1: \documentclass[aps,superscriptaddress]{revtex4}
2:
3: \usepackage{amsmath,graphics}
4:
5: \bibliographystyle{apsrev}
6:
7: \begin{document}
8:
9: \title{A form factor approach to finite temperature correlation functions
10: in $c=1$ CFT}
11: \author{S.~Peysson}
12: \email{speysson@science.uva.nl}
13: \affiliation{Inst. for Theoretical Physics, University of Amsterdam,
14: Valckenierstraat 65, 1018 XE Amsterdam, The Netherlands}
15: \author{K.~Schoutens}
16: \email{kjs@science.uva.nl}
17: \affiliation{Inst. for Theoretical Physics, University of Amsterdam,
18: Valckenierstraat 65, 1018 XE Amsterdam, The Netherlands}
19: \affiliation{Department of Physics, P.O.Box 400714, University of Virginia,
20: Charlottesville, VA 22904-4714, USA}
21:
22: \date{May 06, 2002}
23:
24: \begin{abstract}
25: The excitation spectrum of specific conformal field theories (CFT) with
26: central charge $c=1$ can be described in terms of quasi-particles with
27: charges $Q=-p,+1$ and fractional statistics properties. Using the
28: language of Jack polynomials, we compute form factors of the charge
29: density operator in these CFTs. We study a form factor expansion for
30: the finite temperature density-density correlation function, and find
31: that it shows a quick convergence to the exact result. The low-temperature
32: behavior is recovered from a form factor with $p+1$ particles, while the
33: high-temperature limit is recovered from states containing no more
34: than 3 particles.
35:
36: \end{abstract}
37:
38: \maketitle
39:
40: \section{Introduction}
41:
42: One of the extraordinary features of low-dimensional condensed matter systems
43: is that they possess fractional characteristics.
44: Very soon after the discovery of the fractional quantum Hall effect, it was
45: understood that the excitations of such states of matter carry both
46: fractional charge and fractional exchange statistics.
47: Later, Haldane \cite{HaldanePRL67} proposed an interpretation of these
48: statistics as being
49: exclusion statistics, a concept that can be defined for any dimension.
50: It states that the presence of a particle $i$ restricts the dimension
51: of the Hilbert space that is available to another particle $j$ by
52: an amount $g_{ij}$, called the exclusion statistics parameter. This
53: generalizes the Pauli principle for fermions.
54:
55: The low-energy effective theory for a fractional quantum Hall (fqH)
56: system is a chiral conformal field theory (CFT), and it has been found
57: \cite{vanElburgPRB58} that these edge
58: theories can be analyzed in terms of quasi-particles with fractional
59: charge and statistics. More general chiral RCFT spectra can be
60: built out of fractional statistics excitations through what is known as
61: the ``universal chiral partition function'' \cite{Berkovich99}.
62: In the following, we will be interested in the edge CFT of a principal
63: Laughlin state at filling fraction $\nu=1/p$, where $p$ is an integer.
64: The edge theory is that of a compactified boson at radius $R^2=p$.
65: The chiral Hilbert space of the theory has been understood
66: \cite{vanElburgPRB58} as a collection of multi-particle states built out
67: of fundamental quasi-particle excitations. For $\nu=1/p$, they are edge
68: electrons, of charge $(-e)$, and edge quasi-holes, of charge $(e/p)$,
69: both described by primary fields of the CFT.
70:
71: It was shown \cite{SchoutensPRL79,vanElburgPRB58} that these quasi-particle
72: excitations obey fractional (exclusion) statistics based on the matrix
73: \begin{equation}
74: \label{eq:gmatrix}
75: \boldsymbol{g}=\left(\begin{array}{cc} p & 0 \\ 0 & 1/p \end{array}\right).
76: \end{equation}
77: It means that edge electrons and edge quasi-holes are $p$- and $1/p$-ons,
78: with no mutual exclusion. The $g=p$ and $g=1/p$ particles enjoy duality
79: properties, in that their thermodynamic distribution functions are dual
80: to each other.
81:
82: These properties are reminiscent of the Calogero-Sutherland (CS) model.
83: Indeed, the chiral CFT for $\nu=1/p$ can be identified with the continuum
84: limit of a CS model with interaction parameter equal to $p$
85: \cite{IsoNPB443}.
86: The relation between the fqH basis and the CS basis, defined in terms of
87: Jack symmetric polynomials, has been worked out in \cite{vanElburgJPA33}.
88: The Jack polynomial technology is then available to compute the action of
89: observables on multi-particle states. From there, form factors can be
90: obtained, as has been shown in \cite{vanElburgJPA33}. Proceeding in this
91: manner, one does not have to rely on form factor axioms for integrable
92: field theories (IFT), since explicit computations in a regularized CFT
93: can be performed.
94:
95: It has been proposed that, in integrable field theories,
96: correlation functions at finite temperature can be represented
97: in a form factor expansion. In such an expansion one adds the
98: contributions to the correlator coming from specific multi-particle
99: states, each weighted by an appropriate multi-particle distribution
100: function. There is an ongoing discussion on how precisely these
101: ideas can be implemented for IFTs
102: \cite{LeClairNPB552,DelfinoJPA34,MussardoJPA34,KonikXXX,SaleurNPB567}.
103: In the context of $c=1$ CFTs, a form factor expansion for a correlator
104: can be compared with the exact result obtained thanks to the KZ
105: equations. In \cite{vanElburgJPA33}, the expansion of a specific
106: electron Green function in the $\nu=1/p$ chiral CFT was shown to
107: converge quickly in terms of the number of excitations considered.
108: We report here on progress made in the same direction for the
109: density-density correlation function, paying special attention to
110: low- and high-temperature limits. At the technical level our new
111: results include
112:
113: \begin{description}
114: \item{(i)} an interpretation of the quasi-particle states with non-monotonic
115: ordering of the $m_i$ and $n_j$, in terms of `vanishing' and
116: `scattered states', see eq.~(\ref{eq:scatt}),
117: \item{(ii)} exact results for the action of the density operator on states
118: with up to three particles,
119: \item{(iii)} a precise definition of the irreducible part of form factors
120: with more than a single quasi-hole, in which case the sector
121: index `Q' plays an important role, see eq.~(\ref{eq:defirr-qh}),
122: \item{(iv)} consistency checks using the sum-rule eq.~(\ref{eq:sumrule}),
123: based on the
124: Sugawara form of the Virasoro operator in a CFT with $U(1)$ symmetry.
125: \end{description}
126:
127: Analyzing the form factor expansion for the density-density
128: correlator, we
129: demonstrate that the leading asymptotic behavior for low
130: temperature is given by the `exciton' configuration with
131: a single electron and $p$ quasi-holes. More surprisingly, we
132: show by explicit computation that the high-temperature limit (meaning
133: the $\beta\varepsilon\to 0$ limit) of the Green function
134: $\langle \rho(-\varepsilon)\rho(\varepsilon)\rangle_T$, is recovered from
135: form factors containing
136: no more than 3 particles. This remarkable result is established
137: with the help of the identity eq.~(\ref{eq:ng-identity}) satisfied by
138: the thermodynamic distribution function $\bar{n}_g(\varepsilon)$ for particles
139: satisfying exclusion statistics with parameter $g$.
140:
141: Our computations in this paper can be viewed as a dry run for
142: the computation of non-trivial transport computations in
143: systems that are not CFTs but that do have quasi-particles
144: with well-defined exclusion statistics properties (examples
145: are finite-$N$ CS models or Haldane-Shastry spin chains).
146: For the formalism to become of practical use, it needs
147: further streamlining, possibly by making and exploiting
148: a connection with the axiomatic approach to form factors,
149: based on scattering data in IFT.
150:
151: This paper is organized as follows.
152: In section \ref{sec:basis}, we introduce our notations by recalling the fqH basis construction in \cite{vanElburgJPA33}.
153: We give the basic properties of the excitations and their expressions in terms of Jack polynomials.
