gr-qc0108043/pr.tex
1: %Topology
2: %
3: % by Janna Levin
4: %
5: 
6: \documentstyle[11pt,epsf]{article}
7: \input psfig.sty
8: 
9: \def\double{\baselineskip 24pt \lineskip 10pt} 
10:  
11: 
12: \textheight 8.5in
13: \textwidth 6in
14: \oddsidemargin 0pt
15: \topmargin -30pt
16: 
17: \newcommand{\be}{\begin{equation}}
18: \newcommand{\ee}{\end{equation}}
19: \newcommand{\bea}{\begin{eqnarray}}
20: \newcommand{\eea}{\end{eqnarray}}
21: \newcommand{\ba}{\begin{eqnarray}}
22: \newcommand{\ea}{\end{eqnarray}}
23: \def\cb{{\sc cobe--dmr\,}}
24: \def\ah{{\bf H}^3}
25: \def\es{{\bf S}^3}
26: \def\eu{{\bf E}^3}
27: \def\cq{C(\hat n, \hat n^\prime)}
28: \def\spc{{\xi_{\Phi}}}
29: \def\vx{{\bf \vec x}}
30: \def\vxp{{\bf {\vec x}^\prime}}
31: \def\pow{{{\cal P}_\Phi}}
32: \def\hn{\hat n}
33: \def\hnp{\hat n^\prime}
34: \def\gta{\mathrel{
35: \hbox to 0pt{\lower
36: 3pt\hbox{$\mathchar"218$}\hss}\raise 2.0pt\hbox{$\mathchar"13E$}}}
37: 
38: \def\lta{
39: \mathrel{\hbox to
40: 0pt{\lower 3pt\hbox{$\mathchar"218$}\hss}\raise 2.0pt\hbox{$\mathchar"13C$}}}
41: 
42: 
43: \def\double{\baselineskip 24pt \lineskip 10pt}
44: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
45: 
46: \begin{document}
47: 
48: \begin{titlepage}
49: 
50: \begin{flushright}
51: %gr-qc/\\
52: %DAMTP
53: \end{flushright}
54: 
55: \vspace{.5in}
56: 
57: \begin{center}
58: \huge
59: {\bf Topology and the Cosmic Microwave Background}
60: 
61: \vspace{.5in}
62: %\setcounter{footnote}{0}
63: 
64: \large{
65: Janna Levin}
66: 
67: \normalsize
68: \vspace{.2in}
69: 
70: {\em DAMTP, CMS, Cambridge University,  \\ 
71: Wilberforce Rd, Cambridge CB3 0WA, U.K.\\
72: j.levin@damtp.cam.ac.uk \\
73: and\\
74: Theoretical Physics Group, Imperial College, London}
75: 
76: 
77: \vspace{.5in}
78: 
79: \begin{abstract}
80: 
81: Nature abhors an infinity.  The limits of general 
82: relativity are often signaled by infinities:
83: infinite curvature as in the center of a black hole, the infinite energy
84: of the singular big bang.  We might be inclined to add an infinite universe 
85: to the list of intolerable infinities.  Theories that move beyond general
86: relativity naturally treat space as finite.  In this review we discuss
87: the mathematics of finite spaces and our aspirations to observe the finite
88: extent of the universe in the cosmic background radiation.
89: 
90: 
91: 
92: \end{abstract}
93: 
94: \end{center}
95: 
96: 
97: 
98: \end{titlepage}
99: 
100: 
101: 
102: 
103: \tableofcontents
104: 
105: \newpage
106: 
107: \section{Introduction}
108: \label{introduction}
109: \setcounter{equation}{0}
110: 
111: \def\theequation{\thesection.\arabic{equation}}
112: 
113: Is the universe infinite?  The universe had a birth and will
114: have a death, whether from old age
115: or a more catastrophic demise through
116: a big crunch.  With the advent 
117: of general relativity,  cosmology has accepted space as an
118: evolving geometry.  Although the universe is believed to
119: begin and end a dynamical life, we continue to ascribe 
120: a property to the universe that we would never assign to any other
121: natural object, and that is the property of being spatially infinite.
122: 
123: The universe is home to a plethora of structures 
124: all of which must be finite.  It would be
125: untenable to suggest that any physical structure be infinite, yet 
126: cosmologists often cavalierly assume space itself is infinite.
127: Yet an infinite universe means an infinite number of stars and 
128: galaxies and people.  By the 
129: law of probability an infinite universe
130: accommodates an infinite number of events each happening
131: an infinite number of times \cite{{lum},{css_cqg}}.  Somewhere else
132: in the cosmos, you are there.  In fact there are an infinite number
133: of  you littering space.
134: An infinite number of us living the same lives and
135: an infinite number living slightly different lives and all range of
136: variants.
137: While this concept of infinity is manageable to some, it
138: reeks of something pathological to others.
139: Is the universe a surreal object, unlike any of its inhabitants, or is it 
140: real and physical and limited and finite?
141: 
142: If the universe is finite, then light will wrap around the finite 
143: space decorating the sky with ghost images, sometimes called clone images.
144: In principle we should be able to see copies of ourselves at different
145: ages in different directions.  A kaleidescopic version of `This is your
146: life'.  In practice, the universe does not appear to be
147: so small.  The universe,
148: if finite, must at least be big.  Big enough to house clusters of galaxies
149: and perhaps even bigger than the entire light travel time since the big
150: bang.  Therein lies our question.  If the universe is finite, how small
151: is it, and how do we measure its shape?
152: 
153: The question of the infinite extent of space is often answered
154: dynamically by Einstein's theory. 
155: If the 
156: amount of matter and energy 
157: is underdense then the universe will expand forever, 
158: if overdense then the universe will ultimately
159: recollapse to a big crunch, and if critical then
160: the universe will expand forever.
161: To each of these dynamical possibilities corresponds a geometry 
162: of spacetime:
163: the underdense cosmos is negatively curved and infinite,
164: the overdense cosmos is positively curved and finite 
165: and the critical cosmos is flat and infinite.
166: 
167: Each of these three geometries can support topologies with finite volumes
168: without altering the dynamics or the curvature.  Multiconnected
169: topologies are often overlooked in favor of the simply connected
170: possibilities.  This oversight is encouraged
171: by a limitation of Einstein's theory which elevates gravity to 
172: a theory of geometry but does not provide a theory of topology.
173: If unification of gravity with the other forces ever succeeds,
174: topology will inevitably be integrated as a predictive feature of 
175: any cosmology.  The topology of additional compact dimensions is
176: already understood to be a crucial feature in superstring theories
177: and, in general, 
178: discretization of space into topologically finite bundles is a natural
179: consequence of quantization \cite{lum}.  No one is good enough at 
180: quantum gravity or string theory 
181: to make a prediction for the geometry and shape of 
182: the whole universe.  Still the requirement of extra topologically 
183: compact dimensions has become a commonplace notion in fundamental physics
184: even if it seems esoteric in astronomy.
185: The first suggestion of compact extra dimensions began with Kaluza back
186: in Einstein's day and have been dressed up in the modern guise of 
187: superstring theories since.  Even if we can never see the extent of the
188: large three dimensions, we may be able to look for small extra dimensions
189: to find clues about the geometry of the cosmos.
190: 
191: 
192: In the absence of a predictive theory we can 
193: still explore the mathematical possibilities for a finite universe.
194: This review covers
195: topologies consistent with
196: the standard cosmological 
197: geometries 
198: and  
199: our aspirations to observe
200: the global extent of space in patterns in the oldest historical 
201: relic accessible, the cosmic microwave background (CMB).
202: Methods to determine topology from the CMB range from the geometric such
203: as the search for circles
204: in the sky \cite{{css1},{css2},{css_cqg}}and topological pattern formation 
205: \cite{{conf},{pat},{imogen}}
206: to the brute force statistical methods
207: such as the method of images \cite{{bpsI},{bpsII}}. These methods will
208: be explained in turn.
209: This review is intended both for mathematicians interested in cosmology
210: and cosmologists interested in the mathematics of topology.
211: Not much will be taken for granted and we will take the time
212: to explain even the standard cosmology.
213: 
214: 
215: 
216: Topology
217: can be observed with the CMB or with the distribution of collapsed objects
218: such as quasars or clusters of galaxies.
219: There are reviews on galactic methods
220: \cite{{galreviews1},{galreviews2}}.  
221: To keep the scope of this article manageable we
222: will limit our discussion to searches for artifacts of topology 
223: in the CMB and
224: defer a discussion of galactic methods to the reviews 
225: in Ref.\ \cite{{galreviews1},{galreviews2},{roukreview}}.
226: (see also Refs. \cite{{ghosts1},{ghosts2}}.)
227: Also, much of the
228: history on topology 
229: can be found in an earlier review \cite{lum}.
230: Other accessible reviews can be found in Refs.
231: \cite{{thursweeks},{glennpop}}.
232: Collections of papers for two international workshops can be found in
233: \cite{cqg} and \cite{ctp98}.
234: The observational section in this review will focus on recent
235: methods only, starting where Ref.\ \cite{lum} left off.
236: 
237: \newpage
238: 
239: 
240: \section{Topology and Geometry}
241: \label{mathsection}
242: 
243: %\setcounter{equation}{0}
244: 
245: \def\theequation{\thesection.\arabic{equation}}
246: \label{topintro}
247: 
248: Cosmic topology aims 
249: to deduce the global shape of the universe
250: by experimentally observing
251: a pattern in the distribution of astronomical objects.
252: To do so, it is helpful to first understand the possible topologies from
253: a purely mathematical perspective.  The isotropy of the cosmic
254: microwave background radiation implies that the curvature of the observable
255: universe is very nearly constant, so in the present article we consider
256: only manifolds of constant curvature.  The homogeneity of the observable
257: universe does not, of course, exclude the possibility that the curvature
258: of space varies enormously on a global scale beyond our view, but because
259: that hypothesis is presently untestable we do not pursue it here.
260: By the same reasoning, we are most interested in multiconnected spaces that are
261: small enough to witness, although 
262: theories beyond general relativity, such as string
263: theory, may help us to push beyond those limitations 
264: (see \S \ref{extradsect}).
265: There is a previous Physics Reports on cosmic topology which provides
266: an excellent overview of the mathematical principles
267: \cite{lum}.  The reader is encouraged to consult that article
268: as well as Refs.\ \cite{{wolf},{ellis}}.  We will not repeat these
269: detailed reviews but do survey the same topological methods for 
270: completeness and extend the discussion to include some additional
271: modern methods.
272: 
273: The global shape of any space, including the ultimate outerspace,
274: is characterized by a {\it geometry} and a {\it topology}.
275: The term {\it geometry}, as used by mathematicians,
276: describes the local curves while {\it topology} describes global features which
277: are unaltered by smooth deformations.  
278: A continuous transformation is known as a {\it homeomorphism}, a smooth
279: continuous and invertible map which deforms one manifold into another
280: without cutting or tearing.  General relativity is not invariant
281: under homeomorphisms but is invariant under 
282: {\it diffeomorphisms}, which amount to a change of coordinates.
283: It is this covariance of classical gravity which underlies
284: the principles of general relativity.  Covariance ensures
285: that all observers experience the same laws of physics regardless of 
286: the worldline along which they travel and therefore regardless of the
287: coordinate system within which they interpret the world.  
288: Relativity does not invoke such a principle with respect to topology.
289: For this reason we do not have a fundamental principle to guide us when 
290: contemplating the topology of the universe.  We might hope that
291: such a principle or at least a prediction might precipitate
292: from a complete theory of gravity
293: beyond classical.  In the meantime, we consider any mathematically
294: allowed topology, restricting ourselves to the constant curvature
295: spaces as indicated by observational cosmology.
296: 
297: \subsection{Overview of Geometry}
298: \label{ovgeom}
299: 
300: A cosmological principle asserts that the Earth should not be
301: special in location or perspective.  Cosmological 
302: observations on the largest scales indicate we live in a 
303: space with the symmetries of {\it homogeneity}, space is the same in 
304: all directions, and {\it isotropy}, space looks the same in all directions.
305: Homogeneous and isotropic manifolds have constant curvature.  The
306: $n$-dimensional geometries of constant curvature are the everywhere
307: spherical 
308: ${\bf S}^n$,
309: flat ${\bf E}^n$, and hyperbolic 
310: ${\bf H}^n$ geometries (fig.\ \ref{N.2.1.1}).  
311: One natural consequence of curved geometries worth remembering is that
312: the sum of the interior angles of a triangle are greater than 
313: $\pi$ on ${\bf S}^n$, exactly $\pi$ on ${\bf E}^n$, and less than $\pi$
314: on ${\bf H}^n$.  
315: 
316: \begin{figure}
317: \centerline{\psfig{file=N.2.1.1.ps,angle=-90,width=4.25in}}
318: \vskip 5truept
319: \caption{A homogeneous isotropic space may be spherical (positive curvature),
320: flat (zero curvature), or hyperbolic (negative curvature).}
321: \label{N.2.1.1}
322: \end{figure}
323: 
324: In $2D$, all Riemannian surfaces are homeomorphic to the
325: three constant curvature geometries ${\bf S}^2, {\bf E}^2$,
326: and ${\bf H}^2$.
327: The infinite hyperbolic plane ${\bf H}^2$ cannot be drawn properly
328: in $3D$ and so is difficult to visualize.  It can be drawn schematically
329: as a surface which everywhere has the curvature of a saddle.  A nice
330: way to treat ${\bf H}^2$ is as a pseudosphere embedded in $(2+1)$-Minkowski
331: space.  The pseudosphere has the equation 
332: $-z^2+x^2+y^2=-1$.  The full isometry group of the hyperboloid is
333: $PSL(2,{\bf R})\equiv SL(2,{\bf R})/Z_2$ with 
334: $SL(2,{\bf R})$ the special Lorentz group of real 
335: $2\times 2$ matrices with unit determinant.  The embedding in Minkowski
336: space makes the symmetry group of 
337: ${\bf H}^2$ more obvious.
338: 
339: The sphere ${\bf S}^2$ can be embedded in 
340: Euclidean 3-space
341: as a surface with radius $x^2+y^2+z^2=1$.  The isometry group for the sphere
342: is $O(3)$, all orthogonal $3\times 3$ matrices with the absolute value
343: of the determinant equal to one.
344: 
345: The plane ${\bf E}^2$ is the infinite surface $(x,y)$ with
346: the Minkowski time coordinate $z$ fixed at unity
347: (see fig.\ \ref{embed}).  The Euclidean plane has the
348: full Galilean group of isometries:  translations, reflections, rotations
349: and glide reflections.
350: A glide reflection involves a translation with 
351: a reflection along the line parallel to the translation.
352: 
353: In $3D$, all manifolds are not homeomorphic to the constant
354: curvature manifolds.  Instead there are $8$ homogeneous geometries some
355: of which are anisotropic.  These $8$ geometries were classified by Thurston
356: in the mathematical literature \cite{{thurclass},{Thurston}}.  Cosmologists
357: are more familiar with the equivalent classification of 
358: Bianchi into $8$  homogeneous manifolds \cite{Bianchi}.
359: Out of respect for the cosmological principle we will consider the
360: fully homogeneous and isotropic spaces of constant curvature 
361: ${\es},{\eu}$, and ${\ah}$.  Similar to $2D$, ${\ah}$ can be embedded 
362: as a pseudosphere in $(3+1)$-Minkowski space, $\es$ as the sphere and
363: $\eu $ as the plane at fixed Minkowski time.
364: This isometry group 
365: of $\es$ is $SO(4)$.
366: The isometry group of
367: $\eu $ 
368: %is $R^3\times SO(3)$, which 
369: is the product of translations and
370: the special orthogonal $3\times 3$ matrices.  
371: The isometry group of $\ah$ is
372: $PSL(2,{\bf C})\equiv SL(2,{\bf C})/Z_2$.
373: 
374: \subsection{Overview of Topology}
375: \label{ovtop}
376: 
377: Topology is the pursuit
378: of equivalence classes of spaces; that is, the classification of 
379: homeomorphic manifolds.  To a topologist, all surfaces with one
380: handle for instance are equivalent.  Thus the doughnut and the coffee cup,
381: in addition to providing a dubious breakfast, provide the quintessential
382: example of equivalent topologies, despite their obviously different
383: geometries (i.e. curves).
384: 
385: 
386: \begin{figure}
387: \centerline{\psfig{file=N.1.1.1.ps,angle=-90,width=1.75in}}
388: \vskip 5truept
389: \caption{The flat 2-dimensional torus is defined by abstractly
390: gluing opposite edges of a square.}
391: \label{N.1.1.1}
392: \end{figure}
393: 
394: The doughnut and coffee cup are both manifestations of
395: the 2-dimensional torus (fig.\ \ref{N.1.1.1}), 
396: a prototypical multiply connected
397: space.  The torus has a finite area, yet has no boundary.  When an inhabitant
398: of the torus looks forward (fig.\ \ref{N.1.1.2}), her line of sight wraps around
399: the space and she sees herself from behind.  She has the illusion of seeing
400: another copy of herself directly in front of her (fig.\ \ref{N.1.1.3}).  
401: Indeed she
402: also sees herself when she looks to the side, or along a 45 degree line,
403: or along any line of rational slope.  She thus has the illusion of seeing
404: infinitely many copies of herself (fig.\ \ref{N.1.1.4}).  In other words, even
405: though
406: her universe is finite, she has the illusion of living in an infinite
407: universe containing an infinite lattice of repeating objects.
408: 
409: 
410: \begin{figure}
411: \centerline{\psfig{file=N.1.1.2.ps,angle=-90,width=1.75in}}
412: \vskip 5truept
413: \caption{An inhabitant of the 2-torus looks forward and sees
414: herself from behind.}
415: \label{N.1.1.2}
416: \end{figure}
417: 
418: 
419: \begin{figure}
420: \centerline{\psfig{file=N.1.1.3.ps,angle=-90,width=1.75in}}
421: \vskip 5truept
422: \caption{The inhabitant of the 2-torus has the illusion of seeing
423: a copy of herself.}
424: \label{N.1.1.3}
425: \end{figure}
426: 
427: 
428: \begin{figure}
429: \centerline{\psfig{file=N.1.1.4.ps,angle=-90,width=1.75in}}
430: \vskip 5truept
431: \caption{Indeed, the inhabitant of the 2-torus sees a lattice
432: of images of herself.}
433: \label{N.1.1.4}
434: \end{figure}
435: 
436: 
437: As described below, all multiconnected surfaces can be built by gluing
438: the edges of a fundamental polygon.  The torus is built by identifying
439: opposite edges of a parallelogram.
440: If we begin
441: with a flat rectangle, we can bend this flat 
442: sheet into three-dimensions to glue the top to the bottom and the left
443: edge to the right.  In doing so we have made a torus of revolution
444: as in fig.\ \ref{torus}
445: which is neither flat nor constant curvature.  The curvature
446: of the torus clearly varies over the surface.  This bending of the
447: torus is an artifact of embedding the surface in $3$-dimensions.  A truly
448: flat torus, better known as $T^2=S^1\times S^1$, is content to live in 
449: $2D$ with no such bending and projecting into $3D$.  The flat torus
450: is akin to the video game with periodic boundary conditions where
451: the left edge is identified with the right edge so that a flat explorer
452: could stick their hand out the right edge only to have it appear poking
453: out of the left edge.  The edge in fact is as meaningless to the inhabitant
454: of a flat torus as it is to an inhabitant of the torus of revolution.
455: $T^2$ is truly flat and so has a different geometry from the torus of 
456: revolution although they are topologically equivalent, being characterized
457: by the one handle.
458: The genus $g$ of the manifold is a topological invariant which counts 
459: the number of handles and holes.  The torus has genus $g=1$.
460: 
461: 
462: \begin{figure}
463: \centerline{\psfig{file=torus.ps,width=1.75in}}
464: \vskip 5truept
465: \caption{The torus of revolution has the same topology as the flat torus
466: but different curvature.}
467: \label{torus}
468: \end{figure}
469: 
470: The 2-dimensional Klein bottle (fig.\ \ref{N.1.1.5}) 
471: is similar to the torus, only
472: now the top and bottom of the 
473: rectangle are glued not to each other, but each to itself with
474: a shift of half a unit.  
475: The traditional way to make a Klein bottle is
476: %Another way to make the Klein bottle is 
477: to start with the
478: rectangle, glue top to bottom but glue the left to the right after a 
479: rotation through $\pi$.
480: In the present article we use a different (but equivalent) 
481: construction (namely gluing the top and bottom each to itself with
482: a shift of half a unit)
483: because it will simplify the construction of flat 3-manifolds
484: in section \ref{threed}.
485: As in the torus an inhabitant has the illusion
486: of seeing an infinite lattice of objects, only now the structure of the
487: lattice is different:  it contains glide reflections as well as
488: translations.  
489: The truly flat Klein bottle is topologically equivalent to the
490: bottle immersed in $3D$ and pictured in fig.\ \ref{klein}.  The embedding again
491: leads to a curved surface and in this case to a self-intersection of the
492: surface.  The flat Klein bottle, content to live in $2D$, has no such curvature
493: or self-intersection.
494: 
495: 
496: \begin{figure}
497: \centerline{\psfig{file=N.1.1.5a.ps,angle=-90,width=1.75in}\quad\quad
498: \psfig{file=N.1.1.5b.ps,angle=-90,width=2.5in}}
499: \vskip 5truept
500: \caption{Left: The Klein bottle is similar to the torus, only the square's
501: top and bottom edges are each glued to themselves, not to each other.
502: The lattice of images in the Klein bottle is different from
503: that in the torus because it contains glide reflections as well as
504: translations.}
505: \label{N.1.1.5}
506: \end{figure}
507: 
508: 
509: \begin{figure}
510: %\centerline{\psfig{file=../../bk/pics/klein.ps,angle=-90,width=1.75in}}
511: \centerline{\psfig{file=klein.ps,width=1.75in}}
512: \vskip 5truept
513: \caption{A surface with the topology of the Klein bottle.}
514: \label{klein}
515: \end{figure}
516: 
517: These examples generalize readily to three dimensions.  A cube (fig.\
518: \ref{cube}) with opposite
519: faces glued is a 3-dimensional torus, or 3-torus.
520: Its inhabitant has the illusion of seeing an infinite 3-dimensional
521: lattice of images of himself and of every other other object in the space.
522: Varying the gluings of the faces 
523: varies the structure of the lattice of images.
524: Cosmic topology aims to observe these distinctive lattice-like patterns.
525: 
526: 
527: 
528: \begin{figure}
529: \centerline{\psfig{file=cube.eps,width=1.75in}}
530: \vskip 5truept
531: \caption{3-dimensional lattice of cubes.}
532: \label{cube}
533: \end{figure}
534: 
535: \subsection{Tools of Topology}
536: \label{tools}
537: 
538: In this preliminary discussion several important concepts have already
539: begun to emerge.  The first is the fundamental domain.
540: A {\it fundamental domain} is a polygon or polyhedron from which a manifold
541: may be constructed.  For example, fig.\ \ref{N.1.1.1} shows a square fundamental
542: domain for a 2-torus, and fig.\ \ref{cube} 
543: shows a cube fundamental domain
544: for a 3-torus.
545: A {\it Dirichlet domain} is a special type of fundamental domain.
546: To construct a Dirichlet domain, pick an arbitrary point in the manifold
547: of interest to serve as a {\it basepoint}.  Start inflating a balloon
548: with center at the basepoint (fig.\ \ref{N.1.2.1}) and let the balloon expand
549: uniformly.  Eventually different parts of the balloon will bump into
550: each other.  When this happens, let them press flat against each other,
551: forming a flat (totally geodesic) boundary wall.  Eventually the balloon
552: will fill the whole manifold, at which point it will have the form
553: of a polygon in two dimensions or polyhedron in three dimensions
554: whose faces are the aforementioned boundary walls.  This polyhedron is
555: the Dirichlet domain.  Gluing corresponding faces recovers the original
556: manifold.  
557: Notice that the fundamental rectangular domain for the Klein bottle
558: when glued after a shift of half a unit is not Dirichlet, but when glued
559: after a rotation through $\pi$ is Dirichlet.
560: In some manifolds the Dirichlet domain depends on the choice
561: of basepoint while in others it does not -- more on this later.
562: 
563: 
564: \begin{figure}
565: \centerline{\psfig{file=N.1.2.1.ps,angle=-90,width=3.in}}
566: \vskip 5truept
567: \caption{Start inflating a balloon, with its center point fixed, and let
568: it expand until it fills the entire space.  The resulting polygon
569: (in two dimensions) or polyhedron (in three dimensions) is called
570: a Dirichlet domain.  In some manifolds the shape of the Dirichlet domain
571: depends on the choice of the center point;  in other manifolds it does not.
572: }
573: \label{N.1.2.1}
574: \end{figure}
575: 
576: By identifying the edges of the fundamental domain, a multiconnected space
577: is constructed.
578: Figures \ref{N.1.1.4} and \ref{N.1.1.5}
579: show how an inhabitant of a finite multiply
580: connected space may have the illusion of living in an infinite simply
581: connected space.  This apparent space is called the {\it universal
582: cover} and can be thought of as the simply connected manifold with
583: the constant curvature geometry.  The fundamental domain is cut from this cloth
584: and glued together while preserving the geometry of the universal cover.
585: So the flat torus preserves the geometry of the universal cover while the
586: torus of revolution does not.
587: 
588: 
589: The group of {\it covering transformations} is the group of symmetries
590: of the universal cover.  That is, it is the group of motions of the universal
591: cover that take images of a given object to other images of the same object.
592: For example, in the case of a torus the group of covering transformations is
593: the obvious lattice of translations.  In the case of the Klein bottle the
594: group
595: of covering transformations contains glide reflections as well as
596: translations.
597: 
598: In two-dimensions all topologies have been completely 
599: classified.  We give that classification in \S \ref{twod}.
600: However, in
601: three-dimensions the classification of hyperbolic manifolds remains 
602: incomplete although huge advances were made last century.
603: There is a collection of topological invariants which can be used
604: to recognize the equivalence classes of spaces.  
605: We have already encountered the {\it genus} of the manifold which
606: counts handles.
607: A group structure can 
608: be used to capture handles and thereby detect
609: the connectedness of the space.  The most important
610: of these is the loop group also known as the
611: {\it fundamental group} or the first {\it homotopy group}.  
612: All loops which can be smoothly deformed into each other
613: are called {\it homotopic}.
614: A loop 
615: drawn on the surface of a sphere can be contracted to a point.  In 
616: fact all loops drawn on the surface of a sphere can be contracted a 
617: point.  Since all loops are homotopic, the sphere is 
618: {\it simply connected}
619: and in fact does not have any handles.  If however we draw a loop
620: around the torus of revolution as in fig.\ \ref{torusloops}, 
621: there is no way to smoothly
622: contract that loop to a point, similarly for a loop drawn around the
623: circumference of the toroid
624: also shown in fig.\ \ref{torusloops} .  
625: The torus is therefore {\it multiply connected}.
626: The group of loops
627: is the fundamental group $\pi_1({\cal M})$ and is a topological invariant.
628: In more than $2D$, the fundamental group is not sufficient to determine
629: the connectedness of the manifold.  Nonetheless, the fundamental group
630: is a powerful tool.
631: P\" oincar\'e has conjectured that
632: if the fundamental group is trivial, then
633: the $n$-manifold is topologically equivalent to $S^n$.
634: 
635: 
636: \begin{figure}
637: %\centerline{\psfig{file=../../bk/pics/torusloops.ps,width=2.in}}
638: \centerline{\psfig{file=torusloops.ps,width=2.in}}
639: \vskip 5truept
640: \caption{Two homotopically distinct loops around the torus of revolution.}
641: \label{torusloops}
642: \end{figure}
643: 
644: A closely related structure is the
645: {\it holonomy group}. 
646: The holonomy group can be thought of as the set of instructions
647: for identifying the faces of the fundamental domain.  
648: The holonomies act discretely and without fixed point.
649: The holonomy group
650: is 
651: a subset of the full
652: isometry group of the covering space.  
653: The full symmetries of the universal cover are broken by
654: the identifications and it is customary to represent
655: the compact manifold
656: as a quotient space $G/\Gamma $ where $G$ is the
657: isometry group of the universal cover and $\Gamma $ is the holonomy 
658: group.  
659: The holonomy group is of extreme utility in the game of cosmic
660: topology.  The group provides the set of rules for distributing
661: images throughout the universe and thereby provides
662: the mathematical means by which to determine the geometric distribution
663: of astronomical images in the cosmos.  The group is synonymous with
664: the boundary conditions for the manifold and as we will see this is
665: critical for building a prediction for
666: the spectrum of the cosmic microwave background.
667: 
668: 
669: There can be many different representations for the holonomy group.
670: There is often a simplest presentation $\{\gamma_1...\gamma_{n};r\}$ 
671: where $\gamma_i \in \Gamma$
672: are elements of the group.  Implicitly 
673: $r=1$ following the semicolon and 
674: is a list of all of the relations among the group elements.
675: The most natural presentation for cosmology 
676: has a group element, and so a letter, 
677: for every identification rule.
678: Such a presentation
679: has  many
680: letters and many relations.  
681: 
682: 
683: The holonomies 
684: can map the end-point of an orbit to the start of that orbit so
685: that it becomes periodic in the compact space.
686: In other words, a periodic orbit has
687: $x_{\rm end}=\gamma x_{\rm start}$ where $\gamma $ can be a composite
688: word
689: $\gamma=\prod^n_i\gamma_{k_i}$.  
690: Each word has a corresponding periodic orbit.  
691: %These orbits are not necessarily
692: %unique due to redundancy from the relations which will collapse 
693: %some long words to shorter words.
694: 
695: \begin{figure}
696: %\centerline{\psfig{file=../../top/sr/embed2.eps,width=2.in}}
697: \centerline{\psfig{file=embed2.eps,width=2.in}}
698: \vskip 5truept
699: \caption{The embedding of ${\bf E}^2$ into $(2+1)$-Minkowski
700: space.  The manifold appears
701: as an infinite sheet fixed at $\tau_M=1$.}
702: \label{embed}
703: \end{figure}
704: 
705: As an explicit example we return to the canonical torus.
706: We can make use of the embedding of ${\bf E}^2$ in a $(2+1)$-dimensional 
707: space.
708: Specifically, the $2$-dimensional coordinate
709: is replaced with the $(2+1)$-dimensional coordinate
710: 	\be
711: 	x^a=\pmatrix{\tau_M \cr x \cr y}
712: 	\ee
713: where $\tau_M$ is fixed at unity as in fig. \ref{embed}.  
714: In this coordinate system
715: the holonomy group for the torus is the set of 
716: generators, 
717: 	\be
718: 	T_x=
719: 	\pmatrix{1 & 0 & 0  \cr
720: 	  L_x & 1 & 0\cr
721: 	0 & 0 & 1 },
722: \quad \quad
723: 	T_y=
724: 	\pmatrix{1 & 0 & 0  \cr
725: 	 0&1 & 0 \cr
726: 	L_y &  0  & 1}.
727: 	\ee
728: The boundary condition is then $x^a\rightarrow T^a_b x^b$ 
729: so that identification
730: of the left edge to the right edge results in
731: 	\be
732: 	\pmatrix{\tau_M \cr x\cr y\cr}
733: 	\rightarrow 	\pmatrix{1 & 0 & 0  \cr
734: 	  L_x & 1 & 0\cr
735: 	0 & 0 & 1 }	\pmatrix{\tau_M \cr x\cr y\cr}=
736: 	\pmatrix{\tau_M \cr x+L_x\cr y\cr}
737: 	\label{leftright}
738: 	\ee
739: The holonomy group for the torus in this presentation is
740: $\{T_x,T_y;T_xT_yT_x^{-1}T_y^{-1}\}$.  The relation
741: $T_xT_yT_x^{-1}T_y^{-1}=1$ is simply the commutivity of the
742: generators and prunes the number of unique periodic orbits
743: and therefore the number of clone or ghost images.  Because
744: the generators commute, the periodic orbits (homotopic loops)
745: can be counted symbolically
746: in terms of the number of windings taken around $x$ and around $y$,
747: $(m_x,m_y)$.  
748: The number of loops and therefore the number of
749: clone images grows as a polynomial with length.
750: %in the next section (\S \ref{tessellations}).
751: For example, consider the periodic orbit of fig.\ \ref{fund} 
752: which has $x_{\rm end}=T_yT_x^2x_{\rm start}$
753: and so winding number $(m_x=2,m_y=1)$.
754: 
755: 
756: \begin{figure}
757: %\centerline{\psfig{file=../../top/sr/fund0.eps,width=2.in}}
758: \centerline{\psfig{file=fund0.eps,width=2.in}}
759: \vskip 5truept
760: \caption{A 
761: periodic orbit on the torus
762: which corresponds to 
763: $x_{\rm end}=T_yT_x^2x_{\rm start}$.
764: }
765: \label{fund}
766: \end{figure}
767: 
768: The non-orientable Klein bottle
769: first twists the $z$-faces through $\pi$ before 
770: identification.  The generators of $\Gamma$ for the twisted space
771: are $T_x,R(\pi)T_y$ with $R(\theta)$ the rotation matrix:
772: 	\be
773: 	R(\theta)=\pmatrix{1 & 0 & 0 \cr
774: 	0 & \cos \theta & \sin \theta  \cr
775: 	0 & -\sin \theta & \cos \theta}
776: 	\  \ .
777: 	\ee
778: All of the multiconnected, flat topologies can be built out of a
779: combination of rotations and these translations.
780: 
781: In hyperbolic space, the generators cannot easily be written in functional
782: form but can nonetheless be found with numerical entries in the matrix.
783: The program
784: SnapPea provides the generators explicitly in this embedded coordinate
785: system 
786: for a census of known hyperbolic manifolds and orbifolds
787: \cite{snappea}.
788: (Orbifolds can have singular points.)
789: The number of periodic orbits and therefore of ghost images grows 
790: exponentially in a compact hyperbolic manifold.  The exponential proliferation
791: of images is a manifestation of the chaotic flows supported on compact 
792: hyperbolic spaces and discussed in \S \ref{chaossect}.
793: 
794: 
795: \subsection{Tessellations}
796: \label{tessellations}
797: 
798: The modest tools we have introduced go a long way in constructing
799: cosmological models.  To summarize we will characterize a manifold by:\par
800: 	$\bullet $  The {\bf fundamental domain}
801: -- the shape of space around a given basepoint.\par
802: 	$\bullet $  The {\bf universal cover} -- the simply connected
803: manifold with the same constant curvature geometry.\par
804: 	$\bullet $  The {\bf holonomy group} -- 
805: face pairing isometries, i.e.,
806: the rules for identifying the faces
807: of the fundamental domain.\par
808: 
809: Armed with these tools, we can begin to visualize life in a finite universe.
810: An important visualization technique comes from a tessellation of the
811: universal cover.  Consider the torus again.
812: The flow of an observer or of light can be followed through the compact
813: flat space by conscientiously respecting the boundary conditions and drawing
814: the motion within the fundamental polygon as in fig.\ \ref{fund}.  
815: The Dirichlet domain is the set of points
816: 	\be \{y\in {\cal M}\}\nonumber
817: 	\ee
818: such that 
819: 	\be d(y,x)\le d(y,\gamma x), \forall
820: 	\gamma \in \Gamma \ \  .\label{fd}\ee
821: In $T^2$, light
822: travels along straight lines as do all inertial observers.
823: Another faithful representation of $T^2$ which does not distort the everywhere
824: flat nature of the geometry is provided by a tiling picture.  
825: Beginning with the fundamental polygon, the periodic boundary condition
826: which identifies the left edge of the rectangle to the right edge for
827: instance can be represented by gluing an identical copy of the fundamental
828: polygon left edge to right edge.  
829: This amounts to moving an entire copy of the fundamental domain with
830: $T_x(FD)$.
831: Similarly identical copies can be
832: glued top to bottom 
833: $T_y(FD)$
834: ad infinitum until the entire infinite flat plane
835: is tiled with identical copies of the rectangle.  
836: In general, the generators of the holonomy group can be thought of as defining
837: an alphabet of $2n$ letters, for $n$ generators and their $n$
838: inverses.
839: The alphabet will vary with different representations, as will the grammar.
840: Usually, though not always, the longer the word
841: built out of the letters $\gamma_i\in \Gamma$,
842: the more distant the corresponding image or tiling
843: in the tessellation picture.  
844: Figures \ref{N.1.1.4}, \ref{N.1.1.5}, and \ref{N.1.2.1} 
845: demonstrate tessellations.
846: %If there are
847: %$n$ generators and their inverses,
848: %the number of words $N$ grows as
849: %$(2 n-r)^k<N(k)< 2n^k$ where the number of relations prunes
850: %the allowed number of words.
851: 
852: 
853: These copies are truly
854: identical.  If someone stands in the center of one rectangle
855: he will see images of himself standing at the center of every other
856: rectangle.  If he moves to the left he will see every copy of himself
857: move left.  Because of the finite light travel time, it will take longer
858: to see the more distant images movings.  The light from these more distant
859: images takes a longer path around the compact space before
860: reaching the observer, the source of the very image. 
861: This creates the pattern of images already described so that the universe
862: is an intricate hall of mirrors.
863: 
864: Notice that the tiling completely fills the plane without overlaps or gaps.
865: We can also imagine tiling the plane with a perfect square or even a 
866: parallelogram without overlaps or gaps.  A flat torus can therefore be 
867: made by identifying the edges of these other polygons.  All 
868: have the same topology and the same curvature but some have different
869: metrical features in the sense that the rectangle is truly longer than it is
870: wide.  An observer could line rulers up in space and unambiguously
871: measure that their space was longer than it was wide while an observer
872: in the perfect square would know that their space was equally long and wide.
873: The torus is thus globally anisotropic
874: even though the geometry
875: of the metric is still locally isotropic.  
876: The torus is still globally homogeneous.  
877: The center could be moved arbitrarily and the
878: fundamental polyhedron redrawn around that center to give an equally valid
879: tiling. Most topologies globally break both homogeneity and isotropy.
880: 
881: The holonomies provide the list of rules for identifying the edges of the 
882: polygon.  So the Klein bottle, while it can begin with the same fundamental
883: polygon as the torus and has the same universal cover, it has a different set 
884: of rules for identifying the edges and therefore a different holonomy group.
885: The tiling will be correspondingly different with identical copies 
886: glued top edge to bottom edge but left edge to right only after
887: rotating the tile by $\pi$.  So the tessellation creates the reflected 
888: and translated images specific to the Klein bottle (fig.\ \ref{N.1.1.5}).
889: 
890: \begin{figure}
891: %\centerline{\psfig{file=../../top/sr/tile.eps,angle=-90,width=2.in}}
892: \centerline{\psfig{file=tile.eps,angle=-90,width=2.in}}
893: \vskip 5truept
894: \caption{
895: The compact hypertorus can be represented as an identified
896: parallelepiped.  Alternatively, the compact topology can be 
897: represented
898: by tiling flat space 
899: with identical copies of the fundamental 
900: parallelepiped.  In the tiling picture above
901: only the $(x,y)$ directions are shown.
902: A particular
903: periodic orbit is drawn which corresponds
904: to 
905: $x_{\rm end}=T_yT_x^2x_{\rm start}$.
906: }
907: \label{tile}
908: \end{figure}
909: 
910: 
911: The flat plane can also be tiled by hexagons 
912: (fig.\ \ref{N.1.2.1}) as many a bathroom floor
913: demonstrates.  The hexagon therefore also reproduces a flat torus.
914: However the
915: general parallelogram and hexagon
916: exhaust the possible convex polygons which can
917: tessellate flat space.  If we tried to fill the plane with octagons,
918: we would find that they overlapped at a vertex and could not smoothly
919: fill the plane, as in fig.\ \ref{oct}.  If we glue the octagons together without
920: confining them to the plane, they will create a curling surface 
921: that tries to have a hyperbolic structure.  In fact, the
922: $2D$ hyperbolic plane ${\bf H}^2$ can be tiled by regular octagons. 
923: Recall that the angles of a triangle sum
924: to less than $\pi$ in hyperbolic geometry. 
925: Since the 
926: interior angles of the polygon narrow on the hyperbolic plane, they
927: can be drawn just the right size relative to the curvature scale 
928: so that precisely the right number fit around a vertex to fill the 
929: negatively curved plane without gaps or overlaps.
930: Notice also, that there is no scale in flat space and there is no 
931: intrinsic meaning to the area of the compact torus.  It can be made arbitrarily
932: large or small relative to any unit of measure.  This is not the case when
933: the surface is endowed with a curved metric.  The area is then
934: precisely determined by the topology.  The area of the octagon has to 
935: be precise relative to the curvature scale to properly tessellate the universal
936: cover.  
937: 
938: \begin{figure}
939: \centerline{\psfig{file=oct.eps,width=2.5in}}
940: \vskip 5truept
941: \caption{}
942: \label{oct}
943: \end{figure}
944: 
945: %\subsubsection{Lit of flat 2D spaces}
946: 
947: The topology created by the compact octagon is a double torus
948: $T_2$ where the $2$ refers to the number of handles
949: or genus, g, of the manifold.  Hyperbolic $T_2$ is topologically equivalent
950: to the double doughnut of fig.\ \ref{double} which lives
951: in $R^3$ although again $T_2$ has a different
952: geometry, being flat, while the double doughnut is clearly curved.
953: 
954: 
955: \begin{figure}
956: %\centerline{\psfig{file=../../bk/pics/doubletorii.ps,width=2.5in}}
957: \centerline{\psfig{file=doubletorii.eps,width=2.5in}}
958: \vskip 5truept
959: \caption{}
960: \label{double}
961: \end{figure}
962: 
963: A connected sum of tori can be constructed to make $T_g$,
964: a flat $2D$ compact surface with $g$ handles.  In order to make 
965: a $g$ handled object, a polygon with $4g$ edges is needed.  
966: The polygons
967: have to be drawn larger relative to the curvature scale the more
968: edges they have in order
969: that the interior angles be thin enough 
970: to fit together around a vertex and tile $H^2$.  Consequently
971: a relationship emerges between the area of the surface and the topology.
972: The area of the hyperbolic surface can be related to the 
973: Poincar\'e-Euler characteristic, $\chi=2(1-g)$ through the 
974: Gauss-Bonnet theorem
975: 	\be 
976: 	A=- 2\pi\chi=4\pi(g-1).
977: 	\ee
978: Although the area is a topological invariant in $2D$, the lengths need
979: not be fixed.  For instance, the asymmetric octagon and the
980: regular octagon have the same topology, namely that of the double 
981: doughnut, and the same negative curvature,
982: but they have a different spectrum of periodic orbits.
983: 
984: The tessellation of the universal cover is of particular importance when
985: we consider three-dimensional spaces.  We simply do not have four dimensions
986: available to us in which to bend the three-dimensional volume and that
987: visualization technique fails us.  However, we can still visualize
988: filling the three-dimensional volume with $3D$ tiles which preserve
989: the local geometry of the manifold.  Therefore we rely heavily on the 
990: universal cover, the fundamental polyhedron and the holonomy group
991: when investigating the topologies of these compact spaces.
992: 
993: \subsection{Hyperbolic topologies and Volumes}
994: 
995: All 2-dimensional manifolds have been classified, as have
996: flat and spherical 3-manifolds, but hyperbolic 3-manifolds
997: are not yet fully understood.
998: These continue to resist classification although enormous advances were
999: made last century.
1000: An important realization was the rigid connection between metrical quantities
1001: and topological features in hyperbolic three-manifolds.
1002: Metrical quantities describe lengths and would ordinarily fall under the 
1003: domain of geometry.  Yet in three-dimensions the Mostow
1004: \cite{mp} rigidity
1005: theorem ensures that once the topology is specified, then all metrical 
1006: quantities
1007: are fixed for $3$-D hyperbolic manifolds.  This means that not only the volume
1008: but also the lengths of the shortest geodesics are immutable for a given
1009: topology.
1010: This is to be compared with $2D$ where a given topology can support an
1011: infinite number of metrics as already described.  
1012: 
1013: Although the topology fixes the compact hyperbolic
1014: volume, the volume does not uniquely
1015: specify the topology.  In other words there can be different topologies
1016: with the same volume.
1017: Because of this powerful link between geometry and topology, compact
1018: hyperbolic manifolds can be ordered according to their volume.
1019: A remarkable result is that compact hyperbolic manifolds form 
1020: a countably infinite number of countably infinite sequences ordered
1021: according to volume.  A given sequence shows an accumulation of
1022: manifolds near a limiting volume set by a cusped manifold.  A cusped 
1023: manifold has finite volume but is not compact, having 
1024: infinitely long cusped corners which taper infinitely thin. 
1025: The cusps are topologically a $2$-torus which is conformally shrunk to 
1026: zero down the narrowing throat.  The sequence of compact manifolds
1027: are built systematically through the process of Dehn surgery which is used
1028: in the next section to organize the spectrum of topologies.  Dehn 
1029: surgery is a formal process whereby a torus is drilled out of the cusped 
1030: manifold and replaced with a solid torus.
1031: The surgery is taken along a periodic orbit of $T^2$ identified
1032: by the number of windings $(m_x,m_y)$ the orbit takes around 
1033: the torus before closure.
1034: The program SnapPea names manifolds according to the number of 
1035: windings so that the manifold m003(3,-1) is a compact space constructed by 
1036: Dehn filling the orbifold m003 along the periodic orbit with 
1037: winding numbers (3,-1).
1038: 
1039: The space m003(3,-1) is also known as the Weeks space after the mathematician 
1040: who discovered it and coincidentally has contributed to this article
1041: \cite{weekspace}.
1042: The Weeks space is of particular relevance to cosmology since it is 
1043: the smallest known manifold with a volume 
1044: ${\cal V}\sim 0.94$ in curvature units.  The Weeks space displaced 
1045: the Thurston space m003(-2,3) which has volume ${\cal V}\sim 
1046: 0.98$.
1047: Although the Weeks space is the smallest known manifold it may or may not be
1048: the smallest possible.  The current bound on the minimum volume
1049: for any compact hyperbolic manifold is about 0.3 in units of the
1050: curvature radius.
1051: Contrast this with the
1052: flat hypertorus which can be made infinitesimally small.
1053: Again, the rigidity of compact hyperbolic spaces is at work.
1054: 
1055: Because of the rigidity of $3D$ hyperbolic manifolds, we can also 
1056: characterize a topology by certain volume measures.  Very useful quantities
1057: in cosmological investigations are the {\it in-radius} $r_-$ and the
1058: {\it out-radius} $r_+$.  The in-radius is the radius of the largest geodesic
1059: ball centered at the origin 
1060: which can be inscribed within the Dirichlet domain. 
1061: The out-radius is the radius of the smallest geodesic ball
1062: centered at the origin which can
1063: encompass the Dirichlet domain.  
1064: The disparity between $r_{-}$ and $_+$ can be taken as a rough indication
1065: of the global anisotropy introduced by the topological identification.
1066: The {\it injectivity radius} is the radius of the largest ball,
1067: centered
1068: at the given point, whose interior embeds the manifold.
1069: %half the length of the shortest geodesic
1070: %at a given point of the manifold.  
1071: The Dirichlet domain given by default in SnapPea is centered at a local 
1072: maximum of the injectivity radius.  In this way it often provides the
1073: most symmetric form for the Dirichlet domain.
1074: Moving the observer away from this basepoint will change the appearance of the
1075: global shape of space observationally.
1076: A somewhat ambiguous but sometimes useful quantity is an estimate 
1077: of the 
1078: diameter of the manifold.  Since the manifold is not isotropic
1079: this is not precisely defined.  Nonetheless, one estimation is given by 
1080: 	\be
1081: 	d_{\cal M}=sup_{x,y\in {\cal M}}d(x,y);
1082: 	\ee
1083: that is, it is the largest distance between any two points in the Dirichlet
1084: domain.  
1085: Notice that it is not a geodesic distance so no periodic 
1086: orbit lies along the diameter.  
1087: A more well defined diameter is the maximum
1088: distance between two points in the manifold, where the distance is
1089: measured
1090: in the manifold itself, not in the Dirichlet domain:
1091: 	\be
1092: 	diam=sup_{x,y} (\inf_{{\rm paths \ connecting}\ x\ {\rm to}\ y}
1093: ({\rm length\ of\ path})).
1094: 	\ee
1095: 
1096: The periodic geodesics are quite important and are closely related
1097: both to the eigenmodes in the space and the distribution of ghost
1098: or clone images.  The periodic geodesics can be found with the
1099: holonomy group.  The eigenvectors of the holonomy group span a plane
1100: in the $4D$ embedding which intersects the pseudosphere $\ah$.
1101: The line of intersection occurs along the periodic orbit corresponding
1102: to that group element.  
1103: A composite element can be thought of as a word formed from 
1104: the fundamental alphabet of the presentation.
1105: The fact that the number of words grows exponentially
1106: with length means an exponential growth of long periodic orbits.
1107: The proliferation of periodic orbits, dense in phase space, are the
1108: canonical mark of chaotic dynamics.
1109: This leads us to remark that compact hyperbolic spaces are not just
1110: complicated mathematically but they are truly complex, supporting
1111: chaotic flows of geodesics.  Since most cosmological investigations
1112: try to circumvent the chaotic nature of the motions, we mention
1113: chaos only occasionally as we go along.  We reserve a final section for a 
1114: brief discussion  of this fascinating feature.  A review of 
1115: chaos on the pseudosphere can be found in Ref.\ \cite{bv}.
1116: 
1117: 
1118: 
1119: \newpage
1120: \section{Survey of Compact Manifolds}
1121: \label{overview}
1122: 
1123: 
1124: Having described the topological principles at work and some
1125: examples, we can compile a list of known manifolds, their geometries
1126: and topologies.  The classification is clearly presented in Refs.\
1127: \cite{{lum},{wolf},{ellis}}.  We present a different mathematical
1128: approach for variety.  The classification is organized by 
1129: J. Weeks \cite{jwp} in relation
1130: to four
1131: questions:
1132: 
1133: 
1134: $1$. What constant curvature manifolds exist?\par
1135: 
1136: $2$. How flexible is each manifold?  That is, to what extend can we change
1137: a manifold's shape without changing its topology or violating
1138: the constant curvature requirement?\par
1139: 
1140: $3$. What symmetries does each manifold have?\par
1141: 
1142: $4$. Does a Dirichlet domain for the manifold 
1143: depend on the 
1144: choice of basepoint?\par
1145: 
1146: It turns out that the answers to these questions depend strongly on both
1147: the curvature and the dimension.  The remainder of \S \ref{overview}
1148: answers these
1149: questions in each of the six possible cases:  that is, in spherical, flat,
1150: or hyperbolic geometry in two dimensions (\S \ref{twod}) or three dimensions
1151: (\S \ref{threed}).
1152: 
1153: 
1154: 
1155: %\subsection{Definitions}
1156: %\label{definitions}
1157: 
1158: \subsection{Two-dimensional manifolds}
1159: \label{twod}
1160: 
1161: \subsubsection{Overview of two-dimensional manifolds}
1162: \label{twodover}
1163: 
1164: 
1165: %Another way to classify manifolds is through the process of 
1166: %Dehn surgery.  Since our need for ever more sophisticated
1167: %mathematical technologies is growing we describe the process
1168: %of Dehn surgery here.
1169: 
1170: Every closed surface
1171: may be obtained by starting with a sphere and adding handles 
1172: (fig.\ \ref{N.2.1.2})
1173: or crosscaps (fig.\ \ref{N.2.1.3}).  
1174: Each handle is constructed by removing two disks 
1175: from the sphere and gluing
1176: the resulting boundary circles to each other (fig.\ \ref{N.2.1.2}), and each
1177: crosscap
1178: is constructed by removing one disk and gluing the resulting boundary circle
1179: to itself (fig.\ \ref{N.2.1.3}).
1180: Conway found a particularly simple and elegant proof
1181: of this classification;  for an illustrated exposition see 
1182: Ref.\ \cite{Francis-Weeks}.
1183: 
1184: 
1185: \begin{figure}
1186: \centerline{\psfig{file=N.2.1.2a.ps,angle=-90,width=1.75in}
1187: \psfig{file=N.2.1.2b.ps,angle=-90,width=1.75in}}
1188: \vskip 5truept
1189: \caption{To add a handle to a surface, remove two disks and glue the
1190: remaining boundary circles to each other.}
1191: \label{N.2.1.2}
1192: \end{figure}
1193: 
1194: The interaction between geometry and topology is a primary theme in our
1195: understanding of 2-manifolds.
1196: The promised interaction between geometry and topology is that each closed
1197: topological surface may be given a constant curvature geometry.
1198: Each closed 2-manifold admits a unique geometry.  That is, a surface
1199: which admits a spherical geometry cannot also admit a flat geometry, and so
1200: on.
1201: For an elementary proof see Ref.\ \cite{Shape}.  Infinite 2-manifolds,
1202: by contrast, may admit more than one geometry.  For example, the Euclidean
1203: plane and the hyperbolic plane are topologically the same 2-manifold.
1204: We organize the surfaces according to the total
1205: number of disks that get removed during the construction 
1206: (fig.\ \ref{N.2.1.4}).
1207: %[JANNA:  FIGURE. N.2.1.4 IS A TABLE WHOSE ROWS ARE THE NUMBER OF DISKS
1208: %REMOVED (0, 1, 2,...) AND WHOSE TWO COLUMNS ARE "HANDLES" AND "CROSSCAPS".
1209: %THE ENTRIES IN THE TABLE ARE THE ILLUSTRATIONS FOR THE FOLLOWING
1210: %PARAGRAPHS.]
1211: 
1212: 
1213: \begin{figure}
1214: \centerline{\psfig{file=N.2.1.3a.ps,angle=-90,width=1.75in}
1215: \psfig{file=N.2.1.3b.ps,angle=-90,width=1.75in}}
1216: \vskip 5truept
1217: \caption{To add a cross cap to a surface, remove one disk and glue
1218: the remaining boundary circle to itself by identifying antipodal points.
1219: The construction cannot be physically carried out in $R^3$ without
1220: self-intersections, but abstractly the resulting surface has a seam
1221: that looks locally like the centerline of a M\"obius strip.
1222: }
1223: \label{N.2.1.3}
1224: \end{figure}
1225: 
1226: 
1227: In the case that no disks are removed, we have the sphere.
1228: 
1229: In the case that one disk is removed, we draw the sphere-minus-a-disk
1230: as a hemisphere, so that when the boundary circle gets glued to itself
1231: the spherical geometry continues smoothly across the seam from one side
1232: to the other, giving the resulting sphere-with-a-crosscap a uniform
1233: spherical geometry.
1234: 
1235: 
1236: \begin{figure}
1237: \centerline{\psfig{file=N.2.1.4.ps,angle=-90,width=2.75in}}
1238: \vskip 5truept
1239: \caption{Topological classification of closed 2-manifolds.
1240: Every closed 2-manifold may obtained from a sphere by
1241: adding either handles (left column) or cross caps (right column).
1242: }
1243: \label{N.2.1.4}
1244: \end{figure}
1245: 
1246: In the case that two disks are removed we may have either
1247: a sphere-with-a-handle, better known as a torus, or
1248: a sphere-with-two-crosscaps, better known as a Klein bottle.
1249: We draw the sphere-minus-two-disks as a cylinder.  The cylinder has
1250: an intrinsically flat geometry. 
1251: You can construct one from a sheet of paper
1252: without stretching, and when you glue its top and bottom of the
1253: rectangle to each other
1254: (for the torus) or each to itself (for the Klein bottle) the flat geometry
1255: continues uniformly across the seams.  Thus the torus and the Klein bottle
1256: each has a flat geometry.
1257: 
1258: 
1259: \begin{figure}
1260: \centerline{\psfig{file=N.2.1.5a.ps,angle=-90,width=1.75in}
1261: \psfig{file=N.2.1.5b.ps,angle=-90,width=1.75in}}
1262: \vskip 5truept
1263: \caption{A pair of pants can be made from two hexagons.  For best results
1264: use hexagons cut from the hyperbolic plane, with all interior angles
1265: being right angles.}
1266: \label{N.2.1.5}
1267: \end{figure}
1268: 
1269: In the case that three disks are removed we have a
1270: sphere-with-three-crosscaps,
1271: constructed from a three-way cylinder better known as a {\it pair of
1272: pants}.
1273: The pair of pants may in turn be constructed from two hyperbolic hexagons
1274: (fig.\ \ref{N.2.1.5}), giving it an intrinsically hyperbolic geometry.
1275: When the pants' boundary circles (the cuffs) are glued to themselves
1276: to make the sphere-with-three-crosscaps, the hyperbolic geometry continues
1277: uniformly across the seams.  Thus the sphere-with-three-crosscaps also gets
1278: a hyperbolic geometry.
1279: 
1280: All remaining surfaces -- those for which four or more disks are removed in
1281: the construction -- may be built from pairs of pants, so all get a
1282: hyperbolic
1283: geometry.  In summary, we see that of the infinitely many closed
1284: 2-manifolds,
1285: two admit spherical geometry (the sphere and the sphere-with-crosscap),
1286: two admit flat geometry (the torus and the Klein bottle), and all the rest
1287: admit hyperbolic geometry.
1288: 
1289: The next three sections will examine the flat, spherical, and hyperbolic
1290: closed 2-manifolds in greater detail, in each case answering
1291: the four questions raised in \S \ref{ovtop}.
1292: 
1293: 
1294: \subsubsection{Flat two-dimensional manifolds}
1295: \label{flattwo}
1296: 
1297: Manifolds:  torus and Klein bottle.
1298: 
1299: Flexibility:
1300:    The torus has three degrees of freedom in its construction:  the height
1301:    of the cylinder, the circumference of the cylinder, and the twist
1302:    with which the top gets glued to the bottom.
1303:    The Klein bottle has only two degrees of freedom in its construction:
1304:    the height and circumference of the cylinder.  Each end of the cylinder
1305:    gets glued to itself, identifying diametrically opposite points,
1306:    so there is no twist parameter.
1307: 
1308: 
1309: 
1310: \begin{figure}
1311: \centerline{\psfig{file=N.2.2.1.ps,angle=-90,width=1.75in}}
1312: \vskip 5truept
1313: \caption{The torus has a two-parameter continuous family of symmetries,
1314: consisting of rotations and vertical translations.
1315: }
1316: \label{N.2.2.1}
1317: \end{figure}
1318: 
1319: Symmetry:
1320:    Continuous symmetries.  The torus has a two-parameter continuous family
1321:    of symmetries, consisting of rotations and vertical translations
1322:    (fig.\ \ref{N.2.2.1}).  
1323:    This shows that the torus is {\it globally homogeneous}
1324:    in the sense that any point may be taken to any other point
1325:    by a global symmetry of the manifold.
1326:    The Klein bottle has only a one-parameter continuous family of
1327: symmetries,
1328:    consisting of rotations (fig.\ \ref{N.2.2.2}).  It is globally inhomogeneous.
1329:    When diametrically opposite points on a boundary circle are glued to make
1330:    a crosscap, the seam is a closed geodesic that looks locally like
1331:    the centerline of a M\"obius strip.  Points lying on this centerline are
1332:    fundamentally different from other points in the manifold.
1333: 
1334: 
1335: \begin{figure}
1336: \centerline{\psfig{file=N.2.2.2.ps,angle=-90,width=1.75in}}
1337: \vskip 5truept
1338: \caption{The Klein bottle has continuous rotational symmetry, but
1339: any attempt at vertical translation is blocked by the cross caps.}
1340: \label{N.2.2.2}
1341: \end{figure}
1342: 
1343: 
1344:    Discrete symmetries.  As well as its continuous family symmetries, the
1345: torus
1346:    has a discrete Z/2 symmetry given by a 180 degree rotation interchanging
1347:    the cylinder's top and bottom (fig.\ \ref{N.2.2.3}).  The discrete and
1348: continuous
1349:    symmetries may of course be composed.  The best way to think of this is
1350: that
1351:    the complete symmetry group of the torus consists of two disconnected
1352:    components:  symmetries that interchange the boundary circles and
1353:    symmetries that do not.  In special cases the torus may have additional
1354:    discrete symmetries (fig.\ \ref{N.2.2.4}).
1355:    The Klein bottle's discrete symmetry group is Z/2 + Z/2, generated
1356:    by vertical and horizontal reflections of the cylinder.  Vertical
1357: reflection
1358:    interchanges the two crosscaps.  Horizontal reflection takes each
1359: crosscap
1360:    to itself, but reverses the direction of its centerline.
1361: 
1362: 
1363: \begin{figure}
1364: \centerline{\psfig{file=N.2.2.3a.ps,angle=-90,width=1.in}
1365: \psfig{file=N.2.2.3b.ps,angle=-90,width=1.in}}
1366: \vskip 5truept
1367: \caption{The torus always has a discrete Z/2 symmetry given by a half turn,
1368: even if the cylinder's top and bottom circles are glued to each other
1369: with a twist.  The Klein bottle has a discrete Z/2 + Z/2 symmetry,
1370: which interchanges and/or inverts the cross caps.}
1371: \label{N.2.2.3}
1372: \end{figure}
1373: 
1374: 
1375: \begin{figure}
1376: \centerline{\psfig{file=N.2.2.4.ps,angle=-90,width=4.in}}
1377: \vskip 5truept
1378: \caption{}
1379: \label{N.2.2.4}
1380: \end{figure}
1381: 
1382: 
1383: Dirichlet domain:
1384:    We have seen that the torus is globally homogeneous, so the shape of the
1385:    Dirichlet domain does not depend on the choice of basepoint.
1386:    On the other hand, the torus has three degrees of flexibility, and
1387:    the Dirichlet domain's shape does depend on the shape of the torus.
1388:    Typically the Dirichlet domain is an irregular hexagon, but in
1389:    special cases it may be a square or a regular hexagon.
1390:    The Klein bottle is globally inhomogeneous, so even for a fixed geometry
1391:    of the Klein bottle, the Dirichlet domain's shape depends on the choice
1392:    of basepoint (fig.\ \ref{N.2.2.5}).
1393: 
1394: \begin{figure}
1395: \centerline{\psfig{file=N.2.2.5.ps,angle=-90,width=3.in}}
1396: \vskip 5truept
1397: \caption{In a Klein bottle, the shape of the Dirichlet domain depends
1398: on the choice of basepoint.  For example, a Dirichlet domain centered
1399: at the triple arrowhead is a hexagon, while a Dirichlet domain
1400: centered at a single arrowhead is a rectangle.}
1401: \label{N.2.2.5}
1402: \end{figure}
1403: 
1404: 
1405: 
1406: \subsubsection{Spherical two-dimensional manifolds}
1407: \label{sphertwo}
1408: 
1409: Manifolds:  sphere and sphere-with-crosscap.
1410: 
1411: Flexibility:
1412:    The sphere and the sphere-with-crosscap have no flexibility whatsoever.
1413:    Their geometry is completely determined by their topology.
1414:    Here and throughout we make the assumption that spherical geometry
1415:    refers to a sphere of radius one.  Otherwise these two manifolds would
1416:    have the flexibility of rescaling -- they could be made larger or
1417: smaller.
1418: 
1419: Symmetry:
1420:    The sphere has a symmetry taking any point to any other point, with any
1421:    desired rotation.  These symmetries form a 3-parameter continuous family.
1422:    It also has a single discrete symmetry, given by reflection through
1423:    the origin.
1424:    The sphere-with-crosscap has a 3-parameter continuous family of
1425: symmetries
1426:    taking any point to any other point with any desired rotation and either
1427:    reflected or not reflected.  It has no discrete symmetries.
1428:    Indeed the continuous family already takes a neighborhood of any point
1429:    to a neighborhood of any other point in all possible ways, leaving no
1430:    possibilities for discrete symmetries.
1431: 
1432: Dirichlet domain:
1433:    The sphere and the sphere-with-crosscap both have somewhat degenerate
1434:    Dirichlet domains (try the inflating-the-balloon experiment in each),
1435:    but these manifolds are globally homogeneous so the degenerate Dirichlet
1436:    domains do not depend on the choice of basepoint.
1437: 
1438: 
1439: \subsubsection{Hyperbolic two-dimensional manifolds}
1440: \label{hyptwo}
1441: 
1442: Manifolds:
1443:    There are infinitely many closed hyperbolic 2-manifolds,
1444:    illustrated in fig.\ \ref{N.2.1.4}.
1445: 
1446: Flexibility:
1447:    All closed hyperbolic 2-manifolds are flexible.
1448:    The shape of each manifold is parameterized by the circumferences
1449:    of the cuffs of the pairs of pants used to construct it, and
1450:    by the twists with which distinct cuffs are glued to each other.
1451:    A cuff that is glued to itself to make a crosscap has no twist
1452: parameter.
1453: 
1454: Symmetry:
1455:    Closed hyperbolic 2-manifolds never have continuous families of
1456: symmetries.
1457:   The reason is that each seam (where cuffs are joined together)
1458:    is the shortest loop in its neighborhood,
1459: and therefore cannot be slid
1460:    away from itself.  A seam cannot slide along itself either, because
1461:    a pair of pants admits no continuous rotations.
1462:    A generic closed hyperbolic 2-manifold also has no discrete symmetries,
1463:    but nongeneric ones may have discrete symmetries if the cuff lengths
1464:    and twists are chosen carefully.
1465: 
1466: Dirichlet domain:
1467:    The Dirichlet domain of a closed hyperbolic 2-manifold always depends
1468:    on the choice of basepoint.  For carefully chosen basepoints it may
1469:    exhibit some of the manifold's symmetries.
1470: 
1471: 
1472: \subsection{Three-dimensional manifolds}
1473: \label{threed}
1474: 
1475: 
1476: \subsubsection{Overview of three-dimensional manifolds}
1477: \label{overthree}
1478: 
1479: Unlike two-manifolds, not all three-manifolds admit a constant curvature
1480: geometry.  For a survey of the geometry of 3-manifolds, see 
1481: Ref.\ \cite{Tbull}.
1482: Roughly speaking,  any closed 3-manifold may be cut
1483: into pieces in a canonical way, and each piece admits a geometry that
1484: is homogeneous but not necessarily isotropic.  For a very elementary
1485: exposition of homogeneous anisotropic geometries, see 
1486: Ref.\ \cite{Shape}
1487: %Ch. 17 and 18. 
1488: For a more detailed but still readable exposition see
1489: \cite{pscott}.  The observed isotropy of the microwave background
1490: radiation implies that at least the observable portion of the universe
1491: is approximately homogeneous and isotropic, so cosmologists restrict
1492: their attention to the three homogeneous isotropic geometries:  spherical,
1493: flat and hyperbolic.  However, as noted earlier, an approximately flat
1494: observable universe is also consistent with a very large universe of
1495: varying curvature.
1496: 
1497: 
1498: \subsubsection{Flat three-dimensional manifolds}
1499: \label{flatthree}
1500: 
1501: Manifolds:
1502:    Figure \ref{N.3.2.1} illustrates the ten closed flat 3-manifolds.
1503: 
1504:    The first is the 3-torus introduced earlier -- a cube with
1505:    pairs of opposite faces glued.
1506: 
1507: \begin{figure}
1508: \centerline{\psfig{file=N.3.2.1.big.ps,angle=-90,width=4.5in}}
1509: \vskip 5truept
1510: \caption{The ten closed flat 3-manifolds.}
1511: \label{N.3.2.1}
1512: \end{figure}
1513: 
1514: 
1515:    The second and third manifolds are variations on the 3-torus in which
1516:    the front face is glued to the back face with a half (respectively quarter)
1517: turn,
1518:    while the other two pairs of faces are glued as in the 3-torus.
1519: Other possibilities for gluing the faces of a cube -- for example gluing
1520:    all three pairs of opposite faces with half turns -- fail  to yield flat
1521:    3-manifolds because the cube's corners do not fit together properly
1522:    in the resulting space.  See Ref.\
1523: \cite{Shape} for a more complete
1524:    explanation.
1525: 
1526:    The fourth and fifth manifolds are similar in spirit to the first three,
1527:    only now the cross section is a hexagonal torus instead of a square
1528: torus.
1529:    That is, the cross section is still topologically a torus, but has
1530:    a different shape (recall fig.\ \ref{N.2.2.4}).  The hexagonal cross section
1531: allows
1532:    the front of the prism to be glued to the back with a one-sixth or
1533:    one-third turn, producing two new flat 3-manifolds.  The side faces are
1534: all
1535:    glued to each other in pairs, straight across.
1536: 
1537:    The sixth manifold, the famous Hantzsche-Wendt manifold, is the most
1538:    interesting of all.  Unlike the preceding five manifolds, which we
1539: defined
1540:    by constructing fundamental domains, the most natural way to define the
1541:    Hantzsche-Wendt manifold is to start in the universal cover and define
1542: its
1543:    group of covering transformations.  Specifically, its group of covering
1544:    transformations is the group generated by screw motions about a set
1545:    of orthogonal but nonintersecting axes (indicated by heavy lines
1546:    in fig.\ \ref{N.2.2.4}).  Each screw motion consists of a half turn about an
1547: axis
1548:    composed with a unit translation along that axis.  Note that this group
1549:    of covering transformations does {\it not} take a basic cube to all
1550: other
1551:    cubes in the cubical tiling of 3-space.  Rather, the images of a basic
1552: cube
1553:    fill only half the cubes in the tiling, checkerboard style.  Thus a
1554: complete
1555:    fundamental domain would consist of two cubes, the basic cube and any one
1556:    of its immediate neighbors;  images of the basic cube would fill the
1557: black
1558:    cubes in the 3-dimensional checkerboard, while images of its neighbor
1559: would
1560:    fill the remaining white cubes.  But we would really prefer a fundamental
1561:    domain consisting of a single polyhedron.  To get one we employ the
1562:    balloon construction for a Dirichlet domain introduced in Section N.1.2.
1563:    Let the Dirichlet domain's basepoint be the center of a basic cube.
1564:    As the balloon expands its fills that basic cell entirely, and also fills
1565:    one sixth of each of its six immediate neighbors.  The resulting
1566: Dirichlet
1567:    domain is a rhombic dodecahedron (fig.\ \ref{N.3.2.1}).  The face gluings are
1568: given
1569:    by the original screw motions along the axes.
1570: Note that this construction of the Hantzsche-Wendt manifold
1571:           corrects an error, appearing elsewhere in the cosmological
1572:           literature, which takes the three screw axes to be coincident.
1573:           This erroneous construction leads to a cube with each pair
1574:           of opposite faces glued with a half turn.  The cube's corners
1575:           are therefore identified in four groups of two corners each
1576:           instead of a single group of eight corners.  The resulting
1577:           space has four singular points and is thus an orbifold instead
1578:           of a manifold.
1579: 
1580:    The seventh through tenth manifolds are again defined by fundamental
1581: domains.
1582:    They are similar to the first five manifolds, except now the cross
1583: section
1584:    is a Klein bottle instead of a torus.  That is, the fundamental domain's
1585:    left and right faces are glued to each other as in a torus, but the top
1586: face
1587:    (resp. bottom face) is glued to itself with a horizontal shift of half
1588: the
1589:    width of the fundamental domain, taken modulo the full width, of course.
1590:    This ensures that every cross section is a Klein bottle, which may be
1591:    understood as a cylinder with one crosscap at the top and another at the
1592:    bottom.  The fundamental domain's front face, which is now a Klein
1593: bottle,
1594:    may be glued to its back face in one of four ways:  plain (seventh
1595: manifold),
1596:    interchanging the crosscaps (eighth manifold), inverting the crosscaps
1597:    (ninth manifold), or interchanging and inverting the crosscaps
1598:    (tenth manifold).  Unlike the first six manifolds, these four are all
1599:    nonorientable.
1600: 
1601: Flexibility:
1602:    All ten closed flat 3-manifolds have at least two degrees of flexibility:
1603:    the depth of each fundamental domain (fig.\ \ref{N.3.2.1}) may be chosen
1604:    independently of its width and height.  For some of the manifolds,
1605:    for example the one-sixth turn manifold, this is the only flexibility.
1606:    Others have greater flexibility:  for example the cross section of the
1607:    half turn manifold may be any parallelogram, and the fundamental domain
1608:    for the 3-torus may be any parallelepiped.
1609: 
1610: Symmetry:
1611:    All closed flat 3-manifolds except the Hantzsche-Wendt manifold have
1612:    a continuous one-parameter family of symmetries that pushes into the
1613: page
1614:    in the fundamental domains in fig.\ \ref{N.3.2.1}.  These symmetries take each
1615:    cross section to a cross section further along, with the cross sections
1616:    that fall off the back of the stack reappearing cyclically at the front.
1617:    Some of the manifolds have additional continuous families of symmetries
1618:    as well.  The 3-torus is the most symmetrical, with a 3-parameter family
1619:    of symmetries taking any point to any other point.
1620:    All ten manifolds have discrete symmetries as well, which
1621:    vary from case to case and may depend on the manifold's exact shape
1622:    as well as its topology.  For example, a cubical 3-torus has a 3-fold
1623:    symmetry defined by a one-third turn about one of the cube's long
1624: diagonals,
1625:    but a 3-torus made from an arbitrary parallelepiped lacks this symmetry.
1626: 
1627: Dirichlet domain:
1628:    The 3-torus is exceptional in that its Dirichlet domain does not depend
1629:    on the choice of basepoint.  It is always the same, no
1630: matter
1631:    what basepoint you choose, because there is a symmetry taking any point
1632:    to any other point.  For the nine remaining closed flat 3-manifolds,
1633:    the Dirichlet domain always depends on the choice of basepoint.
1634: 
1635: Observational status:
1636:    All ten closed flat 3-manifolds are of course completely consistent
1637:    with recent evidence of a flat observable universe.
1638:    However, if the fundamental domain is larger than our horizon radius,
1639:    the topology may, in practice, be unobservable.
1640: 
1641: 
1642: \subsubsection{Spherical three-dimensional manifolds}
1643: \label{spherthree}
1644: 
1645: Manifolds:
1646:    This section requires only a minimal understanding of the 3-sphere,
1647:    or hypersphere, defined as the set of points one unit from the origin
1648:    in Euclidean 4-space.  Readers wanting to learn more about the 3-sphere
1649: may
1650:    consult 
1651: Ref.\ \cite{Shape}
1652: % Ch. 14 
1653: for an elementary exposition or
1654: Ref.\ \cite{Thurston}
1655: %section 2.7
1656: for a deeper understanding.
1657: 
1658:    Spherical manifolds fall into two broad categories.  The manifolds
1659:    in the first category have very different properties from those
1660:    in the second.
1661: 
1662:    First category:
1663: 
1664:    The first category manifolds are most easily defined in terms of their
1665:    groups of covering transformations, although it is not hard to work out
1666:    their Dirichlet domains afterwards.  Amazingly enough, their groups
1667:    of covering transformations -- which are symmetries of the 3-sphere --
1668:    come directly from finite groups of symmetries of the ordinary 2-sphere!
1669:    In other words, each finite group of symmetries of the 2-sphere
1670:    (that is, the cyclic, dihedral, tetrahedra, octahedral, and icosahedral
1671:    groups) gives rise to the group of covering transformations of
1672:    a spherical 3-manifold!
1673: 
1674:    The bridge from the 2-sphere to the 3-sphere is provided by the
1675: quaternions.
1676:    The quaternions are a 4-dimensional generalization of the
1677:    familiar complex numbers.  But while the complex numbers have a single
1678:    imaginary quantity $i$ satisfying $i^2 = -1$, the quaternions have three
1679:    imaginary quantities $i, j,$ and $k$ satisfying
1680: 	\ba
1681:        i^2 &=& j^2 = k^2 = -1  \nonumber \\
1682:        ij &=& k = -ji \nonumber \\
1683:        jk &=& i = -kj \nonumber \\
1684:        ki &=& j = -ik
1685: 	\ea
1686:    The set of all quaternions $a + bi + cj + dk$ (for $a,b,c,d$ real)
1687:    defines 4-dimensional Euclidean space, and the set of all unit length
1688:    quaternions, that is, all quaternions $a + bi + cj + dk$ satisfying
1689:  $  a^2 + b^2 + c^2 + d^2 = 1$, defines the 3-sphere.
1690: 
1691:    Before proceeding, it is convenient to note that there is nothing
1692:    special about the three quaternions $i, j,$ and $k$.
1693: 
1694:      Lemma N.3.3.1 (quaternion change of basis).  Let $c$ be a 3x3
1695:      orthogonal matrix, and define
1696: 	\ba
1697:         i^{\prime} &=& c_{ii} i  +  c_{ij} j  +  c_{ik} k \nonumber \\
1698:         j^{\prime} &=& c_{ji} i  +  c_{jj }j  +  c_{jk} k \nonumber \\
1699:         k^{\prime} &=& c_{ki} i  +  c_{kj}j  +  c_{kk} k 
1700: 	\ea
1701:      then $i^{\prime}$, $j^{\prime}$, and $k^{\prime}$ 
1702: satisfy the usual quaternion relations
1703: 	\ba
1704:        i^{\prime 2} &=& j^{\prime 2} = k^{\prime 2} = -1 \nonumber \\
1705:        i^{\prime}j^{\prime} &=& k^{\prime} = -j^{\prime}i^{\prime}\nonumber \\
1706:        j^{\prime}k^{\prime} &=& i^{\prime} = -k^{\prime}j^{\prime}\nonumber \\
1707:        k^{\prime}i^{\prime} &=& j^{\prime} = -i^{\prime}k^{\prime}
1708: 	\ea
1709:    Lemma N.3.3.1 says that an arbitrary purely imaginary quaternion
1710:    bi + cj + dk may, by change of basis, be written as $b^{\prime}i^{\prime}$.
1711:    If the purely imaginary quaternion bi + cj + dk has unit length,
1712:    it may be written even more simply as $i^{\prime}$.  
1713: A not-necessarily-imaginary
1714:    quaternion $a + bi + cj + dk$ may be transformed to 
1715: $a^{\prime} + b^{\prime}i$.
1716:    If it has unit length it may be written as 
1717: $\cos(\theta) + i^{\prime} \sin(\theta)$ for
1718:    some $\theta $. 
1719: 
1720:    The unit length quaternions, which we continue to visualize as the
1721: 3-sphere,
1722:    may act on themselves by conjugation or by left-multiplication.
1723: 
1724:      Lemma N.3.3.2 (conjugation by quaternions).  Let $q$ be an arbitrary
1725:      unit length quaternion.  According to the preceding discussion, we may
1726:      choose a basis $\{1, i^{\prime},j^{\prime},k^{\prime}\}$ such that 
1727: $q = \cos(\theta) + i^{\prime} \sin(\theta)$ for some
1728: $\theta$.
1729:      It is trivial to compute how q acts by conjugation on the basis
1730: $ \{1, i^{\prime},j^{\prime},k^{\prime}\}$.
1731: 	\ba
1732:         (\cos(\theta) + i^{\prime} \sin(\theta)) 1  (\cos(\theta) - i^{\prime} \sin(\theta)) &=& 1\nonumber \\
1733:         (\cos(\theta) + i^{\prime} \sin(\theta)) i^{\prime} (\cos(\theta) - i^{\prime} \sin(\theta)) &=& i^{\prime}\nonumber \\
1734:         (\cos(\theta) + i^{\prime} \sin(\theta)) j^{\prime} (\cos(\theta) - i^{\prime} \sin(\theta)) &=&  j^{\prime} \cos(2\theta) + k^{\prime}
1735: \sin(2\theta)\nonumber \\
1736:         (\cos(\theta) + i^{\prime} \sin(\theta)) k^{\prime} (\cos(\theta) - i^{\prime} \sin(\theta)) &=& -j^{\prime} \sin(2\theta) + k^{\prime}
1737: \cos(2\theta)
1738: \ea
1739:      We see that conjugation always fixes 1 (the north pole), so all
1740:      the action is in the equatorial 2-sphere spanned by 
1741: $\{i^{\prime},j^{\prime},k^{\prime}\}$.
1742:      The equatorial 2-sphere itself rotates about the $i^{\prime}$ axis through
1743:      an angle of $2\theta $.
1744: 
1745:    Compare the preceding action by conjugation to the following action
1746:    by left multiplication.
1747: 
1748:      Lemma N.3.3.3 (left multiplication by quaternions).  Let $q$ be an
1749: arbitrary
1750:      unit length quaternion.  According to the preceding discussion, we may
1751:      choose a basis $\{1, i^{\prime},j^{\prime},k^{\prime}\}$ such that 
1752: $q = \cos(\theta) + i^{\prime} \sin(\theta)$ for some
1753: t.
1754:      It is trivial to compute how q acts by left multiplication on the basis
1755: $   \{1, i^{\prime},j^{\prime},k^{\prime}\}$.
1756: 	\ba
1757:         (\cos(\theta) + i^{\prime} \sin(\theta)) 1  &=&
1758:   1  \cos(\theta) + i^{\prime} \sin(\theta)\nonumber \\
1759:         (\cos(\theta) + i^{\prime} \sin(\theta)) i^{\prime} &=& -1  \sin(\theta) + i^{\prime} \cos(\theta)\nonumber \\
1760:         (\cos(\theta) + i^{\prime} \sin(\theta)) j^{\prime} &=&  j^{\prime} \cos(\theta) + k^{\prime} \sin(\theta)\nonumber \\
1761:         (\cos(\theta) + i^{\prime} \sin(\theta)) k^{\prime} &=& -j^{\prime} \sin(\theta) + k^{\prime} \cos(\theta)
1762: 	\ea
1763:      We see that left multiplication rotates 1 towards $i^{\prime}$ while
1764:      simultaneously rotating $j^{\prime}$ towards $k^{\prime}$.  
1765: The result is a screw motion
1766:      known as a Hopf flow.  What is not obvious from the above computation
1767:      is that the Hopf flow is completely homogeneous -- it looks the same
1768:      at all points of the 3-sphere.
1769: 
1770:    All the tools are now in place to convert finite groups of symmetries
1771:    of the 2-sphere to finite groups of symmetries of the 3-sphere.
1772:    The algorithm is as follows:
1773: 
1774:      Step \#1.  Choose a finite group of symmetries of the 2-sphere.
1775:                (The only such groups are the cyclic groups, the dihedral
1776:                groups, the tetrahedra group, the octahedral group,
1777:                and the icosahedral group.  You might also expect a cubical
1778:                and a dodecahedral group, but they coincide with the
1779:                octahedral and icosahedral groups because the octahedron
1780:                is dual to the cube and the dodecahedron is dual to
1781:                the icosahedron.)
1782: 
1783:      Step \#2.  Write down the quaternions whose action by conjugation
1784:                gives the chosen symmetries.  Note that each symmetry
1785:                corresponds to two quaternions $q$ and $-q$, because
1786:                $q$ and $-q$ act identically when conjugating,
1787:                i.e. $q r q^{-1} = (-q) r (-q)^{-1}$.
1788: 
1789:      Step \#3.  Let the quaternions found in Step \#3 act by multiplication.
1790:                This action defines the group of covering transformations.
1791:                Note that although a quaternion's action by conjugation
1792:                always has a pair of fixed points on the 2-sphere,
1793:                its action by left multiplication on the 3-sphere is
1794:                always fixed point free.
1795: 
1796:    The preceding algorithm is most easily understood in a concrete example.
1797: 
1798:      Example N.3.3.4.  Consider the so-called {\it Klein four group},
1799:      the group of symmetries of the 2-sphere consisting of half turns
1800:      about the $i$-, $j$-, and $k$-axes.  (Remember that we are in the 3-space
1801:      spanned by $\{i,j,k\}$, so the $i$-, $j$-, and $k$-axes play the role of
1802:      the traditional $x$-, $y$-, and $z$-axes.)  By Lemma N.3.3.2 a half turn
1803:      about the $i$-axis corresponds to conjugation by the quaternions 
1804: $\pm i$.  That is, when $\theta = \pi/2$, the quaternion $i = \cos(\pi/2) 
1805: + i \sin(\pi/2)$
1806:      acts by conjugation as a rotation about the $i$-axis through an angle
1807:      $2\theta = \pi$.  
1808: Similarly, the half turn about the $j$-axis corresponds to
1809:      the quaternions $\pm j$, and the half turn about the $k$-axis corresponds
1810:      to $\pm k$.  The trivial symmetry corresponds to conjugation by 
1811: $\pm 1$.
1812:      Thus the complete set of quaternions is $\{\pm 1, \pm i, \pm j, \pm k\}$.
1813:   If we now
1814:      let these quaternions act by left multiplication on the 3-sphere,
1815:      they give a group of covering transformations of order 8.
1816:      (Exercise for the reader:  compute the action of each of those
1817:      eight quaternions on the basis $\{1,i,j,k\}$.)
1818:      It is not hard to see that the fundamental domain is a cube.
1819:      The action of the covering transformations shows you that each face
1820:      of the cube is glued to the opposite face with a one-quarter turn.
1821:      Eight copies of the cube tile the 3-sphere like the eight faces
1822:      of a hypercube.
1823: 
1824: %   Fig.\ \ref{N.3.3.1} applies this algorithm to all finite symmetry groups of
1825: %   the sphere, and illustrates the resulting spherical 3-manifolds.
1826: 
1827: The following table shows the results of applying the algorithm to
1828: each 
1829: finite group of symmetries of the 2-sphere:
1830: 
1831: \vskip 15truept
1832: 
1833: 
1834: {\offinterlineskip
1835: \halign{ \vrule # & \strut\ # & \vrule # &
1836: 	\ # \ & \vrule # \cr
1837: \multispan{5}\hrulefill\cr
1838: height2pt & \omit & & & \cr
1839: & symmetry group of 2-sphere & & spherical 3-manifold &\cr
1840: height2pt & \omit & & &  \cr
1841: \multispan{5}\hrulefill\cr
1842: height2pt & \omit & & & \cr
1843: & cyclic $Z/n$              & & lens space $L(2n,1)$          & \cr
1844: &    & & fundamental domain is lens-shaped solid & \cr
1845: &   & & &\cr
1846: & dihedral $D_n$ & & prism manifold & \cr
1847: &  & & fundamental domain is prism with $2n$-gon base & \cr
1848: &   & & &\cr
1849: & tetrahedral (order 12)  & & octahedral space & \cr
1850: &  & & fundamental domain is octahedron  & \cr 
1851: &  & & 24 of which tile the 3-sphere     &\cr
1852: &  & &	in the pattern of a regular 24-cell & \cr
1853: &   & & &\cr
1854: & octahedral (order 24)  & & snub cube space & \cr
1855: &   & & fundamental domain is snub cube & \cr 
1856: &   & & 48 of which tile the 3-sphere& \cr
1857: &   & & &\cr
1858: & dodecahedral (order 60)  & & P\" oincar\' e dodecahedral space & \cr
1859: &   & & fundamental domain is dodecahedron &\cr
1860: &   & &  120 of which tile the 3-sphere & \cr
1861: &   & & in the pattern of a regular 120-cell & \cr
1862: height2pt & \omit & & &  \cr
1863: \multispan{5}\hrulefill\cr }}
1864: 
1865: 
1866: \vskip 10truept
1867: 
1868:    Second category:
1869: 
1870:    Second category manifolds do not arise from simple left
1871:    multiplication by groups of quaternions.
1872:    The simplest second category manifolds are generic
1873: {\it lens spaces}.  The group of covering
1874:    transformations of a lens space is cyclic, generated by a single
1875:    screw motion.  For the lens space $L(p,q)$, with $p$ and $q$ 
1876: relatively prime
1877:    integers satisfying $p > q > 0$, the screw motion translates a distance
1878: $2\pi/p$
1879:    along a given geodesic, while rotating through an angle $q(2\pi/p)$ about
1880: that
1881:    same geodesic.  The fundamental domain is a lens-shaped solid, $p$ copies
1882:    of which tile the 3-sphere in much the same way that $p$ sectors tile
1883:    the surface of an orange.  In the special case that $q = 1$,
1884:    the screw motion corresponds to left multiplication by the quaternion
1885:    $\cos(2\pi/p) + i \sin(2\pi/p)$, so $L(p,1)$ is a first category manifold.
1886:    Otherwise $L(p,q)$ is a second category manifold.  (Warning:  $L(p, p-1)$
1887:    is the mirror image of $L(p,1)$ and therefore has the properties of a first
1888:    category manifold.  It corresponds to right multiplication
1889:    by $\cos(2\pi/p) + i \sin(2\pi/p)$ instead of left multiplication.)
1890: 
1891: The remaining second category manifolds are defined by simultaneous
1892:       left and right multiplication by quaternions.  That is, the holonomy
1893:       group of each such manifold is a subgroup of a product 
1894: 	$H_1 \times H_2  \times {\pm 1}$,
1895:       where $H_1$ and $H_2$ are finite groups of unit length quaternions.
1896:       An element $(h_1, h_2, \pm1)$ acts according to the rule 
1897: 	$q \rightarrow \pm h_1 q h_2^{-1}$.
1898:       For details, Section 4.4 of \cite{Thurston} and Sections 3 and 4
1899: of cite{newsphere}.
1900: 
1901: Flexibility:
1902:    Spherical 3-manifolds are rigid;  they have no flexibility.
1903:    (As in 2-dimensions we make the assumption that spherical geometry
1904:    refers to a sphere of radius one.  Otherwise all spherical manifolds
1905: would
1906:    have the flexibility of rescaling -- they could be made larger or
1907: smaller.)
1908: 
1909: Symmetry:
1910:    First category manifolds are always globally homogeneous, that is,
1911:    each has a continuous 3-parameter family of symmetries taken any point
1912:    to any other point.
1913: 
1914:       Proposition N.3.3.4 (global homogeneity of first category manifolds).
1915:       For any two points P and Q in a first category manifold M,
1916:       there is a symmetry of M taking P to Q.
1917: 
1918:       Proof.  Think of M as the quotient of the 3-sphere under the action
1919:       of a finite group of unit length quaternions $\{g_1 = 1, g_2, ...,
1920: g_{n}\}$.
1921:       The point P has $n$ images represented by the quaternions
1922:       $\{g_0 x, ..., g_n x\}$, and the point Q has n images represented
1923:       by the quaternions $\{g_0 y, ..., g_n y\}$, for some x and y.
1924:       Right multiplication by $(x^{-1} y)$ is a symmetry of the 3-sphere taking
1925:       the images of P to the images of Q.  More generally, right
1926: multiplication
1927:       takes every set of equivalent quaternions to some other set of
1928: equivalent
1929:       quaternions, so it defines not only a symmetry of the 3-sphere but
1930: also
1931:       a symmetry of the original manifold M.  Q.E.D.
1932: 
1933:    Second category manifolds have smaller continuous families of symmetries.
1934:    For example, the lens space $L(p,q)$ has a 2-parameter family of continuous
1935:    symmetries defined by translations along, and rotations about, its
1936: central
1937:    geodesic.  Both first and second category manifolds may have discrete
1938:    symmetries, which depend on the manifold.
1939: 
1940: Dirichlet domain:
1941:    The Dirichlet domain of a first category manifold never depends
1942:    on the choice of basepoint.  Proof:  By Proposition N.3.3.4 the manifold
1943:    has a symmetry taking any point to any other point, so the manifold
1944:    looks the same at all basepoints, and the Dirichlet domain construction
1945:    must yield the same result.
1946:    By contrast, the Dirichlet domain for a second category manifold
1947:    does depend on the choice of basepoint.
1948: 
1949: Observational status:
1950:    Recent evidence of an approximately flat observable universe implies that
1951:    if the universe is spherical its curvature radius may be 
1952:    comparable to or larger than
1953:    our horizon radius.  So even though all spherical manifolds are viable
1954:    candidates for the topology of the universe, the simplest ones, which
1955: have
1956:    large fundamental domains, would be much larger than the horizon radius
1957:    and couldn't be observed directly.  A more complicated topology, with
1958:    a larger group of covering transformations and a smaller fundamental
1959: domain,
1960:    would be more amenable to direct observation.
1961: 
1962: 
1963: \subsubsection{Hyperbolic three-dimensional manifolds}
1964: 
1965: Manifolds:
1966:    In 1924 Alexander Friedmann published a paper 
1967: \cite{Friedmann1924}
1968: complementing
1969:    his earlier model of an expanding spherical universe with a model
1970:    of an expanding hyperbolic universe.  It is a tribute Friedmann's
1971: foresight
1972:    that he explicitly allowed for the possibility of a closed hyperbolic
1973:    universe at a time when no closed hyperbolic 3-manifolds were known.
1974:    Fortunately examples were not long in coming.  L\"obell published the first
1975:    in 1929 \cite{Löbell1929} and Seifert and Weber published a much simpler
1976:    example soon thereafter \cite{SW1932?}.  Nevertheless, during the first half
1977:    of the twentieth century few closed hyperbolic 3-manifolds were know.
1978:    The situation changed dramatically in the mid 1970's with the work
1979:    of Thurston, who showed that most closed 3-manifolds admit a hyperbolic
1980:    geometry.  For an overview, see 
1981: Ref.\ \cite{Tbull}.  To study
1982:    known low-volume closed hyperbolic 3-manifolds, see Ref.\ 
1983: \cite{snappea}.
1984: 
1985: Flexibility:
1986:    Unlike closed hyperbolic 2-manifolds, closed hyperbolic 3-manifolds
1987:    are rigid;  they have no flexibility.  As usual we assume
1988:    curvature radius one, to suppress the trivial flexibility of rescaling.
1989: 
1990: Symmetry:
1991:    Closed hyperbolic 3-manifolds never have continuous families of
1992: symmetries.
1993:    Truly generic closed hyperbolic 3-manifolds appear to have no discrete
1994:    symmetries either, although the symmetry groups of low-volume manifolds
1995:    are typically small cyclic or dihedral groups.  Nevertheless, Kojima
1996:    has shown that every finite group occurs as the symmetry group of a
1997: closed
1998:    hyperbolic 3-manifold \cite{Kojima??}.
1999: 
2000: Dirichlet domain:
2001:    The Dirichlet domain of a closed hyperbolic 3-manifold always depends
2002:    on the choice of basepoint.  For carefully chosen basepoints, the
2003:    Dirichlet domain may display some of the manifold's symmetries.
2004:    The Dirichlet domain may also display symmetries beyond those
2005:    of the manifold itself;  these so-called hidden symmetries are actually
2006:    symmetries of finite-sheeted covering spaces.
2007: 
2008: \newpage
2009: \section{Standard Cosmology and The Cosmic Microwave Background}
2010: \label{cmbsection}
2011: 
2012: \setcounter{equation}{0}
2013: 
2014: \def\theequation{\thesection.\arabic{equation}}
2015: 
2016: 
2017: In standard big bang theory, the universe is created in a hot energetic state,
2018: a giant primordial soup.  The plasma is opaque to light as photons scatter
2019: off hot charged particles.
2020: As the
2021: universe expands, the soup cools and some 300,000 years into our history, the
2022: cosmos cools enough so that light no longer scatters
2023: efficiently and the universe becomes transparent to radiation.
2024: This transition, known as decoupling, marks the time when the
2025: primordial radiation can free stream throughout the universe unimpeded. 
2026: Decoupling happens quickly but not instantaneously and the small spread in
2027: time will introduce some complication in our attempts to measure topology as
2028: discussed in later sections.  This ancient radiation filling 
2029: all of space has cooled today to a mere $2.73{}^\circ$ K, in the 
2030: microwave range from which it has acquired the name
2031: the cosmic microwave background (CMB).
2032: As the CMB carries information almost from
2033: the beginning of time, it is a
2034: relic of our deep past.
2035: 
2036: The light received at the Earth has traveled 
2037: the same distance in
2038: all directions since last scattering
2039: and therefore defines a sphere.
2040: This sphere is known as the surface of last scatter (SLS)
2041: and is only defined relative to a given observer.  An alien civilization
2042: in a neighboring cluster of galaxies, if coincidentally also
2043: conducting observations of the CMB today, would receive photons from
2044: its own SLS as shown in fig. \ref{alien}.
2045: This will be particularly relevant to
2046: the circles method of detecting topology described in section
2047: \S \ref{obscirc}.
2048: 
2049: 
2050: \begin{figure}
2051: \centerline{\psfig{file=alien.eps,angle=-90,width=1.5in}}
2052: \vskip 5truept
2053: \caption{Two different surfaces of last scattering, each centered on 
2054: a different galaxy.}
2055: \label{alien}
2056: \end{figure}
2057: 
2058: 
2059: The famed CMB
2060: carries much coveted information about the
2061: universe.  It has been hoped that all local geometric quantities such as
2062: the local curvature, the expansion rate, the nature of the
2063: dark matter, and of the luminous matter, can be extracted from this one source. 
2064: Clearly though the data will leave the full set of parameters underdetermined
2065: and a great deal of model dependence is unavoidable.  What exactly the CMB can
2066: tell us is being put to the test with the recent high resolution experiments
2067: such as TOCO \cite{toco}, Boomerang and MAXIMA \cite{exper}.
2068: These experiments are consistent with an approximately flat observable
2069: universe.  If the universe is hyperbolic, its curvature radius may be comparable
2070: to or larger than the horizon radius.  So even though all hyperbolic 
2071: manifolds are viable candidates for the topology of the universe, they
2072: may be too large to see, therein is the challenge.
2073: The CMB can also be used to
2074: decipher the global topology as well as fitting these many local parameters.
2075: Though statistical measures of topology are inherently model dependent,
2076: geometric methods are model independent.
2077: The pattern based searches of circles in the sky and topological pattern
2078: formation require no assumptions about the primordial spectrum nor the nature
2079: of the dark matter.  However, these searches will likely be plagued by
2080: difficulties of their own and it is unclear if these methods will succeed
2081: without a great deal of refitting.
2082: 
2083: 
2084: 
2085: The standard big bang theory also has a powerful successor,
2086: the standard inflationary theory.  The standard inflationary theory suggests
2087: there was a period of accelerated expansion very early in the universe's 
2088: history
2089: which dilutes all fluctuations, all matter, and all energy so that the cosmos
2090: suddenly goes from a tumultuous beginning to a nearly flat, empty, and biggish
2091: universe.  Inflation is exited when it becomes energetically favorable for the
2092: energy driving the expansion to be released into a renewed soup of high energy
2093: particles and light.  This heating of the universe mimics the original 
2094: primordial soup
2095: and the standard history of the universe resumes. 
2096: There are proponents of inflation who 
2097: would revise the big bang still further,
2098: essentially removing the moment of genesis altogether so that inflation is
2099: eternal \cite{{vilenkin},{linde}}.  
2100: The universe can be viewed as a ginger root of inflationary patches
2101: each with its own details.  
2102: As inflation ends in any given patch, the energy is released to heat that
2103: region.
2104: As far as any given observer is 
2105: concerned, the heating epoch on exit from inflation looks like a consequence
2106: of a big bang event.
2107: Since this is all we can ever observe
2108: with the CMB our discussion will apply to any of these early universe 
2109: scripts.  
2110: 
2111: While
2112: theorists still find it tricky to identify a
2113: specific working model of inflation
2114: and to devise a graceful exit to inflation, 
2115: it is the favored story to tell and some of the
2116: reasons for this are good.
2117: Inflation elegantly resolves the mysterious homogeneity and isotropy of our
2118: observable universe.  Without incredibly precise tuning of the initial
2119: conditions, the universe would tend to be extremely lumpy and would
2120: quickly cool 
2121: if underdense or would quickly collapse if overdense.  The observational
2122: fact that the universe is billions of years old and still just marginally
2123: curved, if not outright flat, seems special and so unsettling.  A more rigorous
2124: and compelling argument is made by Guth in his original paper
2125: \cite{guth0} and many subsequent articles \cite{guth1}.
2126: Our homogeneous and isotropic space becomes remarkably unlikely,
2127: and yet here we
2128: are.
2129: 
2130: On the other hand lumpiness seems natural 
2131: since the many regions of the universe would not be
2132: in causal contact and therefore should have very different local properties
2133: such as 
2134: temperature, density etc..  Yet when we look at the CMB we see that in fact the
2135: temperature appears identical to better than 1 part in $10^5$ even for regions
2136: which would be separated far beyond causally connected 
2137: distances.  This would
2138: be like finding two ancient civilizations on opposite sides of the Earth with
2139: nearly identical languages.
2140: The civilizations must have been in causal contact.
2141: Likewise two regions of the universe which seem to have equilibrated to 
2142: precisely the same temperature must have been in causal contact.
2143: Inflation resolves this situation by
2144: taking a small causally connected region and stretching it so large that it
2145: exceeds the extent of the observable universe today.  
2146: The CMB is essentially the same
2147: temperature throughout our patch of the cosmos precisely because it was in
2148: causal contact.  The stretching also naturally renders the universe flat. 
2149: 
2150: 
2151: 
2152: Another motivation for inflation is that, while it makes the universe on
2153: average homogeneous and isotropic, it also naturally generates 
2154: fluctuations about this average. 
2155: These minute perturbations are critical for initiating the collapse
2156: of matter into galaxies and clusters.
2157: These seemingly insignificant fluctuations become the catalysts
2158: for all the order and structure we observe in the universe.
2159: Because of the enormous stretching of scales involved, it manages to place
2160: fluctuations even on the largest possible scales and therefore gives a causal
2161: explanation for the seeds of structure formation.
2162: (Another mechanism for generating the all important initial density 
2163: fluctuations
2164: is via topological defect models although these have lately fallen into
2165: disrepute \cite{topstrings}.)
2166: 
2167: 
2168: To emphasize, the beauty is that inflation both explains the average
2169: homogeneity and isotropy as well as giving a prediction for the 
2170: deviations from homogeneity, observed both in the CMB
2171: and in the form of galaxies.
2172: It seems a good question to ask , Is
2173: topology consistent with inflation?  Precisely as inflation 
2174: drives the curvature
2175: scale beyond observational reach, so too will it drive any topological scale
2176: beyond observational reach.  The universe may still be finite, only
2177: we'll never know it.  Certainly a model of inflation could be concocted which
2178: exited just at the critical moment so that the topology of the universe, and
2179: presumably the curvature too, is just within observation now,
2180: when we also happen to be here to look.  However, such a
2181: contraption would be a perversion against the spirit of inflation as a restorer
2182: of naturalness.  If an observable topology scale and inflation are to be
2183: conjoined, the reasons must be profound and topology must be an integral part
2184: of the inflationary mechanism.  This is conceivable.  For instance the cosmic
2185: Casimir effect is the contribution to the vacuum energy due to any topological
2186: boundary.  If this effect were to dilute as the topology scale expanded then
2187: one can envision inflation driven by this contribution to the vacuum and ending
2188: precisely as the topology scale became some natural big size and the vacuum
2189: energy became some correspondingly natural small size.  To really wrap it all
2190: into one clean package one could even try to explain the seemingly unnatural 
2191: $\Omega_\Lambda=0.7$ this way \cite{imogen}
2192: (see also Refs. \cite{{sokolov},{fagundes1}}.  Attempts to use compact
2193: topology to aid $\Omega<1$ inflation have also been suggested \cite{css3}.
2194: A related mechanism has been put to use in a model with topologically
2195: compact extra dimensions \cite{sst}.
2196: Another recent argument to justify flatness and local isotropy with
2197: topology and without inflation can be found in Ref.\ \cite{bark}.
2198: Barrow and Kodama found that compact topology can severely restrict 
2199: the anisotropies which an infinite universe will allow.  Roughly speaking,
2200: the anisotropic modes will not fit in the compact space.
2201: 
2202: In any case, the inflationary predictions for the statistical distribution
2203: of fluctuations are often used in fixing the initial fluctuations and so will
2204: be relevant here.
2205: This assumption is a weakness of all statistical
2206: constraints on topology and we will discuss these issues in due course.  The
2207: topic for this section is really, regardless of their origin, how do
2208: fluctuations in spacetime translate into temperature fluctuations in the CMB. 
2209: And once this has been established, how does topology alter the standard
2210: predictions leaving 
2211: an archeological imprint for us to uncover billions of years
2212: later.
2213: 
2214: \subsection{Standard Cosmological Equations}
2215: \label{standardeq}
2216: 
2217: This section provides a
2218: quick pedestrian review of the CMB and theories for the generation of its
2219: perturbations.  There is nothing specific to topology in this section
2220: unless explicitly stated.
2221: Many detailed reviews on the CMB have been written and the reader is
2222: referred to these \cite{{mkb},{hu},{ll},{whu}}.  Observing topology in the CMB
2223: is discussed in sections \ref{obstop}, \ref{obsflat}, \ref{obshyp}.
2224: 
2225: The gravitational
2226: field in general relativity is determined by the local Einstein
2227: equations
2228: 	\be
2229: 	G_{\mu \nu}=8\pi G T_{\mu \nu}
2230: 	.
2231: 	\ee
2232: The tensor $G_{\mu \nu}$ describes the curvature and evolution of 
2233: the metric $g_{\mu \nu}$ while $T_{\mu \nu}$ is the energy-momentum
2234: tensor and accounts for the matter fields.
2235: (For thorough discussions on general relativity see \cite{{waldbk},
2236: {mtw},{swein}}.)
2237: It is important to note here
2238: that the Einstein equations are local and therefore only
2239: determine the curvature of spacetime and do not fix the topology.
2240: Furthermore the symmetries of $g_{\mu \nu}$ are also local and are 
2241: nearly always broken in the manifold by topological identifications.
2242: Today the universe appears to be homogeneous and isotropic on the
2243: largest-scales. The CMB gives tremendous confirmation of these
2244: symmetries since the temperature appears identical in every
2245: direction to better than 1 part in $10^5$.  
2246: Homogeneity and isotropy imply that the Earth is not in a privileged position
2247: in the cosmos.
2248: The symmetries of homogeneity and isotropy severely restricts
2249: the class of possible solutions to the Friedman-Robertson-Walker (FRW)
2250: models defined by the gravitational metric
2251: 	\ba
2252: 	ds^2 &=&g_{\mu \nu}dx^\mu dx^\nu \nonumber \\
2253: %	&=&a^2(\eta)\left [-d\eta^2
2254: %	+{1\over 1-\kappa r^2}\left (dr^2+r^2(d\theta^2+\sin^2\theta d\phi^2)
2255: %	\right )\right ]\nonumber \\
2256: 	&=&-dt^2+a^2(\eta)
2257: 	\left [{dr^2\over 1-\kappa r^2}+r^2(d\theta^2+\sin^2\theta d\phi^2)
2258: 	\right  ]
2259: 	\label{stanmet}
2260: 	\ea
2261: where $\kappa=0,-1,1$ corresponds to a flat, negatively curved (hyperbolic)
2262: and positively
2263: curved (elliptical) space respectively (see fig.\ \ref{N.2.1.1}).
2264: Negatively curved space is often termed
2265: ``open'' and positively curved space ``closed'' in reference to their
2266: simply connected geometries.  Since we are interested in constructing
2267: compact, multiconnected 
2268: spaces, we will avoid the ``open'', ``closed'' terminology.
2269: 
2270: The overall scale factor, $a(\eta)$, describing the 
2271: expansion of the universe is determined by the energy of matter
2272: through the remaining Einstein equations.  
2273: We can always operate in comoving units where the manifold is treated as a 
2274: static constant curvature manifold and all of the dynamics is hidden in the 
2275: conformal scale factor $a(\eta)$. 
2276: An alternative expression for the metric is 
2277: 	\be
2278: 	ds^2=a^2(\eta)\left [-d\eta^2 +d\sigma^2 \right ]
2279: 	\ee
2280: with the spatial part of the metric
2281: 	\be
2282: 	d\sigma^2=d\chi^2+f(\chi)(d\theta^2+\sin^2\theta d\phi^2)
2283: 	\label{cosmet}
2284: 	\ee
2285: where we have used conformal time $d\eta=dt/a(t)$ and
2286: 	\be
2287: 	f(\chi) = \left\{ \begin{array}{lll}
2288: 	{\chi}^2 & r=\chi & {\rm flat }  \\
2289: 	 \sinh^2\chi & r=\sinh\chi &
2290: 	{\rm hyperbolic}\\
2291: 	 \sin^2 \chi & r=\sin\chi &
2292: 	{\rm spherical}  \end{array}\right.
2293: 	\ee
2294: See also Appendix 
2295: \ref{AppendixA}. 
2296: %\subsubsection{Evolution equations}
2297: 
2298: The dynamical evolution of the space is given by
2299: the Einstein equation determining the scale factor,
2300: 	\be
2301: 	H^2+\kappa/a^2={8\pi G\over 3}\rho.
2302: 	\label{fried}
2303: 	\ee
2304: The different curvatures correspond to different values of the
2305: global energy density.  Traditionally, $\Omega=\rho/\rho_c$ is defined 
2306: with $\rho_c$ the critical density required to render the universe flat,
2307: so that
2308: $\Omega>1$ corresponds to $\kappa=1$, $\Omega=1$ corresponds to $\kappa=0$,
2309: and $\Omega<1$ corresponds to $\kappa=-1$.
2310: The curvature radius is $R_{\rm curv}=a/|\kappa |^{1/2}$.  
2311: From eqn (\ref{fried}), $8\pi G\rho_c/3=H^2$ so that 
2312: $H^2a^2(\Omega-1)=\kappa$ and the curvature radius can be be expressed in 
2313: terms of $\Omega$ as
2314: 	\be
2315: 	R_{\rm curv}={1\over H\left |\Omega-1\right |^{1/2}} .
2316: 	\ee
2317: When working in comoving
2318: units we take the comoving curvature radius to be unity for curved
2319: space or $\infty$ for 
2320: flat space.  
2321: 
2322: Conservation of energy requires 
2323: 	\be
2324: 	\dot \rho+3H(\rho +p)=0 \label{consv}
2325: 	\ee
2326: for the perfect fluid energy-momentum tensor
2327: $T_\mu^\nu=(\rho,p,p,p )$. 	
2328: Decoupling occurs during the matter dominated error for which 
2329: $p=0$ and so 
2330: $\rho\propto 1/a^3$ by eqn (\ref{consv}).  
2331: The solution for $a(\eta)$ during matter domination is
2332: 	\be
2333: 	a(\eta)\propto \left \{\begin{array}{ll}
2334: 	 \cosh \eta -1 & \kappa =-1 \\
2335: 	 \eta^2/2 & \kappa = 0  \\
2336: 	 1-\cos\eta & \kappa =1\end{array}\right.
2337: 	\label{aevolv}
2338: 	\ee
2339: with the conformal Hubble expansion ${\cal H}=a^{\prime}/a=aH$,
2340: 	\be
2341: 	{\cal H}\propto  \left \{ \begin{array}	{ll}
2342: 	\sinh \eta/(\cosh \eta -1)
2343: 	& \kappa=-1 \\
2344: 	 2/\eta & \kappa=0\\
2345: 	 \sin\eta/(1-\cos\eta) & \kappa=1 \end{array}\right. .
2346: 	\ee
2347: 
2348: In comoving units, the universe is static.  In physical units, the universe 
2349: expands and the cosmic background radiation redshifts.  The physical
2350: wavelength is shifted according to 
2351: 	\be
2352: 	{\lambda(t)\over \lambda_{\rm initial}}={a(t)\over a(t_{\rm initial})}
2353: 	\ee
2354: from which the cosmological redshift is defined:
2355: 	\be
2356: 	{1+z}={a_0\over a(t)}.
2357: 	\ee
2358: A subscript
2359: $0$ will always be used to denote values today.  
2360: The value of $\eta $ during matter domination
2361: can be written in terms of $\Omega_0$ and the redshift 
2362: using $\eta=\int dt/a=\int da/(\dot a a)$ and eqn (\ref{fried}):
2363: 	\be
2364: 	\eta =\left\{\begin{array}{ll}
2365: 	{\rm arccosh}\left (1+{2(1-\Omega_0)\over \Omega_0(1+z)}\right )
2366: 		& \Omega_0<1\\
2367: 	{3\over (1+z)^{1/2}} &
2368: 	\Omega_0=1\\
2369: 	\cos^{-1}\left (1-{2(\Omega_0-1)\over \Omega_0(1+z)}\right )
2370: 		& \Omega_0>1\end{array}\right.
2371: 	\ee
2372: \cite{kolbturner}.
2373: 
2374: If curvature or topology are to be observable, the geometric scales must 
2375: be smaller than the diameter of the SLS.
2376: In flat space, there is no natural scale since $R_{\rm curv}=\infty$ and 
2377: it would seem a random coincidence if the topology scale just happened 
2378: to be observable.  In curved space, one might expect the curvature and 
2379: topology scales to be comparable.  This may not make it any more natural to 
2380: observe geometric effects, but it would mean that if curvature is 
2381: observable, then topology may be also.
2382: 
2383: For the purpose of observational topology it is useful to list
2384: some relevant scales.
2385: The radius of the surface of last scatter is defined as the
2386: the distance light travels between the time of decoupling and today,
2387: 	\be
2388: 	D_\gamma=a(t)\int_{t_{d}}^{t_0} {dt\over a(t)}.
2389: 	\ee
2390: The distance depends on the curvature and evolution of the universe
2391: through eqn (\ref{aevolv}).  
2392: In comoving units the
2393: photon travels a distance
2394: 	\be
2395: 	d_\gamma={D_\gamma\over a(t)}=\int_{t_{d}}^{t_0} {dt\over a(t)}
2396: 	=\int_{\eta_{d}}^{\eta_0}d\eta=\Delta \eta
2397: 	\ee
2398: with $\Delta \eta\equiv \eta_0-\eta_{d}$.
2399: The diameter of the SLS, $2\Delta \eta$, essentially defines the 
2400: extent of the observable universe.  A loose criterion that can be used
2401: to gauge if
2402: topology will influence the CMB is that 
2403: the in-radius be less than $\Delta \eta$.
2404: 
2405: The volume enclosed by the SLS is the integral over
2406: $V_{SLS} =\int \sqrt{-g}drd\theta d\phi $.  For a radius of $\Delta \eta$,
2407: 	\ba
2408: 	V_{SLS} &=& \pi\left (\sinh(2\Delta \eta-2\Delta \eta)\right )
2409: 	 \Omega_0<1 \nonumber \\
2410: 	&= &{4\pi \over 3}\Delta \eta^3 \quad\quad\quad \Omega_0=1\nonumber \\
2411: 	&=&\pi\left (2\Delta \eta-\sin(2\Delta \eta)\right ) \Omega_0>1.
2412: 	\ea
2413: The number of clones of the fundamental domain that can be observed in 
2414: a small compact cosmos can be estimated by the number of copies that can
2415: fit within the SLS: $V_{SLS}/{\cal V}_{\cal M}$.
2416: We will often refer back to these scale comparisons in the
2417: coming discussions.
2418: 
2419: 
2420: \subsection{Fluctuations in the CMB}
2421: \label{fluctcmb}
2422: 
2423: Regardless of the origin, there are small deviations from homogeneity
2424: and isotropy which are treated by perturbing the metric
2425: and matter tensors about the FRW solutions.  The equations of motion for these 
2426: perturbations to linear order are determined by 
2427: $\delta G_{\mu \nu}=8\pi G\delta T_{\mu \nu}$.  While the equations are
2428: extremely complicated there is one quantity, the 
2429: gauge invariant 
2430: potential $\Phi$, which 
2431: is a coordinate invariant combination of 
2432: components of $\delta g_{\mu \nu}$ and 
2433: is of central importance for our considerations.
2434: It is so named since in the Newtonian limit it corresponds to the usual gravitational potential.
2435: 
2436: The hot and cold spots in the CMB are caused by these perturbations in the 
2437: gravitational potential.   The fluctuations Doppler shift the photons
2438: generating an anisotropy both at decoupling and as light traverses the 
2439: time changing gravitational field.  
2440: There are many carefully derived results for the temperature fluctuations in the
2441: literature based on relativistic kinetic theory and relativistic
2442: perturbation theory \cite{{peebk},{ll},{whu}}.  Since these topics comprise multiple review
2443: papers on their own, we will not go through the detailed derivations
2444: but will instead try to provide a cohesive, intuitive motivation for
2445: each of the relevant concepts.  
2446: 
2447: To derive the famous Sachs-Wolfe
2448: effect
2449: \cite{sw}, we followed the pedagogical discussion of White and Hu
2450: \cite{whu}.  Photons gain energy and are
2451: therefore blueshifted as they fall into a potential well and lose
2452: energy and are therefore redshifted as they climb out.  The Doppler
2453: shifts will cancel in a static homogeneous and isotropic space.
2454: However at the time that the photons last scatter, some photons
2455: will be in
2456: potential wells and some will be in potential peaks.  Therefore
2457: this snapshot of the metric fluctuations becomes frozen into the CMB.
2458: It is this snapshot or fossil record we observe.
2459: Energy conservation gives a Doppler shift of 
2460: 	\be
2461: 	\left.{\delta T\over T}\right |_f -\left.{\delta T\over
2462: T}\right |_i
2463: 	=\Phi_f-\Phi_i.
2464: 	\ee
2465: The local gravitational potential 
2466: $\Phi_f$ 
2467: makes an isotropic contribution to ${\delta T/T}$ and
2468: will be left
2469: off hereafter.  
2470: By adiabaticity, $aT=$constant, it follows that
2471: 	\be
2472: 	\left.{\delta T\over T}\right |_i=-{\delta a\over a}.
2473: 	\ee
2474: Assuming flat space for simplicity, $a\propto t^{2/3}$ during matter
2475: domination.  The shift in $a$
2476: is then
2477: 	\be
2478: 	{\delta a\over a}={2\over 3}{\delta t\over t}.
2479: 	\ee
2480: The shift $\delta t$ can be understood as a shift relative to the cosmic
2481: time due to the perturbed gravitational potential.  In a gravitational
2482: potential clocks appear to run slow by
2483: 	\be
2484: 	{\delta t\over t}\sim \Phi 
2485: 	\ee
2486: from which it follows that 
2487: 	\be
2488: 	\left.{\delta T\over T}\right |_f={\Phi_i\over 3}.
2489: 	\ee
2490: This term is the surface Sachs-Wolfe effect from the time of decoupling.
2491: There is an additional contribution to the temperature fluctuation 
2492: known as the 
2493: integrated Sachs-Wolfe (ISW) effect which is accumulated as the photon
2494: traverses space through a decaying potential.  Including this latter
2495: contribution the full Sachs-Wolfe effect is
2496: 	\be
2497: 	{\delta T(\hat n)\over T}={1\over 3}\Phi(\eta_o \hat n)
2498: 	+
2499: 	2\int^{\eta_o}_{\eta_{\rm SLS}}d\eta
2500: 	\Phi^{\prime}(\eta,\vx)
2501: 	\label{sachswolfe}
2502: 	\ee
2503: where a prime denotes differentiation with respect to conformal time
2504: \cite{sw}.  In flat space with ordinary energy, the only contribution
2505: to $\delta T/T$ is from the surface Sachs-Wolfe effect
2506: (the first term in eqn.\ \ref{sachswolfe}).
2507: The ISW
2508: (the second term in eqn.\ \ref{sachswolfe}) makes a significant
2509: contribution to large-scale fluctuations if space is curved and/or
2510: when a cosmological constant dominates the energy density.
2511: In the absence of a significant cosmological constant, the ISW is
2512: important as perturbations decay along the line of sight during
2513: the curvature dominated epoch at 
2514: 	\be
2515: 	1+z\sim (1-\Omega_0)/\Omega_0
2516: 	\label{curvdom}
2517: 	\ee
2518: for underdense cosmologies.  This will prove to be important to 
2519: the issue of topology since the ISW may camouflage the conspicuous
2520: marks of topology imprinted in the surface Sachs-Wolfe effect.
2521: 
2522: 
2523: Therefore the temperature fluctuations observed today can be 
2524: predicted from the primordial fluctuations in the metric.  The shape of
2525: $\Phi(\eta,\vx)$ will depend on the geometry of space as well as some initial
2526: spectrum.
2527: The perturbed Einstein equations 
2528: give the equation of motion
2529: 	\be
2530: 	\Phi^{{\prime} {\prime}}+3{\cal H}(1+c_s^2)\Phi^{\prime} -c^2_s\nabla^2 \Phi
2531: 	+(2{\cal H}^{\prime} +(1+c_s^2){\cal H}^2)\Phi=0
2532: 	\label{feta}
2533: 	\ee
2534: where ${\cal H}=a^{\prime}/a$ and $c_s$ is the speed of sound in the 
2535: cosmological fluid.
2536: In a matter dominated era
2537: $c_s=0$.  The potential is separable and can be written
2538: $\Phi=F(\eta)\Psi(\vec x)$ with $F(\eta)$ the solution to 
2539: eqn (\ref{feta}).  The $\Psi(\vec x)$
2540: can always be expanded in terms of the eigenmodes to the Laplacian;  that is,
2541: 	\be
2542: 	{\Phi(\eta,\vx)}=F(\eta) \int d^3\vec k
2543: 	\ \hat \Phi_{\vec k}\ \psi_{\vec k}(\vec x)
2544: 	\label{elem}
2545: 	\ee
2546: with the eigenmodes $\psi_k$ satisfying
2547: 	\be
2548: 	\nabla^2\psi_{\vec k}=-k^2\psi_{\vec k}.
2549: 	\label{eigeneq}
2550: 	\ee
2551: The Laplacian $\nabla^2=g_{\mu \nu}D^\mu D^\nu$ depends on the 
2552: curvature through the covariant derivatives $D_\mu$.
2553: For now we assume all the $k$'s are continuous eigenfunctions
2554: although this will not be the case when the space is topologically 
2555: identified.
2556: The $\hat \Phi_k$ are initial amplitudes given by the statistical 
2557: profile of the initial fluctuation spectrum.
2558: All assumptions and/or predictions for the initial perturbations are contained
2559: in the $\hat \Phi_{\vec k}$.
2560: 
2561: During inflation quantum fluctuations in the field are amplified by
2562: the accelerated expansion.  These fluctuations about the ground
2563: state of the field theory are known to be Gaussian distributed.  The amplitude of the fluctuations are related to the specific inflationary model
2564: but for the most part are independent of the scale $k$.
2565: So inflation predicts a Gaussian distribution of fluctuations independent
2566: of scale.
2567: Specifically this means the $\hat \Phi_{\vec k}$ are drawn from a random 
2568: Gaussian ensemble consistent with 
2569: 	\begin{equation}
2570: 	\left\langle \hat \Phi _{\vec k}^{*}
2571: 	\hat \Phi _{\vec k}\right\rangle 
2572: 	\propto{\cal P}_\Phi (k)\delta^3 (\vec k-\vec k^{{\prime} })
2573: 	\ \ .   \label{assume}
2574: 	\end{equation}
2575: The angular bracket $<>$ denote an ensemble average and ${\cal P}_\Phi$
2576: is the predicted power spectrum.
2577: De Sitter inflation delivers a flat, Harrison-Zeldovich spectrum
2578: which corresponds to ${\cal P}_\Phi =$constant.
2579: Other forms for ${\cal P}_\Phi\propto k^{n-1}$ have been
2580: derived from specific inflationary models
2581: with the spectral index $n\ne 1$ and give tilted as
2582: opposed to flat spectra.
2583: 
2584: 
2585: 
2586: From the preceding we can fully predict $\Phi$ and
2587: therefore $\delta T/T$ given the curvature of spacetime and the initial 
2588: spectrum.  Armed with this prediction, fluctuations on a given manifold
2589: can be compared to the data.  For data comparison it is 
2590: customary to consider the correlation function.
2591: Although $\delta T(\vx)/T$ permeates space, we only observe 
2592: fluctuations from our SLS at location $\vx=\Delta \eta\hat n$
2593: with $\hat n$ a unit directional vector.  
2594: Since the fluctuations are taken to be Gaussian, the correlation function
2595: contains all of the information about the temperature fluctuations.
2596: The ensemble average
2597: correlation function between any two points on the sky is 
2598: 	\be
2599: 	C(\hn,\hnp)\,  = \,
2600: 	\left\langle 
2601: 	{\delta T\over T}(\hn) {\delta T\over T}(\hnp) \right\rangle.
2602: 	\label{corrf}
2603: 	\ee
2604: The theoretically predicted $C(\hat n,\hat n^{\prime})$
2605: can then be statistically compared to the data to estimate the likelihood
2606: a given model is responsible for the world we live in.
2607: 
2608: Because fluctuations are observed on a sphere it is customary to decompose
2609: the data into spherical harmonics,
2610: 	\be
2611: 	{\delta T\over T}(\hat n)=\sum_{\ell m}a_{\ell m} Y_{\ell m}(\hat n)
2612: 	\label{sphereharm}
2613: 	\ee
2614: where the $Y_{\ell m}$'s are the usual spherical harmonics.  The
2615: spherical harmonics form a complete set of states on the sphere and
2616: are orthogonal so that
2617: 	\be
2618: 	\int_0^{2\pi}d\phi \int_0^{\pi}
2619: 	\sin\theta d\theta Y_{\ell^\prime m^\prime}^*(\theta,\phi)Y_{\ell m}
2620: 	(\theta,\phi)=\delta_{\ell^\prime \ell}\delta_{m^\prime m}
2621: 	\ee
2622: and unit normalized.
2623: Using the orthogonality of the $Y_{\ell m}$,
2624: invert (\ref{sphereharm}) to find
2625: 	\be
2626: 	a_{\ell m}=\int d\Omega Y_{\ell -m}(\hat n){\delta T\over T}
2627: 	(\hat n)
2628: 	\ \ .\label{invert}
2629: 	\ee
2630: If the probability distribution is Gaussian, then the Fourier correlation
2631: function 
2632: 	\be
2633: 	C_\ell=\left < \left |a_{\ell m}\right |^2 \right >
2634: 	\ee
2635: contains
2636: complete information about the fluctuations.  
2637: The $<>$ again denotes an ensemble average.
2638: Either non-Gaussianity or the breaking
2639: of homogeneity and isotropy with topological identifications
2640: will mean that the
2641: $C_\ell$'s do not provide complete information.  We will
2642: emphasize this in section \S \ref{obshyp}.  
2643: 
2644: Assuming for now a simply connected topology, homogeneity and isotropy
2645: of the metric allows one to perform an angular average which is in
2646: effect an average over $m$'s without loss of information.  
2647: Such an average gives an estimator for
2648: the ensemble average angular power spectrum $C_\ell$,
2649: 	\be
2650: 	C_{\ell}=\sum_{m=-\ell}^{m=\ell}|a_{\ell m}|^2/(2l+1).
2651: 	\label{est}	
2652: 	\ee
2653: Using equation (\ref{invert}) gives
2654: 	\be
2655: 	C_{\ell}={1\over (2l+1)}\sum_m\left
2656: 	[\int d\Omega^{\prime} Y_{\ell -m}(\hat n^{\prime})
2657: 	\int d\Omega Y_{\ell m}(\hat n)
2658: 	\left <{\delta T\over T}(\hat n){\delta T\over T}(\hat n^\prime)
2659: 	\right >\right ]
2660: 	\ \ .
2661: 	\label{clong}
2662: 	\ee
2663: The parameter $\ell $ can be associated with the angular size of a given
2664: fluctuation, with $\ell\sim \pi /\theta$.  Large angle fluctuations are 
2665: associated with low $\ell$ and small angle fluctuations are associated
2666: with high $\ell $.
2667: The lower the value of $\ell$, the fewer contributions there are to the sum
2668: (\ref{est}).  
2669: As a result, the estimator of the true ensemble average is poorer
2670: for low $\ell$ than it is for high $\ell$ where there are many 
2671: contributions to the average.  This is known as cosmic variance and can
2672: be included as an error bar.  
2673: For a homogeneous, isotropic space cosmic variance can be estimated as
2674: 	\be
2675: 	C_\ell \sqrt{2/({2 \ell +1})} .\ee
2676: We only have one universe available to measure and we cannot be sure that
2677: our universe is not just a randomly large deviation from average.
2678: It is difficult to interpret the significance of low $\ell$ observations such 
2679: as the lack of power in the quadrupole as observed by \cb.
2680: The ambiguity of cosmic variance
2681: is worsened with topological identifications as emphasized in \cite{bpsII}.
2682: 
2683: 
2684: These are all the tools needed to determine $\delta T/T$ for a specific
2685: theoretical model, the correlation function $C(\hat n,\hat n^\prime)$,
2686: and the angular
2687: average power spectrum $C_\ell$.  These predictions can then 
2688: be compared with data.  
2689: When we come to the influence of topology we will see the increasing importance
2690: of using full sky maps of $\delta T/T$ and the full correlation function
2691: $C(\hat n,\hat n^\prime)$.
2692: 
2693: 
2694: We have described the scalar modes above.
2695: There are also tensor modes which are gravitational waves and vector modes 
2696: which are rotational perturbations and do not grow with time.  The
2697: scalar modes are the only ones which couple to the energy density and
2698: pressure and we restrict ourselves to scalar modes hereafter.
2699: Topology may in fact influence the gravitational wave background but
2700: to date this remains unexplored territory.
2701: 
2702: The primordial fluctuations catalyze the formation of galaxies and
2703: galaxy clusters.  So the minute quantum fluctuations are amplified
2704: into the gigantic structures we see today.  The primordial spectrum
2705: can be tested by measuring the fluctuations in the CMB but there are 
2706: other manifestations.  For instance, the eventual nonlinear growth of
2707: perturbations can be simulated numerically to test theoretical
2708: predictions for structure formation against astronomical
2709: observations.  It would be interesting to know if topological features
2710: could sculpt the distribution of structure as well \cite{lb_fractals}.
2711: 
2712: In the following subsections we find the eigenmode decomposition in 
2713: simply connected spaces.
2714: Perpetuating a cruel prejudice against $\es$ we consider only $\ah $ and
2715: $\eu $.  However, the authors of Ref.\ \cite{newsphere} catalog all of 
2716: the $\es$ topologies and discuss detection strategies based on the
2717: crystallographic methods.
2718: The motivation for neglecting $\es$ topologies in connection with 
2719: the CMB is observational as
2720: there is no outstanding evidence which supports an $\Omega >1$
2721: universe \cite{open}.  
2722: The current debate is whether $\Omega=1$ or whether $\Omega <1$.
2723: The future satellites which aim to refine measurements of the CMB intend
2724: to resolve this debate.
2725: The overriding theoretical prejudice is for $\Omega=1$ consistent with
2726: inflation.  More
2727: recently the supernova data has drummed up enthusiasm for
2728: $\Omega_\Lambda=0.7$ and $\Omega_m=0.3$ so that the total $\Omega=1$
2729: \cite{saul}.  
2730: The recent small angle experiments Boomerang and MAXIMA 
2731: corroborate these values if certain prior assumptions constrain
2732: the statistical analysis.  More quantitatively, the current numbers
2733: are $\Omega =0.90\pm 0.15$, $\Omega_b h_0^2=0.025 \pm 0.010$
2734: $\Omega_{CDM}h_0^2=0.13\pm 0.10$ and a spectral index of
2735: $n=0.99\pm 0.09$ \cite{balbi} where
2736: the subscript $b$ denotes the baryonic 
2737: contribution and the subscript $CDM$ denotes the net contribution
2738: from cold dark matter.
2739: Still, the data are clearly underdetermined
2740: and it is an unresolved issue how many different
2741: models can match the data.
2742: 
2743: In Ref.\ \cite{addr} an argument is made for
2744: a constraint on $\Omega_0$ with some model-independence.
2745: They argue that a constraint is imposed if one
2746: requires that a local maximum detected in
2747: the correlation function of large scale structure 
2748: ($\sim 100-200h^{-1}$Mpc)
2749: occurs at the same comoving positions at different redshifts.
2750: The argument is independent of both CMB and supernova Ia data, and 
2751: favors a hyperbolic universe.
2752: Their result is a cosmology with $\Omega_0=0.9\pm0.15$ (95\% confidence)
2753: (with $Omega_\Lambda$ in the vicinity of $\sim $ 0.65).
2754: 
2755: Instead of entering further into the parameter estimation debate, we will 
2756: review the recent investigations in cosmic topology which have
2757: focused on the flat and the hyperbolic manifolds.  
2758: For an interesting catalog of topologies for more general
2759: Bianchi classes see Ref.\ \cite{bark}
2760: 
2761: 
2762: \subsection{Observing the CMB}
2763: \label{obscmb}
2764: 
2765: 
2766: Many billions of years after last scattering, we build the COsmic
2767: Background Explorer (COBE) satellite to confirm the earlier detections
2768: by Penzias and Wilson of the microwave background radiation.
2769: The all sky map generated by the COBE satellite confirmed that 
2770: the average temperature was homogeneous and isotropic at 
2771: 2.728 ${}^o$ K 
2772: %\pm 0.010
2773: and that the spectrum was extremely thermal, a result
2774: that drew spontaneous applause when presented in 1992.  The data also
2775: confirmed the theoretical expectation that there are in fact minute
2776: fluctuations as predicted by inflation at the $10^{-5}$ level that
2777: appear to be scale invariant.
2778: COBE measures fluctuations on large scales, $\ell \le 30$.  Low $\ell$s
2779: correspond to 
2780: fluctuations on scales far outside
2781: the horizon at the time of decoupling and therefore can only be due to
2782: fluctuations in the metric and density. 
2783: So COBE
2784: which measures $\ell \lta 30$ is a probe of the largest geometric
2785: features and therefore is quite important as a probe of topology as
2786: well.  Causal microphysics becomes important for $\ell \gta 100$.
2787: At these large $\ell$ our calculation of 
2788: $\delta T/T$ in eqn.\ (\ref{sachswolfe})
2789: is insufficient as it neglects microphysical effects and 
2790: a more detailed analysis is required.
2791: Of particular importance are the high $\ell$ Doppler peaks.
2792: Before recombination, sound waves in the baryon-photon fluid induce
2793: additional Doppler shifts which produce a peak in power
2794: on a scale related to the size of the horizon at the time of
2795: decoupling.  The height of the peaks depends on the specific model
2796: parameters.  The location of the first peak depends primarily on curvature, 
2797: $\ell_{\rm peak} \simeq 220 \Omega_0^{-1/2}$, with some small $H_0$ dependence.
2798: \cb does not measure high $\ell$ fluctuations but two important all-sky
2799: satellites will, Microwave Anisotropy probe (MAP) and {\it Planck Surveyor}.
2800: The recent balloon borne experiments locate
2801: $\ell\sim 200$ and therefore contribute more evidence in favor of
2802: $\Omega_0\sim 1$.
2803: 
2804: The task of the Differential Microwave Radiometer (DMR) experiment on 
2805: the COBE satellite was to 
2806: measure the large angle 
2807: anisotropy of the entire sky \cite{{bennet},{smoot}}.
2808: The temperature \cb measures in a given pixel can be written
2809: 	\be
2810: 	\left (\delta T\over T\right )_i=
2811: 	\sum_{\ell m} a_{\ell m} B_\ell Y_{\ell m}(\vec x_i)
2812: 	+n_i
2813: 	\ee
2814: where $B_\ell$ is the experimental beam pattern and the noise in each pixel is 
2815: $n_i$.  The DMR horns are characterized by an imperfect Gaussian beam pattern.
2816: The noise is assumed to be uncorrelated and
2817: Gaussian with mean 
2818: $<n_i>=0$ and variance $ <n_in_j>=\sigma \delta_{ij}$.
2819: 
2820: \begin{figure}[tbp]
2821: \centerline{
2822: \psfig{file=cobe.eps,width=2.5in}
2823: } \vskip 15truept
2824: \caption{The three panels are, from top to bottom, the combined
2825: data before the dipole is subtracted, the data after dipole subtraction but
2826: before the galactic cut, the map minus both dipole and galactic emission.
2827: Taken from the COBE webpage
2828: at http://space.gsfc.nasa.gov/astro/cobe/
2829: }
2830: \label{cobepic}
2831: \end{figure}
2832: 
2833: The experiment consists of three pairs of antennae.  Each pair measures
2834: the temperature difference in two directions separated by $60^\circ$.
2835: Each antenna has a $7^{\circ}$ beam and the data is smoothed on $10^\circ$.
2836: A full sky scanned was performed several times over 4 years.
2837: The \cb data is provided in the form of 6 maps in 3 frequencies, 31,
2838: 53
2839: and 90 GHz \cite{{smoot},{bennet}}.
2840: Each map has $N_p=6144$ of size $(2.6^o)^2$.  
2841: Compressing to resolution 5 pixels, $N_p=1536$ pixels of 
2842: size $(5.2^o)^2$, loses no information and
2843: is sometimes implemented in data analysis.
2844: 
2845: The observed
2846: $C_\ell$s have been determined from the \cb data by Gorski 
2847: \cite{kris}, by
2848: Tegmark \cite{teg}, and by Bond, Jaffe and Knox
2849: \cite{jaffe}.
2850: The monopole ($\ell =0 $) is just the average temperature
2851: itself and can be discarded from the data.  
2852: The largest scale anisotropy is the dipole ($\ell=1$) generated by our solar
2853: system's peculiar velocity relative to the CMB.  Since this is not
2854: cosmological in origin, the entire dipole is discarded from the data.
2855: The first relevant cosmological observations begin with $\ell =2$,
2856: the quadrupole, and it is curious to note that the measured
2857: quadrupole is low.  This may just be cosmic variance in action but compact 
2858: manifolds do happen to predict low power on large scales.  This is discussed
2859: at length in section \S \ref{obsflat}.
2860: The six maps can be 
2861: compressed
2862: into one weighted-sum map
2863: with the monopole and dipole subtracted.  A galactic cut
2864: of a region $\pm 20^o$
2865: around the
2866: Galactic
2867: plane is also needed to eliminate Galactic emission.
2868: This finally messaged data is ready for comparison to theoretical predictions.
2869: 
2870: 
2871: The signal-to-noise of \cb is $\xi=2$.  By comparison, the projected
2872: sensitivity of MAP is $\xi=15$ with a resolution of $0.5^\circ$,
2873: a tenfold improvement in signal-to-noise and 30 in resolution.
2874: Planck has even higher resolution but will launch much later.
2875: 
2876: 
2877: \subsection{Topology and the CMB}
2878: \label{obstop}
2879: 
2880: Several aspects of the fluctuation spectrum are altered when the 
2881: manifold is multiconnected.  The most conspicuous and consistent
2882: signatures of topological identifications are as follows:
2883: (1)  Multiconnectedness destroys
2884: global isotropy for all but the projective space and destroys global
2885: homogeneity for all but the the projective space and the hypertorus.
2886: (2)  The spectrum of fluctuations is discrete reflecting the natural 
2887: harmonics of the finite space.
2888: (3)  Since the fluctuations must fit within
2889: the finite space, there is a cutoff in perturbations with wavelengths 
2890: which exceeds the topological scale in a given direction.
2891: (4)  Geometric patterns are encrypted in the spatial correlations.  The
2892: patterns reflect the repeated occurrence of topologically 
2893: lensed hot and cold spots.
2894: %The most interesting patterns are the 
2895: %emergence of correlated circles in the sky as predicted in Ref. \cite{css1}.
2896: 
2897: 
2898: The proposed methods of scanning the CMB for
2899: evidence of topology can be split into direct
2900: statistical methods and geometric methods.  The direct methods
2901: begin with a theoretical model, compute the temperature fluctuations
2902: as was done in \S \ref{simpflat} and \S \ref{simphyp} for the simply
2903: connected spaces, and determine the likelihood of the model parameters
2904: against the data.  There are shortcomings with this brute force approach.
2905: All conclusions are model dependent and there are an infinite number of
2906: models.  The model is not just the manifold but also the orientation and
2907: the location of the Earth and the initial primordial spectrum.  While
2908: some conclusions can still be drawn there is an appeal to a model independent
2909: attempt at observing topology.  That is where the geometric methods come in.
2910: The geometric methods treat topological lensing much like gravitational
2911: lensing.  One just observes the lensed images and their distribution to 
2912: reconstruct the geometry of the intervening lens.  The most amusing example
2913: is the correlated circles of \S \ref{obscirc} although other patterns
2914: can emerge as well as described in \S \ref{obspat}.
2915: 
2916: Before preceding to review the known approaches to observing topology 
2917: it is worth expanding on point (1) above.
2918: The global anisotropy and inhomogeneity means that the correlation function
2919: depends on the location of the observer and the orientation of the manifold.
2920: Additionally, $C(\hat n,\hat n^\prime)$ depends on both $\hat n$ and 
2921: $\hat n^\prime$ and not just the angle between them.  As a result,
2922: the angular average performed in the definition of the multipole moments
2923: $C_\ell$ discards important topological information.
2924: The information lost in the angular average 
2925: can be quantified
2926: by an enhanced cosmic variance $\left <C_\ell^2 \right>$ 
2927: \cite{bpsII}.
2928: The correlation function can be decomposed into isotropic and 
2929: anisotropic pieces,
2930: 	\be
2931: 	C(\hat n,\hat n^\prime)=C^I(\hat n,\hat n^\prime)
2932: 	+C^A(\hat n, \hat n^\prime).
2933: 	\ee
2934: By isotropy
2935: 	\be
2936: 	C^I(\hat n,\hat n^\prime)=\sum_{\ell} {\ell +1/2\over
2937: 	\ell(\ell +1)}C_\ell P_\ell(\hat n\cdot \hat n^\prime)
2938: 	\ee
2939: and by anisotropy
2940: 	\be
2941: 	C^A(\hat n,\hat n^\prime)=\int d\Omega_{\hat n}
2942: 	\int d\Omega_{\hat n^\prime} C^A(\hat n, \hat
2943: 	n^\prime)P_\ell(\hat n \cdot \hat n^\prime)=0
2944: 	\ee
2945: so that the anisotropic piece is orthogonal to the Legendre
2946: polynomials.
2947: The expectation value of the estimator for the $C_\ell$ depends
2948: solely on the isotropic piece by construction 
2949: 	\be
2950: 	\left < \tilde C_\ell\right >={\ell(\ell+1)\over 8\pi^2}
2951: 	\int  d\Omega_{\hat n} \int  d\Omega_{\hat n^\prime} 
2952: 	C(\hat n, \hat n^\prime)P_\ell(\hat n\cdot \hat n^\prime),
2953: 	\ee
2954: but the variance
2955: 	\be
2956: 	{\rm var}(\tilde C_\ell)=\left < \tilde C_\ell^2\right >-
2957: 	\left < \tilde C_\ell\right >^2 
2958: 	\ee
2959: contains anisotropic pieces.
2960: This can be interpreted as very large error bars due to cosmic 
2961: variance.  As a result, conclusions based on the $C_\ell$'s alone
2962: are weak \cite{{bpsI},{bpsII}}.  
2963: Many of the statistical analyses described below do only
2964: examine the $C_\ell$s and so can only rule out a topology but never 
2965: really confirm topology.
2966: 
2967: However, others have argued that this increased error is not very
2968: large for compact hyperbolic spaces.  An argument by Inoue for instance
2969: expresses the variance as the sum of a geometric variance and the usual cosmic
2970: variance.  The geometric variance contains the additional variance due
2971: to topology and is intrinsically small \cite{inoueprog}.  The smallness of
2972: the geometric variance can be traced to the randomness of the 
2973: eigenmodes on compact hyperbolic models (explored further in \S
2974: \ref{numersol}) which generates what Inoue refers to as
2975: geometric Gaussianity \cite{ktinoue}.
2976: 
2977: \newpage
2978: \section{Observing flat topologies in the CMB}
2979: \label{obsflat}
2980: 
2981: Since the flat spaces are most easily constrained
2982: using the \cb data we begin with them and discuss hyperbolic manifolds
2983: in all their glory separately.
2984: While small universes can be ruled out, it is rather fascinating to 
2985: note that 
2986: large cases are marginally consistent with the data.
2987: After all, the observed quadrupole is low and 
2988: so the infrared truncation in power seen in flat spaces and described
2989: below can actually create a better fit to the data.
2990: However, such a marginal flat space, just 
2991: coinciding with the size of the observable universe is unaesthetic 
2992: at best.  In fact, there is no natural scale for the size of a flat
2993: universe and they are not the favored small universe candidates for this 
2994: reason.  Still, they provide an important and accessible testing ground
2995: for methods of observation and we discuss them next.
2996: 
2997: 
2998: \subsection{Direct methods in flat space}
2999: \label{directflat}
3000: 
3001: 
3002: \subsubsection{Simply connected flat space}
3003: \label{simpflat}
3004: 
3005: In flat, simply connected space, the spectrum of fluctuations is well known.
3006: We need to find the elements of eqn. (\ref{elem}) with $F(\eta)$ 
3007: the solution to eqn. (\ref{feta}).
3008: In flat space ${\cal H}=2/\eta$ 
3009: and the solution to eqn (\ref{feta}) decays as
3010: $\Phi^\prime\propto \eta^{-6}$.
3011: Consequently,
3012: in flat space $F(\eta)$ is effectively constant and there is no ISW effect
3013: during radiation or matter domination.  (There is however an ISW effect
3014: if the universe is dominated by a cosmological constant.  This may turn
3015: out to be important in salvaging finite flat models if the universe is 
3016: accelerating as the recent supernovae observations indicate \cite{saul}.)
3017: The Laplacian in flat space is simply
3018: 	\be
3019: 	\left ({\partial^2_x}+{\partial^2_y}+{\partial^2_z}\right )
3020: 	\psi_{\vec k}= -k^2 \psi_{\vec k}
3021: 	\ee
3022: with solutions 
3023: 	\be
3024: 	\psi_{\vec k}=
3025: 	\exp\left (i{\vec k}\cdot \vx 
3026: 	\right ).
3027: 	\ee
3028: The potential of eqn.\ (\ref{elem}) can then be expanded in terms of these
3029: as 
3030: 	\be
3031: 	{\Phi(\vx)}=\int_{-\infty}^{\infty} d^3 k
3032: 	\hat \Phi_{\vec k}\exp\left (i{\vec k}\cdot \vx 
3033: 	\right )
3034: 	\ee
3035: All of the assumptions about the statistics and shape of the spectrum are
3036: contained in the $\hat \Phi_{\vec k}$.  All other quantities are 
3037: determined by the geometry of the space.  
3038: In accordance with the 
3039: inflationary prediction the $\hat \Phi_{\vec k}$ are
3040: assigned an independent Gaussian probability distribution 
3041: consistent with the flat space normalization
3042: 	\begin{equation}
3043: 	\left\langle \hat \Phi _{\vec k}
3044: 	\hat \Phi^* _{\vec k^\prime}\right\rangle 
3045: 	={\frac{2\pi^2}
3046: 	{k^3}}{\cal P}_\Phi (k)\delta^3 (\vec k-\vec k^{\prime })
3047: 	\ \    \label{assumef}
3048: 	\end{equation}
3049: \cite{mkb}.
3050: 
3051: From eqn. (\ref{sachswolfe}) the Sachs-Wolfe
3052: effect is simply
3053: 	\be
3054: 	{\delta T(\hat n)\over T}={\Phi(\hat n)\over 3}.
3055: 	\ee
3056: 
3057: The correlation function between any two points on the SLS is then, from 
3058: eqn (\ref{corrf}), 
3059: 	\be
3060: 	C(\hn,\hnp)\propto \int
3061: 	{d^3 k\over k^3}
3062: 	{\cal P}_\Phi
3063: 	       \exp\left(i\Delta \eta \vec k \cdot (\hn - \hnp ) 
3064: 	\right )
3065: 	\label{eq:cnn}
3066: 	\ee
3067: up to a normalization.
3068: For a homogeneous and isotropic space, 
3069: the correlation function depends only on the angular
3070: separation between $\hn$ and $\hnp$.  Consequently all the information on 
3071: the theoretical sky is in the angular average.
3072: Using the orthogonality relations of the
3073: Legendre polynomials in eqn (\ref{clong}) gives
3074: 	\ba
3075: 	C_\ell 
3076: 	&= &
3077: 	\frac{1}{4 \pi} \int d \Omega \int d \Omega' C(\hn, \hnp) P_\ell(\mu)
3078: 	\nonumber \\
3079: 	&=& {4\pi\over 25}\int_0^{\infty}{dk\over k}j^2_l(\Delta \eta k)
3080: 	{\cal P}_\Phi.
3081: 	\ea
3082: %where $\mu = \hn \cdot \hnp$.
3083: If we assume the initial fluctuation is a powerlaw
3084: ${\cal P}_\Phi\propto k^{n-1}$, this can be integrated to give
3085: \ba
3086: C_\ell \propto {\Gamma(\ell +(n-1)/2)\over\Gamma(\ell+(5-n)/2)}
3087: {\Gamma((9-n)/2)\over \Gamma((3+n)/2)}
3088: \label{clflat}
3089: \ea
3090: \cite{bondef}.
3091: 
3092: As described in 
3093: \S \ref{obscmb}, 
3094: the \cb data has been analyzed over the years by many groups to 
3095: determine the $C_{\ell}$'s observed.  Statistical likelihood
3096: analyses are then performed to compare the theoretical model
3097: with the \cb data.  The infinite, flat model is generally found
3098: to be consistent with the data for a nearly flat power spectrum
3099: ($n\sim 1$).
3100: The question addressed in the next section, is how well finite
3101: flat models match the data.  To flash forward, the answer is essentially
3102: that finite flat models match the data well if they are comparable in size
3103: to half the observable universe although some studies have put 
3104: even stronger limits as described below.  All limits are subject
3105: to caveats based on the
3106: assumptions made about the perturbation
3107: spectrum -- none of which have yet been tested observationally. 
3108: 
3109: \subsubsection{Compact, flat spaces}
3110: \label{compflat}
3111: 
3112: We have already reviewed the standard decomposition of the temperature
3113: fluctuations in a simply connected space.  When the manifold is compact and
3114: the SLS exceeds the dimensions of the space, then the global topology 
3115: is reflected in the CMB sky.
3116: There are three notable alterations 
3117: to the predicted fluctuations when
3118: the manifold is compact:  (1)  The eigenvalue spectrum is discrete not
3119: continuous.  (2)  There is a cutoff in the power of fluctuations on wavelengths
3120: which exceed the natural size of the space.
3121: (3)  The correlation function $C(\hn,\hnp)$ depends on orientation and
3122: so depends on both $\hn$ and 
3123: $\hnp$ explicitly.  It is not simply a function of the angular separation.
3124: 
3125: 
3126: 
3127: The exact eigenmodes can be found for all of the 6 compact orientable 
3128: flat spaces \cite{{lss},{slsi}}.  This allows a direct statistical 
3129: comparison of the theoretical predictions for compact, orientable spaces 
3130: with the \cb data.  With any statistical comparison of a model with the 
3131: data, the conclusions are model dependent.
3132: The reader should bear in mind that the bounds quoted assume equal-sided 
3133: spaces, $\Omega_\Lambda=0$, and a flat Gaussian distributed
3134: power spectrum.  
3135: 
3136: The earliest bounds on the hypertorus constrained the topology scale to be
3137: $\gta 0.8\Delta \eta$ \cite{sss} which is still less than the diameter of the
3138: SLS.
3139: There could still be as many as eight
3140: copies of our universe within the observable horizon.
3141: The analysis was later extended to the other 
3142: compact, orientable flat spaces where similar bounds were obtained 
3143: \cite{{lss},{slsi}}.  Stronger bounds were put on the hypertorus by comparing
3144: the full covariance matrix against the 2-year \cb data \cite{deO1}.
3145: The length of a side was set to be $\gta 1.2 \Delta \eta $.
3146: 
3147: 
3148: The most conspicuous feature in the fluctuation
3149: spectrum is a suppression of power
3150: on large scales since large fluctuations cannot fit into the finite box.
3151: Such an infrared cutoff can be deduced from Cheeger's inequality \cite{cheeger}
3152: 	\be
3153: 	k_{min}\ge {h_C\over 2},
3154: 	\quad \quad h_C=\inf_X{A(S)\over {\rm min}\left (V(M_1),V(M_2)\right )}
3155: 	,
3156: 	\ee
3157: with the infimum taken over all possible surfaces $S$ that divide the space
3158: into two subspaces ${\cal M}_1$ and ${\cal M}_2$ where 
3159: ${\cal M}={\cal M}_1\cap {\cal M}_2$.  $S$ is the boundary of the two subspaces,
3160: $S=\partial {\cal M}_1=
3161: \partial {\cal M}_2$. 
3162: The isoperimetric constant $h_C$ depends on the geometry more than 
3163: the topology.  Intuitively speaking, in very long thin manifolds
3164: $h_C$ can be quite small leading to a lowered spectrum of eigenvalues.
3165: However, this is highly correlated with thin bottleneck structures and is not
3166: a common feature of spaces with a more regular shape \cite{{cheeger},{buser}}. 
3167: From Cheeger's inequality it follows that
3168: all flat hypertori have $k_{min}\ge 2/L$ with $L$ the longest side of the 
3169: torus
3170: as argued in \cite{bpsI}.  
3171: 
3172: Although the cutoff would seem a good indicator of topology, it just so happens
3173: that
3174: the longest wavelength
3175: fluctuation observed, namely the quadrupole, is in fact low.
3176: Some might even take this as evidence for topology \cite{workshop}.
3177: Cosmic variance is also large on large scales.
3178: Consequently, a fundamental domain the size of the observable universe
3179: is actually consistent with the observed \cb $C_\ell$s \cite{lss}.  
3180: However, since the fundamental domain has a particular orientation on the sky, 
3181: the correlation is not simply a function of the angular separation
3182: between $\hn$ and $\hnp$ as it is in the infinite case.  
3183: Therefore conclusions based on the $C_\ell$s alone are weaker than a 
3184: statistical analysis based on the full $C(\hat n, \hat n^\prime)$.
3185: The strongest bounds on the hypertorus were placed
3186: by comparing the full 
3187: correlation function $C(\hat n, \hat n^\prime)$ to the data.
3188: It was found that $L\gta 1.3 \Delta \eta$ \cite{{deO1},{bpsI},{bpsII}}, which 
3189: is still less than the diameter of the observable universe, 
3190: $\sim 2\Delta \eta$.
3191: Asymmetric spaces were constrained by de Oliviera-Costa et. al. as described
3192: in section \S \ref{geomflat}.
3193: 
3194: %\noindent{\bf Compact Flat Space Spectra}
3195: 
3196: Despite the bounds placed on the equal-sided
3197: compact, flat spaces, they still provide an excellent testing ground for 
3198: geometric measures of topology.  They may also be saved by 
3199: $\Omega_\Lambda\ne 0$ models.  We compile a list of
3200: all the eigenmodes, eigenvalues and relations for the compact flat spaces
3201: for completeness.  
3202: These solutions are taken from Ref. \cite{slsi}.  The spectra were also
3203: found in Ref. \cite{sss} for the other compact spaces, however, the critical
3204: relations were overlooked leading to confusion about the cutoff in the
3205: long wavelength modes.  It was mistakenly concluded because of the missing relations that in the twisted spaces longer
3206: wavelengths could fit in the fundamental domain
3207: since a wave needs to wrap more than once before coming back to 
3208: a fully periodic identification \cite{css2}.
3209: However it was shown in Ref. \cite{slsi}, that these long modes are 
3210: forbidden 
3211: by the relations among the $\hat \Phi_{\vec k}$, and all of the compact 
3212: topologies have roughly the same long wavelength cutoff.
3213: 
3214: As a result of the global topology, 
3215: all of these spaces are anisotropic and all except for the 
3216: hypertorus are inhomogeneous.  Topology can be implemented by
3217: imposing the boundary conditions
3218: 	\be
3219: 	\Phi_{\vec k}(\vx)=\Phi_{\vec k}(g\vx) \quad \forall g\in\Gamma
3220: 	.
3221: 	\ee
3222: Compact topology 
3223: always restricts the eigenvalues to a discrete
3224: spectrum:
3225: 	\be
3226: 	\Phi(\vec x)=\sum_{\vec k}\hat \Phi_{\vec k} e^{i\vec k\cdot \vec x}
3227: 	\ \ .	
3228: 	\label{eigensum}
3229: 	\ee
3230: In additional, relations
3231: are imposed on the $\hat \Phi_{\vec k}$.  
3232: 
3233: 
3234: \begin{figure}
3235: %\centerline{\psfig{file=../../top/flat/paper1/para.eps,width=1.5in}}
3236: \centerline{\psfig{file=para.eps,width=1.5in}}
3237: \caption{Tiling flat space with parallelepipeds.}
3238: \label{para}
3239: \end{figure}
3240: 
3241: 
3242: Demonstrating the discretization 
3243: explicitly for the hypertorus, the identifications of the 
3244: cube are expressed in the three boundary conditions 
3245: $\Phi(x,y,z)=\Phi(x+L_x,y,z)=\Phi(x,y+L_y,z)=\Phi(x,y,z+L_z)$.
3246: Imposing the first boundary condition on $\Phi(\vec x)$ of 
3247: eqn.\ (\ref{eigensum}) gives
3248: $e^{-ik_xx}=e^{-ik_x(x+h)}$ which requires $k_x={2\pi\over l_x}n_x$ with 
3249: $n_x$ an integer.  The other two boundary conditions provide the discrete
3250: spectrum
3251: 	\be
3252: 	k_x={2\pi \over l_x} n_x\quad \quad
3253: 	k_y={2\pi \over l_y} n_y\quad \quad
3254: 	k_z={2\pi \over l_z} n_z
3255: 	\ \ 
3256: 	\ee
3257: with the $n_i$ running over all integers \cite{{zel73},{fanghoujun},
3258: {sss}}.
3259: It is clear that there is a minimum eigenvalue and hence a maximum wavelength
3260: which can fit inside the fundamental domain defined by the parallelepiped
3261: \cite{sss}:
3262: 	\[
3263: 	k_{\rm min}=2\pi \ {\rm min}\left ( {1\over L_i}
3264: 	\right )\quad 
3265: 	{\lambda}_{\rm max} ={\rm max}\left (L_i\right)\  .
3266: 	\nonumber
3267: 	\]
3268: Although global anisotropy is broken by topology, we can still form
3269: the $C_\ell$s to make a comparison, if incomplete, with the \cb 
3270: $C_\ell$s.
3271: Expanding the exponential and 
3272: Legendre polynomials in terms of spherical harmonics, the 
3273: $C_\ell$ eqn.\ (\ref{clong}) becomes
3274: \be
3275: C_\ell \propto
3276: \sum_{\vec k} \sum_{\vec k'}
3277: {{\cal P}_{\vec k}\over k^3}
3278:  j_\ell^2(\Delta \eta k) 
3279: \label{eq:cl}
3280: \ee
3281: Three other spacetimes are constructed from a
3282: parallelepiped.
3283: The first twisted parallelepiped 
3284: has opposite faces identified with one pair of faces
3285: twisted by $\pi $ before gluing.
3286: The eigenmodes are 
3287: $\vec k=2\pi(j / h, w/b,n/2c)$, with the additional relation 
3288: $\hat \Phi_{jwn}=\hat \Phi_{-j -w n}\ e^{i {\pi}n}$.  The minimum 
3289: mode consistent with the relations is $k_{\rm min}=2\pi/L$.  The
3290: $C_\ell$s become
3291: 	\be
3292: C_\ell \propto
3293: \sum_{jwn} 
3294: \frac{{\cal P}(k)}{k^3} 
3295: \frac{j_\ell(\Delta \eta k)^2}
3296: {2 \ell + 1} 
3297: \sum_{m = -\ell}^{\ell}
3298: |Y_{\ell,m}({\bf \hat k})|^2
3299: ( 1 + e^{i \pi (n+m)}).
3300: 	\ee
3301: 
3302: A
3303: parallelepiped with 
3304: one face rotated by $\pi/2$ has
3305: discrete eigenmodes 
3306: $\vec k=2\pi(j / h, w/b,n/4c)$, with the additional relations
3307: $\hat \Phi_{jwn}=\hat \Phi_{w -j n}e^{in\pi/2}
3308: =\hat \Phi_{-w -jn}e^{in\pi}=\hat \Phi_{-w j n}e^{i 3n\pi/2}$.
3309: Again the minimum mode is $k_{\rm min}=2\pi/L$.
3310: The $C_\ell$s are given by
3311: \be
3312: C_\ell  \propto
3313: \sum_{jwn} 
3314: \frac{{\cal P}(k)}{k^3} 
3315: \frac{j_\ell(\Delta \eta k)^2}
3316: {2 \ell + 1} \sum_{m = -\ell}^{\ell}
3317: |Y_{\ell,m}({\bf \hat k})|^2 
3318: ( 1 + e^{i (n+m){\pi}/2} +
3319: e^{i (n+m){\pi}} + e^{i 3(n+m){\pi}/2}).
3320: \ee
3321: 
3322: The last parallelepiped 
3323: has a fundamental domain of volume $2 h b c$ and is thus
3324: a double parallelepiped.
3325: The identification rules involve three rotations through
3326: $\pi$ as shown in \cite{wolf}.
3327: They are
3328: so that $(x,y,z)\rightarrow (x+h,-y,-z)$, 
3329: $(x,y,z)\rightarrow (-x,y+b,-(z+c))$, and 
3330: $(x,y,z)\rightarrow (-(x+h),-(y+b),z+c)$. 
3331: The resultant discrete spectrum is
3332: $\vec k=\pi(j / h, w/b,n/4c)$ with relations
3333: $\hat \Phi_{jwn}
3334: =\hat \Phi_{j-w -n}\ e^{i\pi j}
3335: =\hat \Phi_{-jw -n}\ e^{i\pi(w+n)}
3336: =\hat \Phi_{-j-w n}\ e^{i\pi(j+w+n)}$.
3337: The minimum mode is $k_{\rm min}=\sqrt{2}\pi/L$.
3338: Since the volume of the space is $2L^3$, the
3339: angular averaged long wavelength
3340: is $\sqrt{2}/2^{1/3}\simeq 1.1 $ times the length of the fundamental domain.
3341: The angular power spectrum is
3342: \ba
3343: C_\ell & \propto&
3344: \sum_{jwn} 
3345: \frac{{\cal P}(k)}{k^3} 
3346: \frac{j_\ell(\Delta \eta k)^2}
3347: {2 \ell + 1} \sum_{m = -\ell}^{\ell}
3348: Y_{\ell,m}^*({\bf \hat k}) \times \\ \nonumber
3349: &\big(&
3350: Y_{\ell,m}({\bf \hat k}) (1 + e^{i (m+j+w+n)\pi}) 
3351: + Y_{\ell,-m}({\bf \hat k}) e^{i \ell \pi}
3352: (e^{i (m+j)\pi} + e^{i (w+n)\pi}) \big).
3353: \ea
3354: 
3355: 
3356: \begin{figure}
3357: %\centerline{\psfig{file=../../top/flat/paper1/hex.eps,width=1.5in}}
3358: \centerline{\psfig{file=hex.eps,width=1.5in}}
3359: \caption{Tiling flat space with hexagonal prisms.}
3360: \label{hexag}
3361: \end{figure}
3362: 
3363: 
3364: The last two possible compact flat spaces are based on a
3365: hexagonal tiling. With
3366: the opposite sides of the hexagon identified and the prism faces
3367: rotated relative to each other by 
3368: $2\pi/3$, the potential can be written 
3369: 	\be
3370: 	\Phi =\sum_{n_2 n_3n_z} \hat \Phi_{n_2 n_3 n_z}
3371: 	e^{ik_z z}	 
3372: 	\exp {\left [
3373: 	i{2\pi \over h}\left [
3374: 	n_2\left ( x -{1\over \sqrt{3}} y \right )
3375: 	+n_3\left ( x +{1\over \sqrt{3}} y \right )\right ] \right ]}
3376: 	\label{hexmodes}
3377: 	\ee
3378: with the eigenmodes
3379: $\vec k=2\pi((n_2+n_3) / h, (-n_2+n_3)/b,n_z/3c)$.
3380: The relations on this space are
3381: $\hat \Phi_{n_2, n_3, n_z}
3382: =\hat \Phi_{n_3, -(n_2+n_3), n_z} e^{i2\pi n_z/3}
3383: =\hat \Phi_{-(n_2+n_3), n_2, n_z} e^{i4\pi n_z/3}$
3384: and lead to
3385: \be
3386: C_\ell  \propto
3387: \sum_{n_2n_3n_z} 
3388: \frac{{\cal P}(k)}{k^3} 
3389: \frac{j_\ell(\Delta \eta k)^2}
3390: {2 \ell + 1} \sum_{m = -\ell}^{\ell}
3391: |Y_{\ell,m}({\bf \hat k})|^2 
3392:  ( 1 + e^{i 2(n_z+m){\pi}/3} + e^{i 4(n_z+m){\pi}/3}).
3393: \ee
3394: 
3395: Lastly, the prism faces are glued
3396: after rotation by $\pi/3$.
3397: The potential can still be written as (\ref{hexmodes}), with eigenmodes
3398: $\vec k=2\pi((n_2+n_3) / h, (-n_2+n_3)/b,n_z/6c)$
3399: and a set of relations 
3400: $\hat \Phi_{n_2, n_3, n_z}
3401: =\hat \Phi_{(n_2+n_3), -n_2, n_z} e^{i\pi n_z/3}
3402: = \hat \Phi_{n_3, -(n_2+n_3), n_z} e^{2i\pi n_z/3}
3403: = \hat \Phi_{-n_2,-n_3, n_z} e^{i\pi n_z}
3404: = \hat \Phi_{-(n_2+n_3), n_2, n_z} e^{i4\pi n_z/3}
3405: = \hat \Phi_{-n_3, (n_2+n_3), n_z} e^{i5\pi n_z/3}$.
3406: The $C_\ell$s are given by
3407: \ba
3408: C_\ell & \propto&
3409: \sum_{n_2n_3n_z} 
3410: \frac{{\cal P}(k)}{k^3} 
3411: \frac{j_\ell(\Delta \eta k)^2}
3412: {2 \ell + 1} \sum_{m = -\ell}^{\ell}
3413: |Y_{\ell,m}({\bf \hat k})|^2 \\ \nonumber
3414: &(& 1 + e^{i (n_z+m){\pi}/3} + e^{i 2(n_z+m){\pi}/3} 
3415: + e^{i (n_z+m){\pi}} + e^{i 4 (n_z+m){\pi}/3} 
3416: + e^{i 5 (n_z+m){\pi}/3}).
3417: \ea
3418: The volume of both of these topologies is $h^2 c \frac{\sqrt{3}}{2}$.
3419: 
3420: In summary, both the $\pi$-twisted and 
3421: $\pi/2$-twisted tori have $C_\ell$s that are almost identical to that of the
3422: torus.  
3423: The harmonics do show distinctions 
3424: between the parallelepiped topologies, but unfortunately
3425: the variations fall well within cosmic variances.
3426: In addition to the damping at low $\ell$, harmonics of the discrete 
3427: spectrum create both dips and enhancements extending up to high values of 
3428: $\ell$.  The dips are a natural consequence of the 
3429: discretization and the resultant absence of certain modes in the 
3430: spectrum.  The enhancements are due to the distribution 
3431: of multiple images dictated by the geometry of the fundamental domain.
3432: 
3433: As a result of the low observed quadrupole, 
3434: the predicted cutoff alone is not enough to rule out compact, flat models.
3435: The likelihood of the $C_\ell$ in the full \cb range was 
3436: compared to the relative likelihood of a flat, infinite cosmology in 
3437: Ref. \cite{{sss},{lss},{slsi}}.  
3438: Compact, flat spaces are 
3439: tens of times less likely than their infinite counterparts if the topology scale exceeds about half
3440: the diameter of the SLS.  
3441: Since compact topologies do not give isotropic temperature
3442: fluctuations, this lends ambiguity to any likelihood analysis.  The 
3443: conclusions drawn were quite conservative, ruling out $\ell \gta 0.8 \Delta 
3444: \eta$; that is, the length of a side must be $\gta 0.4$ the diameter
3445: of the observable universe.  Again, it should be emphasized that these 
3446: conclusions are contingent on the correctness of the
3447: assumptions about the initial perturbation spectrum.
3448: 
3449: While the angular power spectrum
3450: is sufficient to constrain symmetric, flat topology it is in general a poor
3451: discriminant. The average over the sky fails to recognize the strong
3452: inhomogeneity and anisotropy manifest in these cosmologies. 
3453: Direct attacks on anisotropic spaces inspires more geometric approaches
3454: as discussed next.
3455: 
3456: 
3457: \subsection{Geometric methods in flat space}
3458: \label{geomflat}
3459: 
3460: %For $T^3$ they performed a comparison of the $C_\ell$ to the 
3461: %data to bound the cell size assuming relative independence on the
3462: %cell orientation for a $20^\circ$ Galaxy cut \cite{deO1}. 
3463: %Asymmetric toroidal universes were ruled constrained in Ref. \cite{deO2}.
3464: %Using the 2 year \cb data set they found that rectangular spaces
3465: %must have topology scales in excess of $3000 h^{-1}$ Mpc at the
3466: %95 \% confidence level.  The rectangles considered had one dimension 
3467: %effectively infinite and either both of the other dimensions compact
3468: %$T^2$ or just one of the other dimensions compact $T^2$.  
3469: 
3470: One of the first geometric, or pattern driven searches for topology was 
3471: initiated in Ref. \cite{deO2}.  For an anisotropic hypertorus,
3472: a symmetry plane or a symmetry axis can be identified in the CMB
3473: \cite{{sokolov},{star},{fang}}.
3474: For these asymmetric spaces, 
3475: a search for patterns was emphasized in Ref. \cite{deO2}.
3476: They develop a statistic that is independent of cell orientations but is
3477: sensitive to the plane and axis symmetries of the rectangular spaces.
3478: For
3479: anisotropic models there is a stronger dependence on the orientation of the
3480: manifold relative to the Galaxy cut.  
3481: The statistic $S(\hat n_i)$ searches for reflection symmetries in a plane
3482: perpendicular to $\hat n_i$ and is defined by
3483: 	\be
3484: 	S(\hat n_i)={1\over N_{pix}}
3485: 	\sum_{j=1}^{N_{pix}}{\left [{\delta T(\hat n_j)\over T}
3486: 	-{\delta T(\hat n_{ij})\over T}\right ]^2\over
3487: 	\sigma^2(\hat n_j)+\sigma^2(\hat n_{ij})}.
3488: 	\ee
3489: $N_{pix}$ is the number of pixels after the Galaxy cut and 
3490: $\sigma (\hat n)$ is the r.m.s. error associated with the pixels
3491: in the direction $\hat n$.  The object
3492: $\hat n_{ij}$
3493: is the reflection of $\hat n_j$ in the plane with normal $\hat n_i$,
3494: 	\be
3495: 	\hat n_{ij}=\hat n_j-2(\hat n_i\cdot \hat n_j)\hat n_i.
3496: 	\ee
3497: Lower values of $S(\hat n_i)$ corresponds to a higher degree of symmetry.
3498: The temperature fluctuation is computed using the eigenmode expansion and 
3499: a flat spectrum.  The resultant $\delta T/T$ depends on 
3500: six parameters: the three spatial orientations fixing
3501: the domain orientation and the three topology scales.
3502: 
3503: The smaller a given direction the more extreme the asymmetry.
3504: In $T^1$ for instance, if the $z$ direction is extremely small,
3505: then the size of fluctuations will be notably smaller in this direction
3506: leading to a map where there is essentially less and less structure
3507: at \cb resolution in this one direction and the structure will be drawn out
3508: on the $(x,y)$ plane.  Similarly for $T^2$ there will essentially only be
3509: structure along the large direction and so the pattern will appear to be rings
3510: of fluctuations along the small directions.
3511: 
3512: %The distribution of their statistic does not appear to depend much on the
3513: %Galaxy cut.
3514: 
3515: Using this statistic, $T^1$ and $T^2$ models were constrained to be greater
3516: than half the radius of the SLS
3517: %$0.5 \Delta \eta$ 
3518: in their small dimensions.  Subsequently, the circle method was
3519: used to study specific asymmetric models \cite{boudcirc}  where again
3520: pessimistic conclusions were reached
3521: as discussed in \S \ref{obscirc}.
3522: 
3523: \newpage
3524: \section{Observing hyperbolic topologies in the CMB}
3525: \label{obshyp}
3526: 
3527: 
3528: 
3529: %\centerline{\bf The Eigenvalue Problem}
3530: 
3531: 
3532: The eigenmode decomposition is well known on simply connected $\ah$ 
3533: and is given in \S \ref{fluctcmb}.  
3534: However, when hyperbolic space is fully compact
3535: the decomposition becomes intractable.  This is a stronger statement
3536: than simply saying the eigenmodes and eigenvalues are {\it difficult}
3537: to find.  It is actually formally impossible to write down the eigenmodes
3538: analytically.  The boundary conditions are so intricate they resist
3539: decomposition \cite{bv}.
3540: 
3541: The absence of an analytic solution to the eigenmode spectrum can be
3542: directly related to the incipient chaos on compact manifolds.
3543: On simply connected $\ah$ geodesics show the first critical ingredient
3544: for the onset of chaos; that is, extreme sensitivity to initial conditions.
3545: The geodesic deviation equation shows that two nearby trajectories 
3546: diverge away from each other exponentially quickly:
3547: 	\be
3548: 	{D^2\zeta^\mu\over ds^2}=-R^\mu_{\alpha\beta \gamma}u^\alpha
3549: 	\zeta^\beta u^\gamma
3550: 	\ee
3551: where $u^\mu=dx^\mu/ds$ and $\zeta^\mu$ is the separation of neighboring
3552: geodesics.   Normal to the geodesic flow this becomes \cite{et}
3553: 	\be
3554: 	{d^2||\zeta_N^\mu\zeta_{N\mu}||\over ds^2}
3555: 	=-2\kappa||\zeta_N^\mu\zeta_{N\mu}||
3556: 	\ee
3557: where $\kappa$ is the curvature.  If $\kappa<0$, these geodesics diverge
3558: exponentially and a coordinate invariant Lyupanov exponent can be 
3559: interpreted as $\lambda=\sqrt{|\kappa|}$.
3560: Still, there is no chaotic motion on infinite $\ah$ since there
3561: is no mixing and no folding of trajectories.  In other words, there is 
3562: no loss of predictability as a result of the exponential deviation.
3563: By contrast, when the space is made topologically compact, the mixing
3564: and folding of trajectories is assured and the flows are well known to 
3565: be fully chaotic.
3566: An entropic measure of the chaotic flow
3567: can be related to the volume of the manifold through 
3568: the Kolmogorov entropy $S_K\propto {\cal V}^{-1/3}$
3569: where ${\cal V}$ is the volume of the spacetime \cite{sinai}.  
3570: Notice that if the space is infinite
3571: this entropy vanishes.
3572: 
3573: 
3574: Chaotic flows on compact 2-dimensional manifolds in particular 
3575: have been studied at great length \cite{{bv},{aurichsteiner},{gutz},
3576: {ott}}.  The cosmological implications of chaos have only been touched 
3577: upon \cite{{css3},{lb_fractals}} and remain a largely unexplored terrain.
3578: By and large, people have tried to obviate the chaotic
3579: flows entirely when analyzing the CMB.  Methods include
3580: brute force numerical determination of the modes, the method of 
3581: images construction of CMB maps, and geometric methods.  We will 
3582: discuss a catalog of such studies.
3583: 
3584: \subsection{Direct methods in hyperbolic space}
3585: \label{directhyp}
3586: 
3587: \subsubsection{Simply connected hyperbolic space}
3588: \label{simphyp}
3589: 
3590: Since all of the numerical methods will rely on the expansion of the
3591: eigenmodes on the universal cover, we summarize those results here.
3592: From the standard expansion (\ref{elem}).  We need to determine each 
3593: of the factors in the decomposition.
3594: In negatively curved space perturbations are time dependent according
3595: to eqn (\ref{feta}) with solution
3596: 	\be
3597: 	F(\eta)={5\left (\sinh^2\eta-3\eta\sinh\eta+4\cosh\eta+4\right
3598: 	)\over \left (\cosh\eta-1\right )^3}.
3599: 	\ee
3600: 
3601: In a negatively curved space the Helmholtz eqn becomes
3602: 	\be
3603: 	{1\over \sinh^2 r}\left [
3604: 	\partial_r\left (\sinh^2r{\partial _r}\right )+
3605: 	{1\over \sin^2\theta} 
3606: 	{\partial_\theta}\left (\sin\theta \partial_\theta
3607: 	\right )+{1\over \sin^2\theta}\partial^2_\theta \right]\psi_{\vec k}
3608: 	=-k^2
3609: 	\psi_{\vec k}.
3610: 	\ee
3611: As emphasized in Ref. \cite{lyth}, a complete orthonormal basis for 
3612: the simply connected space is formed
3613: by modes with real $k^2>1$ and so the eigenvalue range
3614: is $k^2=[1,\infty]$.  These modes vary on a scale below the curvature
3615: radius and are therefore subcurvature modes.  Standard causal theories 
3616: such as inflation will not seed fluctuations on scales $k<1$ although some
3617: other unforeseen mechanism may. 
3618: It is customary to introduce another parameter 
3619: 	\be q^2=k^2-1\ee
3620: with  the 
3621: range $q^2=[0,\infty]$.
3622: The eigenmodes on $\ah $ can be expressed as
3623: 	\be
3624: 	\psi_{q \ell m}=X_{q\ell}(r)Y_{\ell m}(\hat n)
3625: 	\ee
3626: where the $Y_{\ell m}$'s are the usual spherical harmonics and
3627: the $X_{q\ell}$ are the radial functions
3628: 	\be
3629: 	X_{q\ell}={\Gamma(\ell +1+iq)\over \Gamma(iq)}
3630: 	\sqrt{1\over \sinh r}P^{-\ell -{1\over 2}}_{iq-{1\over 2}}(\cosh r)
3631: 	\ee
3632: where $\Gamma$ denotes the Gamma function and $P$ denotes the Legendre functions
3633: \cite{lyth}.	
3634: 
3635: 
3636: 
3637: 
3638: The $\hat\Phi_q$ are drawn from a Gaussian probability distributions
3639: with proper normalization for hyperbolic space
3640: 	\be
3641: 	\left <
3642: 	\hat \Phi_{q\ell m} \hat \Phi^*_{q^\prime\ell^\prime m^\prime}
3643: 	\right >
3644: 	={2\pi^2\over q(q^2+1)}{\cal P}_\Phi(q) \delta({ q- q^\prime})
3645: 	\delta_{\ell\ell^\prime}\delta_{m m^\prime}
3646: 	\ee
3647: 
3648: We can summarize the Sachs-Wolfe
3649: effect as
3650: 	\be
3651: 	{\delta T(\hat n)\over T}=\int d^3{\vec k} \hat\Phi_{\vec k}(\hat n)
3652: 	L_{\vec k}
3653: 	\ee
3654: with both the surface Sachs-Wolfe and the integrated Sachs-Wolfe accounted
3655: for in 
3656: 	\be
3657: 	L_{\vec k}=
3658: 	\left [ {1\over 3}F(\eta_{sls})+2\int_{\eta_{sls}}^{\eta_0}
3659: 	d\eta F^\prime (\eta)\right ] \psi_{\vec k}(\vec x)
3660: 	\ \ .
3661: 	\ee
3662: 
3663: The Fourier multipoles are 
3664: 	\be
3665: 	C_\ell=2\pi^2\int_1^{\infty}{dk\over k(k^2-1)}{\cal P}_\Phi(k)
3666: 	\left [{1\over 5}X_{k\ell}(\Delta \eta)+{6\over 5}
3667: 	\int_{\eta_d}^{\eta_0}dr\ X_{k\ell}(r)F^\prime(\Delta \eta-r)
3668: 	\right ]^2.
3669: 	\ee
3670: The \cb data alone tends to favor $\Omega\sim 0.3-0.4$
3671: (see for instance \cite{krzy}) although both very low 
3672: $\Omega $ and an $\Omega \sim 1$ are still compatible.
3673: For a long time, the COBE results along with other astronomical
3674: observations \cite{spergelcqg} put low $\Omega $ cosmologies
3675: in the limelight.  However, more recent high $\ell $ observations \cite{exper}
3676: in conjunction with the supernovae data \cite{saul} have 
3677: pushed public opinion back towards $\Omega=1$ cosmologies.
3678: The recent high $\ell $ observations only examine small patches of 
3679: the sky.  A full sky map is required to advance topology observations
3680: and so we reserve more commentary until the launch and hopeful
3681: success of the future MAP and {\it Planck} missions.
3682: 
3683: 
3684: \subsubsection{Cusps}
3685: \label{obscusps}
3686: 
3687: There are multiconnected hyperbolic spaces which are not completely
3688: compact (see Ref. \cite{lum} for some examples
3689: such as those of Refs. \cite{{lobell},{seifertweber}}). 
3690: For some of these the motion is not chaotic
3691: and it is possible to find the
3692: eigenmodes.  One topology of particular interest is the toroidal horn
3693: \cite{{sos},{lbbs}}.  The horn is interesting since many
3694: manifolds have cusped corners as stressed by \cite{css2}.  
3695: %A cusp is a finite truncation of the infinitely long horn.
3696: Despite
3697: their frequency in a set of generic manifolds,
3698: it was argued in Ref. \cite{css2} that according to the thick-thin
3699: decomposition,
3700: % \cite{thick},  
3701: it would very improbable for us to live
3702: in the thin part of the manifold.  However, this argument is flawed.
3703: Since the 
3704: cusp narrows exponentially quickly, an observer can live in a fat
3705: part of the manifold and still see photons coming from a constricted
3706: part of the cusp.  
3707: %What is valid as pointed in Ref. \cite{css2} is that
3708: %the fatter the part of the manifold in which we live, the poorer the
3709: %cusp approximation will be to the fully compact manifold which will 
3710: %show the intricacies of the fully chaotic flows.
3711: %[are they right about that?]
3712: However, Olson and Starkman have developed an approximation scheme to study
3713: cusps on full compact manifolds which shows the generic patterns do in 
3714: fact persist.
3715: 
3716: 
3717: \begin{figure}
3718: %\centerline{\psfig{file=../../top/conf/horn.eps,width=2.5in}}
3719: \centerline{\psfig{file=horn.eps,width=2.5in}}
3720: \caption{Embedding of the cusp topology with one of the dimensions
3721: suppressed.}
3722: \label{horn}
3723: \end{figure}
3724: 
3725: The cusp is most easily understood in the upper half space representation
3726: of $\ah $ (see Appendix \ref{AppendixA})
3727: 	\be
3728: 	ds^2=-d\eta^2+dz^2+e^{-2z}\left (dx^2+dy^2\right )
3729: 	\label{coordsys}
3730: 	\ee
3731: with the coordinate transformation
3732: $e^{-z}=\cosh r-\sinh r\cos \theta, e^{-z}x=\sin\theta\cos\phi\sinh r,
3733: e^{-z}y=\sin\theta\sin\phi\sinh r$.
3734: The $(x,y)$ subspace looks like a conformally stretched flat space.
3735: The topological identifications $x\rightarrow x+L_x$ and $y\rightarrow y+L+y$
3736: make a 2-torus which is squeezed in area
3737: by the exponential factor $e^{-2z}$ 
3738: into a tightening cusp.
3739: Since the subspace is conformally flat, there are no tangled geodesics and 
3740: it is possible to decompose the eigenvalues analytically.
3741: 
3742: In the coordinate system (\ref{coordsys}),
3743: the temperature fluctuations can be decomposed as
3744: \begin{equation}
3745: {\frac{\delta T}T}(\hat n_y)=\int_0^\infty dq\ \sum_{n_xn_y}\hat \Phi _{qn_xn_y}
3746: (\hat n)N_{qn_xn_y}L_{qn_xn_y}  \label{it}
3747: \end{equation}
3748: with the normalization 
3749: \begin{equation}
3750: N_{qn_xn_y}=\left( {\frac{2q\sinh (\pi q)}{\pi ^2}}{\frac{(2-\delta
3751: _{n_x0})(2-\delta _{n_y0})}{bh}}\right) ^{1/2}
3752: \end{equation}
3753: and
3754: \be
3755: L_{qn_xn_y}=\left[ {\frac 13}+ 2\int_{\eta _{{\rm i}}}^{\eta
3756: _o}d\eta F^{\prime }(\eta )\right] e^zK_{iq}(Qe^z)\times   
3757:  \pmatrix{ \sin\left({2\pi n_x\over b}x\right )\cr \cos\left({2\pi
3758: n_x\over b}x\right )\cr}\pmatrix{ \sin\left({2\pi n_y\over h}y\right )\cr
3759: \cos\left({2\pi n_y\over h}y\right )\cr}\ \ 
3760: \ee
3761: where $K_{iq}$ is a modified Bessel function with imaginary index. 
3762: The
3763: entire function is included here in the integration over $\eta $. The
3764: argument of the Bessel function is
3765: $Q^2=4\pi ^2(n_x^2/L_x^2+n_y^2/L_y^2)$.  Notice that $q$ is still a continuous
3766: index since the $z$ direction is infinite.  There is no cutoff in 
3767: the range of $q$ but there is a cutoff in the range of $(n_x,n_y)$.
3768: This cutoff results in a severe suppression of power
3769: as an observer lives nearer the cusp.  The suppression appears as a flat
3770: spot in simulations of the \cb sky with concentric rings
3771: of larger and larger structures as the cusp widens \cite{lbbs}.  
3772: 
3773: A standard likelihood analysis of the $C_\ell$s places the size of the 
3774: torus at the location of the Earth to be $\gta \Delta \eta$.  This is 
3775: not a particularly narrow part of the cusp relative to the SLS but nonetheless 
3776: an observer can 
3777: still see exponentially
3778: deep down the throat putting large quiet regions in the CMB maps \cite{lbbs}.
3779: 
3780: The generic property of flat spots 
3781: was tested in Ref. \cite{olsonstark}.  They showed that from a typical
3782: location in a cusped manifold m003, flat spots of angular size $\sim 5^\circ$
3783: would be visible for $\Omega_o=0.3$.  They conjecture that observable
3784: flat spots on this
3785: scale are typical of cusped manifolds.
3786: To handle the fully compact case Olson and Starkman found a means to 
3787: approximate the modes in the cusp without tackling the full eigenmode 
3788: solutions.
3789: They consider a horosphere, a sphere within the Poincar\'e representation
3790: of $\ah$, which is tangent to the cusp and passes through the point
3791: on the SLS.  On the horosphere, the transformation group simplifies and they
3792: are able to isolate modes.
3793: 
3794: The specific space they consider is a cusped manifold, namely m003, constructed
3795: from two tetrahedra with total volume
3796: ${\cal V}\approx 2.0299$.  The restriction to the horosphere results in 
3797: a hexagonal tiling of $\eu$ and the eigenmodes are reminiscent of those
3798: of eqn (\ref{hexmodes}).  Consequently the minimum eigenmode from
3799: the hexagonal tiling provides an estimate for the extent of the flat spot.
3800: Situating the observer at a symmetric point in the manifold they find  
3801: that flat spots in the CMB from cusped regions subtend an average angle
3802: of roughly $5^\circ$ for $\Omega_0=0.3$.  Larger values of $\Omega_0 $
3803: will naturally diminish the angular scale.
3804: They argue that in general cusped manifolds will show flat spots as 
3805: characteristic features in the CMB maps.
3806: 
3807: 
3808: \begin{figure}
3809: %\centerline{\psfig{file=../../top/cusp/top1_k10n25nd15.ps,width=2.5in}}
3810: \centerline{\psfig{file=top1_k10n25nd15.eps,width=2.5in}}
3811: \caption{Flat spot in a map of the sky in the cusp topology with 
3812: scales
3813: $b=h=1$ and $\Omega_0=0.3$}
3814: \label{flatspot}
3815: \end{figure}
3816: 
3817: \subsubsection{Numerical Eigenvalue Solutions}
3818: \label{numersol}
3819: 
3820: The eigenmodes can always be obtained by brute force numerically.
3821: Three groups have developed numerical methods to isolate a list of 
3822: eigenmodes and eigenvalues for a small collection of manifolds
3823: \cite{{inoue},{inouetomita},{inouethesis},{cornsperg},{aurich}}.  
3824: A comparison of the simulated CMB sky with the \cb data shows the 
3825: specific compact hyperbolic manifolds studied were consistent
3826: with the data. However
3827: all of the statistical analyses 
3828: compare only the 
3829: $C_\ell$'s and not the full correlation functions.
3830: Using the method of images to simulate the CMB,
3831: a full statistical analysis is performed 
3832: by Bond, Pogosyan, and Souradeep as described in section \S \ref{obsmi}
3833: where more negative 
3834: conclusions are obtained; namely that the spaces studied were 
3835: inconsistent with COBE except for maybe one orientation of the manifold.
3836: It would be interesting if numerical eigenmode constructions 
3837: would compare a full
3838: statistical analysis to that performed using 
3839: the method of images to check for consistency.
3840: 
3841: \centerline{\bf Eigenmodes of the Thurston Space}
3842: 
3843: Inoue was the first to find precise eigenmodes on a compact hyperbolic
3844: 3-space for the purposes
3845: of simulating a cosmological model.  
3846: (Aurich and Marklof were the first to compute the eigenmodes on an 
3847: orbifold \cite{aurichmarklof}.)
3848: His method was based on the direct 
3849: boundary element method initially developed by Aurich and Steiner for the
3850: study of $2$-dimensional compact hyperbolic spaces.
3851: The numerical method is based on solving the Helmholtz equation
3852: 	\be
3853: 	(\nabla^2+k^2)\psi_{\vec k}(\vx)=0
3854: 	\label{helm}
3855: 	\ee
3856: on a compact manifold, i.e.
3857: with periodic boundary conditions on the universal
3858: covering space.  Various methods for doing so
3859: include the finite element method and the
3860: finite difference method.
3861: He uses the more precise but numerically more time consuming
3862: method called the direct boundary element method
3863: (DBEM) 
3864: also used by Aurich and Steiner to study quantum chaos on 
3865: a 2D compact space \cite{aurichsteiner}.
3866: He isolated 
3867: the first 36 eigenmodes of the 
3868: manifold $m003(-2,3)$ from the SnapPea census, otherwise
3869: known as the Thurston space in homage to the Fields Medalist 
3870: W. Thurston.  
3871: There are two particularly interesting results.
3872: Firstly he finds a cutoff in the low $k$ eigenvalues and therefore
3873: a maximum wavelength at a value
3874: near the average diameter of the space.
3875: Secondly, 
3876: he discovered that 
3877: the expansion coefficients of the eigenmodes ($\xi_{qlm}$ below)
3878: are well described as 
3879: pseudo-random in behavior
3880: \cite{inoue}.
3881: Despite the long-wavelength cutoff, 
3882: the integrated Sachs-Wolfe effect is able to compensate for low 
3883: $\Omega_0$, as expected.  Consequently, 
3884: assuming an 
3885: Harrison-Zeldovich power spectrum,  
3886: he finds that the $C_\ell$'s are consistent 
3887: with \cb for $0.1 \le \Omega_0 \le 0.6$.
3888: 
3889: 
3890: \begin{figure}
3891: \centerline{\psfig{file=thursi.ps,width=2.25in}}
3892: \caption{
3893: Dirichlet domain for the Thurston space}
3894: \label{thurs}
3895: \end{figure}
3896: 
3897: 
3898: As in flat space, the eigenvalues are a discrete subset of the 
3899: continuous eienvalues.  Since, the eigenmodes of a compact manifold
3900: are a
3901: subset of the eigenmodes on the universal cover (\S \ref{simphyp}), 
3902: they can be expanded as
3903: 	\be
3904: 	\psi^{\cal M}_{q }=\sum_{\ell m}
3905: 	\xi_{q\ell m}X_{q\ell}(r)Y_{\ell m}(\hat n).
3906: 	\label{modexpan}
3907: 	\ee
3908: The $\xi_{q\ell m}$ can be thought of us as containing the 
3909: analogue of the relations found explicitly in flat space
3910: in section \S \ref{compflat}.
3911: The challenge of numerically isolating the modes can be reduced to the 
3912: still difficult task of finding the $\xi_{q\ell m}$.
3913: Inoue explicitly finds that the $\xi_{q \ell m}$ behave as random Gaussian 
3914: numbers for ($\ell <19$ and $q>9.94$).  He extrapolates that this continues
3915: to be true, even more so, for all higher modes and introduces this as 
3916: an approximation to determine modes above $k=10$.
3917: A distinction should be made between the random Gaussian
3918: initial conditions
3919: contained in the $\hat \Phi_{\vec q}$ and the random Gaussian behavior of
3920: the $\xi_{\vec q}$.  The former is an 
3921: assumption which may or may not be true depending on the history of 
3922: the universe while the latter is a property of the manifold.
3923: 
3924: The DBEM was first used
3925: to find the pseudo-Gaussian random numbers $\xi_{q\ell m}$
3926: for $k<10$ \cite{inoue} and the analysis was then extended to include the 
3927: first 36 eigenmodes with $k\le 13$ \cite{inouetomita}.  
3928: (As discussed below, it is claimed in Ref. \cite{cornsperg}
3929: that 2 modes were originally overlooked in this method).  
3930: A numerical cutoff in the low $k$ spectrum was found at the value 
3931: $k_1=5.41$.
3932: The corresponding maximum wavelength 
3933: $\lambda_{\rm max}=2\pi/k_1$ can be compared to an approximation for
3934: the diameter of ${\cal M}$ which
3935: he computes as the average of the in-radius and the out-radius.  
3936: The manifold has
3937: in-radius $r_{in}=0.535$, out-radius $r_{out}=0.7485$
3938: (and volume ${\cal V}=0.98139$).
3939: This
3940: yields $k_{\rm min}=4\pi/(r_{+}+r_{-})=4.9$, an agreement to within 10\% of his
3941: $k_1$.
3942: There are no supercurvature modes in the Thurston space.
3943: 
3944: The pseudo-Gaussian behavior can be interpreted 
3945: in a quantum mechanical context \cite{inoue}.
3946: Let the Laplace-Beltrami operator be the Hamiltonian of a quantum system, then
3947: the eigenmode is a wavefunction eigenstate.  
3948: Since the underlying classical dynamics is chaotic, it has been conjectured
3949: that the quantum mechanical system will be governed by the predictions
3950: of random matrix theory (RMT) \cite{{berry},{ott}}.
3951: One such prediction is that the square expansion coefficients are 
3952: Gaussian distributed if expanded with respect to a generic bases.
3953: In particular, the coefficients should be given by the
3954: Gaussian Orthogonal Ensemble (GOE)
3955: 	\be
3956: 	P(x)={1\over \sqrt{2\pi x}}e^{-x/2}.
3957: 	\ee
3958: To check this prediction for the modes (\ref{modexpan}), Inoue
3959: first recasts the eigenmodes
3960: in terms of real independent coefficients
3961: $a_{q\ell m}$ and real functions $R_{q\ell m}$ as 
3962: 	\be
3963: 	\psi^{\cal M}_{q}=\sum_{\ell m}a_{q\ell m}R_{q\ell m}
3964: 	\ee
3965: where the $a_{q\ell m}$ can be expressed in terms of the $\xi_{q\ell m}$
3966: and the $R_{q\ell m}$ can be expressed in terms of the $X_{q\ell}Y_{\ell m}$
3967: \cite{inoue}.
3968: Following \cite{aurichsteiner}, he examines the statistical behavior of
3969: 	\be
3970: 	x={|a_{q\ell m}-\bar a_q|^2\over \sigma_q^2}
3971: 	\ee
3972: where $\bar a_q$ is the average and $\sigma_q^2$ the variance.  
3973: The probability is singular
3974: at $x=0$ and so it is customary to
3975: compare the numerics to the cumulative distributions
3976: 	\be
3977: 	I(x)=\int^x_0P(x)dx={\rm erf}(\sqrt{x/2}).
3978: 	\ee
3979: %The cumulative
3980: %Random matrix theory distribution is computed as
3981: %	\ba
3982: %	I_\mu(x) &=& \int_0^xdx^\prime P_\mu(x^\prime)\nonumber \\
3983: %	&=& {\gamma(\mu/2,\mu x/2)\over \Gamma(\mu/2)}
3984: %	\ea
3985: %where $\gamma(y,z)$ is the incomplete gamma function.  
3986: To test the RMT 
3987: prediction, the numerically determined cumulative distribution are compared
3988: for a goodness of fit.  Very good agreement with the GOE prediction is found
3989: confirming Gaussian behavior for the low lying states.  The Gaussian behavior
3990: is expected for the highly excited states as a consequence of the classical
3991: chaos but Gaussian behavior for the low lying states is less obvious.
3992: Yet this is in fact what his observations confirm.
3993: He also tests the randomness of the $a_{q\ell m}$ and finds that they do 
3994: behave as random variables.
3995: The randomness is not a property of the eigenmodes but 
3996: is due rather to
3997: the almost random distribution of images in the
3998: universal cover.
3999: The number of copies of the fundamental domain inside a sphere with 
4000: radius $\eta_0$ is
4001: 	\be
4002: 	n_1={\pi(\sinh(2\eta_0)-2\eta_0\over {\cal V}}
4003: 	\ee
4004: giving $n_1\sim 29$ for $\eta_0=1.6$.
4005: 
4006: The number of eigenmodes increases as $k^3$ and the number of boundary 
4007: elements increases as $k^2$.  The task of computing high $k$ modes becomes
4008: unmanageable with this method.  
4009: For higher $k$, Ref \cite{inoue} suggests taking the clue
4010: from the behavior of low $k$ modes and approximating the highly-excited states
4011: as random Gaussian numbers with a variance proportional to $q^{-2}$.
4012: Using Weyl's asymptotic formula,
4013: 	\be
4014: 	N[q]={{\cal V}q^3\over 6\pi^3}
4015: 	\quad \quad q\equiv \sqrt{k^2-1}\quad \quad q>>1 
4016: 	\ee
4017: with $N$ an integer.  
4018: Since for a CH space the volume is fixed, Weyl's formula effectively
4019: relates the number of states at a given $k$ with a topological feature
4020: of the space.  It also allows an estimate of the $k_j$.
4021: The spacing between discrete eigenvalues decreases as the inverse of 
4022: $k$ for large $k$ and so approaches a continuous spectrum in the large
4023: limit, as expected.
4024: Weyl's formula 
4025: is well obeyed and 30 of the 36 modes show random Gaussian behavior.
4026: Interestingly,  
4027: six degenerate states are found
4028: which correspond to a nearly symmetric mode 
4029: reflecting the global symmetry of the fundamental domain.
4030: While a linear combination of the 
4031: degenerate modes shows Gaussian
4032: behavior again.
4033: This is typical of classically chaotic systems where the classical chaos
4034: leads to a Gaussian behavior in the quantized eigenmodes although 
4035: occasionally the global symmetry of the space can surface in the eigenmodes
4036: as non-Gaussian behavior. 
4037: (The connection with quantum chaos led
4038: Inoue to use the Selberg Trace Formula to compute eigenvalues for a large
4039: number of CH manifolds \cite{inouetrace}.  This method proved to be quicker
4040: and easier to implement numerically.)
4041: 
4042: 
4043: 
4044: The 
4045: $C_\ell$'s are calculated
4046: %in the range $2\ge \ell \le 18$ 
4047: by sewing together the numerically obtained 
4048: eigenmodes for the eigenvalues in the range $5.4 \le k\le 13$ with 
4049: the above approximation for the expansion coefficients for 
4050: $13\le k<20$. 
4051: With the assumption that ${\cal P}_\Phi=$constant,
4052: the resultant $C_{\ell }$ is compared with the \cb data
4053: to conclude that the spectrum is consistent with \cb for
4054: $0.2 \le \Omega_0\le 0.6$.  
4055: The cutoff in the spectrum due to the minimum
4056: mode $k=5.4$ is buried under the contributions to low $\ell $ power from 
4057: the integrated Sachs-Wolfe effect.  The ISW becomes important during the
4058: curvature dominated epoch at about $1+z\sim (1-\Omega_o)/\Omega_o$.
4059: They find the contribution from $k \le 13$ modes
4060: to $C_\ell $ for $2\le \ell \le 20$ is 7 \%
4061: for $\Omega=0.2$ and 10\% for $\Omega_o=0.4$.  The rest is due to the ISW.
4062: For this reason, the spectra appear to have a gradual peak near the 
4063: long-wavelength cutoff or are nearly
4064: flat down to low $k$ and are less severely constrained 
4065: for $\Omega\ge 0.1$ than the flat
4066: models of section \S \ref{obsflat} \cite{{inoue},{inouetomita}}.
4067: In fact they find very good agreement with
4068: \cb for $\Omega_0=0.6$, better than other FRW models.  In particular,
4069: an alignment of the peak in the \cb data at $\ell\sim 4$ with the peak
4070: due to the cutoff accounts for the better fit.
4071: 
4072: The conclusion cannot be stated as proving that the Thurston space is 
4073: consistent with the observations, only that it does not appear inconsistent.
4074: A comparison of the full correlation function to the data may not survive
4075: consistency and in fact does not according the method of images analysis 
4076: of \cite{{bpsI},{bpsII}}.
4077: 
4078: 
4079: 
4080: \centerline{\bf Eigenmodes of an Orbifold}
4081: 
4082: Aurich was able to compute a huge number of eigenmodes, 
4083: the first 749, in an orbifold with negative curvature.  
4084: Orbifolds can have compact volume but possess points which are
4085: not locally $R^3$.  They can also have rotation group elements among
4086: their isometries, unlike manifolds which have no groups with fixed points
4087: in their isometries.  In the absence of a predictive theory for the topology
4088: of the cosmos, there is
4089: no convincing reason to omit them from the catalog.
4090: Interestingly, they can have even smaller volumes than the minimum bound
4091: on compact hyperbolic spaces.
4092: %$V_{{\cal O}}\sim 0.7173068$ [?].
4093: 
4094: The fundamental cell is a pentahedron which is symmetric along a plane
4095: dividing the fundamental domain into two identical tetrahedra.
4096: This allows a
4097: desymmetrizing of the pentahedron useful for deducing the eigenmodes
4098: following early work \cite{aurichmarklof}.  
4099: There are 9 compact hyperbolic tetrahedra and
4100: the one built into the pentahedral orbifold
4101: is known as $T_8$.  The volume of $T_8$ is
4102: ${\cal V}_{\rm tetrahedron}\simeq 0.3586524$ which is smaller than
4103: the Weeks space.
4104: The smallest compact hyperbolic tetrahedron $T_3$ has volume 
4105: ${\cal V}=0.03588506 $, ten times smaller than $T_8$ 
4106: and smaller even than the existing bound on CH manifolds
4107: ($V_{\cal M}=5/(2\sqrt{3}){\rm arcsinh}^2(\sqrt{3}/5)\sim 0.167$).
4108: Using the boundary element method
4109: 749 eigenmodes on $T_8$ are obtained.
4110: 
4111: Both radiation and matter are included in his analysis which evolves
4112: the metric
4113: perturbations according to the usual adiabatic, linear perturbation theory
4114: and solves $F(\eta)$ numerically.
4115: In his analysis, the eigenmodes are expanded 
4116: as
4117: 	\be
4118: 	\Phi(\eta,\vx)=\sum F_q(\eta)\psi_q(\vx)
4119: 	\label{exp1}
4120: 	\ee
4121: where the $\psi_q$ are the eigenmodes and all of the time dependence is
4122: in $F_{\vec q}(\eta)$.  
4123: Notice that in comparison to the standard
4124: expansion \ref{fluctcmb}, he does not
4125: include the usual randomly seeded fluctuation amplitude $\hat \Phi_k$.  Instead
4126: all of the initial conditions are absorbed into $F(\eta)$ with
4127: 	\be
4128: 	F_q(\eta_i) =\alpha/q^{3/2}
4129: 	\quad \quad F^\prime_q(\eta_i)=0.
4130: 	\label{exp2}
4131: 	\ee
4132: There is however some element of randomness in the sign of the fluctuation.
4133: The eigenfunctions have a freedom in the phase.  Since the phase is always
4134: chosen to be real, this freedom results in a $\pm $ ambiguity in the 
4135: eigenfunction which is chosen randomly.
4136: The constant $\alpha $ is normalized against the \cb data.
4137: This is consistent with the Harrison-Zeldovich flat spectrum in terms of 
4138: normalization but lacks any of the randomness that real fluctuations
4139: would have.  
4140: Because the randomness of the temperature fluctuations about the mean is
4141: neglected, it would not be quite right to
4142: interpret the simulated maps of Ref \cite{aurich} in terms of
4143: an actual universe.  The advantage however is that the maps
4144: show a random nature 
4145: which can only be due to the random character of the modes
4146: themselves.  This is important for the reasons discussed at length in the 
4147: previous
4148: subsection \ref{directhyp}.
4149: 
4150: Since the $C_\ell$ produces a global averaging, 
4151: an ensemble average is generated by the expansion 
4152: (\ref{exp1})-(\ref{exp2}) even if the standard
4153: random initial fluctuations were ignored.  Although
4154: the simulations depend on the location of the observer and the orientation
4155: of the manifold,
4156: this dependency is not examined and the observer
4157: is fixed, offset from the origin.
4158: The $C_\ell $'s are compared to 
4159: the $\ell <30$ \cb data, the Saskatoon data around $\ell \sim 100$
4160: \cite{saskatoon} and the QMAP data \cite{qmap}
4161: above $\ell\sim 80$.
4162: All modes are computed up to $k_{max}=55$.  
4163: %Since there are 749 of these,
4164: %there is quite a large degeneracy of eigenvalues.
4165: Aurich varies $\Omega_0$ between $0.2$ and $0.6$ and finds reasonable
4166: agreement for $\Omega_0\simeq 0.3-0.4$.
4167: For smaller $\Omega_0$ he
4168: finds that the numerical $C_\ell$ increases too quickly with $\ell$.
4169: We suggest that this is evidence of a long wavelength cutoff in the spectrum.
4170: Although for a lower $\Omega_0$, the ISW is important in terms of building
4171: up the low $\ell$ power, it is also true that the suppression is more
4172: severe.  To put it another way,  if the constant $\alpha $ were normalized
4173: to \cb at high $\ell$, as it should be in order
4174: to minimize the effects of cosmic variance, 
4175: then the steep slope would be seen as a suppression of large
4176: angle power due to the finite extent of the space.
4177: 
4178: 
4179: \begin{figure}
4180: \centerline{\psfig{file=Cl_Omega_m30_q00_l60.ps,width=2.25in}}
4181: \caption{
4182: This figure was supplied courtesy of R. Aurich.  
4183: %cite {as}
4184: The figure shows $T_\ell$ measured in micro-Kelvin
4185: versus $\ell $ as computed 
4186: by Aurich and Steiner for an $\Omega_0=0.9$ cosmology with 
4187: $\Omega_\Lambda=0.6$.  The angular power spectrum $C_\ell $ for the
4188: orbifold is shown in comparison to an infinite topology as obtained
4189: by CMBFAST.  The suppression for $\ell > 30$ is only due to the truncation of
4190: the sum over the eigenmodes since only the first 749 eigenmodes have
4191: been taken into account.}
4192: \label{aurcl}
4193: \end{figure}
4194: 
4195: 
4196: The same orbifold was studied again in Ref.\ \cite{as} where both 
4197: a cosmological constant and a smooth dark energy component were 
4198: added to the radiation and matter
4199: energy density.  The total energy is taken to be nearly
4200: but not quite flat to address the recent results from the small-angle
4201: CMB experiments \cite{{exper},{balbi}}.
4202: They still find a suppression in power
4203: for $\ell \lta 10$, as illustrated in figure
4204: \ref{aurcl}.
4205: 
4206: Although the results of Ref. \cite{aurich}
4207: to be at variance with the results of 
4208: \cite{{bpsI},{bpsII}} we emphasize again the limitations of comparing only the 
4209: $C_\ell$'s.  This is expanded upon in section \S \ref{obsmi}.
4210: 
4211: 
4212: 
4213: \centerline{\bf Fast Method for Isolating Eigenmodes}
4214: 
4215: 
4216: A fast numerical method for obtaining the eigenvalues on compact hyperbolic
4217: manifolds was developed by Cornish and Spergel \cite{cornsperg}.  In comparison
4218: to the boundary element method developed by Aurich and Steiner
4219: \cite{aurichsteiner}, it is technically inferior but far quicker and easier to
4220: implement numerically.
4221: The DBEM has only been applied to a few
4222: cosmological spaces, the tetrahedral orbifold of
4223: \cite{aurich}, the Thurston manifold \cite{inoue}, and the 
4224: Weeks manifold to name a few \cite{inouetrace}, to name a few,
4225: while the method of
4226: Ref. \cite{cornsperg} allows them to obtain the lowest eigenvalues and
4227: eigenmodes for 12 manifolds, requiring
4228: only the generators as input.
4229: 
4230: The method is to simply solve
4231: 	\be
4232: 	\Psi_k(x)=\Psi_k(gx)
4233: 	\ee
4234: using a singular value decomposition.
4235: The eigenmodes are again
4236: expanded in terms of the eigenmodes on the universal
4237: cover $\ah$
4238: as in eqn (\ref{modexpan}).
4239: Random points are selected within the fundamental domain and 
4240: all of the images out to some distance are located in the covering space.
4241: For each point $p_j$ there are $n_j$ such images located with the
4242: generators $g_\alpha$.  The
4243: $n_j(n_j+1)/2$ boundary conditions are
4244: 	\be
4245: 	\Psi_k(g_\alpha p_j)=\Psi_k(g_\beta p_j)
4246: 	\quad \quad \alpha \ne \beta
4247: 	\label{bclist}.
4248: 	\ee
4249: Using the expansion of the eigenmodes in terms of the eigenmodes of the
4250: universal cover \ref{simphyp}, they write (\ref{bclist}) 
4251: as a collection of difference
4252: equations which they then solve using a standard 
4253: singular value decomposition
4254: method for handling over constrained systems of equations.
4255: 
4256: 
4257: The example they studied in detail is the Weeks space m003(3,-1).
4258:   They also looked at the Thurston space and
4259: found agreement with the DBEM method applied by Inoue except they found
4260: 2 new modes missed previously but later confirmed \cite{ktinoue}.
4261: Again, for higher $k$, the number of modes obeyed Weyl's formula well.
4262: Also, the modes were well described by random matrix theory which predicts
4263: the expansion coefficients obey a Gaussian
4264: Orthogonal Ensemble.  
4265: 
4266: 
4267: \begin{figure}
4268: \centerline{\psfig{file=weeksi.ps,width=2.25in}}
4269: \caption{
4270: Dirichlet domain for the Weeks space}
4271: \label{weeks}
4272: \end{figure}
4273: 
4274: The Weeks space has a large symmetry group which ensures a large degeneracy of
4275: eigenmodes.
4276: The symmetry group is the Dihedral group of order 6.
4277: They found the first 74 eigenmodes, many of which are degenerate.  The higher
4278: the mode, the more degeneracies in agreement with Weyl's asymptotic formula.
4279: A view of the lowest eigenmode in the Weeks space can be found in 
4280: figure \ref{cornspergfig}.
4281: 
4282: \begin{figure}
4283: \centerline{\psfig{file=escher.eps,angle=-90,width=3.25in}}
4284: \caption{This figure was supplied courtesy of N. Cornish.
4285: %cite {cornsperg}
4286: The lowest eigenmode in the Weeks space is shown.  The space
4287: is drawn in the Poincar\'e representation.  The three panels
4288: represent different slices through the Poincar\'e ball.
4289: The upper figure shows the slice through the fundamental domain
4290: while the lower figure shows the slice through the eigenmode.
4291: Specifically,
4292: the leftmost panel is the $x=0$ slice, the middle is the 
4293: $y=0$ slice, and the rightmost panel is the $z=0$ slice.
4294: }
4295: \label{cornspergfig}
4296: \end{figure}
4297: 
4298: 
4299: 
4300: 
4301: Implementing
4302: their eigenmodes in a simulation of a cosmological
4303: model, they compare the numerically generated $C_\ell$'s to the \cb
4304: data \cite{cornsperg}. 
4305: Their numerically modeled cosmologies are based on the first 100+ modes of
4306: the following small spaces:
4307: The Weeks space m003(3,-1) with ${\cal V}=0.9427$, the Thurston space
4308: m003(-2,3)
4309: with ${\cal V}=0.9814$,
4310: s718(1,1) with ${\cal V}=2.2726$ and v3509(4,3) with ${\cal V}=6.2392$.  
4311: One realization of the Weeks space is shown in figure \ref{weekscirc}.
4312: Generically,
4313: a kind of statistical isotropy prevails even though the spaces are globally
4314: anisotropic.  Since the expansion coefficients are pseudo-random, Gaussian
4315: distributed numbers which are statistically independent of $\ell $
4316: and $m$, they generate nearly isotropic eigenmodes and lead
4317: to a kind of isotropy across the microwave sky.
4318: The authors suggest inflation may not be need to explain why the universe 
4319: is nearly isotropic.
4320: However inflation is still needed 
4321: to explain why local values are so marginally near flat.
4322: 
4323: 
4324: \begin{figure}
4325: \centerline{\psfig{file=wtot03.eps,width=3.25in}}
4326: \vspace{8mm}
4327: \caption{This figure was supplied courtesy of N. Cornish.
4328: %cite cornsperg
4329: A numerical realization of the CMB in the Weeks topology with
4330: $\Omega_o=0.3$. One pair of matched circles is indicated by the
4331: white lines in this projection.}
4332: \label{weekscirc}
4333: \end{figure}
4334: 
4335: A standard likelihood analysis based on 100
4336: realizations of each topology was performed.
4337: A mode cutoff in the spectrum is again evident and
4338: gets worse as $\Omega_0$ is lowered.  
4339: For all of the manifolds studied, they find that 
4340: 	\be
4341: 	\lambda^q_1=(1.3\rightarrow 1.6)D
4342: 	\label{lq}
4343: 	\ee
4344: where $\lambda^q_1=2\pi/q_1$ (although it is customary 
4345: to define the wavelength corresponding to a given mode as $\lambda^k=2\pi/k$)
4346: which obeys the bound
4347: 	\be
4348: 	{4\tilde D\over D^2(\sinh\tilde D+\tilde D)^2}\le k_1^2\le 1+\left ({2\pi\over
4349: D}\right )^2
4350: 	\ee
4351: with $\tilde D$ the square root of the smallest integer $\ge D^2$.
4352: They do caution that they cannot find modes with $q=[0,1/4]$ using their method,
4353: and so perhaps supercurvature modes lurk.  
4354: However the ISW becomes increasingly more
4355: important as $\Omega_0$ is lowered.  There is an optimal value where the 
4356: two effects compete to create a spectrum with a slight tilt at low $\ell$
4357: and give a better match to the \cb data.  For the 
4358: Weeks' space this occurs at $\Omega_0=0.3$.  
4359: 
4360: Additionally, once the spectrum is normalized to \cb they
4361: find that the size of fluctuations on 
4362: $8h^{-1}$ Mpc is naturally increased above the value of $\sigma_8=0.6$ 
4363: for a simply connected universe with $\Omega=0.3$ to a value of 
4364: $\sigma_8=0.75$ for the Weeks manifold with $\Omega=0.3$.  
4365: The present day cluster abundance seems to imply 
4366: $\sigma_8=0.9\pm 0.1$ for $\Omega_0=0.3$ \cite{pen} and so higher values
4367: are desirable.
4368: 
4369: They find the $C_\ell$'s have
4370: a better fit to the \cb data with a relative likelihood of 
4371: $\sim 20 $ and a better fit to 
4372: the large-scale structure date ($\sigma_8$ increases by 
4373: $\sim 25\%$).  However, as with all of the other statistical analyses they
4374: have applied a weak test by analyzing the angular averaged
4375: $C_\ell$s and not the full
4376: correlation function.  Their conclusions can be interpreted as finding the 
4377: manifolds are not ruled out by the $C_\ell$ alone.
4378: 
4379: 
4380: 
4381: 
4382: 
4383: 
4384: 
4385: 
4386: \subsubsection{The Method of Images}
4387: \label{obsmi}
4388: 
4389: The fluctuations can be simulated 
4390: using the method of images.  The correlation
4391: functions can be calculated without the explicit eigenmodes and eigenvalues.
4392: The procedure is to sum the correlation function on the universal cover 
4393: over all images out to some large radius.  The
4394: more distant images are handled in a continuous approximation.
4395: %with the added refinement of a 
4396: %Cesaro resummation technique.
4397: The eigenspectrum can be calculated but the
4398: correlation
4399: functions are obtained to better accuracy for a given order in the sum.
4400: Bond, Pogosyan and Souradeep implemented a detailed method of images
4401: \cite{{bpsI},{bpsII},{bps_texas},{bps_moriond},{bps_cwru}}.
4402: They emphasize that the full
4403: correlation function $C(\hat n, \hat n^\prime)$ must be compared to the
4404: data for a meaningful statistical analysis.  The philosophy is to perform
4405: a statistical search for patterns.
4406: They find
4407: two principle  effects
4408: (1) anisotropic patterns and (2) a long wavelength 
4409: cutoff in the power spectrum.
4410: The patterns appear as spikes of positive correlation when a point on 
4411: the SLS and one of its images is correlated.  The larger the SLS relative
4412: to the out-radius, the more statistically significant will the patterns 
4413: be.  If the SLS is smaller than the in-radius then naturally the correlation
4414: function
4415: is very close to that for the simply connected space.  
4416: 
4417: The angular correlation function $\cq $ is 
4418: computed from the spatial two-point correlation function
4419: $\spc\equiv \left <\Phi(\vec x, \tau_{LS})\Phi(\vec x^\prime,
4420: \tau_{LS})\right >$ which can be expressed as
4421: 	\begin{equation}
4422: 	\spc=\sum_i\pow(k_i)\sum_{j=1}^{m_i}\Psi_{ij}(\vx)\Psi^*_{ij}(\vxp)
4423: 	\end{equation}
4424: with the sum over a discrete ordered set with multiplicities $m_i$.
4425: Notice that the $\Psi_{ij}$ obey
4426: 	\begin{equation}
4427: 	\left (\nabla^2 + k_i^2\right )\Psi_{ij}=0
4428: 	\end{equation}
4429: as always, however with the slightly different two-index notation.  The 
4430: $j$ index accounts for any degeneracies.
4431: The spatial correlation function $\xi^c$ 
4432: on the compact
4433: manifold ${\cal M}^c$, can be expressed in terms of the spatial 
4434: correlation function $\xi^u$ on the universal cover.  A derivation
4435: of the relation between $\xi^c$ and $\xi^u$
4436: exploits orthonormality and completeness with the following equations:
4437: 	\begin{equation}
4438: 	\int_{{\cal M}}d\vxp \spc^c (\vx,\vxp) \Psi^c_{ij}(\vxp)=\pow(k_i)
4439: 	\Psi^c_{ij}(\vxp)
4440: 	\label{spcc}
4441: 	\end{equation}
4442: 	\begin{equation}
4443: 	\int_{{\cal M}^u}d\vxp \spc^u(\vx,\vxp) 
4444: 	\Psi^u_{j}( k,\vxp)=\pow(k_i)
4445: 	\Psi^u_{j}( k,\vxp).
4446: 	\label{spcu}
4447: 	\end{equation}
4448: Since eigenfunction on the compact space must also be eigenfunctions on
4449: the universal cover (although the converse is not true) it follows that
4450: 	\bea
4451: 	\int_{{\cal M}^u}d\vxp \spc^u(\vx,\vxp) 
4452: 	\Psi^c_{j}( k,\vxp) &=& \pow(k_i)
4453: 	\Psi^c_{j}( k,\vxp)\nonumber\\
4454: 	&=& \int_{{\cal M}}d\vxp \spc^c(\vx,\vxp) 
4455: 	\Psi^c_{ij}( \vxp) \label{e3}.
4456: 	\eea
4457: Combining (\ref{e3}) with (\ref{spcu}) gives
4458: 	\be
4459: 	 \int_{{\cal M}}d\vxp \spc^c(\vx,\vxp) 
4460: 	\Psi^c_{j}( k,\vxp) = 
4461: 	\int_{{\cal M}^u}d\vxp \spc^u(\vx,\vxp) 
4462: 	\Psi^c_{ij}( \vxp) 
4463: 	\label{e4}.
4464: 	\ee
4465: Since 
4466: ${\cal M}$ tessellates
4467: ${\cal M}^u$ we can re-express the left-hand-side of (\ref{e4})
4468: as
4469: 	\begin{equation}
4470: 	\sum_{g \in \Gamma}
4471: 	 \int_{{\cal M}}d\vxp \spc^u(\vx,g \vxp) 
4472: 	\Psi^c_{ij}( \vxp) 
4473: 	=
4474: 		 \int_{{\cal M}}d\vxp
4475: 	\left [\tilde{ \sum_{g \in \Gamma}}\spc^u(\vx,g
4476: \vxp) 
4477: 	\right ]
4478: 	\Psi^c_{ij}( \vxp) 
4479: 	\end{equation}
4480: where in the last step a regularization is needed and denoted
4481: by the tilde above the summation.
4482: Identifying integrands
4483: gives
4484: 	\begin{equation}
4485: 	\spc^c(\vx,\vxp)=\tilde{ \sum_{g \in \Gamma}}\spc^u(\vx,g
4486: \vxp) 
4487: 	\end{equation}
4488: and the $\spc^c$ can be calculated as the sum over images in the universal
4489: cover with only a knowledge of the group elements.
4490: The spatial correlation function on the universal cover is known to be
4491: 	\be
4492: 	\spc^u(\vx,\vxp)\equiv \spc^u(r)=\int^\infty_0{dq q\over (q^2+1)}
4493: 	{\sin(qr)\over q\sinh r}{\cal P}_\Phi(q)
4494: 	\ee
4495: and ${\cal P}_\Phi$ is taken to be a Harrison-Zeldovich spectrum.
4496: This is the core of the method of images.
4497: 
4498: The regularizer requires some additional effort.
4499: The correlation function on the universal cover does not have compact support:
4500: 	\begin{equation}
4501: 	\int_{{\cal M}^u}d\vxp \spc^u(\vx,\vxp) 
4502: 	=\infty.
4503: 	\end{equation}
4504: The need for regularization is not a result of the CH space having a large
4505: number of periodic orbits as incorrectly claimed in \cite{css_cqg}
4506: nor is it dues to the chaotic nature of the trajectories but rather is 
4507: a result of the correlation function not having compact support.
4508: Even flat spaces require regularization.
4509: The regularized spatial correlation function can be written	
4510: 	\begin{equation} \tilde \xi^u_\Phi(\vx,\vxp)
4511: 	\equiv \spc^u(\vx,\vxp)-{1\over V_{{\cal M}}}
4512: 	\int_{g{\cal M}}d{\bf{\vec x}^{\prime\prime}} 
4513: 	\spc^u(\vxp,{\bf{\vec x}^{\prime\prime}}) 
4514: 	\end{equation}
4515: for $g $ such that $\vxp$ lies in $g {\cal M}$.
4516: The regularization ensures
4517: 	\begin{equation}
4518: 	\int_{{\cal M}^u}d\vxp \tilde \xi^u_\Phi(\vx,\vxp)
4519: 	=0.
4520: 	\end{equation}
4521: Finally then,
4522: 	\begin{equation} 
4523: 	\spc^c(\vx,\vxp)=\sum_{g \in \Gamma}
4524: 	\tilde \xi^u_\Phi(\vx,\vxp)
4525: 	=\sum_{g \in \Gamma} \spc^u(\vx,\vxp)-{1\over V_{{\cal M}}}
4526: 	\int_{{\cal M}^u}d{\bf{\vec x}^{\prime\prime}} 
4527: 	\spc^u(\vxp,{\bf{\vec x}^{\prime\prime}}) .
4528: 	\end{equation}
4529: 
4530: 
4531: The regularization prescription is not unique.  
4532: The actual limiting procedure 
4533: utilized in Ref. \cite{bpsI} sums images up to a radius $r_*$ and then 
4534: regularizes by subtracting the
4535: integral of $\spc^u(r)$ over a spherical ball of radius $r_*$:
4536: 	\be
4537: 	\spc^c(\vx,\vxp)=\lim_{r_*\rightarrow \infty}
4538: 	\left [\sum_{r_j<r_*}\spc^u(r_j)-{4\pi\over V_{\cal M}}
4539: 	\int^{r_*}_0 dr \sinh^2 r \spc^u(r)\right ]
4540: 	\ee
4541: with $r_j=d(\vx, g_j \vxp) \le r_{j+1}$.  The value of 
4542: $r_*$ is numerically pushed out to 4 or 5 times the out-radius in order to 
4543: get a convergent result.
4544: This procedure
4545: is simpler and does well for large $r_*$.
4546: It also does not require a detailed dependence on the complicated shape
4547: of the fundamental domain.
4548: 
4549: An example of the maps they generate using the method of images is shown
4550: in figure \ref{bpsfig}.
4551: 
4552: \begin{figure}
4553: \centerline{\psfig{file=sch.eps,width=3.25in}}
4554: \vskip 5truept
4555: \caption{This figure was supplied courtesy of T. Souradeep.
4556: %cite {bps}
4557: The full CMB sky is represented by the two hemispherical caps -- one 
4558: in the direction of the South Galactic Pole (SGP) and the other in the
4559: direction of the North Galactic Pole (NGP).  The label SHC refers
4560: to a small compact hyperbolic model, namely m004(-5,1).
4561: As well as showing the fluctuations, this figure shows correlated circle
4562: pairs explained in more detain in a following section.}
4563: \label{bpsfig}
4564: \end{figure}
4565: 
4566: 
4567: 
4568: They are also able to estimate the density of states and therefore obtain
4569: a rough estimate of the power spectrum using the method of images.  This is
4570: in addition to their primary result of having computed the correlation 
4571: function.
4572: When they do estimate the density of states they always find an infrared cutoff
4573: as expected for long wavelength modes.
4574: 
4575: This is consistent with the expectations.
4576: For a 3-dimensional CH space there is a bound on Cheeger's isoperimetric 
4577: constant of \cite{{chavel},{berard}}
4578: 	\be
4579: 	k_{min}\ge h_C/2\ge {1\over d_{\cal M}}\left [ 2\int^{1/2}_0
4580: 	dt \cosh^2(t)\right ]^{-1} =0.92/d_{\cal M}
4581: 	\ee
4582: and so there are no supercurvature modes for $d_{\cal M}<0.92 $.
4583: The suppression of power is covered in part by the ISW and so is less prominent
4584: than in flat models.
4585: 
4586: 
4587: Given their numerical calculation,
4588: they do a Bayseian probability analysis comparing a few
4589: compact hyperbolic cosmologies to the \cb data.
4590: Their analysis represents by far the most complete of the statistical 
4591: tests that has been performed and in principle is the most complete
4592: test that can be performed.
4593: They find the compact hyperbolic models they study are inconsistent
4594: with the \cb data for most orientations although for a 
4595: small set of special orientations they do find a better
4596: fit than the standard infinite models.
4597: Their results are qualitatively independent of the matter content.  
4598: The analysis magnifies the inadequacy of using 
4599: the $C_{\ell}$ alone.  In particular they point to the 
4600: large cosmic variance that results from 
4601: the break down associated with isotropy.
4602: 
4603: With the assumption of Gaussianity both in the noise and in the raw
4604: $\delta T/T$, the probabilistic comparison used is
4605: 	\be
4606: 	{\cal P}(\delta |C_T)={1\over (2\pi)^{N_p/2}\|C_T\|^{1/2}}
4607: 	e^{{1\over 2}\delta\dagger C_T^{-1}\delta}
4608: 	\ee
4609: where 	$ C_T=C(\hat n,\hat n^\prime)$ and $\delta $ is the data.
4610: To numerically
4611: obtain the full correlation function requires an evaluation of the
4612: spatial correlation function at $N_p(N_p+1)N_L^2/2$ pairs of points
4613: where $N_p$ is the number of pixels 
4614: and $N_L$ is is the number of points along the line of sight used to
4615: integrate the ISW.  They find an $N_L\sim 10$ is satisfactory.
4616: They estimate the likelihood function 
4617: 	\be
4618: 	{\cal L}(C_T)\equiv 
4619: 	{\cal P}(\bar \delta |C_T)=\int d\delta 
4620: 	{\cal P}(\bar \delta |\delta)
4621: 	{\cal P}( \delta |C_T)
4622: 	\ee
4623: with $\bar \delta $ the map which maximizes the conditional
4624: probability and the integration is
4625: performed over all realizations of the simulated sky $\delta $.
4626: The resultant likelihood function they use is
4627: 	\be
4628: 	{\cal L}(C_T)=
4629: 	{1\over (2\pi)^{N_p/2}\|C_N+C_T\|^{1/2}}
4630: 	e^{{1\over 2}\bar \delta\dagger(C_N+C_T)^{-1}\bar \delta}
4631: 	\ee
4632: with $C_N$ the noise covariance matrix.
4633: What is actually obtained is model dependent relative likelihoods.
4634: The model-dependent parameters are
4635: ${\cal M}$, the orientation, the location of the observer,
4636: $\Omega_m$, $\Omega_0$ and $\Omega_\Lambda$, and the initial assumptions
4637: regarding the spectral shape and character.
4638: The manifolds studied were m004(-5,1), which is relatively small,
4639: and v3543(2,3), which is comparatively large.  ($\Omega_\Lambda=0$
4640: always.  Adding $\Omega_\Lambda$ relaxes the constraints.)
4641: They varied $\Omega_m=\Omega_0$ over three values arranged so that 
4642: the SLS was comparable to the out-radius and varied over 24 different
4643: orientations.
4644: They do leave the observer at the maximum of the injectivity radius.
4645: This location gives the most symmetric perceived shape to the
4646: Dirichlet
4647: domain which one might expect to lead to the most conservative bounds.
4648: It seems fair to assume that
4649: moving the observer to a thinner region in the manifold
4650: for instance will only amplify asymmetries and the constraint on 
4651: the correlation function should only be more severe.
4652: However, others have argued that locating the Earth away from the local
4653: maximum of the injectivity radius can increase the likelihood fit of
4654: a CH model to the data.  For instance, 
4655: to check the effect of the inhomogeneity,
4656: Inoue moved the observing point for the same model m004(-5,1)=Thurston 
4657: manifold and performed the Baysean analysis as Bond, Pogosyan, and
4658: Souradeep did.  He found
4659: that there are a few choices of the 
4660: position and orientation for which the likelihood is much 
4661: larger than that of the infinite counterpart \cite{inoueprog}
4662: (as did \cite{bpsII}).
4663: The best-fit positions are scattered in the
4664: manifold and far from the local maximum of the injectivity
4665: radius. 
4666: 
4667: In any case, the results of \cite{{bpsI},{bpsII}}
4668: are presented as the relative likelihood of a given
4669: model
4670: in comparison with a simply connected hyperbolic CDM model with 
4671: the same $\Omega_m=\Omega_0$.
4672: They uniformly find that the compact hyperbolic models have very
4673: small relative likelihoods as compared with the simply connected models.
4674: The rare exception occurred near $\Delta\eta\approx r_{+}$ with a
4675: particular
4676: orientation.  The statistical significance is unclear since a
4677: fortuitous
4678: alignment of the measured fluctuations with simulated
4679: topological images could enhance the likelihood when taken over 
4680: all realizations which reinforce this correlation.
4681: The authors
4682: defer conclusions to future tests such as the circle method.
4683: Although again, Inoue argues the statistical significance {\it is}
4684: clear.  If the $C_\ell$s fit the data poorly, there will be no 
4685: orientation of the manifold that leads to an alignment of measured
4686: fluctuations with topological images.  Put another way, it may be
4687: unfair to suggest that a good fit to the data is a fortuitous
4688: alignment instead of acknowledging this good fit to be a good fit.
4689: 
4690: Bond, Pogosyan, and Souradeep 
4691: draw the general conclusion that $d_{\cal M}/2 > 0.7 \Delta\eta$.
4692: The compact models they tested are excluded at the $3\sigma$ level
4693: with the exception of those with very special orientations.
4694: By contrast, if they were to only analyze the statistical likelihood
4695: of the $C_\ell$ they would mistakenly conclude that the compact models
4696: were preferred at the $1\sigma$ level
4697: over the simply connected cosmology.  They emphasize
4698: that
4699: the error bars on the $C_\ell$'s are huge because of the exaggerated
4700: cosmic variance in the topologically connected models.
4701: It may also be worth noting here that there is some argument over
4702: the interpretation of these error bars (see Ref. \cite{inoueprog}).
4703: 
4704: Due to the global breaking of homogeneity, they find the variance is 
4705: spatially dependent.  This is another reason to be weary of
4706: conclusions based on $C_\ell$ alone.  They found characteristic loud
4707: spots,
4708: that is regions in the sky with larger variance than others.
4709: Loud regions correspond to intersections of the SLS through smaller
4710: regions of the fundamental domain.  The regions are geodesicaly small,
4711: that is to say, there are shorter geodesics relative to other regions in
4712: the volume.  Unless the region is very very small, the ISW can obscure
4713: this particular feature.
4714: These loud and quiet regions are familiar from the cusp topology
4715: studied in Ref. \cite{lbbs}.  
4716: There, even the ISW cannot compensate for the deepest regions
4717: of the cusp.
4718: 
4719: Another attitude to take would be to argue that if we do live in 
4720: a compact hyperbolic manifold, some orientation is necessarily going
4721: to be much more likely, namely the right one.
4722: The special status of the `correct' orientation for our manifold,
4723: assuming it is compact hyperbolic, was taken seriously
4724: by one of the pioneers in cosmic
4725: topology, Helio Fagundes \cite{{morehf},{morehf2}}.  He reasoned that the
4726: particularly significant hot and cold spots in the COBE maps 
4727: found by Cay\'on and Smoot \cite{cas} may shed light on the orientation 
4728: of the manifold. These loud spots may be patches of high and low regions
4729: in the physical density, as opposed to being just statistical fluctuations.
4730: Since the density fluctuations eventually evolve into large-scale structure,
4731: then the hot and cold spots should correspond to physical 
4732: superclusters
4733: and voids respectively.  Fagundes used this idea to try to match sources
4734: in catalogs of galaxy superclusters and voids and thereby fix the orientation
4735: and location of the earth in the manifold.  His algorithm 
4736: begins with a point $P^\prime$ centered on one of the spots
4737: isolated in Ref.\ \cite{cas} as illustrated in the schematic figure
4738: \ref{scheme}.  The point $P^\prime \in SLS$. He then 
4739: maps ghost images of $P^\prime=\gamma P$ for $\gamma \in \Gamma$
4740: so the $\gamma$ are composite words built from the
4741: face-pairing matrices.  He tries to alien the image points with 
4742: voids and superclusters for different orientations of the space and 
4743: different basepoints, i.e. different locations of the earth.
4744: He performed these scans for
4745: the ten smallest compact hyperbolic manifolds known
4746: \cite{hw}.
4747: 
4748: \begin{figure}
4749: \centerline{\psfig{file=scmbFig1.ps,width=2.in}}
4750: \vskip 5truept
4751: \caption{
4752: %cite {morehf}
4753: This figure was supplied courtesy of H. Fagundes.  It shows the
4754: fundamental polyhedron as well as a copy of the fundamental polyhedron
4755: that intersects the spherical surface of last scattering.  The point of
4756: intersection identifies the original location of a CMB spot in the universe.}
4757: \label{scheme}
4758: \end{figure}
4759: 
4760: 
4761: \subsection{Geometric methods in hyperbolic space}
4762: \label{geomhyp}
4763: 
4764: Given the likelihood analysis of the previous section, should we generically
4765: conclude that all compact hyperbolic spaces must be large relative to the
4766: observable universe?  The answer is really ``no'' 
4767: since there is so much model
4768: dependence in any of the direct methods of the previous section.  We 
4769: cannot be confident about our assumptions for the initial fluctuation
4770: spectrum, the choice of manifold, the orientation, location of observer,
4771: local parameter values or even the statistics themselves.
4772: Remembering that 
4773: there are an infinite number of manifolds to consider and 
4774: the ambiguities in the model parameters, statistical conclusions 
4775: are limited to the specific.
4776: These attributes beg for a template independent method of searching
4777: for topology.  Much like gravitational lensing, we might hope to simply
4778: look at the sky and see evidence of topological lensing without prior
4779: assumptions about the shape of the lens.  There is hope for such model
4780: independent observations as exemplified 
4781: in the circles in the sky described below.
4782: Generically, we distinguish these
4783: geometric methods which search for
4784: patterns from statistical methods which rely on a specific model.
4785: 
4786: 
4787: \subsubsection{Circles in the Sky}
4788: \label{obscirc}
4789: 
4790: Possibly the nicest geometrical observation made thus far has been 
4791: the prediction of
4792: circles in the sky of Cornish, Spergel, and Starkman 
4793: \cite{{css1},{css2},{css_cqg}}.
4794: It has quickly become a popular topic in 
4795: conversations on
4796: the topology of the universe.  
4797: The circles are most easily seen in the tiling representation of a
4798: compact
4799: space.  Each copy of the Earth will 
4800: come complete
4801: with its own surface of last scatter.  If the two images of the Earth are near
4802: enough,
4803: these identical copies of the SLS will intersect.
4804: The intersection of two 
4805: spheres occurs along a circle.  
4806: Since the observers at the center of the intersecting
4807: spheres are actually just copies of one observer,  the circle of
4808: intersection must always come in pairs as
4809: illustrated in fig.\ \ref{circles}.  To emphasize, 
4810: we will not look up in the sky
4811: and see intrinsic circles in the microwave; that is,
4812: the circle pairs do not have 
4813: identical temperatures along a circle but rather the temperature
4814: varies identically when taken along correlated circles
4815: \cite{css_cqg}.  An illustration of the location of circle pairs
4816: in a finite flat torus is shown in figure \ref{toruscirc} from 
4817: Ref.\ \cite{cornweeks}.
4818: 
4819: 
4820: \begin{figure}
4821: \centerline{\psfig{file=circles.eps,width=2.in}}
4822: \vskip 5truept
4823: \caption{}
4824: \label{circles}
4825: \end{figure}
4826: 
4827: 
4828: \begin{figure}
4829: \centerline{\psfig{file=hemi.eps,angle=0,width=3.25in}}
4830: \voffset -0.5truein
4831: \vskip 5truept
4832: \caption{This figure was supplied courtesy of N. Cornish
4833: %cite cornweeks
4834: and shows the northern and southern hemispheres of the microwave
4835: sky in flat hypertorus.  There are 13 matched circle pairs indicated
4836: by black lines.
4837: }
4838: \label{toruscirc}
4839: \end{figure}
4840: 
4841: The most important aspect of the circles approach is that it applies to all
4842: multiconnected topologies and does not require a template nor any
4843: a priori assumptions
4844: about a model.
4845: All compact spaces will have circles if any part of the geometry
4846: is smaller than the SLS.  The radius, number, and
4847: distribution of the circles will vary from space to space and hence
4848: the topology can be reconstructed from the circle pairs.
4849: A distribution of clone images in the Thurston space can be seen in 
4850: figure \ref{thurcirc}.
4851: Since the shape of the Dirichlet domain is not a topological invariant
4852: but rather depends on the location and orientation of the observer,
4853: even the Earth's location in the universe can be deduced.
4854: 
4855: 
4856: \
4857: \begin{figure}
4858: \vspace{60mm}
4859: 
4860: \special{psfile=thur.eps angle=-90 hscale=55
4861: vscale=55 voffset=250 hoffset=35}
4862: \special{psfile=thurston3d.ps angle=0 hscale=65
4863: vscale=65 voffset=-170 hoffset=20}
4864: 
4865: \vspace{2mm}
4866: 
4867: %\begin{figure}
4868: %\centerline{\psfig{file=thur.eps,angle=-90,width=2.5in}}
4869: %\centerline{\psfig{file=thurston3d.ps,angle=0,width=2.5in}}
4870: %\vskip 5truept
4871: \caption{This figure was supplied courtesy of N. Cornish
4872: %cite css_cqg
4873: and illustrates
4874: just how many circles are to be expected.
4875: A fundamental domain for the Thurston space is shown.  The distribution
4876: of points mark our clones out to a distance of three times the curvature
4877: radius.
4878: The large points are within one
4879: curvature radius, the medium sized points are within two
4880: curvature radii and the small points are within three curvature
4881: radii.  Each of these clones will have a clone surface of last scatter.
4882: The clones which generate circle pairs will
4883: depending on the size of surface of last scatter compared to the
4884: clone distance and therefore depends on the value of $\Omega_0$.}
4885: \label{thurcirc}
4886: \end{figure}
4887: 
4888: 
4889: A statistical scan of the sky must be performed to draw the
4890: correlated circles out of the maps.  The circle pairs will be
4891: completely hidden.  To pull them out, consider two rings
4892: each with angular radius $\alpha$ and with relative phase $\phi_*$
4893: centered on arbitrary points 
4894: $\vx $ and ${\bf \vec y}$.  To test whether these arbitrary rings are 
4895: in fact correlated circles of intersection,
4896: the comparison statistic
4897: 	\be
4898: 	S(\phi_*)={\left < 2T_1(\pm \phi)T_2(\phi+\phi_*)\right > \over
4899: 	\left < T_1(\phi)^2+T_2(\phi+\phi_*)^2\right >}
4900: 	\label{circstat}
4901: 	\ee
4902: has been proposed
4903: where $<>=\int_0^{2\pi} d\phi $ 
4904: with S range $[-1,1]$
4905: \cite{css_cqg}.  Perfectly matched circles have $S=1$ while an ensemble
4906: of uncorrelated circles will have a mean value of $S=0$.
4907: (Roukema includes the small scale Doppler effect in a slight alteration of the
4908: statistic \cite{boudcirc}.)
4909: Orientable topologies will have clockwise-anticlockwise correlations while
4910: non-orientable topologies will have a mixture of clockwise-clockwise and
4911: clockwise-anticlockwise correlations.
4912: In flat space, matched circle pairs have angular radius
4913: 	\be
4914: 	\alpha={\rm arccos}\left ({X\over 2\Delta \eta}\right )
4915: 	\ee
4916: with $X$ the distance between the Earth and its image.
4917: In hyperbolic geometry
4918: 	\be
4919: 	\alpha={\rm arccos}\left({\cosh X-1\over \sinh X\tanh \Delta \eta}
4920: 	\right ).
4921: 	\ee
4922: There are no circle pairs if the image is too far for the spheres
4923: to intersect and the
4924: expressions are invalid for $X>\Delta \eta$.
4925: The number of images grows exponentially with $X$ in hyperbolic space
4926: so that
4927: most circles have small
4928: radii, although the statistic works best for large circles. 
4929: 
4930: Noise will degrade the circle
4931: statistic so that $S\ne 1$.  The experimental noise
4932: in each pixel can be approximated as random Gaussian noise 
4933: 	\be
4934: 	P(n)={1\over \sigma_n
4935: 	\sqrt{2\pi}}e^{-n^2/2\sigma_n^2}
4936: 	\ee
4937: with variance $\sigma_n$.  The true temperature fluctuations
4938: are also taken to have a Gaussian distribution with variance $\sigma_s$.
4939: The probability for the matched circle is then
4940: 	\be
4941: 	P^m(S)dS={\Gamma(N)2^{-N+1}\over \Gamma(N/2)^2}
4942: 	{(1+2\xi^2)^{N/2}\over
4943: 	(1+(1-S)\xi^2)^N}(1-S^2)^{N/2-1}dS
4944: 	\ee
4945: with $\Gamma(N)$ the gamma function and $\xi=\sigma_s/\sigma_n$ is
4946: the signal-to-noise of the detector while $N$ is the number of pixels
4947: [?].  For large $N$, the distribution has a maximum
4948: 	\be
4949: 	S^m_{\rm max}={\xi^2\over 1+\xi^2}+{\cal O}(N^{-1}).
4950: 	\ee
4951: The higher the resolution and the signal-to-noise, the better this
4952: statistic will fare.
4953: If the experimental angular resolution is $\delta \theta $, then there are 
4954: $N\simeq 2\pi\sin\alpha/\delta \theta$ data points around each circle of
4955: angular radius $\alpha$.  For
4956: \cb, the signal-to-noise ratio is $\xi=2$, $\delta \theta =10^o$ and 
4957: $N\simeq 36\sin\alpha $ pixels while for MAP, $\xi\simeq 15$, 
4958: $\delta \theta=0.2^o$ 
4959: and $N\simeq 1800\sin\alpha $
4960: in its highest frequency channel.
4961: Using MAP
4962: parameters, only circles with $\alpha >4^o$ are detectable.
4963: 
4964: To make a relative comparison with a probability distribution 
4965: for unmatched circles they advocate
4966: 	\be
4967: 	P^u(S)dS=
4968: 	{\Gamma(N)2^{-N+1}\over \Gamma(N/2)^2}
4969: 	(1-S^2)^{N/2-1}dS
4970: 	\ee
4971: which is centered at $S=0$ with a FWHM$\simeq(8\ln 2/n)^{1/2}$.
4972: The unmatched probability distribution $P^u$ was derived
4973: by assuming the temperature at each point is an independent Gaussian
4974: random variable. There are some weaknesses in this assumption since
4975: there are known correlations in the CMB.
4976: 
4977: Cornish and Spergel applied
4978: the circles test to maps generated with their numerically
4979: isolated eigenmodes of section \ref{directhyp}.
4980: For a realization of the Weeks space they found the uncorrelated ISW 
4981: clouded the statistic resulting in a poor match
4982: for circle pairs.  To combat this pollution they remove all power in
4983: modes below $\ell=21$ and find a substantial improvement in the match
4984: with values of $S\approx 0.9$.  Their expectations for the future
4985: satellite missions are high.
4986: 
4987: 
4988: The circles method was first applied to simulated maps by Bond,
4989: Pogosyan
4990: and Souradeep using the method of images of section \S \ref{obsmi}.
4991: From
4992: their correlation function they are able to make full sky maps.
4993: Doing so they search for circles using the cross-correlation
4994: coefficient
4995: between fluctuations along two circles $C_1$ and $C_2=g C_1$,
4996: 	\be
4997: 	\rho_{12}\equiv
4998: 	{\left < \delta T(\vx)\delta T(g \vx)\right >
4999: 	\over \left [\left <\delta T(\vx)^2 \right >
5000: 	\left <\delta T(g \vx)^2 \right >\right ]^{1/2}
5001: 	}
5002: 	\ee
5003: with $\vx\in  C_1$ and $g \vx\in C_2$.
5004: A statistical average is implied by the angular brackets.  For one
5005: given realization an integration over $\vx $ along the circles is
5006: performed
5007: instead.  They find $\rho_{12}$ is in the range $0.6-0.95$ for
5008: $\Omega_0=0.9$ and m004(-5,1)
5009: and $0.2-0.6$ for $\Omega_0=0.6$ for v3543(2,3).  
5010: Notice that $\Omega_0$ is selected for each manifold so that 
5011: ${\cal V}$ is comparable to the volume of the SLS.
5012: The matches they find are good even at \cb resolution.
5013: The correlations
5014: along circles gets worse as the ISW contribution is enhanced at 
5015: low $\Omega_0$.
5016: 
5017: In addition to the circular intersections of the boundary of the SLS,
5018: there are
5019: intersections of the volume.
5020: Let S represent the collection of points contained within the SLS.  Then the
5021: intersection
5022: of $g S  \cup S$ defines a lens-shaped region
5023: (see their figure) in the volume
5024: and must be identical
5025: to the intersection of $S\cup g^{-1} S$.
5026: As long as $r_{-}<\Delta \eta$, then there will be lens-shaped
5027: intersections 
5028: of the volumes and
5029: circular intersections of the copies of the SLS.
5030: Since the anisotropy occurs along the entire line of sight,
5031: Bond, Pogosyan and Souradeep 
5032: consider the correlations in the full lens-shaped volume.  
5033: Based on 
5034: the \cb constraints obtained via the method of images
5035: as described in section \S \ref{obsmi} they took
5036: the topology scale
5037: comparable
5038: to the SLS.  They argue
5039: that correlations must then be very near the faces of
5040: the Dirichlet domain and so the correlations provide 
5041: a quick sketch of the
5042: shape
5043: of the universe \cite{bpsII}.
5044: 
5045: The volume intersections are relevant since, contrary to intuition, 
5046: they found the ISW is not entirely uncorrelated.
5047: While there are an infinite number of lines of sight which share
5048: at least one common point, there are also pairs of lines which have
5049: segments in common and so an enhanced correlation.  Every pair of 
5050: lines of sight which are directed toward the center of matched
5051: circles,
5052: necessarily contain segments of identical points
5053: (see fig.) and leads to correlated patterns.
5054: %[still confusing because the photon is not traveling along a constant
5055: %time hypersurface.  in other words might be a identical points at a
5056: %given
5057: %time slice but the ISW does not happen at a fixed time slice, not 
5058: %instantaneous.]
5059: These substantial
5060: anti-correlated features tend to lie at the centers of matched 
5061: circles.
5062: The anticorrelation comes from the interference term between the 
5063: surface Sachs-Wolfe effect and the ISW.
5064: As described in section \S \ref{obsmi}, their likelihood
5065: comparison of a handful 
5066: of spaces to the existing data led to pessimistic conclusions.
5067: 
5068: 
5069: Another
5070: application of the circles method using \cb
5071: sought to identify an asymmetric flat 3-space
5072: \cite{boudcirc}. \cb is not ideal for detecting circles in hyperbolic
5073: space since the ISW accounts for most of the power in the
5074: \cb range of detection.  However, for flat spaces 
5075: with no cosmological constant the sky is determined by the
5076: surface Sachs-Wolfe effect alone without the obscuring effects of the ISW.
5077: Despite the low resolution, \cb might still see circles if we live
5078: in a compact flat space.
5079: 
5080: Roukema considered a specific asymmetric torus
5081: which was put forth as a candidate
5082: model to match cluster observations
5083: \cite{{roukemaedge},{roukemablanloeil}}.  
5084: Although he used the recent circles prediction to test his hypothesis,
5085: the symmetries of the space and the geometric approach are reminiscent
5086: of the earlier work of \cite{deO2} described in section \S \ref{geomflat}.
5087: Hot X-ray bright gas in large galaxy clusters were used to search for
5088: topological images \cite{roukemaedge}.
5089: Two clusters at redshifts
5090: $z\sim 0.4$ very nearly form a right angle with the Coma cluster
5091: with very nearly equal arms.  
5092: On the basis of this geometric
5093: relation, they take a toroidal geometry for the universe to explain 
5094: these clusters as topological
5095: images of the Coma cluster.  
5096: The distance from the Coma cluster to CL 09104+4109
5097: and the distance from Coma to the cluster
5098: RX J1347.5-1145 are both $\approx 960h^{-1}$ Mpc for 
5099: zero cosmological constant.  
5100: The third dimension 
5101: is taken to be larger than the diameter of the SLS and hence topological
5102: effects from this direction are essentially unobservable.
5103: The size of the small dimension was taken to be roughly 
5104: $\Delta \eta/13.2$.  Using the circles statistic,
5105: the candidate was ruled out at the 94 \% confidence level, provided
5106: that the cosmological constant is zero.
5107: Specifically, the statistic Roukema used was
5108: 	\be
5109: d\equiv \left<
5110:          {  \left({\delta T \over T}\right)_i 
5111:                  - \left({\delta T \over T}\right)_j 
5112:      \over { 
5113: \left\{ \left[\Delta \left({\delta T \over T}\right)\right]_i^2 +
5114:       \left[\Delta \left({\delta T \over T}\right)\right]_j^2 
5115: \right\}^{1/2}  
5116: } } \right>
5117: 	\label{bstat}
5118: 	\ee
5119: where $[\Delta(\delta T/T)]_i$ are the observational error estimates 
5120: on the temperature fluctuations 
5121: $(\delta T/T)_i$ as estimated by the COBE team.
5122: This statistic directly tests the consistency 
5123: of temperature values within observational error bars \cite{boudcirc}.
5124: If the ISW effect is treated as noise as in Ref.\ \cite{boudcirc}, 
5125: the application of the
5126: circles principle using the statistic \ref{bstat}
5127: can still enable rejection of a specific partly compact model  
5128: or
5129: show that a model one tenth of the horizon diameter is consistent
5130: with the COBE data.
5131: 
5132: 
5133: In general, there are obstacles to implementing the circle method in practice.
5134: For one, the much emphasized ISW effect is not correlated
5135: in this way and so can obscure the circle pairs. Other problematic
5136: effects include the velocity and thickness of the SLS.  
5137: In Ref. \cite{boudcirc}, it was noted that 
5138: many circles are partially lost along with 
5139: $20^\circ$ galactic cut and that
5140: both detector noise and foreground contamination posed 
5141: difficulties for the circle detections.
5142: Realistically it may not be possible to observe circles even if they 
5143: are there and this will be the challenge faced in realistic
5144: analyses of the future satellite data.
5145: 
5146: 
5147: \subsubsection{Pattern Formation}
5148: \label{obspat}
5149: 
5150: Developmental biology and condensed matter physics have long exploited
5151: pattern formation induced by periodic boundary conditions.  Since compact
5152: topologies can be understood as a set of intricate boundary conditions,
5153: pattern formation has a cosmological analogue.  The 
5154: emergence of patterns in the sky are not so clear since many modes
5155: are competing for attention.  The superposition of many otherwise
5156: distinct geometric patterns can lead to something apparently random.
5157: The task of geometric methods is to separate the patterns out of the sky.
5158: The statistic of eqn.\ (\ref{circstat}) manages to draw out circles.
5159: Other patterns 
5160: in addition to the circles can emerge as described in this section.
5161: 
5162: The patterns are best siphoned off a CMB map by scanning for correlations
5163: \cite{pat}.  As an example consider a map of the antipodal correlation
5164: 	\be
5165: 	A(\hat n)=C_{\cal M}(\hat n,\hat n^\prime)
5166: 	\ee
5167: which measures the correlation of fluctuations received from opposite 
5168: points on the SLS \cite{pat}.
5169: In a simply connected space opposite sides of the SLS should have no 
5170: communication between them and a
5171: map of antipody would generate nothing
5172: more than a monopole.  It is possible that accidental correlations appear
5173: at random but no geometric structure would emerge.
5174: By contrast, if the universe is topologically connected, then 
5175: points on the manifold which seem to be far apart in the tiling picture
5176: may actually be quite close together in the fundamental domain.  Therefore
5177: opposite points on the SLS may be strongly correlated, may in fact be
5178: the same point.  This is another example of ghost images but in an 
5179: antipodal map collections of ghosts are caught and a picture of the 
5180: symmetries of the space emerges \cite{pat}.
5181: 
5182: \begin{figure}
5183: \centerline{\psfig{file=close.eps,width=2.in}}
5184: \vskip 5truept
5185: \caption{}
5186: \label{close}
5187: \end{figure}
5188: 
5189: 
5190: The size of a spot can be estimated at the Silk damping 
5191: scale below which 
5192: fluctuations have smoothed some.
5193: The angular size of these spots are too small for \cb to have detected
5194: but will be visible to the high resolution MAP and {\it Planck
5195: Surveyor}.
5196: 
5197: Although the search for pattern formation in correlated maps is model
5198: independent, a zoo illustrating the variety of structures compact
5199: manifolds produce can be built with some simple approximations.
5200: The correlation between two points 
5201: on a compact manifold 
5202: can be estimated as the correlation they would have on the universal
5203: cover given their minimum separation:
5204: \begin{equation}
5205: C_{{\cal M}}({\hat n},{\hat n}^{\prime })\approx C^U\left[ d_{{\rm min}}( 
5206: \vec x({\hat n}),\vec x^{\prime }({\hat n}^{\prime }))\right] \ \ ,
5207: \label{eq:approx}
5208: \end{equation}
5209: where $C^U$ is the correlation function on the universal cover
5210: and $d_{{\rm min}}$ is the minimum distance between the two points in the
5211: topological space.
5212: The estimate is 
5213: effectively the lowest order term in the method of images.  It is inadequate
5214: for use in a likelihood analysis but is sufficient for predicting the types
5215: of patterns which emerge from topological lensing.
5216: The images of a given point out 
5217: to order $m$ are found with the generators of the identifications as
5218: \begin{equation}
5219: \vec y_{k_m,..,k_i}=\prod_i^mg_{k_i}\vec x^{\prime }(\hat n^{\prime })\ \ .
5220: \end{equation}
5221: The image point which lands closest to $\vec x(\hat n)$ determines $d_{{\rm  
5222: min}}$. 
5223: 
5224: As an example, we show the antipodal map for a $2\pi/3$-twisted hexagonal
5225: prism in fig.\ \ref{hexpat2}.
5226: For the 
5227: antipodal map we prefer the orthographic projection which shows the genuine
5228: shape of the surface of last scattering
5229: instead of the Aitoff projection customary in 
5230: $\delta T({\hat n})/T$ maps.
5231: Notice the clear hexagonal face drawn out by the correlated map.
5232: Since antipody is symmetric under $\pi$, the back is a copy of the front.
5233: For the $\pi/3$-twisted hexagon, if the space is small enough,
5234: pairs of circles appear
5235: in $A(\hat n)$, as
5236: shown in the right-most panel of fig.\ \ref{hexpat2}.  These are the circles
5237: in the sky of section \S \ref{obscirc}.
5238: 
5239: \begin{figure}[tbp]
5240: \centerline{
5241: %\psfig{file=../../top/pat/hex_.6_pat.ps,width=2.2in}
5242: %\quad\quad
5243: %\psfig{file=../../top/pat/hex2_.24_1_pat.ps,width=2.2in}
5244: \psfig{file=hex_.6_pat.eps,width=2.2in}
5245: \quad\quad
5246: \psfig{file=hex2_.24_1_pat.eps,width=2.2in}
5247: } \vskip 15truept
5248: \caption{Left:  Orthographic projection of ${\rm A({\hat n})}$ for a hexagonal
5249: prism with $L= 0.6 \Delta \eta$. 
5250: Right: ${\rm A({\hat n})} $ for $\pi/3$-twisted hexagonal prism.
5251: The length of the prism
5252: direction
5253: is $.24$ while $L=1$. There are circles }
5254: \label{hexpat2}
5255: \end{figure}
5256: 
5257: 
5258: 
5259: 
5260: A compact hyperbolic space shows distinct patterns.  
5261: The compact icosahedron known as
5262: the Best space after the mathematician who identified the manifold
5263: provides the best testing ground \cite{best}.
5264: The map of $A(\hat n)$ is shown in the left of
5265: fig.\ \ref{bestant}.  Antipody outlines
5266: pairs of identified triangular faces and also locates
5267: circles.  Clearly a symmetry group for the Best space is located in this
5268: map.
5269: Another correlation function is also shown which compares one point on the
5270: SLS to the rest of the sphere.  The point selected is a copy of the origin
5271: and so reflects the most symmetric observation of the fundamental domain.
5272: Another example is given by the antipodal map for the Thurston space in 
5273: fig.\ \ref{thursant}.
5274: Some of the correlated features such as the arcs 
5275: in fig.\ \ref{thursant} may be secondary correlations
5276: and it is not clear they will ever be bright enough to be observed.
5277: 
5278: 
5279: \begin{figure}[tbp]
5280: \centerline{\psfig{file=besti.ps,width=2.25in}}\vskip 15truept
5281: \caption{Dirichlet domain for the Best space.
5282: }
5283: \label{best}
5284: \end{figure}
5285: 
5286: \begin{figure}[tbp]
5287: %\centerline{{\psfig{file=../../top/spots/best_.3_pat.eps,width=2.5in}}
5288: %\quad\quad
5289: %\psfig{file=../../top/pat/best_2.3_corr.eps,width=2.5in}}\vskip 15truept
5290: \centerline{{\psfig{file=best_pat.eps,width=1.8in}}
5291: \quad\quad
5292: \psfig{file=best_2.3_corr.eps,width=2.5in}}\vskip 15truept
5293: \caption{Left: $A(\hat n)$ for the Best space with 
5294: $\Omega=0.3$.  Right: The 
5295: point-to-sphere correlation.}
5296: \label{bestant}
5297: \end{figure}
5298: 
5299: 
5300: \begin{figure}[tbp]
5301: %\centerline{{\psfig{file=../../top/imogen/thurs_ant.eps,width=2.5in}}}
5302: \centerline{{\psfig{file=thurs_ant.eps,width=2.5in}}}
5303: \vskip 15truept
5304: \caption{Antipody in the Thurston manifold.}
5305: \label{thursant}
5306: \end{figure}
5307: 
5308: In fairness, it is difficult to know if any of these patterns will really
5309: be measurable in a realistic experiment with physical complications such
5310: as the thickness of the surface of last scatter, additional Doppler effects,
5311: noise, fictitious correlations etc..  
5312: To read the correlations from the future data, real space statistics will need
5313: to be developed which handle smoothings, subtractions of low order multipoles,
5314: noise and fictitious correlations.  While some statistics have been promoted,
5315: a realistic approach will likely develop only when the data is actually
5316: available.
5317: 
5318: 
5319: \newpage
5320: \section{Beyond Standard Cosmology}
5321: \label{extradsect}
5322: 
5323: \def\theequation{\thesection.\arabic{equation}}
5324: 
5325: \setcounter{equation}{0}
5326: 
5327: \subsection{The twin paradox and compact time}
5328: 
5329: So far we have ignored the issue of time.  We have explicitly considered
5330: spacelike hypersurfaces $\Sigma$
5331: of constant curvature foliated by a natural
5332: conformal time.  Compactification of these surfaces leads to the
5333: pictures we have described without compactifying time.  However,
5334: an observer moving on an inertial worldline which is not at rest
5335: with respect to the cosmic expansion will perceive an identification
5336: which mixes spacetime coordinates 
5337: as dictated by the Lorentz transformations $\Lambda $ so that
5338: $\bar x=\Lambda x$ with $x=(\eta,\vec x)$ comoving coordinates.
5339: As a result, a compactification of $\Sigma$ in the comoving coordinates
5340: of the form $(\eta,\vec x)\rightarrow (\eta,\vec x+\vec L)$ for instance
5341: will
5342: result is an identification of 
5343: a time shift as measured by the non-comoving observer of
5344: $(1-\beta^2)^{-1/2}\left (\eta -\vec \beta \cdot \vec x \right )
5345: \rightarrow (1-\beta^2)^{-1/2}\left (\eta -\vec \beta \cdot (\vec x
5346: +\vec L)\right )$ where $\beta$ is the velocity relative to 
5347: $\Sigma$.
5348: 
5349: 
5350: 
5351: \begin{figure}
5352: %\centerline{\psfig{file=../../top/sr/fund2.eps,width=2.in}}
5353: \centerline{\psfig{file=fund2.eps,width=2.in}}
5354: \vskip 5truept
5355: \caption{An observer $O$ at rest with respect to the compact spacelike
5356: hypersurface $\Sigma$ and an observer $\bar O$ moving at constant
5357: velocity with respect to $\Sigma$ on a
5358: periodic orbit of the torus.  The orbit
5359: corresponds to 
5360: $x_{\rm end}=T_yT_x^2x_{\rm start}$.
5361: }
5362: \label{fund2}
5363: \end{figure}
5364: 
5365: As an explicit demonstration, consider the twin paradox
5366: \cite{{jdbi},{jefftwin},{uzan}}.  Let $O$ be at
5367: rest in a comoving frame where a flat spacelike hypersurface is compactified
5368: into a torus.  Let $\bar O$ be an inertial observer on a periodic
5369: orbit with respect to $O$
5370: as in fig.\ \ref{fund2}.  
5371: The periodic orbit will obey the boundary
5372: condition
5373: $\bar x=\Lambda x \rightarrow \Lambda \gamma x$ where $\gamma $ is 
5374: the corresponding word.
5375: From \S \ref{tools} we have $\gamma=T_y T_x^2$
5376: (to include the physical time in the embedded coordinates
5377: and in $T_x, T_y$ see Ref.\ \cite{jdbi}).
5378: According to $\bar O$, both space and time points have been identified.
5379: As a result it becomes impossible for $\bar O$ to synchronize her clocks
5380: \cite{peters}.
5381: The lack of synchronicity
5382: will be given by the time component of $\Lambda(1-\gamma)x$ \cite{jdbi}.
5383: 
5384: Since both $O$ and $\bar O$ are inertial, by the principles of 
5385: relativity, each should believe the other's clocks run slower and therefore
5386: each could expect the other to be younger at their reunion 
5387: leading to a paradox.
5388: However, the topological identification breaks the general invariance
5389: and selects a preferred frame, namely the frame in which the topological
5390: identification is purely on $\Sigma$.  In that frame clocks can be
5391: synchronized and the volume of space looks smallest.  The 
5392: observer on the periodic orbit, though inertial, will discover that
5393: their clocks cannot be synchronized and this additional
5394: shift leads them to compute 
5395: an older age for their twin.
5396: Both agree the twin at rest with respect to $\Sigma $ is older.
5397: 
5398: This simple thought experiment emphasizes that 
5399: compact topology selects a
5400: preferred frame, namely the frame in which the 
5401: universe looks smallest and in which observers can synchronize 
5402: their clocks
5403: \cite{peters} (see also Refs. \cite{lh}).
5404: To generalize to curved space,
5405: $\Lambda$ can be replaced by an appropriate diffeomorphism and 
5406: the spacetime topology generalizes to
5407: ${\cal M}_c=R\otimes G/\Gamma$.
5408: 
5409: %If quantum gravity can explain the creation of a universe from nothing,
5410: %one might intuitively expect small volume universes to be more likely,
5411: %although this is a difficult prediction to formalize.  In any case, 
5412: Finite 
5413: spaces reverse some of Copernicus' philosophical advances by selecting
5414: a preferred location at the center of the space, a preferred
5415: observer at rest with respect to the compactification and a preferred
5416: time.  While Copernicus may have removed us from the center of the universe, 
5417: topology 
5418: puts some observers back there.
5419: %selects some observer, if not actually us, as special.
5420: 
5421: This also raises the question of compactifying time outright.
5422: Probably time could be compactified so that there were closed
5423: timelike curves that were not causality violating.  Very 
5424: restrictive possibilities would result since only events which could
5425: repeat ad infinitum would obey the boundary conditions.
5426: It would be hard to envision a compact time model being consistent
5427: with the laws of thermodynamics, except perhaps on a cosmological
5428: timescale.  The compact time scale would have to be much shorter or
5429: much larger than biological timescales or
5430: no children could sensibly be 
5431: born since they would somehow have to grow young again.  
5432: A universe could go through a big
5433: bang and eventual big crunch only to repeat the history of the
5434: cosmos with another big bang.  The same galaxies, stars and the same
5435: people would be born, live and die.  Fated to repeat their
5436: paths ad infinitum.  Quantum gravity may reset the initial conditions
5437: at each big bang allowing new galaxies, new stars and new organisms
5438: to form, sparing us from a relentless cosmic boredom.
5439: The fanciful possibility of a compact time magnifies the already strange
5440: and distinct nature of time.
5441: 
5442: \subsection{Extra dimensions}
5443: \label{stringsection}
5444:  
5445: 
5446: 
5447: 
5448: 
5449: 
5450: 
5451: 
5452: 
5453: 
5454: %\subsection{Quantum Creation of a Universe}
5455: %\label{qcsection}
5456: 
5457: Topology it has been suggested is a discrete feature of space
5458: \cite{lum}.  As such, it may be better integrated in a quantized
5459: theory of gravity while there is no prediction for topology in 
5460: classical relativity.  
5461: The earliest attempts at creating a finite universe grew out
5462: of semiclassical quantum cosmology \cite{{carlip},{gibbons},{hfq},
5463: {hfq2}}.
5464: It may seem intuitive that a smaller universe should be easier to 
5465: create from nothing than a larger universe.  Therefore the probability
5466: for creating a small, finite cosmos may be relatively high, if only
5467: we could compute the wave function.  However, technical and conceptual
5468: difficulties dog the semiclassical approach such as defining a measure
5469: on the space of states, normalizing the wave function etc.
5470: A convincing statement about the topology of the universe will very likely
5471: require a fully quantized theory of gravity.
5472: 
5473: Interestingly, additional dimensions
5474: have featured prominently in attempts to 
5475: quantize gravity.  These extra dimensions are always topologically
5476: compact and very small.
5477: While no quantum gravity theory is yet able to predict the topology 
5478: of space, the possibility that compact
5479: internal dimensions will have topologically compact external
5480: (that is, large) dimensions is certainly alluring.
5481: Ultimately, a fundamental theory should predict the global topology
5482: of the entire manifold whether it be $3D$ or $11D$.  In the meantime
5483: the hierarchy between small and large dimensions remains mysterious
5484: although some recent suggestions have created a bit of a stir
5485: \cite{randallco}.
5486: 
5487: Kaluza-Klein theories introduced extra compact dimensions in an
5488: attempt to unify fundamental theories of physics \cite{kk}.
5489: Upon compactification, the radii of the small dimensions behave as scalar
5490: and tensor field theories.  More modern string theories naturally
5491: invoke extra dimensions in a manner reminiscent of these early
5492: Kaluza-Klein models.
5493: Recent fervor in string phenomenology has involved compact extra dimensions
5494: of moderate to large size
5495: in an attempt to explain the hierarchy problem in standard particle
5496: physics.  The hierarchy problem questions the 
5497: disparity in scales from the Planck mass of $10^{19}$ GeV to the 
5498: mass of the electron at a few
5499: eV.  A unification of fundamental theories has to 
5500: naturally justify this span over $25$ decades of energy scales.
5501: As in Ref.\ \cite{glennco}  we consider a spacetime with 
5502: $4+{\cal N}$ dimensions of the form ${\cal M}=R\otimes M^3\otimes
5503: {\cal M}^{\cal N}$ where $R$ represents time,
5504: $M^3$ is a constant curvature Friedman-Robertson-Walker metric and 
5505: ${\cal M}^{\cal N}=G^{\cal N}/\Gamma$ is an 
5506: ${\cal N}$-dimensional compact internal space.  The 
5507: $(4+{\cal N})$-dimensional gravitational action becomes
5508: 	\be
5509: 	A=\int d^{(4+{\cal N})}x\sqrt{-g^{(4+{\cal N})}}
5510: 	{\cal R}^{(4+{\cal N})} M_{*}^2.
5511: 	\ee
5512: In the simplest Kaluza-Klein picture, integration over the compact
5513: extra dimensions leads to the $(3+1)$-dimensional action we experience
5514: 	\be
5515: 	A\sim \int d^4x \sqrt{-g^{(4)}} {\cal R}^{(4)}M_{pl}^2
5516: 	+ ...
5517: 	\ee
5518: plus additional dynamical terms
5519: where $M_{pl}^2=M_*^{2+{\cal N}}R^{{\cal N}}$ and $R$ is the radius
5520: of the internal dimensions.  The hierarchy between 
5521: $M_{pl}$ and $M_*$ is large if 
5522: $RM_*$ is large \cite{glennco}.  The disparity in energy scales 
5523: then becomes a dynamical and geometric question.
5524: 
5525: The modern ideas inspired by string 
5526: theory involve the localization of matter to a $3$-brane nested
5527: in the full $(4+{\cal N})$-dimensional space.
5528: The scale $M_*$ is expected to be $\sim $TeV as predicted
5529: by supersymmetry.  Since ordinary matter is confined to a $3$-brane, we would
5530: be unaware of these extra dimensions regardless of their size unless we 
5531: try very hard to look for them.  
5532: Some laboratory experiments are underway to probe any additional
5533: dimensions which may be lurking there.
5534: 
5535: The topology of the compact extra dimensions is not well understood.
5536: There are only a few requirements the internal dimensions must 
5537: satisfy in order to break supersymmetry at a physically sensible scale.
5538: The internal spaces fall loosely under the
5539: broad category of Calabi-Yau manifolds.
5540: % \cite{calabi}.  
5541: Nearly all topologies investigated so far are 
5542: modeled on flat geometries.  An outgrowth of cosmic topology has been the
5543: suggestion that these internal dimensions be compact and negatively 
5544: curved \cite{glennco}.  
5545: The hyperbolic internal spaces have some advantages over flat space 
5546: including a less demanding tuning of geometric parameters
5547: and a suppression of 
5548: astrophysically harmful graviton modes from the extra dimensions
5549: \cite{glennco}.
5550: Another advantage of the compact hyperbolic extra dimensions
5551: over compact flat extra dimensions may be that chaotic mixing on finite
5552: hyperbolic manifolds could explain the smoothness and flatness of the
5553: large dimensions as argued in Ref.\ \cite{sst}.  We briefly discuss
5554: chaos in the next section.
5555: 
5556: The profound connection between small dimensions and large
5557: have only begun to be forged.  As many times in the past,
5558: cosmology provides a unique terrain in which to test fundamental 
5559: theories \cite{hormar}.
5560: 
5561: \subsection{Chaos}
5562: \label{chaossect}
5563: 
5564: \def\theequation{\thesection.\arabic{equation}}
5565: 
5566: \setcounter{equation}{0}
5567: 
5568: We have touched upon the chaotic motions of particles on a compact 
5569: hyperbolic space.  Chaos refers to the thorough mixing of orbits 
5570: which show an extreme sensitivity to initial conditions.
5571: Geodesics deviate on a surface of negative curvature as mentioned
5572: in \S \ref{obshyp} so that they are extremely sensitive to 
5573: initial conditions.  The motion becomes fully ergodic upon compactification
5574: of the surface.  Formally, chaos in these Hamiltonian systems
5575: means the geodesics equations are nonintegrable; there is no smooth
5576: analytic function which can interpolate between orbits with different
5577: initial conditions.  
5578: 
5579: Despite the resistance of chaotic systems to conventional 
5580: integration methods a great deal about the structure of phase
5581: space can be determined.  Much like thermodynamics, chaotic systems
5582: can be understood in terms of a set of states dense in the phase space.
5583: For chaos this set is provided by the collection of unstable
5584: periodic orbits which grow exponentially with length.
5585: All aperiodic orbits can be understood
5586: in terms of this special subset.
5587: The periodic orbits pack themselves
5588: into a fractal set in order to fit within the finite phase space.
5589: Recall the fractional dimension maintains the set
5590: at a finite volume but allows for an infinite area.  In this way
5591: fractals try to maximize the information content, so to speak,
5592: while minimizing the volume.  The explicit fractal structure
5593: on a compact octagon was isolated in Ref.\ \cite{jdbi}
5594: 
5595: Pursuing the analogy with thermodynamics, 
5596: entropy can be defined as
5597: 
5598: 	\be
5599: 	h(\mu)=\lim_{k\rightarrow \infty}{1\over k}
5600: 	\sum_i^{N(k)}\mu\ln(1/\mu(k))
5601: 	\label{genen}
5602: 	\ee
5603: The calculation goes like this.  Draw a fundamental domain.  Now use
5604: the generators $g_1...g_n$ to tile $\ah$
5605: with copies.  The measure at order $k$ is then defined as 
5606: the fraction of the total volume where each tile has
5607: the same volume:
5608: 	\be
5609: 	\mu(k)={1\over N(k)}
5610: 	\ee
5611: and $N(k)$ is the number of unique tiles.
5612: The sum 
5613: in eqn.\ (\ref{genen}) reduces
5614: to 
5615: 	\be
5616: 	h(\mu)= \lim_{k\rightarrow \infty}{1\over k} \ln(N(k))
5617: 	=H_T.
5618: 	\ee
5619: This is equivalently the topological
5620: entropy, a symbolic entropy which counts the 
5621: number of accessible states; that is, closed loops or equivalently 
5622: tiles in the tessellation.
5623: 
5624: Recall that
5625: periodic orbits can then be counted symbolically with the homotopy group.
5626: We do not distinguish between loops
5627: of varying lengths if they are homotopic.
5628: The spectrum of periodic orbits so defined are therefore a 
5629: topological feature.
5630: The number of unique words that can be built of length 
5631: $k$ out of a $n$-letter alphabet with $r$ relations 
5632: is equivalent to the number of neighbors at order $k$ in the 
5633: tiling. 
5634: Not all
5635: of these neighbors will be unique.  The repeats are accounted for by
5636: the relations and we know that the number of neighbors at order $k$
5637: is bounded by
5638: 	\be 
5639: 	(2n)k\ge	N(k)\le (2n)^k
5640: 	\label{Nk}
5641: 	\ee
5642: the lower bound being simple spaces such as the flat nonchaotic torus and the
5643: upper bound the unpruned maximally chaotic case.
5644: The topological entropy of the torus is zero while the entropy of an 
5645: unpruned hyperbolic space is $H_T=\ln(2n)$.  The compact octagon was
5646: shown to have a topological entropy of $H_T=\ln 7$ \cite{jdbi}.
5647: 
5648: Another kind of entropy, the metric entropy introduced in 
5649: \S \ref{obshyp}, 
5650: describes how quickly mixing takes place in a chaotic
5651: system and can be expressed as the sum of the positive Lyapunov
5652: exponents: $h=\sum \lambda $.
5653: 
5654: 
5655: This is just a taste of a rich area in dynamical systems theory 
5656: and the reader is referred to the many excellent texts on the 
5657: subject \cite{{ott},{gutz}}.
5658: 
5659: 
5660: 
5661: Although cosmologists gingerly avoid this complex feature of the 
5662: dynamics, chaos does have profound and unavoidable consequences.
5663: One of the original motivations for a compact topology exploited the chaotic
5664: motions.  The chaotic mixing could lead to a dilution of initial
5665: anisotropies leading to the symmetric universe we observe today.
5666: These initial attempts failed since the space could not 
5667: be made small enough to allow for sufficient mixing and still be the
5668: large cosmos we observe today \cite{ellis}.
5669: Variants on this idea fuse the chaotic mixing with an 
5670: $\Omega<1$ inflation model where an early episode of chaotic mixing 
5671: provides the moderate initial conditions needed to 
5672: permit a subsequent inflationary phase to succeed \cite{css3}.
5673: 
5674: We have already encountered the 
5675: nonintegrability in trying to find eigenmodes on compact hyperbolic
5676: cosmologies.  For the initial spectrum of fluctuations, researchers
5677: continue to use the assumption of a flat Gaussian spectrum as motivated by 
5678: inflation.  Still, since the consistency of inflation and an observably small
5679: topology is shaky, a more consistent approach would be to 
5680: examine a distribution of initial fluctuations on a compact space
5681: in the absence of inflation.
5682: The numerical results of \S \ref{directhyp} do suggest
5683: that a Gaussian flat spectrum is in fact natural on
5684: compact hyperbolic spaces, even in the absence of inflation.
5685: While this work is very suggestive, a primordial quantum system has
5686: never been very thoroughly thought through.
5687: Because of the importance of this issue we take a moment to discuss
5688: quantum chaos.
5689: The quantum system is 
5690: relevant to our discussion regardless since the expansion of fluctuations 
5691: will be analogous
5692: to the expansion of the semiclassical wave function.  The stationary 
5693: Schr\"odinger equation is the usual Helmholtz eqn (\ref{helm}).
5694: 
5695: 
5696: The quantization
5697: of the chaotic system is still not fully understood but interesting 
5698: features have been conjectured and confirmed.  
5699: Reminiscent of the Feynman path integral approach, the wave function can 
5700: be thought of as the sum of classical trajectories.  Since the classical 
5701: trajectories chaotically mix in phase space, it has been conjectured that
5702: the wave function would be a random Gaussian function of the eigenvalue
5703: $q$ with a spectrum given by the Wigner distribution function.  
5704: Wigner's function is an attempt to generalize 
5705: Boltzmann's formula for the 
5706: classical distribution of a statistical system 
5707: to a quantum system \cite{wigner}.  
5708: It provides a description
5709: of general attributes of wave functions \cite{gutz} and shows how the
5710: function tends to distribute itself over classical regions of phase space.
5711: Notice
5712: the similarity between this suggestion and the method-of-images as well
5713: as the numerical results of Ref \cite{{inoue},{cornsperg}}.
5714: 
5715: The nature of the discrete spectrum of energy levels can also be related
5716: to the classical chaos.
5717: As already argued in the numerical work on the eigenmodes of a compact space,
5718: the number of eigenstates can be related to the volume of the space through
5719: Weyl's asymptotic formula.  Additional features, such as the spacing between
5720: energy levels, can be directly associated with the underlying chaos.
5721: 
5722: 
5723: Despite this conjecture of randomness, 
5724: distinct remnants of the underlying chaotic
5725: dynamics have been found in the form of scars \cite{heller}.  Scars are 
5726: regions of enhanced intensity in high $k$ states along periodic orbits.
5727: It is hard to imagine the very high $k$ modes having cosmological significance
5728: since it is the low $k$ modes which probe the largest cosmological distances.
5729: Scaring is expected to be less prominent in low $k$ modes since the width
5730: of the enhancement will be correspondingly more diffuse.  
5731: Although diffuse, scars on large scales could provide the small
5732: catalyst needed to order structure on large-scale.
5733: Scarring on a $2D$ double doughnut was investigated in 
5734: references \cite{aurichsteiner}.  The underlying fractal
5735: structure of periodic orbits on the double
5736: doughnut was studied in detail in Ref.\ \cite{jdbi}.  There it was suggested
5737: that the filamentary structure we observe in the distribution of galaxies
5738: may be a consequence of this slight enhancement of the seeds of 
5739: structure formation.
5740: 
5741: For other important chaos articles see Refs.\ \cite{{ellistavakol},  
5742: {detchaos},{gott},{lockhart}}
5743: 
5744: 
5745: %\vspace{1in}
5746: 
5747: \newpage
5748: \section{Summary}
5749: 
5750: The creation of the universe is still not well understood.
5751: Clarity on the earliest moments will likely come only with a 
5752: fully functional quantum theory of gravity.  In the meantime,
5753: we know that space is curved and evolving and most also possess
5754: {\it some} topology. If the curvature of the universe falls within
5755: the observable horizon, then topology may also.
5756: The CMB provides the deepest probe of the universe on the largest-scales
5757: and we have reviewed the many ideas on how to extract the topology
5758: of space from maps of the
5759: microwave sky.  The methods fall into two primary categories:
5760: direct statistical methods and generic geometric methods.
5761: The salient features in CMB maps which reveal topology are
5762: (1) a discretization of the sizes of hot and cold spots,
5763: (2) a cutoff in the spectrum for wavelengths too big to fit within
5764: the finite space
5765: and (3) an anisotropic and inhomogeneous distribution of 
5766: correlations corresponding to repeated ghost images of the same
5767: spots.
5768: Data from the future satellite missions MAP and 
5769: {\it Planck Surveyor} are needed to determine if the CMB does in fact
5770: encode such features.  When the new high resolution maps are in hand,
5771: we will have to face the potentially prohibitive difficulty of foreground
5772: contaminations, the thickness of the surface last scattering, and fortuitous
5773: correlations.
5774: If we are lucky enough to surmount these observational trials, we 
5775: may be able to see the entire shape of space.
5776: 
5777: Still, as many fear, the topology scale may naturally be far beyond 
5778: the observable universe.  If this is the case, we can turn to  
5779: physics on the smallest scales to learn something about what we will never
5780: see on the largest scales.  If an ultimate theory of gravity beyond
5781: Einstein's is able to predict the geometry and topology of small
5782: extra dimensions, there is every reason to hope we will learn, if 
5783: only indirectly, the geometry and topology of the large dimensions.
5784: 
5785: 
5786: 
5787: 
5788: \section*{Acknowledgements}
5789: I appreciate the generous contributions of suggestions, ideas, 
5790: and figures from the topology/cosmology
5791: community.
5792: I am grateful to R. Aurich, J.D. Barrow,
5793: J.R. Bond, N.J. Cornish, H. Fagundes, K.T. Inoue,
5794: J-P. Luminet, B. Roukema, D. Spergel and G. Starkman
5795: for useful discussions.
5796: I am especially grateful to Jeff Weeks for his mathematical insight
5797: and his direct contributions to this review.  I thank
5798: Theoretical Physics Group at 
5799: Imperial College for their hospitality.
5800: This work is supported by PPARC.
5801: 
5802: 
5803: \addcontentsline{toc}{section}{Acknowledgements}
5804: 
5805: 
5806: \newpage
5807: 
5808: \addcontentsline{toc}{section}{References}
5809: 
5810: \begin{thebibliography}{999}
5811: 
5812: \bibitem{lum} M. Lachieze-Rey and J. -P. Luminet, Phys. Rep.
5813: {\bf 254}, 136, (1995).
5814: 
5815: \bibitem{css_cqg} N.J.Cornish, D.N.Spergel and G.D.Starkman,
5816: Class. Quant. Grav. {\bf 15} 2657 (1998).
5817: 
5818: \bibitem{galreviews1} 
5819: J.-P. Uzan, R. Lehoucq, and J.-P. Luminet,
5820: Proceedings of the XIXth Texas meeting, Paris 14-18 December 1998,
5821: Eds. E. Aubourg, T. Montmerle, J. Paul and P. Peter, article-no: 04.25.
5822: 
5823: \bibitem{galreviews2} J.-P. Luminet and B.F. Roukema,
5824: Proceedings of Cosmology School held at Cargese, Corsica, August 1998
5825: astro-ph/9901364.
5826: 
5827: \bibitem{roukreview}
5828: B.F. Roukema, Bull.Astr.Soc.India, 28, 483 (astro-ph/0010185) (2000);
5829: B.F. Roukema,  Marcel Grossmann IX Conference, eds Ruffini  et al.  (2001)
5830: (astro-ph/0010189).
5831: 
5832: \bibitem{ghosts1} L. Z. Fang and H. Sato, Comm. Theor. Phys,
5833: (China) {\bf 2}, 1055, (1983); H. V. Fagundes, {\em Astrophys. J.}
5834: {\bf 291}, 450, (1985); {\it ibid.}, Astrophys. J. {\bf 338 },
5835: 618, (1989); H. V. Fagundes  and U. F. Wichoski,  Astrophys. J. Lett.
5836: {\bf 322}, L5, (1987).
5837: 
5838: \bibitem{ghosts2} H. V. Fagundes, Phys. Rev. Lett. {\bf 70}, 1579,
5839: (1993); R. Lehoucq, M. Lachiese-Rey and J. -P. Luminet,
5840: Astron. Astrophys. {\bf 313}, 339, (1996); B. F. Roukema and
5841: A. C. Edge, Mon. Not. Roy. Astron. Soc. {\bf 292}, 105, (1997);
5842: B. F. Roukema and V. Blanloeil, Class. Quantum Grav. {\bf 15}, 2645,
5843: (1998);  B.F. Roukema,
5844: MNRAS, 283, 1147 (1996);
5845: B.F. Roukema \& S. Bajtlik, MNRAS, 308, 309 (1999).
5846: 
5847: 
5848: 
5849: \bibitem{thursweeks} W.P.Thurston and J.R.Weeks, Sci. Am. July 94
5850: (1984).
5851: 
5852: \bibitem{glennpop} J.P.Luminet, G.D.Starkman and J.R. Weeks, Sci. Am., April
5853: (1999).
5854: 
5855: \bibitem{cqg}  The entire issue of
5856: Class. Quant. Grav. {\bf 15} 2589 (1998).
5857: 
5858: \bibitem{ctp98}   V. Blanl{\oe}il, B.F. Roukema, editors, 
5859:  Proceedings of the Cosmological Topology in Paris 1998 meeting,
5860:  astro-ph/0010170 (2000). 
5861: 
5862: 
5863: \bibitem{css1} N.J.Cornish, D.N.Spergel and G.Starkman, gr-qc/9602039.
5864: 
5865: \bibitem{css2} N.J.Cornish, D.N.Spergel and G.Starkman,
5866: Phys. Rev. D {\bf 57} (1998) 5982.
5867: 
5868: \bibitem{conf}   J. Levin,
5869: E. Scannapieco and J. Silk, Class. Quant. Grav. {\bf 15}, 2689,
5870: (1998).
5871: 
5872: \bibitem{pat}  J. Levin, E. Scannapieco, G. Gasperis, J. Silk and
5873: J. D. Barrow,
5874: {\em Phys. Rev. } { D} {\bf 58} (1998) article 123006.
5875: 
5876: \bibitem{imogen}  I. Heard and J. Levin,
5877: Proceedings for ``Cosmological Topology'' in Paris 
5878: (CTP98), astro-ph/9907166.
5879: 
5880: \bibitem{bpsI} J. R. Bond, D. Pogosyan and T. Souradeep,
5881: Phys. Rev. D. {\bf 62} (2000) article 043005.
5882: 
5883: \bibitem{bpsII} J. R. Bond, D. Pogosyan and T. Souradeep,
5884: Phys. Rev. D. {\bf 62} (2000) article 042006.
5885: 
5886: \bibitem{wolf} J. A. Wolf, {\it Space of Constant Curvature (5th
5887: ed.)}, (Publish or Perish, Inc., 1994).
5888: 
5889: \bibitem{ellis} G.F.Ellis, Q.J.R.Astron. Soc. {\bf 16} 245 (1975);
5890: G. F. R. Ellis, Gen. Rel. Grav. {\bf 2} 7(1971).
5891: 
5892: \bibitem{thurclass} 
5893: W.P. Thurston, Bull. Am. Math. Soc. {\bf 6} (1982) 357.
5894: 
5895: \bibitem{Thurston}
5896: W.P. Thurston, ``Three-dimensional geometry and topology''
5897: (Ed: Silvio Levy, Princeton University Press, Princeton, N.J. 1997).
5898: 
5899: \bibitem{Bianchi} L. Bianchi, Mem. Soc. It. Della. Sc. (Dei. XL)
5900: 11, 267 (1897).
5901: 
5902: \bibitem{snappea} J. R. Weeks, {\it SnapPea: A computer program for
5903: creating and studying hyperbolic 3-manifolds}, Univ. of Minnesota
5904: Geometry Center (freely available at http://www.northnet.org/weeks).
5905: 
5906: \bibitem{mp} G.D. Mostow, Ann. Math.,  Studies {\bf 78}
5907: Princeton University Press, Princeton, 1973);
5908: G. Prasad, Invent. Math. {\bf 21} 255 (1973).
5909: 
5910: \bibitem{weekspace} J.R. Weeks, PhD thesis, Princeton University (1985).
5911: 
5912: \bibitem{bv}  N.L. Balazs and A. Voros, {\em Phys. Rep.} {\bf 143}
5913: (1986)
5914: 109.
5915: 
5916: \bibitem{jwp} The mathematical section was contributed largely by 
5917: Jeff Weeks who unfortunately was unable to coauthor this paper;
5918: J. Weeks, private communication.
5919: 
5920: \bibitem{Francis-Weeks} 
5921: G. Francis and J. Weeks, American Mathematical Monthly  106 (1999) 393-399.
5922: 
5923: \bibitem{Shape} 
5924: J. Weeks, The Shape of Space, Marcel Dekker Inc., 1985.
5925: 
5926: \bibitem{Tbull} W.P. Thurston, Bull. Am. Math. Soc. {\bf 6}
5927: 357 (1982).
5928: 
5929: \bibitem{pscott} 
5930: P. Scott, London Math. Soc. {\bf 15} 401 (1983).
5931: 
5932: \bibitem{Friedmann1924}
5933: A. Friedmann, Zeitschrift f\"ur Physik {\bf 21} 326 (1924).
5934: 
5935: \bibitem{Löbell1929} 
5936:  F. L\"obell, Ber. d. S\"achs. Akad. d. Wiss. {\bf 83} 167 (1931).
5937: 
5938: 
5939: \bibitem{SW1932?}
5940: C. Weber and H. Seifert,  
5941: Math. Zeitschrift {\bf 37} 237 (1933).
5942: 
5943: 
5944: \bibitem{Kojima??}.
5945: S. Kojima, Isometry Transformations of Hyperbolic 3-Manifolds,
5946: Topology Appl., {\bf 29} 297 (1988).
5947: 
5948: \bibitem{toco}  Miller et al. (TOCO experiment), ApJ 524, L1 (1999).
5949: 
5950: \bibitem{exper}  P. de Bernardis, et. al, astro-ph/0011469;
5951: J.R. Bond, et. al., astro-ph/0011378; A.H. Jaffe, et.al. astro-ph/0007333;
5952: S. Hanany, et. al. astro-ph/0005123.
5953: 
5954: \bibitem{vilenkin}
5955: A. Vilenkin, Phys. Rev. D {\bf 27}{2848}(1983).
5956: % The Birth of Inflationary Universes.
5957: 
5958: \bibitem{linde} 
5959: A.D. Linde, Mod. Phys. Lett.{\bf A1}{81} (1986);
5960: A.D. Linde, Physics Letters {\bf 175B}{395}(1986).
5961: 
5962: \bibitem{guth0} A.H. Guth, Phys. Rev. D. {\bf 23} 347 (1981).
5963: 
5964: \bibitem{guth1} A.H. Guth, Phys. Rep. 333 (2000) 555.
5965: 
5966: \bibitem{topstrings} C. Contaldi, astro-ph/0005115.
5967: 
5968: \bibitem{sokolov} I. Y. Sokolov, JETP Lett. {\bf 57}, 617, (1993).
5969: 
5970: \bibitem{fagundes1} D. Muller, H.V. Fagundes, and R. Opher 
5971: {\it in press} Phys. Rev. D
5972: gr-qc/0103014.
5973: 
5974: \bibitem{css3}   N.J.Cornish, D.N.Spergel and G.Starkman,
5975: Phys. Rev. Lett. {\bf 77} (1996) 215.
5976: 
5977: \bibitem{sst} G.D. Starkman, D. Stojkovic, and M. Trodden,
5978: hep-th/0106143 (2001).
5979: 
5980: \bibitem{bark} J.D. Barrow and H. Kodama, 
5981: Class. Quant. Grav. {\bf 18} 1753 (2001); J.D. Barrow and H. Kodama,
5982: gr-qc/0105049 (2001).
5983: 
5984: \bibitem{mkb} V.F. Mukhanov, H.A. Feldman, R.H. Brandenberger, Phys. Re;. {\bf 215}, 203 (1992).
5985: 
5986: \bibitem{hu} W. Hu, in ``The Universe at High-z, Large Scale Structure and 
5987: the Cosmic Microwave Background'', eds. E. Martinez-Gonzalez and J.L. Sanz
5988: (Springer Verlag) 1995.
5989: 
5990: \bibitem{ll} A.R. Liddle and D.H. Lyth, Phys. Rep. {\bf 231}, 1 (1993).
5991: 
5992: \bibitem{whu} M. White and W. Hu, Astronomy and Astrophysics .
5993: 
5994: 
5995: \bibitem{waldbk} R. Wald, {\it General Relativity}.
5996: 
5997: \bibitem{mtw} Misner, Thorne, and Wheeler, {\it Gravitation}.
5998: 
5999: \bibitem{swein} S. Weinberg, {\it Gravitation and Cosmology}.
6000: 
6001: \bibitem{kolbturner} E.W. Kolb and M.S. Turner,
6002: {\it The Early Universe} (Addison-Wesley, Reading, 1994).
6003: 
6004: \bibitem{peebk} P.J.E. Peebles, {\it The Large-Scale Structure of the 
6005: Universe} (Princeton University Press, 1980).
6006: %Princeton, New Jersey) 
6007: 
6008: \bibitem{sw} R.K. Sachs and A.M. Wolfe, Apj {\bf 147}, 73 (1967).
6009: 
6010: \bibitem{lb_fractals} J.Levin and J.D.Barrow, Class. Quantum Grav.
6011: {\bf 17} L61
6012: (2000).
6013: 
6014: \bibitem{newsphere}  E. Gausmann, R. Lehoucq, J.-P. Luminet, J.-P. Uzan,
6015: and J. Weeks, gr-qc/0106033.
6016: 
6017: \bibitem{open} A. Dekel, D. Burstein and S. D. M. White, 1996, in
6018: {\it Critical Dialogues in Cosmology}, ed. N.Turok, World Scientific;
6019: N. Bahcall
6020: and X. Fan, {\it preprint} (astro-ph/9804082).
6021: 
6022: \bibitem{inoueprog} K.T.Inoue, 
6023: Progress of Theoretical Physics, {\bf 106} 39 (2001). 
6024: 
6025: \bibitem{ktinoue} K.T.Inoue, Phys. Rev. {\bf D} 103001, 1-15 (2000).
6026: 
6027: \bibitem{saul} S. Perlmutter, et. al., Ap. J. {\bf 517} 565 (1999).
6028: 
6029: \bibitem{balbi} A. Balbi, et. al., astro-ph/0005124.
6030: 
6031: \bibitem{addr} B.F. Roukema, G.A. Mamon, S. Bajtlik, submitted Astron.\&
6032: Astroph., astro-ph/0106135 (2001).
6033: 
6034: \bibitem{bennet} C.L. Bennett, et al. 
6035: ApJ {\bf 391} (1992) 466.
6036: 
6037: \bibitem{smoot}  G.F. Smoot et al., Ap. J. {\bf 360} 685 (1990);
6038: G.F. Smoot et al., Ap. J. {\bf 396} L1 (1992).
6039: 
6040: \bibitem{kris} K. Gorski, 
6041: Proc. Moriond XVI, 
6042: ed. F.R. Bouchet et. al. (Gif-Sur-Yvette:  Editions Fronti\`ers) (1997).
6043: 
6044: \bibitem{teg} M. Tegmark {\em Phys. Rev.}
6045: {\bf D 55} (1997) 5895.
6046: 
6047: \bibitem{jaffe} J.R. Bond, A. Jaffe and L. Knox,
6048: preprint CfPA-97-TH-11; {\it ibid.} in preparation;
6049: J.R. Bond and A. Jaffe,
6050: Proc. Moriond XVI, 
6051: ed. F.R. Bouchet et. al. (Gif-Sur-Yvette:  Editions Fronti\`ers) (1997).
6052: 
6053: \bibitem{bondef}  J.R. Bond and G.P. Efstathiou, MNRAS {\bf 226} 655 (1987).
6054: 
6055: \bibitem{lss}J. Levin, E. Scannapieco, and J. Silk, Phys. Rev.
6056: D {\bf 58} (1998) article 103516.
6057: 
6058: \bibitem{slsi}  E. Scannapieco, J. Levin, and J. Silk, MNRAS
6059:  303 (1999) 797.
6060: 
6061: \bibitem{sss}  D. Stevens, D. Scott and J. Silk, {\em Phys. Rev. Lett.} {\bf 
6062: 71} (1993) 20.
6063: 
6064: 
6065: \bibitem{deO1}  A. de Oliveira-Costa and G.F. Smoot, Ap. J. {\bf 448}
6066: 447 (1995).
6067: 
6068: \bibitem{cheeger} J. Cheeger, in {\it Problems in Analysis, (A
6069: Symposium in honor of S. Bochner)}, (Princeton University Press,
6070: 1970).
6071: 
6072: \bibitem{buser} P. Buser, Proc. Symp. Pure Math. {\bf 36}, 29, (1980).
6073: 
6074: \bibitem{workshop} Entire Volume of
6075: Class. Quant. Grav. {\bf 15} (1998).
6076: 
6077: 
6078: \bibitem{zel73} Ya B. Zel'dovich, Comm. astrophys. Space Sci.
6079: {\bf 5} 169 (1973).
6080: 
6081: \bibitem{fanghoujun} L.Z. Fang and M. Houjun, Mod. Phys. Lett. A
6082: {\bf 2} 229 (1987).
6083: 
6084: \bibitem{deO2}  A. de Oliveira Costa, G. F. Smoot and A. A. Starobinsky,
6085: Astrophys. J. {\bf 468}, 457 (1996).
6086: 
6087: \bibitem{star} A. A. Starobinsky, JETP Lett. {\bf 57}, 622 (1993).
6088: 
6089: \bibitem{fang} L.Z. Fang, Mod. Phys. Lett. A {\bf 8} 2615 (1993).
6090: 
6091: \bibitem{boudcirc} B.F. Roukema, MNRAS, 312, 712 (2000);
6092: B.F. Roukema, Class. Quant. Grav {\bf 17} 3951 (2000).
6093: 
6094: \bibitem{et} G.F.R. Ellis and R. Tavakol, in {\it Deterministic
6095: Chaos in General Relativity}, Eds. D. Hobbill, A. Burd, and A. Coley,
6096: (Plenum Press, New York, 1994).
6097: 
6098: \bibitem{sinai} Y.G. Sinai, Sov. Math. Dok., {\bf 1} 335 (1960).
6099: 
6100: \bibitem{aurichsteiner}  
6101: R. Aurich and F. Steiner, Physica D {\bf 39}, 169,
6102: (1989); R.Aurich and F.Steiner, Physica D {\bf 64} 185 (1993).
6103: 
6104: \bibitem{inouetrace}
6105: K.T.Inoue,
6106: Classical and Quantum Gravity, {\bf 18} 629 (2001),
6107: math-ph/0011012.
6108: 
6109: \bibitem{gutz}  M.C. Gutzwiller, {\it Chaos in Classical and Quantum
6110: Mechanics} (Springer-Verlag New York Inc, New York, 1990);
6111: M.C. Gutzwiller, {\it J. Math. Phys.}, {\bf 12}
6112: (1971) 343.
6113: 
6114: \bibitem{ott} E. Ott, {\em Chaos in dynamical systems}, (CUP, 
6115: Cambridge, 1993).
6116: 
6117: \bibitem{lyth}  D.H.Lyth and A. Woszczyna, Phys. Rev. D {\bf 52} 3338 (1995).
6118: 
6119: \bibitem{krzy} K.M. Gorski, B. Ratra, N. Sugiyama, and A.J. Banday, Ap. J.
6120: {\bf 444} L65 (1995).
6121: 
6122: 
6123: \bibitem{spergelcqg}  
6124: D. N. Spergel, Class. Quant. Grav. {\bf 15}, 2589, (1998).
6125: 
6126: 
6127: \bibitem{lobell} F. L\"obell, Ber, S\"achs. Akad. Wiss. Leipzig {\bf 83} (1931).
6128: 
6129: \bibitem{seifertweber}  H.Seifert and C.Weber, Math. Z. {\bf 37} 237 (1933).
6130: 
6131: 
6132: \bibitem{sos}  D. D. Sokolov and A. A. Starobinskii, {\em Sov. Astron. }{\bf 
6133: 19} (1976) 629.
6134: 
6135: \bibitem{lbbs}  J. Levin, J.D. Barrow, E.F. Bunn and J. Silk,
6136: {\em Phys. Rev. Lett.} {\bf 79} (1997) 974 .
6137: 
6138: 
6139: %\bibitem{thick} 
6140: 
6141: \bibitem{olsonstark} D. Olson and G.D. Starkman,
6142: Class. Quant. Grav. {\bf 17} 3093 (2001).
6143: 
6144: \bibitem{inoue} K.T.Inoue, Class. Quantum Grav. {\bf 16} 3071 (1999).
6145: 
6146: %\bibitem{inoue2} K.T.Inoue, astro-ph/9903446.
6147: 
6148: \bibitem{inouetomita} K.T.Inoue, K. Tomita, and N.Sugiyama,
6149: MNRAS {\bf 314} L21 (2000).
6150: 
6151: \bibitem{inouethesis} K.T.Inoue, Ph.D. Thesis, astro-ph 0103158.
6152: 
6153: \bibitem{cornsperg} N.J.Cornish and D.N.Spergel, math.DG/9906017;
6154: N.J.Cornish and D.N.Spergel, Phys. Rev. D. {\bf 62}
6155: (2000) article 087304.
6156: 
6157: \bibitem{aurich}
6158: R. Aurich, Ap. J. {\bf 524} 497 (1999).
6159: 
6160: \bibitem{aurichmarklof} R. Aurich and J. Marklof, Physica {\bf D 92},
6161: 101 (1996).
6162: 
6163: \bibitem{berry}  M.V. Berry, {\it J. Phys. A} {\bf 12} (1977) 2083;
6164: M.V. Berry in {\it Chaotic Behavior of Deterministic
6165: Systems}, eds. G. Iooss, R. Hellman, and R. Stora, Les Houches
6166: Proc. 36 (North-Holland, NY, 1981).
6167: 
6168: \bibitem{saskatoon} C.B.Netterfield, M.J.Devlin, N.Jarosik, L.A.Page
6169: and E.J.Wollack, Astrophys. J. {\bf 474}, 47 (1997).
6170: 
6171: \bibitem{qmap} A. de Oliveira-Costa, M.J.Devlin, T.Herbit, A.D.Miller,
6172: C.B.Netterfield, L.A.Page, and M.Tegmark, Astrophys. J. Lett.
6173: {\bf 509}, L77, (1998).
6174: 
6175: \bibitem{as} R. Aurich and F. Steiner, MNRAS {\bf 323} 1016 (2001).
6176: 
6177: \bibitem{pen} U.-L. Pen, Astrophys. J. {\bf 498} 60 (1998).
6178: 
6179: \bibitem{bps_texas} J. R. Bond, D. Pogosyan and T. Souradeep, in: {\it
6180: Proceedings of the XVIIIth Texas Symposium on Relativistic
6181: Astrophysics}, ed.\ A. Olinto, J. Frieman, and D.~N. Schramm, (World
6182: Scientific, 1997).
6183: 
6184: \bibitem{bps_moriond} J. R. Bond, D. Pogosyan and T. Souradeep, in:
6185: {\it
6186: Proc. of XXXIIIrd Recontre de Moriond}, ``Fundamental Parameters in
6187: Cosmology'', Jan. 17-24, 1998, Les Arc, France.
6188: 
6189: \bibitem{bps_cwru} J. R. Bond, D. Pogosyan and T. Souradeep,
6190: Class. Quant. Grav. {\bf 15}, 2671, (1998).
6191: 
6192: 
6193: \bibitem{chavel} I. Chavel, {\it Eigenvalues in Riemannian geometry},
6194: (Academic Press, 1984).
6195: 
6196: 
6197: \bibitem{berard} P. H. Berard, {\it Spectral Geometry: Direct and
6198: Inverse Problems}, Lec. Notes in Mathematics, {\bf 1207},
6199: (Springer-Verlag, Berlin, 1980).
6200: 
6201: \bibitem{morehf} H.V. Fagundes, Ap. J. {\bf 470} 43 (1996).
6202: 
6203: \bibitem{morehf2} H.V. Fagundes, astro-ph/0007443.
6204: 
6205: \bibitem{cas}  L. Cay\'on and G. Smoot, Ap. J. {\bf 452} 494 (1995).
6206: 
6207: \bibitem{hw}  C.D. Hodgson and J.R.Weeks, Experim. Math. 3, 261 (1994).
6208: 
6209: \bibitem{cornweeks} N.J. Cornish and J. Weeks, Notices of the American
6210: Mathematical Society, astro-ph/9807311.
6211: 
6212: \bibitem{roukemaedge}
6213: B.F. Roukema and A.C. Edge, MNRAS {\bf 292}, 105 (1997).
6214: 
6215: \bibitem{roukemablanloeil}
6216: B.F. Roukema and V. Blan loeil, Class. Quant. Grav., {\bf 15}, 2645 (1998).
6217: 
6218: \bibitem{best} L.A.Best, Can. J. Math. {\bf 23} No. 3, 451 (1971).
6219: 
6220: \bibitem{jdbi} J.D. Barrow and J. Levin, Phys. Rev. A (2001).
6221: 
6222: \bibitem{jefftwin} J. Weeks, unpublished.
6223: 
6224: \bibitem{uzan}
6225: J.P. Uzan, J.P. Luminet, R. Lehoucq, and P. Peter, physics/0006039.
6226: 
6227: \bibitem{peters} P.C. Peters, Am. J. Phys. {\bf 51} (1983) 791;
6228: P.C. Peters, Am. J. Phys. {\bf 54} (1986) 334.
6229: 
6230: \bibitem{lh} J.R. Lucas and P.E. Hodgson, {\it Spacetime and Electromagnetism}
6231: (Oxford University Press: 1990) pp 76-83.
6232: 
6233: \bibitem{carlip}  S. Carlip, Class. Quant. Grav. {\bf 15}
6234: 2629 (1998) and references therein.
6235: 
6236: \bibitem{gibbons} G. W. Gibbons, Nucl. Phys. {bf B 472}, 683, (1996);
6237: G. W. Gibbons, Class. Quantum Grav. {\bf 15}, 2605, (1998).
6238: 
6239: \bibitem{hfq} H.V. Fagundes and S.S. e Costa, General Relativity and
6240: Gravitation {\bf 31} 863 (1999).
6241: 
6242: \bibitem{hfq2}  
6243: S.S. e Costa and H.V. Fagundes, General Relativity and
6244: Gravitation {\bf 33} in press (2001).
6245: 
6246: \bibitem{randallco}  L. Randall and R. Sundrum,
6247: Phys. Rev. Lett. {\bf 83} 3370 (1999).
6248: 
6249: \bibitem{kk}  Th. Kaluza, Situngsber. Preuss. Akad. Wiss. Phys. Math. Kl.
6250: 966 (1921); O. Klein, Z. Phys. {\bf 37} 895 (1929).
6251: 
6252: 
6253: \bibitem{glennco}  N. Kaloper, J. March-Russell, G.D. Starkman,
6254: and M. Trodden,
6255: Phys. Rev. Lett. {\bf 85} 928 (2001).
6256: 
6257: %\bibitem{calabi}
6258: 
6259: \bibitem{hormar} G. T. Horowitz and D. Marolf, J.High Energy
6260: Phys. {\bf 9807}, 014, (1998).
6261: 
6262: \bibitem{wigner} E.P. Wigner, Phys. Rev. {\bf 40} (1932) 749.
6263: 
6264: \bibitem{heller}  E.J. Heller, {\it Phys. Rev. Lett.} {\bf 53}
6265: (1984) 1515.
6266: 
6267: 
6268: \bibitem{lockhart} C.M. Lockhart, B. Misra, I. Prigogine,
6269: Phys. Rev. D. {\bf 25} 921 (1982).
6270: 
6271: \bibitem{gott} J.Richard Gott, III Mon. Not. R. astr. Soc. {\bf 192}
6272: 153 (1980).
6273: 
6274: \bibitem{detchaos}{\it Deterministic Chaos in General Relativity}
6275: eds. D. Hobill et al., Plenum Press, Ne w York, 1994.
6276: 
6277: \bibitem{ellistavakol} G.F.R. Ellis and R. Tavakol, Class. Quantum Grav.  {\bf 11}
6278: 675 (1994).
6279: 
6280: \bibitem{ellisschreiber}  G.F.R. Ellis and G. Schreiber, Phys. Lett. A {\bf
6281: 115} 97 (1986).
6282: 
6283: 
6284: 
6285: \end{thebibliography}
6286: 
6287: \newpage
6288: \appendix
6289: \section{Appendix:  Representations of Hyperbolic space}
6290: \label{AppendixA}
6291: 
6292: Because of the recent emphasis on hyperbolic models, we include a
6293: section here on other useful coordinate models for manipulating
6294: $\ah $.
6295: We have already seen two different coordinate systems
6296: for the metric eqns. (\ref{stanmet}) and (\ref{cosmet}).
6297: Another important coordinate system comes from embedding
6298: $\ah $ in the $(3+1)$-Minkowski space as discusssed in \S \ref{ovgeom}.
6299: The Minkowski metric is
6300: 	\be
6301: 	d\sigma^2=-dx_0^2+dx_1^2+dx_2^2+dx_3^2
6302: 	\ee
6303: The $3D$-hypersurface is constrained to have pseudoradius
6304: 	\be
6305: 	-x_0^2+x_1^2+x_2^2+x_3^2=1
6306: 	\ee
6307: and the curvature radius has been taken to $1$.  In Minkowski coordinates
6308: the generators $\gamma \in \Gamma$ take the convenient
6309: form of $O(4)$, orthogonal
6310: $4\times 4$ matrices.  
6311: These coordinates can be related to those of eqn.\ (\ref{cosmet})
6312: with the transformation
6313: 	\ba
6314: 	x^\mu &=& \pmatrix{x^0\cr x^1\cr x^2\cr x^3}
6315: 	=\pmatrix{\cosh \chi \cr
6316: 	\sinh \chi \sin\theta \cos \phi \cr
6317: 	\sinh \chi \sin\theta \sin \phi \cr
6318: 	 \sinh\chi \cos \theta }.
6319: 	\ea
6320: 
6321: There are several other useful representations of the hyperbolic plane
6322: including the Poincar\'e 
6323: ball with
6324: 	\be
6325: 	\vec x=\tanh(r/2)\hat n
6326: 	\ee
6327: with $\hat n$ the usual unit vector in spherical coordinates.
6328: The Poincar\'e ball model maps $\ah$ to the 
6329: open ball $\{\vx  \in \eu | \vec x\cdot \vec x < 1\}$
6330: with 	
6331: 	\be
6332: 	d\sigma^2={4d\vx\cdot d\vx\over \left(1-{\vx\cdot \vx}\right )}.
6333: 	\ee
6334: It is useful to know the geodesic distance is
6335: 	\be
6336: 	d(\vx,\vxp)={\rm arccosh}\left [1+{2\left |\vx-\vxp\right |
6337: 	\over (1-|\vx|^2)(1-|\vxp|^2)}\right ].
6338: 	\ee
6339: All geodesics intersect the boundary orthogonally and are therefore semicircles
6340: or straight lines which are the diameters of the ball.
6341: 
6342: There is also the upper-half space representation
6343:  $\{\vx  \in \eu | x_3>0\}$,
6344: 	\ba
6345: 	d\sigma^2 &=& {\left (dx^2+dy^2 dz^2\right )\over z^2}\nonumber \\
6346: 	&\equiv &	d\rho^2+e^{-2\rho}\left (dx^2+dy^2\right )
6347: 	\label{upperhalf}
6348: 	\ea
6349: with $e^\rho=z$.  This coordinate system is particularly useful 
6350: for cusped manifolds which are discussed in \S \ref{obscusps}.
6351: 
6352: Finally there is the three-dimensional Klein model with coordinates
6353: 	\be
6354: 	\vec x=\tanh\chi\hat n
6355: 	\ee
6356: which is used often in constructing the Dirichlet domain in three-dimensions
6357: and mapping the periodic geodesics.  Geodesics are mapped into 
6358: straight lines in this model.
6359: 
6360: 
6361: \end{document}
6362: 
6363: 
6364: 
6365: \bibitem{meyerhoff} 
6366: 
6367: 
6368: 
6369: