cond-mat0310405/a.tex
1: \newif\ifpdf
2: \ifx\pdfoutput\undefined
3:    \pdffalse
4: \else
5:    \pdfoutput=1
6:    \pdftrue
7: \fi
8: 
9: \ifpdf
10: \documentclass[pdftex]{ws-rv9x6}
11: \else
12: \documentclass{ws-rv9x6}
13: \fi
14: 
15: \ifpdf
16:     \usepackage[pdftex]{graphicx}
17:     \pdfcompresslevel=9
18:     \usepackage{color}
19:     \usepackage[pdftex,
20:                pdftitle={Two-dimensional quantum antiferromagnets},
21:                pdfauthor={Gregoire Misguich and Claire Lhuillier},
22:                breaklinks=true,
23:                colorlinks=false,
24:                ]{hyperref}
25: \else
26:     %\usepackage[dvips]{graphicx}
27:     \usepackage[hypertex,breaklinks=true,colorlinks=false]{hyperref}
28: \fi
29: 
30: 
31: \usepackage{rotating_rv}
32: 
33: %______________________________________________________________________________
34: \newcommand{\La}{\line (1,0  ){12}}
35: \newcommand{\Lb}{\line (3,5 ){6}}
36: \newcommand{\Lc}{\line (-3,5 ){6}}
37: \newcommand{\Ld}{\line (-1,0){12}}
38: \newcommand{\Le}{\line (-3,-5){6}}
39: \newcommand{\Lf} {\line(3,-5){6}}
40: \newcommand{\C} {\circle*{4}}
41: 
42: \newcommand{\pA}{\put(-6,-10)}
43: \newcommand{\pB}{\put(6,-10)}
44: \newcommand{\pC}{\put(12,0)}
45: \newcommand{\pD}{\put(6,10)}
46: \newcommand{\pE}{\put(-6,10)}
47: \newcommand{\pF}{\put(-12,0)}
48: \newcommand{\pZ}{\put(0,0)}
49: 
50: \newcommand{\Hex}{\pA{\C}\pB{\C}\pC{\C}\pD{\C}\pE{\C}\pF{\C}}
51: 
52: \newcommand{\pG}{\put( 18,-10)}
53: \newcommand{\pH}{\put( 18,10)}
54: \newcommand{\pI}{\put(0,20)}
55: \newcommand{\pJ}{\put(-18,10)}
56: \newcommand{\pK}{\put(-18,-10)}
57: \newcommand{\pL}{\put(0,-20)}
58: 
59: \newcommand{\pM}{\put(24,20)}
60: 
61: \newcommand{\KagHex}{\pA{\C}\pB{\C}\pC{\C}\pD{\C}\pE{\C}\pF{\C}}
62: \newcommand{\KagStar}{\KagHex\pG{\C}\pH{\C}\pI{\C}\pJ{\C}\pK{\C}\pL{\C}}
63: 
64: \newcommand{\Z}{\mathbb{Z}_2}
65: %______________________________________________________________________________
66: 
67: \begin{document}
68: %______________________________________________________________________________
69: 
70: %\title{Two-dimensional quantum antiferromagnets}
71: 
72: \chapter*{TWO-DIMENSIONAL QUANTUM ANTIFERROMAGNETS}
73: \markboth{G. Misguich and C. Lhuillier}{Two-dimensional
74: quantum antiferromagnets}
75: 
76: 
77: \author{Gr\'egoire {\sc Misguich}}
78: \address{Service de Physique Th\'eorique, CEA/DSM/SPhT\\
79: Unit\'e de recherche associ\'ee au CNRS\\
80: CEA/Saclay 91191 Gif-sur-Yvette C\'edex, France\\
81: E-mail: gmisguich@cea.fr}
82: 
83: \author{Claire {\sc Lhuillier}}
84: \address{Laboratoire de Physique Th\'eorique des Liquides\\
85: Universit\'e P. et M. Curie and UMR 7600 of CNRS\\
86: Case 121, 4 Place Jussieu, 75252 Paris C\'edex, France\\
87: E-mail: claire.lhuillier@lptl.jussieu.fr}
88: 
89: \date{October 17$^{\rm th}$, 2003}
90: \setcounter{tocdepth}{4} % Pour avoir les subsubsections dans la TOC
91: \tableofcontents
92: 
93: 
94: %______________________________________________________________________________
95: %______________________________________________________________________________
96: 
97: \section{Introduction}
98: 
99: In this review  we present some theoretical advances  in the field  of
100: quantum     magnetism   in    two-dimensional     (2D)   systems.    The
101: spin-$\frac{1}{2}$   next  neighbor 2-dimensional  Heisenberg model on
102: Bravais  lattices (square, triangular) is N\'eel ordered\footnote{This
103: generic kind of order,  with a macroscopic sublattice magnetization is
104: called in  the following {\it magnetic  } long-ranged order  (LRO), in
105: contrast  to  other ordered  phases   where  the  long-ranged  ordered
106: correlations concern    $S=0$    scalar   observables   (on    dimers,
107: quadrumers...)}     at $T=0$.\cite{manousakis91,bllp94}   Frustration,
108: small coordination  number, competition  between interactions can lead
109: to specific quantum phases without magnetic  long-ranged order.  Since
110: a  decade  this subject is  a highly  debated  issue  in  the field of
111: quantum magnetism.   It  was revived by  the   discovery of high-$T_c$
112: superconductivity in  the  doped   cuprates and   fueled   by numerous
113: experimental studies of 2D antiferromagnetic insulators.\cite{HFM2000}
114: 
115: Section~\ref{sec:j1j2}  is  devoted to   the  first academic  model of
116: quantum  frustration:  the spin-$\frac{1}{2}$ Heisenberg model  on the
117: square     lattice   with   first-   and second-neighbor  interactions
118: ($J_1$--$J_2$ model).  This  model is one of the  most studied in  the
119: field and this  section  is a short guide   to the literature, with  a
120: special emphasis on the various methods used for this problem.
121: 
122: Section~\ref{sec:VBC} deals  with general properties  of  valence-bond
123: crystals (VBC) and   related  states, the  simplest  phase which  is 
124: commonly realized in frustrated spin systems without magnetic LRO.
125: 
126: In  section~\ref{sec:largeN}  we present  large-$N$ generalizations of
127: the Heisenberg model.  This approach was extensively developed by Read
128: and Sachdev from the   early 90's and has  been  the first to give  an
129: insight into  the  alternative between  VBC  and related phases, which
130: have long ranged  order in local  singlet patterns (whence the name of
131: crystals), and  resonating valence-bond  (RVB) spin-liquids (SL) which
132: are translationally invariant phases where the  quantum coherence is a
133: central issue.
134: 
135: Section~\ref{sec:QDM}  presents some   results  of  quantum dimer  models
136: (QDM).  These models are effective approaches to the quantum phases of
137: antiferromagnets which are  dominated by short-range valence-bonds (or
138: singlets).  They have received recently  some special attention and provide
139: useful insights onto the phenomenology of VBC and RVB SL.
140: 
141: In Section~\ref{sec:MSE} we review  some  results concerning models  with
142: multiple-spin    exchange (MSE)    (also    called    ring   exchange)
143: interactions. These interactions are  now recognized to be  present in
144: several physical systems  and appear to play  an important role in the
145: stabilization of RVB liquid ground-states.
146: 
147: The  last section  is devoted to  the Heisenberg  model  on the kagome
148: lattice  (and  related models).  Despite of  an  important activity on
149: this subject,  the   understanding of the  low-energy   physics of the
150: spin-$\frac{1}{2}$  kagome   antiferromagnet remains   a   challenging
151: problem and we discuss some of the important results and questions.
152: 
153: We should warn the readers that this review is quite ``inhomogeneous''
154: and   cannot,   of    course,     replace    textbooks.\cite{fradkinbook,auerbachbook,tsvelikbook,sachdevbook}
155: While some parts deal  with some rather recent works  (QDM or MSE  for
156: instance), some others are devoted to older results which we think are
157: still  of importance for   current research (beginning  of the  section
158: $J_1$--$J_2$,  large-$N$).  The final part  devoted to kagome reflects
159: our own views and some unpublished material
160:  on still unsettled issues. Some parts are intended to be
161: more pedagogical and concrete (QDM and  beginning of large-$N$ section)
162: while some others contain more qualitative discussions of the physical
163: issues (end of the section $J_1$--$J_2$, VBC, kagome).
164: 
165: 
166: %______________________________________________________________________________
167: \section{$J_1$--$J_2$ model on the square lattice}
168: \label{sec:j1j2}
169: 
170: We consider the following Heisenberg model on the square lattice:
171: \begin{equation}
172: 	{\cal H} =  2 J_1 \sum_{\langle ij\rangle}  {\vec S}_i\cdot{\vec S}_j
173: 	+  2 J_2 \sum_{\langle\langle ij \rangle\rangle}  {\vec S}_i\cdot{\vec S}_j
174: \label{J1J2}
175: \end{equation}
176: where $\langle ij\rangle$ and $\langle\langle ij \rangle\rangle$ denote
177: pairs of nearest and next-nearest neighbors respectively.
178: Although  quite simple in appearance,  this spin model realizes several
179: interesting   phenomena which are    relevant   to a  large class   of
180: 2D  frustrated    quantum magnets: classical degeneracy,
181: order by disorder, destruction of some long-ranged order by quantum
182: fluctuations, break down of the spin-wave expansion, opening of a spin
183: gap and (possibly ?)   spontaneous translation symmetry breaking, etc.
184: For this reason  we start  with  a general overview of  some important
185: results  concerning this  system.  We  will   focus on  the properties
186: related  to {\em frustration}.  A  review  on the non-frustrated model
187: ($J_2=0$) can be found in Ref.\cite{manousakis91}
188: 
189: \subsection{Classical ground-state and spin-wave analysis}
190: It  is  easy  to find    {\em    some} classical ground-state  of    a
191: translation   invariant  Heisenberg  model  on   a Bravais lattice
192: because the energy can always be minimized by a planar helix
193: \begin{equation}
194: 	\vec{S}_{\bf    r}=
195: 		\vec{e}_1 \cos ({\bf q} \cdot  {\bf r})
196: 	+   \vec{e}_2 \sin ({\bf q}\cdot {\bf r})
197: \end{equation}
198: provided  that the pitch  ${\bf q}$  minimizes  the Fourier  transform
199: $J({\bf q})$ of the coupling.\cite{lt47}  In the case of the $J_1$--$J_2$
200: model one has
201: \begin{equation}
202: 	J({\bf q})=2J_1\left(\cos(q_x)+\cos(q_y)\right)
203: 	+2J_2\left(\cos(q_x+q_y)+\cos(q_x-q_y)\right)
204: \end{equation}
205: 
206: \begin{itemize}
207: 
208: \item $J_2<0.5J_1$:
209: $J({\bf q})$ has a single minimum at $(\pi,\pi)$. It corresponds to the
210: ``usual'' N\'eel state.
211: 
212: \item $J_2>0.5J_1$:
213: $J({\bf q})$ has two isolated minima at $(0,\pi)$ and $(\pi,0)$.  They
214: correspond to ferromagnetic   lines  (resp.  columns)  arranged  in an
215: antiferromagnetic   way.  These   states are    sometimes  called {\em
216: collinear}  (in {\em real} space).  From these planar helix states one
217: can build many other ground-states  by rotating globally all the spins
218: of one sublattice with  respect to the other.   Although this costs no
219: energy  for classical  spins at  zero  temperature, it is known (order
220: from disorder, see  below) that configurations  where both sublattices
221: have  their staggered magnetization {\em collinear  in spin space} are
222: selected by thermal or quantum fluctuations.
223: 
224: \item $J_2=0.5J_1$:
225: $J({\bf q})$  has  lines of minima  around  the edges of the  Brillouin
226: zone. At this point the classical ground-state  is highly degenerate :
227: We can write $\mathcal{H}={\rm cst}+J_2 \sum
228: \left(S_1+S_2+S_3+S_4\right)^2$ where the   sum runs over   all square
229: plaquettes and any  state where each plaquette  has a vanishing  total
230: spin minimizes the classical energy.
231: \end{itemize}
232: 
233: Even   at the  lowest    order   in $1/S$, zero-temperature    quantum
234: corrections to the  sublattice magnetization (order parameter) diverge
235: around $J_2=0.5J_1$  (Chandra    and  Dou{\c  c}ot\cite{cd88}).   Such
236: large-$S$ approximation usually tends to overestimate the stability of
237: magnetic phases, therefore this breakdown around $J_2\sim 0.5J_1$ is a
238: strong  evidence for the existence  of quantum  disordered phase(s) in
239: this region of parameter space.
240: 
241: \subsection{Order by disorder ($J_2>J_1/2$)}
242: 
243: The concept of ``order by disorder'' was introduced in 1980 by Villain
244: and  co-workers\cite{vbcc80} in the study  of a frustrated Ising model
245: on the square lattice. In this model the next neighbor couplings along
246: the rows are ferromagnetic as well as those on  the odd columns (named
247: $A$ in the following).  The couplings  on the even columns (named $B$)
248: are  antiferromagnetic.  At  $T=0$   the ground-state has   no average
249: magnetization   and  is   disordered.    This  changes when    thermal
250: fluctuations are introduced:    a $B$-chain sandwiched  between  two A
251: chains  with parallel  spins  has {\em lower  excitations}  than a $B$
252: chain between two $A$-chains with  anti-parallel spins.  This gives  a
253: larger Boltzmann weight  to ferrimagnetically ordered states.  Villain
254: {\it  et al.}    have  exactly  shown    that  the system is    indeed
255: ferrimagnetic at  low temperature.  They were  also able  to show that
256: site dilution (non-magnetic  sites) selects the  same ordered pattern,
257: whence the name  of ``order by disorder''.
258: 
259: A somewhat less drastic     phenomenon has been observed  in   quantum
260: systems.  It is the  selection    of particular long-ranged    ordered
261: quantum states among a larger  family of  ordered solutions which  are
262: classically degenerate at $T=0$. 
263: \footnote{In Villain's model
264: the  system is truly   disordered at $T=0$  and  an ordered solution is
265: entropically   selected   at finite   temperature.     In  the quantum
266: $J_1$--$J_2$ model above,  the  classical solutions can  adopt various
267: ordered patterns:  quantum fluctuations select  among these  patterns
268: the most ordered one, that is the situation  with the highest symmetry
269: and the smallest  degeneracy.   The ultimate  effect of these  quantum
270: fluctuations can be the destruction of the  N\'eel order in favor of a
271: fully quantum ground-state with $\mathcal{O}(1)$ degeneracy.}
272: Consider a  spin system in which the
273: molecular field  created by the  spins of one  sublattice on the other
274: cancels, which is  the case when $J_2>0.5J_1$. Shender\cite{shender82}
275: showed that  if  fluctuations are included,   the system  will  select
276: states in  which all spins are  collinear to each other.  This follows
277: from the fact that (moderate)  fluctuations at one site are orthogonal
278: to the mean value of the magnetization at that site and the system can
279: gain    some magnetic  exchange energy   by   making such fluctuations
280: coplanar on neighboring sites, that is to  making the spins collinear.
281: Such a selection of order by quantum  fluctuations (and dilution)
282: was discussed by Henley\cite{h89} and appears also quite straightly
283: in a spin-wave expansion.\cite{mdjr90}
284: 
285: This  selection of the   $(\pi, 0)$ or  $(0,\pi)$ order spontaneously
286: breaks a four-fold lattice symmetry.  An Ising order parameter is thus
287: generated.  It  takes two values  depending whether  the ferromagnetic
288: correlations are locally arranged  horizontally or vertically. Chandra
289: and co-workers\cite{ccl90a} have  studied this mechanism and predicted
290: the  existence of   a   finite  temperature   Ising  phase  transition
291: independent  of   the    subsequent  development  of     a  sublattice
292: magnetization.  This result has been questioned recently\cite{szosh03}
293: and  the  transition    has    not been observed   so      far in  the
294: spin-$\frac{1}{2}$ model.\cite{mbp03,szosh03}  It    has however  been
295: confirmed by some   recent    Monte Carlo simulations of     the  {\em
296: classical}   Heisenberg model.\cite{wm03}  Very  similar phenomena are
297: present in the   $J_1$--$J_2$  quantum Heisenberg  model  on  the {\em
298: triangular} lattice.\cite{jdgb90,cj92,k93,lblp95}
299: 
300: Melzi   {\it   el   a.}\cite{melzi00,melzi01}  have  studied a   quasi
301: 2D spin-$\frac{1}{2}$ system  which  is believed to be  a
302: $J_1$--$J_2$ square  lattice Heisenberg antiferromagnet.  They found some
303: evidence (splitting   of NMR  lines)  for a   collinear ($(\pi,0)$  or
304: $(0,\pi)$)              magnetic        ordering.              Several
305: estimates\cite{melzi01,rosner02,rosner03,mbp03} indeed  point to $J_2>J_1$    in
306: this compound.
307: 
308: \subsection{Non-magnetic region ($J_2\simeq J_1/2$)}
309: 
310: Consider the two classical ``Ising  states'' corresponding to the wave
311: vectors $(\pi,\pi)$ and $(\pi,0)$.    They  can be taken  as   (crude)
312: variational states for the Hamiltonian Eq.~\ref{J1J2}.  Their energies
313: (per  site) are  $E_{\pi,\pi}=-J_1+J_2$   and  $E_{\pi,0}=-J_2$.    As
314: discussed above, these states cross at $J_2=\frac{1}{2}J_1$.  However,
315: one  can also consider  any  first-neighbor singlet (or  valence-bond)
316: covering of the lattice  as   another variational  state.  In such   a
317: completely dimerized state the  expectation  value of the  energy  per
318: site is $E_{dimer}=-\frac {3}{4}J_1$,   which is below the  two  Ising
319: states around $J_2\simeq J_1/2$.  Of course this very simple argument
320: does not  prove anything  since  ``dressing'' these classical  states
321: with quantum fluctuations (spin   flips in the N\'eel-like  states  or
322: valence-bond motions in the  dimerized wave-functions) will  lower the
323: energies of all these trial  states and it  is absolutely not clear
324: which  one may eventually win.   Nevertheless,  this shows in a simple
325: way why non-magnetic states  ({\it i.e} rotationally invariant or spin
326: singlet) such as dimerized states can be a route to minimize the energy in
327: a   frustrated   magnet.\footnote{Klein\cite{klein82}  introduced   a
328: general procedure to generate local and $SU(2)$ symmetric Hamiltonians
329: for which any first-neighbor dimerized state is an exact ground-state.
330: These Hamiltonians   are simply defined  as  sums of  projectors which
331: annihilate all dimer  coverings.   The Majumdar-Gosh\cite{mg69} chain
332: is the simplest example of a ``Klein model''.}
333: 
334: 
335: \subsubsection{Series expansions}
336: 
337: High-order  series  expansions  can   be   a powerful  technique   to
338: investigate frustrated quantum magnets. The general method to generate
339: zero-temperature perturbation expansions  in quantum many-body systems
340: was described  by Singh {\it et  al.}\cite{sgh88} and Gelfand {\it et
341: al.}\cite{gsh90}   For instance,  one   can  consider  the following
342: anisotropic model:
343: \begin{eqnarray}
344: 	{\cal H}(\lambda) =  &&2 J_1 \sum_{<ij>}
345: 	\left[
346: 	S^z_i S^z_j
347: 	+ \lambda \left(S^x_i S^x_j + S^y_i S^y_j \right)
348: 	\right] \nonumber \\
349: 	&& +  2 J_2 \sum_{<<ij>>}
350: 	\left[
351: 	S^z_i S^z_j
352: 	+ \lambda \left(S^x_i S^x_j + S^y_i S^y_j \right)
353: 	\right]
354: \end{eqnarray}
355: ${\cal H}(\lambda=0)$ is a classical Ising model which ground-state is
356: known. The series expansion about  the Ising limit amounts to  compute
357: expectation values in the ground-state $\left|\lambda\right>$ of ${\cal
358: H}(\lambda)$ in powers of $\lambda$:
359: \begin{equation}
360: 	\frac{
361: 	\left<\lambda\right| {\hat O} \left|\lambda\right>
362: 	}{
363: 	\left<\lambda | \lambda\right>}=\sum_n a_n \lambda^n,
364: \end{equation}
365: (energy  gaps, dispersion  relations and susceptibilities  can also be
366: computed in the same approach).  The calculation of $a_n$ requires the
367: enumeration  and evaluation  of  the {\em connected  clusters} of size
368: $\sim n$, whose number grows exponentially with $n$.  Depending on the
369: quantity ${\hat  O}$  and  on the model,    orders from  7 to   20 can
370: typically be  obtained   on present computers.   The   series  is then
371: extrapolated to  $\lambda=1$  by  standard  Pad\'e, Dlog   Pad\'e   or
372: integrated differential approximations.  Such a series expansion about
373: the  Ising limit was done  by Weihong~{\it et al.}\cite{woh91} for the
374: first  neighbor   square-lattice   antiferromagnet.     Oitmaa     and
375: Weihong\cite{ow96}  extended  the  series  to the $J_1$--$J_2$  model,
376: where  each $a_n$ is   now  a polynomial   in  $J_1$ and  $J_2$.   The
377: disappearance of N\'eel order in the Heisenberg model manifests itself
378: by   a  vanishing sublattice magnetization as    well as some singular
379: behavior of  the series for  $\lambda_c<1$.   The results indicate the
380: absence of N\'eel long-ranged  order in the strongly frustrated region
381: $0.4\le  J_2/J_1\le  0.6$.   Such  an   expansion can  locate  with  a
382: reasonable accuracy  the phase boundary  of the N\'eel ordered regions
383: but provides no direct information  on the nature of the  non-magnetic
384: phase.
385: 
386: To study the model around $J_2\simeq  J_1/2$, several other expansions
387: have been carried  out.  A dimer  expansion about an exactly dimerized
388: model was       done       by  Gelfand~{\it    et    al.},\cite{gsh89}
389: Gelfand,\cite{gelfand90} Singh {\it   et   al.}\cite{swho99} and Kotov
390: {\it et al.}.\cite{kosw00}  In this approach $J_1$  and $J_2$ are set
391: to zero everywhere except on isolated bonds arranged in a columnar way
392: and all  the other  couplings are  treated perturbatively.   At zeroth
393: order  the  ground-state is simply  a  product of singlets.   In these
394: calculations the dimerized   phase remains stable  in the intermediate
395: region.  Singh  {\it et al.}\cite{swho99}  also performed a different
396: kind of zero-temperature series expansion.  They  started from a model
397: of isolated  4-spin plaquettes in order to  check a prediction made by
398: Zhitomirsky  and    Ueda\cite{zu96}   that   such   plaquettes  could
399: spontaneously  form around $J_2\simeq  J_1/2$ to produce a state which
400: is   invariant    under $\pi/4$  lattice   rotations.     Although the
401: ground-state energy they  obtained is very  close to the  one obtained
402: from   the  dimerized limit  (within   error bars of the extrapolation
403: procedure) they observed an instability in the plaquette scenario (the
404: triplet gap vanishes   before  reaching the isotropic   square-lattice
405: model)   which suggests  that plaquette  order  is not  the issue (the
406: analysis of  the exact numerical  spectra for 36  sites confirmed this
407: result\cite{cbps01}).
408: 
409: 
410: Sushkov {\it et al.}\cite{sow02}  (improved numerical results compared
411: to Ref.\cite{sow01})  computed the susceptibility  $\chi_D$ associated
412: with the appearance of columnar dimer  order in the $(\pi,\pi)$ N\'eel
413: phase  by   a series   expansion   about  the   Ising limit.   Such  a
414: susceptibility  seems    to       diverge  at    $J_2/J_1=g_{c1}\simeq
415: 0.405\pm0.04$.   On the other hand  the  disappearance of the magnetic
416: LRO is   observed (through the N\'eel  order  parameter or through the
417: anisotropy   in   spin space    of   the  spin-spin  correlations)  at
418: $J_2/J_1=g_{c2}\simeq 0.39\pm0.02$.  This point could a priori be {\em
419: different} from $g_{c1}$.  In such a case the system would first break
420: the $\pi/4$ lattice rotation  symmetry at $g_{c1}$, while magnetic LRO
421: remains (gapless  spin waves).   Only  at $g_{c2}>g_{c1}$  the $SU(2)$
422: rotation  symmetry is restored and the  magnetic excitations acquire a
423: gap.  From  field  theoretical  arguments based  on  effective actions
424: valid close to  the critical points,  Sushkov {\it et al.}\cite{sow02}
425: argue   that  the proximity  (or  possible  equality)  of $g_{c1}$ and
426: $g_{c2}$ is a general  feature  in frustrated magnets which  originate
427: from the coupling of triplet and singlet excitations.
428: 
429: Sushkov  {\it   et al.}\cite{sow01}  computed  susceptibility $\chi_P$
430: associated  to plaquette  order by an   expansion around the dimerized
431: limit, assuming that  the system  has  columnar dimer  LRO. The result
432: shows a  divergence of  $\chi_P$ when  $J_2/J_1\to g_{c3}=0.5\pm0.02$.
433: From these results Sushkov {\it et al.} suggested that the translation
434: symmetry along the columns is broken  down at $g_{c3}$ (giving rise to
435: an   eight-fold  degenerate ground-state in  the  thermodynamic limit)
436: before  the $(\pi,0)-(0,\pi)$   magnetically ordered phase  appears at
437: $g_{c4}\simeq0.6$. This picture  is qualitatively consistent  with the
438: spin-spin    correlations computed in  a  $10\times10$   system with a
439: density matrix renormalization group (DMRG) algorithm.\cite{cls00}
440: 
441: Due to the relatively short series (typically of order 7) involved and
442: the uncertainties in the extrapolation procedures, such results should
443: be confirmed  by  other methods  but this succession  of quantum phase
444: transitions  represents a very interesting scenario.   We note that if
445: the model has a fully symmetric  liquid ground-state in some parameter
446: range, it should be difficult to  capture from series expansions about
447: limits where some lattice symmetries are explicitly broken.
448: 
449: \subsubsection{Exact diagonalizations}
450: Exact diagonalizations have a priori no bias, and were used very early
451: in      this   field.\cite{dm89,fkksrr90,pgbd91} Large-size  computations and
452: sophisticated finite size scaling  analysis are nevertheless needed to
453: extract   significant      information.        Schulz      {\it     et
454: al.}\cite{schulz} performed extensive exact  diagonalizations
455: of  the $J_1$--$J_2$ model  for system sizes up  to  $36$ sites.  They
456: analyzed the behavior of  several quantities such as structure factors
457: (N\'eel order  parameter), ground-state  energy, spin-wave  velocities
458: (obtained  from the   finite  size  corrections   to the  ground-state
459: energy), spin  stiffness and  uniform susceptibility.  Their analysis,
460: including  quantitative  comparisons   with  non-linear  sigma   model
461: predictions,\cite{nlsm}  concluded to the absence
462: of N\'eel long-ranged order in  the strongly frustrated region $0.4\le
463: J_2/J_1\le0.6$.   There, they   show  enhanced   columnar  dimer-dimer
464: correlations  as  well as  chiral ones but  the size  effects were not
465: clear  enough to discriminate between short  or  long-ranged order for
466: these order parameters.
467: 
468: 
469: 
470: \subsubsection{Quantum Monte Carlo}
471: 
472: Quantum Monte Carlo (QMC) methods have been extensively applied to the
473: $J_1$--$J_2$  model  in the low  frustration   regime giving an highly
474: accurate description of the  N\'eel phase (Sandvik\cite{s97} and Refs.
475: therein).  In the non-magnetic  and highly frustrated regime a  simple
476: QMC approach  is ineffective  due  to the  so-called sign problem. The
477: fixed node  approach is the  first answer  to this  problem: the exact
478: imaginary time propagator   $e^{-\tau\mathcal{H}}$ used to  filter out
479: the ground-state from a variational guess $|\psi_g\rangle$ is replaced
480: by  an    approximate  propagator,  which   has  the    same  nodes as
481: $|\psi_g\rangle$.  The quality of the result depends on the quality of
482: the nodal regions of $|\psi_g\rangle$.  Various schemes have been used
483: to    try to  go  beyond  this  limitation: stochastic reconfiguration
484: (Sorella\cite{s98}), eventually   associated  to  a   few    Lanczos
485: iterations.\cite{s01,cbps01} An alternative method has been devised by
486: du Croo de Jongh {\it et  al.},\cite{cls00} where the guiding function
487: is replaced  by the result of  a  DMRG calculation.\cite{w92,w93} Both
488: methods have their  own bias.  Using  the first of them, Capriotti and
489: Sorella\cite{cbps01}    concluded  that  for   $J_2/J_1\sim  0.45$   a
490: Gutzwiller-projected   BCS  wave-function $|p\;BCS\rangle$  was an
491: excellent guiding wave-function:
492: \begin{eqnarray}
493: 	|p\;BCS\rangle&=&\hat\Pi\left|BCS\right>  \\
494: 	|BCS\rangle&=&\exp{\left(
495: 		\sum_{i,j}f_{i,j}
496: 		c^\dag_{i\uparrow}c\dag_{j\downarrow}
497: 	\right)}|0\rangle
498: \end{eqnarray}
499: where      $\left|0\right>$    is     the        fermion        vacuum,
500: $c^\dag_{i\uparrow}c\dag_{j\downarrow}$ creates  a     valence-bond on
501: sites  ($i,j$)  and   $\hat\Pi$  projects   out  states   with  double
502: occupancy. The pairing  amplitude $f_{i,j}$ (often called gap function
503: $\Delta_k$)\footnote{After the  Gutzwiller projection $\Delta_k$ is no
504: longer the observable gap.\cite{gl91}} is optimized with a Monte Carlo
505: algorithm in order to minimize the energy.  Capriotti and Sorella gave
506: convincing indications that their wave-function is quite accurate. The
507: best variational energies are obtained in the frustrated region with a
508: pairing amplitude    which     mixes   $d_{x^2-y^2}$ and      $d_{xy}$
509: symmetries. In particular it reproduces the correct nodal structure of
510: the ground-state in the frustrated region at least for moderate system
511: sizes where the  variational result can be  checked against the  exact
512: result.   This is a   subtle and non-trivial  information for  systems
513: which do not  obey the Marshall's sign rule  as this frustrated model.
514: They  concluded from these results   that  the system probably had  no
515: long-ranged order neither in dimer-dimer correlations nor in four-spin
516: plaquette correlations.  On  the other hand,  du Croo de Jongh {\it et
517: al.} argued in favor of columnar dimerized phase  which also break the
518: translation  symmetry along  the columns (plaquette-like  correlations
519: similar to those found by series expansions\cite{sow01}).
520: 
521: The comparison of the results of these different approaches shows that
522: this  problem  remains a  very challenging  one.    The  model in  the
523: frustrated regime  is probably never very  far from a quantum critical
524: point and in these conditions none of the available methods seems able
525: to discriminate between  a VBC with tiny  gaps both in the singlet and
526: triplet  sectors, a critical phase  with a quasi  order  in dimers and
527: gapless singlet  excitations, or a true SL  with gaps in any sector of
528: spin but no long ranged order in any observable. As we will explain in
529: the following sections some other frustrated models are happily deeper
530: in the strong  coupling regime  and  exhibit quantum phases which  are
531: easier to characterize.
532: 
533: %______________________________________________________________________________
534: 
535: \section{Valence-bond crystals}\label{sec:VBC}
536: \setcounter{footnote}{0}
537: %______________________________________________________________________________
538: \subsection{Definitions}
539: Among the different quantum  solutions to overcome frustration the VBC
540: is  the simplest scenario.   In this phase,  neighboring spins arrange
541: themselves in  a regular pattern  of singlets: dimers,\footnote{whence
542: the  name   Spin Peierls  phase sometimes  given   to the  VBC phase.} 
543: quadrumers or 2$n$-mers  $S=0$ plaquettes.  The  stability of this phase
544: comes from the extreme stability of  small $S=0$ clusters (recall that
545: the  energy  of a  singlet  of two spins  $\frac{1}{2}$  is -3/4 to be
546: compared to the energy of two classical (or Ising) spins which is only
547: -1/4), and eventually from the  fact that frustrated bonds between two
548: different singlets do not contribute to the total energy.
549: 
550: In a  VBC phase there is no  $SU(2)$ symmetry breaking, no long-ranged
551: order in spin-spin correlations,  but long-ranged order in dimer-dimer
552: or  larger singlet units.  Except  at  a quantum  critical point,  all
553: excitations  of a VBC are gapped.   Depending on the lattice geometry,
554: such a wave  function can  spontaneously  break some  lattice symmetry
555: ({\em  spontaneous VBC}) or can  remain fully symmetric ({\em explicit
556: VBC}).   In a  strict sense,  the name  VBC  should be  reserved  for
557: systems with a spontaneous lattice  symmetry breaking. However,  since
558: these two kinds  of systems share   many similarities we  will discuss
559: both in this section. 
560: 
561: When the Hamiltonian has some  inequivalent bonds and an integer  spin
562: in the unit cell (even number  of spin-$\frac{1}{2}$ for instance) the
563: system can take  full advantage of the  strong bonds and  minimize the
564: effects of the frustrating ones.  In that case  the {\em explicit} VBC
565: is  the ``natural'' strong coupling solution.   One can build a simple
566: Hamiltonian in which the bonds  which are not  occupied by the singlet
567: objects  are   turned off.  The   resulting  model is a   set of small
568: decoupled clusters (dimers  or larger plaquettes) and the ground-state
569: is a trivial product  of singlets.  Importantly, this strong  coupling
570: limit has  the same lattice symmetry  as the original one.  Going back
571: to   the  original Hamiltonian {\em  no   quantum  phase transition is
572: encountered when  going from  the trivial  singlet  product up to real
573: interacting ground-state}. Models with an half-odd-integer spin in the
574: unit cell cannot realize a VBC unless they {\em spontaneously} enlarge
575: their unit  cell.   In these  situations  there is no   unique elected
576: position for the 2$n$-mers and a symmetry  breaking must take place in
577: order to form a VBC.  Examples of these two kinds of VBC will be given
578: below.
579: 
580: 
581: %______________________________________________________________________________
582: 
583: \subsection{
584: One-dimensional and quasi one-dimensional examples (spin-$\frac{1}{2}$
585: systems)}
586: 
587: 
588: One of the  simplest example of  (spontaneous) VBC is observed in  the
589: $J_1$--$J_2$ model  on  the chain for   $J_2/J_1 >  (J_2/J_1  )_c \sim
590: 0.24$.\cite{mg69,ss81,h82,a89,ys97}  For $J_2/J_1=0.5$   the    doubly
591: degenerate   ground-states        are  exact    products            of
592: dimers:\cite{mg69,auerbachbook}
593: \begin{equation}
594: |MG_{\pm}> =\prod_{n=1}^{N/2} |(2n, 2n \pm 1)>.
595: \label{mg} 
596: \end{equation}
597: Here   and in the  following we  call ``dimer''  a  pair of spins in a
598: singlet state, and note it:
599: \begin{equation}
600: |(i,j)> =\frac{1}{\sqrt2} \left[ |i, +>|j, ->  - |i, -> |j, +>\right].
601: \label{dimere} 
602: \end{equation}
603: 
604: For  all $J_2/J_1 >  (J_2/J_1 )_c $,
605: the ground-states are products
606: of dimers, dressed by fluctuations of valence bonds, dimer long-ranged
607: order persists in all the range of  parameters.  This model has gapful
608: excitations  which  can be   described     as pairs   of    scattering
609: spin-$\frac{1}{2}$       solitons      separating    the     two exact
610: ground-states\cite{ss81} (these fractionalized      excitations    are
611: specific of the 1D chain).
612: 
613: The  Heisenberg chain  with alternating  strong and  weak  bonds (Spin
614: Peierls  instability),  has indeed  a  unique  ground-state where  the
615: dimers  are mainly  located on  the strong  bonds. In  that  case, the
616: spin-$\frac{1}{2}$   excitations  are   confined  by   the  underlying
617: potential    and   the   true    excitations   are    gapful   integer
618: magnons~(\cite{nt97a,asrp99} and refs. therein). It is an explicit VBC.
619: 
620: A    two-leg ladder with   AF rung   exchange has   also a  unique VBC
621: ground-state  and  gapped magnons as  excitations.\cite{dr96}  On  the
622: other hand   Nersesyan  and Tsvelik\cite{nt97,km98} have   proposed an
623: example  of frustrated  ladder,  with  a   spontaneously dimerized
624: ground-state, and gapful  excitations.  Excitations of this last model
625: are identified as pairs of singlet and triplet domain walls connecting
626: the two ground-states, they form a continuum.
627: 
628: As  can be seen  from this  rapid and  non exhaustive enumeration, VBC
629: ground-states are relatively  frequent in  frustrated  one-dimensional
630: spin-$\frac{1}{2}$  models.  All   these systems are   gapful  but the
631: excitations could be of different nature emerging as modes (associated
632: to integer spin excitations) or continuum of pairs of excitations that
633: could be  fractionalized   (it  is then specific   of  one dimensional
634: systems) or not.
635: %______________________________________________________________________________
636: 
637: \subsection{Valence Bond Solids}
638: \label{ssec:VBS}
639: The  VBS  wave-function was introduced  by  Affleck, Kennedy, Lieb and
640: Tasaki (AKLT).\cite{aklt87,aklt88} It  can be constructed whenever the
641: spin $S$ on a site  is a multiple  of one half the coordination number
642: $z$: $2S=0$ mod $z$.   Let us consider the  simplest case $2S=z$.   In
643: that case the local spin $S$ can  be seen as the symmetric combination
644: of $2S$ (fictitious) spin-$\frac{1}{2}$.    Now on  each bond  of  the
645: lattice    one can    make    a  singlet    between   two   fictitious
646: spins-$\frac{1}{2}$.  Such a  product  of singlets does not  belong to
647: the  physical Hilbert space  of the  original spin-$S$  model but to a
648: much larger space.  The VBS wave-function is defined as the projection
649: of the singlet-product state onto the physical space.  This projection
650: amounts  to symmetrize (for all  lattice sites) the wave-function with
651: respect to the fictitious spins to force them into a physical spin-$S$
652: state. A VBS can be viewed as  an {\em explicit  VBC of the fictitious
653: spins}.  Simple  Hamiltonians with  short ranged and $SU(2)$-symmetric
654: interactions for   which  the VBS is  an  exact   ground-state can  be
655: constructed (sum of projectors\cite{aklt87,aklt88}).  By  construction
656: the VBS wave-function  is a  spin  singlet and {\em  breaks no lattice
657: symmetry}.  By extension  we may say that a   system is in  a VBS {\em
658: phase} if  its ground-state can be  adiabatically transformed into the
659: VBS wave-function without  crossing a phase  transition.  As the  VBC,
660: models in  the VBS phase have a  gap  to all excitations\footnote{This
661: may however not always  be  true when  the coordination number  of the
662: lattice is large.\cite{aklt87} In such cases  the VBS wave-function is
663: still a spin  singlet but has  long-ranged spin-spin correlations.  We
664: do  not consider such  cases  here.}   but  their wave-functions   are
665: slightly more   complex and their order parameter   is non-local.  The
666: order of VBS is associated to long-ranged singlet-singlet correlations
667: in the {\em fictitious spins}.  Expressing such observable in terms of
668: the physical  spins leads to a  non-local quantity  called {\em string
669: order parameter}.\cite{nr89,kt92} Contrary to  explicit VBC,  VBS have
670: fractionalized degrees of freedom at  the edges of the system with
671: open boundary conditions.  These are simply associated to the unpaired
672: fictitious  spins.  To our  knowledge  these properties  have not been
673: explored in  quantum 2D systems.
674: 
675: The spin-1 Heisenberg chain is the prototype of VBS in 1D.\footnote{In
676: 1D, some authors call ``Haldane systems''  all the spin-gapped phases,
677: whatever  their true nature: VBC  or VBS.}  Such a  state has a unique
678: ground-state, a  gap in the   excitations and exponentially decreasing
679: spin-spin and dimer-dimer\cite{j02}  correlations. See  the chapter by
680: P.~Lecheminant in this volume for more details  about the VBS phase of
681: the spin-1 chain.
682: 
683: A spin-$\frac{3}{2}$ specific $SU(2)$-invariant model on the honeycomb
684: lattice\cite{aklt87,aklt88} is another  example of 2D VBS.  The spin-1
685: Heisenberg  model on  the kagome  lattice   was proposed  to realize a
686: VBS-like  ground-state\cite{h00}  in  which  singlets  form   on every
687: hexagon without any spontaneous symmetry breaking
688: (hexagonal singlet solid).\footnote{Each kagome   site belongs to  two
689: hexagons.      Each   physical spin-1 can     be     split  into   two
690: spin-$\frac{1}{2}$, each of them being  involved in the formation of a
691: singlet on one neighboring  hexagon.}  A similar approach  was carried
692: out for  the spin-1  pyrochlore  Heisenberg model.\cite{yk00}  In that
693: case a  lattice distortion was invoked  to lift the degeneracy between
694: the two singlet states of each tetrahedron.
695: 
696: 
697: %______________________________________________________________________________
698: 
699: \subsection{Two-dimensional examples of VBC}
700: \subsubsection{Without spontaneous lattice symmetry breaking}
701: 
702: 
703: Two spin-$\frac{1}{2}$ experimental examples of 2D (explicit) VBC have
704: recently                   attracted                        attention:
705: CaV$_4$O$_9$\cite{CaV4O9,uksl96,am96b,tku96,sr96,szksu96,wgsoh97}
706: and
707: SrCu$_2$(BO$_3$)$_2$.\cite{k99,nkoum99,mu99,k00,kk00,msku00,tmu01,lws02,kthb02,mu03}
708: Both  of  them have a   lattice derived from  the   square lattice: in
709: CaV$_4$O$_9$ the  V$^{4+}$ ions are on a  1/5  depleted square  lattice and
710: SrCu$_2$(BO$_3$)$_2$ is   an illustration  of  the  Shastry Sutherland
711: model\cite{ss81a} . A lattice embedding the main couplings of these two physical problems
712: is drawn in  Fig.~\ref{qss}.   Interactions   are of  the
713: Heisenberg type.\footnote{Small Dzyaloshinsky-Moriya interactions have
714: been identified in SrCu$_2$(BO$_3$)$_2$.\cite{cz02}}
715: 
716: \begin{figure} 
717: \begin{center}
718: \includegraphics[height=4cm]{QSS}
719: \end{center}
720: \caption[99]{
721: The $\frac{1}{5}$-depleted lattice and the
722: Shastry  Sutherland  lattice.  The
723: strong  bonds  of the  Shastry Sutherland  model  are the   bonds $J'$
724: (dotted dashed lines): they can  accommodate orthogonal dimers  ($J=K$
725: acts as  a perturbation  in  the  real SrCu$_2$(BO$_3$)$_2$).    The
726: lattice formed by the  strongest exchanges in CaV$_4$O$_9$ is obtained
727: with $J=0$.    The phase diagram  of   this model\cite{lws02} contains
728: collinear N\'eel phases, dimer and 4-spin plaquette VBC.}
729: \label{qss}
730: \end{figure}
731: 
732: VBC are obvious  ground-states in the strong  coupling limits of each
733: problem.
734: 
735: The  exact ground-state of the Shastry  Sutherland model is built from
736: singlets   on   the  $J'$  bonds.\cite{ss81a,mu99}   For  $J'/J\gtrsim
737: 1.55\pm0.05 $ the  model has a gap in  the spectrum of excitations and
738: is in a dimerized VBC phase.
739: \footnote{
740: Consider a decomposition  of the  Shastry-Sutherland lattice as  edge-
741: and  corner-sharing  triangles.    For  $J'=2J$  the Hamiltonian    is
742: accordingly written  as a  sum of $J(\vec{S}_1+\vec{S}_2+\vec{S}_3)^2$
743: terms for each  triangle ($J'$-bonds are  shared by two triangles) and
744: each such  term is   minimized by the   dimerized state.   This  shows
745: rigorously that  the dimerized  state is  the ground-state as  soon as
746: $J'/J\gtrsim 2$.}  For $J'/J\lesssim 1.15\pm0.05$ the system is in the
747: $(\pi,\pi)$    N\'eel  state  of   the   square   lattice~(results  of
748: zero-temperature series expansion\cite{zoh02}).  The possibility of an
749: intermediate phase,  possibly with helical  short-ranged correlations,
750: has  been actively discussed in the literature.\cite{am96a,kk00,cms01}
751: S.~Miyahara and  K.~Ueda have recently written  a review of the theory
752: of    the    orthogonal     dimer    Heisenberg   spin  model      for
753: SrCu$_2$(BO$_3$)$_2$.\cite{mu03}
754: 
755: The $\frac{1}{5}$-depleted Heisenberg square lattice model ($J=0$) has
756: been studied  as   a  function of   the  ratio  of the  two  different
757: couplings: bonds within   a plaquette ($K$)   and dimer bonds   ($J'$)
758: between   plaquettes.     At isotropic   coupling   ($J'=K$) collinear
759: long-ranged N\'eel order survives  the depletion, (the decrease in the
760: order parameter is roughly $50\%$\cite{tku96}).   A small unbalance in
761: couplings drives the system either  in a 4-spin plaquette VBC ($K>J'$)
762: or in a dimer  VBC ($K<J'$). Both  (explicit) VBC  phases have a  spin
763: gap.  A  recent generalization  of these  models  by L\"auchli {\em et al}
764: encompasses  both the
765: $\frac{1}{5}$-depleted Heisenberg square lattice model and the Shastry
766: Sutherland model\cite{lws02} (see  Fig.~\ref{qss}).   Its phase diagram
767: exhibits collinear   N\'eel   phases ($(\pi,  \pi)$   or  $(0,  \pi)$)
768: separated from the  VBC   phases by second  order  phase  transitions.
769: Transition between the two  VBC phases which have different symmetries
770: occurs via  a  first order phase transition.\footnote{A  recent $Sp(N)$
771: study of the Shastry Sutherland model\cite{cms01} suggests that a spin
772: liquid phase with deconfined spinons might appear in  such a model. No
773: evidence of such a phase has emerged from the $SU(2)$ studies.}
774: 
775: Excitations in these models come from the  promotion of local singlets
776: to triplet excitations.  In 2D the ordered dimer background provides a
777: confining   force  for  the   spin-$\frac{1}{2}$ excitations.  Indeed,
778: separating two  unpaired spins  (that  is two {\em  spinons})  creates
779: disruption in the ordered  pattern all the  way from the first  to the
780: second.  The  energy cost is  thus  proportional to the length  of the
781: string of defaults and  both spin-$\frac{1}{2}$ excitations remains in
782: fact confined.   Only integer spin  excitations are expected.   On the
783: other hand in   these strongly coupled models  single-triplet hoppings
784: can be difficult and correlated motions might be important, leading to
785: a  large   zoology of  excited modes   (see Ref.\cite{tmu01} and Refs.
786: therein).  This potential frustration of the triplet motion favors the
787: appearance   of     magnetization      plateaus     in             VBC
788: models.\cite{fo99,k99,mt00,mt00b,mu00,mjg01}  This  aspect was briefly
789: discussed in the lecture notes published by the authors.\cite{lm02}
790: 
791: %______________________________________________________________________________
792: 
793: \subsubsection{With spontaneous lattice symmetry breaking}
794: 
795: In  the previous models  the (explicit)  VBC phases  do not  break any
796: lattice  symmetry.  They can be directly  related  to the geometry and
797: relative strength of the couplings.  In more symmetric situations with
798: frustration, spontaneously  symmetry breaking VBC  can appear as a way
799: to overcome  this frustration by  taking full advantage of the quantum
800: fluctuations.  This is probably the  case in the $J_1$--$J_3$ model on
801: the square lattice for intermediate  $J_3/J_1 \sim 0.6$,\cite{ll96} in
802: the  $J_1$--$J_2$ model   on  the hexagonal  lattice  for intermediate
803: $J_2/J_1\sim  0.4$,\cite{fsl01}  and in the  Heisenberg  model  on the
804: checker-board    lattice.\cite{canals02,fsl03,sfl02,bh02,baa03,tsma03} In the
805: two   first cases  the  ground-states are   dressed  columnar VBC   of
806: dimers. Translation and $C_4$  (resp.   $C_3$ only) symmetries of  the
807: lattice are spontaneously  broken.  The ground-state is   4 (resp.  3)
808: times  degenerate.  Spin-spin correlations decrease exponentially with
809: the  system size.  All  excitations   are  gapped.  Contrary to    the
810: $J_1$--$J_2$    model     on     the     square     lattice,     exact
811: diagonalizations\cite{ll96,fsl01,fsl03} give a rather  straightforward
812: information on  these systems where the  correlation lengths are small enough
813: (far enough from  the critical points which limit  the boundary of the
814: VBC phases).
815: 
816: The spin-$\frac{1}{2}$ Heisenberg model  on the checker board lattice,
817: which  can  also  be  seen  as  a planar  lattice  of  corner sharing
818: tetrahedrons (see Fig.~\ref{ckb}),  has received the largest attention
819: for different reasons.\cite{mts01,pc02,canals02,bh02,sfl02,fsl03,baa03,tsma03}
820: \begin{figure}
821: \begin{center}
822: \includegraphics[height=4cm]{checkerboard2}
823: \hspace*{0.5cm}
824: \includegraphics[height=4cm]{checkerboard}
825: \end{center}
826: \caption[99]{Left:  the checkerboard  lattice.  The spins  sit at  the
827: vertices shown  by bullets, all couplings are  identical, ${\bf u}_1$,
828: ${\bf u}_2$  are the unit vectors  of the Bravais  lattice. Right: the
829: ground-states of the Heisenberg model on the checker board lattice are
830: translational  symmetry  breaking  VBC  of 4-spin  plaquettes  on  the
831: uncrossed squares.}
832: \label{ckb}
833: \end{figure}
834: The problem  has   classically  a  continuous  local degeneracy:   the
835: Hamiltonian can  be rewritten as the sum  of the squares of  the total
836: spin of each tetrahedron, and every  configuration with a zero spin on
837: each tetrahedron is a  ground-state.  Classically this problem  shares
838: this  property      with   the Heisenberg   model     on   the kagome,
839: pyrochlore\cite{mc98,mc98a},    garnet\cite{pp01}   and     pyrochlore
840: slab\cite{ka02}  lattices (these lattices    made  of corner   sharing
841: ``simplexes'' with  2, 3  or 4  spins  each were  dubbed ``bisimplex''
842: lattices by Henley\cite{henley2000}).
843: 
844: The quantum spin-$\frac{1}{2}$
845: antiferromagnet  on the  kagome  lattice has been  found  to  be quite
846: specific with a  small gap (if any) toward  triplet excitations and an
847: anomalous density  of  gapless  low-lying singlets   excitations  (see
848: \S\ref{sec:kagome}).
849: 
850: The quantum  scenario on the  checkerboard lattice is quite different.
851: The ground-state is a (dressed) product  of 4-spin $S=0$ plaquettes on
852: uncrossed squares: this  state  breaks translational symmetry  but not
853: $C_4$  (the  point group being  defined  at  the  center of  an  empty
854: plaquette).  It has a two fold degeneracy  in the thermodynamic limit:
855: this is easily seen in the  symmetries and finite  size scaling of the
856: low  lying   levels of  exact spectra.\cite{fsl03}  The  choice of the
857: 4-spin $S=0$  states on the  uncrossed squares corresponds to the most
858: stable local configuration.   The product of  $S=0$ quadrumers is only
859: weakly renormalized in the exact ground-state.\cite{fsl03}
860: 
861: 
862: Excitations of this model  have been studied in different  approaches,
863: exact diagonalizations\cite{f03}, series  expansions,\cite{bh02}  real
864: space renormalization transformation.\cite{baa03} All the  excitations
865: (singlet and  triplet) are gapped.   The triplet excitations originate
866: from the triplet excitation of an uncrossed plaquette, they are weakly
867: dispersive with  a  large  gap.   The singlet   excitations  cannot be
868: described so simply: from exact diagonalizations  data one can suspect
869: that some  of these  excitations come  from the  reorganization of two
870: adjacent triplet  on crossed squares.   The Contractor Renormalization
871: (CORE)  method  of Berg  {\it et  al.}\cite{baa03} on   the other hand
872: suggests that  these  excitations are  domain  walls between   the two
873: translated plaquette-VBC ground-states.
874: 
875: %______________________________________________________________________________
876: 
877: \subsection{Methods}
878: 
879: Spin    waves\cite{uksl96} and  Schwinger bosons\cite{am96a,am96b} are
880: simple  approximations  to   study  the  phase  diagram  of  a quantum
881: frustrated  magnet.  But the  first  method only gives an  approximate
882: knowledge  on the range of  existence of the N\'eel  phases, and it is
883: rather difficult  to  include the effect  of  fluctuations  beyond the
884: mean-field  approximation within  the Schwinger-Boson  formulation. As
885: discussed   in   the   section     devoted to  large-$N$    approaches
886: (\S\ref{sec:largeN}), spontaneous  lattice symmetry  breaking are very
887: likely  to be due to  topological effects (Berry  phase of instantons)
888: which cannot be   captured  by the mean-field  state.    This probably
889: explains why no spontaneous VBC has ever been found (to our knowledge)
890: in  Schwinger-Boson   calculations.\footnote{It  seems   however  that
891: spontaneous VBC naturally arise in large-$N$ approaches with fermionic
892: representation of  $SU(N)$ when $1/N$  corrections are considered (see
893: section
894: \ref{ssec:spin2QDM}).}  For  unfrustrated problems,  as the Heisenberg
895: model on    the  $\frac{1}{5}$-depleted   square  lattice,\cite{tku96}
896: QMC  is considered  as  the   method which  can  give
897: benchmarks for other approaches.
898: 
899: Although  VBC  are   naturally  described     with   the help       of
900: spin-$\frac{1}{2}$ valence-bonds, the appearance of such states in the
901: low-spin  limit can  sometimes   be anticipated  from  an  appropriate
902: large-$S$ approach.   An example is given in  the work of Tchernyshyov
903: {\it   et     al.}\cite{tsma03} (see  also   a    previous    paper by
904: Henley\cite{henley2000})  on the  checkerboard  Heisenberg  model.  In
905: this  model, when both   couplings (square lattice bonds and  diagonal
906: bonds)  are    equal,     ground-state  has    a   continuous    local
907: degeneracy. However  leading $1/S$ corrections select collinear states
908: out  of  this    huge  manifold.\cite{henley2000} There   remains   an
909: exponential number of such  states and they  do not have  any magnetic
910: order.  However    they exhibit long-ranged  {\em   bond}  order and a
911: spontaneous symmetry  breaking\cite{tsma03} which is  analogous to the
912: one observed numerically in the spin-$\frac{1}{2}$ case.
913: 
914: For frustrated problems, exact diagonalizations can be useful tools in
915: situations where the system  is not too  close from a critical  point,
916: that is when  the correlations length is  not  too large.   Successful
917: applications of   exact  diagonalizations  methods  to   2D Heisenberg
918: magnets realizing a VBC include studies of  the $J_1$--$J_3$ model on the
919: square  lattice,\cite{ll96}    $J_1$--$J_2$  model   on    the  hexagonal
920: lattice,\cite{fsl01}     Heisenberg model       on    the checkerboard
921: lattice.\cite{fsl03}    In such  situations    analysis of  the  quasi
922: degeneracy of the  low-lying levels    of  the spectra and  of   their
923: finite-size scaling gives an unbiased  and direct informations on  the
924: symmetry breakings in the thermodynamic limit.\cite{fsl03} However the
925: boundaries of  the phases  and  the quantum critical  points cannot be
926:  accurately determined  with  this method.   The  series expansions
927: described in the previous section appears to be a powerful approach to
928: determine  phase  boundaries.\cite{gwsoh96,wgsoh97} If  the  spin-spin
929: correlation length is   not very  short, as it   is  the case in   the
930: $J_1$--$J_2$    model       on        the   square       lattice      for
931: $J_2/J_1\sim0.5$,\cite{cbps01}  it  is very  difficult  to decide from
932: exact diagonalizations between a VBC, a critical  phase or an RVB SL.
933: 
934: Concerning excitations, exact  diagonalizations give the gaps in  each
935: sector and provide a crude approximation of the dispersion laws of the
936: first  excitations. The large-scale nature  of the excitations (as for
937: examples  domain  walls  excitations) can  escape   this method.   The
938: semi-analytical  approaches which  can be  used  for the study  of the
939: excitations of the    VBC, all use  as  a  basic departure point   the
940: excitations of a local cluster  of spins conveniently renormalized  by
941: perturbation\cite{szksu96,bh02}       or    effective  renormalization
942: technique.\cite{baa03}    Contrary  to   exact  diagonalizations these
943: methods are not limited by finite-size effects  but the results can be
944: biased by the departure point.\cite{bh02,baa03}
945: 
946: %______________________________________________________________________________
947: 
948: \subsection{Summary of the properties of VBC phases}
949: The   generic  features of  VBC   (whatever the dimensionality  of the
950: lattice) are:
951: 
952: \begin{itemize}
953: 	\item  A    spin gap,   no   $SU(2)$  symmetry   breaking  and
954: 	short-ranged spin-spin correlations,
955: 	
956: 	\item Long-ranged order   in dimer-dimer  and/or larger  $S=0$
957: 	plaquettes.  The coupling of this order to lattice distortions
958: 	is probable in experimental realizations of spontaneous VBC.
959: 
960: 	\item In {\em spontaneous}   VBC  phases the  ground-state  is
961: 	degenerate.  From  the theoretical point  of view the discrete
962: 	symmetry of the order parameter of the VBC which spontaneously
963: 	breaks  a   lattice symmetry  may   give  birth   to a  finite
964: 	temperature Ising-like  transition.\cite{ccl90a}  Simultaneity
965: 	between this transition  and a  possible structural transition
966: 	is  likely when  the  couplings of  the  spins  to the lattice
967: 	degrees of freedom (phonons) is considered.\cite{bm02}
968: 
969: 	\item VBC have gapped excitations, in the $S=0$ sector as well
970: 	as  in other $S$  sectors.  A wide zoology  of  modes is to be
971: 	expected  as well  as  continua  associated to  multi-particle
972: 	excitations or  scattering of domain walls (in   the case of a
973: 	spontaneous symmetry  breaking of  the ground-state).   In two
974: 	dimensions  all   these excitations  have  integer  spins (the
975: 	ordered   back-ground   inducing     a  confinement   of   the
976: 	spin-$\frac{1}{2}$ excitations)
977: \end{itemize}
978: 
979:  Frustration on the square  lattice  or more  generally  on bipartite
980:   lattices is often  overcome by
981:   VBC phases.  No  theorem  forbids   the appearance  of VBC   in
982: triangular  geometries but  there is up  to  now  no examples of  such
983: phases   in pure  spin-$\frac{1}{2}$  models  (in   Sec.~\ref{sec:QDM}
984: examples will be given within the framework of quantum dimer models).
985: 
986: It     has  been advocated     in    the  large-$N$    approaches (see
987: section~\ref{sec:largeN}) that, at least  in two dimensions, collinear
988: spin-spin correlations  generically lead  to    VBC or VBS and    only
989: non-collinear spin-spin correlations  can give birth   to RVB SL  with
990: unconfined spin-$\frac{1}{2}$  excitations.  The present  knowledge of
991: $SU(2)$ phase diagrams supports this prediction.  The VBC found so far
992: numerically in $SU(2)$   spin  models appear   to  be in  regions   of
993: parameter space where the  spin-spin correlations are characterized by
994: some   short-ranged   collinear order in     the large-$S$ limit.  The
995: $J_1$--$J_2$   model on  the  honeycomb    lattice   has a   classical
996: incommensurate phase  in the regime of  high frustration and there are
997: some evidences that  in the quantum phase  diagram the collinear phase
998: is separated from the columnar VBC phase  by a RVB SL.\cite{fsl01} The
999: multiple-spin exchange  (MSE) model on  the triangular lattice is also
1000: believed to be  a RVB SL\cite{mlbw99}  and the corresponding classical
1001: ground-states   generically  have  non-coplanar spin   configurations.
1002: Capriotti {\it et al.}\cite{cbps01} argued that the spin-$\frac{1}{2}$
1003: square   lattice $J_1$-$J_2$ model could be   a RVB SL.  If confirmed,
1004: this would be the first counter-example to  the general rule explained
1005: above (The  Heisenberg  model on  the pyrochlore  lattice  might be an
1006: other counter-example\cite{baa03}).
1007: 
1008: %_____________________________________________________________________
1009: 
1010: \section{Large-$N$ methods}\label{sec:largeN}
1011: %_____________________________________________________________________
1012: 
1013: Introduced  by   Affleck,\cite{a85}  Affleck  and  Marston,\cite{am88}
1014: Arovas       and      Auerbach\cite{aa88}      and       Read      and
1015: Sachdev\cite{rs89,rs90,rs91} in the  context of spin models, large-$N$
1016: approaches    are   powerful    methods    to   investigate    quantum
1017: antiferromagnets.  When $N$ is taken  to infinity many of these models
1018: can be solved  by saddle point methods and  finite-$N$ corrections can
1019: be, at least in principle, explored in a controlled way.  A success of
1020: these  approaches is  that they  can describe  the phenomenology  of a
1021: large variety of phases encountered  in quantum magnets : Magnetic LRO
1022: (possibly  with  order  by  disorder  selection)  as  well  as  phases
1023: dominated  by short-ranged  valence-bonds: VBC,  VBS and  RVB liquids.
1024: One       crucial       result       (due      to       Read       and
1025: Sachdev\cite{rs89,rs90,rs91,sr91,s93})  concerning  these three  later
1026: phases  is  that  the  analysis  of  finite-$N$  corrections  to  some
1027: large-$N$ formulations  ($Sp(N)$ for  instance, see below)  provides a
1028: general criterion to  decide which of these three  phase should appear
1029: in a given model.
1030: 
1031: This criterion is the following in 2D: if the (large-$N$ equivalent of
1032: the) ``spin'' $S$ at each site matches the lattice coordination number
1033: $z$ by $2S=0$~mod~$z$ a VBS is to be expected.  If  it is not the case
1034: (as for a spin-$\frac{1}{2}$  model on the  square lattice) one should
1035: look at the local spin-spin correlations.  If  they are reminiscent of
1036: a   collinear   order, a VBC   with  spontaneous  translation symmetry
1037: breaking is  expected whereas  non-collinear short-ranged  correlation
1038: generically  give rise to a RVB  phase without any broken symmetry and
1039: deconfined spinon excitations.  These results are of course based on a
1040: large-$N$  generalization of the original  spin model and  there is no
1041: guaranty at all that these rules should always  apply to $SU(2)$ models.
1042: To our knowledge they have however not  been manifestly found in error
1043: up  to now.\footnote{The   spin-$\frac{1}{2}$  kagome  antiferromagnet
1044: might however be such an  example.  See section~\ref{sec:kagome}.}  In
1045: the following we will  present some of  the important reasoning  steps
1046: leading to this result.
1047: 
1048: 
1049: \subsection{Bond variables}
1050: The $SU(2)$  algebra of a spin $S$ at one site
1051: can be  represented by $2$ species of  particles $a^\dag_\sigma$ (with
1052: $\sigma=\uparrow,\downarrow$),  provided  that  the  total  number  of
1053: particles on one site is constrained to be $a^\dag_\uparrow a_\uparrow
1054: +  a^\dag_\downarrow  a_\downarrow=2S$.   The raising  operator  $S^+$
1055: (resp.  $S^-$) is simply represented by $a^\dag_\uparrow a_\downarrow$
1056: (resp.  $a^\dag_\downarrow a_\uparrow$). These particles can be chosen
1057: to   be   fermions   (Abrikosov   fermions)   or   bosons   (Schwinger
1058: bosons). These particles  carry a magnetization $\pm\frac{1}{2}$ since
1059: $S^z=\frac{1}{2}(a^\dag_\uparrow           a_\uparrow-a^\dag_\downarrow
1060: a_\downarrow)$.  For this reason  they are often called {\em spinons}.
1061: The  Heisenberg  interaction  is   a  quartic  interaction  for  these
1062: particles:
1063: \begin{equation}
1064: 	{\vec S}_i\cdot{\vec S}_j =
1065: 	S^2 - \frac{1}{2} A_{ij}^\dagger A_{ij}
1066: 	\label{eq:SSAij}
1067: \end{equation}
1068: with the bond operator $A_{ij}$ defined by:
1069: \begin{equation}
1070: 	A_{ij} = a_{j\downarrow} a_{i\uparrow}
1071: 	- a_{j\uparrow} a_{i\downarrow}
1072: 	\label{eq:ASU2}
1073: \end{equation}
1074: Acting on the  vacuum, $A_{ij}^\dagger$ creates a spin  singlet on the
1075: bond $(ij)$. Physically $A_{ij}^\dagger A_{ij}$ measures the number of
1076: singlets  on   that  bond   and  Eq.~\ref{eq:SSAij}  shows   that  the
1077: antiferromagnetic Heisenberg  interaction just tries  to maximize that
1078: number.
1079: 
1080: The idea of large-$N$ methods is to generalize the $SU(2)$ symmetry of
1081: the spin$-S$ algebra to a larger group $SU(N)$ (or $Sp(N)$) by letting
1082: the index  $\sigma$ go from  $1$ to  $N$  (or $2N$).  The $SU(N)$  (or
1083: $Sp(N)$) generalization of the Heisenberg  model is solved by a saddle
1084: point calculation of the action.  The $N=\infty$ limit is very similar
1085: to a mean-field decoupling    of  the four-body interaction   of   the
1086: physical    $SU(2)$       model:       $A_{ij}^\dagger    A_{ij}\simeq
1087: A_{ij}^\dagger\left<A_{ij}\right>+\left<A_{ij}^\dagger\right>A_{ij}-\left|\left<A_{ij}\right>\right|^2$.
1088: 
1089: \subsection{$SU(N)$}
1090: The generalization of the  Heisenberg model to  such a  symmetry group
1091: depends  only  on the choice  of   the  irreducible representation  of
1092: $SU(N)$  according to which the  ``spin'' operators transform (and not
1093: on    the  choice    of fermions   or      bosons  to implement    the
1094: representation). For $SU(2)$ this amounts to specify the magnitude $S$
1095: of  the spin. For $SU(N)$  irreducible  representations are labeled by
1096: Young tableaux.  The case of a general  rectangular tableau with $n_c$
1097: columns and $m$ rows was discussed by Read and Sachdev\cite{rs89b} and
1098: $n_c$ appears  to  continue  to play a   role similar  to $2S$  in the
1099: large-$N$ phase  diagrams.\footnote{Taking the limit $N\to\infty$ with
1100: $m$ fixed  of order 1 and $n_c\sim  N$ is most conveniently  done with
1101: bosons $b^\dagger_{\alpha p}$  where $\alpha=1\cdots N$ is the $SU(N)$
1102: index, while $p=1\cdots m$ label the  different ``colors''.  There are
1103: therefore $N m$ kinds of bosons. On the other hand it is convenient to
1104: use $n_c$ ``colors'' of fermions (still with an $SU(N)$ index) to deal
1105: with   the case $n_c\sim\mathcal{O}(1)$   and    $m\sim N$.    Bosonic
1106: representations   with   $n_c\sim  N$  are   appropriate  to  describe
1107: magnetically ordered  phases\cite{aa88}    but cannot  be  used    for
1108: frustrated models (such  a representation is  not self-conjugate).  On
1109: the other  hand fermionic  representations,  such as the  $m=N/2$  and
1110: $n_c=1$ one,\cite{am88,rokhsar90} can be used on  any lattice but they
1111: do not display magnetically ordered phases and tend to favor dimerized
1112: states.\cite{am88,rs89b,rokhsar90}} In this  review we will focus on a
1113: slightly different large-$N$ generalization of the $SU(2)$ model which
1114: is both able to deal with frustration and magnetic states.
1115: 
1116: \subsection{$Sp(N)$}
1117: To perform a large-$N$ extension of {\em frustrated} Heisenberg models
1118: one has  to use either fermions\cite{am88} or  bosons with an $Sp(N)$
1119: symmetry.  The  latter  seems to  produce phase diagrams  that closely
1120: resemble the $SU(2)$ problems and we will focus on this representation
1121: which was introduced by  Read and Sachdev.\cite{rs91} The presentation
1122: below           is       largely      inspired        from       their
1123: papers.\cite{rs89,rs90,rs91,sr91,s93}
1124: 
1125: We now have  $2N$ flavors of bosons  at each site: $b_{i,\sigma}$ with
1126: $\sigma=1..2N$ and we define an $Sp(N)$-invariant bond operator:
1127: \begin{equation}
1128: 	A_{ij} = \sum_{\sigma,\sigma'=1..2N}
1129: 	\mathcal{J}_{\sigma,\sigma'} \;b_i^\sigma \;b_j^{\sigma'}
1130: \end{equation}
1131: where the  $2N\times2N$ antisymmetric   tensor $\mathcal{J}$ is  block
1132: diagonal
1133: \begin{equation}
1134: 	\mathcal{J}=\left[
1135: 	\begin{array}{ccc}
1136: 	\begin{array}{cc}
1137: 	0&1\\-1&0
1138: 	\end{array}&	&\\
1139: 		&\ddots	&\\
1140: 		&	&\begin{array}{cc}0&1\\-1&0\end{array}
1141: 	\end{array}
1142: 	\right]	
1143: \end{equation}
1144: and   generalizes the  $SU(2)$   antisymmetric tensor  $\epsilon_{ij}$
1145: ($SU(2)$ is identical to $Sp(1)$).    Up to some constant the  $Sp(N)$
1146: Hamiltonian is
1147: \begin{equation}
1148: 	\mathcal{H}=-\frac{1}{N} \sum_{ij} J_{ij} A_{ij}^\dagger A_{ij}
1149: \end{equation}
1150: with the constraints
1151: \begin{equation}
1152: 	\forall i\;\;
1153: 	\sum_{\sigma=1}^{2N} b^{\dag}_{i\sigma}\;b_{i\sigma}=n_c
1154: 	\label{eq:constraint}
1155: \end{equation}
1156: $n_c=2S$ in the $SU(2)$ case and  $n_c/N=\kappa$ will be kept constant
1157: when  taking  the  large-$N$  limit.  The  partition function  can be
1158: represented by an imaginary time functional integral:
1159: \begin{equation}	
1160: 	Z = \int \mathcal{D}[\lambda_i,b_i^\sigma,b_i^{\dag\sigma}]
1161: 	\exp{\left(-\int_0^\beta (L_0+\mathcal{H})d\tau\right)}
1162: 	\label{eq:Z}
1163: \end{equation}
1164: \begin{eqnarray}
1165: 	L_0&=&\sum_{i \sigma}
1166: 		b^{\dag\sigma}_i ( \partial_\tau+i\lambda_i) b_{i\sigma}
1167: 		-iN\kappa\sum_i \lambda_i
1168: 	\label{eq:lL0}
1169: \end{eqnarray}
1170: The $\lambda_i(\tau)$   are  Lagrange multipliers  that   enforce  the
1171: constraint  Eq.~\ref{eq:constraint} at  every  site.   Bond degrees of
1172: freedom  $Q_{ij}$  are introduced   in order  to  decouple the  bosons
1173: (Hubbard-Stratonovitch).  The partition function is now
1174: \begin{equation}	
1175: 	Z = \int \mathcal{D}[Q_{ij},\bar{Q}_{ij},\lambda_i,b_i^\sigma,b_i^{\dag\sigma}]
1176: 	\exp{\left(-\int_0^\beta ( L_0 + L_1) d\tau\right)}
1177: 	\label{eq:ZQ}
1178: \end{equation}
1179: with
1180: \begin{equation}	
1181: 	L_1 = \sum_{ij} \left[
1182: 		N \frac{ | Q_{ij} |^2}{J_{ij}}
1183: 		- (A_{ij}^\dagger Q_{ij} + {\rm h.c})
1184: 		\right]\;,
1185: \end{equation}
1186: so that   a   Gaussian  integration  of   the    $Q_{ij}$  gives  back
1187: Eq.~\ref{eq:Z}. The bond variable  are $Sp(N)$ invariant and they  can
1188: take  non-zero  expectation  values  at   a mean-field level   without
1189: breaking the $Sp(N)$ symmetry.  As we explain  below, they are however
1190: not {\em gauge-invariant}.
1191: 
1192: \subsubsection{Gauge invariance}
1193: An   important property   of  Eq.~\ref{eq:ZQ}   is  the $U(1)$   gauge
1194: invariance associated to the following transformations:
1195: \begin{eqnarray}
1196: 	b_{i\sigma}&\to&b_{i\sigma}e^{i\phi_i} \label{eq:phi}\\
1197: 	Q_{ij} &\to& Q_{ij} e^{i(\phi_i+\phi_j)} \label{eq:Qtheta}\\
1198: 	\lambda_i &\to& \lambda_i -\partial_\tau \phi_i \label{eq:lambdatheta}
1199: \end{eqnarray}
1200: where  $\phi_i(\tau)$ are arbitrary site- and imaginary-time-dependent
1201: angles.  This gauge invariance comes from the  the conservation of the
1202: local boson number and reflects the fact the magnitude of the spin
1203: is constant at each site. If we focus on the  phase degrees of freedom
1204: of  the bond variables, the    Eq.~\ref{eq:ZQ} describes a system   of
1205: charge-1    bosons     coupled    to    a    $U(1)$     lattice  gauge
1206: theory.\cite{wilson74,bdi74}  These gauge  degrees  of freedom play a
1207: crucial  role in  the analysis  of  the fluctuations  about mean-field
1208: solutions.
1209: 
1210: {\em Effective action for the bond variables} --- The boson degrees of
1211: freedom can be integrated out to give an effective action for the bond
1212: variables:
1213: \begin{equation}	
1214: 	Z = \int \mathcal{D}[Q_{ij},\bar{Q}_{ij},\lambda_i]
1215: 	\exp{\left(-S^{\rm eff} \right)}
1216: \end{equation}
1217: \begin{equation}	
1218: 	S^{\rm eff}=N \int_0^\beta \left[
1219: 		\sum_{ij} \frac{ | Q_{ij} |^2}{J_{ij}}
1220: 		-i\kappa \sum_i \lambda_i
1221: 	\right]
1222: 	- N  {\rm Tr} \log G
1223: 	\label{eq:Seff}
1224: \end{equation}
1225: where $G^{-1}$  is  the  quadratic form which  couples   the bosons in
1226: Eq.~\ref{eq:ZQ} (propagator).  It depends on the bond variables and on
1227: $\lambda_i$.             We     may          write            formally
1228: $G^{-1}=\partial_\tau-i\lambda-Q$. The term $N {\rm Tr} \log G$ is the
1229: free energy    of the  bosons  in  presence  of   the bond  fields. By
1230: construction the action $S^{\rm  eff}$ is gauge-invariant with respect
1231: to  the transformations of  Eqs.~\ref{eq:phi}--\ref{eq:lambdatheta}.
1232: So far  this   is an exact   formulation of  the  original  model  for
1233: arbitrary $N$.
1234: 
1235: 
1236: \subsubsection{Mean-field ($N=\infty$  limit)  } Since $N$  factorizes  (no
1237: flavor index  is left in Eq.~\ref{eq:Seff}),  $Z$  is dominated by the
1238: saddle  point of $S^{\rm eff}$ when  $N$ is  large.  For simple models
1239: such as the  first-neighbor antiferromagnet on  the cubic lattice (any
1240: space dimension), the saddle point can be determined analytically. The
1241: $N=\infty$  limit    is almost   equivalent   to   the Schwinger-boson
1242: mean-field theory.\cite{schwingerboson,ceccatto} This can otherwise be
1243: done numerically.\footnote{To our   knowledge, all  the saddle  points
1244: considered  so far are static  (expectation values of the $Q_{ij}$ are
1245: time-independent) and  the  corresponding $\left<Q_{ij}\right>$  could
1246: all be made  real with an  appropriate gauge transformation.  There is
1247: no  chiral order  and  the  time-reversal  symmetry is  unbroken.  The
1248: (oriented) sum of  the complex phases of the  bond variables  around a
1249: plaquette defines a  $U(1)$ flux.  This flux is  related to  the solid
1250: angle formed by the spins  and it vanishes  in collinear as well as in
1251: coplanar states.  In such cases the phases  can be therefore be gauged
1252: away and the $\left<Q_{ij}\right>$ can be  made real.  For this reason
1253: complex bond variables are  usually not observed.\cite{ot03} } In this
1254: large-$N$  limit, two kinds of mean-field   solutions can appear.  For
1255: large  enough  $\kappa$ the  bosons condense  at some wave-vector, the
1256: spectrum of the  mean-field Hamiltonian is gapless.  This  corresponds
1257: physically to N\'eel long-range order.  On the other hand, for smaller
1258: $\kappa$  (smaller ``spin'') the  mean field Hamiltonian is gapped and
1259: the ground-state preserves  the $Sp(N)$ symmetry.  Fluctuations around
1260: the saddle  point are not   expected to change drastically  the N\'eel
1261: ordered phases  but  they play  an important  role in   the physics of
1262: $Sp(N)$ symmetric phases.  The following  is a brief discussion of the
1263: effects of fluctuations in these non-magnetic phases.
1264: 
1265: 
1266: \subsubsection{Fluctuations about the mean-field solution} At the mean-field
1267: level described  above some  $Q_{ij}$ acquire a  (static in  all known
1268: cases) non-zero expectation value: $\left<Q_{ij}\right>=\bar{Q}_{ij}$.
1269: For  this reason  such a  state spontaneously  breaks the  local gauge
1270: invariance        of       Eqs.~\ref{eq:phi},\ref{eq:Qtheta}       and
1271: \ref{eq:lambdatheta}.   However  this does  not  mean  that the  gauge
1272: degrees are  all gapped and  do not play  any role at low  energy.  In
1273: fact, as remarked by Read and Sachdev, depending on the {\em geometry}
1274: of  the  lattice  defined   by  the  bonds  where  $Q_{ij}\ne0$,  some
1275: long-wavelength gapless  gauge excitations survive  and the associated
1276: fluctuations  must  be  taken   into  account.   More  precisely,  the
1277: fluctuations  of  the  bond  variables  about  the  saddle  point  are
1278: decomposed into an amplitude and a phase
1279: \begin{eqnarray}
1280: 	Q_{ij}=\left(\bar{Q}_{ij} + q_{ij}\right)\exp(i\theta_{ij})
1281: 	\label{eq:qtheta}
1282: \end{eqnarray}
1283: and we  expand $S_{\rm eff}$ with these  new variables. Two cases must
1284: then be considered:
1285: \begin{itemize}
1286: \item[i)]  The  lattice  made  of  the  sites  connected  by  non-zero
1287: $\bar{Q}_{ij}$ bonds is bipartite.  This is  automatically the case if
1288: the original lattice defined  by bonds where the exchange $J_{ij}\ne0$
1289: is  bipartite.  This can   also be true  if the  original lattice is a
1290: non-bipartite lattice but some bonds have $\bar{Q}_{ij}=0$ so that the
1291: remaining lattice  is bipartite.  This  is the case,  for instance, in
1292: the  $J_1$--$J_2$  model on  the   square lattice,\cite{rs91}  in some
1293: regions    of   the     phase  diagram   of   the   Shastry-Sutherland
1294: model\cite{cms01}    and  in  on      the   checkerboard    Heisenberg
1295: model.\cite{mts01} Such configurations of the bond variables give {\em
1296: collinear}   spin  structures: spin-spin  correlations  can  either be
1297: long-ranged (large $\kappa$, N\'eel phase) or short-ranged but in both
1298: cases the magnetic structure factor is  peaked at a simple wave-vector
1299: ${\bf k}_0$   such that $2{\bf k}_0$ is    a reciprocal lattice vector
1300: ($k_0=(\pi,0)$,   $k_0=(0,\pi)$    or     $k_0=(\pi,\pi)$ in    square
1301: geometries).
1302: \item[ii)]
1303: The  lattice  made of the sites   connected by non-zero $\bar{Q}_{ij}$
1304: bonds  is   not  bipartite.  This    happens  in some   phases  of the
1305: $J_1$--$J_2$--$J_3$  model  on the square  lattice,\cite{rs91}  on the
1306: triangular  or         kagome    lattices\cite{sachdev92},   in    the
1307: Shastry-Sutherland model      for  some    values of     the  exchange
1308: parameters\cite{cms01}     and    on    an    anisotropic   triangular
1309: lattice.\cite{cmm01}  Such mean-field  states  generically have planar
1310: but non-collinear spin-spin correlations.
1311: \end{itemize}
1312: It is simple to check that case i) preserve a {\em global} continuous
1313: symmetry while such a symmetry is absent in ii). Consider the
1314: following {\em global} gauge transformation in case i) :
1315: \begin{eqnarray}
1316: 	b_{i\sigma}\to b_{i\sigma}e^{i\phi} \;&&\;
1317: 	b_{j\sigma}\to b_{j\sigma}e^{-i\phi} \label{eq:phiij} \\
1318: 	Q_{ii'} \to Q_{ii'} e^{2i\phi}  \;&&\;
1319: 	Q_{jj'} \to Q_{jj'} e^{-2i\phi}\label{eq:Q2}\\
1320: 	Q_{ij} &\to& Q_{ij} \label{eq:Qij}
1321: \end{eqnarray}
1322: where  $i,i'$  belongs  to   sublattice  $A$  and  $j,j'$  belongs  to
1323: sublattice $B$.   This transformation  does not change  the mean-field
1324: parameters   $\bar{Q}_{ij}$.   The  only   fields  affected   by  this
1325: transformation   are  those   connecting   two  sites   on  the   same
1326: sublattice.\footnote{In  a gauge theory  language the  fields $Q_{ii'}$
1327: (resp.  $Q_{jj'}$) of Eq.~\ref{eq:Q2} transform like a charge-2 scalar
1328: for  the $U(1)$  gauge field.   Instead, from  Eq.~\ref{eq:phiij}, the
1329: bosons (spinons)  carry a  charge 1.}  They  have a  zero expectation
1330: values in  case i) (or  do not even  exist if the physical  lattice is
1331: itself bipartite).  For this reason  it is possible to make low-energy
1332: (and long-wavelength) gapless gauge excitations about the saddle-point
1333: by   replacing   the   global   staggered  phase   shift   $\phi$   of
1334: Eq.~\ref{eq:phiij}  by  a  slowly  varying  (staggered)  $\theta_{ij}$
1335: (Eq.~\ref{eq:qtheta}).  A  gradient expansion of  the effective action
1336: performed  at the  appropriate points  in the  Brillouin zone  for the
1337: phase   fluctuations  only  involves   gradients  of   $\theta$.   The
1338: corresponding action is that of $U(1)$ lattice gauge theory coupled to
1339: charge-1 boson (spinons).
1340: 
1341: 
1342: \subsubsection{Topological effects - instantons and spontaneous dimerization}
1343: So far only small fluctuations around the saddle point were considered
1344: and   the  contribution   of  topologically   non-trivial  gauge-field
1345: configurations  were  ignored.   Consequently,  the magnitude  of  the
1346: ``spin'' $n_c/2$  was a continuous parameter and  {\em the information
1347: about the integer  or half-integer (for instance) character  of $S$ as
1348: disappeared}.   From  Haldane's  work   on  quantum  spin  chains  and
1349: non-linear sigma  models\cite{haldane83} it  is well known  that Berry
1350: phases  in  spin  systems  give  rise  to  topological  terms  in  the
1351: low-energy effective action which can play a crucial role depending on
1352: the parity of $2S$.
1353: 
1354: In non-linear sigma models in  2+1 dimensions the Berry phase vanishes
1355: for  configurations which  are  smooth  on the  scale  of the  lattice
1356: spacing\cite{haldane88,noHopfTerm2D88}  (unlike  the 1+1  dimensional
1357: case).  However  ``hedgehog'' space-time singularities\cite{haldane88}
1358: give  non-trivial  Berry phases.   Read  and  Sachdev  found that  the
1359: closely  related  instantons  of  the effective  $U(1)$  gauge  theory
1360: described  before\footnote{In  that  gauge  theory associated  to  the
1361: phases of the  link variables an instanton corresponds  to a tunneling
1362: event during  which the total  magnetic field piercing the  lattice is
1363: changed by $\pm2\pi$.}  also play a crucial role in the physics of the
1364: $Sp(N)$ (as well as $SU(N)$) spin models.
1365: 
1366: The Berry phase associated to such a singular configuration depends on
1367: details   of  the  lattice  geometry.    In  the   short-range ordered
1368: $(\pi,\pi)$ phase of the   square lattice antiferromagnet  this  Berry
1369: phase is  a multiple of $in_c\pi/2$.   Although dealing  with a gas of
1370: interacting ($1/r$ Coulomb-like  potential) instantons is a  difficult
1371: problem (see  Ref.~\cite{rs90}  and references therein),  we can guess
1372: that  the  physics will   depend on  $n_c\;{\rm mod}\;4$.  A  detailed
1373: analysis\cite{rs90} shows  that   when   $n_c\ne0\;{\rm mod}\;4$   the
1374: instantons  condense and   {\em   spontaneously  break  the    lattice
1375: translation symmetry}.  This generates a static electric field for the
1376: $U(1)$   gauge field.   Since  the  electric field  is  coupled to the
1377: difference  of amplitudes of the  bond  variables, such state acquires
1378: spatially inhomogeneous expectation values of the bond variables, {\em
1379: it  is a  VBC  and spinons are confined   in pairs}.  In the $J_1$--$J_2$
1380: model around $J_2/J_1\simeq0.5$ the   mean-field state is  short-range
1381: ordered with correlations peaked at $(\pi,\pi)$.  A columnar dimerized
1382: state is  predicted by Read  and Sachdev   from this analysis  of  the
1383: fluctuations, in agreement  with  a number  of numerical works  on the
1384: $SU(2)$ $J_1$--$J_2$ spin-$\frac{1}{2}$ model.
1385: 
1386: In a recent work by Harada, Kawashima and Troyer\cite{hkt03} the phase
1387: diagram of the (first neighbor - unfrustrated) $SU(N)$ antiferromagnet
1388: on  the  square  lattice  with $n_c=1$ was  found   to be in  complete
1389: agreement with Read and Sachdev's predictions.  They showed by quantum
1390: Monte Carlo  simulations   that for $N\leq4$   the  systems  is N\'eel
1391: ordered whereas  it  is a columnar   VBC for $N>5$.   This provides an
1392: additional support to the field theory  arguments described above.  It
1393: also underlines that   the mechanism of spontaneous  symmetry breaking
1394: and formation of a VBC  may  come from  quantum fluctuations only  and
1395: that frustration is not  always required (although it clearly enhances
1396: quantum fluctuations).
1397: 
1398: On the other hand when $n_c=0\;{\rm mod}\;4$ the  analysis of Read and
1399: Sachdev shows that fluctuations   should not bring any broken  lattice
1400: symmetry.  Spinons are also  confined and this state closely resembles
1401: the     valence-bond   solid (VBS)  proposed     by   Affleck {\it  et
1402: al.}\cite{aklt87} as a possible ground-state when the spin $S$ matches
1403: the  coordination  number  $z$   according  to $2S=0$   mod  $z$  (see
1404: \S\ref{ssec:VBS}).
1405: 
1406: 
1407: \subsubsection{Deconfined phases}
1408: Now  we suppose    that,  starting from   a mean-field  solution  with
1409: collinear correlations (case i),   a  parameter of the  original  spin
1410: model is varied  so that the mean field  solution is changed  and some
1411: bonds $Q_{ii'}$ ($i$ and $i'$ belong to the same sublattice) acquire a
1412: non zero expectation   value (case ii).  In  the  framework of  square
1413: lattice  antiferromagnets,  a finite  third-neighbor coupling  ($J_3$)
1414: would  be needed.\cite{rs91}  From   the   point   of  view  of    the
1415: long-wavelength gauge fluctuations (related  to the continuum limit of
1416: the  phases  $\theta_{ij}$)   discussed   above,  the appearance    of
1417: $\bar{Q}_{ii'}\ne0$ is  equivalent  to the condensation   of a (Higgs)
1418: boson of charge 2.   This is a spontaneous  break  down of  the global
1419: $U(1)$ staggered symmetry of Eqs.~\ref{eq:phiij}--\ref{eq:Qij} down to
1420: a    $\Z$ one   since the   field   $Q_{ii}$ is   not invariant  under
1421: Eq.~\ref{eq:Q2}  except if  $\phi=0$  or $\pi$.   Based on  results of
1422: Fradkin  and Shenker\cite{fs79}   concerning  confinement  in  compact
1423: lattice gauge theories coupled to matter, Read and Sachdev argued that
1424: this Higgs mechanism  suppresses the low-energy gauge fluctuations and
1425: liberate the spinons.  This  confinement transition is described by  a
1426: $\Z$ gauge theory.  The suppression  of the $U(1)$ gauge  fluctuations
1427: also forbids  the condensation of instantons   discussed above and the
1428: ground-state remains unique\footnote{Except for a discrete topological
1429: degeneracy.} and  {\em   bond  variables    have  uniform  expectation
1430: values.  It is a genuine SL  without any broken symmetry and
1431: deconfined spinons.}
1432: 
1433: %_____________________________________________________________________
1434: 
1435: \section{Quantum Dimer Models}\label{sec:QDM}
1436: \setcounter{footnote}{0}
1437: %_____________________________________________________________________
1438: 
1439: In  a previous  section   (\S\ref{sec:VBC})  we  showed  that  pairing
1440: spins-$\frac{1}{2}$   into singlets at  short   distances is a  rather
1441: natural    way     to     overcome   frustration     in     Heisenberg
1442: antiferromagnets.  QDM   are    defined  in the     Hilbert   space of
1443: nearest-neighbor valence-bond (or dimer) coverings of the lattice.  By
1444: construction these models focus  on the dynamics  in the singlet space
1445: and     ignore magnetic   (gapped      magnons   or gapped    spinons)
1446: excitations. For  this reason they  are (a priori)  not appropriate to
1447: describe  the physics of spin  systems  where magnetic excitations are
1448: gapless.
1449: 
1450: The Hamiltonian of a QDM usually contains kinetic as well as potential
1451: energy terms for these dimers.  Such Hamiltonians can often be simpler
1452: than   their   spin parents  and  are   amenable  to  several analytic
1453: treatments  because of their    close  relations to  classical   dimer
1454: problems,\cite{k61,f61,k63}    Ising     models   and     $\Z$   gauge
1455: theory.\cite{msf02,msp02,ms02}    These   models  can   offer   simple
1456: descriptions    of       VBC\cite{rk88}   as      well      as     RVB
1457: liquids.\cite{ms01,msp02} It is  in particular possible to write  down
1458: some QDM    that  have a   simple   and exact VBC    ground-state with
1459: spontaneous broken symmetries (such as Rokhsar and Kivelson's model on
1460: the square lattice\cite{rk88}  with attractive potential energy only -
1461: in which case the exact ground-state is very simple).  Simple solvable
1462: QDM    which    have  a   dimer-liquid   ground-state   can   also  be
1463: constructed.\cite{msp02}
1464: 
1465: \subsection{Hamiltonian}
1466: 
1467: \newcommand{\hh}{\begin{picture}(13,9)(-2,2)
1468: 	\put (0,0) {\line (1,0) {8}}
1469: 	\put (8,8) {\line (-1,0) {8}}
1470: 	\put (0,0) {\circle*{3}}
1471: 	\put (0,8) {\circle*{3}}
1472: 	\put (8,0) {\circle*{3}}
1473: 	\put (8,8) {\circle*{3}}
1474: 	\end{picture}
1475: }
1476: \newcommand{\vv}{\begin{picture}(13,9)(-2,2)
1477: 	\put (0,0) {\line (0,1) {8}}
1478: 	\put (8,8) {\line (0,-1) {8}}
1479: 	\put (0,0) {\circle*{3}}
1480: 	\put (0,8) {\circle*{3}}
1481: 	\put (8,0) {\circle*{3}}
1482: 	\put (8,8) {\circle*{3}}
1483: 	\end{picture}
1484: }
1485: 
1486: The first QDM was introduced by Rokhsar and Kivelson.\cite{rk88} It is
1487: defined   by  an   Hamiltonian  acting   in  the   Hilbert   space  of
1488: first-neighbor dimer  (valence-bonds) coverings of  the square lattice
1489: and reads:
1490: \begin{equation}
1491: \mathcal{H}=\sum_{\rm Plaquette}
1492: \left[
1493: -J\left(\left|\vv\right>\left<\hh\right| +{\rm H.c.}\right)
1494: +V\left(\left|\vv\right>\left<\vv\right|+\left|\hh\right>\left<\hh\right|\right)
1495: \right]
1496: \label{eq:sQDM}
1497: \end{equation}
1498: 
1499: Flipping two parallel dimers around a square plaquette is the simplest
1500: dimer move on the square lattice and the $J$ terms precisely represent
1501: such dynamics.  The $V$   terms are diagonal in   the dimer basis  and
1502: account for an    attraction or  repulsion between    nearest-neighbor
1503: dimers.   These are the two  {\em  most local}  terms (respecting  all
1504: lattice  symmetries) which   can  be   considered.\footnote{They  were
1505: originally  derived\cite{rk88} as the lowest  order  terms of a formal
1506: overlap     expansion  (see \S\ref{ssec:spin2QDM}     below)   of  the
1507: Heisenberg. In that  calculation $J\sim x^4$  and $V\sim x^8$.  Notice
1508: that  a  three-dimer kinetic term     (extending over two  neighboring
1509: plaquettes)  is generated  at order    $x^6$ and  is not included   in
1510: Eq.~\ref{eq:sQDM}.}
1511: 
1512: 
1513: \subsection{Relation with spin-$\frac{1}{2}$ models}
1514: \label{ssec:spin2QDM}
1515: 
1516: There exists  different interesting  mappings between frustrated  Ising
1517: models  and QDM.\cite{msc00,ms02} Here,     however, we focus   on the
1518: relations between  QDM and $SU(N)$ (or $Sp(N)$)  spin models  in which
1519: dimers are related to singlet valence-bonds.
1520: 
1521: {\em Overlap expansion}. --- A valence-bond state (product of two-spin
1522: singlets -  belongs  to the spin-$\frac{1}{2}$  Hilbert space)  can be
1523: associated  to any dimer   covering.\footnote{There is however  a sign
1524: ambiguity (a valence-bond is antisymmetric under  the exchange of both
1525: spins)  that can be fixed by  choosing  an orientation on every bond.} 
1526: Two such valence-bond states $\left|a\right>$ and $\left|b\right>$ are
1527: not orthogonal but,  as first discussed by  Sutherland\cite{s88} their
1528: overlap decays exponentially with the length $L$ of the loops of their
1529: transition    graphs     (defined    in     \S\ref{sssec:trgr})     as
1530: $|\left<a|b\right>|=2^{1-L/2}$.   Rokhsar   and    Kivelson\cite{rk88}
1531: introduced  a    formal    expansion  parameter   $x$    and  replaced
1532: $|\left<a|b\right>|$  by   $2x^{L}$.   Their  idea  is that   although
1533: $x=\frac{1}{\sqrt{2}}$ for physical $SU(2)$ spins, the physics of some
1534: models may     be captured  by  the   first  orders  of  a  small  $x$
1535: expansion. Truncating this expansion to order $x^n$ gives an effective
1536: Hamiltonian which contains {\em  local}  terms involving at most   $n$
1537: dimers.\footnote{The {\em signs}  of the non-diagonal  (kinetic) terms
1538: of the effective QDM  obtained by such small  $x$ expansion depends on
1539: the   sign  convention which  was  chosen   to  map   valence-bonds to
1540: dimers. An important  question  is to know  whether,  at least at  the
1541: lowest non-trivial  order, a sign convention  giving the same sign for
1542: all kinetic terms exists (as in  Eq.~\ref{eq:sQDM}).  This is the case
1543: on the square\cite{rk88}  and triangular lattices.\cite{ms01}} In this
1544: approach the dimer states of the QDM  are in one-to-one correspondence
1545: with    {\em  orthogonalized}   valence-bonds   in   the spin  Hilbert
1546: space.\footnote{This implicitly assumes  that the  valence-bond states
1547: are  linearly  independent.  This  can be  demonstrated on  the square
1548: lattice and appears   to be the   case on  the triangular  and  kagome
1549: lattices for large enough sizes. The operator which orthogonalizes the
1550: valence-bond basis  into  the  dimer  basis is   $\Omega^{-1/2}$ where
1551: $\Omega_{a,b}=\left<a|b\right>$ is the overlap matrix.}
1552: 	 
1553: {\em Fluctuations about large-$N$ saddle points}.--- From the argument
1554: above it could   seem that the connection between   spin-$\frac{1}{2}$
1555: models and QDM relies on a variational approximation: the spin Hilbert
1556: space  is   restricted     to   the   nearest-neighbor    valence-bond
1557: subspace. This connection is in fact probably deeper, as some theories
1558: describing  fluctuations  about  some  large-$N$ saddle points  ($1/N$
1559: corrections) are equivalent to  (generalized)  QDM.  This mapping  was
1560: discussed by  Read   and Sachdev\cite{rs89b} for  representation  with
1561: $m=1$ (number  of   rows  in  the  Young    tableau of  the    $SU(N)$
1562: representation), $N\to\infty$ and $n_c\sim\mathcal{O}(1)$, it leads to
1563: a {\em generalized} QDM where $n_c$ dimers emanate  from each site.  A
1564: QDM  also    describes $1/N$  corrections      in  the case  of    the
1565: fermionic\footnote{Young  tableau with    $n_c=1$ column  and  $m=N/2$
1566: rows.}   $SU(N)$ generalization\cite{am88,rokhsar90} of the Heisenberg
1567: model:
1568: \begin{eqnarray}
1569:   \mathcal{H}&=&\frac{1}{N}\sum_{ij} J_{ij}:B_{ij}^\dagger B_{ij}: \nonumber \\
1570:   &=& -\frac{1}{N}\sum_{ij} J_{ij}\; B_{ij}^\dagger B_{ij} +{\rm cst}
1571:   \label{eq:HSUNf} \\
1572: 	{\rm where}\;\;B_{ij}&=& \sum_{\sigma=1}^N c_{j\sigma}^\dagger c_{i\sigma}
1573: \end{eqnarray}
1574: and where  the $c_{i\sigma}$ are  $N$ flavors of  fermions satisfying
1575: a  constraint similar to Eq.~\ref{eq:constraint} : 
1576: \begin{equation}
1577:   \sum_{\sigma=1}^N c_{i\sigma}^\dagger c_{i\sigma}=N/2
1578:   \label{eq:constraintF}
1579: \end{equation}
1580: Rokhsar   showed\cite{rokhsar90}  that   in  the   $N\to\infty$  limit
1581: ``dimerized states'' (or  Peierls states) becomes exact  ground-states
1582: of Eq.~\ref{eq:HSUNf} for a large class of models.\footnote{ Let $J_0$
1583: be  the largest value of  the $J_{ij}$. Each  dimerization (no site is
1584: left   empty) where only bonds  where  $J_{ij}=J_0$ are  occupied is a
1585: ground-state.  Here a  ``dimer''   between to neighbors  $i$  and  $j$
1586: consists of a  $SU(N)$  singlet made  with $N$  fermions (one of  each
1587: flavor) hoping back and forth between $i$ and  $j$.  It is constructed
1588: from            $\prod_{\sigma=1}^N          \left(c_{i\sigma}^\dagger
1589: +c_{j\sigma}^\dagger\right)  \left|0\right>$.   by projecting  out the
1590: components  which do  not  satisfy  Eq.~\ref{eq:constraintF}.   Notice
1591: however that in  the $N\to\infty$ limit  the relative  fluctuations of
1592: the total number of fermion on each site are of order $1/\sqrt{N}$ and
1593: can  be neglected.}  Quite  naturally, $1/N$ corrections will induce a
1594: dynamics into this subspace of  dimerized states; it can be  described
1595: by a  QDM (with kinetic  energy terms only  at this order).  At lowest
1596: order, on the square lattice, a kinetic term identical to the $J$ term
1597: in Eq.~\ref{eq:sQDM} is generated and  naturally favors a columnar  or
1598: resonating-plaquette  crystal (in agreement with  a number of works on
1599: the spin-$\frac{1}{2}$ model).  The  same arguments were discussed for
1600: the kagome    lattice.\cite{mz91}  In  that  case the    leading $1/N$
1601: corrections to  the fermionic saddle  point generate three-dimer moves
1602: around     hexagons and      stabilize  a   crystal     of  resonating
1603: hexagons.\footnote{To our knowledge however there is no clear evidence
1604: of such ordering in  the spin-$\frac{1}{2}$ case.} This  formalism was
1605: also applied to the checkerboard  model\cite{canals02} to conclude  to
1606: the presence of a VBC phase.
1607: 
1608: \subsection{Square lattice}
1609: 
1610: The phase diagram of the  Rokhsar and Kivelson's square lattice QDM is
1611: shown Fig.~\ref{fig:sQDM}.  Since  a change in the signs  of the basis
1612: dimer   configurations   can   change  $J$   into   $-J$   (see
1613: Ref.~\cite{rk88})   we  will  choose   $J>0$  without   lost  of
1614: generality.
1615: \begin{figure}
1616:   \begin{center}
1617:     \includegraphics[height=3cm]{QDMsquare}
1618:     \caption[99]{Schematic  phase diagram  of the  the  square lattice
1619:       QDM.  The  possible location of an  intermediate plaquette phase
1620:       is taken from the work of Leung {\it et al}.\cite{lcr96}}
1621:     \label{fig:sQDM}
1622:   \end{center}
1623: \end{figure}
1624: 
1625: \subsubsection{Transition graphs  and topological sectors}
1626: \label{sssec:trgr}
1627: 
1628: In     order to understand the  particularities    of QDM on bipartite
1629: lattices it  is useful to   describe how the  set of  dimer  coverings
1630: splits into {\em  topological  sectors}. To do   so we first   have to
1631: define transition graphs: the transition  graph of two dimer coverings
1632: $c$ and $c'$ is obtained  by superimposing $c$ and  $c'$ on the top of
1633: each other; it  defines a set  of non-intersecting loops covering  the
1634: lattice.  On  each bond  where the dimers  of $c$   and  $c'$ match, a
1635: trivial loop of  length 2 is obtained.   When the lattice is bipartite
1636: (two sublattices $A$ and $B$) these loops can be {\em oriented} in the
1637: following way: any dimer belonging to $c$ is  oriented from $A$ to $B$
1638: and dimers  of $c'$ are  oriented $B\to A$.   The  transition graph is
1639: then  made of  loops of  the   type $A\to  B\to  A\to  B\cdots$.  With
1640: periodic  boundary     conditions    two   winding  numbers\cite{rk88}
1641: $-L/2\leq\Omega_x,\Omega_y\leq    L/2$   are  associated  to    such a
1642: transition graph ($L\times  L$ sites).  $\Omega_x$ (resp.  $\Omega_y$)
1643: is the net number of  topologically non-trivial loops (clockwise minus
1644: counterclockwise) encircling  the     torus in the    $x$ (resp.  $y$)
1645: direction.
1646: 
1647: Dimer coverings can   be   grouped into different   {\em   topological
1648: sectors}. By definition two dimer coverings belong  to the same sector
1649: if they can be transformed into each other by repeated actions of {\em
1650: local} dimer moves (the  transition graph associated to each  movement
1651: does not  wind  around the whole  system  if it has  periodic boundary
1652: conditions).  On the square lattice  two-dimer moves are sufficient to
1653: connect any two  states in the  same sector;  that  is the Hamiltonian
1654: Eq.~\ref{eq:sQDM}  is ergodic  within  each topological  sector.  In a
1655: torus geometry, $c$ and $c'$ belongs to the same sector if and only if
1656: their transition graph has winding numbers $\Omega_x=\Omega_y=0$.  The
1657: different topological sectors can  be labeled by their winding numbers
1658: with respect to some  reference columnar configuration.  Their  number
1659: is of order $\mathcal{O}(L^2)$ for a system of linear size $L$.
1660: 
1661: 
1662: \subsubsection{Staggered VBC for $V/J> 1$}
1663: When $V$ is sufficiently large the system tries to minimize the number
1664: of    parallel   dimers.     The    staggered   configuration    shown
1665: Fig.~\ref{fig:sQDM}  has  no such  {\em  flippable  plaquette}. It  is
1666: always  a zero-energy  eigenstate  of Eq.~\ref{eq:sQDM}  and becomes  a
1667: ground-state  for $J\geq  V$.   It breaks  several lattice  symmetries
1668: (four-fold degenerate) and is a VBC.
1669: 
1670: The  expectation value  of the  energy per  plaquette  satisfies ${\rm
1671: min}(0,V-J)\leq E_0/N_p\leq {\rm max}(0,V+J)$.  For $V/J>1$ this gives
1672: $0 \leq E_0/N_p$ and any  zero-energy state saturates this lower bound
1673: and is a therefore a  ground-state.  One should however notice that it
1674: is possible  to make  zero-energy domain walls  in this VBC  since the
1675: state   shown  Fig.~\ref{fig:stagg2}  is   also  annihilated   by  the
1676: Hamiltonian. No local  dimer movement can take place  in the staggered
1677: VBC  (with or  without domain  walls).  Each  of these  states  form a
1678: topological   sector   with   a   single   configuration   which   has
1679: $\left|\Omega_x\right|+\left|\Omega_y\right|=L/2$  with  respect to  a
1680: columnar state.
1681: 
1682: \begin{figure}
1683:   \begin{center}
1684:     \includegraphics[height=3cm]{stagg2}
1685:     \caption[99]{Domain wall in a staggered VBC.}
1686:     \label{fig:stagg2}
1687:   \end{center}
1688: \end{figure}
1689: 
1690: \subsubsection{Columnar crystal for $V<0$}
1691: 
1692: When parallel dimers  attract each other the system  tries to maximize
1693: the number  of flippable plaquettes. Columnar  configurations as shown
1694: on the left side of Fig.~\ref{fig:sQDM} do maximize this number.  Such
1695: a VBC is exactly realized  for $V<0$ and $J=0$. Elementary excitations
1696: consist of a pair of (say) vertical dimers in a background of vertical
1697: columns  of horizontal  dimers. Such  excitations are  gapped ($\Delta
1698: E=2|V|$) and this VBC phase will  survive to the inclusion of a finite
1699: $J$  term.  Notice  that unlike  the  staggered VBC  presented in  the
1700: previous paragraph  the columnar dimer  configuration is not  an exact
1701: eigenstate  when $J\ne0$.   The exact  diagonalizations data  of Leung
1702: {\it et al}\cite{lcr96} suggest  that the columnar phase may disappear around
1703: a critical value $V/J\simeq -0.2$.
1704: 
1705: \subsubsection{Plaquette phase}
1706: 
1707: When the kinetic energy dominates  ($|V|\ll J$) the system will try to
1708: maximize   the   number    of   resonating   plaquettes   $\left|   \|
1709: \right>+\left|=\right>$. This  can be achieved  through the resonating
1710: plaquette  crystal shown   Fig.~\ref{fig:sQDM} and  the numerical work
1711: (exact  diagonalizations  up  to  $8\times8$ sites)  of  Leung {\it et
1712: al}\cite{lcr96} suggest  that this phase   is realized in  an interval
1713: $-0.2\leq V/J \leq 1$. Although  this model would  not suffer from the
1714: fermion sign  problem we are  not aware of  any QMC simulation on this
1715: model.
1716: 
1717: \subsubsection{Rokhsar-Kivelson point}
1718: 
1719: The point $J=V$ (Rokhsar-Kivelson (RK) point) plays a special role. As
1720: remarked  by  Rokhsar   and  Kivelson\cite{rk88}  the  equal-amplitude
1721: superposition of all dimerizations in a given topological sector is an
1722: exact ground-state. When $J=V$ the Hamiltonian can be written as a sum
1723: of projectors:
1724: \begin{eqnarray}
1725:   \mathcal{H}_{J=V=1}&=&\sum_{p} \left|\Psi_p\right>\left<\Psi_p\right| \label{eq:sRK}\\
1726:   \left|\Psi_p\right>&=&\left|\vv\right> - \left|\hh\right>
1727: \end{eqnarray}
1728: The linear superposition  of all dimer coverings belonging  to a given
1729: sector $\Omega$
1730: \begin{equation}
1731:   \left|0\right>=\sum_{c\in\Omega}\left| c\right>
1732: \label{eq:RK}
1733: \end{equation}
1734: is annihilated  by Eq.~\ref{eq:sRK}  and is therefore  a ground-state.
1735: The  argument  is the  following.   Consider  a  plaquette $p$  and  a
1736: configuration $\left|  c\right>$. If $\left|  c\right>$ has one  or no
1737: dimer at all  on the edges of $p$  we have $\left<\Psi_p |c\right>=0$.
1738: If two dimers  are present, then there exists  a configuration $\left|
1739: c'\right>$ in the same sector which only differ from $\left| c\right>$
1740: by a  two-dimer flip on $p$.   In such a case  the combination $\left|
1741: c\right>+\left|c'\right>$       is      again       orthogonal      to
1742: $\left|\Psi_p\right>$. This shows that $H\left| 0\right>=0$.
1743: 
1744: When open boundary    conditions are considered    (this restricts the
1745: topological  sector to   $\Omega_x=\Omega_y=0$)  the RK  state is  the
1746: linear     combination      of      an    exponential    number     of
1747: configurations.\footnote{This   is also   true for   periodic boundary
1748: conditions provided the two   winding numbers do   not scale like  the
1749: linear size $L$.}  This is  very different from the crystalline states
1750: considered so far where some  periodic configurations were favored and
1751: it closely resembles Anderson's RVB picture.   As we shall see this RK
1752: state is not a ``true'' liquid on the square lattice since dimer-dimer
1753: correlations are    not  short-ranged  but   algebraically    decaying
1754: ($\sim1/r^2$)  with distance.     The  calculation    of   dimer-dimer
1755: correlations in  the  RK  state (Eq.~\ref{eq:RK})  maps  onto   a {\em
1756: classical    }   dimer  problem   solved  by  Kasteleyn,\cite{k61,k63}
1757: Fisher\cite{f61}    and Fisher and   Stephenson.\cite{fs63}  From this
1758: Rokhsar  and Kivelson\cite{rk88}  constructed gapless excitations  (in
1759: single-mode approximation) which dispersion relation vanishes as ${\bf
1760: k}^2$ at small momentum (the origin is taken at $(\pi,\pi)$).
1761: Quoting Rokhsar and  Kivelson,\cite{rk88}  these excitations  (dubbed
1762: ``resonons'') are the {\em Goldstone mode  of the gauge symmetry which
1763: allows the  phases of the different   topological sectors to  be varied
1764: without changing the energy}.  Another   mode of gapless   excitations
1765: (around $(\pi,0)$ and  $(0,\pi)$  in the Brillouin zone),  specific to
1766: the fact that  the  ground-state has critical (algebraic)  dimer-dimer
1767: correlations, was recently discovered.\cite{ms03}
1768: 
1769: The QDM  on the square  lattice is thus  believed to be  ordered (VBC)
1770: everywhere  except  at  the  RK  point  ($J=V$)  where  it  has  quasi
1771: long-ranged (critical) dimer-dimer  correlations. 
1772: 
1773: \subsection{Hexagonal lattice}
1774: 
1775: The QDM on  the honeycomb lattice was  studied by Moessner, Sondhi and
1776: Chandra,\cite{msc01} in particular  with Monte Carlo simulations.  The
1777: phase   diagram is very similar  to  the square lattice-case discussed
1778: above.  It  possesses three crystalline  phases  (Fig.~\ref{fig:hQDM})
1779: and has algebraically decaying dimer-dimer correlations at the Rokhsar
1780: Kivelson  point (where the ground-state  in  each sector has the  same
1781: form  as Eq.~\ref{eq:RK}).  
1782:  The  absence of liquid phase  (with
1783: exponentially   decaying 2$n$-mer-2$n$-mer  correlations)   is  believed  to
1784: prevail in  bipartite lattices.
1785: This relation  between the absence of a
1786: deconfined dimer  liquid  phase
1787: \footnote{
1788: Bipartiteness  seems to  forbid deconfinement but  not short-ranged
1789: dimer-dimer correlations. The 4-8 lattice (squares and octogons) is an
1790: example  where dimer-dimer correlations   are short-ranged.   We thank
1791: R.~Moessner  for   pointing   this  to    us.  On this   lattice   the
1792: equal-amplitude superposition of all coverings  would be similar to an
1793: explicit VBC wave-function (thus  confining). Such situations are only
1794: possible when the number of sites is even in the  unit cell.}
1795: and the  bipartite character  of   the lattice  as been discussed   by
1796: several  authors\cite{hkms03} and is related  to the  possibility of a
1797: {\em height   representation}\cite{height_representation}    of  dimer
1798: coverings when the lattice is bipartite.\footnote{Consider a bipartite
1799: lattice with coordination number $z$.  For  each dimer covering we can
1800: associate integers  (heights) on  the  dual  lattice by  the following
1801: rule.  Set the  height to be zero  on a plaquette  at the origin.  The
1802: height  is  then defined  on the  whole  lattice  by turning clockwise
1803: (resp.      counterclockwise)  around  sites    of  the $A$-sublattice
1804: (resp. $B$-sublattice) and changing the  height by $z-1$ when crossing
1805: a  dimer and by  $-1$ when crossing  an empty bond.    It is simple to
1806: check the difference  of  heights $\delta h(x)=h_1(x)-h_2(x)$  between
1807: two dimerizations is constant   inside each loop of their   transition
1808: graph.  Notice that  the loops of a  transition graph can be naturally
1809: oriented on a bipartite lattice.   Then, $\delta h(x)$ changes by $+z$
1810: (resp.  $-z$) when crossing a clockwise (resp.  counterclockwise) loop
1811: of the  transition  graph.   Columnar dimerizations  have  an averaged
1812: height  which is flat  and staggered ones have  the maximum tilt.  The
1813: winding numbers $(\Omega_x,\Omega_y)$ correspond to the average height
1814: difference between both sides of the sample.  The kinetic energy terms
1815: of  Eqs.~\ref{eq:sQDM} and   \ref{eq:hQDM}  change the height   of the
1816: corresponding  plaquette  by $\pm z$  and  the potential terms ($V>0$)
1817: favor tilted configurations.}
1818: 
1819: \newcommand{\ha}{
1820:   \begin{picture}(32,14)(-16,-2)
1821:     \Hex\pA{\La}\pC{\Lc}\pE{\Le}
1822:   \end{picture}
1823: }
1824: \newcommand{\hb}{
1825:   \begin{picture}(32,14)(-16,-2)
1826:     \Hex
1827:     \pB{\Lb}\pD{\Ld}\pF{\Lf}
1828:   \end{picture}
1829: }
1830: 
1831: \begin{eqnarray}
1832:   \mathcal{H}=&-J&\sum_h \left(
1833:   \left|\ha\right>\left<\hb\right| +{\rm H.c.}
1834:   \right) \nonumber \\
1835:   &+V&\sum_h\left(
1836:   \left|\ha\right>\left<\ha\right|+\left|\hb\right>\left<\hb\right|
1837:   \right)
1838: \label{eq:hQDM}
1839: \end{eqnarray}
1840: 
1841: Fouet   {\it  et   al.}\cite{fsl01}   studied  the  spin-$\frac{1}{2}$
1842: $J_1$--$J_2$--$J_3$    model   on the   hexagonal   lattice  by  exact
1843: diagonalizations and  found evidences of a  staggered VBC of  the type
1844: predicted for  $V/J>1$ by Moessner  {\it  et al.}\cite{msc01}  in  the
1845: QDM. Other phases  (N\'eel ordered phase  and a possible short-range RVB
1846: SL) are also present in the spin-$\frac{1}{2}$ model.\cite{fsl01}
1847: 
1848: 
1849: 
1850: \begin{figure}
1851:   \begin{center}
1852:     \includegraphics[height=3cm]{QDMhex}
1853:     \caption[99]{  Phase  diagram of  the  hexagonal  QDM obtained  by
1854:       Moessner  {\it et  al.}\cite{msc01} Although  the VBC  shown for
1855:       $V<0$ do not  have all dimers parallel to  the same direction it
1856:       is equivalent  to the columnar  VBC found in the  square lattice
1857:       QDM because it maximizes  the number of flippable plaquettes. It
1858:       also  corresponds  to  the  ordering pattern  predicted  in  the
1859:       large-$N$ theory of Ref.~\cite{rs90} As for the VBC obtained for
1860:       $V/J>1$, it is the  hexagonal counterpart of the {\em staggered}
1861:       VBC  of  the  square  lattice  (no  flippable  plaquette,  exact
1862:       eigenstate and maximum tilt in a height representation).}
1863:     \label{fig:hQDM}
1864:   \end{center}
1865: \end{figure}
1866: 
1867: \subsection{Triangular lattice}
1868: \label{ssec:QDMTri}
1869: 
1870: \newcommand{\rhomb}{
1871:   \pA{\C}\pB{\C}\pZ{\C}\pC{\C}
1872:  }
1873: \newcommand{\rhombF}{
1874:   \begin{picture}(18,10)(-5,-6)
1875:     \rhomb
1876:     \pA{\La}\pB{\Lb}\pZ{\Le}\pC{\Ld}
1877:   \end{picture}
1878: }
1879: \newcommand{\rhombH}{
1880:   \begin{picture}(22,10)(-8,-6)
1881:     \rhomb
1882:     \pA{\La}\pC{\Ld}
1883:   \end{picture}
1884: }
1885: \newcommand{\rhombV}{
1886:   \begin{picture}(22,10)(-8,-6)
1887:     \rhomb
1888:     \pB{\Lb}\pZ{\Le}
1889:   \end{picture}
1890: }
1891: 
1892: The most  local dimer Hamiltonian  on the triangular  lattice contains
1893: kinetic and potential two-dimer terms  on each rhombus; it was studied
1894: by Moessner and Sondhi:\cite{ms01}
1895: \begin{eqnarray}
1896:   \mathcal{H}=&-J&\sum_r \left(
1897:   \left|\rhombV\right>\left<\rhombH\right| +{\rm H.c.}
1898:   \right) \nonumber \\
1899:   &+V&\sum_r\left(
1900:   \left|\rhombV\right>\left<\rhombV\right|+\left|\rhombH\right>\left<\rhombH\right|
1901:   \right)
1902: \label{eq:tQDM}
1903: \end{eqnarray}
1904: where  the sums run over  all  rhombi $r$  of the  lattice (with three
1905: possible orientations).   This model was  shown to  possess (at least)
1906: three crystalline phases,  including a   columnar and staggered    one
1907: (Fig.~\ref{fig:QDMtri}) as in the two previous examples. An additional
1908: VBC  (with resonating diamonds plaquettes)  with a large unit cell (12
1909: sites) was also   predicted around $V=0$.   When $V<0$  and $J=0$  the
1910: ground-state  is highly  degenerate  since it  is   possible, from  an
1911: ordered columnar configuration,   to shift all  the dimers  along  any
1912: straight line   without changing the  number  of  flippable plaquettes
1913: (contrary to the square lattice case).   However, an infinitesimal $J$
1914: is expected to lift this degeneracy and to order the ground-state in a
1915: columnar way.
1916: 
1917: The phenomenology of these ordered phases is that  of usual VBC and we
1918: refer to the original paper\cite{ms01} for details.
1919: 
1920: \begin{figure}
1921: \begin{center}
1922: \includegraphics[width=8.5cm]{QDMTri}
1923: \caption[99]{Phase diagram of the triangular lattice QDM
1924: obtained by Moessner and Sondhi.\cite{ms01}}
1925: \label{fig:QDMtri}
1926: \end{center}
1927: \end{figure}
1928: 
1929: \subsubsection{RVB liquid at the RK point}
1930: 
1931: The new physics of this model appears through  the existence of a {\em
1932: liquid} phase in the interval $0.7\lesssim V/J \leq 1$. As for the two
1933: previous  QDM  the ground-states are  exactly  known at   the RK point
1934: $J=V$. As before dimer-dimer correlations are obtained exactly at this
1935: point by  a Pfaffian calculation\cite{ms01,fms02,iif02} but the result
1936: shows a   {\em  finite} correlation length.    From  their Monte Carlo
1937: simulations Moessner {\it et al.}\cite{ms01}  argued that the spectrum
1938: is   gapped  at the  RK   point and  that  this   gap persists down to
1939: $V/J\simeq 2/3$,  that is  over  a  finite range   of  coupling.  This
1940: picture   is consistent with the   exact diagonalizations performed on
1941: this model.\cite{ifiitb02,mlms02}
1942: 
1943: \subsubsection{Topological order}
1944: 
1945: When the lattice is not bipartite the  loops of a transition graph can
1946: no longer be oriented.   The winding numbers $\Omega_x$ and $\Omega_y$
1947: are now  defined as the (positive)  number  of non-trivial loop around
1948: $x$  and $y$  in the transition  graph  with a reference configuration
1949: (say columnar).   These two integers are not  conserved by local dimer
1950: moves. They are however {\em conserved modulo  two}, which leaves only
1951: four   sectors.\footnote{The        two-dimer moves     included    in
1952: Eq.~\ref{eq:tQDM} are not  sufficient to guaranty ergodicity with each
1953: of the  four  sectors.  Staggered   states  (12-fold degenerate  - not
1954: $\mathcal{O}(L)$ like on   the  square  lattice) have   no   flippable
1955: plaquette  but  can be    connected  to other states with   four-dimer
1956: moves.\cite{ms01}} Consequently {\em the  dimer liquid ground-state is
1957: four-fold degenerate} at the RK  point.  This degeneracy holds exactly
1958: at the RK point even on finite-size samples but it is expected to hold
1959: in the thermodynamic limit in the whole liquid phase ($0.7\lesssim V/J
1960: \leq 1$).
1961: 
1962: Conventional  orders are  often associated  to a  spontaneously broken
1963: symmetry and  lead to  ground-state degeneracies in  the thermodynamic
1964: limit.  The  four-fold degeneracy discussed above is  the signature of
1965: some kind of order, called topological
1966: order.\cite{wen91}  The peculiarity  of this  order  is that it is not
1967: associated  to any local  order  parameter: a local observable  cannot
1968: decide whether a given dimerization is in one  sector or another.  The
1969: existence of topological   order   is intimately associated   to   the
1970: fractionalized nature of the  elementary excitations (see  below).  In
1971: the case of a RVB dimer liquid these excitations have been known to be
1972: {\em  Ising vortices}  for   a long  time\cite{k89,rc89}  (dubbed {\em
1973: visons}  in the   recent  literature\cite{sf00,sf01}).   We  will  now
1974: discuss these excitations in more  details in the  framework of a  QDM
1975: which realizes the same dimer liquid phase but for  which not only the
1976: ground-state but all the eigenstates are known exactly.
1977: 
1978: 
1979: \subsection{Solvable QDM on the kagome lattice}
1980: \label{ssec:QDMKag}
1981: 
1982: An exactly    solvable QDM on  the kagome   lattice  was introduced by
1983: D.~Serban,  V.~Pasquier and one  of  us\cite{msp02}.  It offers a very
1984: simple and explicit realization of  the ideas discussed above (visons,
1985: topological order etc.).
1986: 
1987: \subsubsection{Hamiltonian}
1988: The kagome lattice  QDM introduced in  Ref.~\cite{msp02} contains only
1989: kinetic terms and has no external parameter. The Hamiltonian reads:
1990: \begin{eqnarray}
1991:   \mathcal{H}&=&-\sum_h \sigma^x(h) \label{eq:kQDM}\\
1992: {\rm where }\;\;
1993:   \sigma^x(h)&=&\sum_{\alpha=1}^{32}
1994:   \left|d_\alpha(h) \right>\left< \bar{d}_\alpha(h)\right|
1995:   +
1996:   {\rm H.c}
1997:   \label{eq:sigmax}
1998: \end{eqnarray}
1999: The sum runs over the  32 loops on  the lattice which enclose a single
2000: hexagon    and around      which     dimers can   be     moved    (see
2001: Table~\ref{tab:kloops}  for  the 8 inequivalent loops).   The shortest
2002: loop is the  hexagon itself, it involves  3 dimers.  4, 5 and 6-dimers
2003: moves are also possible  by including 2, 4  and 6 additional triangles
2004: (the  loop length must be  even).  The largest loop is  the star.  For
2005: each loop $\alpha$   we associate the  two  ways dimers can be  placed
2006: along                that      loop:  $\left|d_\alpha(h)\right>$   and
2007: $\left|\bar{d}_\alpha(h)\right>$.    Notice that $\sigma^x(h)$ measures
2008: the relative  phases  of dimer configurations displaying  respectively
2009: the   $d_\alpha(h)$    and $\bar{d}_\alpha(h)$   patterns   in the wave
2010: function.
2011: 
2012: 
2013: \begin{table}
2014:    \tbl{
2015: The 8 different classes of loops which can  surround an hexagon of the
2016: kagome lattice.  Including all possible symmetries we find 32 possible
2017: loops. The first column indicates the number of dimers involved in the
2018: coherent motion around the hexagon.  \label{tab:kloops}}
2019: { \begin{tabular}{|c|c|c|c|}
2020:       \hline%_______________________________
2021:       3&
2022:       \begin{picture}(50,34)(-24,-15)
2023: 	\pA{\La}\pB{\Lb}\pC{\Lc}\pD{\Ld}\pE{\Le}\pF{\Lf}\KagHex
2024:       \end{picture} &  & \\
2025:       \hline %_______________________________
2026:       4&
2027:       \begin{picture}(50,34)(-24,-13)
2028: 	\pA{\La}\pB{\La}\pG{\Lc}\pC{\Lc}\pD{\Ld}\pE{\Ld}\pJ{\Lf}\pF{\Lf}
2029: 	\KagHex\pG{\C}\pJ{\C}
2030:       \end{picture} &
2031:       % - - - -
2032:       \begin{picture}(50,34)(-24,-13)
2033: 	\pA{\La}\pB{\La}\pG{\Lc}\pC{\Lc}\pD{\Ld}\pE{\Le}\pF{\Le}\pK{\La}
2034: 	\KagHex\pG{\C}\pK{\C}
2035:       \end{picture} &
2036:       % - - - -
2037:       \begin{picture}(50,34)(-24,-13)
2038: 	\pA{\La}\pB{\La}\pG{\Lc}\pC{\Lb}\pH{\Ld}\pD{\Ld}\pE{\Le}\pF{\Lf}
2039: 	\KagHex\pG{\C}\pH{\C}
2040:       \end{picture}\\
2041:       \hline %_______________________________
2042:       5&
2043:       \begin{picture}(50,34)(-24,-13)
2044: 	\pA{\La}\pB{\La}\pG{\Lc}\pC{\Lb}\pH{\Ld}\pD{\Ld}
2045: 	\pE{\Ld}\pJ{\Lf}\pF{\Le}\pK{\La}
2046: 	\KagHex\pG{\C}\pJ{\C}\pK{\C}\pH{\C}
2047:       \end{picture} &
2048:       % - - - -
2049:       \begin{picture}(50,44)(-24,-23)
2050: 	\pA{\Lf}\pB{\Lb}\pL{\Lb}\pC{\Lb}\pH{\Ld}\pD{\Ld}
2051: 	\pE{\Ld}\pJ{\Lf}\pF{\Le}\pK{\La}
2052: 	\KagHex\pL{\C}\pJ{\C}\pK{\C}\pH{\C}
2053:       \end{picture} &
2054:       % - - - -
2055:       \begin{picture}(50,44)(-24,-23)
2056: 	\pA{\Lf}\pB{\La}\pL{\Lb}\pC{\Lc}\pG{\Lc}\pD{\Ld}
2057: 	\pE{\Ld}\pJ{\Lf}\pF{\Le}\pK{\La}
2058: 	\KagHex\pL{\C}\pJ{\C}\pK{\C}\pG{\C}
2059:       \end{picture}\\
2060:       \hline %_______________________________
2061:       6&
2062:       \begin{picture}(50,50)(-24,-23)
2063: 	\pA{\Lf}\pB{\La}\pL{\Lb}\pC{\Lb}\pG{\Lc}\pD{\Lc}
2064: 	\pE{\Ld}\pJ{\Lf}\pF{\Le}\pK{\La}\pH{\Ld}\pI{\Le}
2065: 	\KagStar
2066:       \end{picture}&  & \\
2067:       \hline %_______________________________
2068:     \end{tabular}
2069: }
2070: \end{table}
2071: 
2072: \subsubsection{RK ground-state}
2073: As  for   the  QDM discussed   previously  the  ground-state   of this
2074: Hamiltonian   is the  equal  amplitude   superposition  of  all  dimer
2075: coverings belonging    to   a given  topological  sector   (as  on the
2076: triangular  lattice there  are four sectors).   This can be
2077: readily shown by writing  $\mathcal{H}$ as a sum  of projectors:
2078: \begin{eqnarray}
2079:   \mathcal{H}&=&-N_h+\sum_h \sum_{\alpha=1}^{32}
2080:   \left[\;\left|d_\alpha(h) \right>-\left| \bar{d}_\alpha(h)\right>\;\right]
2081:   \left[\;\left<d_\alpha(h) \right|-\left< \bar{d}_\alpha(h)\right|\;\right]
2082: \end{eqnarray}
2083: where $N_h$ is the number of hexagons  on the lattice.  When expanding
2084: the products the diagonal terms give a simple constant since
2085: \begin{equation}
2086:   \sum_{\alpha=1}^{32} \left|d_\alpha \right>\left<d_\alpha\right|
2087:   +
2088:   \left| \bar{d}_\alpha\right>\left<\bar{d}_\alpha\right| = 1
2089: \end{equation}
2090: This  reflects  the fact that,  for  any  dimerization, the  dimers on
2091: hexagon $h$  match {\em one and only  one} of the $2\times32$ patterns
2092: $\left\{d_\alpha,\bar{d}_\alpha\right\}$. 
2093: 
2094: Unlike  the  square    or  triangular case,    the   RK  ground-states
2095: $|0\rangle=\sum_{c\in\Omega}|c\rangle$ are   not  degenerate with some
2096: staggered VBC.\footnote{Because resonances loops  of  length up to  12
2097: are present the  dynamics is ergodic in each  of  the four topological
2098: sectors.\cite{msp02}}   This    means  that    the    Hamiltonian   of
2099: Eq.~\ref{eq:kQDM} is not at  a phase transition to a  VBC.  As we will
2100: explain it  is {\em inside} a  liquid RVB phase.
2101: 
2102: The RK wave-function can be viewed as dimer condensate.  It is similar
2103: to  the ground-state  of liquid $^4$He  which   has the  same positive
2104: amplitude      for  any       configuration     and    its    permuted
2105: images.\cite{feynman52-53} An important  difference, however,  is that
2106: the  QDM state is incompressible  and cannot sustain acoustic phonons.
2107: This can be related to  the fact that the $U(1)$  symmetry of the Bose
2108: liquid is absent in the QDM on non-bipartite lattices.  It is replaced
2109: instead by   a  discrete  $\Z$  gauge  symmetry  (see \S\ref{sssec:Z2}
2110: below).
2111: 
2112: 
2113: \subsubsection{Ising pseudo-spin variables}
2114: 
2115: The kinetic energy operators $\sigma^x$ defined in Eq.~\ref{eq:sigmax}
2116: commute with each  other. This is obvious when  two such operators act
2117: on remote  hexagons but it   also  holds for  neighboring ones.   This
2118: property can   easily be  demonstrated  with the   help of  the  arrow
2119: representation   of  dimer     coverings  introduced    by    Zeng and
2120: Elser.\cite{ez93} This mapping of  kagome  dimerizations to arrows  on
2121: the   bonds     of   the  honeycomb       lattice  is      illustrated
2122: Fig.~\ref{fig:arrow}.  Each  arrow has   two possible  directions:  it
2123: points toward the interior of  one of  the two neighboring  triangles.
2124: If site $i$ belongs to a dimer $(i,j)$ its arrow must point toward the
2125: triangle the site $j$ belongs to.  A dimer covering can be constructed
2126: from any   arrow configuration provided   that the number  of outgoing
2127: arrows is odd (1 or 3) on every triangle.
2128: 
2129: \begin{figure}
2130:   \begin{center}
2131:     \includegraphics[height=4.5cm,width=6cm]{KagArrow}
2132:     \caption[99]{A  dimer  covering  on  the kagome  lattice  and  the
2133: 	corresponding arrows. Dashed lines: honeycomb lattice.}
2134:     \label{fig:arrow}
2135:   \end{center}
2136: \end{figure}
2137: 
2138: The operators $\sigma^x$ have a  particularly simple meaning in  terms
2139: of the arrow  degrees  of freedom: $\sigma^x(h)$  flips  the  6 arrows
2140: sitting  on $h$.\footnote{Flipping all   the arrows around  any closed
2141: loop (such  as around an   hexagon)  preserves  the local   constraint
2142: imposed on  arrow configurations.    Flipping   the arrows around    a
2143: topologically non-trivial loop changes the topological sector.}  It is
2144: then clear that the $\sigma^x$ commute and that $\sigma^x(h)^2=1$.  In
2145: fact these operators can  be used as  Ising pseudo-spin  variables and
2146: the  Hamiltonian now   describes  non-interacting pseudo-spins  in   a
2147: uniform   magnetic field  pointing  in  the   $x$ direction.   In  the
2148: ground-state we have $\sigma^x(h)=1$ on every hexagon.
2149: 
2150: \subsubsection{Dimer-dimer correlations}
2151: The  ground-state is the most possible  disordered dimer liquid as the
2152: dimer-dimer  correlations strictly  vanish    beyond  a few    lattice
2153: spacings.  Such  correlations  can    be  computed by    the  Pfaffian
2154: method. On the kagome lattice the determinant  of the Kasteleyn matrix
2155: (which is directly related to the partition  function of the classical
2156: dimers problem) is exactly constant in Fourier space.\cite{hw88} Since
2157: dimer-dimer correlations  are obtained from  the  Fourier transform of
2158: the inverse  of  this determinant, they  turn out  to be strictly zero
2159: beyond a few lattice spacings (as soon as the two bonds  do not touch a
2160: common  triangle).\cite{msp02} This result can also   be obtained by a
2161: simpler  argument\cite{msp02,msp03b} using  the  $\sigma^x$ operators.
2162: This result is related  to the kagome geometry.\footnote{The model  of
2163: Eq.~\ref{eq:kQDM}   can  be  generalized   to any    lattice   made of
2164: corner-sharing triangles.\cite{msp02}}  This absence   of  long-ranged
2165: dimer-dimer correlations  demonstrates that the   RK state is  a dimer
2166: liquid and that it breaks no lattice symmetry.
2167: 
2168: On the  triangular lattice,   even at high    temperature, dimer-dimer
2169: correlations decay exponentially  with distance but these correlations
2170: remain {\em finite}  at  any distance.  On   the  square lattice  such
2171: correlations are  even  larger  because they  decay  only as  a  power
2172: law. This means that the  infinite hard-core dimer repulsion makes QDM
2173: non-trivial  even at infinite temperature; dimers  cannot be free when
2174: they are fully-packed.  From this point of view we see that the kagome
2175: lattice is particular: it is as close as possible to a free dimer gas,
2176: except for non-trivial correlations over  a few lattice spacings. This
2177: is a reason why  dimer coverings on the  kagome lattice can be handled
2178: with independent pseudo-spin  variables and why  the RK state  on this
2179: lattice is the most possible disordered RVB liquid.
2180: 
2181: \subsubsection{Visons excitations}
2182: \label{sssec:vison}
2183: 
2184: The $\sigma^x$ operators  can be simultaneously diagonalized but  they
2185: must satisfy the global constraint  $\prod_h \sigma^x(h)=1$ since this
2186: product flips every  arrow {\em twice}.  It  must  therefore leave all
2187: dimerizations unchanged.   The  lowest  excitations  have therefore an
2188: energy $4$ above the ground-state and they are made of a {\em pair} of
2189: hexagons $a$ and $b$  in a $\sigma^x(a)=\sigma^x(b)=-1$ state. $a$ and
2190: $b$  are the    locations     of  two   Ising   vortices   (or    {\em
2191: visons}\cite{sf00,sf01}).   As  remarked before  this   means that the
2192: relative  phases   of  the  configurations   with   $d_\alpha(h)$  and
2193: $\bar{d}_\alpha(h)$ patterns  have now changed sign. The corresponding
2194: wave-function  is  obtained in the   following way.  Consider a string
2195: $\Omega$ which goes from $a$ to $b$ (see Fig.~\ref{fig:Vison}) and let
2196: $\Omega(a,b)$ be the operator which measures  the parity $\pm1$ of the
2197: number of  dimers crossing that   string.  $\Omega(a,b)$ commutes with
2198: all  $\sigma^x(h)$,    except  for   the     ends   of the     string:
2199: $\sigma^x(a)\Omega(a,b)=-\Omega(a,b)\sigma^x(a)$. A dimer move changes
2200: the sign of  $\Omega(a,b)$ if and only if  the associated loop crosses
2201: the  string an   odd  number of  times,  which  can  only  be done  by
2202: surrounding one  end of  the  string.  This shows  that  $\Omega(a,b)$
2203: flips the   $\sigma^x$ in $a$ and  $b$.\footnote{Up   to a global sign
2204: (reference         dependent)   $\Omega(a,b)$       is  equal       to
2205: $\sigma^z(a)\sigma^z(b)$  where  the $\sigma^z$   operators are  those
2206: introduced by  Zeng and Elser.}  As  the  RK ground state $|0\rangle$,
2207: $\Omega(a,b)|0\rangle$  is   a  linear   combination  of   all   dimer
2208: configurations belonging to one sector. However the amplitudes are now
2209: $1$ and $-1$ depending on the number of dimers crossing $\Omega$. This
2210: wave-function  therefore has nodes, it is  an excited  state of energy
2211: $4$ with  two vortices  in $a$ and   $b$.  It is   easy to see that  a
2212: different choice $\Omega'$ for the string connecting $a$ and $b$ gives
2213: the same state up to a global sign which depends  on the parity of the
2214: number of kagome sites enclosed by $\Omega\cup\Omega'$.
2215: 
2216: These vortex excitations carry a $\Z$ charge  since attempting to put
2217: two  vortices on  the same  hexagon does  not change  the state.  Such
2218: excitations are not local  in terms of  the dimer degrees  of freedom.
2219: Indeed,  determining the sign of a  given dimerization in a state with
2220: two visons which  are far apart requires  the  knowledge of the  dimer
2221: locations along the whole string connecting the  two vortex cores.  In
2222: this model the visons appear to be static and non-interacting. This is
2223: a particularity of  this  solvable model  but the  existence of gapped
2224: vison excitations  is   believed  to be  a   robust  property   of RVB
2225: liquids. In more  realistic models the  visons will acquire a dynamics
2226: and  a  dispersion  relation  but will   remain gapped.\footnote{It is
2227: possible to add potential energy  terms to Eq.~\ref{eq:kQDM} to  drive
2228: the system outside of the liquid phase and this transition corresponds
2229: to a vison condensation.}  They will also  have some interactions with
2230: each  other but  should remain {\em    deconfined}.  This property  is
2231: particularly clear in the  kagome QDM: visons are  necessarily created
2232: by pairs but the energy is independent of their relative distances.
2233: 
2234: \begin{figure}
2235:   \begin{center}
2236: 	\includegraphics[height=2.5cm]{VisonPairA}
2237: 	\includegraphics[height=2.5cm]{VisonPairB}
2238: 
2239:   \caption[99]{A pair of visons (located in $a$ and $b$) is created by
2240:   applying  to the  RK wave-function  a factor  $(-1)$  for each dimer
2241:   crossing   the string $\Omega$. The  dimerization  shown there on the left
2242:    appears  in the  linear superposition  of  the two-vison state
2243:   with the sign $-1$ whereas the one on the right has the sign $+1$.}
2244: 
2245:   \label{fig:Vison} \end{center}
2246: \end{figure}
2247: 
2248: The  Ising vortices  also offer a  simple picture  of the  topological
2249: degeneracy.  Consider  a  ground-state $|+\rangle$ of  the model which
2250: lives in the sector where the  winding number $\Omega_y$ (with respect
2251: to  some  arbitrary  but  fixed    dimerization) is  even.     Another
2252: ground-state $|-\rangle$ is obtained in the odd-$\Omega_y$ sector. Now
2253: consider the combination $|0\rangle=|+\rangle+|-\rangle$ and apply the
2254: operator $\Omega(0,L_x)$ corresponding  to  a closed loop  surrounding
2255: the torus  in the $x$ direction. This   amounts to creating  a pair of
2256: nearby visons at  the origin, taking  one of them  around the torus in
2257: the $x$  direction and annihilating them.  This  can also be viewed as
2258: the creation  of a  vison in  one hole  of  the torus (with  no energy
2259: cost).         It       is      simple      to       check        that
2260: $\Omega(0,L_x)|0\rangle=|+\rangle-|-\rangle$   (up to an    irrelevant
2261: global sign).  This provides  a simple relation between the vison-pair
2262: creation  operator and the  existence   of two topologically  distinct
2263: ground-states $|+\rangle+|-\rangle$ and $|+\rangle-|-\rangle$.
2264: 
2265: \subsubsection{Spinons deconfinement}
2266: \label{sssec:QDMdeconf}
2267: 
2268: We assume that dimers represent  ``dressed'' singlet valence-bonds, as
2269: in  the overlap expansion  (\S\ref{ssec:spin2QDM}).  Since the Hilbert
2270: space is   made   of   fully-packed dimer  coverings  the     model of
2271: Eq.~\ref{eq:kQDM} only describes spin-singlet states.  However, as any
2272: QDM,   it  can   be extended  to   include  static  holes or  spinons.
2273: Configurations  with unpaired sites (spinon or  holon) are now allowed
2274: but the kinetic terms of the original Hamiltonian which loop passes on
2275: an empty site  gives zero.  Consider a  system with two static spinons
2276: in $x$ and   $y$.    As  on  the square\cite{rk88}   and    triangular
2277: lattices\cite{ms01}  at    the  RK   point the      exact ground-state
2278: $\left|x,y\right>$ remains the sum  of   all dimer coverings and   the
2279: ground-state energy is  independent  of the  distance between the  two
2280: spinons (except at  very short  distance  if they belong  to a  common
2281: hexagon).  This is  a first indication that RVB  {\em spin} liquid has
2282: deconfined  spin-$\frac{1}{2}$ excitations   (spinons).   In  the  QDM
2283: language these excitations are simply unpaired sites in a dimer liquid
2284: background. Such  unpaired sites are necessarily  created by pairs but
2285: they can then propagate freely (no attractive potential) when they are
2286: sufficiently far apart.
2287: 
2288: Another calculation allows  to test the  deconfinement properties of a
2289: dimer liquid.  We   consider the state   $\left|\psi\right>=\sum_{{\bf
2290: r}\ne0}\left|0,{\bf r}\right>$ where  $\left|0,{\bf r}\right>$ is  the
2291: (un-normalized) state with  two spinons  in $0$ and  ${\bf  r}$.   The
2292: probability  to find a spinon  in ${\bf r}$ in the $\left|\psi\right>$
2293: can be obtained by the relatively  involved calculation of the monomer
2294: correlation\footnote{Ratio of number of dimer coverings with two holes
2295: in  $0$ and ${\bf r}$  to the number  without  hole.}  with Pfaffians.
2296: One   the  square   lattice   this  probability   goes   to   zero  as
2297: $1/\sqrt{r}$.\cite{fs63} This shows that the second spinon is (quasi-)
2298: confined  in the vicinity  of  the first  one  on  the square  lattice
2299: because escaping  far away represents a large  ``entropy'' cost in the
2300: dimer background.  On the triangular  lattice it goes exponentially to
2301: a constant.\cite{fms02} This result  is a signature of  deconfinement.
2302: In fact the  same  signature  can be  obtained on  the  kagome lattice
2303: without  any technical  calculation since the   monomer correlation is
2304: exactly $1/4$ at any distance.\cite{msp03c}
2305: 
2306: If unpaired  sites are  allowed  one can describe spinons   or holons.
2307: Unfortunately in the presence of simple kinetic energy terms for these
2308: objects the model can no longer be solved.  However one can consider a
2309: static spinon  and its  interaction with  visons: when the   spinon is
2310: adiabatically taken around a vison the dimers are shifted along a path
2311: encircling the vison.  Because the  vison wave-function is  particularly
2312: simple in this model it   is easy to  check  that this multiplies  the
2313: wave-function  by  a  factor  $-1$.    This is   the signature  of   a
2314: long-ranged statistical  interaction\cite{rc89,k89} between visons and
2315: spinons (or holons).  In more realistic models,  as long as the visons
2316: are gapped  excitations the spinons  are expected to be deconfined. On
2317: the  other hand if  the  visons condense their long-ranged statistical
2318: interaction with spinons frustrates  their  motion. This is no  longer
2319: true if they propagate   in {\em pairs},   in which case they are  not
2320: sensitive  any more to  visons (see Ref.~\cite{msp02} for an extension
2321: of the present QDM with  a  vison condensation). This simple  physical
2322: picture illustrates the relation between vison condensation and spinon
2323: confinement.
2324: 
2325: \subsubsection{$\Z$ gauge theory}
2326: \label{sssec:Z2} 
2327: 
2328: The forces responsible for confinement are usually associated to gauge
2329: fields and their fluctuations.   Whereas $U(1)$ compact gauge theories
2330: are  generically confining in  $2+1$ dimensions,\cite{polyakov87,fs79}
2331: $\Z$    gauge    theories   are    known    to   possess    deconfined
2332: phases.\cite{kogut79} For this reason  some attention has been paid to
2333: the  connexions between  $\Z$  theories and  fractionalized phases  in
2334: 2D electronic systems.\cite{sf00}
2335: 
2336: It is known\cite{msf02} that QDM can be obtained as special limits of
2337: $\Z$ gauge  theories, the gauge variable  being the dimer  number on a
2338: bond.  However, on the kagome lattice this connexion can be made exact
2339: and completely  explicit since  there is a  one to  one correspondence
2340: between   dimer   coverings    and   physical   states   ({\it   i.e.}
2341: gauge-invariant) of  a $\Z$ gauge theory.\cite{msp02}  In this mapping
2342: the  gauge  fields are  Ising  variables living  on  the  link of  the
2343: honeycomb lattice  ({\it i.e.} kagome sites) and  are constructed from
2344: the  arrows described  previously.  As  for the  constraints  of gauge
2345: invariance they correspond to the odd parity of the number of outgoing
2346: arrows on  every triangles. The  $\sigma^x$ operator used to  define a
2347: solvable QDM  translate into a gauge-invariant  plaquette operator for
2348: the gauge  degrees of  freedom (product of  the Ising  gauge variables
2349: around an hexagon). With this mapping the visons appear to be vortices
2350: in the  $\Z$ gauge field  and the solvable model  of Eq.~\ref{eq:kQDM}
2351: maps to the deconfined phase of the gauge theory.
2352: 
2353: \subsection{A QDM with an extensive ground-state entropy}
2354: \label{ssec:QDMmu}
2355: 
2356: So far we   have discussed QDM that  realize  either spontaneous VBC,
2357: critical states  or RVB liquids.   We wish to  mention here that these
2358: three  scenarios may  not  be   the  only possible  ground-states  for
2359: QDM. In particular,  a  QDM on the kagome  lattice  with an extensive
2360: ground-state  entropy was  recently  discussed.\cite{msp03} This model
2361: was  introduced from  the observation  that  the  dimer kinetic energy
2362: terms  arising   from   an  overlap  expansion (\S\ref{ssec:spin2QDM})
2363: generally have  non trivial {\em signs}  as soon as resonance loops of
2364: {\em different lengths} are considered.  It was then realized that such
2365: signs (which make  the QDM  no longer  appropriate  for QMC simulations)   can lead to  qualitatively  new phases, different
2366: from      VBC  or RVB   liquids.     The    Hamiltonian introduced  in
2367: Ref.~\cite{msp03}   is           similar      to       that         of
2368: Eqs.~\ref{eq:kQDM}-\ref{eq:sigmax} except that each resonance     loop
2369: $\alpha$   is now included with  a  {\em sign} $(-1)^{n_\alpha}$ where
2370: $n_\alpha=3,\cdots,6$ is the number of dimers involved:
2371: \begin{equation}
2372: 	\mathcal{H}=\sum_h (-1)^{n_\alpha} \left[
2373: 	\left|d_\alpha(h) \right>\left< \bar{d}_\alpha(h)\right|
2374:   	+
2375: 	\left| \bar{d}_\alpha(h)\right>\left<d_\alpha(h)\right|
2376: 	\right]
2377: 	\label{eq:muQDM}
2378: \end{equation}
2379: These signs are precisely those appearing in the overlap expansion (at
2380: the  order of one   hexagon)  of the  Heisenberg  model on  the kagome
2381: lattice.  This expansion was  carried out by Zeng and Elser\cite{ze95}
2382: in an  insightful   paper which laid  the  basis of the analysis  of the
2383: kagome antiferromagnet in the first neighbor valence-bonds subspace.
2384: 
2385: Although  not exactly solvable,  the Hamiltonian of Eq.~\ref{eq:muQDM}
2386: was shown to be a dimer  liquid (short-ranged dimer-dimer correlation)
2387: and to have a huge ground-state degeneracy $\sim 2^{N/6}=1.122^N$ ($N$
2388: is  the   number of  kagome sites).    In addition,  several numerical
2389: indications pointed to a critical behavior of this system,\cite{msp03}
2390: with    a     possible     algebraic     decay    of     energy-energy
2391: correlations.\footnote{Notice   that  a   one-dimensional   analog  of
2392: Eq.~\ref{eq:muQDM} can be defined and exactly maps onto the Ising chain
2393: in  transverse field at its  critical point.}  It  was argued that the
2394: effective   QDM    describing      the singlet     dynamics   of   the
2395: spin-$\frac{1}{2}$ Heisenberg  antiferromagnet on  the kagome  lattice
2396: could be {\em  close} (in parameter  space) to Eq.~\ref{eq:muQDM}.  If
2397: correct, this sheds light  on the very large  density of singlet states
2398: observed at low energy in  the numerical  spectra  of that spin  model
2399: (see Sec.~\ref{sec:kagome}).
2400: 
2401: 
2402: 
2403: 
2404: %_______________________________________________________________________________
2405: \section{Multiple-spin exchange models}\label{sec:MSE}
2406: 
2407: %_______________________________________________________________________________
2408: 
2409: \subsection{Physical realizations of multiple-spin interactions}
2410: 
2411: \subsubsection{Nuclear magnetism of solid $^3$He}
2412: 
2413: Solid $^3$He was the first magnetic system in  which the importance of
2414: MSE interactions   was recognized.\cite{thouless65,rhd83,cf85}  Due to
2415: the large zero-point motion  of the atoms  about their mean  positions,
2416: tunneling events during which 2, 3 or 4 atoms exchange their positions
2417: in a cyclic way  are frequent.  These  processes generate an effective
2418: interaction between the (nuclear) spins which can be written
2419: \begin{equation}
2420:   \mathcal{H}=\sum_P -J_P (-1)^P\left( P_{spin} + P_{spin}^{-1}\right)
2421: \end{equation}
2422: where the  sum runs over  permutations  $P$,  $J_P>0$ is the  exchange
2423: frequency  of the associated tunneling   process (in  real space)  and
2424: $P_{spin}$ acts on the  Hilbert space of spin-$\frac{1}{2}$ located on
2425: the site of  the crystal. The sign $-(-1)^P$  depends of the signature
2426: of the permutation  $P$ and is a consequence  of  the Pauli principle.
2427: For a {\em cyclic}  permutation involving $n$ spins  this sign is just
2428: $(-1)^n$  and  is  responsible   for  the ferromagnetic  character  of
2429: processes involving  an odd number   of spins.  For spin-$\frac{1}{2}$
2430: particles, two  and three-spin exchange  terms  reduce to the familiar
2431: Heisenberg interaction:
2432: \begin{eqnarray}
2433:   P_{12}=2\vec{S}_1\cdot\vec{S}_2+\frac{1}{2} \label{eq:P2} \\
2434:   P_{123}+P_{321}=P_{12}+P_{23}+P_{31}-1 \label{eq:P3}
2435: \end{eqnarray}
2436: but this is no longer true for $n\geq4$:
2437: \begin{equation}
2438: P_{1234}+P_{4321}=P_{12}P_{34}+P_{14}P_{23}-P_{13}P_{24}+P_{13}+P_{24}-1
2439: \end{equation}
2440: which can be  expressed (thanks to Eq.~\ref{eq:P2}) as  a sum of terms
2441: with two and four Pauli matrices. 
2442: 
2443: $^3$He can form solid  atomic mono-layers  with a triangular  geometry
2444: when adsorbed on a graphite substrate at ultra low temperatures (milli
2445: Kelvin range).  This 2D  magnet has been  studied for a long time (see
2446: Refs.~\cite{greywall,gr95}  and references therein) and the importance
2447: of MSE   interactions  involving up    to  six  atoms has  now    been
2448: recognized.\cite{roger84,rbbcg98} The exchange frequencies of the most
2449: important processes have been   computed by Path Integral  Monte Carlo
2450: (PIMC)\cite{cj87,bcl92,ceperley95,bc99}  (analytic  WKB   calculations
2451: have  also been carried  out\cite{roger84,ah00})  as a function of the
2452: density.    The proposed  MSE   Hamiltonian  describing the   magnetic
2453: properties of this 2D quantum crystal reads
2454: \begin{eqnarray}\label{Hmulti}
2455: \mathcal{H}=(J_2-2J_3) \sum_{
2456:         \begin{picture}(17,10)(-2,-2)
2457:                 \put (0,0) {\line (1,0) {12}}
2458:                 \put (0,0) {\circle*{5}}
2459:                 \put (12,0) {\circle*{5}}
2460:         \end{picture}
2461: } P_{12}
2462: +J_4 \sum_{
2463: \begin{picture}(26,15)(-2,-2)
2464:         \put (0,0) {\line (1,0) {12}}
2465:         \put (6,10) {\line (1,0) {12}}
2466:         \put (0,0) {\line (3,5) {6}}
2467:         \put (12,0) {\line (3,5) {6}}
2468:         \put (6,10) {\circle*{5}}
2469:         \put (18,10) {\circle*{5}}
2470:         \put (0,0) {\circle*{5}}
2471:         \put (12,0) {\circle*{5}}
2472:         \end{picture}
2473: } \left( P_{1\ldots 4}+{\rm H.c}\right)\\
2474: -J_5 \sum_{
2475: \begin{picture}(26,15)(-2,-2)
2476:         \put (0,0) {\line (1,0) {24}}
2477:         \put (6,10) {\line (1,0) {12}}
2478:         \put (0,0) {\line (3,5) {6}}
2479:         \put (18,10) {\line (3,-5) {6}}
2480:         \put (6,10) {\circle*{5}}
2481:         \put (18,10) {\circle*{5}}
2482:         \put (0,0) {\circle*{5}}
2483:         \put (12,0) {\circle*{5}}
2484:         \put (24,0) {\circle*{5}}
2485:         \end{picture}
2486: } \left( P_{1\ldots 5}+{\rm H.c}\right) \nonumber
2487: +J_6 \sum_{
2488: \begin{picture}(26,30)(-2,-15)
2489:         \put (6,10) {\line (1,0) {12}}
2490:         \put (6,-10) {\line (1,0) {12}}
2491:         \put (0,0) {\line (3,5) {6}}
2492:         \put (0,0) {\line (3,-5) {6}}
2493:         \put (18,10) {\line (3,-5) {6}}
2494:         \put (18,-10) {\line (3,5) {6}}
2495:         \put (6,10) {\circle*{5}}
2496:         \put (6,-10) {\circle*{5}}
2497:         \put (18,10) {\circle*{5}}
2498:         \put (18,-10) {\circle*{5}}
2499:         \put (0,0) {\circle*{5}}
2500:         \put (12,0) {\circle*{5}}
2501:         \put (24,0) {\circle*{5}}
2502:         \end{picture}
2503: } \left( P_{1\ldots 6}+{\rm H.c}\right)
2504: \label{eq:H26}
2505: \end{eqnarray}
2506: where Eq.~\ref{eq:P3} was used to absorb the  three-spin terms into an
2507: effective      first-neighbor      Heisenberg     exchange   $J_2^{\rm
2508: eff}=J_2-2J_3$. At high density the hard-core potential between Helium
2509: atoms only leaves three-body  exchanges possible  ($J_3\gg J_{n\ne3}$)
2510: and Eq.~\ref{eq:H26} reduces    to a first neighbor   Heisenberg  {\em
2511: ferromagnet},\cite{drh80} as observed   experimentally for  the  first
2512: time by Franco {\it et al.}\cite{frg86} in high-density layers. On the
2513: other hand the  second layer solidifies  at  lower density and  higher
2514: order exchange terms  cannot be  ignored.\footnote{The first layer  is
2515: then so  dense that exchange is strongly  suppressed.  The first layer
2516: can  also   be replaced  by   an   $^4$He  or  HD  mono-layer.}   PIMC
2517: simulations\cite{bc99} and  high-temperature  fits of the experimental
2518: data\cite{rbbcg98}  showed that   the  relative strength   of two- and
2519: four-spin terms   if  roughly  $J_2^{\rm    eff}/J_4\sim -2$  in   the
2520: low-density second layer solid.  The $J_2$--$J_4$ model was studied by
2521: exact diagonalizations in this region of parameter space and evidences
2522: for  a   short-ranged RVB SL   phase   with no  broken  symmetry  were
2523: obtained.\cite{mblw98,mlbw99,lmsl00}   The    ultra-low    temperature
2524: measurements      of   specific     heat\cite{kmyf97}    and   uniform
2525: susceptibility\cite{cthrbbg01} are  not incompatible  with such a spin
2526: liquid phase but the spin gap,  if any, has not  yet been observed 
2527: and should be smaller than 100$\mu$K.
2528: 
2529: 
2530: \subsubsection{Wigner crystal}
2531: 
2532: The   Wigner crystal is  another    fermionic solid with a  triangular
2533: geometry where MSE  interactions can play an  important role.  At very
2534: low density the  Coulomb   energy  dominates, the crystal   is  almost
2535: classical and MSE interactions  are very small.   Exchange frequencies
2536: $J_P$   can be  computed in  this   regime by  a  semi-classical (WKB)
2537: approximation\cite{roger84,kh00,hk01}  and,  as  for  the high density
2538: solid of  $^3$He,  three-body  exchanges dominate  and  give  rise  to
2539: ferromagnetism. However, at higher  density and close to melting, PIMC
2540: calculations  of the  exchange  frequency\cite{bcc01} showed  that the
2541: magnetism  may be described by  a MSE model with parameters ($J_2^{\rm
2542: eff}$  and $J_4$)  close to those   where the triangular  MSE model is
2543: expected to be a RVB  SL.  Unlike the  $^3$He case, the  particles
2544: (electrons) are  charged and  an external magnetic  field has  also an
2545: orbital effect, it introduces complex phases in the exchange energies:
2546: $P+P^{-1}\to   e^{i\alpha}P+e^{-i\alpha}P^{-1}$  where    the    angle
2547: $\alpha=2\pi
2548: \phi/\phi_0$  is  proportional to  the  magnetic  flux $\phi$  passing
2549: through the area  enclosed by the exchange trajectory  and $\phi_0$ is
2550: the  unit  flux  quantum.  This  can  give  rise  to very  rich  phase
2551: diagrams\cite{bcc01,hk01}  where complex  MSE terms  compete  with the
2552: Zeeman  effect  (see  Ref.~\cite{ok98}  for  some  early  experimental
2553: attempts to explore this physics).
2554: 
2555: \subsubsection{Cuprates}
2556: 
2557: The possibility  of significant  four-spin exchange  around  square Cu
2558: plaquettes in copper oxide compounds  was first suggested by Roger and
2559: Delrieu.\cite{rd89} They interpreted  the  anomalously large width  of
2560: Raman scattering spectra as a signature of  four-spin exchange in this
2561: copper oxide superconductor.  The importance of these MSE interactions
2562: in CuO$_2$ planes  ($J_4\sim 0.25 J_2$) has  then been emphasized by a
2563: number   of  groups and   in   different  materials  and  by different
2564: experimental                      and                      theoretical
2565: approaches.\cite{sugai90,coldea01,mr02,kk03} Four-spin  plaquette ring
2566: exchange    also   plays      a     significant   role     in   ladder
2567: compounds.\cite{mkebm00,bmmnu99,sku01,gkt03}   For  instance, exchange
2568: parameters    with  values    $J_{rung}=J_{leg}=110$~meV           and
2569: $J_{ring}=16.5$~meV  were  proposed  for La$_6$Ca$_8$Cu$_{24}$O$_{41}$
2570: based     on    the       dispersion      relation    of      magnetic
2571: excitations.\cite{mkebm00,bmmnu99}
2572: 
2573: \subsection{Two-leg ladders}
2574: \label{ssec:ladderMSE}
2575: 
2576: Numerous  works   were  devoted   to  ladder  models   with  four-spin
2577: interactions.  These include  general bi-quadratic interactions as well
2578: as models  with ring-exchange  terms.  We will  only discuss  here the
2579: simplest of these MSE models:
2580: \begin{eqnarray}
2581:   \mathcal{H}=&J&\sum_n \left(
2582:     \vec{S}_{n,1}\cdot \vec{S}_{n,2}
2583:     +\vec{S}_{n,1}\cdot \vec{S}_{n+1,1}
2584:     +\vec{S}_{n,2}\cdot \vec{S}_{n+1,2}
2585:     \right) \nonumber \\
2586:     +&K&\sum_\Box \left(
2587:     P_{1234} +H.c
2588:     \right)
2589:     \label{eq:ladder}
2590: \end{eqnarray}
2591: Thanks  to several  studies\cite{bmmnu99,mvm02,hmh03,lst03}  the phase
2592: diagram of  this Hamiltonian  is now rather  well understood  and five
2593: different phases were identified.
2594: \begin{itemize}
2595: 
2596: \item   {\bf   Ferromagnetic  phase}.   The   ground-state  is   fully
2597: polarized. This phase includes  the $(J=-1,K=0)$ and the  $(J=0,K=-1)$
2598: points.
2599: 
2600: \item {\bf  Rung-singlet phase}.  This phase  includes the ground-state of
2601: the ladder without  MSE term $(J=1,K=0)$.  The  spectrum is gapped and
2602: the   ground-state  is  unique.    A moderate  $K/J\gtrsim0.23\pm0.03$
2603: destroys this  phase\cite{bmmnu99,hh01,hn02,lst03} in favor of the VBC
2604: below.
2605: 
2606: \item {\bf Staggered VBC} with dimers  on the legs.  In one of the two
2607: degenerate ground-states the dimerized bonds are $(2n,1)-(2n+1,1)$ and
2608: $(2n+1,2)-(2n+2,2)$.       The      VBC     disappears             for
2609: $K/J\gtrsim0.5$.\cite{lst03} Such  a staggered VBC was first predicted
2610: in  the framework   of   a ladder  with  bi-quadratic interaction   by
2611: Nersesyan   and   Tsvelik.\cite{nt97} Using     Matrix-Product Ansatz,
2612: Kolezhuk    and  Mikeska\cite{km98}    constructed  models which   are
2613: generalizations    of  Eq.~\ref{eq:ladder}  and   which    have  exact
2614: ground-state with long-ranged  staggered dimer correlations.   In this
2615: phase the   magnetic excitations are  very  different from  the magnon
2616: excitations of the rung-singlet  phase above. Here the  excitations do
2617: not form well-defined quasi-particles but a  continuum made of pairs of
2618: domain walls connecting two dimerized ground-states.\cite{nt97,km98}
2619:  
2620: \item  {\bf   Scalar  chirality   phase}.   The  order   parameter  is
2621: $\langle\vec{S}_{n,1}\cdot(\vec{S}_{n,2}\times\vec{S}_{n+1,2})\rangle$
2622: and it spontaneously breaks the time-reversal symmetry and translation
2623: invariance.  The  ground-state is two-fold  degenerate up to  the next
2624: transition at $K/J\simeq2.8\pm0.3$.\cite{lst03} There exists a duality
2625: transformation\cite{hmh03,mhnh03} which  maps the  scalar    chirality
2626: order parameter onto   the   dimer   order   parameter of   the    VBC
2627: above.\footnote{The Hamiltonian of Eq.~\ref{eq:ladder} is self-dual at
2628: $2K=J$.}    Applying    such a  transformation     to  the  exact  VBC
2629: ground-states   mentioned   above,  models   with  an  exactly   known
2630: ground-state       and   scalar      chirality     LRO   can        be
2631: constructed.\cite{hmh03,mhnh03} Although  chiral  SL  have  been  much
2632: discussed  in  the  literature, this is   to  our knowledge  the first
2633: realization of such a phase in a SU(2) symmetric spin-$\frac{1}{2}$ model.
2634: 
2635: 
2636: \item   {\bf  Short-ranged   ordered   phase  with   vector-chirality}
2637: correlations.        The         strongest        correlations     are
2638: $\langle(\vec{S}_{n,1}\times\vec{S}_{n,2})\cdot(\vec{S}_{n',1}\times\vec{S}_{n',2})\rangle$
2639: but they  remain short-ranged.    The   spectrum is gapped  and    the
2640: ground-state  is unique.  This   phase includes  the pure $K=1$  model
2641: where  $J=0$.  This phase   is  related by the duality  transformation
2642: discussed    above to   the   rung-singlet   phase.\cite{mhnh03}  This
2643: transformation    indeed relates   the  N\'eel  correlations  $\langle
2644: (\vec{S}_{n,1}-\vec{S}_{n,2})\cdot(\vec{S}_{n',1}-\vec{S}_{n',2})\rangle$
2645: (which  are  the strongest  ones  in the   rung-singlet phase)  to the
2646: vector-chirality correlations.  Close    to the ferromagnetic    phase
2647: ($J<0$) one observes   a crossover to  a  region where the   strongest
2648: correlations are  ferromagnetic spin-spin correlations along  the legs
2649: and antiferromagnetic along the rungs.\cite{lst03}
2650: \end{itemize}
2651: \subsection{MSE model on the square lattice}
2652: The phase  diagram  of  the Hamiltonian  \ref{eq:ladder}  on the  {\em
2653: two-dimensional    square-lattice}  has   been   recently   studied by
2654: L\"auchli\cite{lauchli03}  by  exact      diagonalizations.   N\'eel,
2655: ferromagnetic, columnar VBC and staggered  VBC phases were identified,
2656: as   in the  ladder   model  above.    In  addition,  a nematic  phase
2657: characterized by  long-ranged vector chirality correlations (alternating
2658: spin currents)  was  found  around  the $K=1$, $J=0$  point.   To  our
2659: knowledge this is the first  microscopic realization of a nematic order
2660: in a two-dimensional spin-$\frac{1}{2}$ model.
2661: 
2662: \subsection{RVB phase of the triangular $J_2$--$J_4$ MSE}
2663: \label{ssec:RVBMSE}
2664: 
2665: Because  of its relevance  to solid $^3$He  films and Wigner crystals,
2666: the MSE model on the  triangular lattice has been  the subject of many
2667: studies.\cite{km97,mkn97,mbl98,ksmn98,mblw98,mlbw99,lmsl00}  We   will
2668: discuss here  some properties  of the  simplest MSE  model with up  to
2669: four-spin cyclic exchange interactions ($J_2  -2J_3$ and $J_4$ only in
2670: Eq.~\ref{eq:H26}).         The   classical           phase     diagram
2671: (Fig.  \ref{fig:phasediagMSE}) of this  model has been studied by Kubo
2672: and   collaborators\cite{km97,mkn97}  and the  quantum   one has  been
2673: roughly scanned  in  Ref.\cite{lmsl00}: we  will mainly focus   on the
2674: short-ranged RVB  spin liquid (see Fig.~\ref{fig:phasediagMSE}), which
2675: is probably  {\it the first RVB  SL encountered in  an SU(2)-symmetric
2676: spin model}.
2677: 
2678: \begin{figure} 
2679: \begin{center}
2680: \includegraphics[width=8cm]{MSE}
2681: \end{center}
2682: \caption[99]{
2683: Classical (top)  and  quantum (bottom)  phases   of the $J_2-J_4$  MSE
2684: Hamiltonian.      The classical  model   was    studied  by  Kubo  and
2685: Momoi\cite{km97} and is based on a variational approach.  The quantum
2686: phase diagram  is  the  simplest scenario  compatible with  the  exact
2687: diagonalizations data of Refs.~\cite{mblw98,mlbw99,lmsl00}.  While not
2688: completely understood, in  the type-II spin-liquid  region the spectra
2689: are characterized by a very  large number of singlet excitations below
2690: the first triplet state. This is not the case in the RVB phase.}
2691: \label{fig:phasediagMSE}
2692: \end{figure}
2693: 
2694: \subsubsection{Non-planar classical ground-states}
2695: 
2696: It     is  well-known that      an  Heisenberg  model  (with  possible
2697: second-neighbors,  third-neighbors, ...   interactions)  on a  Bravais
2698: lattice  always admits a planar  helical ground-state at the classical
2699: level. This is no longer  true when  MSE  are present and finding  the
2700: classical  ground-state for arbitrary  $J_2$  and $J_4$ is an unsolved
2701: problem.  A  mean-field phase diagram was  obtained  for the classical
2702: model\cite{km97} but   very   few exact  results  are  known.   In the
2703: neighborhood of $J_4=1$ $J_2=0$ the classical ground-state is known to
2704: be  a  four-sublattice  configurations   with  magnetizations pointing
2705: toward the  vertices of  a  regular tetrahedron.\cite{km97}  This is a
2706: quite interesting model where the  ground-state spontaneously breaks a
2707: discrete Ising  symmetry associated to the sign  of the triple product
2708: $\vec{S}_1\cdot(\vec{S}_2\times\vec{S}_3)$   around a triangle.   This
2709: broken symmetry  gives rise to  a finite-temperature  phase transition
2710: which has  been observed in  Monte Carlo simulations.\cite{mkn97} This
2711: phenomena is similar to    the transition predicted in  the  $(\pi,0)$
2712: phase of the $J_1$--$J_2$ model on the square lattice.\cite{ccl90a}
2713: 
2714: \subsubsection{Absence of N\'eel LRO}
2715: 
2716: 
2717: The   classical  ground-states   at   $J_4=1,J_2=0$  are   tetrahedral
2718: configurations.  Although  this phase appears to be  stable within the
2719: framework    of   linear    spin-wave    calculations\cite{mkn97}   or
2720: Schwinger-Boson mean-field  theory,\cite{mbl98} exact diagonalizations
2721: indicate   that  the   magnetic   LRO  is   washed   out  by   quantum
2722: fluctuations.\cite{mkn97,mlbw99}   The   chiral  order  predicted  to
2723: survive at  long distances and finite  temperatures\cite{mkn97} in the
2724: classical  system for  $J_2=0$  is also  likely  to be  washed out  by
2725: quantum fluctuations.\cite{mlbw99}
2726: 
2727: When $J_2=1$ a relatively small  amount of $J_4\sim 0.1$ is sufficient
2728: to   destroy  the   three-sublattice  N\'eel   LRO  realized   by  the
2729: first-neighbor Heisenberg model.\cite{lmsl00}  The nature of the phase
2730: on  the  other  side  of  this  transition  is  not  settled  but  the
2731: finite-size spectra display a  large density of singlet excitations at
2732: low   energy    which   could   be   reminiscent    of   the   kagome
2733: situation.\cite{lmsl00}
2734: 
2735: From  exact diagonalizations (up  to 36 sites) no  sign  of N\'eel LRO
2736: could be found at $J_4=1,J_2=-2$.\cite{mblw98,mlbw99} In addition, the
2737: finite-size analysis showed  that the spin-spin  correlation length is
2738: quite short  at $J_2=-2$ and  $J_4=1$ and a spin  gap of  the order of
2739: $\Delta\sim 0.8$ exists at this  point.  Much of the numerical  effort
2740: to  elucidate the nature of the  MSE  ground-state was concentrated on
2741: this   point  because it    is  close to   the parameters  realized in
2742: low-density $^3$He films (when higher order exchanges are neglected).
2743: 
2744: 
2745: \subsubsection{Local singlet-singlet correlations  -
2746:   absence of  lattice   symmetry breaking}
2747: 
2748: Having  excluded the possibility of  a  N\'eel ordered ground-state at
2749: $J_4=1,J_2=-2$ it is  natural to look for  a possible VBC.  Because of
2750: the complexity of the  MSE Hamiltonian it   is not clear what kind  of
2751: spatial  order should  be favored.  From   the analysis of dimer-dimer
2752: correlations (see Fig.~\ref{fig:ddcorrel})  it  appears that  parallel
2753: valence-bonds repel each-other at  short distance.  This is similar to
2754: what is observed in the staggered phase of the $J_2$--$J_4$ MSE ladder
2755: and  square-lattice   models.  For  this   reason  it appears  that  a
2756: plausible VBC would be the staggered VBC encountered in the triangular
2757: QDM  for  $V>J$  (\S\ref{ssec:QDMTri}).   However this  scenario seems
2758: difficult   to    reconcile   with  the    weakness     of dimer-dimer
2759: correlations.\cite{mlbw99} In  addition, the low-energy singlet states
2760: and their quantum numbers\cite{StagVBC}  do  not reflect the   12-fold
2761: quasi-degeneracy   that should  be present    if  the  system  was  to
2762: spontaneously break some lattice symmetry according to a staggered VBC
2763: pattern.  Small systems usually  {\em  favor} ordered  phases  because
2764: low-energy and long-wavelength  fluctuations that could destabilize an
2765: ordered state are reduced  compared to larger  systems.  From the fact
2766: that the finite-size spectra do not show the signatures of a staggered
2767: VBC symmetry breaking it is unlikely that the  MSE model could develop
2768: a VBC of this kind in the thermodynamic limit.
2769: 
2770:  
2771: \begin{figure} 
2772: \begin{center}
2773: \includegraphics[width=8cm]{DD28_m210}
2774: \end{center}
2775: \caption[99]{Dimer-dimer correlations in the ground-state
2776: of  the $J_2$--$J_4$ MSE  model on the triangular  lattice  (28 sites) at
2777: $J_2=-2$, $J_4=1$ (from  Ref.~\cite{mlbw99}). Numbers are proportional
2778: to          $\langle\hat{d}_0\hat{d}_x\rangle        -\langle\hat{d}_0
2779: \rangle\langle\hat{d}_x\rangle$  where the  operator       $\hat{d}_x$
2780: projects onto the singlet space of the bond $x$ and $\hat{d}_0$ refers
2781: to the reference  bond $(1,28)$. These  results shows a clear tendency
2782: for repulsion between parallel dimers.}
2783: \label{fig:ddcorrel}
2784: \end{figure}
2785: 
2786: \subsubsection{Topological degeneracy and Lieb-Schultz-Mattis Theorem}
2787: 
2788: Because no  VBC   phase could be  identified   in  the  MSE model  at
2789: $J_4=1,J_2=-2$ the numerical  data were compared  with the predictions
2790: of an RVB liquid scenario.
2791: 
2792: In one  dimension    a  famous   theorem due   to  Lieb,   Schultz  and
2793: Mattis\cite{lsm61} (LSM) states  that in   a one-dimensional spin system
2794: with an half-integer spin in the unit cell there is at least on excited
2795: state collapsing to the ground-state in the thermodynamic limit
2796: (periodic boundary conditions).
2797: There are in fact several arguments suggesting that this theorem might,
2798: at least to some extent,
2799: also  apply to higher dimensions.\cite{al86,o00,mlms02,hastings03,o03}
2800: If  that is the case  a gapped system with  an odd integer spin in the
2801: unit cell must have a degenerate  ground-state.  The simplest scenario
2802: to  explain this degeneracy is a  translation  symmetry breaking.  One
2803: could   think that   this  would  rule  out  the  possibility of   any
2804: (translation invariant) RVB liquid  in such models.  This is incorrect
2805: because a ground-state degeneracy  can have a  topological origin on a
2806: system with  periodic boundary  conditions,  as  we discussed in   the
2807: framework of QDM (\S\ref{sec:QDM}).  Such  a phase is characterized by
2808: a four-fold topological ground-state degeneracy  when the system is on
2809: a torus.   That    degeneracy  allows the  system   to  fulfill  LSM's
2810: requirement   without     any spontaneous       translation   symmetry
2811: breaking.\cite{mlms02}
2812: 
2813: On a finite-size system the topological degeneracy is only approximate
2814: but  some constraints   exist for the  quantum  numbers  (momentum  in
2815: particular)  of the quasi-degenerate  multiplet.\cite{mlms02} A system
2816: with periodic boundary conditions with an even  number of sites but an
2817: {\em odd  number of rows}  is expected to  have two ground-states with
2818: differ by a momentum $\pi$ in  the direction parallel  to the rows, in
2819: close analogy to the  the LSM theorem in  dimension one. The numerical
2820: spectra of the MSE model exhibit a set  of three singlet energy levels
2821: collapsing  onto     the  ground-state  when  the   system     size is
2822: increased\cite{mlbw99} and their    quantum number  turn  out  to   be
2823: consistent with  the   constraints   derived from the      general RVB
2824: picture.\cite{mlms02}
2825: 
2826: \subsubsection{Deconfined spinons}
2827: 
2828: The  SL phase described above is   expected to have deconfined spinons
2829: ($S=\frac{1}{2}$ excitations).  These excitations should show up as an
2830: incoherent   continuum    in   the   spin-spin    dynamical  structure
2831: factor.  However such a feature would  probably be rather difficult to
2832: observe on small 2D lattices, in particular due to the small number of
2833: inequivalent ${\bf k}$-vectors  in  the Brillouin zone.  On  the other
2834: hand, the binding energy of two spinons can  be evaluated by comparing
2835: the ground-state  energy and the first   magnetic excitation energy on
2836: even and odd samples.  In the case  of the MSE model at $J_4=1,J_2=-2$
2837: the results show the existence of  a bound-state (it is more favorable
2838: to put  two spinons in the  same  small sample than  in separate ones,
2839: which is not surprising) but this {\em does  not mean that the spinons
2840: are confined} (contrary  to  the conclusions  of  Ref.~\cite{mlbw99}).
2841: Interestingly this binding energy seems to  go to zero for the largest
2842: available sizes~(Fig.~\ref{spinon}):  this might  indicate the absence
2843: of attraction between   spinons for  large  enough  separation and  an
2844: asymptotic deconfinement.
2845: 
2846: \begin{figure}
2847: \begin{center}
2848: \includegraphics[height=4cm,clip=true]{spinonmse}
2849: \caption[99]{Spinons binding energy as a function of the system size in
2850: the MSE model at $J_2=-2$ and  $J_4=1$. The vertical bars correspond to
2851: the  range    of    values found     for  different  sample    shapes.
2852: }\label{spinon}
2853: \end{center}
2854: \end{figure}
2855: 
2856: It is important to  stress here that  the RVB SL phase discussed  here
2857: (and    its   QDM    counterparts    of     \S\ref{ssec:QDMTri}    and
2858: \S\ref{ssec:QDMKag})  is not the only  way to  spinon deconfinement in
2859: 2D. There is  at least another scenario, inherited from  one
2860: dimension, which is the  {\em sliding Luttinger liquid}.  Indeed,  the
2861: Luttinger liquid behavior and the one-dimensional critical behavior of
2862: magnetic  chains  seem  to  be  robust  to  small (or  moderate)  {\em
2863: frustrating transverse couplings between chains},  as observed both in
2864: theoretical\cite{nge98,ahllt98,efkl00,be01,vc01,sp02,ssl02}        and
2865: numerical  approaches.\cite{sfl02}  This  regime between  one  and two
2866: dimensions which may have been observed in Cs$_2$CuCl$_4$\cite{cttt01}
2867: is the subject of a number of recent studies.\cite{be01,vc01}
2868: 
2869: 
2870: \subsection{Other models with MSE interactions}
2871: 
2872: Multiple-spin interactions are present in a number of models that were
2873: found   to exhibit fractionalization   or  an RVB liquid ground-state.
2874: Well known examples of MSE interactions with an Ising symmetry are $\Z$
2875: gauge theories, where the gauge  invariant plaquette term is a product
2876: of    Pauli  matrices $\prod_i   \sigma^z_i$.    Such  theories have a
2877: deconfined phase in 2+1 dimension\cite{kogut79} and their relevance to
2878: fractionalized phases of  2D  electronic systems has  been
2879: pointed  out by  Senthil and Fisher.\cite{sf00}  The connexion between
2880: $\Z$ gauge theories and QDM was mentioned in
2881: \S\ref{sssec:Z2}. Some MSE  spin models with  an Ising  symmetry and a
2882: fractionalized  ground-state were  discussed by Kitaev,\cite{kitaev97}
2883: Nayak and   Shtengel.\cite{ns01} In the other  limit  of a  $U(1)$ (or
2884: $XY$) symmetry several models  have been studied.  Recent examples are
2885: based  upon  the   spin-$\frac{1}{2}$  four-spin   XY  ring   exchange
2886: interaction\cite{bfg02,pbf02,sdss02,sm02}
2887: \begin{equation}
2888:   \mathcal{H}=-K\sum_{\langle ijkl\rangle} \left(
2889:   S_i^+S_j^-S_k^+S_l^- + {\rm H.c}
2890: \right)
2891: \label{eq:4XY}
2892: \end{equation}
2893: which is the $XY$ analog of the $SU(2)$ MSE interaction $P_{1234}+{\rm
2894: H.c}$. 
2895: 
2896: %______________________________________________________________________________
2897: 
2898: \section{Antiferromagnets on the kagome (and related) lattices}
2899: \label{sec:kagome}
2900: 
2901: 
2902: In RVB SL phases described in \S\ref{ssec:QDMTri}, \S\ref{ssec:QDMKag}
2903: and \S\ref{ssec:RVBMSE} {\em singlet excitations (visons) are gapped}.
2904: In the  $\Z$ gauge  theory  approach this  gap  is essential  for  the
2905: consistency  of the   theory.\cite{sf00,sf01} The   gauge-field  quasi
2906: particles  are  vortices   of the    gauge field  and   carry  a  unit
2907: $\mathbb{Z}_2$ gauge flux but no spin.  As illustrated in
2908: \S\ref{sssec:QDMdeconf} in the framework  of a simple QDM, visons have
2909: long-ranged interaction    with spinons.  If   the   spectrum of these
2910: visons has a gap  then the spinons  are unconfined and the phase is
2911: ``fractionalized''.  If  they   condense, the  long range  interaction
2912: between them and the spinons frustrate the motions of the latter which
2913: remained confined.  The  gap    in  the singlet sector     (above  the
2914: topological degeneracy)   is thus a  crucial  ingredient of  these RVB
2915: SL. At  the  end  of this   section we  will  show how    the spin-$\frac{1}{2}$
2916: first-neighbor Heisenberg model   on the kagome lattice  represents an
2917: enigma with  respect to this scheme  (as well as  the QDM described in
2918: Sec.~\ref{ssec:QDMmu}).  We  begin  the section by   a review  of  the
2919: properties of various models on  the kagome lattice, which topology is
2920: the source of an extreme  frustration, and, in  many models, source of
2921: an extensive degeneracy of the ground-state.
2922: 
2923: \subsection{Miscellaneous models on the kagome lattice}
2924: 
2925: There  has been  a large    number  of studies  devoted to   different
2926: antiferromagnetic  models  on the  kagome  lattice.  The next neighbor
2927: Ising model on such a  lattice is disordered, its  entropy per site is
2928: very large $S_{\rm   kag}^{\rm  Ising}= 0.502$,  more   than half  the
2929: independent spin value, much  larger than the triangular lattice value
2930: $S_{\rm tri}^{\rm Ising}=  0.323$   and   of  the order of     Pauling
2931: approximation      for    independent       triangles          $S_{\rm
2932: Pauling}=0.501$.\cite{p38} This suggests that the correlations in this
2933: system   are very  weak:   the   model   remains disordered at     all
2934: temperatures.\cite{kn53,hr92} Moessner  and Sondhi have  studied  this
2935: Ising model   in  a transverse magnetic   field  (the simplest way  to
2936: include some quantum fluctuations  in the model):  the model  fails to
2937: order for any transverse field, at any temperature.\cite{msc00,ms01b}
2938: 
2939: The nearest neighbor classical Heisenberg  model on the kagome lattice
2940: has  also a  huge ground-state degeneracy.    This property is easy to
2941: understand and  holds on different lattices  with corner sharing units
2942: such as the checkerboard  lattice or the three  dimensional pyrochlore
2943: lattice (Moessner and Chalker\cite{mc98,mc98a}). On all these lattices
2944: the nearest neighbor Heisenberg Hamiltonian  can be written as the sum
2945: of the square of the total spin $\vec{S}_{\alpha}$ of individual units
2946: $\alpha$ (a   tetrahedron in the  2-d and  3-d pyrochlore cases  and a
2947: triangle   for the     kagome  lattice),   which   share only      one
2948: vertex. Classical ground-states are obtained whenever $\forall\alpha\;
2949: \vec{S}_{\alpha}=\vec{0}$.    This     condition fixes   the  relative
2950: positions of the three classical spins of a  triangle at $120$ degrees
2951: from each   other  in a  plane.   But  it does  not  fix  the relative
2952: orientation of the plane of   a triad with  respect  to the planes  of
2953: triads on corner sharing triangles:  the model has a continuous  local
2954: degeneracy\cite{chs92,hr92}    at  $T=0$.\footnote{ Counting  the {\em
2955: planar}  ground-states amounts to determine in  how  many ways one can
2956: associate  one of the three  letters $A$, $B$ and  $C$ to each site so
2957: that each triangle  has spins along  the three different orientations.
2958: This  already  represents  an extensive  entropy.\cite{baxter70,hr92}}
2959: Thermal                fluctuations        select             coplanar
2960: configurations.\cite{chs92,hr92,rcc93}  The possibility of  long-range
2961: order in  spin-spin  correlations at   very low temperature   has been
2962: discussed without any definitive conclusion.\cite{hr92,rb93} The order
2963: parameter of  the planar  phase   is defined  by the   local  helicity
2964: (sometimes called vectorial chirality)~:
2965: \begin{equation}
2966: \vec{\zeta}=
2967: 	\vec{S}_1\times\vec{S}_2
2968: 	+\vec{S}_2\times\vec{S}_3
2969: 	+\vec{S}_3\times\vec{S}_1
2970: \end{equation}
2971: where the  three sites   define a triangle.   This   kind of order  is
2972: sometimes called  {\em nematic} by  analogy  to liquid  crystals.  The
2973: existence   of  such an  order    parameter might  be related   to the
2974: instability of the classical Heisenberg model on the kagome lattice to
2975: Dzyaloshinsky-Moriya interactions.\cite{e02,ecl02} The classical model
2976: has    a    large  density    of    low-lying  excitations    at   low
2977: temperature.\cite{k94}
2978: 
2979: 
2980: \subsection{
2981: Spin-$\frac{1}{2}$  Heisenberg  model on    the   kagome lattice:  an    extreme
2982: play-ground for ``quantum fluctuations''}
2983: 
2984: The  nearest neighbor spin-$\frac{1}{2}$ quantum  Heisenberg model on the kagome
2985: lattice   has        equally    been     the     object      of   many
2986: studies\cite{e89,ce92,s92,le93,ze95,lblps97,web98,m98,smlbpwe00,mm01,dmnm03}
2987: From these studies the following important facts have emerged :
2988: 
2989: \subsubsection{Ground-state energy per spin}
2990: The Heisenberg model on the kagome lattice has an extremely low energy
2991: per  bond ($\langle{\bf  S}_i.{\bf  S}_j\rangle\simeq  -0.44)$ $  \sim
2992: 87\%$ of the energy per bond in  an isolated triangle. On this lattice
2993: the energy per bond  of  the spin-$\frac{1}{2}$  system  is much lower than  the
2994: classical energy $\frac {E_{qu.}}{E_{cl.}}  \sim  1.74$, a ratio  much
2995: larger than   in any other  2D magnet,  that can  only be
2996: compared to the value obtained for the Bethe chain (1.77).  The kagome
2997: lattice   is  the 2D lattice   which  offers the largest
2998: stabilization due to quantum fluctuations.
2999: 
3000: \subsubsection{Correlations}
3001: 
3002: The  ground-state  is disordered.\cite{ze90}  Within  accuracy of  the
3003: finite-size numerical computations, spin-spin correlations,\cite{le93}
3004: dimer-dimer          correlations          (Fig.~\ref{dim-dimkagMSE}),
3005: chirality-chirality correlations\cite{ce92} are short-ranged, which is
3006: consistent   with   the    previous   point   and   series   expansion
3007: results.\cite{sh92}
3008: \begin{figure}
3009: \begin{center}
3010: \includegraphics[height=6cm]{dimer-dimerMSEKAG}
3011: \caption[99]{Dimer-dimer correlations in the ground-state of the
3012: kagome  Heisenberg   model (black  triangles)  and   in the  MSE model
3013: discussed    in     \S\ref{ssec:RVBMSE}    (open     symbols)   versus
3014: distance. Although the decrease of these correlations is weaker in the
3015: KH  model  than  in the  MSE  model,   it seems  nevertheless  roughly
3016: exponential in the  first two decades,  as  the spin-spin correlations
3017: are.\cite{ce92,le93}}\label{dim-dimkagMSE}
3018: \end{center}
3019: \end{figure}
3020: 
3021: \subsubsection{Spin-gap}
3022: There is plausibly a spin-gap  of the order of  1/20$^{\rm th}$ of the
3023: coupling  constant.\cite{web98} In view of     the smallness of   this
3024: spin-gap with regards to the available sizes caution is necessary. The
3025: above  conclusion  was drawn from  the  value of the {\em microscopic}
3026: spin  gap    $E_0(S=S_{min} +1)-E_0(S=S_{min})$ on    exact spectra of
3027: samples with up to 36 spins,\cite{web98} where  $E_0(S)$ is the lowest
3028: energy in the $S$ sector, and $S_{min}=$  0 or $\frac{1}{2}$ depending
3029: on the  parity of the total  number of spins.  The finite-size effects
3030: on these results are an order of magnitude  smaller than in a N{\'e}el
3031: ordered antiferromagnet. Nevertheless   they are still  not negligible
3032: for these sizes. An  alternative determination of  the spin-gap can be
3033: obtained  along the following  line.   The lowest exact eigenstate  in
3034: each total  spin sector $S$ of a  $N$ spins  sample defines the energy
3035: per spin  $e$  at  $T=0$  as  a function  of   its magnetization  $m =
3036: S/(N/2)$.  For  low positive values of  the magnetization, one can fit
3037: $e(m)$ to    the phenomenological  law:\footnote{This phenomenological
3038: form cannot  extend  below $m=0$  and beyond  $m=1/3$ because  at both
3039: magnetizations an angular point appears in $e(m)$ with a discontinuity
3040: of    the      first    derivative    signaling     a    magnetization
3041: plateau.\cite{h01,cghp02} This   is discussed in  more details  in our
3042: lectures notes.\cite{lm02}}
3043: \begin{equation}
3044: e(m) = e(0) + a m + b m^2/2 +{\cal O}(m^3)
3045: \label{eofm}
3046: \end{equation}
3047: $a$ and  $b$ are intensive  quantities which depend  on $N$ but should
3048: converge to some (possibly zero) value when $N\to\infty$.  This point
3049: of view is important because for an  infinite system the thermodynamic
3050: function  $e(m)=\lim_{N\to\infty}E(S=mN/2)/N$ can be measured (through
3051: the zero    temperature        magnetization    curve),    but      not
3052: $E(S\sim\mathcal{O}(1))$.  The physical significance of Eq.~\ref{eofm}
3053: is clear: $a$ measures half  the spin gap  (in a thermodynamic  sense,
3054: that is  the magnetic field $H_c$  required to magnetize the system at
3055: zero temperature) and
3056: \begin{equation}
3057: b=\frac{\partial^2 e}{\partial m^2} = \chi^{-1}
3058: \end{equation}
3059: where  $\chi$ is  the   homogeneous susceptibility of  the  medium for
3060: fields larger than the  critical field $H_c=a$.  This determination of
3061: the {\em thermodynamic} spin gap leads to a renormalization of the raw
3062: data\cite{web98} for small  sizes (see Fig.~\ref{gapkago}).  A  linear
3063: extrapolation  versus $1/N$  (which should give  a lower  bound of the
3064: spin-gap)  leads  to   the     value  $0.06$  for   the     spin   gap
3065: ($e_{\infty}=-0.4365$  and $\chi_{\infty}=0.34$).  This  determination
3066: is in  agreement with the direct extrapolation  of the microscopic gap
3067: $E_0(S=S_{min}  +1)-E_0(S=S_{min})$.    For   $N=36$ at  the  smallest
3068: non-zero  magnetization,  the linear  term of equation~(\ref{eofm}) is
3069: $90\%$  of the quadratic term: this  is an  estimate  of the degree of
3070: confidence on the existence of a spin-gap.
3071:  
3072: \begin{figure}
3073: \begin{center}
3074: \includegraphics[height=5cm]{gapkago}
3075: \caption[99]{Finite size scaling of the spin gap in the spin-$\frac{1}{2}$
3076: Heisenberg model ${\cal H} =  \sum_{<i,j>} {\bf S}_i.{\bf S}_j$ on the
3077: kagome lattice.  Triangles (resp. squares) are  the raw results of the
3078: microscopic   spin  gap in the  even    (resp.  odd) samples.  Bullets
3079: represent the  ``thermodynamic'' spin gap,   they are obtained by  the
3080: procedure described in   the text  (errors  bars  come  from  the  rms
3081: uncertainty in the fits).}\label{gapkago}
3082: \end{center}
3083: \end{figure}
3084: 
3085: 
3086: 
3087: \subsubsection{An exceptional density of low lying excitations
3088:  in the singlet sector}
3089: 
3090: Whatever the ultimate fate of the spin gap a big surprise emerges from
3091: the exact spectra:  the probable absence of gap  in the singlet sector
3092: and  the   anomalous density  of  low  energy  states  adjacent to the
3093: ground-state.  Let us first comment  the second point in details: even
3094: on the smallest size spectra the low lying states appear contiguous to
3095: the  ground-state and the spectra are  extremely dense.  The number of
3096: singlet  levels in the spin-gap  (taken as a natural energy band-width
3097: of the problem) increases exponentially fast with $N$ as $\sim 1.15^N$
3098: as    far   as  the     $N\leq36$    systems   are   concerned    (see
3099: Fig.~\ref{numberostate}). This property remains  (and can be checked on
3100: larger systems) when   the  Hilbert space   is restricted  to that  of
3101: first-neighbor valence-bond coverings.\cite{mm01,msp03}
3102: 
3103: \begin{figure}
3104: \begin{center}
3105: \includegraphics[height=5cm]{numberostate1}
3106: \caption[99]{Logarithm of the number of singlet states in the spin
3107: gap versus sample size (black   squares).  The short dashed and   long
3108: dashed curves display the  theoretical law (Eq.~\ref{law}  with $d=2$)
3109: (short dashes: $p=1$, long dashes: $p=2$).  }\label{numberostate}
3110: \end{center}
3111: 
3112: \end{figure}
3113: 
3114: 
3115: Some remarks  are necessary to  fully appreciate  this  property.  The
3116: $2^N$  states of the  system are stretched on an   energy scale of the
3117: order  of  $NJ$    where  $J$ is    the   coupling  constant   of  the
3118: Hamiltonian. This implies that on most of the  spectrum the density of
3119: states increases exponentially with $N$. If we specialize to the $S=0$
3120: sector as we will  do below, the  picture  is not very different:  the
3121: number of states is $ C^N_{\frac{N}{2}}-C^N_{\frac{N}{2}-1} \sim {\cal
3122: O}(\frac{2^N}{N})$ and here too,  in most of  the spectrum the density
3123: is exponentially increasing with  $N$. But in  all the phases  that we
3124: have studied up to now, the nature of the  ground-state and of the low
3125: lying excitations leads to  a different behavior  at the bottom of the
3126: spectrum.  The ground-state  degeneracy  is ${\cal O}(1)$ in
3127: VBC,  VBS  and in the  RVB  SL  (of the  type  discussed in
3128: \S\ref{ssec:QDMKag} and \S\ref{ssec:RVBMSE} for  instance) and  it is
3129: ${\cal O}(N^p)$  in N{\'e}el ordered  states with $p$ sublattices.  In
3130: all these situations the low-lying  excitations are described as modes
3131: or quasi-particles.
3132: %In an  energy range of  order  $\mathcal{O}(J)$ one counts $N^{\beta}$
3133: %levels associated with single-particle excitations.  This always leads
3134: %to density of states  increasing as a power  law as a function of $N$.
3135: %Inclusion of multi-particle excitations can be done in an average way:
3136: Let us
3137: suppose that single-particle excitations have a dispersion law:
3138: \begin{equation}
3139: \omega({\bf k}) \sim k^p.
3140: \label{dispersionlaw}
3141: \end{equation}
3142: In a   d-dimensional system the total   internal energy increases with
3143: temperature as $E \sim N T^{(p+d)/p}$, the specific heat as $C_v\sim N
3144: T^{d/p}$ as   well as  the  entropy  ${\cal S}\sim   N T^{d/p}$.  In a
3145: micro-canonical point of view   the density of  states $\rho(E)$  or a
3146: large  enough  system is simply  related  to   the entropy  by  ${\cal
3147: S}=\log(\rho(E))$. Since ${\cal S}\sim N T^{d/p}=N (E/N)^{d/(p+d)}$ we
3148: obtain
3149: \begin{equation}
3150: \log (\rho(E)) \propto N^{\frac{p}{p+d}}
3151: \label{law}
3152: \end{equation}
3153: 
3154: As an example, let us consider the Rokhsar  Kivelson QDM on the square
3155: lattice at   the  quantum critical    point (Sec.~\ref{sec:QDM}):  the
3156: dispersion law  of the so-called resonons is  quadratic
3157: around ${\bf Q}=  (\pi,\pi)$,  the logarithm  of the number  of states
3158: increases as $N^{1/2}$ (long dashes of Fig.~\ref{numberostate}).
3159: 
3160: Even with such many-particle excitations  one would expect a number of
3161: levels increasing  more    slowly  than in   the   numerical  spectra.
3162: Infinitely     soft low-energy modes     ($p \rightarrow \infty$), are
3163: necessary  to recover a density  of low-lying levels  growing as $\sim
3164: \alpha^N$.  It is still unclear if we can  do a connection between the
3165: ``zero modes'' of the classical model at $T=0$ and  this picture.  And
3166: we cannot  completely   indulge   ourselves in   saying   that quantum
3167: fluctuations are unable  to  lift the classical degeneracy  as quantum
3168: fluctuations seem to open a spin gap.
3169: 
3170: A  physical consequence  of   this exceptional density  of  low  lying
3171: singlets can be observed in the  specific heat: at low temperature the
3172: specific heat of this  spin system is  unusually large, with  a double
3173: peak    structure,\cite{ze90,ey94,nm95,tr96} insensitive to relatively
3174: large magnetic fields.\cite{smlbpwe00} This is easily understood if we
3175: suppose that in this energy range there is  a large density of singlet
3176: states.\cite{smlbpwe00}   This result   is   to  be compared   to  the
3177: experimental   results of   Ramirez    {\it   et   al.}\cite{rhw00} on
3178: SrCr$_{9p}$Ga$_{12-9p}O_{19}$   (notice   however    that  it   is   a
3179: spin-$\frac{3}{2}$   compound  and  that  numerical calculations  were
3180: performed on the  spin-$\frac{1}{2}$ model)  where the specific   heat
3181: around $5$~K has an extremely low sensitivity to magnetic fields up to
3182: $10\;$~Tesla, whereas the homogeneous  susceptibility in this range of
3183: temperature  is probably  very  low if  we notice  that it turns  down
3184: around $50 K$.\cite{mklmch00}
3185: 
3186: \subsubsection{Absence of gap in the singlet sector}
3187: 
3188: Up  to  $N=36$ sites there is   no evidence of   a possible gap in the
3189: singlet    spectrum: this is    an exceptional  phenomenon in  quantum
3190: mechanics of small systems where discretization is usually the rule.
3191: 
3192: It  has been   advocated  in recent  papers\cite{sma02,ns03}  that the
3193: ground-state of this model could  break the translational symmetry and
3194: be a VBC.   The first proposed crystal\cite{sma02}  is made of resonating  {\em
3195: stars} with 6 dimers.  The corresponding unit  cell has 12 spins.  The
3196: second  VBC\cite{ns03}   is  made   of  resonating  (trimerized)  {\em
3197: hexagons} and was already discussed as the most reasonable crystal (if
3198: any is realized in this system, which we think is not clear at all) by
3199: Marston and  Zeng\cite{mz91} and Zeng  and Elser.\cite{ze95} This later
3200: VBC has a unit cell with 36 sites. In both  scenarios it is the energy
3201: gain obtained by {\em local  resonances} (involving respectively 6 and
3202: 3 valence-bonds) which drives the system toward a  VBC. From the energy
3203: point  of view  the  star  VBC is   (by far)  less realistic  since it
3204: involves a much longer  resonance loop.\footnote{Notice also that  the
3205: gap associated to the star VBC is predicted to be of the order of 1/10
3206: of $J$,\cite{sma02} which is 10 times larger than the largest distance
3207: between two consecutive levels in  the spectrum of the $N=36$ sample.}
3208: This  resonance  loop involving 6 valence-bonds  around  a star  has a
3209: vanishing amplitude at  the  lowest non-trivial  order of the  overlap
3210: expansion  in   the   RVB  subspace,  as  was    shown  by Zeng    and
3211: Elser\cite{ze95}.
3212: To our  opinion there is no  reason to think  that the physics  of the
3213: kagome model can be described from the limit of weakly coupled stars.
3214: On the   other  hand, in  the approximation  where  only the  shortest
3215: resonance  loops are present,  the model was  indeed found to be in the
3216: {\em  hexagon} VBC phase,  as re-discovered  recently from a different
3217: point   of view.\cite{ns03} A crucial  (numerical)  result of Zeng and
3218: Elser\cite{ze95}  is  however that  this VBC {\em   melts} when higher
3219: order resonances loops are included.
3220: 
3221: 
3222: We have studied  the 36-sites sample  which can accommodate these  two
3223: VBC: the low-lying levels of the spectrum  do not give a clear picture
3224: of the supposed  to be VBC.   The eigenstates with the quantum numbers
3225: corresponding to these two VBC are not the lowest energy levels in the
3226: spectrum.   No   gap is seen  in   the spectrum, the  largest distance
3227: between two consecutive states  is $10^{-2}$ and seems distributed  at
3228: random, whereas the average distance between two consecutive states in
3229: the 50 first  states is: $2.5 10^{-3}$.  It  has been argued that long
3230: wave-length  quantum fluctuations (almost  absent in the $N=36$ sample
3231: which contains only  two  resonating hexagons  or three  stars)  could
3232: eventually restore the order.  We think  that this is incorrect.  In a
3233: VBC it  is {\em local resonances}  which favor the crystal  and longer
3234: resonances  which tend to  reduce the order  parameter and which could
3235: eventually destabilize it.  Long-wavelength   will tend to restore   a
3236: larger ground-state symmetry (reducing   the degeneracy).   From  this
3237: point of view it is unlikely that a VBC  pattern will appear in larger
3238: systems if it  is not apparent  in the smallest systems (provided that
3239: boundary conditions do not frustrate the  corresponding VBC).  We thus
3240: consider that up to now numerical results do not support the claims of
3241: a star or hexagon VBC ground-state in this model.
3242: 
3243: If the  hexagon VBC was however  realized the important question would
3244: probably be why the associated gap is  so tiny, certainly smaller than
3245: $10^{-3}$. A natural scenario would be  that of a very close proximity
3246: to  a quantum critical  point, but  that critical  point remains to be
3247: identified.  A third possibility would be that of a system with a true
3248: extensive entropy at zero  temperature.   This is  less likely in  our
3249: opinion since (up to now)  only  models with  simple (if not  trivial)
3250: local symmetries were found to  have such ground-state entropy  (Ising
3251: antiferromagnet on the triangular lattice for instance or the QDM of
3252: \S\ref{ssec:QDMmu}) and this entropy is usually sent to (possibly
3253: small but) finite temperatures by almost any perturbation.
3254:  
3255: \subsubsection{Anomalous density of states in other spin sectors}
3256: An  anomalous density of low lying  states is  equally observed in the
3257: spin $\frac{1}{2}$ sector (where the law could be fitted  to $N 1.15^N$), in the
3258: spin 1 sector as well  as in other sectors  with larger total spin. It
3259: should be noticed that such a density of states implies the absence of
3260: an intrinsic energy scale for the low  lying excitations: a phenomenon
3261: that    has  been   observed    in   inelastic    neutron   scattering
3262: (Ref.~\cite{mmpfa00} and  Refs.   therein)  and theoretically  in  the
3263: imaginary  part of  the  dynamic susceptibility  calculated within the
3264: dynamical   mean field theory (Georges  {\it  et al.}\cite{gsf01}).  A
3265: high spin susceptibility just above the spin  gap is not excluded
3266: in the spin-$\frac{1}{2}$.\cite{ls02}
3267: 
3268: The global picture of  this phase is thus that  of a SL with
3269: no  long-ranged correlations in   any local observable,  and  an large
3270: entropy of singlets   at $T\ll J$, which is   a manifestation  of  the
3271: extraordinary large density of states in each $S$ subspace.
3272: 
3273: \subsection{
3274: Next-neighbor   RVB   description  of  the  spin-$\frac{1}{2}$  kagome
3275: antiferromagnet} Considering a supposed-to-be large spin-gap, Zeng and
3276: Elser\cite{ze95}    proposed a description     of the ground-state and
3277: low-lying excitations  of  the  kagome model  in  the  basis  of  next
3278: neighbor  valence bonds.   They analyzed  in  this  context the  dimer
3279: dynamics and showed on a $N=36$ sample that  the hexagon VBC --favored
3280: by  the shortest (three-)dimer moves--  melts  when introducing higher
3281: order tunneling.  Mila and Mambrini\cite{m98,mm01} confirmed that this
3282: reduced Hilbert space of next neighbor  valence bonds captures some of
3283: the most  perplexing features  of this   magnet and  specifically  the
3284: absence of (measurable) gap in  the singlet sector and the exponential
3285: number of singlets.   One of us, D.~Serban and V.~Pasquier\cite{msp03}
3286: have  elaborated on  this  work and on   Zeng and Elser's approach and
3287: proposed a  QDM  with an  extensive  zero-point entropy   and critical
3288: (energy-energy\footnote{Dimer-dimer correlations are  short-ranged  in
3289: this  model.})  correlations  (see   \S\ref{ssec:QDMmu}).    All these
3290: results  point to an absence of   an intrinsic low-energy scale.  This
3291: feature  is typical of a   critical state, but  as  seen in the  above
3292: discussion, the simple RK  picture does not  seem to fit nicely to the
3293: exact diagonalization  data: may be the  available sizes are too small
3294: or the behavior of  this    quantum system corresponds to    something
3295: completely new.   Some recent  numerical results\cite{dmnm03} (in  the
3296: full spin-$\frac{1}{2}$ Hilbert space as well as  in the RVB subspace)
3297: showed that  (static)  non-magnetic impurities  (holes)  experience an
3298: unexpected {\em   repulsion} in this   system and that  no significant
3299: magnetic moment is created  in the vicinity of  the impurities. It has
3300: been argued\cite{sv00} that static  non-magnetic impurities are useful
3301: to detect   a    possible spinon  deconfinement  in    two-dimensional
3302: antiferromagnets. From this point  of view the results mentioned above
3303: suggest such a deconfinement.
3304: 
3305: 
3306: \subsection{Experiments in compounds with kagome-like lattices}
3307: 
3308: The  low  temperature specific heat  of SrCr$_9$Ga$_{12}$0$_{19}$ (the
3309: magnetic chromium  ion has   a spin 3/2)  is apparently   dominated by
3310: singlet states.\cite{rhw00}  The magnetic excitations of this compound
3311: as seen by muons can be  described as spins $\frac{1}{2}$ itinerant in
3312: a sea of singlets.\cite{ukkll94} The non-linear spin susceptibility of
3313: SrCr$_9$Ga$_{12}$0$_{19}$ exhibits a very large increase at about 5 K,
3314: reminiscent of  spin glasses,\cite{rec90} but  neutrons and muons show
3315: that  a very significant fraction  of the spins  are  not frozen below
3316: this   temperature     and       exhibit      still   very       rapid
3317: fluctuations.\cite{lbar96} The same  phenomena  have been observed  in
3318: two jarosites that are equally  good models of kagome antiferromagnets
3319: with half-odd-integer spin per unit cell.\cite{kklllwutdg96,whmmt98}
3320: 
3321: \subsection{ "Haldane's conjecture"}
3322: Whereas the classical Heisenberg model on the kagome, checkerboard and
3323: pyrochlore  lattices    share  the  properties   of  local  continuous
3324: degeneracy and disorder at $T=0$, their quantum counterparts are quite
3325: different.   As it   has   been explained  in Sec.~\ref{sec:VBC},  the
3326: Heisenberg model  on the checkerboard  lattice has an ordered VBC with
3327: gaps     to  all excitations.  In    contrast    to the   case of  the
3328: $S=\frac{1}{2}$ Heisenberg model on the kagome lattice, recent results
3329: from Hida\cite{h00} show that there is a large  gap to all excitations
3330: in the $S=1$ Heisenberg model on this same  lattice, in agreement with
3331: experiments.\cite{aoyhimw94,wkyoya97,wwyoaonn98}
3332: 
3333: Less  is known on  the ground-state   of the Heisenberg  model on  the
3334: three-dimensional pyrochlore  lattice:   Canals and Lacroix\cite{cl98}
3335: have shown that the spin-spin  correlations are short ranged and  they
3336: have observed  that on a  16-sites spectrum the first  excitations are
3337: singlet ones.  Fouet  {\it et al.}   have  computed the  spectrum of a
3338: 32-sites pyrochlore sample~\cite{fsl03},  this work confirms  that the
3339: first excitations are still  singlets     for this size.  There     is
3340: plausibly a degeneracy of the ground-state  in the thermodynamic limit
3341: but    no evidence of  a  closing  of the   gap  above: the system may
3342: therefore  be a VBC.  A  first description of  the  singlet sector was
3343: proposed by Harris, Berlinsky and Bruder\cite{hbb91} and was developed
3344: further by Tsunetsugu\cite{t01,t02}.   It  starts  from the  limit  of
3345: weakly-coupled tetrahedron  (and thus breaks some spatial symmetries).
3346: The ground-state of an isolated tetrahedron is two-fold degenerate and
3347: can  be described by an Ising  pseudo-spin.   An effective Hamiltonian
3348: describing  the   interactions   between    these   pseudo-spins   was
3349: written\cite{hbb91} and  analyzed  in  a semi-classical approximation.
3350: It  was  concluded that  the system maybe    a VBC. It   was argued by
3351: Tsunetsugu  that  a    soft    mode  could exist   in    the   singlet
3352: sector.\cite{t02}.  The CORE approach of Berg {\it et al.}\cite{baa03}
3353: seems more appropriate  to deal with  these systems where dimerization
3354: is probably the dominant  phenomenon. They treated larger units (block
3355: of four tetrahedron) and  concludes to the existence  of a VBC  with a
3356: larger unit cell than the one predicted before\cite{hbb91,t01,t02} and
3357: a small singlet gap.
3358: 
3359: 
3360: All these results  seem  to    confirm a  2D version  of     Haldane's
3361: conjecture:  among  these  frustrated  systems with  local  continuous
3362: degeneracies  in the  classical  limit,  the spin-$\frac{1}{2}$ kagome
3363: antiferromagnet is the only system to have an half-odd integer spin in
3364: the unit cell.  It is maybe not by chance that it is the only one with
3365: gapless excitations.\footnote{The SCGO compound  with 7 spins 3/2  by
3366: unit   cell  belongs to   the  same  category    and  as  also gapless
3367: excitations.\cite{rhw00}} The spin-$\frac{1}{2}$ Heisenberg model   on
3368: the checkerboard lattice or on  the pyrochlore lattice and the  spin-1
3369: Heisenberg model on the kagome lattice  have integer spins in the unit
3370: cell and  quantum fluctuations lead to  gapful excitations. It is also
3371: interesting to note  that     these results are  consistent    with  a
3372: generalization     of     the      LSM   theorem       in    dimension
3373: two.\cite{al86,o00,mlms02,hastings03,o03}
3374: 
3375: An  interesting    analysis of Von    Delft  and Henley\cite{vdh92-93}
3376: supports this   conjecture.   These  authors  studied   the collective
3377: tunneling of a small cluster of spins  between two spin configurations
3378: that  are degenerate in   the  classical limit.   They found  that for
3379: half-odd-integer  spins the tunneling  amplitude for  a cluster of six
3380: spins around an  hexagon (and on other  larger loops)  is exactly zero
3381: because      of    destructive    interferences    between   different
3382: symmetry-related instantons.  For small integer spins the interference
3383: is constructive and the tunnel amplitude  and the tunnel splitting are
3384: large: this is consistent with  numerical results which gives a  large
3385: gap for the $S=1$ kagome  antiferromagnet, and small  gaps (if any) in
3386: the spin-$\frac{1}{2}$ system.
3387: %_____________________________________________________________________
3388: %_____________________________________________________________________
3389: \section{Conclusions}
3390: \setcounter{footnote}{0}
3391: We  conclude  by summarizing  some   properties  --and  related  open
3392: questions-- of the different phases  discussed in  this review.
3393: 
3394: The      properties  of       these   phases  are     summarized    in
3395: Table~\ref{tab:conclusion}.  Semi-classical  phases with   N\'eel long
3396: ranged order, magnons as gapless  excitations, do exist in spin-$\frac{1}{2}$ 2D
3397: systems  with  moderate  frustration:  the  Heisenberg   model on  the
3398: triangular lattice is  the most  explicit  example, with a  sublattice
3399: magnetization about one  half of the classical value.\cite{bllp94} The
3400: ground-state degeneracy is in the thermodynamic limit a power of $N$.
3401: 
3402: An increased  frustration,  lower coordination number  or smaller spin
3403: lead  to  quantum phases, with a  ground-state  of higher symmetry, no
3404: long  ranged  order  in  spin-spin correlations,   a spin gap  and the
3405: restored $SU(2)$ symmetry. Two  main alternatives are then opened: the
3406: VBC or VBS  phases on one hand,  the RVB SL  on the other.  RVB SL and
3407: VBC   (as  well as   VBS) first  requires     the formation  of  local
3408: singlets.  When a particular local   resonance  pattern dominates  the
3409: dynamics of the Hamiltonian the system will try to maximize the number
3410: of occurrence of this pattern.  This is usually  achieved by a regular
3411: arrangement, that is a VBC. When  no such pattern dominates the system
3412: may form   a  translation invariant RVB   SL.  In  the  first case the
3413: ground-state  can  be     qualitatively described   by   one   ordered
3414: configuration of singlets dressed by small fluctuations. In the RVB SL
3415: the  amplitudes   of   the  wave-function   are  distributed  over  an
3416: exponentially large number of configurations. These ground-states lead
3417: to  very  different excitations:  $\Delta  S=1$ gapped  magnons in the
3418: first case (and  $\Delta S=0$ domain-wall excitations), gapped $\Delta
3419: S=0$ visons  and  gapped $\Delta S=\frac{1}{2}$  unconfined spinons in
3420: the   second case.
3421: 
3422:     In    agreement  with  the   large-$N$   results
3423: (Sec.~\ref{sec:largeN}), VBC or VBS  phases appear in general
3424: in quantum situations where the large-$S$ classical limit  displays collinear order
3425: \footnote{We have discussed some possible counter examples in section 3. 6},
3426:  whereas up to
3427: now RVB SL phases have only be encountered in range of parameters were
3428: the classical solutions are non collinear (MSE\cite{mblw98,mlbw99}
3429:  and $J_1$--$J_2$ on the
3430: honeycomb lattice\cite{fsl01}).
3431: 
3432: These states obey the 2D extension  of LSM theorem:  if $2S$ is odd in
3433: the unit  cell and if excitations are  gapped there is  a ground-state
3434: degeneracy in the thermodynamic limit (with periodic boundary
3435: conditions). However  the origin of the degeneracy  differs  in the two
3436: types   of  quantum phases.    In the   VBC  phases  the degeneracy is
3437: associated to spontaneously broken translation symmetry whereas in the
3438: RVB SL the   degeneracy  has a   topological origin.  In  the VBS  (or
3439: explicit VBC) the ground-state is unique but  $2S$ is even in the unit
3440: cell. As in  one dimension the LSM theorem  seems to play an important
3441: role and a  formal proof of its  validity in 2D  is perhaps very close,
3442:  if
3443: not   achieved.\cite{hastings03} A  RVB SL     may have been   observed
3444: numerically  in  a   spin-$\frac{1}{2}$   model  on   the    hexagonal
3445: lattice.\cite{fsl01} From  the  point of   view  of large-$N$ and  QDM
3446: approaches  a   topological  degeneracy   is   expected  in  such    a
3447: SL. Interestingly  that degeneracy is not  imposed by the 2D extension
3448: of LSM's theorem and has not been detected.\cite{fsl01}
3449: 
3450: 
3451: These paradigms    are relatively well   understood,  at least  on the
3452: qualitative level.  They also appear naturally  in the broader context
3453: of the   classification  of  Mott  insulators.\cite{sachdev03} However
3454: several kinds of  2D frustrated  magnets do  not fall in  these simple
3455: classes and many open  questions remain.
3456: 
3457: %As a first example, we mention the possibility of  a state RVB SL with
3458: %an {\em integer spin} in the unit cell.\footnote{Such a state may have
3459: %been observed numerically in the spin-$\frac{1}{2}$ $J_1$--$J_2$ model
3460: %on the hexagonal lattice.\cite{fsl01}}   The 2D extension of  the  LSM
3461: %argument fails to predict a (topological) degeneracy in that case.
3462: 
3463: 
3464: This review was restricted to $SU(2)$ invariant Hamiltonians.  Whereas
3465: the Ising limit has been much studied, the differences between quantum
3466: XY and Heisenberg models have received much less attention.
3467: 
3468: Chiral  SL   have  not been   discussed  in   this review.   They  are
3469: characterized by a broken time-reversal symmetry. This possibility has
3470: been    studied               intensely                   since    the
3471: 80's\cite{wwz89,kl87-89,ywg93,fradkinbook}.     To  the  best   of our
3472: knowledge   it  has not    yet  been  identified  in  a   realistic 2D
3473: model.\footnote{An example was  recently discovered in  one dimension:
3474: the scalar-chirality  phase of the spin ladder  model with  4-spin MSE
3475: interactions (see \S\ref{ssec:ladderMSE}).}
3476: 
3477: The issue of  quantum phase transitions in  frustrated antiferromagnet
3478: is also an active topic  that is not presented  in this review.   Many
3479: properties of these critical points are  still unknown, not to mention
3480: the fascinating problems associated with (quenched) disorder.
3481: 
3482: Limited by place (and competence) we have not discussed in details the
3483: works done   on spatially anisotropic  models. This  field which is in
3484: between one (review   by P.~Lecheminant  in  this  book) and 2D    is
3485: extremely flourishing tackled by bosonization and large-$N$ methods.
3486: 
3487: To conclude  we  would like  to emphasize  that  new analytical and/or
3488: numerical methods  are highly desirable to  proceed in the analysis of
3489: the two emblematic  problems by which  we have opened  and closed this
3490: review:  the $J_1$--$J_2$    model  on the  square   lattice   and the
3491: spin-$\frac{1}{2}$  Heisenberg  model on   the kagome (and pyrochlore)
3492: lattices.   In  both  of these  problems   a consensus  remains  to be
3493: obtained.
3494: 
3495: 
3496: \begin{sidewaystable}
3497: \tbl{Different phases encountered in $SU(2)$-symmetric frustrated models in 2D
3498: \label{tab:conclusion}
3499: }{
3500: \begin{tabular}{|c|c|c|c|c|c|c|c|}
3501: \hline %__________________________________________________________________________________________________________________________________________________________
3502: 		&		&			&			&		&		&			&			\\
3503: Phase		& $2S$/cell	& Order			& Degeneracy 		& Broken sym.	& Excitations 	& Thermo. 		& Examples		\\
3504: 		&		&			&			&		&		&			&			\\
3505: \hline %__________________________________________________________________________________________________________________________________________________________
3506: N\'eel AF	&		&			&			& $SU(2)$	&Gapless magnons& $C_v\sim T^2$	& Spin-$\frac{1}{2}$ triangular	\\
3507: p-sublattice	& any		& spin-spin LRO		&$\mathcal{O}(N^p)$	& Translations	&(spin waves)	& $\chi\sim {\rm cst}$	& Heisenberg AF 	\\
3508: 		&		&			&			& Point group	&		&			& 			\\
3509: \hline %__________________________________________________________________________________________________________________________________________________________
3510: VBC
3511: (\S\ref{sec:VBC})& odd 		& singlet-singlet  LRO	& $>1$    		& Translations  &Gapped magnons	& $C_v$ and $\chi$ 	& Honeycomb $J_1$--$J_2$\\
3512: (spontaneous) 	&     		& {\tiny colli.
3513: 					spin-spin SRO}	&			&  Point group  &           	& activated		& Checkerboard		\\
3514: 		&		&			&			&		&		&			& Square $J_1$--$J_2$ ?	\\
3515: \hline %__________________________________________________________________________________________________________________________________________________________
3516: VBC (\S\ref{sec:VBC})
3517: 		& even		&   None		& 1			& None 		&Gapped magnons	& $C_v$ and $\chi$ 	& SrCu$_2$(BO$_3$)$_2$	\\
3518: (explicit)	&		&			&        		&      		&		& activated  		& CaV$_4$O$_9$        	\\ 
3519: \hline
3520: VBS (\S
3521: \ref{ssec:VBS})	& even		& ``String'' LRO   	& 1      		& None 		&Gapped magnons	& $C_v$ and $\chi$ 	& AKLT	Hamiltonians	\\
3522: 		&       	&			&        		&      		&Edge excitations& activated   		& $S=1$ kagome AF ?	\\ 
3523: \hline %__________________________________________________________________________________________________________________________________________________________
3524: 	  	& 	    	& Topological         	& 4       		&		&Gapped spinons	& $C_v$ and $\chi$ 	& MSE (\S\ref{sec:MSE})\\
3525: RVB SL		& odd    	& {\tiny non-colli. SRO}& (torus) 		& None		&Gapped visons  & activated        	& QDM on triangular	\\
3526: (\S\ref{ssec:QDMTri}
3527: \S\ref{ssec:QDMKag},
3528: \S\ref{ssec:RVBMSE})
3529: 		&		&			&			&		&		&			& and kagome lattices	\\
3530: %\hline
3531: %RVB SL		&  even		& 			& 1			& None          &Gapped spinons	& $C_v$ and $\chi$ 	&          \\
3532: %		&		& {\tiny non-colli. SRO}&			&               &Gapped visons	& activated        	&          \\
3533: \hline %__________________________________________________________________________________________________________________________________________________________
3534: Kagome (\S\ref{sec:kagome})
3535: 		& 3		& None ?		& $\sim 1.15^N$ ?	& None ?	&Gapped
3536: 												  triplets ?	& $C_v\sim T^\alpha$ ? &			\\
3537: Heisenberg AF	&		&			&			&		&Gapless
3538: 												  singlets	& $\chi$ activated	&			\\
3539: \hline %__________________________________________________________________________________________________________________________________________________________
3540: \end{tabular}}
3541: \begin{tabnote}
3542: This  table   summarizes  the  properties  of   some  important phases
3543: encountered in 2D frustrated magnets.  $S$ is the value of the spin on
3544: each site.  ``Order''    refers  to the   nature of  the   long-ranged
3545: correlations (if any).  The ground-state degeneracy in the limit of an
3546: infinite system  (with periodic boundary   conditions) is indicated in
3547: the  fourth column,  except  for   RVB  SL   it  is  related  to   the
3548: spontaneously  broken  symmetries  mentioned    in the  next   column.
3549: Elementary   excitations  and  the  low-temperature   behavior of  the
3550: specific heat ($C_v$) and uniform susceptibility ($\chi$) are given in
3551: column six.  The  last column gives  some  examples of theoretical  or
3552: experimental realizations of these phases. The six families of systems
3553: presented here of course do not exhaust all possibilities. The results
3554: which   are    plausible     but  still   debated    (concerning   the
3555: spin-$\frac{1}{2}$ Heisenberg antiferromagnet on the kagome lattice in
3556: particular) are  indicated by question  marks.  Some  authors classify
3557: all the systems  with gapped    excitations  in a loose category    of
3558: ``quantum  disordered systems'',  alluding to  the  absence of  N\'eel
3559: long-ranged order.  It is a rather  unhappy appellation for VBC (which
3560: obviously have some order) and in fact for most of the quantum systems
3561: with   a   gap.  In   classical   statistical  physics ``disorder'' is
3562: associated to  entropy, which is not  the case in these gapped systems
3563: at $T=0$.
3564: \end{tabnote}
3565: \end{sidewaystable}
3566: %_____________________________________________________________________
3567: %_____________________________________________________________________
3568: \section{Acknowledgments}
3569: 
3570: It  would have been  difficult   to thank  individually  many of   our
3571: colleagues cited in references with whom we  had numerous and fruitful
3572: discussions -- many thanks to all of  them.  It is  also a pleasure to
3573: thank our close collaborators   B.~Bernu, V.~Pasquier, D.~Serban   and
3574: P.~Sindzingre, this review owes much to them.  We are also grateful to
3575: C.~Henley, A.~Honecker,   R.~Moessner  and  O.~Tchernyshyov for  their
3576: insightful comments on the manuscript.
3577: 
3578: %_____________________________________________________________________
3579: %_____________________________________________________________________
3580: 
3581: \begin{thebibliography}{100}
3582: 
3583: \bibitem{manousakis91}
3584: E. Manousakis,
3585: \href{http://link.aps.org/abstract/RMP/v63/p1}{Rev. Mod. Phys. {\bf 63}, 1 (1991)}.
3586: %The  spin-1/2 Heisenberg antiferromagnet on a  square  lattice and its
3587: %application to the cuprous oxides
3588: 
3589: \bibitem{bllp94}
3590: B. Bernu, P. Lecheminant, C. Lhuillier and L. Pierre,
3591: \href{http://link.aps.org/abstract/PRB/v50/e10048}{Phys. Rev. B {\bf 50}, 10048 (1994)}.
3592: 
3593: \bibitem{HFM2000}
3594: Proceedings  of  the {\em Highly Frustrated Magnetism  2000} conference,
3595: published in J. Can. Phys {\bf 79}, (2001).
3596: 
3597: 
3598: \bibitem{fradkinbook}
3599: E.   Fradkin, {\em   Field  Theories  of  Condensed Matter   Systems},
3600: Addison-Wesley (1998).
3601: 
3602: \bibitem{auerbachbook}
3603: A. Auerbach, {\em Interacting electrons and Quantum Magnetism},
3604: Springer-Verlag, Berlin Heidelberg New York, 1994.
3605: 
3606: \bibitem{tsvelikbook}
3607: A.~M.~Tsvelik, {\em Quantum Field Theory in Condensed Matter Physics}
3608: Cambridge University Press  (1996).
3609: 
3610: \bibitem{sachdevbook}
3611: S.~Sachdev, {\em Quantum Phase  Transitions},    Cambridge  U.  Press, New
3612: York (1999).
3613: 
3614: \bibitem{lt47}
3615: %Errata: Theory of Dipole Interaction in Crystals
3616: J. M. Luttinger and L.~Tisza, \href{http://link.aps.org/abstract/PR/v70/e954}{Phys. Rev. {\bf 70}, 954 (1946)}.
3617: 
3618: \bibitem{cd88}
3619: P. Chandra and B. Dou{\c c}ot, \href{http://link.aps.org/abstract/PRB/v38/e9335}{Phys. Rev. B {\bf 38}, 9335 (1988)}.
3620: 
3621: 
3622: \bibitem{vbcc80}
3623: J. Villain, R. Bidaux, J. P. Carton and R. Conte,
3624: J. Phys. (Paris) {\bf 41}, 1263 (1980).
3625: 
3626: 
3627: \bibitem{shender82}
3628: E.~Shender, Sov. Phys. JETP {\bf 56}, 178 (1982).
3629: %Antiferromagnetic garnets with fluctuationally interacting sublattices.
3630: 
3631: \bibitem{h89}
3632: C.~L. Henley, \href{http://link.aps.org/abstract/PRL/v62/e2056}{Phys. Rev. Lett. {\bf 62}, 2056 (1989)}.
3633: %Ordering due to disorder in a frustrated vector antiferromagnet
3634: 
3635: \bibitem{mdjr90}
3636: A. Moreo, E.  Dagotto, T. Jolic{\oe}ur and  J. Riera,
3637: \href{http://link.aps.org/abstract/PRB/v42/e6283}{Phys. Rev. B {\bf 42}, 6283 (1990)}.
3638: 
3639: \bibitem{ccl90a}
3640: P. Chandra, P. Coleman, and  A. Larkin,
3641: \href{http://link.aps.org/abstract/PRL/v64/e88}{Phys. Rev. Lett. {\bf 64}, 88 (1990)}.
3642: 
3643: \bibitem{szosh03}
3644: R.~R.~P.~Singh, W.~Zheng, J.~Oitmaa, O.~P.~Sushkov, C.~J.~Hamer,
3645: \href{http://arxiv.org/abs/cond-mat/0303075}{cond-mat/0303075}.
3646: 
3647: %A closer  look at  symmetry breaking  in  the  collinear phase  of the
3648: %$J_1-J_2$ Heisenberg Model
3649: 
3650: \bibitem{mbp03}
3651: G.~Misguich,  B.~Bernu   and  L.~Pierre,
3652: \href{http://link.aps.org/abstract/PRB/v68/e113409}{Phys. Rev. B {\bf 68}, 113409 (2003)}.
3653: %Determination     of the exchange    energies  in   Li2VOSiO4  from  a
3654: %high-temperature  series   analysis  of  the  square   lattice   J1-J2
3655: %Heisenberg model
3656: 
3657: \bibitem{wm03}
3658: C. Weber and F. Mila, private communication.
3659: 
3660: \bibitem{jdgb90}
3661: T. Jolic{\oe}ur,  E. Dagotto, E.   Gagliano and S.   Bacci,
3662: \href{http://link.aps.org/abstract/PRB/v42/e4800}{Phys. Rev. B {\bf 42}, 4800 (1990)}.
3663: 
3664: \bibitem{cj92}
3665: A. Chubukov and T. Jolic{\oe}ur, \href{http://link.aps.org/abstract/PRB/v46/e11137}{Phys. Rev. B {\bf 46}, 11137 (1992)}.
3666: 
3667: \bibitem{k93}
3668: S.~E.~Korshunov, \href{http://link.aps.org/abstract/PRB/v47/e6165}{Phys. Rev. B {\bf 47}, 6165 (1993)}. 
3669: 
3670: \bibitem{lblp95}
3671: P.~Lecheminant, B.~Bernu, C.~Lhuillier  and L.~Pierre,
3672: \href{http://link.aps.org/abstract/PRB/v52/e6647}{Phys. Rev. B {\bf 52}, 6647 (1995)}.
3673: 
3674: %$J_1-J_2$  quantum    Heisenberg antiferromagnet   on   the triangular
3675: %lattice:  A    group-symmetry   analysis   of   order    by   disorder.
3676: 
3677: 
3678: \bibitem{melzi00}
3679: %R.~Melzi,  P.~Carretta,    A.~Lascialfari,     M.~Mambrini, M.~Troyer,
3680: %P.~Millet  and F.~Mila, \href{http://link.aps.org/abstract/PRL/v85/e1318}{Phys. Rev. Lett. {\bf 85}, 1318 (2000)}.
3681: R.~Melzi {\it et   al.}, \href{http://link.aps.org/abstract/PRL/v85/e1318}{Phys. Rev. Lett. {\bf 85}, 1318 (2000)}.
3682: 
3683: %Li$_2$VO(Si,  Ge)O$_4$,   a   Prototype   of a  Two-Dimensional
3684: %Frustrated Quantum Heisenberg Antiferromagnet
3685: 
3686: \bibitem{melzi01}
3687: R. Melzi  {\it et   al.}, \href{http://link.aps.org/abstract/PRB/v64/e024409}{Phys. Rev. B {\bf 64}, 024409 (2001)}.
3688: 
3689: %Magnetic and
3690: %thermodynamic  properties of Li2VOSiO$_4$:  A  two-dimensional S = 1/2
3691: %frustrated antiferromagnet on a square lattice
3692: 
3693: \bibitem{rosner02}
3694: H.   Rosner    {\it et   al.},
3695: \href{http://link.aps.org/abstract/PRL/v88/e186405}{Phys. Rev. Lett. {\bf 88}, 186405 (2002)}.
3696: 
3697: %Realization  of a  Large $J_2$  Quasi-2D  Spin-Half Heisenberg System:
3698: %Li2VOSiO$_4$.
3699: 
3700: \bibitem{rosner03}
3701: H. Rosner   {\it et   al.},   \href{http://link.aps.org/abstract/PRB/v67/e014416}{Phys. Rev. B {\bf 67}, 014416 (2003)}.
3702: 
3703: %High-temperature   expansions  for   the  $J_1$-$J_2$  Heisenberg    models:
3704: %Applications to ab  initio  calculated models for  Li$_2$VOSiO$_4$ and
3705: %Li$_2$VOGeO$_4$.
3706: 
3707: \bibitem{klein82}
3708: D. J. Klein,
3709: \href{http://stacks.iop.org/0305-4470/15/661}
3710: {J. Phys. A: Math. Gen. {\bf 15}, 661 (1982)}.
3711: 
3712: %Exact ground states for a class of antiferromagnetic Heisenberg models
3713: %with short-range interactions.
3714: 
3715: \bibitem{mg69}
3716: C.~K.~Majumdar and D.~K.~Ghosh, J. Math. Phys. {\bf 10}, 1399 (1969).
3717: %On next-nearest-neighbor interaction in linear chain. II.
3718: 
3719: 
3720: \bibitem{sgh88}
3721: R.~R.~P.~Singh, M.~P.~Gelfand and D.~A.~Huse,
3722: \href{http://link.aps.org/abstract/PRL/v61/e2484}{Phys. Rev. Lett. {\bf 61}, 2484 (1988)}.
3723: 
3724: %Ground States of Low-Dimensional Quantum Antiferromagnets.
3725: 
3726: \bibitem{gsh90}
3727: M.~P.~Gelfand, R.~R.~P.~Singh and D.~A.~Huse, J. Stat. Phys. {\bf 59},
3728: 1093 (1990).
3729: 
3730: %Perturbation expansions for quantum many-body systems. 
3731: 
3732: \bibitem{woh91}
3733: Zheng Weihong, J. Oitmaa, and C. J. Hamer,
3734: \href{http://link.aps.org/abstract/PRB/v43/e8321}{Phys. Rev. B {\bf 43}, 8321 (1991)}.
3735: 
3736: %Square-lattice Heisenberg antiferromagnet at T=0.
3737: 
3738: \bibitem{ow96}
3739: J. Oitmaa and Zheng  Weihong, \href{http://link.aps.org/abstract/PRB/v54/e3022}{Phys. Rev. B {\bf 54}, 3022 (1996)}.
3740: 
3741: %Series expansion for   the $J_1-J_2$ Heisenberg antiferromagnet on   a
3742: %square lattice.
3743:     
3744: \bibitem{gsh89}
3745: M. P. Gelfand, R. R. P.  Singh and D. A.  Huse,
3746: \href{http://link.aps.org/abstract/PRB/v40/e10801}{Phys. Rev. B {\bf 40}, 10801 (1989)}.
3747: 
3748: %Zero-temperature   ordering   in two-dimensional  frustrated   quantum
3749: %Heisenberg antiferromagnets.
3750: 
3751: \bibitem{gelfand90}
3752: M. P. Gelfand.  \href{http://link.aps.org/abstract/PRB/v42/e8206}{Phys. Rev. B {\bf 42}, 8206 (1990)}.
3753: 
3754: %Series investigations  of  magnetically  disordered ground   states in
3755: %two-dimensional frustrated quantum antiferromagnets.
3756: 
3757: \bibitem{swho99}
3758: R.~R.~P.~Singh, Zheng Weihong, C.~J.~Hamer, and  J.~Oitmaa,
3759: \href{http://link.aps.org/abstract/PRB/v60/e7278}{Phys. Rev. B {\bf 60}, 7278 (1999)}.
3760: 
3761: %Dimer  order with  striped   correlations in the  $J_1-J_2$ Heisenberg
3762: %model.
3763: 
3764: \bibitem{kosw00}
3765: V.~N.~Kotov, J.~Oitmaa, O.~Sushkov  and  Zheng Weihong,
3766: Phil. Mag. B  {\bf 80}, 1483 (2000).
3767: 
3768: %\href{http://arxiv.org/abs/cond-mat/9912228}{cond-mat/9912228}
3769: %Spontaneous  dimer   order, excitation  spectrum,   and  quantum phase
3770: %transitions in the $J_1$--$J_2$ Heisenberg model.
3771: 
3772: \bibitem{zu96}
3773: M. E.   Zhitomirsky  and  K.  Ueda,
3774: \href{http://link.aps.org/abstract/PRB/v54/e9007}{Phys. Rev. B {\bf 54}, 9007 (1996)}.
3775: 
3776: %Valence-bond crystal  phase  of a  frustrated  spin-1/2 square-lattice
3777: %antiferromagnet.
3778: 
3779: \bibitem{cbps01}
3780: L.  Capriotti,      F.     Becca,  A. Parola,    and      S.  Sorella,
3781: \href{http://link.aps.org/abstract/PRL/v87/e097201}{Phys. Rev. Lett. {\bf 87}, 097201 (2001)}.
3782: 
3783: %Resonating  Valence Bond Wave  Functions for  Strongly Frustrated Spin
3784: %Systems.
3785: 
3786: \bibitem{sow02}
3787: O.~P.~Sushkov, J.~Oitmaa, and   Zheng Weihong,
3788: \href{http://link.aps.org/abstract/PRB/v66/e054401}{Phys. Rev. B {\bf 66}, 054401 (2002)}.
3789: 
3790: % Critical  dynamics of singlet and  triplet excitations
3791: %in strongly frustrated spin systems
3792: 
3793: \bibitem{sow01}
3794: O. P. Sushkov, J. Oitmaa, and Zheng Weihong,
3795: \href{http://link.aps.org/abstract/PRB/v63/e104420}{Phys. Rev. B {\bf 63}, 104420 (2001)}.
3796: 
3797: %Quantum phase transitions in the two-dimensional $J_1-J_2$ model.
3798: 
3799: \bibitem{cls00}
3800: M.~S.~L. du~Croo~de  Jongh,  J.~M.~J.~Van   Leeuwen,  and  W.~Van  Saarloos,
3801: \href{http://link.aps.org/abstract/PRB/v62/e14844}{Phys. Rev. B {\bf 62}, 14844 (2000)}.
3802: 
3803: \bibitem{dm89}
3804: E.   Dagotto and  A.  Moreo,  \href{http://link.aps.org/abstract/PRB/v39/e4744}{Phys. Rev. B {\bf 39}, 4744 (1989)},
3805: \href{http://link.aps.org/abstract/PRL/v63/e2148}{Phys. Rev. Lett. {\bf 63}, 2148 (1989)}.
3806: 
3807: \bibitem{fkksrr90}
3808: %F.~Figueirido,  A.~Karlhede, S.~Kivelson,  S.~Sondhi, M.~Rocek  and D.~S.~Rokhsar
3809: F.~Figueirido {\it et al.}, \href{http://link.aps.org/abstract/PRB/v41/e4619}{Phys. Rev. B {\bf 41}, 4619 (1990)}.
3810: 
3811: 
3812: \bibitem{pgbd91}
3813: %Static    and   dynamical  correlations   in   a  spin-1/2  frustrated
3814: %antiferromagnet.
3815: D.~Poilblanc, E.~Gagliano, S.~Bacci and E.~Dagotto,
3816: Phys. Rev. B, {\bf 43} 10970 (1991).
3817: 
3818: \bibitem{schulz}
3819: %\bibitem{sz92}
3820: H.~J.~Schulz, T.~A.~L.~Ziman, \href{http://publish.edpsciences.com/abstract/EPL/v18/p355}{Europhys. Lett. {\bf 18}, 355 (1992)}.
3821: %\bibitem{es95}
3822: T.~Einarsson and H.~J.~Schulz, \href{http://link.aps.org/abstract/PRB/v51/e6151}{Phys. Rev. B {\bf 51}, 6151 (1995)}.
3823: 
3824: %Direct  Calculation  of    the  Spin  Stiffness in   the  $J_1$--$J_2$
3825: %Heisenberg Antiferromagnet.
3826: 
3827: %\bibitem{szp96}
3828: H.~J.~Schulz, T.~A.~L.~Ziman, D.~Poilblanc, J. Physique I {\bf 6}, 675
3829: (1996).
3830: 
3831: %Magnetic  order and disorder   in  the frustrated quantum   Heisenberg
3832: %antiferromagnet in two dimensions.
3833: 
3834: 
3835: \bibitem{nlsm}
3836: %\bibitem{chn89}
3837: S. Chakravarty, B.~I.~Halperin and   D.~R.~Nelson,
3838: \href{http://link.aps.org/abstract/PRB/v39/e2344}{Phys. Rev. B {\bf 39}, 2344 (1989)}.
3839: %Two-dimensional quantum Heisenberg antiferromagnet at low temperatures.
3840: %\bibitem{nz89}
3841: H.~Neuberger and T.~Ziman, \href{http://link.aps.org/abstract/PRB/v39/e2608}{Phys. Rev. B {\bf 39}, 2608 (1989)}.
3842: %\bibitem{f89}
3843: D.~Fisher, \href{http://link.aps.org/abstract/PRB/v39/e11783}{Phys. Rev. B {\bf 39}, 11783 (1989)}.
3844: %\bibitem{ej91}
3845: T.~Einarsson and H.~Johannesson, \href{http://link.aps.org/abstract/PRB/v43/e5867}{Phys. Rev. B {\bf 43}, 5867 (1991)}.
3846: %Effective-action approach to the frustrated Heisenberg antiferromagnet
3847: %in two dimensions.
3848: %\bibitem{hn93a}
3849: P.  Hasenfratz and F.~Niedermayer,
3850: Z.  Phys. B.  Condens. Matter {\bf 92}, 91 (1993).
3851: 
3852: \bibitem{s97}
3853: A.~W. Sandvik, \href{http://link.aps.org/abstract/PRB/v56/e11678}{Phys. Rev. B {\bf 56}, 11678 (1997)}.
3854: 
3855: \bibitem{s98}
3856: S. Sorella, \href{http://link.aps.org/abstract/PRL/v80/e4558}{Phys. Rev. Lett. {\bf 80}, 4558 (1998)}.
3857: 
3858: \bibitem{s01}
3859: S. Sorella, \href{http://link.aps.org/abstract/PRB/v64/e024512}{Phys. Rev. B {\bf 64}, 024512 (2001)}.
3860: 
3861: \bibitem{w92}
3862: S. White, \href{http://link.aps.org/abstract/PRL/v69/e2863}{Phys. Rev. Lett. {\bf 69}, 2863 (1992)}.
3863: 
3864: \bibitem{w93}
3865: S. White, \href{http://link.aps.org/abstract/PRB/v48/e10345}{Phys. Rev. B {\bf 48}, 10345 (1993)}.
3866: 
3867: \bibitem{gl91}
3868: T. Giamarchi and C. Lhuillier, \href{http://link.aps.org/abstract/PRB/v43/e12943}{Phys. Rev. B {\bf 43}, 12943 (1991)}.
3869: 
3870: %Phase diagrams of the   two-dimensional Hubbard and  t-J models  by  a
3871: %variational Monte Carlo method
3872: 
3873: \bibitem{ss81}
3874: B. Shastry and B. Sutherland, \href{http://link.aps.org/abstract/PRL/v47/e964}{Phys. Rev. Lett. {\bf 47}, 964 (1981)}.
3875: 
3876: \bibitem{h82}
3877: F. Haldane, \href{http://link.aps.org/abstract/PRB/v25/e4925}{Phys. Rev. B {\bf 25}, 4925 (1982)}.
3878: 
3879: \bibitem{a89}
3880: I. Affleck, J. Phys. Cond. Matt. {\bf 1},  3047  (1989).
3881: 
3882: \bibitem{ys97}
3883: H. Yokoyama and Y. Saiga, \href{http://jpsj.ipap.jp/link?JPSJ/66/3617}{J. Phys. Soc. Jpn. {\bf 66}, 3617 (1997)}.
3884: 
3885: \bibitem{nt97a}
3886: T. Nakamura and S. Takada, \href{http://link.aps.org/abstract/PRB/v55/e14413}{Phys. Rev. B {\bf 55}, 14413 (1997)}.
3887: 
3888: \bibitem{asrp99}
3889: D. Augier, E. Sorensen, J. Riera, and D. Poilblanc,
3890: \href{http://link.aps.org/abstract/PRB/v60/e1075}{Phys. Rev. B {\bf 60}, 1075 (1999)}.
3891: 
3892: \bibitem{dr96}
3893: E. Dagotto and T.~M. Rice, Science {\bf 271},  618  (1996).
3894: 
3895: %\bibitem{nt97}
3896: %A.~A. Nersesyan and A.~M. Tsvelick, \href{http://link.aps.org/abstract/PRL/v78/e3939}{Phys. Rev. Lett. {\bf 78}, 3939 (1997)}.
3897: 
3898: %\bibitem{km98}
3899: %A.~K. Kolezhuk and H.-J. Mikeska, \href{http://link.aps.org/abstract/PRL/v80/e2709}{Phys. Rev. Lett. {\bf 80}, 2709 (1998)}.
3900: 
3901: \bibitem{nt97}
3902: A.~A.~Nersesyan and A.~M.~Tsvelik,
3903: \href{http://link.aps.org/abstract/PRL/v78/e3939}{Phys. Rev. Lett. {\bf 78}, 3939 (1997)}.
3904: %One-Dimensional Spin-Liquid without Magnon Excitations.
3905: 
3906: \bibitem{km98} A.~K.~Kolezhuk and H.-J.  Mikeska,
3907: Int. J. Mod. Phys. B {\bf 12}, 2325 (1998).
3908: \href{http://link.aps.org/abstract/PRL/v80/e2709}{Phys. Rev. Lett. {\bf 80}, 2709 (1998)}.
3909: %Non-Haldane Spin-Liquid Models with Exact Ground States
3910: \href{http://link.aps.org/abstract/PRB/v56/e11380}{Phys. Rev. B {\bf 56}, 11380 (1997)}.
3911: %Models with  exact ground states  connecting smoothly the  S=1/2 dimer
3912: %and S=1 Haldane phases of one-dimensional spin chains
3913: 
3914: 
3915: \bibitem{aklt87}
3916: I.~Affleck, T.~Kennedy, E.~Lieb, and H.~Tasaki,
3917: \href{http://link.aps.org/abstract/PRL/v59/e799}{Phys. Rev. Lett. {\bf 59}, 799 (1987)}.
3918: 
3919: \bibitem{aklt88}
3920: I. Affleck, T. Kennedy, E.~H. Lieb, and H. Tasaki,
3921: Commun. Math. Phys. {\bf  115},  477  (1988).
3922: 
3923: \bibitem{nr89}
3924: M. den Nijs and K. Rommelse, \href{http://link.aps.org/abstract/PRB/v40/e4709}{Phys. Rev. B {\bf 40}, 4709 (1989)}.
3925: 
3926: \bibitem{kt92}
3927: T. Kennedy and H. Tasaki, \href{http://link.aps.org/abstract/PRB/v45/e304}{Phys. Rev. B {\bf 45}, 304 (1992)}.
3928: 
3929: \bibitem{j02}
3930: T. Jolic{\oe}ur, private communication.
3931: 
3932: \bibitem{h00}
3933: K. Hida, \href{http://jpsj.ipap.jp/link?JPSJ/69/4003}{J. Phys. Soc. Jpn. {\bf 69}, 4003 (2000)}.
3934: 
3935: \bibitem{yk00}
3936: %Spin-Driven Jahn-Teller Distortion in a Pyrochlore System
3937: Y.~Yamashita and K.~Ueda, 
3938: \href{http://link.aps.org/abstract/PRL/v85/e4960}{Phys. Rev. Lett. {\bf 85}, 4960 (2000)}.
3939: 
3940: \bibitem{CaV4O9}
3941: %\bibitem{ki94}
3942: N.~Katoh and M.~Imada, \href{http://jpsj.ipap.jp/link?JPSJ/63/4529}{J. Phys. Soc. Jpn. {\bf 63}, 4529 (1994)}.
3943: %\bibitem{ttw94}
3944: M.~Troyer, H.~Tsunetsugu, and D.~Wuertz, \href{http://link.aps.org/abstract/PRB/v50/e13515}{Phys. Rev. B {\bf 50}, 13515 (1994)}.
3945: %\bibitem{tnyk95}
3946: S.~Taniguchi {\it et~al.}, \href{http://jpsj.ipap.jp/link?JPSJ/64/2758}{J. Phys. Soc. Jpn. {\bf 64}, 2758 (1995)}.
3947: %\bibitem{khst96}
3948: K.~Kodama {\it et~al.}, \href{http://jpsj.ipap.jp/link?JPSJ/65/1941}{J. Phys. Soc. Jpn. {\bf 65}, 1941 (1996)}.
3949: %\bibitem{fo96}
3950: Y.~Fukumoto and A.~Oguchi, \href{http://jpsj.ipap.jp/link?JPSJ/65/1440}{J. Phys. Soc. Jpn. {\bf 65}, 1440 (1996)}.
3951: %\bibitem{khsk97}
3952: K.~Kodama {\it et~al.}, \href{http://jpsj.ipap.jp/link?JPSJ/66/793}{J. Phys. Soc. Jpn. {\bf 66}, 793 (1997)}.
3953: %\bibitem{my96}
3954: T.~Miyasaki and D.~Yoshioka, \href{http://jpsj.ipap.jp/link?JPSJ/65/2370}{J. Phys. Soc. Jpn. {\bf 65}, 2370 (1996)}.
3955: %\bibitem{oyiu97}
3956: T.~Ohama, H.~Yasuoka, M.~Isobe, and Y.~Ueda,
3957: \href{http://jpsj.ipap.jp/link?JPSJ/66/23}{J. Phys. Soc. Jpn. {\bf 66}, 23 (1997)}.
3958: 
3959: \bibitem{uksl96}
3960: K. Ueda, H. Kontani, M. Sigrist, and P.~A. Lee, \href{http://link.aps.org/abstract/PRL/v76/e1932}{Phys. Rev. Lett. {\bf 76}, 1932 (1996)}.
3961: 
3962: \bibitem{am96b}
3963: %Spin gap in CaV4O9: A large-S approach
3964: M.~Albrecht and F. Mila, \href{http://link.aps.org/abstract/PRB/v53/e2945}{Phys. Rev. B {\bf 53}, 2945 (1996)}.
3965: 
3966: \bibitem{tku96}
3967: M. Troyer, H. Kontani, and K. Ueda, \href{http://link.aps.org/abstract/PRL/v76/e3822}{Phys. Rev. Lett. {\bf 76}, 3822 (1996)}.
3968: 
3969: 
3970: 
3971: \bibitem{sr96}
3972: S. Sachdev and N. Read, \href{http://link.aps.org/abstract/PRL/v77/e4800}{Phys. Rev. Lett. {\bf 77}, 4800 (1996)}.
3973: 
3974: \bibitem{szksu96}
3975: O.~A. Starykh {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v77/e2558}{Phys. Rev. Lett. {\bf 77}, 2558 (1996)}.
3976: 
3977: 
3978: 
3979: \bibitem{wgsoh97}
3980: Zheng Weihong {\it et~al.}, \href{http://link.aps.org/abstract/PRB/v55/e11377}{Phys. Rev. B {\bf 55}, 11377 (1997)}.
3981: 
3982: \bibitem{k99}
3983: H. Kageyama {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v82/e3168}{Phys. Rev. Lett. {\bf 82}, 3168 (1999)}.
3984: 
3985: \bibitem{nkoum99}
3986: H. Nojiri {\it et~al.}, \href{http://jpsj.ipap.jp/link?JPSJ/68/2906}{J. Phys. Soc. Jpn. {\bf 68}, 2906 (1999)}.
3987: 
3988: \bibitem{mu99}
3989: S. Miyahara and K. Ueda, \href{http://link.aps.org/abstract/PRL/v82/e3701}{Phys. Rev. Lett. {\bf 82}, 3701 (1999)}.
3990: 
3991: \bibitem{k00}
3992: H. Kageyama {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v84/e5876}{Phys. Rev. Lett. {\bf 84}, 5876 (2000)}.
3993: 
3994: \bibitem{kk00}
3995: A. Koga and N. Kawakami, \href{http://link.aps.org/abstract/PRL/v84/e4461}{Phys. Rev. Lett. {\bf 84}, 4461 (2000)}.
3996: 
3997: \bibitem{msku00}
3998: E. M\"uller-Hartmann, R.~R.~P. Singh, C. Knetter, and G.~S. Uhrig,
3999: \href{http://link.aps.org/abstract/PRL/v84/e1808}{Phys. Rev. Lett. {\bf 84}, 1808 (2000)}.
4000: 
4001: \bibitem{tmu01}
4002: K. Totsuka, S. Miyahara, and K. Ueda, \href{http://link.aps.org/abstract/PRL/v86/e520}{Phys. Rev. Lett. {\bf 86}, 520 (2001)}.
4003: 
4004: \bibitem{lws02}
4005: A. L\"auchli, S. Wessel, and M. Sigrist, \href{http://link.aps.org/abstract/PRB/v66/e014401}{Phys. Rev. B {\bf 66}, 014401 (2002)}.
4006: 
4007: \bibitem{kthb02}
4008: %K. Kodama, M. Takigawa , M. Horvatic, C.
4009: %Berthier,  H. Kageyama, Y. Ueda, S. Miyahara, F. Becca
4010: %and F. Mila
4011: K. Kodama {\it et al.}, Science {\bf 298},395 (2002).
4012: 
4013: \bibitem{mu03}
4014: S.~Miyahara and K.~Ueda, J. Phys.: Condens. Matter {\bf 15}, R327-R366 (2003).
4015: 
4016: \bibitem{ss81a}
4017: B. Shastry and B. Sutherland, Physica B (Amsterdam) {\bf 108},  1069  (1981).
4018: 
4019: \bibitem{cz02} O.~Cepas, T.~Ziman, Proceedings of  the conference in
4020: Fukuoka,   Nov.  2001,   to   appear  in   Fukuoka  University   Press
4021: [\href{http://arxiv.org/abs/cond-mat/0207191}{cond-mat/0207191}].     O.~Cepas,    T.~Sakai,     T.~Ziman,
4022: Progr. Theor. Phys. Suppl. {\bf 145}, 43 (2002).
4023: 
4024: \bibitem{zoh02}
4025: W. Zheng, J. Oitmaa, and C. J. Hamer
4026: \href{http://link.aps.org/abstract/PRB/v65/e014408}{Phys. Rev. B {\bf 65}, 014408 (2002)}.
4027: 
4028: \bibitem{am96a}
4029: %First-order transition between  magnetic order and valence  bond order
4030: %in a 2D frustrated Heisenberg model
4031: M.~ Albrecht and F. Mila, \href{http://publish.edpsciences.com/abstract/EPL/v34/p145}{Europhys. Lett. {\bf 34}, 145 (1996)}.
4032: 
4033: \bibitem{cms01}
4034: C.~H.~Chung, J.~B.~Marston, and S.~Sachdev,
4035: \href{http://link.aps.org/abstract/PRB/v64/e134407}{Phys. Rev. B {\bf 64}, 134407 (2001)}.
4036: 
4037: %Quantum  phases of the Shastry-Sutherland antiferromagnet: Application
4038: %to SrCu2(BO3)2
4039: 
4040: 
4041: \bibitem{fo99}
4042: Y. Fukumoto and A. Oguchi, \href{http://jpsj.ipap.jp/link?JPSJ/68/3655}{J. Phys. Soc. Jpn. {\bf 68}, 3655 (1999)}.
4043: 
4044: \bibitem{mt00}
4045: T. Momoi and K. Totsuka, \href{http://link.aps.org/abstract/PRB/v61/e3231}{Phys. Rev. B {\bf 61}, 3231 (2000)}.
4046: 
4047: \bibitem{mt00b}
4048: T. Momoi and K. Totsuka, \href{http://link.aps.org/abstract/PRB/v62/e15067}{Phys. Rev. B {\bf 62}, 15067 (2000)}.
4049: 
4050: \bibitem{mu00}
4051: S. Miyahara and K. Ueda, \href{http://link.aps.org/abstract/PRB/v61/e3417}{Phys. Rev. B {\bf 61}, 3417 (2000)}.
4052: 
4053: \bibitem{mjg01}
4054: G. Misguich, T. Jolic{\oe}ur, and S.~M.~Girvin,
4055: \href{http://link.aps.org/abstract/PRL/v87/e097203}{Phys. Rev. Lett. {\bf 87}, 097203 (2001)}.
4056: 
4057: \bibitem{lm02}
4058: C. Lhuillier and G. Misguich, in {\em High Magnetic Fields}, edited by
4059: C.   Berthier,  L. Levy, and  G.  Martinez  (Springer,  Berlin, 2002),
4060: pp.~161--190, \href{http://arxiv.org/abs/cond-mat/0109146}{cond-mat/0109146}.
4061: 
4062: \bibitem{ll96}
4063: P.~W. Leung and N. Lam, \href{http://link.aps.org/abstract/PRB/v53/e2213}{Phys. Rev. B {\bf 53}, 2213 (1996)}.
4064: 
4065: \bibitem{fsl01}  J.-B.~Fouet,  P.~Sindzingre,   and  C.~Lhuillier,
4066: \href{http://publish.edpsciences.com/abstract/EPJB/v20/p241}{Eur. Phys. J. B {\bf 20}, 241 (2001)}.
4067: 
4068: %An investigation  of the quantum $J_1-J_2-J_3$ model  on the honeycomb
4069: %lattice
4070: 
4071: \bibitem{canals02}
4072: B.~Canals, \href{http://link.aps.org/abstract/PRB/v65/e184408}{Phys. Rev. B {\bf 65}, 184408 (2002)}.
4073: 
4074: \bibitem{fsl03}
4075: J.-B.~Fouet,     M.~Mambrini,    P.~Sindzingre,  and     C.~Lhuillier,
4076: \href{http://link.aps.org/abstract/PRB/v67/e054411}{Phys. Rev. B {\bf 67}, 054411 (2003)}.
4077: 
4078: \bibitem{sfl02}
4079: P. Sindzingre, J.~B. Fouet, and C. Lhuillier,
4080: \href{http://link.aps.org/abstract/PRB/v66/e174424}{Phys. Rev. B {\bf 66}, 174424 (2002)}.
4081: 
4082: \bibitem{bh02}
4083: W. Brenig  and  A. Honecker,
4084: \href{http://link.aps.org/abstract/PRB/v65/e140407}{Phys. Rev. B {\bf 65}, 140407R (2002)}.
4085: 
4086: \bibitem{baa03}
4087: E.~Berg, E.~Altman, and A.~Auerbach,
4088: \href{http://link.aps.org/abstract/PRL/v90/e147204}{Phys. Rev. Lett. {\bf 90}, 147204 (2003)}.
4089: 
4090: \bibitem{tsma03}
4091: %Bond order from disorder in the planar pyrochlore magnet
4092: O.~Tchernyshyov, O.~A.~Starykh, R.~Moessner and A.~G.~Abanov,
4093: \href{http://arxiv.org/abs/cond-mat/0301303}{cond-mat/0301303}.
4094:  
4095: \bibitem{mts01}
4096: R. Moessner, Oleg Tchernyshyov and S.~L.~Sondhi,
4097: \href{http://arxiv.org/abs/cond-mat/0106286}{cond-mat/0106286}.
4098: %Planar pyrochlore, quantum ice and sliding ice
4099: 
4100: \bibitem{pc02}
4101: S. Palmer and J.~.T.~Chalker, \href{http://link.aps.org/abstract/PRB/v64/e094412}{Phys. Rev. B {\bf 64}, 094412 (2002)}.
4102: 
4103: \bibitem{mc98}
4104: R. Moessner and J.~T. Chalker, \href{http://link.aps.org/abstract/PRB/v58/e12049}{Phys. Rev. B {\bf 58}, 12049 (1998)}.
4105: 
4106: \bibitem{mc98a}
4107: R. Moessner and J.~T. Chalker, \href{http://link.aps.org/abstract/PRL/v80/e2929}{Phys. Rev. Lett. {\bf 80}, 2929 (1998)}.
4108: 
4109: 
4110: \bibitem{pp01}
4111: %Classical  Heisenberg  antiferromagnet  on a  garnet lattice:  A Monte
4112: %Carlo simulation
4113: O.~A.~Petrenko and D.~McK.~Paul,
4114: \href{http://link.aps.org/abstract/PRB/v63/e024409}{Phys. Rev. B {\bf 63}, 024409 (2001)}.
4115: 
4116: 
4117: \bibitem{ka02}
4118: %Chiral   Kosterlitz-Thouless transition in  the frustrated  Heisenberg
4119: %antiferromagnet on a pyrochlore slab
4120: Hikaru Kawamura and Takuya Arimori,
4121: \href{http://link.aps.org/abstract/PRL/v88/e077202}{Phys. Rev. Lett. {\bf 88}, 077202 (2002)}
4122: 
4123: \bibitem{henley2000}
4124: %Effective  Hamiltonians and  dilution effects  in  Kagome and  related
4125: %anti-ferromagnets
4126: C.~L.~Henley, Can. J. Phys. (Canada) {\bf 79}, 1307 (2001).
4127: [\href{http://arxiv.org/abs/cond-mat/0009130}{cond-mat/0009130}]
4128: 
4129: \bibitem{f03}
4130: J.-B.~Fouet, Ph.D. thesis, Universit\'e Cergy Pontoise, 2003.
4131: 
4132: \bibitem{gwsoh96}
4133: M.~P.~Gelfand {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v77/e2794}{Phys. Rev. Lett. {\bf 77}, 2794 (1996)}.
4134: 
4135: 
4136: \bibitem{bm02}
4137: %Peierls-Like  Transition Induced  by Frustration in  a Two-Dimensional
4138: %Antiferromagnet
4139: F. Becca and F. Mila, \href{http://link.aps.org/abstract/PRL/v89/e037204}{Phys. Rev. Lett. {\bf 89}, 037204 (2002)}.
4140: 
4141: \bibitem{mlbw99}
4142: %Spin-Liquid   phase of the  Multiple-Spin  Exchange Hamiltonian on the
4143: %Triangular Lattice.
4144: G. Misguich, C. Lhuillier, B. Bernu and C. Waldtmann,
4145: \href{http://link.aps.org/abstract/PRB/v60/e1064}{Phys. Rev. B {\bf 60}, 1064 (1999)}.
4146: 
4147: 
4148: \bibitem{a85}
4149: I. Affleck, \href{http://link.aps.org/abstract/PRL/v54/e966}{Phys. Rev. Lett. {\bf 54}, 966 (1985)}.
4150: %Large-n Limit of SU(n) Quantum "Spin" Chains
4151: 
4152: \bibitem{am88}
4153: I. Affleck and J. Marston, \href{http://link.aps.org/abstract/PRB/v37/e3774}{Phys. Rev. B {\bf 37}, 3774 (1988)}.
4154: 
4155: 
4156: \bibitem{aa88}
4157: D. Arovas and A.~Auerbach, \href{http://link.aps.org/abstract/PRB/v38/e316}{Phys. Rev. B {\bf 38}, 316 (1988)}.
4158: 
4159: \bibitem{rs89}
4160: N. Read and S. Sachdev, \href{http://link.aps.org/abstract/PRL/v62/e1694}{Phys. Rev. Lett. {\bf 62}, 1694 (1989)}.
4161: 
4162: \bibitem{rs90}
4163: N.~Read and S.~Sachdev,
4164: \href{http://link.aps.org/abstract/PRB/v42/e4568}{Phys. Rev. B {\bf 42}, 4568 (1990)}.
4165: 
4166: \bibitem{rs91}
4167: N. Read and S. Sachdev,
4168: \href{http://link.aps.org/abstract/PRL/v66/e1773}{Phys. Rev. Lett. {\bf 66}, 1773 (1991)}.
4169: %Large-$N$ expansion for frustrated quantum antiferromagnets.
4170: 
4171: \bibitem{sr91}
4172: S. Sachdev and N. Read, Int. J. Mod. Phys. {\bf 5}, 219 (1991).
4173: % Large-N Expansion for frustrated and doped quantum antiferromagnets
4174: 
4175: \bibitem{s93}  S.  Sachdev  in  {\it  Low  Dimensional  Quantum  Field
4176: Theories   for  Condensed   Matter  Physicists}   edited  by   Y.  Lu,
4177: S. Lundqvist, and G. Morandi, World Scientific, Singapore, 1995.
4178: \href{http://arxiv.org/abs/cond-mat/9303014}{cond-mat/9303014}.
4179: 
4180: 
4181: \bibitem{rs89b}
4182: N. Read and S. Sachdev, Nucl. Phys. B {\bf 316}, 609 (1989).
4183: 
4184: \bibitem{rokhsar90}
4185: D. Rokhsar, \href{http://link.aps.org/abstract/PRB/v42/e2526}{Phys. Rev. B {\bf 42}, 2526 (1991)}.
4186: %Quadratic quatum antiferromagnets in the fermionic large-N limit
4187: 
4188: \bibitem{wilson74}
4189: K. G. Wilson,
4190: \href{http://link.aps.org/abstract/PRD/v10/e2445}{Phys. Rev. D {\bf 10}, 2445 (1974)}.
4191: 
4192: %Confinement of quarks
4193: 
4194: \bibitem{bdi74}
4195: 
4196: R. Balian, J. M. Drouffe, and C. Itzykson,
4197: \href{http://link.aps.org/abstract/PRD/v10/e3376}{Phys. Rev. D {\bf 10}, 3376 (1974)},
4198: \href{http://link.aps.org/abstract/PRD/v11/e2098}{Phys. Rev. D {\bf 11}, 2098 (1975)},
4199: \href{http://link.aps.org/abstract/PRD/v11/e2098}{Phys. Rev. D {\bf 11}, 2098 (1975)}.
4200: 
4201: % Gauge fields on a lattice. I. General outlook
4202: 
4203: \bibitem{schwingerboson}
4204: D.~P.~Arovas and A.~Auerbach, \href{http://link.aps.org/abstract/PRB/v38/e316}{Phys. Rev. B {\bf 38}, 316 (1988)}.
4205: A.~Auerbach and D.~P.~Arovas, \href{http://link.aps.org/abstract/PRL/v61/e617}{Phys. Rev. Lett. {\bf 61}, 617 (1988)}.
4206: J.~E.~Hirsch and S.~Tang, \href{http://link.aps.org/abstract/PRB/v39/e2850}{Phys. Rev. B {\bf 39}, 2850 (1989)}.
4207: 
4208: \bibitem{ceccatto}
4209: H.~A.~Ceccatto, C.~J.~Gazza, and A.~E.~Trumper,
4210: \href{http://link.aps.org/abstract/PRB/v47/e12329}{Phys. Rev. B {\bf 47}, 12329 (1993)}.
4211: A.~E.~Trumper, L.~O.~Manuel, C.~J.~Gazza, and
4212: H.~A.~Ceccatto \href{http://link.aps.org/abstract/PRL/v78/e2216}{Phys. Rev. Lett. {\bf 78}, 2216 (1997)}.
4213: 
4214: \bibitem{ot03}
4215: We thank O.~Tchernyshyov  for pointing us this physical interpretation
4216: of the $U(1)$ flux.
4217: 
4218: 
4219: \bibitem{sachdev92}
4220: S. Sachdev, \href{http://link.aps.org/abstract/PRB/v45/e12377}{Phys. Rev. B {\bf 45}, 12377 (1992)}.
4221: 
4222: \bibitem{cmm01}
4223: C.~H.~Chung and J.~B.~Marston and Ross H.~McKenzie,
4224: J. Phys.: Condens. Matter {\bf 13}, 5159  (2001).
4225: [\href{http://arxiv.org/abs/cond-mat/0012216}{cond-mat/0012216}]
4226: 
4227: \bibitem{haldane83}
4228: F.~D.~M.~Haldane, Phys. Lett. {\bf 93A}, 464 (1983);
4229: %Continuum   dynamics    of the   1-D    Heisenberg    antiferromagnet:
4230: %identification with the O(3) nonlinear sigma model
4231: %
4232: 
4233: \href{http://link.aps.org/abstract/PRL/v50/e1153}{Phys. Rev. Lett. {\bf 50}, 1153 (1983)}.
4234: %Nonlinear   Field Theory  of  Large-Spin Heisenberg  Antiferromagnets:
4235: %Semiclassically  Quantized Solitons  of the One-Dimensional  Easy-Axis
4236: %N?el State
4237: %
4238: 
4239: \bibitem{haldane88}
4240: F.~D.~M.~Haldane, \href{http://link.aps.org/abstract/PRL/v61/e1029}{Phys. Rev. Lett. {\bf 61}, 1029 (1988)}.
4241: 
4242: %O(3) Nonlinear sigma   Model and the  Topological  Distinction between
4243: %Integer- and Half-Integer-Spin Antiferromagnets in Two Dimensions
4244: 
4245: 
4246: \bibitem{noHopfTerm2D88}
4247: X.~G.~Wen and A.~Zee, \href{http://link.aps.org/abstract/PRL/v61/e1025}{Phys. Rev. Lett. {\bf 61}, 1025 (1988)}.
4248: E.~Fradkin and M.~Stone, \href{http://link.aps.org/abstract/PRB/v38/e7215}{Phys. Rev. B {\bf 38}, 7215 (1988)}.
4249: T.~Dombre and N.~Read, \href{http://link.aps.org/abstract/PRB/v38/e7181}{Phys. Rev. B {\bf 38}, 7181 (1988)}.
4250: 
4251: 
4252: \bibitem{hkt03}
4253: K.~Harada, N.~Kawashima, and M.~Troyer,
4254: \href{http://link.aps.org/abstract/PRL/v90/e117203}{Phys. Rev. Lett. {\bf 90}, 117203 (2003)}.
4255: 
4256: %N\'el and  Spin-Peierls Ground States  of Two-Dimensional SU(N) Quantum
4257: %Antiferromagnets
4258: 
4259: 
4260: \bibitem{fs79}
4261: E. Fradkin and S. H. Shenker, \href{http://link.aps.org/abstract/PRD/v19/e3682}{Phys. Rev. D {\bf 19}, 3682 (1979)}.
4262: %Phase diagram of lattice gauge theories with Higgs fields
4263: 
4264: \bibitem{k61}
4265: P. W. Kasteleyn, Physica {\bf 27}, 1209 (1961).
4266: 
4267: \bibitem{f61}
4268: M. E. Fisher, \href{http://link.aps.org/abstract/PR/v124/e1664}{Phys. Rev. {\bf 124}, 1664 (1961)}.
4269: %Statistical Mechanics of Dimers on a Plane Lattice
4270: 
4271: 
4272: \bibitem{k63}
4273: P.~W.~Kasteleyn,
4274: %Dimer statistics and phase transitions.
4275: J. of Math. Phys. {\bf 4}, 287 (1963).
4276: 
4277: \bibitem{msf02}
4278: R. Moessner, S.~L. Sondhi, E.~Fradkin,
4279: %Short-ranged RVB physics, quantum dimer models and Ising gauge theories.
4280: \href{http://link.aps.org/abstract/PRB/v65/e024504}{Phys. Rev. B {\bf 65}, 024504 (2002)}.
4281: 
4282: \bibitem{msp02}
4283: G. Misguich, D.~Serban, V.~Pasquier,
4284: \href{http://link.aps.org/abstract/PRL/v89/e137202}{Phys. Rev. Lett. {\bf 89}, 137202 (2002)}.
4285: 
4286: \bibitem{ms02}
4287: %Ising and dimer models in two and three dimensions
4288: R. Moessner, S. L. Sondhi, \href{http://link.aps.org/abstract/PRB/v68/e054405}{Phys. Rev. B {\bf 68}, 054405 (2003)}.
4289: 
4290: \bibitem{rk88}
4291: D. S. Rokhsar and S. A. Kivelson,
4292: %Superconductivity and the quantum hard-core dimer gas.
4293: \href{http://link.aps.org/abstract/PRL/v61/e2376}{Phys. Rev. Lett. {\bf 61}, 2376 (1988)}.
4294: 
4295: 
4296: \bibitem{ms01}
4297: R. Moessner and S.~L. Sondhi, \href{http://link.aps.org/abstract/PRL/v86/e1881}{Phys. Rev. Lett. {\bf 86}, 1881 (2001)}.
4298: %Resonating Valence Bond Phase in the Triangular Lattice Quantum Dimer Model
4299: 
4300: 
4301: \bibitem{msc00}
4302: R. Moessner, S.~L. Sondhi, and P. Chandra,
4303: \href{http://link.aps.org/abstract/PRL/v84/e4457}{Phys. Rev. Lett. {\bf 84}, 4457 (2000)}.
4304: 
4305: 
4306: \bibitem{s88}
4307: B. Sutherland, \href{http://link.aps.org/abstract/PRB/v37/e3786}{Phys. Rev. B {\bf 37}, 3786 (1988)}.
4308: 
4309: \bibitem{mz91} J. B. Marston, C. Zeng,
4310: %Spin-Peierls and spin-liquid phases of kagome quantum antiferromagnets.
4311: J. Appl. Phys. {\bf 69}, 5962 (1991).
4312: 
4313: \bibitem{lcr96}
4314: P. W. Leung, K. C. Chiu and K. J. Runge,
4315: \href{http://link.aps.org/abstract/PRB/v54/e12938}{Phys. Rev. B {\bf 54}, 12938 (1996)}.
4316: 
4317: %Columnar  dimer and  plaquette resonating-valence-bond  orders  in the
4318: %quantum dimer model
4319: 
4320: 
4321: \bibitem{fs63}
4322: %Statistical  Mechanics  of  Dimers  on  a  Plane  Lattice.  II.  Dimer
4323: %Correlations and Monomers
4324: M. E. Fisher and J. Stephenson \href{http://link.aps.org/abstract/PR/v132/e1411}{Phys. Rev. {\bf 132}, 1411 (1963)}.
4325: 
4326: \bibitem{ms03}
4327: R. Moessner and  S.~L.   Sondhi,  \href{http://arxiv.org/abs/cond-mat/0307592}{cond-mat/0307592}.  See  Appendix   B
4328: concerning the square-lattice QDM .
4329: 
4330: \bibitem{msc01}
4331: R.~Moessner, S.~L.~Sondhi  and P.~Chandra,
4332: \href{http://link.aps.org/abstract/PRB/v64/e144416}{Phys. Rev. B {\bf 64}, 144416 (2001)}.
4333: 
4334: %Phase diagram of the hexagonal lattice quantum dimer model
4335: 
4336: \bibitem{hkms03}
4337: D.~A.~Huse,  W.~Krauth, R.~Moessner and S.~L.~Sondhi, \href{http://arxiv.org/abs/cond-mat/0305318}{cond-mat/0305318}
4338: and references therein.
4339: 
4340: \bibitem{height_representation}
4341: 
4342: L.~S.~Levitov, \href{http://link.aps.org/abstract/PRL/v64/e92}{Phys. Rev. Lett. {\bf 64}, 92 (1990)},
4343: %Equivalence  of  the  dimer  resonating-valence-bond  problem  to  the
4344: %quantum roughening problem
4345: 
4346: %-------
4347: C.~L.~Henley, J. Stat. Phys. {\bf 89}, 483 (1997).
4348: 
4349: %Relaxation time for a dimer covering with height representation 
4350: 
4351: \bibitem{fms02}
4352: P. Fendley, R.~Moessner and S.~L.~Sondhi,
4353: \href{http://link.aps.org/abstract/PRB/v66/e214513}{Phys. Rev. B {\bf 66}, 214513 (2002)}.
4354: 
4355: %Classical dimers on the triangular lattice
4356: 
4357: \bibitem{iif02}
4358: A.  Ioselevich, D.~A.~Ivanov  and M.~V.~Feigelman,
4359: \href{http://link.aps.org/abstract/PRB/v66/e174405}{Phys. Rev. B {\bf 66}, 174405 (2002)}.
4360: %Ground-state  properties of  the Rokhsar-Kivelson  dimer model  on the
4361: %triangular lattice
4362: 
4363: \bibitem{ifiitb02}
4364: L.~B.~Ioffe, M.~V.~Feigel'man, A.~Ioselevich, D.~Ivanov,  M.~Troyer,
4365: G.~Blatter, Nature {\bf 415}, 503 (2002).
4366: 
4367: \bibitem{mlms02}
4368: G.~Misguich,   C.~Lhuillier,   M.~Mambrini,   P.~Sindzingre,
4369: \href{http://publish.edpsciences.com/abstract/EPJB/v26/p167}{Eur. Phys. J. B {\bf 26}, 167 (2002)}.
4370: 
4371: %Degeneracy   of  the   ground-state   of  antiferromagnetic   spin-1/2
4372: %Hamiltonians
4373: 
4374: 
4375: \bibitem{wen91}
4376: X. G. Wen, \href{http://link.aps.org/abstract/PRB/v44/e2664}{Phys. Rev. B {\bf 44}, 2664 (1991)}.
4377: 
4378: %Mean-field  theory of spin-liquid  states with  finite energy  gap and
4379: %topological orders
4380: 
4381: \bibitem{k89}
4382: S. Kivelson, \href{http://link.aps.org/abstract/PRB/v39/e259}{Phys. Rev. B {\bf 39}, 259 (1989)}.
4383: 
4384: \bibitem{rc89}
4385: N. Read and B. Chakraborty, \href{http://link.aps.org/abstract/PRB/v40/e7133}{Phys. Rev. B {\bf 40}, 7133 (1989)}.
4386: 
4387: \bibitem{sf00}
4388: T. Senthil and M.~P.~A.~Fisher,
4389: \href{http://link.aps.org/abstract/PRB/v62/e7850}{Phys. Rev. B {\bf 62}, 7850 (2000)}.
4390: %Z2 gauge  theory of electron fractionalization  in strongly correlated
4391: %systems
4392: 
4393: \bibitem{sf01}
4394: T. Senthil and M. P. A. Fisher,
4395: \href{http://link.aps.org/abstract/PRL/v86/e292}{Phys. Rev. Lett. {\bf 86}, 292 (2001)};
4396: \href{http://link.aps.org/abstract/PRB/v63/e134521}{Phys. Rev. B {\bf 63}, 134521 (2001)}. 
4397: 
4398: \bibitem{feynman52-53}
4399: R.~P.~Feynman, \href{http://link.aps.org/abstract/PR/v90/e1116}{Phys. Rev. {\bf 90}, 1116 (1952)},
4400: \href{http://link.aps.org/abstract/PR/v91/e1291}{Phys. Rev. {\bf 91}, 1291 (1953)}.
4401: 
4402: \bibitem{ez93}
4403: V. Elser and C. Zeng.
4404: %kagome spin-1/2 antiferromagnets in the hyperbolic plane.
4405: \href{http://link.aps.org/abstract/PRB/v48/e13647}{Phys. Rev. B {\bf 48}, 13647 (1993)}. 
4406: 
4407: \bibitem{hw88}
4408: A.~J.~Phares and   F.~J.~Wunderlich,
4409: Nuovo Cimento   B {\bf  101}, 653 (1988).
4410: \bibitem{msp03b}
4411: See {\S}V.E.6 of Ref.\cite{msp03}
4412: 
4413: \bibitem{msp03c}  This  follows from  the  independence  of the  arrow
4414: variables, see {\S}V.B of Ref.\cite{msp03}
4415: 
4416: %Quantum dimer model with  extensive ground-state entropy on the kagome
4417: %lattice
4418: 
4419: \bibitem{polyakov87}
4420: A.~M.~Polyakov, {\it Gauge Fields and Strings}, (Harwood Academic, New York, 1987).
4421: 
4422: 
4423: 
4424: \bibitem{kogut79}
4425: %An introduction to lattice gauge theory and spin systems
4426: J. B. Kogut, \href{http://link.aps.org/abstract/RMP/v51/p659}{Rev. Mod. Phys. {\bf 51}, 659 (1979)}.
4427: 
4428: \bibitem{msp03} G.~Misguich,  D.~Serban and V.~Pasquier,
4429: \href{http://link.aps.org/abstract/PRB/v67/e214413}{Phys. Rev. B {\bf 67}, 214413 (2003)}. 
4430: 
4431: \bibitem{ze95}
4432: C.~Zeng    and   V.~Elser.
4433: %Quantum     dimer  calculations   on  the
4434: %spin-$\frac{1}{2}$ kagome Heisenberg  antiferromagnet.
4435: \href{http://link.aps.org/abstract/PRB/v51/e8318}{Phys. Rev. B {\bf 51}, 8318 (1995)}.
4436: 
4437: \bibitem{thouless65}
4438: D. J. Thouless, Proc. Phys. Soc. London {\bf 86}, 893 (1965).
4439: %Exchange in solid 3He and the Heisenberg Hamiltonian.
4440: 
4441: \bibitem{rhd83}  M. Roger,  J.  H. Hetherington,  and  J. M.  Delrieu,
4442: \href{http://link.aps.org/abstract/RMP/v55/p1}{Rev. Mod. Phys. {\bf 55}, 1 (1983)}.
4443: %Magnetism in solid 3He.
4444: 
4445: \bibitem{cf85} M. C. Cross and D.  S. Fisher,
4446: \href{http://link.aps.org/abstract/RMP/v57/p881}{Rev. Mod. Phys. {\bf 57}, 881 (1985)}.
4447: %Magnetism in solid  3He: Confrontation between  theory and experiment.
4448: \bibitem{greywall}   D.~S.~Greywall   and   P.~A.~Busch,
4449: \href{http://link.aps.org/abstract/PRL/v62/e1868}{Phys. Rev. Lett. {\bf 62}, 1868 (1989)}.
4450: %Heat capacity of 3He  adsorbed on graphite at millikelvin temperatures
4451: %and near third-layer promotion.
4452: D.~S.~Greywall,  \href{http://link.aps.org/abstract/PRB/v41/e1842}{Phys. Rev. B {\bf 41}, 1842 (1990)}.
4453: %Heatcapacity  of  multilayers  of  3He  adsorbed on  graphite  at  low
4454: %millikelvin temperatures.
4455: \bibitem{gr95}
4456: H. Godfrin and R.~E.~Rapp, Adv. Phys. {\bf 44}, 113 (1995).
4457: %Two-dimensional nuclear magnets.
4458: 
4459: \bibitem{roger84}
4460: M.~Roger, \href{http://link.aps.org/abstract/PRB/v30/e6432}{Phys. Rev. B {\bf 30}, 6432 (1984)}.
4461: %Multiple exchange in 3He and in the Wigner solid.
4462: 
4463: \bibitem{rbbcg98} M.  Roger, C. B\"auerle,  Yu.~M.~Bunkov, A.-S.~Chen, and
4464: H.~Godfrin, \href{http://link.aps.org/abstract/PRL/v80/e1308}{Phys. Rev. Lett. {\bf 80}, 1308 (1998)}.
4465: 
4466: %Multiple-Spin   Exchange  on  a   Triangular Lattice:   A Quantitative
4467: %Interpretation of Thermodynamic  Properties  of Two-Dimensional  Solid
4468: %3He.
4469: 
4470: 
4471: \bibitem{cj87}
4472: D.~M.~Ceperley and G.  Jacucci,    \href{http://link.aps.org/abstract/PRL/v58/e1648}{Phys. Rev. Lett. {\bf 58}, 1648 (1987)}.
4473: 
4474: %Calculation of exchange frequencies  in bcc 3He with the path-integral
4475: %Monte Carlo method.
4476: 
4477: \bibitem{bcl92}  B.  Bernu,  D.  Ceperley,  and  C.  Lhuillier,  J.  Low
4478: Temp. Phys. {\bf 89}, 589 (1992).
4479: 
4480: 
4481: \bibitem{ceperley95}
4482: D.~M.~Ceperley, \href{http://link.aps.org/abstract/RMP/v67/p279}{Rev. Mod. Phys. {\bf 67}, 279 (1995)}.
4483: %Path integrals in the theory of condensed helium.
4484: 
4485: \bibitem{bc99} B.  Bernu and D.  Ceperley, in {\it Quantum Monte Carlo
4486: Methods in  Physics and Chemistry},  edited by M. P.   Nightingale and
4487: C. J. Umrigar (Kluwer, Dordrecht, The Netherlands, 1999).
4488: 
4489: \bibitem{ah00}
4490: H.~Ashizawa and D.~S.~Hirashima, \href{http://link.aps.org/abstract/PRB/v62/e9413}{Phys. Rev. B {\bf 62}, 9413 (2000)}.
4491: %WKB calculation of multiple spin exchange in monolayer solid 3He.
4492: 
4493: \bibitem{drh80}  J.~M.~Delrieu, M.~Roger,  J.~H.~Hetherington,  J. Low
4494: Temp.  Phys.  {\bf 40}, 71 (1980).
4495: 
4496: %Exchange and magnetic order in hcp helium-3 and adsorbed helium-3 with
4497: %triangular lattice.
4498: 
4499: \bibitem{frg86}   H.~Franco,   R.~E.~Rapp   and   H.~Godfrin,
4500: \href{http://link.aps.org/abstract/PRL/v57/e1161}{Phys. Rev. Lett. {\bf 57}, 1161 (1986)}.
4501: 
4502: %Nuclear Ferromagnetism of Two-Dimensional 3He.
4503: 
4504: \bibitem{mblw98}   G.~Misguich,   B.~Bernu,  C.~Lhuillier,   and
4505: C.~Waldtmann, \href{http://link.aps.org/abstract/PRL/v81/e1098}{Phys. Rev. Lett. {\bf 81}, 1098 (1998)}.
4506: 
4507: \bibitem{lmsl00}
4508: W.~LiMing,  G.~Misguich, P.~Sindzingre, and  C.~Lhuillier,
4509: \href{http://link.aps.org/abstract/PRB/v62/e6372}{Phys. Rev. B {\bf 62}, 6372 (2000)}.  
4510: 
4511: 
4512: %From N\'eel  long-range order to  spin  liquids  in the  multiple-spin
4513: %exchange model.
4514: 
4515: \bibitem{kmyf97}  K.~Ishida,  M.~Morishita, K.~Yawata,  and  H.~Fukuyama,
4516: \href{http://link.aps.org/abstract/PRL/v79/e3451}{Phys. Rev. Lett. {\bf 79}, 3451 (1997)}.
4517: 
4518: %Low Temperature Heat-Capacity Anomalies in Two-Dimensional Solid 3He.
4519: 
4520: \bibitem{cthrbbg01}
4521: %E.~Collin,   S.~Triqueneaux,  R.~Harakaly,   M.~Roger,   C.~B\"auerle,
4522: %Yu.~M.~Bunkov, and H.~Godfrin
4523: E.~Collin {\it et al.}, \href{http://link.aps.org/abstract/PRL/v86/e2447}{Phys. Rev. Lett. {\bf 86}, 2447 (2001)}.
4524: 
4525: %Quantum   Frustration in the  "Spin   Liquid" Phase of Two-Dimensional
4526: %$^3$He.
4527: 
4528: \bibitem{kh00}
4529: M. Katano and D.~S.~Hirashima, \href{http://link.aps.org/abstract/PRB/v62/e2573}{Phys. Rev. B {\bf 62}, 2573 (2000)}.
4530: %Multiple-spin exchange in a two-dimensional Wigner crystal.
4531: 
4532: \bibitem{hk01}          D.~S.~Hirashima          and          K.~Kubo,
4533: \href{http://link.aps.org/abstract/PRB/v63/e125340}{Phys. Rev. B {\bf 63}, 125340 (2001)}.
4534: 
4535: \bibitem{bcc01}   B.   Bernu,  L.   Candido,   and   D.  M.   Ceperley
4536: \href{http://link.aps.org/abstract/PRL/v86/e870}{Phys. Rev. Lett. {\bf 86}, 870 (2001)}.
4537: 
4538: %Exchange Frequencies in the 2D Wigner Crystal.
4539: 
4540: \bibitem{ok98} T. Okamoto  and S. Kawaji,
4541: \href{http://link.aps.org/abstract/PRB/v57/e9097}{Phys. Rev. B {\bf 57}, 9097 (1998)}.
4542: 
4543: %Magnetism in a  Wigner solid and  the Aharonov-Bohm effect: Experiment
4544: %and theory.
4545: 
4546: \bibitem{rd89} M. Roger and J. M. Delrieu,
4547: \href{http://link.aps.org/abstract/PRB/v39/e2299}{Phys. Rev. B {\bf 39}, 2299 (1989)}.
4548: 
4549: %Cyclic  four-spin  exchange   on a   two-dimensional   square lattice:
4550: %Possible applications in high-T$_c$ superconductors.
4551: 
4552: \bibitem{sugai90}
4553: S.~Sugai {\it et al.},
4554: %M.~Sato,  T.~Kobayashi,   J.~Akimitsu,
4555: %T.~Ito, H.~Takagi, S.~Uchida, S.~Hosoya, T.~Kajitani and T.~Fukuda,
4556: \href{http://link.aps.org/abstract/PRB/v42/e1045}{Phys. Rev. B {\bf 42}, 1045 (1990)}.
4557: 
4558: %High-energy spin  excitations  in the insulating  phases of high-T$_c$
4559: %superconducting cuprates and La$_2$NiO$_4$.
4560: 
4561: \bibitem{coldea01}
4562: R.~Coldea  {\it  et   al.},
4563: %S.~M.~Hayden,  G.~Aeppli,   T.~G.~Perring,  C.~D.~Frost,  T.~E.~Mason,
4564: %S.-W.~Cheong, and Z.~Fisk,
4565: \href{http://link.aps.org/abstract/PRL/v86/e5377}{Phys. Rev. Lett. {\bf 86}, 5377 (2001)}.
4566: 
4567: %Spin Waves and Electronic Interactions in La$_2$CuO$_4$. 
4568: 
4569: \bibitem{mr02}
4570: E.~M\"uller-Hartmann and A.~Reischl,
4571: \href{http://publish.edpsciences.com/abstract/EPJB/v28/p173}{Eur. Phys. J. B {\bf 28}, 173 (2002)}.
4572: 
4573: %Derivation  of   effective spin models  from  a  three  band model for
4574: %CuO$_2$-planes.
4575: 
4576: \bibitem{kk03} A.~A.~Katanin, and A.~P.~Kampf,
4577: \href{http://link.aps.org/abstract/PRB/v67/e100404}{Phys. Rev. B {\bf 67}, 100404R (2003)}.
4578: 
4579: %Theoretical  analysis of magnetic  Raman  scattering in La$_2$CuO$_4$:
4580: %Two-magnon intensity with the inclusion of ring exchange.
4581: 
4582: \bibitem{mkebm00}
4583: 
4584: M.   Matsuda {\it et al.},
4585: %K.   Katsumata,   R.~S.~Eccleston,
4586: %S. Brehmer and H.-J.~Mikeska,
4587: \href{http://link.aps.org/abstract/PRB/v62/e8903}{Phys. Rev. B {\bf 62}, 8903 (2000)}.
4588: %Magnetic  excitations  and  exchange  interactions in  the  spin-($\frac{1}{2}$)
4589: %two-leg ladder compound La$_6$Ca$_8$Cu$_{24}$O$_{41}$.
4590: 
4591: \bibitem{bmmnu99}
4592: S.~Brehmer {\it et al.},
4593: %H.-J.~Mikeska,  M.~M\"uller, N.~Nagaosa and S.~Uchida,
4594: \href{http://link.aps.org/abstract/PRB/v60/e329}{Phys. Rev. B {\bf 60}, 329 (1999)}.
4595: 
4596: %Effects  of  biquadratic  exchange  on   the  spectrum of   elementary
4597: %excitations in spin ladders.
4598: 
4599: \bibitem{sku01}  K.   P.  Schmidt,  C.   Knetter  and  G.   S.  Uhrig,
4600: \href{http://publish.edpsciences.com/abstract/EPL/v56/p877}{Europhys. Lett. {\bf 56}, 877 (2001)}.
4601: 
4602: %Raman response in antiferromagnetic two-leg S=1/2 Heisenberg ladders.
4603: 
4604: \bibitem{gkt03}
4605: A.~G\"oßling {\it et al.},
4606: %U.~Kuhlmann, C.~Thomsen A.~L\"offert, C.~Gross, and W.~Assmus,
4607: \href{http://link.aps.org/abstract/PRB/v67/e052403}{Phys. Rev. B {\bf 67}, 052403 (2003)}.
4608: %Magnetic excitations
4609: 
4610: \bibitem{mvm02}
4611: M. M\"uller, T. Vekua, and H.-J. Mikeska,
4612: \href{http://link.aps.org/abstract/PRB/v66/e134423}{Phys. Rev. B {\bf 66}, 134423 (2002)}.
4613: %Perturbation theories  for the  S=1/2  spin ladder with  a four-spin
4614: %ring exchange
4615: 
4616: \bibitem{hmh03}
4617: T. Hikihara, T. Momoi, and X. Hu,
4618: \href{http://link.aps.org/abstract/PRL/v90/e087204}{Phys. Rev. Lett. {\bf 90}, 087204 (2003)}.
4619: 
4620: %Spin-Chirality Duality in a Spin Ladder with Four-Spin Cyclic Exchange.
4621: 
4622: \bibitem{lst03} A. L\"auchli,  G. Schmid, and M. Troyer,
4623: \href{http://link.aps.org/abstract/PRB/v67/e100409}{Phys. Rev. B {\bf 67}, 100409R (2003)}.
4624: %Phase diagram of a spin ladder with cyclic four-spin exchange.
4625: 
4626: \bibitem{hh01}
4627: Y.  Honda and T.  Horiguchi, \href{http://arxiv.org/abs/cond-mat/0106426}{cond-mat/0106426}.
4628: 
4629: %Quantum phase transition by  cyclic four-spin exchange interaction for
4630: %S=1/2 two-leg spin ladder.
4631: 
4632: \bibitem{hn02}
4633: K. Hijii and K.  Nomura \href{http://link.aps.org/abstract/PRB/v65/e104413}{Phys. Rev. B {\bf 65}, 104413 (2002)}.
4634: 
4635: %Universality  class of  an S=$\frac{1}{2}$  quantum spin  ladder  system with
4636: %four-spin exchange
4637: 
4638: \bibitem{mhnh03}  T. Momoi,  T.~Hikihara,  M.~Nakamura,
4639: Xiao      Hu,   \href{http://link.aps.org/abstract/PRB/v67/e174410}{Phys. Rev. B {\bf 67}, 174410 (2003)}.
4640: %Scalar chiral ground states of spin ladders with four-spin exchanges.
4641: 
4642: 
4643: \bibitem{lauchli03}
4644: A.  L\"auchli,   talk given at  the   {\em Highly Frustrated Magnetism
4645: 2003} conference, Grenoble, France (August 2003).
4646: 
4647: 
4648: \bibitem{km97}  K. Kubo  and T.  Momoi, Z.  Phys. B  {\bf 103},  485 (1997).
4649: %Ground  state  of a  spin  system  with  two- and  four-spin  exchange
4650: %interactions on the triangular lattice
4651: 
4652: \bibitem{mkn97}
4653: T.  Momoi, K.   Kubo, and K.  Niki,
4654: \href{http://link.aps.org/abstract/PRL/v79/e2081}{Phys. Rev. Lett. {\bf 79}, 2081 (1997)}.
4655: 
4656: %Possible Chiral Phase Transition in Two-Dimensional Solid 3He.
4657: 
4658: \bibitem{mbl98}   G.~Misguich,   B.~Bernu,   and   C.~Lhuillier,
4659: \href{http://dx.doi.org/10.1023/A:1022249501044}{J.   Low  Temp. Phys.
4660: {\bf 110},  327 (1998)}.
4661: 
4662: %The  multiple-spin exchange phase diagram   on the triangular lattice:
4663: %Schwinger-boson analysis
4664: 
4665: \bibitem{ksmn98}  K.  Kubo,  H.  Sakamoto,  T.~Momoi,  and  K.~Niki,
4666: \href{http://dx.doi.org/10.1023/A:1022249501044}{J.                 Low
4667: Temp. Phys.  {\bf 111}, 583 (1998)}.
4668: 
4669: %A Possible Magnetic Phase with Scalar Chirality in Solid 3He Layers.
4670: 
4671: 
4672: \bibitem{StagVBC}
4673: A 16- and  a 32-site triangular  lattices which  do not frustrate  the
4674: staggered VBC were    investigated.    In both  cases   some  of   the
4675: irreducible representations of the  space group required to break  the
4676: appropriate     lattice    symmetries    are   very   high    in   the
4677: spectrum. C. Lhuillier and G. Misguich (unpublished).
4678: 
4679: \bibitem{lsm61}
4680: E.~H.~Lieb, T.~D.~Schultz,  D.~C.~Mattis., Ann. Phys. (N.Y)  {\bf 16},
4681: 407 (1961).
4682: 
4683: \bibitem{al86}
4684: I. Affleck and E. Lieb, Lett. Math. Phys. {\bf 12},  57  (1986).
4685: 
4686: \bibitem{o00}
4687: M. Oshikawa, \href{http://link.aps.org/abstract/PRL/v84/e1535}{Phys. Rev. Lett. {\bf 84}, 1535 (2000)}.
4688: 
4689: \bibitem{hastings03}
4690: M.~B.~Hastings, \href{http://arxiv.org/abs/cond-mat/0305505}{cond-mat/0305505}.
4691: 
4692: \bibitem{o03}
4693: M.~Oshikawa,
4694: \href{http://link.aps.org/abstract/PRL/v90/e236401}{Phys. Rev. Lett. {\bf 90}, 236401 (2003)}. 
4695: 
4696: \bibitem{nge98}
4697: A.~A. Nersesyan, A.~O. Gogolin, and F.~H.~L. Essler,
4698: \href{http://link.aps.org/abstract/PRL/v81/e910}{Phys. Rev. Lett. {\bf 81}, 910 (1998)}.
4699: 
4700: \bibitem{ahllt98}
4701: P. Azaria {\it et ~al.}, \href{http://link.aps.org/abstract/PRL/v81/e1694}{Phys. Rev. Lett. {\bf 81}, 1694 (1998)}.
4702: 
4703: \bibitem{efkl00}
4704: V.~J.~Emery, E. Fradkin, S.~A.~Kivelson and T.~C.~Lubensky,
4705: \href{http://link.aps.org/abstract/PRL/v85/e2160}{Phys. Rev. Lett. {\bf 85}, 2160 (2000)}.
4706: 
4707: \bibitem{be01}
4708: M.  Bocquet, F. Essler, A.~M.~Tsvelik and A.~O.~Gogolin,
4709: \href{http://link.aps.org/abstract/PRB/v64/e094425}{Phys. Rev. B {\bf 64}, 094425 (2001)},
4710: M.~Bocquet, \href{http://link.aps.org/abstract/PRB/v65/e1884415}{Phys. Rev. B {\bf 65}, 1884415 (2001)}.
4711: 
4712: 
4713: \bibitem{vc01}
4714: A. Vishwanath and D. Carpentier, \href{http://link.aps.org/abstract/PRL/v86/e676}{Phys. Rev. Lett. {\bf 86}, 676 (2001)}.
4715: 
4716: \bibitem{sp02}
4717: S. Sachdev and K. Park, Annals of Physics (N.Y.), {\bf 58}, 298 (2002).
4718: 
4719: \bibitem{ssl02}
4720: O.~A. Starykh and R.~R.~P. Singh and  G.~C. Levine,
4721: \href{http://link.aps.org/abstract/PRL/v88/e167203}{Phys. Rev. Lett. {\bf 88}, 167203 (2002)}.
4722: 
4723: \bibitem{cttt01}
4724: R. Coldea, D.~A. Tennant, A.~M. Tsvelik, and Z. Tylczynski,
4725: \href{http://link.aps.org/abstract/PRL/v86/e1335}{Phys. Rev. Lett. {\bf 86}, 1335 (2001)}.
4726: 
4727: \bibitem{kitaev97}
4728: A.  Kitaev, Annals Phys. {\bf 303}, 2 (2003).
4729: [\href{http://arXiv.org/abs/quant-ph/9707021}{quant-ph/9707021}]
4730: 
4731: \bibitem{ns01}  C. Nayak  and  K.  Shtengel,
4732: \href{http://link.aps.org/abstract/PRB/v64/e064422}{Phys. Rev. B {\bf 64}, 064422 (2001)}.
4733: 
4734: %Microscopic models   of  two-dimensional magnets   with fractionalized
4735: %excitations
4736: 
4737: \bibitem{bfg02}
4738: L.  Balents, M.  P.  A. Fisher, and S.  M.  Girvin,
4739: \href{http://link.aps.org/abstract/PRB/v65/e224412}{Phys. Rev. B {\bf 65}, 224412 (2002)}.
4740: 
4741: %Fractionalization in an easy-axis kagome antiferromagnet
4742: 
4743: \bibitem{pbf02}  A.~Paramekanti, L.~Balents, and  M.~P.~A.~Fisher
4744: \href{http://link.aps.org/abstract/PRB/v66/e054526}{Phys. Rev. B {\bf 66}, 054526 (2002)}.
4745: 
4746: %Ring  exchange,  the exciton  Bose  liquid,  and  bosonization in  two
4747: %dimensions
4748: 
4749: \bibitem{sdss02}  A.~W.~Sandvik,  S.~Daul,  R.~R.~P.~Singh,  and
4750:  D.~J.~Scalapino, \href{http://link.aps.org/abstract/PRL/v89/e247201}{Phys. Rev. Lett. {\bf 89}, 247201 (2002)}.
4751: 
4752: %Striped Phase in a Quantum XY Model with Ring Exchange
4753: 
4754: \bibitem{sm02} T.  Senthil and O.  Motrunich,
4755: \href{http://link.aps.org/abstract/PRB/v66/e205104}{Phys. Rev. B {\bf 66}, 205104 (2002)}.
4756: %Microscopic  models for fractionalized  phases in  strongly correlated
4757: %systems.
4758: 
4759: \bibitem{p38}
4760: L. Pauling,  in {\em The nature of the chemical bond} (Cornell University
4761:   Press, Ithaca, 1938).
4762: 
4763: \bibitem{kn53}
4764: K. Kano and S. Naya, Prog. in Theor. Phys. {\bf 10},  158  (1953).
4765: 
4766: \bibitem{hr92}
4767: D. Huse and A. Rutenberg, \href{http://link.aps.org/abstract/PRB/v45/e7536}{Phys. Rev. B {\bf 45}, 7536 (1992)}.
4768: 
4769: \bibitem{ms01b}
4770: R. Moessner and S.~L. Sondhi, \href{http://link.aps.org/abstract/PRB/v63/e224401}{Phys. Rev. B {\bf 63}, 224401 (2001)}.
4771: 
4772: \bibitem{chs92}
4773: J. Chalker, P.~C.~W.~Holdsworth, and E.~F.~Shender,
4774: \href{http://link.aps.org/abstract/PRL/v68/e855}{Phys. Rev. Lett. {\bf 68}, 855 (1992)}.
4775: 
4776: \bibitem{baxter70}
4777: R. J. Baxter, J. Math. Phys. {\bf 11}, 784 (1970).
4778: 
4779: \bibitem{rcc93}
4780: I. Richtey, P.~Chandra, and P.~Coleman, \href{http://link.aps.org/abstract/PRB/v47/e15342}{Phys. Rev. B {\bf 47}, 15342 (1993)}.
4781: 
4782: \bibitem{rb93}
4783: J. Reimers and A.~Berlinsky, \href{http://link.aps.org/abstract/PRB/v48/e9539}{Phys. Rev. B {\bf 48}, 9539 (1993)}.
4784: 
4785: \bibitem{e02}
4786: M.  Elhajal,           Ph.~D.     thesis,     Universit\'e      Joseph
4787: Fourier. Grenoble. France, 2002.
4788:  
4789: \bibitem{ecl02}
4790: M.~Elhajal, B.~Canals, and C.~Lacroix,
4791: \href{http://link.aps.org/abstract/PRB/v66/e014422}{Phys. Rev. B {\bf 66}, 014422 (2002)}.
4792: 
4793: \bibitem{k94}
4794: A. Keren, \href{http://link.aps.org/abstract/PRL/v72/e3254}{Phys. Rev. Lett. {\bf 72}, 3254 (1994)}.
4795: 
4796: \bibitem{e89}
4797: V. Elser, \href{http://link.aps.org/abstract/PRL/v62/e2405}{Phys. Rev. Lett. {\bf 62}, 2405 (1989)}.
4798: 
4799: \bibitem{ce92}
4800: J. Chalker and J. Eastmond, \href{http://link.aps.org/abstract/PRB/v46/e14201}{Phys. Rev. B {\bf 46}, 14201 (1992)}.
4801: 
4802: \bibitem{s92}
4803: S. Sachdev, \href{http://link.aps.org/abstract/PRB/v45/e12377}{Phys. Rev. B {\bf 45}, 12377 (1992)}.
4804: 
4805: \bibitem{le93}
4806: P. Leung and V. Elser, \href{http://link.aps.org/abstract/PRB/v47/e5459}{Phys. Rev. B {\bf 47}, 5459 (1993)}.
4807: 
4808: 
4809: \bibitem{lblps97}
4810: P. Lecheminant {\it et~al.}, \href{http://link.aps.org/abstract/PRB/v56/e2521}{Phys. Rev. B {\bf 56}, 2521 (1997)}.
4811: 
4812: \bibitem{web98}
4813: C. Waldtmann {\it et~al.}, \href{http://publish.edpsciences.com/abstract/EPJB/v2/p501}{Eur. Phys. J. B {\bf 2}, 501 (1998)}.
4814: 
4815: \bibitem{m98}
4816: F. Mila, \href{http://link.aps.org/abstract/PRL/v81/e2356}{Phys. Rev. Lett. {\bf 81}, 2356 (1998)}.
4817: 
4818: \bibitem{smlbpwe00}
4819: P. Sindzingre {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v84/e2953}{Phys. Rev. Lett. {\bf 84}, 2953 (2000)}.
4820: 
4821: \bibitem{mm01}
4822: M. Mambrini and F. Mila, \href{http://publish.edpsciences.com/abstract/EPJB/v17/p651}{Eur. Phys. J. B {\bf 17}, 651 (2001)}.
4823: 
4824: \bibitem{dmnm03}
4825: S.~Dommange, M.~Mambrini, B.~Normand and F.~Mila, \href{http://arxiv.org/abs/cond-mat/0306299}{cond-mat/0306299}.
4826: 
4827: \bibitem{ze90}
4828: C. Zeng and V. Elser, \href{http://link.aps.org/abstract/PRB/v42/e8436}{Phys. Rev. B {\bf 42}, 8436 (1990)}.
4829: 
4830: \bibitem{sh92}
4831: R. Singh and D. Huse, \href{http://link.aps.org/abstract/PRL/v68/e1766}{Phys. Rev. Lett. {\bf 68}, 1766 (1992)}.
4832: 
4833: \bibitem{h01}
4834: K. Hida, \href{http://jpsj.ipap.jp/link?JPSJ/70/3673}{J. Phys. Soc. Jpn. {\bf 70}, 3673 (2001)}.
4835: 
4836: \bibitem{cghp02}
4837: D.~C.~Cabra, M.~D.~Grynberg, P.~C.~W.~Holdsworth,  P.~Pujol,
4838: \href{http://link.aps.org/abstract/PRB/v65/e094418}{Phys. Rev. B {\bf 65}, 094418 (2002)}.
4839: 
4840: \bibitem{ey94}
4841: N. Eltsner and A.~P. Young, \href{http://link.aps.org/abstract/PRB/v50/e6871}{Phys. Rev. B {\bf 50}, 6871 (1994)}.
4842: 
4843: \bibitem{nm95}
4844: T. Nakamura and S. Miyashita, \href{http://link.aps.org/abstract/PRB/v52/e9174}{Phys. Rev. B {\bf 52}, 9174 (1995)}.
4845: 
4846: \bibitem{tr96}
4847: P.~Tomczak and J.~Richter, \href{http://link.aps.org/abstract/PRB/v54/e9004}{Phys. Rev. B {\bf 54}, 9004 (1996)}.
4848: 
4849: \bibitem{rhw00}
4850: A.~P. Ramirez, B. Hessen, and M.~Winkelmann, \href{http://link.aps.org/abstract/PRL/v84/e2957}{Phys. Rev. Lett. {\bf 84}, 2957 (2000)}.
4851: 
4852: \bibitem{mklmch00}
4853: P. Mendels {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v85/e3496}{Phys. Rev. Lett. {\bf 85}, 3496 (2000)}.
4854: 
4855: \bibitem{sma02}
4856: A.~V.~Syromyatnikov and S.~V.~Maleyev,
4857: \href{http://link.aps.org/abstract/PRB/v66/e132408}{Phys. Rev. B {\bf 66}, 132408 (2002)}.
4858: 
4859: \bibitem{ns03}
4860: P. Nikolic and T. Senthil, \href{http://arxiv.org/abs/cond-mat/0305189}{cond-mat/0305189}.
4861: 
4862: \bibitem{mmpfa00}
4863: T. Mondelli {\it et~al.}, Physica B {\bf 284},  1371  (2000).
4864: 
4865: \bibitem{gsf01}
4866: A. Georges, R. Siddhartan and S. Florens, \href{http://link.aps.org/abstract/PRL/v87/e277203}{Phys. Rev. Lett. {\bf 87}, 277203 (2001)}.
4867: 
4868: \bibitem{ls02}
4869: C. Lhuillier and P. Sindzingre,  in {\em Quantum properties of Low dimensional
4870:   antiferromagnets}, edited by Y.~Ajiro and J.~P.~Boucher (Kyushu University
4871:   Press, Fukuoka, Japan, 2002), p.\ 111.%, iSBN 4 87378 740 8.
4872: 
4873: \bibitem{sv00}
4874: S.~Sachdev and M.~Vojta,
4875: Proceedings of the XIII International Congress on Mathematical Physics, July 2000, London.
4876: A.~Fokas, A.~Grigoryan, T.~Kibble, and B.~Zegarlinski eds, International Press, Boston (2001)
4877: [\href{http://arxiv.org/abs/cond-mat/0009202}{cond-mat/0009202}].
4878: 
4879: \bibitem{ukkll94}
4880: Y. Uemura {\it et~al.}, \href{http://link.aps.org/abstract/PRL/v73/e3306}{Phys. Rev. Lett. {\bf 73}, 3306 (1994)}.
4881: 
4882: \bibitem{rec90}
4883: A. Ramirez, G.~P. Espinosa,  and A.~S.~Cooper,
4884: \href{http://link.aps.org/abstract/PRL/v64/e2070}{Phys. Rev. Lett. {\bf 64}, 2070 (1990)}.
4885: 
4886: \bibitem{lbar96}
4887: S.-H. Lee {\it et~al.}, \href{http://publish.edpsciences.com/abstract/EPL/v35/p127}{Europhys. Lett. {\bf 35}, 127 (1996)}.
4888: 
4889: \bibitem{kklllwutdg96}
4890: A. Keren {\it et~al.}, \href{http://link.aps.org/abstract/PRB/v53/e6451}{Phys. Rev. B {\bf 53}, 6451 (1996)}.
4891: 
4892: \bibitem{whmmt98}
4893: A.~S. Wills {\it et~al.}, \href{http://publish.edpsciences.com/abstract/EPL/v42/p325}{Europhys. Lett. {\bf 42}, 325 (1998)}.
4894: 
4895: \bibitem{aoyhimw94}
4896: K. Awaga {\it et~al.}, \href{http://link.aps.org/abstract/PRB/v49/e3975}{Phys. Rev. B {\bf 49}, 3975 (1994)}.
4897: 
4898: \bibitem{wkyoya97}
4899: N. Wada {\it et~al.}, \href{http://jpsj.ipap.jp/link?JPSJ/66/961}{J. Phys. Soc. Jpn. {\bf 66}, 961 (1997)}.
4900: 
4901: \bibitem{wwyoaonn98}
4902: I. Watanabe {\it et~al.}, \href{http://link.aps.org/abstract/PRB/v58/e2438}{Phys. Rev. B {\bf 58}, 2438 (1998)}.
4903: 
4904: \bibitem{cl98}
4905: B. Canals and C. Lacroix, \href{http://link.aps.org/abstract/PRL/v80/e2933}{Phys. Rev. Lett. {\bf 80}, 2933 (1998)}.
4906: 
4907: \bibitem{hbb91}
4908: A.~B.~Harris, A.~J.~Berlinsky and C.~Bruder,
4909: \href{http://link.aip.org/link/?JAP/69/5200/1}{J. Appl. Phys. {\bf 69},
4910: 5200 (1991)}.
4911: 
4912: \bibitem{t01}
4913: H. Tsunetsugu, \href{http://jpsj.ipap.jp/link?JPSJ/70/640}{J. Phys. Soc. Jpn. {\bf 70}, 640 (2001)}.
4914: 
4915: \bibitem{t02}
4916: H. Tsunetsugu, \href{http://link.aps.org/abstract/PRB/v65/e024415}{Phys. Rev. B {\bf 65}, 024415 (2002)}.
4917: 
4918: \bibitem{vdh92-93}
4919: J.~V. Delft and C.~L. Henley, \href{http://link.aps.org/abstract/PRL/v69/e3236}{Phys. Rev. Lett. {\bf 69}, 3236 (1992)},
4920: \href{http://link.aps.org/abstract/PRB/v48/e965}{Phys. Rev. B {\bf 48}, 965 (1993)}.
4921: 
4922: \bibitem{sachdev03}
4923: S.~Sachdev, Annals Phys. {\bf 303}, 226 (2003) [\href{http://arxiv.org/abs/cond-mat/0211027}{cond-mat/0211027}]
4924: and  Rev. Mod. Phys. {\bf 75}, 913 (2003) [\href{http://arxiv.org/abs/cond-mat/0211005}{cond-mat/0211005}].
4925: 
4926: \bibitem{wwz89}
4927: X. Wen, F. Wilczek, and A. Zee, \href{http://link.aps.org/abstract/PRB/v39/e11413}{Phys. Rev. B {\bf 39}, 11413 (1989)}.
4928: 
4929: \bibitem{kl87-89}
4930: V. Kalmeyer and R. Laughlin, \href{http://link.aps.org/abstract/PRL/v59/e2095}{Phys. Rev. Lett. {\bf 59}, 2095 (1987)},
4931: \href{http://link.aps.org/abstract/PRB/v39/e11879}{Phys. Rev. B {\bf 39}, 11879 (1989)}.
4932: 
4933: \bibitem{ywg93}
4934: K. Yang, L. Warman, and S.~M.~Girvin, \href{http://link.aps.org/abstract/PRL/v70/e2641}{Phys. Rev. Lett. {\bf 70}, 2641 (1993)}.
4935: 
4936: 
4937: \end{thebibliography}
4938: \end{document}
4939: 
4940: