1: %% LyX 1.3 created this file.  For more info, see http://www.lyx.org/.

2: %% Do not edit unless you really know what you are doing.

3: \documentclass[12pt,english]{article}

4: \usepackage[T1]{fontenc}

5: \usepackage[latin1]{inputenc}

6: \usepackage{a4wide}

7: \usepackage{array}

8: \usepackage{amsmath}

9: \usepackage{graphicx}

10: \usepackage{amssymb}

11: 

12: \makeatletter

13: 

14: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.

15: %% Because html converters don't know tabularnewline

16: \providecommand{\tabularnewline}{\\}

17: 

18: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Textclass specific LaTeX commands.

19:  \newcommand{\lyxaddress}[1]{

20:    \par {\raggedright #1 

21:    \vspace{1.4em}

22:    \noindent\par}

23:  }

24: 

25: \usepackage{babel}

26: \makeatother

27: \begin{document}

28: 

29: \title{On the boundary form factor program}

30: 

31: 

32: \author{Z. Bajnok$^{1}$, L. Palla$^{2}$, and G. Takács$^{1}$}

33: 

34: \maketitle

35: 

36: \lyxaddress{\begin{center}$^{1}$\emph{\small Theoretical Physics Research Group,

37: Hungarian Academy of Sciences, }\\

38: \emph{\small 1117 Budapest, Pázmány Péter sétány 1/A, Hungary} \\

39: $^{2}$\emph{\small Institute for Theoretical Physics, Eötvös University,

40: }\\

41: \emph{\small 1117 Budapest, Pázmány Péter sétány 1/A, Hungary}\end{center}}

42: 

43: \begin{abstract}

44: Boundary form factor axioms are derived for the matrix elements of

45: local boundary operators in integrable 1+1 dimensional boundary quantum

46: field theories using the analyticity properties of correlators via

47: the boundary reduction formula. Minimal solutions are determined for

48: the integrable boundary perturbations of the free boson, free fermion

49: (Ising), Lee-Yang and sinh-Gordon models and the two point functions

50: calculated from them are checked against the exact solutions in the

51: free cases and against the conformal data in the ultraviolet limit

52: for the Lee-Yang model. In the case of the free boson/fermion the

53: dimension of the solution space of the boundary form factor equation

54: is shown to match the number of independent local operators. We obtain

55: excellent agreement which proves not only the correctness of the solutions

56: but also confirms the form factor axioms. 

57: \end{abstract}

58: 

59: \section{Introduction}

60: 

61: The bootstrap program aims to classify and explicitly solve 1+1 dimensional

62: integrable quantum field theories by constructing all of their Wightman

63: functions. The first stage is the S-matrix bootstrap: the scattering

64: matrix, connecting asymptotic \emph{in} and \emph{out} states, is

65: determined from its properties such as factorizability, unitarity,

66: crossing symmetry and the Yang-Baxter equation (YBE) supplemented

67: by the maximal analyticity assumption. The result is the complete

68: on-shell solution of the theory, i.e. the spectrum of excitations

69: and their scattering amplitudes, which can be related to some independent

70: definition of the model as a perturbed conformal field theory or a

71: Lagrangian QFT (for reviews see \cite{Sbstr,Sanal}). The second step

72: is the form factor bootstrap, which allows one to determine matrix

73: elements of local operators between asymptotic states using their

74: analytic properties originating from the already known S-matrix. Supposing

75: maximal analyticity leads to a set of solutions each of which corresponds

76: to a local operator of the theory. The form factors are then used

77: to build the correlation (Wightman) functions via their spectral representations,

78: yielding a complete off-shell description of the theory (see \cite{Smirnov,Arndt}

79: for reviews). 

80: 

81: The first step of an analogous bootstrap program for 1+1 dimensional

82: integrable \emph{boundary} quantum field theories, the boundary R-matrix

83: bootstrap, has been developed for several theories. In boundary theories

84: the asymptotic states are connected by the reflection R-matrix, which

85: satisfies unitarity and boundary crossing unitarity; for integrable

86: boundary QFT, it also satisfies the boundary YBE (BYBE) and boundary

87: bootstrap requirements. These equations supplemented by maximal analyticity

88: assumptions make possible to determine the reflection matrices and

89: provide the complete information about the theory on the mass shell

90: \cite{GZ}. 

91: 

92: For the second step matrix elements of local operators between asymptotic

93: states have to be computed. In a boundary quantum field theory there

94: are two types of operators, the bulk and the boundary operators, where

95: their names indicate their localization point. Due to the broken translational

96: invariance one point functions of bulk operators may acquire nontrivial

97: space dependence behaving analogously to the two point functions in

98: a bulk theory. Indeed this one point function can be calculated in

99: the crossed channel, where the role of time and space is changed and

100: the spatial boundary appears as a temporal one represented as an initial

101: (boundary) state in the matrix element. Inserting a complete system

102: of the bulk Hilbert space a spectral representation for the one point

103: functions can be obtained in terms of the bulk form factors and the

104: matrix element of the boundary state \cite{Muss1pt,DPTW1pt}. Truncating

105: this expansion at finite intermediate states provides a convergent

106: large distance expansion. However, matrix elements of boundary operators

107: cannot be computed in this way and the purpose of the present paper

108: is to develop a technique to compute their correlation functions. 

109: 

110: In this paper we initiate the second step of the boundary bootstrap

111: program, namely the boundary form factor program for calculating the

112: matrix elements of local boundary operators between asymptotic states.

113: We derive their analytic structure from that of the R-matrix which,

114: when supplemented by the assumption of maximal analyticity, leads

115: to their determination. In the bulk case, it was shown in \cite{cardy_mussardo}

116: that the solution space of the form factor equations can be brought

117: into one-to-one correspondence with the operator content of the model.

118: Based on this, we expect that the classification of the solutions

119: of the boundary form factor axioms provides information on the boundary

120: operator content of the theory, which in the ultraviolet limit is

121: in a one-to-one correspondence with the Hilbert space of the model.

122: Using the explicit form of the boundary form factors the spectral

123: representation for the boundary correlation functions can be obtained.

124: 

125: The paper is organized as follows: first we define the boundary form

126: factors by introducing asymptotic \emph{in} and \emph{out} multi-particle

127: states, which are related by the multi-particle reflection matrix.

128: Simple crossing relations are presented from which the form factor

129: axioms follow easily, and then the axioms are verified by some consistency

130: requirements. We outline a general strategy to solve theories with

131: diagonal bulk scattering and boundary reflection amplitudes, and to

132: compare the resulting two-point functions with their ultraviolet limits.

133: This idea is applied to integrable boundary perturbations of several

134: models, such as the free boson model, free fermion (alias Ising) field

135: theory, the scaling Lee-Yang model and sinh-Gordon theory. Appendix

136: A contains a heuristic derivation of the crossing relations from the

137: boundary reduction formula \cite{BBT}, while in Appendix B we present

138: a formal derivation of the boundary form factor axioms from the boundary

139: version of the Faddeev-Zamolodchikov algebra. 

140: 

141: 

142: \section{Boundary form factors}

143: 

144: 

145: \subsection{Definitions}

146: 

147: The Hilbert space of a boundary quantum field theory consists of multi-particle

148: states, which can be labeled by the particle species and the corresponding

149: particle energies. To simplify the notations we restrict ourselves

150: to theories containing only one particle type with a given mass $m$.

151: In 1+1 dimensions it is convenient to work with the rapidity variable

152: $\theta_{i}$; the energy $E_{i}$ of the particle can be written

153: as $E_{i}=m\cosh\theta_{i}$, while the momentum is $p_{i}=m\sinh\theta_{i}$.

154: Following the evolution of the multi-particle state in time to $t\to-\infty$

155: the particles get far away form each other and from the boundary,

156: therefore forming an \emph{in} state which is equivalent to a free

157: multi-particle state and is denoted as%

158: \footnote{In general, particles in an interacting two dimensional quantum field

159: theory have an effective fermionic statistics with the sole exception

160: of free bosonic theories, for which it is necessary to allow equality

161: in the ordering of the particle rapidities.%

162: } \[

163: \vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}\quad;\qquad\theta_{1}>\theta_{2}>\dots>\theta_{n}>0\]

164: Positivity of all incoming rapidities is a consequence of the assumption

165: that the boundary is at the right end of the half line and it is a

166: major difference from the bulk situation. This difference is essential

167: because it influences the analyticity domain of matrix elements. 

168: 

169: For $t\to+\infty$ all the scatterings and reflections are terminated,

170: the particles are again far away from each other and from the boundary

171: forming the \emph{out} state, 

172: 

173: \[

174: \vert\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'}\rangle_{out}\quad;\qquad\theta_{1}^{'}<\theta_{2}^{'}<\dots<\theta_{m}^{'}<0\]

175: which is again equivalent to a free state. By the standard assumption

176: of asymptotic completeness, the two sets of states form a complete

177: basis separately and are connected by the reflection matrix, which

178: is the boundary analogue of the S matrix. In an integrable theory,

179: due to the infinite number of conserved charges, there is no particle

180: creation ($n=m$), the set of rapidities changes only sign $\theta_{i}^{'}=-\theta_{i}$,

181: and the reflection matrix factorizes into the product of pairwise

182: bulk scatterings and individual reflections \begin{equation}

183: \vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}=\prod_{i<j}S(\theta_{i}-\theta_{j})S(\theta_{i}+\theta_{j})\prod_{i}R(\theta_{i})\vert-\theta_{1},-\theta_{2},\dots,-\theta_{n}\rangle_{out}\label{Rmatrix}\end{equation}

184: where $S(\theta_{i}-\theta_{j})$ connects the two particle asymptotic

185: \emph{in} and \emph{out} states in the bulk theory (without the boundary)

186: 

187: \begin{tabular}{cc}

188: \begin{tabular}{c}

189: $\vert\theta_{1},\theta_{2}\rangle_{in}^{bulk}=S(\theta_{1}-\theta_{2})\vert\theta_{2},\theta_{1}\rangle_{out}^{bulk}\qquad\qquad$depicted

190: as\tabularnewline

191: \tabularnewline

192: \end{tabular}&

193: \begin{tabular}{c}

194: \tabularnewline

195: \includegraphics[%

196:   height=2cm]{s.eps}\tabularnewline

197: \end{tabular}\tabularnewline

198: \end{tabular}

199: 

200: It is defined originally for $\theta_{1}>\theta_{2}$ but can be analytically

201: continued for complex rapidity parameters such that the extended function

202: (denoted the same way) is meromorphic and satisfies unitarity and

203: crossing symmetry \[

204: S(\theta)S(-\theta)=1\quad,\qquad S(i\pi-\theta)=S(\theta)\]

205: It might have poles on the imaginary axis at locations $\theta=iu_{j}$

206: with residue $-i\textrm{res}_{\theta=iu_{j}}S(\theta)=\Gamma_{j}^{2}$,

207: some of which correspond to bound states. 

208: 

209: The amplitude $R(\theta)$ connects the one particle asymptotic states

210: in the boundary theory 

211: 

212: \begin{tabular}{cc}

213: \begin{tabular}{c}

214: $\vert\theta\rangle_{in}=R(\theta)\vert-\theta\rangle_{out}\qquad\qquad$depicted

215: as \tabularnewline

216: \tabularnewline

217: \end{tabular}&

218: \begin{tabular}{c}

219: \tabularnewline

220: \hspace{2cm}\includegraphics[%

221:   height=2cm]{r.eps}\tabularnewline

222: \end{tabular}\tabularnewline

223: \end{tabular}

224: 

225: It can also be extended from the fundamental domain $\theta>0$ to

226: a meromorphic function on the whole complex $\theta$ plane satisfying

227: unitarity and boundary crossing unitarity \[

228: R(\theta)R(-\theta)=1\quad,\qquad R(i\pi-\theta)S(2\theta)=R(\theta)\]

229: $R(\theta)$ may have poles at imaginary locations $\theta=iv_{j}$

230: ($0<v_{j}<\pi/2$), some corresponding to excited boundary states.

231: If the interpolating field has a nontrivial vacuum expectation value

232: then generally there is also a pole at $\theta=i\pi/2$ with residue

233: \begin{equation}

234: -i\mathop{\textrm{Res}}_{\theta=\frac{i\pi}{2}}R(\theta)=\frac{g^{2}}{2}\label{eq:Defg}\end{equation}

235: 

236: 

237: The boundary form factor is defined as the matrix element of some

238: local boundary operator, $\mathcal{O}(t)$, between asymptotic states

239: \begin{eqnarray*}

240: \,_{out}\langle\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'}\vert\mathcal{O}(t)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in} & =\\

241:  &  & \hspace{-2cm}F_{mn}^{\mathcal{O}}(\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'};\theta_{1},\theta_{2},\dots,\theta_{n})e^{-imt(\sum\cosh\theta_{i}-\sum\cosh\theta_{j}^{'})}\end{eqnarray*}

242: These form factors are defined only for $\theta_{1}>\theta_{2}>\dots>\theta_{n}>0$

243: and $\theta_{1}^{'}<\theta_{2}^{'}<\dots<\theta_{m}^{'}<0$. We can

244: introduce other form factors as \[

245: \,_{out}\langle\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'}\vert\mathcal{O}(t)\vert-\theta_{1},-\theta_{2},\dots,-\theta_{n}\rangle_{out}=F_{mn}^{\mathcal{O}}(\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'};-\theta_{1},-\theta_{2},\dots,-\theta_{n})\]

246:  and consider them as a continuation of the original ones in the rapidities.

247: Expressing these form factors (via the boundary reduction formula

248: \cite{BBT}) in terms of correlation functions an analytic continuation

249: can be performed for any (even) complex values of the rapidity parameters.

250: As a result the generalized form factors are meromorphic functions

251: of the rapidity parameters, and we shall assume that their poles always

252: have physical origins (maximal analyticity assumption). From the crossing

253: formula \begin{equation}

254: F_{mn}^{\mathcal{O}}(\theta_{1}^{'},\theta_{2}^{'},\dots,\theta_{m}^{'};\theta_{1},\theta_{2},\dots,\theta_{n})=F_{m-1n+1}^{\mathcal{O}}(\theta_{2}^{'},\dots,\theta_{m}^{'};\theta_{1}^{'}+i\pi,\theta_{1},\theta_{2},\dots,\theta_{n})+\textrm{disc}\label{eq:crossing}\end{equation}

255: derived in Appendix A, we can express all the form factors in terms

256: of the elementary form factors\[

257: \,_{out}\langle0\vert\mathcal{O}(0)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}=F_{n}^{\mathcal{O}}(\theta_{1},\theta_{2},\dots,\theta_{n})\]

258: It is important to notice that the boundary form factors $F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})$,

259: in contrast to the bulk case, do depend in general on all the rapidities

260: $\theta_{i}$, not just on their differences, since in the presence

261: of a boundary Lorentz invariance is broken. 

262: 

263: 

264: \subsection{Axioms}

265: 

266: In the Appendices we derive all the following properties of the matrix

267: elements of local boundary operators valid in any integrable boundary

268: quantum field theory. Following the general philosophy in the bulk

269: case \cite{Smirnov} we take them as axioms defining the local operators

270: via their matrix elements. 

271: 

272: I. Permutation:

273: 

274: \begin{center}\[

275: F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{i},\theta_{i+1},\dots,\theta_{n})=S(\theta_{i}-\theta_{i+1})F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{i+1},\theta_{i},\dots,\theta_{n})\]

276: \includegraphics[%

277:   height=3cm,

278:   keepaspectratio]{bd1.eps}~~~\includegraphics[%

279:   height=3cm,

280:   keepaspectratio]{bd3.eps}\end{center}

281: 

282: II. Reflection:\[

283: F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n-1},\theta_{n})=R(\theta_{n})F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n-1},-\theta_{n})\]

284: 

285: 

286: \begin{center}\includegraphics[%

287:   height=3cm,

288:   keepaspectratio]{bd1.eps}~~~ \includegraphics[%

289:   height=33mm,

290:   keepaspectratio]{bd2.eps}\end{center}

291: 

292: III. Crossing reflection: \[

293: F_{n}^{\mathcal{O}}(\theta_{1},\theta_{2},\dots,\theta_{n})=R(i\pi-\theta_{1})F_{n}^{\mathcal{O}}(2i\pi-\theta_{1},\theta_{2},\dots,\theta_{n})\]

294: 

295: 

296: \begin{center}\includegraphics[%

297:   height=3cm,

298:   keepaspectratio]{bd1.eps}~~~\includegraphics[%

299:   height=3cm,

300:   keepaspectratio]{bd4.eps}\end{center}

301: 

302: The singularity structure of the form factors is determined on physical

303: grounds and can be axiomatized as follows:

304: 

305: IV. Kinematical singularity\[

306: -i\mathop{\textrm{Res}}_{\theta=\theta^{'}}F_{n+2}^{\mathcal{O}}(\theta+i\pi,\theta^{'},\theta_{1},\dots,\theta_{n})=\left(1-\prod_{i=1}^{n}S(\theta-\theta_{i})S(\theta+\theta_{i})\right)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\]

307: or equivalently described as \[

308: -i\mathop{\textrm{Res}}_{\theta=\theta^{'}}F_{n+2}^{\mathcal{O}}(-\theta+i\pi,\theta^{'},\theta_{1},\dots,\theta_{n})=\left(R(\theta)-\prod_{i=1}^{n}S(\theta-\theta_{i})R(\theta)S(\theta+\theta_{i})\right)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\]

309: 

310: 

311: \begin{center}\includegraphics[%

312:   height=3cm,

313:   keepaspectratio]{bdk1.eps}~~~\includegraphics[%

314:   height=3cm,

315:   keepaspectratio]{bdk2.eps}\includegraphics[%

316:   height=3cm,

317:   keepaspectratio]{bdk3.eps}\end{center}

318: 

319: V. Boundary kinematical singularity

320: 

321: \[

322: -i\mathop{\textrm{Res}}_{\theta=0}F_{n+1}^{\mathcal{O}}(\theta+\frac{i\pi}{2},\theta_{1},\dots,\theta_{n})=\frac{g}{2}\Bigl(1-\prod_{i=1}^{n}S\bigl(\frac{i\pi}{2}-\theta_{i}\bigr)\Bigr)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\]

323: 

324: 

325: \begin{center}\includegraphics[%

326:   height=3cm,

327:   keepaspectratio]{bdbk1.eps}~~~\includegraphics[%

328:   height=3cm,

329:   keepaspectratio]{bdbk2.eps}\includegraphics[%

330:   height=3cm,

331:   keepaspectratio]{bdbk3.eps}\end{center}

332: 

333: VI. Bulk dynamical singularity \[

334: -i\mathop{\textrm{Res}}_{\theta=\theta^{'}}F_{n+2}^{\mathcal{O}}(\theta+iu,\theta^{'}-iu,\theta_{1},\dots,\theta_{n})=\Gamma F_{n+1}^{\mathcal{O}}(\theta,\theta_{1},\dots,\theta_{n})\]

335: 

336: 

337: \begin{center}\includegraphics[%

338:   height=3cm,

339:   keepaspectratio]{bdd1.eps}~~~\includegraphics[%

340:   height=3cm,

341:   keepaspectratio]{bdd2.eps}\end{center}

342: 

343: VII. Boundary dynamical singularity

344: 

345: \[

346: -i\mathop{\textrm{Res}}_{\theta=iv}F_{n+1}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n},\theta)=\tilde{g}\tilde{F}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\]

347: 

348: 

349: .

350: 

351: \begin{center}\includegraphics[%

352:   height=3cm,

353:   keepaspectratio]{bdbd1.eps}~~~\includegraphics[%

354:   height=3cm,

355:   keepaspectratio]{bdbd2.eps}\end{center}

356: 

357: We note that equations similar to some of ours have been obtained

358: earlier studying boundary form factors in specific spin chains. Using

359: a concrete realization for the Hilbert space and the operators, these

360: equations were extracted originally for the XXZ and XYZ models in

361: \cite{XXZ} and extended for other spin chains in \cite{XXZGen}.

362: By extending the bulk free field representation for the boundary sine-Gordon

363: model the analogues of XXZ equations were obtained in \cite{SGff}.

364: In all these approaches, however, there is no analogue of the axiom

365: V, without which the equations do not determine completely the form

366: factors as can be seen on the example of the sinh-Gordon model. In

367: contrast, in our approach the form factor axioms are firmly established

368: from first principles of local quantum field theory, thus they are

369: valid in a general setting. As a further result of our systematic

370: approach the axioms found form a complete system ready to be solved. 

371: 

372: 

373: \subsection{Consistency checks}

374: 

375: Before proceeding to concrete examples we perform a few consistency

376: checks of the axioms. First we note that they are self-consistent

377: in the sense that for specific rapidities the $n+2$ particle form

378: factor can be connected to the $n$ particle form factor either by

379: the kinematical singularity equations or by using twice the boundary

380: kinematical equations, and the two procedures give the same result.

381: Indeed taking double residue in the first case, first at $\theta=\theta^{'}$

382: and then at $\theta=i\frac{\pi}{2}$ gives \begin{eqnarray*}

383: i\mathop{\textrm{Res}}_{\theta=\frac{i\pi}{2}}i\mathop{\textrm{Res}}_{\theta^{'}=\theta}F_{n+2}^{\mathcal{O}}(-\theta+i\pi,\theta^{'},\theta_{1},\dots,\theta_{n}) & =\\

384:  &  & \hspace{-4cm}\left(-i\mathop{\textrm{Res}}_{\theta=\frac{i\pi}{2}}R(\theta)\right)\left(1-\prod_{i=1}^{n}S(\frac{i\pi}{2}-\theta_{i})S(\frac{i\pi}{2}+\theta_{i})\right)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\end{eqnarray*}

385: Taking now the residue at $\theta=\frac{i\pi}{2}$ first then at $\theta^{'}=\frac{i\pi}{2}$

386: and using that $S(0)=-1$ gives \begin{eqnarray*}

387: i\mathop{\textrm{Res}}_{\theta=\frac{i\pi}{2}}i\mathop{\textrm{Res}}_{\theta^{'}=\frac{i\pi}{2}}F_{n+2}^{\mathcal{O}}(-\theta+i\pi,\theta^{'},\theta_{1},\dots,\theta_{n}) & =\\

388:  &  & \hspace{-4cm}\frac{g^{2}}{4}\left(1+\prod_{i=1}^{n}S(\frac{i\pi}{2}-\theta_{i})\right)\left(1-\prod_{i=1}^{n}S(\frac{i\pi}{2}-\theta_{i})\right)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\end{eqnarray*}

389:  The two different orders of taking the residues differ by a factor

390: of $2$ since in the first case after taking the residue at $\theta^{'}=\theta$

391: we get a factor $f(2\theta-i\pi)$ which has a zero at $\theta=i\frac{\pi}{2}$

392: (due to $S(0)=-1$ the bulk minimal form factor vanishes at the threshold:

393: $f(0)=0$). In the second case after taking the first residue a factor

394: $f(\theta-i\frac{\pi}{2})$ appears. When expanding around $\theta=i\frac{\pi}{2}$

395: to take the second residue there appears a factor $2$ due to the

396: difference in the arguments of this particular factor (all other terms

397: are identical in the two cases). Combining the crossing symmetry of

398: the S-matrix with the definition of $g$ (\ref{eq:Defg}) the two

399: expressions are easily seen to be equivalent.

400: 

401: It is worth emphasizing that in the boundary kinematical singularity

402: axiom it is the particle-boundary coupling constant $g$ which appears

403: although the residue of the reflection factor determines only $g^{2}$.

404: There are known examples where in two physically different situations

405: the fundamental reflection amplitudes are the same and the two cases

406: are distinguished only by the sign of $g$ (e.g. the boundary Lee-Yang

407: model with $1$ boundary and with $\Phi$ boundary with a particular

408: value of the boundary coupling \cite{DTWbct} -- see in more details

409: in Sec. 3.3). Because of axiom V the solutions of the form factor

410: axioms are different for the two cases, as shown in detail in Sec.

411: 3.3.

412: 

413: As a second test we relate the two disconnected physical domains (\emph{in/out})

414: of the definition of the form factor. By permuting successively each

415: rapidity to the last position, applying a reflection and permuting

416: back to their original position we obtain that \[

417: F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})=\prod_{i<j}S(\theta_{i}+\theta_{j})\prod_{i}R(\theta_{i})\prod_{i<j}S(\theta_{i}-\theta_{j})F_{n}^{\mathcal{O}}(-\theta_{1},\dots,-\theta_{n})\]

418: The product appearing is nothing but the multi-particle R-matrix,

419: (\ref{Rmatrix}), which connects the \emph{in} and \emph{out} states.

420: 

421: Finally we use the fact that the reflection matrix can be considered

422: as a special form factor (of the identity operator $Id$) whose analytic

423: properties are well known. By definition\[

424: F_{2}^{Id}(\theta^{'}+i\pi,\theta)=\,_{out}\langle\theta^{'}\vert Id\vert\theta\rangle_{in}=R(\theta)\,_{out}\langle\theta^{'}\vert Id\vert-\theta\rangle_{out}=R(\theta)\delta(\theta+\theta^{'})\]

425: Now using the permutation property and tricks as above we have\[

426: F_{2}^{Id}(\theta^{'}+i\pi,\theta)=S(i\pi+\theta^{'}-\theta)F_{2}^{Id}(\theta,\theta^{'}+i\pi)=S(i\pi+\theta^{'}-\theta)R(i\pi+\theta^{'})\delta(\theta+\theta^{'})\]

427: which, due to the boundary crossing unitarity, is equivalent to the

428: previous expression. 

429: 

430: 

431: \subsection{General solution}

432: 

433: In this section we describe the general procedure we use to obtain

434: the solutions of the form factor equations in the various specific

435: models. In doing so we emphasize the similarities and the differences

436: between the boundary and bulk form factors and also separate the (boundary)

437: operator dependent parts from the ones that depend on the specific

438: field theory considered but are independent of the operators in question.

439: 

440: 

441: \subsubsection{One particle form factors}

442: 

443: In sharp contrast to the bulk case, in the boundary theory, the boundary

444: operators in general may have non trivial one particle form factors

445: (1PFF). Since the multi-particle form factors are recursively determined,

446: the 1PFF-s are very important inputs to these recursions, and their

447: determination is necessarily the first step. The equations for the

448: 1PFF read: \begin{equation}

449: F_{1}(\theta)=R(\theta)F_{1}(-\theta)\quad;\quad F_{1}(i\pi+\theta)=R(-\theta)F_{1}(i\pi-\theta),\label{eq:1pff}\end{equation}

450:  where the reflection amplitude $R(\theta)$ is analytic in the physical

451: strip $0\leq\Im m(\theta)\leq\pi/2$ (apart from the presence of finitely

452: many discrete poles on the imaginary axis), and from general considerations

453: using the reduction formulae we know that $F_{1}(\theta)$ is analytic

454: for $0\leq\Im m(\theta)\leq\pi$. Note that if $F_{1}(\theta)$ is

455: a solution of (\ref{eq:1pff}) then $F_{1}(\theta)\Psi(\theta)$ is

456: also a solution provided \[

457: \Psi(\theta)=\Psi(-\theta),\qquad\Psi(i\pi+\theta)=\Psi(i\pi-\theta),\]

458:  i.e. if $\Psi$ is even and $2\pi i$ periodic. Therefore one can

459: take $\Psi(\theta)=\psi(y)$ with $y=e^{\theta}+e^{-\theta}$.

460: 

461: To construct solutions to (\ref{eq:1pff}) we reduce them to a problem

462: already solved in the bulk form factor bootstrap. To this end we write

463: $F_{1}(\theta)=g_{1}(\theta)g_{2}(i\pi-\theta)$ and suppose that

464: \begin{equation}

465: g_{1}(\theta)=R(\theta)g_{1}(-\theta)\quad;\quad g_{1}(i\pi+\theta)=g_{1}(i\pi-\theta),\label{eq:g1}\end{equation}

466:  which are nothing else but the bulk two particle form factor equations

467: \cite{FMS}, where the reflection amplitude, $R(\theta)$, plays the

468: role of the S-matrix. Furthermore, plugging this product form $F_{1}$

469: into (\ref{eq:1pff}) reveals, that $g_{2}$ must also solve (\ref{eq:g1}).

470: Thus a solution to (\ref{eq:1pff}) can be constructed as \[

471: F_{1}(\theta)=g(\theta)g(i\pi-\theta),\]

472:  where $g(\theta)$ is an appropriate solution of (\ref{eq:g1}).

473: 

474: To obtain a solution of (\ref{eq:g1}) we use the following theorem

475: \cite{KW}. If the function $h(\theta)$ is meromorphic in the physical

476: strip $0\leq\Im m(\theta)<\pi$ with possible poles at $i\alpha_{1},\dots,i\alpha_{l}$

477: and zeros at $i\beta_{1},\dots,i\beta_{k}$ and grows as at most a

478: polynomial in $\exp(|\theta|)$ for $|\Re e\,\theta|\rightarrow\infty$,

479: furthermore it satisfies \[

480: h(\theta)=M(\theta)h(-\theta);\quad M(\theta)=\exp\left\{ \int_{0}^{\infty}dtf(t)\sinh\left(\frac{t\theta}{i\pi}\right)\right\} ;\quad h(i\pi-\theta)=h(i\pi+\theta);\]

481:  then it is uniquely defined up to normalization as \[

482: h(\theta)=\frac{\prod_{j=1}^{k}\sinh\left(\frac{1}{2}(\theta-i\beta_{j})\right)\sinh\left(\frac{1}{2}(\theta+i\beta_{j})\right)}{\prod_{j=1}^{l}\sinh\left(\frac{1}{2}(\theta-i\alpha_{j})\right)\sinh\left(\frac{1}{2}(\theta+i\alpha_{j})\right)}\exp\left\{ \int_{0}^{\infty}dtf(t)\frac{\sin^{2}\left(\frac{i\pi-\theta}{2\pi}t\right)}{\sinh x}\right\} .\]

483:  Since the reflection amplitudes are usually expressed as products

484: of the blocks $(x_{i})$ \begin{equation}

485: R(\theta)=\prod_{i}(x_{i});\qquad(x_{i})=\frac{\sinh(\frac{\theta}{2}+i\frac{\pi x_{i}}{2})}{\sinh(\frac{\theta}{2}-i\frac{\pi x_{i}}{2})},\label{eq:blockdef}\end{equation}

486:  to use this theorem we need the integral representation of one single

487: block $(x)$,: \[

488: -(x)=\exp\left\{ 2\int_{0}^{\infty}\frac{dt}{t}\frac{\sinh t(1-x)}{\sinh t}\sinh\left(\frac{t\theta}{i\pi}\right)\right\} .\]

489:  Then, if $R(\theta)$ consists of an even number of blocks, the \emph{minimal}

490: solution (with \textsl{no zeroes and poles}) to eq.(\ref{eq:1pff})

491: can be written as \[

492: r_{\textrm{min}}^{e}(\theta)=\exp\left\{ 2\int_{0}^{\infty}\frac{dt}{t}\frac{\sum_{i}\sinh t(1-x_{i})}{\sinh^{2}t}\left(1-\cosh\frac{t}{2}\cos\frac{t}{\pi}\left(\frac{i\pi}{2}-\theta\right)\right)\right\} .\]

493:  If $R(\theta)$ contains an extra minus sign, or is the product of

494: an odd number of blocks, $R=-\prod(-(x_{i}))$, then the factor $g(\theta)$

495: necessarily contains a zero at the origin which is implemented by

496: putting an extra $\sinh\frac{\theta}{2}$ into it; thus in this case

497: $r_{\textrm{min}}^{o}(\theta)=\sinh\theta\, r_{\textrm{min}}^{e}(\theta)$.

498: In the following an important role is played by the appropriate modification

499: of the \textsl{\emph{minimal}} 1PFF denoted by $r(\theta)$ \[

500: r(\theta)=r_{\textrm{min}}(\theta)\times\frac{\mathrm{zeroes}}{\mathrm{poles}}\]

501:  where the last factor denotes an appropriate number of zeroes and

502: poles at the right places (usually the same as in $R(\theta)$).

503: 

504: Thus the general solution of eq.(\ref{eq:1pff}) can be written as

505: \[

506: F_{1}(\theta)=r(\theta)Q_{1}(y),\qquad y=e^{\theta}+e^{-\theta},\]

507:  where the choice of $Q_{1}(y)$ is restricted by the analyticity

508: and the possible asymptotics of $F_{1}$. It is the $Q_{1}(y)$ in

509: the 1PFF that carries the dependence on the boundary operator $\mathcal{O}$.

510: Note in particular that if $Q_{1}(y)$ corresponds to the operator

511: $\mathcal{O}$ then $\tilde{Q}_{1}(y)\sim y^{N}Q_{1}(y)$ with $N$

512: integer $N\geq1$, describes the 1PFF of the operator $\partial_{\tau}^{N}\mathcal{O}$.

513: 

514: 

515: \subsubsection{Two-particle form factors}

516: 

517: The next step is to investigate the two-particle form factors (2PFF).

518: The novel feature compared to the 1PFF is that their equations contain

519: also the bulk S-matrix. It is worthwhile to go through the analysis

520: in some detail since it is straightforward to write down the general

521: form of the $n$-particle form factors once that of the 2PFF-s is

522: obtained. The equations for the 2PFF-s have the form \begin{eqnarray}

523: F_{2}(\theta_{1},\theta_{2}) & = & S(\theta_{1}-\theta_{2})F_{2}(\theta_{2},\theta_{1}),\quad(a)\qquad F_{2}(\theta_{1},\theta_{2})=R(\theta_{2})F_{2}(\theta_{1},-\theta_{2})\quad(b)\nonumber \\

524: F_{2}(i\pi+\theta_{1},\theta_{2}) & = & R(-\theta_{1})F_{2}(i\pi-\theta_{1},\theta_{2}).\quad(c)\label{eq:2pff}\end{eqnarray}

525:  Note that if $F_{2}(\theta_{1},\theta_{2})$ is a solution to these

526: equations then so is $F_{2}(\theta_{1},\theta_{2})H(\theta_{1},\theta_{2})$

527: provided $H$ is a symmetric, even and $2i\pi$ periodic function.

528: 

529: To construct solutions to eq.(\ref{eq:2pff}) we write \[

530: F_{2}(\theta_{1},\theta_{2})=f(\theta_{1}-\theta_{2})\Psi(\theta_{1},\theta_{2})\]

531:  where $f(\theta)$ is the minimal bulk two particle form factor \cite{FMS},

532: i.e. the minimal solution of \[

533: f(\theta)=S(\theta)f(-\theta),\qquad f(i\pi+\theta)=f(i\pi-\theta).\]

534:  Plugging this $F_{2}$ into (\ref{eq:2pff}a) reveals that $\Psi$

535: must be symmetric $\Psi(\theta_{1},\theta_{2})=\Psi(\theta_{2},\theta_{1})$.

536: The most convenient way to satisfy (\ref{eq:2pff}b) is that $\Psi$

537: has the form \[

538: \Psi(\theta_{1},\theta_{2})=f(\theta_{1}+\theta_{2})r(\theta_{1})r(\theta_{2})\Phi(\theta_{1},\theta_{2})\]

539:  where $\Phi$ is symmetric and even $\Phi(\theta_{1},\theta_{2})=\Phi(\theta_{1},-\theta_{2})$.

540: Finally this $F_{2}$ satisfies eq.(\ref{eq:2pff}c) also if $\Phi(i\pi-\theta_{1},\theta_{2})=\Phi(i\pi+\theta_{1},\theta_{2})$.

541: The conditions on $\Phi$ can be satisfied simply by writing $\Phi(\theta_{1},\theta_{2})=\phi(y_{1},y_{2})$

542: where $\phi$ is a symmetric function of the $y_{i}$-s ($y_{i}=e^{\theta_{i}}+e^{-\theta_{i}}$,

543: $i=1,2$). Thus the general form of the 2PFF, compatible with eq.(\ref{eq:2pff})

544: is \[

545: F_{2}(\theta_{1},\theta_{2})=r(\theta_{1})r(\theta_{2})f(\theta_{1}-\theta_{2})f(\theta_{1}+\theta_{2})\phi(y_{1},y_{2}),\qquad\phi(y_{1},y_{2})=\phi(y_{2},y_{1}).\]

546:  Different choices of the boundary operator $\mathcal{O}$ correspond

547: to different functions $\phi(y_{1},y_{2})$ in this expression.

548: 

549: 

550: \subsubsection{Multi-particle form factors}

551: 

552: From the explicit form of the 2PFF it is clear that the general form

553: of the multi-particle form factors can be written in the following

554: form: \begin{equation}

555: F_{n}(\theta_{1},\theta_{2},\dots,\theta_{n})=G_{n}(\theta_{1},\theta_{2},\dots,\theta_{n})\prod_{i=1}^{n}r(\theta_{i})\prod_{i<j}f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j}),\label{eq:GenAnsatz}\end{equation}

556:  where $f(\theta)$ is the minimal bulk two particle form factor.

557: As a consequence of the form factor equations $G_{n}$ is a $2\pi i$

558: periodic, symmetric and even function of the rapidities: $\theta_{i}$,

559: i.e. it is symmetric in the variable $y_{i}=2\cosh\theta_{i}$. When

560: the bulk S-matrix is nontrivial, the bulk kinematical singularity

561: equations \begin{equation}

562: -i\mathop{\textrm{Res}}_{\theta^{'}=\theta}F_{n+2}(\theta^{'}+i\pi,\theta,\theta_{1},\dots,\theta_{n})=(1-\prod_{i=1}^{n}S(\theta-\theta_{i})S(\theta+\theta_{j}))F_{n}(\theta_{1},\dots,\theta_{n})\label{eq:ksing}\end{equation}

563:  give recursive relations linking $G_{n}$ to $G_{n+2}$. (Note that

564: these singularities are \textsl{absent} in the two particle case).

565: The advantage of using the $y_{i}$-s becomes clear if one tries to

566: describe the bulk kinematical singularities: since $y(i\pi+\theta)=-y(\theta)$,

567: thus including a (symmetric) factor $y_{i}+y_{j}$ in the denominator

568: automatically accounts for the pole. Therefore in the following we

569: put \[

570: G_{n}(\theta_{1},\theta_{2},\dots,\theta_{n})=\frac{Q_{n}(y_{1},y_{2}\dots,y_{n})}{\prod\limits _{i<j}(y_{i}+y_{j})},\]

571:  (with $Q_{n}$ being a symmetric function of $y_{1},\dots,y_{n}$)

572: and then eq.(\ref{eq:ksing}) give recursive relations between the

573: functions $Q_{n}$. Clearly the actual form of these recursive relations

574: varies from model to model since they depend on the bulk S-matrix.

575: The form of the recursions depends also on the choice of the 1PFF

576: $r(\theta)$; it is useful to choose an $r(\theta)$ which gives the

577: simplest possible recursion. Writing the 2PFF in the same form as

578: the $n$-particle one \[

579: \phi(y_{1},y_{2})=\frac{Q_{2}(y_{1},y_{2})}{y_{1}+y_{2}}\]

580:  then the absence of kinematical singularities requires $Q_{2}(y,-y)=0$. 

581: 

582: If the bulk S-matrix is nontrivial and the reflection factor has a

583: pole at $\frac{i\pi}{2}$ then the form factors with odd and even

584: particle number are connected by the boundary kinematical singularity

585: equation:\[

586: -i\mathop{\textrm{Res}}_{\theta=0}F_{n+1}^{\mathcal{O}}(\theta+\frac{i\pi}{2},\theta_{1},\dots,\theta_{n})=\frac{g}{2}\Bigl(1-\prod_{i=1}^{n}S\bigl(\frac{i\pi}{2}-\theta_{i}\bigr)\Bigr)F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\]

587: The corresponding pole in the $n$ particle form factor can be included

588: as \[

589: G_{n}(\theta_{1},\theta_{2},\dots,\theta_{n})=\frac{Q_{n}(y_{1},y_{2}\dots,y_{n})}{\prod_{i}y_{i}\,\prod\limits _{i<j}(y_{i}+y_{j})},\]

590: and the boundary kinematical singularity equation relates $Q_{n}$

591: to $Q_{n+1}$.

592: 

593: The even and the odd particle form factors are also related if the

594: bulk S-matrix has a \lq\lq self fusing'' pole describing the 2

595: particle $\rightarrow$ 1 particle process, which parallels the bulk

596: situation (this happens e.g. in the Lee-Yang model). (In this case

597: it is customary to include this pole also in $f(\theta)$). Since

598: the fusing angle in this process is necessarily $2\pi/3$, one finds

599: from bootstrap that in this case the dynamical singularities imply

600: \begin{equation}

601: -i\mathop{\textrm{Res}}_{\theta^{'}=\theta}F_{n+2}(\theta^{'}+\frac{i\pi}{3},\theta-\frac{i\pi}{3},\theta_{1},\dots,\theta_{n})=\Gamma F_{n+1}(\theta,\theta_{1},\dots,\theta_{n}),\label{eq:dsing}\end{equation}

602:  where $\Gamma$ is related to the residue of the S-matrix at the

603: self fusing pole: $-i\textrm{res}_{\theta=\frac{2\pi i}{3}}S(\theta)=\Gamma^{2}$.

604: 

605: An important restriction follows on the form factor functions from

606: requiring a power law bounded ultraviolet behaviour for the two point

607: correlator of the boundary operators $\langle0|\mathcal{O}(\tau)\mathcal{O}(0)|0\rangle$:

608: the growth of the function $F_{n}(\theta_{1},\dots,\theta_{n})$ must

609: be bounded by some exponential of the rapidity as $\theta\rightarrow\infty$

610: (i.e. the form factors only grow polynomially with particle energy).

611: This can be shown using an argument identical to that in the bulk

612: case \cite{counting_ops}. If $r\left(\theta\right)$ and $f\left(\theta\right)$

613: are specified in a way to include all poles induced by the dynamics

614: of the model, then it follows that the functions $Q_{n}$ must be

615: \textsl{polynomials} of the $y_{i}$. Therefore in the following we

616: only look for explicit \textsl{polynomial} solutions of the various

617: recursion equations. This is a posteriori confirmed since we find

618: as many polynomial solution of the boundary form factor equation as

619: many independent local operator exist in the theories. 

620: 

621: 

622: \subsection{Two-point function}

623: 

624: Once an appropriate solution of the form factor axioms is found it

625: can be used to describe correlators of boundary operators. The two-point

626: function of the boundary operator $\mathcal{O}$ can be computed by

627: inserting a complete set of states \begin{equation}

628: \langle0\vert\mathcal{O}(t)\mathcal{O}(0)\vert0\rangle=\sum_{n=0}^{\infty}\frac{1}{(2\pi)^{n}}\int_{\theta_{1}>\theta_{2}>\dots>\theta_{n}>0}d\theta_{1}d\theta_{2}\dots d\theta_{n}e^{-imt\sum_{i}\cosh\theta_{i}}F_{n}F_{n}^{+}\label{eq:2pt}\end{equation}

629: where time translation invariance was used and the form factors are

630: \[

631: F_{n}=\langle0\vert\mathcal{O}(0)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}=F_{n}^{\mathcal{O}}(\theta_{1},\theta_{2},\dots,\theta_{n})\]

632: and \[

633: F_{n}^{+}=\,_{in}\langle\theta_{1},\theta_{2},\dots,\theta_{n}\vert\mathcal{O}(0)\vert0\rangle=F_{n}^{\mathcal{O}}(i\pi+\theta_{n},i\pi+\theta_{n-1},\dots,i\pi+\theta_{1})\]

634:  which, for unitary theories, is the complex conjugate of the previous

635: one: $F_{n}^{+}=F_{n}^{*}$. In the Euclidean $(\tau=it)$ version

636: of the theories the form factor expansion of the correlator for large

637: separations converges rapidly since multi-particle terms are exponentially

638: suppressed. 

639: 

640: The identification between solutions of the form factor equations

641: and operators of the theory is a central issue. One possible way is

642: to analyze the behaviour of the boundary correlators for short distances.

643: Although on general grounds one may expect the form factor expansion

644: to converge rapidly only in the infrared (large volume) regime, the

645: examples from the various bulk theories, where the form factor expansion

646: converges even in the UV domain (see e.g. \cite{Z1}), suggest that

647: similar behaviour may happen in the boundary setting as well. If the

648: theory can be described as a relevant perturbation of a conformal

649: field theory, then in the UV domain the two-point function must follow

650: a behaviour dictated by this limiting theory. The short distance singularity

651: exponent is related to the scaling dimension of the operator $\mathcal{O}$

652: and can be calculated from the asymptotic growth of the form factors. 

653: 

654: 

655: \section{Model studies}

656: 

657: In this section we carry out a detailed investigation of the solutions

658: of the form factor equations in four different models. The first two

659: models (the massive scalar field with linear boundary interaction

660: and the Ising model interacting with a boundary magnetic field \cite{GZ})

661: are free in the bulk and the correlation functions are known explicitly,

662: thus the form factors obtained from the explicit field theoretic solutions

663: can be compared directly to the solutions of the form factor equations.

664: In both cases we find that the space of appropriate \textsl{polynomial}

665: solutions of the FF equations can be identified with the space of

666: local boundary operators obtained from the explicit construction.

667: In the case of the Ising model we also show how the conformal dimensions

668: of the various operators of the ultraviolet limiting BCFT can be obtained

669: from the solutions of the FF equations. The third and fourth models,

670: namely the scaling Lee-Yang and the sinh-Gordon models with integrability

671: preserving boundaries are among the simplest boundary integrable theories.

672: In contrast to the previous cases they cannot be solved directly so

673: one has to rely upon the solution of the form factor equations. Since

674: these models contain nontrivial bulk interactions the recursion relations

675: connecting the multi-particle form factors are no longer trivial,

676: and in these cases we investigate their solutions in detail.

677: 

678: 

679: \subsection{Massive scalar with linear boundary interaction}

680: 

681: 

682: \subsubsection{Direct calculation}

683: 

684: The free massive scalar field $\Phi(x,t)$ restricted to the negative

685: half-line $x\leq0$ subject to linear boundary condition\begin{equation}

686: \partial_{x}\Phi(x,t)|_{x=0}=-\lambda(\Phi(0,t)-\Phi_{0}).\label{eq:robin}\end{equation}

687:  can be described by the following Lagrangian:\[

688: \mathcal{L}=\Theta(-x)\left(\frac{1}{2}(\partial_{t}\Phi)^{2}-\frac{1}{2}(\partial_{x}\Phi)^{2}-\frac{m^{2}}{2}\Phi^{2}\right)-\delta(x)\frac{\lambda}{2}(\Phi-\Phi_{0})^{2},\]

689: This one parameter family of linear boundary condition interpolates

690: between Neumann $\partial_{x}\Phi\vert_{x=0}=0$ (for $\lambda=0$)

691: and Dirichlet $\Phi\vert_{x=0}=\Phi_{0}$ (for $\lambda\to\infty$)

692: boundary conditions. Since for any $\lambda$ we are dealing with

693: a free theory it can be solved explicitly. The mode decomposition

694: of the field is \[

695: \Phi(x,t)=Ae^{mx}+\int_{0}^{\infty}\frac{dk}{\omega(k)}\Bigl\{ a(k)e^{-i\omega(k)t}\bigl(e^{ikx}+R(k)e^{-ikx}\bigr)+a^{+}(k)e^{i\omega(k)t}\bigl(e^{-ikx}+R(-k)e^{ikx}\bigr)\Bigr\}\]

696:  where $A=\frac{\lambda}{m+\lambda}\Phi_{0}$ and \[

697: R(k)=\frac{k-i\lambda}{k+i\lambda}\]

698:  is the reflection factor on the boundary at $x=0$. The creation/annihilation

699: operators are normalized as \[

700: [a(k),a^{+}(k^{'})]=2\pi\omega(k)\delta(k-k^{'})\quad,\quad k\,,\: k^{'}>0\]

701: The boundary two-point function can be calculated easily\[

702: \langle0|\Phi(0,t)\Phi(0,t^{'})|0\rangle=A^{2}+\int_{0}^{\infty}\frac{dk}{2\pi\omega(k)}e^{-i\omega(k)(t-t^{'})}\left(1+R(k)\right)\left(1+R(-k)\right)\]

703:  By comparing this expression to the form factor expansion of the

704: two-point function (\ref{eq:2pt}), the form factor of the elementary

705: field can be extracted: \[

706: \langle0|\Phi(0,t)|\theta\rangle=e^{-i\omega(k)t}\left(1+R(k)\right)\]

707: The same result can be obtained by taking the general (space-dependent)

708: two point function \begin{eqnarray*}

709: \langle0|T\left(\Phi(x,t)\Phi(x^{'},t^{'})\right)|0\rangle & = & A^{2}e^{m(x+x^{'})}\\

710:  &  & +\int\frac{d^{2}k}{(2\pi)^{2}}\frac{i}{k^{2}-m^{2}+i\epsilon}e^{-ik_{0}(t-t^{'})}\left(e^{-ik_{1}(x-x^{'})}+R(k)e^{+ik_{1}(x+x^{'})}\right)\end{eqnarray*}

711:  and using the boundary reduction formula \cite{BBT} \begin{eqnarray*}

712: \langle0|\Phi(x,t)|\theta\rangle & = & 2i\int_{-\infty}^{0}dx^{'}\int_{-\infty}^{\infty}dt^{'}\: e^{i\omega(\theta)t^{'}}\cos(p(\theta)x^{'})\left\{ \partial_{t^{'}}^{2}-\partial_{x^{'}}^{2}+m^{2}+\delta(x^{'})\partial_{x^{'}}\right\} \\

713:  &  & \hspace{8cm}\langle0|T\left(\Phi(x,t)\Phi(x^{'},t^{'})\right)|0\rangle\end{eqnarray*}

714:  where $\omega(\theta)=m\cosh(\theta)$ and $p(\theta)=m\sinh(\theta)$.

715: Performing explicitly the calculation yields\begin{equation}

716: \langle0|\Phi(x,t)|\theta\rangle=e^{-i\omega(\theta)t}(e^{ip(\theta)x}+R(\theta)e^{-ip(\theta)x})\label{eq:Phivev}\end{equation}

717: which for the form factor of the operator $\Phi(x=0,t)$ reads as 

718: 

719: \[

720: \langle0|\Phi(0,t)|\theta\rangle=e^{-i\omega(\theta)t}\left(1+R(\theta)\right)\]

721:  Introducing $\tau=-it$ one also gets\[

722: \langle0|\partial_{\tau}^{n}\Phi(0,0)|\theta\rangle=\omega(\theta)^{n}\left(1+R(\theta)\right),\quad n>0.\]

723:  Clearly these operators have no multi-particle matrix elements. It

724: is important to realize that $\partial_{x}\Phi(0,0)$ is not an independent

725: operator since the boundary condition eq.(\ref{eq:robin}) relates

726: it to $\Phi(0,0)$, thus the set of independent boundary operators

727: having only one particle matrix elements is given by $\partial_{\tau}^{n}\Phi(0,0)$.

728: To obtain multi-particle matrix elements one has to consider \[

729: \langle0|:\Phi(x_{1},t_{1})\dots\Phi(x_{k},t_{k}):|\theta_{1}\dots\theta_{k}\rangle.\]

730:  Using the analogous boundary reduction formula and the Wick theorem

731: we obtain \begin{eqnarray}

732:  &  & \langle0|:\Phi(x_{1},t_{1})\dots\Phi(x_{k},t_{k}):|\theta_{1}\dots\theta_{k}\rangle=\label{eq:Phisvev}\\

733:  &  & \left\{ e^{-i\omega(\theta_{1})t_{1}}\left(e^{ip(\theta_{1})x_{1}}+R(\theta_{1})e^{-ip(\theta_{1})x_{1}}\right)\right\} \dots\left\{ e^{-i\omega(\theta_{k})t_{k}}\left(e^{ip(\theta_{k})x_{k}}+R(\theta_{k})e^{-ip(\theta_{k})x_{k}}\right)\right\} +\dots,\nonumber \end{eqnarray}

734:  where the ellipses at the end represent additional terms which make

735: it completely symmetric in all coordinates. From this expression one

736: can extract the form factor of the most general boundary operator

737: of the theory\[

738: \langle0|:\partial_{\tau_{1}}^{n_{1}}\Phi(0,0)\dots\partial_{\tau_{k}}^{n_{k}}\Phi(0,0):|\theta_{1}\dots\theta_{k}\rangle=\omega(\theta_{1})^{n_{1}}\left(1+R(\theta_{1})\right)\dots\omega(\theta_{k})^{n_{k}}\left(1+R(\theta_{k})\right)+\dots\]

739:  where again a complete symmetrization in the $\theta_{i}$ rapidities

740: is understood. Checking the leading asymptotic behaviour of these

741: form factors gives that for all $\theta_{i}\sim\theta$ large they

742: grow as $e^{N\theta}$, where $N=n_{1}+\dots+n_{k}$ is the total

743: number of derivatives in the expression. We note that we have as many

744: operators for a given $N$ as many partition $N$ has into the numbers

745: $1,2,\dots,k$. This can be seen by writing $N=N_{1}+2N_{2}+\dots+kN_{k}$

746: and associating to it the operator with $n_{1}=N_{1}+N_{2}\dots+N_{k}$,

747: $n_{2}=N_{2}+\dots+N_{k}$ $\dots$ $n_{k}=N_{k}$ derivatives.

748: 

749: The Dirichlet boundary condition ($R=-1$) can be obtained in the

750: $\lambda\to\infty$ limit. Clearly $\Phi\vert_{x=0}=\Phi_{0}$ is

751: a c-number and the Dirichlet boundary condition does not connect the

752: operator $\partial_{x}\Phi\vert_{x=0}$ to $\Phi\vert_{x=0}$. We

753: can extract, however, the form factors of the operator $\partial_{x}\Phi(0,t)$

754: from that of $\Phi(0,t)$ by taking the $\lambda\to\infty$ limit

755: carefully in (\ref{eq:Phivev}): \[

756: \langle0|\partial_{x}\Phi(0,t)|\theta\rangle=e^{-i\omega(\theta)t}2ip(\theta)\]

757: and for its derivatives \[

758: \langle0|\partial_{\tau}^{n}\partial_{x}\Phi(0,0)|\theta\rangle=\omega(\theta)^{n}2ip(\theta),\quad n>0.\]

759:  This can be extended similarly to the most general operator as \[

760: \langle0|:\partial_{\tau}^{n_{1}}\partial_{x}\Phi(0,0)\dots\partial_{\tau}^{n_{k}}\partial_{x}\Phi(0,0):|\theta_{1}\dots\theta_{k}\rangle=\omega(\theta_{1})^{n_{1}}2ip(\theta_{1})\dots\omega(\theta_{k})^{n_{k}}2ip(\theta_{k})+\dots\]

761:  where again a complete symmetrization in the $\theta_{i}$ rapidities

762: is understood. Checking the leading asymptotic behaviour of these

763: form factors gives that for all $\theta_{i}\sim\theta$ large they

764: grow as $e^{N\theta}$, where $N=k+n_{1}+\dots+n_{k}$ is the total

765: number of derivatives in the expression. 

766: 

767: 

768: \subsubsection{Solving the form factor equations}

769: 

770: The bulk $S$-matrix of the theory together with the reflection factor

771: are \[

772: S(\theta)=1,\quad R(\theta)=\frac{\sinh\theta-i\frac{\lambda}{m}}{\sinh\theta+i\frac{\lambda}{m}}=-\left(1+\frac{B}{2}\right)\left(-\frac{B}{2}\right),\quad\frac{\lambda}{m}=\sin\frac{\pi B}{2},\]

773: where the block notation (\ref{eq:blockdef}) is used to express $R(\theta)$.

774: As a consequence of $S=1$ the minimal bulk form factor is trivial:

775: $f(\theta)=1$. To determine the 1PFF we note that this reflection

776: factor is identical to the two particle S-matrix of the sinh-Gordon

777: model if the above identification of parameters is done. Therefore

778: eq.(\ref{eq:g1}) in this case is identical to the equation for the

779: minimal bulk form factor of the sinh-Gordon model. Choosing for $g(\theta)$

780: the solution given in \cite{FMS} (described in detail in the sinh-Gordon

781: section) gives \[

782: r(\theta)=2g(\theta)g(i\pi-\theta)=\frac{2\sinh\theta}{\sinh\theta+i\frac{\lambda}{m}}=1+R(\theta)\]

783:  Clearly this corresponds to the form factor of the operator $\Phi(0,0)-A$. 

784: 

785: Now we demonstrate that the number of independent solutions of the

786: form factor equations coincides with the number of local boundary

787: operators. In this case the general Ansatz (\ref{eq:GenAnsatz}) takes

788: the following form\[

789: F_{n}(\theta_{1},\dots,\theta_{n})=P_{n}(y_{1},\dots,y_{n})\prod_{i}r(\theta_{i})\]

790:  where $y_{i}=2\cosh\theta_{i}$ as before. Since the bulk $S$-matrix

791: is trivial there are no bulk/boundary kinematical singularities and

792: $P_{n}$ is a completely symmetric polynomial in the $y_{i}$-s. One

793: can count the independent solutions of the BFF equations by counting

794: the possible solutions for $P_{n}$. If $P_{n}$ has degree $N$ then

795: the solutions are given by the partitions of $N$ into the numbers

796: $1,2,\dots n$ in the following way. Since the completely symmetric

797: polynomials of $n$ variable are generated by the $\sigma_{i}$-s

798: (elementary symmetric polynomials of degree $i$) one can write:\[

799: \prod_{i=1}^{n}(x+x_{i})=\sum_{i=1}^{n}\sigma_{i}\, x^{n-i};\quad P_{n}\propto\sigma_{1}^{k_{1}}\sigma_{2}^{k_{2}}\dots\sigma_{n}^{k_{n}};\quad N=\sum ik_{i}\]

800:  It is clear that this space has the same dimension as the space of

801: boundary operators having only $n$ particle matrix elements with

802: asymptotic growth $e^{N\theta}$ . 

803: 

804: The Dirichlet ($\lambda\to\infty$) limit for the operator $\partial_{x}\Phi(0,0)$

805: can be obtained using its relation to $\Phi(0,0)$ via the boundary

806: condition (\ref{eq:robin}) as we did in the Lagrangian framework. 

807: 

808: 

809: \subsection{Ising model with boundary magnetic field }

810: 

811: 

812: \subsubsection{Direct calculation}

813: 

814: The Ising model with a boundary magnetic field can be described by

815: a free massive Majorana fermion perturbed at the boundary \cite{GZ}.

816: In Minkowskian formalism the Dirac equation can be obtained form the

817: Lagrangian:\[

818: \mathcal{L}=\frac{1}{2}\bar{\Psi}(i\gamma^{\mu}\partial_{\mu}-m)\Psi\]

819: The gamma matrices are chosen as \[

820: \gamma^{0}=\left(\begin{array}{cc}

821: 0 & -i\\

822: i & 0\end{array}\right)\quad;\quad\gamma^{1}=\left(\begin{array}{cc}

823: 0 & -i\\

824: -i & 0\end{array}\right)\]

825: in order for the Dirac equation to be real:\[

826: \left(\begin{array}{cc}

827: -m & \partial_{x}+\partial_{t}\\

828: \partial_{x}-\partial_{t} & -m\end{array}\right)\Psi=0\]

829: Thus the Majorana condition corresponds to taking $\Psi$ real. Using

830: the component notation \[

831: \Psi=\left(\begin{array}{c}

832: \psi_{+}\\

833: \psi_{-}\end{array}\right)\]

834: the Lagrangian of the boundary field theory takes the form 

835: 

836: \[

837: \mathcal{L}=-i\Theta(-x)\left(\frac{1}{2}\psi_{+}(\partial_{t}-\partial_{x})\psi_{+}-\frac{1}{2}\psi_{-}(\partial_{t}+\partial_{x})\psi_{-}-m\psi_{+}\psi_{-}\right)-i\delta(x)U_{B}\]

838: where \[

839: U_{B}=\frac{1}{2}\psi_{+}\psi_{-}+\frac{1}{2}a\dot{a}+\frac{1}{2}ha(\psi_{+}+\psi_{-})\]

840: The operator $a$ is a boundary fermion $a^{2}=1$, which implements

841: the boundary condition \[

842: \partial_{t}(\psi_{+}-\psi_{-})=\frac{h^{2}}{2}(\psi_{+}+\psi_{-})\]

843: Since the theory is free we can solve it explicitly. The mode expansion

844: of the fermionic fields are \begin{eqnarray*}

845: \psi_{\pm}(x,t) & = & \int_{0}^{\infty}\frac{d\theta}{2\pi}\Biggl\{ b(\theta)e^{-i\omega(\theta)t}\biggl(u_{\pm}(\theta)e^{ik(\theta)x}+R(\theta)u_{\pm}(-\theta)e^{-ik(\theta)x}\biggr)\\

846:  &  & \hspace{2cm}+b^{+}(\theta)e^{i\omega(\theta)t}\biggl(v_{\pm}(\theta)e^{-ik(\theta)x}+R(-\theta)v_{\pm}(-\theta)e^{ik(\theta)x}\biggr)\Biggr\}\end{eqnarray*}

847: where $u_{\pm}(\theta)=v_{\pm}^{*}(\theta)=\sqrt{m}e^{\mp\frac{i\pi+2\theta}{4}}$

848: are the spinor amplitudes, $R(\theta)$ is nothing else but the one-particle

849: reflection factor \[

850: R(\theta)=i\tanh\left(\frac{i\pi}{4}-\frac{\theta}{2}\right)\frac{\sinh\theta+i\kappa}{\sinh\theta-i\kappa}\qquad,\quad\kappa=1-\frac{h^{2}}{2m}\]

851:  and the creation/annihilation operators are normalized as \[

852: \{ b(\theta),b^{+}(\theta^{'})\}=2\pi\delta(\theta-\theta^{'})\]

853: The boundary two point function can be calculated explicitly: \begin{equation}

854: \langle0|\psi_{\pm}(0,t)\psi_{+}(0,t^{'})|0\rangle=\int_{0}^{\infty}\frac{d\theta}{2\pi}e^{-i\omega(\theta)(t-t^{'})}\left(u_{\pm}(\theta)+R(\theta)u_{\pm}(-\theta)\right)\left(v_{+}(\theta)+R(-\theta)v_{+}(-\theta)\right)\label{eq:Fermcorrexact}\end{equation}

855: which, as compared to the form factor expansion (\ref{eq:2pt}), gives

856: \begin{equation}

857: \langle0\vert\psi_{\pm}(0,t)\vert\theta\rangle=e^{-i\omega(\theta)t}\left(u_{\pm}(\theta)+R(\theta)u_{\pm}(-\theta)\right)\label{eq:Fermionffexact}\end{equation}

858: These two operators are not independent since they are related by

859: the boundary condition and so there is only one boundary fermion field,

860: say $\psi_{+}$. As a result, the algebraically independent operators

861: at the boundary are the fermion field and its derivatives $\partial_{\tau}^{n}\psi_{+}$.

862: Note that $\partial_{x}\psi_{+}|_{x=0}$ is not an independent field,

863: as it is determined by the Dirac equation in terms of $\partial_{\tau}\psi_{+}|_{x=0}$

864: and $\psi_{-}|_{x=0}$. As a consequence of the fermionic nature of

865: the field the most general boundary operator has the form $\partial_{\tau}^{n_{1}}\psi_{+}\partial_{\tau}^{n_{2}}\psi_{+}\dots\partial_{\tau}^{n_{k}}\psi_{+}$

866: where $n_{1}>n_{2}>\dots>n_{k}$ (the inequalities are strict, in

867: contrast to the bosonic case discussed earlier). $N=n_{1}+n_{2}+\dots+n_{k}$

868: is called the level of the operator, and operators at level $N$ can

869: be brought in one-to-one correspondence with partitions of $N$ into

870: the numbers $1,2,\dots,k$. For a partition \[

871: N=kN_{k}+(k-1)N_{k-1}+\dots+2N_{2}+N_{1}\]

872: we associate the operator above with $n_{k}=N_{k};\, n_{k-1}=N_{k}+N_{k-1};\dots;\, n_{1}=N_{k}+N_{k-1}+\dots+N_{1}$. 

873: 

874: (Had we included also the operator $a$ we would have had to perform

875: a GSO type projection, leaving only non-fermionic operators. This

876: would amount to keeping all operators with an even number of fermion

877: factors plus all odd ones multiplied with a factor $a$, but this

878: would lead to the same number of operators.)

879: 

880: 

881: \subsubsection{Solution of the FF bootstrap}

882: 

883: Using again the block notation (\ref{eq:blockdef}) the $S$-matrix

884: of the theory and the reflection factor are \cite{GZ}\[

885: S(\theta)=-1\quad,\qquad R_{x}(\theta)=[x]\left(-\frac{1}{2}\right)\quad,\qquad[x]=(x)(1-x)\]

886:  where $x$ is related to the boundary magnetic field as \[

887: \sin\pi x=1-\frac{h^{2}}{2m}=\kappa\]

888:  For $h=0$ we recover the free boundary condition with \[

889: R_{\textrm{free}}(\theta)=\left(\frac{1}{2}\right)\]

890: which has a pole at $i\frac{\pi}{2}$ corresponding to the fact that

891: the ground state is doubly degenerate. In contrast to the generic

892: situation the pole at $i\frac{\pi}{2}$ is a \emph{dynamical} pole

893: and not a kinematical one (since the field has no vacuum expectation

894: value at all). The $h\rightarrow\infty$ limit corresponds to the

895: fixed boundary condition (when the Ising spin takes a fixed value

896: at the boundary), and the reflection factor is \[

897: R_{\textrm{fixed}}(\theta)=\left(-\frac{1}{2}\right)\]

898: which has no pole in the physical strip at all. The minimal one particle

899: form factor for the fixed case can be calculated directly using the

900: recipe in Section 2 \[

901: r_{\textrm{fixed}}(\theta)=\frac{\sinh\theta}{\sinh(\frac{\theta}{2}+i\frac{\pi}{4})}\]

902: For the free case we include the dynamical singularity into the 1PFF

903: \[

904: r_{\textrm{free}}(\theta)=-2i\frac{\sinh\theta}{\cosh\theta}\sinh\left(\frac{\theta}{2}+i\frac{\pi}{4}\right)\]

905:  The simplest solution which interpolates between these two cases

906: and has a pole exactly at the location of the boundary dynamical singularity

907: of the reflection factor is \begin{equation}

908: r(\theta)=\frac{\sinh\theta}{\sinh(\frac{\theta}{2}+i\frac{\pi}{4})}\frac{\cosh\theta+i(1-\kappa)}{\sinh\theta-i\kappa}.\label{eq:rising}\end{equation}

909:  This expression is the same we obtained from the exact solution of

910: the model (\ref{eq:Fermionffexact}). The minimal bulk two particle

911: form factor is simply \[

912: f(\theta)=\sinh\frac{\theta}{2}.\]

913:  Since $R(\theta)$ has no kinematical pole at $i\frac{\pi}{2}$,

914: boundary kinematical singularities are absent, and since the bulk

915: $S$-matrix is $-1$ there are no bulk kinematical singularities either.

916: Thus we look for the $n$ particle form factors in the form (\ref{eq:GenAnsatz})

917: \[

918: F_{n}(\theta_{1},\dots,\theta_{n})=P_{n}(y_{1},\dots,y_{n})\prod_{i}r(\theta_{i})\prod_{i<j}f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j}),\]

919:  where $y_{i}=2\cosh\theta_{i}$ and $P_{n}$ is a completely symmetric

920: polynomial in the $y_{i}$-s. Taking into account the special form

921: of $f(\theta)$ the form factor simplifies to \begin{equation}

922: F_{n}(\theta_{1},\dots,\theta_{n})=P_{n}(y_{1},\dots,y_{n})\prod_{i}r(\theta_{i})\prod_{i<j}(y_{i}-y_{j})\label{eq:isingff}\end{equation}

923:  The independent solutions are counted in the same way as in the bosonic

924: case, i.e. by the partitions of $N$ into the numbers $1,2,\dots,n$

925: and are generated by the $\sigma_{i}$-s. It is clear that the dimension

926: of the space they span is the same as the one of the boundary operators

927: obtained from the direct calculation. 

928: 

929: Since the UV limit of this theory is a boundary conformal field theory

930: one can go further than in the bosonic case and calculate the UV dimension

931: of the various boundary operators. As the form factor equations are

932: not coupled we can choose a basis among operators consisting of those

933: having matrix elements only with a certain fixed number of particles.

934: 

935: If the operator has only one-particle matrix element then its correlator

936: is \begin{equation}

937: \langle0|\mathcal{O}(\tau)\mathcal{O}(0)|0\rangle=\int_{0}^{\infty}\frac{d\theta}{2\pi}|F_{1}^{\mathcal{O}}(\theta)|^{2}e^{-m\cosh\theta\tau}\label{eq:isingtpf}\end{equation}

938:  where $F_{1}^{\mathcal{O}}(\theta)=r(\theta)P_{1}(y)$. Plugging

939: (\ref{eq:rising}) into (\ref{eq:isingtpf}) we obtain the exact correlator

940: (\ref{eq:Fermcorrexact}). If the operator $\mathcal{O}$ goes to

941: a scaling operator in the UV limit ($\tau\rightarrow0$) then the

942: exact correlator has the short distance asymptotics $\tau^{-2\Delta}$,

943: where $\Delta$ is the appropriate scaling dimension in the ultraviolet

944: BCFT. In (\ref{eq:isingtpf}) the singularity comes from the large

945: $\theta$ asymptotics of the form factor. If $|F_{1}^{\mathcal{O}}(\theta)|^{2}$

946: diverges as $y^{n}$ here, then the corresponding weight is $\Delta=\frac{n}{2}$.

947: Taking the simplest solution $P_{1}(y)=1$ the weight is $\Delta=\frac{1}{2}$

948: which corresponds to the boundary fermion field. Choosing $P_{1}(y)=\sigma_{1}^{n}(y)$

949: corresponds to the $n$-th derivative of this operator which has weight

950: $n+\frac{1}{2}$.

951: 

952: Similarly we can analyze an operator having $n$-particle matrix element

953: only. The corresponding correlator is \[

954: \langle0|\mathcal{O}_{(n)}(\tau)\mathcal{O}_{(n)}(0)|0\rangle=\int_{0}^{\infty}\frac{d\theta_{1}}{2\pi}\dots\int_{0}^{\infty}\frac{d\theta_{n}}{2\pi}\frac{1}{n!}|F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})|^{2}e^{-m(\cosh\theta_{1}+\dots+\cosh\theta_{n})\tau}\]

955:  The operator which has the mildest UV behaviour corresponds to $P_{n}=1$.

956: The corresponding form factor square for large $\theta$-s behaves

957: as \[

958: |F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})|^{2}\propto\mathrm{exp}(\theta(n+n(n-1))=e^{\theta n^{2}},\]

959:  thus the UV dimension of $\mathcal{O}_{(n)}$ is $\Delta=\frac{n^{2}}{2}$.

960: The explicit boundary operator which has nonzero matrix elements only

961: with $n$ particle states and has the mildest UV behaviour is \[

962: \psi_{+}\partial_{\tau}\psi_{+}\dots\partial_{\tau}^{n-1}\psi_{+}\]

963: with dimension $\Delta=\frac{n}{2}+\frac{n(n-1)}{2}$; therefore it

964: can be associated to $\mathcal{O}_{(n)}$.

965: 

966: To match the descendent operators, note that to any partition of $N=k_{1}+2k_{2}+\dots+n\cdot k_{n}$

967: there exists a solution of the form factor equations with $P_{n}^{N}=\sigma_{1}^{k_{1}}\dots\sigma_{n}^{k_{n}}$.

968: The number of such polynomials is the same as the number of descendants

969: of $\mathcal{O}_{(n)}$ at level $N$: to the given partition we can

970: associate the operator \[

971: \partial_{\tau}^{k_{n}}\psi_{+}\partial_{\tau}^{1+k_{n-1}+k_{n}}\psi_{+}\dots\partial_{\tau}^{n-1+k_{1}+\dots+k_{n}}\psi_{+}\]

972: Conversely, given an operator of the form \[

973: \partial_{\tau}^{p_{1}}\psi_{+}\dots\partial_{\tau}^{p_{n}}\psi_{+}\qquad,\qquad0\leq p_{1}<p_{2}\dots<p_{n}\]

974: of weight $N+\frac{n^{2}}{2}$, one can define a partition as $k_{n}=p_{1}$,

975: $k_{n-1}=p_{2}-k_{n}-1$, $\dots$ and thus associate a polynomial

976: solution of the form factor equations with appropriate asymptotic

977: behaviour. It is important to emphasize that we do not claim that

978: the form factor related to $P_{n}^{N}$ belongs to the operator above,

979: what we have shown is only that the dimension of the space of operators

980: with certain scaling dimension is the same as the dimension of the

981: solution of the form factor equations.

982: 

983: 

984: \subsection{The boundary scaling Lee-Yang model}

985: 

986: The scaling Lee-Yang model with boundary is a combined bulk and boundary

987: perturbation of the boundary version of the ${\mathcal{M}}(2/5)$

988: minimal model \cite{DPTW1pt,DPTW1}. In the bulk the perturbation

989: is given by the unique relevant spinless field $\phi$, at the boundary

990: the perturbation depends on which of the two possible conformal boundary

991: conditions is present in the unperturbed model. One, denoted by $\textrm{1}$

992: in \cite{DPTW1}, does not have any relevant boundary fields - thus

993: can have no boundary perturbation either -, while the other, denoted

994: $\Phi$ in \cite{DPTW1}, has a single relevant boundary field $\varphi$

995: with scaling dimension $-1/5$. In this latter case the general perturbed

996: action is \[

997: {\mathcal{A}}_{\lambda,\Phi(h)}={\mathcal{A}}_{\Phi}+\lambda\int\limits _{-\infty}^{\infty}dy\int\limits _{-\infty}^{0}dx\phi(x,y)+h\int\limits _{-\infty}^{\infty}dy\varphi(y),\]

998:  where ${\mathcal{A}}_{\Phi}$ denotes the action for ${\mathcal{M}}(2/5)$

999: with the $\Phi$ boundary condition imposed at $x=0$, and $\lambda$

1000: and $h$ denote the bulk and boundary couplings respectively. The

1001: action of ${\mathcal{A}}_{\lambda,\textrm{1}}$ is similar, but the

1002: last term on the right hand side is missing. If $\lambda>0$ then

1003: in all cases the bulk behaviour is described by an integrable massive

1004: theory having only a single particle type with the following S matrix

1005: \cite{CM}: \[

1006: S(\theta)=-\left(\frac{1}{3}\right)\left(\frac{2}{3}\right)=-\left[\frac{1}{3}\right]\quad;\quad(x)=\frac{\sinh\left(\frac{\theta}{2}+\frac{i\pi x}{2}\right)}{\sinh\left(\frac{\theta}{2}-\frac{i\pi x}{2}\right)}.\]

1007:  The pole at $\theta=\frac{2\pi i}{3}$ corresponds to the {}``$\varphi^{3}$

1008: property'', i.e. the particle appears as a bound state of itself.

1009: The minimal bulk two particle form factor which has only a zero at

1010: $\theta=0$ and a pole at $\theta=\frac{2\pi i}{3}$ in the strip

1011: $0\leq\Im m(\theta)<\pi$ has the form \cite{Z1}: \[

1012: f(\theta)=\frac{y-2}{y+1}v(i\pi-\theta)v(-i\pi+\theta)\quad,\quad y=e^{\theta}+e^{-\theta},\]

1013:  where \[

1014: v(\theta)=\exp\left\{ 2\int_{0}^{\infty}\frac{dt}{t}e^{i\frac{\theta t}{\pi}}\frac{\sinh\frac{t}{2}\sinh\frac{t}{3}\sinh\frac{t}{6}}{\sinh^{2}t}\right\} .\]

1015: 

1016: 

1017: In the boundary theory with the perturbed $\Phi$ boundary, the reflection

1018: amplitude of the particle depends on the strength of the coupling

1019: constant of the boundary field as \cite{DPTW1} \[

1020: R(\theta)_{\Phi}=R_{0}(\theta)R(b,\theta)=\left(\frac{1}{2}\right)\left(\frac{1}{6}\right)\left(-\frac{2}{3}\right)\left[\frac{b+1}{6}\right]\left[\frac{b-1}{6}\right],\]

1021:  where the dimensionless parameter $b$ is related to the dimensionful

1022: $h$ as \[

1023: h(b)=\sin\left(\bigl(b+\frac{1}{2}\bigr)\frac{\pi}{5}\right)m(\lambda)^{6/5}h_{crit},\qquad h_{crit}=-\pi^{\frac{3}{5}}2^{\frac{4}{5}}5^{\frac{1}{4}}\frac{\sin\frac{2\pi}{5}}{\sqrt{\Gamma(\frac{3}{5})\Gamma(\frac{4}{5})}}\left(\frac{\Gamma(\frac{3}{5})}{\Gamma(\frac{3}{5})}\right)^{\frac{6}{5}},\]

1024:  and $m(\lambda)$ is the mass of the particle giving the overall

1025: scale in the infrared description. In the case of the $\textrm{1}$

1026: boundary the reflection amplitude is the parameter independent expression

1027: \[

1028: R(\theta)_{1}=\left(\frac{1}{2}\right)\left(\frac{1}{6}\right)\left(-\frac{2}{3}\right).\]

1029:  Note that $R(\theta)_{1}$ is identical to $R(\theta)_{\Phi}$ at

1030: $b=0$ and so both have a pole at $\theta=i\pi/2$ coming from the

1031: $\left(\frac{1}{2}\right)$ block but their $g$ factors differ in

1032: a sign \cite{DTWbct}. 

1033: 

1034: 

1035: \subsubsection{Lee-Yang model with perturbed $\Phi$ boundary }

1036: 

1037: We consider first the Lee-Yang model with perturbed $\Phi$ boundary.

1038: The 1PFF corresponding to $R(\theta)_{\Phi}$ is chosen as \[

1039: r(\theta)=\frac{i\sinh\theta}{(\sinh\theta-i\sin\frac{\pi(b+1)}{6})(\sinh\theta-i\sin\frac{\pi(b-1)}{6})}u(\theta),\]

1040:  where \[

1041: u(\theta)=\exp\left\{ \int_{0}^{\infty}\frac{dt}{t}\left[\frac{1}{\sinh\frac{t}{2}}-2\cosh\frac{t}{2}\cos\left[\left(\frac{i\pi}{2}-\theta\right)\frac{t}{\pi}\right]\frac{\sinh\frac{5t}{6}+\sinh\frac{t}{2}-\sinh\frac{t}{3}}{\sinh^{2}t}\right]\right\} .\]

1042:  Note that $r\sim1$ at $y\rightarrow\infty$, and $r(\theta)$ satisfies

1043: the $r(\theta+i\pi)=r(\theta)^{*}$ reality condition for real $\theta$.

1044: The general $n$-particle form factors have the form \begin{equation}

1045: F_{n}(\theta_{1},\dots,\theta_{n})=H_{n}Q_{n}(y_{1},\dots,y_{n})\prod_{i}\frac{r(\theta_{i})}{y_{i}}\prod_{i<j}\frac{f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j})}{y_{i}+y_{j}},\label{Ansatz}\end{equation}

1046:  where we separated a normalizing factor $H_{n}$ from the polynomials

1047: $Q_{n}$. The various $F_{n}$-s are related to each other by both

1048: the kinematical and the dynamical singularity equations, since the

1049: S-matrix is nontrivial and also has a $\varphi^{3}$ pole with $\Gamma=i2^{\frac{1}{2}}3^{\frac{1}{4}}$.

1050: In addition, these $F_{n}$-s also have to satisfy the equation coming

1051: from the residue of the pole at $\theta=i\pi/2$: \begin{equation}

1052: -i\mathop{\textrm{Res}}_{\theta=i\pi/2}F_{n+1}(\theta,\theta_{1},\dots,\theta_{n})=\frac{g_{\Phi}}{2}\left(1-\prod\limits _{j=1}^{n}S(i\frac{\pi}{2}-\theta_{j})\right)F_{n}(\theta_{1},\dots,\theta_{n}).\label{eq:ipipol}\end{equation}

1053:  Our strategy is to solve the recursion equations coming from the

1054: first two conditions first and check whether the solutions also satisfy

1055: the third requirement (\ref{eq:ipipol}). By choosing the normalizing

1056: factors $H_{n}$ and introducing the useful quantities $\beta_{k}$

1057: \begin{equation}

1058: H_{n}=N\left(\frac{i3^{\frac{1}{4}}}{2^{\frac{1}{2}}v(0)}\right)^{n}\qquad\beta_{k}(b)=2\cos\frac{\pi}{6}(b+k),\quad k\in\mathrm{Z},\label{eq:Hn_betak}\end{equation}

1059:  the overall normalization $N$ drops out and the recursion equations

1060: coming from the dynamical (resp. kinematical) singularities read\begin{eqnarray}

1061: Q_{2}(y_{+},y_{-}) & = & (y^{2}-\beta_{-3}^{2})Q_{1}(y),\nonumber \\

1062: Q_{n+2}(y_{+},y_{-},y_{1},\dots,y_{n}) & = & Q_{n+1}(y,y_{1},\dots,y_{n})\,(y^{2}-\beta_{-3}^{2})\prod_{i=1}^{n}(y+y_{i}),\quad n>0;\label{eq:rec2a}\end{eqnarray}

1063: \begin{equation}

1064: Q_{n+2}(-y,y,y_{1},\dots,y_{n})=Q_{n}(y_{1},\dots,y_{n})\,(y^{2}-\beta_{-1}^{2})(y^{2}-\beta_{1}^{2})\, P_{n},\label{eq:rec2b}\end{equation}

1065:  where \begin{equation}

1066: P_{n}=\frac{1}{2(y_{+}-y_{-})}\left[\prod_{i=1}^{n}(y_{i}-y_{-})(y_{i}+y_{+})-\prod_{i=1}^{n}(y_{i}+y_{-})(y_{i}-y_{+})\right],\label{eq:Pequ}\end{equation}

1067:  and \begin{equation}

1068: y_{+}=\omega x+\omega^{-1}x^{-1};\quad y_{-}=\omega^{-1}x+\omega x^{-1};\quad x=e^{\theta};\quad\omega=e^{i\frac{\pi}{3}},\quad y=x+x^{-1}.\label{eq:ypmeq}\end{equation}

1069: Next we present the minimal solution of these recursion equations

1070: up to $n=3$. The solution is called minimal, if the leading overall

1071: degrees of the $F_{n}$-s in all of the $y$ variables are the smallest

1072: possible ones. Of course the solution also depends on the input function

1073: $Q_{1}(y_{1})$. Since $F_{1}$ can have no pole at $\theta=i\pi/2$

1074: while $r(\theta)/y$ has one, $Q_{1}$ must be chosen to cancel this

1075: pole; the choice with the minimal degree is $Q_{1}(y_{1})=y_{1}=\sigma_{1}(y_{1})$.

1076: Using this as input, we find from (\ref{eq:rec2a} ,\ref{eq:rec2b})

1077: the \textsl{unique} solution \begin{eqnarray}

1078: Q_{1}(y_{1})=\sigma_{1},\qquad\qquad\qquad\quad Q_{2}(y_{1},y_{2})=\sigma_{1}(\sigma_{2}+3-\beta_{-3}^{2}),\nonumber \\

1079: Q_{3}(y_{1},y_{2},y_{3})=\sigma_{1}\left[\sigma_{1}(\sigma_{2}+\beta_{-1}^{2})(\sigma_{2}+\beta_{1}^{2})-(\sigma_{2}+3)(\sigma_{1}\sigma_{2}-\sigma_{3})\right].\label{eq:lysol1}\end{eqnarray}

1080:  The remarkable property of this solution is that it contains no free

1081: parameters. A simple counting of the various powers shows that the

1082: leading overall degree of $F_{1}$, $F_{2}$, and $F_{3}$ vanish. 

1083: 

1084: To check eq.(\ref{eq:ipipol}) we need the following relations following

1085: from the explicit solution (\ref{eq:lysol1}) and from the various

1086: identities among the $\beta_{k}$-s: \begin{eqnarray*}

1087: Q_{2}(0,y_{2}) & = & \sigma_{1}(y_{2})(3-\beta_{-3}^{2}),\\

1088: Q_{3}(0,y_{2},y_{3}) & = & \beta_{1}\beta_{-1}\sigma_{1}(y_{2},y_{3})Q_{2}(y_{2},y_{3})=(3-\beta_{-3}^{2})\sigma_{1}(y_{2},y_{3})Q_{2}(y_{2},y_{3}).\end{eqnarray*}

1089:  Indeed using them in eq.(\ref{eq:ipipol}) leads to a consistency

1090: condition on the ratio of the $H_{n}$-s: \[

1091: \frac{H_{n+1}}{H_{n}}r(i\frac{\pi}{2})(3-\beta_{-3}^{2})=-2i\sqrt{3}g_{\Phi},\qquad n=1,2.\]

1092:  Since \[

1093: r(i\frac{\pi}{2})=\frac{4u(i\frac{\pi}{2})}{i(\sqrt{3}-\beta_{-3})^{2}},\quad\textrm{and}\quad g_{\Phi}=i2(3)^{1/4}(2-\sqrt{3})^{1/2}\frac{\sqrt{3}+\beta_{-3}}{\sqrt{3}-\beta_{-3}},\]

1094:  the $b$ dependence cancels from the consistency condition and using

1095: the actual form of the $H_{n}$-s leads to \[

1096: \frac{u(i\frac{\pi}{2})}{\sqrt{2}v(0)}=\sqrt{3}\,(2-\sqrt{3})^{1/2},\]

1097:  which we checked numerically up to 7 digits.

1098: 

1099: To test these form factors numerically against the predictions of

1100: conformal field theory, we take the spectral representation of the

1101: boundary two-point function \begin{eqnarray*}

1102: \langle0\vert\mathcal{O}(t)\mathcal{O}(0)\vert0\rangle & = & \sum_{n=0}^{\infty}\int_{0}^{\infty}\frac{d\theta_{1}\dots d\theta_{n}}{n!(2\pi)^{n}}F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})^{+}F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})e^{-mt\sum_{i=1}^{n}\cosh\theta_{i}}\\

1103:  & = & \sum_{n=0}^{\infty}(-)^{n}\int_{0}^{\infty}\frac{d\theta_{1}\dots d\theta_{n}}{n!(2\pi)^{n}}\vert F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})\vert^{2}e^{-mt\sum_{i=1}^{n}\cosh\theta_{i}}\end{eqnarray*}

1104:  which we truncate to the first few terms in the boundary form factor

1105: expansion. Since the minimal solution of the form factor problem has

1106: the mildest UV behaviour it is natural to assume, that in the UV it

1107: corresponds to the boundary field $\varphi$. Therefore $\langle0\vert\mathcal{O}(t)\mathcal{O}(0)\vert0\rangle$

1108: obtained from the FF expansion must be compared to the short distance

1109: expansion: \[

1110: \langle0\vert m^{\frac{1}{5}}\varphi(t)m^{\frac{1}{5}}\varphi(0)\vert0\rangle=-(mt)^{\frac{2}{5}}+(mt)^{\frac{1}{5}}C_{\varphi\varphi}^{\varphi}\langle m^{\frac{1}{5}}\varphi\rangle+\dots\]

1111: where appropriate powers of $m$ were inserted to make the expression

1112: dimensionless and the fusion coefficient is \[

1113: C_{\varphi\varphi}^{\varphi}=-\sqrt{\frac{1+\sqrt{5}}{2}}\sqrt{\frac{\Gamma(\frac{1}{5})\Gamma(\frac{6}{5})}{\Gamma(\frac{3}{5})\Gamma(\frac{4}{5})}}\]

1114: while the ($b$-dependent, dimensionless) vacuum expectation value

1115: \[

1116: \langle m^{\frac{1}{5}}\varphi\rangle=-\frac{5}{6h_{crit}}\frac{\cos(\frac{\pi b}{6})}{\cos(\frac{\pi}{10}(2b+1))}\]

1117:  is given explicitly in \cite{DPTW1}. In analogy with the bulk case

1118: \cite{DPTW1pt} we choose the normalization factor $N$ in (\ref{eq:Hn_betak})

1119: as the vacuum expectation value of the boundary field \begin{equation}

1120: N=\langle m^{\frac{1}{5}}\varphi\rangle\label{eq:LYnorm}\end{equation}

1121:  With this choice the boundary form factor expansion gives \begin{eqnarray*}

1122: \langle0\vert\mathcal{O}(t)\mathcal{O}(0)\vert0\rangle & = & \vert F_{0}^{\mathcal{O}}\vert^{2}-\int_{0}^{\infty}\frac{d\theta}{2\pi}\vert F_{1}^{\mathcal{O}}\vert^{2}e^{-mt\cosh\theta}\\

1123:  &  & +\int_{0}^{\infty}\frac{d\theta_{1}d\theta_{2}}{2(2\pi)^{2}}\vert F_{2}^{\mathcal{O}}(\theta_{1},\theta_{2})\vert^{2}e^{-mt(\cosh\theta_{1}+\cosh\theta_{2})}\\

1124:  &  & -\int_{0}^{\infty}\frac{d\theta_{1}d\theta_{2}d\theta_{3}}{6(2\pi)^{3}}\vert F_{3}^{\mathcal{O}}(\theta_{1},\theta_{2},\theta_{3})\vert^{2}e^{-mt(\cosh\theta_{1}+\cosh\theta_{2}+\cosh\theta_{3})}+\dots\end{eqnarray*}

1125: The two expansions are compared on the next figure 

1126: 

1127: \begin{center}\includegraphics[%

1128:   width=15cm]{bly.eps}\end{center}

1129: 

1130: \noindent where $b=-1.05$ and the dimensionless correlation function

1131: is plotted against $mt$. The predicted UV behaviour is given by the

1132: continuous line and the numerically determined form factor expansion

1133: truncated at 1, 2 and 3 particle intermediate states is denoted by

1134: the symbols $\square$, $\times$ and $\circ$, respectively. 

1135: 

1136: We checked that the agreement between the form factor expansion truncated

1137: at three particles and the UV CFT prediction holds for various values

1138: of the parameter $b$: indeed as we change $b$ the two curves move

1139: together. The agreement above also confirms our choice (\ref{eq:LYnorm})

1140: for the normalization of the form factors. Based on all these we associate

1141: the boundary operator corresponding to the minimal solution of the

1142: form factor axioms to the one, that in the UV limit becomes the $\varphi$

1143: field of the boundary Lee-Yang model. 

1144: 

1145: 

1146: \subsubsection{Lee-Yang model with the $\textrm{1}$ boundary}

1147: 

1148: The 1PFF corresponding to the parameter free $R(\theta)_{1}$ is chosen

1149: as \[

1150: r_{1}(\theta)=i\sinh\theta\,\, u(\theta),\]

1151:  where $u(\theta)$ is the same as in the previous subsection. Note

1152: that $r_{1}$ also satisfies $r_{1}(\theta+i\pi)=r_{1}(\theta)^{*}$

1153: but its asymptotic behaviour $r_{1}\sim y^{2}$ at $y\rightarrow\infty$

1154: is different from that of the $r$ in the previous case. Since $R(\theta)_{1}$

1155: also has a pole at $\theta=i\pi/2$ we introduce a similar Ansatz

1156: as in (\ref{Ansatz}) \begin{equation}

1157: F_{n}(\theta_{1},\dots,\theta_{n})=H_{n}^{1}Q_{n}(y_{1},\dots,y_{n})\prod_{i}\frac{r_{1}(\theta_{i})}{y_{i}}\prod_{i<j}\frac{f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j})}{y_{i}+y_{j}},\label{Ansatz1}\end{equation}

1158: with $H_{n}^{1}=4^{n}H_{n}$ where $H_{n}$ is the same as in (\ref{eq:Hn_betak}).

1159: Then one finds the following recursion equations for the $Q_{n}$-s

1160: from the dynamical (resp. kinematical) singularity equations: \begin{eqnarray}

1161: Q_{2}(y_{+},y_{-}) & = & Q_{1}(y),\nonumber \\

1162: Q_{n+2}(y_{+},y_{-},y_{1},\dots,y_{n}) & = & Q_{n+1}(y,y_{1},\dots,y_{n})\,\prod_{i=1}^{n}(y+y_{i}),\quad n>0;\label{eq:rec21a}\end{eqnarray}

1163: \begin{equation}

1164: Q_{n+2}(-y,y,y_{1},\dots,y_{n})=Q_{n}(y_{1},\dots,y_{n})\, P_{n},\label{eq:rec21b}\end{equation}

1165:  where the various symbols are the same as in eq.(\ref{eq:rec2a},\ref{eq:rec2b}).

1166: Up to $n=4$ the \textsl{unique} minimal solution of these recursion

1167: equations with the input $Q_{1}(y_{1})=\sigma_{1}$ is \[

1168: Q_{1}(y_{1})=\sigma_{1},\quad Q_{2}(y_{1},y_{2})\sim\sigma_{1},\quad Q_{3}(y_{1},y_{2},y_{3})\sim\sigma_{1}^{2},\quad Q_{4}(y_{1},y_{2},y_{3},y_{4})\sim\sigma_{1}^{2}(\sigma_{2}+3).\]

1169:  It is easy to show that the leading overall degree of the first four

1170: form factors $F_{1},\dots,F_{4}$ is two. This indicates that the

1171: operator that corresponds to this set is different from the one encountered

1172: in the case of the perturbed $\Phi$ boundary. Therefore in the conformal

1173: limit this operator is different from the $\varphi$ field and this

1174: is in accord with the fact that only the identity operator and its

1175: descendents can live on the conformal boundary condition $1$. Based

1176: on the asymptotics of the form factors for large $\theta$ the corresponding

1177: operator has ultraviolet dimension $2$ and can be identified with

1178: the unique such operator in the conformal vacuum module which is the

1179: $L_{-2}$ descendent of the identity. \emph{}This identification is

1180: further confirmed by comparing the numerically obtained truncated

1181: form factor expansion to the conformal two-point function. 

1182: 

1183: Since \[

1184: r_{1}(i\frac{\pi}{2})=-u(i\frac{\pi}{2}),\quad\textrm{and}\quad g_{1}=-i2(3)^{1/4}(2-\sqrt{3})^{1/2},\]

1185:  (see also \cite{DTWbct}) one can readily show that these four form

1186: factors also satisfy the equation coming from the residue of the pole

1187: at $\theta=i\pi/2$. 

1188: 

1189: 

1190: \subsection{The boundary sinh-Gordon model}

1191: 

1192: The sinh-Gordon theory in the bulk is defined by the Lagrangian%

1193: \footnote{Note that the parameter $b$ is used here with a different meaning

1194: compared to the former case of the boundary Lee-Yang model. %

1195: }:\[

1196: \mathcal{L}=\frac{1}{2}(\partial_{\mu}\Phi)^{2}-\frac{m^{2}}{b^{2}}(\cosh b\Phi-1)\]

1197: It can be considered as the analytic continuation of the sine-Gordon

1198: model for imaginary coupling $\beta=ib$. The S-matrix of the model

1199: is \[

1200: S(\theta)=-\left(1+\frac{B}{2}\right)\left(-\frac{B}{2}\right)=-\left[-\frac{B}{2}\right]\qquad;\quad B=\frac{2b^{2}}{8\pi+b^{2}}\]

1201: The minimal bulk two particle form factor belonging to this S-matrix

1202: is \cite{FMS} \[

1203: f(\theta)=\mathcal{N}\exp\left[8\int_{0}^{\infty}\frac{dx}{x}\sin^{2}\left(\frac{x(i\pi-\theta)}{2\pi}\right)\frac{\sinh\frac{xB}{4}\sinh(1-\frac{B}{2})\frac{x}{2}\sinh\frac{x}{2}}{\sinh^{2}x}\right],\]

1204:  and it satisfies \begin{equation}

1205: f(\theta)f(\theta+i\pi)=\frac{\sinh\theta}{\sinh\theta+i\sin\frac{\pi B}{2}}\,\,.\label{eq:ffip}\end{equation}

1206: The sinh-Gordon theory can be restricted to the negative half-line,

1207: but the integrability is maintained only by imposing either the Dirichlet\[

1208: \Phi(0,t)=\Phi_{0}^{D}\]

1209: or the two parameter family of perturbed Neumann\[

1210: V_{B}(\Phi(0,t))=M_{0}\cosh\left(\frac{b}{2}(\Phi(0,t)-\Phi_{0})\right)-M_{0}\]

1211: boundary conditions. The latter interpolates between the Neumann and

1212: the Dirichlet boundary conditions, since for $M_{0}=0$ we recover

1213: the Neumann, while for $M_{0}\to\infty$ the Dirichlet boundary condition

1214: with $\Phi_{0}^{D}=\Phi_{0}$. The reflection factor which depends

1215: on two continuous parameters can be written as \[

1216: R(\theta)=R_{0}(\theta)R(E,F,\theta)=\left(\frac{1}{2}\right)\left(\frac{1}{2}+\frac{B}{4}\right)\left(-1-\frac{B}{4}\right)\left[\frac{E-1}{2}\right]\left[\frac{F-1}{2}\right]\]

1217:  in terms of the parameterization used in \cite{shGEF}. They are

1218: related to the parameters of the Lagrangian as \begin{eqnarray*}

1219: \cos\frac{E}{16}(b^{2}+8\pi)\cos\frac{F}{16}(b^{2}+8\pi) & = & \frac{M_{0}}{M_{crit}}\cosh\frac{b\Phi_{0}}{2}\\

1220: \sin\frac{E}{16}(b^{2}+8\pi)\sin\frac{F}{16}(b^{2}+8\pi) & =- & \frac{M_{0}}{M_{crit}}\sinh\frac{b\Phi_{0}}{2}\end{eqnarray*}

1221: where $M_{crit}=m\sqrt{\frac{2}{b^{2}\sinh(b^{2}/8)}}$. Note that

1222: for generic values of the parameters ($E\ne0$, $F\ne0$) this reflection

1223: factor has a pole at $\theta=i\pi/2$ coming from the $\left(\frac{1}{2}\right)$

1224: factor. Imposing Dirichlet boundary condition instead of the general

1225: one corresponds to removing the $F$ dependent factor from $R(\theta)$.

1226: Then the remaining parameter $E$ is related to the $\Phi_{0}$ boundary

1227: value of the sinh-Gordon field as $E=i8b\Phi_{0}/(b^{2}+8\pi)$.

1228: 

1229: 

1230: \subsubsection{Sinh-Gordon model with $\Phi_{0}=0$ Dirichlet b.c.}

1231: 

1232: This case is interesting because $E=0$ implies that the pole at $\theta=i\pi/2$

1233: is \textsl{absent} in this case. Therefore the equation coming from

1234: the residue of this pole is also absent and the form factors are less

1235: restricted. The 1PFF corresponding to the reflection amplitude on

1236: the $E=0$ Dirichlet boundary is \[

1237: r_{0}(\theta)=\frac{\sinh\theta}{\sinh\theta+i}u(\theta,B),\]

1238:  where \[

1239: u(\theta,B)=\exp\left[-2\int_{0}^{\infty}\frac{dx}{x}\left[\cos(\frac{i\pi}{2}-\theta)\frac{x}{\pi}-1\right]\frac{\cosh\frac{x}{2}}{\sinh^{2}x}\left(\sinh\frac{xB}{4}+\sinh(1-\frac{B}{2})\frac{x}{2}+\sinh\frac{x}{2}\right)\right].\]

1240:  Note that $r_{0}\sim y$ at $y\rightarrow\infty$ and has no pole

1241: at $\theta=i\pi/2$. At $B=0$ - which corresponds to a free theory

1242: - $u(\theta,0)$ can be integrated explicitly yielding $r_{0}(\theta)|_{B=0}=(-i\sinh\theta)/2$;

1243: and this, apart from a trivial normalizational phase coincides with

1244: the 1PFF for a free scalar with Dirichlet b.c. (As discussed in Section

1245: 3.1.1 in this case $\partial_{x}\Phi(0,0)$ is the operator having

1246: one particle matrix element only). 

1247: 

1248: We write the $n$ particle form factors in the general form: \[

1249: F_{n}(\theta_{1},\dots,\theta_{n})=H_{n}Q_{n}(y_{1},\dots,y_{n})\prod_{i}r_{0}(\theta_{i})\prod_{i<j}\frac{f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j})}{y_{i}+y_{j}}.\]

1250:  Since there is no self fusing pole in the $S$ matrix of the sinh-Gordon

1251: model, the $F_{n}$-s are related only by the kinematical singularity

1252: equations. Choosing the ratio of the $H_{n}$-s appropriately the

1253: recursion equations originating from here take the form: \[

1254: Q_{n+2}(-y,y,y_{1},\dots,y_{n})=-Q_{n}(y_{1},\dots,y_{n})\, P_{n},\]

1255:  where $P_{n}$ is given by eq.(\ref{eq:Pequ},\ref{eq:ypmeq}) with

1256: $\omega=e^{i\pi\frac{B}{2}}$.

1257: 

1258: As $r_{0}$ has no pole at $\theta=i\pi/2$ , one can have a minimal

1259: solution of this recursion equations starting with $Q_{1}=1$ which

1260: has non vanishing form factors for \emph{odd} particle numbers. We

1261: calculated up to $n=5$ and found that the solution is uniquely given

1262: by \[

1263: Q_{1}(y_{1})=1,\qquad Q_{3}(y_{1},y_{2},y_{3})=-\sigma_{1},\]

1264:  \[

1265: Q_{5}(y_{1},\dots,y_{5})=\sigma_{1}[\sigma_{3}\sigma_{2}-(\omega+\omega^{-1})^{2}\sigma_{5}+(\omega-\omega^{-1})^{4}\sigma_{1}-(\omega-\omega^{-1})^{2}(\sigma_{3}+\sigma_{1}\sigma_{2})],\]

1266:  with all the $F_{1}$, $F_{3}$ and $F_{5}$ form factors having

1267: leading overall degree one. There is a unique local operator with

1268: this property, namely $\partial_{x}\Phi$. 

1269: 

1270: Of course one can also find non vanishing solutions with \emph{even}

1271: particle numbers also starting with a non trivial $Q_{2}$. Since

1272: $F_{2}$ can have no kinematical singularity, the minimal choice is

1273: $Q_{2}(y_{1},y_{2})=\sigma_{1}$. With this input we obtained again

1274: a unique solution \[

1275: Q_{2}(y_{1},y_{2})=\sigma_{1},\qquad Q_{4}(y_{1},\dots,y_{4})=\sigma_{1}^{2}(\sigma_{2}-(\omega-\omega^{-1})^{2}),\]

1276:  where both the $F_{2}$ and the $F_{4}$ have leading overall degree

1277: two. 

1278: 

1279: For $\Phi_{0}=0$ ($E=0$) the $\Phi\rightarrow-\Phi$ reflection

1280: symmetry of the bulk sinh-Gordon model survives also in the boundary

1281: theory. Therefore the boundary operators can be classified as even

1282: or odd ones, having only non-vanishing even or odd particle form factors,

1283: respectively. Thus the second form factor family can be identified

1284: with the operator $(\partial_{x}\Phi)^{2}$. 

1285: 

1286: 

1287: \subsubsection{Sinh-Gordon model with $\Phi_{0}\ne0$ Dirichlet b.c.}

1288: 

1289: For $\Phi_{0}\ne0$ ($E\ne0$) the reflection factor acquires a pole

1290: at $\theta=i\pi/2$ due to the fact that the field has a nontrivial

1291: vacuum configuration. At the same time the reflection symmetry of

1292: the bulk sinh-Gordon model is violated in the boundary theory. Therefore

1293: the boundary operators cannot be classified into representations of

1294: this symmetry, and the equation coming from the residue of the pole

1295: at $\theta=i\pi/2$ connects the form factors with even and odd particle

1296: numbers. Note that now this equation plays an essential role as it

1297: is the only one that relates these two sets of form factors to each

1298: other.

1299: 

1300: The 1PFF corresponding to the reflection amplitude on the $E\ne0$

1301: Dirichlet boundary is \[

1302: r_{E}(\theta)=\frac{\sinh\theta}{\sinh\theta-i\sin\gamma}u(\theta,B)\quad,\qquad\qquad\gamma=\frac{\pi}{2}(E-1),\]

1303:  where $u(\theta,B)$ is the same as in the previous case. \emph{}Note

1304: that $r_{E}\sim y$ at $y\rightarrow\infty$. \emph{}

1305: 

1306: Writing the $n$ particle form factors in the general form \[

1307: F_{n}(\theta_{1},\dots,\theta_{n})=H_{n}Q_{n}(y_{1},\dots,y_{n})\prod_{i}\frac{r_{E}(\theta_{i})}{y_{i}}\prod_{i<j}\frac{f(\theta_{i}-\theta_{j})f(\theta_{i}+\theta_{j})}{y_{i}+y_{j}},\]

1308: and choosing the ratio of the $H_{n}$-s appropriately the recursion

1309: equations originating from the kinematical singularity equation take

1310: the form: \begin{equation}

1311: Q_{n+2}(-y,y,y_{1},\dots,y_{n})=(y^{2}-4\cos^{2}\gamma)Q_{n}(y_{1},\dots,y_{n})\, P_{n},\label{eq:shGkin}\end{equation}

1312:  where $P_{n}$ is given by eq.(\ref{eq:Pequ},\ref{eq:ypmeq}) with

1313: $\omega=e^{i\pi\frac{B}{2}}$. Next we show how the equation coming

1314: from the residue of the pole at $\theta=i\pi/2$: \begin{equation}

1315: -i\mathop{\textrm{Res}}_{\theta=i\pi/2}F_{n+1}(\theta,\theta_{1},\dots,\theta_{n})=\frac{g_{E}}{2}\left(1-\prod\limits _{j=1}^{n}S(i\frac{\pi}{2}-\theta_{j})\right)F_{n}(\theta_{1},\dots,\theta_{n}),\label{eq:ipipol2}\end{equation}

1316:  helps to eliminate the arbitrariness in the minimal solution of the

1317: recursion equations.

1318: 

1319: $Q_{1}(y_{1})=\sigma_{1}$ is the minimal choice that guarantees that

1320: $F_{1}$ has no pole at $\theta=i\pi/2$. Using this in the recursion

1321: equation (\ref{eq:shGkin}) gives that the most general $Q_{3}$ has

1322: the form: \[

1323: Q_{3}(y_{1},y_{2},y_{3})=-\sigma_{1}^{2}(\sigma_{2}+4\cos^{2}\gamma)+(A+B\sigma_{1})(\sigma_{1}\sigma_{2}-\sigma_{3}),\]

1324:  where $A$ and $B$ are arbitrary constants. Eq.(\ref{eq:ipipol2})

1325: leads to the following relation between $Q_{3}$ and $Q_{2}$: \[

1326: H_{3}r_{E}(i\frac{\pi}{2})Q_{3}(0,y_{2},y_{3})=g_{E}2\sin\frac{B\pi}{2}H_{2}\sigma_{1}(y_{2},y_{3})Q_{2}(y_{2},y_{3}).\]

1327:  Since \[

1328: Q_{3}(0,y_{2},y_{3})=\sigma_{1}\left\{ -\sigma_{1}(\sigma_{2}+4\cos^{2}\gamma)+(A+B\sigma_{1})\sigma_{2}\right\} ,\]

1329:  the expression in the curly bracket should be proportional to $Q_{2}$.

1330: This observation fixes the values of $A$ and $B$: $Q_{2}$ has to

1331: be proportional to $\sigma_{1}$ to guarantee that $F_{2}$ has no

1332: kinematical singularity and this requirement is met only if $A=0$,

1333: while $F_{2}$ has a leading degree not exceeding that of $F_{1}$

1334: and $F_{3}$ if $B=1$. Thus with these two requirements one obtains

1335: a parameter free solution starting with $Q_{1}=\sigma_{1}$; up to

1336: $n=4$ it has the form: \[

1337: Q_{2}(y_{1},y_{2})\sim-4\cos^{2}\gamma\sigma_{1},\quad Q_{3}(y_{1},y_{2},y_{3})=-\sigma_{1}\sigma_{3}-4\cos^{2}\gamma\sigma_{1}^{2},\]

1338:  \[

1339: Q_{4}(y_{1},\dots,y_{4})\sim-4\cos^{2}\gamma(\sigma_{1}\sigma_{3}+4\cos^{2}\gamma\sigma_{1}^{2})(\sigma_{2}+4\sin^{2}\frac{\pi B}{2}).\]

1340:  Note that both $Q_{2}$ and $Q_{4}$ vanish for $\gamma=-\pi/2$

1341: ($E=0$). Furthermore for $E=0$ one also has $Q_{1}/y_{1}=1$, and

1342: $Q_{3}/(y_{1}y_{2}y_{3})=-\sigma_{1}$, thus the solution goes over

1343: smoothly into the one with $E=0$. Since \[

1344: r_{E}(i\frac{\pi}{2})=\frac{1}{1-\sin\gamma},\qquad{\textrm{and}}\qquad g_{E}=\frac{2(1+\cos\frac{\pi B}{4}+\sin\frac{\pi B}{4})}{\sqrt{\sin\frac{\pi B}{2}}}\frac{\cos\gamma}{1-\sin\gamma}\,,\]

1345:  the $\gamma$ dependence cancels from the ratios of $H_{3}/H_{1}$

1346: and $H_{4}/H_{2}$ when we use eq.(\ref{eq:ipipol2}) for $n=1,2,3$:

1347: \[

1348: -i\frac{H_{4}}{H_{2}}=-i\frac{H_{3}}{H_{1}}=\left(1+\cos\frac{\pi B}{4}+\sin\frac{\pi B}{4}\right)^{2}4\sin\frac{\pi B}{2}.\]

1349: In the $b\to0$ limit these ratios vanish, therefore the higher form

1350: factors decouple in accord with the fact that the kinematical singularity

1351: axiom becomes trivial for the free field theory. 

1352: 

1353: 

1354: \section{Conclusion}

1355: 

1356: In this paper we treated the form factor bootstrap for boundary operators

1357: in integrable boundary quantum field theory. Although there have been

1358: earlier treatment of form factors for specific (mainly lattice) models

1359: \cite{XXZ,XXZGen,SGff}, none of these has actually given a complete

1360: formulation similar to the axiomatic approach by Smirnov for the bulk

1361: case \cite{Smirnov}. The present work initiates an extension of this

1362: axiomatic program to boundary fields. 

1363: 

1364: We have given a complete axiomatization of the properties of boundary

1365: form factors, derived from first principles of quantum field theory

1366: (unitarity and the boundary extension of the LSZ reduction formulae).

1367: In particular, the axiom describing boundary kinematic singularities

1368: is an entirely new result of this paper, as this has never been treated

1369: before in any previous study. We have shown that these axioms are

1370: consistent with many known aspects of integrable boundary field theory.

1371: In particular, the relation between the residue of the reflection

1372: factor at $i\pi/2$ and the one-particle contribution to the boundary

1373: state, noted previously in the context of finite size effects, was

1374: confirmed once more as a necessary condition for the consistency between

1375: the boundary and the bulk kinematical axiom (the only exception to

1376: this relation is when the bulk is free, but then the two axioms are

1377: trivial). Therefore it seems that this particular relation is a consequence

1378: of integrability and the existence of a nontrivial bulk scattering

1379: matrix.

1380: 

1381: We then proceeded to give a systematic method to solve the boundary

1382: form factor axioms for the case of diagonal scattering. The solution

1383: is a natural generalization of the bulk case, but necessitates the

1384: introduction of a minimal boundary form factor function in addition

1385: to the already known minimal bulk form factor. The periodicity, permutation

1386: and reflection axioms can then be solved by a general Ansatz, and

1387: the residue axioms can be recast as recursion relations for certain

1388: \emph{polynomial} functions which characterize the form factor solution

1389: completely.

1390: 

1391: In particular, we treated the case of the free boson and the free

1392: fermion (noncritical Ising model with boundary magnetic field), where

1393: the polynomial solutions of the form factor axioms were shown to be

1394: identical to the explicit solutions obtained from the field theory,

1395: and it was also shown that the polynomial solutions of the form factor

1396: bootstrap match the full boundary operator content expected from the

1397: Lagrangian approach. 

1398: 

1399: As example for the interacting case, we first treated the Lee-Yang

1400: model, where the boundary kinematical singularity axiom makes its

1401: first appearance, and it is very important in order to distinguish

1402: between boundary conditions that have different conformal limits.

1403: We have also computed the spectral expansion of the two-point correlation

1404: function for the operator with the lowest conformal dimension and

1405: have shown that it matches perfectly with the ultraviolet expansion

1406: of the same correlation function obtained from boundary conformal

1407: field theory.

1408: 

1409: Our second interacting example is the sinh-Gordon model, with Dirichlet

1410: boundary condition (an extension to the general case is in principle

1411: straightforward, but we decided to treat only the Dirichlet case to

1412: keep it short and simple). The boundary conditions with zero and with

1413: nonzero value of the field on the boundary are differentiated again

1414: by the boundary kinematical axiom, and we have shown that the results

1415: of the boundary form factor bootstrap fit perfectly well with expectations

1416: from the Lagrangian approach.

1417: 

1418: An open question is to find and classify non-minimal solutions of

1419: the form factor equations and interpret them in terms of the local

1420: boundary operator algebra of the underlying field theory, extending

1421: the method presented for the bulk sinh-Gordon model in \cite{koubek_mussardo}.

1422: In particular it is interesting to find out whether the counting of

1423: operators in the conformal limit can be matched with the full set

1424: of solutions in the interacting case. 

1425: 

1426: It is obvious that the results presented in this paper can be applied

1427: directly to any integrable boundary quantum field theory for which

1428: the factorized scattering theory is known, and that they formulate

1429: a well-defined program to determine form factors and correlation functions

1430: of boundary operators, similar to the approach used in the bulk case.

1431: We have also shown how to solve the axioms for theories with diagonal

1432: bulk and boundary scattering.

1433: 

1434: It is an interesting problem to extend these results to the case of

1435: nondiagonal scattering (with boundary sine-Gordon theory as the most

1436: prominent example). The extension of the axioms is straightforward:

1437: they must be decorated by multiplet indices, just like in the bulk

1438: case, although here we avoided to give this extension explicitly to

1439: keep the exposition simple. However, solving them will probably encounter

1440: much more difficulties, and just as in the bulk, new methods must

1441: be devised for the task, like the boundary extension of the Lukyanov

1442: free field representation in \cite{SGff}.

1443: 

1444: The comparison to the Lagrangian and perturbed conformal field theory

1445: description would be greatly facilitated by establishing sum rules

1446: for the spectral representation of the boundary correlators, similar

1447: to the $c$-theorem \cite{c-th} and $\Delta$-theorem \cite{delta-th}

1448: in the bulk case, and is one of the most important problems left open

1449: by the present work. Another promising open direction is to consider

1450: possible applications of boundary form factors and correlation functions

1451: in the area of boundary quantum field theory and condensed matter.

1452: 

1453: 

1454: \subsection*{Acknowledgments}

1455: 

1456: We are grateful to F.A. Smirnov and G. Watts for useful discussions.

1457: This research was partially supported by the EC network {}``EUCLID'',

1458: contract number HPRN-CT-2002-00325, and Hungarian research funds OTKA

1459: T043582, K60040 and TS044839. GT was also supported by a Bolyai J\'anos

1460: research scholarship.

1461: 

1462: \appendix

1463: 

1464: \section{Heuristic derivation of the FF axioms}

1465: 

1466: We present some heuristic arguments, along the lines of \cite{Smirnov},

1467: for the derivation of boundary form factor axioms using the boundary

1468: reduction formula \cite{BBT}. 

1469: 

1470: We analyze the analyticity properties of the form factor\[

1471: F_{n}^{\mathcal{O}}:=F_{n}^{\mathcal{O}}(\theta_{1},\theta_{2},\dots,\theta_{n})=\langle0\vert\mathcal{O}(0)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}\]

1472:  as a function of the variable $\theta_{1}$. 

1473: 

1474: We follow the notations of \cite{BBT}: The asymptotic creation/annihilation

1475: operators can be expressed in terms of the free asymptotic fields

1476: as \begin{eqnarray}

1477: a_{in}(\theta) & = & 2i\int_{-\infty}^{0}dx\cos(k(\theta)x)e^{i\omega(\theta)t}{\mathop{\partial}^{\leftrightarrow}}_{t}\Phi_{in}(x,t)\label{aa+}\\

1478: a_{in}^{+}(\theta) & = & -2i\int_{-\infty}^{0}dx\cos(k(\theta)x)e^{-i\omega(\theta)t}{\mathop{\partial}^{\leftrightarrow}}_{t}\Phi_{in}(x,t)\;.\nonumber \end{eqnarray}

1479:  where $k(\theta)=m\sinh\theta$ and $\omega(\theta)=m\cosh\theta$.

1480: The \emph{in} state is a free state and we have \begin{equation}

1481: \langle0\vert\mathcal{O}(0)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}=\langle0\vert\mathcal{O}(0)a_{in}^{+}(\theta_{1})\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\label{matel}\end{equation}

1482: We use (\ref{aa+}) together with \[

1483: \mathcal{O}(0)\Phi_{in}(x,t)=[\mathcal{O}(0),\Phi_{in}(x,t)]+\Phi_{in}(x,t)\mathcal{O}(0)\]

1484: to obtain \begin{equation}

1485: F_{n}^{\mathcal{O}}=\textrm{disc.}-2i\int_{-\infty}^{0}dx\cos(k(\theta_{1})x)e^{-i\omega(\theta_{1})t}{\mathop{\partial}^{\leftrightarrow}}_{t}\,\langle0\vert[\mathcal{O}(0),\Phi_{in}(x,t)]\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\;.\label{disc1}\end{equation}

1486: where the disconnected part is \[

1487: \textrm{disc.}=\langle0\vert a_{in}^{+}(\theta_{1})\mathcal{O}(0)\vert\theta_{2},\dots,\theta_{n}\rangle_{in}=\langle0\vert a_{in}^{+}(\theta_{1})\vert0\rangle F_{n-1}^{\mathcal{O}}(\theta_{2},\dots,\theta_{n})\]

1488: Note that in theories with nonzero vacuum expectation values of the

1489: field $\Phi$ the matrix element $\langle0\vert a_{in}^{+}(\theta_{1})\vert0\rangle$

1490: is nonzero and can be written as \[

1491: \langle0\vert a_{in}^{+}(\theta_{1})\vert0\rangle=\frac{g}{2}2\pi\delta(\theta_{1}-\frac{i\pi}{2})\]

1492: which corresponds to the one particle term in the boundary state in

1493: the crossed channel \cite{GZ}. It was conjectured in \cite{DPTW1pt}

1494: and later confirmed using TBA arguments \cite{BLusch} that the one

1495: particle contribution to the boundary state has a coefficient equal

1496: to $\frac{g}{2}$ rather than $g$ as suggested in \cite{GZ}. In

1497: the channel we use here this translates directly into the equation

1498: above. 

1499: 

1500: Supposing that the \emph{in} field can be expressed in terms of the

1501: interacting field as $\Phi(x,t)\to Z^{1/2}\Phi_{in}(x,t)$ for $t\to-\infty$

1502: and that $[\mathcal{O}(0),\Phi(x,0)]=0$, the connected part can be

1503: written in the form \[

1504: \textrm{conn.}=iZ^{-1/2}2\int_{-\infty}^{0}dx\int_{-\infty}^{0}dt\partial_{t}\biggl\{\cos(k(\theta_{1})x)e^{-i\omega(\theta_{1})t}{\mathop{\partial}^{\leftrightarrow}}_{t}\,\langle0\vert[\mathcal{O}(0),\Phi(x,t)]\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\biggr\}\]

1505: Performing the usual partial integration while taking care of the

1506: surface term we obtain\begin{equation}

1507: \textrm{conn.}=iZ^{-1/2}2\int d^{2}xe^{-i\omega(\theta_{1})t}\cos(k(\theta_{1})x)\Theta(-t)\langle0\vert[\mathcal{O}(0),J(x,t)]\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\label{redrin}\end{equation}

1508: where $J(x,t)=\{\square+m^{2}+\delta(x)\partial_{x}\}\Phi(x,t)$ and

1509: the integration goes over the entire spacetime. The range of the integration

1510: is the interior of the past light cone due to the presence of $\Theta(-t)$

1511: and of the vanishing of $[\mathcal{O}(0),J(x,t)]$ on space-like intervals.

1512: The analytic properties of the integral are determined by the exponent

1513: for large negative times. The exponent decreases if $\Im m(\omega(\theta_{1}))>0$

1514: thus the \emph{in} form factor ($\theta_{1}>\theta_{2}>\dots>\theta_{n}>0$)

1515: allows an analytical continuation into the domain:\[

1516: 0<\Im m(\theta_{1})<\pi\]

1517: 

1518: 

1519: Repeating the same procedure for the \emph{out} matrix elements \[

1520: F_{n}^{\mathcal{O}}(-\theta_{1},-\theta_{2},\dots,-\theta_{n})=\langle0\vert\mathcal{O}(0)\vert-\theta_{1},-\theta_{2},\dots,-\theta_{n}\rangle_{out}\]

1521: we obtain the domain of analytical continuation: $0<\Im m(-\theta_{1})<\pi$. 

1522: 

1523: To derive the crossing relation we consider the following matrix element\[

1524: F_{1n-1}^{\mathcal{O}}:=F_{1n-1}^{\mathcal{O}}(\theta_{1}\vert\theta_{2},\dots,\theta_{n})=\,_{in}\langle\theta_{1}\vert\mathcal{O}(0)\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\]

1525: Applying the reduction formula to the particle with rapidity $\theta_{1}$

1526: (\ref{aa+}) we obtain\[

1527: F_{1n-1}^{\mathcal{O}}=\textrm{disc}-2i\int_{-\infty}^{0}dx\cos(k(\theta_{1})x)e^{i\omega(\theta_{1})t}{\mathop{\partial}^{\leftrightarrow}}_{t}\,\langle0\vert[\mathcal{O}(0),\Phi_{in}(x,t)]\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\]

1528: where the disconnected part (supposing $\theta_{1}\geq\theta_{2}$)

1529: is \[

1530: \textrm{disc.}=\langle0\vert\mathcal{O}(0)a_{in}^{+}(\theta_{1})\vert\theta_{2},\dots,\theta_{n}\rangle_{in}=\,_{in}\langle\theta_{1}\vert\theta_{2}\rangle_{in}F_{n-2}^{\mathcal{O}}(\theta_{3},\dots,\theta_{n})\]

1531: Performing a partial integration as before the result for the connected

1532: component is \begin{equation}

1533: \textrm{conn.}=iZ^{-1/2}2\int d^{2}xe^{i\omega(\theta_{1})t}\cos(k(\theta_{1})x)\Theta(-t)\langle0\vert[\mathcal{O}(0),J(x,t)]\vert\theta_{2},\dots,\theta_{n}\rangle_{in}\label{redlin}\end{equation}

1534:  which has an analytic continuation for \[

1535: -\pi<\Im m(\theta_{1})<0\]

1536:  Comparing (\ref{redrin}) with (\ref{redlin}) and using that $m\cosh(\theta_{1}+i\pi)=-m\cosh\theta_{1}$

1537: the crossing relation (\ref{eq:crossing}) is proved. Similar result

1538: can be obtained for an \emph{out} state and the $-\theta_{1}<-\theta_{2}<\dots<-\theta_{n}<0$

1539: range of the parameters. 

1540: 

1541: The reflection property (Axiom II) can be shown by considering \[

1542: F_{n}^{\mathcal{O}}(\theta_{1},\theta_{2},\dots,\theta_{n})=\langle0\vert\mathcal{O}(0)\vert\theta_{1},\theta_{2},\dots,\theta_{n}\rangle_{in}\]

1543:  and crossing all particles except the one with rapidity $\theta_{n}$

1544: to the left. Now inserting a complete set of \emph{out} states we

1545: have \[

1546: _{in}\langle\dots\vert\mathcal{O}(0)\vert\theta_{n}\rangle_{in}=\sum_{out}\,_{in}\langle\dots\vert A(0)\vert n\rangle_{out}\,_{out}\langle n\vert\theta_{n}\rangle_{in}\]

1547: where only the first two terms are nonzero: \[

1548: _{in}\langle\dots\vert\mathcal{O}(0)\vert\theta_{n}\rangle_{in}=\,_{in}\langle\dots\vert A(0)\vert0\rangle\langle0\vert\theta_{n}\rangle_{in}+\sum_{\theta}\langle\dots\vert A(0)\vert\theta\rangle_{out}\,_{out}\langle\theta\vert\theta_{n}\rangle_{in}\]

1549: The connected part gives the required $R$ factor while the disconnected

1550: one combined with the disconnected part in (\ref{disc1}) and the

1551: permutation property gives the boundary kinematical singularity. 

1552: 

1553: The permutation property in the bulk case is usually derived from

1554: very similar argumentation we used above for showing the reflection

1555: property. Note, however, that the same result can be obtained from

1556: the analysis of the singularity structure of the Green functions:

1557: the part, which is responsible for the discontinuity in the form factor

1558: by changing two neighboring rapidities, is related to the bulk $S$-matrix.

1559: The permutation property in the boundary case (Axiom I) can be derived

1560: only from the second approach, namely from a detailed investigation

1561: of the singularity structure of the Green functions. By extending

1562: the result on the two point function in \cite{BBT} one can show that

1563: multi point functions have momentum preserving parts identical to

1564: their bulk counterparts and exactly these parts contribute only, when

1565: two neighboring (both positive) rapidities are changed, and cause

1566: the same discontinuity in the form factor we met in the bulk case.

1567: 

1568: The kinematical singularity equation (Axiom IV) can be obtained (using

1569: the permutation and reflection axioms) from the analysis of the disconnected

1570: components in the crossing relations as obtained for the \emph{in}

1571: and for the \emph{out} states: \[

1572: F_{1n-1}^{\mathcal{O}}(\pm\theta_{1}\vert\pm\theta_{2},\dots,\pm\theta_{n})=F_{n}^{\mathcal{O}}(i\pi\pm\theta_{1},\pm\theta_{2},\dots,\pm\theta_{n})+2\pi\delta(\theta_{1}-\theta_{2})F_{n-2}^{\mathcal{O}}(\pm\theta_{3},\dots,\pm\theta_{n})\]

1573: 

1574: 

1575: Although our derivation of the boundary form factor axioms is heuristic

1576: to some extent we expect that the formulation of the same ideas in

1577: a rigorously defined quantum field theoretical framework would lead

1578: to the proper and mathematically founded derivation (but note that

1579: this has not been performed yet in the bulk case either). 

1580: 

1581: 

1582: \section{Formal derivation of the FF axioms }

1583: 

1584: Here we show how the boundary form factor axioms can be formally derived

1585: from a boundary analogue of the Faddeev-Zamolodchikov algebra. 

1586: 

1587: In the bulk case one introduces creation $Z^{*}(\theta)$ and annihilation

1588: $Z(\theta)$ operators corresponding to asymptotic states. They are

1589: defined for real rapidities $\theta\in R$ and satisfy the following

1590: defining relations \begin{eqnarray}

1591: Z^{*}(\theta_{1})Z^{*}(\theta_{2}) & = & S(\theta_{1}-\theta_{2})Z^{*}(\theta_{2})Z^{*}(\theta_{1})\nonumber \\

1592: Z(\theta_{1})Z(\theta_{2}) & = & S(\theta_{1}-\theta_{2})Z(\theta_{2})Z(\theta_{1})\nonumber \\

1593: Z(\theta_{1})Z^{*}(\theta_{2}) & = & S(\theta_{2}-\theta_{1})Z^{*}(\theta_{2})Z(\theta_{1})+2\pi\delta(\theta_{1}-\theta_{2})\label{eq:ZFDEF}\end{eqnarray}

1594:  Extending $Z,Z^{*}$ to imaginary rapidities (treating $\theta$

1595: as a $2\pi i$ periodic complex variable) we encounter singularities

1596: in their products at $\theta_{1}=\theta_{2}\pm i\pi$ with residues\[

1597: -i\mathop{\textrm{Res}}_{\theta_{1}=\theta_{2}+i\pi}Z^{*}(\theta_{1})Z^{*}(\theta_{2})=1\]

1598: and \[

1599: -i\mathop{\textrm{Res}}_{\theta_{1}=\theta_{2}+i\pi}Z(\theta_{1})Z(\theta_{2})=1\]

1600: These can be formulated by postulating the crossing property \[

1601: Z(\theta)=Z^{*}(\theta+i\pi)\]

1602:  and taking into account the defining relations (\ref{eq:ZFDEF}).

1603: We note that using this identification all of the defining relations

1604: (\ref{eq:ZFDEF}) can be combined into a single one\[

1605: Z^{*}(\theta_{1})Z^{*}(\theta_{2})=S(\theta_{1}-\theta_{2})Z^{*}(\theta_{2})Z^{*}(\theta_{1})+2\pi\delta(\theta_{1}-\theta_{2}-i\pi)\]

1606:  In the boundary case the generators $Z,Z^{*}$ are defined only for

1607: positive values of the rapidity arguments and additionally two new

1608: formal generators are introduced creating the boundary vacuum as follows\[

1609: \vert0\rangle_{B}=B^{*}\vert0\rangle\quad,\qquad\,_{B}\langle0\vert=\langle0\vert B\]

1610: We introduce two new relations \[

1611: Z^{*}(\theta)B^{*}=R(\theta)Z^{*}(-\theta)B^{*}\]

1612: and \[

1613: BZ(\theta)=BZ(-\theta)R(-\theta)\]

1614: which describe how we can extend the generators for negative rapidities.

1615: By analytically continuing in the rapidity again we have singularities

1616: in the operator products \[

1617: -i\mathop{\textrm{Res}}_{\theta=i\frac{\pi}{2}}Z^{*}(\theta)B^{*}=\frac{g}{2}B^{*}\]

1618:  and \[

1619: i\mathop{\textrm{Res}}_{\theta=-i\frac{\pi}{2}}BZ(\theta)=\frac{g}{2}B\]

1620: They would correspond to particles with real rapidity in the crossed

1621: channel (exchanging time and space coordinates). These new relations

1622: can again be summarized in a single one\[

1623: Z^{*}(\theta)B^{*}=R(\theta)Z^{*}(-\theta)B^{*}+2\pi\delta(\theta-\frac{i\pi}{2})\frac{g}{2}B^{*}\]

1624: together with its formal conjugate \[

1625: BZ(\theta)=BZ(-\theta)R(-\theta)+2\pi\delta(\theta+\frac{i\pi}{2})\frac{g}{2}B\]

1626:  We claim that representing the form factor of the boundary operator

1627: $\mathcal{O}(0)$ as \[

1628: F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})=\langle0\vert B\,\mathcal{O}(0)\, Z^{*}(\theta_{1})\dots Z^{*}(\theta_{n})B^{*}\vert0\rangle\]

1629: and supposing locality \[

1630: [\mathcal{O}(0),Z^{*}(\theta)]=0\]

1631: we can recover all the non-singularity type form factor axioms immediately.

1632: For deriving the singularity axioms we have to observe that singularity

1633: appears not only from a single term. E.g. in the boundary kinematical

1634: singularity axiom, the form factor $F_{n}^{\mathcal{O}}(\theta_{1},\dots,\theta_{n})$

1635: exhibits a singularity in $\theta_{1}$ at $i\frac{\pi}{2}$ coming

1636: from two places: the operator product of both $B$ and $B^{*}$ with

1637: $Z^{*}(\theta_{1})$ is singular. Supposing that they appear in additive

1638: terms of the form factor we can obtain the desired formula.

1639: 

1640: Finally, we note that formulating the boundary FZ algebra in the spirit

1641: of \cite{Max} might lead to a more rigorous derivation of our axioms. 

1642: 

1643: \begin{thebibliography}{10}

1644: \bibitem{Sbstr}G. Mussardo\emph{, Phys.Rept}. \textbf{218}, (1992) 215-379. 

1645: \bibitem{Sanal}P. E. Dorey, \emph{Exact S matrices}, Lecture Notes in Physics, Springer,

1646: eds. Zal\'an Horv\'ath and L\'aszl\'o Palla, E\"otv\"os Summer

1647: School in Physics: Conformal Field Theories and Integrable Models,

1648: Budapest, Hungary, 13-18 Aug 1996. 

1649: \bibitem{Smirnov}F.A. Smirnov: \texttt{Form-factors in completely integrable models

1650: of quantum field theory}, \emph{Adv. Ser. Math. Phys}. \textbf{14}

1651: (1992) 1-208. 

1652: \bibitem{Arndt}H. Babujian, A. Fring, M. Karowski, A. Zapletal, \emph{Nucl.Phys.}

1653: \textbf{B538} (1999) 535-586.

1654: \bibitem{GZ}S. Ghoshal and A.B. Zamolodchikov, \emph{Int. J. Mod. Phys.} \textbf{A9}

1655: (1994) 3841-3886 (Erratum-ibid. \textbf{A9} 4353), hep-th/9306002. 

1656: \bibitem{Muss1pt}G. Mussardo, \emph{Spectral Representation of Correlation Functions

1657: in two-dimensional Quantum Field Theories,} hep-th/9405128.

1658: \bibitem{DPTW1pt}P. Dorey, M. Pillin, R. Tateo and G. Watts, \emph{Nucl. Phys.} \textbf{B594}

1659: (2001) 625-659. 

1660: \bibitem{cardy_mussardo}J.L. Cardy and G. Mussardo, \emph{Nucl. Phys.} \textbf{B340} (1990)

1661: 387-402.

1662: \bibitem{BBT}Z. Bajnok, G. B\"{o}hm and G. Tak\'{a}cs, \emph{J. Phys.} \textbf{A35}

1663: (2002) 9333-9342, hep-th/0207079.\\

1664:  Z. Bajnok, G. B\"{o}hm and G. Tak\'{a}cs, \emph{Nucl. Phys.} \textbf{B682}

1665: (2004) 585-617, hep-th/0309119. 

1666: \bibitem{XXZ}M. Jimbo, R. Kedem, H. Konno, T. Miwa, R. Weston, \emph{Nucl. Phys.}

1667: \textbf{B448} (1995) 429-456. 

1668: \bibitem{XXZGen}Y.-H. Quano, \emph{Int.J.Mod.Phys.} \textbf{A15} (2000) 3699-3716,

1669: \emph{J. Phys.} \textbf{A33} (2000) 8275, \emph{J.Phys.} \textbf{A34}

1670: (2001) 8445-8464. 

1671: \bibitem{SGff}B. Hou, K. Shi, Y. Wang, W.-l. Yang, \emph{Int. J. Mod. Phys.} \textbf{A12}

1672: (1997) 1711-1741.

1673: \bibitem{DTWbct}P. Dorey, R. Tateo and G. Watts, \emph{Phys. Lett.} \textbf{B448}

1674: (1999) 249-256.

1675: \bibitem{FMS}A. Fring, G. Mussardo, P. Simonetti, \emph{Nucl.Phys.} \textbf{B393}

1676: (1993) 413-441. 

1677: \bibitem{KW}M. Karowski and P. Weisz, \emph{Nucl. Phys.} \textbf{B139} (1978)

1678: 455-476. 

1679: \bibitem{counting_ops}G. Delfino and G. Mussardo, \emph{Nucl. Phys.} \textbf{B455} (1995)

1680: 724-758, hep-th/9507010.

1681: \bibitem{Z1}Al.B. Zamolodchikov, \emph{Nucl. Phys.} \textbf{B348} (1991) 619-641. 

1682: \bibitem{DPTW1}P. Dorey, A. Pocklington, R. Tateo and G. Watts, \emph{Nucl. Phys.}

1683: \textbf{B525} (1998) 641-663. 

1684: \bibitem{CM}J.L. Cardy and G. Mussardo, \emph{Phys. Lett.} \textbf{B225} (1989)

1685: 275-278. 

1686: \bibitem{shGEF}E. Corrigan, A. Taormina, \emph{J.Phys.} \textbf{A33} (2000) 8739.

1687: \bibitem{koubek_mussardo}A. Koubek and G. Mussardo, \emph{Phys. Lett.} \textbf{B311} (1993)

1688: 193-201, hep-th/9306044.

1689: \bibitem{c-th}A. B. Zamolodchikov, \emph{Pis'ma Zh Eksp. Theor. Fiz.} \textbf{43}

1690: (1986) 565. (\emph{JETP Lett.} \textbf{43} (1986) 730.)

1691: \bibitem{delta-th}G. Delfino, P. Simonetti and J-L. Cardy, \emph{Phys. Lett.} \textbf{B387}

1692: (1996) 327-333.

1693: \bibitem{BLusch}Z. Bajnok, L. Palla and G. Takács, \emph{Nucl. Phys.} \textbf{B716}

1694: (2005) 519-542.

1695: \bibitem{Max}M.R. Niedermaier, \emph{Nucl.Phys.} \textbf{B440} (1995) 603-646;

1696: Erratum-ibid. \textbf{B456} (1995) 755.

1697: \end{thebibliography}

1698: 

1699: \end{document}

1700: