1: \documentclass[12pt,letterpaper]{JHEP3}
2: \usepackage{amsmath}
3: \usepackage{cite}
4: \usepackage[vcentermath]{youngtab}
5: \usepackage{epic}
6:
7: %% TABLEAUX.TEX
8: %% This macro file is for producing a ``Young Tableau'' which is
9: %% an array of little squares sometimes used in mathematical physics.
10: %% For instance, the command $\tableau{6 3 2}$ will produce a tableau
11: %% with 6 squares in the top row, 3 in the next, and 2 in the last.
12: %% OOOOOO
13: %% This tableau will look like OOO but made of squares instead of Os
14: %% OO
15: %% Any number of rows may be present, each having a nonzero number of
16: %% squares.
17: %%
18: %% A tableau is math mode material, so use $ or $$ to enclose it.
19: %%
20: %% The size and line-thickness of the little boxes are controlled by
21: %% the dimension parameters --
22: %% \tableauside=1.0ex %(size)
23: %% \tableaurule=0.4pt %(line-thickness)
24: %% Change them if you want.
25: %%
26: %% -- Doug Eardley 9/19/8
27: %%
28: \newdimen\tableauside\tableauside=1.0ex
29: \newdimen\tableaurule\tableaurule=0.4pt
30: \newdimen\tableaustep
31: \def\phantomhrule#1{\hbox{\vbox to0pt{\hrule height\tableaurule width#1\vss}}}
32: \def\phantomvrule#1{\vbox{\hbox to0pt{\vrule width\tableaurule height#1\hss}}}
33: \def\sqr{\vbox{%
34: \phantomhrule\tableaustep
35: \hbox{\phantomvrule\tableaustep\kern\tableaustep\phantomvrule\tableaustep}%
36: \hbox{\vbox{\phantomhrule\tableauside}\kern-\tableaurule}}}
37: \def\squares#1{\hbox{\count0=#1\noindent\loop\sqr
38: \advance\count0 by-1 \ifnum\count0>0\repeat}}
39: \def\tableau#1{\vcenter{\offinterlineskip
40: \tableaustep=\tableauside\advance\tableaustep by-\tableaurule
41: \kern\normallineskip\hbox
42: {\kern\normallineskip\vbox
43: {\gettableau#1 0 }%
44: \kern\normallineskip\kern\tableaurule}%
45: \kern\normallineskip\kern\tableaurule}}
46: \def\gettableau#1 {\ifnum#1=0\let\next=\null\else
47: \squares{#1}\let\next=\gettableau\fi\next}
48:
49: \tableauside=1.35ex \tableaurule=0.4pt
50:
51:
52:
53: \title{Holography and entropy bounds\\ in the plane wave matrix model}
54:
55: \author{%
56: Raphael Bousso and Aleksey L. Mints \\
57: Center for Theoretical Physics, Department of Physics\\
58: University of California, Berkeley, CA 94720-7300, U.S.A.\\
59: {\em and}\\
60: Lawrence Berkeley National Laboratory, Berkeley, CA 94720-8162, U.S.A.\\
61: E-mail: \email{bousso@lbl.gov, mints@socrates.berkeley.edu}}
62:
63: \abstract{%
64: As a quantum theory of gravity, Matrix theory should provide a
65: realization of the holographic principle, in the sense that a
66: holographic theory should contain one binary degree of freedom per
67: Planck area. We present evidence that Bekenstein's entropy
68: bound, which is related to area differences, is manifest in the
69: plane wave matrix model. If holography is implemented in this way,
70: we predict crossover behavior at strong coupling when the energy
71: exceeds $N^2$ in units of the mass scale.}
72:
73: \preprint{\hepth{0512201} \\ UCB-PTH-05/32 \\ LBNL-58947}
74:
75: %\keywords{keywords go here}
76:
77: \begin{document}
78:
79: \section{Introduction}
80:
81: The holographic principle~\cite{Tho93,Sus95,CEB2} requires that the
82: surprising entropy bounds apparent in nature~\cite{CEB1} should be
83: manifest in quantum gravity. For asymptotically Anti-de~Sitter
84: spacetimes this was realized by the Maldacena
85: conjecture~\cite{Mal97} in which $A$ binary degrees of freedom of an
86: $SU(N)$ Yang-Mills theory suffice to describe a region of surface area
87: $A$~\cite{SusWit98}, in Planck units. This holographic behavior
88: follows from a UV/IR relation that places a UV cutoff on the field
89: theory as the size of bulk regions is decreased~\cite{PeePol98}.
90:
91: When the bulk region becomes as small as the AdS curvature scale, the
92: field theory reaches its lowest possible UV cutoff. Only the
93: zero-modes on the $S^3$ are retained, and the theory reduces to matrix
94: quantum mechanics. At this point holography is still manifest
95: with $N^2$ degrees of freedom describing a region of area $N^2$. But
96: for smaller regions it is not clear how to eliminate any further
97: degrees of freedom from the field theory. For similar reasons
98: holographic state counting is quite obscure in the matrix models
99: proposed to describe M-theory on asymptotically flat space~\cite{BFSS}
100: or on the plane wave in eleven dimensions~\cite{BMN}.
101:
102: In this paper, we propose that the entropy bound that becomes manifest
103: in matrix quantum mechanics is not the covariant bound relating
104: entropy to surface area. Rather, it is the Bekenstein
105: bound~\cite{Bek81}, which states that the entropy of a system will not
106: exceed its mass times its linear size. We find evidence that this
107: bound is saturated by the matrix degrees of freedom available for the
108: description of weakly curved geometries.
109:
110: The Bekenstein bound appears to be connected to the holographic
111: principle in that it arises from the generalized covariant entropy
112: bound~\cite{FMW} in a certain weak-gravity limit~\cite{Bou03}. In its
113: regime of validity it is much tighter than the covariant bound. Thus,
114: it may govern the emergence of local weakly gravitating regions from a
115: more fundamental theory~\cite{Bou04}, similar to the more general role
116: envisaged for the covariant bound. However, it has suffered from
117: ambiguities in its definition of entropy (recent discussions of this
118: problem include
119: Refs.~\cite{Pag00c,Bek00b,Bou03a,MarRoi04,MarMin03,Pag04,Bek04}).
120:
121: Attempts to rectify this situation led to the proposal~\cite{Bou03b}
122: to formulate Bekenstein's bound in terms of the discrete light-cone
123: quantization (DLCQ) of theories on backgrounds with a null Killing
124: vector: The entropy in a sector with $N$ units of momentum should not
125: exceed $2\pi^2N$:\footnote{In Ref.~\cite{Bou03} the Bekenstein bound
126: is derived as the integrated loss of cross-sectional area of
127: parallel light-rays focused by a matter system (see also
128: Ref.~\cite{BouFla03,StrTho03}). It arises not in its original form
129: ($S\leq 2\pi M R$, where $R$ is the largest linear dimension), but
130: in the stronger form $S\leq \pi M \Delta x$, where $\Delta x$ is the
131: width of the system along an arbitrary direction in its rest frame.
132: It is more natural to work in the lightcone frame picked out by the
133: light-sheet. One then has $S\leq\pi P_- \Delta x^-$, where the null
134: coordinate $x^-$ and the longitudinal momentum $P_-$ are defined as
135: usual~\cite{Bou03b}. Ambiguities in defining the spatial extent of
136: quantum states can be suppressed by compactifying the null
137: direction. Then the momentum is quantized ($P_- = 2\pi N/\Delta
138: x^-$), and the bound becomes $S\leq 2\pi^2 N$. Here $S$ is the
139: logarithm of the number of discrete states~\cite{Bou03a} in the
140: sector with $N$ units of momentum.}
141: \begin{equation}
142: S\leq 2\pi^2 N~.
143: \label{eq-bek}
144: \end{equation}
145:
146: It is natural to test this proposal in the context of M-theory.
147: M-theory is {\em defined\/} in terms of its DLCQ as a $U(N)$
148: matrix model~\cite{Sus97}. Moreover, we expect that it represents a
149: consistent quantum theory of gravity. This means that violations, if
150: found, cannot be ascribed to an artificial choice of Lagrangian.
151:
152: In particular, we shall study M-theory on the eleven-dimensional plane
153: wave background. Unlike Matrix theory in flat space~\cite{BFSS}, the
154: spectrum of the plane wave matrix model is discrete. It is known
155: exactly~\cite{DSV1} at large boosts $\mu$, where the quantum mechanics
156: is weakly coupled. However, the curvature of the gravity dual is
157: strong in this regime, and the Bekenstein bound is not expected to
158: apply.
159:
160: At small $\mu$, when the plane wave is weakly curved, the matrix model
161: is strongly coupled. However, there are still infinite towers of
162: protected states. Without a cutoff on the lightcone energy the
163: protected states alone would seem to contribute an infinite amount of
164: entropy at any value of $N$, apparently violating the Bekenstein
165: bound.
166:
167: However, both the full spectrum at weak coupling and the spectrum of
168: protected states at arbitrary coupling undergo crossover behavior
169: when the energy in units of $\mu$ becomes of order $N^2$, i.e., for
170: \begin{equation}
171: E_{\rm cross}\sim N^2~.
172: \end{equation}
173: The entropy of the full spectrum (at weak coupling) behaves as
174: \begin{equation}
175: S\sim E~~(E\lesssim N^2)~;~~~ \qquad \qquad
176: S\sim N^2\log\frac{E}{N^2}~~(E\gtrsim N^2)~.
177: \label{eq-prefull}
178: \end{equation}
179: The entropy of the protected states behaves as
180: \begin{equation}
181: S_{\rm p}\sim (EN)^{1/3}~~(E\lesssim N^2)~;~~~~~
182: S_{\rm p}\sim N\log\frac{E}{N^2}~~(E\gtrsim N^2)~.
183: \label{eq-preprot}
184: \end{equation}
185:
186: At strong coupling, where the Bekenstein bound should apply, only the
187: protected states are guaranteed to be present. We shall assume that
188: they dominate the entropy in the microcanonical ensemble, at least in
189: the vicinity of the crossover energy. Then the above results show
190: that the system undergoes a kind of phase transition (smoothed out by
191: finite $N$) at $E\sim N^2$, both at strong and at weak coupling. On
192: the gravity side, one expects this type of behavior to be associated
193: with a nonperturbative modification of the background~\cite{AhaMar03}
194: analogous to the Hawking-Page transition in Anti-de~Sitter
195: space~\cite{HawPag83,Wit98a}.
196:
197: It would be interesting to develop a concrete proposal for the gravity
198: interpretation of this transition in the 11D plane wave. No black
199: hole solutions asymptotic to the 11D plane wave have been found, and
200: it is possible that the eleven-dimensional spacetime interpretation
201: breaks down entirely at the crossover energy.
202:
203: States associated with a strong perturbation of the background do not
204: contribute to the Bekenstein bound. Thus, under the stated
205: assumptions, the entropy entering the bound is at most the crossover
206: entropy,
207: \begin{equation}
208: S_{\rm p, cross} \sim N~.
209: \end{equation}
210: Comparison with Eq.~(\ref{eq-bek}) finds the Bekenstein bound
211: saturated, but not exceeded.
212:
213: This paper is organized as follows. In Sec.~\ref{sec-weak} we
214: consider the plane wave matrix model at weak coupling. Adapting
215: general arguments for a Hagedorn/deconfinement phase transition in
216: weakly coupled Yang-Mills
217: theory~\cite{Sun00,AhaMar03,AhaMar04,AhaMar05a,AhaMar05b,Sus97,
218: LiMar98,HadRam04,Sem04,FurSch03}, we find crossover behavior at $E\sim N^2$,
219: Eq.~(\ref{eq-prefull}). At this point the energy becomes large enough
220: to excite all matrix degrees of freedom; the finite,
221: quantum-mechanical nature of the finite $N$ system is revealed, and is
222: reflected in a much slower growth of entropy.
223:
224: Because this result applies in a regime of strong curvature, it is not
225: directly relevant to the Bekenstein bound. However, it lends
226: additional support to the evidence we find in Sec.~\ref{sec-encp} for
227: a similar transition at strong coupling. We consider the spectrum of
228: protected multiplets. It receives no corrections at strong coupling
229: when the model is conjectured to describe a weakly curved background.
230: We find that it also undergoes a transition at $E\sim N^2$, as shown
231: in Eq.~(\ref{eq-preprot}).
232:
233: Our results are suggestive but not conclusive. In
234: Sec.~\ref{sec-discuss} we clarify the additional assumptions required
235: to interpret the crossover entropy $S_{\rm p, cross}\sim N$ of
236: protected states as a manifestation of the Bekenstein bound. In
237: particular, one must assume that the crossover seen for protected
238: states is representative of the full theory at strong coupling.
239:
240: In Appendix~\ref{sec-n1} we discuss a subtlety arising for the center
241: of mass degrees of freedom. Appendix~\ref{sec-protected} summarizes
242: properties of protected states.
243:
244: \section{Entropy and crossover in the free theory}
245: \label{sec-weak}
246:
247: In this section we introduce the plane wave matrix model and discuss
248: the crossover behavior of the entropy at energy $N^2$ at weak
249: coupling when the full spectrum is known exactly.
250:
251: \subsection{The plane wave matrix model}
252: \label{sec-pwmm}
253:
254: The $U(N)$ plane wave matrix model is given by the Hamiltonian
255: \begin{eqnarray}
256: H &=& R\ {\rm Tr} \left( {1 \over 2} \Pi_A^2 - {1 \over 4} [X_A, X_B]^2
257: - {1 \over 2} \Psi^\top \gamma^A [X_A, \Psi] \right) \cr
258: &+& {R \over 2} {\rm Tr} \left( \left({\mu\over
259: 3R}\right)^2 X_i^2 + \left({\mu \over 6R}\right)^2 X_a^2
260: \right. \cr
261: && \qquad \qquad \left. +
262: i {\mu \over 4R} \Psi^\top \gamma^{123} \Psi + i {2\mu \over 3R}
263: \epsilon^{ijk} X_i X_j X_k \right)~.
264: \label{eq-model}
265: \end{eqnarray}
266: Indices $A\ldots$ run from 1 to 9; $i\ldots$ run from 1 to 3; and
267: $a\ldots$ run from 4 to 9. This model was proposed~\cite{BMN} to
268: describe M-theory on the maximally supersymmetric plane wave
269: background of eleven-dimensional supergravity,
270: \begin{eqnarray}
271: ds^2 &=& - 2 dx^+ dx^- + (dx^A)^2 - \left( {\mu^2
272: \over 9} x^i x_i + {\mu^2 \over 36} x^a x_a\right) (dx^+)^2~, \\
273: F_{123+} &=& \mu~.
274: \label{eq-metric}
275: \end{eqnarray}
276:
277: The parameter $\mu$ in the metric is a coordinate artifact; it can be
278: set to any value by rescaling $x^\pm$. The M-theory limit of the
279: matrix model (\ref{eq-model}) is obtained by taking $N\to\infty$ while
280: holding $N/R$ fixed. In this limit the model must also become
281: independent of $\mu$.
282:
283: At finite $N$ the matrix model is expected to describe the DLCQ of
284: M-theory, in the sector with $N$ units of longitudinal
285: momentum~\cite{Sus97}. In this case, the coordinate $x^-$ is
286: periodically identified with period $2\pi R$. For finite $N$ the
287: boost-invariant quantity $\mu/R$ is a physical parameter that
288: distinguishes qualitatively different regimes.
289:
290: To see this, it is useful to think of $\mu$ as the curvature radius of
291: the transverse dimensions as measured in the frame in which the
292: periodically identified hypersurfaces have spatial distance $R$. To
293: obtain a good geometric description we require both $\mu\ll 1$ and
294: $R\gg 1$; hence, $\mu/R\ll 1$. For large $\mu/R$ the matrix model
295: does not correspond to a classical background in any frame; in
296: particular, we do not expect the Bekenstein bound to apply.
297:
298: In the limit $\mu\to\infty$ at fixed $N$ and $R$ the plane wave
299: matrix model (\ref{eq-model}) becomes free. For each partition of $N$
300: there is a superselection sector with its own $1/2$ BPS vacuum,
301: corresponding to a collection of concentric fuzzy spheres~\cite{BMN}.
302: Let us focus on the $X=0$ sector given by the trivial partition
303: $N=1+1+\ldots+1$. As we shall see, it exhibits the most rapid growth
304: of entropy.
305:
306: \subsection{$X=0$ sector}
307:
308: The $X=0$ sector has Hamiltonian
309: \begin{equation}
310: H = \frac{\mu}{3} \sum_i A^{\dagger}_i A_i + \frac{\mu}{6}
311: \sum_a A^{\dagger}_a A_a + \frac{\mu}{4} \sum_{I\alpha}
312: \psi_{I\alpha}^{\dagger} \psi_{I\alpha}
313: \label{eq-ham}
314: \end{equation}
315: and contains rank $N$ matrix creation operators
316: \begin{equation}
317: A_i^\dagger=\sqrt{\frac{\mu}{6R}} X^i -
318: i\sqrt{\frac{3R}{2\mu}}\Pi^i~,~~~
319: A_a^\dagger= \sqrt{\frac{\mu}{12R}} X^a -
320: i\sqrt{\frac{3R}{\mu}}\Pi^a~,
321: \label{eq-ops}
322: \end{equation}
323: as well as fermionic operators $\psi_{I\alpha}$~\cite{DSV1}.
324:
325: Each creation operator contributes of order $\mu$ to the lightcone
326: energy, so the dimensionless energy $E\equiv -P_+/\mu$ measures the
327: number of quanta. Physical states must be gauge-invariant. Hence,
328: they correspond to products of traces of products of creation
329: operators.
330:
331: It is important that there is more than one kind of matrix operator.
332: There is one operator for each transverse direction that the system
333: can move in, so their number, $q$, is of order ten. In general
334: matrices do not commute, so a state with $E$ quanta can be made in
335: $q^E$ different ways\footnote{This is a slight simplification, even
336: leaving aside the issue of trace relations discussed below. At
337: least one trace must be taken, and of course multiple traces are
338: allowed as well. Thus, we have undercounted states since the
339: sequence of $E$ operators can be sprinkled with traces. However,
340: this cannot multiply the number of states by more than the number of
341: partitions of $E$, $p(E)\approx \exp(\sqrt{\frac{2}{3}}\pi
342: \sqrt{E})$. Hence it will correct the entropy (\ref{eq-striv}) at
343: most by a subleading term of order $\sqrt{E}$. We have also
344: overcounted states because states related by cyclic exchange of the
345: creation operators within a trace are not independent. This will
346: reduce the entropy at most by a term of order $\log E$, which again
347: is negligible compared to Eq.~(\ref{eq-striv}).} leading to an
348: entropy $S\sim E\log q$. Since $q$ is not very large, one
349: has~\cite{Sun00,AhaMar03}
350: \begin{equation}
351: S\sim E~.
352: \label{eq-striv}
353: \end{equation}
354:
355: This Hagedorn behavior cannot persist indefinitely. At sufficiently
356: high energy the entropy is dictated by the thermodynamics of a
357: quantum mechanical system with about $N^2$ degrees of freedom:
358: \begin{equation}
359: S\sim N^2\log \frac{E}{N^2}~.
360: \label{eq-shigh}
361: \end{equation}
362: At a crossover energy of order $N^2$ the two expressions for the
363: entropy match.
364:
365: The argument for Hagedorn growth, Eq.~(\ref{eq-striv}), breaks down
366: because trace relations can lead to identifications between
367: states~\cite{AhaMar03}. Apparently\footnote{There is tentative direct
368: evidence for this. One can show that all single-trace states of
369: length up to $N$ are independent, even for $q>1$~\cite{BouMin05a}.
370: For $q=1$ all traces longer than $N$ decompose into products of
371: traces. Hence, for $q>1$ trace relations also set in at lengths
372: exceeding $N$, at least for some traces. In a typical partition of
373: $E$ most traces have length of order $\sqrt{E}$, which will exceed
374: $N$ once $E$ becomes larger than $N^2$. Hence, this is the point at
375: which we expect a typical state in the Hagedorn spectrum to become
376: identified with other states by trace relations. This argument
377: assumes that trace relations are not dominated by relations
378: involving products of more than one trace on both sides of the
379: equality.} this effect becomes important only at $E\sim N^2$.
380:
381: \subsection{Full spectrum}
382:
383: Other sectors exhibit a less rapid growth of states. Consider the
384: ``irreducible'' vacuum, which corresponds to the partition of $N$ into
385: just one term. It describes a single fuzzy membrane of momentum
386: $N/R$. At small energies the entropy scales as $S_{\rm M2}\sim
387: E^{2/3}$, as it should for a 2+1 dimensional object. Thus, it grows
388: more slowly than (\ref{eq-striv}) below. Matching to the asymptotic
389: density of states, Eq.~(\ref{eq-shigh}), reveals that the crossover
390: occurs at energies of order $N^3$.
391:
392: The differing crossover energies are easily understood as follows.
393: Crossover happens when we notice that the matrix is finite, i.e., for
394: energies large enough to excite all matrix degrees of freedom. In the
395: trivial vacuum there are approximately $qN^2$ matrix elements. Each
396: corresponds to a creation operator that increases $E$ by about 1.
397: Thus, with energies of order $N^2$ they can all become excited. In
398: the irreducible vacuum there are again $N^2$ oscillators, but with
399: different energies: There are, roughly, $i$ operators with mass
400: $i\mu$, $1\leq i\leq N$~\cite{DSV1}. Hence, the crossover energy is
401: of order $N^3$, the energy required to excite all oscillators.
402:
403: As soon as we turn on any non-zero coupling ($\mu$ large but finite),
404: the different partitions of $N$ cease to define superselection
405: sectors. All sectors mix in the microcanonical ensemble. The entropy
406: will be dominated by the sector with the most rapid growth, the $X=0$
407: vacuum. Hence, the transition to the thermodynamic behavior
408: (\ref{eq-shigh}) will set in at energies of order $N^2$.
409:
410: The crossover entropy will also be of order $N^2$. It is worth
411: stressing again that the weakly coupled matrix model does not admit an
412: interpretation as a weakly curved geometry. Hence, we are not in a
413: regime where the Bekenstein bound can be tested; we have merely noted
414: that the model undergoes crossover behavior at energy $N^2$.
415:
416:
417: \section{Entropy and crossover of protected states}
418: \label{sec-encp}
419:
420: In this section we discuss the spectrum of protected states, which
421: will be exact at all values of the coupling. We show that it also
422: undergoes a transition at energy $N^2$ when the entropy is about $N$.
423:
424: \subsection{Strong coupling and protected states}
425:
426: In any given sector the coupling becomes strong for
427: \begin{equation}
428: \left(\frac{R}{\mu N_{\rm max}}\right)^3 N_{\rm max} > 1~,
429: \end{equation}
430: where $N_{\rm max}$ is the largest term in the partition of
431: $N$~\cite{DSV1}. As $\mu$ is decreased this will happen first for
432: the $X=0$ vacuum, at $\mu/R\sim 1$.
433:
434: The information we have about the spectrum at strong coupling comes
435: from quantities which are protected for arbitrary positive values of
436: $\mu$~\cite{DSV2}. In particular, multiplets in the weight $(0,n,0)$
437: representation of the $SU(4|2)$ symmetry group are exactly protected.
438: The lightcone energy of states in these multiplets is of order $n\mu$:
439: \begin{equation}
440: E\sim n~;
441: \end{equation}
442: it does not receive corrections.
443:
444: Denoted by Young supertableaux, the protected representations are
445: rectangular with $n$ columns and 2 rows, e.g.:
446: \begin{equation}
447: {\tiny \yngSLASH(11,11)}~.
448: \end{equation}
449: The degeneracy of each multiplet is of order $n^4$. The entire
450: multiplet can be obtained by acting with supersymmetry generators on
451: its primary states, which are associated with the $SO(6)$ transverse
452: directions $X^a$ (see Appendix~\ref{sec-protected}).
453:
454: \subsection{$X=0$ sector}
455:
456: We will first discuss the entropy of protected states in individual
457: superselection sectors of the free theory ($\mu\to\infty$). As
458: explained in Appendix~\ref{sec-n1}, we suppress the center of mass
459: degrees of freedom. Then the single-membrane sector contains neither
460: operators nor states of weight $(0,n,0)$.
461:
462: Let us focus instead on the $X=0$ sector, given in Eqs.~(\ref{eq-ham})
463: and (\ref{eq-ops}). The primaries for $(0,n,0)$ representations are
464: of the form
465: \begin{equation}
466: S^{a_1...a_n}~ {\rm T\!R}_N \big(A^\dagger_{a_1} \cdots A^\dagger_{a_n}\big)
467: |0\rangle ~.
468: \end{equation}
469: where $S^{a_1...a_n}$ is a totally symmetric traceless tensor of
470: $SO(6)$, and we use the notation ${\rm T\!R}_N$ to indicate that an
471: arbitrary number of traces may be sprinkled among the $n$ operators,
472: subject to the constraint that each trace contain at most $N$
473: operators. (Because of the complete symmetrization, traces longer
474: than $N$ are guaranteed not to be independent. In order to exclude
475: the $U(1)$ one should also require that each trace include at least
476: two operators, but this will give a negligible correction to the
477: entropy.)
478:
479: Ordering the traces by their length, we may represent each primary by
480: a Young diagram with $n$ boxes and at most $N$ rows, where each column
481: represents one trace. These auxiliary diagrams do not indicate a
482: group representation; they simply keep track of all the different ways
483: one can {\em produce\/} the $(0,n,0)$ representation, and thus of its
484: degeneracy in the spectrum.
485:
486: Hence, the number of protected representations of energy $E$ is given
487: by the number of restricted partitions of $E$,
488: \begin{equation}
489: p_N(E) \approx \frac{1}{4\sqrt{3}E}
490: \exp\left[ \pi\sqrt{\frac{2E}{3}} -
491: \frac{\sqrt{6E}}{\pi}\exp\left(\frac{-\pi N}{\sqrt{6E}}\right)
492: \right]~.
493: \label{eq-pn}
494: \end{equation}
495: This formula is valid for $1\ll E\ll N^2$. The entropy of protected
496: states will thus grow as
497: \begin{equation}
498: S_{{\rm p}~(X=0)} \sim 2\pi \sqrt{\frac{E}{6}}
499: \end{equation}
500: until a crossover threshold around $E\sim N^2$, when
501: \begin{equation}
502: S_{{\rm p, cross}~(X=0)} \sim N~.
503: \label{eq-spc}
504: \end{equation}
505: The contribution $4\log E$ from the degeneracy of states within each
506: protected multiplet is negligible.
507:
508: One can understand this result from a number of perspectives.
509: Naively, one might expect crossover behavior for $p_N(E)$ around
510: $E=N$ when the restriction to at most $N$ rows first becomes
511: nontrivial. However, what matters for the entropy is how the
512: restriction acts on {\em typical\/} partitions. For large $E$ the
513: vast majority of unrestricted partitions contain about $\sqrt{E}$ rows
514: ranging in length up to $\sqrt{E}$; that is, the Young diagram
515: associated to a typical partition looks very roughly like a triangle
516: of height and width $\sqrt{E}$.\footnote{More precisely, by computing
517: expected occupation numbers of $N\to\infty$ oscillators in
518: Eq.~(\ref{eq-zpn}) below one obtains the limiting curve
519: $e^{x/T}+e^{y/T}=1$, where $x$ and $y$ are coordinates along the
520: edges of the Young diagram, and $T=\sqrt{6E}/\pi$. This means that
521: the expected height and width of the diagram is $T\log T$. Hence it
522: will begin to exceed $N$ slightly earlier than for $T\sim N$. This
523: slows down the growth of the degeneracy but only enough to modify
524: order one prefactors in the entropy. One still has $S\sim\sqrt{E}$
525: until $T\sim N$~\cite{BalDeb05}.} This means that the restriction to
526: at most $N$ rows will not become important until $E\sim N^2$, when a
527: typical partition would prefer to have more than $N$ rows. Therefore,
528: we expect crossover behavior in the number of states for $E\sim N^2$,
529: not $E\sim N$.
530:
531: Let us gain more insight into the behavior of the entropy beyond the
532: crossover. The partition function for protected representations is
533: identical to that for $N$ bosons in a harmonic oscillator or for a
534: system of $N$ harmonic oscillators of frequency $1,2,\ldots, N$, with
535: zero point energies removed:
536: \begin{equation}
537: Z(T) = \prod_{j=1}^N \frac{1}{1-e^{-j/T}}~.
538: \label{eq-zpn}
539: \end{equation}
540: Computing the entropy and energy from $Z(T)$ for $1\ll T\ll N$ yields
541: Eq.~(\ref{eq-pn})~\cite{TraMur03}.
542:
543: For $T\to\infty$ this partition function yields $S_{{\rm
544: p}\,(X=0)}=2N+N\log (T/N)$ and $E\sim NT$, consistent with the
545: thermodynamics of a quantum mechanical system with $N$ degrees of
546: freedom. Hence
547: \begin{equation}
548: S_{{\rm p}\,(X=0)}= 2N + N\log \frac{E}{N^2}~.
549: \label{eq-hent}
550: \end{equation}
551: The onset of this asymptotic behavior is for $T=N$, when the
552: temperature is large enough to excite the frequency $N$ oscillator.
553: This corresponds to a lower bound $E\sim N^2$ for the asymptotic
554: regime, consistent with the upper limit obtained for the $T<N$ regime.
555: The crossover entropy, $S_{\rm p, cross}\sim N$, is also consistent
556: with Eq.~(\ref{eq-spc}).
557:
558:
559: \subsection{Full protected spectrum}
560:
561: At any finite coupling, and in particular at strong coupling, all
562: sectors of the free theory must be included in the microcanonical
563: ensemble. So far we have considered only the protected states of the
564: $X=0$ vacuum ($N=1+\ldots+1$). The other vacua, corresponding to
565: nontrivial partitions of $N$, will contribute additional protected
566: states.
567:
568: Consider a general partition of $N$, $(N_1^{M_1}\cdots N_l^{M_l})$.
569: In this notation $M_i$ denotes how many times $N_i$ appears in the
570: sum, i.e., $N=\sum_{i=1}^l M_i N_i$. Each term $N_i^{M_i}$
571: corresponds to a stack of $M_i$ coincident fuzzy spheres of individual
572: momenta $N_i/R$. Each stack contributes a rank $M_i$ protected matrix
573: creation operator. Thus, for the purposes of protected states, the
574: momentum of the stack is irrelevant; only its ``height'' $M_i$
575: matters.
576:
577: The counting problem is not only how $E$ can be partitioned into
578: traces, but at the same time how $N$ can be partitioned into stacks,
579: whose height controls the maximum length of traces and whose number
580: controls the number of different types of traces that can be
581: constructed.
582:
583: To compute the total number of protected states it is convenient to
584: introduce a chemical potential $\nu$ conjugate to $N$. Then the
585: problem is an extension of the canonical partition function generating
586: the partitions of $E$. A stack $N_i^{M_i}$ forms in response to the
587: chemical potential $\nu$, in analogy to an oscillator of frequency $j$
588: being excited to occupation number $m_j$ in response to a
589: temperature $T=1/\beta$.
590:
591: The grand canonical partition function for a single stack of fuzzy
592: spheres is thus
593: \begin{equation}
594: Z_k(\beta,\nu) = \sum_{N,E=0}^\infty e^{-\nu k N}\, e^{-\beta E}\,
595: p_N(E)~,
596: \end{equation}
597: where $k$ is the number of units of longitudinal momentum of each
598: sphere in the stack. Using Eq.~(\ref{eq-zpn}) this
599: becomes~\cite{LinMal05}
600: \begin{equation}
601: Z_k(\beta,\nu) = \prod_{n=0}^\infty\frac{1}{1-e^{-\nu k}\,
602: e^{-\beta n}}~.
603: \end{equation}
604:
605: A general vacuum contains stacks of fuzzy spheres of arbitrary
606: momenta, described by the product of these partition functions
607: (analogous to the product of partition functions of harmonic
608: oscillators of different frequencies):
609: \begin{equation}
610: Z(\beta,\nu) = \prod_{k=1}^\infty Z_k(\beta,\nu)~.
611: \label{eq-zbm}
612: \end{equation}
613: For small $\beta$ and $\nu$ this evaluates to
614: \begin{equation}
615: \log Z(\beta,\nu) \approx \frac{c}{\beta\nu}~,
616: \end{equation}
617: with $c\approx 1.20206$.
618:
619: The entropy is obtained from the integral transform
620: \begin{equation}
621: e^{S_{\rm p}(N,E)} = \int d\nu\, d\beta\, e^{\nu N+\beta E}\, Z(\beta,\nu)~,
622: \end{equation}
623: where in the saddlepoint approximation
624: \begin{equation}
625: N=\frac{c}{\beta\nu^2}~,~~~ E=\frac{c}{\beta^2 \nu}~.
626: \label{eq-sadp}
627: \end{equation}
628: This yields a density of states $\frac{1}{2} (\frac{c}{N^2E^2})^{1/3}
629: \exp[3(cNE)^{1/3}]$, so that to leading order~\cite{LinMal05}
630: \begin{equation}
631: S_{\rm p}(N,E) = 3(cNE)^{1/3}~.
632: \label{eq-fent}
633: \end{equation}
634:
635: Using Eq.~(\ref{eq-sadp}), the assumption of small $\beta$ and $\nu$
636: translates into the regime of validity
637: \begin{equation}
638: \sqrt{N}\ll E\ll N^2~.
639: \end{equation}
640: A lower crossover is at $E^2\sim N$, when the entropy becomes
641: completely dominated by the number of vacua (the partitions of $N$).
642: We are interested in the higher crossover at $E\sim N^2$. At this
643: point the entropy becomes dominated by the states in the $X=0$ sector.
644: From Eq.~(\ref{eq-fent}) we find again that the entropy at this
645: crossover energy is of order $N$:
646: \begin{equation}
647: S_{\rm p, cross} \sim N~.
648: \label{eq-allcr}
649: \end{equation}
650:
651: To approach the crossover from the other side ($E\gg N^2$), we
652: evaluate the partition function (\ref{eq-zbm}) for $\nu\gg 1$:
653: \begin{equation}
654: \log Z(\beta,\nu) \approx \frac{1}{\beta e^\nu}~,
655: \end{equation}
656: with saddlepoint
657: \begin{equation}
658: E=\frac{1}{\beta^2 e^\nu}~,~~~N = \frac{1}{\beta e^\nu}~.
659: \end{equation}
660: Hence the asymptotic behavior of the entropy at very high energies is
661: \begin{equation}
662: S_{\rm p} = 2N+N\log\frac{E}{N^2}~.
663: \end{equation}
664: For $E\sim N^2$, this matches up with Eq.~(\ref{eq-fent}) at the
665: crossover entropy (\ref{eq-allcr}).
666:
667:
668: \section{Assumptions and implications}
669: \label{sec-discuss}
670:
671: Our analysis shows that the Bekenstein bound, in the DLCQ form
672: $S\lesssim N$, is satisfied and approximately saturated in the plane
673: wave matrix model under the following assumptions:
674: \begin{enumerate}
675: \item{The protected states correctly estimate the entropy going into
676: the Bekenstein bound, at least for energies approaching the
677: crossover scale $E=N^2$ from below.}
678: \item{For energies above the crossover the matrix model does not
679: describe a weakly gravitating system.}
680: \end{enumerate}
681: In this section we discuss why these assumptions are needed, point out
682: that the first lends plausibility to the second, and speculate about
683: their implications for the matrix model.
684:
685: Applicability of the Bekenstein bound requires a good semiclassical
686: background. This is why we have studied the spectrum for small values
687: of $\mu$, where curvatures are small. Because the matrix model is
688: strongly coupled there, we were only able to discuss protected states.
689:
690: But a weak background is not enough. The states included in the
691: Bekenstein bound must themselves be weakly gravitating, in the sense
692: that they are incapable of focussing light significantly over a
693: distance set by their own spatial size~\cite{Bou03}.
694:
695: Some protected states, even in a weakly curved background, may well
696: have large backreaction. (In the context of AdS, an example are
697: single gravitons with energy above the Planck energy.) Thus, not all
698: protected states necessarily contribute to the entropy in the
699: Bekenstein bound. On the other hand, there may be some unprotected
700: states (in the AdS analogy, say, multiple weak gravitons) which have
701: small backreaction and do contribute.
702:
703: Such effects will not be important if they modify the entropy only by
704: factors of order one, or only far from the crossover energy. But if
705: they are significant near $E\sim N^2$, they can affect one or both of
706: our conclusions (that the Bekenstein bound is satisfied, and that it
707: is saturated). This is why the first assumption is needed. Without
708: the second assumption, the Bekenstein bound would be violated by
709: states with $E\gg N^2$.
710:
711: We are hopeful that the validity of both assumptions can be clarified
712: thanks to recent progress in the physical understanding of the plane
713: wave states and their geometry at strong coupling. Excitations of M5
714: branes appear in the spectrum of protected states~\cite{M5}, and their
715: multiparticle states are likely to play a role in this analysis.
716: Since the geometry near a single M5 brane is not smooth, it will be
717: interesting to consider solutions with coincident M5 brane
718: configurations~\cite{LinMal05} in this context. We leave this to
719: future work.
720:
721: If the first assumption is correct, it will have an interesting
722: implication that makes the second assumption more plausible: The
723: crossover behavior at $E\sim N^2$ (in units of $\mu$), which is quite
724: well understood at weak coupling, will persist at strong coupling.
725: This may seem surprising: When $\mu\ll M_{\rm Pl}$, why should the
726: characteristic scale continue to be $\mu$, rather than, say, the
727: Planck scale? Consider a scattering problem with impact parameter
728: much shorter than the transverse curvature radius, $(R/\mu)^{1/2}$
729: (which is large in Planck units in the strong coupling regime). Such
730: processes should be described as in the flat space matrix model.
731: Therefore, $\mu$ should be dynamically irrelevant, and all interesting
732: scales should arise from the Planck mass and powers of $N$.
733:
734: However, this logic applies only to short timescales. In the flat
735: space case ($\mu=0$ exactly) two gravitons that scatter can move off
736: to infinity uninhibited by the quartic interaction since off-diagonal
737: excitations are frozen out at large separation. But gravitons that
738: scatter in the plane wave will eventually come to notice that they
739: live in a confining background and are really in a bound state. This
740: takes a time of order $1/\mu$, which diverges in the $\mu\to 0$ limit.
741: None of these bound states survive at $\mu=0$, so we do not expect the
742: BFSS limit to be smooth in this sense. This argument suggests that
743: $\mu$ remains an important scale in the microcanonical ensemble at all
744: nonzero values of $\mu$ no matter how small.
745:
746: Phase transitions in Yang-Mills theory have been analyzed in detail in
747: Ref.~\cite{AhaMar03}. In terms of their classification, our analysis
748: suggests that the plane wave matrix model will {\em not\/} behave
749: like, say, ordinary $d=4$ Yang-Mills theory, whose transition
750: temperature can be set by two different scales. ($T_{\rm crit}\sim
751: 1/R$ for compactification on a small three-manifold of size $R^3$, but
752: $T\sim\Lambda_{\rm QCD}$ for $R\gg 1/\Lambda_{\rm QCD}$, when coupling
753: is strong.) Rather the matrix model should behave like $d=4$, ${\cal
754: N}=4$ super-Yang-Mills theory on a three-sphere, whose phase
755: structure is expected to be the same at strong and weak coupling. (The
756: critical temperature is of order $1/R$ and the crossover energy is
757: $N^2/R$, where $R$ is the size of the $S^3$.)
758:
759: It is interesting to ask about the gravity dual to the matrix model
760: above the crossover energy. The deconfinement phase transition of
761: ${\cal N}=4$, $d=4$ super-Yang-Mills theory is related via AdS/CFT to
762: the Hawking-Page transition in AdS~\cite{Wit98a,Wit98b}. More
763: generally, it has been suggested that the high energy phase of a
764: large-$N$ Yang-Mills theory always corresponds to the presence of a
765: black object in the dual closed string background~\cite{AhaMar03}.
766: This amounts to a nonperturbative modification of the background.
767:
768: Here we are dealing with an M-theory background, but it is tempting to
769: speculate that it becomes similarly modified. Conceivably, the bulk
770: dual of the crossover could be more drastic: the gravity
771: interpretation may break down entirely. Unlike for AdS, no black hole
772: or black string solutions asymptotic to the eleven-dimensional plane
773: wave are known. Moreover, it is doubtful that the canonical ensemble
774: can be defined any more sensibly than it can for flat
775: space.\footnote{We thank M.~Van Raamsdonk for stressing this point to
776: us.} (To put flat space at a finite temperature requires nonzero
777: constant energy density and thus infinite energy; in a theory with
778: gravity this invalidates the background.)
779:
780: In AdS/CFT gravity turns off for $N\to\infty$, where the crossover
781: becomes a sharp phase transition. In this limit there is no longer a
782: good gravity description above the critical temperature. Unlike AdS,
783: the M-theory limit on the plane wave {\em requires\/} $N\to\infty$.
784: This is another reason to suspect that there may be no sensible
785: eleven-dimensional gravity dual beyond the crossover energy.
786:
787: For the purposes of obtaining a cutoff on the entropy entering the
788: Bekenstein bound, it only matters that such states can no longer be
789: described as small perturbations of the plane wave background. Any of
790: the scenarios we have described---black hole formation, or a complete
791: breakdown of the geometric interpretation---would guarantee this.
792:
793: \acknowledgments
794:
795: We would like to thank M.~Aganagic, O.~DeWolfe, B.~Freivogel,
796: C.~Keeler, H.~Lin, J.~Maldacena, S.~Shenker, B.~Tweedie and especially
797: M.~Van Raamsdonk for helpful discussions. This work was supported by
798: the Berkeley Center for Theoretical Physics, by a CAREER grant of the
799: National Science Foundation, and by DOE grant DE-AC03-76SF00098.
800:
801:
802:
803:
804: \appendix
805:
806:
807: \section{The $N=1$ sector}
808: \label{sec-n1}
809:
810: In the geometric regime the matrix model is strongly coupled, except
811: in the case $N=1$ when all interaction terms vanish. This sector is
812: present for all values of $N$ since $U(N)=U(1)\times SU(N)$. It
813: describes the decoupled dynamics of the center of mass. In this
814: appendix we discuss the spectrum of the $U(1)$ sector. We will argue
815: that it does not contribute to the entropy in the Bekenstein bound
816: because all of its states are gauge copies of each other under
817: diffeomorphisms.
818:
819: The Hamiltonian and creation operators for the $N=1$ sector can be obtained
820: from Eqs.~(\ref{eq-ham}) and (\ref{eq-ops}) as a special case. It is
821: the Hamiltonian for a particle in a nine-dimensional harmonic
822: oscillator, namely a graviton of longitudinal momentum $1/R$ in
823: linearized 11D supergravity~\cite{KimYos03}. Its frequency is set by
824: the curvature parameter of the plane wave, $\mu$, and its effective
825: mass is given by the longitudinal momentum, $1/R$. Hence, the
826: characteristic length scale of the oscillator is $(R/\mu)^{1/2}$. The
827: spectrum is given by infinite towers of bosonic states generated by
828: the creation operators $A^\dagger$, with an additional multiplicity of
829: $2^8$ from the fermionic operators $\psi^\dagger$.
830:
831: Application of the Bekenstein bound requires not only that the
832: background be weakly curved, but also that the backreaction of the
833: system we study be negligible. In particular, the light-sheet in the
834: $x^-$ direction should not focus much over a distance $2\pi R$. The
835: total area loss of the light-sheet is $N$~\cite{Bou03}. This
836: corresponds to negligible focussing only if the transverse area of the
837: system (here, the graviton) is much larger than $N$:
838: \begin{equation}
839: A\gg N~.
840: \label{eq-an}
841: \end{equation}
842: In the ground state, the graviton occupies a transverse area of order
843: $(R/\mu)^{9/2}$, so weak backreaction requires
844: \begin{equation}
845: \mu/R\ll 1~.
846: \label{eq-br1}
847: \end{equation}
848: This condition is automatically satisfied in a weakly curved
849: background; see Sec.~\ref{sec-pwmm}.
850:
851: The spread of the graviton wave function in the transverse directions
852: will be of order $\sqrt{RE/\mu}$. The area loss along $x^-$ is $N$
853: independently of $E$, so the backreaction becomes weaker for excited
854: states. The point is that the spreading of the wave function over a
855: nine-dimensional area overcompensates for the larger energy leading
856: to a decrease in energy density and in backreaction. Hence, the
857: spectrum will not be truncated due to large backreaction.
858:
859: Under the criteria offered so far, it would appear that all states of
860: the nine-dimensional oscillator contribute to the entropy as long as
861: we choose $\mu/R\ll 1$. Their number is infinite\footnote{The entropy
862: grows with energy as $S_{N=1}(E) \sim 9 \log E$, but there is no
863: cutoff on $E$.}, so the Bekenstein bound would seem to be violated
864: already for $N=1$.
865:
866: However, this is clearly wrong for the simple reason that the states
867: we have considered are all the same. It is easiest to see this in the
868: (overcomplete) basis of coherent states. All coherent states can be
869: mapped to the ground state by an $SU(4|2)$ transformation, i.e., by a
870: symmetry transformation that leaves the form (\ref{eq-metric}) of the
871: plane wave metric invariant. The required generators~\cite{BlaOlo02}
872: commute with $\partial_-$, so the (discrete) longitudinal momentum is
873: not affected by this transformation. Hence, the Bekenstein bound is
874: trivially satisfied for $N=1$.
875:
876: The same argument will apply to the center-of-mass motion for larger
877: values of $N$. In the matrix model this corresponds to the $U(1)$
878: factor that decouples from the $SU(N)$ degrees of freedom. We discard
879: the $U(1)$ states for the same reason that we would not count
880: different boosts of the same system as distinct bound states in flat
881: space.
882:
883:
884: \section{Protected representations}
885: \label{sec-protected}
886:
887: Here we give a brief summary of the form and properties of
888: supersymmetrically protected states in the plane wave matrix model.
889: For more details see, e.g., Refs.~\cite{DSV2,M5,LinMal05}.
890:
891: The symmetry algebra of the eleven-dimensional plane wave is a basic
892: classical Lie superalgebra
893: \begin{equation}
894: SU(4|2) \supset SU(4) \times SU(2) \times U(1)_H \sim SO(6) \times
895: SO(3) \times U(1)_H~.
896: \end{equation}
897: The bosonic subalgebra $SO(6) \times SO(3)$ describes the symmetry of
898: the nine-dimensional transverse plane, and the Hamiltonian is the
899: $U(1)_H$ hypercharge. One can also think of this symmetry group as
900: arising in the Penrose limit of AdS/CFT. On the AdS side, the Penrose
901: limit of $AdS_7 \times S^4$ gives the plane wave. On the CFT side,
902: the corresponding contraction of the $\mathcal{N}=4$ superconformal
903: group $SU(2,2|4)$ gives the supergroup $SU(4|2)$.
904:
905: At $\mu\to\infty$, all superrepresentations in the plane wave matrix
906: model are tensor representations. Hence they are described by Young
907: tableaux. To distinguish a supertableau from a bosonic tableau,
908: slashed boxes are used. Any $SU(4|2)$ superrepresentation can be
909: completely decomposed into representations of the bosonic subgroup
910: $SU(4) \times SU(2)$. For example, the superrepresentation that is
911: the sole building block of the entire $U(1)$ spectrum has the
912: following decomposition
913: \begin{equation}
914: \label{eq-superdecomp}
915: {\tiny \yngSLASH(1,1)} \; = \;
916: \left( \tableau{1 1}, 1 \right) \; \oplus \; \;
917: \left( \tableau{1}, \tableau{1} \right) \; \; \oplus \;
918: \left( 1, \tableau{2} \right)~.
919: \end{equation}
920:
921: Conversely, starting with a highest weight bosonic representations
922: $\vert \psi \rangle$, the full superrepresentation can be recovered by
923: acting with the $2^8$ combinations of fermionic lowering operators.
924: If the resulting states are all independent, the superrepresentation
925: is called ``typical''; otherwise, ``atypical''.
926:
927: A special set of multiplets are nonperturbatively protected from
928: receiving corrections to their energy as the parameter $\mu$ of the
929: matrix model is modified. The Young supertableaux of these ``doubly
930: atypical'' representations have two rows of equal length
931: \begin{equation}
932: \label{eq-DAform}
933: \cdot \, , \; {\tiny \yngSLASH(1,1) } \, , \;
934: {\tiny \yngSLASH(2,2) } \, , \; \ldots, \;
935: \underbrace{ {\tiny \yngSLASH(6,6) } }_n \, , \; \ldots
936: \end{equation}
937: and dimension
938: \begin{equation}
939: \rm{dim}_{\rm{DA}}(n)=\frac{1}{3}\left( 4n^4+16n^3+20n^2+8n+3 \right)~.
940: \end{equation}
941: Their bosonic decomposition is (for $n \geq 2$)
942: \begin{eqnarray}
943: \label{eq-DAdecomp}
944: \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!
945: \underbrace{ \tiny{ \yngSLASH(6,6) } }_n \; &=& \;
946: \Big( \underbrace{ \tableau{6 6}}_n \, ,
947: 1 \Big)_{{\bf \frac{1}{4}}\,\rm{BPS} } \; \oplus \;
948: \Big( \underbrace{ \tableau{6 5}}_n \, ,
949: \tableau{1} \Big)_{{\bf \frac{1}{8}}\,\rm{BPS}} \; \oplus \;
950: \Big( \underbrace{ \tableau{5 5}}_{n-1} \, ,
951: \tableau{2} \Big)_{{\bf \frac{1}{8}}\,\rm{BPS}} \nonumber \\
952: &\oplus& \; \Big( \underbrace{ \tableau{6 4}}_{n} \, ,
953: \tableau{1 1} \Big)_{{\bf \frac{1}{4}}\,\rm{BPS}} \; \oplus \;
954: \Big( \underbrace{ \tableau{5 4}}_{n-1} \, ,
955: \tableau{2 1} \Big)_{{\bf \frac{1}{8}}\,\rm{BPS}} \; \oplus \;
956: \Big( \underbrace{ \tableau{4 4}}_{n-2} \, ,
957: \tableau{2 2} \Big)_{{\bf \frac{1}{4}}\,\rm{BPS}}~,
958: \end{eqnarray}
959: where, following Ref.~\cite{DSV2}, each bosonic subrepresentations has
960: been labeled by the fraction of the 32 supersymmetries it preserves.
961: The energy of all constituent states is of order $E \sim n$.
962:
963: The nonstandard supersymmetry algebra on the plane wave allows for BPS
964: states with nonzero energy. This follows from the commutation
965: relations between the supersymmetry generators and the Hamiltonian
966: $H$. Schematically,
967: \begin{eqnarray}
968: \left\{ Q^{\dagger}, Q \right\} &\sim& H - \mu M^{ij} - \mu M^{ab}~,
969: \nonumber \\
970: \left[ H, Q \right] &\sim& \mu Q~,
971: \end{eqnarray}
972: where the $SO(3)$, $SO(6)$ rotation generators $M^{ij}$, $M^{ab}$
973: allow for positive definite $H$ when $\left\{ Q^{\dagger}, Q
974: \right\}=0$.
975:
976: \bibliographystyle{board}
977: \bibliography{all}
978: \end{document}
979:
980: