Supersymmetric Vertex Models with Domain Wall Boundary Conditions

0 downloads 0 Views 171KB Size Report
Jan 5, 2007 - [5] V.E. Korepin and P. Zinn-Justin, Thermodynamic limit of the ... model: Algebraic Bethe ansatz approach and ground state structure, Nucl.
arXiv:hep-th/0701042v1 5 Jan 2007

Supersymmetric Vertex Models with Domain Wall Boundary Conditions Shao-You Zhao a and Yao-Zhong Zhang a,b a

Department of Mathematics, University of Queensland, Brisbane, QLD 4072, Australia b

Physikalisches Institut, Universit¨at Bonn, D-53115 Bonn, Germany E-mail: [email protected], [email protected]

Abstract By means of the Drinfeld twists, we derive the determinant representations of the partition functions for the gl(1|1) and gl(2|1) supersymmetric vertex models with domain wall boundary conditions. In the homogenous limit, these determinants degenerate to simple functions.

Keywords: Drinfeld twist; Domain wall boundary condition; Supersymmetric vertex model.

1

Introduction

The six vertex model on a finite square lattice with the so-called domain wall (DW) boundary condition was first proposed by Korepin in [1], where recursion relations of the partition functions of the were derived. In [2, 3] it was found that the partition functions of the DW model can be represented as determinants. Taking the homogenous limit of the spectral parameters, Sogo found that the partition function satisfies the Toda differential equations [4]. Then by using the equations Korepin and Zinn-Justin obtained the bulk free energy of the system [5]. The determinant representations of DW partition functions are uesful in solving some pure mathematical problems, such as the problem of alternating sign matrices [6]. By using the fusion method, Caradoc, Foda and Kitanine obtained the determinant expressions of the partition functions for the spin-k/2 vertex models with the DW boundary conditions [7]. In [8], Bleher and Fokin obtained the large N asymptotics of the DW six-vertex model in the disordered phase. Supersymmetric integrable models based on superalgebras are important for describing strongly correlated electronic systems of high Tc superconductivity [9, 10, 11, 12, 13, 14, 15]. In this paper, we investigate the gl(1|1) and gl(2|1) supersymmetric vertex models on a N × N square lattice with DW boundary conditions. By using the approach of Drinfeld twists [16, 18, 19], we derive the determinant representations of the DW partition functions for the two systems. We find that the partition functions degenerate to simple functions in the homogenous limit. The framework of the paper is as follows. In section 2, we obtain the determinant representation of the partition function of the gl(1|1) vertex model with DW boundary condition. Then we derive the homogenous limit of the partition function. In section 3, we solve the gl(2|1) supersymmetric vertex model with the DW boundary conditions, by deriving its DW partition functions exactly. In section 4, we present some discussions.

2

gl(1|1) vertex model with DW boundary condition

In this section, we study the DW boundary condition for the gl(1|1) vertex model on a N ×N square lattice.

2

2.1

Description of the model

Let V be the 2-dimensional irreducible gl(1|1)-module and R ∈ End(V ⊗ V ) the R-matrix associated with this representation. In this paper, we choose the FB grading for V , i.e. [1] = 1, [2] = 0. The the R-matrix satisfies the graded Yang-Baxter equation (GYBE) R12 R13 R23 = R23 R13 R12 ,

(2.1)

where Rij ≡ Rij (λi , λj ) with spectral parameters specified by λi . Explicitly,   c12 0 0 0  0 a12 b12 0   R12 (λ1 , λ2 ) =   0 b12 a12 0  , 0 0 0 1

(2.2)

where

λ1 − λ2 , λ1 − λ2 + η λ1 − λ2 − η = c(λ1 , λ2 ) ≡ λ1 − λ2 + η

b12 = b(λ1 , λ2 ) ≡

a12 = a(λ1 , λ2 ) ≡ c12

η , λ1 − λ2 + η (2.3)

and η is the crossing parameter which could be normalized to 1 in the rational cases we are considering. In the following, we study the gl(1|1) 6-vertex model on a 2-d N × N square lattice. At any site of the lattice, there may be a vacuum φ or a single fermion ρ. A vertex configuration in the lattice is constructed by two nearest particle states in a horizontal line and two nearest particle states in a vertical line. For the present model, there are altogether 6 possible weights corresponding to a vertex configuration. The possible configurations and their corresponding Boltzmann weights wi are given as follows:

3

φ φ

ρ φ

φ

ρ

ρ

ρ

ρ ρ

φ

φ

φ

ρ

φ

ρ

w1

w2

w3

w4

ρ

φ

ρ

φ

φ

ρ

φ

ρ

w5

w6

Figure 1. Vertex configurations and their Boltzmann weights for the gl(1|1) vertex model.

φ φ φ φ φ

φ

φ

φ

z1 z2 . . . zN l1

ρ

l2 .. .

ρ

lN

ρ

ρ

ρ

ρ

ρ

ρ

Figure 2. DW boundary condition for the gl(1|1) vertex model.

One may check that these weights preserve the fermion numbers. If we assign horizontal lines parameters {λj } and vertical lines parameters {zk }, and let w1 = 1,

w2 = c(λ, z),

w3 = w4 = a(λ, z),

w5 = w6 = b(λ, z),

(2.4)

then the Boltzmann weights correspond to the elements of the gl(1|1) R-matrix (2.2). We now propose the gl(1|1) vertex model with DW boundary condition. Similar to [1, 3], if the ends of the square lattice satisfy the special boundary condition, that is states of all 4

left and top ends are vacuum φ while states of all right and bottom ends are single fermion ρ, we then call the boundary condition the DW boundary condition. A configuration with the DW boundary condition is shown in figure 2. In this figure, λi and zj are parameters associated with the horizontal and vertical lines, respectively. The partition function for this system is then given by ZN =

6 XY

wini ,

(2.5)

i=1

where the summation is over all possible configurations, ni is the number of configurations with the Boltzmann weight wi . By means of the R-matrix, the partition function may be rewritten as ZN =

N −1 Y

ρ(N −i)

i=0

N −1 Y

φ(N −j)

j=0

N Y N Y

Rkl (λk , zl )

N Y

ρ(j)

j=1

k=1 l=1

N Y

φ(i) .

(2.6)

i=1

Here ρ(i) (φ(i) ) stands for the state ρ (φ) in ith horizontal (vertical) line. Following Korepin, we use the transfer matrix to arrange the partition function. Define the monodromy matrix Tk (λk ) along the horizontal lines Tk (λk ) = Rk,N (λk , zN )Rk,N −1 (λk , zN −1 ) . . . Rk,1(λk , z1 ) ≡



A(λk ) B(λk ) C(λk ) D(λk )



.

(2.7)

(k)

therefore, in terms of Tk (λk ) the partition function is equal to ZN =

N −1 Y i=0

=

N −1 Y

h1|(N −i)

N −1 Y

h0|(N −j)

j=0

N Y

Tk (λk )

N Y

|1i(j)

j=1

k=1

h1|(N −i) C(λ1 )C(λ2 ) . . . C(λN )

i=0

N Y

N Y

|0i(i)

i=1

|0i(i) .

(2.8)

i=1

Here we have introduced the notation   0 , |0i = 1

|1i =



1 0



(2.9)

to denote the particle states φ, ρ, respectively.

2.2

Exact solution of the partition function

In this subsection, we will compute the partition function (2.8) for the gl(1|1) vertex model with DW boundary conditions by using the Drinfeld twist approach. 5

Let us remark that even though we believe that Drinfeld twists do exist for all algebras, so far only those related to A-type (super) algebras and to XY Z model have been explictly constructed [17, 18, 19, 20, 21]. Let σ be any element of the permutation group SN . We then define the following lowertriangular matrix [19, 20] F1,...N =

X∗

X

N Y

α

σ(j) σ P(σ(j)) S(c, σ, ασ )R1...N ,

(2.10)

σ∈SN ασ(1) ...ασ(N) j=1

α α where P(k) has the elements (P(k) )mn = δαm δαn at the kth space with root indices α = 1, 2, P∗ the sum is taken over all non-decreasing sequences of the labels ασ(i) :

ασ(i+1) ≥ ασ(i) ,

if σ(i + 1) > σ(i),

ασ(i+1) > ασ(i) ,

if σ(i + 1) < σ(i)

and the c-number function S(c, σ, ασ ) is given by ) ( N  1 X 1 − (−1)[ασ(k) ] δασ(k) ,ασ(l) ln(1 + cσ(k)σ(l) ) . S(c, σ, ασ ) ≡ exp 2 l>k=1

(2.11)

(2.12)

σ In (2.10), R12...N is the N-fold R-matrix which can be decomposed in terms of elementary

R-matrices (2.2) by the decomposition law ′



σσ σ . R1...N = Rσσ′ (1...N ) R1...N

(2.13)

We showed in [18] that the F -matrix is nondegenerate and satisfies the relation σ Fσ(1...N ) (zσ(1) , . . . , zσ(N ) )R1...N (z1 , . . . , zN ) = F1...N (z1 , . . . , zN ).

(2.14)

The nondegeneracy of the F -matrix means that its column vectors form a complete basis, which is called the F -basis. The nondegeneracy also ensures that the F -matrix is invertible. The inverse is given by [19] −1 ∗ F1...N = F1...N

Y

∆−1 ij

(2.15)

i σ(i).

(2.18)

Working in the F -basis, the entries of the monodromy matrix (2.7) can be simplified to symmetric forms, e.g. the lower entry C(λ) becomes [19] −1 ˜ C(λ) ≡ F12...N C(λ)F12...N N X 12 = b(λ, zi )E(i) ⊗j6=i diag (2a(λ, zj ), 1)(j) ,

(2.19)

i=1

αβ where E(l) (α, β = 1, 2; l = 1, 2, . . . , N) are generators of the superalgebra gl(1|1) at the site

l. From (2.19), one can see that in the F-basis all compensating terms (polarization clouds) ˜ in the original expression of C(λ) in terms of local generators disappear from C(λ). Applying the F -matrix and its inverse to the states |0i(1) ⊗ · · · ⊗ |0i(N ) and h1|(N ) ⊗ · · · ⊗ h1|(1) , we have F1...N |0i(1) ⊗ · · · ⊗ |0i(N ) = |0i(1) ⊗ · · · ⊗ |0i(N ) −1 h1|(N ) ⊗ · · · ⊗ h1|(1) F1...N = h1|(N ) ⊗ · · · ⊗ h1|(1)

Y (2a(zi , zj ))−1 . i