gr-qc0207111/d.tex
1: \documentclass[11pt]{article}
2: \usepackage{amsmath,amsthm,latexsym,amssymb,amsfonts}
3: 
4: \addtolength{\textwidth}{2cm}
5: \addtolength{\oddsidemargin}{-1cm}
6: 
7: \newtheorem{lemm}{Lemma}[section]
8: \newtheorem{prop}[lemm]{Proposition}
9: \newtheorem{coro}[lemm]{Corollary}
10: \newtheorem{defi}[lemm]{Definition}
11: \newtheorem{assu}[lemm]{Assumption}
12: 
13: \newcommand{\scripta}{\mathfrak{A}} 
14: \newcommand{\hilb}{\mathcal{H}}
15: \newcommand{\Hilb}{\mathfrak{h}}
16: \newcommand{\one}{\text{\bf 1}}
17: \newcommand{\X}{X}
18: \newcommand{\E}{E}
19: \newcommand{\abar}{\overline{\mathcal{A}}}
20: \newcommand{\isom}{\cong}
21: \newcommand{\mo}[1]{^{\text{mod}}#1}  
22: \DeclareMathOperator{\porder}{\mathcal{P}}
23: \DeclareMathOperator{\im}{Im}
24: \DeclareMathOperator{\re}{Re} 
25: \DeclareMathOperator{\Span}{span}
26: \DeclareMathOperator{\cyl}{Cyl}
27: \DeclareMathOperator{\dom}{dom}  
28: \DeclareMathOperator{\sutwo}{SU(2)}
29: \newcommand{\rep}[1]{\pi\left(#1\right)}
30: \newcommand{\comm}[2]{\left[#1\,,\,#2\right]}
31: \newcommand{\F}[2]{F_{#1,#2}} 
32: \newcommand{\scpr}[2]{\left\langle#1\,,\, #2 \right\rangle} 
33: \newcommand{\vvec}[1]{\underline{#1}}
34: \newcommand{\norm}[1]{\left\lVert #1 \right\rVert}     
35: \newcommand{\betr}[1]{\left\lvert #1 \right\rvert}    
36: 
37: 
38: \begin{document}
39: \title{Some Comments on the Representation Theory of the Algebra
40:   Underlying Loop Quantum Gravity}
41: \author{
42: Hanno Sahlmann\thanks{email address: sahlmann@aei-potsdam.mpg.de}\\
43:        MPI f\"ur Gravitationsphysik, Albert-Einstein-Institut, \\
44:            Am M\"uhlenberg 1, 14476 Golm near Potsdam, Germany}
45: \date{{\small PACS No. 04.60, Preprint AEI-2002-056}}
46: 
47: \maketitle
48: 
49: \begin{abstract}
50: Important characteristics of the loop approach to quantum gravity 
51: are a specific choice of the algebra $\scripta$ of observables and 
52: of a representation of $\scripta$ on a measure space over the space
53: of generalized connections. This representation is singled out by its 
54: elegance and diffeomorphism covariance. 
55: 
56: Recently, in the context of the quest for \textit{semiclassical
57: states}, states of the theory in which the quantum gravitational
58: field is close to some classical geometry, it was realized that it
59: might also be worthwhile to study different representations of the
60: algebra $\scripta$ of observables.   
61: 
62: The content of the present note is the observation that under some
63: mild assumptions, the mathematical structure of representations of 
64: $\scripta$ can be analyzed rather effortlessly, to a certain extent: 
65: Each representation can be labeled by sets of functions and measures 
66: on the space of (generalized) 
67: connections that fulfill certain conditions.  
68: 
69: These considerations are however mostly of mathematical nature. 
70: Their physical content remains to be clarified, and physically
71: interesting examples are yet to be constructed. 
72: \end{abstract}
73: %---------------------------------------------------------------------------
74: \section{Introduction}
75: %---------------------------------------------------------------------------
76: Loop quantum gravity (LQG for short) is a promising approach to the
77: problem of finding a quantum theory of gravity, and has led to many
78: interesting insights. It is based on the formulation of gravity as a
79: constrained canonical system in terms of the Ashtekar variables
80: \cite{Ashtekar:1986yd}, a canonical pair of an SU(2)-connection (in its
81: real formulation) and a triad field.
82: 
83: One of the interesting features of LQG (and perhaps one reason for its 
84: success) is its specific choice of basic observables: The configuration 
85: variables are holonomies along curves in the spacial slices of the
86: spacetime, the basic momentum variables are integrals of a triad field 
87: over surfaces in the spacial slices of the spacetime. 
88: This is in contrast to ordinary quantum field theories, where both the 
89: configuration and the momentum observables are three dimensional
90: integrals of the basic field and its conjugate momentum. 
91: The choice of basic variables in LQG is, however, well motivated, since 
92: in contrast to other possibilities, these variables can be
93: defined without recurse to a fixed classical background geometry, and
94: it furthermore leads to well defined operators for interesting
95: geometric quantities such as area and volume. 
96: 
97: A quantum theory for this type of basic variables was first given by Rovelli
98: and Smolin in \cite{Rovelli:1990za}. Since then, much work has gone
99: into extracting the essence of this quantization and putting it onto 
100: firm mathematical ground. Key ideas in this context were the use of 
101: $C^*$-algebraic methods \cite{Ashtekar:1992kc} and projective limit techniques
102: \cite{Ashtekar:1995mh,Ashtekar:1995wa} resulting in what is now called the 
103: \textit{connection representation}. This representation is based on a
104: Hilbert space which is an $L^2$ space over the space of connections
105: with respect to a certain measure, the Ashtekar-Lewandowski measure. 
106: The holonomies act as multiplication operators and the integrated 
107: triad fields as certain vector fields. 
108: Due to its diffeomorphism invariance and mathematical elegance, this
109: representation is considered the fundamental representation of LQG.  
110: 
111: That it might nevertheless be interesting to also consider 
112: representations other than the AL-representation was realized when
113: attempts were made to construct states for LQG in which the
114: quantum gravitational field behaves almost classical. 
115: The first proposal in this direction was contained in 
116: \cite{Arnsdorf:1999wn}. There, the goal was to find states for LQG
117: that have semiclassical properties for spacetimes with non-compact
118: spacial slices.   
119: Representations that are inequivalent to the AL-representation also 
120: seem to arise if one implements the ideas \cite{Bombelli:2000ua} about
121: the use of statistical geometry for the construction of semiclassical
122: states. 
123: Finally, in a series of recent works \cite{V2,V3,Ashtekar:2001xp}, 
124: measures on the space of
125: generalized connections were constructed that derive from the Gaussian
126: measure of ordinary (background dependent) free quantum field theory. 
127: 
128: The representation theory for the holonomy algebra is well understood
129: and many representations inequivalent to the AL-representation have 
130: been considered in the literature. 
131: Less attention has been paid to the question of what happens when one
132: also takes the integrated triads into consideration. 
133: The main observation of the present note is that due to the 
134: structure of its commutation relations, representations of the
135: combined algebra of holonomies and integrated triads can, without
136: effort, be analyzed to a certain extent: 
137: Each representation can be labeled by sets of functions and measures 
138: on the space of (generalized) connections that fulfill certain
139: conditions. 
140: 
141: However, a cautionary remark is in order here. The considerations of the
142: present note are mostly of mathematical nature. Truly interesting,  
143: albeit difficult, tasks would be to state physically 
144: motivated criteria for singling out interesting classes of
145: representations, actually constructing such representations, and
146: understanding the physical content of representations constructed by 
147: mathematical considerations. None of this will be addressed in this
148: note. However, we hope that it can be used as a starting point 
149: when approaching those questions motivated by physics. 
150: 
151: To finish this introduction, we should mention that the occurrence of
152: inequivalent representations of the observable algebra is well known
153: from quantum field theory and quantum statistical mechanics
154: \cite{Haag:1992hx}: In that context, it was realized
155: that the choice of representation for the observable algebra contains 
156: important physical information: Roughly speaking, whereas the
157: algebraic structure of the theory encodes the physical
158: system one is considering, the chosen representation carries the
159: global information about the physical state, the system is in. It
160: might for example decide whether the system is in a ground or in a
161: thermal state or whether the state carries a global charge.  
162: Since the change of the global properties of a state of the system 
163: is not always physically realizable (it might necessitate an infinite
164: amount of energy or the creation of charges) the emergence of inequivalent
165: representations is quite natural. 
166: The considerations of representations different from the
167: AL-representation in the quest for the semiclassical
168: regime of LQG fits quite nicely into this general picture.   
169: %---------------------------------------------------------------------------
170: \section{CQG briefing}
171: %---------------------------------------------------------------------------
172: It is sometimes helpful to quantize
173: a given classical system in two steps. The first consists in
174: associating to each member of a chosen set of classical observables,
175: an operator in some (abstract) *-algebra $\scripta$, such that
176: \begin{itemize}
177: \item The Poisson structure of the classical observables is mirrored
178:   as closely as possible by the commutators within the algebra
179:   ("Poisson brackets go to commutators").
180: \item Complex conjugate on classical observables are mapped to
181:   conjugates under the $*$-operation on $\scripta$.
182: \end{itemize}
183: The importance of the second condition lies in the fact that it
184: ensures that real classical quantities will be associated with
185: symmetric operators, which in turn have spectrum on the real line and
186: real expectation values. If this would not be the case, the
187: interpretation of the resulting quantum theory would be completely
188: obscure.
189:  
190: The second step consists in choosing a
191: $*$-representation of the algebra $\scripta$, thus enabling one to
192: compute expectation values and hence make physical
193: predictions.
194: 
195: The purpose of this section is to look at the
196: first of these two steps in the context of LQG. It has been
197: extensively studied there and the choice of the set of classical
198: observables as well as the corresponding $*$-algebra which is made can
199: be regarded as the very essence of LQG.
200: In this section, we
201: will briefly review these developments to make the paper self
202: contained as well as fix the notation. Thus the pedagogical value of
203: this section will be very low. For an extensive recent review
204: see \cite{Thiemann:2001yy}.
205: 
206: As a first step recall that the canonical pair in LQG is a SU(2)
207: connection one-form $A$ and a frame field $E_I$ with a nontrivial
208: density weight. Both of these take values on a spacial slice $\Sigma$
209: of the four-manifold $M$.  Being a one-form, $A$ can be integrated
210: naturally (that is, without recurse to background structure) along
211: curves $e$ in $\Sigma$, to form \textit{holonomies}
212: \begin{equation*}
213: h_e[A]=\porder\exp\left[i\int_e A_a ds^a \right].
214: \end{equation*}
215: 
216: It turns out to be convenient to consider functions of $A$ which are 
217: slightly more general.
218: \begin{defi}
219:     A \textit{graph} in $\Sigma$ is a collection of analytic, oriented curves 
220:     in $\Sigma$ which intersect each other at most in their 
221:     endpoints.
222: 
223:     A function $c$ depending on connections $A$ on $\Sigma$ just in
224:     terms of their holonomies along the edges of a graph, i.e. 
225:     \begin{equation*}
226:       c[A]\equiv c(h_{e_1}[A],h_{e_2}[A],\ldots,h_{e_n}[A] ), \qquad
227:       e_1,e_2,\ldots, e_n\quad\text{ edges of some }\gamma, 
228:     \end{equation*}
229:     where $c(g_1,\ldots, e_n)$ viewed as a function on $\sutwo^n$ is
230:     \textit{continuous}, will be called \textit{cylindrical}. 
231: \end{defi}
232: Analyticity of the edges is required to exclude certain pathological 
233: intersection structures of the edges with surfaces which would render
234: the Poisson brakets which will be introduced below ill-defined.
235:   
236: It turns out that the set of cylindrical functions can be equipped with 
237: a norm (essentially the sup-norm for functions on SU(2)$^n$) such that 
238: its closure $\cyl$ with respect to that norm is a
239: commutative C$^*$-algebra. We will not spell out the details of this
240: construction but refer the reader to the beautiful presentations
241: \cite{Ashtekar:1995mh,Ashtekar:1995wa}. We note furthermore that by changing the word
242: ``continuous'' 
243: in the above definition to ``$n$ times differentiable'', we can define
244: subsets $\cyl^n$ of $\cyl$ and, most importantly for us, 
245: \begin{equation*}
246:   \cyl^\infty\doteq \bigcap_n\cyl^n, 
247: \end{equation*}
248: the space of smooth cylindrical functions. 
249: 
250: The density weight of $E$ on the other hand is such that, using an
251: additional real (co-)vector field $f^i$, it can be naturally integrated
252: over oriented surfaces $S$ to form a quantity
253: \begin{equation*}
254: E_{S,f}= \int_S E^a_i f^i \epsilon_{abc}\,dx^b\,dx^c
255: \end{equation*}
256: analogous to the electric flux through $S$. 
257: 
258: One of the defining choices 
259: of LQG is to base the quantization precisely on the elements of $\cyl$ 
260: and the fluxes $E_{S,f}$
261: as classical observables.
262: From the Poisson brackets of $A$ and
263: $E$ on can compute the Poisson brackets for the $c$, $E_{S,f}$. 
264: Call $\gamma$ adapted to $S$ if all $p$ in $S\cap\gamma$ are vertices 
265: of $\gamma$. 
266: Let $c$ be a function cylindrical on $\gamma$ and $S$ some analytical 
267: surface. Without restriction of generality we assume that $\gamma$ is 
268: adapted to $S$\footnote{There is always $\gamma'$ that contains
269:   $\gamma$ such that $\gamma'$ is 
270: adapted to $S$. A $c$ cylindrical on $\gamma$ is clearly also
271: cylindrical on $\gamma'$. For details see \cite{Ashtekar:1997eg}.}. Then   
272: \begin{equation*}
273:   \left\{E_{S,f},c\right\} = \frac{\kappa}{2}\sum_{p\in S\cap\gamma}
274:   \sum_{e_{p}} \kappa(e_{p})f_{i}X_{e_{p}}^{i}[c],
275: \end{equation*}
276: where the second sum is over the edges of $\gamma$ adjacent to $p$, 
277: \begin{equation*}
278:     \kappa(e_{p})=\begin{cases}1&\text{ if $e$ lies above $S$}\\
279: 0&\text{ if $e$ is tangential to $S$}\\ 
280: -1&\text{ if $e$ lies below $S$}\end{cases},
281: \end{equation*}
282: and $X^{I}_{e_{p}}$ is the $I$th left-invariant (right-invariant) 
283: vector field on SU(2) acting on the argument of $c$ corresponding to 
284: the holonomy $h_{e_{p}}$ if $e_{p}$ is pointing away from (towards) $S$.
285: $\kappa$ is the coupling constant of gravity. 
286: 
287: Surprisingly, the Poisson brackets of the $E_{S,f}$ among themselves
288: do not vanish as one would expect for the momentum observables. This
289: poses two questions: Can one nevertheless give some well defined
290: ``Poisson bracket goes to commutator''-prescription to associate algebra
291: elements to classical observables? And: Can one understand where this
292: non-commutativity of the momentum observables come from? As shown in
293: \cite{Ashtekar:1998ak}, both questions can be answered affirmatively.
294: We do not want to repeat the discussion of \cite{Ashtekar:1998ak} here
295: but just give its
296: result, condensed in a definition of the algebra $\scripta$ on which
297: the quantum theory will be based, as well as the association of
298: classical observables to algebra elements. Let 
299: \begin{equation*}
300:     X_{S,f}[c]:=\frac{il_P^2}{2}\sum_{p\in S\cap\gamma} \sum_{e_{p}}
301:     \kappa(e_{p})f_{i}X_{e_{p}}^{i}[c],
302: \end{equation*}
303: where we have used the notation introduced above. 
304: \begin{defi}
305:   Let $\scripta$ be the algebra generated by the cylindrical functions
306:   $\cyl$, together with the derivations $X_{S,f}$ on $\cyl$. On
307:   $\scripta$ a $*$-operation is given by the usual complex conjugation
308:   on $\cyl$ and the trivial $(X_{S,f})^*:=X_{S,f}$ on $\X$.
309: 
310:   The association of the classical functionals $\cyl, \E$ 
311:   with elements of $\scripta$ is then given by
312: \begin{align*}
313:   c &\mapsto c,\\
314:   E_{S,f} &\mapsto X_{S,f},\\
315:   \left\{E_{S,f},E_{S',f'}\right\} &\mapsto i\hbar^{-1}\comm{X_{S,f}}{X_{S',f'}},
316: \end{align*}
317: and higher order Poisson brackets of elements of $\E$ are mapped to
318: the higher order commutators of the corresponding derivations.
319: \end{defi} 
320: Note that since $\scripta$ is generated by the elements of 
321: $\cyl$ and $\X$, a representation $\pi$ of $\scripta$ is completely 
322: determined once the representors $\rep{\cyl}$ and $\rep{\X}$ are known.  
323: %---------------------------------------------------------------------------
324: \section{Remarks on the representation theory of $\scripta$}
325: %---------------------------------------------------------------------------
326: In the present section we will make some simple observations on the 
327: structure of representations of $\scripta$. Before we proceed to the
328: details, let us give a brief outline of what we are going to do.
329: 
330: We assume that a representation of $\scripta$ is given. As a first
331: step, we appeal to the  powerful machinery available
332: for representation of C$^*$-algebras, to decompose
333: the representation space into subspaces on which $\cyl$ acts cyclic. 
334: Then we look at the action of the representors of the $E_{S,f}$ with
335: respect to this decomposition. Since the respective operators will be
336: unbounded, we have to make an assumption that gets possible domain
337: problems out of the way: We assume that the $\rep{E_{S,f}}$
338: share a certain dense set in their domains. Then we make the central 
339: observation that under this assumption, the action of the 
340: $\rep{E_{S,f}}$ is rather
341: simple: Roughly speaking it is the sum of a derivative defined by
342: $X_{S,f}$ and a multiplication operator.
343: As a consequence, we can show that each representation of $\scripta$
344: is uniquely determined by a set of measures and functions on the space 
345: of (generalized) connections fulfilling certain compatibility
346: conditions.   
347: Since despite our assumption the considerations might appear
348: exceedingly general, we will finish by giving a useful corollary of
349: our results in a rather simple case. 
350: 
351: Before we start our analysis of the representations of $\scripta$ we
352: want to recall some basic facts about the representation theory of
353: $\cyl$. As was realized in \cite{Ashtekar:1992kc,Ashtekar:1995mh}, many powerful
354: results are at hand because $\cyl$ is an unital Abelian
355: C$^*$-algebra.
356: Firstly we recall that, due to a theorem of Gelfand (see for example
357: \cite{Bratteli}), 
358: since $\cyl$ is Abelian, it is isomorphic, via some 
359: isomorphism $\iota$, to the algebra of continuous
360: functions on the \textit{spectrum} $\overline{\mathcal{A}}$, a compact 
361: Hausdorff space, of $\cyl$. From this and the Riesz-Markov Theorem (see 
362: for example \cite{Reed})
363: it follows that every positive linear functional
364: on $\cyl$ is given by a positive Baire measure on $\abar$.
365: The converse trivially holds true: Every positive Baire measure on
366: $\abar$ gives a positive functional on $\cyl$.
367: 
368: Now let $(\pi,\hilb)$ be a \textit{cyclic} representation of $\cyl$.  
369: Since the representation is cyclic, it 
370: defines a positive linear functional $\omega$ on $\cyl$ and certainly 
371: is unitarily equivalent to the GNS representation coming from $\omega$. 
372: Moreover as concluded above $\omega$ must be given by a positive Baire 
373: measure on $\abar$. Vice versa, every cyclic representation of $\cyl$ is given
374: by a positive Baire measure on $\abar$. 
375: Thus, we conclude that the cyclic representations of $\cyl$ are all of 
376: the form 
377: \begin{equation}
378: \label{eq7}
379: \hilb_\nu=L^2(\overline{\mathcal{A}}, d\mu_\nu),\qquad \pi_\nu(c)=\iota(c),
380: \end{equation} 
381: where $\mu_\nu$ is some positive Baire measure on $\abar$. 
382: Note that we did not have to assume continuity of the
383: cyclic representation. Rather, continuity follows automatically from
384: cyclicity.
385: 
386: It is important for the rest of this work, that because of their
387: structure \eqref{eq7}, for cyclic representations 
388: the $\rep{c}$ play a double role: On the one hand they are
389: operators on the representation space, on the other hand they are
390: $L^2$ functions. Let us note the following Lemma which will be
391: useful, 
392: later on:
393: \begin{lemm}
394: Let $(\pi,\hilb)$ be some cyclic representation of $\cyl$. 
395: Then $\rep{\cyl^{\infty}}$ is dense in $L^2(\overline{\mathcal{A}},d\mu)$. 
396: \end{lemm}
397: \begin{proof}[Sketch of the proof]
398: Since we do not want to introduce the projective limit machinery that
399: is used to define the closure $\cyl$ of the set of cylindrical
400: functions, we will only sketch the proof. The details can however be
401: easily fixed using the methods of
402: \cite{Ashtekar:1995mh,Ashtekar:1995wa}.
403: 
404: The idea for the proof is that functions in $\cyl^\infty$ can
405: essentially be viewed as subset of the continuous functions on a
406: compact space.    
407: They are separating points and the constant functions are
408: among them, so the Stone-Weierstrass Theorem (see for example \cite{Stone}) 
409: applies, showing that they are dense in $\cyl$ (wrt. its C$^*$ norm). 
410: Now cyclicity of the representation just means that $\rep{\cyl}$ is
411: dense in $\hilb$, whence $\rep{\cyl^{\infty}}$ is dense in $\hilb$ as
412: well. 
413: \end{proof}
414: Let now a representation $(\pi,\hilb)$ of $\scripta$ 
415: be given. It is well known that every representation of a C$^*$-algebra is a
416: direct sum of cyclic representations (see for example \cite{Conway}). 
417: Applying this to the representation $\pi\rvert_{\cyl}$ of $\cyl$ yields 
418: \begin{equation*}
419: \hilb \isom \bigoplus_\nu \hilb_\nu, \quad \quad \pi \isom \oplus_\nu \pi_\nu. 
420: \end{equation*}
421: where the $(\pi_{\nu},\hilb_{\nu})$ are cyclic and therefore 
422: \begin{equation*}
423:     \hilb_\nu\isom L^2(\overline{\mathcal{A}}, d\mu_\nu),\qquad 
424:     \pi_\nu(c)\isom \iota(c).
425: \end{equation*}   
426: To simplify notation in what follows, we will take all isometries as 
427: identities. Furthermore, we denote by $I_{\nu}$ the canonical 
428: inclusion
429: \begin{equation*}
430:     I_{\nu}: \hilb_{\nu}\hookrightarrow\hilb 
431: \end{equation*}
432: and by $P_{\nu}$ the canonical projection followed by the inverse of 
433: $I_{\nu}$
434: \begin{equation*}
435:     P_{\nu}:\hilb\rightarrow \hilb_{\nu}.
436: \end{equation*}
437: Now we have to analyze the action of the operators $\rep{E_{S,f}}$ on
438: $\hilb$. This gets complicated by the fact that they represent vector
439: fields and will therefore be unbounded operators. To get these
440: complications out of the way, we will make an assumption on $\pi$. To this 
441: end, let us define the following subspace of $\hilb$:
442: \begin{equation*}
443:     \Hilb:=\Span\left[ \bigcup_{\nu}I_{\nu}\left(\cyl^\infty\right)\right].
444: \end{equation*}
445: Note that $\Hilb$ is dense in $\hilb$ because $\cyl^\infty$ is dense in 
446: $\hilb_{\nu}$. 
447: With this definition at hand, we can state our 
448: assumption:\footnote{Note that this assumption \textit{does not} automatically 
449: follow from the perhaps more natural one that
450: $\{\rep{c}\,|\,c\in\cyl\}$ should be contained in the domains of the
451: $E_{S,f}$.}   
452: \begin{assu}
453:   \label{ass2} 
454:   The representation $\pi$ should be such that
455:   $\Hilb \subset \dom\left(\rep{E_{S,f}}\right)$ for all surfaces
456:   $S$ and co-vector fields $f$ on $S$.
457: \end{assu}
458: Under this assumption, the action of the $\rep{E_{S,f}}$ can be 
459: computed rather explicitely: Let $c$ be a cylindrical function. Then 
460: \begin{align*}
461:     \rep{E_{S,f}}\left[I_{\nu}(c)\right] 
462:     &=\rep{E_{S,f}}\rep{c}\left[I_{\nu}(\one)\right]\\
463:     &=\comm{\rep{E_{S,f}}}{\rep{c}}\left[I_{\nu}(\one)\right]
464:     +\rep{c}\rep{E_{S,f}}\left[I_{\nu}(\one)\right]\\
465:     &=I_{\nu}\left(X_{S,f}[c]\right) + 
466:     \sum_{\iota}I_{\iota}\left(cF^{\iota\nu}_{S,f}\right)
467: \end{align*}
468: where we have made the definition 
469: \begin{equation*}
470:     F^{\iota\nu}_{S,f}
471:     :=P_{\iota}\left(\rep{E_{S,f}}\left[I_{\nu}(\one)\right]\right).
472:     \in \hilb_{\iota}
473: \end{equation*}
474: Thus the action of the fluxes on $\Hilb$ and hence on $\hilb$ is 
475: completely determined by the $F_{S,f}^{\iota\nu}$. Let us exhibit some 
476: further properties of this family:
477: 
478: Because of Assumption \ref{ass2}, we have that\footnote{Note that from 
479: Assumption \ref{ass2} we directly get
480: $c\sum_{\iota}I_{\iota}\left(F^{\iota\nu}_{S,f}\right)\in\hilb$ for
481: cylindrical functions $c$. But since $\cyl$ also contains the constant 
482: functions, \eqref{dom} follows.} 
483: \begin{equation}
484: \label{dom}
485:   \tag{dom}
486:     \sum_{\iota}I_{\iota}\left(F^{\iota\nu}_{S,f}\right)\in\hilb. 
487: \end{equation}
488: More properties come from the fact that the $\rep{E_{S,f}}$ represent 
489: the $E_{S,f}$. First of all, for co-vector fields $f,f''$ on a 
490: surface $S$ and $f'$ on $S'$   
491: \begin{gather}
492: \label{rep}
493: \tag{rep}
494: \begin{split}
495:     F^{\iota\nu}_{S,f+f''}&=F^{\iota\nu}_{S,f}+F^{\iota\nu}_{S,f''},\\
496:     F^{\iota\nu}_{S,f}+F^{\iota\nu}_{S',f'}
497:     &=F^{\iota\nu}_{S\setminus S',f}+F^{\iota\nu}_{S'\setminus S,f'} 
498:     +F^{\iota\nu}_{S\cap S',f\pm f'},
499: \end{split}
500: \end{gather}
501: where in the second line $S \cap S'$ is given the orientation of $S$
502: and the sign  
503: depends on the relative orientation of $S$ and $S'$ on their
504: intersection.
505:  
506: Further relations come from the fact that $\pi$ is a *-representation: 
507: For $c,c'\in\cyl$ let 
508: \begin{equation*}
509:   \Delta_{S,f}^{(\iota)}(c,c'):=\scpr{X_{S,f}[c]}{c'}_{\hilb_\iota}
510:  -\scpr{c}{X_{S,f}[c']}_{\hilb_\iota}
511: \end{equation*}
512: denote the divergence of the vector field $X_{S,f}$ with respect to
513: the measure $\mu_\iota$. As the $\rep{E_{S,f}}$ have to be symmetric, we
514: have 
515: \begin{align*}
516:     \scpr{\rep{E_{S,f}}I_{\nu}(c)}{I_{\iota}(c')}
517:     &=\scpr{I_{\nu}(c)}{\rep{E_{S,f}}I_{\iota}(c')}\\
518:     \Leftrightarrow \delta_{\nu\iota}\Delta^{\iota}_{S,f}(c,c')
519:     &=\scpr{c}{F^{\nu\iota}_{S,f}c'}_{\hilb_{\nu}}
520:     -\scpr{F^{\iota\nu}_{S,f}c}{c'}_{\hilb_{\iota}}
521: \end{align*}
522: which can easily seen to be equivalent to 
523: \begin{equation}
524: \label{div}
525: \tag{div}
526: \begin{split}
527:     \Delta_{S,f}^{(\iota)}(c,c')
528:     &=2i\scpr{c}{\im\left(F_{S,f}^{\iota\iota}\right)c'},\\
529:     F^{\nu\iota}\,d\mu_{\nu}&=\overline{F^{\iota\nu}}\,d\mu_{\iota} 
530:     \qquad \text{for } \iota\neq \nu.
531: \end{split}
532: \end{equation}
533: Let us summarize our findings 
534: \begin{prop}
535: \label{pr1}
536: Any representation $(\pi,\hilb)$ of $\scripta$, fulfilling our
537: Assumption \ref{ass2}, determines
538: \begin{itemize}
539:   \item A family of positive measures $\{\mu_{\nu}\}$ on $\abar$, 
540:   \item A family of functions $\{F^{\iota\nu}_{S,f}\}$, where  
541: $F^{\iota\nu}_{S,f}\in L^2(\abar,d\mu_{\iota})$, 
542: \end{itemize}  
543: such that \eqref{dom}, \eqref{rep}, \eqref{div} are fulfilled.  
544: \end{prop}
545: It is probably more interesting to note that also the converse holds true:
546: \begin{prop}
547:   \label{pr2}
548:   Let a family of measures $\{\mu_{\nu}\}$ on $\abar$ and a family of 
549:   functions $\{F^{\iota\nu}_{S,f}\}$, where  
550:   $F^{\iota\nu}_{S,f}\in L^2(\abar,d\mu_{\nu})$ that fulfill 
551:   \eqref{dom}, \eqref{rep}, \eqref{div}. 
552: From this data, one can construct a representation $\pi$ of
553: $\scripta$ that fulfills the Assumption \ref{ass2}. 
554: \end{prop}
555: \begin{proof}
556: The proof is quite obvious: Let $\{\mu_{\nu}\}$,
557: $\{F^{\iota\nu}_{S,f}\}$ fulfilling \eqref{dom}, \eqref{rep},
558: \eqref{div} be given. 
559: The representation space is defined as 
560: \begin{equation*}
561:   \hilb \doteq\bigoplus_\nu L^2(\abar, d\mu_\nu)
562: \end{equation*}
563: whence
564: \begin{equation*}
565:   \rep{c}\oplus_\nu f_\nu \doteq \oplus_\nu c f_\nu, \qquad f_\nu\in
566:   L^2(\abar, d\mu_\nu). 
567: \end{equation*}
568: Now
569: \begin{equation}
570: \label{eq5}
571:   \rep{E_{S,f}}[\oplus_\nu c_\nu]\doteq 
572:   \oplus_\nu\left(X_{S,f}[c_\nu]+c_\nu\sum_\iota
573:     F^{\nu\iota}_{S,f}\right),\qquad c_\nu \in \cyl^{\infty} 
574: \end{equation}
575: defines operators that are well defined on $\Hilb$ because of
576: \eqref{dom}, hence Assumption \ref{ass2} is fulfilled. 
577: Moreover they are symmetric because of \eqref{div} and give a
578: representation of the $E_{S,f}$ since the commutator with the
579: representors of cylindrical functions is right and
580: \eqref{rep} holds. 
581: Finally, \eqref{eq5} completely determines the action of the $\rep{E_{S,f}}$
582: on $\Hilb$ and hence on $\hilb$. 
583: \end{proof}
584: The above results may appear exceedingly general. 
585: Let us therefore reduce consideration to representations in which
586: $\cyl$ acts cyclic and state the following corollary which is perhaps
587: closer to applications than the above general results: 
588: \begin{prop}
589:   \label{pr3}
590:   Let a \textit{cyclic} representation $(\pi,\hilb)$ of $\cyl$ be given. 
591:   Then a necessary and sufficient condition for $\pi$ to be
592:   extendable to a representation, fulfilling Assumption \ref{ass2}, 
593:   of the whole $\scripta$ is that for
594:   each surface $S$ and co-vector field $f$ on $S$ there exists a
595:   constant $C_{S,f}$ such that
596:   \begin{equation}
597:     \label{eq6}
598:     \betr{\Delta_{S,f}(c,\one)} \leq C_{S,f}\norm{c}_{\hilb}\qquad\text{ for 
599:       all }c\in \cyl^{\infty}
600:   \end{equation}
601:   where the sesquilinear form $\Delta_{S,f}$ is given by
602:   \begin{equation*}
603:     \Delta_{S,f}(c,c')\doteq \scpr{\rep{X_{S,f}[c]}}{\rep{c'}}_{\hilb}-
604:     \scpr{\rep{c}}{\rep{X_{S,f}[c']}}_{\hilb}, \qquad c,c'\in\cyl.
605:   \end{equation*}
606: \end{prop}
607: \begin{proof}
608: Let us first prove necessity: Let a representation 
609: $(\tilde{\pi},\hilb)$ of $\scripta$ be given such that
610: $\tilde{\pi}\rvert_{\cyl}=\pi$. Application of Proposition \ref{pr1}
611: then yields a measure $\mu$ and a family of functions
612: $\{F_{S,f}\}$ satisfying \eqref{dom}, \eqref{rep}, \eqref{div}. 
613: Thus we can finish by noting that \eqref{div}, \eqref{dom} imply
614: \eqref{eq6}.
615:   
616: Sufficiency can be proved by straightforward construction: 
617: Let a cyclic representation $(\pi,\hilb)$ of $\cyl$, fulfilling the
618: condition \eqref{eq6}, be given. 
619: Because of cyclicity, $\hilb$ is isomorphic to $L^2(\abar,d\mu)$ for
620: some positive regular measure $\mu$. Moreover, $\rep{\cyl^\infty}$ is
621: dense in $\hilb$. Therefore 
622: the Riesz Representation Theorem (see for example \cite{Reed}) shows that
623: \eqref{eq6} implies that $\Delta_{S,f}(c,\one)$ is given by an element
624: $\widetilde{F}_{S,f}$ of $\hilb$,
625: i.e. $\Delta_{S,f}(c,\one)=\langle\rep{c},\widetilde{F}_{S,f}\rangle$.
626: Using the fact that $\hilb$ is an $L^2$ space, one can easily see that 
627: $\Delta_{S,f}(c,c')$ is determined by $\widetilde{F}_{S,f}$ as well
628: and that $\re(\widetilde{F}_{S,f})=0$. 
629: Set $F_{S,f}\doteq \widetilde{F}_{S,f}/2$. The $F_{S,f}$ fulfill
630: \eqref{dom}, \eqref{rep} because the $\widetilde{F}_{S,f}$ do. 
631: Moreover, the $F_{S,f}$ satisfy \eqref{div}. So one can construct a
632: representation of $\scripta$ from the data $\mu,\{F_{S,f}\}$ by
633: Proposition \ref{pr2}. 
634: \end{proof}
635: %--------------------------------------------------------------------
636: \section{Discussion}
637: %--------------------------------------------------------------------
638: Let us start the discussion of the above results by describing the
639: simplest case, the AL-representation. In that representation, $\cyl$ 
640: acts cyclic, the corresponding measure on $\abar$ is the 
641: Ashtekar-Lewandowski measure $\mu_{\text{AL}}$ constructed in 
642: \cite{Ashtekar:1995mh} and the $F_{S,f}$ are all equal to zero. 
643: 
644: Next, we remark that Assumption \ref{ass2} precludes the possibility that the
645: generalized divergences $\Delta_{S,f}$ \textit{are} given by functions which
646: however \textit{are not} $L^2$. This case does not seem unnatural, 
647: so it might appear too restrictive to exclude it. Note however that
648: admitting that case as well would mean that not all of $\cyl$ 
649: (especially not the constant functions) would be contained in the
650: domains of the $\rep{E_{S,f}}$. On the other hand the
651: cylindrical functions are the only ones we have direct control on,
652: and removing some of them would most likely leave us with a set that
653: is not dense anymore. Thus it would be extremely difficult to work 
654: with such more general representations.  
655: 
656: Also, we would like to make some remarks concerning Proposition
657: \ref{pr3}: As we saw, it is simple to derive that result. It turns out 
658: to be much more difficult to actually come up with an example for a
659: measure on $\abar$ fulfilling the condition, other then the
660: AL-measure. All measures constructed so far, with the remarkable
661: exception of the AL-measure, violate \eqref{eq6}. The interested
662: reader is referred to \cite{2} for a closer investigation of this
663: subject. 
664: A class of representations that avoids this problem is the one
665: obtained by using the AL-measure but having the $F_{S,f}$ real and not 
666: equal to zero. In such representations, the 
667: $\rep{E_{S,f}}$ have non vanishing expectation values. 
668:  
669: As the last remark showed, this note merely provides a starting point
670: for the analysis of the representations of $\scripta$, and much more
671: difficult and interesting problems remain to be tackled.   
672: Nevertheless, we hope that this note is a useful preparation for that 
673: task, and we would like to come back to some of the questions related
674: to it in future publications. 
675: \\
676: \\
677: \\ 
678: {\bfseries \Large Acknowledgements}
679: \\
680: \\
681: It is a pleasure to thank Thomas Thiemann for numerous valuable
682: discussions and suggestions concerning the present work. We are also
683: grateful to Fotini Markopoulou for many very helpful discussions 
684: on conceptual issues in quantum gravity. 
685: Financial support from the Studienstiftung des Deutschen Volkes and
686: the Max Planck-Institut f\"ur Gravitationsphysik are gratefully
687: acknowledged. 
688: \bibliographystyle{JHEP-2}
689: \bibliography{../paper/lqg2}        
690: \end{document}