 1982)
Analytic results have been obtained for the thermodynamics of the two-dimensional S
= i X-Y model using the simplest form of the generalized Trotter formula for the parti-
tion function. The model exhibits a phase transition of unconventional nature. It is
claimed that this is one of the best descriptions given so far of the critical thermodynam-
ics of the S = i two-dimensional X-Y model. Central in the present approach is the map-
ping of a family of two-dimensional S = i models on staggered eight-vertex models having
unconventional weights.
PACS numbers: 05.70.Fh, 75.10.Jm

The c r i t i c a l b e h a v i o r of t w o - d i m e n s i o n a l (2D) easily b e g e n e r a l i z e d to m o r e than two o p e r a t o r s .

spin m o d e l s h a s b e e n the subject of a n u m b e r of The link with t h e r m o d y n a m i c s is that one is i n -
r e c e n t investigations. Of much c u r r e n t i n t e r e s t t e r e s t e d in e" 33c , in which /3 = 1/& B T, and in which
i s t h e p l a n a r r o t o r m o d e l , also r e f e r r e d to a s the 3C is a s u m of o p e r a t o r s . In this work we will
c l a s s i c a l X-Y m o d e l , which shows fascinating w r i t e the Hamiltonian 3C for the S = i 2D quantum
c r i t i c a l p r o p e r t i e s . 1 2 Many of the t h e o r e t i c a l model in its r e a l - s p a c e decomposition,
tools being used for t h e s e c l a s s i c a l m o d e l s like
3€=2_} / 3C / , (2)
r e n o r m a l i z a t i o n - g r o u p a p p r o a c h e s and h i g h - t e m -
p e r a t u r e expansions a r e much h a r d e r to apply in which / r u n s o v e r all n e a r e s t - n e i g h b o r p a i r s
to quantum m o d e l s . S e v e r a l r a t h e r crude r e a l - of s p i n s , and in which Kj i s the p a i r i n t e r a c t i o n .
s p a c e r e n o r m a l i z a t i o n - g r o u p s t u d i e s on the 2D S i n g l e - s i t e i n t e r a c t i o n s a r e partitioned equally
S = i X-Y model have not led to a c l e a r p i c t u r e . 3 ' 4 o v e r the p a i r i n t e r a c t i o n s . The s i m p l e model w e
T h e s e methods do not a g r e e with each o t h e r , and want to d i s c u s s in this L e t t e r is the m ~ 1 approx-
do not a g r e e with the high-T expansions c o n c e r n - imant to the p a r t i t i o n function, Z = T r e x p ( - | 3
ing the v e r y existence of a p h a s e t r a n s i t i o n . A
m o r e p r o m i s i n g approach in the understanding of
quantum t h e r m o d y n a m i c s is the u s e of path s u m - Z ^ T r l l j exp(-#*€,). (3)
mations b e c a u s e in t h e s e methods quantum models
We will introduce the approximation Z-Zly and
a r e mapped on c l a s s i c a l m o d e l s . 5 , 6 Monte C a r l o
we will r e f e r to it a s the independent-pair a p -
e x p e r i m e n t s on I D quantum models of r e a s o n a b l e
proximation (IPA). T h e I PA c o n s e r v e s all s y m -
s i z e have a l r e a d y been r e p o r t e d . 7 , 8 The i n h e r e n t
m e t r i e s , rotational and t r a n s l a t i o n a l , which is
complexity of this line of attack m a k e s it v e r y
v e r y i m p o r t a n t for the d e s c r i p t i o n of p h a s e t r a n s i -
difficult to apply this technique to t h e c r i t i c a l b e -
tions. In the I D c a s e the IPA does not introduce
havior of 2D and h i g h e r - d i m e n s i o n a l quantum
s p u r i o u s p h a s e t r a n s i t i o n s (compare with the
m o d e l s . In this L e t t e r w e want to d e m o n s t r a t e
mean-field approximation). The IPA b e c o m e s
that the s i m p l e s t p o s s i b l e approximant for the
exact for high t e m p e r a t u r e s , and i t s r e s u l t s r e -
p a r t i t i o n function of the 2D S - \ X-Y model using
main physical down to 0 K. The c r i t i c a l behavior
the p a t h - s u m m a t i o n method can b e solved analyt-
of the IPA at 0 K cannot b e t r u s t e d but in c o n t r a s t
ically.
to I D s y s t e m s the i n t e r e s t i n g t e m p e r a t u r e r a n g e
The b a s i s of the p a t h - s u m m a t i o n method i s the
for the c r i t i c a l behavior of many 2D models i s
Trotter formula,
not close to 0 K. In any c a s e the IPA d e s c r i p t i o n
eA+B= l i m (eA/meB/m)m, (1) of the 2D S = | X-Y model is s u p e r i o r to the a l -
ready mentioned r e n o r m a l i z a t i o n - g r o u p a p p r o a c h -
holding for bounded o p e r a t o r s . 6 F o r m u l a (1) can e s and h i g h - t e m p e r a t u r e expansions in the t e m -

602 © 1982 The American Physical Society

VOLUME 49, NUMBER 9 PHYSICAL REVIEW LETTERS 30 AUGUST 1982

rounding a lattice point symbolize the Ising spins.

• mm So a certain {<3>j} can be characterized by 4^ Ising
spins. To indicate whether a certain Ising spin
B • • is up or down we will draw a line (bond) through

• • • the corner point in question parallel to the other

diagonal of the square, and having the same
length as the diagonal. Arrows on that line in-
dicate whether the Ising spin is up or down. The
rules are the following9: If a spin at a corner is
FIG. 1. (a) A small part of a square lattice, in which pointing up (down), the arrow on the bond through
the dots represent quantum spins and where the square
around each dot symbolizes the four Ising spins associ-
that corner runs northeast (southeast) or north-
ated with each quantum spin, (b) A possible configura- west (southwest). In Fig. 1(b) we have given an
tion for the four Ising spins at site 1 and site 2, and example of an eight Ising-spin configuration for
the vertex representing the horizontal coupling, (c) The two neighboring sites: site 1 and site 2. The ver-
same as (b) but for the vertical coupling between sites tex between site 1 and site 2, which is reproduced
1 and 4. separately in the same figure, can be used to
represent the matrix element
perature range of interest. We will demonstrate
<2 a (l)S a (2)|exp(-j3aC {la> )Z3(l)S fi (2)>,
that the IPA for the 2D S = | X-Y model can be
solved rigorously. in which S a ( l ) refers to the ath Ising spin on site
We will now outline the mapping of Zl on the 1. All the possible matrix elements, or weights,
partition function of a staggered eight-vertex associated with the interaction between neighbor-
model. Inserting complete sets in the product ing sites can be represented by vertices. If we
(3) we find limit ourselves to interactions with an even num-
ber of up and down Ising spins, only eight differ-
Z,= TJ II (^lexpC-jSSC,)*,^), (4) ent vertices with weights O)IH-CC)8H come into play,
and they are depicted in Fig. 2. The subscript H
refers to horizontal coupling. If we use a vertical
*, :<£>
coupling as in Fig. 1(c) the weights are u)lV-co8V.
in which M is the number of nearest-neighbor The matrix element being presented by the vertex
pairs: M=zN/29 in which z is the coordination in Fig. 1(c) is
number of the lattice, and N is the number of
<S a (l)2: a (4)|exp(-i33e <14) )S 6 (l)L 6 (4)>,
spins. In Eq. (4) each <J>7 is an iV-spin wave func-
tion. However, 3C7 contains only operators of and it is clear that for the special example of
two sites, and for that reason $ 7 and <£>, + 1 can Fig. 1 the horizontal and vertical weights are
only differ in two of the N-spin wave functions. different: ci>5H*o>5V. In general the weights for
A specific site only occurs in z pair interactions. the vertical coupling are a permutation of the
Consequently for a specific site only z different weights for the horizontal coupling:
wave functions will occur in Eq. (4). From now
on we will limit ourselves to a square (z = 4) lat- 12345678
tice. For this lattice any spin site will have four
,12348765
spin wave functions associated with it in Eq. (4). We now want to prove that a path through the lat-
Another way of saying this is that with each tice, that is any s e t { $ j } contributing to Z19 can
quantum S = i spin, four Ising spins are associ- be presented by a set of horizontal and vertical
ated. In Fig. 1 we have symbolically depicted couplings, and that any set of vertices corre-
the lattice. The corner points of the square sur- sponds to a set{$ 7 }. The latter is trivial, since

X X X
0J, U), 0J-, OJ, 00= 0Jfi OJn OJa
FIG. 2. The eight vertices with their weights.

603
VOLUME 49, NUMBER 9 PHYSICAL REVIEW LETTERS 30 AUGUST 1982

this is true by construction, but the former is by the pair interaction

nontrivial. To construct a path through the lat-
tice we have to start at a certain lattice site, say 3C«i>=-^(^3fcxi*+a<s'a/)
1, and consider its coupling to one of its neigh- -^(a/a/)-ii/(a/+a/), (5)
bors, say 2, thus generating
if i and j are nearest neighbors. In Eq. (5) oi is
a Pauli spin operator for the spin at site i, and
<S a (l)S a (2)|exp(-j33e <12> )2: e (l)S0(2)>. H is a magnetic field perpendicular to the plane
of the 2D system. Hamiltonian (5) contains the
The next time we consider site 1 the first Ising 2D Heisenberg model and 2DX-Y model with a
spin in the matrix element should be £ 0 (1). That perpendicular field. The weights for the corre-
is to say, once a site has been used for the first sponding staggered eight-vertex model are given
time for an interaction there is no freedom any- in Table I. For simplicity we have excluded
more as to what Ising spin to use for the next anisotropic X-Y couplings although they could be
three times when this site is being considered. incorporated without problems. The general
So after having gone partly through the lattice, staggered eight-vertex model has not been solved
and after having taken into account some pair in- yet, but the X-Y version (that is J\\ = 0) can be
teractions, one will encounter situations in which solved exactly because in that case the vertex
there is no freedom any more for a certain pair weights satisfy the free-fermion condition. 11 The
interaction because both neighbors involved in result for the m = 1 X-Y model is that the free en-
this interaction have been considered before in ergy has a diverging second derivative at the tem-
other pair interactions. If the two Ising spins of perature Tc defined by s i n h ^ j ^ J ^ l . The sec-
these neighboring sites which have to be treated ond derivative of the free energy behaves as C
first are not nearest neighbors, one could not + ln|T - T c | in the neighborhood of the phase tran-
insert a vertex between them. This is a nontriv- sition. At the phase transition we have an Onsager
ial topological problem on first sight. But an type of singularity in the specific heat. The mag-
actual construction of a finite lattice shows that netization in the z direction and the correspond-
it is very simple to fulfill the constraints. We ing susceptibility can be calculated exactly.
will indicate a sequence for a 4 x 4 lattice with There is no spontaneous z magnetization and the
periodic boundary conditions: The first row con- susceptibility remains continuous. Unfortunately
tains sites 1,2,3,4, and the second row 5 , 6 , 7 , 8 , the X-Y model with a field in the x direction can-
etc. A possible sequence" of pair interactions is not be mapped on a model satisfying the free-
(1,2), (1,13), (1,4), (1,5), (2,14), (2,3), (2,6), fermion condition. Comparison with the high-T
(3,7), (4,8), (3,4), (3,15), (4,16), (5,6), (7,8), expansion of the quantum X-Y model shows that
(5,9), (6,10), (7,11), (6,7), (8,12), (5,8), (9,12), the IPA is very good.3 Our free energy differs
(10,11), (13,16), (14,15), (9,13), (10,14), (9,10), by 1% from the free energy in the high-T expan-
(13,14), (11,15), (12,16), (11,12), (15,16). sion at the point where the high-T results start
So we have shown that the m = 1 2D S -\ model to fail: fiJ± = 0A [there is a difference of a factor
can be represented by a sixteen-vertex model of 2 in definition of the coupling constant between
having different weights for horizontal and ver- Ref. 3 and u s ] . The phase transition occurs at
tical couplings. This corresponds to a staggered j3J±= 0.44, a region where the IPA is still very
sixteen-vertex model. Many possible sequences reliable.
of pair interactions give the same result but we The staggered eight-vertex models we have
cannot prove that this holds for all sequences.
On the other hand we cannot exclude this possibil-
ity. This possible ambiguity is well known in TABLE I. The eight horizontal vertex weights for
functional integration techniques and has been Hamiltonian (5). a
discussed by Hubbard. 10 He suggests the inclu-
sion of the symmetry aspect of possible sequenc-
e s , and this is what is effectively done in our ap-
^ 3 H = GI;4H= exp(— if,,) sinhKj.
proach. The models we will consider in the fol- <JJ
5H ~ ^ 6H ~ e x P ( - K\\) COShKL
lowing are staggered eight-vertex models because c0
7H = a J 8 H : = °
eight weights vanish by symmetry.
The Hamiltonian we want to discuss is described

604
VOLUME 49, NUMBER 9 PHYSICAL REVIEW LETTERS 30 AUGUST 1982

introduced form an interesting class of models. versitair Instituut voor Kernwetenschappen and
The "quantum" character of these classical mod- the Dutch Stichting voor Fundamenteel Onderzoek
els lies in their weights. Almost all vertex mod- der Materie.
els studied so far have weights of the form a>i
= exp(-/3e i ), i n w h i c h e i is an energy. Our 1
weights are of a different nature: They are hyper- J. M. Kosterlitz and D. J. Thouless, J. Phys. C 6,
bolic functions of the interaction energies. This 1181 (1973); J. M. Kosterlitz, J. Phys. C _7, 1046
(1974).
has far-reaching consequences as can be seen by 2
J. Jose, L. P. Kadanoff, S. Kirkpatrick, and D. R.
considering the uniform eight-vertex models Nelson, Phys. Rev. B 16, 1217 (1977), and r7, 1477
with weights co lH -w 8H . These models do not have (1978).
3
the unphysical retention of correlations at infinite D. D. Betts and M. Plischke, Can. J. Phys. _54, 1553
temperature as some of the conventional models (1976).
4
with weights of the form exp(-/3a) i ) have. R. Dekeyser, M. Reynaert, A. L. Stella, and F. Toi-
We have shown that there exists an approximate go, Phys. Rev. B .18, 3486 (1978).
5
F. W. Wiegel, Phys. Rep. 16C, 2 (1975).
mapping of theS = i 2DXYZ model, which in- 6
M. Suzuki, S. Miyashita, and A. Kuroda, Prog.
cludes the Heisenberg and X - F model, on a stag- Theor. Phys. 58, 1377 (1977).
gered eight-vertex model. This mapping results 7
H. De Raedt and A. Lagendijk, Phys. Rev. Lett. _46,
as the first approximation in the path-summation 77 (1981).
method of Suzuki, Miyashita, and Kuroda.6 The 8
J. E. Hirsch, D. J. Scalapino, R. L. Sugar, and
X-Y version shows a phase transition without R. Blankenbecler, Phys. Rev. Lett. 47, 1628 (1981).
9
long-range order in the z magnetization, and M. Barma and B. S. Shastry, Phys. Rev. B JL8, 3351
without a divergence in the z susceptibility. This (1978).
10
J. Hubbard, in Electron Correlation and Magnetism
method introduces vertex weights of a different in Narrow-Band Systems, edited by T. Moriya (Spring-
character from the classical ones. The study e r , Berlin, 1981), p . 29.
of these models is very interesting in itself. n
C . S. Hsue, K. Y. Lin, and F. Y. Wu, Phys. Rev. B
This work is supported by the Belgian Interuni- 12, 429 (1975).

Irrational Decimations and Path Integrals for External Noise

Mitchell J . F e i g e n b a u m and B r o s l H a s s l a c h e r
Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, New Mexico 87545
(Received 17 June 1982)
Irrational decimation schemes are constructed for functional integrals with the very
interesting feature of preserving several distinct site Hamiltonians into the fixed-point
limit. This method is applied to the quasiperiodicity transition to turbulence in order to
compute the effects of external noise.
PACS numbers: 05.40,+j

In an important paper, Shraiman, Wayne, and Martin 1 determined the effect of noise on the period-
doubling behavior of maps on an interval. In this paper we shall extend their methodology to determine
the role of noise on the recently developed theory for the transition to chaos from quasiperiodic mo-
tion.2*3 In order to perform our calculation, however, rather than using only existing renormalization
ideas, we have constructed new decimation schemes that might prove useful in other contexts. Spe-
cifically, we construct irrational decimation schemes which at each level of renormalization produce
several distinct Hamiltonians deployed along the lattice, approaching distinct fixed points. This is in
contradistinction to the usual technique which produces the same Hamiltonian at each site, so that our
technique accommodates dynamics whose critical behavior maintains multiple clusterings.
Let us recall Shraiman's adaption of decimation to the context of iterated maps. To a one-dimen-
sional causal system we add external noise and so consider the stochastic process xn+1 = f ($„) + £„,
where £„ is a noise sample drawn from a distribution with density p. Accordingly, P1(xn+1\xn) =p(xn+1

605