PHYSICAL REVIEW B VOLUME 12, NUMBER 5 1 SEPTEMBFR 197$

Bicritical and tricritical phenomena in uniaxial ferroma~ets

M. R. H. Khajehpour, ~ Yung-Li Wang, ~ and Robert A, Kromhout*
Department of Physics, The Florida State University, Tallahassee, Florida 32306
(Received 24 March 1975)
The phase diagrams of a spin-1 uniaxial ferromagnetic model with both exchange anisotropy and single-ion
anisotropy (D) are studied in the mean-field approximation. In the absence of an external magnetic field a
bicritical point (BCP) is found in the (D, T) plane for the transverse (K) to parallel (J) coupling ratio
K/J) 0.462. As the exchange anisotropy is decreased the BCP moves towards the temperature axis. In the
limit of the isotropic exchange the BCP is located on the T axis. In the presence of a field along the parallel
(transverse) direction two symmetric lines of tricritical points (TCP) are generated in the (D, H, (H, ), T)
space. These lines meet at the BCP. For a significantly wide range of D/J and K/J values TCP prevails in
the (H, (H, ), 7) plane. Application of a transverse (parallel) field generates the wing coexistence surfaces.
The whole thermodynamic space is experimentally accessible. Outside the above ranges, one may still have a
TCP in the (H&, T) plane, but not in the (H„T) phase diagram. Some special features of the (H„T)phase
diagram are also discussed.

I. INTRODUCTION two order parameters; e. g. , the parallel and the

transverse (in flop phase) sublattice magnetizations
The study of the systems exhibiting multicritical in the uniaxial antiferromagnets in a field, the po-
phenomena has recently attracted much attention. larization and the magnetization in ferromagnetic-
While some types of multicritical behavior have ferroelectric, ' the crystalline and superfluid (diag-
been known to exist for many years the recent in- onal and off-diagonal long-range orders) in He, 8
tense interest in the subject reflects two great ad- and two types of magnetic ordering in the magnetic
vancements: a fairly basic and thorough understand- mixed crystals. ~'7 Theories based on the mean-
ing of ordinary critical phenomena through the fieM approximation ' and the phenomenological
hypothesis of thermodynamic scaling and the re- Landau-type expansions ' ' have been applied to
normalization-group approach, and the develop- the above systems. Liu and Fisher and Bruce and
ment of high-precision techniques and ingenious de- Aharony have given a succinct Landau-type theory
of the bi- and tetracritical points. A similar theory
signs for measurements, which make the critical
region amenable to experimental scrutiny. The for tricritical systems is developed by Baush. "
In
study of the multicritical phenomena also allows us this rather general and qualitatively correct con-
to examine the validity of the existing concepts of text, bi-, tri-, and tetracritical points are defined
homogeneity and universality or the possibility of as the points where two, three, and four critical
their extension. lines meet, respectively. While the coexistence of
By now quite a few physical systems are known two "pure" phases (defined by the iwo order param-
to exhibit tricriticality. 2 The existence of bicritical eters) along a line (flop line) is necessary to have
points (BCP) in the uniaxially anisotropic antifer- a BCP, at temperatures below a tetracritical point
romagnets in auniform magnetic fields and in the fer- an intermediate or "mixed" phase lies between the
romagnetic-fer roe lectric systems4 has been known tyro "pure" phases. Another approach~2 indicates
f or some time, although the implicationsof the bicrit- that this picture may not be complete if the full
ical singularities have notbeen appreciated until quite thermodynamic field space is considered.
recently. 5 Tetracritical phenomena are newcomers Fisher and Nelson' have postulated a scaling
on the scene. Their existence has been reported in hypothesis for BCP of a uniaxially anisotropic anti-
certain magnetic alloyss and in mixed crystals. 7 ferromagnet in a field. A similar homogeneity
A multicritical point is characterized by the hypothesis was applied to a model by Chang et al.
existence of several phase transitions (which may and Ha, rbus et aE. "
The renormalization-group
be of different natures) in its vicinity. Near a tri- technique' applied to a classical n-vector model of
critical point the interplay is between first-order a uniaxial antiferromagnet in a field, confirms the
and second-order phase transitions, and a single scaling hypothesis, and leads to an important con-
order parameter is sufficient to describe both tran- sequence": for n~ n*(d) [n*(d) ) 3 for d= 3] the bi-
sitions. In systems exhibiting bi- and tetracritical critical exponents are the same as those of a fully
phenomena we have more than one kind of ordered isotropic Heisenberg system.
state. The phase transitions are from ordered The application of the scaling hypothesis near a
states to paramagnetic phase as well as between the BCP has iwo important consequences': (i) additional
ordered phases. Therefore the behavior of these scaling laws hold among the bicritical exponents,
systems depends upon the competition between the and (ii) the critical lines and the flop line have a

1849
1850 KHA JEHPOUR, WANG, AND KROMHOUT 12
common tangent at BCP. ion anisotropy
In the models discussed in Refs. 5, 10, and 13,
the ordering fields are either physically fictitious R= QZ„S,(i)S,(j) QK„S,(i) S,(j) ~

&Q&
fields or experimentally uncontrollable. There-
fore one cannot investigate the vicinity of the BCP
in all directions experimentally. The features in
+~ Q [S,(i)1' — 4 g S(i),
. (2 1)

the vicinity of a BCP are of great interest, For where 4 and K are assumed positive. The trans-
example, it is shown- by Chang et al. ' that four verse coupling provides the possibility for the sys-
tricritical lines meet at a BCP in-a generalized tem to order in the basal plane perpendicular to the
metamagnetic model, The ordering fieMs to gen- z axis for certairi ranges of temperature and an-
erate these lines in all the above models are the isotropy. In the mean-field approximation (MFA)
staggered fields. the Hamiltonian (2. 1) can be written
In this paper we discuss an anisotropic uniazial
ferromagnetic model which exhibits, in addition to X„=-2(S,)S, -~2K(S, )S, +D(S,)
the ordinary critical phenomena, bicriticality at
—h S+(S,) yK(Si),
~
(2. 2)
H= 0 and tricriticality in a finite field. The Hamil-
tonian contains a single-ion term of the form where K=K(0)/&(0), h= pH/J(0), D=&/J(0), ~(0)
.
D(S,) The exchange interaction is also assumed =g& J,&,K(0) =g& K, &, and energies are measured
to be anisotropie. A magnetic field along either in units of &(0).
the parallel or the transverse direction generates It should be noted that while the basal plane is
two tricritieal lines which terminate at the BCP.
isotropic we have chosen a direction for the order-
Consequently, in a region near the BCP one expects
ing and have labeled it with the perpendicular sign;
a variety of critical (either Ising-like or xy-like), (S,) is the component of (8) on the basal plane.
bicritical, and tricritical behaviors, as well as the One can write Eq. (2. 2) in an even simpler form
crossovers between these behaviors.
Another interesting feature of the model is that X„=D(Sg)3 n, S, —n—~S, + eo, (2. 3)
when the anisotropy D is treated as a constant,
(H, , H, = 0, T) and (H, = 0, H„T) phase diagrams where
exhibit a tricritical point (TCP) for a wide range of n = h, +2(S, ),
the values of the parameters of the system. Here
one has a situation where the TCP and the wings n~= h, +2(S~), (2. 4)
are all within an experimentally accessible and eo-(S, ) +K(S, ) .
contro11able thermodynamic field space. This same
feature is shown in the Blume-Capel model in a The energy eigenvalues of X are the roots of the
transverse field (which is a special case of our secular equation
more general model) and has been discussed by the
authors elsewhere. '7 Xq —2DX~q+ (D —n~ n~ = 0
n~)Xq+D— . (2. 5)
In order to have a general picture of the phase
diagrams, we have carried out an investigation of. Equation (2. 5) can be solved by standard methods
this model in the mean-field approximation (MFA). to yield the result
The model Hamiltonian is discussed in Sec. II. We
have obtained the zero-field partition function as a
double expansion in the order parameters. The
zero-field phase diagrams in the whole range of the
=sD+s(D'+3n,'+3n', )'"cos[y+ 3m(k 1)]+e„-
transverse coupling && 0 is obtained in See. III, (2. 6)
where 4=1, 2, 3 and,
The range of values of K where the phase diagram
contains a BCP is found. The study of the system 1 2D' —18(n,'+n', )D+27n', D
in a finite magnetic field is the subject of Sec. IV. (D'+ 3n'+ 3n')"'
The behaviors of the system in a parallel and a
transverse field are treated separately, and the Equations (2. 6) and (2. 7} are used for numerical
corresponding phase diagrams are worked out, The calculations. For the special cases of H, = 0 (or
ranges of the parameters D and positive K where H, = 0) a simple expansion in (S, ) (or (S, )) can be
trieritieal behavior exists are found. Finally in easily obtained from Eq. (2. 5) in the vicinity of a
Sec. V the main points of the work are summarized. second-order phase transition. In general one. can
use Eq. (2. 6) to find a set of self-consistent equa-
P

II. MODEL HAMILTONIAN

tions for the parallel and the transverse magnetiza-
We consider a spin-1 uniaxial ferromagnet with tions.
an anisotropic exchange interaction and a single- . In the limit of zero external field a perturbation
 BICRITICAL AND TRICRITICAI PHENOMENA IN ~ ~ ~

expansion of energy eigenvalues about (S,) = 0 (or beeri vrorked out in Ref. 3.8. %e refer the reader
(S, ) =0) is a convenient approach to discuss the to that paper for the details of calculation.
second-o&rder phase transitions. To do this one as- %'hen a transverse fieM is applied, again. two
sumes that (S, } (or (S,)) is finite: phases are possible, a canted phase and a phase
with moments only in the basal plane. In this case,
E, =e, + 24, X — X + ~ ~ ~
+so, (2. 8a) (S, ) is the appropriate order parameter, and one
16 $8 expands the partition function in powers of (S, ).
s +2~
4
E xa
&
+1 ( ) o
~ + +&to (2. 8b) III. PHASE TRANSITIONS IN ZERO EXTERNAL FIELD

D' (1 Using Eg. (2. 10), one can write the Landau free
energy
2 l, E'+ e.j
F = - (1/P) lnZ,
x —
D' j.
+ ~+~
4 s, s (e, s]~X +~ ~ ~
+so (2. 8c)
f=f fX yfrP+ftPX +fg fX + ~ ~ ~,
(3. 1.)
(3. 2)

c, =D(1 wq), f=PF& fo=PFo= —lnZo, (S. Sa)

(2. 8) yD S (3. 3b)

4K S, '
2Z(S, )
D 4

("."-")
Zo
Although it may appear that Etis. (2. 8a) and
)-
(2. 8b) blow up as (S, 0, in the calculation of the ~ =-'. (3. 3d)
partition function one can show that all divergent
g23
terms from the two energy levels cancel each (S. Se)
Z 2Z
other, and so the free energy is actually an analytic
function of both (S, ) and (S, ). The final result can 1 Z',
be written as a double expansion of the partition fo Z 2Z 5)l
' (3, Sf)
function Z, for small values of (S, ) and (S, )'.
The coefficients fo, f4, and fo are functions of y
g ~agg
f
only. The dependence of on K is only through the
coefficient f~. One should note that knowledge of
= e 'o(Zo+ Z, X'+ Z, q'+ Zp X'+ Z,q'+Z, X'), sixth-order terms is needed only for checking the
stabibty at TCP.
(2. 10} The Landau free energies of the form (3.. 2) have
vrhere
been discussed by Liu and Fisher %'egner
Zo(y) =1+2e ", and
more recently by Bruce and Aharony. ~0 %e just
Zj(r) =r(1 e "), - (2. lib) mention the results for the case when f, & 0, which
is of relevance to our model. The continuous tran-
Zo(r) =r'e ', (2. 1lc) sition from the phase with g&0, X=O (Z phase) to
z, (r)=r[l-e (1 +r+o)r], (2, lid) the paramagnetic phase (P) happens when fo = 0; f~,
f& & 0,
Z4(r) =Jr'e ", (2. lie)
Zo(r) + 'r)],
=r [or 1+e '(1— - (2. 1 1f)
.
f, =b[D-D. (r)]=lr'[t- (r)], . (S. 4)

1 Z(0)
—
P= =
t kT D.(r) =4Z, /rz. t.(r) =4Z. /r*z. ,
, (S. 8)
One gets the same final results if one assumes (S,) i e. , at D.=D,{y) [or t=t, (y)].
to be finite and expands the energy eigenvalues in The transition from the & &0, q = 0 phase (T
(S. ). phase) to. the paramagnetic phase occurs at f~=O,
fn a field parallel to the e axis, (S~) is the ap- fo, fo o,'
propriate order parameter. A simple classical
calculation at &=0 shows that two phases are possi- f, = 'r[d d. (r)]=-(r'/«-)[t t.'(r)1, - (S. 8)
ble: (i) a phase with moments parallel to the applied
field, snd (ii) a canted phase in which the e com-
d=D/K (3. Va)
ponent vanishes chen h, tends to zero. The per-
turbation calculation corresponding to this ease has
1852 KHA JEHPOUR, WANG, AND KROMHOUT |2
to paramagnetic phase [Fig .1(d)]. The Z- P line
before meeting the T-P transition line has gone
2.Q-
through its TCP.
I 0- 'I 0- ~
I 0 i.o N The determination of the flop line in all above
Z
cases is done by comparing the two free energies
0.5 I.O 0,5 I.O
corresponding to the two phases of Z and & order-
-I,O -
z -I.O - -1.0 -I.O- ing. At ~= 0 the flop transition takes place when
D = 4(WK- K}.
(o) (b) (c) (d)
The coordinates of BCP in the D-T plane for any
FIG. 1. Phase diagrams in (D, t) plane for different value of K& 0. 462 can be found from Fig. 2. In this
K values. The BCP is the intersection of the two critical figure D, (y) and d, (y) are plotted as a function of y.
lines (solid) and flop line (dashed), T and Z deisgnate Since K=D, /d, at BCP, one should find the value of
the phases with orderings in the basal plane and along
y such that the ordinates for the two curves has a
the s axis, respectively. (a) K=1.25, (b) K=1. 00, (c)
K=0. 6, and (d) K=0. 35. ratio equal to K. Then t, can be found from y, = P,D,
= D~/t~.
One should note the similarity of the phase dia-
grams in the D-t plane, with those of an anisotropic
c
4 ~S p

c
4K' antiferromagnet in a parallel external field. As in
yg ~
yg Z
the case of an antiferromagnet, the flop line sepa-
The paramagnetic phase is stable against a transi- rates two phases one with xy symmetry (T phase)
tion to the above phases when both f~ and fa are and the other Ising-like (Z phase). Phase diagrams
positive. In general, a phase with both &+0 and similar to Fig. 1(b) (near the flop line) have been
g&0 would be stable only if suggested by Bruce and Aharony'o for structural
phase transitions in perovskites, with basically the
a'=(2') -f4fs'0 (3 9) same symmetries as mentioned above.
Numerical calculation shows that for all values of IV. PHASE TRANSITIONS IN A FINITE FIELD
the parameter y &0, g&0. Consequently, the canted
phase (h= 0) is never stable for our model. The A. E&0.462
transition between @&0, X=0 phase (Z phase) and l. .Field along s uxis
& &0, g= 0 phase (T phase) is of first order (flop
transition) and occurs whenever In the presence of a magnetic field along the s
axis two possible phases exist: one with moments
fs = ( fs/f4)"f2 (3. 9) along the field and the other with canted magnetic
provided f, and f4 are positive. One can easily show moments, The second-order phase transition be-
that f, is always non-negative (i. e. , a T-P phase tween the two phases can be easily understood by
transition is always of second order) while f, taking (S, ) as the order parameter and expanding
changes sign at D = 0. 924 and t= 0. 67, (which is the the free energy as a Taylor series in powers of
Blume-Capel TCP). Equation (3. 9), is the equa- (S, ). The equation for the critical line is found by
tion of the flop line close to the point where it meets setting the coefficient of the second-order term
with the two second-order transition lines. When equal to zero, with the condition that the coefficient
f4&0 the iwo critical lines f~=O and f2=0 and the
flop line meet at the bicritical point (BCP) (D, , t~). Dc dc
When f4& 0 the phase diagram ceases to contain a
BCP. This happens when K& 0. 462. At K=0. 462,
f, ~fz =f4= 0. The possible phase diagrams in the
D-~ plane are shown in Fig. One can distinguish l. 2.0
the following cases:
K&1. In this case the flop line is below the t
axis, and is bent slightly downwards [Fig 1(a)]. .
K= 3. . The flop line is along the D= 0 axis, and
BCP is located on temperature axis at /=+3. For
L)&0one finds the &phase and a&0, the Z phase
[»g 1(b)].
0, 462& K& 1, %e have a BCP in the D & 0 region 0.0' - y'
of D-t plane, as shown in Fig. 1(c}. The flop line 0.0 0.4 0.8 l.2 l.6 2.0
bends upward. FIG. 2. Plot of Dc Eq. (3 4) a" dc Ej (3 6) a
0. 462 &K&0. In this case the critical line f~=0 functions of y =D/t. This plot is used to obtain the BCP
meets a first-order phase transition line of Z phase for K& 0. 462.
 BICRI TICAL AND TRIC Rr THECAL I HZ NOME XA IN. ..

FIG. 3. Schematic phase diagrams in the (D, h.„t) and (D, h~, t) hyperplanes (K & 0. 462). The hatched regions are the
coexistence surfaces. R2 and R& denote the two-dimensional critical surface of order 2, and the one-dimensional criti-
cal surface of order 3 (tricritical line), respectively (Ref. 12). In each phase diagram two R~ surfaces meet at the
BCP ( Rp)

of the fourth-order term is positive: the phase transitions.

If the anisotropy parameter D is also fixed, then
D- n, M' we can study the phase diagram in the (h„ f) plane.
2K D+a,
These phase diagrams are convenient for the study
1 1 1 gD of TCP. Application of an ordering field h, , will
D- o.~ D+n~ 2K
(4. 1)
generate the wings. One should note that here the
whole relevant thermodynamic field space (h, , h, , f}
where
is exper imentally accessible.
n, =2(S, ), +I, (4. 2) To discuss the phase diagrams, -especially close
to BCP, we consider two typical values of K, E= 0.6
(S ) (
Bag flag)y( sn~ --saq+ BD) (4. 3) (~0. 462) and K=0. 35 (& 0. 462). The latter case
The coefficient of the fourth-order term is rather will be discussed after the K=O, 6 case is fully ex-
involved and is given in Ref. 18. The numerical plained in both parallel and transverse fields.
computation of this coefficient is straightforward. The three-dimensional phase diagram in (D, f, k, }
The simultaneous vanishing of the second- and space is sketched in, Fig. 3(a). Three cross sec.-
fourth-order coefficients gives the tricritical point tions of this phase diagram at different tempera-
(TCP). tures of t=0, t=0. 65, and t=.0. 8 are shown in
In discussing the phase diagrams, we shall al- Fig. 4. At t=, 0, we have a quenched moment state
ways consider the parameter E as fixed. If then we at 5, = 0, if 8 & 4K. By increasing the external field
treat the anisotropy parameter D as a field, we the system undergoes a phase transition from the
would have a thermodynamic field space for de- quenched to the canted phase (curve a). The order
scription of the phase diagrams consisting of the parameter is (S, ). One can easily show that thi. s
three variables D, t, and h, (putting the ordering transition is always of second order' for all values
field k, =O). Then the phase diagrams are the ex- of K, and the equation of this curve-is given by
tension of those considered in Sec. ID, to the h, &0 I, =[D(D-4')]'" . (4. 4)
region. The (D, f, h, ) field space is necessary for
understanding the nature of BCP and the geometry The line b on Fig. 4 is the locus of transition points
of phase diagrams close to it. The variation of this from the canted phase to the paramagnetic phase as
geometry with the transverse coupling K will pro- 5, is increased. The order parameter for this
vide us with the complete qualitative description of transition is again the moment (S,), which vanishes
1854 KHA JEHPOUR, WANG, AND KROMHOUT

An interesting case is K=1. 0. From Eqs. (4. 4)

and (4. 6) one can find the corresponding curves a
and b at t=0 while Eq. (4. 6) gives the TCP at t=O
at (D=O, h, =O). At finite t, a similar situation
prevails, i. e. , the balloonlike surface in Fig. 3(a)
2.0 is totally a critical surface and the two tricritical
lines we had for E& 1 coincide with the flop line,
now on the t axis.
I.O
BCP', 2. Field in the basal plane

0.5 I.O
I

l. 5
%ith a fieM applied in the transverse direction,
while Pg, =0, we have a three-dimensional phase
FIG. 4. (D, h, ) cross sections of Fig. 3(a) for X=0. 6 diagram in the (D, h, , f) field space, as given sche-
at different temperatures. The bvo t=0. 0 lit, s a and b matically in Fig. S(b). The surface of the second-
meet at infinity and enclose the canted phase. Solid lines order transition can be obtained by a perturbative
and dashed lines represent, respectively, the second- expansion similar to the one discussed in Ref. 18.
order and the first-order phase boundaries. The dotted Again we would have the free energy expanded in
line is the projection of the line of tricritical points in even powers of (S,). Vanishing of the second-order
Fig. 3(a) on the (D, h ) plane. term in the expansion gives the critical point, while
the fourth-order term is positive. The equation of
surface of second-order transitions thus obtained is
at the transition point if it is of the second order. (4/zo)(bo e "o+ bg e "~ + h, e "2) = 1, (4 6)
The second-order part of curve b is given by
where
h, =D+2(K-1) . (4. 6)
o= —
D
a
kl
bi =-
X4g
&2 (4 9)
The lines of first- and second-order transitions join X&)

at TCP located by x = (D'+ 4u')'" (4. 10)

2K'+2(D 1)K- D=—0 . o. , =h, +2K(S,), , (4. 11)
The two curves a and b meet at D=h, =~. At any (4. 12)
finite temperature the two lines join at finite D and
h, (curve c). As temperature increases the area
shrinks to a point at BCP. As shown in Fig. 3, the
two lines of TCP ( 8,) bounding the critical surfaces so= 8 "~ .
( Rz) (notation due to Chang, Hankey, and Stanley'~)
meet each other at the BCP. (S,)o is the transverse moment when (S,) =0 and is
Some (h„ f) cross sections of the phase diagram given by the following self-consistent equation:
Fig. 3(a) for different D values are shown in Fig. 6.
The projection of the line of TCP's is also included
in these diagrams. Such (h„ f) phase diagrams de-
serve a special interest: since (S, ) is the order
parameter, its conjugate h, can be utilized to set
up a three-dimensional phase diagram in (h, , h„ f) 0.25--- D=I.O

space. The accessibility of the whole thermody-

0.20- D=095
namic field space is a point we want to emphasize.
To have a TCP in (h„ f) plane, the parameter K D=0.9
should be greater than 0. 462, and D should be in O. l 5-
0=0.85
~a

the following range:

O. IO-
2E —2K
' (4. f)
1-2K 0.5-
where, D~ satisfies the equation
O. I 0.3 0.5 0.7 0.9
4K- DE 4K+ 2DE
(4. 7a) FIG. 5. (h„t) phase diagrams for K=0. 6 and various
D~
values of the anisotropy D. The projection of the tricrit-
For K=O. 6, D should lie in the range 0. &3&a&2. 4 ical line in Fig. 3(a) on the (h~, t) plane is displayed by
to have a TCP in an applied field h, . the dotted line,
 BICRITICAL AND TRICRITICAL PHENOMENA lN. .. 1855

0.8 ', BCP I. 5—

l.4—
0.6
I. 3 P

OA l. 2— /
/

0.2 I. I
/

0.0 I.O '

—,

hq hz
O. l 0.3 0.5 0.7 O. l 0.2 0.3 0.4 0.5
FIG. 6. P, h~) phase diagrams for K=0. 6. These
phase diagrams are cross sections of Fig. 3(b). For t FIG, 8. (D, hg) phase diagrams for K=0. 35. The first-
& 0. 924, the phase diagram consists only of a second- order l.ines (dashed) Pq and PR terminate the second-
order line. The dotted line is the locus of the tricritical order (solid) boundary of the canted phase at point P.
points. As t increases the canted phase region decreases.

projections of the tricritical lines in the (D-h, )

4ao sinh(-,' Px) plane are also shown in Fig. 6. As the temper-
(4. 14)
( x[2 cosh(-,' Px) + e-""j
' ature increases these lines get shorter, and finally
vani. sh (at t =0. 924). The (h, , f) phase diagrams,
Again the equation of the fourth-order coefficient is Fig. 7, contain a TCP, for D within a certain
quite involved although its numerical calculation is range. Here again we have the possibility of ob-
straightforward. serving the wing coexistence surfaces in the ac-
%hen D=O one can obtain easily the equation. of cessible field space of (h, , h„ t). For the particu-
the second-order line: lar value of K= 0. 6, D should lie between 0. 44 and
(4. 15) 0. 83 to have a TCP.
It should be noted that the range of values of D
(4. 16) where a TCP can be seen on the phase diagram is
(4. 1V) somewhat wider tha. n that given by Eq. (4. 7). For
example for K=O. 6, if 0. 44&D&2. 4, one has the
8, (y) is the Brillouin function. possibili. ty of getting a YCP by applying one of the
In Fig. 2(b), the two coexistence surfaces (shaded) two fields h, or h,
intersect each other at the flop line which ends up
in the BCP. The two tricritical lines ( R~) bounding B. 0(X&0.462
the critical surfaces R~, also meet at the BCP. In Sec. DI we noted that at H = 0 we do not have a
Figure 6 shows some constant temperature cross BC P on the D- t phase diagram if 0 & K & 0. 462,
sections of this phase diagram (for K=O. 6). The Fig. 1(d). When D& DO=4(MK —K), the system ini-
tially orders in the basal plane, while it orders
along z axis for D&DO. For a typical value of
hg K=O. 35, DO=0. 96. When D&DO (D&DO) one can
apply a field along a longitudinal (transverse) di-
rection, without it being conjugate to the proper or-
I la ~ ~ ~

0.25-
~ ~
~
0=0.5
der parameter.
0.20- A phase diagram in the (D-h, ) plane at constant t
for K=O. 35 and D & Do is shown in Fig. 8; it con-
O. I5-
sists, for all temperatures, of two coexistence
O. IO- lines QP (between the canted and the paramagnetic
phases) and PR (between the states with longitudinal
0.05- moments only), and a second-order transition line
0=0.75
a Sp. These three li. nes intersect one another at
O. l 0.3 0.5 0.7 0.9 poi. nt I'. As temperature is raised, the point I'
FIG. 7. (h~, t) phase diagrams for K'=0. 6 and various moves down and the area of the region of canted
values of anisotropy D. Solid l. ines and dashed lines phase decreases. This area shrinks to zero at
represent, respectively, the second, and the first-order f =0. 6V (independent of the value K). The phase
phase boundaries. The dotted line is the projection of diagram exhibits no TCP for any value of K i.n this
the tricritical line on the (h~, T) plane. range.
1856 KHA JEHPOUR, WANG, AND KROMHOUT

"zx t0 (i) The model characterized by the Hamiltonian

Q. 6 (2. 1) exhibits two kinds of ordering, and therefore
I
two kinds of critical behavior are expected in zero
TCP field, depending on the magnitudes of the anisotropy
parameter D and transverse coupling K The two
kinds of ordering are labeled as Z (Ising-like) and
T (xy-like). They are separated by a CXS (coex-
istence surface), the flop line, in the D-f phase
diagram.
0, 2— I (ii) The phase diagram in the D tpl-ane (h =0)
Al
]
/
contains a BCP, if the transverse coupling
Q. l K & 0. 462. BCP is the terminus of the flop line
K*0,55 K&OA K-"0.5
CXS. As the transverse coupling is increased the
00 E BCP moves towards the temperature axis, and in
O. I 0.5 0.5 0.7 the limit of isotropic exchange it lies along the t
FIG. 9. t) phase diagrams for D=O. 98 and three
axis.
(h~,
different values of the transverse coupling: K=O. 35, (iii) In the presence of an external magnetic field
0. 4, and 0. 5. For K&0. 462 no TCP exists but a critical (along either the parallel or the transverse direc-
end point C at (0. 27, 0. 58). tion) there exists a, TCP in the (h, f) plane, provided
that K and D are within certain ranges which are by
no means narrow.
(iv) In the four-dimensional field space of (D, h, ,
The h; t phase diagrams at D = 0. 98 & Dp are A„ f), four tricritical lines approach the BCP when
shown in Fig. 9, for several values of K For 0. 462 &K& 1.
Z=O. 35 (K=0. 4), there exist two lines of first- At h =0, the flop line is the locus of the coexis-
order phase transitions AC (BC for %=0. 4) and AF tence of four ferromagnetic phases. To distinguish
(BG) and a. critical line AD (BZ), which intersect these phases we should apply the infinitesimal fields
at the point A (B). The slopes of the lines at the +5, -0 -
and +5, 0. We designate these phases by
intersection point are different in general, although S', Z", T', and T . The signs + and —indicate
they appear to be equal in the drawing. The line the direction of the infi. nitesimal fields. In the
AC (BC) describes a phase change between the two notation of Chang, Hankey, and Stanley' the flop
states with longitudinal moments only, and is in line is an X, CXS. Since a (a+I)-dimensional CXS
f act, a part of a cross section of the wing CXS in ~x„, (p being the number of the phases coexisting)
the Slums-Capel model (cf. Ref. 19). It is also is bounded by a critical surface (CRS) of dimension
clear from the figure that as E increases, this co- d and order p, our BCP is a critical point of order
existence line shortens, and the canted phase region '
4, Bo. Chang et al. and Hankey et gl, have
'
increases. At E=O. 462 it completely disappears. carried out a rather general study of the critical
%hen a&Do, the system exhibits tricriticality, points of higher order. Based on an Ising model
only in an applied field in a transverse direction. with variable interplanar interactions, they have-
,
The D-h, phase diagram for K&0. 462, Fig. 10, re- demonstrated that a R~ critical point is the inter-
sembles the D-h, diagrams for K &0. 462. For section of several tricritical lines. A similar situ-
K=O. 35 the phase diagram exhibits a change in the
nature of phase transition for systems with B value
in the range 0. 56 &D& 0. 924, and the conj esponding
tricritical temperature range 0 & t, & 0. 67. «ince in
this range of temperature D-k, phase diagrams lie O-
very close together, we have only drawn the T = 0. 4
case in Fig. 10. TCI' for several values of temper-
I.
.
~TCP at t=0.65

ature are also depicted on the diagram.

The limiting case of K=O corresponds to the 0.5
Blume-Capel model in a transverse field which has
been discussed in detail elsewhere. '
V. CONCLUSION
0.0
We have studied an S =1 ferromagnetic model 0.0 0.5 I.O
with both exchange and single-ion anisotropy. A FIG. 10. (D, k, ) phase diagrams for E=O. 35 and two
summary of the main features of the model as re- values of temperature, The phase diagram resembles
vealed by an MFA calculation is as follows: F ig, 6,
12 BICRITICAL AND TRICRITICAI PHENOMENA IN ~ ~ ~ 1857

ation prevails at the BCP of our model. restriction delimits a complete scan of the vicinity
The assumption of homogeneity near a BCP (or a of the BCP in all directions. The model discussed
Ro) has important consequences. In addition to the in this paper, though not suffering. from this re-
bicritical scaling laws, the tangency of the critical striction, is subjected to another limitation in the
lines to the flop line may be of special importance. study of BCP: A controlled variation of D in a real
Also of interest is the competition between critical, magnet, if not impossible, is rather difficult (al-
bicritical, and tricritica) behaviors in the vicinity though there are evidences that D can be varied by
of the BCP, which results in crossovers between applying a pressure to the crystal ).
these different criticalities. Hankey et al. , ' Perhaps the most interesting feature of this model
based on the simultaneous validity of the scaling is the TCP behavior. Since the ordering field is
groups of critical points of order 2 (ordinary criti- experimentally accessible, the wing CXS's are
cal point) and 3 (tricritical point) with that of or- wholly in the experimental fi, eld space. The range
der 4, have arrived at the "double and triple scal- of the parameters Kand D to permit a TCP in an
ing functions" for the singular part of the Gibbs external field is significantly large. Therefore, a
potential, and have discussed the corresponding search for a system with the model Hamiltonian
crossover regions. Most of the systems exhibiting (2. I) is highly desirable.
bi criticality (uniaxial antiferromagnets and an- A study of this model in the light of the recent
isotropically stressed perovskites' ) suffer from works by Fisher and Nelson, Nelson, Kosterlitz,
the restriction that the ordering field is either fic- and Fisher, and Bruce and Aharony wi. ll be pub-
ti.tious or uncontrollable in the laboratory. This lished later.

Supported in part by the Committee on Faculty Research Commun. 12, 779 (1973).
Support. K. S. Liu and M. E. Fisheg, J. Low. Temp. Phys. 10,
~Supported in part by National Science Foundation under 665 (1973).
Grant No. GH-40174. 9F. Wegner, Solid State Commun. 12, 785 (1973).
'Ph, Kohnstamm, in HandMch dew Pky sik, Vol. 10, A. D. Bruce and A. Aharony, phys. Rev. B 11, 478
edited by H. Geiger and K. Scheel (Springer, Berlin, (1975).
1926), p. -223; I. R. Krichevskii et al. , Russ. J. Phys. ~~8. Ba,usch, Z. Phys. 254, 81 (1972).
Chem. 37, 1046 (1963); G. S. Radyshevskaya et al.', ' T. S. Chang, A. Hankey, H. E. Stanley, Phys. Rev.
J. Gen. Chem. USSR 32, 673 (1968); see also R. B. B 8, 346 (1973); A. Hankey, T.' S. Chang, and H. E.
Griffith and B, Qidom, Phys. Rev. A 8, 2173 (1973), Stanley ibiH' 8 1178 (1973)
and references cited therein. F. Harbus, A. Hankey, H. E. Sta, nley, and T. S.
2E. H. Graf, M. Lee, J. D. Reppy, Phys. Bev. Lett. Chang, Phys. Rev. B 8, 273 (1973).
19, 417 (1967); I. S. Jacobs and P. E. Lawrence, Phys. D. Nelson, J, M, Kosterlitz, and N. E. Fisher, Phys.
Rev. 164, 866 {1967); G. GoeQner and H. Meyer, Phys. Rev, Lett. 33, 813 {1974).
Rev. Lett. 26, 1534 {1971);D. P. Landau, B, E. Keen, n*(d) = (4+3. 176 e')/(1+1, 294 e) up to order c (e =4-d).
B. Schneider, and W. P. Wolf, Phys. Bev, B 3, 2310 See Befs. 5 and 14.
{1971);M. Blume, V. J. Emery, and R. B. Griffith, ~~The real. nature of the staggered fields in certain anti-
Phys. Rev. A 4, 1071 {1971). ferromagnets is discussed in M. :Blume, L. M. Corliss,
3C. J. Gorter and T. Van Peski-Tinbergen, Physica J. M. Hastings, and E. Schiller, phys. Rev. Lett. 32,
(Utr. ) 22, 273 {1956). 554 {1974).
4G. A. Smolenski, Fiz. Tverd. Tela 4, 1095 (1962) fSov. M. R, H. Khajehpour, R. A. Kromhout, and Yung-Li
Phys. — Solid State 4, 807 {1962)]. Wang, J. Phys. A 8, 913 (1975).
M. E. Fisher and D. B. Nelson, Phys. Bev. Lett. 32, Yung-Li Wang and M. R. H. Khajehpour, Phys. Rev.
1350 (1974); A. Aharony and A. D, Bruce, ibid, 33, B 6, 1778 (1972),
1350 (1974). 9M, Blume, J. V. Emery, and B, B, Griffith, Phys.
6Ch. Wissel, Phys. Status SolidiB51, 699 (1972), and Rev. A 4, 1071 (1971).
references cited therein. D. Meier, M. Kornezos, and S. A, Friedberg, ~P
~H. Weitzel, , Z. Kristallogr. 131, 239 (1970); H. A. Conf. Proc. 2, 1316 {1974).
Qbermayer, H. Dachs, and H. Schroke, Solid State