Eur. Phys. J. B 18, 275–282 (2000)
THE EUROPEAN
PHYSICAL JOURNAL B
EDP Sciences
c Società Italiana di Fisica
Springer-Verlag 2000
Phase transitions in the mixed Ashkin-Teller model
S. Bekhechi, A. Benyoussef a , A. El kenz, B. Ettaki, and M. Loulidi
Laboratoire de Magnétisme et de Physique des Hautes Énergies, Département de Physique, BP 1014, Faculté des Sciences,
Rabat, Morocco
Received 5 July 1999 and Received in final form 4 July 2000
Abstract. By using a mean-field approximation (MFA) and Monte-Carlo (MC) simulations, we have studied the effect on the phase diagrams of mixed spins (σ = 1/2 and S = 1) in the Ashkin-Teller model
(ATM) on a hypercubic lattice. By varying the strength describing the four spin interaction and the single
ion potential, we have obtained by these two methods quite rich phase diagrams with several multicritical
points. This model exhibits a new partially ordered phase hSi which does not exist neither in the spin-1/2
ATM nor in the spin-1 ATM. While MFA yields phase diagrams which are sometimes qualitatively incorrect, accurate results are obtained from MC simulations. From the critical exponents which have been
calculated using finite-size scaling ideas, we have shown that all phase transitions are Ising-like except for
the paramagnetic-Baxter critical surface on which the critical exponents vary continuously, by varying only
the strength of the coupling interaction independently of the value of the single ion potential.
PACS. 75.10.Hk Classical spin models
1 Introduction
The Ashkin-Teller model [1] (A.T.M.) is a generalization of the Ising model to four component systems. It
may be considered as a superposition of two Ising models, which are described by variables σi and Si sitting on
each of the sites on a hypercubic lattice. Within each Ising
model, there is a two spin nearest-neighbour interaction
with strength K2 . In addition, the different Ising models are coupled by a four spin interaction with strength
K4 . Different methods have been applied to study the
critical behaviour of this model [2–9]. All these methods
yield three different phases: a paramagnetic (P) phase
in which neither σ nor S nor anything’s else is ordered
(hσi = hSi = hσSi = 0); a Baxter phase in which σ and
S independently order in a ferromagnetic fashion and also
hσSi is unequal to zero; and a third phase called PO1
in which σS is ordered ferromagetically, hσSi =
6 0, but
hσi = hSi = 0. Apart from the variational approaches
which give a tricritical point, the other accurate methods
yield only a line of critical points which connects the Ising
critical point at one end to the four state Potts critical
point at the other end, and along this line the exponents
vary continuously [2,9]. A good physical realization for
this model is the compound of Selenium adsorbed on a Ni
surface. In this context, Per Bak et al. [2] have shown by
using symmetry considerations that the order parameters
σ and S transform as two-dimensional representations of
the 4 pmm [10] symmetry group of the Ni(100) surface,
and the phase transition belongs to the universality class
of the xy model with cubic anisotropy as does the transition in the Ashkin-Teller model. The order parameter
a
e-mail: benyous@fsr.ac.ma
hσSi transforms as a one-dimensional representation of
the Ni(100) symmetry group and so the melting line is of
Ising character.
One of the most interesting and challenging phenomena is the appearance of other new partially ordered
phases in the ATM, such as: (i) the hσi phase defined by
hσi =
6 0 and hSi = hσSi = 0 in the three-dimensional antiferromagnetic ATM [3]. (ii) The hσi and hSi phases which
are connected by a symmetry operation to the hσSi phase
in the bidimensional anisotropic ATM [11–13]. (iii) The
PO2 phase defined by (hσi = hSi =
6 0; hσSi = 0) found
recently in the spin-1 Ashkin Teller model [14,15].
The synthesis of single-chain and double-chain ferrimagnets is now becoming standard, and attempts to
synthesize higher-dimensional polymeric ferrimagnets are
starting to give very encouraging results. Some of the
materials investigated are 2D organometallic ferrimagnets [16], 2D networks of the mixed-metal material
{[P(Ph)4 ][MnCr(ox)3 ]}n where Ph is phenyl and ox is oxalate [17]. The intense activity related to the synthesis of
ferrimagnetic materials requires a parallel effort in theoretical study. Mixed Ising spin systems provide a good
model for studying ferrimagnetism [18,19].
In this paper we are mainly interested in the study
of the isotropic mixed spins (σ = 1/2, S = 1) AshkinTeller model by using MFA and MC simulations. This
model might be thought of either as describing ferrimagnets or adsorption phenomena. The paper is organized
as follow. In the next section, the model is introduced
and the ground state diagram is presented. Section 3 is
devoted to the MFA description and the results are presented. Section 4 contains the formalism of the MC simulation and finite-size scaling theory. Our numerical results
276
The European Physical Journal B
h·iAF indicate the thermal average of spin variables respectively in the ferromagnetic and antiferromagnetic phases.
(ii) For D/K2 < −3z/8: We have two critical values of the single ion potential D, Dc1 /K2 = −z(1 −
K4 /2K2 )/4 and Dc2 /K2 = −z(1 + K4 /4K2 )/2, such
that if D/K2 < Dc1 /K2 and D/K2 > Dc2 /K2 the
Baxter2 and Baxter1 phases are stable respectively. If
Dc1 /K2 < D/K2 < Dc2 /K2 the spins σi are parallel
while the spins Si are equal to zero then we have hσiF 6= 0,
and hSiF = hσSiF = 0 we obtain the phase called “hσi”.
%D[WHU
..
V!
%D[WHU
3 Mean-field approximation
'.
Fig. 1. Ground state phase diagram.
for the phase diagrams and the critical exponents are presented in Section 5. Finally, in Section 6 we conclude.
2 Model and ground state diagram
The mean-field theory represents the infinite dimensional
limit of statistical systems since it neglects correlations
between different spins. However it is interesting to study
the mean-field behaviour of the anisotropic A.T.M. so that
we may effectively bracket the three-dimensional system
between the mean- field and the d = 2 behaviour.
To write the mean-field equations let hσ , hS and hσS
denote the molecular fields associated with the order parameters hσi, hSi and hσSi respectively, and j(i) represents the set of all nearest neighbours to the site i:
The model Hamiltonian used is given by:
X
(σi σj + Si Sj )
H = −K2
hσ =
j(i)
hi,ji
− K4
X
hi,ji
σi σj Si Sj − D
X
Si2 ,
X
hSj(i) iK2 ;
hS =
(1)
where the spins Si = ±1, 0 and σi = ±1/2, are localized on the sites of a hypercubic lattice. The first term
describes the bilinear interactions between the σ and S
spins at sites i and j, with the interaction parameter K2 .
The second term describes the four spin interaction with
strength K4 , and on each site there is a single ion potential
D. All these interactions are restricted to the z nearestneighbours pairs of spins.
In order to calculate the ground state energy, we express the Hamiltonian as a sum of the contributions of the
nearest-neighbours spins. So the contribution of a pair S1 ,
S2 and σ1 , σ2 is:
2D 2
(S1 + S22 ).
z
(2)
By comparing the values of Ep for different configurations
we obtain the following structure of phase diagram shown
in Figure 1:
(i) For D/K2 ≥ −3z/8: if K4 /K2 > −1 the Baxter1 phase
is stable since both spins σi and Si are aligned, otherwise
if K4 /K2 < −1 the spins σi are antiparallel while the spins
Si are parallel then we have: hσiF = hSiAF = hσSiF = 0
and hσiAF 6= 0, hSiF 6= 0 and hσSiAF 6= 0 which characterize the phase called Baxter2. The symbols h·iF and
(3)
j(i)
i
Ep = −K2 (σ1 σ2 + S1 S2 ) − K4 σ1 σ2 S1 S2 −
X
hσj(i) iK2 ;
and
hσS =
X
hσj(i) Sj(i) iK4 .
j(i)
The effective Hamiltonian of the system is:
X
X
X
!
.
(4)
1
Z0 = 2 exp(βhS + βD) cosh (βhσ + βhσS )
2
1
+ 2 exp(−βhS + βD) cosh (βhσ − βhσS )
2
1
+ 2 cosh βhσ
2
(5)
H0 = − hσ
σi + hS
i
i
Si + hσS
σi Si
i
It generates the following partition function:
where β = 1/KBT is the inverse temperature.
The variational principle for the free energy per site is
described by:
F ≤Φ=
−1
ln(Z0 ) + hH − H0 i
β
(6)
S. Bekhechi et al.: Phase transitions in the mixed Ashkin-Teller model
and the order parameters which are the spin averages are
given by:
277
V6!)
mσ = hσi
+ sinh β2 hσ
!
2∆
$
%$
!
(7)
exp β(−hS + D) sinh
2∆
β
2 (hS
− hσS )
,
β
(hS + hσS )
2
β
β
+ exp β(−hS + D) cosh (hS − hσS ) + cosh hσ .
2
2
However the total free energy can be written as:
X
+ K4
ln Z0 + K2
X
X
hi,ji
hσi Si ihσj Sj i.
hσi ihσj i +
X
hi,ji
..
∆ = exp β(hS + D) cosh
i
V!
exp β(hS + D) sinh β2 (hS + hσS )−
%D[WHU
%D[WHU
mσS = hσSi
!
%$
∆
Φ=−
&
$
exp β(hS + D) cosh β2 (hS + hσS )−
exp β(−hS + D) cosh β2 (hS − hσS )
=
&
&
mS = hSi
=
&
3DUD
7.
=
V6!$)
exp β(hS + D) sinh β2 (hS + hσS )+
exp β(−hS + D) sinh β2 (hS − hσS )
hSi ihSj i
(8)
hi,ji
Usually the solutions of equation (7) combined with
equation (3) are not unique. We choose the ones that minimize the free energy (Eq. (8)) and then represent the pure
phases. If the order parameters are continuous, the transition is of the second- order, while if they are discontinuous,
the transition is of first-order.
Our MF results are presented in the plane
(K4 /K2 , T /K2 ) for D/K2 = −2.0, Figure 2. In order
to describe the different entities in the phase diagrams,
we will use the Griffiths notations for the multicritical
points [12,20]. At high temperature, the ordered phases
(Baxter1 and Baxter2) are separated from the paramagnetic phase by the partially ordered phases hσSiF and
hσSiAF at high absolute values of K4 /K2 . It appears also
that lines of first order which are linked by triple points
A3 , join the second order ones by tricritical point C and
multicritical point BA2 at high and low absolute values
of K4 /K2 . At low temperature, the ordered phase hσi
which exists at the ground state is separated from the two
other ordered (Baxter1 and Baxter2) and paramagnetic
phases by first and second order phase transitions respectively. Since the MFA provides a qualitative picture of the
phase diagram, we will present detailed studies using MC
simulations.
Fig. 2. Phase diagram for D/K2 = −2 as obtained from mean
field approximation. Solid and dashed lines denote second and
first order phase transitions respectively. C, A3 and BA2 denote respectively tricritical points, triple points and critical end
points.
4 Monte-Carlo simulations
The system studied is a L × L square lattice with even
values of L, containing N = L2 spins, and we use
the well-known Metropolis algorithm [21] with periodic
boundary conditions to update the lattice configurations.
Monte-Carlo (MC) simulations are performed for d = 2
with systems of sizes L = 8, 16, 24, 32 and 64. We use
95 000 to 700 000 MC steps to calculate the thermodynamic quantities after discarding 5 000–50 000 sweeps for
thermal equilibrium. Most of the phase diagrams presented here are obtained with L = 32 and are compared
with those derived from MFA.
We can evaluate the stationary phase diagram of the
model and its associated critical exponents by using the
finite-size scaling concepts [22] applied to some thermodynamic properties of the system. The physical quantities
of use are the magnetizations |Mα |(α = σ, S, σS), and are
estimated by:
|Mα |=h|Mα |i=
1 XX
αi (c) with α = σ, S, σS, (9)
Np c i
where i runs over the lattice sites, c runs over the configurations obtained to update the lattice over one sweep of
the N spins of the lattice (one Monte-Carlo step, MCS)
counted after the system reaches thermal equilibrium, and
p is the number of the MCS.
In order to measure the phase boundaries we will
find useful the measurement of fluctuations (variance of
the order-parameters) in Mα defined by the magnetic
susceptibility:
χα =
N
hMα2 i − h|Mα |i2
KB T
with
α = σ, S, σS.
(10)
278
The European Physical Journal B
The fourth-order cumulant Uα (α = σ, S, σS) are defined by:
hMα4 i
·
3hMα2 i2
M0 (L
1/ν
ε),
χαL (T ) = Lγ/ν χ0 (L1/ν ε),
UαL (T ) = U0 (L
1/ν
ε),
Uα′ (T ) = L1/ν U0′ (L1/ν ε),
7.
%
%
%$
%D[WHU
(12)
%$
(14)
Fig. 3. Phase diagram for D/K2 = −2 as obtained from MC
simulations, where diamonds and plus denote second and first
order phase transitions respectively. There is a multicritical
point B3 and a critical end point B2 A.
D
V!
6!
V6!
V!
6!
V6!
7.
7.
V!
6!
V6!
5.1 Phase diagrams
0DJQHWL]DWLRQV
5 Results and discussion
E
6XVFHSWLELOLWLHV
A rich variety of phase transitions is observed by varying
the strength of the coupling parameters.
- For D/K2 = −2.0, our MC results are presented in
the plane (K4 /K2 , T /K2 ), Figure 3. At high temperature, the ordered phases (Baxter1 and Baxter2) are separated from the paramagnetic phase by the partially ordered phases hσSiF and hσSiAF at high absolute values of
K4 /K2 . At intermediate values of K4 /K2 , the MC results
present a new partially ordered phase called hSi whereas
in the MF ones this latter phase is absent (see Fig. 2).
This new phase, defined by hσi = hσSi = 0 and hSi =
6 0,
does not exist neither in the spin-1/2 [3] nor in the spin1 [13] AT models. The variation of the order parameters
and their susceptibilities are shown in Figure 4. The MC
results are obtained from the maxima in the susceptibility
and are qualitatively better than those of the MF ones
and all the MC transition lines are of second order linked
by multicritical points B3 and B2 A, along all these lines
..
(15)
so that Uα′ (Tc ) = L1/ν U0′ (0). Then we can find the critical
exponent ν from the log-log plot of UL′ (T c) versus L.
A first-order transition is signalled by hysteresis and
discontinuous jumps in the internal energy (E = hHi)
and/or the order parameter (as will be shown in the
next section). The first and second order transitions may
be distinguished by the buildup of the magnetization
by quenching the system from a disordered state (corresponding to an equilibrium configuration at very high
temperature) to a temperature just below the transition
temperature [21,23]. Due to the presence of long-lived
metastable states, a two-step relaxation process is expected in the case of a first order transition, whereas
a smooth buildup is observed in the case of continuous
transition.
Baxter 1
V!
(13)
where ε = T − Tc .
If we derive equation (14) with respect to the temperature T , we obtain the scaling relation:
%
6!
%
0DJQHWL]DWLRQV
|Mα |L (T ) = L
%
6!
Finally, we will use finite-size-scaling theory [21–23]
to analyse our results. Following this approach, in the
neighbourhood of the infinite critical point Tc , the above
quantities obey for sufficiently large L:
−β/ν
3DUD
V6!$)
(11)
6XVFHSWLELOLWLHV
Uα = 1 −
V6!)
7.
V!
6!
V6!
7.
Fig. 4. Plot of order parameters hσi, hSi and hσSi for D/K2 =
−2.0 from MC simulations, showing the existence of two new
partially ordered phases at high temperature. The corresponding inset shows associated susceptibilities. a) K4 /K2 = 5, existence of the hσSi phase where at T1 = 2.13, hσi = hSi = 0 but
hσSi =
6 0 whereas at T2 = 2.26 we have hσi = hSi = hσSi = 0.
b) K4 /K2 = 2, existence of the hSi phase where at T1′ = 1.29,
hσi = hσSi = 0 but hSi =
6 0 whereas at T2′ = 1.34 we have
hσi = hSi = hσSi = 0.
S. Bekhechi et al.: Phase transitions in the mixed Ashkin-Teller model
V
0 7
3DUD
279
, QFUHD VLQ J 7-
D
'HFUHD VLQ J 7-
%D[WHU
&
%$
, QFUHD VLQ J 7-
E
'HFUHD VLQ J 7-
0 7
7.
V
V!
'.
we didn’t observe any hysteresis in the order parameters
when crossing the boundaries (an example is shown below). Whereas in the MF phase diagram it appears also
lines of first order which are linked by triple points A3
and join the second order ones by tricritical point C and
multicritical point BA2 at high and low absolute values of
K4 /K2 . At low temperature, the ordered phase hσi which
exists at the ground state is separated from the two other
ordered (Baxter1 and Baxter2) and paramagnetic phases
by first and second order critical lines respectively. As we
can see in Figure 3, the hSi phase region is small for this
value of D/K2 . Then in order to enlarge this region we
have fixed the four spin interaction K4 /K2 and varied the
crystal field D/K2 . Also we note that only MC results are
presented, since the MF ones are qualitatively similar with
few differences in the multicritical points and the order of
magnitude in the temperature transition which is higher
as seen in the proceeding paragraph.
- For K4 /K2 = 0 (we do not present the plots here) the
mixed ATM is decoupled into the two well known independent spin-1/2 Ising model and spin-1 Blume-Capel model.
The structure of the phase diagram is obtained by superposing the phase diagrams of these models. At high temperature and for all values of D/K2 > −2, we obtain the
hSi phase which is sandwiched between the Baxter1 and
the paramagnetic phases. Its region of stability is enlarged
for higher values of D/K2 . Whereas for D/K2 < −2, we
have only the hσi and paramagnetic phases separated by
a second order transition. At low temperature the two
ordered phases hσi and Baxter1 are separated by a first
order transition.
- By increasing the strength of the four spin interaction K4 /K2 > 0, the hSi region decreases and appears
only at a high crystal field D/K2 , until a critical value
of K4 /K2 = 4, is reached. Then the hSi phase disappears and we have only a critical line with tricritical point
which separates the Baxter1 phase (hσi = hSi = hσSi)
from the paramagnetic phase, Figure 5. Along the firstorder line strong hysteresis was observed. The tricritical
, QFUHD VLQ J 7-
F
'HFUHD VLQ J 7-
V
0 7
Fig. 5. Phase diagram for K4 /K2 = 4.0 as obtained from MC
simulations. Diamonds and pluses denote second and first order
phase transitions respectively. In the phase diagram, there is a
critical end point BA2 and a tricritical point C denoted by a
square and located at (D/K2 = −3.65 ± 0.01, T /K2 = 1.13 ±
0.01).
7
Fig. 6. Plot of order parameters hσi vs. T /K2 for K4 /K2 = 4.
(a) Typical hysteresis observed when crossing the first order
transition boundary for D/K2 = −3.9, we show also the discontinuity indicative of a first order transition. (b) By increasing the critical field, D/K2 = −3.68, the hysteresis becomes
very small and the behaviour of the order parameter is between
first and second order transitions. At D = −3.3 the hysteresis
behaviour disappears and we have also a continuity in the order
parameter which is characteristic of a second order transition.
point was determined when the hysteresis disappears, Figures 6; it occurs, for L = 32, at (D/K2 = −3.65 ± 0.01,
T /K2 = 1.13 ± 0.01). We note that the location of this
tricritical point is not performed by a finite size scaling.
When the four coupling becomes very strong,
K4 /K2 > 4, Figure 7, the phase diagram exhibits the
other partially ordered phase hσSiF defined by hσi =
hSi = 0 and hσSi =
6 0, instead of the hSi phase. The
former phase which appears at high crystal field is the
same as the known partially ordered phase that occurs in
the usual spin-1/2 ATM, since for large D the Hamiltonian (Eq. (1)) is reduced to the spin-1/2 ATM for which
the hσSi phase is favoured for strong values of the four
spin interactions K4 . By decreasing the strength of the
crystal field D/K2 , the transition between the Baxter 1
and the paramagnetic phases becomes of first order with
a tricritical point C.
5.2 Critical behaviour
In the previous section we have described the general characteristics of different varieties of phase diagrams. Since
it is important to determine the universality class of the
critical boundaries we will focus ourselves to calculate the
critical exponents.
In order to calculate the critical exponents of the
Baxter1-disorder critical line and locate better the critical
280
The European Physical Journal B
V6!
D
/
0 /7
V
3DUD
%
7.
&
/
/
; /7
/
V
E
%D[WHU
/
/
%$
/
/
V!
/
/
'.
/
F
/
/
V
8 /7
Fig. 7. Phase diagram for K4 /K2 = 5.0 as obtained from
MC simulations. Diamonds and pluses denote second and first
order phase transitions respectively. B3 , C and BA2 denote
respectively multicritical point, tricritical point and critical end
point.
/
/
7
temperature Tc of the model for K4 /K2 = 4 and D/K2 =
−2 we plot in Figures 8 the magnetization, the susceptibility and the cumulant defined by equations (10–12) as
a function of temperature T , for several values of L. The
scaling relation for the fourth-order cumulant shows that,
at the critical temperature, all curves of Uα (T ) must intercept themselves at Tc for whatever value of L. From
the latter figure we estimate the value of Tc as being
1.83 ± 0.01. With Tc determined we could now evaluate
the critical exponents of the model. In Figure 9, we exhibit, at the critical temperature Tc , the log-log plot of
the staggered magnetization, Mα (Tc ), the susceptibility
χα (T ) and the derivative of the cumulant Uα′ (Tc ) versus L.
From the slope of the straight line, which is the best fit
to the data points, and using equations (13, 14, 15), we
can obtain the value of the stationary critical ratios β/ν,
γ/ν and 1/ν which are associated respectively to Mα (Tc ),
χα (T ) and Uα′ (Tc ). They are given in the table below.
We can also estimate the ratio γ/ν by a log-log plot of
the maximum value of the susceptibility versus L that
is also scaled as Lγ/ν (Fig. 9b). The value we obtain is
γ/ν = 1.745 ± 0.002. By the same way the critical temperature for D/K2 = 0.5 was located at Tc = 2.34 ± 0.01
and the critical exponents (see Tab. 1) are given from the
finite size Scaling analysis.
At the tricritical point, with Tc determined (Tc /K2 =
1.13 ± 0.01) our best fits give νt = 0.556 ± 0.013,
Figure 10a.
Another way to find the critical exponent ν is to use
the location of the susceptibility peak as the finite lattice
critical temperature Tc (L), then Tc (L) − Tc = L−1/ν . The
results obtained (Fig. 10b), νt = 0.55 ± 0.013, are consistent with the previous ones. All our results show that for
K4 /K2 = 4, the critical line belongs to the Ising critical
and tricritical universality classes (β/ν = 1/8, γ/ν = 7/4,
ν = 1 [24,25], and νt = 5/9 [26,27]).
Fig. 8. Plots of the temperature variation T /K2 for K4 /K2 =
4.0 and D/K2 = −2, for various choices of L, of: (a) the
fourth-order cumulant. The critical temperature Tc is determined when all curves of Uα (T ) intercept themselves. We estimate the value of Tc as being 1.83 ± 0.01. (b) Location of
the susceptibility peak as the finite lattice critical temperature
Tc (L) = 1.85 ± 0.01 : Tc (L) → Tc when L increase. (c) The
order parameters hσi (Similar behaviours are obtained for hSi
and hσSi).
For K4 /K2 = 0.25, D/K2 = −2, the model
(Eq. (1)) becomes equivalent to the four state Potts
model as confirmed in Figure 11 from the estimate of
ν, ν = 0.664 ± 0.01, which is close to the four-state Potts
critical exponents ν = 2/3 [28]. By increasing K4 /K2
the critical exponent ν varies until it reaches the value of
the critical exponent Ising model ν = 1 for K4 /K2 → ∞ .
We note that the critical exponent ν depends only on the
value of K4 /K2 . It remains independent of D/K2 which
just extends the critical exponents from points (in the case
of spin 1/2 ATM) to lines with tricritical exponents in
addition as shown in reference [15] (spin-1 ATM). We conclude that the critical exponents vary continuously on the
Baxter- paramagnetic critical surface in the same manner
as in the spin 1/2 ATM (Ref. [9]) independently of the
ratio D/K2 .
6 Conclusion
In this paper, we have shown by using MFA and MC simulations that the mixed ATM presents as well as the partially ordered phase hσSi another new partially ordered
phase hSi which does not exist neither in the spin-1/2
ATM nor in spin-1 A.T.M. The MC simulations give rich
S. Bekhechi et al.: Phase transitions in the mixed Ashkin-Teller model
281
Table 1. The critical exponents for K4 /K2 = 4.
Critical exponents:
D/K2 = −2
D/K2 = 0.5
Exact results [24, 25]
β/ν
0.124 ± 0.003
0.126 ± 0.003
1/8
γ/ν
1.765 ± 0.008
1.719 ± 0.04
7/4
ν
0.994 ± 0.007
0.992 ± 0.02
1
D
Q
V
6ORSH
OQ8 7 F
OQ8
V
7F
D
6ORSH
Q
W
E
6ORSH
JQ
JQ
7
F
/
E
6ORSH Q
W
OQ 7
V
F
F
V
OQ0 7 F
/ 7
7
F
6ORSH
V
OQ;
V
6ORSH EQ
OQ/
Fig. 9. Finite-size dependence of critical behaviour for
K4 /K2 = 4.0 and D/K2 = −2 in log-log plots: (a) U ′ (σ)(Tc )
vs. L. The straight line is the best fit to the data points which
gives: ν = 1.021 ± 0.002. (b) The susceptibility χ(σ)(Tc ) at
T = Tc vs. L. The straight line is the best fit to the data points
which gives: γ/ν = 1.765 ± 0.008. The susceptibility χ(σ)(TL )
at its maximum vs. L. The straight line is the best fit to the
data points which gives: γ/ν = 1.745 ± 0.002. (c) The magnetization M (σ)(Tc ) vs. L. From the slope of the straight, which is
the best fit to the data points, we obtain β/ν = 0.124 ± 0.003.
OQ/
Fig. 10. At the critical point Tc for K4 /K2 = 4.0 and D/K2 =
−2 a finite-size dependence in log-log plots: (a) U ′ (Tc ) vs. L.
The straight line is the best fit to the data points which gives:
νt = 0.556 ± 0.013. (b) (Tc (L) − Tc ) vs. L. The straight line is
the best fit to the data points which gives: νt = 0.55 ± 0.013.
6ORSH Q
OQ 7 F / 7 F
phase diagrams with second and first order phase transitions which are more accurate and qualitatively better than those obtained by MFA. In the parameter space
(K4 /K2 , D/K2 , T /K2 ) the phase diagram presents rich
varieties of phase transitions with surfaces of first and
second order phase transitions which are bounded by
lines of tricritical, triple and multicritical points. We have
shown that all second order phase transitions are Isinglike while at the paramagnetic-Baxter phase transition
the critical exponents vary continuously, by varying the
strength of K4 /K2 between the Ising and 4-state Potts
critical exponents independently of the ratio D/K2 .
OQ/
Fig. 11. (Tc (L)−Tc ) vs. L for K4 /K2 = 0.25 and D/K2 = −2.
The straight line is the best fit to the data points which gives:
ν = 0.664 ± 0.01.
282
The European Physical Journal B
This work was supported by the PARS grant N◦ : Physique
035. One of the authors, (S.B.) thanks the Abdus Salam International Center For Theoretical Physics (ICTP) of Trieste for
financial support.
References
1. J. Ashkin, E. Teller, Phys. Rev. 64, 178 (1943).
2. P. Bak, P. Kleban, W.N. Unertel, J. Ochab, G. Akinci,
N.C. Barlet, T.L. Einstein, Phys. Rev. Lett. 54, 1539
(1985).
3. R.V. Ditzian, J.R. Banavar, G.S. Grest, L.P. Kadanoff,
Phys. Rev. B 22, 2542 (1980).
4. R.V. Ditzian, Phys. Lett. A 38, 451 (1972).
5. F.J. Wegner, J. Phys. C 5, L131 (1972).
6. J.R. Banavar, D. Jasnow, D.P. Landau, Phys. Rev. B 20,
3820 (1979).
7. H.J.F. Knops, J. Phys. A 8, 1508 (1975).
8. J.A. Plascak, F.C. Sà Barreto, J. Phys. A 19, 2195 (1986).
9. S. Wiesman, E. Domany, Phys Rev. E 48, 4080 (1993); S.
Wiesman, E. Domany, Phys. Rev. E 51, 3074 (1994).
10. D. Mukamel, E. Domany, M.E. Fisher, Phys. Rev. Lett.
37, 565 (1976).
11. F.Y. Wu, K.Y. Lin, J. Phys. C 7, L181 (1974).
12. .S. Bekhechi, A. Benyoussef, A. El kenz, B. Ettaki, M.
Loulidi, Physica A 264, 503 (1999).
13. M. Badehdah, S. Bekhechi, A. Benyoussef, M. Touzani,
Physica B 291, 394 (2000).
14. M. Loulidi, Phys. Rev. B 55, 11611 (1997).
15. M. Badehdah, S. Bekhechi, A. Benyoussef, B. Ettaki, Phys.
Rev. B 59, 6052 (1999).
16. H. Tamaki, Z.J. Zhong, N. Matsumoto, S. Kida, M.
Koikawa, N. Achiwa, Y. Hashimoto, H. Okawa, J. Am.
Chem. Soc. 114, 6974 (1992).
17. S. Decurtins, H.W. Schmalle, H.R. Oswald, A. Linden, J.
Ensling, P. Gütlich, A. Hauser, Inorg. Chim. Acta. 65, 216
(1994).
18. G. Ming, C. Zhang Yang, Phys. Rev. B 48, 9452 (1993).
19. G.M. Buendia, M.A. Novothy, J. Phys. Cond. Matter. 9,
5951 (1997).
20. R.B. Griffiths, Phys. Rev. B 12, 345 (1975).
21. K. Binder, in Applications of the Monte-Carlo Method in
statistical physics, edited by K. Binder (Springer-Verlag,
Berlin, 1988).
22. V. Privman, Finite-Size Scaling and Numerical Simulation
of Statistical Systems (World Scientific, Singapore, 1990).
23. D.P. Landau, K. Binder, Phys. Rev. B 17, 2328 (1978);
J.D. Kimel, S. Black, P. Carter, Y.L. Wang, Phys. Rev. B
35, 3347 (1987); S. Bekhechi, A. Benyoussef, Phys. Rev.
B 56, 13954 (1997).
24. H.E. Stanley, Introduction to phase Transitions and Critical Phenomena (Oxford University Press, 1971).
25. B.M. McCoy, T.T. Wu., The Two dimensional Ising Model
(Harvard University Press, Cambridge, MA, 1973).
26. M.P.M. den Nijs, J. Phys. A 12, 1857, (1979).
27. B. Nienhuis, E.K. Riedel, M. Schick, J. Phys. A 13, L 189,
(1980).
28. B. Nienhuis, in Phase Transition and Critical Phenomena,
edited by C. Domb, L. Lebowitz (Academic Press, New
York, 1987).