1: \documentclass[12pt]{article}
2: \usepackage{a4wide}
3: \usepackage{amssymb}
4: \usepackage{graphicx}
5: %\voffset-1cm
6: \begin{document}
7: {\renewcommand{\thefootnote}{\fnsymbol{footnote}}
8: %\hfill IGPG--yy/m--n\\
9: %\medskip
10: %\hfill gr--qc/yymmnnn\\
11: %\medskip
12: \begin{center}
13: {\LARGE Canonical Relativity\\[2mm] and the Dimensionality of the World\footnote{In: {\em Relativity and the Dimensionality of the
14: World}, Ed.\ V.~Petkov (Springer, 2007), pp.\ 137--152}}\\
15: \vspace{1.5em}
16: Martin Bojowald\\
17: %\footnote{e-mail address: {\tt bojowald@gravity.psu.edu}}\\
18: \vspace{0.5em}
19: Institute for Gravitation and the Cosmos,\\
20: The Pennsylvania State
21: University,\\
22: 104 Davey Lab, University Park, PA 16802, USA\\
23: \vspace{1.5em}
24: \end{center}
25: }
26:
27: \setcounter{footnote}{0}
28:
29: \newcommand{\case}[2]{{\textstyle \frac{#1}{#2}}}
30: \newcommand{\lP}{l_{\mathrm P}}
31:
32: \newcommand{\md}{{\mathrm{d}}}
33:
34: \newcommand*{\R}{{\mathbb R}}
35: \newcommand*{\N}{{\mathbb N}}
36: \newcommand*{\Z}{{\mathbb Z}}
37: \newcommand*{\Q}{{\mathbb Q}}
38: \newcommand*{\C}{{\mathbb C}}
39:
40:
41: \begin{abstract}
42: Different aspects of relativity, mainly in a canonical formulation,
43: relevant for the question ``Is spacetime nothing more than a
44: mathematical space (which describes the evolution in time of the
45: ordinary three-dimensional world) or is it a mathematical model of a
46: real four-dimensional world with time entirely given as the fourth
47: dimension?'' are presented. The availability as well as clarity of the
48: arguments depend on which framework is being used, for which
49: currently special relativity, general relativity and some schemes of
50: quantum gravity are available. Canonical gravity provides means to
51: analyze the field equations as well as observable quantities, the
52: latter even in coordinate independent form. This allows a unique
53: perspective on the question of dimensionality since the space-time
54: manifold does not play a prominent role. After re-introducing a
55: Minkowski background into the formalism, one can see how distinguished
56: coordinates of special relativity arise, where also the nature of time
57: is different from that in the general perspective. Just as it is of
58: advantage to extend special to general relativity, general relativity
59: itself has to be extended to some theory of quantum gravity. This
60: suggests that a final answer has to await a thorough formulation and
61: understanding of a fundamental theory of space-time. Nevertheless, we
62: argue that current insights into quantum gravity do not change the
63: picture of the role of time obtained from general relativity.
64: \end{abstract}
65:
66: \section{Introduction}
67:
68: When faced by the question of whether the world is three- or
69: four-dimensional, the quick answer by a modern-day physicist will most
70: likely be ``four.'' This is indeed what relativity tells us formally
71: where space and time are essentially interchangeable: Lorentz
72: transformations, or their physical manifestations of Lorentz
73: contraction and time dilatation, show that space and time not only
74: play similar roles but can even be transformed into each other. Just
75: as we can rotate a body in three dimensions to observe all its
76: extensions, thereby transforming, e.g., its height in width, we can boost
77: an object\footnote{By ``object'' we will mean a physical system defined by
78: a set of observable properties such that it can be recognized at
79: different occurrences in space and in time. Objects will not be
80: idealized to be point-like or event-like in order to remain unbiased
81: toward the question of dimensionality. Thus, non-vanishing extensions
82: of objects in space as well as time are allowed in order to take into
83: account the necessary unsharpness of measurements needed to verify the
84: defining properties.} so as to, at least to a certain extent, replace
85: space-like by time-like extension and vice versa. The qualification
86: ``to a certain extent'' is necessary because even in relativity space
87: and time are not quite the same but distinguished by the signature of
88: the space-time metric. By itself, this difference in signature is not
89: sufficient reason to deny time the same ontological status as space.
90:
91: There are, however, differences between the usual treatment of space
92: and time in physics going beyond relativity, although they are usually
93: presupposed in relativistic discussions. In order to answer the
94: question of the dimensionality of the world from the viewpoint of
95: relativity, such hidden assumptions have to be uncovered and analyzed,
96: or avoided altogether. Some of these issues lie at the forefront of
97: current physics and still await explanation. For instance, while we
98: can, and have to, limit objects to finite spatial extensions, we have
99: no means to limit their time extensions safe for transformations such
100: as particle decay or other reactions. Even though objects may change
101: in time, they never cease to exist completely. There seems to be a
102: simple reason for that: conservation laws. We simply cannot limit an
103: object's extension in time because, e.g., its energy must be
104: conserved. Thus, the object could be transformed into something else
105: of the same energy but not removed completely. Such laws are derived
106: as consequences of symmetries which first give local conservation laws
107: in terms of currents. Going from local to global conservation laws, as
108: they are required for an explanation of the persistence of objects in
109: time, is a further step and requires additional assumptions. As the
110: following more detailed discussion shows, conservation laws cannot be
111: used to explain the difference of spatial and time-like extensions,
112: for the derivation itself distinguishes space from time.
113:
114: One starts with a local equation such as\footnote{We follow the
115: abstract index notation common in general relativity (see, e.g.,
116: \cite{Wald}). Indices $a,b,\ldots$ refer to space-time while indices
117: $i,j,\ldots$ used later refer only to space. Repeated indices
118: occurring once raised and once lowered are summed over the
119: corresponding range $0,1,2,3$ for space-time indices and $1,2,3$ for
120: space indices. The covariant derivative compatible with a given
121: space-time metric $g_{ab}$ is denoted by $\nabla_a$ which for
122: Minkowski space reduces to the partial derivatives $\partial_a$ in
123: Cartesian coordinates.} $\nabla^aT_{ab}=0$ for the energy-momentum
124: tensor $T_{ab}$. If the space-time metric is sufficiently symmetric
125: and allows a Killing vector field $\xi^a$ satisfying ${\cal
126: L}_{\xi}g_{ab}=\nabla_a\xi_b+\nabla_b\xi_a=0$, the current
127: $j_a=T_{ab}\xi^b$ is conserved: $\nabla^aj_a=0$. At this stage, the
128: only difference between space and time enters through the signature of
129: the metric. A global conservation law is then derived by integrating
130: the local conservation equation over a space-time region bounded by
131: two spatial surfaces $\Sigma_1$ and $\Sigma_2$ and some boundary $B$
132: which could be at infinity. Stokes theorem then shows that the
133: quantity $\int j_a\md S^a$ is the same on $\Sigma_1$ and $\Sigma_2$
134: and thus conserved, {\em provided that all fields fall off
135: sufficiently rapidly toward the boundary $B$}. Thus, one already has
136: to assume physical objects to be of finite spatial extent before
137: obtaining a global conservation law, while there is no such
138: restriction for the time-like extension.\footnote{In this discussion
139: we understood, as usually, that space-time is Minkowski as in special
140: relativity. The energy conservation argument in our context works
141: better if one considers instead a universe model with compact spatial
142: slices, such as an isotropic model with positive spatial curvature, or
143: a compactification of isotropic models of non-positive curvature. This
144: requires one to go beyond special relativity, as we will do later on
145: for other reasons, and to allow non-zero curvature or non-trivial
146: topology. From our perspective, closed universe models are
147: conceptually preferred because space is already finite without
148: boundary such that non-spacelike boundaries are not needed to derive
149: global conservation laws from local ones.} If fields do not vanish at
150: $B$, one interprets $\int_B j_a \md S^a$ as the flux into or out of
151: the spatial region evolving from $\Sigma_1$ to $\Sigma_2$. However,
152: this different interpretation of $\int j_a\md S^a$ as conserved
153: quantity on $\Sigma_1$ and $\Sigma_2$ and as flux on $B$ treats space
154: and time differently, corresponding to the non-relativistic
155: decomposition of energy-momentum in energy and momentum. This
156: different treatment is not implied by the theory but put in by
157: interpreting its objects. The issue of a limited spatial extent versus
158: unconstrainable duration of objects thus remains and has to be faced
159: even before coming to conservation laws.
160:
161: \begin{figure}
162: \begin{center}
163: \includegraphics[height=4cm,keepaspectratio]{Conserve.eps}
164: \caption{Causal diagram of space-time region integrated over to derive
165: global conservation laws.}
166: \end{center}
167: \end{figure}
168:
169: For this reason, it seems to be potentially misleading to consider
170: objects in space-time such as point particles or their worldlines to
171: address the dimensionality of the world, for there are already
172: presuppositions about space and time involved. Indeed, from this
173: perspective a worldline, or the world-region of an extended object,
174: seems inadequate for a relativistic treatment. It would be more
175: appropriate to use only either space-time events or bounded
176: four-dimensional world-regions of extended objects. This already
177: indicates a possible answer to the question of dimensionality:
178: events
179: %\footnote{Space-time events are the quantities for which the
180: %philosophical idea of space and time as {\em principia
181: %individuationis} --- entities which are themselves not physical but
182: %required to individualize physical objects --- most clearly arises.}
183: are zero-dimensional and to be considered as idealizations just as
184: their analogs of point particles. The only option then is to consider
185: bounded world-regions\footnote{Indeed, to analyze an object by
186: whatever means we not only need to capture it at one time --- which is
187: virtually impossible, anyway --- but also hold and observe it for some
188: time. Observations thus always refer to some finite extension in time
189: during which we must be able to recognize the system. A good example
190: can be found in particle physics where too short decay times imply
191: that particles appear rather as resonances without sharp values for
192: all their properties. This is a consequence of uncertainty and thus
193: quantum theory which we will come back to later. Even though common
194: terminology often assigns the object status to an isolated system at a
195: given time, evolving and possibly changing, observations always
196: consider world-regions which could be assigned the object status as
197: well. This non-traditional use of the term ``object'' is probably
198: discouraged because it is too observer-dependent: it is the observer
199: who decides when to end the experiment and thus determines the
200: time-like extension of the space-time region. However, while the
201: classical world allows us to draw sharp spatial boundaries and thus
202: seems to imply individualized spatial objects, this is no longer
203: possible in a quantum theory. Drawing the line around spatial objects
204: is then a matter of choice, too, comparable to limiting the duration
205: of an observation.} as physical objects, which are four-dimensional.
206:
207: It is difficult to follow these lines toward a clear-cut argument for
208: the four-dimensionality of the world due to our limited understanding
209: of the nature of time. An alternative approach is to disregard objects
210: in space-time and rather consider the relativistic physics of
211: space-time itself. For this, we need general relativity which,
212: compared to special relativity, has the added advantage of removing
213: the background structure given by assuming Minkowski space-time. As
214: backgrounds can be misleading, if possible one should consider the
215: more general situation and then see how special situations can be
216: re-obtained.
217:
218: The following sections collect possible ingredients which can be
219: helpful in the context of dimensionality. There are different
220: formulations of general relativity, covariant and canonical ones,
221: which apparently reflect the possible interpretations of a
222: four-dimensional versus a three-dimensional world: While covariant
223: field equations are given on a space-time manifold, the canonical
224: formulation starts with a slicing of space-time in a family of spatial
225: slices. Canonical fields are then defined on space and evolve in time,
226: suggesting a three-dimensional world with an external time parameter
227: \cite{ADM}. Nonetheless, the question of dimensionality cannot be
228: answered easily because, for one thing, the formulations are
229: mathematically equivalent. In what follows, we will mainly use
230: canonical relativity so as to see if it indeed points to a
231: three-dimensional world or, as the covariant formulation, a
232: four-dimensional one. Covariant formulations are also best suited to
233: understand the relativistic kinematics and dynamics. After this
234: general exposition we will specialize the formalism to Minkowski space
235: in order to see which freedom is eliminated in special relativity
236: compared to general relativity and how this can change the picture of
237: time. We end with a brief discussion of dynamical consequences of
238: general relativity as well as some comments on quantum aspects.
239:
240: \section{Canonical Relativity}
241:
242: The signature of the metric also has implications for the form of
243: relativistic field equations on a given space-time which are
244: hyperbolic rather than elliptic. This means that a reasonable set-up
245: for solving these equations is by an initial value problem: for given
246: initial values on space at an initial time one obtains a unique
247: solution. For our purposes, this aspect is not decisive because we
248: could interpret initial values as corresponding to objects placed in
249: space before starting an observation, thus corresponding to a
250: three-dimensional world, or simply as labels to distinguish solutions
251: which themselves play the role of objects of a four-dimensional
252: world.\footnote{In fact, this is clearly brought forward by the
253: canonical formulation where one can either specify states by a phase
254: space given by initial data on the initial surface, or by the
255: so-called covariant phase space consisting of entire solutions to the
256: field equations. In both cases, the phase space is endowed with a
257: symplectic structure, and the formulations are equivalent.} The choice
258: is then just a matter of convenience.
259:
260: \subsection{ADM Formulation}
261:
262: For the field equations of the metric itself the situation is more
263: complicated and crucially different (see \cite{CauchyProblem} for the
264: general relativistic initial value problem). Einstein's equations
265: correspond to ten field equations for the ten components of the
266: space-time metric $g_{ab}$, a symmetric tensor. However, there are
267: only six evolution equations containing time derivatives only of some
268: components while the remaining equations are elliptic and do not
269: contain time derivatives. Although there is no fixed coordinate
270: system, it is meaningful to distinguish between time and space
271: derivatives because, due to the signature of the metric, they are
272: related to vector fields of negative and positive norm squared,
273: respectively. Time evolution is described by an arbitrary timelike
274: vector field $t^a$ while spatial slices of space-time are introduced
275: as level surfaces $\Sigma_t\colon t={\rm const}$ of a time
276: function $t$ such that $t^a\partial_at=1$. The space-time metric
277: $g_{ab}$ induces a spatial metric $h_{ab}(t)$ on each slice $\Sigma_t$
278: as well as covariant spatial derivatives. The spatial metric is most
279: easily expressed as
280: \begin{equation}\label{metric}
281: h_{ab}=g_{ab}+n_an_b
282: \end{equation}
283: where $n_a$ is the unit future-pointing timelike co-normal to a
284: slice. These are only six independent components because $h_{ab}$ is
285: degenerate from the space-time point of view: $n^ah_{ab}=0$. These are
286: also precisely the components of the space-time metric whose time
287: derivatives\footnote{Time derivatives are understood as Lie
288: derivatives with respect to the time evolution vector field $t^a$.}
289: appear in Einstein's equations.
290:
291: \begin{figure}
292: \begin{center}
293: \includegraphics[height=5cm,keepaspectratio]{Split.eps}
294: \caption{Decomposition of the time evolution vector field $t^a$ into
295: the shift vector $N^a$ and a normal contribution $Nn^a$.}
296: \end{center}
297: \end{figure}
298:
299: At this point, we may view the equations as describing the evolution
300: of a three-dimensional quantity $h_{ab}$ in an external time parameter
301: $t$. The remaining four space-time metric components encode the
302: freedom in choosing the time evolution vector field, which can be
303: parameterized as $t^a=Nn^a+N^a$ with components usually called lapse
304: function $N$ and shift vector $N^a$ such that $n_aN^a=0$. They are
305: indeed metric components since (\ref{metric}) implies
306: \begin{eqnarray}
307: g^{ab} &=&-n^an^b+h^{ab}=
308: -\frac{1}{N^2}(t^a-N^a)(t^b-N^b)+h^{ab}\nonumber\\
309: &=& -\frac{1}{N^2}t^at^b+\frac{1}{N^2} (t^aN^b+N^at^b)+
310: h^{ab}-\frac{1}{N^2}N^aN^b \label{invmetric}.
311: \end{eqnarray}
312: The time-time component of the inverse space-time metric is thus
313: $-N^{-2}$ while time-space components are $N^{-2}N^a$. These
314: components enter the field equations, too, but they are not dynamical
315: in the sense that they would have evolution equations determining
316: their time derivatives. In addition to the six evolution equations for
317: $h_{ab}$,
318: \begin{equation}
319: \dot{\pi}^{ab}=f[h_{ab},\pi^{ab},N,N^a]
320: \end{equation}
321: for the momenta $\pi^{ab}[\dot{h}_{cd},N,N^c]$ conjugate to $h_{ab}$,
322: there are then four constraint equations
323: \begin{equation}
324: C[h_{ab},\pi^{ab}]=0 \quad\mbox{and}\quad C^a[h_{bc},\pi^{bc}]=0
325: \end{equation}
326: which are of elliptic nature and restrict the values of the dynamical
327: fields at any spatial slice (independently of $N$ and $N^a$).
328:
329: A possible interpretation is that there are six fields $h_{ab}$ on
330: space which change in time as governed by the evolution equations,
331: depending on four prescribed but arbitrary auxiliary fields $N$ and
332: $N^a$. This would be a mixed viewpoint as far as dimensionality is
333: concerned because $h_{ab}$ would look like spatial objects while $N$
334: and $N^a$ would have to be {\em prescribed} as functions of space and
335: time but are not evolving in time. The system is thus rather
336: four-dimensional since $N$ and $N^a$ have to be functions on a
337: four-dimensional space and determine the evolution of $h_{ab}$ for
338: which only initial values on space are needed. One can save the
339: three-dimensional interpretation by considering $N$ and $N^a$ as
340: external functions for the evolution system of $h_{ab}$, but this has
341: the drawback that there would be no predictivity and no
342: uniqueness of solutions in terms of initial values for the dynamical
343: fields, as solutions also depend on choices of $N$ and $N^a$.
344:
345: Here, the completely four-dimensional view is much more attractive: We
346: not only have to choose the functions $N$ and $N^a$ but can also
347: supplement the dynamical equations for $h_{ab}$ by evolution equations
348: for space-time {\em coordinates}. This is indeed possible, for if we
349: choose four functions $N$ and $N^a$ and ask that they play the
350: role of space-time metric components as they enter the canonical
351: equations (\ref{invmetric}) we have the transformation laws
352: \begin{eqnarray}
353: -\frac{1}{N(t,x^i)^2} &=& q^{bc}(x')\partial'_bt\partial'_ct\\
354: N(t,x^i)^{-2}N^j(t,x^i)&=&q^{cd}(x')\partial'_ct\partial'_dx^j
355: \end{eqnarray}
356: from an arbitrary (inverse) space-time metric $q^{ab}$ to the new
357: functions. These transformations can be interpreted as evolution
358: equations for the space-time coordinates which are then fixed by the
359: choice of $N$ and $N^a$ in terms of coordinates on an initial spatial
360: surface.\footnote{Solutions $x(x')$ depend on the auxiliary metric
361: $q^{ab}$ as well, as they should since no change in coordinates is
362: necessary if $q^{ab}$ had already been of canonical form with the
363: chosen $N$ and $N^a$.} With this interpretation, we obtain, for given
364: initial values, unique solutions to our evolution equations up to
365: changes of coordinates, corresponding to a change in the free
366: functions $N$ and $N^a$.
367:
368: The functions $N$ and $N^a$ which must be defined on a
369: four-dimensional manifold thus determine coordinates $x^a$ such that
370: canonical field equations for $h_{ab}$ result. In general, there is no
371: way to split this globally into a time coordinate $t$ and
372: space-coordinates $x^i$ which one would need for a three-dimensional
373: evolution picture of $h_{ab}$. Thus, a four-dimensional interpretation
374: results. This attractive viewpoint is available only if we take as
375: physical object on which field equations are imposed the entire
376: space-time and not just the metric on spatial slices. We clearly have
377: to consider general relativity which gives field equations for the
378: metric and allows us to perform arbitrary coordinate changes, not just
379: Lorentz transformations. Here, getting rid of the Minkowski background
380: of special relativity, corresponding to a synchronization of rigid
381: clocks and rulers throughout space-time under the assumption of the
382: absence of a gravitational field, is required.
383:
384: \subsection{Relational Observables}
385:
386: Since we obtain a unique solution to the Einstein equations only up to
387: arbitrary changes of space-time coordinates, predictivity requires
388: observable quantities to be coordinate independent, too. While
389: coordinates are usually used in explicit calculations, values of
390: observable quantities must not change when transforming to different
391: coordinates. Abstractly, one can also formulate the concept of an
392: observable in an explicitly coordinate-free manner leading to
393: relational observables: evolution is then measured not with respect to
394: coordinate time but with respect to other geometrical or matter
395: quantities. While this is appealing conceptually, it can be hard to do
396: explicitly.\footnote{``Such a question can, we are assured, always be
397: answered from a sufficient set of initial data, though the performance
398: of this task may call for considerable mathematical agility.''
399: \cite{BergmannTime}.} For instance, in a cosmological situation one can measure
400: how the value of a matter field changes with respect to a change in
401: the total spatial scale or volume. Since in such a picture coordinates
402: are eliminated, an alternative view on the question of dimensionality
403: is possible. It also allows us to show, as we will see, how Minkowski
404: space is recovered and what is special about special relativity.
405:
406: Coordinate changes on a manifold imply transformations for fields such
407: as $g_{ab}$ on that manifold. Observable quantities then must be
408: expressions formed by the fields of a theory being invariant under any
409: change of coordinates. Simple examples are integrals of
410: densities\footnote{A density is a mathematical object transforming in
411: the same manner as $\sqrt{\det g}$ under changes of coordinates such
412: that its coordinate integration is well-defined.} over the whole
413: space-time manifold, but they are too special and do not give one
414: access to local properties. A more general, abstract way of
415: constructing observables is as follows:\footnote{This idea goes back
416: to Komar and Bergmann \cite{KomarObs,BergmannTime} and has more
417: recently been elaborated and used in, e.g.,
418: \cite{GeomObs1,GeomObs2,MiniQuant,PartialCompleteObs,PartialCompleteObsII,DittrichThesis,GPSCoord,EffectiveObs}.}
419: We use the group of transformations of our basic fields corresponding
420: to coordinate changes $x^a\mapsto x'{}^a(x^b)$, which in our case is
421: the group of space-time diffeomorphisms.\footnote{Space-time
422: diffeomorphisms are in general not in one-to-one correspondence with
423: coordinate changes. For our purposes, local considerations are
424: sufficient where this identification can be made. A local coordinate
425: change is then infinitesimally given by $x^a\mapsto x^a+\xi^a(x^b)$
426: where the vector field $\xi^a$ is of compact support, and the same
427: vector field generates a diffeomorphism.} In an explicit realization,
428: group elements would have infinitely many labels corresponding to four
429: functions on space-time, or a space-time vector field $\xi^a(x)$. A
430: relational observable requires one to choose a quantity $f$ to be
431: measured with respect to as many other functionals $\Phi^a_x$ of the
432: basic fields as there are parameters of the group.\footnote{For a
433: specific example, $f$ could be a matter field and $\Phi^a_x$ the
434: spatial volume $\det h_{ab}$ in an isotropic cosmological model. Here,
435: the infinite number of variables $\Phi^a_x$ is reduced to only one by
436: the high degree of symmetry. This corresponds to the fact that only
437: spatially constant time reparameterizations respect the
438: symmetry. Thus, the label $a$ disappears because spatial coordinates
439: cannot be changed in a relevant manner (they can be rescaled in some
440: cases, without affecting the basic fields), and $x$ disappears due to
441: spatially constant reparameterizations. We will come back to possible
442: reductions in the number of independent variables from the
443: counterintuitive infinite size in the following subsection.} These
444: functionals $\Phi^a_x$ will be called internal
445: variables,\footnote{Often, ``internal time'' or ``clock variables''
446: are used in this context, as these quantities are commonly employed to
447: discuss the problem of time. However, since they do not only refer to
448: time and it is even unclear in which sense time is involved, we prefer
449: a neutral term.} for gravity labeled by the space-time index $a$ and
450: a point $x$ in space-time. This corresponds to the freedom in labels
451: of the diffeomorphism group. From $f$ and $\Phi^a_x$ we construct an
452: observable\footnote{The notation, similar to that in
453: \cite{DittrichThesis,PartialCompleteObs}, is quite loaded and
454: indicates that $F[\Phi^a_x]$ is a relational object telling us how $f$
455: changes under changes of the internal variables $\Phi^a_x$. The answer
456: depends parametrically on infinitely many real numbers $\phi^a_x$: for
457: each fixed set of these parameters, $F[\Phi^a_x]_{\phi^a_x}$ gives
458: coordinate independent information on the relational behavior as a
459: functional of the basic field $g_{ab}$.} $F[\Phi^a_x]_{\phi^a_x}$ as a
460: functional of the basic fields parameterized by time values $\phi^a_x$
461: as real numbers: to compute the evaluation
462: $F[\Phi^a_x]_{\phi^a_x}(g_{ab})$ of the observable on a given set of
463: basic fields $g_{ab}$ we first find a coordinate transformation for
464: which $g'_{ab}$ becomes such that $\Phi^a_x(g'_{ab})=\phi^a_x$ equal
465: the chosen time parameters. The value of the observable is then
466: defined to be the original function $f(g'_{ab})$ evaluated in this
467: transformed set of basic fields. For any set of time variables
468: $\phi^a_x$ one obtains a functional of the basic fields $g_{ab}$. This
469: clearly results in an observable independent of the system of
470: coordinates, and is well-defined at least for certain ranges of the
471: fields and parameters involved.\footnote{Global issues, as always in
472: general considerations for general relativity, are much more difficult
473: to handle.}
474:
475: The observable, interpreted as measuring the change of $f$
476: relationally with respect to the internal variables rather than with
477: respect to coordinates does, however, depend on the parameters $\phi^a_x$
478: which crucially enter the construction. One is rather dealing with a
479: family of observables labeled by these parameters. While one obtains
480: an observable for each fixed set of parameters, its interpretation
481: would be complicated and loose any dynamical information of
482: change. This is probably one of the clearest indications for the
483: dimensionality of the world from a mathematical point of view: What we
484: are constructing directly are relational observables depending on
485: parameters $\phi^a_x$, roughly corresponding to a set of
486: worldlines. While this can be restricted to fixed
487: parameters,\footnote{In fact, even though the $\phi^a_x$ are sometimes
488: called ``time parameters,'' only for one value of $a$ does it really
489: correspond to an infinity of times while the remaining parameters are
490: space parameters. Having also space-parameters is actually an
491: advantage in light of our earlier discussion where entire
492: world-regions rather than any kind of lower-dimensional object were
493: preferred. Such a world-region is then spanned by suitable ranges of
494: all the parameters $\phi^a_x$.} it would be a secondary step. Moreover,
495: if all parameters are fixed, also spatial dependence is eliminated; in
496: such a case we end up only with non-local observables. The primary
497: observable quantities are thus not spatial at all but rather give, in
498: an intricate, relational manner, a four-dimensional world.
499:
500: On second thought, there seems to be a problem because we have
501: infinitely many parameters. From special relativity, or any kind
502: of non-relativistic physics, we would expect only one time parameter
503: in addition to three space parameters as independent
504: variables.\footnote{This refers strictly only to one observer. In
505: special relativity one considers time and space coordinates between
506: boosted observers. For a given observer, the synchronization
507: conditions of special relativity imply that there is only one time and
508: three space parameters.} On the other hand, special relativity is
509: obtained from general relativity by introducing a background given by
510: Minkowski space-time. Physically, this corresponds to synchronizing
511: all clocks to measure time (and using a fixed set of rulers to measure
512: lengths). When all clocks are synchronized, there is only one time
513: parameter, and so it is not surprising after all that general
514: relativity, lacking a synchronization procedure, requires infinitely
515: many parameters $\phi^a_x$ for its observable quantities. The
516: mathematical situation is thus in agreement with our physical
517: expectations. We will now make this more explicit by showing how a
518: Minkowski background can be re-introduced.
519:
520: \subsection{Recovering the Minkowski Background}
521:
522: The synchronization procedure can be implemented directly for general
523: relational observables, clearly showing the reduction from infinitely
524: many parameters to only one time coordinate. This brings us to the
525: promised recovery of special relativity by re-introducing the
526: Minkowski background and illustrates the relation between the
527: infinitely many parameters of relational observables and the finite
528: number of coordinates in Minkowski space. We make use of expressions
529: derived recently for general relativity
530: \cite{PartialCompleteObs,PartialCompleteObsII,DittrichThesis}. We
531: assume that four internal field variables $\Phi^a(x)$ have been
532: chosen, having conjugate momenta $\Pi_a$ in a canonical formulation,
533: which in a space-time region we are interested in are monotonic
534: functions of $x^b$. For simplicity, we assume that these variables are
535: four scalar fields which are already present in the theory, rather
536: than more complicated functionals of basic fields such as curvature
537: scalars used in \cite{KomarObs,BergmannTime}. Moreover, we ignore
538: their dynamics, i.e.\ assume that there are no potentials, since our
539: aim here is to reconstruct the non-dynamical Minkowski
540: space-time. Geometrically, the momenta are given by the (density
541: weighted) derivatives of the internal variables along the unit normal
542: to spatial slices,
543: \begin{equation}\label{momentum}
544: \Pi^a=\sqrt{\det h} n^b\partial_b \Phi^a\,.
545: \end{equation}
546: This determines the rate by which the fields change from slice
547: to slice. In a region of monotonic fields, we can thus view
548: $x^a\mapsto \Phi^a(x)$ as a coordinate transformation and transform our
549: metric accordingly, observing (\ref{metric}) and (\ref{momentum}):
550: \begin{equation}
551: g'{}^{ab} = \partial_c\Phi^a\partial_d\Phi^b(h^{cd}-n^cn^d)=
552: \partial_c\Phi^a\partial_d\Phi^bh^{cd}-\Pi^a\Pi^b/\det h
553: \end{equation}
554: or, splitting into time and space components,
555: \begin{eqnarray}
556: g'{}^{00} &=& \partial_i\Phi^0\partial_j\Phi^0 h^{ij}
557: -\Pi^0\Pi^0/\det h \label{transa}\\
558: g'{}^{i0} &=& \partial_j\Phi^0\partial_k\Phi^ih^{jk}-
559: \Pi^i\Pi^0/\det h\label{transb}\\
560: g'{}^{ij} &=& \partial_k\Phi^i\partial_l\Phi^j h^{kl}-
561: \Pi^i\Pi^j/\det h\,.\label{transc}
562: \end{eqnarray}
563:
564: First, we suppress the components $\Phi^i$ to bring out the role of
565: time which will be played by $\Phi^0$. We thus assume that the spatial
566: metric $h^{ij}$ is already given by $\delta^{ij}$ as in Minkowski
567: space in its standard coordinate representation. Under the remaining
568: transformation corresponding to $\Phi^0$, the original spatial metric
569: $h^{ij}$ is transformed to the new spatial metric $g'{}^{ij}$ which to
570: preserve Minkowski space should also equal $\delta^{ij}$. Since we
571: suppressed the spatial parameters $\Phi^i$, we need to require
572: $\Pi^i=0$ such that the spatial coordinate system is fixed in
573: time.\footnote{Thus, our set of rulers does not change in time.} Time
574: synchronization then implies that $\Phi^0$ does not depend on spatial
575: coordinates, so also $g'{}^{i0}=0$ is of Minkowski form. For the final
576: component of the metric $g'{}^{ab}$ we obtain $g'{}^{00}=-(\Pi^0)^2$
577: which is of Minkowski form for $\Pi^0=1$. These conditions can be
578: summarized by saying that spatial coordinates do not change in time
579: ($\Pi^i=0$), and time progresses at the same constant pace everywhere
580: ($\Pi^0=1$).
581:
582: For instance from the construction of relational observables in
583: \cite{PartialCompleteObs} it follows that with such a choice of clock
584: variables a relational observable takes the form
585: \begin{equation} \label{ObsFixed}
586: F[\Phi^a_x]_{\phi^0}=\sum_{k=0}^{\infty}\frac{1}{k!} \dot{f}(\Phi^i)
587: (\phi^0-\Phi^0)^k= f(\Phi^i,\phi^0-\Phi^0)
588: \end{equation}
589: where the dot refers to the change in $f$ under a change of the time
590: field $\Phi^0$. If we identify space-time coordinates with $\Phi^a$, any
591: function will be observable since the background is completely fixed
592: for a given observer.
593:
594: This refers to one observer who has performed a full
595: synchronization. If we change the observer, we obtain the usual
596: Lorentz transformations and between two different observers time does
597: certainly not proceed at the same pace. To see this, we now allow all
598: four functions $\Phi^a$ to be non-trivial. We want to describe a
599: situation where one synchronized observer is given by a system of the
600: form just derived, such that $h^{ij}=\delta^{ij}$, whose internal
601: variables we now call $\Psi^a$. From there, we transform to a new
602: system of internal variables $\Phi^a(\Psi^b)$ such that also the
603: metric $g'{}^{ab}$ is Minkowski for the new synchronized
604: observer. Thus, the right hand sides of (\ref{transa}),
605: (\ref{transb}), (\ref{transc}) must be $\Psi^a$-independent and only
606: linear functions $\Phi^a$ are allowed:
607: \begin{eqnarray}
608: \Phi^i &=& \omega^i_j \Psi^j+\alpha^i \Psi^0\\
609: \Phi^0 &=& \beta_i\Psi^i+\gamma \Psi^0\,.
610: \end{eqnarray}
611: Derivatives in (\ref{transa}), (\ref{transb}), (\ref{transc}) are now
612: taken with respect to $\psi^i$, and
613: $\Pi^a=\partial\Phi^a/\partial\Psi^0$. From (\ref{transa}) we then
614: obtain $g'{}^{00}=-1=\beta_i\beta_j
615: \delta^{ij}-\gamma^2$ such that
616: \begin{equation} \label{gamma}
617: \gamma=\sqrt{1+|\beta|^2}
618: \end{equation}
619: where $|\cdot|$ denotes the norm of vectors in $h_{ij}=\delta_{ij}$. From
620: Eq.~(\ref{transb}) in the form
621: \[
622: g'{}^{i0} = \partial_j\Phi^0\partial_k\Phi^i\delta^{jk}- \Pi^i\Pi^0
623: \]
624: we have $0=\beta_j\omega^i_k\delta^{jk}-\alpha^i\gamma$ such that
625: \begin{equation} \label{alpha}
626: \alpha^i=\frac{\beta_j\omega^i_k \delta^{jk}}{\gamma}\,.
627: \end{equation}
628: Finally, Eq.~(\ref{transc}) in the form
629: \[
630: g'{}^{ij} = \partial_k\Phi^i\partial_l\Phi^j\delta^{kl}-\Pi^i\Pi^j
631: \]
632: implies
633: \begin{equation}\label{rot}
634: \delta^{ij}= \omega^i_k\omega^j_l\delta^{kl}-\alpha^i\alpha^j\,.
635: \end{equation}
636: Defining
637: \begin{equation}
638: \rho^i_j:=\omega^i_j-\frac{\alpha^i\beta_j}{1+\gamma}\,,
639: \end{equation}
640: for which we have
641: \begin{eqnarray*}
642: \rho^i_k\rho^j_l\delta^{kl} &=& \omega^i_k\omega^j_l\delta^{kl}-
643: \frac{\omega^i_k\alpha^j\beta^k}{1+\gamma}-
644: \frac{\omega^j_k\alpha^i\beta^k}{1+\gamma}+
645: \frac{\alpha^i\beta_k\alpha^j\beta^k}{(1+\gamma)^2}\\
646: &=& \omega^i_k\omega^j_l\delta^{kl}- \alpha^i\alpha^j
647: \end{eqnarray*}
648: using (\ref{gamma}) and (\ref{alpha}), shows that the freedom in
649: $\omega^i_j$ is given by an orthogonal matrix $\rho^i_j$. Thus, only
650: a vector $\beta^i$ and a rotation $\rho^i_j$ can be chosen freely to
651: specify a transformation. The remaining coefficients $\alpha^i$ and
652: $\gamma$ are then fixed by (\ref{alpha}) and (\ref{gamma}). This is
653: easily recognized as the usual coefficients of Lorentz transformations
654: if we only identify $\beta^i=v^i/\sqrt{1-v^2/c^2}$ and use
655: $\rho^i_j$ as the rotational part of the transformation.
656:
657: Allowing different synchronized observers, observable functions as in
658: (\ref{ObsFixed}) have to be Lorentz invariant and are not arbitrary.
659: Completely arbitrary, non-synchronized observers then require the
660: general relativistic situation with complicated relational expressions
661: for observables.
662:
663: From our perspective, this shows that the usual space and time
664: parameters one has in special relativity are what is left after fixing
665: all but finitely many of the infinitely many parameters $\phi^a_x$. These
666: infinitely many parameters occur automatically when one attempts to
667: write observables in a relational manner. In general, none of these
668: parameters is distinguished as a possible time parameter to describe
669: the evolution of a three-dimensional world. In the relational picture,
670: thus, only the four-dimensional option is available.
671:
672: \section{Challenges and Resolutions}
673:
674: The canonical structure of relativity and an analysis of what is
675: observable thus gives good reasons for the four-dimensionality of the
676: world. Some difficulties certainly remain because, for one thing, we
677: considered only local regions and had to assume that we can find
678: functions $\Phi^a_x$ which are monotonic there. In order to describe the
679: whole space-time in this manner we would need globally monotonic
680: functions which may be difficult to find in general. For strictly
681: physical purposes such a global description is also an
682: over-idealization because all observations we can ever make are
683: restricted to some bounded region of space-time, however big this
684: region may be in cosmological observations. There are more severe
685: potential challenges to this picture, one resulting from properties of
686: general relativity not considered so far, and the other resulting from
687: quantum theory.
688:
689: \subsection{Singularities}
690:
691: Locally, solutions to Einstein's field equations always exist and
692: determine the space-time metric as well as manifold. This played a
693: crucial role in our arguments given so far where we wanted to
694: eliminate backgrounds and consider dynamical space-time. These
695: equations are, however, non-linear and so global aspects are more
696: difficult to control. One consequence is that most solutions which we
697: think are relevant for what we observe are singular when extrapolated
698: in general relativity. They allow one to describe space-time only for
699: a finite amount of proper time for some, and in some cases all,
700: observers after which the classical theory breaks down
701: \cite{SingTheo}. This is usually accompanied by a divergence of
702: curvature, but in any case represents a finite boundary to space-time.
703:
704: If the theory does not allow us, even in principle, to extend
705: solutions arbitrarily far in one direction, it may be difficult to
706: view this direction as a dimension of the world. Here, the
707: three-dimensional viewpoint seems more suitable because we would
708: simply have to deal with space and objects in time, described by the
709: theory for some finite range of time. To be sure, there are also
710: solutions where space is finite, but even if there are such boundaries
711: space-time can usually be extended and they are thus
712: artificial.\footnote{There can also be boundaries to space arising
713: from singularities where space-time cannot be extended in spatial
714: directions. Such time-like singularities, however, do not generically
715: arise in relevant cosmological or black hole solutions and thus can be
716: ignored here. In homogeneous cosmological models, from which most of
717: the cosmological intuition is derived, such time-like singularities
718: are ruled out by the assumption of homogeneity (be it a precise or
719: approximate symmetry) while for black holes time-like singularities
720: arise for negative mass where the singular behavior is even welcome to
721: rule out negative mass and argue for the stability of Minkowski
722: space. Other black hole solutions where time-like singularities arise,
723: such as the Reissner--Nordstrom solution for electrically charged
724: black holes in vacuum, are unstable to the addition of matter. Generic
725: singularities are then space-like or null
726: \cite{SpacelikeSing,NullSing}.} This is not the case with
727: singularities. If we are interested in a four-dimensional
728: interpretation, then, we will have to deal with fundamental
729: limitations to the extension of four-dimensional objects, including
730: space-time itself.
731:
732: \subsection{Quantum Aspects}
733:
734: Just as it was helpful to embed special relativity into general
735: relativity for a wider viewpoint, the classical description is itself
736: incomplete not the least because it leads to space-time
737: singularities. This requires a corresponding extension of general
738: relativity to a quantum theory of gravity. But even before this stage
739: is reached, quantum properties do have a bearing on some of the
740: arguments that can be used to decide on the dimensionality of the
741: world. For instance, the four-dimensional interpretation is
742: advantageous because it embodies the fact that we have to recognize an
743: object in order to denote it as such, showing that the time extension
744: plays a central role in assigning object status. Such a recognition is
745: not possible in quantum mechanics where identical particles are
746: indistinguishable. We can then never be sure that a particle we
747: recognize is the same one we saw before, and so assigning object
748: status to worldlines or world-regions would not make sense unless all
749: identical particles are subsumed in one and only one object.
750:
751: \subsection{Resolutions}
752:
753: These puzzles are resolved easily if one just considers suitable
754: combinations of quantum theory with special and general relativity,
755: respectively. Combining quantum theory with special relativity leads
756: to quantum field theory where indeed the particle concept is weakened
757: compared to the classical or quantum mechanical picture. There is not
758: a collection of individual but indistinguishable particles, but a
759: field whose excitations may in some cases be interpreted as
760: particles. Thus indeed, one is treating all identical particles as one
761: single object, the corresponding field, and any problem with
762: recognizability is removed automatically. The field is a function on
763: space-time or a world-region, a four-dimensional object.\footnote{In
764: algebraic quantum field theory one considers algebras associated with
765: world-regions of ``diamond'' shape as the basic objects, so also here
766: it is bounded regions in space-time determining what objects in the
767: theory are.}
768:
769: Singularities of general relativity pose a more complicated problem,
770: but there are indications that they, too, are automatically dealt with
771: when the underlying classical theory, this time general relativity, is
772: combined with quantum theory. While the classical space-time picture
773: breaks down at a singularity, several recent investigations have shown
774: that quantum geometry continues to be well-defined, albeit in a
775: discrete manner
776: \cite{Sing,HomCosmo,Spin,BHInt,ModestoConn,SphSymmSing}. One can then
777: extend the classical space-time through a quantum region, or view
778: space-time as fundamentally described by a quantum theory of gravity
779: which reduces to general relativity in certain limits when curvature
780: is not too large.
781:
782: Indeed, background independent versions of quantum gravity are not
783: formulated on a space-time manifold such that the question of whether
784: the three- or four-dimensional viewpoint should be taken does not
785: really arise at all. One is either dealing with space-time objects
786: directly, such as in discrete path integral approaches, or employs a
787: canonical quantization where the central object is a wave function on
788: the space of geometries and observables are relational as discussed
789: before. In quantum gravity, the four-dimensional, relational viewpoint
790: is thus even more natural than in classical gravity. It is also
791: crucial for the results on non-singular behavior which are based on
792: the relational behavior of wave functions or other quantities which
793: now play the role of basic objects. Extensions beyond classical
794: singularities can then be provided by considering the range of
795: suitable internal variables and their quantizations: The relational
796: dependence can, and in all cases studied so far will, continue through
797: stages where one would classically encounter a singularity. This is
798: much more robust than looking at possible modifications of field
799: equations and corresponding extensions of space-times in coordinate
800: form which have turned out to be non-generic if available at all.
801:
802: \section{Conclusions}
803:
804: The question of whether a theoretical object is just a mathematical
805: construct or a real physical thing is always difficult to address in
806: physics. Often, the answer depends on what theory is used,
807: which itself depends on current available knowledge. Not just the
808: theoretical structure needs to be understood well but also its
809: ontological underpinning. This is notoriously difficult if space and
810: time are involved, and often hidden assumptions already enter
811: constructions.
812:
813: In such a situation, it is best to make use of as flexible a framework
814: as possible and to eliminate any background structure. Thus, we
815: focused on general relativistic dynamics rather than special
816: relativistic kinematics. We have highlighted some relevant
817: consequences using a canonical formulation. Canonical formulations are
818: often perceived as not being preferable because they break manifest
819: covariance. However, they also offer well structured mathematical
820: formulations and can be particularly illuminating for the dynamical
821: behavior.
822:
823: In particular, canonical techniques allow, or even require one to discuss
824: observables in a coordinate free manner. This leads to a relational
825: description where no coordinates are used but instead field values are
826: related to values of other fields to retrieve observable
827: information. Usually, these quantities take the form of families of
828: functionals parameterized by real numbers (most generally, infinitely
829: many ones). In contrast to coordinates, these parameters do not
830: distinguish between space and time and even the signature of a
831: space-time metric is irrelevant. This formulation is then of the most
832: democratic form and removes the danger of being misled by the
833: different forms of space and time coordinates. A difference between
834: those parameters arises in special situations such as when a Minkowski
835: background is re-introduced. This illustrates, again, that background
836: structures are to be eliminated as far as possible.
837:
838: In addition to this extension from special to general relativity it is
839: believed that a further one is necessary to combine it with quantum
840: theory. A theory of quantum gravity in a reliable and completely
841: convincing form is not yet available, but from what we know it does
842: not seem to change much of the arguments presented here. It can even
843: eliminate potential problems such as that of singularities. At a
844: kinematical level, one can still imagine different possibilities
845: concerning the dimensionality\footnote{Background independent
846: formulations seem to agree on a lower dimensional kinematical nature
847: on microscopic scales
848: \cite{RS:Spinnet,ALMMT,SurfaceSum,LowDimDynTriag,FractalAsSafe}.} but
849: one still has the full parameter families corresponding to a
850: four-dimensional world when it comes to observables.
851:
852: There are also conceptual advantages of a four-dimensional
853: understanding. If the world and its objects are four-dimensional, they
854: are simply there and do not need to become. There is then no need to
855: explain their origin, eliminating a difficult physical and
856: philosophical question.\footnote{Also for this question, quantum
857: gravity or cosmology seems to be affirmative: Initial conditions for
858: quantum cosmological solutions, which have traditionally been imposed
859: by intuitively motivated choices \cite{tunneling,nobound}, can arise
860: directly from the dynamical laws \cite{DynIn,Essay}. Thus, although
861: completely unique scenarios are difficult to construct,
862: four-dimensional dynamics can automatically select solutions and to
863: some degree eliminate additional physical input to formulate an
864: origin.}
865:
866: \bibliographystyle{../preprint}
867: \bibliography{../Bib/QuantGra}
868:
869: \end{document}
870: