1: \documentclass[12pt,titlepage]{article}

2: \usepackage{graphicx}

3: \usepackage{psfrag}

4: \usepackage{amssymb}

5: %\documentclass[12pt]{article}

6: \usepackage{epsfig}

7: %\usepackage{showkeys}

8: %\usepackage{src}

9: % boldface math; use in math mode, e.g.: $\bm{a}\cdot\bm{b}$

10: \usepackage{bm}

11: \textheight=220truemm

12: \textwidth=160truemm

13: \setlength{\oddsidemargin}{0truemm}

14: \setlength{\evensidemargin}{0truemm}

15: \setlength\arraycolsep{2pt}

16: \setcounter{topnumber}{8}

17: \font\small=cmr8 scaled \magstep0

18: \font\smallish=cmr8 scaled \magstep1

19: \font\grande=cmr10 scaled \magstep4

20: \font\medio=cmr10 scaled \magstep2

21: \outer\def\beginsection#1\par{\medbreak\bigskip

22:       \message{#1}\leftline{\bf#1}\nobreak\medskip

23: \vskip-\parskip

24:       \noindent}

25: \def\obdot{\hskip-8pt \vbox to 11pt{\hbox{..}\vfill}}

26: \def\obbdot{\hskip-8pt \vbox to 14pt{\hbox{..}\vfill}}

27: \def\odot{\hskip-6pt \vbox to 6pt{\hbox{..}\vfill}}

28: \renewcommand\topfraction{1}

29: \newcommand{\eq}{\begin{equation}}

30: 

31: \newcommand{\eqx}{\end{equation}}

32: \newcommand{\eqn}{\begin{eqnarray}}

33: 

34: \newcommand{\eqnx}{\end{eqnarray}}

35: \newcommand{\bi}{\begin{itemize}}

36: 

37: \newcommand{\ei}{\end{itemize}}

38: 

39: \newcommand{\ad}{{a^{\dagger}}}

40: \newcommand{\fd}{f^{\dagger}}

41: \newcommand{\Ad}{{A^{\dagger}}}

42: \newcommand{\Qd}{{Q^{\dagger}}}

43: \newcommand{\bx}{\bar{x}}

44: \newcommand{\nn}{\nonumber}

45: \newcommand{\ra}{\rangle}

46: \newcommand{\la}{\langle}

47: \newcommand{\ar}{$\mapsto$}

48: 

49: \begin{document}

50: \titlepage

51: 

52: \begin{flushright}

53: \vspace{5mm}

54: CERN-PH-TH/2006-143\\

55: TPJU-09/2006

56: \end{flushright}

57: \vspace{15mm}

58: \begin{center}

59: 

60: 

61: \grande{ A supersymmetric

62: matrix model: \\

63: II. Exploring higher-fermion-number sectors}

64: 

65: \vspace{15mm}

66: 

67: \large{G. Veneziano}

68: 

69: \vspace{5 mm}

70: 

71:  {\sl Theory Division, CERN, CH-1211 Geneva 23, Switzerland }

72: 

73: {\sl and}

74: 

75: {\sl Coll\`ege de France, 11 place M. Berthelot, 75005 Paris, France}

76: \vspace{5 mm}

77: 

78: 

79:    \large{J. Wosiek}

80: 

81:    \vspace{5 mm}

82: 

83:    {\sl M. Smoluchowski Institute of Physics, Jagellonian University}

84: 

85: {\sl Reymonta 4, 30-059 Krak\'{o}w, Poland}

86: 

87: \end{center}

88: \vskip 10mm

89: \centerline{\medio  Abstract}

90: \vskip 4mm

91: \noindent

92: Continuing our previous analysis of a supersymmetric quantum-mechanical  matrix model,  we study in detail the properties of its  sectors with fermion number $F=2$ and $3$.

93: We confirm all

94: previous expectations, modulo the appearance, at strong coupling,  of {\it two} new bosonic ground states  causing a further

95: jump  in Witten's index  across  a previously identified critical

96: 't Hooft coupling $\lambda_c$. We are able to elucidate the origin of these new SUSY vacua

97: by considering the

98: $\lambda \rightarrow \infty$ limit  and a strong coupling expansion around it.

99: \vspace{5mm}

100: 

101: 

102: \vfill

103: \begin{flushleft}

104: CERN-PH-TH/2006-143 \\

105: TPJU-09/2006\\

106: July  2006\\

107: \end{flushleft}

108: 

109: 

110: \newpage

111: 

112: \section{Introduction}

113: 

114: In a previous paper \cite{VW1} (see also \cite{Adriano} for motivations and details) we have introduced and discussed a  new Hamiltonian approach

115: to large-$N$ (planar) theories and have illustrated its effectiveness in a simple supersymmetric 

116: quantum-mechanics  model. It is

117:  the $N \rightarrow \infty$ limit of an $N \times N$ matrix model

118: defined by the following supersymmetric charges and 

119: Hamiltonian: 

120: \eq Q= {\rm Tr} [f \ad(1+g\ad)] =  {\rm Tr} [f A^{\dagger}],

121: \;\;\; \Qd=   {\rm Tr} [\fd (1+g a) a]=   {\rm Tr} [\fd A]\, . \eqx \eq

122: H=\{Q^{\dagger},Q\} = H_B+H_F  \, ,\label{h1} \eqx \eq H_B=  {\rm Tr}

123: [\ad a + g(\ad^2 a + \ad a^2) + g^2 \ad^2 a^2] \, ,\label{h2} \eqx

124: \eqn

125: H_F&=&  {\rm Tr}  [\fd f + g ( \fd f (\ad+a) + \fd (\ad+a) f) \nonumber \\

126: & + & g^2 ( \fd a f \ad + \fd a \ad f + \fd f \ad a + \fd \ad f a)] \, ,\label{h3}

127: \eqnx

128: where bosonic and fermionic destruction and creation operators satisfy:

129: \eq

130: [a_{ij},\ad_{kl}]=\delta_{il}\delta_{jk}  \label{com} \, , \, \, \{f_{ij} \fd_{kl}\}=\delta_{il}\delta_{jk} \, ,

131: \eqx

132: all other (anti)commutators being zero.

133: 

134: This dynamical  system turned out to be quite interesting per se. Since its

135: Hamiltonian conserves fermion number $F =  {\rm Tr}[\fd f ]$, it can be analysed

136: $F$-sector by $F$-sector, with SUSY connecting sectors differing

137: by one unit of $F$. The task of pairing states in supermultiplets is facilitated by

138: the introduction of another conserved operator \cite{VW1}: \eq C

139: \equiv [\Qd, Q]    \, , ~  [C, H] =0 \, , ~ C^2 = H^2 \, . \label{C}

140: \eqx Eigenstates of $H$ with eigenvalue $E>0$ can be classified

141: according to their ``$C$-parity", i.e. according to whether $C =

142: \pm E$. We may also consider the combination: \eq C_F = (-1)^F ~

143: \frac{C}{E} \, , \eqx a good quantum number for each  SUSY

144: doublet.  States with $C=+E ~(-E)$ are  annihilated by $\Qd~ (Q)$.

145: All the states  in the $F=0,1$ sectors turn out to have $C_F= -1$

146: but, in higher-$F$ sectors,  there also are some ``unnatural"

147: states  with $C_F = +1$:  these are important for the full matching

148: of bosons and fermions.

149: 

150: 

151: In \cite{VW1} we analysed in detail the $F=0$ and $F=1$ sectors of the model. They turn out to provide  a complete  (although highly reducible) representation of SUSY and to exhibit a number of interesting features. We briefly summarize them hereafter:

152: \begin{itemize}

153: \item

154: There is a (discontinuous) phase transition, as a function of the

155: 't Hooft coupling $ \lambda \equiv g^2 N $, at $ \lambda =

156: \lambda_c =  1 $.  The mass/energy gap, which is present at $ \lambda < 1

157: $, disappears at $ \lambda = 1 $

158:  and reappears at $ \lambda >1 $.

159: \item

160: The Witten index \cite{WQM} (restricted to these two sectors) jumps by one unit at  $ \lambda = 1 $ as a result of the appearance, on top of the trivial Fock vacuum,

161:  of a new, normalizable zero-energy state at $ \lambda > 1 $. As a consequence, all higher

162:   supermultiplets rearrange at the transition point.

163: \item The system exhibits an  exact strong--weak duality in its spectrum, with  fixed point at $\lambda_c$: all excited eigenenergies at

164: $\lambda$ and $1/\lambda$  are connected by

165: a simple relation.

166: \item

167: The spectrum of the model can be computed analytically in terms of the zeros of a hypergeometric function.

168: This makes all the above properties explicit.

169: In particular, the details of the critical behaviour at $\lambda_c=1$ can be analysed.

170: \end{itemize}

171: 

172: In a subsequent paper \cite{OVW1}  some mathematical implications of the pairing of states at arbitrary

173:   $F$ and $B$ for the combinatorics of binary necklaces have been discussed.

174: Here

175:  we extend the analysis of \cite{VW1} to the $F=2$ and $F=3$ sectors.

176: As expected, unlike those with  $F=0$ and $F=1$, these two higher-$F$  sectors do {\it not} provide

177: a complete representation of SUSY (although they contain many of them). In other words, while all

178: $F=2$ states find their SUSY partner in the $F=3$ sector, the reverse is not true: some $F=3$ states

179: have no partner with  $F=2$; their partners  are  expected to lie, rather, in the $F=4$ sector.

180: The quantum number $C$, introduced above,  distinguishes the $F=3$ states that have SUSY partners with

181: $F=2$ from those with $F=4$

182: companions.

183: Thus the linear combinations of $F=3$ fermions that should be

184: degenerate with $F=2$ bosons can be neatly identified, at least for sufficiently weak coupling,

185: and these expectations can be compared with actual numerical calculations.

186: 

187: While we  find no surprises at weak coupling ($\lambda < 1$), the numerical spectrum at strong

188: coupling ($\lambda >1$) leads to something  unexpected: {\it two new} zero-energy bosonic states pop up,

189:  causing Witten's index to jump by two units at the $\lambda = 1$ phase transition.

190: 

191: The outline of the paper is as follows:

192: in the next section we construct the single-trace (planar) basis with $F=2,3$ and compare  it with

193: the  $F=0,1$ case discussed in  \cite{VW1,Adriano}.  We also derive explicit expressions for the leading-order  matrix elements of the Hamiltonian.

194: The increasing complexity (and some emerging regularities), as we move to higher fermionic sectors,

195: will be emphasized. Section 3 contains the detailed discussion of the physics based on

196: the numerical diagonalization of the planar Hamiltonian. Section 4 provides  an analytic construction

197: of the new susy vacua, first at infinite 't Hooft coupling, and then in the whole strong coupling phase via an expansion in $1/ \lambda$. We end by summarizing the main results of this work.

198: 

199: \section{$F = 2,3 $ Fock states and matrix elements at large $N$}

200: Since the Hamiltonian and SUSY charges are simple polynomials in creation and annihilation (c/a) operators,

201: it is advantageous to work in the gauge-invariant eigenbasis of the occupation number operators $B= {\rm Tr} [\ad a]$

202: and $F= {\rm Tr} [\fd f]$.

203: At infinite $N$, this basis substantially simplifies. Only  states created by single traces of fermionic and bosonic

204: operators give  leading-$N$ contributions. For this reason, the $F=0$ and $F=1$ Hilbert spaces were spanned by  basis vectors

205: labelled by just a single integer $n$ --the number of bosonic quanta. Extension to higher fermionic sectors is straightforward

206: but requires a little care. A generic state with two fermions has a form

207: \eqn

208: |n_1,n_2\ra = \frac{1}{{\cal N}_{n_1 n_2}} {\rm Tr} [\ad^{n_1}\fd\ad^{n_2}\fd]|0\ra ,\;\;\;n_1,n_2 \geq 0 \, ,

209: \eqnx

210: with a known normalization constant to be discussed shortly. Owing to the cyclic symmetry of the trace,

211: states differing by the interchange of $n_1$ and $n_2$ are linearly dependent (in fact $|n_1,n_2\ra = - |n_2,n_1\ra$

212: in this case). This fermionic minus sign has yet another consequence: there are no states with $n_1=n_2$. Therefore

213: the $F=2$ basis can be taken to be

214: \eqn

215: |n_1,n_2\ra,\;\;\; 0\leq n_1 < n_2 .

216: \eqnx

217: 

218: Similarly the basis states with three fermions are taken as

219: \eqn

220: |n_1,n_2,n_3\ra = \frac{1}{{\cal N}_{n_1 n_2 n_3}} {\rm Tr} [\ad^{n_1}\fd\ad^{n_2}\fd \ad^{n_3}\fd ]|0\ra,\;\;\;\;

221:  0\leq n_1 ,\;\;\;\;\ n_1 \leq n_2,\;\;\;\;\ n_1 \leq n_3.  \label{bf3}

222: \eqnx

223: Notice that cyclic symmetry does not imply any ordering of $n_2$ and $n_3$. Also, the cases where all (or some) of the bosonic occupation numbers coincide are not excluded in this sector.

224: All the above states are orthogonal in the planar limit, i.e. the off-diagonal elements of the inner-product matrix are subleading.

225: 

226: Matrix elements of the Hamiltonian (\ref{h1}--\ref{h3}) also simplify considerably at infinite $N$.

227: The ``planar rules"  for  obtaining  the leading contributions were formulated and illustrated in detail in

228: Refs.\cite{VW1} and \cite{Adriano}. In short: one uses  Wick's theorem, keeping only the colour contractions

229: that  give the maximal number of colour-index  loops, hence the highest power of $N$.

230: They come from the planar configurations of all c/a operators, as in the celebrated case of Feynman diagrams

231: \cite{'tH}.

232: 

233: The simplest application of the above rules is the calculation of the normalization factors. One obtains for $F=2$

234: \eqn

235: {\cal N}_{n_1 n_2} = \sqrt{N^{n_{{\rm tot}}}},

236: \eqnx

237: where $n_{tot}=n_1+n_2+2$ is the total number of quanta in a given state $|n_1,n_2\ra$.

238: 

239: There is more structure in the three-fermion sector:

240: \eqn

241: {\cal N}_{n_1 n_2 n_3} = \sqrt{d} \sqrt{N^{n_{{\rm tot}}}},\;\;\;n_{{\rm tot}}=n_1+n_2+n_3+3,

242: \eqnx

243: with the degeneracy factor

244: \eqn

245: d=\left\{ \begin{array}{cc}

246:       3 & {\rm if} \;\; n_1=n_2=n_3, \\

247:       1 & {\rm otherwise}.

248:       \end{array}

249:       \right.

250: \eqnx

251: 

252: We are now ready to calculate the complete matrix elements of various terms of the Hamiltonian (\ref{h1}--\ref{h2}).

253: The algebra is straightforward, though a little tedious. We stress that the above normalization

254: factors are crucial. 

255: Our final result  for the $F=2$ sector reads:

256: \eqn

257: \langle n_1,n_2|H|n_1,n_2\rangle = (n_1+n_2+2)(1+b^2) - b^2 (2 - \delta_{n_1,0}) - 2 b^2 \delta_{n_2,n_1+1}, \\

258: \langle n_1+1,n_2|H|n_1,n_2\rangle = b (n_1+2) = \langle n_1,n_2|H|n_1+1,n_2\rangle,\\

259: \langle n_1,n_2+1|H|n_1,n_2\rangle = b (n_2+2) = \langle n_1,n_2|H|n_1,n_2+1\rangle.\\

260: \langle n_1+1,n_2-1|H|n_1,n_2\rangle  = \langle n_1,n_2|H|n_1+1,n_2-1\rangle = 2 b^2 (1-

261:  \delta_{n_2,n_1+1}) \, , \label{last}

262: \eqnx where, as in  \cite{VW1}, we have introduced, for

263: convenience, $b \equiv \sqrt{\lambda}$. Note the

264: $\delta_{n_2,n_1+1}$ exception in the first and last equations.

265: For this configuration of the bosonic occupation numbers 

266: Eq. (\ref{last}) gives rise to a diagonal matrix element. The

267: $\delta_{n_2,n_1+1}$ function prevents double counting of this

268: contribution. Notice also that it comes with the opposite sign

269: to that suggested in Eq. (\ref{last}). This is because the resulting

270: state $\langle n_1+1,n_2-1|$  equals  minus the state $\langle

271: n_2-1,n_1+1|$, which belongs to our basis.

272: 

273: In the $F=3$ sector we obtain:

274: \eqn

275: \langle n_1,n_2, n_3|H|n_1,n_2,n_3\rangle = (n_1+n_2+n_3 +3)(1+b^2) 

276: - b^2 (3 - \delta_{n_1,0} -

277: \delta_{n_2,0}- \delta_{n_3,0})

278: \label{diag3} , \\

279: \langle n_1+1,n_2,n_3|H|n_1,n_2,n_3\rangle = \langle n_1,n_2,n_3|H|n_1+1,n_2,n_3\rangle = b (n_1+2) \Delta

280: \label{off31} \, , \\

281: \langle n_1+1,n_2-1,n_3|H|n_1,n_2, n_3\rangle  =  \langle n_1,n_2,n_3|H|n_1+1,n_2-1, n_3\rangle b^2 \Delta

282: \label{off32} \, ,

283: \eqnx

284: as well as cyclic permutations of (\ref{off31}), (\ref{off32}), where $\Delta=1/\sqrt{3}$ if $n_1=n_2=n_3$ and $\Delta=\sqrt{3}$ if the final state is of this form;

285: otherwise $\Delta=1$.

286: Similarly to the $F=2$ case, for some sets of $n_i$, Eq. (\ref{off32}) gives rise to diagonal

287: elements. However, the $F=3$ Hilbert space is considerably richer that the $F=2$ one and many other ``degenerate

288: cases" occur here as well.

289: Instead of dealing explicitly with all these ``exceptional" configurations, we adopt a simple, unifying rule:

290: \begin{center}

291: \begin{minipage}{12cm}

292: All contributions listed in (\ref{diag3}--\ref{off32}) should be added to the appropriate locations

293: in the $H_{IJ}$ table ($I$ and $J$ are the linear, composite indices labelling all states satisfying

294: (\ref{bf3})).

295: \end{minipage}

296: \end{center}

297: 

298: This rule also covers other ``degenerate" situations. For example,

299: for $n_1=n_2=n_3$ the ``plus cyclic" qualifier implies that the same matrix element receives an additional

300: factor 3. According to our rule, however, the same elementary matrix element should be added three times, which is of course equivalent.

301: 

302: 

303: \section{Results}

304: 

305: As for  the  $F=0$ and $F=1$ cases, the complete spectrum with two and three fermions was obtained by diagonalizing numerically the

306: above-mentioned  Hamiltonian matrices. To this end, we have introduced a cutoff $B_{{\rm max}}$ that limits the total number of

307: bosonic quanta in the system:

308: \eqn

309: B \leq B_{{\rm max}} \, .

310: \eqnx

311: Such a cutoff was found very useful in many applications \cite{JW}, \cite{CW}, since a) it can be easily implemented

312: in our bases, b) it preserves many symmetries in more complex models, and c) the spectra converge

313:  well with increasing $B_{{\rm max}}$.

314: 

315: \subsection{Dynamical supermultiplets}

316: The situation is not different in the present case. At $\lambda=2.0$, for example, the lower

317: (i.e. first 20) levels have converged to five decimal

318: digits for cutoffs $B_{{\rm max}}=40~(30) $ for $F=2~ (3) $. Figure 1 compares the spectra

319: in all four fermionic sectors, displaying clearly similarities and differences between the $F=0, 1$ and the

320:  $F=2, 3$ cases. Supersymmetry, which is broken by the cutoff, is nicely restored -- there is an excellent

321:  boson fermion degeneracy already at the above or higher values of $B_{{\rm max}}$. Interestingly, while all $F=0$ states had their

322:  supersymmetric counterparts in the $F=1$ sector and {\em vice versa}, for higher fermionic numbers this is not so.

323:  Every state with  $F=2$ has its partner in the $F=3$ sector, but the reverse is not true:  some states with  $F=3$ are ``unpaired" in Fig. 1 and, consequently, must have their counterparts in the $F=4$ sector. This is already expected

324:  at  the level of counting basis states in respective Hilbert spaces;  however, Fig. 1 provides a clear dynamical

325:  confirmation of this structure.

326: 

327: 

328: \begin{figure}[tbp]

329: %\begin{center}

330: \psfrag{yyy}{${\scriptstyle E}$ }\psfrag{ttt}{${\scriptstyle\lambda=2.0}$}

331: \psfrag{xx1}{${\scriptstyle F=0\hspace*{2.5cm}F=1}$}\psfrag{tt1}{${\scriptstyle\lambda=0.5}$}

332: \psfrag{fff}{${\scriptstyle F=2\hspace*{2.5cm}F=3}$}

333: \epsfig{width=8cm,file=levels01.eps}

334: \epsfig{width=8cm,file=levels23.eps}

335: %\end{center}

336: 

337: \vskip-4mm \caption{Low-lying bosonic and fermionic levels in the

338: first four fermionic sectors.} \label{fig:susyf01}

339: \end{figure}

340: 

341: Apart from supersymmetry, the spectra in higher-$F$ channels are obviously much richer than

342:  those for $F=0,1$. In fact in all earlier cases we know of, the energy levels were almost

343:  equidistant in the  large-$N$ limit \cite{MP},\cite{MO}. This also happened for  $F=0,1$ in our model, possibly suggesting

344:   some generic simplification of planar dynamics, but is no longer true with more fermions. In fact, some

345:  levels are so close that they seem to merge into one because of the poor resolution of the graph

346:  (e.g. the levels \# 4,5 with $F=2$ and \# 4,5 with $F=3$\footnote{This splitting has been clearly established and is

347:  {\em not} a finite cutoff effect.}).  Interestingly, both members of the $\# 4, 5$, $F=2$ doublet find their

348:  partners  within the $F=3$ sector.  However, this is by no means a general rule (cf. \# 10,11 with $F=3$ ).

349: 

350: \subsection{Phase transition, rearrangement and new SUSY vacua}

351: One of the remarkable features of our system, seen also

352: analytically in the $F=0,1$ sectors, was the phase transition at

353: $\lambda = \lambda_c=1$. At this point the spectrum became

354: gapless, and a critical slowing down of the cutoff dependence was

355: observed. The same phenomenon occurs in the $F=2$ and $F=3$

356: sectors. In Fig. 2 we show the $\lambda$ dependence of the few

357: lowest levels, with two and three gluinos. The critical slowing

358: down of the convergence is clearly observed. Close to the critical

359: point, supersymmetry is visibly broken, at fixed $B_{{\rm max}}$, while

360: away from $\lambda=1$ supermultiplets are well formed.

361: 

362: Interestingly, SUSY partners rearrange themselves across the phase-transition point,

363: as in the $F=0,1$ case.  A novel feature, however, is that, in the strong coupling phase,  {\it two} zero-energy states appear in the $F=2$ sector. This is to be compared with the $F=0$ case where the  trivial Fock vacuum (present at all

364: $\lambda$) is  accompanied by a second $E=0$ state in the strong coupling phase.

365: 

366: The claim that  all the curves in Fig. 2 converge  to zero at

367: $\lambda=1$ may seem a little premature.  However, we have repeated

368: our  analysis for a few values of $B_{{\rm max}}$ and established that

369: this is the most likely scenario. This point is now being carefully

370: and quantitatively studied  by P. Korcyl. The analytic results

371: discussed in the next section also confirm this conclusion.

372: 

373: 

374: 

375: \begin{figure}[tbp]

376: \psfrag{xxx}{${\scriptstyle\sqrt{\lambda}}$} \psfrag{yyy}{${\scriptstyle E}$ }

377: \psfrag{ttt}{${\scriptstyle F=2\;\; {\rm and}\;\; F=3}$}

378: \begin{center}

379: \epsfig{width=8cm,file=rearrange23.eps}

380: \end{center}

381: \vskip-4mm

382: \caption{Rearrangement of the $F=2$ (red) and $F=3$ (black) levels while passing through the critical

383: coupling $\lambda_c=1$; $B_{{\rm max}} = 30~ (25)$ for $F=2 ~(3)$.}

384: \label{fig:cfplot}

385: \end{figure}

386: 

387: \subsection{Supersymmetry fractions}

388: 

389: Apart from the unbalanced vacua and boson--fermion degeneracies of higher states, supersymmetry manifests itself

390: in yet a different but important way. Namely, the supersymmetry charges $Q$ and $\Qd$ transform members of the supermultiplets

391: into each other. Since our method provides not only the eigenenergies, but also the eigenstates, we can directly

392: check these relations. To this end define the ``supersymmetry fractions" \cite{CW}

393: \eqn

394: q_{mn}=\sqrt{\frac{2}{E_n+E_m}}\la F=3,E_m|\Qd|F=2,E_n\ra,

395: \eqnx

396: which are the, suitably normalized,  coordinates of the supersymmetric image of  the $n$-th, $F=2$, eigenstate in the $F=3$ sector.

397: For unbroken supersymmetry, $q_{mn}=\delta_{mn}$ (with an appropriate labelling of the $F=3$ states). For finite cutoff, however,

398: the supermultiplets are not yet well formed and $q's$ are not unity \footnote{Even with  unbroken supersymmetry,

399: supersymmetry fractions may not be exactly 1. This happens when there is an exact degeneracy {\em between supermultiplets}.

400: In such a situation $q's$ measure directly the mixing angles of the members of supermultiplets.}.

401: Figure 3 shows the cutoff dependence of the first few supersymmetry fractions. For low cutoffs they can vary rather irregularly until  the partner is rapidly found  and  convergence takes place  up to a small noise.

402: 

403: \begin{figure}[tbp]

404: \psfrag{yyy}{$q_n+n-1$}\psfrag{xxx}{$B_{max}$}

405: \begin{center}

406: \epsfig{width=8cm,file=sfrac.eps}

407: \end{center}

408: \vskip-4mm \caption{First five supersymmetry fractions, defined as

409: $q_n={\rm Max}_m(|q_{mn}|)$, for a range of cutoffs.} \label{fig:susyf01}

410: \end{figure}

411: 

412: \subsection{Restricted Witten index}

413: 

414: The effect of the new SUSY vacua on the Witten index is best seen if we restrict

415: the sum

416: \eqn

417: W(T,\lambda)= \sum_i (-1)^{F_i} e^{-T E_i}

418: \eqnx

419: to the states in the $F=2$ and $F=3$ sectors and exclude the $C=-E, F=3$ states. In practice we have used

420: the supersymmetry fractions and identified the SUSY partners for each of the $F=2$ states.

421: This procedure, performed necessarily at finite cutoff, is unambiguous away from the transition point. Close to

422: $\lambda_c$, however,  supersymmetry is badly broken at any finite $B_{{\rm max}}$, and a definition of  what is meant by the

423: ``energy of the supersymmetric partner" must be given. We made the following choice:

424: \eqn

425: W_R(T,\lambda)= \sum_i  \left( e^{-T E_i} -  e^{-T \bar{E}_i} \right),\;\;\;

426: \bar{E}_i=\frac{\sum_f E_f |q_{fi}|^2}{\sum_f |q_{fi}|^2} \, ,

427: \eqnx

428: where the $i(f)$ indices run over all states in the $F=2~(3)$ sectors. In another words, we take for the

429: energy of the partner the weighted average over all $F=3$ states with the weight given by the

430: supersymmetric fractions.

431: Away from $\lambda_c$ this selects the true supersymmetric partners and

432: automatically excludes  the $C=-E, F=3$ states.

433: The index defined in this way is a smooth function of the 't Hooft coupling (see Fig. 4) showing

434: the onset of the discontinuity caused by the new vacua in the strong coupling phase.

435: 

436: \begin{figure}[tbp]

437: \psfrag{yyy}{${\scriptstyle I_W(6)}$}\psfrag{xxx}{${\scriptstyle \sqrt{\lambda}}$}

438: \begin{center}

439: \epsfig{width=8cm,file=wi.eps}

440: \end{center}

441: \vskip-4mm \caption{Behaviour of the restricted Witten index, at $T=6$, around the phase transition.

442: The various curves correspond to $8 \le B_{{\rm max}} \le 14 $.} \label{fig:susyf01}

443: \end{figure}

444: 

445: \section{Explicit construction of the strong-coupling vacua}

446: The (numerical)  appearance of two new SUSY vacua above $ \lambda =1$ was quite unexpected.

447: In order to understand their origin we first consider the limit of infinite  't Hooft coupling.

448: As we will  see,  the necessity of  two new vacua comes out

449: very simply at  $ \lambda = \infty$. The corresponding eigenvectors also greatly simplify in that limit

450: (they have projections only on

451:  one or two Fock states).

452:  We will then  show, by an explicit construction,  that those  vacua

453: persist even at finite (but sufficiently large)   $ \lambda$, although the corresponding eigenvectors will contain an infinite superposition of Fock states.

454: 

455: We wish to mention that the  $ \lambda \rightarrow \infty$ limit also simplifies the discussion of our model at arbitrary $F$, since the Hamiltonian becomes block-diagonal with finite-dimensional blocks.

456: Furthermore, some restricted supertraces can be used to identify (and/or guess) the blocks

457: where  new SUSY vacua are expected. This result was exploited in Ref. \cite{OVW1} and will be

458: described in more detail elsewhewe. 

459: 

460: 

461: \subsection{SUSY vacua at infinite $ \lambda$}

462: 

463: 

464: The limit $ \lambda  \rightarrow \infty$ of our Hamiltonian is actually infinite. However, we can  define

465: appropriately rescaled SUSY charges that have a finite limit and that, consequently, define a finite rescaled Hamiltonian. Of course eigenvectors will be invariant under such an overall rescaling.

466: The strong-coupling limit of our rescaled SUSY charges reads simply:

467: \eqn

468:  Q_{\rm{strong}}=\lim_{\lambda\rightarrow\infty} \frac{1}{\sqrt{\lambda}} Q =\frac{1}{\sqrt{N}} {\rm Tr} (\ad^2 f),\;\;\;\;

469: Q_{\rm{strong}}^{\dagger}=\lim_{\lambda\rightarrow\infty} \frac{1}{\sqrt{\lambda}} \Qd =\frac{1}{\sqrt{N}} {\rm Tr} (a^2 \fd).

470: \eqnx

471: 

472: The peculiar property of these charges is that they change $F$ and $B$ but preserve the

473: combination $2F+B$ (unlike the zero-coupling charges that preserve $F+B$).

474: In other words, if we represent our base states on a grid having coordinates $F$ and $F+B$, the

475:  infinite $ \lambda$ charges connect states along a 45-degree diagonal (while the zero-coupling charges connect  states on lines parallel to the $F$ axis).

476:  A necessary condition for having no zero-energy states along these diagonals is that the corresponding supertrace be zero. Conversely, if the supertrace is non-zero, there must be some unpaired SUSY vacua lying on those diagonals.

477: 

478:  A very simple inspection immediately shows that the latter is the case for the diagonal that contains

479:  the state with $F=0$ and $B=1$ (the strong coupling vacuum of \cite{VW1}) as well as for those that contain either the $F=2, B=1$ or the  $F=2, B=3$ states.

480:  The latter are our new SUSY vacua at  infinite $ \lambda$.

481: 

482:  Let us now consider the  infinite $ \lambda$ limit of the (rescaled)  Hamiltonian (\ref{h1})--(\ref{h3}):

483: \eqn

484:  H_{\rm{strong}}=\lim_{\lambda\rightarrow\infty} \frac{1}{\lambda} H =

485:  \frac{1}{N}{\rm Tr} [ \ad^2 a^2 + \fd a f \ad + \fd a \ad f + \fd f \ad a + \fd \ad f a] \, .

486: \eqnx

487:  This result can be simplified further,  providing a  good illustration

488:  of our planar rules.

489: Namely, the third and  fourth terms must be brought to the normal form. Explicitly

490: \eqn

491: \fd_{ij} a_{jk} \ad_{kl} f_{li} &=& \fd_{ij}(\ad_{kl}a_{jk} + \delta_{jl}\delta_{kk}) f_{li}=N {\rm Tr} [\fd f],\\

492: \fd_{ij} f_{jk} \ad_{kl} a_{li} &=& \fd_{ij} \ad_{kl} f_{jk}  a_{li} = 0 \, ,

493: \eqnx

494: since the neglected terms do not yield a single trace.

495: The strong-coupling Hamiltonian thus reads

496: \eqn

497: H_{{\rm SC}}= \fd f +

498:  \frac{1}{N}{\rm Tr} [\ad^2 a^2 + \ad \fd a f + \fd\ad f a] \, .  \label{hstr}

499: \eqnx

500: Remarkably, it conserves {\em both} $F$ {\em and} $B$.

501: 

502: 

503: 

504: \subsection{ Strong-coupling vacua from infinite-coupling vacua}

505: 

506: In the $F=0$ sector we were able to give an explicit construction of the non-trivial vacuum  at $\lambda >1$. Actually, the new zero-energy state could be written down for any value of

507: $\lambda$ but was only normalizable in the strong-coupling region. We may ask whether a similar

508: analysis can be  carried out in the $F=2$ sector.  The $F=0$ and  $F=2$ sectors share the property that all their states are annihilated by $Q$. Therefore, null states  with  $F=0,2$ are characterized by the fact that they are  annihilated by $\Qd$. Indeed, the $F=0$ null state can  be easily obtained in this way by writing:

509: \eqn

510: Q &=& Q_w + Q_s \, , \, ~ \Qd = \Qd_w + \Qd_s  \, , \, \nonumber \\

511: Q_w &=&  {\rm Tr}  [f \ad]  \, , \,  ~ Q_s = g {\rm Tr}  [f \ad^2]

512: \;\;\; \Qd_w=  {\rm Tr}  [\fd  a]  \; , \;\; \Qd_s =  g {\rm Tr} [\fd a^2] \, ,

513: \eqnx

514: where the labels $w$, $s$ refer to the weak and strong coupling forms of the supersymmetric charges, respectively. Note that $\{Q_w,Q_s\} =0$ and that, if we define $H_w = \{Q_w,\Qd_w\}$, $H_s = \{Q_s,\Qd_s\}$, we also have $[H_w, Q_w] = [H_s, Q_s]=0 $, and similarly with $Q \rightarrow \Qd$.

515: 

516: Given the action of  $\Qd_w$ and $\Qd_s$ on  $F=0$ states:

517: \eq

518: \Qd_w |n \rangle = \sqrt{n} |n-1,F=1 \rangle \, , \, ~ \Qd_s |n \rangle = b \sqrt{n} |n-2, F=1 \rangle \, ,

519: \eqx

520: we see immediately that:

521: \eq

522: |v, F=0 \rangle  = \sum_{n=1}^{\infty}\frac{ ( -b)^{-n}}{\sqrt{n}}   |n \rangle

523: \eqx

524: is annihilated by $\Qd$. Clearly, it is normalizable only for $b >1$.

525: 

526: A very similar construction works  for one of the two strong-coupling ground states.

527: Let us look indeed for a state consisting of a linear superposition of $F=2$ states of the form:

528: \eq

529: |v, F=2\rangle _1 = \sum_{n=1}^{\infty}  c_n  |0, n \rangle \, .

530: \eqx

531: This time we find:

532: \eq

533: \Qd_s |0,1 \rangle =0 ~,~  \Qd_w |0, n \rangle = b^{-1} \Qd_s|0, n+1 \rangle \, ,

534: \eqx

535: and therefore a zero-energy state is simply:

536: \eq

537: |v, F=2\rangle _1   = \sum_{n=1}^{\infty} ( -b)^{-n}  |0, n \rangle \, .

538: \eqx

539: Note that the above state is written as a Taylor series in

540: $1/b$ starting from one of the two infinite-coupling ground states:

541: \eq

542: |v, F=2\rangle _1^{\infty} = |0,1 \rangle \, .

543: \eqx

544: Like its analogue in the $F=0$ sector,  this null state is only normalizable  if $b>1$.

545: 

546: For a second independent $E=0$ state,  let us start our series from the other infinite-coupling vacuum:

547: \eq

548:  |v, F=2\rangle _2^{\infty}\equiv|0,3\rangle - 2 |1,2\rangle \rightarrow \, |v, F=2\rangle _2

549: \label{SCnull}

550: \eqx

551: and use the following theorem:

552: 

553: \begin{center}

554: \begin{minipage}{12cm}

555:  Given an $F=2$ state annihilated by $\Qd_s$ (e.g.  (\ref{SCnull})) the state:

556: \eq

557: |v, F=2\rangle _2 \equiv \left(1 +  Q_s H_s^{-1} \Qd_w\right)^{-1} |v, F=2\rangle _2^{\infty}

558: \eqx

559: is annihilated by $\Qd$ and is therefore a null state.

560: \end{minipage}

561: \end{center}

562: 

563: 

564: The proof is easily carried out by expanding the fraction in powers of $1/b$ (NB: $Q_s H_s^{-1} = O(1/b) $)  and by noting that the first term is annihilated (by construction)  by $\Qd_s$ while the other terms cancel pairwise.  Indeed:

565: \eqn

566: &&  \Qd_s \left(Q_s H_s^{-1} \Qd_w\right)^{n +1} = ( \Qd_w  + H_s^{-1}Q_s \Qd_w \Qd_s ) \left(Q_s H_s^{-1} \Qd_w\right)^n \nonumber \\ &=& \Qd_w  \left(Q_s H_s^{-1} \Qd_w\right)^n +

567: H_s^{-1}Q_s \Qd_w Q_s H_s^{-1} \Qd_w \Qd_s \left(Q_s H_s^{-1} \Qd_w\right)^{n-1}\nonumber \\

568: &=& \dots =

569:  \Qd_w  \left(Q_s H_s^{-1} \Qd_w\right)^n \, ,

570: \eqnx

571: where we have used $\Qd_w^2 =0$ and have  iterated the procedure until $\Qd_s$ annihilates the state (\ref{SCnull}).

572: 

573: The above construction  can be generalized to the case in which the states in the chosen sector are not

574: necessarily annihilated by $Q_w$ and/or $Q_s$. This is relevant in sectors with $F>2$.

575: The claimed generalization is that the following state is annihilated by $\Qd$ {\it and } $Q$:

576: \eq

577: |v \rangle  \equiv \left(1 +  Q_s H_s^{-1} \Qd_w + \Qd_s H_s^{-1} Q_w \right)^{-1} |v \rangle^{\infty}\,.

578: \eqx

579: 

580:  The proof (left as an excercise) is facilitated by the following two identities:

581:  \eqn

582:  \Qd_w \Qd_s  \left( Q_s H_s^{-1} \Qd_w + \Qd_s H_s^{-1} Q_w \right)^n &=& \Qd_w  \left( Q_s H_s^{-1} \Qd_w  \right)^n \Qd_s  \nonumber \\

583:  Q_w Q_s  \left( Q_s H_s^{-1} \Qd_w + \Qd_s H_s^{-1} Q_w \right)^n &=& Q_w  \left( \Qd_s H_s^{-1} Q_w  \right)^n Q_s \, .

584:  \eqnx

585: 

586: 

587: 

588: \section{Summary}

589: 

590: In this paper we have extended our previous work  on the $F =0, 1$ sectors  of a supersymmetric matrix model \cite{VW1}  to sectors with fermion number $F=2, 3$.

591:  In Section 1 we have summarized our previous results:  hereafter we will do the same for the two new sectors, underlining similarities and differences.

592: 

593: \bi

594: 

595: \item   As in the $F=0,1$ case,  here too  there is a phase transition at  $\lambda = \lambda_c=1$: while the spectrum is discrete below and above $\lambda_c$, the energy-gap

596: disappears (the spectrum becomes continuous) exactly at

597: $\lambda_c$.

598:  There is also, in both cases, a critical slowing down of convergence (as a function of

599:  the cutoff $B_{{\rm max}}$)  in the vicinity of $\lambda_c$.

600: 

601: \item Again in analogy with  what was found in the $F=0,1$ sectors, supersymmetry, which is broken by the cutoff, is quickly restored for

602: $\lambda\ne\lambda_c$,  giving  exact degeneracy of fermionic and bosonic

603: eigenstates. This degeneracy is indeed SUSY-driven:  with the aid of suitably defined

604: ``supersymmetry fractions", we have verified that the degenerate eigenstates are supersymmetric

605: images of each other.

606: 

607: \item While in the low-coupling phase there are no zero-energy eigenstates in these sectors,

608: {\it two} such states appear for $\lambda > \lambda_c$ in the  $F=2$ sector (with a consequent jump of Witten's index by two units). This is similar (but not identical)

609:  to the $F=0,1$ case, where the empty Fock state (a zero-energy eigenstate for all

610:  values of $\lambda$) is accompanied  by another, non-trivial SUSY vacuum, just  in the

611:  strong-coupling phase. Such a  ``popping up" of new SUSY vacua is only possible,  at finite cutoff,

612: thanks to explicit supersymmetry breaking around $\lambda_c$.

613:  In fact our numerical results

614:  reveal, as in the $F=0,1$ case, the rearrangement of members of all supermultiplets while crossing

615:  the transition point at any finite value of  $B_{{\rm max}}$.

616:  At infinite cutoff (i.e.  in complete absence of SUSY breaking) this rearrangement is made possible by the disappearance of the energy gap at $\lambda_c$.

617: 

618: 

619: \item A new feature of multifermion states is the {\em intertwining}

620: of the $F= 2, 3$ and $F=3, 4$ supermultiplets: while all $F=2$ states have their partners

621: in the $F=3$ sector, some $F=3$ states remain unpaired: they must form new supersymmetry doublets

622: with the $F=4$ states. This behaviour  must repeat itself {\em ad infinitum} in $F$ since

623: the sizes of bases, at fixed $B$, are monotonically growing with $F$.

624: This was not the case for $F=0,1$, where all doublets were complete.

625: In other words, beginning with three ``gluinos", eigenstates with given $F$

626: may (and will) carry both signs of the $C$-parity quantum number (\ref{C})  introduced in \cite{VW1}.

627: 

628: \item This last point implies that defining the Witten index for the $F=2,3$ sectors

629: requires more care than in the $F=0,1$ case. Namely, one should sum only over

630: complete supermultiplets and not over all $F=2$ and $F=3$ states. 

631: This was done using once more

632: the supersymmetry fractions and the resulting object does indeed exhibit a discontinuity

633: at $\lambda=\lambda_c$, corresponding to the number of new vacua appearing across $\lambda_c$. 

634: 

635: \item  We did not find, in these higher-$F$ sectors,  any evidence for the  strong-weak duality

636: discovered in \cite{VW1}.

637: 

638: \item  Unlike in the $F=0,1$ sectors, we were not able to solve analytically for the spectrum.

639: Indeed,  the structure of the spectra with $F=2,3$  is much more

640: complex than the one with $F=0,1$. In particular, the eigenenergies are no longer approximately equidistant,

641: as was   the case for $F=0,1$ and  for other matrix models discussed

642: in the literature. Thus approximate equidistance  cannot be considered as a generic property of infinite $N$ dynamics.

643: 

644: \item

645: Finally, by considering the $\lambda \rightarrow \infty$ limit of our model and a strong-coupling expansion around it, we have understood the origin

646: of (and analytically constructed)  the new  bosonic zero-energy states at strong coupling.

647: The argument shows that the occurrence of such bosonic vacua should extend to arbitrary values

648: of $F$.

649: In the infinite-coupling limit the (rescaled) Hamiltonian becomes block-diagonal with finite-size

650: blocks characterized by a fixed number of fermions {\it and} bosons. This limit  has interesting

651: implications for the combinatorics of binary necklaces \cite{OVW1} and for the

652: spectrum of (non-supersymmetric) spin-chain models, as explained in a forthcoming paper.

653: \ei

654: 

655: 

656: \section*{Acknowledgements}

657: We would like to thank E. Onofri for useful discussions. This work is partially supported by the

658: grant No. P03B 024 27

659: (2004 - 2007) of the Polish Ministry of Education and Science.

660: 

661: \begin{thebibliography}{99}

662: \bibitem{VW1} G. Veneziano and J. Wosiek, JHEP  {\bf 0601}  (2006) 156  \ [hep-th/0512301].

663: \bibitem{Adriano} G. Veneziano and J. Wosiek, to appear in Adriano Di Giacomo's Festschrift (2006)\ [hep-th/0603045].

664: \bibitem{WQM} E. Witten, Nucl. Phys. {\bf B185} (1981) 513; {\bf B202} (1983) 253.

665: \bibitem{OVW1} E. Onofri, G. Veneziano and J. Wosiek, {\em Supersymmetry and Combinatorics},  \ [math-ph/0603082].

666: \bibitem{'tH} G. 't Hooft, Nucl. Phys. {\bf B72} (1974) 461; see also G. Veneziano,

667: Nucl. Phys. {\bf B117} (1976) 519.

668: \bibitem{JW} J. Wosiek, Nucl.\ Phys. {\bf B644 } (2002) 85 \ [hep-th/0203116].

669: \bibitem{CW} M. Campostrini and J. Wosiek, Nucl.\ Phys. {\bf B703 } (2004) 454 \ [hep-th/0407021].

670: \bibitem{MP} E. Marinari and G. Parisi, Phys. Lett. {\bf B240} (1990) 375.

671: \bibitem{MO} G. Marchesini and E. Onofri, Phys. Lett. {\bf B240} (1990) 375.

672: \end{thebibliography}

673: 

674: \end{document}

675: 

676: 

677: