1: \documentclass[showpacs,twocolumn,floatfix]{revtex4}

2: 

3: \usepackage{amsmath,amssymb,mathrsfs}

4: \usepackage{latexsym}

5: \usepackage{graphicx,psfrag} 

6: \usepackage[dvips]{color}

7: 

8: % boldsymbol (requires amsmath)

9: \newcommand{\bs}[1]{\boldsymbol{#1}}

10: \newcommand{\nn}{\nonumber}

11: \newcommand{\mb}[1]{\mathbf{#1}}

12: % 1/2

13: \newcommand{\half}{$\frac{1}{2}$ }

14: % commutator and anticommutator

15: \newcommand{\comm}[2]{\left[#1,#2\right]}

16: \newcommand{\anticomm}[2]{\left\{#1,#2\right\}}

17: % notation for vacuum, an empty set inside a ket

18: \newcommand{\vac}{\left|\,0\,\right\rangle}

19: % define the bra and ket, and braket

20: \newcommand{\ket}[1]{\left|#1\right\rangle}

21: \newcommand{\bra}[1]{\left\langle#1\right|}

22: \newcommand{\braket}[2]{\bigl\langle#1\bigl|\bigr.#2\bigr\rangle}

23: % simplifies using the up and down arrows to denote spin

24: \newcommand{\up}{\uparrow}

25: \newcommand{\dw}{\downarrow}

26: \newcommand{\cas}{\mathcal{C}^2_{\text{SU($n$)}}}

27: \newcommand{\casi}[2]{\mathcal{C}^2_{\text{SU(3)}}(#1,#2)}

28: \newcommand{\casin}[1]{\mathcal{C}^2_{\text{SU($n$)}}(#1)}

29: \newcommand{\casinn}[2]{\mathcal{C}^2_{\text{SU($#1$)}}(#2)}

30: \newcommand{\dimn}[2]{\text{dim}_{\text{SU($#1$)}}(#2)}

31: 

32: \def\ie{\emph{i.e.},\ }

33: \def\eg{\emph{e.g.}\ }

34: \def\ea{\emph{et al.}}

35: 

36: \def\b{\text{b}}

37: \def\r{\text{r}}

38: \def\g{\text{g}}

39: \def\y{\text{y}}

40: \def\c{\text{c}}

41: \def\m{\text{m}}

42: 

43: \allowdisplaybreaks[1]\begin{document}

44: \title{Valence bond solids for SU($\bs{n}$) spin chains:\\

45: exact models, spinon confinement, and the Haldane gap}

46: \author{Martin Greiter and Stephan Rachel}

47: \affiliation{Institut f\"ur Theorie der Kondensierten Materie,\\

48:   Universit\"at Karlsruhe, Postfach 6980, 76128 Karlsruhe, Germany}

49: \pagestyle{plain}

50: 

51: \begin{abstract}

52:   To begin with, we introduce several exact models for SU(3) spin

53:   chains: (1) a translationally invariant parent Hamiltonian involving

54:   four-site interactions for the trimer chain, with a three-fold

55:   degenerate ground state.  We provide numerical evidence that the

56:   elementary excitations of this model transform under representation

57:   $\bs{\bar 3}$ of SU(3) if the original spins of the model transform

58:   under rep.\ $\bs{3}$.  (2) a family of parent Hamiltonians for

59:   valence bond solids of SU(3) chains with spin reps.\ $\bs{6}$,

60:   $\bs{10}$, and $\bs{8}$ on each lattice site.  We argue that of

61:   these three models, only the latter two exhibit spinon confinement

62:   and hence a Haldane gap in the excitation spectrum.  We generalize

63:   some of our models to SU($n$).  Finally, we use the emerging rules

64:   for the construction of VBS states to argue that models of

65:   antiferromagnetic chains of SU($n$) spins in general possess a

66:   Haldane gap if the spins transform under a representation

67:   corresponding to a Young tableau consisting of a number of boxes

68:   $\lambda$ which is divisible by $n$.  If $\lambda$ and $n$ have no

69:   common divisor, the spin chain will support deconfined spinons and

70:   not exhibit a Haldane gap.  If $\lambda$ and $n$ have a common

71:   divisor different from $n$, it will depend on the specifics of the

72:   model including the range of the interaction.

73: \end{abstract}

74: 

75: \pacs{75.10.Jm, 75.10.Pq, 75.10.Dg, 32.80.Pj}

76: 

77: %75.10.Jm Quantized spin models

78: %75.10.Pq Spin chain models

79: %75.10.Dg Spin Hamiltonians

80: 

81: \maketitle

82: 

83: \section{Introduction}

84: 

85: Quantum spin chains have been a most rewarding subject of study almost

86: since the early days of quantum mechanics, beginning with the

87: invention of the Bethe ansatz in 1931~\cite{bethe31zp205} as a method

88: to solve the $S=\frac{1}{2}$ Heisenberg chain with nearest-neighbor

89: interactions.  The method led to the discovery of the Yang--Baxter

90: equation in 1967~\cite{yang67prl1312, baxter90}, and provides the

91: foundation for the field of integrable

92: models~\cite{KorepinBogoliubovIzergin93}.  Faddeev and

93: Takhtajan~\cite{faddeev-81pla375} discovered in 1981 that the

94: elementary excitations (now called spinons) of the spin-$1/2$

95: Heisenberg chain carry spin $1/2$ while the Hilbert space is spanned

96: by spin flips, which carry spin 1.  The fractional quantization of

97: spin in spin chains is conceptually similar to the fractional

98: quantization of charge in quantized Hall

99: liquids~\cite{laughlin83prl1395, stone92}.  In 1982,

100: Haldane~\cite{haldane83pla464, haldane83prl1153} identified the O(3)

101: nonlinear sigma model as the effective low-energy field theory of

102: SU(2) spin chains, and argued that chains with integer spin possess a

103: gap in the excitation spectrum, while a topological term renders

104: half-integer spin chains gapless~\cite{affleck90proc,Fradkin91}.

105: 

106: The general methods---the Bethe ansatz method and the use of effective

107: field theories including

108: bosonization~\cite{GogolinNersesyanTsvelik98,Giamarchi04}---are

109: complemented by a number of exactly solvable models, most prominently

110: among them the Majumdar--Ghosh (MG) Hamiltonian for the

111: $S=\frac{1}{2}$ dimer chain~\cite{majumdar-69jmp1399}, the AKLT model

112: as a paradigm of the gapped $S=1$ chain~\cite{affleck-87prl799,

113:   affleck-88cmp477}, and the Haldane--Shastry model

114: (HSM)~\cite{haldane88prl635, shastry88prl639, haldane91prl1529,

115:   haldane-92prl2021}.  The HSM is by far the most sophisticated among

116: these three, as it is not only solvable for the ground state, but

117: fully integrable due to its Yangian symmetry~\cite{haldane-92prl2021}.

118: The wave functions for the ground state and single-spinon excitations

119: are of a simple Jastrow form, elevating the conceptual similarity to

120: quantized Hall states to a formal equivalence.  Another unique feature

121: of the HSM is that the spinons are free in the sense that they only

122: interact through their half-Fermi statistics~\cite{haldane91prl937,

123:   ha-93prb12459, essler95prb13357, greiter-05prb224424,greiter06prl},

124: which renders the model an ideal starting point for any perturbative

125: description of spin systems in terms of interacting

126: spinons~\cite{greiter-06prl}.  The HSM has been generalized from SU(2)

127: to SU($n$)~\cite{kawakami92prb1005, kawakami92prbr3191, ha-92prb9359, 

128:   bouwknegt-96npb345, schuricht-05epl987, schuricht-06prb235105}.

129: 

130: For the MG and the AKLT model, only the ground states are known

131: exactly.  Nonetheless, these models have amply contributed to our

132: understanding of many aspects of spin chains, each of them through the

133: specific concepts captured in its ground state~\cite{%use of MG:

134:   jullien-83baps344, affleck89jpcm3047, okamoto-92pla433, eggert96prb9612, 

135:   white-96prb9862,sen-07prb104411,

136:   knabe88jsp627,fannes-89epl633,

137:   freitag-91zpb381,%received 12.02.91

138:   kluemper-91jpal955,%received 03.05.91

139:   kennedy-92prb304,%received 29.07.91

140:   kluemper-92zpb281,%received 13.02.92

141:   kluemper-93epl293,batchelor-94ijmpb3645,

142:   schollwock-96prb3304, kolezhuk-96prl5142, kolezhuk-02prb100401,

143:   normand-02prb104411,lauchli-06prb144426}.

144: The models are specific to SU(2) spin chains.  We will review both 

145: models below.

146: 

147: In the past, the motivation to study SU($n$) spin systems with $n>2$

148: has been mainly formal.  The Bethe ansatz method has been generalized

149: to multiple component systems by

150: Sutherland~\cite{sutherland75prb3795}, yielding the so-called nested

151: Bethe ansatz. In particular, this has led to a deeper understanding

152: of quantum integrability and the applicability of the Bethe

153: ansatz~\cite{choi-82pla83}.  Furthermore, the nested Bethe ansatz was

154: used to study the spectrum of the SU($n$)

155: HSM~\cite{kawakami92prb1005,ha-93prb12459}.  It has also been applied

156: to SU(2) spin chains with orbital degeneracy at the SU(4) symmetric

157: point~\cite{Li-99prb12781,Gu-02prb092404}.

158: %

159: Most recently, Damerau and Kl\"umper obtained highly accurate

160: numerical results for the thermodynamic properties of the SU(4)

161: spin--orbital model~\cite{damerau-06jsm12014}.

162: %

163: SU($n$) Heisenberg models have been studied recently by Kawashima and

164: Tanabe~\cite{kawashima-07prl057202} with quantum Monte Carlo, and by

165: Paramekanti and Marston~\cite{Paramekanti-06cm0608691} using

166: variational wave functions.

167: 

168: The effective field theory description of SU(2) spin chains by Haldane

169: yielding the distinction between gapless half-integer spin chains with

170: deconfined spinons and gapped integer spin chains with confined

171: spinons cannot be directly generalized to SU($n$), as there is no

172: direct equivalent of the CP$^1$ representation used in Haldane's

173: analysis.

174: %

175: The critical behavior of SU($n$) spin chains, however, has been

176: analyzed by Affleck in the framework of effective field

177: theories~\cite{affleck86npb409, affleck88npb582}.

178: 

179: An experimental realization of an SU(3) spin system, and in particular

180: an antiferromagnetic SU(3) spin chain, however, might be possible in

181: an optical lattice of ultracold atoms in the not-too-distant future.

182: The ``spin'' in these systems would of course not relate to the

183: rotation group of our physical space, but rather relate to SU(3)

184: rotations in an internal space spanned by three degenerate ``colors''

185: the atom may assume, subject to the requirement that the number of

186: atoms of each color is conserved.  A possible way to realize such a

187: system experimentally is described in Appendix~\ref{sec:exp}.

188: Moreover, it has been suggested recently that an SU(3) trimer state

189: might be realized approximately in a spin tetrahedron

190: chain~\cite{chen-05prb214428,chen-06prb174424}.

191: 

192: Motivated by both this prospect as well as the mathematical challenges

193: inherent to the problem, we propose several exact models for SU(3)

194: spin chains in this article.  The models are similar in spirit to the

195: MG or the AKLT model for SU(2), and consist of parent Hamiltonians and

196: their exact ground states.  There is no reason to expect any of these

197: models to be integrable, and none of the excited states are known

198: exactly.  We generalize several of our models to SU($n$), and use the

199: emerging rules to investigate and motivate which SU($n$) spin chains

200: exhibit spinon confinement and a Haldane gap.

201: 

202: The article is organized as follows.  Following a brief review of the

203: MG model in Sec.~\ref{sec:mg}, we introduce the trimer model for SU(3)

204: spin chains in Sec.~\ref{sec:trimer}.  This model consists of a

205: translationally invariant Hamiltonian involving four-site

206: interactions, with a three-fold degenerate ground state, in which

207: triples of neighboring sites form SU(3) singlets (or trimers).  In

208: Sec.~\ref{sec:su3rep}, we review the representations of SU(3), which

209: we use to verify the trimer model in Sec.~\ref{sec:trimercont}.  In

210: this section, we further provide numerical evidence that the

211: elementary excitations of this model transform under representation

212: $\bs{\bar 3}$ of SU(3) if the original spins of the model transform

213: under rep.\ $\bs{3}$.  We proceed by introducing Schwinger bosons in

214: Sec.~\ref{sec:schwinger} and a review of the AKLT model in

215: Sec.~\ref{sec:aklt}.  In Sec.~\ref{sec:vbs}, we formulate a family of

216: parent Hamiltonians for valence bond solids of SU(3) chains with spin

217: reps.\ $\bs{6}$, $\bs{10}$, and $\bs{8}$ on each lattice site, and

218: proof their validity.  We argue that only the rep.\ $\bs{10}$ and the

219: rep.\ $\bs{8}$ model, which are in a wider sense generalizations of

220: the AKLT model to SU(3), exhibit spinon confinement and hence a

221: Haldane-type gap in the excitation spectrum.  In Sec.~\ref{sec:sun},

222: we generalize three of our models from SU(3) to SU($n$).

223: %

224: In Sec.~\ref{sec:conf}, we use the rules emerging from the numerous

225: VBS models we studied to investigate which models of SU($n$) spin

226: chains in general exhibit spinon confinement and a Haldane gap.  In

227: this context, we first review a rigorous theorem due to Affleck and

228: Lieb~\cite{affleck-86lmp57} in Sec.\ \ref{sec:afflecklieb}.  In Sec.\

229: \ref{sec:generalcriterion}, we argue that the spinons in SU($n$) spin

230: chains with spins transforming under reps.\ with Young tableaux

231: consisting of a number of boxes $\lambda$ which is divisible by $n$

232: are always confined.  In Sec.\ \ref{sec:examples}, we construct

233: several specific examples to argue that if $\lambda$ and $n$ have a

234: common divisor different from $n$, the model will be confining only if

235: the interactions are sufficiently long ranged.  Specifically, the

236: models we study suggest that if $q$ is the largest common divisor of

237: $\lambda$ and $n$, the model will exhibit spinon confinement only if

238: the interactions extends at least to the $\frac{n}{q}$-th neighbor on

239: the chain.  If $\lambda$ and $n$ have no common divisor, the spinons

240: will be free and chain will not exhibit a Haldane gap.  We briefly

241: summarize the different categories of models in Sec.\ \ref{sec:sum},

242: and present a counter-example to the general rules in Sec.\

243: \ref{sec:counter}.  We conclude with a brief summary of the results 

244: obtained in this article in 

245: Sec.\ \ref{sec:conclusion}.

246: 

247: A brief and concise account of the SU(3) VBS models we elaborate here 

248: has been given previously~\cite{greiter-07prb060401}.

249: 

250: 

251: \section{The Majumdar--Ghosh model}

252: \label{sec:mg}

253: 

254: Majumdar and Ghosh~\cite{majumdar-69jmp1399} noticed in 1969 that on a

255: linear spin $S=\frac{1}{2}$ chain with an even number of sites, the

256: two valence bond solid or dimer states

257: \begin{eqnarray}

258: \big|\psi_{\text{MG}}^{\textrm{even}\rule{0pt}{5pt}\atop 

259: \textrm{(odd)}}\big\rangle 

260: &=&

261: \prod_{i\ \textrm{even}\atop (i\ \textrm{odd})}  

262: \Bigl(c^\dagger_{i\up} c^\dagger_{i+1\dw} - c^\dagger_{i\dw} c^\dagger_{i+1\up}

263: \Bigr) \vac =\nonumber \\

264: &=&\left\{ \begin{array}{lc}

265: \ket{\hbox{\begin{picture}(102,8)(-4,-3)

266: \put(13,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

267: \put(41,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

268: \put(69,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

269: \put(6,0){\circle{4}}

270: \put(20,0){\circle{4}}

271: \put(34,0){\circle{4}}

272: \put(48,0){\circle{4}}

273: \put(62,0){\circle{4}}

274: \put(76,0){\circle{4}}

275: \put(90,0){\circle{4}}

276: %\put(104,0){\circle{4}}

277: %\put(118,0){\circle{4}}

278: \end{picture}}} 

279: &\quad\textrm{``even''}\quad\\

280: %\ket{\hbox{\begin{picture}(132,8)(-4,-3)

281: \ket{\hbox{\begin{picture}(102,8)(-4,-3)

282: %\put(4,0){\makebox(0,0)[r]{\rule{4.pt}{ 0.8pt}}}

283: \put(27,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

284: \put(55,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

285: \put(83,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

286: %\put(111,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

287: \put(6,0){\circle{4}}

288: \put(20,0){\circle{4}}

289: \put(34,0){\circle{4}}

290: \put(48,0){\circle{4}}

291: \put(62,0){\circle{4}}

292: \put(76,0){\circle{4}}

293: \put(90,0){\circle{4}}

294: %\put(104,0){\circle{4}}

295: %\put(118,0){\circle{4}}

296: \end{picture}}} 

297: &\quad\textrm{``odd''}\quad\rule{0pt}{15pt}

298: \end{array}\right.\rule{0pt}{25pt}

299: \label{eq:psimg}

300: \end{eqnarray}

301: where the product runs over all even sites $i$ for one state and over

302: all odd sites for the other, are exact zero-energy ground

303: states~\cite{broek80pla261} of the parent Hamiltonian

304: \begin{equation}

305: H_{\text{MG}} = 

306: \sum_i \left(\boldsymbol{S}_i \boldsymbol{S}_{i+1} +

307: \frac{1}{2}\boldsymbol{S}_i \boldsymbol{S}_{i+2} +\frac{3}{8}\right),

308:  \label{eq:hmg}

309: \end{equation}

310: where

311: \begin{equation}

312: \boldsymbol{S}_i=\frac{1}{2}\,\sum_{\tau,\tau'=\up,\dw}\,

313: c_{i\tau}^{\dagger} \boldsymbol{\sigma}_{\tau\tau'}^{\phantom{\dagger}}

314: c_{i\tau'}^{\phantom{\dagger}}, 

315: \label{eq:s}

316: \end{equation}

317: and $\boldsymbol{\sigma}=(\sigma_x,\sigma_y,\sigma_z)$ is the vector

318: consisting of the three Pauli matrices.

319: 

320: The proof is exceedingly simple.  We rewrite

321: \begin{equation}

322: H_{\textrm{MG}}=\frac{1}{4}\sum_i H_i,\ \

323: H_i=\bigl(\boldsymbol{S}_i +\boldsymbol{S}_{i+1} 

324: +\boldsymbol{S}_{i+2}\bigr)^2 -\frac{3}{4}.

325: \end{equation}

326: Clearly, any state in which the total spin of three neighboring spins

327: is $\frac{1}{2}$ will be annihilated by $H_i$.  (The total spin can

328: only be $\frac{1}{2}$ or $\frac{3}{2}$, as

329: $\bf\frac{1}{2}\otimes\frac{1}{2}\otimes\frac{1}{2}=

330: \frac{1}{2}\oplus\frac{1}{2}\oplus\frac{3}{2}$.)  In the dimer states

331: above, this is always the case as two of the three neighboring spins

332: are in a singlet configuration, and %; the total spin is hence

333: $\bf 0\otimes\frac{1}{2}=\frac{1}{2}$.  Graphically, we may express

334: this as

335: \begin{equation}

336: H_i \ket{\hbox{\begin{picture}(40,8)(0,-3)

337: %\put(7,0){\line(1,0){12}}

338: \put(13,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

339: \put(6,0){\circle{4}}

340: \put(20,0){\circle{4}}

341: \put(34,0){\circle{4}}

342: \end{picture}}}

343: = H_i \ket{\hbox{\begin{picture}(40,8)(0,-3)

344: %\put(7,0){\line(1,0){12}}

345: \put(27,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

346: \put(6,0){\circle{4}}

347: \put(20,0){\circle{4}}

348: \put(34,0){\circle{4}}

349: \end{picture}}}=0.

350: \end{equation}

351:  As $H_i$ is positive definite, the two zero-energy eigenstates of

352: $H_{\textrm{MG}}$ are also ground states. 

353: 

354: Is the Majumdar--Ghosh or dimer state in the universality class

355: generic to one-dimensional spin-\half liquids, and hence a

356: useful paradigm to understand, say, the nearest-neighbor Heisenberg

357: chain?  The answer is clearly no, as the dimer states (\ref{eq:psimg})

358: violate translational symmetry modulo translations by two lattice

359: spacings, while the generic liquid is invariant.  

360: 

361: Nonetheless, the dimer chain shares some important properties of this

362: generic liquid.  First, the spinon excitations---here domain walls

363: between ``even'' and ``odd'' ground states---are deconfined.  (To

364: construct approximate eigenstates of

365: $H_{\textrm{MG}}$, we take momentum superpositions

366: of localized domain walls.)  Second, there are (modulo the overall

367: two-fold degeneracy) only $M+1$ orbitals available for an individual

368: spinon if $2M$ spins are condensed into dimers or valence bond

369: singlets.  This is to say, if there are only a few spinons in a long

370: chain, the number of orbitals available to them is roughly half the

371: number of sites.  This can easily be seen graphically:

372: \begin{center}

373: \begin{picture}(222,25)(-26,-17)

374: \put(-12,0){\makebox(0,0)[r]{\rule{10.pt}{ 0.8pt}}}

375: \put(13,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

376: \put(41,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

377: %\put(55,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

378: %\put(69,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

379: \put(83,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

380: %\put(97,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

381: \put(111,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

382: %\put(125,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

383: \put(139,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

384: %\put(153,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

385: \put(181,0){\makebox(0,0){\rule{10.pt}{ 0.8pt}}}

386: %\put(204,0){\makebox(0,0)[l]{\rule{7.pt}{ 0.8pt}}}

387: \put(-24,0){\circle{4}}

388: \put(-10,0){\circle{4}}

389: \put(6,0){\circle{4}}

390: \put(20,0){\circle{4}}

391: \put(34,0){\circle{4}}

392: \put(48,0){\circle{4}}

393: \put(62,0){\circle{4}}

394: \put(76,0){\circle{4}}

395: \put(90,0){\circle{4}}

396: \put(104,0){\circle{4}}

397: \put(118,0){\circle{4}}

398: \put(132,0){\circle{4}}

399: \put(146,0){\circle{4}}

400: \put(160,0){\circle{4}}

401: \put(174,0){\circle{4}}

402: \put(188,0){\circle{4}}

403: %\put(202,0){\circle{4}}

404: \put(62,1){\makebox(0,0){\vector(0,1){14}}}

405: \put(160,1){\makebox(0,0){\vector(0,1){14}}}

406: \put(14,-12){\makebox(0,0){\small even}}

407: \put(111,-12){\makebox(0,0){\small odd}}

408: \end{picture}

409: \end{center}

410: If we start with an even ground state on the left, the spinon to its

411: right must occupy an even lattice site and vice versa.  The resulting

412: state counting is precisely what one finds in the Haldane--Shastry

413: model, where it is directly linked to the half-Fermi statistics of the

414: spinons~\cite{haldane91prl937}.

415: 

416: The dimer chain is further meaningful as a piece of a general

417: paradigm.  The two degenerate dimer states (\ref{eq:psimg}) can be

418: combined into an $S=1$ chain, the AKLT chain, which serves as a

419: generic paradigm for $S=1$ chains which exhibit the Haldane

420: gap~\cite{haldane83pla464, haldane83prl1153, affleck89jpcm3047},

421: and provides the intellectual background for several of the exact models

422: we introduce further below.  Before doing so, however, we will now

423: introduce the trimer model, which constitutes an SU(3) analog of the MG 

424: model.

425: 

426: 

427: \section{The trimer model}

428: \label{sec:trimer}

429: 

430: \subsection{The Hamiltonian and its ground states} 

431: \label{sec:trimerstated}

432: 

433: Consider a chain with $N$ lattice sites, where $N$ has to be

434: divisible by three, and periodic boundary conditions (PBCs). On each

435: lattice site we place an SU(3) spin which transforms under the

436: fundamental representation $\bs{3}$, \ie the spin can take

437: the values (or colors) blue (b), red (r), or green (g). The trimer

438: states are obtained by requiring the spins on each three neighboring

439: sites to form an SU(3) singlet, which we call a trimer and sketch it

440: by

441: \begin{picture}(38,8)(1,-2)\linethickness{0.8pt}

442: \put(6,0){\circle{4}}\put(8,0){\line(1,0){10}}\put(20,0){\circle{4}}

443: \put(22,0){\line(1,0){10}}\put(34,0){\circle{4}}

444: \end{picture}.

445: The three %orthogonal 

446: linearly independent trimer %ground 

447: states on the chain are given by

448: \begin{eqnarray}\label{trimerstates.3}

449: \ket{\psi_{\text{trimer}}^{(\mu)}}

450: \!&\!=\!&\!\left\{

451: \begin{array}{l}

452: |\hbox{

453: \setlength{\unitlength}{1pt}

454: \begin{picture}(80,8)(4,-2)\linethickness{0.8pt}

455: \put(6,0){\circle{4}}\put(8,0){\line(1,0){10}}\put(20,0){\circle{4}}

456: \put(22,0){\line(1,0){10}}\put(34,0){\circle{4}}\put(48,0){\circle{4}}

457: \put(50,0){\line(1,0){10}}\put(62,0){\circle{4}}\put(64,0){\line(1,0){10}}

458: \put(76,0){\circle{4}}

459: \end{picture}

460: }\rangle\,\equiv\,\ket{\psi_{\text{trimer}}^{(1)}},\\[3mm]

461: |\hbox{

462: \setlength{\unitlength}{1pt}

463: \begin{picture}(80,8)(4,-2)\linethickness{0.8pt}

464: %\put(0,0){\line(1,0){4}}

465: \put(6,0){\circle{4}}\put(20,0){\circle{4}}

466: \put(22,0){\line(1,0){10}}\put(34,0){\circle{4}}\put(36,0){\line(1,0){10}}

467: \put(48,0){\circle{4}}\put(62,0){\circle{4}}\put(64,0){\line(1,0){10}}

468: \put(76,0){\circle{4}}

469: %\put(78,0){\line(1,0){4}}

470: \end{picture}

471: }\rangle\,\equiv\,\ket{\psi_{\text{trimer}}^{(2)}},\\[3mm]

472: |\hbox{

473: \setlength{\unitlength}{1pt}

474: \begin{picture}(80,8)(4,-2)\linethickness{0.8pt}

475: %\put(0,0){\line(1,0){4}}

476: \put(6,0){\circle{4}}\put(8,0){\line(1,0){10}}\put(20,0){\circle{4}}

477: \put(34,0){\circle{4}}\put(36,0){\line(1,0){10}}\put(48,0){\circle{4}}

478: \put(50,0){\line(1,0){10}}\put(62,0){\circle{4}}\put(76,0){\circle{4}}

479: %\put(78,0){\line(1,0){4}}

480: \end{picture}

481: }\rangle\,\equiv\,\ket{\psi_{\text{trimer}}^{(3)}}\,.

482: \end{array}\right.\\[-2mm]\nn

483: \end{eqnarray}

484: Introducing operators $c_{i\sigma}^{\dagger}$ which create a fermion

485: of color $\sigma$ ($\sigma=\b,\r,\g$) at lattice site $i$, the trimer

486: states can be written as

487: \begin{equation}

488:   \label{eq:trimer}

489:   \ket{\psi_{\text{trimer}}^{(\mu)}\!}=\hspace{-15pt}\prod_{\scriptstyle{i} \atop 

490:     \left(\scriptstyle{\frac{i-\mu}{3}\,{\rm integer}}\right)}

491:   \hspace{-5pt}\Bigl(

492:   \sum_{\scriptstyle{(\alpha,\beta,\gamma)}=\atop\scriptstyle{{\pi}(\b,\r,\g)}} 

493:   \hspace{-5pt}\hbox{sign}({\pi})\,

494:   c^{\dagger}_{i\,\alpha}\,c^{\dagger}_{i+1\,\beta}\,c^{\dagger}_{i+2\,\gamma}

495:   \Bigr)\!\vac\!,

496: \end{equation}

497: where $\mu=1,2,3$ labels the three degenerate ground states, and $i$ runs 

498: over the lattice sites subject to the constraint that $\frac{i-\mu}{3}$

499: is integer.  The sum extends over all six permutations $\pi$ of 

500: the three colors b, r, and g, \ie

501: \begin{equation}

502:   \label{eq:permsum}

503:   \begin{split}  

504:   &\sum_{\scriptstyle{(\alpha,\beta,\gamma)}=\scriptstyle{{\pi}(\b,\r,\g)}} 

505:   \hspace{-5pt}\hbox{sign}({\pi})\,

506:   c^{\dagger}_{i\,\alpha}\,c^{\dagger}_{i+1\,\beta}\,c^{\dagger}_{i+2\,\gamma}\\

507:   &\hspace{18pt}=c^{\dagger}_{i\,\b}c^{\dagger}_{i+1\,\r}c^{\dagger}_{i+2\,\g}

508:    +c^{\dagger}_{i\,\r}c^{\dagger}_{i+1\,\g}c^{\dagger}_{i+2\,\b}

509:    +c^{\dagger}_{i\,\g}c^{\dagger}_{i+1\,\b}c^{\dagger}_{i+2\,\r}\\

510:   &\hspace{18pt}-c^{\dagger}_{i\,\b}c^{\dagger}_{i+1\,\g}c^{\dagger}_{i+2\,\r}

511:    -c^{\dagger}_{i\,\g}c^{\dagger}_{i+1\,\r}c^{\dagger}_{i+2\,\b}

512:    -c^{\dagger}_{i\,\r}c^{\dagger}_{i+1\,\b}c^{\dagger}_{i+2\,\g}.\hspace{5pt}

513: \end{split}

514: \end{equation}

515: 

516: The SU(3) generators at each lattice site $i$ are in analogy to

517: \eqref{eq:s} defined as

518: \begin{equation}

519: J^a_i=\frac{1}{2}\,\sum_{\sigma,\sigma'=\b,\r,\g}\,

520: c_{i\sigma}^{\dagger} \lambda^a_{\sigma\sigma'}

521: c_{i\sigma'}^{\phantom{\dagger}}, \quad a=1,\ldots,8,

522: \label{eq:J_a}

523: \end{equation}

524: where the $\lambda^a$ are the Gell-Mann matrices (see

525: App.~\ref{app:conventions}).  The operators \eqref{eq:J_a} satisfy the

526: commutation relations

527: \begin{equation}

528: \comm{J^a_{i}}{J^b_{j}}=\delta_{ij}\;f^{abc}J^c_{i},

529: \quad a,b,c=1,\ldots,8,

530: \label{eq:su3-spincommforsites}

531: \end{equation}

532: (we use the Einstein summation convention) with $f^{abc}$ the

533: structure constants of SU(3) (see App.~\ref{app:conventions}).  We

534: further introduce the total SU(3) spin of $\nu$ neighboring sites

535: $i,\ldots,i+\nu-1$,

536: \begin{equation}\label{eq:defJnu}

537: \bs{J}^{(\nu)}_i = \sum_{j=i}^{i+\nu-1}\,\bs{J}_{j},

538: \end{equation}

539: where $\bs{J}_i$ is the eight-dimensional vector formed by its

540: components \eqref{eq:J_a}.  The parent Hamiltonian for the trimer

541: states \eqref{eq:trimer} is given by

542: \begin{equation}

543:   \label{ham.trimer}

544:   H_{\textrm{trimer}}\, =\, \sum_{i=1}^N

545:   \left(\,\left(\bs{J}^{(4)}_i\right)^{\!4}\,

546:     -\,\frac{14}{3}\left(\bs{J}^{(4)}_i\right)^{\!2}\,+

547:     \,\frac{40}{9}\,\right).

548: \end{equation}

549: The $\bs{J}_i\bs{J}_j$ terms appear complicated in terms

550: of Gell-Mann matrices, but are rather simply when written out

551: using the operator $P_{ij}$, which permutes the SU($n$) spins

552: (here $n=3$) on sites $i$ and $j$:

553: \begin{equation}

554:   \label{eq:permut}

555:   \bs{J}_i\bs{J}_j = \frac{1}{2}\left( P_{ij} - \frac{1}{n} \right).

556: \end{equation}

557: 

558: To verify the trimer Hamiltonian \eqref{ham.trimer}, as well as for

559: the valence bond solid (VBS) models we propose below, we will need a

560: few higher-dimensional representations of SU(3).  We review these in

561: the following section.

562: 

563: \section{Representations of SU(3)}

564: \label{sec:su3rep}

565: 

566: \subsection{Young tableaux and representations of SU(2) }

567: 

568: \begin{figure}[tb]

569: \setlength{\unitlength}{8pt}

570: \begin{picture}(28,10)(0,0)

571: \linethickness{0.3pt}

572: \multiput(0.5,9)(1,0){2}{\line(0,1){1}}

573: \multiput(0.5,9)(0,1){2}{\line(1,0){1}}

574: \put(0.5,9){\makebox(1,1){1}}

575: \put(1,1){\line(0,1){4}}

576: \multiput(1,2)(0,2){2}{\circle*{0.5}}

577: \put(0,5){\makebox(1,1){$S^z$}}

578: \put(1.5,3.5){\makebox(2,1){$\ket{\up}$}}

579: \put(1.5,1.5){\makebox(2,1){$\ket{\dw}$}}

580: 

581: \put(3.5,9){\makebox(1,1){$\otimes$}}

582: 

583: \multiput(6.5,9)(1,0){2}{\line(0,1){1}}

584: \multiput(6.5,9)(0,1){2}{\line(1,0){1}}

585: \put(6.5,9){\makebox(1,1){2}}

586: \put(7,1){\line(0,1){4}}

587: \multiput(7,2)(0,2){2}{\circle*{0.5}}

588: \put(6,5){\makebox(1,1){$S^z$}}

589: \put(7.5,3.5){\makebox(2,1){$\ket{\up}$}}

590: \put(7.5,1.5){\makebox(2,1){$\ket{\dw}$}}

591: 

592: \put(9,9){\makebox(1,1){=}}

593: 

594: \multiput(11.5,8)(1,0){2}{\line(0,1){2}}

595: \multiput(11.5,8)(0,1){3}{\line(1,0){1}}

596: \put(11.5,9){\makebox(1,1){1}}

597: \put(11.5,8){\makebox(1,1){2}}

598: \put(12,2){\line(0,1){2}}

599: \put(12,3){\circle*{0.5}}

600: \put(11,4){\makebox(1,1){$S^z$}}

601: \put(13.8,2.5){\makebox(5,1){$\tfrac{1}{\sqrt{2}}\bigl(

602: \ket{\up\dw}-\ket{\dw\up}\bigr)$}}

603: 

604: \put(15.75,9){\makebox(1,1){$\oplus$}}

605: 

606: \multiput(20,9)(1,0){3}{\line(0,1){1}}

607: \multiput(20,9)(0,1){2}{\line(1,0){2}}

608: \put(20,9){\makebox(1,1){1}}

609: \put(21,9){\makebox(1,1){2}}

610: \put(21,0){\line(0,1){6}}

611: \multiput(21,1)(0,2){3}{\circle*{0.5}}

612: \put(20,6){\makebox(1,1){$S^z$}}

613: \put(21.6,0.5){\makebox(2,1){$\ket{\dw\dw}$}}

614: \put(22.8,2.5){\makebox(5,1){$\tfrac{1}{\sqrt{2}}\bigl(

615: \ket{\up\dw}+\ket{\dw\up}\bigr)$}}

616: \put(21.6,4.5){\makebox(2,1){$\ket{\up\up}$}}

617: \end{picture}

618: \caption{Tensor product of two $S=\frac{1}{2}$ spins with Young

619:   tableaux and weight diagrams of the occurring SU(2) representations.

620:   $S^z$ is the diagonal generator.}

621: \label{fig:su2-1}

622: \end{figure}

623: \begin{figure}[tb]

624: \setlength{\unitlength}{8pt}

625: \begin{picture}(32,8)(2,-1)

626: \linethickness{0.3pt}

627: \put(3,6){\line(1,0){1}}

628: \put(3,5){\line(1,0){1}}

629: \put(3,5){\line(0,1){1}}

630: \put(4,5){\line(0,1){1}}

631: \put(3,5){\makebox(1,1){1}}

632: \put(4.6,5.3){$\otimes$}

633: \put(6,6){\line(1,0){1}}

634: \put(6,5){\line(1,0){1}}

635: \put(6,5){\line(0,1){1}}

636: \put(7,5){\line(0,1){1}}

637: \put(6,5){\makebox(1,1){2}}

638: \put(3,4){$\underbrace{\phantom{iiiiiiiiii}}$}

639: \put(3,2.5){\line(1,0){1}}

640: \put(3,1.5){\line(1,0){1}}

641: \put(3,0.5){\line(1,0){1}}

642: \put(3,0.5){\line(0,1){2}}

643: \put(4,0.5){\line(0,1){2}}

644: \put(3,1.5){\makebox(1,1){1}}

645: \put(3,0.5){\makebox(1,1){2}}

646: \put(2.1,-1){$S=0$}

647: \put(5,1.8){$\oplus$}

648: \put(7,2.5){\line(1,0){2}}

649: \put(7,1.5){\line(1,0){2}}

650: \put(7,1.5){\line(0,1){1}}

651: \put(8,1.5){\line(0,1){1}}

652: \put(9,1.5){\line(0,1){1}}

653: \put(7,1.5){\makebox(1,1){1}}

654: \put(8,1.5){\makebox(1,1){2}}

655: \put(6.8,-0.7){$S=1$}

656: \put(7.6,5.3){$\otimes$}

657: \put(9,6){\line(1,0){1}}

658: \put(9,5){\line(1,0){1}}

659: \put(9,5){\line(0,1){1}}

660: \put(10,5){\line(0,1){1}}

661: \put(9,5){\makebox(1,1){3}}

662: \put(11.6,5.3){$=$}

663: \put(14,6){\line(1,0){1}}

664: \put(14,5){\line(1,0){1}}

665: \put(14,4){\line(1,0){1}}

666: \put(14,3){\line(1,0){1}}

667: \put(14,3){\line(0,1){3}}

668: \put(15,3){\line(0,1){3}}

669: \put(14,5){\makebox(1,1){1}}

670: \put(14,4){\makebox(1,1){2}}

671: \put(14,3){\makebox(1,1){3}}

672: \put(16.1,5.3){$\oplus$}

673: \put(18,6){\line(1,0){2}}

674: \put(18,5){\line(1,0){2}}

675: \put(18,4){\line(1,0){1}}

676: \put(18,4){\line(0,1){2}}

677: \put(19,4){\line(0,1){2}}

678: \put(20,5){\line(0,1){1}}

679: \put(18,5){\makebox(1,1){1}}

680: \put(19,5){\makebox(1,1){3}}

681: \put(18,4){\makebox(1,1){2}}

682: \put(17.5,2.2){$S=\frac{1}{2}$}

683: \put(21.1,5.3){$\oplus$}

684: \put(23,6){\line(1,0){2}}

685: \put(23,5){\line(1,0){2}}

686: \put(23,4){\line(1,0){1}}

687: \put(23,4){\line(0,1){2}}

688: \put(24,4){\line(0,1){2}}

689: \put(25,5){\line(0,1){1}}

690: \put(23,5){\makebox(1,1){1}}

691: \put(23,4){\makebox(1,1){3}}

692: \put(24,5){\makebox(1,1){2}}

693: \put(22.5,2.2){$S=\frac{1}{2}$}

694: \put(26.1,5.3){$\oplus$}

695: \put(28,6){\line(1,0){3}}

696: \put(28,5){\line(1,0){3}}

697: \put(28,5){\line(0,1){1}}

698: \put(29,5){\line(0,1){1}}

699: \put(30,5){\line(0,1){1}}

700: \put(31,5){\line(0,1){1}}

701: \put(28,5){\makebox(1,1){1}}

702: \put(29,5){\makebox(1,1){2}}

703: \put(30,5){\makebox(1,1){3}}

704: \put(28,2.2){$S=\frac{3}{2}$}

705: \put(14.5,4.5){\makebox(0,0){\line(1,2){1.8}}}

706: \put(14.5,4.5){\makebox(0,0){\line(1,-2){1.8}}}

707: \end{picture}

708: \caption{Tensor product of three $S=\frac{1}{2}$ spins with Young

709:   tableaux.  For SU($n$) with $n>2$, the tableau with three boxes on

710:   top of each other exists as well.}

711: \label{fig:su2-2}

712: \end{figure}

713: Let us begin with a review of Young tableaux and the representations

714: of SU(2).  The group SU(2) has three generators $S^a$, $a=1,2,3$,

715: which obey the algebra

716: \begin{equation}

717: \comm{S^a}{S^b}=i\epsilon^{abc}S^c,

718: \label{eq:su2algebra}

719: \end{equation}

720: where repeated indices are summed over and $\epsilon^{abc}$ is the

721: totally antisymmetric tensor.  The representations of SU(2) are

722: classified by the spin $S$, which takes integer or half-integer

723: values.  The fundamental representation of SU(2) has spin

724: $S=\frac{1}{2}$, it contains the two states $\ket{\up}$ and

725: $\ket{\dw}$.  Higher-dimensional representations can be constructed as

726: tensor products of fundamental representations, which is conveniently

727: accomplished using Young tableaux (see

728: \emph{e.g.}~\cite{InuiTanabeOnodera96}).  These tableaux are

729: constructed as follows (see Figs.~\ref{fig:su2-1} and~\ref{fig:su2-2}

730: for examples).  For each of the $N$ spins, draw a box numbered

731: consecutively from left to right.  The representations of SU(2) are

732: obtained by putting the boxes together such that the numbers assigned

733: to them increase in each row from left to right and in each column

734: from top to bottom.  Each tableau indicates symmetrization over all

735: boxes in the same row, and antisymmetrization over all boxes in the

736: same column.  This implies that we cannot have more than two boxes on

737: top of each other.  If $\kappa_i$ denotes the number of boxes in the

738: $i$th row, the spin is given by $S=\frac{1}{2}(\kappa_1-\kappa_2)$.

739: 

740: To be more explicit, let us consider the tensor product

741: $\bf\frac{1}{2}\otimes\frac{1}{2}\otimes\frac{1}{2}$ depicted in

742: Fig.~\ref{fig:su2-2} in detail.  We start with the state

743: $\ket{\tfrac{3}{2},\tfrac{3}{2}}=\ket{\up\up\up}$, and hence find

744: \begin{equation}

745: \ket{\tfrac{3}{2},\tfrac{1}{2}}

746: =\frac{1}{\sqrt{3}}\,S^-\ket{\tfrac{3}{2},\tfrac{3}{2}}=

747: \frac{1}{\sqrt{3}}\bigl(

748: \ket{\up\up\dw}+\ket{\up\dw\up}+\ket{\dw\up\up}\bigr).

749: \label{eq:s=3/2state}

750: \end{equation}

751: The two states with $S=S^z=\frac{1}{2}$ must be orthogonal to

752: \eqref{eq:s=3/2state}.  A convenient choice of basis is

753: \begin{equation}

754: \label{chiralbasis}

755: \begin{split}

756: \ket{\tfrac{1}{2},\tfrac{1}{2},+}&=

757: \frac{1}{\sqrt{3}}\bigl(

758: \ket{\up\up\dw}+\omega\ket{\up\dw\up}+\omega^2\ket{\dw\up\up}\bigr),\\[2mm]

759: \ket{\tfrac{1}{2},\tfrac{1}{2},-}&=

760: \frac{1}{\sqrt{3}}\bigl(

761: \ket{\up\up\dw}+\omega^2\ket{\up\dw\up}+\omega\ket{\dw\up\up}\bigr),

762: \end{split}

763: \end{equation}

764: where $\omega=\exp\bigl(i\frac{2\pi}{3}\bigr)$. The tableaux tell us

765: primarily that two such basis states exist, not what a convenient choice

766: of orthonormal basis states may be.

767: 

768: The irreducible representations of SU(2) can be classified through the

769: eigenvalues of the Casimir operator given by the square of the total

770: spin $\bs{S}^2$.  The special feature of $\bs{S}^2$ is that it

771: commutes with all generators $S^a$ and is hence by Schur's

772: lemma~\cite{Cornwell84vol2} proportional to the identity for any

773: finite-dimensional irreducible representation.

774: The eigenvalues are given by

775: \begin{displaymath}

776:   \bs{S}^2=\mathcal{C}^2_{\text{SU}(2)}=S(S+1).

777: \end{displaymath}

778: 

779: \begin{figure}[t]

780: \includegraphics[scale=0.18]{vbs3.eps}\hspace{5mm}

781: \includegraphics[scale=0.18]{vbs3bar.eps}

782: \caption{(Color online) a) Weight diagram of the fundamental SU(3)

783:   representation $\bs{3}=(1,0)$. b) Weight diagram of the complex conjugate

784:   representation $\bs{\bar{3}}=(0,1)$.

785:   %(reproduced with permission from Ref.~\cite{SchurichtGreiter05colepl}). 

786:   $J^3$ and $J^8$ denote the diagonal generators, $I^+$, $U^+$, and

787:   $V^+$ the raising operators. }

788: \label{fig:33barweight}

789: \end{figure}

790: \begin{figure}[tb]

791: \setlength{\unitlength}{8pt}

792: \begin{picture}(32,8)(2,-1)

793: \linethickness{0.5pt}

794: \put(3,6){\line(1,0){1}}

795: \put(3,5){\line(1,0){1}}

796: \put(3,5){\line(0,1){1}}

797: \put(4,5){\line(0,1){1}}

798: \put(3,5){\makebox(1,1){1}}

799: \put(4.6,5.3){$\otimes$}

800: \put(6,6){\line(1,0){1}}

801: \put(6,5){\line(1,0){1}}

802: \put(6,5){\line(0,1){1}}

803: \put(7,5){\line(0,1){1}}

804: \put(6,5){\makebox(1,1){2}}

805: \put(3,4.4){$\underbrace{\phantom{iiiiiiiiii}}$}

806: \put(3,2.9){\line(1,0){1}}

807: \put(3,1.9){\line(1,0){1}}

808: \put(3,0.9){\line(1,0){1}}

809: \put(3,0.9){\line(0,1){2}}

810: \put(4,0.9){\line(0,1){2}}

811: \put(3,1.9){\makebox(1,1){1}}

812: \put(3,0.9){\makebox(1,1){2}}

813: \put(3,-1){\makebox(1,1){$\bs{\bar{3}}$}}

814: \put(4.6,1.5){$\oplus$}

815: \put(6,2.4){\line(1,0){2}}

816: \put(6,1.4){\line(1,0){2}}

817: \put(6,1.4){\line(0,1){1}}

818: \put(7,1.4){\line(0,1){1}}

819: \put(8,1.4){\line(0,1){1}}

820: \put(6,1.4){\makebox(1,1){1}}

821: \put(7,1.4){\makebox(1,1){2}}

822: \put(6,-1){\makebox(2,1){$\bs{6}$}}

823: \put(7.6,5.3){$\otimes$}

824: \put(9,6){\line(1,0){1}}

825: \put(9,5){\line(1,0){1}}

826: \put(9,5){\line(0,1){1}}

827: \put(10,5){\line(0,1){1}}

828: \put(9,5){\makebox(1,1){3}}

829: \put(11.6,5.3){$=$}

830: \put(14,6){\line(1,0){1}}

831: \put(14,5){\line(1,0){1}}

832: \put(14,4){\line(1,0){1}}

833: \put(14,3){\line(1,0){1}}

834: \put(14,3){\line(0,1){3}}

835: \put(15,3){\line(0,1){3}}

836: \put(14,5){\makebox(1,1){1}}

837: \put(14,4){\makebox(1,1){2}}

838: \put(14,3){\makebox(1,1){3}}

839: \put(14,1){\makebox(1,1){$\bs{1}$}}

840: \put(16.1,5.3){$\oplus$}

841: \put(18,6){\line(1,0){2}}

842: \put(18,5){\line(1,0){2}}

843: \put(18,4){\line(1,0){1}}

844: \put(18,4){\line(0,1){2}}

845: \put(19,4){\line(0,1){2}}

846: \put(20,5){\line(0,1){1}}

847: \put(18,5){\makebox(1,1){1}}

848: \put(19,5){\makebox(1,1){3}}

849: \put(18,4){\makebox(1,1){2}}

850: \put(18,1){\makebox(2,1){$\bs{8}$}}

851: \put(21.1,5.3){$\oplus$}

852: \put(23,6){\line(1,0){2}}

853: \put(23,5){\line(1,0){2}}

854: \put(23,4){\line(1,0){1}}

855: \put(23,4){\line(0,1){2}}

856: \put(24,4){\line(0,1){2}}

857: \put(25,5){\line(0,1){1}}

858: \put(23,5){\makebox(1,1){1}}

859: \put(23,4){\makebox(1,1){3}}

860: \put(24,5){\makebox(1,1){2}}

861: \put(23,1){\makebox(2,1){$\bs{8}$}}

862: \put(26.1,5.3){$\oplus$}

863: \put(28,6){\line(1,0){3}}

864: \put(28,5){\line(1,0){3}}

865: \put(28,5){\line(0,1){1}}

866: \put(29,5){\line(0,1){1}}

867: \put(30,5){\line(0,1){1}}

868: \put(31,5){\line(0,1){1}}

869: \put(28,5){\makebox(1,1){1}}

870: \put(29,5){\makebox(1,1){2}}

871: \put(30,5){\makebox(1,1){3}}

872: \put(28,1){\makebox(3,1){$\bs{10}$}}

873: \end{picture}

874: \caption{Tensor product $\bs{3}\otimes\bs{3}\otimes\bs{3}$ with Young

875:   tableaux.}

876: \label{fig:youngdiagram}

877: \end{figure}

878: 

879: \subsection{Representation theory of SU(3) }

880: 

881: The group SU(3) has eight generators $J^a$, $a=1,\ldots,8$, which

882: obey the algebra

883: \begin{equation}

884: \comm{J^a}{J^b}=f^{abc}J^c,

885: \label{eq:su3algebra}

886: \end{equation}

887: where the structure constants $f^{abc}$ are given in

888: App.~\ref{app:conventions}.  For SU(3) we have two diagonal

889: generators, usually chosen to be $J^3$ and $J^8$, and six generators

890: which define the ladder operators $I^\pm=J^1\pm iJ^2$, $U^\pm=J^6\pm

891: iJ^7$, and $V^\pm=J^4\pm iJ^5$, respectively.  An explicit realization

892: of \eqref{eq:su3algebra} is, for example, given by the $J^a$'s as

893: expressed in terms of Gell-Mann matrices in \eqref{eq:J_a}.  This

894: realization defines the fundamental representation $\bs{3}$ of SU(3)

895: illustrated in Fig.~\ref{fig:33barweight}a.  It is three-dimensional,

896: and we have chosen to label the basis states by the colors blue (b),

897: red (r), and green (g).  The weight diagram depicted in

898: Fig.~\ref{fig:33barweight}a instructs us about the eigenvalues of the

899: diagonal generators as well as the actions of the ladder operators on

900: the basis states.

901: 

902: \begin{figure}[tb]

903: \setlength{\unitlength}{8pt}

904: \begin{picture}(16,4)(0,0)

905: \linethickness{0.3pt}

906: \multiput(0,4)(0,1){2}{\line(1,0){16}}

907: \put(0,3){\line(1,0){11}}

908: \put(0,2){\line(1,0){5}}

909: \multiput(0,2)(1,0){2}{\line(0,1){3}}

910: \multiput(4,2)(1,0){2}{\line(0,1){3}}

911: \put(6,3){\line(0,1){2}}

912: \multiput(10,3)(1,0){2}{\line(0,1){2}}

913: \put(12,4){\line(0,1){1}}

914: \multiput(15,4)(1,0){2}{\line(0,1){1}}

915: \put(11,0.7){\makebox(5,1){$\underbrace{\hspace{38.8pt}}$}}

916: \put(5,0.7){\makebox(6,1){$\underbrace{\hspace{46.7pt}}$}}

917: \put(11,-0.4){\makebox(5,1){$\mu_1\, $boxes}}

918: \put(5,-0.4){\makebox(6,1){$\mu_2\, $columns}}

919: \end{picture}

920: \caption{Dynkin coordinates $(\mu_1,\mu_2)$ for a given Young tableau.

921:   The columns containing three boxes represent additional SU(3)

922:   singlet factors, which yield equivalent representations and hence

923:   leave the Dynkin coordinates $(\mu_1,\mu_2)$ unchanged.}

924: \label{fig:dynkin}

925: \end{figure}

926: 

927: All other representations of SU(3) can be constructed by taking tensor

928: products of reps.\ $\bs{3}$, which is again most conveniently

929: accomplished using Young tableaux (see Fig.~\ref{fig:youngdiagram} for

930: an example).  The antisymmetrization over all boxes in the same column

931: implies that we cannot have more than three boxes on top of each

932: other.  Each tableaux stands for an irreducible representation of

933: SU(3), which can be uniquely labeled by their highest weight or Dynkin

934: coordinates $(\mu_1,\mu_2)$~\cite{Cornwell84vol2,Georgi82}

935: (see Fig.~\ref{fig:dynkin}).  For example, the fundamental

936: representation $\bs{3}$ has Dynkin coordinates (1,0).  Note that all

937: columns containing three boxes are superfluous, as the

938: antisymmetrization of three colors yields only one state.  In

939: particular, the SU(3) singlet has Dynkin coordinates (0,0).  In

940: general, the dimension of a representation $(\mu_1,\mu)$ is given by

941: $\frac{1}{2}(\mu_1+1)(\mu_2+1)(\mu_1+\mu_2+2)$. The labeling using

942: bold numbers refers to the dimensions of the representations alone.

943: Although this labeling is not unique, it will mostly be sufficient for

944: our purposes.  A representation $\bs{m}$ and its conjugated

945: counterpart $\bs{\overline{m}}$ are related to each other by

946: interchange of their Dynkin coordinates.

947: 

948: \subsection{Examples of representations of SU(3)}

949: 

950: We now consider some specific representations of SU(3) in detail.  As

951: starting point we use the fundamental representation $\bs{3}$ spanned

952: by the states $\ket{\b}$, $\ket{\r}$, and $\ket{\g}$.  The second

953: three-dimensional representation $\bs{\bar{3}}$ is obtained by

954: antisymmetrically coupling two reps.\ $\bs{3}$. The Dynkin coordinates

955: of the rep.\ $\bs{\bar{3}}$ are (0,1), \ie the reps.\ $\bs{3}$ and

956: $\bs{\bar{3}}$ are complex conjugate of each other.  An explicit

957: basis of the rep.\ $\bs{\bar{3}}$ is given by the colors yellow (y),

958: cyan (c), and magenta (m),

959: \begin{eqnarray}

960: \ket{\y}\!\!&=&\!\!\frac{1}{\sqrt{2}}\bigl(\ket{\r\g}-\ket{\g\r}\bigr),

961: \nonumber\\

962: \ket{\c}\!\!&=&\!\!\frac{1}{\sqrt{2}}\bigl(\ket{\g\b}-\ket{\b\g}\bigr),\\

963: \ket{\m}\!\!&=&\!\!\frac{1}{\sqrt{2}}\bigl(\ket{\b\r}-\ket{\r\b}\bigr).

964: \nonumber

965: \end{eqnarray}

966: The weight diagram is shown in Fig.~\ref{fig:33barweight}.b.  The

967: generators are given by \eqref{eq:J_a} with $\lambda^a$ replaced by

968: $-(\lambda^a)^*$, where $^*$ denotes complex conjugation of the matrix

969: elements~\cite{Georgi82}. In particular, we find

970: $I^+\ket{\y}=-\ket{\c}$, $U^+\ket{\c}=-\ket{\m}$, and

971: $V^+\ket{\y}=-\ket{\m}$.

972: 

973: \begin{figure}[t]

974: \setlength{\unitlength}{0.4pt}

975: \begin{picture}(320,320)(0,0)

976: \linethickness{0.3pt}

977: \put(0,166){\line(1,0){300}}

978: \put(150,0){\line(0,1){294}}

979: \multiput(50,224)(100,0){3}{\circle*{10}}

980: \multiput(100,137)(100,0){2}{\circle*{10}}

981: \put(150,50){\circle*{10}}

982: \put(320,300){\large\bf 6}

983: \put(155,290){$J^8$}

984: \put(295,150){$J^3$}

985: \put(262,225){$\ket{\b\b}$}

986: \put(5,225){$\ket{\r\r}$}

987: \put(68,245){$\tfrac{1}{\sqrt{2}}\bigl(\ket{\b\r}+\ket{\r\b}\bigr)$}

988: \put(210,115){$\tfrac{1}{\sqrt{2}}\bigl(\ket{\b\g}+\ket{\g\b}\bigr)$}

989: \put(-55,115){$\tfrac{1}{\sqrt{2}}\bigl(\ket{\r\g}+\ket{\g\r}\bigr)$}

990: \put(160,35){$\ket{\g\g}$}

991: \end{picture}

992: \caption{Weight diagram of the representation $\bs{6}=(2,0)$.

993:   The weight diagram of the conjugate representation

994:   $\bs{\bar{6}}=(0,2)$ is obtained by reflection at the

995:   origin~\cite{Cornwell84vol2}.}

996: \label{fig:6weight}

997: \end{figure}

998: 

999: The six-dimensional representation $\bs{6}$ has Dynkin coordinates

1000: (2,0), and can hence be constructed by symmetrically coupling two

1001: reps.\ $\bs{3}$. The basis states of the rep.\ $\bs{6}$ are shown in

1002: Fig.~\ref{fig:6weight}.  The conjugate representation $\bs{\bar{6}}$

1003: can be constructed by symmetrically coupling two reps.\

1004: $\bs{\bar{3}}$.

1005: 

1006: \begin{figure}[t]

1007: \setlength{\unitlength}{0.4pt}

1008: \begin{picture}(460,300)(0,0)

1009: \linethickness{0.3pt}

1010: \put(0,137){\line(1,0){300}}

1011: \put(150,0){\line(0,1){274}}

1012: \multiput(100,224)(100,0){2}{\circle*{10}}

1013: \multiput(50,137)(100,0){3}{\circle*{10}}

1014: \put(150,137){\circle{20}}

1015: \multiput(100,50)(100,0){2}{\circle*{10}}

1016: \put(160,130){\line(3,-2){120}}

1017: \put(320,280){\large\bf 8}

1018: \put(155,270){$J^8$}

1019: \put(295,120){$J^3$}

1020: \put(210,235){$\ket{\b\m}$}

1021: \put(50,235){$\ket{\r\m}$}

1022: \put(263,158){\makebox(3,1){$\ket{\b\c}$}}

1023: \put(37,158){\makebox(3,1){$\ket{\r\y}$}}

1024: \put(210,30){$\ket{\g\c}$}

1025: \put(55,30){$\ket{\g\y}$}

1026: \put(285,63){$\tfrac{1}{\sqrt{2}}\bigl(\ket{\b\y}-\ket{\r\c}\bigr)$}

1027: \put(285,28){$\tfrac{1}{\sqrt{6}}\bigl(\ket{\b\y}+\ket{\r\c}-

1028: 2\ket{\g\m}\bigr)$}

1029: \end{picture}

1030: \caption{Weight diagram of the adjoint representation $\bs{8}=(1,1)$.

1031:   The state with $J^3=J^8=0$ is doubly

1032:   degenerate~\cite{Cornwell84vol2}.  Note that two reps.\ $\bs{8}$ can

1033:   be constructed by combining three fundamental reps.\ $\bs{3}$

1034:   (colors), just as two reps.\ $\bs{\frac{1}{2}}$ can be constructed

1035:   by combining three SU(2) spins (cf. \eqref{chiralbasis}).  The

1036:   states in the diagram span a basis for one of these

1037:   representations.}

1038: \label{fig:8weight}

1039: \end{figure}

1040: 

1041: Let us now consider the tensor product

1042: $\bs{3}\otimes\bs{\bar{3}}=\bs{1}\oplus\bs{8}$.

1043: % included in Fig.~\ref{fig:youngdiagram}.  

1044: The weight diagram of the so-called adjoint representation

1045: $\bs{8}=(1,1)$ is shown in Fig.~\ref{fig:8weight}. The states can be

1046: constructed starting from the highest weight state $\ket{\b\m}$,

1047: yielding $I^-\ket{\b\m}=\ket{\r\m}$, $U^-\ket{\b\m}=-\ket{\b\c}$,

1048: $V^-\ket{\b\m}=\ket{\g\m}-\ket{\b\y}$, and so on. This procedure

1049: yields two linearly independent states with $J^3=J^8=0$.  The

1050: representation $\bs{8}$ can also be obtained by coupling of the reps.\

1051: $\bs{6}$ and $\bs{3}$, as can be seen from the Young tableaux in

1052: Fig.~\ref{fig:youngdiagram}.  On a more abstract level, the adjoint

1053: representation is the representation we obtain if we consider the

1054: generators $J^a$ themselves basis vectors.

1055: In the weight diagram shown in Fig.~\ref{fig:8weight}, the generators

1056: $J^3$ and $J^8$ correspond to the two states at the origin, whereas

1057: the ladder operators $I^\pm$, $U^\pm$, and $V^\pm$ correspond to the

1058: states at the six surrounding points.  In the notation of

1059: Fig.~\ref{fig:8weight}, the singlet orthogonal to $\bs{8}$ is given by

1060: $\tfrac{1}{\sqrt{3}}\bigl(\ket{\b\y}+\ket{\r\c}+\ket{\g\m}\bigr)$.

1061: 

1062: The weight diagrams of four other representations relevant to our

1063: purposes below are shown in Figs.~\ref{fig:10weight}

1064: to~\ref{fig:27weight}.

1065: 

1066: \begin{figure}[t]

1067: \setlength{\unitlength}{0.3pt}

1068: \begin{picture}(420,390)(50,0)

1069: \linethickness{0.3pt}

1070: \put(0,224){\line(1,0){400}}

1071: \put(200,0){\line(0,1){364}}

1072: \multiput(50,311)(100,0){4}{\circle*{13}}

1073: \multiput(100,224)(100,0){3}{\circle*{13}}

1074: \multiput(150,137)(100,0){2}{\circle*{13}}

1075: \put(200,50){\circle*{13}}

1076: \put(420,367){\large\bf 10}

1077: \put(205,357){$J^8$}

1078: \put(395,205){$J^3$}

1079: \put(355,280){$\ket{\b\b\b}$}

1080: \put(-5,280){$\ket{\r\r\r}$}

1081: \put(260,115){$\tfrac{1}{\sqrt{3}}\bigl(\ket{\b\g\g}+\ket{\g\b\g}+

1082: \ket{\g\g\b}\bigr)$}

1083: \put(210,30){$\ket{\g\g\g}$}

1084: \end{picture}

1085: \caption{Weight diagram of the representation $\bs{10}=(3,0)$.

1086:   The weight diagram of the conjugate representation

1087:   $\bs{\overline{10}}=(0,3)$ is obtained by reflection at the

1088:   origin~\cite{Cornwell84vol2}.}

1089:   \label{fig:10weight}

1090: \end{figure}

1091: 

1092: \begin{figure}[t]

1093: \setlength{\unitlength}{0.18pt}

1094: \begin{picture}(550,470)(0,-87)

1095: \linethickness{0.3pt}

1096: \put(0,195){\line(1,0){400}}

1097: \put(200,0){\line(0,1){361}}

1098: \multiput(100,311)(100,0){3}{\circle*{15}}

1099: \multiput(50,224)(100,0){4}{\circle*{15}}

1100: \multiput(150,224)(100,0){2}{\circle{30}}

1101: \multiput(100,137)(100,0){3}{\circle*{15}}

1102: \put(200,137){\circle{30}}

1103: \multiput(150,50)(100,0){2}{\circle*{15}}

1104: \put(420,371){\large\bf 15}

1105: \end{picture}

1106: \begin{picture}(550,550)(-50,0)

1107: \linethickness{0.3pt}

1108: \put(0,282){\line(1,0){500}}

1109: \put(250,0){\line(0,1){448}}

1110: \multiput(50,398)(100,0){5}{\circle*{15}}

1111: \multiput(100,311)(100,0){4}{\circle*{15}}

1112: \multiput(150,224)(100,0){3}{\circle*{15}}

1113: \multiput(200,137)(100,0){2}{\circle*{15}}

1114: \put(250,50){\circle*{15}}

1115: \put(520,454){\large\bf 15'}

1116: \end{picture}

1117: \caption{Weight diagram of the representations $\bs{15}=(2,1)$ and

1118:   $\bs{15'}=(4,0)$.  

1119:   }

1120:   \label{fig:15weight}

1121: \end{figure}

1122: 

1123: \begin{figure}[t]

1124: \setlength{\unitlength}{0.18pt}

1125: \begin{picture}(550,460)(50,0)

1126: \linethickness{0.3pt}

1127: \put(0,224){\line(1,0){500}}

1128: \put(250,0){\line(0,1){448}}

1129: \multiput(150,398)(100,0){3}{\circle*{15}}

1130: \multiput(100,311)(100,0){4}{\circle*{15}}

1131: \multiput(50,224)(100,0){5}{\circle*{15}}

1132: \multiput(100,137)(100,0){4}{\circle*{15}}

1133: \multiput(150,50)(100,0){3}{\circle*{15}}

1134: \multiput(200,311)(100,0){2}{\circle{30}}

1135: \multiput(150,224)(100,0){3}{\circle{30}}

1136: \multiput(200,137)(100,0){2}{\circle{30}}

1137: \put(250,224){\circle{40}}

1138: \put(520,440){\large\bf 27}

1139: \end{picture}

1140: \caption{Weight diagram of the self-conjugate representation

1141:   $\bs{27}=(2,2)$. The state with $J^3=J^8=0$ is three-fold

1142:   degenerate~\cite{Cornwell84vol2}.}

1143:   \label{fig:27weight}

1144: \end{figure}

1145: 

1146: It is known that the physical properties of SU(2) spin chains

1147: crucially depend on whether on the lattice sites are integer or

1148: half-integer spins.  A similar distinction can be made for SU(3)

1149: chains, as elaborated in Sec.~\ref{sec:conf}.  The distinction integer

1150: or half-integer spin for SU(2) is replaced by a distinction between

1151: three families of irreducible representations of SU(3): either the

1152: number of boxes in the Young tableau is divisible by three without

1153: remainder (\eg $\bs{1}$, $\bs{8}$, $\bs{10}$, $\bs{27}$), 

1154: %, $\bs{28}$, $\bs{35}$), 

1155: with remainder one (\eg $\bs{3}$, $\bs{\bar 6}$, $\bs{15}$, $\bs{15'}$),

1156: %, $\bs{\overline{21}}$, $\bs{\overline{24}}$), or

1157: or with remainder two 

1158: (\eg $\bs{\bar 3}$, $\bs{6}$, $\bs{\overline{15}}$, $\bs{\overline{15}'}$). 

1159: %$\bs{21}$, $\bs{24}$).  

1160: 

1161: While SU(2) has only one Casimir operator, SU(3) has two.  The

1162: quadratic Casimir operator is defined as

1163: \begin{equation}

1164: \bs{J}^2=\sum_{a=1}^8 J^aJ^a,

1165: \label{eq:casimir}

1166: \end{equation}

1167: where the $J^a$'s are the generators of the representation. As

1168: $\bs{J}^2$ commutes with all generators $J^a$ it is proportional to

1169: the identity for any finite-dimensional irreducible representation.

1170: The eigenvalue in a representation with Dynkin

1171: coordinates $(\mu_1,\mu_2)$ is~\cite{Cornwell84vol2}

1172: \begin{equation}

1173: \bs{J}^2=\mathcal{C}^2_{\text{SU($3$)}}(\mu_1,\mu_2)=

1174: \frac{1}{3}\bigl(\mu_1^2 + \mu_1\mu_2 + \mu_2^2 + 3\mu_1 + 3\mu_2\bigr).

1175: \label{eq:casimirev}

1176: \end{equation}

1177: We have chosen the normalization in \eqref{eq:casimir} according to

1178: the convention

1179: \begin{displaymath}

1180:   \mathcal{C}^2_{\text{SU($n$)}}(\text{adjoint representation})=n,

1181: \end{displaymath}

1182: which yields $\mathcal{C}^2_{\text{SU($3$)}}(1,1)=3$ for the

1183: representation $\bs{8}$.  Note that the quadratic Casimir operator

1184: cannot be used to distinguish between a representation and its

1185: conjugate.  This distinction would require the cubic Casimir 

1186: operator~\cite{Cornwell84vol2}, which we will not need for any of

1187: the models we propose below.

1188: 

1189: 

1190: \section{The trimer model (continued)}

1191: \label{sec:trimercont}

1192: \addtocounter{subsection}{1}

1193: 

1194: \subsection{Verification of the model}

1195: \label{sec:trimerverified}

1196: 

1197: We will now proceed with the verification of the trimer Hamiltonian

1198: \eqref{ham.trimer}.  Since the spins on the individual sites transform

1199: under the fundamental representation $\bs{3}$, the SU(3) content of

1200: four sites is

1201: \begin{equation}\label{rep3power4}

1202: \mb{3}\otimes\mb{3}\otimes\mb{3}\otimes\mb{3}\,=\,

1203: 3\cdot\mb{3}\,\oplus\,2\cdot\mb{\bar 6}\,\oplus\,3\cdot\mb{15}\,

1204: \oplus\,\mb{15'},

1205: \end{equation}

1206: \ie we obtain representations $\bs{3}$, $\bs{\bar 6}$, and two

1207: non-equivalent 15-dimensional representations with Dynkin coordinates

1208: $(2,1)$ and $(4,0)$, respectively.  All these representations can be

1209: distinguished by their eigenvalues of the quadratic Casimir operator,

1210: which is given by $\left(\bs{J}^{(4)}_i\right)^2$ if the four spins

1211: reside on the four neighboring lattice sites \hbox{$i,\ldots,i+3$.}

1212: 

1213: For the trimer states \eqref{trimerstates.3}, the situation

1214: simplifies as we only have the two possibilities 

1215: \begin{displaymath}

1216: \begin{array}{rcl}

1217: &

1218: \begin{picture}(55,8)(0,-2)\linethickness{0.8pt}

1219: \put(8,0){\line(1,0){10}}\put(6,0){\circle{4}}\put(20,0){\circle{4}}

1220: \put(22,0){\line(1,0){10}}\put(34,0){\circle{4}}\put(48,0){\circle{4}}

1221: %\put(50,0){\line(1,0){4}}

1222: \end{picture}

1223: ~&~\hat =~~~\mb{1}\,\otimes\,\mb{3}~=~\mb{3},\label{eq:one}\\[1mm]

1224: &

1225: \begin{picture}(55,8)(0,-2)\linethickness{0.8pt}

1226: %\put(0,0){\line(1,0){4}}

1227: \put(8,0){\line(1,0){10}}\put(6,0){\circle{4}}

1228: \put(20,0){\circle{4}}\put(36,0){\line(1,0){10}}\put(34,0){\circle{4}}

1229: \put(48,0){\circle{4}}

1230: %\put(50,0){\line(1,0){4}}

1231: \end{picture}~&~

1232: \hat =~~~\mb{\bar 3} \,\otimes\,\mb{\bar 3}~=~\mb{3}\,\oplus\,\mb{\bar 6},

1233: \end{array}

1234: \end{displaymath}

1235: which implies that the total SU(3) spin on four neighboring sites can

1236: only transform under representations $\mb{3}$ or $\mb{\bar{6}}$. The

1237: eigenvalues of the quadratic Casimir operator for these

1238: representations are $4/3$ and $10/3$, respectively.  The auxiliary

1239: operators

1240: \begin{equation}

1241:   \label{eq:auxop}

1242:   H_i=\left(\Bigl(\bs{J}^{(4)}_i\Bigr)^2-\frac{4}{3}\right)

1243:   \left(\Bigl(\bs{J}^{(4)}_i\Bigr)^2-\frac{10}{3}\right)

1244: \end{equation} 

1245: hence annihilate the trimer states for all values of $i$, while they

1246: yield positive eigenvalues for $\bs{15}$ or $\bs{15'}$, \ie all other

1247: states.  Summing $H_i$ over all lattice sites $i$ yields

1248: \eqref{ham.trimer}.  We have numerically confirmed by exact

1249: diagonalization of \eqref{ham.trimer} for chains with $N=9$ and 12

1250: lattice sites that the three states \eqref{eq:trimer} are the only

1251: ground states.

1252: 

1253: Note that the representation content of five neighboring sites in 

1254: the trimer chains is just the conjugate of the above, as

1255: \begin{displaymath}

1256: \begin{array}{rcl}

1257: &

1258: \begin{picture}(65,8)(0,-2)\linethickness{0.8pt}

1259: \put(8,0){\line(1,0){10}}\put(6,0){\circle{4}}\put(20,0){\circle{4}}

1260: \put(22,0){\line(1,0){10}}\put(34,0){\circle{4}}\put(48,0){\circle{4}}

1261: \put(50,0){\line(1,0){10}}\put(62,0){\circle{4}}

1262: %\put(64,0){\line(1,0){4}}

1263: \end{picture}

1264: ~&~\hat =~~\mb{1}\,\otimes\,\mb{\bar 3}~=~\mb{\bar 3}\,,\\[1mm]

1265: &

1266: \begin{picture}(65,8)(0,-2)\linethickness{0,8pt}

1267: %\put(0,0){\line(1,0){4}}

1268: \put(6,0){\circle{4}}\put(20,0){\circle{4}}

1269: \put(36,0){\line(1,0){10}}\put(34,0){\circle{4}}\put(48,0){\circle{4}}

1270: \put(22,0){\line(1,0){10}}\put(62,0){\circle{4}}

1271: %\put(64,0){\line(1,0){4}}

1272: \end{picture}

1273: ~&~\hat =~~\mb{3}\,\otimes\,\mb{1}\,\otimes\,\mb{3}~=~\mb{\bar 3}\,\oplus\,

1274: \mb{6}.\hspace{12pt}

1275: \end{array}

1276: \end{displaymath}

1277: Since the quadratic Casimirs of conjugate representations have

1278: identical eigenvalues, $\mathcal{C}^2_{\text{SU($3$)}}(\mu_1,\mu_2)=

1279: \mathcal{C}^2_{\text{SU($3$)}}(\mu_2,\mu_1)$, we can construct another

1280: parent Hamiltonian for the trimer states \eqref{eq:trimer} by simply

1281: replacing $\bs{J}^{(4)}_i$ with $\bs{J}^{(5)}_i$ in

1282: \eqref{ham.trimer}.  This Hamiltonian will have a different spectrum.

1283: In comparison to the four-site interaction Hamiltonian

1284: \eqref{ham.trimer}, however, it is more complicated while bearing no

1285: advantages.  We will not consider it further.  

1286: 

1287: \subsection{Elementary excitations}

1288: \label{sec:trimerexcitations}

1289: 

1290: \begin{figure}[t]

1291: \begin{minipage}[b]{.45\linewidth}

1292: \includegraphics[width=\linewidth]{loc_exc_3.eps}

1293: \\ (a)

1294: \end{minipage}\hfill

1295: \begin{minipage}[b]{.495\linewidth}

1296: \includegraphics[width=\linewidth]{loc_exc_bar3.eps}

1297: \\ (b)

1298: \end{minipage}

1299: \caption{Couplings used in the numerical studies to create (a) the

1300:   localized rep.\ $\bs{3}$ trial state and (b) the localized rep.\

1301:   $\bs{\bar{3}}$ trial state.}

1302: \label{fig:loc}

1303: \end{figure}

1304: 

1305: Let us now turn to the low-lying excitations of \eqref{ham.trimer}.

1306: In analogy with the MG model, it is evident that the SU(3) spinon

1307: or ``coloron'' excitations correspond to domain walls between the

1308: degenerate ground states.  For the trimer model, however, there

1309: are two different kinds of domain walls, as illustrated by:

1310: \begin{eqnarray}

1311: \setlength{\unitlength}{2pt}

1312: &\begin{picture}(90,12)(4,-2)

1313: \linethickness{0.8pt}

1314: \put(6,0){\circle{4}}

1315: \put(8,0){\line(1,0){10}}

1316: \put(20,0){\circle{4}}

1317: \put(22,0){\line(1,0){10}}

1318: \put(34,0){\circle{4}}

1319: \put(48,0){\circle{4}}

1320: \put(47.5,8){\makebox(1,1){$\mb{3}$}}

1321: \put(62,0){\circle{4}}

1322: \put(64,0){\line(1,0){10}}

1323: \put(76,0){\circle{4}}

1324: \put(78,0){\line(1,0){10}}

1325: \put(90,0){\circle{4}}

1326: \end{picture}&

1327: \label{eq:trial3}\\[3mm]

1328: &\begin{picture}(104,12)(4,-2)

1329: \linethickness{0.8pt}

1330: \put(6,0){\circle{4}}

1331: \put(8,0){\line(1,0){10}}

1332: \put(20,0){\circle{4}}

1333: \put(22,0){\line(1,0){10}}

1334: \put(34,0){\circle{4}}

1335: \put(48,0){\circle{4}}

1336: \put(50,0){\line(1,0){10}}

1337: \put(54.5,8){\makebox(1,1){$\mb{\bar 3}$}}

1338: \put(62,0){\circle{4}}

1339: \put(76,0){\circle{4}}

1340: \put(78,0){\line(1,0){10}}

1341: \put(90,0){\circle{4}}

1342: \put(92,0){\line(1,0){10}}

1343: \put(104,0){\circle{4}}

1344: \end{picture}&

1345: \label{eq:trial3bar}

1346: \end{eqnarray}

1347: The first domain wall \eqref{eq:trial3} connects ground state $\mu$ to

1348: the left to ground state $\mu+1$ to the right, where $\mu$ is defined

1349: modulo 3 (see \eqref{eq:trimer}), and consists of an individual SU(3)

1350: spin, which transforms under representation $\bs{3}$.  The second

1351: domain wall \eqref{eq:trial3bar} connects ground state $\mu$ with 

1352: ground state $\mu+2$.  It consists of two antisymmetrically coupled

1353: spins on two neighboring sites, and hence transforms under

1354: representation $\bs{\bar{3}}$.  As we take momentum superpositions of

1355: the localized domain walls illustrated above, we expect one of them,

1356: but not both, to constitute an approximate eigenstate of the trimer

1357: model.  The reason we do not expect both of them to yield a valid

1358: excitation is that they can decay into each other, \ie if the rep.\

1359: $\bs{3}$ excitation is valid the rep.\ $\bs{\bar{3}}$ domain wall

1360: would decay into two rep.\ $\bs{3}$ excitations, and vice versa.  The

1361: question which of the two excitations is the valid one, \ie whether

1362: the elementary excitations transform under $\bs{3}$ or $\bs{\bar{3}}$

1363: under SU(3) rotations, can be resolved through numerical studies.  We

1364: will discuss the results of these studies now.

1365: 

1366: \begin{table}[t]

1367: \begin{center}

1368: \begin{tabular}{c@{\hspace{10pt}}|@{\hspace{10pt}}c@{\hspace{10pt}}c@{\hspace{10pt}}c@{\hspace{10pt}}c}

1369: \hline\hline

1370: mom&$E_\text{tot}$&&\%&~over-~\\

1371: $[2\pi/N]$&exact&trial&~~off~~&lap\\

1372: \hline

1373: 0    &2.9735& 4.5860&54.2&0.9221\\

1374: 1, 12&6.0345&10.2804&70.4&0.5845\\

1375: 2, 11&9.0164&17.2991&91.9&0.0\\

1376: 3, 10&6.6863&13.1536&96.7&0.0\\

1377: 4, 9 &3.0896& 5.0529&63.5&0.8864\\

1378: 5, 8 &4.8744& 7.5033&53.9&0.8625\\

1379: 6, 7 &8.5618&16.6841&94.9&0.1095\\

1380: \hline\hline

1381: \end{tabular}

1382: \end{center}

1383: \caption{Energies of the rep.\ $\bs{3}$ trial states \eqref{eq:trial3}

1384:   in comparison to the 

1385:   exact excitation energies of the trimer model \eqref{ham.trimer} and

1386:   their overlaps for an SU(3) spin chain with $N=13$ sites.}

1387: \label{tab:rep3}

1388: \end{table}

1389: 

1390: \begin{figure}[t]

1391: \begin{center}

1392: \includegraphics[scale=0.35,angle=270]{13prb.eps}

1393: \caption{Dispersion of the rep.\ $\bs{3}$ trial states

1394:   \eqref{eq:trial3} in comparison to the exact excitation energies of

1395:   \eqref{ham.trimer} for a chain with $N=13$.  The lines are a guide

1396:   to the eye.}

1397: \label{fig:spectrum3}

1398: \end{center}

1399: \end{figure}

1400: 

1401: The rep.\ $\bs{3}$ and the rep.\ $\bs{\bar{3}}$ trial states require

1402: chains with \mbox{$N=3\cdot\text{integer}+1$} and

1403: \mbox{$N=3\cdot\text{integer}+2$} sites, respectively; we chose $N=13$

1404: and $N=14$ for our numerical studies.  To create the localized domain

1405: walls \eqref{eq:trial3} and \eqref{eq:trial3bar}, we numerically

1406: diagonalized auxiliary Hamiltonians with appropriate couplings, as

1407: illustrated in Fig.~\ref{fig:loc}.  From these localized excitations,

1408: we constructed momentum eigenstates by superposition, and compared them

1409: to the exact eigenstates of our model Hamiltonian \eqref{ham.trimer}

1410: for chains with the same number of sites.  The results are shown in

1411: Tab.~\ref{tab:rep3} and Fig.~\ref{fig:spectrum3} for the rep.\

1412: $\bs{3}$ trial state, and in Tab.~\ref{tab:rep3bar} and

1413: Fig.~\ref{fig:spectrum3bar} for the rep.\ $\bs{\bar 3}$ trial state.

1414: 

1415: \begin{table}[t]

1416: \begin{center}

1417: \begin{tabular}{c@{\hspace{10pt}}|@{\hspace{10pt}}c@{\hspace{10pt}}c@{\hspace{10pt}}c@{\hspace{10pt}}c}

1418: \hline\hline

1419: mom&$E_\text{tot}$&&\%&~over-~\\

1420: $[2\pi/N]$&exact&trial&~~off~~&lap\\

1421: \hline

1422: 0   &2.1013&2.3077&9.8&0.9953\\

1423: 1, 13&4.3677&4.8683&11.5&0.9864\\

1424: 2, 12&7.7322&8.7072&12.6&0.9716\\

1425: 3, 11&6.8964&7.7858&12.9&0.9696\\

1426: 4, 10&3.2244&3.5415&9.8&0.9934\\

1427: 5, 9 &2.2494&2.4690&9.7&0.9950\\

1428: 6, 8 &5.4903&6.1016&11.1&0.9827\\

1429: 7   &7.4965&8.5714&14.3&0.9562\\

1430: \hline\hline

1431: \end{tabular}

1432: \caption{Energies of the rep.\ $\bs{\bar 3}$ trial states \eqref{eq:trial3bar}

1433:   in comparison to the 

1434:   exact excitation energies of the trimer model \eqref{ham.trimer} and

1435:   their overlaps for an SU(3) spin chain with $N=14$ sites.}

1436: \label{tab:rep3bar}

1437: \end{center}

1438: \end{table}

1439: 

1440: \begin{figure}[t]

1441: \begin{center}

1442: \includegraphics[scale=0.35,angle=270]{14prb.eps}

1443: \caption{Dispersion of the rep.\ $\bs{\bar{3}}$ trial states

1444:   \eqref{eq:trial3bar} in comparison to the exact excitation energies

1445:   of \eqref{ham.trimer} for a chain with $N=14$.  The lines are a

1446:   guide to the eye.}

1447: \label{fig:spectrum3bar}

1448: \end{center}

1449: \end{figure}

1450: 

1451: The numerical results clearly indicate that the rep.\ $\bs{\bar 3}$

1452: trial states \eqref{eq:trial3bar} are valid approximations to the

1453: elementary excitations of the trimer chain, while the rep.\ $\bs{3}$

1454: trial states \eqref{eq:trial3} are not.  We deduce that the elementary

1455: excitations of the trimer chain \eqref{ham.trimer} transform under

1456: $\bs{\bar{3}}$, that is, under the representation conjugated to the

1457: original SU(3) spins localized at the sites of the chain.  Using the

1458: language of colors, one may say that if a basis for the original spins

1459: is spanned by blue, red, and green, a basis for the excitations is

1460: spanned by the complementary colors yellow, cyan, and magenta.  This

1461: result appears to be a general feature of SU(3) spin chains, as it was

1462: recently shown explicitly to hold for the Haldane--Shastry model as

1463: well~\cite{ bouwknegt-96npb345, schuricht-05epl987, schuricht-06prb235105}.

1464: 

1465: Note that the elementary excitations of the trimer chain are

1466: deconfined, meaning that the energy of two localized representation

1467: $\bs{\bar{3}}$ domain walls or colorons %spinons

1468: \eqref{eq:trial3bar} does not depend on the distance between them.

1469: The reason is simply that domain walls connect one ground state with

1470: another, without introducing costly correlations in the region between

1471: the domain walls.  In the case of the MG model and the trimer model

1472: introduced here, however, there is still an energy gap associated with

1473: the creation of each coloron, %spinon,

1474: which is simply the energy cost associated with the domain wall.

1475: 

1476: In most of the remainder of this article, we will introduce a family

1477: of exactly soluble valence bond models for SU(3) chains of various

1478: spin representations of the SU(3) spins at each lattice site.  To

1479: formulate these models, we will first review Schwinger bosons for

1480: both SU(2) and SU(3) and the AKLT model.

1481: 

1482: 

1483: \section{Schwinger Bosons}

1484: \label{sec:schwinger}

1485: 

1486: Schwinger bosons~\cite{schwinger65proc,Auerbach94} constitute a way

1487: to formulate spin-$S$ representations of an SU(2) algebra.  The spin

1488: operators

1489: \begin{equation}

1490: \begin{array}{rcccl}

1491: S^x + iS^y &=& S^+ &=& a^\dagger b, \\\rule{0pt}{12pt} 

1492: S^x - iS^y &=& S^- &=& b^\dagger a, \\\rule{0pt}{12pt}

1493: && S^z  &=& \frac{1}{2}(a^\dagger a - b^\dagger b), 

1494: \label{eq:schw}

1495: \end{array}

1496: \end{equation}

1497: are given in terms of boson creation and annihilation

1498: %where $a^\dagger,b^\dagger$ (and $a,b$) are boson creation (and annihilation)

1499: operators which obey the usual commutation relations

1500: \begin{equation}

1501: \begin{array}{c}

1502: \comm{a}{a^\dagger}=\comm{b}{b^\dagger}=1, \\

1503: \comm{a^{\phantom{\dagger}}\!\!}{b}=\comm{a}{b^\dagger}

1504: =\comm{a^\dagger}{b}=\comm{a^\dagger}{b^\dagger}=0\rule{0pt}{16pt}.

1505: \end{array}

1506: \label{eq:schwb}

1507: \end{equation}

1508: It is readily verified with (\ref{eq:schwb}) that $S^x$, $S^y$, and

1509: $S^z$ satisfy \eqref{eq:su2algebra}.

1510: The spin quantum number $S$ is given by half the number of bosons,  

1511: \begin{equation}

1512: 2S=a^\dagger a + b^\dagger b,

1513: \label{eq:schwt}

1514: \end{equation}

1515: and the usual spin states (simultaneous eigenstates of $\boldsymbol{S}^2$

1516: and $S^z$) are given by

1517: \begin{equation}

1518: \ket{S,m} = \frac{(a^\dagger)^{S+m}}{\sqrt{(S+m)!}} \frac{(b^\dagger)^{S-m}}

1519: {\sqrt{(S-m)!}} \vac.

1520: \label{eq:schs}

1521: \end{equation}

1522: In particular, the spin-\half states are given by

1523: \begin{equation}

1524: \ket{\up}=c_{\up}^\dagger \vac =a^\dagger \vac ,\qquad 

1525: \ket{\dw}=c_{\dw}^\dagger \vac =b^\dagger \vac ,

1526: \label{eq:schwfun}

1527: \end{equation}

1528: \ie $a^\dagger$ and $b^\dagger$ act just like the fermion creation

1529: operators $c^\dagger_\up$ and $c^\dagger_\dw$ in this case.  The

1530: difference shows up only when two (or more) creation operators act on

1531: the same site or orbital.  The fermion operators create an

1532: antisymmetric or singlet configuration (in accordance with the Pauli

1533: principle), 

1534: \begin{equation}

1535: \ket{0,0} = c_{\up}^\dagger c_{\dw}^\dagger \vac ,

1536: \label{eq:schwfer}

1537: \end{equation}

1538: while the Schwinger bosons create a totally symmetric or

1539: triplet (or higher spin if we create more than two bosons) configuration,

1540: \begin{eqnarray}

1541: \ket{1,1} &=& \textstyle{\frac{1}{\sqrt{2}}} (a^\dagger)^{2}\vac 

1542: ,\nonumber\\[1mm]

1543: \ket{1,0} &=& a^\dagger b^\dagger \vac , \\[1mm]

1544: \ket{1,-1} &=& \textstyle{\frac{1}{\sqrt{2}}} (b^\dagger)^{2}\vac 

1545: .\nonumber

1546: \end{eqnarray}

1547: 

1548: The generalization to SU($n$) proceeds without incident.  We content

1549: ourselves here by writing the formalism out explicitly for SU(3).  In

1550: analogy to \eqref{eq:schw}, we write the SU(3) spin operators

1551: \eqref{eq:J_a} 

1552: \begin{equation}

1553: \begin{array}{rcccl}

1554: J^1 + iJ^2 &=& I^+ &=& b^\dagger r, \\\rule{0pt}{12pt} 

1555: J^1 - iJ^2 &=& I^- &=& r^\dagger b, \\\rule{0pt}{12pt} 

1556: &&            J^3 &=& \frac{1}{2}(b^\dagger b - r^\dagger r),\\\rule{0pt}{12pt}

1557: J^4 + iJ^5 &=& V^+ &=& b^\dagger g, \\\rule{0pt}{12pt} 

1558: J^4 - iJ^5 &=& V^- &=& g^\dagger b, \\\rule{0pt}{12pt} 

1559: J^6 + iJ^7 &=& U^+ &=& r^\dagger g, \\\rule{0pt}{12pt} 

1560: J^6 - iJ^7 &=& U^- &=& g^\dagger r, \\\rule{0pt}{12pt} 

1561: &&             J^8 &=& 

1562: \frac{1}{2\sqrt{3}}(b^\dagger b + r^\dagger r - 2g^\dagger g),

1563: \label{eq:schwsu3}

1564: \end{array}

1565: \end{equation}

1566: in terms of the boson annihilation and creation operators 

1567: $b,b^\dagger$ (blue), $r,r^\dagger$ (red), and $g,g^\dagger$ (green)

1568: satisfying

1569: \begin{equation}

1570: %\begin{array}{c}

1571: \comm{b}{b^\dagger}=\comm{r}{r^\dagger}=\comm{g}{g^\dagger}=1 %, \\

1572: %\text{all other commutators}=0.\rule{0pt}{16pt}

1573: %\end{array}

1574: \label{eq:schwbsu3}

1575: \end{equation}

1576: while all other commutators vanish.  Again, it is readily verified

1577: with \eqref{eq:schwbsu3} that the operators $J^a$ satisfy

1578: \eqref{eq:su3algebra}.  The basis states spanning the fundamental

1579: representation $\bs{3}$ may in analogy to \eqref{eq:schwfun} be

1580: written using either fermion or boson creation operators:

1581: \begin{equation}

1582: \begin{array}{c}

1583: \ket{\b}=c_{\b}^\dagger \vac=b^\dagger \vac ,\\\rule{0pt}{12pt} 

1584: \ket{\r}=c_{\r}^\dagger \vac=r^\dagger \vac ,\\\rule{0pt}{12pt}

1585: \ket{\g}=c_{\g}^\dagger \vac=g^\dagger \vac .

1586: \end{array}

1587: \label{eq:schwfunsu3}

1588: \end{equation}

1589: We write this abbreviated

1590: \begin{equation}

1591: \begin{array}{rcccccccl}

1592: \bs{3}&=&(1,0)&=&

1593: \setlength{\unitlength}{8pt}

1594: \begin{picture}(1,1)(0,0.1)

1595: \put(0,1){\line(1,0){1}}

1596: \put(0,0){\line(1,0){1}}

1597: \put(0,0){\line(0,1){1}}

1598: \put(1,0){\line(0,1){1}}

1599: \end{picture}

1600: &\hat =&c_{\alpha}^\dagger\vac &=& \alpha^\dagger \vac. %\nonumber

1601: \end{array}

1602: \end{equation}

1603: The fermion operators can be used to combine spins transforming under

1604: the fundamental representation $\bs{3}$ antisymmetrically, and hence 

1605: to construct the representations 

1606: %labeled by Young tableaux in which the boxes are arranged in a

1607: %vertical column, in particular

1608: \begin{equation}

1609: \begin{array}{rcccccl}

1610: \bs{\bar 3}&=&(0,1)&=&

1611: \setlength{\unitlength}{8pt}

1612: \begin{picture}(1,2)(0,0.5)

1613: \put(0,2){\line(1,0){1}}

1614: \put(0,1){\line(1,0){1}}

1615: \put(0,0){\line(1,0){1}}

1616: \put(0,0){\line(0,1){2}}

1617: \put(1,0){\line(0,1){2}}

1618: \end{picture}

1619: &\hat =&c_{\alpha}^\dagger c_{\beta}^\dagger \vac,\\\rule{0pt}{12pt}

1620: \bs{1}&=&(0,0)&=&

1621: \setlength{\unitlength}{8pt}

1622: \begin{picture}(1,3)(0,1)

1623: \put(0,3){\line(1,0){1}}

1624: \put(0,2){\line(1,0){1}}

1625: \put(0,1){\line(1,0){1}}

1626: \put(0,0){\line(1,0){1}}

1627: \put(0,0){\line(0,1){3}}

1628: \put(1,0){\line(0,1){3}}

1629: \end{picture}

1630: &\hat =&c_{\b}^\dagger c_{\r}^\dagger c_{\g}^\dagger \vac.

1631: \end{array}

1632: \vspace{6pt}

1633: \label{eq:schwketsu3}

1634: \end{equation}

1635: The Schwinger bosons, by contrast, combine fundamental representations

1636: $\bs{3}$ symmetrically, and hence yield representations labeled by

1637: Young tableaux in which the boxes are arranged in a horizontal row,

1638: like

1639: \begin{equation}

1640: \begin{array}{rcccccl}

1641: \bs{6}&=&(2,0)&=&

1642: \setlength{\unitlength}{8pt}

1643: \begin{picture}(2,1)(0,0.1)

1644: \put(0,1){\line(1,0){2}}

1645: \put(0,0){\line(1,0){2}}

1646: \put(0,0){\line(0,1){1}}

1647: \put(1,0){\line(0,1){1}}

1648: \put(2,0){\line(0,1){1}}

1649: \end{picture}

1650: &\hat =&\alpha^\dagger \beta^\dagger \vac,\\\rule{0pt}{14pt}

1651: \bs{10}&=&(3,0)&=&

1652: \setlength{\unitlength}{8pt}

1653: \begin{picture}(3,1)(0,0.1)

1654: \put(0,1){\line(1,0){3}}

1655: \put(0,0){\line(1,0){3}}

1656: \put(0,0){\line(0,1){1}}

1657: \put(1,0){\line(0,1){1}}

1658: \put(2,0){\line(0,1){1}}

1659: \put(3,0){\line(0,1){1}}

1660: \end{picture}

1661: &\hat =&\alpha^\dagger \beta^\dagger \gamma^\dagger \vac,\\\rule{0pt}{14pt}

1662: \bs{15'}&=&(4,0)&=&

1663: \setlength{\unitlength}{8pt}

1664: \begin{picture}(4,1)(0,0.1)

1665: \put(0,1){\line(1,0){4}}

1666: \put(0,0){\line(1,0){4}}

1667: \put(0,0){\line(0,1){1}}

1668: \put(1,0){\line(0,1){1}}

1669: \put(2,0){\line(0,1){1}}

1670: \put(3,0){\line(0,1){1}}

1671: \put(4,0){\line(0,1){1}}

1672: \end{picture}

1673: &\hat =&\alpha^\dagger \beta^\dagger \gamma^\dagger \delta^\dagger \vac, 

1674: %\ \text{\sl etc.},

1675: \end{array}

1676: \label{eq:schwbossu3}

1677: \end{equation}

1678: where $\alpha, \beta, \gamma, \ldots \in \{b,r,g\}$.

1679: Unfortunately, it is not possible to construct representations

1680: like 

1681: \begin{displaymath}

1682: \bs{8}\ =\ (1,1)\ =\

1683: \setlength{\unitlength}{8pt}

1684: \begin{picture}(2,1.2)(0,0.7)

1685: \put(0,2){\line(1,0){2}}

1686: \put(0,1){\line(1,0){2}}

1687: \put(0,0){\line(1,0){1}}

1688: \put(0,0){\line(0,1){2}}

1689: \put(1,0){\line(0,1){2}}

1690: \put(2,1){\line(0,1){1}}

1691: \end{picture}

1692: \vspace{2pt}

1693: \end{displaymath}

1694: by simply taking products of anti-commuting or commuting creation 

1695: or annihilation operators.

1696:  

1697: 

1698: \section{The AKLT model}

1699: \label{sec:aklt}

1700: 

1701: Using the SU(2) Schwinger bosons introduced in the previous section, we

1702: may rewrite the Majumdar--Ghosh states \eqref{eq:psimg} as

1703: \begin{equation}

1704: \big|\psi_{%\scriptscriptstyle

1705: \text{MG}}^{\textrm{even}\rule{0pt}{5pt}\atop 

1706: \textrm{(odd)}} \big\rangle \, =

1707: \underbrace{\displaystyle

1708: \prod_{i\ \textrm{even}\atop (i\ \textrm{odd})}  

1709: \Bigl(a^\dagger_{i} b^\dagger_{i+1} - b^\dagger_{i} a^\dagger_{i+1}\Bigr)

1710: }_{\displaystyle

1711: \equiv\Psi_{%\scriptscriptstyle

1712: \text{MG}}^{\textrm{even}\rule{0pt}{5pt}\atop 

1713: \textrm{(odd)}} \big[a^\dagger,b^\dagger\big]

1714: } \,\vac

1715: \label{eq:mgschw}

1716: \end{equation}

1717: This formulation was used by Affleck, Kennedy, Lieb, and

1718: Tasaki~\cite{affleck-88cmp477, affleck-87prl799} to propose a family

1719: of states for higher spin representations of SU(2).  In particular, they 

1720: showed that the valance bond solid (VBS) state

1721: \begin{eqnarray}

1722: \ket{\psi_{%\scriptscriptstyle

1723: \text{AKLT}}} \, &=&

1724: \prod_i\,\Bigl(a^\dagger_{i} b^\dagger_{i+1} - b^\dagger_{i} a^\dagger_{i+1}\Bigr)

1725: \,\vac =\nonumber\\

1726: &=&

1727: \Psi_{%\scriptscriptstyle

1728: \textrm{MG}}^{\scriptscriptstyle\textrm{even}}

1729: \big[a^\dagger,b^\dagger\big] \,\cdot\, 

1730: \Psi_{%\scriptscriptstyle

1731: \textrm{MG}}^{\scriptscriptstyle\textrm{odd}}

1732: \big[a^\dagger,b^\dagger\big] \,\vac =\nonumber\\ 

1733: \label{eq:psiaklt}

1734: &=&

1735: \setlength{\unitlength}{1pt}

1736: \Big|

1737: \begin{picture}(128,20)(-7,4)

1738: \linethickness{0.8pt}

1739: \multiput(0,10)(14,0){9}{\circle{4}}

1740: \multiput(0,2)(14,0){9}{\circle{4}}

1741: \thicklines

1742: \multiput(2,10)(28,0){4}{\line(1,0){10}}

1743: \multiput(16,2)(28,0){4}{\line(1,0){10}}

1744: %\put(-6,2){\line(1,0){4}}

1745: %\put(114,10){\line(1,0){4}}

1746: \thinlines

1747: \put(37,-3){\line(1,0){10}}

1748: \put(37,-3){\line(0,1){18}}

1749: \put(37,15){\line(1,0){10}}

1750: \put(47,-3){\line(0,1){18}}

1751: \put(42,-3){\line(0,-1){9}}

1752: \put(42,-20){\makebox(0,0)[c]{\small projection onto spin $S=1$}}

1753: \end{picture}\Big\rangle\\[5mm]

1754: & &\nonumber

1755: \end{eqnarray}

1756: is the exact zero-energy ground state of the spin-1 extended 

1757: Heisenberg Hamiltonian

1758: \begin{equation}

1759: H_{\text{AKLT}} = 

1760: \sum_i \left(\boldsymbol{S}_i \boldsymbol{S}_{i+1} +

1761: \frac{1}{3}\bigl(\boldsymbol{S}_i \boldsymbol{S}_{i+1}\bigr)^2 

1762: +\frac{2}{3}\right) 

1763: \label{eq:haklt}

1764: \end{equation}

1765: with periodic boundary conditions.  Each term in the sum

1766: (\ref{eq:haklt}) projects onto the subspace in which the total spin of

1767: a pair of neighboring sites is $S=2$.  The Hamiltonian

1768: (\ref{eq:haklt}) thereby lifts all states except (\ref{eq:psiaklt}) to

1769: positive energies.  The VBS %valance bond solid

1770: state (\ref{eq:psiaklt}) is a generic paradigm as it shares all the

1771: symmetries, but in particular the Haldane spin

1772: gap~\cite{haldane83pla464, haldane83prl1153, affleck89jpcm3047}, of

1773: the spin-1 Heisenberg chain.  It even offers a particularly simple

1774: understanding of this gap, or of the linear confinement potential

1775: between spinons responsible for it, as illustrated by the cartoon:

1776: %(see Sec.~\ref{sec:10VBS}):

1777: \begin{center}

1778: \setlength{\unitlength}{1pt}

1779: \begin{picture}(196,45)(0,0)

1780: \linethickness{0.8pt}

1781: \multiput(0,35)(14,0){14}{\circle{4}}

1782: \multiput(14,25)(14,0){14}{\circle{4}}

1783: \thicklines

1784: \multiput(2,35)(28,0){2}{\line(1,0){10}}

1785: \multiput(72,35)(28,0){2}{\line(1,0){10}}

1786: \multiput(142,35)(28,0){2}{\line(1,0){10}}

1787: \multiput(16,25)(28,0){6}{\line(1,0){10}}

1788: \put(184,25){\line(1,0){10}}

1789: %\put(-6,25){\line(1,0){4}}

1790: \thinlines

1791: \put(56,36){\makebox(0,0){\vector(0,1){14}}}

1792: \put(126,36){\makebox(0,0){\vector(0,1){14}}}

1793: \put(91,11){\vector(-1,0){35}}

1794: \put(91,11){\vector(1,0){35}}

1795: \put(91,0){\makebox(0,0){\small energy cost $\propto$ distance}}

1796: \end{picture}\\[5mm]

1797: \end{center}

1798: Our understanding~\cite{greiter02prb134443,greiter02prb054505} of the

1799: connection between the confinement force and the Haldane gap is that

1800: the confinement effectively imposes an oscillator potential for the

1801: relative motion of the spinons.  We then interpret the zero-point

1802: energy of this oscillator as the Haldane gap in the excitation

1803: spectrum.%, as illustrated in Fig.\,\ref{fig:osci}.

1804: 

1805: The AKLT state can also be written as a matrix

1806: product~\cite{kluemper-91jpal955,kluemper-92zpb281,kluemper-93epl293}.

1807: We first rewrite the valence bonds

1808: \begin{displaymath}

1809: \Bigl(a^\dagger_{i} b^\dagger_{i+1} - b^\dagger_{i} a^\dagger_{i+1}\Bigr)=

1810: \Bigl(a^\dagger_{i}, b^\dagger_{i}\Bigr)

1811: \left(\!\begin{array}{c} b^\dagger_{i+1}\\-a^\dagger_{i+1}\end{array}\!\right),

1812: \end{displaymath}

1813: and then use the outer product to combine the two vectors at each site

1814: into a matrix

1815: %\vspace{2pt}

1816: \begin{equation}

1817:   \label{eq:akltmatrix}

1818:   \begin{array}{rcl}

1819:     M_i &\equiv& 

1820:   \left(\!\begin{array}{c} b^\dagger_{i}\\-a^\dagger_{i}\end{array}\!\right)

1821:   \Bigl(a^\dagger_{i}, b^\dagger_{i}\Bigr) \;\vac_i 

1822: %  \\ &=&\left(\!\begin{array}{cc} 

1823: %      a^\dagger_{i} b^\dagger_{i} & {b^\dagger_{i}}^2\\

1824: %      -{a^\dagger_{i}}^2 & a^\dagger_{i} b^\dagger_{i}

1825: %    \end{array}\!\right)\vac_i 

1826:   \\[18pt] &=&\left(\!\begin{array}{cc} 

1827:        \ket{1,0}_i& \sqrt{2}\ket{1,-1}_i\\[4pt]

1828:       -\sqrt{2}\ket{1,1}_i &-\ket{1,0}_i 

1829:     \end{array}\!\right).\\

1830:   \end{array}

1831: \end{equation}

1832: %\vspace{2pt}

1833: %where the index $i$ on the states denotes the site.  

1834: Assuming PBCs, \eqref{eq:psiaklt} may then be written as the trace of

1835: the matrix product

1836: \begin{equation}

1837:   \label{eq:akltm}

1838:   \ket{\psi_{\text{AKLT}}}=\text{tr}\biggl( \prod_i M_i \biggr).

1839: \end{equation}

1840: 

1841: In the following section, we will propose several exact models of

1842: VBSs for SU(3).

1843: 

1844: 

1845: \section{SU(3) valence bond solids}

1846: \label{sec:vbs}

1847: 

1848: To begin with, we use SU(3) Schwinger bosons

1849: introduced in Sec.~\ref{sec:schwinger} to rewrite the trimer states

1850: \eqref{eq:trimer} as

1851: \begin{eqnarray}

1852:   \ket{\psi_\text{trimer}^{(\mu)}\!}

1853:   &=&\hspace{-5pt}\prod_{\scriptstyle{i} \atop 

1854:     \left(\scriptstyle{\frac{i-\mu}{3}\,{\rm integer}}\right)}

1855:   \hspace{-5pt}\Bigl(

1856:   \sum_{\scriptstyle{(\alpha,\beta,\gamma)}=\atop\scriptstyle{{\pi}(b,r,g)}} 

1857:   \hspace{-5pt}\hbox{sign}({\pi})\,

1858:   \alpha^{\dagger}_{i}\,\beta^{\dagger}_{i+1}\gamma^{\dagger}_{i+2}

1859:   \Bigr)\vac \nonumber \\[2pt]

1860:   &\equiv&\,\Psi^{\mu}\!\left[

1861:     b^{\dagger},r^{\dagger},g^{\dagger}\right]\,\vac,

1862:   \label{eq:trimerschw}

1863: \end{eqnarray}

1864: where, as in \eqref{eq:trimer}, $\mu=1,2,3$ labels the three

1865: degenerate ground states, $i$ runs over the lattice sites subject to

1866: the constraint that $\frac{i-\mu}{3}$ is integer, and the sum extends

1867: over all six permutations $\pi$ of the three colors b, r, and g.

1868: This formulation can be used directly to construct VBSs for SU(3) 

1869: spin chains with spins transforming under representations $\bs{6}$ 

1870: and $\bs{10}$ on each site.  

1871: 

1872: \subsection{The representation $\bs{6}$ VBS}

1873: 

1874: We obtain a representations $\bs{6}$ VBS from two trimer states by

1875: projecting the tensor product of two fundamental representations

1876: $\bs{3}$ onto the symmetric subspace, \ie onto the $\bs{6}$ in the

1877: decomposition $\bs{3}\,\otimes\,\bs{3}=\bs{\bar{3}}\,\oplus\,\bs{6}$.

1878: Graphically, this is illustrated as follows:

1879: \begin{equation}

1880: \setlength{\unitlength}{1pt}

1881: \begin{picture}(128,48)(4,-22)

1882: \linethickness{0.8pt}

1883: \multiput(0,10)(14,0){9}{\circle{4}}

1884: \multiput(14,0)(14,0){9}{\circle{4}}

1885: \thicklines

1886: \put(2,10){\line(1,0){10}}

1887: \put(16,10){\line(1,0){10}}

1888: \put(44,10){\line(1,0){10}}

1889: \put(58,10){\line(1,0){10}}

1890: \put(86,10){\line(1,0){10}}

1891: \put(100,10){\line(1,0){10}}

1892: \put(16,0){\line(1,0){10}}

1893: \put(30,0){\line(1,0){10}}

1894: \put(58,0){\line(1,0){10}}

1895: \put(72,0){\line(1,0){10}}

1896: \put(100,0){\line(1,0){10}}

1897: \put(114,0){\line(1,0){10}}

1898: \thinlines

1899: % \put(65,-7){\line(1,0){10}}

1900: % \put(65,-7){\line(0,1){24}}

1901: % \put(65,17){\line(1,0){10}}

1902: % \put(75,-7){\line(0,1){24}}

1903: % \put(70,-15){\line(0,1){8}}

1904: \put(65,-5){\framebox(10,20)}

1905: \put(70,-15){\line(0,1){10}}

1906: \put(70,-22){\makebox(1,1){\small projection onto rep.\ $\bs{6}=(2,0)$}}

1907: \put(70,25){\makebox(1,1){\small one site}}

1908: \end{picture}

1909: \label{fig:6state}

1910: \end{equation}

1911: This construction yields three linearly independent $\mb{6}$ VBS

1912: states, as there are three ways to choose two different trimer states

1913: out of a total of three.  These three VBS states are readily written

1914: out using \eqref{eq:trimerschw},

1915: \begin{equation}

1916: \label{eq:6VBS}

1917: \ket{\psi_{\bs{6}\, \text{VBS}}^{(\mu)}}=

1918: \Psi^{\mu}\!\left[b^{\dagger},r^{\dagger},g^{\dagger}\right]\cdot

1919: \Psi^{\mu+1}\!\left[b^{\dagger},r^{\dagger},g^{\dagger}\right]\,

1920: \vac

1921: \end{equation}

1922: for $\mu=1$, 2, or 3.

1923: If we pick four neighboring sites on a chain with any of these states, 

1924: the total SU(3) spin of those may contain the representations

1925: \begin{displaymath}

1926: \setlength{\unitlength}{1pt}

1927: \begin{picture}(56,12)(0,2)

1928: \linethickness{0.8pt}

1929: \multiput(0,10)(14,0){4}{\circle{4}}

1930: \multiput(0,0)(14,0){4}{\circle{4}}

1931: \thicklines

1932: \put(2,10){\line(1,0){10}}

1933: \put(16,10){\line(1,0){10}}

1934: %\put(44,10){\line(1,0){4}}

1935: %\put(-2,0){\line(-1,0){4}}

1936: \put(16,0){\line(1,0){10}}

1937: \put(30,0){\line(1,0){10}}

1938: \end{picture}

1939: \hat =~~~\bs{3}\,\otimes\,\bs{3}~=~\bs{\bar 3}\,\oplus\,\bs{6}

1940: \end{displaymath}

1941: or the representations

1942: \begin{displaymath}

1943:   \setlength{\unitlength}{1pt}

1944: \begin{picture}(56,12)(0,2)

1945: \linethickness{0.8pt}

1946: \multiput(0,10)(14,0){4}{\circle{4}}

1947: \multiput(0,0)(14,0){4}{\circle{4}}

1948: \thicklines

1949: \put(2,10){\line(1,0){10}}

1950: \put(16,10){\line(1,0){10}}

1951: \put(2,0){\line(1,0){10}}

1952: \put(30,0){\line(1,0){10}}

1953: \end{picture}

1954: \hat =~~~\bs{\bar 3}\,\otimes\,\bs{\bar 3}\,\otimes\,\bs{3}~

1955: =~2\cdot\bs{\bar 3}\,\oplus\,\bs{6}\,\oplus\,\bs{\overline{15}},

1956: \end{displaymath}

1957: \ie the total spin transforms under $\bs{\bar{3}}$, $\bs{6}$, or

1958: $\bs{\overline{15}}=(1,2)$, all of which are contained in the product

1959: \begin{multline}

1960:   \label{eq:6666}

1961:   \bs{6}\,\otimes\,\bs{6}\,\otimes\,\bs{6}\,\otimes\,\bs{6}=\\

1962:   3\cdot\bs{\overline{3}}\,\oplus\,6\cdot\bs{6}\,\oplus\,

1963:   7\cdot\bs{\overline{15}}\,\oplus\,3\cdot\bs{\overline{15}'}\,\oplus\,

1964:   3\cdot\bs{21}\,\\ \oplus\,8\cdot\bs{24}\,\oplus\,

1965:   6\cdot\bs{\overline{42}}\,\oplus\,\bs{45}\,\oplus\,

1966:   6\cdot\bs{60}\,\oplus\,3\cdot\bs{63}

1967: \end{multline}

1968: and hence possible for a representation $\bs{6}$ spin chain in

1969: general.  The corresponding Casimirs are given by

1970: $\casi{0}{1}=\frac{4}{3}$, $\casi{2}{0}=\frac{10}{3}$, and

1971: $\casi{1}{2}=\frac{16}{3}$.  This leads us to propose the parent

1972: Hamiltonian

1973: \begin{equation}

1974:   \label{eq:ham6}

1975:   H_{\bs{6}\,\text{VBS}}=\sum_{i=1}^N H_i

1976: \end{equation}

1977: with 

1978: \begin{equation}

1979:   \label{eq:hi6}

1980:   H_i=\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{4}{3}\right)

1981:   \!\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{10}{3}\right)

1982:   \!\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{16}{3}\right).

1983: \end{equation}

1984: Note that the operators $J_i^a$, $a=1,\ldots,8$, are now given by

1985: $6\times 6$ matrices, as the Gell-Mann matrices only provide the

1986: generators \eqref{eq:J_a} of the fundamental representation $\bs{3}$.

1987: Since the representations $\bs{\bar 3}$, $\bs{6}$, and

1988: $\bs{\overline{15}}$ possess the smallest Casimirs in the expansion

1989: \eqref{eq:6666}, $H_i$ and hence also $H_{\bs{6}\,\text{VBS}}$ are

1990: positive semi-definite (\ie have only non-negative eigenvalues).  The

1991: three linearly independent states \eqref{eq:6VBS} are zero-energy

1992: eigenstates of \eqref{eq:ham6}.  

1993: 

1994: To verify that these are the only ground states, we have numerically

1995: diagonalized \eqref{eq:ham6} for $N=6$ and $N=9$ sites.  For $N=9$, we

1996: find zero-energy ground states at momenta $k=0,3,\ \text{and}\ 6$ (in

1997: units of $\frac{2\pi}{N}$ with the lattice constant set to unity).

1998: Since the dimension of the Hilbert space required the use of a LANCZOS

1999: algorithm, we cannot be certain that there are no further ground

2000: states.  We therefore diagonalized \eqref{eq:ham6} for $N=6$ as well,

2001: where we were able to obtain the full spectrum.  We obtained five zero

2002: energy ground states, two at momentum $k=0$ and one each at

2003: $k=2,3,4$.  One of the ground states at $k=0$ and

2004: the $k=2,4$ ground states constitute the space of momentum eigenstates

2005: obtained by Fourier transform of the space spanned by the three

2006: $\bs{6}$ VBS states \eqref{eq:6VBS}.  The remaining two states at

2007: $k=0,3$ are the momentum eigenstates formed by superposition of the

2008: state

2009: %\pagebreak

2010: \begin{equation}

2011: \setlength{\unitlength}{1pt}

2012: \begin{picture}(120,36)(-14,-12)

2013: \linethickness{0.8pt}

2014: \put(0,0){\circle{4}}

2015: \put(0,10){\circle{4}}

2016: \put(98,0){\circle{4}}

2017: \put(98,10){\circle{4}}

2018: \multiput(14,10)(14,0){6}{\circle{4}}

2019: \multiput(14,0)(14,0){6}{\circle{4}}

2020: \thicklines

2021: \put(2,0){\line(1,0){5}}

2022: \put(7,0){\line(1,0){5}}

2023: \put(30,0){\line(1,0){10}}

2024: \put(58,0){\line(1,0){10}}

2025: \put(16,1){\line(4,3){10}}

2026: \put(44,1){\line(4,3){10}}

2027: \put(72,1){\line(4,3){10}}

2028: \put(86,0){\line(1,0){5}}

2029: \put(91,0){\line(1,0){5}}

2030: %\put(100,1){\line(4,3){3}}

2031: \put(-7,5){\line(4,3){5}}

2032: \qbezier(15,12)(28,25)(41,12)

2033: \qbezier(43,12)(56,25)(69,12)

2034: \qbezier(99,12)(101,14)(103,15)

2035: \thinlines

2036: \put(7,-8){\dashbox{2}(84,32)}

2037: \put(-14,-15){\makebox(0,0){\small $i=$}}

2038: \put(0,-15){\makebox(0,0){\small $6$}}

2039: \put(14,-15){\makebox(0,0){\small $1$}}

2040: \put(28,-15){\makebox(0,0){\small $2$}}

2041: \put(42,-15){\makebox(0,0){\small $3$}}

2042: \put(56,-15){\makebox(0,0){\small $4$}}

2043: \put(70,-15){\makebox(0,0){\small $5$}}

2044: \put(84,-15){\makebox(0,0){\small $6$}}

2045: \put(98,-15){\makebox(0,0){\small $1$}}

2046: \end{picture}

2047: \label{fig:finite6state}

2048: \end{equation}

2049: and the same translated by one lattice spacing.  It is readily seen

2050: that these two states are likewise zero energy eigenstates of

2051: \eqref{eq:ham6} for $N=6$ sites.  The crucial difference, however, is

2052: that the $\bs{6}$ VBS states \eqref{eq:6VBS} remain zero-energy

2053: eigenstates of \eqref{eq:ham6} for all $N$'s divisible by three, while

2054: the equivalent of \eqref{fig:finite6state} for larger $N$ do not.  We

2055: hence attribute these two additional ground states for $N=6$ to the

2056: finite size, and conclude that the three states

2057: \eqref{eq:6VBS} are the only zero-energy ground states of

2058: \eqref{eq:ham6} for general $N$'s divisible by three.

2059: 

2060: Excitations of the $\bs{6}$ VBS model are given by domain walls

2061: between two of the ground states \eqref{eq:6VBS}.  As in the trimer

2062: model, two distinct types of domain walls exist, which transform

2063: according to representations $\bs{\bar{3}}$ and $\bs{3}$:

2064: \begin{equation}\label{eq:6VBSexc}

2065: \setlength{\unitlength}{1pt}

2066: \begin{picture}(220,45)(4,-15)

2067: \linethickness{0.8pt}

2068: \multiput(0,10)(14,0){3}{\circle{4}}

2069: \multiput(42,10)(14,0){2}{\circle*{4}}

2070: \multiput(70,10)(14,0){6}{\circle{4}}

2071: \multiput(154,10)(14,0){1}{\circle*{4}}

2072: \multiput(168,10)(14,0){3}{\circle{4}}

2073: \multiput(14,0)(14,0){15}{\circle{4}}

2074: \thicklines

2075: \put(2,10){\line(1,0){10}}

2076: \put(16,10){\line(1,0){10}}

2077: \put(44,10){\line(1,0){10}}

2078: \put(72,10){\line(1,0){10}}

2079: \put(114,10){\line(1,0){10}}

2080: \put(128,10){\line(1,0){10}}

2081: \put(86,10){\line(1,0){10}}

2082: \put(170,10){\line(1,0){10}}

2083: \put(184,10){\line(1,0){10}}

2084: \put(16,0){\line(1,0){10}}

2085: \put(30,0){\line(1,0){10}}

2086: \put(58,0){\line(1,0){10}}

2087: \put(72,0){\line(1,0){10}}

2088: \put(100,0){\line(1,0){10}}

2089: \put(114,0){\line(1,0){10}}

2090: \put(142,0){\line(1,0){10}}

2091: \put(156,0){\line(1,0){10}}

2092: \put(184,0){\line(1,0){10}}

2093: \put(198,0){\line(1,0){10}}

2094: \put(12,-12){\makebox(0,0){\small $\Psi^1\!\cdot\!\Psi^2$}}

2095: \put(101,-12){\makebox(0,0){\small $\Psi^2\!\cdot\!\Psi^3$}}

2096: \put(198,-12){\makebox(0,0){\small $\Psi^1\!\cdot\!\Psi^2$}}

2097: \put(49,25){\makebox(0,0){$\bs{\bar 3}$}}

2098: \put(154,25){\makebox(0,0){$\bs{3}$}}

2099: \thinlines

2100: %

2101: \put(23,-5){\dashbox{2}(52,20)}

2102: \put(121,-6){\dashbox{2}(52,20)}

2103: \put(135,-4){\dashbox{1}(52,20)}

2104: %

2105: \end{picture}

2106: \end{equation}

2107: %\nopagebreak

2108: It is not clear which excitation has the lower energy, and it appears

2109: likely that both of them are stable against decay.  Let us first look

2110: at the rep.\ $\bs{\bar 3}$ excitation.  The four-site Hamiltonian

2111: \eqref{eq:hi6} annihilates the state for all $i$'s except the four

2112: sites in the dashed box in \eqref{eq:6VBSexc}, which contains

2113: the representations

2114: \begin{multline}%{displaymath}

2115:   \bs{\bar 3}\,\otimes\,\bs{\bar 3}\,\otimes\,\bs{\bar 3}\,

2116:   \otimes\,\bs{3}\,\otimes\,\bs{3}\\ =

2117:   6\cdot\bs{\bar 3}\,\oplus\,5\cdot\bs{6}\,\oplus\,6\cdot

2118:   \bs{\overline{15}}\,\oplus\,\bs{\overline{15}'}\,\oplus\,2\cdot\bs{24}\,

2119:   \oplus\,\bs{\overline{42}}

2120:   \nonumber

2121: \end{multline}%{displaymath}

2122: \ie the representations $\bs{\overline{15}'}=(0,4)$, $\bs{24}=(3,1)$

2123: twice, and $\bs{\overline{42}}=(2,3)$ with Casimirs

2124: $\frac{28}{3}$, $\frac{25}{3}$ and $\frac{34}{3}$, respectively, in

2125: addition to representations annihilated by $H_i$.  For the

2126: rep.\ $\bs{3}$ excitation sketched on the right in \eqref{eq:6VBSexc},

2127: there are two sets of four neighboring sites not annihilated by $H_i$

2128: as indicated by the dashed and the dotted box.  Each set contains the

2129: representations

2130: \begin{displaymath}

2131:   \bs{\bar 3}\,\otimes\,\bs{3}\,\otimes\,\bs{3}\,\otimes\,\bs{3}\,=

2132:   \,3\cdot\bs{\bar 3}\,\oplus\,3\cdot\bs{6}\,\oplus\,2\cdot\bs{\overline{15}}\,

2133:   \oplus\,\bs{24}

2134: \end{displaymath}

2135: \ie only the rep.\ $\bs{24}$ in addition to representations

2136: annihilated by $H_i$.  For our parent Hamiltonian \eqref{eq:ham6}, it

2137: hence may well be that the rep.\ $\bs{3}$ anti-coloron %spinon

2138: has the lower energy, but it is all but clear that the rep.\ $\bs{\bar

2139:   3}$ has sufficiently higher energy to decay.  For general

2140: representation $\bs{6}$ spin chains, it may depend on the specifics of

2141: the model which excitation is lower in energy and whether the

2142: conjugate excitation decays or not.

2143: 

2144: Since the excitations of the rep.\ $\bs{6}$ VBS chain are merely

2145: domain walls between different ground states, there is no confinement

2146: between them.  We expect the generic antiferromagnetic rep.\ $\bs{6}$

2147: chain to be gapless, even though the model we proposed here has a gap

2148: associated with the energy cost of creating a domain wall.

2149: 

2150: \subsection{The representation 10 VBS}\label{sec:10VBS}

2151: 

2152: Let us now turn to the $\bs{10}$ VBS chain, which is a direct

2153: generalization of the AKLT chain to SU(3).  By combining the three

2154: different trimer states \eqref{eq:trimerschw} for $\mu=1,2,\

2155: \text{and}\ 3$ symmetrically,

2156: \begin{eqnarray}

2157: \ket{\psi_{\bs{10}\,\text{VBS}}}\!&\!=\!&\!

2158: \Psi^1\!\left[b^{\dagger}\!,r^{\dagger}\!,g^{\dagger}\right]

2159: \!\cdot\!\Psi^2\!\left[b^{\dagger}\!,r^{\dagger}\!,g^{\dagger}\right]

2160: \!\cdot\!\Psi^3\!\left[b^{\dagger}\!,r^{\dagger}\!,g^{\dagger}\right]\vac

2161: \nonumber \\[5pt]

2162: \!&\!=\!&\!

2163: \prod_i\Bigl(

2164: \sum_{\scriptstyle{(\alpha,\beta,\gamma)}=\atop\scriptstyle{{\pi}(b,r,g)}} 

2165: \hspace{-5pt}\hbox{sign}({\pi})\,

2166: \alpha^{\dagger}_{i}\,\beta^{\dagger}_{i+1}\gamma^{\dagger}_{i+2}

2167: \Bigr)\vac , 

2168: \label{state.10VBS}

2169: \end{eqnarray}

2170: we automatically project out the rep.\ $\bs{10}$ in the decomposition

2171: $\bs{3}\otimes\bs{3}\otimes\bs{3}=\bs{1}\oplus2\!\cdot\bs{8}\oplus\bs{10}$

2172: generated on each lattice site by the three trimer chains.  This

2173: construction yields a unique state, as illustrated:

2174: \begin{equation}

2175: \setlength{\unitlength}{1pt}

2176: \begin{picture}(180,56)(4,-6)

2177: \linethickness{0.8pt}

2178: \multiput(0,35)(14,0){12}{\circle{4}}

2179: \multiput(14,25)(14,0){12}{\circle{4}}

2180: \multiput(28,15)(14,0){12}{\circle{4}}

2181: \thicklines

2182: \put(2,35){\line(1,0){10}}

2183: \put(16,35){\line(1,0){10}}

2184: \put(44,35){\line(1,0){10}}

2185: \put(58,35){\line(1,0){10}}

2186: \put(86,35){\line(1,0){10}}

2187: \put(100,35){\line(1,0){10}}

2188: \put(16,25){\line(1,0){10}}

2189: \put(30,25){\line(1,0){10}}

2190: \put(58,25){\line(1,0){10}}

2191: \put(72,25){\line(1,0){10}}

2192: \put(100,25){\line(1,0){10}}

2193: \put(114,25){\line(1,0){10}}

2194: \put(30,15){\line(1,0){10}}

2195: \put(44,15){\line(1,0){10}}

2196: \put(72,15){\line(1,0){10}}

2197: \put(86,15){\line(1,0){10}}

2198: \put(114,15){\line(1,0){10}}

2199: \put(128,15){\line(1,0){10}}

2200: %

2201: \put(128,35){\line(1,0){10}}

2202: \put(142,35){\line(1,0){10}}

2203: \put(142,25){\line(1,0){10}}

2204: \put(156,25){\line(1,0){10}}

2205: \put(156,15){\line(1,0){10}}

2206: \put(170,15){\line(1,0){10}}

2207: %

2208: \thinlines

2209: \put(51,10){\framebox(10,30){}}

2210: \put(56,2){\line(0,1){8}}

2211: \put(56,-5){\makebox(1,1){\small projection onto $\bs{10}=(3,0)$}}

2212: \put(56,48){\makebox(1,1){\small one site}}

2213: \put(121,10){\dashbox{2}(24,30)}

2214: \end{picture}

2215: \label{fig:10state}

2216: \end{equation}

2217: In order to construct a parent Hamiltonian, note first that the total

2218: spin on two (neighboring) sites of a rep.\ $\bs{10}$ chain is given by

2219: \begin{equation}

2220: \bs{10}\otimes\bs{10}

2221: =\bs{\overline{10}}\oplus\bs{27}\oplus\bs{28}\oplus\bs{35}.

2222: \label{twosites.10}

2223: \end{equation}

2224: On the other hand, the total spin of two neighboring sites for

2225: the $\bs{10}$ VBS state can contain only the representations

2226: \begin{equation}

2227:   \bs{\bar{3}}\otimes\bs{\bar{3}}\otimes\bs{3}\otimes\bs{3}=

2228:   2\cdot\bs{1}\oplus 4\cdot\bs{8}

2229:   \oplus\bs{10}\oplus\bs{\overline{10}}\oplus\bs{27},

2230:   \label{twosites.10VBS}

2231: \end{equation}

2232: as can be seen easily from the dashed box in the cartoon above.  (Note

2233: that this result is independent of how many sites we include in the

2234: dashed box.)  After the projection onto rep.\ $\bs{10}$ on each lattice

2235: site, we find that only reps.\ $\bs{\overline{10}}=(0,3)$ and

2236: $\bs{27}=(2,2)$ occur for the total spin of two neighboring sites for the

2237: $\bs{10}$ VBS state.  With the Casimirs $\casi{0}{3}=6$ and

2238: $\casi{2}{2}=8$ we obtain the parent Hamiltonian

2239: \begin{equation}

2240:   \label{ham.10VBS}

2241:   H_{\bs{10}\,\text{VBS}}

2242:   =\sum_{i=1}^N\,\left(\bigl(\bs{J}_i\bs{J}_{i+1}\bigr)^2

2243:     +5\,\bs{J}_i\bs{J}_{i+1}+6\right),

2244: \end{equation}

2245: the operators $J_i^a$, $a=1,\ldots,8$, are now $10\times 10$ matrices,

2246: and we have used $\bs{J}^2_i=6$.  $H_{\bs{10}\,\text{VBS}}$ is

2247: positive semi-definite and annihilates the $\bs{10}$ VBS state

2248: \eqref{state.10VBS}.  We assume that \eqref{state.10VBS} is the only

2249: ground state of \eqref{ham.10VBS}.

2250: 

2251: The Hamiltonian \eqref{ham.10VBS} provides the equivalent of the AKLT

2252: model~\cite{affleck-87prl799,affleck-88cmp477}, whose unique ground state is

2253: constructed from dimer states by projection onto spin 1, for SU(3)

2254: spin chains.  Note that as in the case of SU(2), it is sufficient to

2255: consider linear and quadratic powers 

2256: %(\ie $\bs{S}_i\bs{S}_{i+1}$ or $\bs{J}_i\bs{J}_{i+1}$) 

2257: of the total spin of only two neighboring

2258: sites.  This is a general feature of the corresponding SU($n$) models,

2259: as we will elaborate in the following section.

2260: 

2261: Since the $\bs{10}$ VBS state \eqref{state.10VBS} is unique, we cannot

2262: have domain walls connecting different ground states.  We hence expect

2263: the coloron and anti-coloron excitations to be confined in pairs, as

2264: illustrated below.  The state between the excitations is no longer

2265: annihilated by \eqref{ham.10VBS}, as there are pairs of neighboring

2266: sites containing higher-dimensional representations, as indicated by

2267: the dotted box below.  As the number of such pairs increases linearly

2268: with the distance between the excitation, the confinement potential

2269: depends linearly on this distance.

2270: \begin{equation}

2271: \setlength{\unitlength}{1pt}

2272: \begin{picture}(210,70)(0,0)

2273: \linethickness{0.8pt}

2274: \multiput(0,45)(14,0){15}{\circle{4}}

2275: \multiput(42,45)(14,0){2}{\circle*{4}}

2276: \multiput(154,45)(14,0){1}{\circle*{4}}

2277: \multiput(14,35)(14,0){15}{\circle{4}}

2278: \multiput(28,25)(14,0){15}{\circle{4}}

2279: \multiput(182,25)(14,0){3}{\circle{4}}

2280: \thicklines

2281: \put(2,45){\line(1,0){10}}\put(16,45){\line(1,0){10}}

2282: \put(44,45){\line(1,0){10}}\put(72,45){\line(1,0){10}}

2283: \put(86,45){\line(1,0){10}}\put(114,45){\line(1,0){10}}

2284: \put(128,45){\line(1,0){10}}\put(170,45){\line(1,0){10}}

2285: \put(184,45){\line(1,0){10}}

2286: \put(16,35){\line(1,0){10}}\put(30,35){\line(1,0){10}}

2287: \put(58,35){\line(1,0){10}}\put(72,35){\line(1,0){10}}

2288: \put(100,35){\line(1,0){10}}\put(114,35){\line(1,0){10}}

2289: \put(142,35){\line(1,0){10}}\put(156,35){\line(1,0){10}}

2290: \put(184,35){\line(1,0){10}}\put(198,35){\line(1,0){10}}

2291: \put(30,25){\line(1,0){10}}\put(44,25){\line(1,0){10}}

2292: \put(72,25){\line(1,0){10}}\put(86,25){\line(1,0){10}}

2293: \put(114,25){\line(1,0){10}}\put(128,25){\line(1,0){10}}

2294: \put(156,25){\line(1,0){10}}\put(170,25){\line(1,0){10}}

2295: \put(198,25){\line(1,0){10}}\put(212,25){\line(1,0){10}}

2296: \thinlines

2297: \put(105,11){\vector(-1,0){49}}

2298: \put(105,11){\vector(1,0){49}}

2299: \put(105,0){\makebox(0,0){\small energy cost $\propto$ distance}}

2300: %

2301: \put(49,56){\makebox(0,0){$\bs{\bar 3}$}}

2302: \put(154,56){\makebox(0,0){$\bs{3}$}}

2303: \put(49,67){\makebox(0,0){\small coloron}}

2304: \put(154,67){\makebox(0,0){\small anti-coloron}}

2305: \put(93,20){\dashbox{1}(24,30)}

2306: \end{picture}\label{10VBSstate.exc}

2307: \end{equation}

2308: In principle, it would also be possible to create three colorons (or

2309: three anti-colorons) rather than a coloron--anti-coloron pair, but as

2310: all three excitations would feel strong confinement forces, we expect

2311: the coloron--anti-coloron pair to constitute the dominant low energy

2312: excitation.  The confinement force between the pair induces a

2313: linear oscillator potential for the relative motion of the

2314: constituents.  The zero-point energy of this oscillator gives rise to

2315: a Haldane-type energy gap

2316: (see~\cite{greiter02prb134443,greiter02prb054505} for a similar

2317: discussion in the two-leg Heisenberg ladder), which is independent of

2318: the model specifics.  We expect this gap to be a generic feature of

2319: rep.\ $\bs{10}$ spin chains with short-range antiferromagnetic

2320: interactions.

2321: 

2322: 

2323: \subsection{The representation 8 VBS}

2324: \label{sec:8VBS}

2325: 

2326: To construct a representation $\bs{8}$ VBS state, consider first a chain

2327: with alternating representations $\bs{3}$ and $\bs{\bar{3}}$ on

2328: neighboring sites, which we combine into singlets.  This can be done in

2329: two ways, yielding the two states  

2330: \begin{equation}

2331: \setlength{\unitlength}{1pt}

2332: \begin{picture}(210,22)(0,-12)

2333: \linethickness{0.8pt}

2334: \multiput(5,0)(30,0){3}{\circle{4}}

2335: \multiput(20,0)(30,0){3}{\circle{6}}

2336: \multiput(138,0)(30,0){3}{\circle{4}}

2337: \multiput(153,0)(30,0){3}{\circle{6}}

2338: %

2339: \thicklines

2340: \multiput(7,0)(30,0){3}{\line(1,0){10}}

2341: \multiput(156,0)(30,0){2}{\line(1,0){10}}

2342: %

2343: \put(136,0){\line(-1,0){4}}

2344: \put(216,0){\line(1,0){3}}

2345: %

2346: \thinlines

2347: \put(109,0){\makebox(0,0){and}}

2348: \put(224,-3){\makebox(0,0){.}}

2349: \put(35,-12){\makebox(0,0){\small $\bs{3}$}}

2350: \put(50,-12){\makebox(0,0){\small $\bs{\bar{3}}$}}

2351: \end{picture}

2352: \nonumber

2353: %\label{fig:33bar}

2354: \end{equation}

2355: We then combine a $\bs{3}$--$\bs{\bar{3}}$ state with an identical one

2356: shifted by one lattice spacing.  This yields representations

2357: $\bs{3}\otimes\bs{\bar{3}}=\bs{1}\oplus\bs{8}$ at each site.  The

2358: $\bs{8}$ VBS state is obtained by projecting onto %the adjoint

2359: representation $\bs{8}$.  Corresponding to the two

2360: $\bs{3}$--$\bs{\bar{3}}$ states illustrated above, we obtain two

2361: linearly independent $\bs{8}$ VBS states, $\Psi^\text{L}$ and

2362: $\Psi^\text{R}$, which may be visualized as

2363: \begin{equation}

2364: \setlength{\unitlength}{1pt}

2365: %\begin{picture}(210,50)(0,-20)

2366: \begin{picture}(210,60)(0,-24)

2367: \linethickness{0.8pt}

2368: \multiput(5,10)(30,0){3}{\circle{4}}

2369: \multiput(20,10)(30,0){3}{\circle{6}}

2370: \multiput(138,10)(30,0){3}{\circle{4}}

2371: \multiput(153,10)(30,0){3}{\circle{6}}

2372: \multiput(5,0)(30,0){3}{\circle{6}}

2373: \multiput(20,0)(30,0){3}{\circle{4}}

2374: \multiput(138,0)(30,0){3}{\circle{6}}

2375: \multiput(153,0)(30,0){3}{\circle{4}}

2376: %

2377: \thicklines

2378: \multiput(7,10)(30,0){3}{\line(1,0){10}}

2379: \multiput(156,10)(30,0){2}{\line(1,0){10}}

2380: \multiput(22,0)(30,0){2}{\line(1,0){10}}

2381: \multiput(141,0)(30,0){3}{\line(1,0){10}}

2382: %

2383: \put(136,10){\line(-1,0){4}}

2384: \put(216,10){\line(1,0){3}}

2385: \put(2,0){\line(-1,0){3}}

2386: \put(82,0){\line(1,0){4}}

2387: %

2388: \thinlines

2389: \put(109,5){\makebox(0,0){and}}

2390: \put(224,2){\makebox(0,0){.}}

2391: %

2392: \put(44,-6){\framebox(12,22)}

2393: \put(50,-6){\line(0,-1){9}}

2394: \put(50,-22){\makebox(1,1){\small projection onto $\bs{8}=(1,1)$}}

2395: \put(50,25){\makebox(1,1){\small one site}}

2396: \put(162.5,-6){\dashbox{2}(26,22)}

2397: \end{picture}

2398: \label{fig:8VBS}

2399: \end{equation}

2400: These states transform into each other under space reflection or color 

2401: conjugation (interchange of $\bs{3}$ and $\bs{\bar{3}}$).

2402: % $(\bs{3},\bs{\bar{3}})\to (\bs{\bar{3}},\bs{3})$.  

2403: 

2404: 

2405: It is convenient to formulate the corresponding state vectors as a

2406: matrix product.  Taking (\b,\r,\g ) and (\y,\c,\m ) as bases for the

2407: reps.\ $\bs{3}$ and $\bs{\bar{3}}$, respectively, the singlet bonds in

2408: $\Psi^\text{L}$ above can be written

2409: \begin{multline}

2410:   \Bigl(\ket{\b}_i \ket{\y}_{i+1} + \ket{\r}_i \ket{\c}_{i+1}

2411:   + \ket{\g}_i \ket{\m}_{i+1}\Bigr)\\ =

2412:   \Bigl(\ket{\b}_i,\ket{\r}_i,\ket{\g}_i\Bigr)

2413:   \left(\!

2414:     \begin{array}{c} \ket{\y}_{i+1}\\\ket{\c}_{i+1}\\\ket{\m}_{i+1}\end{array}

2415:   \!\right)\!.

2416: \nonumber

2417: \end{multline}

2418: We are hence led to consider matrices composed of the outer product

2419: of these vectors on each lattice site,

2420: \begin{displaymath}

2421:   %\tilde 

2422:   M_i^{\bs{1}\oplus\bs{8}} 

2423:   = \left(\!

2424:     \begin{array}{c} \ket{\y}_{i}\\\ket{\c}_{i}\\\ket{\m}_{i}\end{array}

2425:     \!\right)\!

2426:   \Bigl(\ket{\b}_i,\ket{\r}_i,\ket{\g}_i\Bigr).

2427: \end{displaymath}\\

2428: In the case of the AKLT model reviewed above, the Schwinger bosons take

2429: care of the projection automatically, and we can simply assemble

2430: these matrices into a product state.  For the $\bs{8}$ VBS,

2431: however, we need to enforce the projection explicitly.  This is

2432: most elegantly accomplished using the Gell-Mann matrices, yielding the

2433: projected matrix

2434: \begin{equation}

2435:   \label{eq:m8}

2436:   M_i%^{\bs{8}}

2437:   =\frac{1}{2}\sum_{a=1}^8 \lambda^a\, 

2438:   \text{tr}\Bigl(\lambda^a M_i^{\bs{1}\oplus\bs{8}}\Bigr).

2439: \end{equation}

2440: Here we have simply used the fact that the eight Gell-Mann matrices,

2441: supplemented by the unitary matrix, constitute a complete basis for

2442: the space of all complex $3\times 3$ matrices.  By omitting the unit

2443: matrix in the expansion \eqref{eq:m8}, we effectively project out

2444: the singlet state.  

2445: Written out explicitly, we obtain 

2446: \begin{widetext}

2447: \begin{equation}

2448:   \label{eq:m8explictly}

2449:   M_i%^{\bs{8}}

2450:   =\left(\!\begin{array}{ccc} 

2451:   \frac{2}{3}\ket{\b\y}_i-\frac{1}{3}\ket{\r\c}_i-\frac{1}{3}\ket{\g\m}_i

2452:   &\ket{\r\y}_i&\ket{\g\y}_i\\[4pt]

2453:   \ket{\b\c}_i

2454:   &-\frac{1}{3}\ket{\b\y}_i+\frac{2}{3}\ket{\r\c}_i-\frac{1}{3}\ket{\g\m}_i

2455:   &\ket{\g\c}_i\\[4pt]

2456:   \ket{\b\m}_i&\ket{\r\m}_i

2457:   &-\frac{1}{3}\ket{\b\y}_i-\frac{1}{3}\ket{\r\c}_i+\frac{2}{3}\ket{\g\m}_i 

2458:   \\[4pt]    

2459:     \end{array}\!\right).

2460: \end{equation}

2461: \end{widetext}

2462: Assuming PBCs, the $\bs{8}$ VBS state $\Psi^\text{L}$ (illustrated in

2463: on the left \eqref{fig:8VBS}) is hence given by the trace of the

2464: matrix product

2465: \begin{equation}

2466:   \label{eq:8VBSL}

2467:   \ket{\psi_{\bs{8}\,\text{VBS}}^\text{L}}

2468:   =\text{tr}\biggl( \prod_i M_i%^{\bs{8}}

2469:   \biggr).

2470: \end{equation}

2471: To obtain the state $\Psi^\text{R}$ (illustrated on the right in

2472: \eqref{fig:8VBS}) we simply have to transpose the matrices in the

2473: product,

2474: \begin{equation}

2475:   \label{eq:8VBSR}

2476:   \ket{\psi_{\bs{8}\,\text{VBS}}^\text{R}}

2477:   =\text{tr}\biggl( \prod_i M_i%^{\bs{8}}

2478:   ^\text{T} \biggr).

2479: \end{equation}

2480: 

2481: Let us now formulate a parent Hamiltonian for these states.

2482: If we consider two lattice sites on an SU(3) chain with a representation

2483: $\bs{8}$ on each lattice site in general, we find the full SU(3) content

2484: \begin{equation}

2485: \bs{8}\,\otimes\,\bs{8}\,=\,\bs{1}\,\oplus\,2\cdot\bs{8}\,

2486: \oplus\,\bs{10}\,\oplus\,\bs{\overline{10}}\,\oplus\,\bs{27}

2487: \end{equation}

2488: with $\bs{10}=(3,0)$, $\bs{\overline{10}}=(0,3)$, and $\bs{27}=(2,2)$.

2489: On the other hand, for the $\bs{8}$ VBS states only the

2490: representations $\bs{3}\otimes\bs{\bar 3}\,=\,\bs{1}\oplus\bs{8}$ can

2491: occur for the total spin of two neighboring sites, as the two sites

2492: always contain one singlet (see dashed box in \eqref{fig:8VBS} on the

2493: right above).  With the Casimirs $\casi{0}{0}=0$ and $\casi{1}{1}=3$

2494: for representations $\bs{1}$ and $\bs{8}$, respectively, we construct

2495: the parent Hamiltonian

2496: \begin{equation}

2497: \label{ham.8VBS}

2498: H_{\bs{8}\,\text{VBS}}=\sum_{i=1}^N\left(\bigl(\bs{J}_i\bs{J}_{i+1}\bigr)^2+

2499: \frac{9}{2}\,\bs{J}_i\bs{J}_{i+1} + \frac{9}{2}\right),

2500: \end{equation}

2501: where the operators $J_i^a$, $a=1,\ldots,8$, are now $8\times 8$

2502: matrices, and we have used the Casimir $\bs{J}_i^2=3$ on each site.

2503: $H_{\bs{8}\,\text{VBS}}$ is positive semi-definite, and annihilates the

2504: states $\Psi^\text{L}$ and $\Psi^\text{R}$.  We have numerically

2505: verified for chains with $N=3$, 4, 5, and 6 lattice sites that

2506: $\Psi^\text{L}$ and $\Psi^\text{R}$ are the only ground states of

2507: \eqref{ham.8VBS}.

2508: 

2509: Naively, one might assume the $\bs{8}$ VBS model to support deconfined

2510: spinons or colorons, which correspond to domain walls between the

2511: two ground states $\Psi^\text{L}$ and $\Psi^\text{R}$.  A closer look

2512: at the domain walls, however, shows that this is highly unlikely,

2513: as each domain wall is a bound state of either two anti-colorons or two

2514: colorons, as illustrated below.

2515: \begin{equation}

2516: \setlength{\unitlength}{1pt}

2517: %\begin{picture}(210,50)(0,-20)

2518: \begin{picture}(210,64)(0,-24)

2519: \linethickness{0.8pt}

2520: \multiput(5,10)(30,0){7}{\circle{4}}

2521: \multiput(20,10)(30,0){7}{\circle{6}}

2522: \multiput(35,0)(30,0){7}{\circle{6}}

2523: \multiput(20,0)(30,0){7}{\circle{4}}

2524: \multiput(65,10)(30,0){1}{\circle*{4}}

2525: \multiput(50,0)(30,0){1}{\circle*{4}}

2526: \multiput(170,10)(30,0){1}{\circle*{6}}

2527: \multiput(155,0)(30,0){1}{\circle*{6}}

2528: %

2529: \thicklines

2530: \multiput(7,10)(30,0){2}{\line(1,0){10}}

2531: \multiput(83,10)(30,0){3}{\line(1,0){10}}

2532: \multiput(187,10)(30,0){1}{\line(1,0){10}}

2533: \multiput(22,0)(30,0){1}{\line(1,0){10}}

2534: \multiput(68,0)(30,0){3}{\line(1,0){10}}

2535: \multiput(172,0)(30,0){2}{\line(1,0){10}}

2536: %

2537: \thinlines

2538: %\put(50,0){\line(3,2){13}}

2539: \put(22,-15){\makebox(0,0){\small $\Psi^\text{L}$}}

2540: \put(112,-15){\makebox(0,0){\small $\Psi^\text{R}$}}

2541: \put(202,-15){\makebox(0,0){\small $\Psi^\text{L}$}}

2542: %

2543: \put(50,22){\makebox(0,0){\small $\bs{3}$}}

2544: \put(65,22){\makebox(0,0){\small $\bs{3}$}}

2545: \put(155,22){\makebox(0,0){\small $\bs{\bar{3}}$}}

2546: \put(170,22){\makebox(0,0){\small $\bs{\bar{3}}$}}

2547: %

2548: \put(57.5,34){\makebox(0,0){\small anti-colorons}}

2549: \put(162.5,34){\makebox(0,0){\small colorons}}

2550: \end{picture}

2551: \label{fig:8VBSdomainwall}

2552: \end{equation}

2553: There is no reason to assume that the domain wall depicted above as

2554: two anti-colorons in fact corresponds to a single coloron, as it

2555: appears to be the case for the trimer chain.  There we created a

2556: domain wall corresponding to a single coloron by removing one of the

2557: rep.\ $\bs{3}$ spins from a trimer, leaving the remaining rep.\

2558: $\bs{3}$ spins coupled antisymmetrically as in the ground state.

2559: %There, however, we created a domain wall corresponding to a single

2560: %coloron by removing one of the rep.\ $\bs{3}$ spins from a trimer.  

2561: If we were to combine the two reps.\ $\bs{{3}}$ into a rep.\

2562: $\bs{\bar{3}}$ in \eqref{fig:8VBSdomainwall}, we would not reproduce a

2563: correlation present in the ground state, but enforce a new

2564: correlation.  The correct interpretation of the domain wall between

2565: $\Psi^\text{L}$ and $\Psi^\text{R}$ is hence that of a bound state between

2566: two linearly confined anti-colorons.  The origin of the confining

2567: potential is illustrated below.

2568: \begin{equation}

2569: \setlength{\unitlength}{1pt}

2570: \begin{picture}(175,70)(0,-30)

2571: \linethickness{0.8pt}

2572: \multiput(5,10)(30,0){6}{\circle{4}}

2573: \multiput(20,10)(30,0){5}{\circle{6}}

2574: \multiput(35,0)(30,0){5}{\circle{6}}

2575: \multiput(20,0)(30,0){6}{\circle{4}}

2576: \multiput(125,10)(30,0){1}{\circle*{4}}

2577: \multiput(50,0)(30,0){1}{\circle*{4}}

2578: %

2579: \thicklines

2580: \multiput(7,10)(30,0){4}{\line(1,0){10}}

2581: \multiput(143,10)(30,0){1}{\line(1,0){10}}

2582: \multiput(22,0)(30,0){1}{\line(1,0){10}}

2583: \multiput(68,0)(30,0){4}{\line(1,0){10}}

2584: %

2585: \thinlines

2586: \put(22,-15){\makebox(0,0){\small $\Psi^\text{L}$}}

2587: \put(157,-15){\makebox(0,0){\small $\Psi^\text{R}$}}

2588: %

2589: \put(50,23){\makebox(0,0){\small $\bs{3}$}}

2590: \put(125,23){\makebox(0,0){\small $\bs{3}$}}

2591: \put(50,34){\makebox(0,0){\small anti-coloron}}

2592: \put(125,34){\makebox(0,0){\small anti-coloron}}

2593: %

2594: \put(87.5,-14){\vector(-1,0){37.5}}

2595: \put(87.5,-14){\vector(1,0){37.5}}

2596: \put(87.5,-25){\makebox(0,0){\small energy cost $\propto$ distance}}

2597: \end{picture}

2598: \label{fig:8VBS3bar3bar}

2599: \end{equation}

2600: As in the $\bs{10}$ VBS, the confinement induces a linear oscillator

2601: potential for the relative motion of the anti-colorons. The zero-point

2602: energy of this oscillator corresponds to a Haldane-type gap in the

2603: spectrum.  The ground state

2604: wave function of the oscillator is symmetric, and hence corresponds to

2605: a symmetric combination of $\bs{3}\otimes\bs{3}$, \ie rep.\ $\bs{6}$.

2606: The antisymmetric combination $\bs{\bar{3}}$ corresponds to the first

2607: excited state of the oscillator, which we expect to cost more than

2608: twice the energy of the symmetric state~\cite{greiter02prb054505}.

2609: This statement holds for the pair of colorons in

2610: \eqref{fig:8VBSdomainwall} as well.

2611: 

2612: \begin{figure}[t]

2613: \includegraphics[angle=270,scale=0.53]{rep8spec.eps}

2614: %%BoundingBox: 176 43 574 530 (new)

2615: \begin{picture}(0,0)

2616: \put(13,5){\makebox(0,0){$k~[2\pi/N]$}}

2617: \put(-120,110){\rotatebox{90}{$E_{\bs{8}\,\text{VBS}}$}}

2618: \put(105,61){\makebox(0,0)[r]{{\scriptsize singlet}}}

2619: \put(105,53){\makebox(0,0)[r]{{\scriptsize rep.\,$\bs{8}$}}}

2620: \put(105,46){\makebox(0,0)[r]{{\scriptsize rep.\,$\bs{10}$}}}

2621: \put(105,38){\makebox(0,0)[r]{{\scriptsize rep.\,$\bs{27}$}}}

2622: \end{picture}

2623: \caption{%(Color online)

2624:   Spectrum of the $\bs{8}$ VBS Hamiltonian \eqref{ham.8VBS} for $N=8$

2625:   sites obtained by exact diagonalization.  (The lines are merely

2626:   guides to the eye.)  The ``magnon'' excitation transforming under

2627:   rep.\ $\bs{8}$ of SU(3) has the lowest energy, followed by a singlet

2628:   excitation, as expected from the discussion in the text.  The well

2629:   defined modes at low energies provide strong evidence of

2630:   coloron--anti-coloron bound states as compared to deconfined domain

2631:   walls, and hence support our conclusion that the $\bs{8}$ VBS

2632:   exhibits a Haldane gap due to spinon confinement.}

2633: \label{fig:8VBSspec}

2634: \end{figure}

2635: 

2636: The domain walls, however, are not the only low energy excitations.

2637: In either of the ground states, we can create coloron--anti-coloron

2638: bound states, which make no reference to the other ground state, as

2639: illustrated below.

2640: \begin{equation}

2641: \setlength{\unitlength}{1pt}

2642: \begin{picture}(175,70)(0,-30)

2643: \linethickness{0.8pt}

2644: \multiput(5,10)(30,0){6}{\circle{4}}

2645: \multiput(20,10)(30,0){6}{\circle{6}}

2646: \multiput(35,0)(30,0){5}{\circle{6}}

2647: \multiput(20,0)(30,0){5}{\circle{4}}

2648: \multiput(125,0)(30,0){1}{\circle*{6}}

2649: \multiput(50,0)(30,0){1}{\circle*{4}}

2650: 

2651: %

2652: \thicklines

2653: \multiput(7,10)(30,0){6}{\line(1,0){10}}

2654: \multiput(142,0)(30,0){1}{\line(1,0){10}}

2655: \multiput(22,0)(30,0){1}{\line(1,0){10}}

2656: \multiput(68,0)(30,0){2}{\line(1,0){10}}

2657: %

2658: \thinlines

2659: \put(22,-15){\makebox(0,0){\small $\Psi^\text{L}$}}

2660: \put(157,-15){\makebox(0,0){\small $\Psi^\text{L}$}}

2661: %

2662: \put(50,23){\makebox(0,0){\small $\bs{3}$}}

2663: \put(125,23){\makebox(0,0){\small $\bs{\bar{3}}$}}

2664: \put(50,34){\makebox(0,0){\small anti-coloron}}

2665: \put(125,34){\makebox(0,0){\small coloron}}

2666: %

2667: \put(87.5,-14){\vector(-1,0){37.5}}

2668: \put(87.5,-14){\vector(1,0){37.5}}

2669: \put(87.5,-25){\makebox(0,0){\small energy cost $\propto$ distance}}

2670: \end{picture}

2671: \label{fig:8VBS33bar}

2672: \end{equation}

2673: The oscillator model tells us again that the ``symmetric'' combination

2674: of $\bs{3}\otimes\bs{\bar{3}}$, \ie rep.\ $\bs{8}$, has the lowest

2675: energy, which we expect to be comparable, if not identical, to the

2676: energy required to create each of the domain walls above.  

2677: The %antisymmetric combination, \ie the

2678: singlet $\bs{1}$ will have an energy comparable to that of a domain

2679: wall transforming under either $\bs{\bar{3}}$ or $\bs{3}$.  In any

2680: event, we expect the $\bs{8}$ VBS model to display a Haldane gap due to

2681: coloron confinement.

2682: 

2683: The excitation spectrum of \eqref{ham.8VBS} for a chain with $N=8$

2684: sites and PBCs is shown in Fig.~\ref{fig:8VBSspec}.  The spectrum shows

2685: that the lowest excitation transforms under rep.~$\bs{8}$, as expected

2686: from \eqref{fig:8VBS33bar}, with a singlet and then another rep~$\bs{8}$

2687: following at slightly higher energies.  It is tempting to interpret 

2688: those three levels as the lowest levels of the coloron--anti-coloron

2689: oscillator \eqref{fig:8VBS33bar}, but then there should be another 

2690: singlet at a comparable spacing above.  The fact that the spacings

2691: between these excitations are significantly smaller than the energy

2692: of the first exited state, however, would be consistent with such an 

2693: interpretation, as the spinons in VBS models always have a local energy 

2694: cost associated with their creation, which is specific to these models

2695: and not related to the universal Haldane gap stemming from confinement

2696: forces.

2697: 

2698: Most importantly, the spectrum provides strong evidence in favor of

2699: our assumption that the domain walls are not elementary excitations,

2700: but bound states of either two colorons or two anti-colorons, and

2701: hence that the lowest energy excitations of finite chains are

2702: coloron--anti-coloron bound states as illustrated in

2703: \eqref{fig:8VBS33bar}.  The assumption is crucial for our conclusion

2704: that the model exhibits a Haldane gap.  If the low energy sector of

2705: the model was determined by two deconfined domain walls, we would see

2706: a continuum of states in the spectrum, similar to the spectrum seen in

2707: spin $S=\frac{1}{2}$ chains of SU(2).  The well defined low-energy

2708: modes in Fig.~\ref{fig:8VBSspec}, however, look much more like the

2709: spinon--spinon bound state excitations seen in $S=1$ chains

2710: or two-leg $S=\frac{1}{2}$ Heisenberg ladders.  In particular, if we

2711: assume that the individual domain walls transform under reps.\

2712: $\bs{6}$ and $\bs{\bar{6}}$, we expect excitations transforming under

2713: the representations contained in

2714: $\bs{6}\otimes\bs{\bar{6}}=\bs{1}\oplus\bs{8}\oplus\bs{27}$ to be

2715: approximately degenerate.  Fig.~\ref{fig:8VBSspec} shows clearly that

2716: such a multiplet is not present a the lowest energies.

2717: 

2718: 

2719: 

2720: 

2721: \section{SU($\bs{n}$) models}

2722: \label{sec:sun}

2723: 

2724: In this section, we generalize three of the models proposed for SU(3)

2725: spin chains, the trimer model, the symmetric representation $\bs{10}$ VBS, 

2726: and the matrix product state $\bs{8}$ VBS to the case of 

2727: SU($n$) spin chains.

2728: 

2729: \subsection{The $\bs{n}$-mer model}

2730: 

2731: Consider an  SU($n$) spin chain with $N$ sites, where $N$ is a multiple of $n$,

2732: with a spin transforming according 

2733: to the fundamental representation $\bs{n}$ of  SU($n$) at each lattice site,

2734: \begin{equation}

2735: \begin{array}{rcccccccl}

2736: \bs{n}&=&(1,0,\ldots,0)&=&

2737: \setlength{\unitlength}{8pt}

2738: \begin{picture}(1,1)(0,0.1)

2739: \put(0,1){\line(1,0){1}}

2740: \put(0,0){\line(1,0){1}}

2741: \put(0,0){\line(0,1){1}}

2742: \put(1,0){\line(0,1){1}}

2743: \end{picture}

2744: &\hat =&c_{\sigma}^\dagger\vac, 

2745: \end{array}

2746: \end{equation}

2747: where $\sigma$ denotes a ``flavor'', $\sigma\in\{{\rm f}_1,\ldots,{\rm

2748:   f}_n\}$, and $c_{\sigma}^\dagger$ creates a fermion of flavor

2749: $\sigma$.

2750:   

2751: The SU($n$) generators at site $i$ are in analogy to

2752: \eqref{eq:s} and \eqref{eq:J_a} defined as

2753: \begin{equation}

2754: J^a_i=\frac{1}{2}\,\sum_{\sigma,\sigma'={\rm f}_1,\ldots,{\rm f}_n}\,

2755: c_{i\sigma}^{\dagger} V^a_{\sigma\sigma'}

2756: c_{i\sigma'}^{\phantom{\dagger}}, \quad a=1,\ldots,n^2-1,

2757: \label{eq:J_an}

2758: \end{equation}

2759: where the $V^a$ denote the $n^2\!-\!1$ SU($n$) Gell-Mann

2760: matrices~\cite{macfarlane-68cmp77}.  The generators are normalized

2761: through the eigenvalue the quadratic Casimir operator takes in the

2762: adjoint representation, $\bs{J}^2=\casin{1,0,\ldots,0,1}=n$.

2763: 

2764: To determine the eigenvalues of the quadratic Casimir for general

2765: representations of SU($n$), a significant amount of representation

2766: theory is required~\cite{Humphreys87}.  We content ourselves here

2767: by providing the formulas up to $n=6$ in App.~\ref{quadraticcasimirs}.

2768: 

2769: In analogy to the trimer states \eqref{eq:trimer}, we construct 

2770: the $n$-mer states of an SU($n$) spin chain %are simply constructed 

2771: by combining sets of $n$ neighboring spins into a singlet,

2772: \begin{equation}

2773:   \label{eq:nmer}

2774:   \ket{\psi_{n\text{-mer}}^{(\mu)}\!}=\hspace{-15pt}\prod_{\scriptstyle{i} \atop 

2775:     \left(\scriptstyle{\frac{i-\mu}{n}\,{\rm integer}}\right)}

2776:   \hspace{-5pt}\Bigl(

2777:   \sum_{\scriptstyle{(\sigma_1,\ldots,\sigma_n)}=

2778:     \atop\scriptstyle{{\pi}({\rm f}_1,\ldots,{\rm f}_n)}} 

2779:   \hspace{-10pt}\hbox{sign}({\pi})\,

2780:   \prod_{\kappa=1}^n c^{\dagger}_{i-1+\kappa\;\sigma_\kappa}

2781:   \Bigr)\!\vac\!,

2782: \end{equation}

2783: where $\mu=1,\ldots,n$ labels the $n$ degenerate ground states, and

2784: $i$ runs over the lattice sites subject to the constraint that

2785: $\frac{i-\mu}{n}$ is integer.  The sum extends over all $n!$

2786: permutations $\pi$ of the $n$ flavors ${\rm f}_1,\ldots,{\rm f}_n$.

2787: 

2788: In order to identify a parent Hamiltonian, consider the total SU($n$)

2789: spin on $n+1$ neighboring sites for the $n$-mer states.  Following the

2790: rules of combining representations labeled by Young tableaux (see \eg

2791: \cite{Cornwell84vol2,Georgi82}), it is not difficult to see that the

2792: total spin will only contain representations given by tableaux with

2793: $n+1$ boxes and two columns, \ie tableaux of the form

2794: \begin{displaymath}

2795: \setlength{\unitlength}{8pt}

2796: \begin{picture}(5,7.5)

2797: \put(0,0){\line(0,1){7.5}}

2798: \put(1,0){\line(0,1){7.5}}

2799: \put(2,3.5){\line(0,1){4}}

2800: %

2801: \put(0,0){\line(1,0){1}}

2802: \put(0,1){\line(1,0){1}}

2803: \put(0,2.5){\line(1,0){1}}

2804: \put(0,3.5){\line(1,0){2}}

2805: \put(0,4.5){\line(1,0){2}}

2806: \put(0,6.5){\line(1,0){2}}

2807: \put(0,7.5){\line(1,0){2}}

2808: \put(4.5,5.5){\makebox(0,0)

2809: {$\left.\begin{array}{c}~\\~\end{array}\right\}{\small ~\nu~\text{rows}}$}}

2810: \end{picture}

2811: \end{displaymath}

2812: with $1\le\nu\le\frac{n+1}{2}$.  The 

2813: eigenvalues of the quadratic Casimir operator for these

2814: representations are

2815: \begin{equation}

2816: f_n(\nu)=\frac{1}{2n}\Bigl(n^2(2\nu -1) -2n(\nu -1)^2 -1 \Bigr).

2817: \end{equation}

2818: An educated guess for a parent Hamiltonian for the $n$-mer chain

2819: hence appears to be

2820: \begin{equation}

2821: H_{\text{trial}}=\sum_{i=1}^N H_i\quad

2822: \label{eq:MGSUni}

2823: \end{equation}

2824: with

2825: \begin{equation}

2826: H_i=\prod_{\nu=1}^{\lfloor\frac{n+1}{2}\rfloor}

2827: \left(\!\left(\bs{J}_i^{(n+1)}\right)^2 - f_n(\nu)\right),

2828: \label{eq:MGSUn}

2829: \end{equation}

2830: where $\lfloor \; \rfloor$ denotes the floor function, \ie $\lfloor

2831: x\rfloor$ is the largest integer $l\le x$, and we use the notation

2832: introduced in \eqref{eq:defJnu}.  

2833: 

2834: This construction yields the MG model~\cite{majumdar-69jmp1399} for

2835: SU(2), the trimer model \eqref{ham.trimer} for SU(3), and a valid

2836: parent Hamiltonian for the four degenerate 4-mer states for SU(4).

2837: For $n\ge 5$, however, the decomposition of the tensor product

2838: $\bs{n}^{\otimes (n+1)}$ contains irreducible representations

2839: corresponding to Young tableaux with more than two columns, whose

2840: Casimirs are equal or smaller than a number of Casimirs included in

2841: the list $f_n(\nu)$, $\nu=1,2,\ldots,\lfloor\frac{n+1}{2}\rfloor$.  If

2842: the Casimir of such an ``undesired'' representation not included in

2843: the list is smaller than an odd number of Casimirs included in the

2844: list, we obtain negative eigenvalues for $H_i$, and it is not a priori

2845: clear any more that the Hamiltonian \eqref{eq:MGSUni} is positive

2846: semi-definite.  An obvious cure to this problem is to write

2847: \begin{equation}

2848:   H_{n\text{-mer}}=\sum_{i=1}^N H_i^2,\quad

2849:   \label{eq:hisquare}

2850: \end{equation}

2851: with $H_i$ as in \eqref{eq:MGSUn}.  This does, however, not cure

2852: potential problems arising from undesired representations which 

2853: %happen to 

2854: share the eigenvalues of the Casimir with one of the

2855: representations from the list, as it happens to be the case for $n=5$.

2856: The Hamiltonian \eqref{eq:hisquare} likewise annihilates these

2857: representations, giving rise to a remote possibility that the $n$-mer

2858: states \eqref{eq:nmer} are not the only ground states of

2859: \eqref{eq:hisquare}.  The potential relevance of these problems has to

2860: be investigated for each $n$ separately.

2861: 

2862: \subsection{The representation $\bs{(n,0,\ldots,0)}$ VBS}

2863: 

2864: As a generalization of the AKLT model for SU(2) and the ${\bs{10}}$ VBS

2865: model for SU(3) discussed above, we now consider a VBS for an SU($n$) chain

2866: of spins transforming under the symmetric representation

2867: \begin{equation}

2868:   \label{eq:n000}

2869:   \begin{array}{rcccl}

2870:     (n,0,\ldots,0)&=&

2871:     \setlength{\unitlength}{8pt}

2872:     \begin{picture}(5,2)(0,3.15)

2873:       \put(0,4){\line(1,0){5}}

2874:       \put(0,3){\line(1,0){5}}

2875:       \put(0,3){\line(0,1){1}}

2876:       \put(1,3){\line(0,1){1}}

2877:       \put(4,3){\line(0,1){1}}

2878:       \put(5,3){\line(0,1){1}}

2879:       \put(2.5,1.5){\makebox(0,0)%

2880:         {$\underbrace{\quad\text{~~~}\qquad}_{\text{\small $n$~boxes}}$}}

2881:    \end{picture}

2882:     &\hat =& b_{\sigma_1}^\dagger b_{\sigma_2}^\dagger\ldots b_{\sigma_n}^\dagger

2883:     \vac,\\[2pt] \nonumber

2884: \end{array}

2885: \end{equation}

2886: where each $b_{\sigma}^\dagger$, $\sigma\in\{{\rm f}_1,\ldots,{\rm

2887:   f}_n\}$, is an SU($n$) Schwinger boson.  The VBS state is obtained

2888: by combining $n$ $n$-mer states \eqref{eq:nmer}, one for each

2889: $\mu=1,\dots,n$, in that the total spin on each lattice site is

2890: projected onto the symmetric representation $(n,0,\ldots,0)$.  This

2891: yields

2892: \begin{equation}

2893:   \label{eq:n000VBS}

2894:   \ket{\psi_{(n,0,\ldots,0)\, \text{VBS}}}=

2895:   \prod_{i} \Bigl(\!\!

2896:   \sum_{\scriptstyle{(\sigma_1,\ldots,\sigma_n)}=

2897:     \atop\scriptstyle{{\pi}({\rm f}_1,\ldots,{\rm f}_n)}} 

2898:   \hspace{-12pt}\hbox{sign}({\pi})

2899:   \prod_{\kappa=1}^n b^{\dagger}_{i-1+\kappa,\sigma_\kappa}

2900:   \Bigr)\!\vac\!.

2901: \end{equation}

2902: 

2903: Let us now construct a parent Hamiltonian for the symmetric VBS

2904: \eqref{eq:n000VBS}.  The total SU($n$) spin of two neighboring sites

2905: of a representation $(n,0,\ldots,0)$ spin chain in general

2906: contains all the representations corresponding to Young tableaux 

2907: with $2n$ boxes and at most two rows, \ie all tableaux of the form

2908: \begin{displaymath}

2909: \setlength{\unitlength}{8pt}

2910: \begin{picture}(9,5)(0,1.5)

2911: \put(0,6){\line(1,0){9}}

2912: \put(0,5){\line(1,0){9}}

2913: \put(0,4){\line(1,0){5}}

2914: \put(0,4){\line(0,1){2}}

2915: \put(1,4){\line(0,1){2}}

2916: \put(4,4){\line(0,1){2}}

2917: \put(5,4){\line(0,1){2}}

2918: \put(6,5){\line(0,1){1}}

2919: \put(8,5){\line(0,1){1}}

2920: \put(9,5){\line(0,1){1}}

2921: \put(2.5,2.5){\makebox(0,0)%

2922: {$\underbrace{\quad\text{~~~}\qquad}_{\text{\small $n-\nu$~columns}\qquad\ }$}}

2923: \put(7,2.5){\makebox(0,0)%

2924: {$\underbrace{\quad\text{~\,~}\quad}_{\quad\text{\small $2\nu$ boxes}}$}}

2925: %\put(17.5,22.5){\makebox(0,0){$\cdots$}}

2926: %\put(17.5,27.5){\makebox(0,0){$\cdots$}}

2927: \end{picture}

2928: \end{displaymath}

2929: The eigenvalues of the quadratic Casimir operator for these

2930: representations are given by

2931: \begin{equation}

2932:   \label{eva:aklt}

2933:   g_n(\nu)\equiv \casin{2\nu,n\!-\!\nu,0,\dots,0}=2n^2-4n+\nu(\nu+1). 

2934: \end{equation}

2935: 

2936: On the other hand, the total SU($n$) spin of two neighboring sites of

2937: the representation $(n,0,\ldots,0)$ VBS \eqref{eq:n000VBS} has to be

2938: contained in the product

2939: \begin{equation}

2940:   \label{eq:sunprod}

2941:   \setlength{\unitlength}{8pt} 

2942:   \begin{picture}(1.2,1.7)(0,0.6)

2943:     \multiput(0,0)(1,0){2}{\line(0,1){2}}

2944:     \multiput(0,0)(0,1){3}{\line(1,0){1}}

2945:   \end{picture}

2946:   ^{\otimes n-1}\otimes\

2947:   \begin{picture}(1,1)(0,0.1)

2948:     \multiput(0,0)(1,0){2}{\line(0,1){1}}

2949:     \multiput(0,0)(0,1){2}{\line(1,0){1}}

2950:   \end{picture}

2951:   \ \otimes\

2952:   \begin{picture}(1,1)(0,0.1)

2953:     \multiput(0,0)(1,0){2}{\line(0,1){1}}

2954:     \multiput(0,0)(0,1){2}{\line(1,0){1}}

2955:   \end{picture}

2956: \end{equation}

2957: As we project the spin on each lattice site onto the representation

2958: $(n,0,\ldots,0)$, only two these representations remain:

2959: \begin{displaymath}

2960:  \setlength{\unitlength}{8pt}

2961: \begin{picture}(19,2.5)(0,.6)

2962: \linethickness{0.3pt}

2963: \multiput(0,0)(1,0){2}{\line(0,1){2}}

2964: \multiput(4,0)(1,0){2}{\line(0,1){2}}

2965: \multiput(6,1)(1,0){2}{\line(0,1){1}}

2966: \multiput(0,1)(0,1){2}{\line(1,0){7}}

2967: \put(0,0){\line(1,0){5}}

2968: \put(8,0.5){\makebox(3,1){and}}

2969: \multiput(12,0)(1,0){2}{\line(0,1){2}}

2970: \multiput(16,0)(1,0){3}{\line(0,1){2}}

2971: %\multiput(12,0)(1,0){7}{\line(0,1){2}}

2972: \multiput(12,0)(0,1){3}{\line(1,0){6}}

2973: \end{picture}

2974: \end{displaymath}

2975: The eigenvalues of the quadratic Casimir operator are given by

2976: $g_n(0)=2n(n-2)$ and $g_n(1)=2(n-1)^2$, respectively.  Hence, using

2977: $\bs{J}^2_i=n(n-1)$, we obtain the parent Hamiltonian 

2978: %\begin{multline}

2979: \begin{equation}

2980:   \label{ham.aklt.sun}

2981:   \begin{split}

2982:   &H_{{\scriptstyle (n,0,\ldots,0)}\, {\scriptstyle \text{VBS}}}\\

2983: %  H_{\text{SU($n$)\hspace{2pt}VBS}}

2984:   &=\sum_{i=1}^N

2985:   \Bigl(\bigl(\bs{J}_i\bs{J}_{i+1}\bigr)^2 + (2n-1)\,\bs{J}_i\bs{J}_{i+1}+

2986:   n(n-1)\Bigr).

2987:   \end{split}

2988: \end{equation}

2989: %\end{multline}

2990: Since $g_n(0)\ge 0$ for $n\ge 2$ and $g_n(\nu)$ is a strictly

2991: increasing function of $\nu$, the Hamiltonian \eqref{ham.aklt.sun} is

2992: positive semi-definite.  For $n=2$, we recover the AKLT model

2993: \eqref{eq:haklt}; for $n=3$, we recover the $\bs{10}$ VBS model

2994: \eqref{ham.10VBS}.

2995: 

2996: \subsection{An example of a matrix product state}

2997: \label{sec:mp}

2998: 

2999: In principle, a matrix product VBS can be formulated on all SU($n$)

3000: chains with spins transforming under the symmetric combination of any

3001: representation $\bs{m}$ and its conjugate representation

3002: $\bs{\overline{m}}$.  Unless the rep.~$\bs{m}$ is self-conjugate,

3003: %\ie $\bs{m}=\bs{\overline{m}}$, 

3004: we obtain two inequivalent states, which transform into each other

3005: under space reflection.  The construction of these is analogous to the

3006: $\bs{8}$ VBS introduced above, and likewise best illustrated as

3007: \begin{equation}

3008: \setlength{\unitlength}{1.333pt}

3009: \begin{picture}(170,52)(0,-24)

3010: \linethickness{0.8pt}

3011: \multiput(5,10)(30,0){3}{\makebox(0,0){\small $\bs{m}$}}

3012: \multiput(20,10.6)(30,0){3}{\makebox(0,0){\small $\bs{\overline{m}}$}}

3013: \multiput(5,0.6)(30,0){3}{\makebox(0,0){\small $\bs{\overline{m}}$}}

3014: \multiput(20,0)(30,0){3}{\makebox(0,0){\small $\bs{m}$}}

3015: \multiput(128,10)(30,0){2}{\makebox(0,0){\small $\bs{m}$}}

3016: \multiput(143,10.6)(30,0){1}{\makebox(0,0){\small $\bs{\overline{m}}$}}

3017: \multiput(128,0.6)(30,0){2}{\makebox(0,0){\small $\bs{\overline{m}}$}}

3018: \multiput(143,0)(30,0){1}{\makebox(0,0){\small $\bs{m}$}}

3019: %

3020: \thicklines

3021: \multiput(9,10)(30,0){3}{\line(1,0){7}}

3022: \multiput(24,0)(30,0){2}{\line(1,0){7}}

3023: \multiput(147,10)(30,0){1}{\line(1,0){7}}

3024: \multiput(132,0)(30,0){1}{\line(1,0){7}}

3025: %

3026: \put(1,0){\line(-1,0){3}}

3027: \put(84,0){\line(1,0){3}}

3028: \put(124,10){\line(-1,0){3}}

3029: \put(162,0){\line(1,0){3}}

3030: %

3031: \thinlines

3032: \put(104,5){\makebox(0,0){and}}

3033: \put(169,2){\makebox(0,0){.}}

3034: %

3035: \put(44,-6){\framebox(12,22)}

3036: \put(50,-6){\line(0,-1){7}}

3037: \put(50,-20){\makebox(1,1){\small projection onto the symmetric}}

3038: \put(50,-28){\makebox(1,1){\small combination in 

3039: $\bs{m}\otimes\bs{\overline{m}}$}}

3040: \put(50,23){\makebox(1,1){\small one site}}

3041: \put(122.5,-6){\dashbox{2}(26,22)}

3042: \end{picture}

3043: \vspace{10pt}

3044: \label{fig:mmbarVBS}

3045: \end{equation}

3046: The thick lines indicate that we combine pairs of neighboring

3047: representations $\bs{m}$ and $\bs{\overline{m}}$ into singlets.  On

3048: each lattice site, we project onto the symmetric combination of

3049: $\bs{m}$ and $\bs{\overline{m}}$, as indicated.  By ``symmetric

3050: combination'' we mean that if representations $\bs{m}$ and

3051: $\bs{\overline{m}}$ of SU($n$) have Dynkin coordinates

3052: $(\mu_1,\mu_2,\ldots,\mu_{n-1})$ and

3053: $(\mu_{n-1},\mu_{n-2},\ldots,\mu_1)$, respectively, we combine them

3054: into the representation with Dynkin coordinates

3055: $(\mu_1\!+\!\mu_{n-1},\mu_2\!-\!\mu_{n-2},\ldots,\mu_{n-1}\!+\!\mu_1)$.

3056: In other words, we align the columns of both tableaux horizontally,

3057: and hence obtain a tableau with twice the width, without ever adding a

3058: single box vertically to a column of the tableaux we started with.

3059: %

3060: The states \eqref{fig:mmbarVBS} we obtain are translationally

3061: invariant and we expect the parent Hamiltonians to involve

3062: nearest-neighbor interactions only.

3063: 

3064: In this subsection, we will formulate the simplest SU($n$) model

3065: of this general family.  We take $\bs{m}$ to be the representation

3066: formed by antisymmetrically combining $\kappa\le \frac{n}{2}$ fundamental 

3067: representations,

3068: \begin{displaymath}

3069: \bs{m}\;=\;[\kappa]\;\equiv\; (0,\ldots,0,1,0,\ldots,0)\;=\;

3070: \setlength{\unitlength}{8pt}

3071: \begin{picture}(7,3)(0.7,5.2)

3072: \put(1,3.5){\line(0,1){4}}

3073: \put(2,3.5){\line(0,1){4}}

3074: %

3075: \put(1,3.5){\line(1,0){1}}

3076: \put(1,4.5){\line(1,0){1}}

3077: \put(1,6.5){\line(1,0){1}}

3078: \put(1,7.5){\line(1,0){1}}

3079: \put(5,5.5){\makebox(0,0){$\left.\begin{array}{c}~\\~\end{array}\right\}

3080: {\small ~\kappa~\text{boxes}}$\ ,}}

3081: \put(-7,3.6){\line(0,1){0.8}}

3082: \put(-7,2.6){\makebox(0,0){\small $\kappa$-th entry}}

3083: \end{picture}\\[20pt]

3084: \end{displaymath}

3085: which implies that we consider a model with the representation

3086: corresponding to a Young tableaux with a column with $n-\kappa$ boxes

3087: to the left of a column with $\kappa$ boxes at each lattice site:\\

3088: %[\kappa ,n\!-\!\kappa]\;\equiv\;

3089: %(0,\ldots,0,1,0,\ldots,0,1,0,\ldots,0)\;=\; 

3090: \setlength{\unitlength}{8pt}

3091: \begin{picture}(8,8)(-14,0)

3092: \put(-7,3.8){$[\kappa ,n\!-\!\kappa]\;\equiv$}

3093: \put(0,0){\line(0,1){7.5}}

3094: \put(1,0){\line(0,1){7.5}}

3095: \put(2,3.5){\line(0,1){4}}

3096: %

3097: \put(0,0){\line(1,0){1}}

3098: \put(0,1){\line(1,0){1}}

3099: \put(0,2.5){\line(1,0){1}}

3100: \put(0,3.5){\line(1,0){2}}

3101: \put(0,4.5){\line(1,0){2}}

3102: \put(0,6.5){\line(1,0){2}}

3103: \put(0,7.5){\line(1,0){2}}

3104: \put(4.5,5.5){\makebox(0,0){$\left.\begin{array}{c}~\\~\end{array}\right\}

3105: {\small ~\kappa~\text{rows}}$}}

3106: %\put(9,3.6){.}

3107: \end{picture}%\\[26pt]

3108: 

3109: The construction of the parent Hamiltonian is similar to the $n$-mer

3110: model above.  The total spin on two neighboring lattice sites can only

3111: assume representations contained in $\bs{m}\otimes\bs{\overline{m}}$,

3112: \ie representations corresponding to tableaux of the form

3113: \begin{displaymath}

3114: [\nu ,n-\nu]\;=\;

3115: \setlength{\unitlength}{8pt}

3116: \begin{picture}(8.5,4)(-0.3,3.9)

3117: \put(0,0){\line(0,1){7.5}}

3118: \put(1,0){\line(0,1){7.5}}

3119: \put(2,4){\line(0,1){3.5}}

3120: %

3121: \put(0,0){\line(1,0){1}}

3122: \put(0,1){\line(1,0){1}}

3123: \put(0,3){\line(1,0){1}}

3124: \put(0,4){\line(1,0){2}}

3125: \put(0,5){\line(1,0){2}}

3126: \put(0,6.5){\line(1,0){2}}

3127: \put(0,7.5){\line(1,0){2}}

3128: \put(4.5,5.75){\makebox(0,0)

3129: {$\left.\begin{array}{c}~\\[-0.6pt]~\end{array}\right\}

3130: {\small ~\nu~\text{rows}}$}}

3131: \end{picture}\\[26pt] 

3132: \end{displaymath}

3133: with $0\le\nu\le\kappa$.  The eigenvalues of the quadratic

3134: Casimir operator for these representations are

3135: \begin{equation}

3136:   h_n(\nu)=\nu\,(n-\nu+1).

3137: \end{equation}

3138: The obvious proposal for a parent Hamiltonian is hence 

3139: \begin{equation}

3140: H=\sum_{i=1}^N H_i, \quad %\text{with}\quad

3141: H_i=\prod_{\nu=0}^{\lfloor\frac{n}{2}\rfloor}

3142: \left(\!\left(\bs{J}_i^{(2)}\right)^2 - h_n(\nu)\right),

3143: \label{eq:mmbarham}

3144: \end{equation}

3145: where $\lfloor x\rfloor$ denotes again the floor function.  This

3146: Hamiltonian singles out the matrix product state \eqref{fig:mmbarVBS}

3147: as unique ground states for $n\le 5$, but suffers from the same

3148: shortcomings as \eqref{eq:MGSUni} with \eqref{eq:MGSUn} for $n\ge 6$.

3149: 

3150: 

3151: \section{Spinon confinement and the Haldane gap}

3152: \label{sec:conf}

3153: 

3154: \subsection{The Affleck--Lieb theorem}

3155: \label{sec:afflecklieb}

3156: 

3157: For the generic SU(2) spin chain, Haldane~\cite{haldane83pla464,

3158:   haldane83prl1153} identified the O(3) nonlinear sigma model as the

3159: effective low energy field theory of SU(2) spin chains, and argued

3160: that chains with integer spin possess a gap in the excitation spectrum,

3161: %due to spinon confinement, 

3162: while a topological term renders

3163: half-integer spin chains gapless~\cite{affleck90proc,Fradkin91}.  The exact

3164: models for SU(2) spin chains we reviewed above, the MG and the AKLT

3165: chain, serve a paradigms to illustrate the general principle.

3166: Unfortunately, the effective field theory description of Haldane

3167: yielding the distinction between gapless half integer spin chains with

3168: deconfined spinons and gapped integer spin chains with confined

3169: spinons cannot be directly generalized to SU($n$) chains, as there is

3170: no direct equivalent of the CP$^1$ representation used in Haldane's

3171: analysis.

3172: 

3173: Nonetheless, there is a rigorous and rather general result for

3174: antiferromagnetic chains of SU($n$) spins transforming under a

3175: representation corresponding to a Young tableau with a number of boxes

3176: not divisible by $n$: Affleck and Lieb~\cite{affleck-86lmp57} showed

3177: that if the ground state is non-degenerate, and the Hamiltonian

3178: consists nearest-neighbor interactions only, than the gap in the

3179: excitation spectrum vanishes as $1/N$ (where $N$ is the number of

3180: sites) in the thermodynamic limit.  This result is fully consistent

3181: with the picture suggested by the models introduced above.  Like the

3182: MG model, the trimer model and the representation $\bs{6}$ VBS have

3183: degenerate ground states and interactions which extend beyond the

3184: nearest neighbor, which implies that the theorem is not directly

3185: applicable.

3186: 

3187: On physical grounds, however, the statement that a given model is

3188: gapless (\ie the excitation gap vanishes in the thermodynamic limit)

3189: implies that the spinons are deconfined.  The reason is simply that if

3190: there was a confinement force between them, the zero-point energy

3191: associated with the quantum mechanical oscillator of the relative

3192: motion between the spinons would inevitably yield an energy gap.  The

3193: MG, the trimer and the $\bs{6}$ VBS model constitute pedagogically valuable

3194: paradigms of deconfined spinons.  Since the excitations in these

3195: models are literally domain walls between different ground states, no

3196: long-range forces can exist between them.

3197: 

3198: %\subsection{A category of models with spinon confinement}

3199: \subsection{A general criterion for spinon confinement}

3200: \label{sec:generalcriterion}

3201: 

3202: More importantly, however, the exact models we introduced above

3203: provide information about the models of SU($n$) spin chains with

3204: representations corresponding to Young tableaux with a number of boxes

3205: divisible by $n$, \ie models for which the Affleck--Lieb theorem is

3206: not applicable.  We have studied two SU(3) models belonging to this

3207: family 

3208: %, which we refer to as the singlet family since it contains the

3209: %singlet rep.~$\bs{1}$, 

3210: in Sec.~\ref{sec:vbs}, the rep.~$\bs{10}$ VBS

3211: and the rep.~$\bs{8}$ VBS, and found that both have confined

3212: spinons or colorons and hence display a Haldane-type gap in the

3213: spectrum.  

3214: 

3215: In this section, we will argue that 

3216: %\begin{quote}

3217: %\emph{

3218: models of antiferromagnetic chains of SU($n$) spins transforming under

3219: a representation corresponding to a Young tableau consisting of a

3220: number of boxes $\lambda$ divisible by $n$ generally possess a

3221: Haldane-type gap due to spinon confinement forces.

3222: %}

3223: %\end{quote}

3224: 

3225: We should caution immediately that our argument is based on several

3226: assumptions, which we consider reasonable, but which we are ultimately

3227: unable to prove.

3228: 

3229: The first, and also the most crucial, is the assumption that the

3230: question of whether a given model supports free spinon excitations can

3231: be resolved through study of the corresponding VBS state.  This

3232: assumption definitely holds for SU(2) spin chains, where the MG model

3233: for $S=\frac{1}{2}$ indicates that the spinons are free, while the

3234: AKLT model for $S=1$ serves as a paradigm for spinon confinement and

3235: hence the Haldane gap.  The general conclusions we derived from our

3236: studies of the SU(3) VBSs above rely on this assumption.  The

3237: numerical results we reported on the rep.~$\bs{8}$ VBS provide

3238: evidence that this assumption holds at least for this model.

3239: 

3240: Let us consider an SU($n$) spin chain with spins transforming under

3241: a representation corresponding to Young tableau consisting of $L$ columns

3242: with $\kappa_1\le\kappa_2\le\ldots\le\kappa_L<n$ boxes each,

3243: %\pagebreak

3244: \begin{equation}

3245:   \label{eq:rep}

3246:   [\kappa_1,\kappa_2,\ldots,\kappa_L]\;\equiv\;

3247: \setlength{\unitlength}{8pt}

3248: \begin{picture}(9,4)(0,3)

3249: \put(0,0){\line(1,0){1}}

3250: \multiput(0,1)(0,1){2}{\line(1,0){2}}

3251: %\put(2,2.25){\line(1,0){2.75}}

3252: \put(2,2.45){\line(1,0){2.75}}

3253: \put(4.75,3.75){\line(1,0){1.75}}

3254: \put(6.5,5){\line(1,0){2.25}}

3255: \put(7.75,6){\line(1,0){3}}

3256: \put(0,7){\line(1,0){10.75}}

3257: \put(0,6){\line(1,0){2}}

3258: \multiput(0,0)(1,0){2}{\line(0,1){7}}

3259: \put(2,1){\line(0,1){6}}

3260: %\put(4.75,3.75){\line(0,-1){1.5}}

3261: \put(4.75,3.75){\line(0,-1){1.3}}

3262: \put(6.5,5){\line(0,-1){1.25}}

3263: \multiput(9.75,6)(1,0){2}{\line(0,1){1}}

3264: \multiput(7.75,5)(1,0){2}{\line(0,1){2}}

3265: \put(0.55,-1){\makebox(0,0){\small $\kappa_L$}}

3266: %\put(9.35,4){\makebox(0,0){\small $\kappa_3\kappa_2\kappa_1$}}

3267: \put(8.25,4){\makebox(0,0){\small $\kappa_3$}}

3268: \put(10,4.3){\makebox(0,0){\small $\kappa_2\kappa_1$}}

3269: %\put(13,-0.8){\makebox(0,0){\small boxes}}

3270: \end{picture} \\[27pt] \phantom{oooops} 

3271: \end{equation}

3272: with a total number of boxes

3273: \begin{displaymath}

3274:   \lambda=\sum_{l=1}^L \kappa_l

3275: \end{displaymath}

3276: divisible by $n$.  We denote the Dynkin coordinates of this

3277: representation by $(\mu_1,\mu_2,\ldots,\mu_{n-1})$, which implies

3278: \begin{displaymath}

3279:   \sum_{i=1}^{n-1}\mu_i=L.

3280: \end{displaymath}

3281: Note that this representation is, by definition, given by the 

3282: maximally symmetric component of the tensor product of the individual

3283: columns,

3284: \begin{equation}

3285:   \label{eq:sym}

3286:   [\kappa_1,\kappa_2,\ldots,\kappa_L]=\mathcal{S}

3287:   \big([\kappa_1]\otimes [\kappa_2]\otimes\ldots\otimes [\kappa_L ]\big).

3288: \end{equation}

3289: For convenience, we denote the $\left({n \atop \kappa}\right)$

3290: dimensional representation $[\kappa]$ in this subsection as

3291: $\bs{\kappa}_l\equiv [\kappa_l]=

3292: \setlength{\unitlength}{1.25pt}

3293: \begin{picture}(10,3)(0,-2)

3294: \put(5,0){\circle{8}}

3295: \put(5,0){\makebox(0,0){\small $l$}}

3296: \end{picture}$.

3297: 

3298: Since $\lambda$ is divisible by $n$, it will always be possible to

3299: obtain a singlet from the complete sequence of representations

3300: $\bs{\kappa}_1,\bs{\kappa}_2,\ldots,\bs{\kappa}_L$ by combining them

3301: antisymmetrically.  To be precise, when we write that we combine

3302: representations $\bs{\kappa_1}$ and $\bs{\kappa_2}$ antisymmetrically, we

3303: mean we obtain a new representation $[\kappa_1+\kappa_2]$ by

3304: stacking the two columns with $\kappa_1$ and $\kappa_2$ boxes on

3305: top of each other if $\kappa_1+\kappa_2<n$, and a new representation 

3306: $[\kappa_1+\kappa_2-n]$ if $\kappa_1+\kappa_2\ge n$.  In equations,

3307: we write this as

3308: \begin{equation}

3309:   \label{eq:asym}

3310:   \mathcal{A}

3311:   \big([\kappa_1]\otimes [\kappa_2]\big)\equiv

3312:   \begin{cases}

3313:     [\kappa_1+\kappa_2]&\text{for}\ \kappa_1+\kappa_2<n\\

3314:     [\kappa_1+\kappa_2-n]&\text{for}\ \kappa_1+\kappa_2\ge n

3315:   \end{cases}

3316: \end{equation}

3317: Following the notation used above, we indicate the

3318: antisymmetric combination of representations $\bs{\kappa}_l$ by a line

3319: connecting them.  In particular, we depict the singlet formed by

3320: combining $\bs{\kappa}_1,\bs{\kappa}_2,\ldots,\bs{\kappa}_L$ on

3321: different lattice sites as

3322: \begin{equation}

3323: \label{eq:kappasinglet}

3324:   \setlength{\unitlength}{1.25pt}

3325: \begin{picture}(90,0)(0,-1)

3326: \linethickness{0.8pt}

3327: \multiput(5,0)(20,0){3}{{\circle{8}}}

3328: \multiput(85,0)(20,0){1}{{\circle{8}}}

3329: \put(5,0){\makebox(0,0){\small 1}}

3330: \put(25,0){\makebox(0,0){\small 2}}

3331: \put(45,0){\makebox(0,0){\small 3}}

3332: \put(65,0){\makebox(0,0){\small \ldots}}

3333: \put(85,0){\makebox(0,0){\small $L$}}

3334: %

3335: \thicklines

3336: \multiput(15,0)(20,0){2}{\makebox(0,0){\line(2,0){12}}}

3337: \put(49,0){\line(1,0){8}}

3338: \put(81,0){\line(-1,0){8}}

3339: %

3340: \end{picture}

3341: \vspace{10pt}

3342: \end{equation}

3343: The understanding here is that we combine them in the order indicated

3344: by the line, \ie in \eqref{eq:kappasinglet} we first combine

3345: $\bs{\kappa}_1$ and $\bs{\kappa}_2$, then we combine the result with

3346: $\bs{\kappa}_3$, and so on.  Depending on the order of the

3347: representations $\bs{\kappa}_l$ on the line, we obtain different, but

3348: not necessarily orthogonal, singlets.  We assume that it is irrelevant

3349: whether we combine the representations starting from the left or from

3350: the right of the line, as the resulting state will not depend on it.

3351: 

3352: In general, it will be possible to construct a number $\le\lambda/n$

3353: of singlets out of various combinations of the $\bs{\kappa}$'s, one

3354: for each block of $\bs{\kappa}$'s for which the values of $\kappa$ add

3355: up to $n$ as we combine the representation in the order described above.  

3356: In this case, we will be able to construct one VBS for

3357: each singlet, and subsequently combine them at each site symmetrically

3358: to obtain the desired representation \eqref{eq:rep}.  The argument

3359: for spinon confinement we construct below will hold for each of the

3360: individual VBSs, and hence for the combined VBS as well.  It is hence

3361: sufficient for our purposes to consider situations where the entire

3362: sequence $\bs{\kappa}_1,\bs{\kappa}_2,\ldots,\bs{\kappa}_L$

3363: is needed to construct a singlet.

3364: 

3365: A possible VBS ``ground state'' for a representation corresponding to a

3366: Young tableau with $L=4$ columns is depicted below.

3367: \begin{equation}

3368:   \setlength{\unitlength}{1.25pt}

3369: %  \setlength{\unitlength}{1.3pt}

3370: \begin{picture}(160,70)(0,-15)

3371: \linethickness{0.8pt}

3372: \multiput(0,40)(20,0){9}{{\circle{7}}}

3373: \multiput(0,40)(20,0){9}{\makebox(0,0){\small 1}}

3374: \multiput(0,30)(20,0){9}{{\circle{7}}}

3375: \multiput(0,30)(20,0){9}{\makebox(0,0){\small 2}}

3376: \multiput(0,20)(20,0){9}{{\circle{7}}}

3377: \multiput(0,20)(20,0){9}{\makebox(0,0){\small 3}}

3378: \multiput(0,10)(20,0){9}{{\circle{7}}}

3379: \multiput(0,10)(20,0){9}{\makebox(0,0){\small 4}}

3380: %

3381: \thicklines

3382: \multiput(3.5,38.25)(20,0){8}{\line(2,-1){13}}

3383: \multiput(3.5,28.25)(20,0){8}{\line(2,-1){13}}

3384: \multiput(3.5,18.25)(20,0){8}{\line(2,-1){13}}

3385: %

3386: \thinlines

3387: \put(54,4){\framebox(12,42)}

3388: \put(60,4){\line(0,-1){7}}

3389: \put(60,-10){\makebox(1,1){\small projection onto representation}}

3390: %\put(60,-18){\makebox(1,1){\small $[\kappa_1,\kappa_2,\ldots,\kappa_L]$}}

3391: \put(60,-18){\makebox(1,1){\small $[\kappa_1,\kappa_2,\kappa_3,\kappa_4]$}}

3392: \put(60,55){\makebox(1,1){\small one site}}

3393: \end{picture}

3394: %\vspace{10pt}

3395: %\nonumber

3396: \label{eq:VBS1}

3397: \end{equation}

3398: In general, there are as many inequivalent VBS ``ground states'' as

3399: there are inequivalent ways to order the representations

3400: $\bs{\kappa}_1,\bs{\kappa}_2,\ldots,\bs{\kappa}_L$, \ie

3401: the number of inequivalent VBS ``ground states'' is given by

3402: \begin{displaymath}

3403:   \frac{L!}{\mu_1!\cdot\mu_2!\cdot\ldots\cdot\mu_{n-1}!}.

3404: \end{displaymath}

3405: To give an example, the following VBS

3406: \begin{equation}

3407:   \setlength{\unitlength}{1.25pt}

3408: %  \setlength{\unitlength}{1.3pt}

3409: \begin{picture}(160,70)(0,-15)

3410: \linethickness{0.8pt}

3411: \multiput(0,40)(20,0){9}{{\circle{7}}}

3412: \multiput(0,40)(20,0){9}{\makebox(0,0){\small 2}}

3413: \multiput(0,30)(20,0){9}{{\circle{7}}}

3414: \multiput(0,30)(20,0){9}{\makebox(0,0){\small 1}}

3415: \multiput(0,20)(20,0){9}{{\circle{7}}}

3416: \multiput(0,20)(20,0){9}{\makebox(0,0){\small 3}}

3417: \multiput(0,10)(20,0){9}{{\circle{7}}}

3418: \multiput(0,10)(20,0){9}{\makebox(0,0){\small 4}}

3419: %

3420: \thicklines

3421: \multiput(3.5,38.25)(20,0){8}{\line(2,-1){13}}

3422: \multiput(3.5,28.25)(20,0){8}{\line(2,-1){13}}

3423: \multiput(3.5,18.25)(20,0){8}{\line(2,-1){13}}

3424: %

3425: \thinlines

3426: \put(54,4){\framebox(12,42)}

3427: \put(60,4){\line(0,-1){7}}

3428: \put(60,-10){\makebox(1,1){\small projection onto representation}}

3429: %\put(60,-18){\makebox(1,1){\small $[\kappa_1,\kappa_2,\ldots,\kappa_L]$}}

3430: \put(60,-18){\makebox(1,1){\small $[\kappa_1,\kappa_2,\kappa_3,\kappa_4]$}}

3431: \put(60,55){\makebox(1,1){\small one site}}

3432: \end{picture}

3433: %\vspace{10pt}

3434: %\nonumber

3435: \label{eq:VBS2}

3436: \end{equation}

3437: is inequivalent to the one above if and only if $\kappa_1\ne\kappa_2$.

3438: Note that all these VBS ``ground states'' states are translationally

3439: invariant.  We expect some of these states, but not all of them,

3440: degenerate in energy for the appropriate parent Hamiltonian, and have

3441: accordingly written ``ground states'' in quotation marks.  For example,

3442: if we form a SU(4) VBS with representation 

3443: $[\kappa_1,\kappa_2,\kappa_3]=[1,1,2]$, the

3444: combination 

3445: \begin{equation}

3446:   \setlength{\unitlength}{1.25pt}

3447: \begin{picture}(50,0)(0,-1)

3448: \linethickness{0.8pt}

3449: \multiput(5,0)(20,0){3}{{\circle{8}}}

3450: \put(5,0){\makebox(0,0){\small 1}}

3451: \put(25,0){\makebox(0,0){\small 3}}

3452: \put(45,0){\makebox(0,0){\small 2}}

3453: %

3454: \thicklines

3455: \multiput(15,0)(20,0){2}{\makebox(0,0){\line(2,0){12}}}

3456: \end{picture}\nonumber

3457: \vspace{4pt}

3458: \end{equation}

3459: might yield a state with a lower energy for the appropriate

3460: Hamiltonian than the state formed by combining

3461: \begin{equation}

3462:   \setlength{\unitlength}{1.25pt}

3463: \begin{picture}(50,0)(0,-1)

3464: \linethickness{0.8pt}

3465: \multiput(5,0)(20,0){3}{{\circle{8}}}

3466: \put(5,0){\makebox(0,0){\small 1}}

3467: \put(25,0){\makebox(0,0){\small 2}}

3468: \put(45,0){\makebox(0,0){\small 3}}

3469: %

3470: \thicklines

3471: \multiput(15,0)(20,0){2}{\makebox(0,0){\line(2,0){12}}}

3472: \end{picture}. \nonumber

3473: \vspace{4pt}

3474: \end{equation}

3475: Simple examples where we have only two inequivalent VBS ground states

3476: are provided by the matrix product states discussed in

3477: Secs.~\ref{sec:8VBS} and \ref{sec:mp}.

3478: 

3479: We will now argue that the elementary excitations of the corresponding

3480: VBS models are always confined.  To this end, we first observe that

3481: any domain wall between translationally invariant ``ground states''

3482: consists of a total of $m\cdot n$ representations $\bs{\kappa}_l$ ($m$

3483: integer).  To illustrate this, consider a domain wall between the

3484: ``ground states'' depicted in \eqref{eq:VBS2} and \eqref{eq:VBS1}:

3485: \begin{equation}

3486:   \setlength{\unitlength}{1.25pt}

3487: %  \setlength{\unitlength}{1.3pt}

3488: \begin{picture}(170,49)(20,-2)

3489: \linethickness{0.8pt}

3490: \multiput(20,40)(20,0){2}{{\circle{7}}}

3491: \multiput(20,40)(20,0){2}{\makebox(0,0){\small 2}}

3492: \multiput(80,40)(20,0){6}{{\circle{7}}}

3493: \multiput(80,40)(20,0){6}{\makebox(0,0){\small 1}}

3494: \multiput(20,30)(20,0){3}{{\circle{7}}}

3495: \multiput(20,30)(20,0){3}{\makebox(0,0){\small 1}}

3496: \multiput(100,30)(20,0){5}{{\circle{7}}}

3497: \multiput(100,30)(20,0){5}{\makebox(0,0){\small 2}}

3498: \multiput(20,20)(20,0){4}{{\circle{7}}}

3499: \multiput(20,20)(20,0){4}{\makebox(0,0){\small 3}}

3500: \multiput(120,20)(20,0){4}{{\circle{7}}}

3501: \multiput(120,20)(20,0){4}{\makebox(0,0){\small 3}}

3502: \multiput(20,10)(20,0){5}{{\circle{7}}}

3503: \multiput(20,10)(20,0){5}{\makebox(0,0){\small 4}}

3504: \multiput(140,10)(20,0){3}{{\circle{7}}}

3505: \multiput(140,10)(20,0){3}{\makebox(0,0){\small 4}}

3506: %

3507: %{\color{blue}

3508: \put(60,40){{\circle*{7}}}

3509: \put(80,30){{\circle*{7}}}

3510: \put(100,20){{\circle*{7}}}

3511: \put(120,10){{\circle*{7}}}

3512: %}

3513: {\color{white}

3514: \put(60,40){\makebox(0,0){\small\bf 2}}

3515: \put(80,30){\makebox(0,0){\small\bf 2}}

3516: \put(100,20){\makebox(0,0){\small\bf 3}}

3517: \put(120,10){\makebox(0,0){\small\bf 4}}

3518: }

3519: % 

3520: \thicklines

3521: \multiput(23.5,38.25)(20,0){2}{\line(2,-1){13}}

3522: \multiput(23.5,28.25)(20,0){2}{\line(2,-1){13}}

3523: \multiput(23.5,18.25)(20,0){2}{\line(2,-1){13}}

3524: %

3525: %{\color{red}

3526: \put(63.5,28.25){\line(2,-1){13}}

3527: \put(63.5,18.25){\line(2,-1){13}}

3528: %}

3529: \multiput(83.5,38.25)(20,0){5}{\line(2,-1){13}}

3530: \multiput(83.5,28.25)(20,0){5}{\line(2,-1){13}}

3531: \multiput(83.5,18.25)(20,0){5}{\line(2,-1){13}}

3532: \end{picture}

3533: %\vspace{10pt}

3534: \nonumber

3535: \end{equation}

3536: In the example, the domain wall consists of reps.~$\bs{\kappa}_2$,

3537: $\bs{\kappa}_2$, $\bs{\kappa}_3$, and $\bs{\kappa}_4$.  If the

3538: translationally invariant states on both sites are true ground states,

3539: the domain wall is likely to correspond to two elementary excitations:

3540: a rep.~$\bs{\bar{\kappa}}_1$ spinon consisting of an antisymmetric

3541: combination of a $\bs{\kappa}_2$, a $\bs{\kappa}_3$, and a

3542: $\bs{\kappa}_4$, as indicated by the line in the drawing, and another

3543: rep.~$\bs{\kappa}_2$ spinon.  The reason we assume that

3544: $\bs{\kappa}_2$, $\bs{\kappa}_3$, and $\bs{\kappa}_4$ form a

3545: rep.~$\bs{\bar{\kappa}}_1$ is simply that this combination is present

3546: in both ground states on either side, and hence bound to be the

3547: energetically most favorable combination.  The second

3548: $\bs{\kappa}_2$, however, is not part of this elementary excitation,

3549: as combining it antisymmetrically with the others (\ie the

3550: $\bs{\bar{\kappa}}_1$) would induce correlations which are not present

3551: in the ground state.  We hence conclude that the second

3552: $\bs{\kappa}_2$ is an elementary excitation as well.  The domain wall

3553: depicted above consists of a spinon transforming under

3554: rep.~$\bs{\bar{\kappa}}_1$ and an anti-spinon transforming under

3555: rep.~$\bs{\kappa}_2$.

3556: 

3557: The next step in our argument is to notice that the spinon and the

3558: complementary particle created simultaneously which may either be its

3559: anti-particle or some other spinon, are confined through a linear

3560: potential.  To see this, we pull them apart and look at the state

3561: inbetween (color online):

3562: \begin{equation}

3563:   \setlength{\unitlength}{1.25pt}

3564: %  \setlength{\unitlength}{1.3pt}

3565: \begin{picture}(180,73)(0,-10)

3566: \linethickness{0.8pt}

3567: \multiput(0,40)(20,0){1}{{\circle{7}}}

3568: \multiput(0,40)(20,0){1}{\makebox(0,0){\small 2}}

3569: \multiput(40,40)(20,0){8}{{\circle{7}}}

3570: \multiput(40,40)(20,0){8}{\makebox(0,0){\small 1}}

3571: \multiput(0,30)(20,0){2}{{\circle{7}}}

3572: \multiput(0,30)(20,0){2}{\makebox(0,0){\small 1}}

3573: \multiput(40,30)(20,0){4}{{\circle{7}}}

3574: \multiput(40,30)(20,0){4}{\makebox(0,0){\small 2}}

3575: \multiput(140,30)(20,0){3}{{\circle{7}}}

3576: \multiput(140,30)(20,0){3}{\makebox(0,0){\small 2}}

3577: \multiput(0,20)(20,0){7}{{\circle{7}}}

3578: \multiput(0,20)(20,0){7}{\makebox(0,0){\small 3}}

3579: \multiput(160,20)(20,0){2}{{\circle{7}}}

3580: \multiput(160,20)(20,0){2}{\makebox(0,0){\small 3}}

3581: \multiput(0,10)(20,0){8}{{\circle{7}}}

3582: \multiput(0,10)(20,0){8}{\makebox(0,0){\small 4}}

3583: \multiput(180,10)(20,0){1}{{\circle{7}}}

3584: \multiput(180,10)(20,0){1}{\makebox(0,0){\small 4}}

3585: %

3586: %{\color{blue}

3587: \put(20,40){{\circle*{7}}}

3588: \put(120,30){{\circle*{7}}}

3589: \put(140,20){{\circle*{7}}}

3590: \put(160,10){{\circle*{7}}}

3591: %}

3592: {\color{white}

3593: \put(20,40){\makebox(0,0){\small\bf 2}}

3594: \put(120,30){\makebox(0,0){\small\bf 2}}

3595: \put(140,20){\makebox(0,0){\small\bf 3}}

3596: \put(160,10){\makebox(0,0){\small\bf 4}}

3597: }

3598: % 

3599: %{\color{blue}

3600: \thicklines

3601: \put(123.5,28.25){\line(2,-1){13}}

3602: \put(143.5,18.25){\line(2,-1){13}}

3603: %}

3604: \thicklines

3605: \multiput(3.5,38.25)(20,0){1}{\line(2,-1){13}}

3606: \multiput(3.5,28.25)(20,0){2}{\line(2,-1){13}}

3607: \multiput(3.5,18.25)(20,0){2}{\line(2,-1){13}}

3608: %

3609: {\color{red}

3610: \multiput(43.5,38.25)(20,0){3}{\line(2,-1){13}}

3611: \multiput(43.5,28.25)(20,0){4}{\line(2,-1){13}}

3612: \multiput(43.5,18.25)(20,0){4}{\line(2,-1){13}}

3613: %

3614: % TWO ALTERNTIVES

3615: %

3616: \put(90,25){\makebox(0,0){\line(2,3){15.9}}}

3617: %\qbezier(43.5,31.75)(50,35)(70,35)

3618: %\qbezier(96.4,38.25)(90,35)(70,35)

3619: }

3620: %

3621: \multiput(123.5,38.25)(20,0){3}{\line(2,-1){13}}

3622: \multiput(123.5,28.25)(20,0){3}{\line(2,-1){13}}

3623: \multiput(123.5,18.25)(20,0){3}{\line(2,-1){13}}

3624: %

3625: \thinlines

3626: \put(20,56){\makebox(0,0){\small $\bs{\kappa}_2$-spinon}}

3627: \put(140,56){\makebox(0,0){\small $\bs{\bar{\kappa}}_1$-spinon}}

3628: %\put(20,56){\makebox(0,0){\small $[\kappa_2]$-spinon}}

3629: %\put(140,56){\makebox(0,0){\small $[n\!-\!\kappa_1]$-spinon}}

3630: \put(80,-2){\vector(-1,0){60}}

3631: \put(80,-2){\vector(1,0){60}}

3632: \put(80,-8){\makebox(0,0){\small energy cost $\propto$ distance}}

3633: \end{picture}

3634: %\vspace{10pt}

3635: \label{eq:kappaconf}

3636: \end{equation}

3637: The state between spinon and anti-spinon is not translationally

3638: invariant.  In the example, the unit cell of this state is depicted in

3639: red and consists of two regular bonds with three ``singlet lines''

3640: between the sites, one stronger bond with four lines (which cross in

3641: the cartoon), and one weaker bond with only two lines.  If we assume

3642: that the two states \eqref{eq:VBS2} and \eqref{eq:VBS1} on both sides

3643: are true ground states, it is clear that the irregularities in the

3644: strength of the bond correlations will cause the state between the

3645: spinon and anti-spinon to have a higher energy than either of them.

3646: This additional energy cost will induce a linear confinement potential

3647: between the spinons, and hence a linear oscillator potential for the

3648: relative motion of the particles.  As in the models studied above, the

3649: Haldane gap corresponds to the zero-point energy of this oscillator.

3650: 

3651: If one of the ``ground states'' to the left or to the right of the

3652: domain wall is not a true ground state, but a translationally

3653: invariant state corresponding to a higher energy than the ground

3654: state, there will be a confining force between this domain wall and

3655: another domain wall which transforms the intermediate ``ground state''

3656: with a higher energy back into a true ground state.  This force will

3657: be sufficient to account for a Haldane gap, regardless of the strength

3658: or existence of a confinement force between the constituent particles

3659: of each domain wall.

3660: 

3661: The lowest-lying excitations of a representation

3662: $[\kappa_1,\kappa_2,\ldots,\kappa_L]$ spin chain, however, will in

3663: general not be domain walls, but spinons created by breaking up one of

3664: the singlets \eqref{eq:kappasinglet} in a ground state.  This is 

3665: illustrated below for the ground state \eqref{eq:VBS1}:

3666: \begin{equation}

3667:   \setlength{\unitlength}{1.25pt}

3668: %  \setlength{\unitlength}{1.3pt}

3669: \begin{picture}(170,49)(20,-2)

3670: \linethickness{0.8pt}

3671: \multiput(20,40)(20,0){2}{{\circle{7}}}

3672: \multiput(20,40)(20,0){2}{\makebox(0,0){\small 1}}

3673: \multiput(80,40)(20,0){6}{{\circle{7}}}

3674: \multiput(80,40)(20,0){6}{\makebox(0,0){\small 1}}

3675: \multiput(20,30)(20,0){3}{{\circle{7}}}

3676: \multiput(20,30)(20,0){3}{\makebox(0,0){\small 2}}

3677: \multiput(100,30)(20,0){5}{{\circle{7}}}

3678: \multiput(100,30)(20,0){5}{\makebox(0,0){\small 2}}

3679: \multiput(20,20)(20,0){4}{{\circle{7}}}

3680: \multiput(20,20)(20,0){4}{\makebox(0,0){\small 3}}

3681: \multiput(120,20)(20,0){4}{{\circle{7}}}

3682: \multiput(120,20)(20,0){4}{\makebox(0,0){\small 3}}

3683: \multiput(20,10)(20,0){5}{{\circle{7}}}

3684: \multiput(20,10)(20,0){5}{\makebox(0,0){\small 4}}

3685: \multiput(140,10)(20,0){3}{{\circle{7}}}

3686: \multiput(140,10)(20,0){3}{\makebox(0,0){\small 4}}

3687: %

3688: %{\color{blue}

3689: \put(60,40){{\circle*{7}}}

3690: \put(80,30){{\circle*{7}}}

3691: \put(100,20){{\circle*{7}}}

3692: \put(120,10){{\circle*{7}}}

3693: %}

3694: {\color{white}

3695: \put(60,40){\makebox(0,0){\small\bf 1}}

3696: \put(80,30){\makebox(0,0){\small\bf 2}}

3697: \put(100,20){\makebox(0,0){\small\bf 3}}

3698: \put(120,10){\makebox(0,0){\small\bf 4}}

3699: }

3700: % 

3701: \thicklines

3702: \multiput(23.5,38.25)(20,0){2}{\line(2,-1){13}}

3703: \multiput(23.5,28.25)(20,0){2}{\line(2,-1){13}}

3704: \multiput(23.5,18.25)(20,0){2}{\line(2,-1){13}}

3705: %

3706: %{\color{red}

3707: \put(63.5,28.25){\line(2,-1){13}}

3708: \put(63.5,18.25){\line(2,-1){13}}

3709: %}

3710: \multiput(83.5,38.25)(20,0){5}{\line(2,-1){13}}

3711: \multiput(83.5,28.25)(20,0){5}{\line(2,-1){13}}

3712: \multiput(83.5,18.25)(20,0){5}{\line(2,-1){13}}

3713: \end{picture}

3714: %\vspace{10pt}

3715: \nonumber

3716: \end{equation}

3717: In the example, we have created a spinon transforming under

3718: rep.~$\bs{\bar{\kappa}}_1$ and its anti-particle, a spinon

3719: transforming under rep.~$\bs{\kappa}_1$.  This is, however, irrelevant

3720: to the argument---we may break the singlet in any way we like.

3721: The important feature is that we obtain, by construction, at least two

3722: excitations, and that these are confined.  In our specific example, the 

3723: confining potential is equivalent to the confining potential in

3724: \eqref{eq:kappaconf} above (color online):

3725: \begin{equation}

3726:   \setlength{\unitlength}{1.25pt}

3727: %  \setlength{\unitlength}{1.3pt}

3728: \begin{picture}(180,78)(0,-15)

3729: \linethickness{0.8pt}

3730: \multiput(0,40)(20,0){1}{{\circle{7}}}

3731: \multiput(0,40)(20,0){1}{\makebox(0,0){\small 1}}

3732: \multiput(40,40)(20,0){8}{{\circle{7}}}

3733: \multiput(40,40)(20,0){8}{\makebox(0,0){\small 1}}

3734: \multiput(0,30)(20,0){2}{{\circle{7}}}

3735: \multiput(0,30)(20,0){2}{\makebox(0,0){\small 2}}

3736: \multiput(40,30)(20,0){4}{{\circle{7}}}

3737: \multiput(40,30)(20,0){4}{\makebox(0,0){\small 2}}

3738: \multiput(140,30)(20,0){3}{{\circle{7}}}

3739: \multiput(140,30)(20,0){3}{\makebox(0,0){\small 2}}

3740: \multiput(0,20)(20,0){7}{{\circle{7}}}

3741: \multiput(0,20)(20,0){7}{\makebox(0,0){\small 3}}

3742: \multiput(160,20)(20,0){2}{{\circle{7}}}

3743: \multiput(160,20)(20,0){2}{\makebox(0,0){\small 3}}

3744: \multiput(0,10)(20,0){8}{{\circle{7}}}

3745: \multiput(0,10)(20,0){8}{\makebox(0,0){\small 4}}

3746: \multiput(180,10)(20,0){1}{{\circle{7}}}

3747: \multiput(180,10)(20,0){1}{\makebox(0,0){\small 4}}

3748: %

3749: %{\color{blue}

3750: \put(20,40){{\circle*{7}}}

3751: \put(120,30){{\circle*{7}}}

3752: \put(140,20){{\circle*{7}}}

3753: \put(160,10){{\circle*{7}}}

3754: %}

3755: {\color{white}

3756: \put(20,40){\makebox(0,0){\small\bf 1}}

3757: \put(120,30){\makebox(0,0){\small\bf 2}}

3758: \put(140,20){\makebox(0,0){\small\bf 3}}

3759: \put(160,10){\makebox(0,0){\small\bf 4}}

3760: }

3761: % 

3762: %{\color{blue}

3763: \thicklines

3764: \put(123.5,28.25){\line(2,-1){13}}

3765: \put(143.5,18.25){\line(2,-1){13}}

3766: %}

3767: \thicklines

3768: \multiput(3.5,38.25)(20,0){1}{\line(2,-1){13}}

3769: \multiput(3.5,28.25)(20,0){2}{\line(2,-1){13}}

3770: \multiput(3.5,18.25)(20,0){2}{\line(2,-1){13}}

3771: %

3772: {\color{red}

3773: \multiput(43.5,38.25)(20,0){3}{\line(2,-1){13}}

3774: \multiput(43.5,28.25)(20,0){4}{\line(2,-1){13}}

3775: \multiput(43.5,18.25)(20,0){4}{\line(2,-1){13}}

3776: %

3777: % TWO ALTERNTIVES

3778: %

3779: \put(90,25){\makebox(0,0){\line(2,3){15.9}}}

3780: %\qbezier(43.5,31.75)(50,35)(70,35)

3781: %\qbezier(96.4,38.25)(90,35)(70,35)

3782: }

3783: %

3784: \multiput(123.5,38.25)(20,0){3}{\line(2,-1){13}}

3785: \multiput(123.5,28.25)(20,0){3}{\line(2,-1){13}}

3786: \multiput(123.5,18.25)(20,0){3}{\line(2,-1){13}}

3787: %

3788: \thinlines

3789: \put(20,56){\makebox(0,0){\small $\bs{\kappa}_1$-spinon}}

3790: \put(140,56){\makebox(0,0){\small $\bs{\bar{\kappa}}_1$-spinon}}

3791: \put(80,-2){\vector(-1,0){60}}

3792: \put(80,-2){\vector(1,0){60}}

3793: \put(80,-8){\makebox(0,0){\small energy cost $\propto$ distance}}

3794: \end{picture}

3795: %\vspace{10pt}

3796: \nonumber

3797: \end{equation}

3798: We leave it to the reader to convince him- or herself that the

3799: conclusions regarding confinement drawn from the simple examples

3800: studied here hold in general.

3801: 

3802: \subsection{Models with confinement through interactions 

3803: extending beyond nearest neighbors }

3804: \label{sec:examples}

3805: 

3806: Let us briefly summarize the results obtained.  The SU($n$) models we

3807: have studied so far fall into two categories.  The models belonging to

3808: the first---the trimer chain, the $\bs{6}$ VBS, and the $n$-mer

3809: chain---have $n$ degenerate ground states, which break translational

3810: invariance up to translations by $n$ lattice spacings.  The Young

3811: tableaux describing the representations of SU($n$) at each site

3812: consist of a number of boxes $\lambda$ which is smaller than $n$, and

3813: hence obviously not divisible by $n$.  The models support deconfined

3814: spinon excitations, and hence do not possess a Haldane gap in the

3815: spectrum.  The Hamiltonians of these models require interactions

3816: between $n+1$ neighboring sites along the chain.  Even though the

3817: Affleck--Lieb theorem is not directly applicable to the models we

3818: constructed above, it is applicable to SU($n$) spin chains with spins

3819: transforming under the same representations.  Like the VBS models, the

3820: theorem suggests that there is no Haldane gap in this family of

3821: models.

3822: 

3823: 

3824: The models belonging to the second category---the $\bs{10}$ VBS, the

3825: $\bs{8}$ VBS, the representation $(n,0,\ldots,0)$ VBS, and the

3826: $\bs{m}$-$\bs{\overline{m}}$ matrix product state---have

3827: translationally invariant ground states.  The ground states are unique

3828: for some models, like the $\bs{10}$ and the $(n,0,\ldots,0)$ VBS, and

3829: degenerate for others, like the $\bs{8}$ VBS.  The Young tableaux

3830: describing the representations of SU($n$) at each site consist of a

3831: number of boxes $\lambda$ which is divisible by $n$.  The

3832: Affleck--Lieb theorem is not applicable to models of this category.

3833: The spinon excitations for this category of models are subject to

3834: confinement forces, which give rise to a Haldane gap.  The parent

3835: Hamiltonians for these models require interactions between

3836: nearest-neighbor sites only.

3837: 

3838: 

3839: At first glance, this classification might appear complete.  Further

3840: possibilities arise, however, in SU($n$) spin chains where number of

3841: boxes $\lambda$ the Young tableau consists of and $n$ have a common

3842: divisor different from $n$, which obviously requires that $n$ is not a

3843: prime number.  In this case, it is possible to construct VBS models in

3844: which the ground state breaks translational invariance only up to

3845: translations by $n/q$ lattice spacings, where $q$ denotes the largest

3846: common divisor of $\lambda$ and $n$ such that $q<n$.  This implies

3847: that the ground state of the appropriate, translationally invariant

3848: Hamiltonian will be $n/q$-fold degenerate.  In the examples we will

3849: elaborate on below, the parent Hamiltonians for these models require

3850: interactions between $\frac{n}{q}+1$ sites, a feature we conjecture to

3851: hold in general.  The spinon excitations of these models are confined,

3852: even though the Affleck--Lieb theorem states that the nearest-neighbor

3853: Heisenberg chain of SU($n$) spins transforming under these

3854: representations is gapless.  (We implicitly assume here that the 

3855: ground states of the SU($n$) nearest-neighbor Heisenberg chains are

3856: non-degenerate.)  Let us illustrate the general features of

3857: this third category of models with a few simple examples.

3858: 

3859: 

3860: (1) Consider an SU(4) chain with spins transforming under the

3861: 10-dimensional representation

3862: \begin{displaymath}

3863:   (2,0,0)\ =\

3864:   \setlength{\unitlength}{8pt}

3865:   \begin{picture}(2,1)(0,0.15)

3866:     \multiput(0,0)(1,0){3}{\line(0,1){1}}

3867:     \multiput(0,0)(0,1){2}{\line(1,0){2}}

3868:   \end{picture}\ \ .

3869: \end{displaymath}

3870: Following the construction principle of the $\bs{6}$ VBS of SU(3), we

3871: find that the two degenerate VBS states illustrated through

3872: \begin{equation}

3873: \setlength{\unitlength}{1pt}

3874: \begin{picture}(162,48)(4,-22)

3875: \linethickness{0.8pt}

3876: \multiput(0,10)(14,0){12}{\circle{4}}

3877: \multiput(0,0)(14,0){12}{\circle{4}}

3878: \thicklines

3879: \multiput(2,10)(14,0){3}{\line(1,0){10}}

3880: \multiput(58,10)(14,0){3}{\line(1,0){10}}

3881: \multiput(114,10)(14,0){3}{\line(1,0){10}}

3882: \put(-2,0){\line(-1,0){4}}

3883: \multiput(2,0)(14,0){1}{\line(1,0){10}}

3884: \multiput(30,0)(14,0){3}{\line(1,0){10}}

3885: \multiput(86,0)(14,0){3}{\line(1,0){10}}

3886: \multiput(142,0)(14,0){1}{\line(1,0){10}}

3887: \put(156,0){\line(1,0){4}}

3888: \thinlines

3889: %\put(37,-7){\framebox(10,24)}

3890: \put(37,-5){\framebox(10,20)}

3891: \put(107,-5){\dashbox{2}(38,20)}

3892: \put(42,-5){\line(0,-1){8}}

3893: \put(42,-20){\makebox(1,1){\small projection onto rep.\ $(2,0,0)$}}

3894: \put(42,23){\makebox(1,1){\small one site}}

3895: \end{picture}

3896: \label{eq:200state}

3897: \end{equation}

3898: are exact zero-energy ground states of 

3899: \begin{equation}

3900:   \label{eq:SU4model}

3901: %  H_{\text{SU}(4)}^{\text{10-VBS}}

3902:   H_{(2,0,0)\,\text{VBS}}

3903:   =\sum_{i=1}^N

3904:   \left(\Bigl(\bs{J}_i^{(3)}\Bigr)^4-12\Bigl(\bs{J}_i^{(3)}\Bigr)^2+

3905:     \frac{135}{4}\right),

3906: \end{equation}

3907: which contains next-nearest neighbor interactions.  

3908: % Note that \eqref{eq:200state} breaks translational invariance only up

3909: % to translations by two lattice spacings.

3910: 

3911: The example illustrates the general rule stated above. The largest

3912: common divisor of $n=4$ and $\lambda=2$ is $q=2$.  This implies

3913: $\frac{n}{q}=2$ and hence two degenerate VBS states which break

3914: translational invariance up to translations by two lattice spacings.

3915: The parent Hamiltonian requires interaction between three neighboring

3916: sites.

3917: 

3918: According to the Affleck--Lieb theorem, models of antiferromagnetic

3919: SU(4) chains of representation $(2,0,0)$ with nearest-neighbor

3920: Heisenberg interactions and non-degenerate ground states are gapless

3921: in the thermodynamic limit, which implies that the spinons are

3922: deconfined.  In all the models we have studied in previous sections,

3923: the conclusions drawn from the Affleck--Lieb theorem were consistent

3924: with those drawn from our exact models.  For the present model,

3925: however, they are not consistent.

3926: 

3927: Specifically, we strongly conjecture that the spinons in the $(2,0,0)$

3928: VBS are confined.  This conjecture is based on the reasonable

3929: assumption that the elementary excitations of the model transform as

3930: either the fundamental representation $\bs{4}=(1,0,0)$ of SU(4) or its

3931: conjugate representation $\bs{\bar 4}=(0,0,1)$.  This assumption

3932: implies that a single domain wall in one of the $4$-mer chains used to

3933: construct the VBS state \eqref{eq:200state} shifts this chain by one

3934: lattice spacing.  If we assume a ground state to the left of the

3935: spinon, the state on to the right will have a higher energy for the

3936: next-nearest neighbor Hamiltonian \eqref{eq:SU4model}, as illustrated

3937: below.  

3938: \begin{equation}

3939: \setlength{\unitlength}{1pt}

3940: \begin{picture}(170,60)(4,-25)

3941: \linethickness{0.8pt}

3942: \multiput(0,10)(14,0){13}{\circle{4}}

3943: \multiput(0,0)(14,0){13}{\circle{4}}

3944: \multiput(0,10)(14,0){3}{\circle*{4}}

3945: \multiput(140,0)(14,0){3}{\circle*{4}}

3946: \thicklines

3947: \multiput(2,10)(14,0){2}{\line(1,0){10}}

3948: \multiput(44,10)(14,0){3}{\line(1,0){10}}

3949: \multiput(100,10)(14,0){3}{\line(1,0){10}}

3950: \multiput(156,10)(14,0){1}{\line(1,0){10}}

3951: \put(-2,0){\line(-1,0){4}}

3952: \multiput(2,0)(14,0){1}{\line(1,0){10}}

3953: \multiput(30,0)(14,0){3}{\line(1,0){10}}

3954: \multiput(86,0)(14,0){3}{\line(1,0){10}}

3955: \multiput(142,0)(14,0){2}{\line(1,0){10}}

3956: \put(170,10){\line(1,0){4}}

3957: \thinlines

3958: \put(84,-14){\vector(-1,0){70}}

3959: \put(84,-14){\vector(1,0){70}}

3960: \put(84,-22){\makebox(0,0){\small energy cost $\propto$ distance}}

3961: \put(14,22){\makebox(0,0){$\bs{\bar 4}$}}

3962: \put(14,32){\makebox(0,0){\small spinon}}

3963: \put(154,22){\makebox(0,0){$\bs{\bar 4}$}}

3964: \put(154,32){\makebox(0,0){\small spinon}}

3965: \put(65,-5){\dashbox{1}(38,20)}

3966: \end{picture}

3967: \label{eq:200spinons}

3968: \end{equation}

3969: To recover the ground state, a second domain wall is required

3970: nearby, which is bound to the first by a linear potential.

3971: 

3972: Our conclusion is not in contradiction with the rigorous result of

3973: Affleck and Lieb, as they explicitly restrict themselves to models

3974: with nearest-neighbor interactions.  If we had only nearest-neighbor

3975: interactions, the energy expectation value in the region between the

3976: domain walls would not be higher than in the ground state, as one can

3977: see easily from the cartoon above---the sequence of alternating links

3978: would merely shift from (strong, weak, strong, weak) to (strong,

3979: strong, weak, weak).

3980: 

3981: (2) The situation is similar for an SU(6) chain with spins transforming

3982: under the 56-dimensional representation

3983: \begin{displaymath}

3984:   (3,0,0,0,0)\ =\

3985:   \setlength{\unitlength}{8pt}

3986:   \begin{picture}(3,1)(0,0.15)

3987:     \multiput(0,0)(1,0){4}{\line(0,1){1}}

3988:     \multiput(0,0)(0,1){2}{\line(1,0){3}}

3989:   \end{picture}\ \ .

3990: \end{displaymath}

3991: With $n=6$ and $\lambda=3$, we have again $\frac{n}{q}=2$.

3992: Accordingly, we find that the two VBS states illustrated through

3993: \begin{equation}

3994: \setlength{\unitlength}{1pt}

3995: \begin{picture}(162,58)(4,-22)

3996: \linethickness{0.8pt}

3997: \multiput(0,20)(14,0){12}{\circle{4}}

3998: \multiput(0,10)(14,0){12}{\circle{4}}

3999: \multiput(0,0)(14,0){12}{\circle{4}}

4000: \thicklines

4001: \multiput(2,20)(14,0){5}{\line(1,0){10}}

4002: \multiput(86,20)(14,0){5}{\line(1,0){10}}

4003: \put(-2,10){\line(-1,0){4}}

4004: \put(-2,0){\line(-1,0){4}}

4005: \multiput(2,10)(14,0){1}{\line(1,0){10}}

4006: \multiput(30,10)(14,0){5}{\line(1,0){10}}

4007: \multiput(114,10)(14,0){3}{\line(1,0){10}}

4008: \put(156,10){\line(1,0){4}}

4009: \multiput(2,0)(14,0){3}{\line(1,0){10}}

4010: \multiput(58,0)(14,0){5}{\line(1,0){10}}

4011: \multiput(142,0)(14,0){1}{\line(1,0){10}}

4012: \put(156,0){\line(1,0){4}}

4013: \thinlines

4014: %\put(37,-7){\framebox(10,24)}

4015: \put(37,-5){\framebox(10,30)}

4016: \put(107,-5){\dashbox{2}(38,30)}

4017: \put(42,-5){\line(0,-1){8}}

4018: \put(42,-20){\makebox(1,1){\small projection onto rep.\ $(3,0,0,0,0)$}}

4019: \put(42,33){\makebox(1,1){\small one site}}

4020: \end{picture}

4021: \label{eq:30000state}

4022: \end{equation}

4023: are exact ground states of a parent Hamiltonian containing up to

4024: next-nearest-neighbor interactions only, and that the spinon excitations

4025: are confined.

4026: 

4027: (3) As a third example, consider an SU(6) spin chain with spins

4028: transforming under the 21-dimensional representation

4029: \begin{displaymath}

4030:   (2,0,0,0,0)\ =\

4031:   \setlength{\unitlength}{8pt}

4032:   \begin{picture}(2,1)(0,0.15)

4033:     \multiput(0,0)(1,0){3}{\line(0,1){1}}

4034:     \multiput(0,0)(0,1){2}{\line(1,0){2}}

4035:   \end{picture}\ \ .

4036: \end{displaymath}

4037: This implies $\frac{n}{q}=3$.  We find that the three VBS states

4038: illustrated by

4039: \begin{equation}

4040: \setlength{\unitlength}{1pt}

4041: \begin{picture}(162,48)(4,-22)

4042: \linethickness{0.8pt}

4043: \multiput(0,10)(14,0){12}{\circle{4}}

4044: \multiput(0,0)(14,0){12}{\circle{4}}

4045: \thicklines

4046: \multiput(2,10)(14,0){5}{\line(1,0){10}}

4047: \multiput(86,10)(14,0){5}{\line(1,0){10}}

4048: \put(-2,0){\line(-1,0){4}}

4049: \multiput(2,0)(14,0){2}{\line(1,0){10}}

4050: \multiput(44,0)(14,0){5}{\line(1,0){10}}

4051: \multiput(128,0)(14,0){2}{\line(1,0){10}}

4052: %\multiput(170,10)(14,0){2}{\line(1,0){10}}

4053: \put(156,0){\line(1,0){4}}

4054: %\put(198,10){\line(1,0){4}}

4055: \thinlines

4056: %\put(37,-7){\framebox(10,24)}

4057: \put(37,-5){\framebox(10,20)}

4058: \put(93,-5){\dashbox{2}(52,20)}

4059: \put(42,-5){\line(0,-1){8}}

4060: \put(42,-20){\makebox(1,1){\small projection onto rep.\ $(2,0,0,0,0)$}}

4061: \put(42,23){\makebox(1,1){\small one site}}

4062: \end{picture}

4063: \label{eq:20000state}

4064: \end{equation}

4065: are exact ground states of a parent Hamiltonian involving the

4066: quadratic Casimir of total spin of four neighboring sites,

4067: \begin{equation}

4068:   \label{eq:ham20000}

4069:   H_{(2,0,0,0,0)\,\text{VBS}}=\sum_{i=1}^N H_i

4070: \end{equation}

4071: with 

4072: \begin{equation}

4073:   \label{eq:hi20000}

4074:   H_i=\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{32}{3}\right)

4075:   \!\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{44}{3}\right)

4076:   \!\left(\!\left(\!\bs{J}^{(4)}_i\!\right)^2 \! - \frac{50}{3}\right).

4077: \end{equation}

4078: These VBS states break translational symmetry only up to translations

4079: by three lattice spacings.  The spinons of this model are again

4080: confined through a linear potential, as illustrated below.

4081: \begin{equation}

4082: \setlength{\unitlength}{1pt}

4083: \begin{picture}(176,60)(4,-25)

4084: \linethickness{0.8pt}

4085: \multiput(0,10)(14,0){15}{\circle{4}}

4086: \multiput(0,0)(14,0){15}{\circle{4}}

4087: \multiput(0,10)(14,0){5}{\circle*{4}}

4088: \multiput(126,0)(14,0){5}{\circle*{4}}

4089: \thicklines

4090: \put(-2,0){\line(-1,0){4}}

4091: \multiput(2,10)(14,0){4}{\line(1,0){10}}

4092: \multiput(72,10)(14,0){5}{\line(1,0){10}}

4093: \multiput(156,10)(14,0){2}{\line(1,0){10}}

4094: \put(184,10){\line(1,0){4}}

4095: \multiput(2,0)(14,0){2}{\line(1,0){10}}

4096: \multiput(44,0)(14,0){5}{\line(1,0){10}}

4097: \multiput(128,0)(14,0){4}{\line(1,0){10}}

4098: %\multiput(156,0)(14,0){1}{\line(1,0){10}}

4099: %\put(198,0){\line(1,0){4}}

4100: \thinlines

4101: \put(91,-14){\vector(-1,0){63}}

4102: \put(91,-14){\vector(1,0){63}}

4103: \put(91,-22){\makebox(0,0){\small energy cost $\propto$ distance}}

4104: \put(28,22){\makebox(0,0){$\bs{\bar 6}$}}

4105: \put(28,32){\makebox(0,0){\small spinon}}

4106: \put(154,23){\makebox(0,0){$\bs{\bar 6}$}}

4107: \put(154,33){\makebox(0,0){\small spinon}}

4108: \put(107,-5){\dashbox{1}(52,20)}

4109: \end{picture}

4110: \label{eq:20000spinons}

4111: \end{equation}

4112: The conclusions we have drawn for this VBS model rest on the

4113: assumption that the quadratic Casimirs of the representations

4114: contained in the tensor product shown in the dashed box in

4115: \eqref{eq:20000state} as well as in the tensor product one obtains if

4116: one shifts this box by one lattice spacing to the left or to the right

4117: are smaller than the largest Casimir contained in the tensor product

4118: shown in the dotted box in \eqref{eq:20000spinons}.  We have verified the

4119: validity of this assumption for the (2,0,0,0,0) VBS model we

4120: considered here, but not shown it to hold for similar models with

4121: larger $n$ or $\lambda$.

4122: 

4123: (4) Finally, consider an SU(6) spin chain

4124: with spins transforming under the 70-dimensional representation

4125: \begin{displaymath}

4126:   (1,1,0,0,0)\ =\

4127:   \setlength{\unitlength}{8pt}

4128:   \begin{picture}(2,1)(0,-0.35)

4129:     \multiput(0,0)(1,0){3}{\line(0,1){1}}

4130:     \multiput(0,0)(1,0){2}{\line(0,-1){1}}

4131:     \multiput(0,0)(0,1){2}{\line(1,0){2}}

4132:     \multiput(0,-1)(0,1){1}{\line(1,0){1}}

4133:   \end{picture}\ \ .

4134: \end{displaymath}

4135: Thus we have once again $\frac{n}{q}=2$.  In a notation

4136: similar to the one introduced for the $\bs{8}$ VBS, 

4137: \begin{displaymath}

4138: %  (0,1,0,0,0)\ =\

4139:   \setlength{\unitlength}{8pt}

4140:   \begin{picture}(1.8,1)(0,0.15)

4141:     \multiput(0,0)(1,0){2}{\line(0,1){1}}

4142:     \multiput(0,0)(0,1){2}{\line(1,0){1}}

4143:   \end{picture}\hat =

4144:   \setlength{\unitlength}{1pt}

4145:   \begin{picture}(10,10)(0,0)

4146:     \put(6,2.5){\circle{4}}

4147:   \end{picture},\ \

4148:   \setlength{\unitlength}{8pt}

4149:   \begin{picture}(1.8,1)(0,-0.35)

4150:     \multiput(0,0)(1,0){2}{\line(0,1){1}}

4151:     \multiput(0,0)(1,0){2}{\line(0,-1){1}}

4152:     \multiput(0,-1)(0,1){3}{\line(1,0){1}}

4153:   \end{picture}\hat =

4154:   \setlength{\unitlength}{1pt}

4155:   \begin{picture}(11,10)(0,0)

4156:     \put(6,2.5){\circle{6}}

4157:   \end{picture} ,

4158: \end{displaymath}

4159: the two degenerate VBSs are illustrated by

4160: \begin{equation}

4161: \setlength{\unitlength}{1pt}

4162: \begin{picture}(162,48)(4,-22)

4163: \linethickness{0.8pt}

4164: \multiput(5,10)(60,0){3}{\circle{4}}

4165: \multiput(50,10)(60,0){3}{\circle{4}}

4166: \multiput(20,0)(60,0){3}{\circle{4}}

4167: \multiput(35,0)(60,0){3}{\circle{4}}

4168: \multiput(20,10)(60,0){3}{\circle{6}}

4169: \multiput(35,10)(60,0){3}{\circle{6}}

4170: \multiput(5,0)(60,0){3}{\circle{6}}

4171: \multiput(50,0)(60,0){3}{\circle{6}}

4172: %

4173: \thicklines

4174: \multiput(7,10)(60,0){3}{\line(1,0){10}}

4175: \multiput(38,10)(60,0){3}{\line(1,0){10}}

4176: \multiput(23,10)(60,0){3}{\line(1,0){9}}

4177: \multiput(37,0)(60,0){3}{\line(1,0){10}}

4178: \multiput(8,0)(60,0){3}{\line(1,0){10}}

4179: \multiput(53,0)(60,0){2}{\line(1,0){9}}

4180: \put(2,0){\line(-1,0){4}}

4181: \put(173,0){\line(1,0){4}}

4182: %

4183: \thinlines

4184: \put(44,-6){\framebox(12,22)}

4185: \put(50,-6){\line(0,-1){9}}

4186: \put(50,-22){\makebox(1,1){\small projection onto rep.\ $(1,1,0,0,0)$}}

4187: \put(50,25){\makebox(1,1){\small one site}}

4188: \put(104.5,-6){\dashbox{2}(41,22)}

4189: %\put(104.5,-6){\dashbox{2}(41,24)}

4190: %\put(89.5,-8){\dashbox{2}(41,24)}

4191: \end{picture}

4192: \label{eq:11000state}

4193: \end{equation}

4194: are exact ground states of a parent Hamiltonian involving the

4195: quadratic Casimir of the total spin of three neighboring sites,

4196: \begin{equation}

4197:   \label{eq:ham11000}

4198:   H_{(1,1,0,0,0)\,\text{VBS}}=\sum_{i=1}^N H_i

4199: \end{equation}

4200: with 

4201: \begin{equation}

4202:   \label{eq:hi11000}

4203:   H_i=\left(\!\left(\!\bs{J}^{(3)}_i\!\right)^2 \! - 20\right)

4204:   \!\left(\!\left(\!\bs{J}^{(3)}_i\!\right)^2 \! - 70\right)

4205:   \!\left(\!\left(\!\bs{J}^{(3)}_i\!\right)^2 \! - 540\right).

4206: \end{equation}

4207: The states \eqref{eq:11000state} are not the only VBSs one can form.

4208: Other possibilities like

4209: \begin{equation}

4210: \setlength{\unitlength}{1pt}

4211: \begin{picture}(174,12)(4,2)

4212: \linethickness{0.8pt}

4213: \multiput(5,10)(60,0){3}{\circle{6}}

4214: \multiput(50,10)(60,0){3}{\circle{6}}

4215: \multiput(20,0)(60,0){3}{\circle{6}}

4216: \multiput(35,0)(60,0){3}{\circle{6}}

4217: \multiput(20,10)(60,0){3}{\circle{4}}

4218: \multiput(35,10)(60,0){3}{\circle{4}}

4219: \multiput(5,0)(60,0){3}{\circle{4}}

4220: \multiput(50,0)(60,0){3}{\circle{4}}

4221: %

4222: \thicklines

4223: \multiput(8,10)(60,0){3}{\line(1,0){10}}

4224: \multiput(37,10)(60,0){3}{\line(1,0){10}}

4225: \multiput(22,10)(60,0){3}{\line(1,0){11}}

4226: \multiput(38,0)(60,0){3}{\line(1,0){10}}

4227: \multiput(7,0)(60,0){3}{\line(1,0){10}}

4228: \multiput(52,0)(60,0){2}{\line(1,0){11}}

4229: \put(3,0){\line(-1,0){4}}

4230: \put(172,0){\line(1,0){4}}

4231: %

4232: \thinlines

4233: \put(74.5,-6){\dashbox{2}(41,22)}

4234: \end{picture}

4235: \nonumber

4236: \end{equation}

4237: or

4238: \begin{equation}

4239: \setlength{\unitlength}{1pt}

4240: \begin{picture}(178,16)(4,-2)

4241: \linethickness{0.8pt}

4242: \multiput(5,10)(30,0){6}{\circle{4}}

4243: \multiput(20,10)(30,0){6}{\circle{6}}

4244: \multiput(5,0)(30,0){6}{\circle{6}}

4245: \multiput(20,0)(30,0){6}{\circle{4}}

4246: %

4247: \thicklines

4248: \multiput(7,10)(30,0){6}{\line(1,0){10}}

4249: \multiput(23,10)(60,0){3}{\line(1,0){10}}

4250: \multiput(8,0)(30,0){6}{\line(1,0){10}}

4251: \multiput(52,0)(60,0){2}{\line(1,0){10}}

4252: \put(2,0){\line(-1,0){4}}

4253: \put(172,0){\line(1,0){4}}

4254: %

4255: \thinlines

4256: \put(104.5,-6){\dashbox{2}(41,24)}

4257: \put(89.5,-8){\dashbox{2}(41,24)}

4258: \end{picture},

4259: \nonumber

4260: \end{equation}

4261: however, contain additional representations for the total SU(6) spin

4262: of three neighboring sites, and are hence expected to possess

4263: higher energies.

4264: 

4265: Spinon excitations transforming under the 6-dimensional rep.\

4266: $(0,0,0,0,1)$ are linearly confined to spinons transforming under the

4267: 15-dimensional rep.\ $(0,0,0,1,0)$:

4268: \begin{equation}

4269: \setlength{\unitlength}{1pt}

4270: \begin{picture}(180,60)(4,-25)

4271: \linethickness{0.8pt}

4272: \multiput(5,10)(60,0){3}{\circle{4}}

4273: \multiput(50,10)(60,0){3}{\circle{4}}

4274: \multiput(20,0)(60,0){3}{\circle{4}}

4275: \multiput(35,0)(60,0){3}{\circle{4}}

4276: \multiput(20,10)(60,0){3}{\circle{6}}

4277: \multiput(35,10)(60,0){3}{\circle{6}}

4278: \put(185,10){\circle{6}}

4279: \multiput(5,0)(60,0){3}{\circle{6}}

4280: \multiput(50,0)(60,0){3}{\circle{6}}

4281: %

4282: \multiput(5,10)(30,0){1}{\circle*{4}}

4283: \multiput(20,10)(15,0){2}{\circle*{6}}

4284: \multiput(170,0)(15,0){1}{\circle*{6}}

4285: \multiput(155,0)(30,0){2}{\circle*{4}}

4286: %

4287: \thicklines

4288: \multiput(7,10)(60,0){3}{\line(1,0){10}}

4289: \put(172,10){\line(1,0){10}}

4290: \multiput(52,10)(60,0){2}{\line(1,0){11}}

4291: \multiput(23,10)(60,0){3}{\line(1,0){9}}

4292: \multiput(8,0)(60,0){3}{\line(1,0){10}}

4293: \multiput(37,0)(60,0){3}{\line(1,0){10}}

4294: \multiput(53,0)(60,0){2}{\line(1,0){9}}

4295: \put(173,0){\line(1,0){10}}

4296: \put(2,0){\line(-1,0){4}}

4297: \put(188,10){\line(1,0){4}}

4298: %

4299: \thinlines

4300: \put(20,25){\makebox(0,0){\small $(0,0,0,0,1)$}}

4301: \put(20,36){\makebox(0,0){\small spinon }}

4302: \put(170,25){\makebox(0,0){\small $(0,0,0,1,0)$}}

4303: \put(170,36){\makebox(0,0){\small spinon}}

4304: \put(74.5,-6){\dashbox{2}(41,22)}

4305: \put(95,-14){\vector(-1,0){75}}

4306: \put(95,-14){\vector(1,0){75}}

4307: \put(95,-22){\makebox(0,0){\small energy cost $\propto$ distance}}

4308: \end{picture}

4309: \label{eq:11000spinons}

4310: \end{equation}

4311: The VBS configuration we have drawn between the two spinons in

4312: \eqref{eq:11000spinons} constitutes just one of several possibilities.

4313: We expect, however, that this possibility corresponds to the lowest

4314: energy among them for the Hamiltonian \eqref{eq:ham11000}.  This

4315: concludes our list of examples.

4316: 

4317: The models introduced in this subsection are interesting in that they

4318: provide us with examples where spinon confinement, and with the

4319: confinement the existence of a Haldane gap, is caused by interactions

4320: extending beyond nearest neighbors.  The conclusion drawn from the

4321: Affleck--Lieb theorem for SU($n$) models with spins transforming under

4322: representations we have labeled here as the ``third category'' hence

4323: appear to hold for models with nearest-neighbor interactions only, to

4324: which the theorem is applicable.  For these models, the theorem states

4325: that the spectrum is gapless, which according to our understanding

4326: implies that the models support deconfined spinon excitations.  The

4327: examples we have studied, by contrast, suggest that models with

4328: longer-ranged interactions belonging to this category exhibit

4329: confinement forces between the spinon excitations and hence possess a

4330: Haldane gap.

4331: 

4332: It is worth noting that even though in the examples we elaborated here

4333: $\lambda <n$, we expect our conclusions to hold for models with

4334: $\lambda >n$ as well.  To see this, let $m>0$ be an integer such that

4335: $n m<\lambda <n(m+1)$.  We can now construct a first VBS with spinon

4336: confinement using $n m$ boxes of the Young tableau and combine it with

4337: a second by projection on each side with a second VBS constructed from

4338: the remaining $\lambda'= \lambda-n m$ boxes.  Since the spinons

4339: of the first VBS are always confined and hence gapped, the final VBS

4340: will support deconfined spinons if and only if the second VBS will

4341: support them, which in turn will depend on the largest common divisor

4342: $q'$ of $\lambda'$ and $n$ as well as the range of the interaction.

4343: Since the largest common divisor $q$ of $n$ and $\lambda$ is

4344: equal to $q'$, there is no need to think in terms of $\lambda'$ and $q'$.

4345: The conclusions regarding confinement and the Haldane gap will not

4346: depend on the distinction between $\lambda$ and $\lambda'$.

4347: 

4348: \subsection{The different categories of models}

4349: \label{sec:sum}

4350: 

4351: In this section, we used the rules

4352: emerging from the numerous examples we studied to argue that models of

4353: SU($n$) spin chains in general fall into three categories.  The

4354: classification depends on the number of boxes $\lambda$ the Young

4355: tableau corresponding to the representation of the individual spins

4356: consists off, as follows:

4357: \begin{enumerate}

4358: \item If $\lambda$ and $n$ have no common divisor, the models will

4359:   support free spin excitations and hence not exhibit a Haldane gap.

4360: 

4361:   The general reasoning here is simply that the VBS states in this

4362:   category break translational invariance up to translations by $n$

4363:   lattice spacings, and that there are (at least) $n$ degenerate

4364:   VBS ground states to each model.  Spinons transforming under

4365:   representations of Young tableaux with an arbitrary number of boxes

4366:   can be accommodated in domain walls between these different ground

4367:   states.  Consequently, the spinons are deconfined.  

4368: 

4369: \item If $\lambda$ is divisible by $n$, the general argument we have

4370:   constructed in Sec.\ \ref{sec:generalcriterion} indicates that the

4371:   model will exhibit spinon confinement and hence a Haldane gap.

4372: 

4373: \item \label{three}

4374:   If $\lambda$ and $n$ have a common divisor different from $n$,

4375:   the examples studied in Sec.\ \ref{sec:examples} suggest that the

4376:   question of whether the spinons are confined or not depends on the

4377:   details of the interactions.  If $q$ is the largest common divisor

4378:   of $\lambda$ and $n$, interactions ranging to the $\frac{n}{q}$-th

4379:   neighbor were required for spinon confinement in the models we

4380:   studied.  The Affleck-Lieb theorem~\cite{affleck-86lmp57}, on the

4381:   other hand, tells us that SU($n$) chains with nearest-neighbor

4382:   Heisenberg interactions belonging to this category are gapless if

4383:   the ground states are non-degenerate.

4384: \end{enumerate}

4385: 

4386: Note that the second category is really just the special case $q=n$

4387: of the third:  with $\frac{n}{q}=1$, nearest-neighbor interactions

4388: are already sufficient for spinon confinement and a Haldane gap.

4389: 

4390: The conclusion that interactions ranging to the $\frac{n}{q}$-th

4391: neighbor are required for spinon confinement and a Haldane gap,

4392: however, is not universally valid.  A counter-example is provided by

4393: the extended VBSs (XVBSs) introduced by Affleck, Arovas, Marston, and

4394: Rabson~\cite{affleck-91npb467}.  In this model, each site effectively

4395: takes the role of two neighboring sites when a VBS is constructed, and

4396: nearest-neighbor interactions are already sufficient to cause spinon

4397: confinement.  We briefly review this model in the following

4398: subsection.

4399: 

4400: 

4401: \subsection{A counter-example: the SU(4) representation $\bs{6}$

4402: extended VBS}

4403: \label{sec:counter}

4404: 

4405: In an article devoted to quantum antiferromagnets with spins transforming under

4406:  the self-conjugate representations of SU($2n$),

4407: Affleck \ea~\cite{affleck-91npb467} introduced an extended VBS for the

4408: six-dimensional SU(4) representation $(0,1,0)$.  This %six-dimensional 

4409: representation is, with regard to the number of boxes the corresponding

4410: Young tableau consists of, not distinguishable from the symmetric 

4411: representation $\bs{10}$ considered in Sec.~\ref{sec:examples}, as both

4412: tableaux consist of two boxes:

4413: \begin{equation}

4414: \setlength{\unitlength}{0.8pt}  

4415: \begin{picture}(116,36)(-6,-6)

4416: \put(-6,10){\line(1,0){10}} 

4417: \put(-6,20){\line(1,0){10}} 

4418: \put(-6,10){\line(0,1){10}} 

4419: \put(4,10){\line(0,1){10}} 

4420: \put(14,15){\makebox(1,1){$\otimes$}}

4421: \put(24,10){\line(1,0){10}} 

4422: \put(24,20){\line(1,0){10}} 

4423: \put(24,10){\line(0,1){10}} 

4424: \put(34,10){\line(0,1){10}} 

4425: \put(47,15){\makebox(1,1){$=$}}

4426: \put(60,10){\line(1,0){20}} 

4427: \put(60,20){\line(1,0){20}} 

4428: \put(60,10){\line(0,1){10}} 

4429: \put(70,10){\line(0,1){10}} 

4430: \put(80,10){\line(0,1){10}} 

4431: \put(90,15){\makebox(1,1){$\oplus$}}

4432: % \put(100,10){\line(1,0){10}} 

4433: % \put(100,20){\line(1,0){10}} 

4434: % \put(100,30){\line(1,0){10}} 

4435: % \put(100,10){\line(0,1){20}} 

4436: % \put(110,10){\line(0,1){20}} 

4437: \put(100,5){\line(1,0){10}} 

4438: \put(100,15){\line(1,0){10}} 

4439: \put(100,25){\line(1,0){10}} 

4440: \put(100,5){\line(0,1){20}} 

4441: \put(110,5){\line(0,1){20}} 

4442: \put(-1,-2){\makebox(1,1){$\bs{4}$}}

4443: \put(29,-2){\makebox(1,1){$\bs{4}$}}

4444: \put(70,-2){\makebox(1,1){$\bs{10}$}}

4445: \put(105,-7){\makebox(1,1){$\bs{6}$}}

4446: \end{picture}

4447: \end{equation}

4448: Since the two boxes are combined antisymmetrically for rep.~$\bs{6}$,

4449: a VBS constructed along the lines of Sec.~\ref{sec:examples} would no

4450: longer provide a paradigm for antiferromagnetic spin chains of the

4451: corresponding representation in general.  Affleck

4452: \ea~\cite{affleck-91npb467} have constructed an extended VBS, % (XVBS),

4453: which is illustrated in the following cartoon:

4454: \begin{equation}

4455: \label{gs.su4.xvbs}

4456: \setlength{\unitlength}{1pt}

4457: \begin{picture}(170,20)(0,0)

4458:  \multiput(0,10)(30,0){6}{\circle{4}}

4459:  \multiput(8,10)(30,0){6}{\circle{4}}

4460:  \multiput(4,10)(30,0){6}{\circle{16}}

4461: % \multiput(-4,4)(30,0){6}{\dashbox{2}(16,12)}

4462: % \multiput(-4,4)(30,0){6}{\framebox(16,12)}

4463: \thicklines

4464:  \multiput(2,10)(60,0){3}{\line(1,0){4}}

4465:  \multiput(10,10)(30,0){5}{\line(1,0){18}}

4466:  \put(-8,10){\line(1,0){6}}

4467:  \put(160,10){\line(1,0){7}}

4468: \end{picture}

4469: \end{equation}

4470: Here each small circle represents a fundamental representation

4471: $\bs{4}$ of SU(4) (a box in the Young tableau), and each large circle a

4472: lattice site.  The lines connecting four dots indicate that these four

4473: fundamental representations are combined into an SU(4) singlet.  The

4474: total spin of two neighboring sites in this state may assume the

4475: representations

4476: \begin{equation}

4477:   \bs{4}\otimes\bs{\bar 4}=\bs{1}\oplus\bs{15},

4478: \end{equation}

4479: while combining two reps.~$\bs{6}$ on neighboring sites in general

4480: yields

4481: \begin{equation}

4482:   \bs{6}\otimes\bs{6}=\bs{1}\oplus\bs{15}\oplus\bs{20}.

4483: \end{equation}

4484: To construct a parent Hamiltonian for the XVBS \eqref{gs.su4.xvbs}, it

4485: is hence sufficient to sum over projectors onto rep.~$\bs{20}$ on all

4486: pairs of neighboring sites.  Using our conventions (Affleck \ea

4487: ~\cite{affleck-91npb467} have normalized the eigenvalues of the

4488: quadratic Casimir operator to $\casinn{4}{\textrm{adjoint rep.}}=8$),

4489: the parent Hamiltonian takes the form

4490: \begin{equation}

4491:   \label{ham.su4.xvbs}

4492:   H_{\textrm{XVBS}}=\sum_{i=1}^N \left( \bs{J}_i\bs{J}_{i+1} +

4493:      \frac{1}{3}\left(\bs{J}_i\bs{J}_{i+1}\right)^2+\frac{5}{12} \right),

4494: \end{equation}

4495: where the operators $J_i^a$, $a=1,\ldots,15$, are given by $6\times 6$

4496: matrices.  Note that the ground state is two fold degenerate, as it 

4497: breaks translational symmetry modulo translations by two lattice spacings.

4498: 

4499: We conjecture that the lowest lying excitation is a

4500: bound state consisting of two fundamental reps $\bs{4}$, which most

4501: likely are combined antisymmetrically into a rep.~$\bs{6}$:

4502: \begin{equation*}

4503:   \setlength{\unitlength}{1pt}

4504: \begin{picture}(170,20)(0,0)

4505:  \multiput(0,10)(30,0){6}{\circle{4}}

4506:  \multiput(8,10)(30,0){6}{\circle{4}}

4507:  \multiput(68,10)(22,0){2}{\circle*{4}}

4508:  \multiput(4,10)(30,0){6}{\circle{16}}

4509: % \multiput(-4,4)(30,0){6}{\dashbox{2}(16,12)}

4510: % \multiput(52,0)(60,0){3}{\line(1,0){10}}

4511: \thicklines

4512:  \multiput(32,10)(90,0){2}{\line(1,0){4}}

4513:  \multiput(10,10)(30,0){5}{\line(1,0){18}}

4514:  \put(-8,10){\line(1,0){6}}

4515:  \put(160,10){\line(1,0){7}}

4516: \end{picture}

4517: \end{equation*}

4518: The SU(4) rep.~$\bs{6}$ XVBS provides us with an example where a

4519: nearest-neighbor Hamiltonian is sufficient to induce spinon

4520: confinement and a Haldane gap, even though the largest common divisor

4521: of $n=4$ and $\lambda=2$ is $q=2$, \ie interactions including

4522: $\frac{n}{q}=2$ neighbors would be required following the rules

4523: derived from the examples in Sec.~\ref{sec:examples}.  The reason for

4524: this discrepancy is that in the XVBS model considered here, each site

4525: effectively takes the role of two neighboring sites.  

4526: Affleck \ea~\cite{affleck-91npb467,marston-89prb11538} conjecture

4527: that the ground state of the SU(4) rep.~$\bs{6}$ nearest-neighbor 

4528: Heisenberg model is, like the XVBS reviewed here, two-fold degenerate,

4529: which implies that the Affleck-Lieb theorem is not applicable.

4530: 

4531: This example is valuable in showing that it is advisable to explicitly

4532: construct the VBS for a given representation of SU($n$) in order to

4533: verify the applicability of the general rules motivated above.

4534: 

4535: \section{Conclusion}

4536: \label{sec:conclusion}

4537: 

4538: In the first part of this article, we have formulated several exact

4539: models of SU(3) spin chains.  We introduced a trimer model and

4540: presented evidence that the elementary excitations of the model

4541: transform under the SU(3) representations conjugate to the

4542: representation of the original spin on the chain.  We then introduced

4543: three SU(3) valence bond solid chains with spins transforming under

4544: representations $\bs{6}$, $\bs{10}$, and $\bs{8}$, respectively.  We

4545: argued that of these four models, the coloron excitations are confined

4546: only in the $\bs{10}$ and the $\bs{8}$ VBS models, and that only those

4547: models with confined spinons exhibit a Haldane gap.  We subsequently

4548: generalized three of our models to SU($n$), and investigated again

4549: which models exhibit spinon confinement.  

4550: 

4551: Finally, we used the rules emerging from the numerous examples we

4552: studied to argue that models of SU($n$) spin chains in general fall

4553: into three categories with regard to spinon confinement and the

4554: Haldane gap.  These are summarized in the previous subsection.

4555: The results rely crucially on the assumption that the conclusions we

4556: obtained for the VBS models we studied are of general validity.

4557: This assumption certainly holds for the corresponding SU(2) models,

4558: and appears reasonable on physical grounds.  Ultimately, however,

4559: it is only an assumption, or at best a hypotheses.

4560: 

4561: On a broader perspective, we believe that the models we have studied

4562: provide further indication that SU($n$) spin chains are an

4563: equally rich and rewarding subject of study as SU(2) spin chains have 

4564: been since Bethe.

4565: 

4566: 

4567: 

4568: 

4569: \section*{ACKNOWLEDGMENTS}

4570: 

4571: We would like to thank Peter Zoller and Peter W\"olfle, but 

4572: especially Dirk Schuricht and Ronny Thomale for many highly inspiring 

4573: discussions of various aspects of this work.  We further wish to

4574: thank Dirk Schuricht for valuable suggestions on the manuscript.  

4575: One of us (SR) was supported by a Ph.D.\ scholarship of the Cusanuswerk.

4576: 

4577: %\newpage

4578: \appendix

4579: \section{A proposal for an experimental realization of SU(3) spin 

4580: chains in an optical lattice}

4581: \label{sec:exp}

4582: 

4583: In this appendix, we wish to describe a proposal for an experimental

4584: realization of SU(3) spin chains. The most eligible candidate for such

4585: experiments are ultra-cold gases in optical lattices. Recently, these

4586: systems have become an interesting playground for the realization of

4587: various problems of condensed matter physics, such as the phase

4588: transition from a superfluid to a Mott

4589: insulator~\cite{Jaksch-98prl3108,Greiner-02n39}, the fermionic Hubbard

4590: model~\cite{Honerkamp-04prl170403,Honerkamp-04prb094521,rapp-06cm0607138},

4591: and SU(2) spin chains~\cite{Duan-03prl090402,Garcia-04prl250405}.  In

4592: particular, the Hamiltonians for spin lattice models may be engineered

4593: with polar molecules stored in optical lattices, where the spin is

4594: represented by a single-valence electron of a heteronuclear

4595: molecule~\cite{micheli-06np341,brennen-06qp0612180}.

4596: 

4597: In a most naive approach, one might expect to realize an SU(3) spin

4598: (at a site in an optical lattice) by using atoms with three internal

4599: states, like an atom with spin $S=1$.  If we now were to interpret the

4600: $S^z=+1$ state as SU(3) spin ``blue'', the $S^z=0$ state as ''red'',

4601: and the $S^z=-1$ state as ``green'', however, the SU(3) spin would not

4602: be conserved.  The SU(2) algebra would allow for the process

4603: $\ket{+1,-1}\rightarrow\ket{0,0}$, which in SU(3) language corresponds

4604: to the forbidden process $\ket{\b,\g}\rightarrow\ket{\r,\r}$.

4605: 

4606: A more sophisticated approach is hence required.  One way to obtain a

4607: system with three internal states in which the number of particles in

4608: each state (\ie of each color) is conserved 

4609: %(proposed to us by P.\ Zoller) 

4610: is to manipulate an atomic system with total

4611: angular momentum $F=3/2$ (where

4612: $\bs{F}=\bs{S}_{\text{el}}+\bs{L}_{\text{orb}}+\bs{S}_{\text{nuc}}$

4613: includes the internal spin of the electrons, the orbital angular

4614: momentum, and the spin of the nucleus) to simulate an SU(3) spin.  The

4615: important feature here is that the atoms have four internal states,

4616: corresponding to

4617: $F^z=-\frac{3}{2},-\frac{1}{2},+\frac{1}{2},+\frac{3}{2}$.  

4618: %, which is the case for \eg $^{137}$Ba.  

4619: For such atoms, one has to suppress the

4620: occupation of one of the ``middle'' states, say the $F^z=-\frac{1}{2}$

4621: state, by effectively lifting it to a higher energy while keeping the

4622: other states approximately degenerate.  This can be accomplished

4623: through a combination of an external magnetic field and two carefully

4624: tuned lasers, which effectively push down the energies of the

4625: $F^z=-\frac{3}{2}$ and the $F^z=+\frac{1}{2}$ states by coupling these

4626: states to states of (say) the energetically higher $F=5/2$ multiplet

4627: (see Fig.~\ref{fig:exp}).

4628: \begin{figure}

4629: \begin{center}

4630: \includegraphics[width=0.38\textwidth]{exper2.eps}

4631: \caption{Effective lifting of the $F^z=-\frac{1}{2}$ state.}

4632: \label{fig:exp}

4633: \end{center}

4634: \end{figure}

4635: At sufficiently low temperatures, we are hence left with a system with

4636: three internal states $F^z=-\frac{3}{2},+\frac{1}{2},+\frac{3}{2}$,

4637: which we may identify with the colors ``blue'', ``red'', and ``green''

4638: of an SU(3) spin.  In leading order, the number of particles of each

4639: color is now conserved, as required by SU(3) symmetry.

4640: For example, conservation of $F^z$ forbids processes in which a

4641: ``blue'' and a ``green'' particle turn into two ``red'' ones,

4642: $\ket{\b,\g}\rightarrow\ket{\r,\r}$.  Higher order processes of the

4643: kind $\ket{\b,\g,\g}\rightarrow\ket{\r,\r,\r}$ are still possible, but

4644: negligible if the experiment is conducted at sufficiently short time

4645: scales.

4646: 

4647: If one places fermionic atoms with an artificial SU(3) spin engineered

4648: along the lines of this or a related proposal in an optical lattice

4649: and allows for a weak hopping of the atoms on the lattice, one has

4650: developed an experimental realization of an SU(3) Hubbard model.

4651: If the energy cost $U$ of having two atoms on the same lattice site is

4652: significantly larger than the hopping $t$, and the density is one atom

4653: per site, the system will effectively constitute an SU(3)

4654: antiferromagnet.  The dimension of this antiferromagnet will depend on

4655: the optical lattice, which can be one-, two-, or three-dimensional.

4656: 

4657: %\newpage

4658: In principle, the above proposal can be generalized to SU($n$), even

4659: though the experimental obstacles are likely to grow ``exponentially''

4660: with $n$.  Besides, it is far from clear that such an endeavor is

4661: worthwhile, as all the non-trivial properties of SU($n$) are already

4662: present in SU(3) chains (while SU(2) constitutes a special case).

4663: 

4664: %\vfill\newpage

4665: 

4666: \section{Gell-Mann matrices}

4667: \label{app:conventions}

4668: 

4669: The Gell-Mann matrices are given by~\cite{Cornwell84vol2,Georgi82}

4670: \begin{displaymath}

4671: %\begin{array}{r@{}c@{}l@{}r@{}c@{}l@{}r@{}c@{}l}

4672: \begin{array}{c@{}c@{}c@{\hspace{6pt}}c@{}c@{}c@{\hspace{6pt}}c@{}c@{}c}

4673: \lambda^1&=&\!\left(\begin{array}{ccc}0&1&0\\1&0&0\\0&0&0\end{array}

4674: \right)\!\!,&

4675: \lambda^2&=&\!\left(\begin{array}{ccc}0&-i&0\\i&0&0\\0&0&0\end{array}

4676: \right)\!\!,&

4677: \lambda^3&=&\!\left(\begin{array}{ccc}1&0&0\\0&-1&0\\0&0&0\end{array}

4678: \right)\!\!,\\[27pt]

4679: \lambda^4&=&\!\left(\begin{array}{ccc}0&0&1\\0&0&0\\1&0&0\end{array}

4680: \right)\!\!,&

4681: \lambda^5&=&\!\left(\begin{array}{ccc}0&0&-i\\0&0&0\\i&0&0\end{array}

4682: \right)\!\!,&

4683: \lambda^6&=&\!\left(\begin{array}{ccc}0&0&0\\0&0&1\\0&1&0\end{array}

4684: \right)\!\!,\\[27pt]

4685: %\end{array}

4686: %\end{displaymath}

4687: %\begin{displaymath}

4688: %\begin{array}{c@{}c@{}c@{\hspace{6pt}}c@{}c@{}c@{\hspace{6pt}}c@{}c@{}c}

4689: \lambda^7&=&\!\left(\begin{array}{ccc}0&0&0\\0&0&-i\\0&i&0\end{array}

4690: \right)\!\!,&

4691: \lambda^8&=&

4692: \multicolumn{3}{l}{{\displaystyle\frac{1}{\sqrt{3}}}

4693: \!\left(\begin{array}{ccc}1&0&0\\0&1&0\\0&0&-2\end{array}\right)\!\!.}

4694: \end{array}

4695: \end{displaymath}

4696: They are normalized as

4697: $\mathrm{tr}\left(\lambda^a\lambda^b\right)=2\delta_{ab}$ and satisfy

4698: the commutation relations

4699: $\comm{\lambda^a}{\lambda^b}=2f^{abc}\lambda^c.$ The structure

4700: constants $f^{abc}$ are totally antisymmetric and obey Jacobi's

4701: identity %$f^{abc}f^{cde}+f^{bdc}f^{cae}+f^{dac}f^{cbe}=0$.

4702: \begin{displaymath}

4703: f^{abc}f^{cde}+f^{bdc}f^{cae}+f^{dac}f^{cbe}=0.

4704: \end{displaymath} 

4705: Explicitly, the non-vanishing structure constants are given by $f^{123}=i$,

4706: $f^{147}=f^{246}=f^{257}=f^{345}=-f^{156}=-f^{367}=i/2$,

4707: $f^{458}=f^{678}=i\sqrt{3}/2$, and 45 others obtained by permutations of 

4708: the indices.\\[-0.1pt]

4709: 

4710: \section{Eigenvalues of the quadratic Casimir operator}

4711: \label{quadraticcasimirs}

4712: 

4713: The eigenvalues of the quadratic Casimir operator for representations

4714: $\casin{\mu_1,\mu_2,\ldots,\mu_{n-1}}$ of SU($n$) up to $n=6$ are given by:

4715: \begin{itemize}

4716: \item[] $ \casinn{2}{\mu}=\frac{1}{4}\left(\mu^2+2\mu\right)=

4717:        \frac{\mu}{2}\left(\frac{\mu}{2}+1\right)$ 

4718: \item[] $\casinn{3}{\mu_1,\mu_2}=\frac{1}{3} \bigl( \mu_1^2 + \mu_1\mu_2 +

4719:        \mu_2^2 + 3\mu_1 + 3\mu_2 \bigr)$

4720: \item[] $\casinn{4}{\mu_1,\mu_2,\mu_3} =

4721:        \frac{1}{8}\left(3\mu_1^2+4\mu_2^2+

4722:          3\mu_3^2+4\mu_1\mu_2\right.$\\[3pt]

4723:        $ \phantom{\casinn{2}{\mu}} \left.

4724:       + 2\mu_1\mu_3 +4\mu_2\mu_3 + 12\mu_1+ 16\mu_2+12\mu_3 \right)$

4725: \item[] $\casinn{5}{\mu_1,\mu_2,\mu_3,\mu_4} = \frac{1}{5}\left( 2\mu_1^2 

4726:       + 3\mu_2^2 + 3\mu_3^2 + 2\mu_4^2 \right.$\\[3pt] 

4727: $\phantom{\casinn{2}{\mu}} + 3\mu_1\mu_2 + 4\mu_2\mu_3 + 3\mu_3\mu_4 + 

4728: 2\mu_1\mu_3 + \mu_1\mu_4$\\[3pt]

4729: $\phantom{\casinn{2}{\mu}} \left. + 2\mu_2\mu_4 + 10\mu_1

4730: + 15\mu_2 + 15\mu_3 + 10\mu_4 \right)$

4731: \item[] $\casinn{6}{\mu_1,\mu_2,\mu_3,\mu_4,\mu_5}= $\\[3pt]

4732: $\phantom{\casinn{2}{\mu}}\!\!\!\frac{1}{12}\left( 

4733:   5\mu_1^2 + 8\mu_2^2 + 9\mu_3^2 + 8\mu_4^2 + 5\mu_5^2 \right.$\\[3pt]

4734: $\phantom{\casinn{2}{\mu}}+ 8\mu_1\mu_2 + 12\mu_2\mu_3 + 12\mu_3\mu_4 

4735:   + 8\mu_4\mu_5  $\\[3pt]

4736: $\phantom{\casinn{2}{\mu}} + 4\mu_1\mu_4 + 6\mu_1\mu_3  + 8\mu_2\mu_4 + 6\mu_3\mu_5$\\[3pt]

4737: $\phantom{\casinn{2}{\mu}} + 4\mu_2\mu_5 + 2\mu_1\mu_5 + 30\mu_1 + 48\mu_2$\\[3pt]

4738: $\phantom{\casinn{2}{\mu}}\left. + 54\mu_3 + 48\mu_4 + 30\mu_5 \right)$\\

4739: \end{itemize}

4740: The general method to obtain these and further eigenvalues for $n>6$

4741: requires a discussion of representation theory~\cite{Humphreys87} at a level

4742: which is beyond the scope of this article.

4743: 

4744: The dimensionality of a representation $(\mu_1,\mu_2,\ldots,\mu_{n-1})$

4745: is determined by the so-called Hook formula~\cite{Cornwell84vol2} 

4746: %for SU($n$)

4747: \begin{equation}

4748:   \label{eq:hook-formula}

4749:   \text{dim}=\frac{\prod_{i<j}^n\left( \lambda_i - \lambda_j + j - i  \right)}

4750:              {\prod_{i<j}^n\left( j - i  \right)},

4751: \end{equation}

4752: where $\lambda_i=\sum_{j=i}^{n-1}\mu_j$ for $i=1,\ldots,n$. In particular, 

4753: it yields for $n=2,3,4$:

4754: 

4755: %\pagebreak

4756: \begin{itemize}

4757: \item[] $\dimn{2}{\mu}=\mu+1$

4758: \item[] $\dimn{3}{\mu_1,\mu_2}=\frac{1}{2}(\mu_1+1)(\mu_2+1)(\mu_1+\mu_2+2)$

4759: %\pagebreak

4760: \item[] \parbox{\linewidth}{$\dimn{4}{\mu_1,\mu_2,\mu_3}=\frac{1}{12}(\mu_1+1)(\mu_2+1)

4761:          (\mu_3+1)$\\[3pt]

4762:         $\phantom{\text{dimdim}}(\mu_1+\mu_2+2)(\mu_2+\mu_3+2)

4763:          (\mu_1+\mu_2+\mu_3+3)$}

4764: \end{itemize}

4765: 

4766: \begin{thebibliography}{10}

4767: 

4768: \bibitem{bethe31zp205}

4769: H. Bethe, Z. Phys. {\bf 71},  205  (1931), translated in D. C. Mattis, ed.,

4770:   \emph{The Many-Body Problem} (World Scientific, Singapore, 1993).

4771: 

4772: \bibitem{yang67prl1312}

4773: C.~N. Yang, Phys. Rev. Lett. {\bf 19},  1312  (1967).

4774: 

4775: \bibitem{baxter90}

4776: R.~J. Baxter, {\em Exactly Solved Models in Statistical Mechanics} (Academic

4777:   Press, London, 1990).

4778: 

4779: \bibitem{KorepinBogoliubovIzergin93}

4780: V.~E. Korepin, N.~M. Bogoliubov, and A.~G. Izergin, {\em Quantum Inverse

4781:   Scattering Method and Correlation Functions} (Cambridge University Press,

4782:   Cambridge, 1997).

4783: 

4784: \bibitem{faddeev-81pla375}

4785: L.~D. Faddeev and L.~A. Takhtajan, Phys. Lett.~A {\bf 85},  375  (1981).

4786: 

4787: \bibitem{laughlin83prl1395}

4788: R.~B. Laughlin, Phys. Rev. Lett. {\bf 50},  1395  (1983).

4789: 

4790: \bibitem{stone92}

4791: {\em Quantum Hall Effect}, edited by M. Stone (World Scientific, Singapur,

4792:   1992).

4793: 

4794: \bibitem{haldane83pla464}

4795: F.~D.~M. Haldane, Phys. Lett.~A {\bf 93},  464  (1983).

4796: 

4797: \bibitem{haldane83prl1153}

4798: F.~D.~M. Haldane, Phys. Rev. Lett. {\bf 50},  1153  (1983).

4799: 

4800: \bibitem{affleck90proc}

4801: I. Affleck,  in {\em Fields, strings and critical phenomena}, Vol.~XLIX of {\em

4802:   Les {H}ouches lectures}, edited by E. Br\'ezin and J. Zinn-Justin (Elsevier,

4803:   Amsterdam, 1990).

4804: 

4805: \bibitem{Fradkin91}

4806: E. Fradkin, {\em Field Theories of Condensed Matter Systems}, No.~82 in {\em

4807:   Frontiers in Physics} (Westview Press, Bouler, 1991).

4808: 

4809: \bibitem{GogolinNersesyanTsvelik98}

4810: A.~O. Gogolin, A.~A. Nersesyan, and A.~M. Tsvelik, {\em Bosonization and

4811:   Strongly Correlated Systems} (Cambridge University Press, Cambridge, 1998).

4812: 

4813: \bibitem{Giamarchi04}

4814: T. Giamarchi, {\em Quantum Physics in One Dimension} (Oxford University Press,

4815:   Oxford, 2004).

4816: 

4817: \bibitem{majumdar-69jmp1399}

4818: C.~K. Majumdar and D.~K. Ghosh, J.~Math. Phys. {\bf 10},  1399  (1969).

4819: 

4820: \bibitem{affleck-87prl799}

4821: I. Affleck, T. Kennedy, E.~H. Lieb, and H. Tasaki, Phys. Rev. Lett. {\bf 59},

4822:   799  (1987).

4823: 

4824: \bibitem{affleck-88cmp477}

4825: I. Affleck, T. Kennedy, E.~H. Lieb, and H. Tasaki, Commun. Math. Phys. {\bf

4826:   115},  477  (1988).

4827: 

4828: \bibitem{haldane88prl635}

4829: F.~D.~M. Haldane, Phys. Rev. Lett. {\bf 60},  635  (1988).

4830: 

4831: \bibitem{shastry88prl639}

4832: B.~S. Shastry, Phys. Rev. Lett. {\bf 60},  639  (1988).

4833: 

4834: \bibitem{haldane91prl1529}

4835: F.~D.~M. Haldane, Phys. Rev. Lett. {\bf 66},  1529  (1991).

4836: 

4837: \bibitem{haldane-92prl2021}

4838: F.~D.~M. Haldane, Z.~N.~C. Ha, J.~C. Talstra, D. Bernard, and V. Pasquier,

4839:   Phys. Rev. Lett. {\bf 69},  2021  (1992).

4840: 

4841: \bibitem{haldane91prl937}

4842: F.~D.~M. Haldane, Phys. Rev. Lett. {\bf 67},  937  (1991).

4843: 

4844: \bibitem{ha-93prb12459}

4845: Z.~N.~C. Ha and F.~D.~M. Haldane, Phys. Rev.~B {\bf 47},  12459  (1993).

4846: 

4847: \bibitem{essler95prb13357}

4848: F.~H.~L. E{\ss}ler, Phys. Rev.~B {\bf 51},  13357  (1995).

4849: 

4850: \bibitem{greiter-05prb224424}

4851: M. Greiter and D. Schuricht, Phys. Rev.~B {\bf 71},  224424  (2005).

4852: 

4853: \bibitem{greiter06prl}

4854: M. Greiter, Statistical Phases and Momentum Spacings for One-Dimensional

4855:   Anyons, {s}ubmitted to Phys. Rev. Lett.

4856: 

4857: \bibitem{greiter-06prl}

4858: M. Greiter and D. Schuricht, Many-spinon states and the secret significance of

4859:   Young tableaux, {s}ubmitted to Phys. Rev. Lett.

4860: 

4861: \bibitem{kawakami92prb1005}

4862: N. Kawakami, Phys. Rev.~B {\bf 46},  1005  (1992).

4863: 

4864: \bibitem{kawakami92prbr3191}

4865: N. Kawakami, Phys. Rev.~B {\bf 46},  R3191  (1992).

4866: 

4867: \bibitem{ha-92prb9359}

4868: Z.~N.~C. Ha and F.~D.~M. Haldane, Phys. Rev.~B {\bf 46},  9359  (1992).

4869: 

4870: \bibitem{bouwknegt-96npb345}

4871: P. Bouwknegt and K. Schoutens, Nucl. Phys.~B {\bf 482},  345  (1996).

4872: 

4873: \bibitem{schuricht-05epl987}

4874: D. Schuricht and M. Greiter, Europhys. Lett. {\bf 71},  987  (2005).

4875: 

4876: \bibitem{schuricht-06prb235105}

4877: D. Schuricht and M. Greiter, Phys. Rev. B {\bf 73},  235105  (2006).

4878: 

4879: \bibitem{jullien-83baps344}

4880: R. Jullien and F.~D.~M. Haldane, Bull. Am. Phys. Soc. {\bf 28},  344  (1983).

4881: 

4882: \bibitem{affleck89jpcm3047}

4883: I. Affleck, J. Phys.: Condens. Matter {\bf 1},  3047  (1989).

4884: 

4885: \bibitem{okamoto-92pla433}

4886: K. Okamoto and K. Nomura, Phys. Lett. A {\bf 169},  433  (1992).

4887: 

4888: \bibitem{eggert96prb9612}

4889: S. Eggert, Phys. Rev. B {\bf 54},  R9612  (1996).

4890: 

4891: \bibitem{white-96prb9862}

4892: S.~R. White and I. Affleck, Phys. Rev. B {\bf 54},  9862  (1996).

4893: 

4894: \bibitem{sen-07prb104411}

4895: D. Sen and N. Surendran, Phys. Rev. B {\bf 75},  104411  (2007).

4896: 

4897: \bibitem{knabe88jsp627}

4898: S. Knabe, J. Stat. Phys. {\bf 52},  627  (1988).

4899: 

4900: \bibitem{fannes-89epl633}

4901: M. Fannes, B. Nachtergaele, and R.~F. Werner, Europhys. Lett. {\bf 10},  633

4902:   (1989).

4903: 

4904: \bibitem{freitag-91zpb381}

4905: W.-D. Freitag and E. M\"uller-Hartmann, Z. Phys. B {\bf 83},  381  (1991).

4906: 

4907: \bibitem{kluemper-91jpal955}

4908: A. Kl\"umper, A. Schadschneider, and J. Zittartz, J. Phys. A {\bf 24},  L955

4909:   (1991).

4910: 

4911: \bibitem{kennedy-92prb304}

4912: T. Kennedy and H. Tasaki, Phys. Rev. B {\bf 45},  304  (1992).

4913: 

4914: \bibitem{kluemper-92zpb281}

4915: A. Kl\"umper, A. Schadschneider, and J. Zittartz, Z. Phys. B {\bf 87},  281

4916:   (1992).

4917: 

4918: \bibitem{kluemper-93epl293}

4919: A. Kl\"umper, A. Schadschneider, and J. Zittartz, Europhys. Lett. {\bf 24},

4920:   293  (1993).

4921: 

4922: \bibitem{batchelor-94ijmpb3645}

4923: M.~T. Batchelor and C.~M. Yung, Int. J. Mod. Phys. B {\bf 8},  3645  (1994).

4924: 

4925: \bibitem{schollwock-96prb3304}

4926: U. Schollw\"ock, T. Jolicoeur, and T. Garel, Phys. Rev. B {\bf 53},  3304

4927:   (1996).

4928: 

4929: \bibitem{kolezhuk-96prl5142}

4930: A. Kolezhuk, R. Roth, and U. Schollw\"ock, Phys. Rev. Lett. {\bf 77},  5142

4931:   (1996).

4932: 

4933: \bibitem{kolezhuk-02prb100401}

4934: A. Kolezhuk and U. Schollw\"ock, Phys. Rev. B {\bf 65},  100401(R)  (2002).

4935: 

4936: \bibitem{normand-02prb104411}

4937: B. Normand and F. Mila, Phys. Rev. B {\bf 65},  104411  (2002).

4938: 

4939: \bibitem{lauchli-06prb144426}

4940: A. L\"auchli, G. Schmid, and S. Trebst, Phys. Rev. B {\bf 74},  144426  (2006).

4941: 

4942: \bibitem{sutherland75prb3795}

4943: B. Sutherland, Phys. Rev.~B {\bf 12},  3795  (1975).

4944: 

4945: \bibitem{choi-82pla83}

4946: T.~C. Choy and F.~D.~M. Haldane, Phys. Lett.~A {\bf 90},  83  (1982).

4947: 

4948: \bibitem{Li-99prb12781}

4949: Y.~Q. Li, M. Ma, D.~N. Shi, and F.~C. Zhang, Phys. Rev.~B {\bf 60},  12781

4950:   (1999).

4951: 

4952: \bibitem{Gu-02prb092404}

4953: S.~J. Gu and Y.~Q. Li, Phys. Rev.~B {\bf 66},  092404  (2002).

4954: 

4955: \bibitem{damerau-06jsm12014}

4956: J. Damerau and A. Kl\"umper, J. Stat. Mech.  12014  (2006).

4957: 

4958: \bibitem{kawashima-07prl057202}

4959: N. Kawashima and Y. Tanabe, Phys. Rev. Lett. {\bf 98},  057202  (2007).

4960: 

4961: \bibitem{Paramekanti-06cm0608691}

4962: A. Paramekanti and J.~B. Marston, cond-mat/0608691.

4963: 

4964: \bibitem{affleck86npb409}

4965: I. Affleck, Nucl. Phys. B {\bf 265},  409  (1986).

4966: 

4967: \bibitem{affleck88npb582}

4968: I. Affleck, Nucl. Phys.~B {\bf 305},  582  (1988).

4969: 

4970: \bibitem{chen-05prb214428}

4971: S. Chen, C. Wu, S.~C. Zhang, and Y. Wang, Phys. Rev. B {\bf 72},  214428

4972:   (2005).

4973: 

4974: \bibitem{chen-06prb174424}

4975: S. Chen, Y. Wang, W.~Q. Ning, C. Wu, and H.~Q. Lin, Phys. Rev. B {\bf 74},

4976:   174424  (2006).

4977: 

4978: \bibitem{affleck-86lmp57}

4979: I. Affleck and E.~H. Lieb, Lett. Math. Phys. {\bf 12},  57  (1986).

4980: 

4981: \bibitem{greiter-07prb060401}

4982: M. Greiter, S. Rachel, and D. Schuricht, Phys.~Rev.~B {\bf 75},  060401(R)

4983:   (2007).

4984: 

4985: \bibitem{broek80pla261}

4986: P.~M. van~den Broek, Phys. Lett.~A {\bf 77},  261  (1980).

4987: 

4988: \bibitem{InuiTanabeOnodera96}

4989: T. Inui, Y. Tanabe, and Y. Onodera, {\em Group Theory and Its Applications in

4990:   Physics} (Springer, Berlin, 1996).

4991: 

4992: \bibitem{Cornwell84vol2}

4993: J.~F. Cornwell, {\em Group theory in physics} (Academic Press, London, 1984),

4994:   Vol.~II.

4995: 

4996: \bibitem{Georgi82}

4997: H. Georgi, {\em Lie Algebras in Particle Physics} (Addison-Wesley, Redwood

4998:   City, 1982).

4999: 

5000: \bibitem{schwinger65proc}

5001: J. Schwinger,  in {\em Quantum Theory of Angular Momentum}, edited by L.

5002:   Biedenharn and H. van Dam (Academic Press, New York, 1965).

5003: 

5004: \bibitem{Auerbach94}

5005: A. Auerbach, {\em Interacting electrons and quantum magnetism} (Springer, New

5006:   York, 1994).

5007: 

5008: \bibitem{greiter02prb134443}

5009: M. Greiter, Phys. Rev. B {\bf 65},  134443  (2002).

5010: 

5011: \bibitem{greiter02prb054505}

5012: M. Greiter, Phys. Rev. B {\bf 66},  054505  (2002).

5013: 

5014: \bibitem{macfarlane-68cmp77}

5015: A.~J. Macfarlane, A. Sudbery, and P.~H. Weisz, Commum. Math. Phys. {\bf 11},

5016:   77  (1968).

5017: 

5018: \bibitem{Humphreys87}

5019: J.~E. Humphreys, {\em Introduction to Lie Algebras and Representation Theory}

5020:   (Springer, New York, 1987).

5021: 

5022: \bibitem{affleck-91npb467}

5023: I. Affleck, D.~P. Arovas, J.~B. Marston, and D.~A. Rabson, Nucl. Phys. {\bf

5024:   B366},  467  (1991).

5025: 

5026: \bibitem{marston-89prb11538}

5027: J.~B. Marston and I. Affleck, Phys. Rev. B {\bf 39},  11538  (1989).

5028: 

5029: \bibitem{Jaksch-98prl3108}

5030: D. Jaksch, C. Bruder, J.~I. Cirac, C.~W. Gardiner, and P. Zoller, Phys. Rev.

5031:   Lett. {\bf 81},  3108  (1998).

5032: 

5033: \bibitem{Greiner-02n39}

5034: M. Greiner, O. Mandel, T. Esslinger, T.~W. H\"ansch, and I. Bloch, Nature {\bf

5035:   415},  39  (2002).

5036: 

5037: \bibitem{Honerkamp-04prl170403}

5038: C. Honerkamp and W. Hofstetter, Phys. Rev. Lett. {\bf 92},  170403  (2004).

5039: 

5040: \bibitem{Honerkamp-04prb094521}

5041: C. Honerkamp and W. Hofstetter, Phys. Rev. B {\bf 70},  094521  (2004).

5042: 

5043: \bibitem{rapp-06cm0607138}

5044: A. Rapp, G. Zar\'{a}nd, C. Honerkamp, and W. Hofstetter, cond-mat/0607138.

5045: 

5046: \bibitem{Duan-03prl090402}

5047: L.-M. Duan, E. Demler, and M.~D. Lukin, Phys. Rev. Lett. {\bf 91},  090402

5048:   (2003).

5049: 

5050: \bibitem{Garcia-04prl250405}

5051: J.~J. Garc\'{\i}a-Ripoll, M.~A. Martin-Delgado, and J.~I. Cirac, Phys. Rev.

5052:   Lett. {\bf 93},  250405  (2004).

5053: 

5054: \bibitem{micheli-06np341}

5055: A. Micheli, G.~K. Brennen, and P. Zoller, Nature Phys. {\bf 2},  341  (2006).

5056: 

5057: \bibitem{brennen-06qp0612180}

5058: G.~K. Brennen, A. Micheli, and P. Zoller, quant-ph/0612180.

5059: 

5060: \end{thebibliography}

5061: 

5062: \end{document}

5063: 

5064: