arXiv:cond-mat/0405155v3 [cond-mat.stat-mech] 14 Dec 2004 optto ewe iei n lcrsai nrisin- energies electrostatic and resulting kinetic The between competition repulsion. Coulomb a experiencing and tice ecie hre ooshpigi a in for Hamiltonian hopping theory Bose-Hubbard bosons effective charged The describes as [19]. and chains [18] spin higher lattices bosonic optical model in the to atoms optics, proposed quantum was In dynamics and Bose-Hubbard used [17]. it films, arrays widely physics Helium junctions and mesososcopic Josephson superconductors, [16]), In proposed granular also was contexts. describes (see different It [15] many in model. Haldane by Bose-Hubbard first the is bosons read-out. to hard cases, is physics both involve whose processes In local, order (quasi) higher though 9]. 7]; Hamiltonians [8, the 6, the however, unaltered keeping 5, formally matrix [4, scattering Hamiltonian Hamiltonian the the the change change then and alternatively, and unaltered matrix theory continuous transfer the ma- of scattering the trix keep ways: two, complementary far in somewhat circumvented involves was interaction inconvenience This “non- the to neighbors). is leads (that theory Hamiltonians field it the local” gas procedure is Bose The limit continuous continuous 14]. classical whose 13, theory 12, integrable 11, of quantization classical 10, direct a 9, since 8, straightforward studied not Bose 7, is authors 6, continuous 5, many the [4, decades of gas last regularization trans- explain the a lattice we uses In integrable [it Appendix Ansatz the matrix]. Bethe In algebraic fer the & dynamics. QISM body the dynamics two two- many-body a the the Yang-to reduces the of This satisfy it property equation. for- system: Baxter crucial the be of a matrix can scattering exploit body gas that Method Bose [3] Scattering inte- Inverse (QISM) the Quantum called of the therefore within integrability is mulated The and motion the grable. commut- of mutually integrals many, ing infinitely by possesses dimension described model one this In is Schr¨odinger equation). dynamics non-linear the classical 2] [1, corresponding bosons interacting (the strongly for models famous most nte aaimtcltiemdlfrinteracting for model lattice paradigmatic Another the of one is interaction delta with gas Bose The (2) ..Yn nttt o hoeia hsc,SaeUnivers State Physics, Theoretical for Institute Yang C.N. edsrbdb h oegswt et neato.Telatt The interaction. delta with contrasted. gas are Bose models integrable the different are by others described Bose-Hubbard be the like integrable not nvraiyo h n iesoa oegswt et inte delta with gas Bose dimensional one the of Universality ecnie eea oeso neatn oosi n dim one a in bosons interacting of models several consider We .INTRODUCTION I. (1) MATIS-INFM nvri´ iCtna il .Dra6 -52 aai,I Catania, I-95125 6, Doria A. viale Catania, Universit´a di ug Amico Luigi & d dmninllat- -dimensional iatmnod eoooi iih hmce(DMFCI), Chimiche e Fisiche Metodologie di Dipartimento (1) lattice n ldmrKorepin Vladimir and iaino o-nerblt [28]. in- non-integrability as of spacings considered dication this level [27], of statistics distribution Wigner-Dyson obeys the numerical Recent that [26]. suggested char- multi-occupancy studies statistics the bosonic to the due of [24] acteristic coordinate the fails hopping, ansatz neighbor Bethe nearest For estab- [25]. fully For is lished thermodynamics problem. the challenging dimensional hopping a densi- one range infinite still the higher is of at model solution Bose-Hubbard ones, exact the above However study the to ties. like desirable be results systems would Exact physical model [23]. Bose-Hubbard crucial are the Liniger of and exact Lieb the of where results [22] regime “Tonks-Girardeau” called so ustedffrne ewe h existing the dis- sys- between the we of differences Finally behavior the models. the cuss for Bose-Hubbard obtained used by results be phrased of can tems number gas Bose a the particular lat- for In (Bose-Hubbard) integrable Ablowitz-Ladik non models. and density tice quantum 14]) 13, low and 12, 9, universal [8, [4], as Izergin-Korepin gas [7], (Faddeev-Takhtadjan-Tarasov Bose integrable various the of limit of theory field r vldfr3 the- body Gross-Pitaevskii for the many (valid of quantum ory breakdown lattices in the optical cause example, of effects For limit case. dilute one-dimensional are the effects the technol- Quantum in different [17]. strongest a junctions with Josephson using with both and, ogy, realized [20] commonly confinement are optical chains consideration. bosonic under fact models In the by captured systems cal hs rniini usatal lee yteconfining [21]. the by potential quantum altered harmonic substantially superfluid-insulator is the transition of phase intensity criticality laser sys- the the the changing [20]; by applications tuned optical are parameters In tem repulsion. Coulomb the that eauepaedarm At diagram. tem- zero phase the dominate perature that fluctuations quantum duces ytertobtentebsnctneigrate tunneling bosonic the between ratio the controlled Mott by transition a phase predicts quantum model Bose-Hubbard insulator-superfluid the site per bosons nti ae edsusteitgal continuous integrable the discuss we paper this In nti ae efcso n ieso.W remark We dimension. one on focus we paper this In t fNwYr tSoyBok Y11794-3840 NY Brook, Stony at York New of ity d tlwdniyalo hs oescan models these of all density low At . smr hna cdmccs o h physi- the for case academic an than more is 1 = c orcin orsodn othe to corresponding corrections ice ninlltie oeo hmare them of Some lattice. ensional d (2) ytm)addietesse othe to system the drive and systems) ayand taly commensurate raction integrable ubrof number t and lat- 2

q tice regularization of the Bose gas and the non-integrable The density, given by D = −q ρ(u)du, depends on the Bose-Hubbard model. chemical potential since q = q(h) [3]. The Bose gas is R The paper is laid out as follow. In the next section we characterized by different phases depending on the sign summarize a number of properties of the integrable field of the coupling constant c. Let us describe the ground theory of the Bose gas. In section III we apply some of state at zero temperature: these properties to the Bose-Hubbard model. The lattice i) For c = 0 we have free bosons. In the ground state all integrable theories are discussed in section IV. In section the particles have zero momentum (Bose-condensation). V we draw our conclusions. In the Appendix we summa- ii) In the repulsive case c > 0 the Pauli principle is rize the main ideas of the QISM. valid (this is the distinctive feature of the one dimensional bosons). The ground state is a Fermi sphere. iii) In the attractive case c < 0 the ground state is a II. THE BOSE GAS WITH DELTA single large bound state of all the particles. All the par- INTERACTION ticles stay close to one another (in configuration space). It is a droplet. One can say that in this case the phases We shall start with continuous field theory. This will separate. be the continuous limit of all bosonic models considered In the repulsive case at zero temperature the asymp- in this paper. The one dimensional Bose gas is described totics (large distances) of correlation functions is [3, 29] by an integrable non-relativistic model of field theory [1, † −1/θ 3].The Hamiltonian is: Ψ (x)Ψ(0) = Ax (4) h i A cos [2πDx] Ψ†(x)Ψ(x)Ψ†(0)Ψ(0) = D2 + + B , (5) = dx (∂ Ψ†)(∂ Ψ)+ cΨ†Ψ†ΨΨ hΨ†Ψ . h i x2 xθ HBG x x − Z   (1) where the critical exponent θ depends on the density in Here: [Ψ(x), Ψ†(y)] = δ(x y) and [Ψ†(x), Ψ†(y)] = 0. a remarkably simple way through: The coupling constant is denoted− by c, and h is the chem- θ =2 2(q)=4πD/v . ical potential. It is equivalent to a many body quantum Z F mechanical problem with Hamiltonian: The thermal correlation functions can be obtained by re- vF placing: x sinh (πT x/vF ) in these expression. Here N 2 → πT ∂ vF is the Fermi velocity [for the model under considera- HBG = +2c δ(zj zk) . (2) − ∂z2 − tion, it coincides with the velocity of sound]. So in the j=1 j N≥j>k≥1 X X repulsive case the asymptotic of the thermal correlation The energy is E = N u2 h ; the momenta u’s obey functions is j=1 j − Bethe equations: v −1/θ P
Ψ†(x)Ψ(0) = F exp( πT x/v θ) , (6) h iT πT − F N u u + ic   2 iuj L j k 2πT e = − j =1 ...N. (3) † † −2πT x/vF u u ic Ψ (x)Ψ(x)Ψ (0)Ψ(0) T = B2 e 6 j k h i vF kY=j − −   2πT θ Let us consider repulsive case c > 0. In the thermody- + B e−πT θx/vF cos(2πDx) . (7) 3 v namic limit N,L such that N/L = D = const the  F  ground state energy-density→ ∞ [per unit length] is The coefficients A,B,B2,B3 are related to certain form q q factors [29]. E = ǫ (u)ρ(u)du = (u2 h)ρ(u)du . 0 − Z−q Z−q III. APPLICATION TO THE BOSE-HUBBARD The function ρ(u) describes the distribution of momenta MODELS in the ground state. It is defined by the equation: The Hamiltonian for the one dimensional Bose- q . 2πρ(u)=1+ K(u v)ρ(v)dv = (u) , Hubbard model reads −q − Z Z Ns † † where (u) is called dressed charge. The kernel of the HBH = [U(ni 1) µ]ni t(a ai+1 +a ai) , (8) − − − i i+1 Z 2 2 − integral equation is K(x)=2c/(c +x ). For further pur- i=XNs poses it is convenient to define a function τ(u) through † where the operators ni := ai ai count the number of q † bosons at the site i; operators ai, ai obey the canoni- 2πτ(u)= K(u q) dvK(u v)τ(v) . † − − − cal commutation relations [ai,a ]= δij and and 2Ns +1 Z−q j 3 is the number of sites. The parameters t, U of (8) are are interesting by them-self: good examples are Calogero- the hopping amplitude and the strength of the on-site Sutherland, Haldane-Shastry [30], and pairing [31, 32, 33] Coulomb repulsion, respectively, while the chemical po- models. To obtain local Hamiltonians the Lax matrices tential µ fixes the average number of bosons in each have to be modified. This modification induces complica- site. The density of bosons in the lattice is given by tions on the final form of the though local Hamiltonians D = N/(2Ns + 1)∆. Here ∆ is the lattice spacing. The (see the Refs. [4, 5, 6, 7]). All these are solvable by alge- Bose-Hubbard model can be considered as a possible lat- braic Bethe Ansatz (see Ref. [3]); some basic formulas tice regularization of the Bose gas. In fact at small filling of the QISM are summarized in Appendix. Here we will factor ν = N/(2Ns +1)= D∆ the Bose-Hubbard model present these Hamiltonians. Some of them simplify in the can be described by the Bose gas with delta interaction. weak coupling limit. In this limit the resulting Hamilto- In order to obtain the correct commutators, in the con- nians are compared with the Bose-Hubbard model. tinuous limit ∆ 0 one has to re-normalize the opera- → † tors: ai = √∆Ψ(x) and ni = ∆Ψ (x)Ψ(x), x = ∆i. The Bose-Hubbard model then reduces to the Bose gas Izergin-Korepin models

2 HBH = t∆ HBG , The first integrable version of the Bose gas on the lat- tice was constructed in Refs [4, 5]. To write the Hamil- with c = U/(t∆) and h = (µ +2t)/(t∆2). Therefore tonian of the model in a compact way we introduce the − results for the Bose gas can be phrased for systems cap- following notations. The bosons interact differently in tured by the Bose-Hubbard model. The asymptotics of odd and even lattice sites. Let us start with canonical the zero-temperature correlations in the repulsive case Bose operators are [a ,a† ]= δj ; a 0 =0 . (14) j k k j | i a†a = A(i j)−1/θ , (9) h i j i − In order to describe the model, it is convenient to intro- ν 2 A cos [2πν(i j)/∆] duce ’renormalized’ operators ninj = ( ) + 2 + B −θ ,(10) h i ∆ (i j) (i j) −1 − − bj = aj σj . (15) where the critical exponent θ is Here σj is slightly different in odd and even lattice sites: 4D 4νt θ =2 1+ =2 1+ . (11) c U †     1+ ωaj aj if j is even , σj = (16)  q † The thermal correlation functions read 1+ ω(a aj 1) if j is odd .  j − −1/θ q † vF −1/θ where ω = c∆/4. We can express everything is terms of ai aj T = sinh [πT (i j)/vF ] , (12) h i πT − ’renormalized’ operators bj :   2 2πT −2πT (i−j)/vF † † † −1 ninj T = B2 e (13) aj aj = bj bj 1 ωbjbj , (17) h i vF { − }   σ2 = 1 ωb†b −1 for j even , (18) 2πT θ j { − j j } + B e−πTθ(i−j)/vF cos [2πν(i j)/∆] . 3 v − For odd j the expressions are slightly different:  F  a†a = (1 ω)b†b 1 ωb†b −1 , (19) j j − j j{ − j j } IV. INTEGRABLE LATTICE MODELS σ2 = (1 ω) 1 ωb†b −1 for j odd . (20) j − { − j j} We can also write a closed commutation relations for In this section we discuss the integrable bosonic theo- renormalized operators: ries arising as quantum lattice regularization of the field 1 1 † −1 † † − 2 † † − 2 theory of the Bose gas with delta interaction. From bj 1 ωb bj b 1 ωb bj b bj 1 ωb bj { − j } j −{ − j } j { − j } a classical side the Hamiltonian structure [action-angle 1 if j is even = (21) variables] is directly related to the R-matrix. This con- (1 ω)−1 if j is odd . stituted motivation to define the lattice models keeping  − unaltered the R-matrix (for an alternative way to con- In different lattice sites operators bj commute. Now we struct lattice regularization of the Bose gas see the Ap- can present the Hamiltonian of the model as pendix [8, 9]). Pursuing a quantum version of this proce- † 4 2 ω ∆a aj dure, namely directly quantizing the classical Lax matri- H = t + t† + − j . IK −3c∆3 j j 2(1 ω) − 1 ω2 ces, it turns out that the arising Hamiltonians are non- j ! X − − local. Here we should note that non-local Hamiltonians (22) 4

Below, we shall represent operators tj in the form Faddeev-Takhtadjan-Tarasov models

1 −1 In this section we discuss another lattice integrable ver- tj = Oj Cj Oj . (23) −2 sion of the Bose gas. The Hamiltonian was suggested by Faddeev-Takhtadjan-Tarasov in Ref. [7]. The model has The formula for operators O and C are different for odd the same R-matrix as the continuous Bose gas. It looks and even lattice sites. For j odd: like a quantum spin chain with negative spin. Let us first introduce an operator of ’angular-momentum’ J . It O = σ 1+ ωb† b σ , (24) j,j+1 j j+1{ j+1 j+2} j+2 is defined as a solution of the following operator equation: −1 † −1 −2 † −1 C = σ 1+ ωb b σ 1+ ωb b − j j { j−1 j} j−1{ j j 1} × −1 † Jj,j+1(Jj,j+1 +1)=2Sj Sj+1 +2s(s + 1). σ σ − 1 ωb b σ ,. (25) ⊗ j j 1{ − j−1 j+1} j+1 Here the ’spin’ is s = 2/(c∆). The Hamiltonian of the For j even: Faddeev-Takhtadjan-Tarasov− model is

† Oj = σj−2 1+ ωb − bj−1 σj−1 , (26) ′ { j 2 } HF T T = 2κ Γ (Jj,j+1 + 1)/Γ(Jj,j+1 + 1) . −1 † −1 −2 † −1 − C = σ 1+ ωb b − σ 1+ ωb b j j j−1{ j j 1} j { j−1 j } × X 1 ωb† b σ . (27) { − j−1 j+1} j+1 The relation with the lattice bosons becomes transparent using the Holstein-Primakov realization of spins: In the weak coupling limit the Hamiltonian describes a coupling of five neighbors lattice sites j 2 ...j +2. Re- a†ρ + ρ a − x j j j j taining contributions up to second order in c∆ it has the Sj = , √c∆ form † a ρj + ρj aj Sy = i− j , (32) 4 c∆ j √c∆ HIK = 3 (Kj,j−1 nj Kj−1,j+1) (28) −3c∆ 8 − − † j 2 c∆a aj X Sz = 1+ j , (c∆)2 j c∆ 2 + (K2 + h ) , − ! 16 j,j−1 j 2 with ρ = 1+ c∆a†a/4. For small c, the spin s is large. h = v n n + w a a† (29) j αβ j+α j+β αβ j+α j+β The Hamiltonian simplifies (second order in 1/s): α,β=−2 p X h † † 2 + (t + q n ) r a a + s a a + h.c. , κ s † † α α j+α β j+β j+β−1 β j−β j+β H = (a a )(a a ) F T T s3 4 j+1 − j j+1 − j  i j  † † s X where Kl,m = alam + amal (with certain coefficients † 2 † † † 2 2 2 [(aj ) + aj+1aj + (aj+1) ][aj + aj+1aj + aj+1] vαβ , wαβ , tαβ , qαβ). After simple manipulations −16 s Eq. (28) can be written as + a† a†a a s2a†a . (33) 16 j+1 j j+1 j − j j o c∆ c∆ 2 2 H = H g K − + g( ) K + h , The model is solvable by algebraic Bethe Ansatz [7]. Af- IK BH − 8 j 1,j+1 4 j,j−1 j ter simple manipulations Eq.(33) can be written as   (30) H = H (34) where periodic boundary conditions have been consid- F T T BH 1 ered: +V n n + K2 + (n + n )K , j j+1 2 j,j+1 j j+1 j,j+1 j   2 2 X nj = nj+1 and nj = nj+1 ; (31) j j j j where periodic boundary conditions have been consid- X X X X ered (see Eqs.(31)); the parameters in HBH above are 2 the parameters in HBH above are µ/g = (c∆)(1 µ = κ(c∆)(c∆/16 + 1)/4, U = κ(c∆) /64, t = κc∆/8, 5c∆/16)/8, U/g = 5(c∆)2/128, t/g = c∆/8 with− and V = κ(c∆)2/64. Eq. (34) elucidates the rela- g = 4/(3c∆3). The Bose-Hubbard model differs from tion between the Faddeev-Takhtadjan-Tarasov and Bose- Itzergin-Korepin− model in the presence of the quadratic Hubbard models. In the continuous limit, when lattice hopping and in the non local terms. In the continuous spacing ∆ 0 the Faddeev-Takhtadjan-Tarasov (and limit, when the lattice spacing ∆ 0, this lattice model the Bose-Hubbard→ model as well) lattice model turns into turns into a continuous Bose gas (1).→ the continuous Bose gas (1). 5

V. CONCLUSIONS. is the analog of the Casimir operator of Lie algebras and it is another invariant of the Yang-Baxter algebra. In Eq. (37) is the projector We considered several models of interacting bosons. P Faddeev-Takhtadjan-Tarasov model and Izergin-Korepin 00 00 model are different lattice models originally constructed 1 0 e−η 1 0 from the Bose gas continuous field theory. For these mod- = R(η; η) :=  −  . (38) P 2 cosh η 0 1 eη 0 els the lattice contributions do not destroy the integrabil- −  00 00  ity of the original field theory; instead the Bose Hubbard     model contains lattice-effects leading to a non integrable To construct lattice integrable theories for the Bose dynamics. In the weak coupling limit these integrable gas two different procedures (somehow complementary) Hamiltonians and the Bose-Hubbard Hamiltonian can be have been pursued. The first consists in keeping the R- compared explicitly. In particular the model in Eq.(34) matrix of the continuous theory unaltered and changing could serve exact studies of chains of Josephson junctions the Lax operators. Such procedure has been followed in away from the degeneracy point [17]. The lattice effects the text above. One can also change both the R and vanish at low density: all these theories are integrable L operators to simplify the final form of the Hamilto- and described by the Bose gas with delta interaction. In nian. This approach was initiated by P.P. Kulish et al. this regime the known asymptotic of the Bose gas correla- [8, 9] and developed in the papers [12, 13, 14], quantizing tion functions can be used for physical systems captured classical discretization on non-linear Schroedinger equa- by the Bose-Hubbard model (see [38] for a recent applica- tion developed earlier by Ablowitz and Ladik [35]. For tion). The Luttinger liquid field theory in Ref. [36] and the Ablowitz-Ladik model the Lax operators and the R- the Bose gas are characterized by the same set of correla- matrix [8, 9] are : tion functions where the spectral parameters vJ , vF , vN 10 00 are determined by [37] θ =2 vJ /vN , vF = √vN vJ , and ζ q − ′ q ′ i 0 b c 0 v = ǫ (q) duǫ (u)τ(u) /[ (q)] (see section II). Li(ζ) := † , R(φ; η) :=  +  . F 0 −q 0 p q ζ−1 0 c b 0 − Z  i  h i −  00 01  R   Discussions with R. Fazio, P. Kulish and A. Osterloh  (39) are acknowledged. Here exp φ = ζ/ξ and ζ, ξ are spectral parameters, e±η sinh φ b± = , APPENDIX sinh (φ η) − The construction of integrable lattice models of inter- sinh η c = , acting bosons is based on Quantum Inverse Scattering −sinh (φ η) Method (QISM) [3]. Here we shall summarize some fea- − tures of this method. The starting point of the QISM is the quantity η R is a deformation parameter. The Hamiltonian of quantum∈ Ablowitz-Ladik can be defined a local quantum Lax operators Li(u) and a matrix R(u), satisfying the Yang Baxter equation by

Ns R(φ)Li(ζ) Li(ξ)= Li(ξ) Li(ζ)R(φ) . H = q†q + q† q α log(1 + q†q ) . ⊗ ⊗ AL − j j+1 j+1 j − j j − j=XNs h i The monodromy matrix (40) We shall see later that α doesn’t play the role of coupling T (ζ)= L−N (ζ)L−N +1(ζ) ...LN (ζ) , (35) s s s constant. In fact the potential and the kinetic energies also fulfills the Yang-Baxter relation commute (then the potential energy turns into a con- stant on the eigenstates of the kinetic term of (40)). The R(φ)T (ζ) T (ξ)= T (ξ) T (ζ)R(φ) . (36) operators q satisfy the following commutation relation: ⊗ ⊗ j The transfer matrix is defined as t(ζ) := tr T (ζ) where † 2η † i (0) [qi, qj ] = (e 1) 1+ qj qj δj (41) tr means the trace in the auxiliary space. It is a gener- − (0)   ating functional of integrals of motion and of the Hamil- as follows the Yang-Baxter algebra. The Hamilto- tonian since it commutes with itself at different values nian (40) can be rewritten using the trace identities. In of spectral parameters: [t(ζ),t(ξ)] = 0 (t(ζ) is an invari- particular the hopping term of the Ablowitz-Ladik model ant of the Yang-Baxter algebra). The quantum determi- is nant [34] 2Ns+1 † t(ζ) ζ qj qj+1 = lim − , (42) −η − ζ→∞ ζ2Ns−1 detq(T (ζ) := Tr T (ζ) T (e ζ) , (37) j P ⊗ X 6 and [2] C.N.Yang, C.P. Yang. Journ. of Math. Phys. 10, 1115 (1969) −2Ns−1 † t(ζ) ζ [3] V.E. Korepin, N.M. Bogoliubov, and A.G. Izergin, Quan- qj qj+1 = lim − , (43) − ζ→0 ζ1−2Ns tum Inverse Scattering Method and Correlation Func- j X tions, (Cambridge Univ. Press, Cambridge 1993). whereas the interaction is related to the quantum deter- [4] V.E. Korepin and A.G. Izergin, Nucl. Phys. B 205, 401 minant (1982). [5] V.E. Korepin and A.G. Izergin, Doklady Akad. Nauk Ns 259, 76, (1981). † [6] A. Kundu and O. Ragnisco, J. Phys. A 27, 6335 (1994). ln detq(T )= log(1 + qj qj )+(2Ns + 1)η . (44) [7] V.O.Tarasov, L.A. Takhadzhyan and L.D. Faddeev, The- i=−Ns X oretical and Mathematical Physics 57, 163 (1983). In simple words: the hopping term (the kinetic energy) [8] P. Kulish, Lett. Math. Phys. 5, 191 (1981). commutes with the interaction since the transfer ma- [9] V.S. Gerdzikov, M.I. Ivanov, and P.P.Kulish, J. Math. trix and the quantum determinant commute each other. Phys. 25, 25 (1984). [10] N.M. Bogoliubov, R.K.Bullough, Phys. Lett. A 168 , 264 Eigenvectors of the transfer matrix can be constructed (1992) by algebraic Bethe Ansatz (see Refs. [8, 12] or Chapter [11] N.M. Bogoliubov, R.K.Bullough, Journ. Phys. A: Math. VII of the book [3]). Let us denote the off-diagonal ele- Gen. 25 , 4057, (1992) ment of monodromy matrix (35) by C(ζ)= T21(ζ). The [12] N.M. Bogoliubov, R.K.Bullough and G.D.Pang. Phys. eigenvectors of monodromy matrix and Hamiltonian can Rev B 47 , 11495 (1993) be written as [13] N.M. Bogoliubov, R.K.Bullough and Timonen, Phys. Rev. Lett. 72 , 3933 (1994) n [14] N.M. Bogoliubov, A.G.Izergin and N.A. Kitanine, Nucl. χ = C(ζk) 0 . (45) Phys. B 516 , 501, 1998 | i | i k=1 [15] F.D.M. Haldane, Phys. Lett. A 80, 281 (1980) Y [16] M.P.A. Fisher, P.B. Weichman, G. Grinstein, and D.S. Here 0 > is the Fock vacuum q 0 = 0. The eigenvalue | j | i Fisher, Phys. Rev. B 40, 546 (1989). of monodromy matrix θ(ζ) is [17] R. Fazio and H. van der Zant, Phys. Rep. 355, 235 (2001). n − − θ(ζ)= ζ2Ns+1enη sinh(λ λk η) [18] D. Jaksch, C. Bruder, J.I. Cirac, C.W. Gardiner, and P. k=1 sinh(λ−λk ) Zoller, Phys. Rev. Lett. 81, 3108 (1998). n − + ζ−2Ns−1enη Q sinh(λ λk +η) . (46) k=1 sinh(λ−λk ) [19] J. J. Garcia-Ripoll, M. A. Martin-Delgado, and J. I. Cirac, cond-mat/0404566. Q Here exp(λ) = ζ (the same for exp(λk)= ζk). The vari- [20] M. Greiner et al., Nature 415, 39 (2002). ables λk have to satisfy the Bethe equations [21] G.G. Batrouni, et al., Phys. Rev. Lett. 89, 117203 (2002). [22] M. Girardeau, J. Math. Phys. 1, 516 (1960); Phys. Rev. n 139, B500 (1965). 2(2N +1)λ sinh(λa λk + η) e s a = − . (47) [23] E.H. Lieb, R. Seiringer, and J. Yngvason, Phys. Rev. − sinh(λa λk η) k=1 Lett. 91, 150401 (2003). Y − − [24] F.D.M. Haldane, Phys. Lett. A 81, 545 (1981); T.C. The vectors (45) are also eigenvectors of the logarithm Choy, Phys. Lett. A 80, 49 (1980); T.C. Choy and F.D.M. of the quantum determinant (44) with eigenvalues equal Haldane Phys. Lett. A 90, 83 (1990). to (2NS + 1)η. The energy levels [ eigenvalues of the [25] J.-B. Bru and T. Dorlas, J. Stat. Phys. 113, 177 (2003). Hamiltonian] are [26] B. Sutherland, Exactly solvable problems in condensed matter and relativistic field theory, Lecture notes in n physics 242, eds S. Shastry, S.S. Jha, and V. Singh 2(1 e2η) cosh2λ . (48) (Springer Berlin, 1985). − k k=1 [27] A.R. Kolovsky and A. Buchleitner, cond-mat/0403213. X [28] G. Montambaux et al., Phys. Rev. Lett. 70, 497 (1993). The operators qj can be represented in terms of the stan- [29] N.M. Bogoliubov, A.G. Izergin, and N.Yu. Reshetikhin, † n dard Bose operators [an,am]= δm J. Phys. A 20, 5361 (1987). [30] A.P. Polychronakos, in Topological aspects of low di- exp2η(a†a + 1) 1 mensional systems, Les Houches lectures notes (EDP, q = † − a . Springer, 1998). s 1+ a a [31] R.W. Richardson and N. Sherman, Nucl. Phys. 52, 221 (1964); 52, 253 (1964). The correlation functions of the model were evaluated in [32] L. Amico, A. Di Lorenzo, and A. Osterloh, Phys. Rev. [13, 14]. Lett 86, 5759 (2001); Nucl. Phys. B 614, 449 (2001). [33] J. Dukelsky, S. Pittel, and G. Sierra, nucl-th/0405011. [34] A.G. Izergin and V.E. Korepin, Sov. Phys. Dokl. 26 (7), 653 (1981). [35] Ablowitz and Ladik, J. Math. Phys. 16, 598 (1975); ibid. [1] E.H.Lieb and W.Liniger Phys. Rev. 130, 1605 (1963). 17, 1011 (1976). 7

[36] F.D.M. Haldane, Phys. Rev. Lett. 47, 1840 (1981). cond-mat/0310388. [37] F.D. Haldane, Phys. Lett. 81, 153 (1981). [38] C. Kollath, U. Schollw¨ock, J. von Delft, and W. Zwerger,