154: In section \ref{sec:zoo}, we give the expressions for form factors up to 3 particles for the density operator.
155: In section \ref{sec:expansion} we turn to the form factor expansion for the
156: density-density correlation function.
157: Using the results of the preceding section, we show what are the different
158: contributions to the low-temperature and high-temperature limits.
159: Exact numerical results will sustain our arguments in favor of a quick
160: convergence for the full correlation function.
161:
162: \section{Fractional quantum Hall systems and Jack polynomials}
163: \label{sec:basis}
164:
165: \subsection{fqH basis}
166: The effective low-energy theory for a $\nu=1/p$ fqH system is a compactified boson with radius $R^2=p$, which is a chiral $c=1$ CFT.
167: The Hilbert space for this theory is obtained as a collection of sectors of zero modes. The sectors are labeled by an integer $Q$, which is the $U(1)$
168: electric charge measured in units of $e/p$. The partition function is
169: \begin{equation}
170: Z={\rm Tr}\left( q^{L_0} \right) =
171: \sum_{Q=-\infty}^{\infty} \frac{q^{Q^2/2p}}{(q)_\infty}
172: \end{equation}
173: where $(q)_N=\prod_{l=1}^{N} (1-q^l)$.
174:
175: This Hilbert space can be understood as a collection of multi-particle states, the fundamental quasi-particles being the edge electron and the edge quasi-hole, of charge $Q=-p$ and $Q=1$ respectively.
176: They are described by the conformal primary fields
177: \begin{equation}
178: J^{(-p)}(z)=\sum_t J_{-t} z^{t-p/2} \qquad
179: \phi^+(z)=\sum_s \phi_{-s} z^{s-1/2p} \ .
180: \end{equation}
181: The Fourier modes of these operators can be interpreted as creation operators. The independent multi-particle states that generate the chiral Hilbert space were identified to be
182: \begin{align}
183: \label{eq:basis}
184: &|m_M,\ldots m_1;n_N\ldots n_1>^Q \equiv J_{-(2M-1)p/2+Q-m_M} \ldots J_{-p/2+Q-m_1} \, \phi_{-(2N-1)/2p-Q/p-n_N} \ldots \phi_{-1/2p-Q/p-n1} |Q>,\\
185: & \quad m_M \geq \ldots \geq m_1 \geq 0 \qquad n_N \geq \ldots \geq n_1 \geq 0 \qquad (n_1>0 \text{ if } Q<0),
186: \end{align}
187: where $|Q>$ ($Q=-(p-1),\ldots,-1,0$) is the lowest-energy state of charge $Q$. The identification was proven through the equality of the partition functions.
188:
189: \subsection{Fractional statistics and duality}
190: Fractional exclusion statistics is a tool introduced by Haldane \cite{HaldanePRL67} for the analysis of strongly correlated many-body systems.
191: It is only based on the assumption that the Hilbert space is finite-dimensional and extensive, i.e. particles are excitations of the considered condensed matter system, so it is a very generic concept.
192: The statistics are encoded in a matrix $\boldsymbol{g}=(g_{ij})$ corresponding to the reduction of the available Hilbert space for particle of type $i$ by filling a one-particle state by a particle of type $j$.
193: This is then a generalization of the Pauli principle.
194:
195: The thermodynamics for a gas of such particles have been worked out by Isakov, Ouvry and Wu(IOW) \cite{DasnieresDeVeigyPRL72,IsakovMPLB8,WuPRL73}.
196: The one-particle grand canonical partition functions $\lambda_i$ and distribution functions $\bar{n}_i$ are respectively given by the IOW equations
197: \begin{align}
198: &\left( \frac{\lambda_i-1}{\lambda_i} \right) \prod_j \lambda_j^{g_{ij}} = e^{\beta(\mu_i-\varepsilon)}\equiv z_i,\\
199: &\bar{n}_i(\varepsilon) = z_i \frac{\partial}{\partial z_i} \log \prod_j \lambda_j.
200: \end{align}
201:
202: In the case of a fqH at $\nu=1/p$ the statistical matrix is given by (\ref{eq:gmatrix}), where electrons and quasi-holes are non-exclusive to each other.
203: For $p=1$ we recover the Fermi-Dirac distribution functions, and for $p=2$:
204: \begin{equation}
205: \bar{n}_2(\varepsilon)=\frac{1}{2} \left(1-\frac{1}{\sqrt{1+4e^{-\beta(\varepsilon-\mu)}}} \right) \qquad \bar{n}_{1/2} (\varepsilon) = \frac{2}{\sqrt{1+4e^{2\beta(\varepsilon-\mu)}}}.
206: \end{equation}
207: In general for $g$-ons, the distribution has a limiting value for
208: $\varepsilon \to -\infty$ equal to $\bar{n}_g^{\text{max}}=1/g$ and
209: an asymptotic behaviour for $\varepsilon\to\infty$ equal to
210: $e^{-\beta\varepsilon}$.
211:
212: The transport properties of $g$-ons have been worked out in \cite{IsakovPRL83}. In the fqH case, one furthermore has a duality between $g$- and $1/g$-ons, appearing in the identity
213: \begin{equation}
214: g\bar{n}_g(\varepsilon)=1-\bar{n}_{1/g} (-\varepsilon/g)/g \ ,
215: \label{eq:duality}
216: \end{equation}
217: related to the fact that the removal of a $g$-on corresponds to the
218: creation of $g$ $1/g$-ons. This agrees with the physics of edge-to-edge
219: tunneling in the
220: fqH, in which the duality couples weak and strong backscattering.
221: It also means the quasi-particle basis proposed in (\ref{eq:basis}) is
222: not unique.
223: Iso \cite{IsoNPB443} proposed a basis made out of particles of one kind only,
224: filling up 1-particle states with energies extending over both positive and
225: negative values. One shall see that our 'excitation' picture is more
226: practical to compute physical quantities.
227:
228: The duality (\ref{eq:duality}) can be used in the evaluation of
229: thermodynamic quantities.
230: In each $Q$ sector, the partition function decomposes in a product of the
231: partition function for electrons and that for quasi-holes.
232: Then, the specific heat (or central charge) can be written as a sum over the
233: electron and the quasi-hole contribution, and using (\ref{eq:duality}) one
234: finds $c=1$ independent of $p$. Depending on the sign of an imposed
235: Voltage, the Hall conductance is given by an electron or by a quasi-hole
236: expression, of the general form
237: \begin{equation}
238: G=\bar{n}_g^{\text{max}}\frac{q^2}{h} = \frac{1}{p} \frac{e^2}{h} ,
239: \end{equation}
240: where $q=e/p$ or $q=-e$ is the charge of the quasi-particle that carries
241: the current. We shall see that the duality is also instrumental in the
242: computation of form factors and the evaluation of the form factor expansion.
243:
244: \subsection{Correspondence with Jack polynomials}
245: As for the fqH multi-particle basis, it has been shown to be in one-to-one correspondence with the orthogonal eigenbasis of the Calogero-Sutherland (CS) model
246: \begin{equation}
247: H_{CS}=-\sum_{i=1}^N \frac{\partial^2}{\partial x_i^2} + \left(\frac{2\pi}{L}\right)^2 \sum_{i<j} \frac{2\lambda(\lambda-1)}{\sin^2(\pi x_{ij}/L)},
248: \end{equation}
249: in the thermodynamic limit $N\to\infty$ and with interaction parameter $\lambda=p$.
250: This Hamiltonian is best specified using a scalar field $\varphi(z)$ with $\partial\varphi(z)=\sum_n a_n z^{-n-1}$.
251: In terms of this field, the CS Hamiltonian takes the form
252: \begin{equation}
253: H_{CS}=\frac{p-1}{p} \sum_{l=0}^{\infty} (l+1)(i\sqrt{p}a_{-l-1})
254: (i\sqrt{p}a_{l+1}) \, +\frac{1}{3p} \left[ (i\sqrt{p} \partial \varphi)^3 \right]_0.
255: \end{equation}
256: It is then possible to build all the eigenstates of the Hamiltonian using multi-$J$ and -$\phi$ quanta.
257: It is found that the states (\ref{eq:basis}) are not $H_{CS}$ eigenstates, but that they rather act as head states that need to be supplemented by a tail of subleading states.
258: The eigenbasis will then be denoted by
259: \begin{equation}
260: |\{m_i;n_j\}>^Q = |m_i;n_j>^Q + \ldots
261: \label{eq:fqHstates}
262: \end{equation}
263: We refer to \cite{vanElburgJPA33} for further details.
264:
265: From a different perspective, the analysis in \cite{IsoNPB443}
266: has led to a basis of eigenstates of $H_{CS}$, specified as
267: \begin{equation}
268: |\{\mu\},q> = J^{1/p}_{\{\mu'\}}(\{p_n=\sqrt{p}a_{-n}\})|q>
269: \equiv J^{1/p}_{\{\mu'\}}|q>,
270: \label{eq:CSstates}
271: \end{equation}
272: where the $U(1)$ charge $q$ runs over all integers, and $\{\mu\}$ runs over all Young tableaus.
273: $J^{1/p}_{\{\mu'\}}$ is called a Jack symmetric polynomial. They form a basis
274: of the ring of symmetric functions with a given scalar product. Useful
275: details about them are reported in the appendix.
276:
277: A one-to-one correspondence between fqH states
278: (\ref{eq:fqHstates})
279: and CS states
280: (\ref{eq:CSstates})
281: is obtained through the identification \cite{vanElburgJPA33}
282: \begin{equation}
283: |\{m_i;n_j\}>^Q=|(\{m\}+N^M)\cup\{n'\},Q+N-pM>,
284: \label{eq:compositeYoung}
285: \end{equation}
286: with $\{m\}$ the Young tableau built with the $m_i$ quanta and $\{n'\}$ the
287: dual Young tableau built with the $n_j$ quanta.
288:
289: Our strategy will be the following. We will set up the form factor
290: expansion in terms of the fqH basis, but do the actual computation
291: of the form factors using the representation in terms of Jack
292: polynomials, allowing us to exploit what is known about them.
293: In the appendix we present some `Jack polynomial technology' that
294: we have used for computing the various form factors.
295: In the next section we give an overview of the form factors
296: that we obtained.
297:
298: \section{Overview of form factors}
299: \label{sec:zoo}
300:
301: In ref.~\cite{vanElburgJPA33}, form factors for the electron creation
302: operator have been studied.
303: Here we focus on the computation of form factors for the charge density
304: operator in the fqH edge at $\nu=1/p$.
305: Its Fourier modes are $i\rho_m=i\sqrt{p}a_m=ip_{-m}$, so that it has a nice
306: interpretation in the language of Jack polynomials as a power sum.
307: The form factors we are searching for are
308: \begin{equation}
309: <\{\nu\},q|p_{-m}|\{\mu\},q> \ .
310: \label{eq:generalff}
311: \end{equation}
312: We shall obtain them by developing the product of a power sum and a Jack
313: polynomial on the basis of Jack polynomials.
314: In the former works on the Calogero-Sutherland model \cite{LesageNPB435},
315: where only zero-temperature properties were computed, one only considered
316: form factors with either the in or the out state equal to the vacuum.
317: For such cases, only the expansion of power sums on a basis of Jack
318: polynomials is needed. In the case of form factors of the general form
319: \ref{eq:generalff}, which appear in the form factor expansion for
320: finite temperature correlators, it is necessary to use a less traditional
321: approach, combining the expansion of the power sums in elementary symmetric
322: functions and the Pieri formula written in the appendix.
323: This allows in principle to compute \emph{any} form factor.
324: We have obtained closed, analytic expressions for a number of form
325: factors with up to three particles. To obtain them, we have proceeded by
326: evaluating a few simple examples (the smaller $m$'s), conjecturing a
327: general form, and then checking the conjectured form by using the
328: following sum rule
329: \begin{equation}
330: \label{eq:sumrule}
331: \sum_{m \geq 1} {}_N<\{m_i;n_j\}| p_m p_{-m} |\{m_i;n_j\}>_N = p |\mu| \ .
332: \end{equation}
333: The sum rule follows from the so-called Sugawara form of the
334: Virasoro operator $L_0$
335: \footnote{The sum rule (\ref{eq:sumrule}) can also be obtained using only Jack polynomial technology.},
336: \begin{equation}
337: L_0 = \frac{1}{2p} \rho_0^2 + \frac{1}{p} \sum_{m \geq 1} \rho_{-m}\rho_m \ .
338: \end{equation}
339: The sum rule is given in terms of normalized basis states
340: \begin{equation}
341: |\{m_i;n_j\}>_N = N_{\{m_i;n_j\}}^{-\frac{1}{2}} |\{m_i;n_j\}> \ ,
342: \end{equation}
343: with
344: \begin{equation}
345: N_{\{m_i;n_j\}} = \langle \{m_i;n_j\} | \{m_i;n_j\}>
346: = j_{\mu^\prime}^{\frac{1}{p}} \ ,
347: \end{equation}
348: with $\{\mu^\prime\}$ the dual to the composite Young tableau
349: defined in eq.~(\ref{eq:compositeYoung}) and $j$ the inner product
350: given in eq.~(\ref{eq:Jackinner}).
351:
352: We report in the following results for one- to three-particle
353: states, in any $Q$-sector. In each case we give expressions for
354: general $p$, and then specify to the case $p=2$, which is the case
355: analyzed numerically in section \ref{sec:expansion}.
356:
357: \subsection{1 quasi-particle}
358: One finds
359: \begin{equation}
360: \label{eq:ff1p}
361: p_{-m} |m_1>= \pm \sqrt{pg} |m_1-m>,
362: \end{equation}
363: which for $p=2$ gives
364: \begin{align}
365: p_{-m} |\{m_1\}> &= -2 |\{m_1-m\}> \qquad \text{for electrons},\\
366: p_{-m} |\{n_1\}> &= |\{n_1-m\}> \qquad \text{for quasi-holes}.
367: \end{align}
368:
369: \subsection{2 quasi-particles}
370: At the level of two particles, many-body effects start to appear.
371: It means that the density operator is not acting on each particle individually, but rather on both at the same time.
372: \subsubsection{2 electrons or 2 quasi-holes}
373: We obtain the following action
374: \begin{align}
375: \label{eq:ff2p}
376: &p_{-m} |\{m_2,m_1\}> = \pm m \,\sqrt{pg} \sum_{l=0}^{m} G(m_1,m_2;m_1-l,m_2-m+l) |\{m_2-m+l,m_1-l\}>,\\ \nonumber
377: &G(m_1,m_2;m_1'=m_1-l,m_2'=m_2-m+l)= G_{12} \\ \label{eq:Gfunction}
378: &= \sum_{i=0}^{l} \frac{i(-)^i}{l(m-l+i)} \left(\begin{array}{c} m-l+i \\ i \end{array} \right) \left(\begin{array}{c} l \\ i \end{array} \right) \frac{\Gamma(g+i)\Gamma(m_2-m_1+g+1)\Gamma(m_2'-m_1'+g-i)}{\Gamma(g-i)\Gamma(m_2-m_1+g+i+1)\Gamma(m_2'-m_1'+g)} .
379: \end{align}
380: For the expression eq.~(\ref{eq:ff2p}) to make sense, we have to
381: specify the meaning of 2-particle states $|\{m_2,m_1\}\rangle$ with
382: $m_2<m_1$. For the case of electrons, the prescription is
383: \begin{eqnarray}
384: \label{eq:scatt}
385: |\{m_2',m_1'\}>
386: & \Rightarrow & (-)^p \frac{N_{\{m_2',m_1'\}}}{N_{\{m_1'-p,m_2'+p\}}}
387: |\{m_1'-p,m_2'+p\}> \qquad \text{ if } m_2'\leq m_1'-2p \nonumber
388: \\
389: & \Rightarrow & 0 \qquad
390: \text{ if } m_2'-m_1'= -1, -2, \ldots, -2p+1 \ .
391: \end{eqnarray}
392: This rule can be extended to multi-particle states, and to quasi-holes.
393: The latter behave slightly differently: one needs at least $(p+1)$
394: quasi-holes to have scattered states. Clearly, the prefactor
395: $\frac{N_{\{m_2',m_1'\}}}{N_{\{m_1'-p,m_2'+p\}}}$
396: is analogous to the scattering phase that one expects in the
397: continuum limit of this theory.
398:
399: In the case of $p=2$, we have
400: \begin{multline}
401: p_{-m} |\{m_2,m_1\}> = -2 \bigg( |\{m_2-m,m_1\}> + \frac{(m_2-m_1+1)(m_2-m_1+m+3)}{(m_2-m_1+3)(m_2-m_1+m+1)} |\{m_2,m_1-m\}> \\ - 2m \sum_{l=1}^{m-1} \frac{1}{(m_2-m_1+3)(m_2-m_1+2l-m+1)} |\{m_2-m+l,m_1-l\}> \bigg)
402: \end{multline}
403: for electrons, and
404: \begin{multline}
405: p_{-m} |\{n_2,n_1\}> = |\{n_2-m,n_1\}> \\
406: + \frac{\Gamma(n_2-n_1+1/2)\Gamma(n_2-n_1+3/2)\Gamma^2(n_2-n_1+m+1)}{\Gamma(n_2-n_1+m+1/2)\Gamma(n_2-n_1+m+3/2)\Gamma^2(n_2-n_1+1)}) |\{n_2,n_1-m\}> \\
407: + \sum_{l=1}^{m-1}\bigg(\sum_{i=0}^{l} \frac{mi}{l(m-l+i)}(-)^i \left(\begin{array}{c} m-l+i \\ i \end{array} \right) \left(\begin{array}{c} l \\ i \end{array} \right) \frac{\Gamma^2(i+1/2)}{\Gamma^2(1/2)} \\
408: \times
409: \frac{\Gamma(n_2-n_1+3/2)\Gamma(n_2-n_1+2l-m-i+1/2)}{\Gamma(n_2-n_1+i+3/2)\Gamma(n_2-n_1+2l-m+1/2)} \bigg) |\{n_2-m+l,n_1-l\}>
410: \end{multline}
411: for quasi-holes.
412: The factor appearing in front of the second term in the right hand side of
413: these expressions corresponds to the reordering of the particles during the action of the density operator. We shall call it a scattering factor, though it doesn't have the same physical origin as the one appearing in (\ref{eq:scatt}).
414:
415: \subsubsection{1 electron and 1 quasi-hole}
416: \begin{multline}
417: \label{eq:ffmix}
418: p_{-m} |\{m_1;n_1\}> = -p |\{m_1-m;n_1\}> +\frac{(m_1+p(n_1-m)+1)(m_1+p(n_1+1))}{(m_1+p(n_1-m+1))(m_1+pn_1+1)} |\{m_1;n_1-m\}> \\
419: - mp(p-1) \sum_{l=1}^{m-1} \frac{1}{(m_1-m+l+p(n_1-l+1))(m_1+1+pn_1)} |\{m_1-m+l;n_1-l\}>,
420: \end{multline}
421: along with the replacement
422: \begin{equation}
423: \label{eq:replacement}
424: |\{-1;0\}>^Q \equiv \left\lbrace \begin{array}{c} |\{-;-\}>^{-(p-1)} \text{ for } Q=0 \\ |\{0;-\}>^{Q+1} \text{ for } Q<0 \ . \end{array} \right.
425: \end{equation}
426:
427: A $G$-function can also be defined here, defining it as the factor in the
428: last line of (\ref{eq:ffmix}) divided by $m$.
429:
430: \subsection{3 quasi-particles}
431:
432: For the case of form-factors with three quasi-particles, we do not
433: have complete results, but we report the following. We distinguish
434: between terms where a single one, or two, or all three quasi-particles
435: are affected by the action of the density operator
436: \begin{itemize}
437: \item when the density operator is affecting
438: only one or two of the quasi-particles, one obtains the
439: one- or two particle action given above, multiplied by
440: specific scattering factors
441: % \begin{equation}
442: % p_{-m} |\{m_i, \{m_{j\neq i} \} \} =
443: % \ldots\pm m \sqrt{pg_i} \prod_{(k,l)} G_{kl} |\{m_i-m, \{m_{j\neq i} \} \}
444: % \text{ + other contributions}
445: %\end{equation}
446: %that is, the one-particle form factor modulo some scattering factors;
447: %\item the same happens when the density operator
448: %acts on 2 of the 3 quasi-particles;
449: \item when the density operator acts on all three quasi-particles, more
450: complicated many-body effects appear for $p\neq 1$.
451: One has to define a new function
452: \begin{equation}
453: G'_{ij}=\sum_{i=0}^{l-1} (-)^i \left(\begin{array}{c} m-l+i-1 \\ i \end{array} \right) \left(\begin{array}{c} l-1 \\ i \end{array} \right) \frac{\Gamma(g+i)\Gamma(m_2-m_1+g+1)\Gamma(m_2'-m_1'+g-i)}{\Gamma(g-i)\Gamma(m_2-m_1+g+i+1)\Gamma(m_2'-m_1'+g)}
454: \end{equation}
455: for two particles of the same type.
456: Here follow, case by case, the results we could obtain :
457: \begin{description}
458: \item[(2,1).] Acting on $|\{m_2,m_1,m_0=n_1\}>$, the factor
459: multiplying $|\{m_2',m_1',m_0'\}>$ is
460: \begin{align}
461: &m\big[ \delta m_0 G_{01} G_{02} G'_{12} + \delta m_1 G_{01}G_{12} + \delta m_2 G_{02} G_{12} \nonumber \\ &+ (\delta m_0 \delta m_1 + \delta m_0 \delta m_2 + \delta m_1 \delta m_2 /p - \delta m_0 \delta m_1 \delta m_2 /(p-1) ) G_{01} G_{02} G_{12} \big]
462: \end{align}
463: with $\delta m_i = m_i-m_i'$ and the sum of the $\delta m_i$ being equal to $m$. Replacements similar to (\ref{eq:replacement}) have to be made, which is the case for any mixed state.
464: \item[(1,2).] Acting on $|\{n_2,n_1,n_0=m_1\}>$, the factor
465: multiplying $|\{n_2',n_1',n_0'\}>$ is
466: \begin{align}
467: &m\big[ \delta n_0 G_{01} G_{02} G'_{12} / p(1-p) + \delta n_1 G_{01}G_{12} + \delta n_2 G_{02} G_{12} \nonumber \\ &+ (\delta n_1 \delta n_2 - \delta n_0 \delta n_1 \delta n_2 /(p-1) ) G_{01} G_{02} G_{12} \big] \ .
468: \end{align}
469: \item[(3,0).] Acting on $|\{m_3,m_2,m_1\}>$, the factor multiplying
470: $|\{m_3',m_2',m_1'\}>$ is, for the special case $p$=2,
471: \begin{align}
472: &-2m\big[ \delta m_1 G_{12} G_{13} + \delta m_2 G_{12}G_{23} + \delta m_3 G_{13} G_{23} \nonumber \\ &+ (\delta m_1 \delta m_2 -\delta m_2 \delta m_3 +3/2 \delta m_1 \delta m_2 \delta m_3 ) G_{12} G_{13} G_{23} \big] \ . \label{eq:3p}
473: \end{align}
474: \end{description}
475: \end{itemize}
476:
477: In the next section the various form factors presented here will be
478: used in a form factor expansion for the finite-temperature
479: density-density correlation function.
480:
481: \section{Form-factor expansion}
482: \label{sec:expansion}
483:
484: We now turn to the goal of computing a finite-temperature correlation function.
485: Formally, it amounts to compute
486: \begin{equation}
487: \label{eq:corfun}
488: <\mathcal{O}>_T = \frac{1}{Z(T)} \sum_{\Psi\in\mathcal{H}_N} <\Psi|\mathcal{O}|\Psi> \exp(-\beta E_\Psi),
489: \end{equation}
490: where $\mathcal{H}_N$ is the Hilbert space of normalized states.
491: The problem arising with this expression is that it is hard to handle
492: for any given quantum field theory.
493: Indeed, divergences appear in the correlators of the right-hand side
494: and these need to be resummed. In the context of (massive) Integrable
495: Field Theory (IFT), it has been proposed that, using the basis of the
496: asymptotic particle states in the zero-temperature theory, one may be
497: able to rewrite (\ref{eq:corfun}) as a single sum free of divergences
498: \cite{LeClairNPB552}. The resulting formula, called a
499: `form factor expansion', can be evaluated by using scattering data,
500: the thermodynamic Bethe Ansatz (TBA) and the form factor bootstrap
501: (FFB) (see \cite{BalogNPB419} for a discussion of how the TBA
502: is recovered from the FFB). There is an ongoing discussion about
503: how precisely the form factor expansion can be implemented in IFTs,
504: in particular for the case of multi-point correlation functions
505: \cite{LeClairNPB552,DelfinoJPA34,MussardoJPA34,KonikXXX,SaleurNPB567}.
506:
507: For the case of Conformal Field Theory, there has been a similar
508: but independent proposal for writing finite-temperature correlators
509: in terms of quasi-particle form factors and appropriate thermal
510: distribution functions \cite{vanElburgJPA33}. The idea here is
511: that the `fqH basis' (\ref{eq:fqHstates}) of eigenstates of the
512: CS hamiltonian provides the proper notion of `asymptotic particle
513: states', with simple, but non-trivial fractional statistics
514: properties.
515:
516: The motivation for studying the form factor expansion for CFT has
517: been that, if successful, this approach to finite-temperature
518: correlation functions can possibly be extended to models, such
519: as quantum spin chains, that are gapless but that are not CFTs.
520:
521: Comparing the proposed form factor expansions for IFT and CFT
522: (the particular CFT discussed in this paper), one is led to
523: identify a 2-body $S$-matrix of the simple form
524: \begin{equation}
525: \boldsymbol{S}=\exp [ 2i\pi (\boldsymbol{\delta}-\boldsymbol{g})
526: \Theta (\theta) ]
527: \end{equation}
528: in the thermodynamic limit. It is still an open question whether
529: the form factor computed in \cite{vanElburgJPA33} and in
530: this paper can be obtained by means of an axiomatic approach,
531: starting from these scattering data.
532:
533: In this paper, we study the following CFT form factor expansion for
534: one-point functions
535: \begin{align}
536: <\mathcal{O}(\varepsilon)>_T &= \frac{1}{\beta} \sum_{M,N} O^{(M,N)} (\varepsilon) \\
537: O^{(M,N)} (\varepsilon)&= \beta a\sum_{\{m_i;n_j\}} D^{(M,N)}
538: (m,\{m_i;n_j\}) \prod_{i=1}^M \bar{n}_p (a m_i) \prod_{j=1}^N
539: \bar{n}_{1/p} (a n_j) \\
540: D^{(M,N)}(m,\{m_i;n_j\}) &= {}_N<\{m_i;n_j\}| \mathcal{O}(m) |\{m_i;n_j\}>_N + \text{ subfactors} \ .
541: \label{eq:defirr}
542: \end{align}
543: Here $a=2\pi/L$, $\varepsilon=am$, $\mathcal{O}(\varepsilon)=a \mathcal{O}(m)$;
544: the continuum limit is obtained by sending $a\to 0$.
545: $D$ is the irreducible form-factor, which we shall define with precision.
546: In (\ref{eq:defirr}), the subfactors are built from multi-particle states which are subsets of $\{m_i;n_j\}$.
547: Their leading state of the form (\ref{eq:basis}) is the original one on which some creation operators are canceled.
548: It has then to be rewritten with the correct charge sector.
549: Specific examples of our definitions are,
550: \begin{align}
551: \label{eq:defirr-e}
552: D^{(2,0)}(m,\{m_2,m_1\}^Q) \equiv& {}_N^Q<\{m_2,m_1\}| \mathcal{O}(m) |\{m_2,m_1\}>_N^Q \nonumber \\ &- {}_N^Q<\{m_1\}| \mathcal{O}(m) |\{m_1\}>_N^Q - {}_N^Q<\{m_2+p\}| \mathcal{O}(m) |\{m_2+p\}>_N^Q,\\
553: D^{(0,2)}(m,\{n_2,n_1\}^Q) \equiv& {}_N^Q<\{n_2,n_1\}| \mathcal{O}(m) |\{n_2,n_1\}>_N^Q \nonumber \\ &- {}_N^Q<\{n_1\}| \mathcal{O}(m) |\{n_1\}>_N^Q - {}_N^{Q+1(-p)}<\{n_2(+1)\}| \mathcal{O}(m) |\{n_2(+1)\}>_N^{Q+1(-p)},
554: \label{eq:defirr-qh}
555: \end{align}
556: the charge shift $(-p)$ and the momentum shift $(+1)$ being present for $Q$=0.
557:
558: This definition can be interpreted as follows.
559: Each form factor may appear as part of bigger form factors.
560: The way to resum them is to use these irreducible form factors and the one-particle distributions.
561: This expansion has already proved successful in describing the Green function for the $\nu=1/p$ fqH \cite{vanElburgJPA33}.
562: In the following we will concentrate on the density-density correlation function, where $\mathcal{O} (m)= p_m p_{-m}$. Most of the results will be given for any $p$, we will restrict to the $p=2$ case for numerics.
563:
564: Using standard CFT methods, one finds the density-density correlator to be
565: \begin{equation}
566: <\rho(-\varepsilon) \rho(\varepsilon)>_T =
567: \frac{1}{\beta}\, \frac{p\, \beta\varepsilon}{e^{\beta \varepsilon}-1} \ .
568: \end{equation}
569: We shall compare the results of the form-factor expansion with this
570: exact result. We start (section A) by explaining how the procedure works
571: for $p=1$. It already contains all the features and is completely solvable.
572: Then (sections B and C), we treat the general $p$ case, where we evaluate
573: the asymptotics of the form factor expansion for $\beta \varepsilon\to$ 0
574: or $\infty$ in closed form. Finally (section D), we discuss numerical
575: results for the full correlation function for the case $p=2$.
576:
577: \subsection{Case $p$=1}
578:
579: The irreducible form factors are here straightforward to obtain.
580: One finds the following contributions to the density-density correlator
581: \begin{align}
582: O^{(1,0)}&=O^{(0,1)}=\beta\int_\varepsilon^\infty d\varepsilon_1\, \bar{n}_1(\varepsilon_1)=\log \left( \frac{e^{\beta\varepsilon}+1}{e^{\beta\varepsilon}} \right)\\
583: O^{(2,0)}&=O^{(0,2)}=-\beta\int_0^\infty d\varepsilon_1\, \bar{n}_1(\varepsilon_1)\bar{n}_1(\varepsilon+\varepsilon_1)=\frac{1}{e^{\beta\varepsilon}-1}\left[\log 2-e^{\beta\varepsilon} \log \left( \frac{e^{\beta\varepsilon}+1}{e^{\beta\varepsilon}} \right) \right]\\
584: O^{(1,1)}&=\beta\int_0^\varepsilon d\varepsilon_1\, \bar{n}_1(\varepsilon_1)\bar{n}_1(\varepsilon-\varepsilon_1)=\frac{2}{e^{\beta\varepsilon}-1}\log \left( \frac{e^{\beta\varepsilon}+1}{2e^{\beta\varepsilon/2}} \right) \ .
585: \end{align}
586: All the other irreducible form factor vanish.
587: One then easily checks that the sum of all contributions gives the correct answer $\beta\varepsilon/(\exp(\beta\varepsilon)-1)$.
588: The corresponding curves are shown in figure \ref{fig:Figp=1}.
589:
590: \begin{figure}
591: \begin{center}
592: \includegraphics{Figp=1.eps}
593: \end{center}
594: \caption{\label{fig:Figp=1} All the form factor contributions at $p$=1.}
595: \end{figure}
596:
597: Before going to general $p$, let us remark a few facts:
598: \begin{itemize}
599: \item the asymptotics for $\beta\varepsilon\to 0$ are
600: \begin{itemize}
601: \item $\beta\int_0^\infty d\varepsilon\, \bar{n}_1(\varepsilon)$ for 1 electron and 1 quasi-hole
602: \item $-\beta\int_0^\infty d\varepsilon\, \bar{n}_1^2(\varepsilon)$
603: for 2 electrons or 2 quasi-holes
604: \item 0 for 1 electron and 1 quasi-hole;
605: \end{itemize}
606: giving a total of $2\int_0^\infty d\beta\varepsilon\, \bar{n}_1(\varepsilon)(1-\bar{n}_1(\varepsilon))$=$2\bar{n}_1(0)$=1.
607: \item the asymptotics for $\beta\varepsilon\to\infty$ are
608: \begin{itemize}
609: \item $\beta e^{-\beta\varepsilon}$ for 1 electron or 1 quasi-hole
610: \item $\beta(\log 2-1)e^{-\beta\varepsilon}$ for 2 electrons or 2 quasi-holes
611: \item $\beta\varepsilon e^{-\beta\varepsilon}$ for 1 electron and 1 quasi-hole: it is the dominant one, and fits the exact asymptot.
612: \end{itemize}
613: \end{itemize}
614:
615: These features for the asymptots will be shared for general $p$, as we will see now.
616:
617: \subsection{Low-temperature expansion}
618:
619: \begin{figure}
620: \begin{center}
621: \includegraphics{Figlow.eps}
622: \end{center}
623: \caption{\label{fig:low} Comparison of the exact $p=2$ low-temperature asymptot with the contribution from the form factor for a (1,2) state.}
624: \end{figure}
625:
626: We look here at the $\varepsilon\to\infty$ limit, where the correlation
627: function behaves as $p\beta\varepsilon e^{-\beta\varepsilon}$.
628:
629: For a generic multi-particle excitation, this limit will be dominated by
630: excitations for which one quasi-particle has a momentum $m_i$ bigger than
631: $m=\varepsilon/a$, and all the others a much smaller one. The main
632: contribution comes from the term where the density operator acts only
633: on the quasi-particle of momentum $m_i$. This contribution, which appears
634: equally in all the subfactors of the irreducible form factor, is
635: proportional to $\varepsilon^{1-g} \exp(-\beta\varepsilon)$.
636: The contribution from each other particle is of the form
637: \begin{align}
638: \text{contribution}(\varepsilon_{j\neq i}) &\propto \int_0^\infty d\varepsilon_j \bar{n}_{g_j}(\varepsilon_j) (S_{ij}-1)\\
639: &\propto \int_0^\infty d\varepsilon_j \bar{n}_{g_j}(\varepsilon_j) \frac{1}{(\varepsilon_i-\varepsilon)/\sqrt{g_i}\pm \varepsilon_j/\sqrt{g_j}}
640: \end{align}
641: where $S_{ij}$ is a scattering factor. These parts do not affect the leading
642: behavior for $\varepsilon\to\infty$ and one finds a total contribution
643: which falls off faster than $\varepsilon e^{-\beta\varepsilon}$.
644:
645: There is one exception to the above-mentioned rule.
646: The form factor can become big if the intermediate state is the vacuum,
647: that is for an initial neutral excitation consisting of one electron and
648: $p$ quasi-holes. To see this we use the expansion of $p_m$ in terms of
649: Jack polynomials: we remark that for $\lambda=1/p$ this expansion is
650: indeed restricted to states with 1 electron and $p$ quasi-holes.
651: The factor is then quickly found to be
652: \begin{equation}
653: D(1,2)(m=m_1+\sum_{j=1}^p n_j+p;m_1;\{n_j\}) = (\chi^{1/p}_\nu)^2 j^{1/p}_\nu,
654: \end{equation}
655: with $\{\nu\}=(\{n_j+1\},1^{m_1})$.
656: This expression has been evaluated in \cite{HaNPB435}. In the
657: thermodynamic limit $a\to 0$ it gives
658: \begin{multline}
659: O^{(1,p)}(\varepsilon) = \beta p^2\frac{\Gamma(p)\Gamma^p(1/p)}{\prod_{i=1}^{p}\Gamma^2(i/p)} \int d\varepsilon_1 \prod_{i=1}^{p} d\varepsilon_i' \delta(\varepsilon-\varepsilon_1 -\sum_i \varepsilon_i') \frac{\varepsilon^2\varepsilon_1^{p-1} \prod_{i<j} (\varepsilon_j'-\varepsilon_i')^{2/p}}{\prod_i(\varepsilon_1+p\varepsilon_i')^2 \prod_i (\varepsilon_i')^{1-1/p}} \\ \bar{n}_p(\varepsilon_1) \prod_i \bar{n}_{1/p}(\varepsilon_p') \text{ + subdominant terms} \ .
660: \end{multline}
661:
662: Power counting in this formula gives an asymptotic behaviour proportional
663: to $\varepsilon e^{-\beta\varepsilon}$.
664: The coefficient in front is obtained by remarking that
665: \begin{equation}
666: O_T^{(1,p)}(\varepsilon) \simeq O_0^{(1,p)}(\varepsilon) e^{-\beta\varepsilon},
667: \end{equation}
668: For example, for $p=2$
669: \begin{align}
670: O^{(1,2)}(\varepsilon)
671: &\simeq \beta\varepsilon e^{-\beta \varepsilon} 4 \int d\varepsilon_1 d\varepsilon'_1 d\varepsilon'_2 \frac{\varepsilon \varepsilon_1 (\varepsilon'_2-\varepsilon'_1)}{\sqrt{\varepsilon'_1 \varepsilon'_2} (\varepsilon_1+2\varepsilon'_1)^2(\varepsilon_1+2\varepsilon'_2)^2} \Theta(\varepsilon'_2-\varepsilon'_1) \delta (\varepsilon-\varepsilon_1-\varepsilon'_1-\varepsilon'_2) \\
672: &\simeq \beta\varepsilon e^{-\beta \varepsilon} 4 \int_0^\infty d\theta \frac{\cosh (\theta)}{\sinh^3(\theta)} [\sinh(\theta)-\theta]\\
673: &=2 \beta\varepsilon e^{-\beta \varepsilon}
674: \end{align}
675: which is the right answer. The result is shown in figure \ref{fig:low}.
676:
677: \subsection{High-temperature expansion}
678:
679: When contemplating a form-factor expansion for finite temperature
680: correlators, one may worry about the convergence in the
681: high-temperature ($\beta\varepsilon\to 0$) limit, where thermal
682: distribution functions do not effectively suppress many-particle
683: contributions. We will see that in our case the situation
684: is remarkably good: using form factors with up to three
685: quasi-particles, we recover the exact high-temperature limit.
686: To establish this result, we use the following identity
687: for the equilibrium distribution function $\bar{n}_g$
688: \cite{IsakovPRL83}
689: \begin{equation}
690: \bar{n}_g(\varepsilon)
691: + (1-2g) \bar{n}_g^2 (\varepsilon)
692: - g(1-g) \bar{n}_g^3 (\varepsilon)
693: = - \bar{n}_g'(\varepsilon) \ .
694: \label{eq:ng-identity}
695: \end{equation}
696:
697: % \begin{center}
698: % \begin{tabular}{|c|c|c|} \hline
699: % & $g$=2 & $g$=1/2 \\ \hline
700: % $\beta\int d\varepsilon\, \bar{n}_g (\varepsilon)$ & 0.4812 & 0.9624 \\
701: % \hline
702: % $\beta\int d\varepsilon\, \bar{n}_g^2(\varepsilon)$ & 0.0789 & 0.4463 \\
703: % \hline
704: % $\beta\int d\varepsilon\, \bar{n}_g^3(\varepsilon)$ & 0.0159 & 0.2720 \\
705: % \hline
706: % \end{tabular}
707: % \end{center}
708:
709: \subsubsection{1 particle}
710:
711: \begin{figure}
712: \begin{center}
713: \includegraphics{Fig1p.eps}
714: \end{center}
715: \caption{\label{fig:1p} 1-particle contributions at $p$=2.}
716: \end{figure}
717:
718: Direct application of (\ref{eq:ff1p}) gives the irreducible form factor
719: \begin{equation}
720: D^{(1)}(m;m_1) = pg \frac{\Gamma(m_1-m+g)\Gamma(m_1+1)}{\Gamma(m_1-m+1)\Gamma(m_1+g)} \ .
721: \end{equation}
722: The form factor contribution is then
723: \begin{equation}
724: O^{(1)}(\varepsilon\to 0)=pg \,
725: \beta\int d\varepsilon_1\, \bar{n}_g(\varepsilon_1).
726: \end{equation}
727: On figure \ref{fig:1p} are the curves obtained at $p=2$ for one electron and one quasi-hole.
728:
729: \subsubsection{2 particles}
730:
731: \begin{figure}
732: \begin{center}
733: \includegraphics{Fig2p.eps}
734: \end{center}
735: \caption{\label{fig:2p} 2-particle contributions at $p$=2.}
736: \end{figure}
737:
738: Using (\ref{eq:ff2p}) and the definitions for the irreducible form factor (\ref{eq:defirr-e},\ref{eq:defirr-qh}), one obtains for $(m_1,m_2) \gg m$ (2 particles of the same type)
739: \begin{align}
740: D^{(2)}(m;m_2,m_1)&=pg2\sum_{i=1}^m (-)^i \left(\begin{array}{c} m-1 \\ i-1 \end{array} \right) \left(\begin{array}{c} m+i-1 \\ i \end{array} \right) \frac{\Gamma(g+i)\Gamma(m_2-m_1+g-i)\Gamma(m_2-m_1+g+1)}{\Gamma(g-i)\Gamma(m_2-m_1+g)\Gamma(m_2-m_1+g+i)} \\
741: &\simeq pg\delta_{m_2,m_1}(2g-1) \sum_{i=1}^m \frac{(-)^i}{2i-1} \left(\begin{array}{c} m-1 \\ i-1 \end{array} \right) \left(\begin{array}{c} m+i-1 \\ i \end{array} \right)\\
742: &\simeq -pg(2g-1) \delta_{m_2,m_1} \ ,
743: \end{align}
744: leading to
745: \begin{equation}
746: O^{(2)}(\varepsilon\to 0)=
747: -pg(2g-1)\, \beta\int d\varepsilon_1\, \bar{n}^2_g(\varepsilon_1) \ .
748: \end{equation}
749:
750: The curves for two electrons and two quasi-holes at $p=2$ are given in figure \ref{fig:2p}. As can be seen, the limit $\varepsilon \to 0$ is indeed
751: $-12\,\beta\int d\varepsilon_1 \, \bar{n}_2^2(\varepsilon_1)=-0.9472$
752: for 2 electrons, and 0 for 2 quasi-holes.
753:
754: For mixed states (1 electron and 1 quasi-hole), (\ref{eq:ffmix}) leads to the following limit for the irreducible form factor
755: \begin{equation}
756: O^{(1,1)}(\varepsilon\to 0) \sim (p^2+1) \beta\varepsilon \int_\varepsilon d\varepsilon_1 d\varepsilon_1' \frac{1}{(\varepsilon_1+p\varepsilon_1')^2} \bar{n}_p(\varepsilon_1) \bar{n}_{1/p}(\varepsilon_1'),
757: \end{equation}
758: leading to a linear in $\varepsilon$ dependence.
759: So it has a vanishing limit for $\varepsilon\to 0$.
760: This result will be the same for any mixed state.
761:
762: \subsubsection{3 particles}
763:
764: \begin{figure}
765: \begin{center}
766: \includegraphics{Fig3p.eps}
767: \end{center}
768: \caption{\label{fig:3p} 3-particle contribution at $p=2$ in the
769: high-temperature limit. Due to the complexity of the numerical
770: calculations, the discretization used for $\beta\varepsilon$ is
771: $5\times10^{-3}$, a relatively large value which accounts for the
772: under-estimation of the $\varepsilon\to 0$ limit.}
773: \end{figure}
774:
775: Considering the argument given above, we are only interested in the
776: contribution of 3 particles of the same type (3 electrons
777: or 3 quasi-holes). As for 2 particles, we find that the contribution
778: in the limit $\beta\varepsilon\to 0$ comes entirely from the diagonal,
779: where the energies of the three particles are the same. We have not been
780: able to get an analytic expression for the high-temperature limit,
781: but we present the following conjecture
782: \begin{equation}
783: \label{eq:irr3p}
784: O^{(3)} (\varepsilon\to 0) = pg^2(g-1) \,
785: \beta\int d\varepsilon_1\, \bar{n}_g^3(\varepsilon_1) \ .
786: \end{equation}
787:
788: We have strong support for this conjecture. First of all, thanks to
789: (\ref{eq:3p}), we are able to compute the contributions from form
790: factors for 3 electrons at $p=2$ for low values of $m>0$, all
791: representing the limit $\varepsilon=am\to 0$. We reproduce the
792: expected form (\ref{eq:irr3p}) when $m=3$, and we expect that
793: this result is stable for $m\geq 3$. Continuity for bigger
794: values of $m$ (corresponding to finite $\varepsilon=am$) has
795: been checked numerically, as is shown in figure \ref{fig:3p}.
796: [We note that according to eq.~(\ref{eq:irr3p}) the numerical
797: value of the $(3,0)$ contribution at $\varepsilon=0$ will be
798: .1272; the numerical curve displayed in figure \ref{fig:3p}
799: gives a slightly smaller value, due to the fact that a
800: relatively large value of the discretization $a$ had to be used.]
801: Further support for the conjecture comes from the irreducible
802: form factor for 3 quasi-holes for general $p$ at $m$=1, which
803: directly gives the expression (\ref{eq:irr3p}).
804:
805: \subsubsection{Sum of 1, 2 and 3 particle contributions}
806:
807: We are now ready to evaluate the sum of all contributions to
808: the density-density correlator at $\beta\varepsilon\to 0$.
809: Adding the contributions from form factors
810: with 1, 2, or 3 quasi-particles of statistics $g$ results in
811: \begin{equation}
812: pg \beta\int_0^\infty d\varepsilon
813: \left( \bar{n}_g(\varepsilon)
814: + (1-2g) \bar{n}_g^2 (\varepsilon)
815: -g(1-g) \bar{n}_g^3 (\varepsilon) \right)
816: = - pg \beta\int_0^\infty d\varepsilon\, \bar{n}_g'(\varepsilon)
817: = pg \bar{n}_g (0) .
818: \label{eq:distr-id}
819: \end{equation}
820: The density-density correlator has these contributions
821: from electrons ($g=p$) and from quasi-holes ($g=1/p$). Through
822: duality $g\bar{n}_g(0)+1/g\, \bar{n}_{1/g}(0) = 1$, leading to
823: the result
824: \begin{equation}
825: \beta \, <\rho(-\varepsilon) \rho(\varepsilon)>_T
826: \, \stackrel{\beta\varepsilon\to 0}{\to} \, p
827: \end{equation}
828: in agreement with the exact result. We thus find that the high
829: temperature limit is saturated by contributions with up
830: to three quasi-particles, with the 3-particle contributions
831: being absent in the special case $p=1$. In our view, this
832: non-trivial, exact result gives strong support for the validity
833: of the proposed CFT form factor expansion. We mention that,
834: in general, good convergence at the high-temperature end is
835: scarcely seen in form factor expansions.
836:
837: \subsection{Full correlation function}
838:
839: \begin{figure}
840: \begin{center}
841: \includegraphics{Figtot.eps}
842: \end{center}
843: \caption{\label{fig:tot} High temperature expansion. One can compare the exact density-density correlation function at $p=2$ and the contributions from up to 1 and 2 quasi-particles. The agreement is very good in the latter case.}
844: \end{figure}
845:
846: In the two preceding sections, we have proved that the low-
847: and high-temperature asymptotics for the density-density correlation
848: function were recovered in the form factor expansion using only a
849: few quasi-particles. We now observe that the integrated value
850: of the correlator is recovered already at the level of the 1 particle
851: contributions. One only needs the sum rule (\ref{eq:sumrule}) and the
852: procedure for building irreducible form factors to prove this. At the
853: level of 1 particle, one gets
854: \begin{equation}
855: \beta \int_0^\infty d\varepsilon\,
856: O^{(1)} = p \beta \int_0^\infty d\varepsilon\, \,
857: \beta\varepsilon(\bar{n}_p + \bar{n}_{1/p})(\varepsilon)
858: = p \beta \int_0^\infty d\varepsilon\,
859: \frac{\beta\varepsilon}{e^{\beta \varepsilon}-1} \ .
860: \end{equation}
861: At the level of 2 particles,
862: \begin{equation}
863: \beta \int_0^\infty d\varepsilon\, O^{(2)}= -\frac{1}{2} \big[ \beta
864: \int_0^\infty
865: d\varepsilon\, (p \bar{n}_p-\bar{n}_{1/p})(\varepsilon)\big]^2 = 0
866: \end{equation}
867: by duality.
868: For any contribution with more than 2 particles, the integrated curve is 0.
869:
870: All these facts tend to show that the proposed form factor expansion gives
871: a quick convergence for the \emph{full} correlation function.
872: This is indeed the case, as one sees on figure \ref{fig:tot},
873: where the contribution from states with up to 2 particles at
874: $p$=2 have been collected.
875:
876: \section{Conclusion}
877:
878: We have developed a form factor expansion for the density-density
879: correlation function in the fqH edge CFT at $\nu=1/p$.
880: We have used the correspondence between a fqH basis made of dual
881: quasi-particles (electrons and quasi-holes) and the Jack polynomial basis
882: of the Calogero-Sutherland model. We used `Jack polynomial technology'
883: to study form factors, and we obtained analytical expressions for form
884: factors with up to 3 excited particles. Using these form factors in the
885: proposed form factor expansion, we showed that the convergence towards
886: the exact correlation function is quick in terms of the number of excited
887: particles. We demonstrated that in the high- and low- temperature regimes,
888: the expansion reproduces the exact result. This is a clear indication
889: that the expansion is correct.
890:
891: This result opens the way for the computation of correlation functions for
892: non-CFT theories, as soon as their excitations form a gas of fractional
893: statistics particles.
894: This is the case for the Calogero-Sutherland model, and the models derived
895: from it, like the Haldane-Shastry spin chain.
896: Of fundamental interest also is the need to understand the connection
897: between the thermodynamic limit of the CFT form factors and a Thermodynamic
898: Bethe Ansatz method. This would allow for treating such complicated problems
899: as impurities in edge-to-edge tunneling.
900:
901: \vskip 2mm
902:
903: We thank Robert Konik for discussions and comments on the
904: manuscript. This research was supported in part by the
905: Netherlands Organisation for Scientific Research (NWO).
906: KS was supported by the NSF under grant no. DMR-98-02813.
907:
908: \appendix*
909: \section{Jack polynomial technology}
910:
911: Most of the Jack polynomial technology is found in \cite{Macdonald}.
912: We reproduce here the results useful in the present paper.
913:
914: Jack polynomials form a complete and orthogonal basis of the ring of symmetric polynomials with the given inner product:
915: \begin{equation}
916: <p_\mu, p_\nu>=\delta_{\mu,\nu} \lambda^{-l(\mu)} z_\mu
917: \end{equation}
918: where $p_\mu = \prod_{j=1}^{l(\mu)} p_{\mu_j}$, $p_m(\{x_i\})=\sum_i x_i^m$ are the power sums, and $z_\mu = \prod_{i >= 1} i^{m_i}m_i!$.
919:
920: They are defined in a unique way through their development in monomial symmetric functions:
921: \begin{equation}
922: P^\lambda_\mu = \sum_{\nu <= \mu} v_{\mu,\nu}^\lambda m_\nu \qquad \text{with} \quad v_{\mu,\mu}=1.
923: \end{equation}
924:
925: Their norms are
926: \begin{equation}
927: <P^\lambda_\mu,P^\lambda_\mu> = j^{\lambda}_{\{\nu\}}=\prod_{s\in\{\nu\}} \frac{\lambda l(s) +a(s)+1}{\lambda(l(s)+1)+a(s)},
928: \label{eq:Jackinner}
929: \end{equation}
930: where $l(s)$ and $a(s)$ are the leg and the arm of cell $s$ in the tableau $\{\nu\}$.
931:
932: In the paper, we need the expansion of power sums in Jack polynomials:
933: \begin{align}
934: \label{eq:pm/Jack}
935: p_m &= \sum_{\mu\vdash m} \chi^\lambda_\mu J^\lambda_\mu \\
936: \chi^\lambda_\mu &= m \frac{\prod_{s\neq(1,1)} (a'(s)-\lambda l'(s))}{\prod_s (\lambda l(s)+a(s)+1)}
937: \end{align}
938: and in elementary polynomials:
939: \begin{equation}
940: \label{eq:pm/e}
941: p_m=m\sum_{\nu\vdash m}\, (-1)^{m-l(\nu)} \, \frac{(l(\nu)-1)!}{\prod_i m_i(\nu)!} e_\nu,
942: \end{equation}
943: where $e_\nu=e_{\nu_1} \ldots e_{\nu_m}$ and $e_r=J^\lambda_{(1^r)}$ for any $\lambda$. For example,
944: \begin{align}
945: p_1 &= e_1 \\
946: p_2 &= e_{1^2} - 2e_2 \\
947: p_3 &= e_{1^3} - 3e_{21} + 3e_3 \\
948: p_4 &= e_{1^4} - 4e_{21^2} + 2e_{2^2} + 4e_{31} - 4e_4 \\
949: p_5 &= e_{1^5} - 5e_{21^3} + 5e_{2^21} + 5e_{31^2} -5e_{32} -5e_{41} +5e_5 \\
950: p_6 &= e_{1^6} - 6e_{21^4} + 9e_{2^21^2} - 2e_{2^3} + 6e_{31^3} - 12e_{321} + 3e_{3^2} - 6e_{41^2} + 6e_{42} + 6e_{51} - 6e_6 \ .
951: \end{align}
952:
953: Inner products between Jacks and elementary symmetric functions are known through the Pieri formula:
954: \begin{align}
955: \label{eq:Pieri}
956: J^\lambda_\nu e_r &= \sum_\mu \psi'_{\mu/\nu} J^\lambda_\mu,\\
957: \psi'_{\mu/\nu} &= \prod_{C_{\mu/\nu}\backslash R_{\mu/\nu}} \frac{j^\lambda_\mu(s)}{j^\lambda_\nu(s)}
958: \end{align}
959: with $\mu-\nu$ being a vertical $r$-strip (at maximum 1 box per row, for a total of $r$), $C_{\mu/\nu}$ (resp. $R_{\mu/\nu}$) being the union of columns (resp. rows) that intersect $\mu-\nu$.
960:
961: \bibliography{density}
962:
963: \end{document}
964: