1: % August 28

2: \documentclass[12pt]{article}

3: %\usepackage{showkeys}

4: \usepackage{epsf,amssymb,psfrag}

5: %\usepackage{hyperref}

6: 

7: \catcode`\@=11

8: %-------------------------------------------------------------

9: \textwidth 173mm

10: \textheight 235mm

11: \topmargin -50pt

12: \oddsidemargin -0.45cm

13: \evensidemargin -0.45cm

14: %-------------------------------------------------------------

15: \def \thesection {\arabic{section}.}

16: \def \thesubsection {\thesection\arabic{subsection}.}

17: \def \thesubsubsection {\thesubsection\arabic{subsubsection}.}

18: %-------------------------------------------------------------

19: \def \be  {\begin{equation}}

20: \def \ee  {\end{equation}}

21: \def \ba  {\begin{eqnarray}}

22: \def \ea  {\end{eqnarray}}

23: \def \baa {\begin{eqnarray*}}

24: \def \eaa {\end{eqnarray*}}

25: \def \bb  {\begin {thebibliography} }

26: \def \eb  {\end{thebibliography}}

27: %\def \lab #1 {\label{#1}               \mbox{\# ${#1}$}}

28: \def \lab #1 {\label{#1}}

29: \newcommand \ci [1] {\cite{#1}}

30: \newcommand \bi [1] {\bibitem{#1}}

31: \newcommand\re[1]{({\ref{#1}})}

32: \def \qqquad {\qquad\quad}

33: \def \qqqquad {\qquad\qquad}

34: %-------------------------------------------------------------

35: \def \matrix #1 {\left(\begin{array}{cc} #1 \end{array}\right)}

36: \def \Tr {\mathop{\rm Tr}\nolimits}

37: \def \tr {\mathop{\rm tr}\nolimits}

38: \def \Im {\mathop{\rm Im}\nolimits}

39: \def \Re {\mathop{\rm Re}\nolimits}

40: \def \res{\mathop{\rm res}\nolimits}

41: \def \e  {\mathop{\rm e}\nolimits}

42: \newcommand\lr[1]{{\left({#1}\right)}}

43: \newcommand \widebar [1] {\overline{#1}}

44: \newcommand\bin[2]{\left({#1}\atop{#2}\right)}

45: \newcommand \vev [1] {\langle{#1}\rangle}

46: \newcommand \VEV [1] {\left\langle{#1}\right\rangle}

47: \newcommand \ket [1] {|{#1}\rangle}

48: \newcommand \bra [1] {\langle {#1}|}

49: \newcommand \partder [1] {{\partial \over\partial #1}}

50: \newcommand{\as}{\ifmmode\alpha_{\rm s}\else{$\alpha_{\rm s}$}\fi}

51: \newcommand{\asbar}{\ifmmode\bar{\alpha}_{\rm s}\else{$\bar{\alpha}_{\rm s}$}\fi}

52: \def \CO {{\cal O}}

53: \def \CP {{\cal P}}

54: \def \CT {{\cal T}}

55: \def \CM {{\cal M}}

56: \def \CK {{\cal K}}

57: \def \CH {{\cal H}}

58: \def \CI {{\cal I}}

59: \def \CV {{\cal V}}

60: \def \CJ {{\cal J}}

61: \def \CL {{\cal L}}

62: \def \CR {{\cal R}}

63: \def \CD {{\cal D}}

64: 

65: \def \G  {\Gamma}

66: \def \D  {\Delta}

67: 

68: \font\cmss=cmss12 \font\cmsss=cmss10 at 11pt

69: \def\inbar{\,\vrule height1.5ex width.4pt depth0pt}

70: \def\IC{\relax\hbox{$\inbar\kern-.3em{\rm C}$}}

71: \def\IZ{\relax{\hbox{\cmss Z\kern-.4em Z}}}

72: \def\IR{{\hbox{{\rm I}\kern-.2em\hbox{\rm R}}}}

73: \def\R{{\tiny \IR}}

74: \def\IP{{\hbox{{\rm I}\kern-.2em\hbox{\rm P}}}}

75: \def\II{\hbox{{1}\kern-.25em\hbox{l}}}

76: 

77: \def\numberbysection{\@addtoreset{equation}{section}

78:                      \def\theequation{\thesection\arabic{equation}}}

79: \numberbysection

80: 

81: \newcommand \mybf[1] {\mbox{\boldmath$\scriptstyle {#1} $}}

82: \newcommand \Mybf[1] {\mbox{\boldmath$ {#1} $}}

83: \newbox\lett\newdimen\lheight\newdimen\lwidth

84: \def\ontop#1#2{\setbox\lett=\hbox{#2}\lheight\ht\lett

85: \multiply\lheight by 12 \divide\lheight by 10\relax%

86: \lwidth\wd\lett \multiply\lwidth by 8 \divide\lwidth by 10\relax #2\kern-\lwidth%

87: \raise\lheight\hbox{{$\scriptstyle #1$}}\kern.1ex}

88: 

89: %%%%%%%%%%%% Figures labels %%%%%%%%%%%%%%%%

90: 

91: 

92: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

93: 

94: 

95: \begin{document}

96: 

97: \begin{titlepage}

98: \begin{flushright}

99: \begin{tabular}{l}

100: LPT--Orsay--03--61\\

101: RUB-TP2-12/03\\

102: hep-th/0309144

103: \end{tabular}

104: \end{flushright}

105: 

106: \vskip3cm

107: \begin{center}

108:   {\large \bf Baxter $\mathbb{Q}-$operator and Separation of Variables \\[3mm]

109: for the open $SL(2,\mathbb{R})$ spin chain }

110: 

111: \def\thefootnote{\fnsymbol{footnote}}%

112: \vspace{1cm}

113: {\sc S.\'{E}. Derkachov}${}^1$, {\sc G.P.~Korchemsky}${}^2$

114:  and {\sc A.N.~Manashov}${}^3$\footnote{ Permanent

115: address:\ Department of Theoretical Physics,  Sankt-Petersburg State University,

116: St.-Petersburg, Russia}

117: \\[0.5cm]

118: 

119: \vspace*{0.1cm} ${}^1$ {\it

120: Department of Mathematics, St.-Petersburg Technology Institute,\\

121: St.-Petersburg, Russia

122:                        } \\[0.2cm]

123: \vspace*{0.1cm} ${}^2$ {\it

124: Laboratoire de Physique Th\'eorique%

125: \footnote{Unite Mixte de Recherche du CNRS (UMR 8627)},

126: Universit\'e de Paris XI, \\

127: 91405 Orsay C\'edex, France

128:                        } \\[0.2cm]

129: \vspace*{0.1cm} ${}^3$

130:  {\it

131: Institut f\"ur Theoretische Physik II, Ruhr-Universit\"at Bochum,\\

132: D-44780 Bochum, Germany}

133: 

134: \vskip2cm

135: 

136: {\bf Abstract:\\[10pt]} \parbox[t]{\textwidth}{

137: We construct the Baxter $\mathbb{Q}-$operator and the representation of the

138: Separated Variables (SoV) for the homogeneous open $SL(2,\mathbb{R})$ spin chain.

139: Applying the diagrammatical approach, we calculate Sklyanin's integration measure

140: in the separated variables and obtain the solution to the spectral problem for

141: the model in terms of the eigenvalues of the $\mathbb{Q}-$operator. We show that

142: the transition kernel to the SoV representation is factorized into the product of

143: certain operators each depending on a single separated variable. As a

144: consequence, it has a universal pyramid-like form that has been already observed

145: for various quantum integrable models such as periodic Toda chain, closed

146: $SL(2,\mathbb{R})$ and $SL(2,\mathbb{C})$ spin chains.} \vskip1cm

147: 

148: \end{center}

149: 

150: \end{titlepage}

151: 

152: \newpage

153: \tableofcontents

154: 

155: %\newpage

156: \setcounter{footnote}{0}

157: 

158: 

159: \section{Introduction}

160: 

161: Recently it has been found that the evolution equations describing the scale

162: dependence of certain correlation functions in four-dimensional Yang-Mills theory

163: possess a hidden symmetry. Remarkably enough, the emerging integrable structures

164: are well-known in the theory of lattice integrable models \cite{QISM} as

165: corresponding to open Heisenberg spin magnets. In particular, the energy spectrum

166: of the magnet determines the spectrum of the anomalous dimensions of the

167: correlation functions in Yang-Mills theory~\cite{BDM,AB,DKM99}. A unusual feature

168: of these models as compared with conventional magnets studied thoroughly in

169: applications to statistical physics~\cite{Baxter} is that the spin operators are

170: generators of infinite-dimensional representations of the $SL(2,\mathbb{R})$

171: group. This group emerges as subgroup of the full $SO(4,2)$ conformal group of

172: four-dimensional Yang-Mills theory.

173: 

174: Exact solution of the spectral problem for integrable systems with

175: infinite-dimensional quantum space is a nontrivial task. The conventional

176: Algebraic Bethe Ansatz (ABA)~\cite{ABA} is not always applicable to such systems

177: and one has to rely instead on a more elaborated methods like the Baxter

178: $\mathbb{Q}-$operator~\cite{Baxter} and the Separation of Variables

179: (SoV)~\cite{Sklyanin}. Being combined together, the two methods allow one to find

180: the energy spectrum of the model and obtain integral representation for the

181: eigenstates. At present, such program has been carried out for a number of models

182: with periodic boundary conditions. They include periodic Toda chain~\cite{PG,KL},

183: the DST model~\cite{KSS}, noncompact closed $SL(2)$ Heisenberg

184: magnets~\cite{SD,DKM,SoV} and Calogero-Sutherland model~\cite{KMS}. In the

185: present paper, we apply the both methods to the quantum $SL(2,\mathbb{R})$ open

186: Heisenberg spin chain.

187: 

188: A systematic approach to building quantum integrable models with nontrivial

189: boundary conditions (including open Heisenberg spin chains)

190: %within the Quantum Inverse Scattering Method (QISM)

191: has been developed by Sklyanin~\cite{Sklyanin88}. For such models a little

192: progress has been made in constructing the $\mathbb{Q}-$operator and the SoV

193: representation. One of the reasons for this is that the $R-$matrix formulation is

194: more cumbersome in that case as compared to the models with periodic boundary

195: conditions and, in addition, there exist no regular procedure for obtaining the

196: $\mathbb{Q}-$operator.

197: 

198: In this paper, we construct the Baxter $\mathbb{Q}-$operator and representation

199: of the Separated variables for the quantum $SL(2,\mathbb{R})$ open spin chain.

200: Our analysis is based on the Feynman diagram approach described at length in

201: previous publications~\cite{DKM,SoV}. In this approach, one realizes the

202: $\mathbb{Q}-$operator as an integral operator acting on the quantum space of the

203: model and represents its kernel as a certain Feynman diagram. Then, various

204: properties of the $\mathbb{Q}-$operator can be established by making use of a few

205: elementary diagrammatical relations. Using the obtained expressions, we determine

206: the energy spectrum of the open Heisenberg spin chain in terms of the eigenvalues

207: of the $\mathbb{Q}-$operator and obtain integral representation for the

208: eigenfunctions.

209: 

210: The presentation is organized as follows. In Section~2 we define the open

211: Heisenberg magnet with the $SL(2,\mathbb{R})$ spin symmetry and review Sklyanin's

212: formulation of the model. In Section~3 we construct Baxter $\mathbb{Q}-$operator

213: for the homogeneous open spin chain and establish its properties. In Section~4 we

214: present an explicit construction of the unitary transformation to the Separated

215: Variables for the open $SL(2,\mathbb{R})$ spin chain. In particular, we calculate

216: the integration measure defining the scalar product in the SoV representation and

217: discuss its

218: analytical properties. %In Section~5 we establish the relation between the

219: %$Q-$operator and the transition function to the SoV representation.

220: In Section~5 we demonstrate that for the open spin chain with two sites the

221: eigenvalues of the $\mathbb{Q}-$operator coincide with the Wilson orthogonal

222: polynomials. Section~6 contains concluding remarks. Some technical details and

223: description of the diagrammatical technique are given in the Appendix.

224: 

225: \setcounter{equation}{0}

226: \section{Open Heisenberg spin chain}

227: \label{Open}

228: 

229: \subsection{Definition of the model}

230: 

231: The homogeneous open Heisenberg spin chain is a lattice model of $N$ interacting

232: spins $\vec S_n=(S_n^1,S_n^2,S_n^3)$ (with $n=1,...,N$) described by the

233: Hamiltonian

234: \be\label{H}

235: \CH_N = \sum_{n=1}^{N-1} H_{n,n+1}, \qqqquad

236: H_{n,n+1}~=~2\left[\psi(J_{n,n+1})-\psi(2s)\right]\,,

237: \ee

238: where $\psi(x)=d\log\Gamma(z)/dz$ is the Euler $\psi-$function. The pairwise

239: Hamiltonian $H_{n,n+1}$ defines the interaction between two neighboring spins

240: $\vec S_n$ and $\vec S_{n+1}$. It is expressed in terms of the operator

241: $J_{n,n+1}$ related to their sum

242: \be\label{JJ}

243: J_{n,n+1}(J_{n,n+1}-1)=(\vec{S}_n+\vec{S}_{n+1})^2\,.

244: \ee

245: The spin operators in different sites commute with each other and obey the

246: standard commutation relations

247: %and the corresponding quadratic Casimir operator is defined as

248: \be

249: [S_n^a,S_k^b]=i\varepsilon_{abc}\delta_{nk} S_n ^c\,,\qquad \vec S^2_n%\equiv

250: %\sum_{a,b} \delta_{ab} S_n^a S_n^b

251: = s_n(s_n-1)\,.

252: \ee

253: We shall assume for simplicity that the spin chain is homogeneous,

254: $s_1=...=s_N=s$, with real $s\ge 1/2$ the same as in \re{H}.

255: 

256: Notice that the Hamiltonian \re{H} does not involve interaction between the

257: boundary spins $\vec S_1$ and $\vec S_N$. If one added the corresponding

258: two-particle Hamiltonian $H_{N,1}$ to the r.h.s.\ of \re{H}, the resulting

259: Hamiltonian would define a homogeneous closed Heisenberg spin chain. The latter

260: model admits solution within the $R-$matrix approach both by the ABA

261: method~\cite{ABA} and the methods of the Baxter $\mathbb{Q}-$operator~\cite{SD}

262: and SoV~\cite{SoV}. In the present paper we extend the analysis performed in the

263: papers~\cite{SD,SoV} to the case of the open spin chain and apply the method of

264: the Baxter $\mathbb{Q}-$operator to solve

265: %present the solution of

266: the spectral problem for the Hamiltonian \re{H}

267: \be

268: \mathcal{H}_N \Psi_{\mybf{q}}(z_1,\ldots,z_N) =

269: E_{\mybf{q}}\Psi_{\mybf{q}}(z_1,\ldots,z_N)\,.

270: \label{Sch}

271: \ee

272: Here $\Mybf{q}$ denotes the complete set of the quantum numbers parameterizing

273: the energy spectrum  and $z_n$ (with $n=1,\ldots,N$) are the coordinates on the

274: quantum space $V_n$ associated with the $n$th site of the spin chain.

275: 

276: The Hamiltonian $\mathcal{H}_N$ acts on the quantum space of the model

277: $\mathcal{V}_N=\prod_{n=1}^N \otimes V_n$

278: and its energy spectrum depends on the choice of the Hilbert space $V_n$.

279: In what follows we shall assume that

280: $\mathcal{V}_N$ is spanned by functions $\Psi(z_1,\ldots,z_N)\in \mathcal{V}_N$

281: holomorphic in the upper half-plane $\Im z_n>0$ and normalizable with respect to

282: the scalar product

283: \begin{equation}

284: \vev{\Psi_1|\Psi_2}~=~\int\mathcal{D}^N z\,

285: \left(\Psi_1(z_1,\ldots,z_N)\right)^*\,\Psi_2(z_1,\ldots,z_N)\,,

286: \label{norm}

287: \end{equation} where integration measure is defined as $\mathcal{D}^N

288: z=\prod_{n=1}^N \mathcal{D} z_n$ with ($z_n=x_n+iy_n$)

289: \be

290: \mathcal{D}z_n

291: =\frac{2s-1}{\pi}\,d^2z_n \, (2\Im z_n)^{2s-2}\theta(\Im z_n)

292: =\frac{2s-1}{\pi}\,dx_ndy_n\,(2y_n)^{2s-2}\theta(y_n)

293: \label{measure}

294: \ee

295: and integration in \re{norm} goes over the upper half-plane. The spin operators

296: $\vec S_n$ can be realized on this space as differential operators%

297: \footnote{In Yang-Mills theory the spin operators \re{s-rep} define

298: representation of the generators of the collinear $SL(2,\mathbb{R})$ subgroup of

299: the full $SO(4,2)$ conformal group on the space of correlation functions

300: $\vev{0|\Phi_s(z_1 n)\ldots \Phi_s(z_N n)|0}$ of primary fields with conformal

301: spin $s$ and ``living'' on the light-cone $n_\mu^2=0$.}

302: \begin{equation}

303: \label{s-rep}

304: S_n^{+}~=~z_n^2\partial_{z_n}+2 s\,z_n,\ \ \

305: S_n^{-}~=~-\partial_{z_n},\ \ \

306: S_n^{0}~=~z_n\partial_{z_n}+ s\,.

307: \end{equation}

308: where $S_n^\pm =S_n^1\pm iS_n^2$ and $S_n^0=S_n^3$. These operators are

309: anti-hermitian with respect to the scalar product \re{norm}

310: \be

311: (S_n^0)^\dagger = -S_n^0\,,\qquad(S_n^\pm)^\dagger = -S_n^\pm\,.

312: \ee

313: %while the Hamiltonian of the model \re{H} is hermitian, $\mathcal{H}_N^\dagger

314: %=\mathcal{H}_N$.

315: Notice that the quantum space of the model is

316: infinite-dimensional for arbitrary finite $N$. For integer and half-integer $s$,

317: the Hilbert space $V_n$ coincides with the representation space of unitary

318: representation of the $SL(2,\mathbb{R})$ group of the discrete series~\cite{Gelfand}.

319: 

320: 

321: \subsection{$R-$matrix formulation}

322: 

323: The open $SL(2,\mathbb{R})$ Heisenberg spin magnet \re{H} is a completely

324: integrable model. To identify its integrals of motion we follow Sklyanin's

325: approach~\cite{Sklyanin88}. To begin with, one defines the Lax operator for the

326: $SL(2,\mathbb{R})$ magnet

327: \begin{equation}

328: \label{Lax}

329: L_n(u)=u+i (\vec\sigma\cdot  \vec{S}_n)= \left(\begin{array}{cc}u+iS^0_n&

330: iS^-_n\\iS^+_n&u-iS^0_n\end{array}\right),

331: \end{equation}

332: where $\vec \sigma=(\sigma_1,\sigma_2,\sigma_3)$ are the Pauli matrices.

333:  It acts on the tensor product of the auxiliary space and the quantum space

334: in the $n$th site, ${\mathbb C}^2\otimes{V_n}$. Taking the product of $N$ Lax

335: operators along the spin chain in the auxiliary space one defines the operator

336: %\footnote{This operator coincides with the monodromy matrix for the closed spin

337: %chain.}

338: -- the monodromy matrix for the closed spin chain

339: \be\label{Tc}

340: T_N(u)~=~L_1(u)\ldots

341: L_N(u)~=~\left(\begin{array}{cc}a(u)&b(u)\\

342: c(u)&d(u)

343: \end{array}\right),

344: \ee

345: which is a $2\times 2$ matrix with the entries $a(u),\ldots,d(u)$ being operators

346: acting on $\mathcal{V}_N$. It satisfies the Yang-Baxter commutation relations

347: \be\label{YBR}

348: R_{12}(u-v)\ontop1T_N(u)\ontop2T_N(v)=\ontop2T_N(v)\ontop1T_N(u)R_{12}(u-v)\,,

349: \ee

350: where  $\ontop1T_N(u)={T}_N(u)\otimes \II$ and $\ontop2T_N(v)=\II\otimes

351: {T}_N(v)$. The $R-$matrix acts on the tensor product of two auxiliary spaces,

352: $\mathbb{C}^2\otimes\mathbb{C}^2$,

353: \be\label{Rm}

354: R_{12}(u) = u\,\II + i\,P_{12}\,,

355: \ee

356: with $P_{12}$ being the permutation operator. The monodromy matrix for the open

357: spin chain is defined as~\cite{Sklyanin88} \footnote{General definition of the

358: integrable spin chain with nontrivial boundary conditions involves the boundary

359: matrices $K_\pm$~\cite{Ch,Sklyanin88,KS}. The Hamiltonian \re{H} corresponds to

360: the simplest case $K_\pm = \II$.}

361: \be\label{To}

362: \mathbb{T}_N(u)~=~T_N(u)\,T_N^{-1}(-u+i)=\frac1{\rho^{N}(u)}\cdot

363: T_N(u)\,\sigma_2\,T^{t}_N(-u)\,\sigma_2\,,

364: \ee

365: where the c-valued factor $\rho(u)=(u-is)(u+i(s-1))$ absorbs all poles of

366: $\mathbb{T}_N(u)$ and the superscript `$t$' denotes transposition in the

367: auxiliary space. It satisfies the fundamental ``reflection'' Yang-Baxter

368: relation~\cite{Ch,Sklyanin88,KS}

369: \be\label{YBo}

370: \mathbb{\ontop2T}_N(v)\,R_{12}(u+v-i)\,\mathbb{\ontop1T}_N(u)\,R_{12}(u-v)~=~

371: R_{12}(u-v)\,\mathbb{\ontop1T}_N(u)\,R_{12}(u+v-i)\,\mathbb{\ontop2T}_N(v)\,

372: \ee

373: with the same $R-$matrix \re{Rm}. It proves convenient to change a normalization

374: of $\mathbb{T}_N(u)$ as

375: \be\label{nTo}

376: \widehat{\mathbb{T}}_N(u)=\rho^N(-u)\,\mathbb{T}_N(-u)=T_N(-u)\,

377: \sigma_2\,T^t_N(u)\,\sigma_2=

378: \left(\begin{array}{cc}A(u)&B(u)\\

379: C(u)&D(u)

380: \end{array}\right).

381: \ee

382: It follows from \re{To} that $\widehat{\mathbb{T}}_N(u)$ satisfies the relation

383: \be

384: \widehat{\mathbb{T}}_N(-u-i)\,\widehat{\mathbb{T}}_N(u)=\left[(u+is)(u-i(s-1))\right]^{2N}

385: \II\,.

386: \label{q-det}

387: \ee

388: The Yang-Baxter relation \re{YBo} leads to the set of fundamental relations for

389: the operators $A(u),\ldots,D(u)$. For our purposes we will need only two of them

390: \ba

391: && B(u)B(v) = B(v)B(u)\,,

392: \label{BB}

393: \\

394: && B(u)D(v)= \frac{(u+v+i)(u-v-i)}{(u-v)(u+v)}D(v)B(u) + i

395: \left[A(u)+\frac{u+v+i}{u-v}D(u)\right] \frac{B(v)}{u+v}\,. \nonumber

396: \ea

397: The monodromy matrix \re{nTo} satisfies the following relation

398: \be

399: \widehat{\mathbb{T}}_N(u)=\frac{1}{2u-i}\left[ 2u\,\sigma_2

400: \widehat{\mathbb{T}}_N^{\,t}(-u)\sigma_2-i\, \widehat{\mathbb{T}}_N(-u)\right].

401: \label{T-rel}

402: \ee

403: To verify it one starts with the definition of $\widehat{\mathbb{T}}_N(u)$,

404: Eq.~\re{nTo}, interchanges the operators $T_N(-u)$ and $T^t_N(u)$ with a help of

405: the Yang-Baxter relation \re{YBR} and uses the explicit expression \re{Rm} for

406: the $R-$matrix. Substitution of \re{nTo} into \re{T-rel} yields

407: %\ba

408: %&& A(u)=\frac{1}{2u-i}\left[2u

409: %D(-u)-iA(-u)\right]\,,\qquad\frac{B(u)}{2u+i}=\frac{B(-u)}{-2u+i}\,,

410: %\nonumber \\

411: %&& D(u)=\frac{1}{2u-i}\left[2u A(-u)-iD(-u)\right]\,,\qquad

412: %\frac{C(u)}{2u+i}=\frac{C(-u)}{-2u+i}\,.

413: %\label{A-minus}

414: %\ea

415: \be

416: D(u)=\frac{1}{2u-i}\left[2u A(-u)-iD(-u)\right]\,,\qquad

417: \frac{B(u)}{2u+i}=\frac{B(-u)}{-2u+i}\,.

418: \label{A-minus}

419: \ee

420: In the standard manner, the transfer matrix for the open spin chain

421: $\widehat{t}_N(u)$ is defined as the trace of the monodromy matrix \re{nTo} over

422: the auxiliary space

423: %. Finally, one defines the transfer matrix as

424: \ba\label{th}

425: \widehat t_N(u) = \tr\widehat{\mathbb{T}}_N(u)=A(u) + D(u)

426: =\left(1-\frac{i}{2u}\right)\,D(u)~+~\left(1+\frac{i}{2u}\right)\,D(-u)\,,

427: \ea

428: where in the last relation we took into account \re{A-minus}. Following

429: Sklyanin~\cite{Sklyanin88} and making use of the Yang-Baxter relation \re{YBo},

430: one can show that the transfer matrix commutes with itself for different values

431: of the spectral parameter, with the Hamiltonian~(\ref{H}) and with the operator

432: of the total spin $\vec{S}=\sum_{n=1}^N \vec S_n$

433: \be\label{comm-th}

434: [\widehat t_N(u),\widehat t_N(v)]=[\widehat t_N(u),\CH_N]=[\widehat

435: t_N(u),\vec{S}]=0\,.

436: \ee

437: The expansion of $\widehat t_N(u)$ in powers of $u$ generates the integrals of

438: motion of the model. One deduces from \re{th} and \re{nTo} that the transfer

439: matrix is an even polynomial in $u$ of degree $2N$, $\widehat t_N(-u)=\widehat

440: t_N(u)$, which scales at large $u$ as $\widehat t_N(u)\sim 2(-1)^N u^{2N}$. In

441: addition, it follows from \re{To} that $\mathbb{T}_N(i/2)=\II$ leading to

442: \be

443: \widehat t_N(-i/2)=2\rho^N(i/2)=2(s-1/2)^{2N}\,.

444: \ee

445: These properties imply that $t_N(u)-t_N(\pm i/2)$ is proportional to

446: $(u+i/2)(u-i/2)$\\[2mm]

447: \be

448: \widehat t_N(u) = (-1)^N\left(u^2+1/4\right) \left[ 2 u^{2N-2} +\widehat q_2\,

449: u^{2N-4} + \ldots + \widehat q_{N-1} \, u^2 +\widehat q_N\right]+

450: 2\left(s-1/2\right)^{2N}\,.

451: \label{q's}

452: \ee\\[2mm]

453: Here the $\widehat q-$operators are given by polynomials in the spin operators

454: $\vec S_n$, for instance

455: \be

456: \widehat q_2=-4 \vec S^2 + 2Ns(s-1)-\frac12\,.

457: %,\quad \quad ???\widehat q_N=i^{2N} \tr

458: %\left[\prod_{k=1}^N(\sigma\cdot S_k) \prod_{k=1}^N(\sigma\cdot S_{N-k+1})\right].

459: \label{q2}

460: \ee

461: It follows from \re{comm-th} and \re{q's} that $N-1$ operators $\widehat

462: q_2,\ldots,\widehat q_N$ form the family of mutually commuting $SL(2)$ invariant

463: integrals of motion. Since $[\CH_N,\vec{S}]=0$, the remaining $N$th integral of

464: motion is provided by one of the components of the total spin. It is convenient

465: to choose the latter as $iS_-= -i\sum_n \partial_{z_n}$ since its eigenvalues

466: define the total momentum.

467: 

468: Thus, the open Heisenberg spin chain is a completely integrable model and the

469: spectral problem for the Hamiltonian~(\ref{H}) can be reformulated as the

470: spectral problem for the transfer matrix

471: \ba

472: \label{eit}

473: &&\widehat t_N(u)

474: \Psi_{\mybf{q},p}(z_1,\ldots,z_N)=t_N(u)\Psi_{\mybf{q},p}(z_1,\ldots,z_N)\,,

475: \\[2mm]

476: && (iS_--p)\Psi_{\mybf{q},p}(z_1,\ldots,z_N)=0\,, \nonumber

477: \ea

478: where $t_N(u)$ is the eigenvalue of the transfer matrix~(\ref{th}) and

479: $\Mybf{q}=(q_2,\ldots,q_N)$ denotes the eigenvalues of the integrals of motion. A

480: general solution to \re{eit} takes the form

481: \be

482: \Psi_{\mybf{q},p}(z_1,\ldots,z_N)=\int_{-\infty}^\infty dx_0\, \e^{ipx_0}

483: \Psi_{\mybf{q}}(z_1-x_0,\ldots,z_N-x_0)\,,

484: \ee

485: where integration goes along the real axis. The eigenstate

486: $\Psi_{\mybf{q}}(z_1,\ldots,z_N)$ has to diagonalize simultaneously the operators

487: $\widehat q_2,\ldots,\widehat q_N$.% Eq.~\re{q2}.

488: 

489: \setcounter{equation}{0}

490: \section{Baxter $\mathbb{Q}-$operator}

491: \label{QBaxter}

492: 

493: To solve the spectral problem for the open Heisenberg spin chain, Eq.~\re{eit},

494: we apply the method of the Baxter $\mathbb{Q}-$operator. The method relies on the

495: existence of the operator $\mathbb{Q}(u)$ which acts on the quantum space of the

496: model $\CV_N$, depends on the spectral parameter $u$ and satisfies the following

497: defining relations:

498: \begin{itemize}

499: \item Commutativity:

500: \be\label{Q-comm}

501: \left[\mathbb{Q}(u),\mathbb{Q}(v)\right]~=~0\,.

502: \ee

503: \item Q -- t relation:

504: \be\label{tQ}

505: \left[\mathbb{Q}(u),\widehat t_N(u)\right]=0\,.

506: \ee

507: \item Baxter relation:

508: \be\label{Bax-eq}

509: \widehat t_N(u)\,\mathbb{Q}(u)~=~\Delta_{+}(u)\,\mathbb{Q}(u+i)~+~

510: \Delta_{-}(u)\,\mathbb{Q}(u-i)\,,

511: \ee

512: \end{itemize}

513: where $\Delta_\pm(u)$ are some scalar functions of $u$. For the homogeneous open

514: spin chain they are given by

515: \be\label{Deltapm}

516: \Delta_\pm(u)=(-1)^N \frac{2u\mp i}{2u}(u\pm is)^{2N}\,.

517: \ee

518: In this Section, we construct the operator $\mathbb{Q}(u)$ satisfying

519: Eqs.~\re{Q-comm}--\re{Bax-eq} and discuss its properties.

520: 

521: It follows from \re{Q-comm} and \re{tQ} that the Baxter $\mathbb{Q}-$operator and

522: the transfer matrix $\widehat t_N(u)$ share the common set of the eigenstates

523: \be

524: \mathbb{Q}(u)\Psi_{\mybf{q}}(z_1,\ldots,z_N) = Q_{\mybf{q}}(u)

525: \Psi_{\mybf{q}}(z_1,\ldots,z_N)\,.

526: \label{Bax-eig}

527: \ee

528: The eigenstates $\Psi_{\mybf{q}}(z_1,\ldots,z_N)$ are the solutions to the

529: Schr\"odinger equation \re{Sch} whereas the corresponding eigenvalues of the

530: $\mathbb{Q}-$operator, $Q_{\mybf{q}}(u)$, satisfy the Baxter relation \re{Bax-eq}

531: with the transfer matrix $\widehat t_N(u)$, Eq.~\re{q's}, replaced by its

532: eigenvalue. As we will show below, the Baxter $\mathbb{Q}-$operator encodes

533: information about the spectrum of the open spin chain. Namely,

534: %the transfer matrices of the model and the unitary transformation

535: %to the SoV representation  can be expressed in terms of a single operator

536: %$\mathbb{Q}(u)$. As a consequence,

537: having calculated its eigenvalues $Q_{\mybf{q}}(u)$ one would be able to

538: reconstruct the energy spectrum of the model $E_{\mybf{q}}$.

539: 

540: 

541: 

542: \subsection{Gauge transformations}

543: \label{GT}

544: 

545: Our approach to constructing the Baxter $\mathbb{Q}-$operator is based on the

546: representation of $\mathbb{Q}(u)$ as an integral operator acting on the quantum

547: space of the model

548: \be\label{Bax-int}

549: \left[ \mathbb{Q}(u)\Psi\right](z_1,\ldots,z_N) ~=~\int \CD^N  w\,

550: Q_u(z_1\ldots,z_N|\bar w_1,\ldots,\bar w_N)\Psi(w_1,\ldots,w_N)\,,

551: \ee

552: with $\bar w_n=w_n^*$ and the integration measure defined in \re{measure}. To

553: find the explicit expression for the kernel $Q_u(z_1\ldots,z_N|\bar

554: w_1,\ldots,\bar w_N)$ we shall explore the fact that the transfer matrix of the

555: open spin chain, Eq.~\re{th}, is invariant under local gauge transformations of

556: the Lax operators~\cite{PG,SD}

557: \be\label{Lgr}

558: L_n(u)\to \widetilde L_n(u) = M_n^{-1}\,L_n(u)\,M_{n+1}\,,

559: \ee

560: where $M_n$ (with $n=1,\ldots,N+1$) are arbitrary $2\times 2$ matrices with $\det

561: M_n\neq 0$. According to Eqs.~\re{Tc}, \re{To} and \re{nTo} the operators

562: $T_N(u)$ and $\widehat{\mathbb{T}}_N(u)$ are transformed under \re{Lgr} as

563: \ba

564: T_N(u)&\to& \widetilde T_N(u)=M_1^{-1}\, T_N(u)\, M_{N+1},\ \ \ \ \ \ \

565:  %\label{Mc}

566: \nonumber

567: \\

568: \widehat{ \mathbb{T}}_N(u)&\to& \widetilde{\widehat{\mathbb{T}}}_N(u)=M_1^{-1}\,

569: \widehat{ \mathbb{T}}_N(u)\, M_{1}\,,

570: \label{Mo}

571: \ea

572: so that the transfer matrix $\widehat t_N(u)=\tr\widehat{\mathbb{T}}_N(u)$ stays

573: invariant.

574: 

575: The gauge rotation of the Lax operator, Eq.~\re{Lgr}, has been used by Pasquier

576: and Gaudin to construct the Baxter operator for the periodic Toda chain~\cite{PG}

577: and it was later applied to the closed spin chain in Refs.~\cite{SD,SoV}. In the

578: latter case, the transfer matrix equals $\tr T_N(u)$ and in order to preserve its

579: invariance under the transformation \re{Mo} one had to impose periodic boundary

580: conditions $M_{N+1}=M_1$. For the open spin chain the matrices $M_{N+1}$ and

581: $M_1$ can be arbitrary, $M_{N+1}\neq M_1$. \footnote{Notice that the monodromy

582: matrices $\widetilde T_N(u)$ and $\widetilde{\widehat{\mathbb{T}}}_N(u)$ satisfy

583: the Yang-Baxter relations, Eq~\re{YBR} and \re{YBo}, respectively. This follows

584: immediately from the invariance of the $R-$matrix~(\ref{Rm}) under

585: transformations $R\to U\, R\, U^{-1}$ with $U=M\otimes M$.} In spite of this

586: difference, many results obtained in Refs.~\cite{SD,SoV} for the closed spin

587: chain are applicable to the open chain.

588: 

589: To begin with, let us introduce the function~\cite{SD,SoV}

590: \be\label{Y-N}

591: Y_u(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_{N+1})~=~ \prod_{k=1}^N (z_k-\bar

592: w_k)^{-s-iu}\,(z_k-\bar w_{k+1})^{-s+iu}\,,

593: \ee

594: which is a (anti)holomorphic function of the complex variables $\vec

595: z=(z_1,\ldots,z_N)$ and $\vec w=(\bar w_1,\ldots,\bar w_{N+1})$  in the upper

596: half-plane $\Im z_k>0$ and $\Im w_n>0$ (with $\bar w_n=w_n^*$). It satisfies the

597: following relations

598: \ba

599: \widetilde b(u; \bar w_1,\bar w_{N+1})\,Y_u(\vec{z}|\vec{w})&=&0\,,\nonumber \\

600: \widetilde a(u; \bar w_1,\bar

601: w_{N+1})\,Y_u(\vec{z}|\vec{w})&=&(u+is)^N\,Y_{u+i}(\vec{z}|\vec{w})\,,

602: \label{TYc}

603: \\

604: \widetilde d(u; \bar w_1,\bar

605: w_{N+1})\,Y_u(\vec{z}|\vec{w})&=&(u-is)^N\,Y_{u-i}(\vec{z}|\vec{w})\,,\nonumber

606: \ea

607: where the operators $\widetilde a(u),\ldots,\widetilde d(u)$ are defined

608: similarly to \re{Tc} as the entries of the gauge rotated transfer matrix

609: $\widetilde T_N(u)$, Eq.~\re{Mo}, with the $M-$matrices given by

610: \be\label{MMM}

611: M_1=\left(\begin{array}{cc}1&1/\bar w_1\\0&1\end{array} \right)\,,\ \ \ \ \ \ \ \

612: \ \ M_{N+1}=\left(\begin{array}{cc}1& 1/\bar w_{N+1}\\0&1\end{array} \right)\,,

613: \ee

614: with $\bar w_{N+1}\neq \bar w_1$. They are given by linear combinations of the

615: operators $a(u),\ldots,d(u)$, Eq.~\re{Tc}, with the coefficients depending on the

616: gauge parameters $\bar w_{N+1}$ and $\bar w_1$, which are identified as the right

617: arguments of the kernel $Y_u(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_{N+1})$.

618: Similar relations hold between the entries of the monodromy matrices

619: $\widehat{\mathbb{T}}_N(u)$ and $\widetilde{\widehat{\mathbb{T}}}_N(u)$,

620: Eqs.~\re{Mo} and \re{nTo}, so that

621: \be\label{tADB}

622: \widetilde A(u;\bar w_1)+\widetilde D(u;\bar w_1)=A(u)+D(u)\,,\qquad \widetilde

623: B(u;\bar w_1) = B(u) + \mathcal{O}(1/\bar w_1)\,,

624: \ee

625: where we indicated explicitly the dependence on the gauge parameter $\bar w_1$.

626: Eqs.~\re{TYc} play a crucial r\^ole in our subsequent analysis. Their derivation

627: can be found in~\cite{SD,SoV}.

628: 

629: 

630: To proceed further let us express the  entries of the monodromy matrix of the

631: open chain, $\widetilde B(u)$ and $\widetilde D(u)$, in terms of those for the

632: closed spin chain, $\widetilde a(u),\ldots, \widetilde d(u)$. One finds from

633: \re{Mo}, \re{Tc} and \re{nTo}

634: \be

635: \widetilde B(u)=\widetilde b\lr{-u}\widetilde a\lr{u} -\widetilde

636: a\lr{-u}\widetilde b\lr{u}\,,\qquad \widetilde D(u)=\widetilde d\lr{-u}\widetilde

637: a\lr{u} -\widetilde c\lr{-u}\widetilde b\lr{u}\,.

638: \label{B}

639: \ee

640: 

641: \subsection{Kernel of the $\mathbb{Q}-$operator}

642: 

643: We now turn to constructing the kernel of the Baxter $\mathbb{Q}-$operator and

644: consider the following auxiliary operator $\mathbb{G}(u,v): \mathcal{V}_N

645: \longmapsto \mathcal{V}_N$ with the kernel given by the convolution of two

646: $Y-$functions introduced in the subsection~\ref{GT}

647: \ba

648: &&G_{u,v}(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)= \e^{i\pi s(2N-1)}\,\int

649: \mathcal{D} y_2\ldots \int \mathcal{D} y_{N}\,

650: \label{G-def}

651: \\

652: &&\hspace*{30mm} \times\ Y_{u}(z_1,\ldots,z_N|\bar w_1,\bar y_2,\ldots,\bar y_N,

653: \bar w_N)\, Y_{v}(y_2,\ldots,y_N|\bar w_1,\bar w_2,\ldots, \bar w_N)\,.

654: \nonumber

655: \ea

656: Here the integration measure $\mathcal{D} y_n$ is defined in \re{measure} and the

657: prefactor is introduced for the later convenience. Notice that two $Y-$functions

658: in \re{G-def} have a different number of arguments and depend on the same

659: variables $\bar w_1$ and $\bar w_{N}$.

660: 

661: Let us demonstrate that for $v=-u$ the operator $\mathbb{G}(u,v)$ satisfies the

662: relations \re{Q-comm} -- \re{Bax-eq} and, therefore, it can be identified as the

663: Baxter $\mathbb{Q}-$operator for the homogeneous open Heisenberg spin chain

664: \be

665: \mathbb{Q}(u)= \mathbb{G}(u,-u)%= \mathbb{G}(-u,u)

666: \,,

667: \label{Q-kern}

668: \ee

669: or equivalently

670: \ba

671: &&\hspace*{-20mm} Q_u(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)

672: %~=~G_{u,-u}(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)

673: = \e^{i\pi s(2N-1)}\, (z_1-\bar w_1)^{-\beta_u} (z_N-\bar w_N)^{-\alpha_u}

674: \nonumber

675: \\

676: && \times %\int \mathcal{D} y_2\ldots \int \mathcal{D} y_{N}\,

677: \prod_{n=2}^{N} \int \mathcal{D} y_n\, (z_{n-1}-\bar y_n)^{-\alpha_u}(z_n-\bar

678: y_n)^{-\beta_u} (y_n-\bar w_{n-1})^{-\alpha_u}(y_n-\bar w_n)^{-\beta_u}

679: \label{Q-explicit}

680: \ea

681: with $\alpha_u=s-iu$ and  $\beta_u=s+iu$. To prove \re{Q-kern} we apply the

682: diagrammatical approach developed in Ref.~\cite{SoV}. In this approach, one

683: represents the kernel $G_{u,v}(\vec{z}|\vec{w})$, Eq.~\re{G-def}, as the Feynman

684: diagram shown in Fig.~\ref{fig1}. There, the arrow with the index $\alpha$ that

685: goes from $y$ to $z$ represents the factor $(z-\bar y)^{-\alpha}$ (see

686: Eq.~\re{alpha-rep}) while the black blob denotes integration over the position

687: $w$ of the corresponding vertex with the $SL(2,\mathbb{R})$ measure

688: $\mathcal{D}w$ (see Eq.~\re{chain-h} and Fig.~\ref{Chain}).

689: 

690: The operator $\mathbb{G}(u,v)$ is symmetric under interchange of the spectral

691: parameters

692: \be\label{Guv}

693: \mathbb{G}(u,v)=\mathbb{G}(v,u) \,,

694: \ee

695: or equivalently $G_{u,v}(\vec{z}|\vec{w})~=~G_{v,u}(\vec{z}|\vec{w})$. The proof

696: of \re{Guv} is based on the permutation identity shown in Fig.~\ref{comm-f}.

697: Writing $\beta_u=\beta_v + i(u-v)$, one replaces the left-most vertical line in

698: the left diagram in Fig.~\ref{fig1} by two lines with the indices $\beta_v $ and

699: $i(u-v)$. Then, one displaces the line with the index $i(u-v)$ across the diagram

700: to the right with a help of the permutation identity until it merges with the

701: right-most vertical line and changes its index to $\alpha_u+i(u-v)=\alpha_v$ (see

702: Ref.~\cite{SoV} for details). The resulting diagram coincides with the original

703: one but with the spectral parameters interchanged. Furthermore, it follows from

704: \re{G-def} and \re{TYc} that

705: \ba

706: \widetilde a(u; \bar w_1,\bar w_{N})\,{G}_{u,v}(\vec z; \vec w) &=& (u+is)^N

707: {G}_{u+i,v}(\vec z; \vec w)\,,\nonumber

708: \\

709: \widetilde d(u; \bar w_1,\bar w_{N})\,{G}_{u,v}(\vec z; \vec w) &=&

710: (u-is)^N{G}_{u-i,v}(\vec z; \vec w)\,,

711: \label{G-prop}

712: \\

713: \widetilde b(u; \bar w_1,\bar w_{N})\,{G}_{u,v}(\vec z; \vec w) &=& 0\,,

714: \nonumber

715: \ea

716: where $\vec z=(z_1,\ldots,z_N)$ and $\vec w=(\bar w_1,\ldots,\bar w_N)$. Here the

717: operators $\widetilde a(u),\ldots,\widetilde d(u)$ depend on the gauge parameters

718: $\bar w_1$ and $\bar w_N$, which coincide with the corresponding arguments of the

719: kernel $G_{u,v}(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)$.

720: %

721: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

722: \begin{figure}[t]

723: \psfrag{z1}[cc][cc]{$z_1$}\psfrag{z2}[cc][cc]{$z_2$}\psfrag{zn}[cc][cc]{$z_N$}

724: \psfrag{w1}[cc][cc]{$\bar w_1$}\psfrag{w2}[cc][cc]{$\bar w_2$}

725: \psfrag{wn}[cc][cc]{$\bar w_N$} \psfrag{dots}[cc][cc]{$\Mybf{\cdots}$}

726: \psfrag{a}[cc][cc]{$\beta_u$} \psfrag{b}[cc][cc]{$\alpha_u$}

727: \psfrag{c}[cc][rc]{$\beta_v$} \psfrag{d}[cc][rc]{$\alpha_v$}

728: \centerline{\epsfxsize10.0cm\epsfbox{open_b.eps}} \vspace*{0.5cm}

729: \caption[]{Diagrammatical representation of the function $G_{u,v}(\vec z | \vec

730: w)$. For $v=-u$ the diagram defines the kernel of the Baxter

731: $\mathbb{Q}-$operator, Eq.~\re{Q-explicit}.

732: Here $\alpha_x=s-ix$ and $\beta_x=s+ix$.}%

733: \label{fig1}%

734: \end{figure}%

735: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

736: 

737: Let us demonstrate that the operator $\mathbb{Q}(u)$ satisfies the Baxter

738: equation \re{Bax-eq}. To this end, one examines the expression entering the

739: l.h.s.\ of the Baxter equation \re{Bax-eq} and applies Eqs.~\re{th}, \re{tADB}

740: and \re{Q-kern} to get

741: \be

742: \widehat t_N(u)\,\mathbb{Q}(u)= (\widetilde A(u) +\widetilde

743: D(u))\,\mathbb{G}(u,-u) = \left[\frac{2u-i}{2u}\,\widetilde

744: D(u)+\frac{2u+i}{2u}\,\widetilde D(-u)\right]\mathbb{G}(u,-u)\,.

745: \label{tQ=G}

746: \ee

747: Taking into account Eqs.~\re{B}, \re{G-prop} and \re{Guv}, one finds

748: \ba

749: \widetilde D(u)\,\mathbb{G}(u,-u) &=& \widetilde d\lr{-u} \widetilde a\lr{u}

750: \mathbb{G}(u,-u)

751: %\nonumber\\ &=&

752: =(u+is)^N\widetilde d\lr{-u}\mathbb{G}(-u,u+i)

753: \nonumber\\

754: &=&(u+is)^N(-u-is)^N\mathbb{G}(-u-i,u+i)

755: \label{DG}

756: %\ea

757: %In a similar manner,

758: %\ba\label{AG}

759: %\widetilde A_N(u)\,\mathbb{G}(u,-u)&=&(-1)^N (u-is)^{2N}

760: %\lr{1+\frac{i}{2u}}\mathbb{G}(u-i,-u+i)

761: %\nonumber\\

762: %&-& (-1)^{N}\frac{i}{2u}(u+is)^{2N}\mathbb{G}(u+i,-u-i)

763: %\\

764: %\widetilde B_N(u)\,\mathbb{G}(u,-u) &=& (u+is)^N \widetilde

765: %b_N\lr{-u}\,\mathbb{G}(u+i,-u) = 0

766: %\label{BG}

767: \ea

768: Substituting \re{DG} into \re{tQ=G} one concludes that the operator

769: $\mathbb{Q}(u)$ defined in \re{Q-explicit} verifies the Baxter relation

770: \re{Bax-eq}. In addition, one deduces from \re{B} and \re{G-prop} that the kernel

771: of the $\mathbb{Q}-$operator is nullified by the operator $\widetilde B(u)$

772: \be\label{Bnul}

773: \widetilde B(u;w_1)\, Q_u(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)~=~0\,.

774: \ee

775: We will use this property in Sect.~\ref{SoV} to construct the unitary

776: transformation to the SoV representation.

777: 

778: The operator $\mathbb{Q}(u)$

779: %with the kernel defined in \re{Q-explicit}

780: %One deduces from the definition \re{Q-explicit} that the operator $\mathbb{Q}(u)$

781: has the following properties

782: \begin{itemize}

783: \item Parity:

784: \be\label{parity}

785: \mathbb{Q}(u)=\mathbb{Q}(-u)\,.

786: \ee

787: \item Normalization:

788: \be\label{KK}

789: \mathbb{Q}(\pm is)=\mathbb{K}\,.

790: \ee

791: \item Hermiticity:

792: \be\label{hermiticity}

793:       \lr{\mathbb{Q}(u)}^\dagger=\mathbb{Q}(u^*)\,.

794: \ee

795: \item $SL(2)$ invariance:

796: \be\label{SL2-inv}

797:  [\mathbb{Q}(u), \vec S] =0\,,

798: \ee

799: \end{itemize}

800: where $\mathbb{K}$ is the unit operator on the Hilbert space of the model

801: $\mathcal{V}_N$~(see Eq.~\ref{K-ker} in Appendix~\ref{Ap}) and $\vec

802: S=\sum_{k=1}^N \vec S_k$ is the operator of the total spin. Eq.~\re{parity} is a

803: consequence of \re{Q-kern} and \re{Guv}. Eq.~\re{KK} follows from the fact that

804: $\beta_{is}=\alpha_{-is}=0$ so that the corresponding lines in the diagram in

805: Fig.~\ref{fig1} disappear leading to drastic simplification of the kernel.

806: Eq.~\re{hermiticity} follows directly from the definition of the conjugated

807: operator $\lr{\mathbb{Q}(u)}^\dagger$. To verify \re{SL2-inv} one notices that

808: the kernel of the $\mathbb{Q}-$operator, Eq.~\re{Q-explicit}, is transformed

809: under the $SL(2,\mathbb{R})$ transformations as

810: \be

811: {Q}_u(\vec z'|\vec w') = \prod_{k=1}^N (\gamma \bar w_k + \delta)^{2s} (\gamma

812: z_k + \delta)^{2s}\,{Q}_u(\vec z|\vec w)

813: \label{Q-trans}

814: \ee

815: where $z_k'=(\alpha z_k+\beta)/(\gamma z_k+\delta)$ and $\bar w_k'=(\alpha\bar

816: w_k+\beta)/(\gamma\bar w_k+\delta)$ with real $\alpha,\ldots,\delta$ such that

817: $\alpha\delta-\beta\gamma=1$. %As we will show in the next subsection,

818: %\re{Q-trans} also holds for complex $\alpha,\ldots,\delta$.

819: 

820: We are now ready to demonstrate that the $\mathbb{Q}-$operator \re{Q-kern}

821: satisfies the relations \re{tQ} and \re{Q-comm}. To verify \re{tQ}, one performs

822: the Hermitian conjugation of the both sides of the Baxter equation \re{Bax-eq}.

823: Taking into account \re{hermiticity}, one finds that the r.h.s.\ of \re{Bax-eq}

824: goes into $\widehat t_N(u^*)\mathbb{Q}(u^*)$ whereas its l.h.s.\ is replaced by

825: $\lr{ \widehat t_N(u)\mathbb{Q}(u)}^\dagger=\mathbb{Q}(u^*) \widehat t_N(u^*)$.

826: Equating the two expressions one arrives at \re{tQ}. Finally, let us show that

827: the operator \re{Q-kern} satisfies the commutativity condition \re{Q-comm}. The

828: proof can be performed diagrammatically. To this end, one examines the Feynman

829: diagram corresponding to the product

830: $\mathbb{Q}(v)\mathbb{Q}(u)=\mathbb{G}(v,-v)\,\mathbb{G}(u,-u)$ and inserts a

831: pair of lines with the indices $\pm i(u+v)$ into one of the central rhombuses as

832: shown in Fig.~\ref{fig2}. Displacing the two lines horizontally in the opposite

833: directions with a help of the permutation identity (see Fig.~\ref{comm-f}) one

834: obtains the Feynman diagrams shown in Fig.~\ref{fig2} to the right. It differs

835: from the original diagram in that various $\alpha-$ and $\beta-$indices got

836: interchanged and two additional lines with the indices $\pm i(u+v)$ connect the

837: ``end points'', $\bar w_1$ with $z_1$ and $\bar w_N$ with $z_N$. Taking into

838: account the definition of the $\mathbb{G}-$operator, Eq.~\re{G-def} (see

839: Fig.~\ref{fig1}) one finds that the Feynman integral corresponding to this

840: diagram can be written as

841: \be

842: \bigg[\mathbb{Q}(v)\,\mathbb{Q}(u)\bigg](\vec z; \vec w) ={(z_1-\bar

843: w_1)^{i(u+v)} (z_N-\bar w_N)^{-i(u+v)}} \left[\lr{\mathbb{G}(u^*,v^*)}^\dagger

844: \,\mathbb{G}(-u,-v)\right](\vec z; \vec w)\,,

845: \label{Identity}

846: \ee

847: where the kernel of the integral operator $\lr{\mathbb{G}(u^*,v^*)}^\dagger$ is

848: given by $(G_{u^*,v^*}(w_1,\ldots,w_N|\bar z_1,\ldots,\bar z_N))^*$. According to

849: \re{Guv}, the r.h.s.\ of \re{Identity} is invariant under interchanging

850: $u\leftrightarrows v$ thus proving the commutativity relation \re{Q-comm}.

851: \begin{figure}[t]

852: \psfrag{z1}[cc][cc]{$z_1$}\psfrag{z2}[cc][cc]{$z_2$}\psfrag{zN}[cc][cc]{$z_N$}

853: \psfrag{w1}[cc][cc]{$\bar w_1$}\psfrag{w2}[cc][cc]{$\bar w_2$}

854: \psfrag{wN}[cc][cc]{$\bar w_N$} \psfrag{dots}[cc][cc]{$\Mybf{\cdots}$}

855: \psfrag{=}[cc][cc]{$\Mybf{=}$} \psfrag{a1}[cc][rc]{$\beta_v$}

856: \psfrag{b1}[cc][rc]{$\alpha_v$} \psfrag{a2}[cc][rc]{$\beta_u$}

857: \psfrag{b2}[cc][rc]{$\alpha_u$}

858: 

859: \centerline{\epsfxsize17.0cm\epsfbox{identity.eps}} \vspace*{0.5cm}

860: \caption[]{Diagrammatical proof of Eq.~\re{Identity}. The left diagram represents

861: the kernel of the operator $\mathbb{Q}(v)\mathbb{Q}(u)$. The right diagram is

862: obtained by displacing two wavy lines carrying the indices $\pm i(u+v)$ to the

863: right/left with a help of the permutation identity. Here $\alpha_x=s-ix$,

864: $\beta_x=\alpha_{-x}=s+ix$. }

865: \label{fig2}

866: \end{figure}

867: 

868: \subsection{Contour-integral representation for the $\mathbb{Q}-$operator}

869: 

870: In the previous subsection we constructed the Baxter $\mathbb{Q}-$operator for

871: the homogeneous open spin chain, Eqs.~\re{Q-explicit}. As was already mentioned,

872: the $\mathbb{Q}-$operator is diagonalized by the eigenstates of the model,

873: Eq.~\re{Bax-eig}, and the corresponding eigenvalues $Q_{\mybf{q}}(u)$ satisfy

874: \re{Bax-eq}.

875: 

876: The Baxter equation \re{Bax-eq} is a finite-difference functional equation and

877: its solutions are defined up to multiplication by an arbitrary periodic function,

878: $f(u+i)=f(u)$. To fix this ambiguity and determine eigenvalues of the

879: $\mathbb{Q}-$operator, one has to specify analytical properties of

880: $Q_{\mybf{q}}(u)$.

881: %These conditions follow from the explicit form of the operator $\mathbb{Q}(u)$.

882: %We would like to stress that the Baxter equation

883: %\re{Bax-eq} is a universal feature of quantum models with the $SL(2)$ spin

884: %symmetry and it does not depend on the choice of a particular representation of

885: %the $SL(2,\mathbb{R})$ group.

886: They can be identified using the following contour-integral representation for

887: the $\mathbb{Q}-$operator on the quantum space of the model

888: $\Psi(z_1,\ldots,z_N)\in \mathcal{V}_N$

889: \ba\label{Q-cont}

890: &&\left[\mathbb{Q}(u)\Psi\right](z_1,\ldots,z_N)=[B(s+iu,s-iu)]^{-2N+1}\\[3mm]

891: &&\ \ \ \times \int_0^1\prod_{n=1}^N d\sigma_n\,

892: (1-\sigma_n)^{s+iu-1}\,\sigma_n^{s-iu-1} \int_0^1\prod_{k=2}^N d\tau_k\,

893: (1-\tau_k)^{s+iu-1}\,\tau_k^{s-iu-1}\, \Psi(Z_1,\ldots,Z_N)\,,\nonumber

894: \ea

895: where $B(x,y)$ is the Euler beta-function and the $Z-$coordinates are defined as

896: \ba\label{Z}

897: &Z_1&=(1-\sigma_1) z_1+\sigma_1[\tau_2 z_1+(1-\tau_2) z_2]\,\nonumber\\

898: &Z_k&=(1-\sigma_k)[\tau_k z_{k-1}+(1-\tau_k) z_k]~+~

899: \sigma_k [\tau_{k+1} z_{k}+(1-\tau_{k+1}) z_{k+1}]\,, \ \  \  \, (1<k<N)\,,\nonumber\\

900: &Z_N&=(1-\sigma_N)[\tau_N z_{N-1}+(1-\tau_N) z_N]~+~\sigma_N z_{N}\,.

901: \ea

902: To obtain \re{Q-cont}, one uses the integral representation for the

903: $\mathbb{Q}-$operator, Eq.~\re{Q-explicit}, and applies the identity \re{drift}.

904: 

905: Since the function $\Psi(Z_1,\ldots,Z_N)$ is holomorphic in the upper half-plane

906: $\Im z_n>0$, the integral in the r.h.s.\ of \re{Q-cont} is convergent inside the

907: strip $-s < \Re (iu) < s$ in the complex $u-$plane. Analytically continuing the

908: integral outside this strip, one finds that it contains poles of the order $p \le

909: 2N-1$ originating from integration at the vicinity of the end-points $\sigma_n

910: \,, \tau_k\to 0,1$. They are located at $iu = \pm (n-s)$ (with $n$ nonnegative

911: integer) and are compensated by the beta-function prefactor entering \re{Q-cont}.

912: As a result, $[\mathbb{Q}(u)\Psi](z_1,\ldots,z_N)$ does not have poles in $u$

913: and, therefore, the eigenvalues of the Baxter operator, $Q_{\mybf{q}}(u)$, are

914: entire functions of $u$.

915: 

916: Eq.~\re{Q-cont} also allows one to determine asymptotic behaviour of

917: $Q_{\mybf{q}}(u)$ at large $u$. It is given by

918: \be

919: Q_{\mybf{q}}(u) \sim u^{2h} \left[1+ \mathcal{O}(1/u^2)\right],

920: \label{Q-asym}

921: \ee

922: with $h$ nonnegative integer defining the total spin of the model

923: % It is governed by the eigenvalue of the total Casimir operator

924: \be

925: [\vec S^2-(h+Ns)(h+Ns-1)]\Psi_{\mybf{q}}(\vec z)=0\,.

926: \label{S2-eig}

927: \ee

928: To establish \re{Q-asym}, one verifies using \re{Q-cont} that the Baxter operator

929: is invariant under arbitrary $SL(2,\mathbb{C})$ transformations, in particular

930: under the following one $z\mapsto -i\lr{w+i}/\lr{w-i}$

931: \be\label{RtU}

932: \Psi(z_1,\ldots,z_N)\mapsto \widetilde\Psi(w_1,\ldots,w_N)=\prod_{k=1}^N

933: (w_k-i)^{-2s}\Psi\left(-i\frac{w_1+i}{w_1-i},\ldots,-i\frac{w_N+i}{w_N-i}\right)\,,

934: \ee

935: which map the upper half-plane $\Im z_k >0$ into a unit disk $|w_k|<1$. The main

936: advantage of dealing with functions $\widetilde\Psi(w_1,\ldots,w_N)$ holomorphic

937: inside the unit circle is that solutions to \re{S2-eig} have a simple form in

938: that case. Namely, the Hilbert space of the model contains the highest weigths

939: which satisfy \re{S2-eig} and are given by homogeneous translation invariant

940: polynomials of degree $h$,

941: $\widetilde\Psi(w_1,\ldots,w_N)=P_h(w_1-w_2,\ldots,w_{N-1}-w_N)$. Since the

942: Baxter operator \re{Q-cont} remains invariant under \re{RtU}, one can substitute

943: the function $\Psi(\vec z)$ in \re{Q-cont} by such polynomial. Then,

944: $\Psi(Z_1,\ldots,Z_N)$ entering \re{Q-cont} becomes a polynomial in the $\sigma-$

945: and $\tau-$parameters. Integrating term-by-term in the r.h.s.\ \re{Q-cont} one

946: finds that the dominant contribution at large $u$ comes from terms containing a

947: maximum number of $\sigma$'s and $\tau$'s. This number equals $2h$ and leads to

948: the asymptotics \re{Q-asym}.

949: 

950: Given that $Q_{\mybf{q}}(u)$ is an even function of $u$, Eq.~\re{parity}, and

951: making use of \re{hermiticity}, we conclude that the eigenvalues of the

952: $\mathbb{Q}-$operator are real polynomials in $u^2$ of degree $h$

953: \be

954: Q_{\mybf{q}}(u) = a_{\mybf{q}}\prod_{k=1}^h (u^2-\lambda_k^2)\,, \qquad

955: \lr{Q_{\mybf{q}}(u)}^*= Q_{\mybf{q}}(u^*)

956: \label{roots}

957: \ee

958: with the normalization constant $a_{\mybf{q}}$ fixed by the condition

959: $Q_{\mybf{q}}(is)=1$, Eq.~\re{KK}. Substituting \re{roots} into \re{Bax-eq} and

960: putting $u=\lambda_k^2$, one finds that the roots $\lambda_k^2$ satisfy the Bethe

961: equations for the open spin chain~\cite{Sklyanin88}.

962: 

963: \subsection{Relation to the Hamiltonian}

964: 

965: Let us demonstrate that the Hamiltonian of the open spin chain, Eq.~\re{H}, is

966: given by a logarithmic derivative of the Baxter operator evaluated at special

967: values of the spectral parameter $u=\pm is$. Due to \re{KK} the expansion of the

968: $\mathbb{Q}-$operator around $u=\pm is$ can be written as

969: \be\label{Q-H}

970: \left[\mathbb{Q}(\pm is

971:   +\epsilon)\Psi\right](\vec{z})~=~\Psi(\vec{z})~\mp~

972: i\epsilon\, \left[\CH_N\,\Psi\right](\vec{z})~+~{\cal O}(\epsilon^2)\,,

973: \ee

974: with $\CH_N$ being some integral operator. Its explicit form can be found from

975: the contour-integral representation for the $\mathbb{Q}-$operator, \re{Q-cont}.

976: At $u=-is+\epsilon$ the beta-prefactor in the r.h.s.\ of \re{Q-cont} vanishes as

977: $\epsilon^{2N-1}$ but it is compensated by poles coming from integration at the

978: vicinity of $\sigma_n=\tau_k =0$. Carefully separating contribution from this

979: region, one obtains that the operator $\CH_N$ entering \re{Q-H} is given by the

980: sum of two-particle integral operators

981: \be\label{HH}

982: \CH_N =-i\frac{d}{d\epsilon}\ln \mathbb{Q}(- is+\epsilon)\biggl|_{\epsilon=0}=

983: \sum_{n=0}^{N-1} H_{n,n+1}\,,

984: \ee

985: where $H_{n,n+1}$ acts on $\Psi(z_n,z_{n+1})\in V_n\otimes V_{n+1}$ as

986: \be\label{Hn}

987: [H_{12}\,\Psi](z_1,z_2)~=~-\int_0^1\frac{d\tau}{\tau}  {{(1-

988:     \tau)}^{2s-1}}

989: \left[\Psi(z_{12}(\tau),z_2)+\Psi(z_1,z_{21}(\tau))-2\Psi(z_1,z_2) \right]\,,

990: \ee

991: with $z_{ik}(\tau)=(1-\tau)z_i+\tau z_k$. It is straightforward to check that the

992: Hamiltonian $H_{n,n+1}$ commutes with the two-particle spin

993: $\vec{S}_n+\vec{S}_{n+1}$ defined in \re{s-rep} and, therefore, it only depends

994: on the Casimir operator $J_{n,n+1}$, Eq.~\re{JJ}. To find the explicit form of

995: this dependence one applies $H_{12}$ to the state $\Psi(z_1,z_2)=

996: (z_1-z_2)^h/((z_1+i)(z_2+i))^{h+2s}$ with $h$ nonnegative

997: integer.%

998: \footnote{Under conformal mapping \re{RtU} this state is transformed into a

999: homogeneous polynomial of degree $h$, $\widetilde\Psi(w_1,w_2)=(w_1-w_2)^h$.} It

1000: diagonalizes simultaneously the Casimir operator $J_{12}\,\Psi = (h+2s)\Psi$ and

1001: the two-particle kernel $H_{12} \Psi= 2[\psi(h+2s)-\psi(2s)]\Psi$ leading to the

1002: expression

1003: \be\label{Hnn}

1004: H_{n,n+1}~=~2\left[\psi(J_{n,n+1})~-~\psi(2s)\right]\,,

1005: \ee

1006: which coincides with \re{H}.

1007: 

1008: Eq.~\re{HH} establishes the relation between the Hamiltonian of the model \re{H}

1009: and the Baxter $\mathbb{Q}-$operator. Obviously, the same relation holds between

1010: their eigenvalues

1011: \be\label{QE}

1012: E_{\mybf{q}}~=~ \pm i\frac{d}{d\epsilon}\ln Q_{\mybf{q}}(\pm

1013: is+\epsilon)\biggl|_{\epsilon=0}\,.

1014: \ee

1015: Thus, to reconstruct the energy spectrum of the model, one has to find polynomial

1016: solutions to the Baxter equation \re{Bax-eq} and apply \re{QE}.

1017: 

1018: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1019: \setcounter{equation}{0}

1020: \section{Separation of Variables}

1021: \label{SoV}

1022: 

1023: %In the previous section, we constructed the Baxter $\mathbb{Q}-$operator for the

1024: %homogeneous open spin chain and expressed the Hamiltonian of the model as a

1025: %logarithmic derivative of the operator $\mathbb{Q}(u)$ at $u=\pm is$. We argued

1026: %that the eigenvalues of the $\mathbb{Q}-$operator are even polynomials in $u$

1027: %which satisfy the Baxter equation. This allows one to determine the energy

1028: %spectrum of the model.

1029: 

1030: In this section we will construct integral representation for the eigenstates of

1031: the model, Eq.~\re{Sch}, by going over to the representation of the Separated

1032: Variables $(p,\Mybf{x})=(p,x_1,...,x_{N-1})$ (SoV)

1033: \be

1034: \Psi_{\mybf{q},p}\,(z_1,\ldots,z_N)=\int_{\mathbb{R}_{+}^{N-1}}

1035: d^{N-1}\Mybf{x}\,\mu(\Mybf{x})\, U_{p,\mybf{x}}(z_1,\ldots,z_N)\,

1036: \Phi_{\mybf{q}}(\Mybf{x})\,.

1037: \label{SoV-gen}

1038: \ee

1039: Here $\Phi_{\mybf{q}}(\Mybf{x})$ is the eigenfunction of the model in the

1040: separated variables. It is factorized into a product of functions depending on a

1041: single variable $\Phi_{\mybf{q}}(\Mybf{x})\sim Q(x_1)\ldots Q(x_{N-1})$. As will

1042: be shown in this section, $Q(x_k)$ coincides with the eigenvalue of the Baxter

1043: $\mathbb{Q}-$operator. The kernel $U_{p,\mybf{x}}$ of the unitary operator

1044: corresponding to the SoV transformation is defined as

1045: \be

1046: U_{p,\mybf{x}}(z_1,\ldots,z_N)=\vev{z_1,\ldots,z_N|p,\Mybf{x}}\,.

1047: \label{U-ker}

1048: \ee

1049: We will argue below that the separated variables $x_k$ (with $k=1,\ldots,N-1$)

1050: take real positive values so that integration in \re{SoV-gen} goes over

1051: $\Mybf{x}\in \mathbb{R}_+^{N-1}$ with $d^{N-1}\Mybf{x}=d x_1 \ldots dx_{N-1}$ and

1052: $\mu(\Mybf{x})$ being a nontrivial integration measure. Eq.~\re{SoV-gen} defines

1053: the transformation $\Phi_{\mybf{q}}\mapsto\Psi_{\mybf{q},p}$. The inverse

1054: transformation looks as follows

1055: \be

1056: \Phi_{\mybf{q}}(\Mybf{x})\,\delta(p-p') = \vev{p',\Mybf{x}|\Psi_{\mybf{q},p}}

1057: =\int {\cal D}^N z\, \lr{U_{p',\mybf{x}}

1058: (z_1,\ldots,z_N)}^*\Psi_{\mybf{q},p}\,(z_1,\ldots,z_N)\,.

1059: \label{SoV-inv}

1060: \ee

1061: 

1062: To construct the unitary transformation to the SoV representation one has to

1063: specify the complete set of the states $\ket{p,\Mybf{x}}$ and define the

1064: corresponding kernel \re{U-ker}. It is well-known that for the $SL(2)$ spin chain

1065: with periodic boundary conditions, within the framework of the Sklyanin's

1066: approach~\cite{Sklyanin}, the basis vectors $\ket{p,\Mybf{x}}$ can be defined as

1067: eigenvectors of the operator $b(u)$ which is the off-diagonal matrix element of

1068: the monodromy matrix $T_N(u)$, Eq.~\re{Tc}. We will demonstrate that the same

1069: recipe also works for the open spin chain. Namely, the basis vectors

1070: $\ket{p,\Mybf{x}}$ in \re{U-ker} can be defined as the eigenstates of the

1071: operator $B(u)$ entering the expression for monodromy matrix $\mathbb{T}_N(u)$,

1072: Eq.~\re{nTo}.

1073: 

1074: According to \re{nTo}, $B(u)$ is a polynomial in $u$ of degree $2N-1$ with

1075: operator-valued coefficients, $B(u) = 2i(-1)^{N-1} S_- u^{2N-1}+\ldots$. In

1076: addition, it follows from \re{A-minus} that $B(-i/2)=0$ and, moreover,

1077: $B(u)/(2u+i)$ is an even function of $u$. This suggests to remove the

1078: ``kinematic'' zero of $B(u)$ and define the operator

1079: \be\label{hatB}

1080: \widehat B(u) ~=~\frac{B(u)}{2u+i}=(-1)^{N-1}iS_{-}\, \left(u^{2N-2}~+~\widehat

1081: b_2\,u^{2N-4}~+~\ldots~+~\widehat b_N \right),

1082: \ee

1083: with $\widehat b_2,\ldots,\widehat b_N$ being some (commuting) operators. Since

1084: $B(u)=b(-u)a(u)-a(-u)b(u)$ (see Eq.~\re{B}), one finds using $\lr{a(u)}^\dagger =

1085: a(u^*)$ and $\lr{b(u)}^\dagger = b(u^*)$ that $\lr{B(u)}^\dagger=-B(-u^*)$, or

1086: equivalently $(\widehat B(u))^\dagger= \widehat B(u^*)$. Thus, $\widehat B(u)$ is

1087: hermitian operator for real $u$.

1088: 

1089: Following Sklyanin~\cite{Sklyanin}, we identify the eigenstates of the operator

1090: $\widehat B(u)$ as the kernel of the transition operator to the SoV

1091: representation

1092: \be

1093: \widehat{B}(u)\,U_{p,\mybf{x}}(z_1,\ldots,z_N)=(-1)^{N-1}\,p\,(u^2-x_1^2)\cdots(u^2-x_{N-1}^2)\,

1094: U_{p,\mybf{x}}(z_1,\ldots,z_N)\,.

1095: \label{B-pol}

1096: \ee

1097: According to \re{BB}, $[\widehat{B}(u),\widehat{B}(v)]=0$ and, therefore,

1098: $U_{p,\mybf{x}}(z_1,\ldots,z_N)$ does not depend on the spectral parameter $u$.

1099: Due to \re{hatB}, the corresponding eigenvalues are real polynomials in $u^2$ of

1100: degree $N-1$. They can be parameterized by the total momentum $p$ and by the set

1101: of parameters $\Mybf{x}=(x_1,\ldots,x_{N-1})$ which are identified as the

1102: separated variables. Hermiticity of the operator $\widehat{B}(u)$ implies that

1103: $x_k^2$ can be either real, or can appear in complex conjugated pairs,

1104: $x_k^2=(x_j^2)^*$. We will argue in the next section that the separated variables

1105: satisfy a much stronger condition $x_k^2>0$, which together with the symmetry of

1106: \re{B-pol} under $x_k\to -x_k$ allows one to assign to the separated variables

1107: $\Mybf{x}$ \textit{real positive} values. This follows from the requirement that

1108: $U_{p,\mybf{x}}(z_1,\ldots,z_N)$ have to be the eigenstates of the self-adjoint

1109: operator $\widehat{B}(u)$ and, therefore, they have to fulfill the completeness

1110: condition

1111: \be

1112: \int_0^\infty dp \int_{\mathbb{R}_+^{N-1}}

1113: d^{N-1}\Mybf{x}\,\mu(\Mybf{x})\,\lr{U_{p,\mybf{x}}(w_1,\ldots,w_N)}^*\,U_{p,\mybf{x}}(z_1,\ldots,z_N)

1114: %= \mathbb{K}(\vec z |\vec w)

1115: =\prod_{n=1}^N \mathbb{K}(z_n|\bar w_n)

1116: \label{complete}

1117: \ee

1118: where $\mathbb{K}(z |\bar w)=\e^{i\pi s} (z-\bar w)^{-2s}$ is the kernel of the

1119: identity operator (see Eq.~\re{K-ker}).

1120: 

1121: The diagonal element $D(\pm x_k)$ of the monodromy matrix \re{nTo} acts on

1122: $U_{p,\mybf{x}}(w_1,\ldots,w_N)$ as a shift operator%~\cite{?}

1123: \be\label{D-shift}

1124: D(\pm x_k)U_{p,\mybf{x}}(z_1,\ldots,z_N) = \delta(\pm x_k) U_{p,\mybf{x}\pm

1125: i\mybf{e}_k}(z_1,\ldots,z_N)\,.

1126: \ee

1127: Indeed, taking $v=\pm x_k$ in the second fundamental relation \re{BB} and

1128: applying its both sides to $U_{p,\mybf{x}}(z_1,\ldots,z_N)$, one arrives at

1129: \re{D-shift}. The scalar factor $\delta(x_k)$ depends on the normalization of

1130: $U_{p,\mybf{x}}(z_1,\ldots,z_N)$. Applying $U_{p,\mybf{x}}$ to the both sides of

1131: \re{q-det} and taking $u=x_k$ one finds that $\delta(x_k)$ satisfies the relation

1132: \be

1133: \delta(x_k)\delta(-x_k-i)=[(x_k+is)(x_k+i(1-s))]^{2N}\,.

1134: \ee

1135: In \re{D-shift} it was tacitly assumed that the function

1136: $U_{p,\mybf{x}}(z_1,\ldots,z_N)$ can be continued to complex $\Mybf{x}$. Notice

1137: that $U_{p,\mybf{x}\pm i\mybf{e}_k}(z_1,\ldots,z_N)$ is not the eigenfunction of

1138: the operator $\widehat B(u)$ even though it satisfies the differential equation

1139: \re{B-pol}.

1140: 

1141: \subsection{Transition kernel}

1142: 

1143: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1144: \begin{figure}[t]

1145: \psfrag{z1}[cc][cc]{$z_1$}\psfrag{z2}[cc][cc]{$z_2$}\psfrag{zn}[cc][cc]{$z_N$}

1146: %\psfrag{w1}[cc][cc]{$\bar w_1$}

1147: \psfrag{w2}[cc][cc]{$\bar w_2$} \psfrag{wn}[cc][cc]{$\bar w_N$}

1148: \psfrag{dots}[cc][cc]{$\Mybf{\cdots}$} \psfrag{a}[cc][cc]{$\beta_x$}

1149: \psfrag{b}[cc][cc]{$\alpha_x$} \psfrag{c}[cc][rc]{$\alpha_x$}

1150: \psfrag{d}[cc][rc]{$\beta_x$} \centerline{\epsfxsize10.0cm\epsfbox{lamo.eps}}

1151: \vspace*{0.5cm} \caption[]{ The diagrammatical representation of the function

1152: $\Lambda_{x}(z_1,\ldots,z_N|\bar w_2,\ldots,\bar w_N)$.

1153: }%

1154: \label{lamo}%

1155: \end{figure}%

1156: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1157: 

1158: Solving the spectral problem \re{B-pol}, we follow the approach developed in

1159: Ref.~\cite{SoV} in application to the closed spin chain. To begin with, we notice

1160: that the differential equation \re{B-pol} is equivalent to the system of $N$

1161: equations

1162: \begin{equation}

1163: iS_-\,U_{p,\mybf{x}}(z) = p\,U_{p,\mybf{x}}(z)\,,\qquad \widehat{B}(\pm

1164: x_k)\,U_{p,\mybf{x}}(z)=0,\qquad (k=1,...N-1)\,.

1165: \label{BxN}

1166: \end{equation}

1167: Let us consider the second relation and compare it with a similar relation

1168: \re{Bnul} for $u=\pm x_k$. Sending the gauge parameter $w_1$ in \re{Bnul} to

1169: infinity and taking into account \re{tADB} one finds that $B(\pm x_k)$

1170: annihilates the following function %obtained from the kernel of the

1171: %$\mathbb{Q}-$operator as

1172: \be\label{Lambda}

1173: \Lambda_{x}(z_1,\ldots,z_N|\bar w_2\ldots,\bar w_{N})~=~ \lim_{\bar w_1\to\infty}

1174: \bar w_1^{2s}\, Q_x(z_1,\ldots,z_N|\bar w_1,\ldots,\bar w_N)\,.

1175: \ee

1176: Here the additional prefactor is inserted to make the limit finite

1177: \ba

1178: \Lambda_{u}(z_1,\ldots,z_N|\bar w_2\ldots,\bar w_{N}) &=&\e^{i\pi s(2N-1)} \int

1179: \mathcal{D} y_2 \ldots \mathcal{D} y_N\,(y_N-\bar w_N)^{-\beta_u}(z_N-\bar

1180: w_N)^{-\alpha_u}

1181: \label{Lambda-ker}

1182: \\

1183: &&\hspace*{-20mm}\times \prod_{k=2}^N (z_{k-1}-\bar y_k)^{-\alpha_u} (z_k-\bar

1184: y_k)^{-\beta_u}\prod_{n=2}^{N-1} (y_n-\bar w_n)^{-\beta_u}(y_{n+1}-\bar

1185: w_n)^{-\alpha_u}\,. \nonumber

1186: \ea

1187: As before, it is convenient to represent this expression as the Feynman diagram

1188: shown in Fig.~\ref{lamo}. It differs from the Feynman diagram for the

1189: $\mathbb{Q}-$operator (see Fig.~\ref{fig1}) in that two lines attached to the

1190: vertex $\bar w_1$ are removed.

1191: 

1192: By the construction, the $\Lambda-$function satisfies the relation

1193: \be\label{Bn}

1194: \widehat{B}(\pm x_k)\,\Lambda_{x_k}(z_1,\ldots,z_N|\bar w_2\ldots,\bar

1195: w_{N})~=~0\,.

1196: \ee

1197: Let us introduce the integral operator $\Lambda_N(x)$ with the kernel given by

1198: \re{Lambda-ker}. It maps a function of $N-1$ variables

1199: $\Psi_{N-1}(w_2,\ldots,w_N)$ into a function of $N$ variables

1200: $\Psi_N(z_1,\ldots,z_N)$

1201: \ba

1202: \label{Lambda1}

1203: \Psi_N(z_1,\ldots,z_N) &=& [\Lambda_N(u) \, \Psi_{N-1}](z_1,\ldots,z_N) \\

1204: &=& \int \mathcal{D} w_2 \ldots \int \mathcal{D} w_N

1205: \,\Lambda_u(z_1,\ldots,z_N|\bar w_2\ldots,\bar

1206: w_{N})\Psi_{N-1}(w_2,\ldots,w_N)\,, \nonumber

1207: \ea

1208: The operator $\Lambda_N(x)$ defined in this way has a number of remarkable

1209: properties:

1210: \begin{itemize}

1211: \item{Parity:}

1212: \be\label{PL}

1213: \Lambda_N(x)=\Lambda_N(-x)

1214: \ee

1215: \item{Commutativity:}

1216: \be\label{CL}

1217: \Lambda_N(x_1)\Lambda_{N-1}(x_2) = \Lambda_N(x_2)\Lambda_{N-1}(x_1)

1218: \ee

1219: \item{Baxter relation:}

1220: \be

1221: \widehat t_N(x)\,\Lambda_N(x)=\Delta_{+}(x)\,\Lambda_N(x+i)~+~

1222: \Delta_{-}(x)\,\Lambda_N(x-i)

1223: \label{Bax-Lambda}

1224: \ee

1225: \item{Exchange relation:}

1226: \be

1227: \Lambda_N^\dagger(x) \Lambda_N(y) =\varphi(x,y)\cdot

1228: \Lambda_{N-1}(y)\Lambda_{N-1}^\dagger(x)\,,

1229: \label{exchange}

1230: \ee

1231: \end{itemize}

1232: where $x\neq y$ and the scalar function $\varphi(x,y)=\varphi(y,x)$ is defined as

1233: \be

1234: \varphi(x,y)=\e^{4i\pi s}\,a(\alpha_x,\alpha_y)\,a(\beta_x,\beta_y)\,

1235: a(\beta_x,\alpha_y)\,a(\alpha_x,\beta_y)

1236: \label{varphi}

1237: \ee

1238: with $\alpha_x=s-ix$ and $\beta_x=s+ix$,

1239: \be

1240: a(\alpha,\beta)=\e^{-i\pi s}\,

1241: \frac{\Gamma(\alpha+\beta-2s)\Gamma(2s)}{\Gamma(\alpha)\Gamma(\beta)}\,.

1242: \label{a}

1243: \ee

1244: The following comments are in order.

1245: 

1246: Eq.~\re{PL} follows from the parity property of the Baxter $\mathbb{Q}-$operator,

1247: Eq.~\re{parity}. The proof of \re{CL} can be performed diagrammatically with a

1248: help of the permutation identities (see Figs.~\ref{comm-f} and \ref{amp1}). It

1249: goes along the same lines as the proof of commutativity property for the

1250: $\mathbb{Q}-$operator presented at the end of Sect.~3.2. Eq.~\re{Bax-Lambda}

1251: follows immediately from \re{Lambda} and \re{Bax-eq}. The proof of the exchange

1252: relation \re{exchange} is illustrated in Fig.~\ref{fig3}. The product

1253: $\Lambda_N^\dagger(x)\Lambda_N(y)$ corresponds to the left diagram in

1254: Fig.~\ref{fig3}. The left-most vertex in this diagram can be integrated out with

1255: a help of the chain relation (see Fig.~\ref{Chain}) producing a single line with

1256: the index $\alpha_x+\beta_y-2s=-i(x-y)$. Then, one moves this line horizontally

1257: to the right of the diagram by applying the permutation identity (see

1258: Fig.~\ref{comm-f}). Repeating the same steps for the resulting diagram one

1259: finally arrives at the right diagram in Fig.~\ref{fig3} with the additional

1260: prefactor \re{varphi}.

1261: 

1262: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1263: \begin{figure}[t]

1264: \psfrag{=}[cc][cc]{$=$} \psfrag{a1}[cc][cc]{$\beta_x$}

1265: \psfrag{b1}[cc][cc]{$\alpha_x$} \psfrag{a2}[cc][cc]{$\alpha_y$}

1266: \psfrag{b2}[cc][cc]{$\beta_y$}

1267: \centerline{\epsfxsize16.0cm\epsfbox{exchange.eps}} \vspace*{0.5cm}

1268: \caption[]{Exchange relation. Here $\alpha_x=s-ix$ and $\beta_x=s+ix$. }

1269: \label{fig3}

1270: \end{figure}

1271: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1272: 

1273: Taking into account the properties of the $\Lambda-$operator, it becomes

1274: straightforward to write a general solution to the system \re{BxN}

1275: \be

1276: U_{p,\mybf{x}}(z_1,\ldots,z_N)= p^{Ns-1/2} \int \mathcal{D}w_N\, \e^{ip\,w_N}\,

1277: U_{\mybf{x}}(\vec z\,;\bar w_N)\,,

1278: \label{B-ei}

1279: \ee

1280: where $U_{\mybf{x}}(\vec z\,;\bar w_N)$ is factorized into the product of $N-1$

1281: operators

1282: %\footnote{The

1283: %only difference with the closed spin chain is in the explicit form of the operator

1284: %$\Lambda_N(x_1)$. In our case it is given by two layers.}

1285: \be

1286: U_{\mybf{x}}(\vec z\,;\bar w_N)=

1287: \left[\Lambda_N(x_1)\,\Lambda_{N-1}(x_2)\ldots\Lambda_2(x_{N-1})

1288: \right](z_1,\ldots,z_N|\bar w_N)\,,

1289: \label{U}

1290: \ee

1291: with $\vec z=(z_1,\ldots,z_N)$ and the additional factor $p^{Ns-1/2}$ introduced

1292: in \re{B-ei} for the later convenience. Indeed, the first relation in \re{BxN} is

1293: satisfied due to invariance of \re{U} under translations $z_k \to z_k+\epsilon$

1294: and $\bar w_N \to\bar w_N+\epsilon$ with $\epsilon$ real. It follows from \re{PL}

1295: and \re{CL} that $U_{\mybf{x}}(\vec z\,;\bar w_N)$ is an even symmetric function

1296: of $x_1,\ldots,x_{N-1}$. Since $\widehat{B}(\pm x_1)\,U_{\mybf{x}}(\vec z\,;\bar

1297: w_N)=0$ by virtue of \re{U} and \re{Bn}, the second relation in \re{BxN} is

1298: fulfilled for arbitrary $k$. Notice that the kernel $U_{\mybf{x}}(\vec z\,;\bar

1299: w_N)$ satisfies a multi-dimensional Baxter relation

1300: \be

1301: \widehat t_N(x_k)\,U_{\mybf{x}}(\vec z\,;\bar w_N)=\Delta_+(x_k)

1302: U_{\mybf{x}+i\mybf{e}_k}(\vec z\,;\bar w_N)+\Delta_-(x_k)

1303: U_{\mybf{x}-i\mybf{e}_k}(\vec z\,;\bar w_N)\,,

1304: \label{Bax-U}

1305: \ee

1306: where $\Mybf{e}_k$ denotes a unit basis vector in the $\Mybf{x}-$space,

1307: $\Mybf{x}=\sum_k x_k \Mybf{e}_k$. Eq.~\re{Bax-U} follows from the similar

1308: property of the $\Lambda-$operator, Eq.~\re{Bax-Lambda}, and the symmetry of the

1309: kernel under permutations of $x-$variables.

1310: 

1311: Eqs.~\re{B-ei}  and \re{U} define the transition kernel

1312: $U_{p,\mybf{x}}(z_1,\ldots,z_N)$ to the SoV representation for the homogeneous

1313: open spin chain. Remarkably enough, these expressions

1314: %for the transition kernel $U_{p,\mybf{x}}(z_1,\ldots,z_N)$, Eqs.~\re{B-ei} and \re{U}, has

1315: have the same form as for the closed $SL(2)$ spin chain~\cite{SoV}. The only

1316: difference between the two cases is in the definition of the $\Lambda-$operator.

1317: Diagrammatical representation for the transition kernel \re{U} is shown in

1318: Fig.~\ref{opyr}. The corresponding Feynman diagram has a pyramidal form which

1319: reflects the structure of the kernel \re{U}. It consists of $(N-1)-$rows with

1320: each row representing a single $\Lambda-$operator.

1321: 

1322: 

1323: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1324: \begin{figure}[t]

1325: \psfrag{=}[cc][cc]{$=$} \psfrag{a1}[cc][cc]{$\beta_{x_1}$}

1326: \psfrag{b1}[cc][cc]{$\alpha_{x_1}$} \psfrag{a2}[cc][cc]{$\beta_{x_3}$}

1327: \psfrag{b2}[cc][cc]{$\alpha_{x_3}$} \psfrag{a3}[cc][cc]{$\beta_{x_2}$}

1328: \psfrag{b3}[cc][cc]{$\alpha_{x_2}$}

1329: 

1330: %\psfrag{1a}[cc][cc]{$\beta_{-x_1}$} \psfrag{1b}[cc][cc]{$\alpha_{-x_1}$}

1331: %\psfrag{2a}[cc][cc]{$\beta_{-x_2}$} \psfrag{2b}[cc][cc]{$\alpha_{-x_2}$}

1332: %\psfrag{3a}[cc][cc]{$\beta_{-x_3}$} \psfrag{3b}[cc][cc]{$\alpha_{-x_3}$}

1333: 

1334: \psfrag{z1}[cc][cc]{$z_1$}

1335: \psfrag{z2}[cc][cc]{$z_2$}

1336: 

1337: \psfrag{z3}[cc][cc]{$z_3$} \psfrag{z4}[cc][cc]{$z_4$} \psfrag{w1}[cc][cc]{$\bar

1338: w_4$}

1339: 

1340: \centerline{\epsfysize10.0cm\epsfbox{pyramid.eps}} %\vspace*{0.5cm}

1341: \caption[]{The diagrammatic representation of the kernel

1342: $U_{\Mybf{x}}(\vec{z},w_N)$ for $N=4$.}

1343: \label{opyr}

1344: \end{figure}

1345: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1346: 

1347: It remains to verify that the kernel \re{B-ei} satisfies for real $\Mybf{x}$ the

1348: completeness condition \re{complete}. As we will show in Sect.~5, the

1349: transformation to the SoV representation for $N=2$ open spin chain coincides with

1350: the Fourier-Jacobi transform (see, e.g. Ref.~\cite{Koor}). Then, reality

1351: condition for the separated variable $x$ and completeness condition for

1352: $U_{p,x}(z_1,z_2)$ follow immediately from the properties of the Fourier-Jacobi

1353: transform. For $N\ge 3$ some arguments will be presented in the next subsection.

1354: 

1355: \subsection{Integration measure}

1356: 

1357: Let us demonstrate that the transition kernel defined in Eqs.~\re{B-ei} and

1358: \re{U} satisfies the orthogonality condition

1359: \ba

1360: \vev{p',\Mybf{x'}|p,\Mybf{x}} &=& \int \mathcal{D}^N z\,

1361: U_{p,\mybf{x}}(z_1,...,z_N) \lr{U_{p',\mybf{x'}}(z_1,...,z_N)}^* \nonumber

1362: \\

1363: &=& %(2\pi )^N

1364: \delta(p-p') \left\{\delta(\Mybf{x}-\Mybf{x'})+ \cdots \right\}

1365: \frac{\mu^{-1}(\Mybf{x})}{(N-1)!}\,,

1366: \label{U-ort}

1367: \ea

1368: and calculate the integration measure $\mu(\Mybf{x})$. Here

1369: $\delta(\Mybf{x}-\Mybf{x'})\equiv \prod_{k=1}^{N-1}\delta(x_k-x'_k)$ and

1370: ${\Mybf{x}}=(x_1,\ldots,x_{N-1})$ take positive real values, $x_k > 0$. Ellipses

1371: denote the terms with all possible permutations inside the set

1372: $\Mybf{x}=(x_1,\ldots,x_{N-1})$.

1373: 

1374: The calculation of \re{U-ort} repeats similar analysis for the closed spin chain

1375: described at length in Ref.~\cite{SoV}. Substitution of \re{B-ei} into \re{U-ort}

1376: yields

1377: \be

1378: \vev{p',\Mybf{x'}|p,\Mybf{x}}=(pp')^{Ns-1/2} \int \mathcal{D} w_N

1379: \e^{ip\,w_N}\int \mathcal{D} w_N'\e^{-ip'\,\bar w_N'}

1380: \vev{w_N',\Mybf{x}'|w_N,\Mybf{x}}\,,

1381: \label{limit}

1382: \ee

1383: where the notation was introduced for the ket-vector

1384: $\vev{z_1,\ldots,z_N|w_N,\Mybf{x}}=U_{\mybf{x}}(\vec z\,;\bar w_N)$, or

1385: equivalently

1386: \be

1387: \ket{w_N,\Mybf{x}}=\Lambda_N(x_1)\Lambda_{N-1}(x_2)\ldots

1388: \Lambda_2(x_{N-1})\ket{w_N}\,,

1389: \label{L-comp}

1390: \ee

1391: with $\ket{w_N}$ being a ``single-particle state''. We recall that the operator

1392: $\Lambda_k(u)$ maps $(k-1)-$particle state into $k-$particle one, so that a

1393: composition of the $\Lambda-$operators in \re{L-comp} produces the $N-$particle

1394: state. Calculating the scalar product $\vev{w_N',\Mybf{x}'|w_N,\Mybf{x}}$ one

1395: applies systematically the exchange relation \re{exchange} and obtains

1396: \be

1397: \vev{w_N',\Mybf{x}'|w_N,\Mybf{x}}~=~c(\Mybf{x},\Mybf{x}')

1398: \vev{w_N'|\lr{\Lambda_2^\dagger(x_{N-1}')\Lambda_2(x_1)} \ldots

1399: \lr{\Lambda_2^\dagger(x_{1}')\Lambda_2(x_{N-1})}|w_N}\,,

1400: \label{ww0}

1401: \ee

1402: where $c(\Mybf{x},\Mybf{x}')=\prod_{1\le j,\,k\le N-2 \atop j + k \le

1403: N-1}\varphi(x_j,x_k')$. Notice that the exchange relation \re{exchange} holds

1404: only for $x\neq y$. Therefore, calculating \re{ww0} we have tacitly assumed that

1405: $x_j\neq x_k'$ for $j + k \le N-1$, or equivalently that all factors

1406: $\varphi(x_j,x_k')$ are finite. For $N\ge 3$ the matrix element entering \re{ww0}

1407: can be represented as follows

1408: \be\label{ww}

1409: %\vev{w_N',\Mybf{x}'|w_N,\Mybf{x}}=c(\Mybf{x},\Mybf{x}')

1410: %\int \prod_{i=1}^{N-1}\mathcal{D}w_i\,

1411: \int \mathcal{D} w_1 \ldots \int \mathcal{D} w_{N-2}

1412: \vev{w_N',x'_{N-1}|w_1,x_1}\vev{w_1,x'_{N-2}|w_2,x_2}\ldots

1413: \vev{w_{N-2},x'_{1}|w_{N},x_{N-1}}\,,

1414: \ee

1415: where $\vev{w',x'|w,x}=[\Lambda_2^\dagger(x')\,\Lambda_2(x)](w';w)$. Thus, the

1416: calculation of the scalar product \re{limit} for arbitrary $N$ is reduced to the

1417: calculation of $\vev{w',x'|w,x}$ at $N=2$. Given that the separated variables at

1418: $N=2$ take real positive values we deduce from \re{ww0} that the same holds true

1419: for arbitrary $N$.

1420: 

1421: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1422: \begin{figure}[t]

1423: \psfrag{=}[cc][cc]{$=$} \psfrag{a}[cc][cc]{$\beta_{x'}$}

1424: \psfrag{b}[cl][cc]{$\alpha_{x'}$} \psfrag{d}[cc][cc]{$\beta_{x}$}

1425: \psfrag{c}[cl][cc]{$\alpha_{x}$} \psfrag{a1}[cr][cc]{$\beta_{x'}+\epsilon$}

1426: \psfrag{d1}[cr][cc]{$\beta_{x}+\epsilon$}

1427: 

1428: \psfrag{w}[cc][cc]{$w$}\psfrag{w1}[cc][cc]{$w'$}

1429: 

1430: \centerline{\epsfysize7.0cm\epsfbox{osp.eps}} \vspace*{0.5cm} \caption[]{The

1431: scalar product $\vev{w',x'|w,x}$ for $N=2$}

1432: \label{osp}

1433: \end{figure}

1434: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1435: 

1436: 

1437: To calculate the scalar product at $N=2$ we apply the diagrammatical approach of

1438: Ref.~\cite{SoV} and represent the matrix element $\vev{w',x'|w,x}$ as the Feynman

1439: diagram shown in Fig.~\ref{osp}. One expects from \re{U-ort} that

1440: $\vev{w',x'|w,x}\sim \delta(x-x')$, so that the scalar product $\vev{w',x'|w,x}$

1441: should be understood as a distribution. To find its explicit form we regularize

1442: the corresponding Feynman integral by introducing a small parameter $\epsilon$

1443: and shifting the indices of two lines as indicated in Fig.~\ref{osp}. Under such

1444: regularization, the Feynman integral remains finite at $x=x'$ and it can be

1445: calculated exactly with a help of the chain relation and permutation identity

1446: (see Figs.~\ref{Chain} and \ref{comm-f}). The calculation is straightforward and

1447: some details can be found in Ref.~\cite{SoV}. Going over to the momentum

1448: representation and taking the limit $\epsilon\to 0$, one finds (for $x\,,x' >0$)

1449: \be

1450: \int \mathcal{D} w\,\e^{ipw}\, \vev{w',x'|w,x}~ =~2\pi\,\e^{ipw'}\,

1451: p^{-2s}\,\Gamma^5(2s)\,\frac{|\Gamma(i(x+x'))|^2}{|\Gamma(s+ix)\Gamma(s+ix')|^4}

1452: \,\delta(x-x')\,.

1453: \label{Fourier-2}

1454: \ee

1455: Substituting \re{ww0} and \re{ww} into \re{limit} and taking into account

1456: \re{Fourier-2} one obtains after some algebra

1457: \ba

1458: \vev{p',\Mybf{x'}|p,\Mybf{x}}&=&(2\pi)^{N-1} \delta(p-p')

1459: \prod_{k=1}^{N-1}\delta(x_k-x_{N-k}')\cdot\Gamma^{N}(2s)\,\prod_{k=1}^{N-1}

1460: \left[\frac{\Gamma(s-ix_k)\Gamma(s+ix_k)}{\Gamma(2s)}\right]^{-2N}

1461: \nonumber\\[2mm]

1462: & & \hspace*{-20mm}\times\left(\prod_{1\leq j\leq k\leq

1463:   N-1}\frac{x_k+x_j}{\pi}\sinh\pi(x_k+x_j) \prod_{1\leq j<k\leq

1464:   N-1}\frac{x_k-x_j}{\pi}\sinh\pi(x_k-x_j)\right)^{-1}\,.

1465: \label{measure-1}

1466: \ea

1467: %\ba\label{measure-1}

1468: %\vev{p',\Mybf{x'}|p,\Mybf{x}}&=&(2\pi)^{N-1} \Gamma^{N}(2s)\,\left(\prod_{1\leq j<k\leq

1469: %  N-1}\frac{x_k-x_j}{\pi}\sinh\pi(x_k-x_j)\right)^{-1}\,

1470: %\nonumber\\[3mm]

1471: %&

1472: %\times&\left(\prod_{1\leq j\leq k\leq

1473: %  N-1}\frac{x_k+x_j}{\pi}\sinh\pi(x_k+x_j)\right)^{-1}\,

1474: %\prod_{k=1}^{N-1}\left[\frac{\Gamma(s-ix_k)\Gamma(s+ix_k)}{\Gamma(2s)}\right]^{-2N}

1475: %\nonumber\\[3mm]

1476: %&\times&\delta(p-p')\prod_{k=1}^{N-1}\delta(x_k-x_{N-k}')\,.

1477: %\ea

1478: We recall that the calculation was performed under assumption that $x_j\neq x_k'$

1479: for $j + k \le N-1$. Since the kernel $U_{p,\mybf{x}}$ is a symmetric function of

1480: $\Mybf{x}$, $\vev{p',\Mybf{x'}|p,\Mybf{x}}$ should possess the same property.

1481: This allows one to relax the above assumption and replace the product of

1482: delta-functions $\prod_{k=1}^{N-1}\delta(x_k-x_{N-k}')$ in the r.h.s.\ of

1483: \re{measure-1} by the sum $\sum_{\cal S}\delta(\Mybf{x}-{\cal S}\Mybf{x}')$ over

1484: all permutations inside the set $\Mybf{x}'=\lr{x'_1,\ldots,x'_{N-1}}$.

1485: 

1486: Matching \re{measure-1} into \re{U-ort}, one finds the expression for the

1487: integration measure in the SoV representation

1488: \ba\label{mu}

1489: \mu(\Mybf{x})&=&\frac{\Gamma^{-N}(2s)}{(N-1)!(2\pi)^{N-1}}

1490: \prod_{k=1}^{N-1}\left[\frac{\Gamma(s-ix_k)\Gamma(s+ix_k)}{\Gamma(2s)}\right]^{2N}

1491: %\prod_{k=1}^{ N-1}\frac{2x_k}{\pi}\sinh(2\pi x_k)

1492: %\nonumber

1493: \\[3mm]

1494: &&\times\prod_{1\leq j<k\leq

1495:   N-1}\frac{x_k^2-x_j^2}{2\pi^2}\left[\cosh(2\pi x_k)-\cosh(2\pi x_j)\right]\prod_{k=1}^{

1496:   N-1}\frac{2x_k}{\pi}\sinh(2\pi x_k)\,.

1497: \nonumber

1498: \ea

1499: This expression has the following properties. As expected, $\mu(\Mybf{x})$ is an

1500: even function of the separated variables. It takes nonnegative values for real

1501: $\Mybf{x}=(x_1,\ldots,x_{N-1})$ and vanishes on the hyperplanes $x_j=x_k$.

1502: 

1503: After analytical continuation to complex $\Mybf{x}$, the measure

1504: $\mu({\Mybf{x}})$ becomes a meromorphic function of $x_k$ $(k=1,\ldots,N-1$) with

1505: poles of the order $2N$ located along the imaginary axis at $x_k=\pm i(s+n)$ with

1506: $n\in\mathbb{N}$. The measure decreases exponentially fast when one of the

1507: separated variables, say $x_k$, goes to infinity along the real axis

1508: % while remaining $\Mybf{x}-$variables take finite values

1509: \be\label{m-lim}

1510: \mu(\Mybf{x}) ~\sim \,\e^{-2\pi |x_k|}x_k^{2Ns-3}\,.

1511: \ee

1512: %as  $x_k\to\infty$ and $\Im x_k={\rm fixed}$.

1513: One verifies that the measure \re{mu} satisfies the functional relation

1514: \be

1515: \frac{\mu(\Mybf{x}+i\Mybf{e}_k)}{\mu(\Mybf{x})}=\frac{x_k+i}{x_k}

1516: \lr{\frac{x_k+is}{x_k+i(1-s)}}^{2N}\prod_{j\neq

1517: k}\frac{x_k-x_j+i}{x_k-x_j}\frac{x_k+x_j+i}{x_k+x_j}\,,

1518: \label{measure-rec}

1519: \ee

1520: with $\Mybf{e}_k$ defined in \re{Bax-U}.

1521: 

1522: %Eq.~\re{measure-rec} can be also obtained in a pure algebraic way following the

1523: %Sklyanin's approach~\cite{Sklyanin}. Obviously, the relation \re{measure-rec}

1524: %alone would allow one to determine the measure up to an arbitrary periodic

1525: %function.

1526: 

1527: It is instructive to compare \re{mu} with a similar expression for the

1528: integration measure for the \textit{closed} spin chain \cite{SoV}

1529: \be\label{mfi}

1530: \mu_{\rm cl}(\Mybf{x})=\prod_{j,k=1\atop

1531: j<k}^{N-1}{(x_k-x_j)}\,\sinh(\pi(x_k-x_j)) \,

1532: \prod_{k=1}^{N-1}\left[\Gamma(s+ix_k)\Gamma(s-ix_k)\right]^N\,.

1533: \ee

1534: One observes that $\mu_{\rm cl}(\Mybf{x})$ enters as a factor into the expression

1535: for $\mu(\Mybf{x})$, Eq.~\re{mu}.

1536: 

1537: \subsection{Eigenfunctions in the SoV representation}

1538: 

1539: The eigenfunctions $\Psi_{\mybf{q},p}(z_1,\ldots,z_N)$ are orthogonal to each

1540: other for different sets of quantum numbers with respect to the $SL(2)$ scalar

1541: product \re{norm}.  In the SoV representation the same condition looks as follows

1542: \be

1543: \vev{\Psi_{\mybf{q'},p'}|\Psi_{\mybf{q},p}}%_{_{\rm SL(2,\mathbb{R})}}

1544: = \vev{\Phi_{\mybf{q'}}|\Phi_{\mybf{q}}}_{_{\rm SoV}}\delta(p-p')

1545: =\delta(p-p')\,\delta_{\mybf{q},\mybf{q'}}\,,

1546: \label{ortho}

1547: \ee

1548: where the scalar product in the SoV representation is given by

1549: \be

1550: \vev{\Phi_{\mybf{q'}}|\Phi_{\mybf{q}}}_{_{\rm SoV}}= \int_{\mathbb{R}^+_{N-1}}

1551: d^{N-1} \Mybf{x}\,\mu(\Mybf{x}) \lr{\Phi_{\mybf{q}'}(x_1,\ldots,x_{N-1})}^*

1552: \Phi_{\mybf{q}}(x_1,\ldots,x_{N-1})\,.

1553: \label{SoV-sp}

1554: \ee

1555: We recall that the momentum $p$ takes real positive values whereas the spectrum

1556: of the integrals of motion $\Mybf{q}=(q_2,\ldots,q_N)$ is discrete~\cite{DKM99}.

1557: 

1558: To define the eigenfunction in the SoV representation,

1559: $\Phi_{\mybf{q}}(\Mybf{x})$, we substitute $\Psi_{\mybf{q},p}(z_1,\ldots,z_N)$ in

1560: \re{eit} by its integral representation \re{SoV-gen}. Following the standard

1561: procedure~\cite{Sklyanin} and making use of Eqs.~\re{Bax-U} and \re{measure-rec} one can

1562: show that $\Phi_{\mybf{q}}(\Mybf{x})$ satisfies the $(N-1)-$dimensional Baxter

1563: equation

1564: \be

1565: \label{PhiB}

1566: t_{N}(x_k) \,\Phi_{\mybf{q}}(\Mybf{x})~=~ \Delta_+(x_k)

1567: \Phi_{\mybf{q}}(\Mybf{x}+i\Mybf{e}_k)~+~ \Delta_-(x_k)

1568: \Phi_{\mybf{q}}(\Mybf{x}-i\Mybf{e}_k)\,,

1569: \ee

1570: where $t_N(x_k)$ is the eigenvalue of the transfer matrix, Eq.~\re{eit}.

1571: %We would like to stress that deriving this equation one tacitly assumes that the

1572: %function $\Phi_{\mybf{q}}(\Mybf{x})$ can be analytically continued from the real

1573: %$x_k-$axis to the complex $x_k-$plane.

1574: As before, to solve this equation one has to specify additional conditions for

1575: $\Phi_{\mybf{q}}(\Mybf{x})$.

1576: 

1577: Using \re{SoV-inv} one can show that $\Phi_{\mybf{q}}(\Mybf{x})$ is a polynomial

1578: in $\Mybf{x}=(x_1,\ldots,x_{N-1})$. The proof goes along the same lines as

1579: analysis of analytical properties of the Baxter $\mathbb{Q}-$operator in

1580: Sect.~3.3. Namely, substituting the expression for the kernel \re{B-ei} into

1581: \re{SoV-inv} and applying \re{drift}, one can express the r.h.s.\ of \re{SoV-inv}

1582: as a nested contour integral. Analytical properties of the function

1583: $\Phi_{\mybf{q}}(\Mybf{x})$ are in the one-to-one correspondence with the

1584: properties of this integral.

1585: 

1586: It is easy to see that polynomial solutions to \re{PhiB} can be represented in

1587: the factorized form

1588: \be\label{PhiQ}

1589: \Phi_{\mybf{q}}(\Mybf{x})~=~c_{\mybf{q}}\, Q_{\mybf{q}}(x_1)\ldots

1590: Q_{\mybf{q}}(x_{N-1})\,,

1591: \ee

1592: where $Q_{\mybf{q}}(x)$ is the eigenvalue of the Baxter $\mathbb{Q}-$operator,

1593: Eq.~\re{roots}, and the coefficient $c_{\mybf{q}}$ is fixed by the normalization

1594: condition \re{ortho}. Substituting \re{PhiQ} into \re{SoV-sp} and taking into

1595: account that $Q_{\mybf{q}}(x)$ is a real function, Eq.~\re{roots}, we find that

1596: the solutions to the Baxter equation satisfy the orthogonality condition

1597: \be\label{Q-ort}

1598: \int_{\mathbb{R}^{N-1}_+} d^{N-1}\Mybf{x}\,\mu(\Mybf{x})

1599: \prod_{k=1}^{N-1}Q_{\mybf{q}'}(x_k) \, Q_{\mybf{q}}(x_k) ~\sim

1600: \delta_{\mybf{q},\mybf{q}'}\,,

1601: \ee

1602: with the measure given by \re{mu}.

1603: 

1604: Thus, having determined the eigenvalues of Baxter operator $Q_{\mybf{q}}(x)$ one

1605: would be able to restore both the energy and the corresponding eigenfunction,

1606: Eqs.~\re{QE} and \re{SoV-gen}, respectively. The solutions to the Baxter equation

1607: for the $SL(2,\mathbb{R})$ open spin chain have been studied in \cite{DKM99}. It

1608: turns out that the ground state $\Omega_p(\vec z)$ of the model can be found

1609: exactly. This state has the total $SL(2)$ spin $h=0$,

1610: %$\vec{S}^2\,\Omega_p(\vec{z})=Ns(Ns-1)\Omega_p(\vec{z})$,

1611: Eq.~\re{S2-eig}, and has the form%the energy $E_{\mybf{q}}=0$

1612: ~\footnote{In the terminology of the Algebraic Bethe Ansatz, $\Omega_p(\vec z)$

1613: is a pseudovacuum state.}

1614: \be

1615: \Omega_p(\vec z)=\frac{p^{2s-1}}{\Gamma(2s)}\int \mathcal{D} w\, \e^{ipw}

1616: \prod_{k=1}^N (z_k-\bar w)^{-2s}\,.

1617: \label{pseudo}

1618: \ee

1619: Indeed, $\Omega_p(\vec z)$ diagonalizes simultaneously the operators of

1620: two-particle spins, $(J_{n,n+1}-2s)\Omega_p(\vec z)=0$, Eq.~\re{JJ}, and leads

1621: the energy $E_{\mybf{q}}=0$. It is interesting to notice that $\Omega_p(\vec z)$

1622: is also the ground state of the homogeneous $SL(2,\mathbb{R})$  \textit{closed}

1623: spin chain~\cite{SoV}.

1624: 

1625: The state \re{pseudo} diagonalizes the Baxter operator

1626: $\mathbb{Q}(u)\ket{\Omega_p} = \ket{\Omega_p}$, or equivalently

1627: $Q_{\mybf{q}}(u)=1$. As a consequence, it admits the following integral

1628: representation (see Ref.~\cite{SoV})

1629: %~\cite{?}

1630: \footnote{Notice that this relation takes the same form both for the open and

1631: closed spins chain whereas the expressions for the integration measure and the

1632: transition kernel to the SoV representation are different in the two cases.}

1633: \be

1634: \Omega_p(\vec z)=p^{Ns-1/2}\e^{-i\pi s(2N-1)}\int_{\mathbb{R}^{N-1}_+}

1635: d^{N-1}\Mybf{x}\,\mu(\Mybf{x}) U_{p,\mybf{x}}(\vec z)\,.

1636: \label{Omega-SoV}

1637: \ee

1638: Let us calculate the $SL(2)$ scalar product $\vev{\Omega_p|\Omega_{p'}}$ and use

1639: two different expressions for $\Omega_p(\vec z)$, Eqs.~\re{pseudo} and

1640: \re{Omega-SoV}. Equating the two expressions, one finds that the integration

1641: measure \re{mu} satisfies the normalization condition

1642: \be\label{mN}

1643: \int_{\mathbb{R}^{N-1}_+} d\Mybf{x}\,\mu(\Mybf{x}) ~=~\frac{1}{\Gamma(2Ns)}\,.

1644: \ee

1645: 

1646: The relation \re{Omega-SoV} allows one to establish the equivalence between the

1647: SoV and the ABA methods for the $SL(2,\mathbb{R})$ open spin chain. Let

1648: $\lambda_1,\ldots,\lambda_h$ be the Bethe roots, or equivalently zeros of the

1649: polynomial $Q_{\mybf{q}}(x)$, Eq.~\re{roots}. Applying $\widehat

1650: B(\lambda_1)\ldots \widehat B(\lambda_h)$ to the both sides of \re{Omega-SoV} and

1651: taking into account \re{B-pol} we obtain

1652: \be\label{ABASOV}

1653: \Psi_{p,\mybf{q}}(\vec z)=\widehat B(\lambda_1)\ldots \widehat B(\lambda_h) \,

1654: \Omega_p(\vec{z})~=~ c(p)\int_{\mathbb{R}^{N-1}_+} d\Mybf{x}\,\mu(\Mybf{x})\,

1655: U_{p,x}(\vec{z})\prod_{k=1}^{N-1}Q_{\mybf{q}}(x_k)\,,

1656: \ee

1657: where $c(p)$ is the normalization factor. In \re{ABASOV}, the first relation

1658: coincides with the ABA representation for the eigenstate of the model while the

1659: second one defines the same eigenstate in the SoV representation.

1660: 

1661: The explicit form of the eigenfunctions in the SoV representation \re{PhiQ}

1662: suggests that there exists a relation between the Baxter $\mathbb{Q}-$operator

1663: and the transition kernel $U_{p,\mybf{x}}(\vec{z})$~\cite{KuS}. In the case of

1664: the closed spin chain it has been established in Ref.~\cite{SoV}. It turns out

1665: that this relation is universal and it also holds for the open spin chain

1666: \be\label{QU}

1667:  U_{p,\mybf{x}}(\vec{z})~=~[\mathbb{Q}(x_1)\,\ldots\mathbb{Q}(x_{N-1})\,

1668: \Theta_p\,](z_1,\ldots,z_N)\,.

1669: \ee

1670: Here %$\mathbb{Q}(u)$ is the Baxter operator in the model under consideration and

1671: $\Theta_p(\vec z)$ is a certain $\Mybf{x}-$independent function of $\vec z$,

1672: which does not belong to the quantum space of the model.%

1673: \footnote{In Ref.~\cite{SoV}, the function $\Theta_p(\vec z)$ was defined as a

1674: limiting case of the state $\ket{\Omega_{\bar w_0,\bar w_N}}$ belonging to the

1675: Hilbert space of the model.} Since $\mathbb{Q}(is)=\mathbb{K}$, the function

1676: $\Theta_p(\vec z)$ is equal to $U_{p,\mybf{x}}(\vec{z})$ for special values of

1677: the $x-$variables, $x_1=\ldots=x_{N-1}=is$ that we denote as

1678: $U_{p,{is}}(\vec{z})$. The expression for $U_{p,{is}}(\vec{z})$ can be easily

1679: obtained from diagrammatic representation of the kernel (see Fig.~\ref{opyr})

1680: \be\label{exOm}

1681: \Theta_p(\vec{z})~=~U_{p,{is}}(\vec{z})~=~p^{Ns-1/2}\e^{i\pi s(N-1)}\,

1682: \e^{ipz_{N}}\,.

1683: \ee

1684: Notice that $\Theta_p(\vec{z})$ depends only on a single variable $z_N$ and,

1685: therefore, it is not normalizable with respect to the $SL(2)$ scalar product

1686: \re{norm}.

1687: 

1688: The proof of \re{QU} can be performed diagrammatically and it repeats similar

1689: analysis in Ref.~\cite{SoV}. Another way to verify \re{QU} is to use the

1690: following identities

1691: \ba

1692: &&\hspace*{-10mm}\int\mathcal{D}^N\!w \, Q_u(\vec z|\vec w)\,\e^{ipw_{N}} =

1693: \e^{-i\pi s}\int \mathcal{D} w_N\, \Lambda_u(z_{N-1},z_N|\bar w_N) \e^{ipw_{N}}

1694: %\equiv [\Lambda_2(u)\Theta_p\,](z_{N-1},z_N)

1695: \label{Q-red}

1696: \\

1697: &&\hspace*{-10mm}\int\mathcal{D}^N\!w \, Q_u(\vec z|\vec

1698: w)\Lambda_v(w_k,\ldots,w_N|\bar y_{k+1},\ldots,\bar y_N) = \e^{-i\pi

1699: s}[\Lambda_{k+1}(u)\Lambda_k(v)](z_{k-1},\ldots,z_N;\bar y_{k+1},\ldots,\bar

1700: y_N)\,, \nonumber

1701: \ea

1702: with $\vec z=(z_1,\ldots,z_N)$, $\vec w=(\bar w_1,\ldots,\bar w_N)$ and

1703: $\mathcal{D}^N w=\prod_{n=1}^N\mathcal{D} w_n$. We recall that the

1704: $\Lambda-$operator was defined in Eqs.~\re{Lambda1} and \re{Lambda-ker}. To

1705: derive \re{Q-red} one substitutes the expression for $Q_u(\vec z|\vec w)$,

1706: Eq.~\re{Q-explicit}, and integrates over ``free'' vertices $w_1,\ldots,w_{k-1}$

1707: with a help of the identity \re{drift} for $\Psi(w)=1$. Eqs.~\re{Q-red} can be

1708: rewritten symbolically as

1709: \be

1710: \mathbb{Q}(u)\ket{\Theta_p} =  \e^{-i\pi s}\Lambda_2(u)\ket{\Theta_p} \,,\qquad

1711: \mathbb{Q}(u)\Lambda_k(v)=\e^{-i\pi s}\Lambda_{k+1}(u)\Lambda_k(v)\,.

1712: \ee

1713: Applying these relations one verifies that \re{QU} coincides with \re{B-ei} and

1714: \re{U}.

1715: 

1716: \section{Relation to the Wilson polynomials}

1717: 

1718: In this section, we consider the open spin chain with $N=2$ sites. We will

1719: demonstrate that in that case the eigenvalues of the Baxter $\mathbb{Q}-$operator

1720: are given by the Wilson polynomials~\cite{Koekoek} and the unitary transformation

1721: to the SoV representation coincides with the Fourier-Jacobi

1722: transformation~\cite{Koor}.

1723: 

1724: At $N=2$ the Hamiltonian of the open spin chain \re{H} equals

1725: %differs by factor 2 from the Hamiltonian of the closed spin chain,

1726: $\mathcal{H}_2=H_{12}=2[\psi(J_{12})-\psi(2s)]$. Its eigenstates are uniquely

1727: fixed by the values of the momentum $p$ and the total $SL(2)$ spin $h$,

1728: Eq.~\re{S2-eig} and are given by

1729: \be

1730: \Psi_{p,h}(z_1,z_2) = \frac{p^{2s-1}}{\Gamma(2s)}\int \mathcal{D} w\, \e^{ipw}

1731: \frac{(z_1-z_2)^h}{(z_1-\bar w)^{2s+h}(z_1-\bar w)^{2s+h}}\,.

1732: \label{Psi-N=2}

1733: \ee

1734: Substituting \re{Psi-N=2} into \re{Q-cont}, one finds the

1735: eigenvalues of the $N=2$ Baxter operator after some algebra as%

1736: \footnote{As was explained in Sect.~3.3, due to the $SL(2)$ invariance of the

1737: Baxter operator, one can calculate $Q_h(u)$ by substituting

1738: $\Psi(z_1,z_2)=(z_1-z_2)^h$ into \re{Q-cont}.}

1739: \be

1740: Q_h(u)={}_4 F_3\lr{{{-N,N+4s-1,s+iu,s-iu}\atop {2s,2s,2s}}\bigg|1}=\left[

1741: \frac{\Gamma(2s)}{\Gamma(2s+N)}\right]^3W_N(u^2,s,s,s,s)\,,

1742: \ee

1743: where $W_N(u^2)$ is the Wilson polynomial~\cite{Koekoek}. It is interesting to

1744: note that the solution of the $N=2$ Baxter equation for the closed spin chain is

1745: given by the continuous Hahn polynomials (see e.g. Ref.~\cite{SoV}). The

1746: polynomials $Q_h(x)$ are orthogonal on the half-axis $x>0$ with respect to the

1747: scalar product \re{Q-ort} with the measure \re{mu} given by

1748: \be

1749: \mu_{N=2}(x) %= \mu_{\rm closed}(x) \lr{a(s+ix,s+ix) a(s-ix,s-ix)}^{-1}

1750: =\frac1{2\pi}\left|\frac{\Gamma^4(s+ix)}{\Gamma(2ix)\Gamma^3(2s)} \right|^2\,.

1751: \ee

1752: This property is in a perfect agreement with the orthogonality condition for the

1753: Wilson polynomials \cite{Koekoek}.

1754: 

1755: The  eigenvalue of the $N=2$ Hamiltonian corresponding to \re{Psi-N=2} can be

1756: calculated either by replacing the operator $J_{12}$ in the expression for

1757: $\mathcal{H}_2$ by its eigenvalue $(J_{12}-h-2s)\Psi_{p,h}=0$, or by applying

1758: \re{QE}. In this way, one obtains

1759: \be\label{E-2}

1760: E_h = \pm i\frac{d}{d\epsilon}\ln Q_{h}(\pm

1761: is+\epsilon)\biggl|_{\epsilon=0}=2\left[\psi(h+2s)-\psi(2s)\right]\,.

1762: \ee

1763: The ground state corresponds to $h=0$.

1764: 

1765: Let us examine the unitary transformation to the SoV representation \re{SoV-gen}

1766: for $N=2$. It is defined by the transition kernel, Eqs.~\re{B-ei} and \re{U},

1767: which looks at $N=2$ like

1768: \be

1769: \label{U-N=2}

1770: U_{p,x}(z_1,z_2)= p^{2s-1/2} \int \mathcal{D} w_2\, \e^{ipw_2}

1771: \Lambda_x(z_1,z_2|\bar w_2)\,,

1772: \ee

1773: where $\Lambda_x(z_1,z_2|\bar w_2)$ are given by \re{Lambda-ker}

1774: \be

1775: \Lambda_x(z_1,z_2|\bar w_2)=\e^{3i\pi s}\int \mathcal{D} y_2 \,(y_2-\bar

1776: w_2)^{-\beta_x}(z_2-\bar w_2)^{-\alpha_x}(z_1-\bar y_2)^{-\alpha_x}(z_2-\bar

1777: y_2)^{-\beta_x}\,, \nonumber

1778: \ee

1779: with $\beta_x=s+ix$ and $\alpha_x=s-ix$. It is convenient to transform

1780: $U_{p,x}(z_1,z_2)$ to the momentum representation.

1781: 

1782: For arbitrary function $\Psi(z_1,z_2)\in \mathcal{V}_2$ this transformation is

1783: defined as

1784: \be\label{F}

1785: \Psi(z_1,z_2)~=~\frac{1}{\Gamma(2s)}\int_0^\infty dp_1\,dp_2\,

1786: \e^{i(p_1z_1+p_2z_2)} (p_1\,p_2)^{s-1/2}\,\widetilde \Psi(p_1,p_2)\,,

1787: \ee

1788: where the additional factor $(p_1\,p_2)^{s-1/2}$ was introduced to simplify the

1789: expression for the scalar product \re{norm} in the momentum representation (see

1790: Eq.~\re{delta-f})

1791: \be

1792: \vev{\Psi_1 |\Psi_2} = \int_0^\infty dp_1 dp_2\,

1793: \lr{\widetilde\Psi_1(p_1,p_2)}^*\widetilde\Psi_2(p_1,p_2)\,.

1794: \ee

1795: In particular, the $N=2$ eigenstates \re{Psi-N=2} are given in the momentum

1796: representation by the Jacobi polynomials %~\cite{?}%

1797: \footnote{This expression is well-known in QCD as defining the conformal

1798: operators built from two fields with the conformal spin $s$.}

1799: \be

1800: \widetilde \Psi_{p,h}(p_1,p_2) = a_h\, \delta(p-p_1-p_2)

1801: (p_1p_2)^{s-1/2}(p_1+p_2)^h\,

1802: \textrm{P}^{(2s-1,2s-1)}_h\lr{\frac{p_1-p_2}{p_1+p_2}}\,,

1803: \label{mom-eig}

1804: \ee

1805: where $a_h=i^{-h-4s} h!\Gamma(2s)/\Gamma^2(h+2s)$ and delta-function ensures the

1806: momentum conservation. Applying \re{alpha-rep} and performing integration in

1807: \re{U-N=2}, one finds that in the momentum representation the $N=2$ transition

1808: kernel is given by

1809: \be\label{UP}

1810: \widetilde U_{p,x}(p_1,p_2)~=\delta(p-p_1-p_2)\,\frac{\Gamma^2(2s)\e^{i\pi

1811: s}}{|\Gamma(s+ix)|^2} \left(\frac{p}{p_1p_2}\right)^{1/2}

1812: \left(\frac{p_2}{p_1}\right)^s {}_2F_1\lr{{{s-ix,s+ix} \atop

1813: {2s}}\bigg|-\frac{p_2}{p_1}}\,.

1814: \ee

1815: It defines the SoV transformation $\widetilde\Psi_p(p_1,p_2) \mapsto \Phi(x)$

1816: \be\label{FU}

1817: \Phi(x)\delta(p-p')~=\int_0^\infty dp_1\,dp_2 \left (\widetilde

1818: U_{p',x}(p_1,p_2)\right)^*\,\widetilde \Psi_{p}(p_1,p_2)\,,

1819: \ee

1820: with $p\,,x>0$. Substituting \re{UP} into this relation and introducing notations

1821: for $\widetilde \Psi_p(p_1,p_2)=\delta(p-p_1-p_2)

1822: p^{-1/2}\xi^{s-1/2}(1+\xi)f(\xi)$ with $\xi=p_2/p_1$, one finds that \re{FU} is

1823: reduced to

1824: \be\label{FJ}

1825: \Phi(x)~=~ \e^{i\pi s}\, \frac{\Gamma^2(2s)}{|\Gamma(s+ix)|^2}\int_0^\infty

1826: d\xi\, \xi^{2s-1}\,{}_2F_1\lr{{{s-ix,s+ix} \atop {2s}}\bigg|-\xi}\, f(\xi)\,.

1827: \ee

1828: This relation defines the map $f(\xi)\mapsto \Phi(x)$, which is known as the

1829: Fourier-Jacobi or the index hypergeometric transform~\cite{Koor}. Then, the

1830: unitarity of the SoV transformation at $N=2$ follows from the similar property of

1831: the transformation \re{FJ}.

1832: 

1833: Let us replace $\widetilde \Psi_p(p_1,p_2)$ in \re{FU} by the $N=2$ eigenstate

1834: \re{mom-eig}. According to \re{PhiQ} and \re{SoV-inv}, the corresponding

1835: eigenfunction in the separated variables is given by the Wilson polynomial

1836: $\Phi(x)= Q_h(x)\sim W_h(x^2,s,s,s,s)$. Then, \re{FJ} leads to a known

1837: representation for the Wilson polynomials as the index hypergeometric transform

1838: of the Jacobi polynomials~\cite{Koor}.

1839: 

1840: \setcounter{equation}{0}

1841: \section{Concluding remarks}

1842: 

1843: In this paper we have constructed the Baxter $\mathbb{Q}-$operator and the

1844: representation of the Separated Variables for the open homogeneous

1845: $SL(2,\mathbb{R})$ spin magnet. Our analysis relied on the diagrammatical

1846: approach developed in Refs.~\cite{SoV} in application to the closed spin chain.

1847: In this approach, one represents the kernels of the relevant integral operators

1848: ($\mathbb{Q}-$operator, the transition kernel to the SoV representation) as

1849: Feynman diagrams and establishes their various properties with a help of a few

1850: simple diagrammatical identities.

1851: 

1852: We found that the Feynman diagrams for the $\mathbb{Q}-$operator and the

1853: transition kernel to the SoV representation have a remarkably simple form (see

1854: Figs.~\ref{fig1} and \ref{opyr}). In the latter case, the diagram reveals a

1855: universal pyramid-like structure which has been already observed for various

1856: quantum integrable models like periodic Toda chain~\cite{KL}, closed

1857: $SL(2,\mathbb{R})$ and $SL(2,\mathbb{C})$ spin chains~\cite{DKM,SoV} and

1858: Calogero-Sutherland model~\cite{KMS}. This structure is a manifestation of a

1859: general factorization property \re{U} of the transition kernel to the SoV

1860: representation. Namely, the kernel is factorized into the product of

1861: $\Lambda-$operators each depending on a single separated variable. The only

1862: difference between the models mentioned above resides in the explicit form of the

1863: $\Lambda-$operator. The latter can be obtained as a certain limit of the

1864: $\mathbb{Q}-$operator leading to the expression for the transition kernel to the

1865: SoV representation as the product of the $\mathbb{Q}-$operators projected onto a

1866: special reference state. Another advantage of the diagrammatical approach is that

1867: it offers a simple regular way for calculating the integration measure in the SoV

1868: representation.

1869: 

1870: We found that there exists an intrinsic relation between the open spin chains and

1871: Wilson polynomials~\cite{Koekoek}. The latter occupy the top level in the Askey

1872: scheme of hypergeometric orthogonal polynomials~\cite{Askey}. These polynomials

1873: define the eigenvalues of the Baxter operator for open spin chain with $N=2$

1874: sites~\cite{DKM99}.

1875: 

1876: It is straightforward to extend our analysis to the case of inhomogeneous open

1877: $SL(2,\mathbb{R})$ spin chains. One can show that for such models the transition

1878: kernel to the SoV representation is given by the same pyramid-like diagram shown

1879: in Fig.~\ref{opyr} with the only difference that both the indices attached to

1880: various lines and the integration measure corresponding to internal vertices

1881: should be modified appropriately. Another interesting possibility could be to

1882: consider an open spin chain with the $SL(2,\mathbb{C})$ symmetry. Such models

1883: naturally appear in high-energy QCD as describing Regge singularities of

1884: scattering amplitudes with meson quantum numbers~\cite{KK}. In that case, the

1885: quantum space of the model does not possess the highest weight (pseudovacuum

1886: state) and, therefore, the Algebraic Bethe Ansatz is not applicable.

1887: 

1888: \section*{Acknowledgements}

1889: 

1890: 

1891: We are grateful to V.~Braun and F.~Smirnov for useful discussions. This work was

1892: supported in part by the grant 03-01-00837 of the Russian Foundation for

1893: Fundamental Research, by the NATO Fellowship (A.M. and S.D.) and the Sofya

1894: Kovalevskaya programme of Alexander von Humboldt Foundation (A.M.).

1895: 

1896: 

1897: 

1898: \appendix

1899: \renewcommand{\theequation}{\Alph{section}.\arabic{equation}}

1900: \setcounter{table}{0}

1901: \renewcommand{\thetable}{\Alph{table}}

1902: 

1903: \section{Appendix: Feynman diagram technique}

1904: \label{Ap}

1905: 

1906: Here we collect some useful formulae for the $SL(2,\mathbb{R})$ integrals. Their

1907: derivation can be found in Refs.~\cite{SoV}.

1908: 

1909: \noindent%

1910: \begin{figure}[ht]%

1911: \vspace*{-5mm}

1912: \psfrag{a}[cc][cc]{$\alpha$}

1913: \psfrag{w}[cc][cc]{$\bar w$}

1914: \psfrag{z}[cc][cc]{$z$} $\bullet$~Propagator:\vspace*{-2mm}

1915: \be

1916: {\parbox[c]{30mm}{\centerline{\epsfxsize20mm\epsfbox{line.eps}}}}

1917: =\frac{1}{(z-\bar w)^\alpha}=\frac{\e^{-i\pi

1918: \alpha/2}}{\Gamma(\alpha)}\,\int_0^{\infty} dp\,\e^{ip\,(z-\bar w)} \,

1919: p^{\alpha-1}\,,

1920: \label{alpha-rep}

1921: \ee

1922: \vspace*{-5mm}

1923: \end{figure}%

1924: 

1925: \medskip

1926: \noindent$\bullet$~``Chain relation'' (see Fig.~\ref{Chain}):

1927: \be

1928: \int \mathcal{D}w\,{(z-\bar w)^{-\alpha} (w-\bar v)^{-\beta}~=~

1929: {a(\alpha,\beta)}\,(z-\bar v)^{-\alpha-\beta+2s}}\,,\qquad (\alpha+\beta\neq 2s)

1930: \label{chain-h}

1931: \ee

1932: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1933: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1934: \begin{figure}[t]

1935: \psfrag{a}[cc][cc]{$\beta$} \psfrag{b}[cc][cc]{$\alpha$}

1936: \psfrag{ab}[cc][cc]{$\alpha+\beta-2s$} \psfrag{text}[cc][cc]{$=\ \ \ \ \ \ \

1937: a(\alpha,\beta)\ \ \ \ \times $} \centerline{\epsfxsize14.0cm\epsfbox{chain.eps}}

1938: \vspace*{0.5cm} \caption[]{Chain relation.}

1939: \label{Chain}

1940: \end{figure}

1941: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1942: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1943: $\bullet$~Delta-function relation:

1944: \be

1945: \int \mathcal{D} w  \e^{i p w-ip'\bar w} = \delta(p-p')

1946: \,p^{1-2s}\cdot\Gamma(2s)\,,

1947: \label{delta-f}

1948: \ee

1949: $\bullet$~Identity operator:

1950: \be

1951: [\mathbb{K}\cdot\Psi](z) = \int \mathcal{D}w\,\frac{\e^{i\pi s}}{(z-\bar

1952: w)^{2s}}\Psi(w)= \Psi(z)\,.

1953: \label{K-ker}

1954: \ee

1955: $\bullet$~Contour-integral representation:

1956: \be

1957: \int\mathcal{D}w\,\frac{\e^{i\pi s}\Psi(w)}{(z_1-\bar w)^{\alpha_x} (z_2-\bar

1958: w)^{\beta_x}}=\frac{\Gamma(2s)}{\Gamma(\alpha_x)\Gamma(\beta_x)} \int_0^1d\tau

1959: \tau^{\alpha_x-1}(1-\tau)^{\beta_x-1}\Psi(\tau z_1+(1-\tau)z_2)\,.

1960: \label{drift}

1961: \ee

1962: $\bullet$~Fourier integral:

1963: \be

1964: \int \mathcal{D} w\, \frac{\e^{i p w}}{(z-\bar{w})^{\alpha}} =

1965:  \theta(p)\,p^{\alpha-2s}\e^{i p z}\cdot\e^{-i\pi\alpha/2}\frac{\Gamma(2s)}{\Gamma(\alpha)}\,,

1966: \label{p-rep}

1967: \ee

1968: $\bullet$~Permutation identity (see Figs.~\ref{comm-f} and \ref{amp1}):

1969: \be

1970: \label{comm-i}

1971: (z_2-\bar v_2)^{i(x-y)}I(z,\bar v\,;x,y)=(z_1-\bar v_1)^{i(x-y)} I(z,\bar

1972: v\,;y,x)\,,

1973: \ee

1974: where $z=(z_1,z_2)$, $\bar v=(\bar v_1,\bar v_2)$ and

1975: \be

1976: I(z,\bar v\,;x,y)= \int \mathcal{D}w \frac{1}{(w-\bar v_1)^{\alpha_x}(w-\bar

1977: v_2)^{\beta_x} (z_1-\bar w)^{\beta_y}(z_2-\bar w)^{\alpha_y}}\,.

1978: \label{four}

1979: \ee

1980: In these relations, $\alpha_x=s-ix$ and $\beta_x=s+ix$ for arbitrary $x$, the

1981: $a-$function is defined in \re{a} and the integration measure $\mathcal{D} w$ is

1982: given by \re{measure}, $\bar w=w^*$ and $p>0$.

1983: 

1984: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1985: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1986: \begin{figure}[ht]

1987: \psfrag{a1}[cl][cc]{$s+iy$} \psfrag{a2}[cl][cc]{$s-iy$}

1988: \psfrag{a3}[cl][cc]{$s-ix$} \psfrag{a4}[cl][cc]{$s+ix$}

1989: \psfrag{a5}[bc][cc][1][90]{$i(y-x)$} \psfrag{b1}[cr][cc]{$s+iy$}

1990: \psfrag{b2}[cr][cc]{$s-iy$} \psfrag{b3}[cr][cc]{$s-ix$}

1991: \psfrag{b4}[cr][cc]{$s+ix$} \centerline{\epsfxsize10.0cm\epsfbox{comm.eps}}

1992: \vspace*{0.5cm} \caption[]{Permutation identity.}

1993: \label{comm-f}

1994: \end{figure}

1995: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1996: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1997: 

1998: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1999: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2000: \begin{figure}[h!]

2001: \psfrag{a1}[cl][cc]{$s+iy$} \psfrag{a2}[cl][cc]{$s-iy$}

2002: \psfrag{a3}[cl][cc]{$s-ix$} \psfrag{a4}[cl][cc]{$s+ix$}

2003: \psfrag{a5}[bc][cc][1][90]{$i(y-x)$} \psfrag{b1}[cr][cc]{$s+iy$}

2004: \psfrag{b2}[cr][cc]{$s-iy$} \psfrag{b3}[cr][cc]{$s-ix$}

2005: \psfrag{b4}[cr][cc]{$s+ix$} \centerline{\epsfxsize10.0cm\epsfbox{amp2.eps}}

2006: \vspace*{0.5cm} \caption[]{Special case of the permutation identity. It is

2007: obtained {}from Figure~\ref{comm-f} by sending one of the external points to

2008: infinity.}

2009: \label{amp1}

2010: \end{figure}

2011: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2012: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2013: 

2014: \begin{thebibliography}{99}

2015: 

2016: \bibitem{QISM} L.A.~Takhtajan and L.D.~Faddeev, Russ.\ Math.\ Survey\ {\bf 34} (1979) 11;

2017: \\        E.K.~Sklyanin, L.A.~Takhtajan and L.D.~Faddeev,

2018:           Theor.\ Math.\ Phys.\ {\bf 40} (1980) 688;

2019: \\        V.E.~Korepin, N.M.~Bogoliubov and A.G.~Izergin, {\it Quantum

2020:           inverse scattering method and correlation functions\/},

2021:           Cambridge Univ. Press, 1993.

2022: \bibitem{BDM}  V.M.~Braun, S.\'E.~Derkachov, A.N.~Manashov,

2023:           Phys.\ Rev.\ Lett.\ \textbf{81} (1998) 2020.

2024: \bibitem{AB}   A.V.~Belitsky,

2025:           Phys.\ Lett.\ B \textbf{453} (1999) 59;

2026:           Nucl.\ Phys.\ B \textbf{558} (1999) 259;

2027:           Nucl.\ Phys.\ B \textbf{574} (2000) 407.

2028: \bibitem{DKM99}

2029:           S.~\'E.~Derkachov, G.~P.~Korchemsky and A.~N.~Manashov,

2030:           %``Evolution equations for quark gluon distributions in multi-color QCD  and open spin chains,''

2031:           Nucl.\ Phys.\ B {\bf 566} (2000) 203.

2032:           %[arXiv:hep-ph/9909539].

2033:           %%CITATION = HEP-PH 9909539;%%

2034: \bibitem{Baxter}

2035:           R.J.~Baxter, {\it Exactly Solved Models in Statistical

2036:           Mechanics\/}, Academic Press, London, 1982;

2037:           Stud.\ Appl.\ Math.\ {\bf 50} (1971) 51.

2038: \bibitem{ABA}  L.D.~Faddeev,

2039:           %``Algebraic aspects of Bethe Ansatz,''

2040:           Int.\ J.\ Mod.\ Phys.\ A {\bf 10} (1995) 1845. % [hep-th/9404013];

2041:           %``How Algebraic Bethe Ansatz works for integrable model,''

2042:           %hep-th/9605187.

2043: \bibitem{Sklyanin}

2044:           E.K.~Sklyanin, {\it The quantum Toda chain\/},

2045:           Lecture Notes in Physics, vol.\ 226, Springer, 1985, pp.196--233;

2046:           {\it Functional Bethe ansatz\/}, in ``Integrable

2047:           and superintegrable systems'', ed.\ B.A. Kupershmidt, World

2048:           Scientific, 1990, pp.8--33;

2049:           {\it Quantum inverse scattering method. Selected topics\/},

2050:           ``Quantum Group and Quantum Integrable Systems'' (Nankai

2051:           Lectures in Mathematical Physics), ed.\ Mo-Lin Ge, Singapore: World Scientific,

2052:           1992, pp.\ 63--97 [hep-th/9211111];

2053:           Progr.\ Theor.\ Phys.\ Suppl.\ {\bf 118} (1995) 35

2054:           [solv-int/9504001].

2055: \bibitem{PG} V.~Pasquier amd M.~Gaudin, J.~Phys.~A: Math.~Gen.

2056:           {\bf 25} (1992) 5243.

2057: \bibitem{KL}   S.~Kharchev and D.~Lebedev,

2058:           %``Integral representation for the eigenfunctions of quantum periodic Toda  chain,''

2059:           Lett.\ Math.\ Phys.\  {\bf 50} (1999) 53;

2060:           %``Eigenfunctions of $GL(N,\RR)$ Toda chain: The Mellin-Barnes representation,''

2061:           JETP Lett.\  {\bf 71} (2000) 235;

2062:           %``Integral representations for the eigenfunctions of quantum

2063:           % open and periodic Toda chains from QISM formalism,''

2064:           J.\ Phys.\ A: Math. Gen. {\bf 34} (2001) 2247. %[hep-th/0007040].

2065: \bibitem{KSS}

2066:           V.B.~Kuznetsov, M.Salerno and E.K.~Sklyanin,

2067:           %``Quantum Backlund transformation for the integrable DST model,''

2068:           J.\ Phys.\ A {\bf 33} (2000) 171.

2069:           %[arXiv:solv-int/9908002].

2070:           %%CITATION = SOLV-INT 9908002;%%

2071: \bibitem{SD}

2072:           S.\'E. Derkachov, J. Phys. A: Math. Gen. {\bf 32} (1999) 5299.

2073: \bibitem{DKM}S.\'E.~Derkachov, G.P.~Korchemsky and A.N.~Manashov,

2074:           %``Noncompact Heisenberg spin magnets from high-energy QCD.

2075:           %I: Baxter  Q-operator and separation of variables,''

2076:           Nucl.\ Phys.\ B {\bf 617} (2001) 375. %[hep-th/0107193].

2077: \bibitem{SoV} S.\'E.~Derkachov, G.~P.~Korchemsky and A.~N.~Manashov,

2078:           %``Separation of variables for the quantum SL(2,R) spin chain,''

2079:           JHEP {\bf 0307} (2003) 047.

2080:           %[arXiv:hep-th/0210216].

2081: \bibitem{KMS}

2082: V.~B.~Kuznetsov, V.~V.~Mangazeev and E.~K.~Sklyanin,

2083: %``Q-operator and factorised separation chain for Jack's symmetric polynomials,''

2084: arXiv:math.ca/0306242.

2085: %%CITATION = MATH-CA 0306242;%%

2086: 

2087: \bibitem{Sklyanin88}  E.K.~Sklyanin, J.Phys.A: Math.Gen. {\bf 21}

2088:            (1988) 2375.

2089: \bibitem{Gelfand}

2090:           I.M.~Gelfand, M.I.~Graev and N.Ya.~Vilenkin, {\it Generalized functions. Vol.5

2091:           Integral geometry and representation theory\/}, New York, NY Academic Press 1966.

2092: \bibitem{Ch}

2093:     I.V.~Cherednik,

2094:     %``Factorizing Particles On A Half Line And Root Systems,''

2095:     Theor.\ Math.\ Phys.\ {\bf 61} (1984) 977.

2096: 

2097: \bibitem{KS}

2098:     P.P.~Kulish and E.K.~Sklyanin,

2099:     %``Algebraic structures related to the reflection equations,''

2100:     J.\ Phys.\ A {\bf 25} (1992) 5963;

2101:     % [arXiv:hep-th/9209054].

2102:     %%CITATION = HEP-TH 9209054;%%

2103: \\

2104:     P.P.~Kulish and R.~Sasaki,

2105:     %``Covariance properties of reflection equation algebras,''

2106:     Prog.\ Theor.\ Phys.\ {\bf 89} (1993) 741;

2107:     % hep-th/9212007.

2108: \\

2109:     H.J.~de Vega and A.~Gonzalez-Ruiz,

2110:     %``Boundary K matrices for the XYZ, XXZ and XXX spin chains,''

2111:     J.\ Phys.\ {\bf A27} (1994) 6129.

2112: \bibitem{Koor}

2113:           T.H.~Koornwinder, %{\it Special orthogonal polynomial system mapping to each other

2114:           % by Fourier-Jacobi transform},

2115:           Lect.~Notes Math. {\bf 1171} (1985) 174. %-183.

2116: 

2117: \bibitem{KuS} V.B.~Kuznetsov and E.K.~Sklyanin,

2118:           % {\it Few remarks on B\"acklund transformations for many-body systems\/},

2119:           J.\ Phys.\ A {\bf 31} (1998) 2241. %[solv-int/9711010].

2120: 

2121: \bibitem{Koekoek} R.~Koekoek and S.~Swarttouw, {\it The Askey scheme of hypergeometric

2122: orthogonal polynomials and its q-analogues}, Report 94-05, Delft University of

2123: Technology, 1994. % [http://citeseer.nj.nec.com/96840.html].

2124: 

2125: \bibitem{Askey}

2126: G.E.~Andrews, R.~Askey and R.~Roy, \textit{Special functions}, Encyclopedia of

2127: Mathematics and Its Applications, The University Press, Cambridge, 1999.

2128: 

2129: \bibitem{KK}

2130: D.~Karakhanian and R.~Kirschner,

2131: %``Conserved currents of the three-reggeon interaction,''

2132: Phys.\ Atom.\ Nucl.\  {\bf 65} (2002) 1501 [Yad.\ Fiz.\  {\bf 65} (2002) 1539].

2133: %[arXiv:hep-th/9902147].

2134: %%CITATION = HEP-TH 9902147;%%

2135: 

2136: 

2137: 

2138: \end{thebibliography}

2139: 

2140: \end{document}

2141: