HAL Id: jpa-00246975
https://hal.archives-ouvertes.fr/jpa-00246975
Submitted on 1 Jan 1994
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Solvable nineteen-vertex models and quantum spin chains of spin one
Makoto Idzumi, Tetsuji Tokihiro, Masao Arai
To cite this version:
Makoto Idzumi, Tetsuji Tokihiro, Masao Arai. Solvable nineteen-vertex models and quantum spin chains of spin one. Journal de Physique I, EDP Sciences, 1994, 4 (8), pp.1151-1159.
�10.1051/jp1:1994245�. �jpa-00246975�
J. Phys. I IYance 4
(1994)
llsl-l159 AuGusT 1994, PAGE llslClassification Physics Abstracts 64.60C 75.10J -02.90
Solvable nineteen-vertex mortels and quantum spin chains of spin
oneMakoto Idzumi
(*), Tetsuji
Tokihiro and Masao AraiDepartment of Applied Physics, University of Tokyo, Hongo, Bunkyo, Tokyo, i13, Japan
(Received
10 January1994, received in final form 12 April 1994, accepted 15 April1994)
Abstract. We solved the Yang-Baxter equation for the R-matrices of three-state vertex mortels with ice condition, and obtained a complete list of solvable mneteen-vertex mortels and
associated quantum spin Hamiltonians of spm one.
In this brief paper we report on a
complete
list of solvable nineteen-vertex models whichare classical statistical mechanics models on square lattices
(see Fig.
l, which showsvisually
our
assumption
on the Boltzmannweights).
The term solvable means that the Boltzmannweights
of the model or the R-matricesR(u),
withspectral
parameter tt,satisfy
theYang-
Baxter equation
(YBE)
or thestar-triangle
relation.(R(u))u~c
defines afamily
of solvable models. When the YBE issatisfied,
the transfer matrices of the models in thefamily
commute:[T(u), T(u)]
= 0. It means coefficients
T(")
of tt" inT(u)
form a set ofcommuting
operators.It is believed that the existence of such
commuting
operators is a sufficient condition forsolvability
of the model [1, 2]. Infact,
in many cases, there exist methods to obtain an exactexpression for the free energy: for
example,
one canapply
the Bethe Ansatz or the quantuminverse scattering method
[Ii
or some other methods [2]. To ourknowledge, however,
we donot have any
almighty
method to get the free energy eventhough
the R-matrix satisfies theYBE;
we must try toapply
the Bethe Ansatz method or the other case by case(but
how tosolve the models is non the
subject
of thispaper).
We also
give
a list of theintegrable
quantum spin chains ofspin
one, which are related to the above vertex modelsthrough
the transfer matrices. The Hamiltonian is definedby
a
logarithmic
derivative ofT(u).
The set ofmutually commuting
operators(T("))n=o,1,2,...
includes the
spin
Hamiltonian. We have said the spin chain isintegrable
because of the existence of such manycommuting
operators.This paper is
organized
as follows. First we define vertex models and associated quantumspin
chains in ageneral setting,
andthen,
we shall define the nineteen vertex model andassociated spin chain of spin one which we shall consider in this paper. After
that,
we descnbe(*) Present address: Department of Mathematics, Shimane University, Matsue, Shimane 690, Japan.
II52 JOURNAL DE PHYSIQUE I N°8
tl"~> ~ à=~,
,
~ C"~> ~
g~ ~ ~ ~,
P"~,
, ,
~~~
o=~.
Fig. l. Allowed nineteen vertex configurations.
Dur method ta find Dut ail the solvable models. We note that we made fuit use of Mathematica
m the course of calculations. Dur strategy was
e~hausting
ailpossible
cases with thehelp of
thecomputer. Results are summarized in three tables:
complete
lists of solvable nineteen-verte~models and associated solvable quantum spin Hamiltonians
of
spin one are shown in tables I andII;
we aise show related solvable t- J models m table III. There we shallpoint
Dut that somemortels in Dur lists were
already
known. For the methods of solution for the known models, weshall refer ta the
original
papers. As was mentioned above we shall net be interested in the method of solution.A vertex model on a two-dimensional square lattice is defined as follows. Let a spin variable live on each bond of the lattice and take values in some range, say
I;
forexample
I=
(1,
0,-1).
TO each vertex
(or site)
attach a Boltzmannweight
P
j~~OE ~ "
vp
~
oe
N°8 NINETEEN-VERTEX MODELS AND SPIN CHAINS OF SPIN ONE l153
where ~t, v, a,
fl
E I. Thepartition
function isz
~ ~
j~~OEUP configurations vertices
# tr
T~,
where the T is the transfer matrix
j'OEl"'CEN j~~loElj~~20E2 j~~NOEN
ôl" ôN
~
~261 ~362 ~lôN
~l>... ~N
We assumed that the size of the lattice is M x N and that boundaries are connected ta the other sides sa that the lattice is on a torus. In the
thermodynamic brait,
we getZ~Af
where Ao is the maximum
eigenvalue
of the transfer matrix(we
assume here that it is netdegenerate).
Onegoal
of statistical mechanics is to find an exact expression for theAo.
Below,
we suppose that the Boltzmannweights
areparametrized by
some functions of a variable u EC,
calledspectral
parameter, as R(~1).A quantum
spin
chain is associated with the vertex model. The Hamiltonian is definedby
H =
lT111)~ £T1~l)1)1...)1 l~-~.
It is related to the Boltzmann
weights
of the vertex modelthrough
the transfer matrix. We note that the Hamiltonian thus definedhas,
ingeneral,
a verycomplicated
form and can containlong
range interactions.We now assume that the Boltzmann
weights
or the R-matricessatisfy
the(additive) YBE(~
R(~1)j(j(R(~1+ v)((j(R(v)((((
=R(v)j(j(R(~1
+v)j(((R(~1)((((. Il)
From the YBE we get
Riv Îl)iTiv)
@Till))
=
iTill)
@Tiv))Riv Îl)
12)where the T is the
monodromy
matrix definedby
~~~~~Î~~"
ÎÎ ~ ~~~~~~~l
~~~~~~~~~~~~~ÎÎ~~
~2 ;. ~"
Tàking
traces of the bath sides we haveviii)> Tiv)1
= °. 13)
(~ It is known that there exist nonadditive YBES which produce solvable models. An example is the chiral Potts model [4, Si. In this paper we shall net be concerned with such models related to
nonadditive YBES.
l154 JOURNAL DE PHYSIQUE I N°8
Table I. Vertex
weigllts
of solvable 19 vertex mortels.Here,
U e e", ik ++1,
â~ +à -1= 0,
fl~
(à 1)fl (à
1)~= 0.
vertex
weight #1
#2#3 #4 #5 #6
a
e~~
ch~l +chqsh~l
ch~l+chqsh~l
+ ii~1 1~ ~ ~
(ô 1)(1
eV) -eV + e~U~~~
l +
(à
+fl
1)eU 2 eVe~U = a = a 1 + b b + eV
1
g 0 0 0 0 +b +b
o
e°~
= a ch~l
chqsh~l
+13~1 1 + b b + e~p 0
sh~sh~l shqsh~l
14~1 0 0#7 #8 #9 #10
1 ~~1 + ~1~
uj4
u4 ~ sh(~1 +q)sh(~1
+2q)
~ ~ ~
shqsh2q
[11
+ ~12)
~
[iii u4)
~~ Sh~lShl~1n)
~
u3ji
3 +V~~i4) ShnSh2q
2
1
-U(2+U~)
b+3 U3
1 ~1
j(4
U~~~~)~
~~s q
i1~l +
~U~(1 U~) +(
~~)~/~U
b ii)
3 2 s q
1 1
~ ~~ l +
VÎ
~
2~
+( ~~~)
sh~lsh(~1+ q)
2~
~ 2~ 23 +
VÎ
~
~
shqsh2q
U(-1
+ USh~lsh(~1 +
q)
%2
§ (~~l
+ ~l~ ° ~ ~~ShqSh2q
N°8 NINETEEN-VERTEX MODELS AND SPIN CHAINS OF SPIN ONE l155
Table II. Solvable
spin-1
quantumspin
Hamiltonians. ik,à, fi
are as above.coefficients
#1 #2 #3 #4 #5
a O
chq
+shq -chq
+shq
-212 + 13 + 14~ à +
fl
fl IA C)
+ 303chq -3chq (ii
712 + 613) o1 O
chq -chq -12
+ 13 0à
IA
+C)
+ 203chq -chq iii
512 + 413 0e O
chq
+sh~ -chq
+shq
-212 + 13 + 14Il
+1)
aq -O
-chq chq
212 13 -a~t -O
-chq chq
12 13 0#6 #7 #8 #9 #10
~ ~~ ~ ~~~ ~ ~
~~ Î121
~
-2 1 -21 +
8VÎ
~~~~~
2 s
2q
~ ~
~~ s~q
~ ~~ ~
~~
~s/ÎÎ~~
~l
-j(3+2ii+12) ~~~~ -9+4v$+4( ~~~)~/~ a-ii~)(~
S n
~ ~
~~
shlq
~ ~
~~ Î~Î
l156 JOURNAL DE PHYSIQUE I N°8
Table III. Solvable t J mortels. Here,
A,
C,O,
q arearbitrary
constants, and ik = +1.#1-#4 correspond
to tllose in tables I and II.#1 #2 #3 #4
t 0
-shq -shq
-14Jz
2(A C)
0 0 2iiJi 0 0 0 212
V
IA
+C)
+ O2chq
0 ii+13
Expanding T(u)
=£ T(")~1",
we get[T(~), T(")]
= 0. It means
(T("))
forms a set of commut-ing operators, The existence of such a set is
regarded
as a suilicient condition forsolvability.
In many cases, one can find an
expression
for the free energyby
means of some appropriate methods:equation (2)
is astarting point
of the quantum inversescattering
method (1] and equation(3)
is a basic relation of Baxter'scommuting
transfer matrices method [2].Again
weremark that an
almighty
method of solution is notknown;
ingeneral
we must try toapply
the Bethe Ansatz method or the other caseby
case.Let us
specify
the model which we shall consider in this paper. The nineteen-verte~ model in this paper has the range of spin variable(1,
0,-1)
and allowed vertexconfigrations
as shown infigure
1. The otherconfigurations
give zero Boltzmannweights.
For each horizontaledge,
weassigned
aright-arrow
for spin variable +1, a left arrow for -1, andnothing
for 0,respectively.
For each vertical
edge,
weassigned
an up-arrow for +1, a down arrow for-1,
andnothing
for
0, respectively.
We have assumed thefollowing symmetries: Ii)
the ice r~lle(or
the icecondition)
RS)
= o uniess ~t + a = v +p;
(ii)
the usualspin up-down
symmetryj~~OE j~~~>~OE,
HP -V,-fl'
(iii)
an invariance with respect to reflections~~OE ~OE~
~Pv
vp pv a~.
In this sense, it is a direct extension of the well-known six-vertex model (3, 2]. Trie q~lant~lm
spin
Hamiitonianof
spin one is defined as beforethrough
the transfer matrix of the nineteen-vertex model. The ice rule results m a relationl~
~Zj ~, ,
1-e-, the Hamiltonian commutes with the total
projected spin
S~=
£~ Sj.
In thefollowing
wefurther make a standard assumption
li(Ù)
Î~, l~~pô~flôv~.
N°8 NINETEEN-VERTEX MODELS AND SPIN CHAINS OF SPIN ONE l157
Then the Hamiltonian will contain
only nearest-neighbor
interactions~ "
i
~~~~>
~°J-i~J-i llRi~l)iiiiil«-~ ôai+iôi+i
ôoe~~~.We have said that this quantum
spin
chain hasspin
one because to each site a three-dimensional vector space is attached. Later we will see that this isactually
written in terms of thespin-one representation
ofsu(2)
generators.We have solved the YBE
Il)
for the R(~1)given
infigure
1 and found ail solutions. Dur method to find ail R-matricessatisfying
the YBE is quite naive andstraightforward.
We assume thatIi)
the R-matrix hasonly
19 nonzero entries with thesymmetries
described in the lastparagraph (see Fig. l); (ii)
the R-matrix R(~1) is ananalytic
function of ~1 m aneighborhood
of ~1= 0;
(iii) Rio)
=
P;
and(iv)
it satisfies the additive YBEil).
TO find apossible
form of R(~1) let usexpand
it in ~lasRj~)
=
R(o)
~~j~(1) ~ ~2 j~(2) ~
where, by assumption, R(°)
= P.
Substituting
thisexpansion
into theYBE,
we get several relations to be satisfiedby
the coefficient matrices R(~) We wish to salve theseequations
and to find a solutionR(~),
i= 1, 2,.
Luckily
we found that it waspossible.
Infact,
in the course ofcalculations,
we found thatla)
the coefficient matricesR(~),R(~),.
were ail written interms of entries of the
R(~)
and 16) equations for the entries of theR(~)
could be solved: therewere a finite number of
possible
forms ofR(~)
andme e~ha~lsted all
possibilities.
In this way wegot ail
possible
solutionsR(~).
The observed factla)
is known as thetangential star-triangle hypothesis
[6],which,
to ourknowledge,
has net beenproved ngorously (but
is believed to betrue).
We observed that thishypothesis really
holds for the R-matrices we have assumed.Dur results are summarized in tables I-III. In table I we show the
complete
list of solvable nineteen-vertex models(three-state
vertexmodels)
with theprescribed symmetries; they
areclassified into ten groups for convemence. Table II gives the
corresponding
quantum spin Hamiltonians of spm one; this table shows coefficients of the Hamiltonians, which are definedby
N
~
~~J,J+1>
j=1
Hj,j+i =a(S)S)~i
+S)S)~~)
+fl(Sjsj~i)
+i(S)S)~i
+S)S)~~)~
+
ôls]S]+1)~
+fllsj Sj+1)~ [Sjsj+i
+S]S]+1)~ lS]S]+i)~l
+ q
l§,j+1
+~l(Sj
+Sj~i)~, Pj,j+i =(Sj Sj+1)~
+(Sj Sj+i)
-1,where S~, = x,g,z, are the standard spin operators of spm one
(the
spin-one representation ofsu(2) generators)
10
0 0 -1 0 1 0 0~~
V~
~~
~~
~ÎÎ ~~
~~
ÎÎ l~
Let us point eut known and solved models m our tables. First we remark that model
#1
is
trivially
solvable as there is no interaction. The Hamiltoman #10 in table II is the onel158 JOURNAL DE PHYSIQUE I N°8
obtained
by
Zamolodchikov and Fateev (7]. This is a directland non-trivial)
extension of the well-known XXZ HarniltonianIan anisotropic Heisenberg spin Hamiltonian)
to the case of spin one. For a method of solution see theiroriginal
paper. We note that model#7
can beobtained as a
special
case of#10
in the rational limit. Sutherland [8] solved the modelN
H = q
£ (,j+i
n" +1,
j=i N
= q
£((Sj Sj+1)~
+(Sj Sj+i) I),
j=1
which is included in #4 in table
II;
a solution of this modelby
the quantum inversescattering
method is described in reference [9].Takhtajan [loi
andBabujian iii]
solved aspin
chain definedby
~
N
~
i~
~~~ ~~~~ ~~J,J+1 21),
=
£(-Sj Sj+i
+(Sj Sj+1)~ 31),
2
~
J=1
which is induded in
#7,
andKlümper
[12] solvedN
H = a
£(-(Sj Sj+1)~
+I),
j=1
which is induded in
#5.
Model#6
was solved inKlümper
[13]. For the methods of solutionswe refer ta the
original
papers.It seems that Dur models #2, #3, #8, and #9 are new
(We
wish ta thank the referee forthis
information).
Dur list of
spin
Hamiltonians in table II contains solvable t J models asspecial
cases. The t J model is a model oftight-binding
electrons withcorrelation,
and definedby
the so-called t J HamiltonianHt-j = t
£ Pc$cj~P
+h-c-)
<1,J>
+
~j (Jz (si sj)
+ Ji(sis)
+s$s))
+Vninj )
<i,J>
where
ni +
~j c$cw
= 0 or 1 def. of an action of
projection P),
«
8~ %
~ -C$OE$~icia"
~ ~i
Table III
gives
coefficients of the solvable t J Hamiltonians.Among
them the model #4 isinteresting:
for both t = +1, there are solvable cases(Jz
=
Ji,V)
=(2, -1/2), (-2,1/2),
N°8 NINETEEN-VERTEX MODELS AND SPIN CHAINS OF SPIN ONE l159
(2,3/2), (-2, -3/2);
which areisotropic
t J models solvedby
Schlottmann [14]. We note that there are other solvable cases: for both t= +1, these are
(Jz, Ji,V)
=
(2,
-2,-1/2), (-2, 2,1/2), (2, -2,3/2), (-2,
2,-3/2);
these areanisotropic
t J models. We do not knowwhether these
anisotropic
models havealready
been solved or not.Finally
we remark on thesymmetries
we haveimposed
on the R-matrices. Dur interest wasmainly
in a natural extension of the six-vertex model and the XXZ model tohigher spin
cases from theview-points
of condensed matterphysics.
Wethought
that our choice ofsymmetries
were natural for this purpose. If we were interested in a more
general
class of solvablemodels,
we had to relax some conditions on
R;
one can find aninvestigation
m this direction in Perk and Schultz[15],
where some models notpresented
in the present paper were obtained.References
[1] Kulish P-P- and Sklyanin E-K-, Lecture Notes m Pllysics 151
(1982)
fil.[2] Baxter R-J-, Exactly solved models in statistical mechamcs
(Academic
Press, London,1982).
[3] Lieb E-H-, Pllys. Rev. 162
(1967)
162; Pllys. Rev. Lent. 18(1967)
1046, 19(1967)
108.[4] Au-Yang H., Mccoy B-M-, Perk J-H-H-, Tang S. and Yan M., Pllys. Lent. A 123
(1987)
219.[5] Mccoy B-M-, Perk J-H-H-, Tang S. and Sah C.H., Pllys. Lent. A 125
(1987)
9.[fil Jimbo M. and Miwa T., Nucl. Pllys. 8257
(1985)
1.[7] Zamolodchikov A.B. and Fateev V.A., Sov. J. Nucl. Pllys. 32
(1980)
2.[8] Sutherland B., Pllys. Rev. B 12
(1975)
3795.[9] Kuhsh P-P- and Reshetikhin N.Yu, Sov. Phys. JETA 53 (1981) 108.
[loi
Takhtajan L.A., Pllys. Lent. A 87(1982)
479.[Il]
Babujian H-M-, Pllys. Lent. A 90(1982)
479; Nucl. Pllys.8215[FS7] (1983)
317.[12] Kliimper A., Europllys. Lent. 9
(1989)
815.[13] Klümper A., J. Pllys. A: Math. Gen. 23
(1990)
809.[14] Schlottmann P., Pllys. Rev. B 36
(1987)
5177.[15] Perk J-H-H- and Schultz C.L., Proceedings of RIMS Symposium 1981, Jimbo and Miwa Eds.