subsets of Q) is called a Gibbs m easure (w ith H am iltonian H ) if it satisfies th e D obrushin-
where uD is the conditional probability (finite-volume Gibbs measure):
uD (an ) = i = ^ e x p ( - e H (a D |^
dc
) .
Z D,tp
Here,
Z
d
,
v
= ^ 2 e x p ( - в Н (SDI^dc)).
CTD GH
d
Since only neighboring vertices interact, th e GM satisfies th e M arkov property, if a
configuration wWn is given, th en th e configurations in Vn-1, i.e. inside Wn , and in V \ Vn+1,
i.e. outside Wn are (conditionally) independent.
It is known th a t for any в > 0 th e set of all GMs forms a nonem pty convex com pact
subset in the space of all probability m easures (see, for example, [1], [2], [7]).
Fix x 0 G V. For x, y G V we will w rite x < y if th e p a th from x0 to y passes through
x. A vertex y is called a direct descendant of x if y > x and x, y are neighbors. Note th a t
in Г к every vertex x = x 0 has k direct descendants, and x0 has k + 1 descendants.
Consider a special class of GMs, which in th e work [1] are called M arkov chains, and
in the work [4] - splitting GMs (SGMs).
Let h : x M h x = (h0,x, h 1,x, ..., hm,x) G R m+1 be a vector function of x G V \ {x0}.
Consider a probability distribution ^ (n) on QVn:
^ n (an) = Z - 1 e x p ( - в Н (an) + £
hCT(x),x),
(3)
xGWn
where a n G QVn and
Zn = ^ 2 e x p ( - в Н (an) + ^ 2 ha(x),x).
ffn£Hyn
xGWn
A probability distribution ^ (n) is said to be com patible if for any n > 1 and a n-1 G
QVn-1, we have
£
^ (n) (an-1 U
ш
„) = ^ (n-1)(an-1),
(4)
WnG^Wn
where a n-1 U wn G QVn.
In this case, there is a unique m easure ^ on Q such th a t
M W n = a n }) = ^ (n)(an)
for all n and a n G QVn. Such a m easure is called an SGM corresponding to th e H am iltonian
H and to th e function x M hx, x = x0.
The following statem en t gives a condition on hx guaranteeing th a t th e distribution of
^ (n)(an) is com patible.
T h e r o e m 1.[5] The probability distributions ^ (n)(an ) ,n = 1, 2,... determined by the
formula (3) are compatible iff f or any x G V \ {x0}
the following equation holds:
hx = £
F ( h , m , 9 ) ,
(5)
yGS(x)
369
где В = eJe , в = T . Here h*x = (ho,x - hm,x, hi,x -
hm,x,..., hm—i,x -
hm,x) and F ( • , m ,
R m
m
R m is a vector function, i.e.
F (h, m , В) = (F
0
( ( h , m , B ) , ..., Fm- i ( h , m , B ) ) ) ,
h = (ho , h i , .. ., hm- i ) ,
such that
лт- l B^i—
m - l
E ”="c1 B ^ e ^ + Bm
F i ( h , m, B ) = l n ^ j 0 1------------------- ,
i = 0 , . . . , m - 1.
(6)
A
'
E m=0i Вт— ehj + 1 '
’
’
V
^
Let G k is a norm al subgroup of G k.
D e fin itio n 1. A set of vectors h = { h x , x E G k} is called G k-periodic, if h yx = hx for
any x E G k , y E G k. Gk-periodic collection of vectors is called translation-invariant.
D e fin itio n 2. A m easure ^ is called Gk -periodic, if it correspond to a G k-periodic set
of vectors h . Gk-periodic Gibbs m easures are called translation-invariant.
In num erous studies (see, for example, [1], [3-5]), periodic Gibbs m easures for various
models of statistical mechanics have been studied on the Cayley tree. These measures
were m ainly translation-invariant, or Gk )-periodic, where G k ) is a subgroup of G k, which
consist of th e words of even length.
Let k = 2 and Ф = { 0 ,1, 2}, i.e. m = 2. Suppose h 2,x = 0. For translation-invariant
h о
hi
Gibbs measures from (5), using notations e "
2
° = x, e "
2
" = y, we get th e following system
of equations:
x2+dy2+e2
x
e2x2+ey2 + i
y = Ox2 +y2 +e
y
O2x2+Oy2+i
(7)
Note th a t x = 1 satisfies th e first equation of (7) for any B. In [6] th e following is
proved.
T h e o r e m 2. For th e SOS model w ith m = 2 and В = Bcr( « 0.1242) on th e Cayley
tree of order five, there are at least two non-periodic Gibbs measures.
R e m a r k 1. It was proved in [3] th a t for m = 2 and В > В'( ^ 0.53) on the Cayley tree
of order five there exists a unique translation-invariant Gibbs m easure; if В < В', there is a
three translation-invariant Gibbs m easure on a Cayley tree. Note th a t th e described Gibbs
measures in Theorem 2 are not translation-invariant, i.e. different from th e m easures found
in [3].
T h e o r e m 3: For th e SOS model w ith
m = 2 and
В E E( a, b) , there are at least two
non-periodic Gibbs m easures on the Cayley tree of order k = a + b + 2.
R e m a r k 2. Note th a t the set E( a, b) is not empty, since th e case a = 1,b = 2 is
considered in Theorem 2. Also it is easy to cheek, th a t in th e cases a = e,b = 2e, We E N
the set E (a,b) is not empty. May be exist another (a,b) E {(a,b) : a = e,b = 2 e , We E N },
which the set E( a, b) is not empty.
R e f e r e n c e s
1. K. Preston, Gibbs states on countable sets, Mir, M., 1977
2. Christof Kulske, Utkir A. Rozikov R andom Structures and A lgorithm s DOI
10.1002/rsa. 2017 640.
3. A. E. Mazel, Yu. M. Suhov, J. Statist. Phys., 64 (1991), 111-134.
4. R. Fernandez, C ontour ensembles and th e description of G ibbsian probability d istri
butions a t low tem perature, h ttp ://w w w .u n iv -ro u en .fr//L M R S /P erso p ag e/F ern an d ez/
resucont.htm l, 1998.
370
5.
Rozikov U.A. Suhov Y.M. Gibbs m easures for SOS model on a Cayley tree.
Inf.D im .A n.,Q uant.P rob. and R elated Topics. 2006, V9, N3, 471-488.
6. C.Kuelske, U.A.Rozikov, E xtrem ality of translation-invariant phases for s th ree-state
SOS-model on the binary tree. J.S tat.P h y s. (2015) 160:659-680, doi 10.1007/s10955-015-
1279-9.
7. Ya. G. Sinai, T heory of phase transitions: rigorous results, Science, M., 1980.
8. U.A. Rozikov, Gibbs Measures on Cayley Trees W orld Scientific, Singapore, 2013.
Do'stlaringiz bilan baham: