1: \documentclass[twocolumn,amsmath,amssymb,aps,secnumarabic,nofootinbib,%

2:     superscriptaddress,floatfix]{revtex4}

3: \usepackage{amsthm,times,mathptmx,graphicx,pstricks,pst-node}

4: \bibliographystyle{hamsplain}

5: 

6: \newtheorem{theorem}{Theorem}

7: \newtheorem{lemma}[theorem]{Lemma}

8: \newtheorem{corollary}[theorem]{Corollary}

9: \newtheorem{proposition}[theorem]{Proposition}

10: \newtheorem{question}[theorem]{Question}

11: \theoremstyle{remark}

12: \newtheorem*{remark}{Remark}

13: \newcommand{\Hom}{\operatorname{Hom}}

14: \newcommand{\conv}{\operatorname{conv}}

15: \newcommand{\Z}{\mathbb{Z}}

16: \newcommand{\R}{\mathbb{R}}

17: \newcommand{\F}{\mathbb{F}}

18: \newcommand{\HH}{\mathbb{H}}

19: \renewcommand{\hat}{\widehat}

20: \renewcommand{\tilde}{\widetilde}

21: \renewcommand{\d}{\partial}

22: \newcommand{\st}{\mathrm{st}}

23: \newcommand{\hg}{{\hat{g}}}

24: \newcommand{\hS}{{\hat{S}}}

25: \newcommand{\pol}{{}^\triangle}

26: \newcommand{\ppol}{{}^\diamondsuit}

27: \newcommand{\maxfatness}{5.048}

28: \newcommand{\eqdef}{\stackrel{\mathrm{def}}{=}}

29: 

30: \newcommand{\eg}{\emph{e.g.}}

31: \newcommand{\Eg}{\emph{E.g.}}

32: \newcommand{\ie}{\emph{i.e.}}

33: \newcommand{\cf}{\emph{cf.}}

34: \newcommand{\eatline}{\vspace{-\baselineskip}}

35: \newcommand{\blankline}{\vspace{\baselineskip}}

36: 

37: \renewcommand{\figurename}{Figure}

38: \newenvironment{fullfigure}[2]

39:     {\begin{figure}[ht]\begin{center}\def\ffa{#1}\def\ffb{#2}}

40:     {\vspace{\baselineskip}\caption{\ffb.}\label{\ffa}\end{center}\end{figure}}

41: \renewcommand{\tablename}{Table}

42: \renewcommand{\thetable}{\arabic{table}}

43: 

44: \newcommand{\fig}[1]{Figure~\ref{#1}}

45: \newcommand{\thm}[1]{Theorem~\ref{#1}}

46: \renewcommand{\sec}[1]{Section~\ref{#1}}

47: \newcommand{\lem}[1]{Lemma~\ref{#1}}

48: \newcommand{\prop}[1]{Proposition~\ref{#1}}

49: \newcommand{\eq}[2]{\begin{equation}\label{#1}#2\end{equation}}

50: 

51: \psset{linewidth=.5pt,dash=2.5pt 2.5pt,unit=.25in,arrowsize=4pt 3}

52: \SpecialCoor

53: 

54: \newgray{gray90}{.90}

55: \newcommand{\ar}{\pspicture(0,0)(0,0)

56:     \psline[arrows=->](4.3pt,0)(4.4pt,0)\endpspicture}

57: \newcommand{\sgpie}{

58:     \pscustom[linestyle=none,fillstyle=solid,fillcolor=gray90]{

59:     \psarcn(4.899;30){1.464}{255}{165} \psarcn(4.899;60){1.464}{285}{195}

60:     \psarcn(4.899;90){1.464}{315}{225} \psarcn(4.899;120){1.464}{345}{255}

61:     \psarcn(4.899;150){1.464}{15}{285} \psarcn(4.899;210){1.464}{75}{345}

62:     \psarcn(4.899;240){1.464}{105}{15} \psarcn(4.899;270){1.464}{135}{45}

63:     \psarcn(4.899;300){1.464}{165}{75} \psarcn(4.899;330){1.464}{195}{105}}

64:     \psarcn(4.899;30){1.464}{255}{165} \psarcn(4.899;60){1.464}{285}{195}

65:     \psarcn(4.899;90){1.464}{315}{225} \psarcn(4.899;120){1.464}{345}{255}

66:     \psarcn(4.899;150){1.464}{15}{285} \psarcn(4.899;210){1.464}{75}{345}

67:     \psarcn(4.899;240){1.464}{105}{15} \psarcn(4.899;270){1.464}{135}{45}

68:     \psarcn(4.899;300){1.464}{165}{75} \psarcn(4.899;330){1.464}{195}{105}

69:     \qdisk(4;165){.1} \qdisk(4;195){.1} \qdisk(4;225){.1} \qdisk(4;255){.1}

70:     \qdisk(4;285){.1} \qdisk(4;315){.1} \qdisk(4;345){.1} \qdisk(4;15){.1}

71:     \qdisk(4;45){.1} \qdisk(4;75){.1} \qdisk(4;105){.1} \qdisk(4;135){.1}

72:     \rput{120}(3.435;30){\ar} \rput{150}(3.435;60){\ar}

73:     \rput{180}(3.435;90){\ar} \rput{210}(3.435;120){\ar}

74:     \rput{240}(3.435;150){\ar} \rput{300}(3.435;210){\ar}

75:     \rput{330}(3.435;240){\ar} \rput{0}(3.435;270){\ar}

76:     \rput{30}(3.435;300){\ar} \rput{60}(3.435;330){\ar}

77:     \uput[180](4;180){\large $\vdots$}

78:     \uput[0](4;0){\large $\vdots$}}

79: 

80: \begin{document}

81: \title{Fat 4-polytopes and fatter 3-spheres}

82: 

83: \author{David Eppstein}

84: \email[Email: ]{eppstein@ics.uci.edu}

85: \thanks{Supported in part by NSF grant CCR \#9912338}

86: \affiliation{Department of Information and Computer Science,

87:     University of California, Irvine, CA 92697}

88: \author{Greg Kuperberg}

89: \email[Email: ]{greg@math.ucdavis.edu}

90: \thanks{Supported by NSF grant DMS \#0072342}

91: \affiliation{Department of Mathematics,

92:     University of California, Davis, CA 95616}

93: \author{G\"unter M. Ziegler}

94: \email[Email: ]{ziegler@math.tu-berlin.de}

95: \thanks{Supported by a DFG Leibniz grant}

96: \affiliation{Institut f\"ur Mathematik, MA 6-2,

97:     Technische Universit\"at Berlin, D-10623 Berlin, Germany}

98: 

99: \begin{abstract}

100: We introduce the \emph{fatness} parameter of a $4$-dimensional polytope $P$,

101: defined as $\phi(P)=(f_1+f_2)/(f_0+f_3)$.  It arises in an important open

102: problem in $4$-dimensional combinatorial geometry: Is the fatness of convex

103: $4$-polytopes bounded?

104: 

105: We describe and analyze a hyperbolic geometry construction that produces

106: $4$-polytopes with fatness $\phi(P)>\maxfatness$, as well as the first infinite

107: family of $2$-simple, $2$-simplicial $4$-polytopes. Moreover, using a

108: construction via finite covering spaces of surfaces, we show that fatness is

109: not bounded for the more general class of strongly regular CW decompositions of

110: the $3$-sphere.

111: \end{abstract}

112: 

113: \maketitle

114: 

115: \section{Introduction}

116: \label{s:intro}

117: 

118: The characterization of the set ${\cal F}_3$ of $f$-vectors of convex 3-dimensional

119: polytopes (from 1906, due to Steinitz \cite{Steinitz:eulerschen}) is well-known

120: and explicit, with a simple proof: An integer vector $(f_0,f_1,f_2)$ is the

121: $f$-vector of a $3$-polytope if and only if it satisfies

122: \begin{itemize}\setlength{\itemsep}{0pt}

123: \item $f_1=f_0+f_2-2$ (the Euler equation),

124: \item $f_2\le 2f_0-4$ (with equality for simplicial polytopes), and

125: \item $f_0\le 2f_2-4$ (with equality for simple polytopes).

126: \end{itemize}

127: (Recall that by the definition of the $f$-vector, $f_k$ is the number

128: of $k$-faces of the polytope.)

129: This simple result is interesting for several reasons:

130: \begin{itemize}\setlength{\itemsep}{0pt}

131: \item The set of $f$-vectors is the set of \emph{all} the integer points in a

132: closed $2$-dimensional \emph{polyhedral cone} (whose apex is the $f$-vector

133: $f(\Delta_3)=(4,6,4)$ of a $3$-dimensional simplex). In particular, it is

134: \emph{convex} in the sense that ${\cal F}_3=\conv({\cal F}_3)\cap \Z^3$.

135: \item The same characterization holds for convex $3$-polytopes (geometric

136: objects), more generally for strongly regular CW $2$-spheres (topological

137: objects), and yet more generally for Eulerian lattices of length $4$

138: (combinatorial objects \cite{Stanley:eulerian}).

139: \end{itemize}

140: 

141: In contrast to this explicit and complete description of ${\cal F}_3$, our

142: knowledge of the set ${\cal F}_4$ of $f$-vectors of (convex) $4$-polytopes (see

143: Bayer \cite{Bayer:extended} and H\"oppner and Ziegler \cite{HZ:census}) is very

144: incomplete. We know that the set ${\cal F}_4$ of all $f$-vectors of

145: 4-dimensional polytopes has no similarly simple description. In particular, the

146: convex hull of ${\cal F}_4$ is not a cone, it is not a closed set, and not all

147: integer points in the convex hull are $f$-vectors. Also, the $3$-dimensional

148: cone with apex $f(\Delta_4)$ spanned by ${\cal F}_4$ is not closed, and its

149: closure may not be polyhedral.

150: 

151: Only the two extreme cases of simplicial and of simple $4$-polytopes (or

152: $3$-spheres) are well-understood.   Their $f$-vectors correspond to faces of

153: the convex hull of ${\cal F}_4$, defined by the valid inequalities $f_2\ge

154: 2f_3$ and $f_1\ge 2f_0$, and the $g$-Theorem, proved for $4$-polytopes by

155: Barnette \cite{Barnette:inequalities} and for $3$-spheres by Walkup

156: \cite{Walkup:lower}, provides \emph{complete} characterizations of their

157: $f$-vectors. (The $g$-Theorem for general simplicial polytopes was famously

158: conjectured by McMullen \cite{McMullen:simplicial} and proved by Billera and

159: Lee \cite{BL:sufficiency} and Stanley \cite{Stanley:simplicial}.  See

160: \cite[\S8.6]{Ziegler:gtm} for a review.)

161: 

162: But we have no similarly complete picture of other extremal types of

163: $4$-polytopes. In particular, we cannot currently answer the following key

164: question: Is there a constant $c$ such that all  $4$-dimensional convex

165: polytopes $P$ satisfy the inequality

166: \[

167: f_1(P)+f_2(P)\;\le\;c\,(f_0(P)+f_3(P))?

168: \]

169: To study this question, we introduce the \emph{fatness} parameter

170: \[

171: \phi(P)\;\eqdef\;\frac{f_1(P)+f_2(P)}{f_0(P)+f_3(P)}

172: \]

173: of a $4$-polytope $P$.  We would like to know whether fatness is bounded.

174: 

175: For example, the $4$-simplex has fatness $2$, while the $4$-cube and the

176: $4$-cross polytope have fatness $\frac{56}{24}=\frac73$. More generally, if $P$

177: is simple, then we can substitute the Dehn-Sommerville relations

178: \[

179: f_2(P)=f_1(P)+f_3(P)-f_0(P)\qquad f_1(P)=2f_0(P)

180: \]

181: into the formula for fatness, yielding

182: \[

183: \phi(P)=\frac{f_1(P)+f_2(P)}{f_0(P)+f_3(P)}

184:     = \frac{3f_0(P)+f_3(P)}{f_0(P)+f_3(P)}<3.

185: \]

186: 

187: Since every $4$-polytope and its dual have the same fatness, the same upper

188: bound holds for simplicial $4$-polytopes. On the other hand, the ``neighborly

189: cubical'' $4$-polytopes of Joswig and Ziegler \cite{JZ:cubical} have

190: $f$-vectors

191: \[

192: (4,2n,3n-6,n-2)\cdot 2^{n-2},

193: \]

194: and thus fatness

195: \[

196: \phi=\frac{5n-6}{n+2} \to 5.

197: \]

198: In particular, the construction of these polytopes disproved the conjectured

199: flag-vector inequalities of Bayer \cite[pp.~145, 149]{Bayer:extended} and

200: Billera and Ehrenborg \cite[p.~109]{HZ:census}.

201: 

202: The main results of this paper are two lower bounds on fatness:

203: 

204: \begin{theorem} There are convex $4$-polytopes $P$ with fatness

205: $\phi(P)>\maxfatness$. \label{th:fat} \end{theorem}

206: 

207: \begin{theorem} The fatness of cellulated $3$-spheres is not bounded.

208: A $3$-sphere $S$ with $N$ vertices may have fatness as high as

209: $\phi(S) = \Omega(N^{1/12})$.

210: \label{th:fatter} \end{theorem}

211: 

212: We will prove \thm{th:fat} in \sec{s:poly} and \thm{th:fatter} in

213: \sec{s:spheres}, and present a number of related results along the

214: way.

215: 

216: \section{Conventions}

217: \label{s:conven}

218: 

219: Let $X$ be a finite

220: CW complex. If $X$ is identified with a manifold $M$, it is also called a

221: \emph{cellulation} of $M$. The complex $X$ is \emph{regular} if all closed

222: cells are embedded \cite[\S38]{Munkres:algtop}.

223: If $X$ is regular, we define it to

224: be \emph{strongly regular} if in addition the intersection of any

225: two closed cells

226: is a cell.  For example, every simplicial complex is a strongly regular CW

227: complex.  The complex $X$ is \emph{perfect} if the boundary maps of its chain

228: complex vanish.  (A non-zero-dimensional perfect complex is never regular.)

229: 

230: The \emph{$f$-vector} of a cellulation $X$, denoted $f(X)=(f_0,f_1,\ldots)$,

231: counts the number of cells in each dimension:  $f_0(X)$ is the number of

232: vertices, $f_1(X)$ is the number of edges, etc.

233: 

234: If $X$ is $2$-dimensional, we define its fatness as

235: \[

236: \phi(X) \eqdef \frac{f_1(X)}{f_0(X) + f_2(X)}.

237: \]

238: If $X$ is $3$-dimensional, we define its fatness $X$ as

239: \[

240: \phi(X) \eqdef \frac{f_1(X) + f_2(X)}{f_0(X) + f_3(X)}.

241: \]

242: If $P$ is a convex $d$-polytope, then its $f$-vector $f(P)$ is defined to be

243: the $f$-vector of its boundary complex, which is a strongly regular

244: $(d-1)$-sphere. If $d=4$ we can thus consider $\phi(P)$, the fatness of $P$.

245: The faces of $P$ of dimension 0 and 1 are called \emph{vertices} and

246: \emph{edges} while the faces of dimension $d-1$ and $d-2$ are called

247: \emph{facets} and \emph{ridges}. We extend this terminology to general

248: cellulations of $(d-1)$-manifolds.

249: 

250: The \emph{flag vector} of a regular cellulation $X$, and likewise the flag

251: vector of a polytope $P$, counts the number of nested sequences of cells with

252: prespecified dimensions.  For example, $f_{013}(X)$ is the number triples

253: consisting of a $3$-cell of $X$, an edge of the $3$-cell, and a vertex of the

254: edge; if $P$ is a 4-cube, then

255: $$f_{013}(P) = 192.$$

256: 

257: A convex polytope $P$ is \emph{simplicial} if each facet of $P$ is a simplex.

258: It is \emph{simple} if its polar dual $P\pol$

259: is simplicial, or equivalently if the cone of each vertex matches

260: that of a simplex.

261: 

262: 

263: \section{4-Polytopes}

264: \label{s:poly}

265: 

266: In this section we construct families of $4$-polytopes with several

267: interesting properties:

268: 

269: \begin{itemize}\setlength{\itemsep}{0pt}

270: \item They are the first known infinite families of {\em $2$-simple,

271: $2$-simplicial $4$-polytopes}, that is, polytopes in which all $2$-faces and

272: all dual $2$-faces are triangles (all edges are ``co-simple''). \Eg, Bayer

273: \cite{Bayer:extended} says that it would be interesting to have an infinite

274: family.

275: 

276: As far as we know, there were only six such polytopes previously known: the

277: simplex, the hypersimplex (the set of points in $[0,1]^4$ with coordinate sum

278: between $1$ and $2$), the dual of the hypersimplex, the $24$-cell, and a gluing

279: of two hypersimplices (due to Braden \cite{Braden:glued}), and the dual of the

280: gluing. (There are claims in Gr\"unbaum

281: \cite[p.~82, resp. p.~170]{Grunbaum:convex} that results of Shephard, resp.\

282: Perles and Shephard, imply the existence of infinitely many $2$-simple

283: $2$-simplicial $4$-polytopes.  Both claims appear to be incorrect.)

284: 

285: \item They are the fattest known convex $4$-polytopes.

286: 

287: \item They yield finite packings of (not necessarily congruent) spheres in

288: $\R^3$ with slightly higher average kissing numbers than previously known

289: examples \cite{Kuperberg:kissing}.

290: \end{itemize}

291: 

292: Let $Q\subset\R^4$ be a $4$-polytope that contains the origin in its interior.

293: If an edge $e$ of $Q$ is tangent to the unit sphere $S^3\subset\R^4$ at a point

294: $t$, then the corresponding ridge ($2$-dimensional face) $F=e\ppol$ of the

295: polar dual $P=Q\pol$ is also tangent to $S^3$ at $t$.  (Recall that the

296: polar dual is defined as

297: $$Q\pol \eqdef \{p | p \cdot q \le 1 \forall q \in Q\}.)$$

298: Furthermore, the affine

299: hulls of $e$ and $F$ form orthogonal complements in the tangent space of $S^3$,

300: so the convex hull $\conv(e\cup F)$ is an orthogonal bipyramid tangent to $S^3$

301: (\cf, Schulte \cite[Thm.~1]{Schulte:steinitz}).

302: 

303: We will construct polytopes $E$ by what we call the \emph{$E$-construction}.

304: This means that they are convex hulls

305: $$E\eqdef\conv(Q\cup P),$$

306: where $Q$ is a simplicial $4$-dimensional polytope whose edges are tangent to

307: the unit $3$-sphere $S^3$, and $P$ is the polar dual of $Q$.  Thus $P$ is

308: simple and its ridges are tangent to $S^3$. (We call $Q$ \emph{edge-tangent}

309: and $P$ \emph{ridge-tangent}.)

310: 

311: \begin{proposition}

312: If $P$ is a simple, ridge-tangent $4$-polytope, then the $4$-polytope

313: $E=\conv(P\cup Q)$ produced by the $E$-construction is $2$-simple and

314: $2$-simplicial, with $f$-vector

315: \[

316:  f(E) = (f_2(P), 6f_0(P), 6f_0(P), f_2(P)),

317: \]

318: and fatness

319: \[

320:  \phi(E) = 6\,\frac{f_0(P)}{f_2(P)}.

321: \]

322: \label{p:e-construction} \end{proposition}

323: 

324: \begin{proof}

325: Another way to view the $E$-construction is that $E$ is produced from $P$ by

326: adding the vertices of $Q$ sequentially. At each step, we cap a facet of $P$

327: with a pyramid whose apex is a vertex of $Q$. Thus the new facets consist of

328: pyramids over the ridges of $P$, where two pyramids with the same base

329: (appearing in different steps) lie in the same hyperplane (tangent to $S^3$),

330: and together form a bipyramid.  The facets of the final polytope $E$ are

331: orthogonal bipyramids over the ridges of $P$ and are tangent to $S^3$.  Since

332: the 2-faces of $E$ are pyramids over the edges of $P$, $E$ is $2$-simplicial.

333: 

334: The polytope $E$ is $2$-simple if and only if each edge is \emph{co-simple},

335: \ie, contained in exactly three facets of $E$.  The iterative construction of

336: $E$ shows that it has two types of edges: (i) edges of $P$, which are co-simple

337: in $E$ if and only if they are co-simple in $P$, and (ii) edges formed by

338: adding pyramids, which are co-simple if and only if the facets of $P$ are

339: simple.  Since $P$ is simple, its facets are simple and its edges are

340: co-simple, so $E$ is then $2$-simple.

341: 

342: This combinatorial description of $E$ yields an expression for the $f$-vector

343: of $E$ in terms of the flag vector

344: \cite{BJ:dehn,Bayer:extended,HZ:census} of $P$.  Since the facets

345: of $E$ are bipyramids over the ridges of $P$, the following identities hold:

346: \begin{align*}

347: f_3(E) &= f_2(P) & f_2(E) &= f_{13}(P) \\

348: f_1(E) &= f_1(P)+f_{03}(P) & f_0(E) &= f_0(P)+f_3(P).

349: \end{align*}

350: Since $P$ is simple,

351: \[

352: f_{03}(P)=4f_0(P) \qquad f_{13}(P)=3f_1(P).

353: \]

354: These identities together with Euler's equation and $f_1(P)=2f_0(P)$ imply the

355: proposition.

356: \end{proof}

357: 

358: The $f$-vector of $P$ also satisfies

359: \[

360: f_0(P)-f_1(P)+f_2(P)-f_3(P)=0 \qquad f_1(P)=2f_0(P),

361: \]

362: in the second case because $P$ is simple, so the fatness of $E$ can also be

363: written

364: \[

365: \phi(E) = 6\,\Big(1-\frac{f_3(P)}{f_2(P)}\Big) =

366: 6\,\Big(1-\frac{f_0(Q)}{f_1(Q)}\Big).

367: \]

368: Thus maximizing the fatness of $E$ is equivalent to maximizing the ridge-facet

369: ratio $f_2(P)/f_3(P)$, or the average degree of the graph of $Q$.  It also

370: shows that the $E$-construction cannot achieve a fatness of $6$ or more.

371: 

372: In light of \prop{p:e-construction}, we would like to construct edge-tangent

373: simplicial $4$-polytopes. Regular simplicial $4$-polytopes (suitably scaled)

374: provide three obvious examples: the $4$-simplex $\Delta_4$, the cross polytope

375: $C_4\pol$, and the $600$-cell.  From these, the $E$-construction produces the

376: dual of the hypersimplex, the 24-cell, and a new $2$-simple, $2$-simplicial

377: polytope with $f$-vector $(720,3600,3600,720)$ and fatness $5$,

378: whose facets are bipyramids over pentagons.

379: 

380: We will construct new edge-tangent simplicial $4$-polytopes by gluing together

381: (not necessarily simplicial) edge-tangent $4$-polytopes, called \emph{atoms},

382: to form \emph{compounds}.  We must position the polytopes so that their

383: facets match, they remain edge-tangent, and the resulting compound is convex.

384: It will be very useful to interpret the interior of the $4$-dimensional unit

385: ball as the Klein model of hyperbolic $4$-space $\HH^4$, with $S^3$ the sphere

386: at infinity.  (See Iversen \cite{Iverson:hyperbolic} and Thurston

387: \cite[Chap.~2]{Thurston:geometry} for introductions to hyperbolic geometry.) In

388: particular, Euclidean lines are straight in the Klein model, Euclidean

389: subspaces are flat, and hence any intersection of a convex polytope with

390: $\HH^4$ is a convex (hyperbolic) polyhedron.  Even though the Klein model

391: respects convexity, it does not respect angles. However, angles and convexity

392: are preserved under hyperbolic isometries.  There are enough isometries to

393: favorably position certain $4$-polytopes to produce convex compounds.

394: 

395: If a polytope $Q$ is edge-tangent to $S^3$, then it is hyperbolically

396: hyperideal:  Not only its vertices, but also its edges, lie beyond the sphere

397: at infinity, except for the tangency point of each edge.  Nonetheless portions

398: of its facets and ridges lie in the finite hyperbolic realm.  As a hyperbolic

399: object the polytope $Q$ (more precisely, $Q\cap \HH^4$) is convex and has flat

400: facets. The ridge $r$ between any two adjacent facets has a well-defined

401: hyperbolic dihedral angle, which is strictly between $0$ and $\pi$ if (as in

402: our situation) the ridge properly intersects $\HH^4$.  To compute this angle we

403: can intersect $r$ at any point $t$ with any hyperplane $R$ that contains the

404: (hyperbolic) orthogonal complement to $r$ at $t$.  We let $t$ be the tangent

405: point of any edge $e$ of the ridge, and let $R$ be the hyperplane perpendicular

406: to $e$.

407: 

408: \begin{fullfigure}{f:hyplink}{A cone emanating from an ideal point $t$ in

409:     $\HH^3$ in the Poincar\'e model, and a horosphere $S$ incident to $t$.  The

410:     link of $t$ (here a right isosceles triangle) inherits Euclidean geometry

411:     from $S$}

412: \begin{large} \pspicture(-5.5,-5.2)(5.2,6.5)

413: \pscircle(0,0){5}

414: \pscurve[linestyle=dashed](2.5,2.165)(1.294,2.415)(0,2.5)(-1.294,2.415)

415:     (-2.5,2.165)(-3.536,1.768)(-4.33,1.25)(-4.83,0.647)(-5,0)

416: \pscurve[linestyle=dashed](2.5,2.165)(3.536,1.768)(4.33,1.25)(4.83,0.647)(5,0)

417: \pscurve(-5,0)(-4.83,-0.647)(-4.33,-1.25)(-3.536,-1.768)(-2.5,-2.165)

418:     (-1.294,-2.415)(0,-2.5)(1.294,-2.415)(2.5,-2.165)(3.536,-1.768)(4.33,-1.25)

419:     (4.83,-0.647)(5,0)

420: \pscircle(1.5,1.299){2}

421: \pscurve[linestyle=dashed](2.5,2.165)(2.018,2.265)(1.5,2.299)(0.982,2.265)

422:     (0.5,2.165)(0.086,2.006)(-0.232,1.799)(-0.432,1.558)(-0.5,1.299)

423: \pscurve[linestyle=dashed](2.5,2.165)(3.032,1.942)(3.379,1.641)(3.5,1.299)

424: \pscurve(-0.5,1.299)(-0.379,0.957)(-0.032,0.656)(0.5,0.433)(1.153,0.314)

425:     (1.847,0.314)(2.5,0.433)(3.032,0.656)(3.379,0.957)(3.5,1.299)

426: \pscurve[linestyle=dashed](2.5,2.165)(2.121,1.887)(1.751,1.715)(1.398,1.654)

427:     (1.07,1.705)(0.776,1.866)

428: \pscurve(0,4.33)(0.033,3.66)(0.13,3.064)(0.289,2.559)(0.507,2.156)(0.776,1.866)

429: \pscurve[linestyle=dashed](2.5,2.165)(2.171,1.822)(1.95,1.459)(1.841,1.082)

430:     (1.848,0.702)(1.97,0.327)

431: \pscurve(4.33,-1.25)(3.665,-1.02)(3.082,-0.74)(2.597,-0.416)(2.224,-0.057)

432:     (1.97,0.327)

433: \psline[linestyle=dashed](2.5,2.165)(0.5,0.433)

434: \psline(0.5,0.433)(-2.5,-2.165)

435: \pscurve(-2.5,-2.165)(-2.415,-0.971)(-2.165,0.29)(-1.768,1.531)(-1.25,2.667)

436:     (-0.647,3.622)(-0,4.33)

437: \pscurve(4.33,-1.25)(4.183,-0.087)(3.75,1.083)(3.062,2.178)(2.165,3.125)

438:     (1.121,3.859)(0,4.33)

439: \pscurve[linewidth=1.5pt](0.5,0.433)(0.512,0.706)(0.546,0.992)(0.602,1.285)

440:     (0.68,1.579)(0.776,1.866)

441: \pscurve[linewidth=1.5pt](0.5,0.433)(0.774,0.367)(1.065,0.323)(1.366,0.301)

442:     (1.67,0.303)(1.97,0.327)

443: \pscurve[linewidth=1.5pt](0.776,1.866)(0.973,1.581)(1.22,1.244)(1.491,0.894)

444:     (1.751,0.578)(1.97,0.327)

445: \qdisk(4.33,-1.25){1.8pt}

446: \qdisk(-2.5,-2.165){1.8pt}

447: \qdisk(2.5,2.165){1.8pt}

448: \qdisk(0,4.33){1.8pt}

449: \pspolygon[linewidth=1.5pt](-5.3,5)(-5.3,6.3)(-4,5)

450: \pcline[nodesepA=1.3,nodesepB=1,arrows=->](-4.867,5.433)(1.082,0.875)

451: \uput[-75](2.5,2.165){$t$}

452: \rput(1.5,-0.26){$S$}

453: \rput(-3.348,0.1){$\HH^3$}

454: \endpspicture

455: \end{large} \eatline \end{fullfigure}

456: 

457: Within the hyperbolic geometry of $R \cong \HH^3$, every line emanating from

458: the ideal point $t$ is orthogonal to any horosphere incident to $t$. Thus the

459: link of the edge $e$ of $Q$ is the intersection of $Q\cap R$ with a

460: sufficiently small horosphere $S$ at $t$.  Since horospheres have flat

461: Euclidean geometry \cite[p.~61]{Thurston:geometry}, the link $Q \cap S$ is a

462: Euclidean polygon.  Its edges correspond to the facets of $Q$ that contain $e$,

463: and its vertices correspond to the ridges of $Q$ that contain $e$. Thus the

464: dihedral angle of a ridge $r$ of $Q$ equals to the Euclidean angle of the

465: vertex $r \cap S$ of the Euclidean polygon $Q \cap S$. This is easier to see in

466: the Poincar\'e model of hyperbolic space, because it respects angles, than in

467: the Klein model. \fig{f:hyplink} shows an example.

468: 

469: To summarize:

470: 

471: \begin{lemma}

472: A compound of two or more polytopes is convex if and only if each ridge has

473: hyperbolic dihedral angle less than $\pi$, or equivalently, iff each edge

474: link is a convex Euclidean polygon.

475: \label{l:convex} \end{lemma}

476: 

477: Compounds can also have interior ridges with total dihedral

478: angle exactly $2\pi$.  But since all atoms of a compound are

479: edge-tangent, compounds do not have any interior edges or vertices.

480: 

481: If $Q$ is a regular polytope, then $Q \cap S$ is a regular polygon.

482: The following lemma is then immediate:

483: 

484: \begin{lemma}

485: If $Q$ is a regular, edge-tangent, simplicial $4$-polytope,

486: then in the hyperbolic metric of the Klein model, its dihedral angles

487: are $\pi/3$ (for the simplex), $\pi/2$ (for the cross polytope), and

488: $3\pi/5$ (for the $600$-cell).

489: \label{l:regang} \end{lemma}

490: 

491: A hyperideal hyperbolic object, even if it is an edge-tangent convex

492: polytope, can be unfavorably positioned so that it is unbounded as a Euclidean

493: object (\cf, Schulte \cite[p.~508]{Schulte:steinitz}).  Fortunately

494: there is always a bounded position as well:

495: 

496: \begin{lemma} Let be $Q$ an edge-tangent, convex polytope in $\R^d$ whose

497: points of tangency with $S^{d-1}$ do not lie in a hyperplane. Then there is a

498: hyperbolic isometry $h$ (extended to all of $\R^d$) such that $h(Q)$ is

499: bounded.

500: \label{l:bounded} \end{lemma}

501: \begin{proof} Let $p$ lie in the interior of the convex hull of the edge

502: tangencies and let $f$ be any hyperbolic motion that moves $p$ to the Euclidean

503: origin in $\R^d$.  Since the convex hull $K$ of the edge tangencies of $h(Q)$

504: contains the origin, $K\pol$ is a bounded polytope that circumscribes

505: $S^{d-1}$. Since $K\pol$ is facet-tangent where $h(Q)$ is edge-tangent, $h(Q)

506: \subset K\pol$.

507: \end{proof}

508: 

509: In the following we discuss three classes of edge-tangent simplicial convex

510: $4$-polytopes that are obtained by gluing in the Klein model: Compounds of

511: simplices, then simplices and cross polytopes, and finally compounds from cut

512: $600$-cells.  There are yet other edge-tangent compounds involving cross

513: polytopes cut in half (\ie, pyramids over octahedra), $24$-cells, and

514: hypersimplices as atoms, but we will not discuss these here.

515: 

516: \subsection{Compounds of simplices}

517: 

518: In this section we classify compounds whose atoms are simplices. This includes

519: all \emph{stacked} polytopes, which are simplicial polytopes that decompose as

520: a union of simplices without any interior faces other than facets. However

521: compounds of simplices are a larger class, since they may have interior ridges.

522: 

523: \begin{lemma}

524: Any edge-tangent $d$-simplex is hyperbolically regular.

525: \label{l:hypreg} \end{lemma}

526: 

527: \begin{proof} The proof is by induction on $d$, starting from the case $d=2$, where the

528: three tangency points define an ideal triangle in $\HH^2$. All ideal triangles

529: are congruent \cite[p.~83]{Thurston:geometry}.  Since edge-tangent triangles

530: are the polar duals of ideal triangles, they are all equivalent as well. 

531: 

532: If $d>2$, let $B$ be a general edge-tangent $d$-simplex. On the one hand, there

533: exists an edge-tangent simplex $A$ which is regular both in hyperbolic geometry

534: and Euclidean geometry.   On the other hand, given the position of $d$ of the 

535: vertices, there are at most two choices for the last vertex that  produce an

536: edge-tangent simplex, one on each side of the hyperplane spanned by the first

537: $d$. By induction there exists an isometry that takes a face of $B$ to a face

538: of $A$ and the remaining vertex to the same side.  The edge-tangent constraint

539: implies that this isometry takes the last vertex of $B$ to the last vertex of

540: $A$ as well.

541: \end{proof}

542: 

543: \begin{proposition}\label{p:simplices}

544: There are only three possible edge-tangent compounds of $4$-simplices:

545: \begin{itemize}\setlength{\itemsep}{0pt}

546: \item the regular simplex,

547: \item the bipyramid (a compound of two simplices that share a facet), and

548: \item the join of a triangle and a hexagon (a compound of six

549:   simplices that share a ridge).

550: \end{itemize}

551: \end{proposition}

552: 

553: \begin{proof} \fig{f:tjewels} shows all strictly convex polygons with

554: unit-length edges tiled by unit equilateral triangles, or \emph{triangle

555: jewels}.  Since the atoms of an edge-tangent compound of simplices are

556: edge-tangent, they are hyperbolically regular by \lem{l:hypreg}, and their edge

557: links are equilateral triangles. Thus every edge link of a compound of

558: simplices is a triangle jewel.  Any three $4$-simplices in a chain in such a

559: compound share a ridge. In order to create an edge link matching

560: \fig{f:tjewels}, they must extend to a ring of six simplices around the same

561: ridge. Adding any further simplex to these six would create an edge link in the

562: form of a triangle surrounded by three other triangles, which does not appear

563: in \fig{f:tjewels}.

564: \end{proof}

565: 

566: \begin{fullfigure}{f:tjewels}{The 3 possible edge links of edge-tangent

567:     compounds of $4$-simplices}

568: \psset{unit=.8cm} \pspicture(-2.5,-1)(3.5,1)

569: \rput(-1.5,0){\pspicture(-.5,-.6)(.5,.6) % triangle

570: \pspolygon(0,.577)(-.5,-.289)(.5,-.289)

571: \endpspicture}

572: \rput(0,0){\pspicture(-.5,-1)(.5,1) % lozenge

573: \psline(-.5,0)(0,.866)(.5,0)(-.5,0)(0,-.866)(.5,0)

574: \endpspicture}

575: \rput(2,0){\pspicture(-1,-1)(1,1) % hexagon

576: \pspolygon(1;0)(1;60)(1;120)(1;180)(1;240)(1;300)

577: \psline(1;0)(1;180) \psline(1;60)(1;240) \psline(1;120)(1;300)

578: \endpspicture}

579: \endpspicture \eatline \end{fullfigure}

580: 

581: Two of the $E$-polytopes produced by \prop{p:simplices} were previously known.

582: If $Q$ is the simplex, then $E$ is dual to the hypersimplex.  If $Q$ is the

583: bipyramid, then $E$ is dual to Braden's glued hypersimplex.  However, if $Q$ is

584: the six-simplex compound (dual to the product of a triangle and a hexagon),

585: then $E$ is a new $2$-simple, $2$-simplicial polytope with $f$-vector

586: $(27,108,108,27)$.

587: 

588: \prop{p:simplices} also implies an interesting impossibility

589: result.

590: 

591: \begin{corollary}

592: No stacked $4$-polytope with more than $6$ vertices is edge-tangent.

593: \end{corollary}

594: 

595: See Schulte~\cite[Thm.~3]{Schulte:steinitz} for the first examples of

596: polytopes that have no edge-tangent realization.

597: 

598: \subsection{Compounds of simplices and cross polytopes}

599: 

600: Next, we consider compounds of simplices and regular cross polytopes. The edge

601: link of any convex compound of these two types of polytopes must be one of the

602: eleven strictly convex polygons tiled by unit triangles and squares, or

603: \emph{square-triangle jewels} (\fig{f:stjewels}).  See Malkevitch

604: \cite{Malkevitch:tiling} and Waite \cite{Waite:combining} for work on convex

605: compounds of these shapes relaxing the requirement of strict convexity.

606: 

607: \begin{fullfigure}{f:stjewels}{The 11 possible edge links of edge-tangent

608:     compounds of $4$-simplices and cross polytopes}

609: \psset{unit=.8cm} \pspicture(-.75,.75)(10,8.75)

610: \rput(0,8){\pspicture(-.5,-.5)(.5,.5) % triangle

611: \pspolygon(0,.433)(-.5,-.433)(.5,-.433) \endpspicture}

612: \rput(1.5,8){\pspicture(-.5,-.5)(.5,.5) % square

613: \psframe(-.5,-.5)(.5,.5) \endpspicture}

614: \rput(3.25,8){\pspicture(-1,-.5)(1,.5) % lozenge

615: \psline(0,-.5)(.866,0)(0,.5)(0,-.5)(-.866,0)(0,.5) \endpspicture}

616: \rput(5.5,8){\pspicture(-.5,-.5)(1.5,.5) % house

617: \psframe(-.5,-.5)(.5,.5) \psline(.5,.5)(1.366,0)(.5,-.5)

618: \endpspicture}

619: \rput(8.25,8){\pspicture(-1.5,-.5)(1.5,.5) % football

620: \psframe(-.5,-.5)(.5,.5) \psline(.5,.5)(1.366,0)(.5,-.5)

621: \psline(-.5,.5)(-1.366,0)(-.5,-.5) \endpspicture}

622: \rput(.75,6.25){\pspicture(-1,-1)(1,1) % hexagon

623: \pspolygon(1;0)(1;60)(1;120)(1;180)(1;240)(1;300)

624: \psline(1;0)(1;180) \psline(1;60)(1;240) \psline(1;120)(1;300)

625: \endpspicture}

626: \rput(4,6){\pspicture(-1.5,-1)(1.5,1) % six+two

627: \pspolygon(.5,1)(-.5,1)(-1.366,.5)(-1.366,-.5)(-.5,-1)

628:     (.5,-1)(1.366,-.5)(1.366,.5)

629: \psline(-.5,-1)(-.5,1) \psline(.5,-1)(.5,1)

630: \psline(1.366,.5)(.5,0)(-.5,0)(-1.366,.5)

631: \psline(1.366,-.5)(.5,0) \psline(-.5,0)(-1.366,-.5)

632: \endpspicture}

633: \rput(7.75,5.75){\pspicture(-1.5,-1)(1.5,1) % eight+four

634: \pspolygon(.5,1.366)(-.5,1.366)(-1.366,.866)(-1.866,0)(-1.366,-.866)

635:     (-.5,-1.366)(.5,-1.366)(1.366,-.866)(1.866,0)(1.366,.866)

636: \pspolygon(0,-.5)(.866,0)(0,.5)(-.866,0)

637: \psline(1.866,0)(.866,0) \psline(-1.866,0)(-.866,0)

638: \psline(1.366,.866)(.866,0)(1.366,-.866)

639: \psline(-1.366,.866)(-.866,0)(-1.366,-.866)

640: \psline(.5,1.366)(0,.5)(-.5,1.366) \psline(.5,-1.366)(0,-.5)(-.5,-1.366)

641: \psline(0,.5)(0,-.5)

642: \endpspicture}

643: \rput(1.5,3){\pspicture(-2,-2)(2,2) % pie

644: \pspolygon(1.932;15)(1.932;45)(1.932;75)(1.932;105)(1.932;135)(1.932;165)

645:     (1.932;195)(1.932;225)(1.932;255)(1.932;285)(1.932;315)(1.932;345)

646: \pspolygon(1;30)(1;90)(1;150)(1;210)(1;270)(1;330)

647: \psline(1.932;15)(1;30)(1.932;45) \psline(1.932;75)(1;90)(1.932;105)

648: \psline(1.932;135)(1;150)(1.932;165) \psline(1.932;195)(1;210)(1.932;225)

649: \psline(1.932;255)(1;270)(1.932;285) \psline(1.932;315)(1;330)(1.932;345)

650: \psline(1;30)(1;210) \psline(1;90)(1;270) \psline(1;150)(1;330)

651: \endpspicture}

652: \rput(5.25,3.5){\pspicture(-2,-2)(2,2) % seven+three

653: \pspolygon(0,-1.866)(.866,-1.366)(1.366,-.5)(1.366,.5)(.5,1)(-.5,1)

654:     (-1.366,.5)(-1.366,-.5)(-.866,-1.366)

655: \pspolygon(0,-.866)(.5,0)(-.5,0)

656: \psline(1.366,-.5)(.5,0)(.5,1) \psline(-1.366,-.5)(-.5,0)(-.5,1)

657: \psline(-.866,-1.366)(0,-.866)(.866,-1.366)

658: \psline(1.366,.5)(.5,0) \psline(-1.366,.5)(-.5,0) \psline(0,-1.866)(0,-.866)

659: \endpspicture}

660: \rput(8.5,3){\pspicture(-2,-1.5)(0,2) % three+two

661: \pspolygon(-0,.5)(-.5,1.366)(-1.366,.866)(-1.866,0)(-1.366,-.866)

662:     (-.5,-1.366)(-0,-.5)

663: \psline(-.866,0)(-0,.5) \psline(-.866,0)(0,-.5) \psline(-.866,0)(-1.366,.866)

664: \psline(-.866,0)(-1.866,0) \psline(-.866,0)(-1.366,-.866)

665: \endpspicture}

666: \endpspicture \eatline \end{fullfigure}

667: 

668: If $Q$ is a single cross polytope, then $E$ is a 24-cell. We can also glue

669: simplices onto subsets of the facets of the cross polytope. The new dihedral

670: angles formed by such a gluing are $5\pi/6$. The resulting compound is convex

671: as long as no two glued cross polytope facets share a ridge.  We used a

672: computer program to list the combinatorially distinct ways of choosing a subset

673: of nonadjacent facets of the cross polytope; the results may be summarized as

674: follows. In addition to the 24-cell, this yields 20 new $2$-simple,

675: $2$-simplicial polytopes.

676: 

677: \begin{proposition}

678: There are exactly $21$ distinct simplicial edge-tangent compounds composed of

679: one regular $4$-dimensional cross polytope and $k\ge0$ simplices, according to

680: the following table:

681: $$\begin{tabular}{r@{\;}|@{\;}ccccccccc@{\;}|@{\;}c}

682: $k$&0&1&2&3&4&5&6&7&8&Total\\ \hline

683: \#&1&1&3&3&6&3&2&1&1&21\\

684: \end{tabular}$$

685: \label{p:crosspoly} \end{proposition}

686: 

687: We can also confirm that every square-triangle jewel arises as the edge link of

688: an edge-tangent $4$-dimensional compound.  For every jewel other than the one

689: in the center, we can form a convex \emph{edge bouquet} consisting of simplices

690: and cross polytopes that meet at an edge: we replace each triangle by a simplex

691: and each square by a cross polytope.  Since the central jewel has two adjacent

692: squares, its edge bouquet is not convex.  Instead we glue two cross polytopes

693: along a facet so that the 4 ridges of that facet are flush, \ie, their dihedral

694: angle is $\pi$.  Thus we can ``caulk'' each such ridge with three simplices

695: that share the ridge.  The central jewel is the link of 6 of the edges of the

696: resulting compound of 2 cross polytopes and 12 simplices.

697: 

698: Simplices and regular cross polytopes combine to form many other edge-tangent

699: simplicial polytopes and hence $2$-simple, $2$-simplicial

700: polytopes. In particular, these methods lead to the following theorem.

701: 

702: \begin{theorem}

703: There are infinitely many combinatorially distinct $2$-simple, $2$-simplicial

704: facet-tangent $4$-polytopes.

705: \label{th:inf} \end{theorem}

706: 

707: \begin{proof} We glue $n$ cross polytopes end-to-end. Each adjacent pair

708: produces 4 flush ridges that we caulk with chains of three simplices.  The

709: facets to which these simplices are glued are not adjacent and so do not

710: produce any further concavities.

711: \end{proof}

712: 

713: The chain of $n$ cross polytopes has $f$-vector

714: \[

715: (4n+4,18n+6,28n+4,14n+2).

716: \]

717: Filling a concavity adds $(2,9,14,7)$, so after filling the $4(n-1)$

718: concavities we get a simplicial polytope $Q$ with

719: \[

720: f(Q)\;=\;(12n- 4,54n-30,84n-54,42-26),

721: \]

722: which yields a $2$-simple, $2$-simplicial $4$-polytope $E$

723: with

724: \[

725: f(E)\;=\;(54n-30,252n-156,252n-156,54n-30)

726: \]

727: by \prop{p:e-construction}.

728: 

729: \begin{remark} Every $2$-simple, $2$-simplicial $4$-polytopes that we know is

730: combinatorially equivalent to one which circumscribes the sphere.  Are there

731: any that are not?

732: 

733: We do know a few $2$-simple, $2$-simplicial $4$-polytopes which are not

734: $E$-polytopes.  Trivially there is the simplex. There are a few others that

735: arise by the fact that the $24$-cell is the $E$-polytope of a cross polytope

736: in 3 different ways.  Color the vertices of a $24$-cell red, green, and blue,

737: so that the vertices of each color span a cross polytope. If we cap one facet

738: of a cross-polytope by a simplex and apply the $E$-construction, the result is

739: a $24$-cell in which 6 facets that meet at 1 vertex are replaced by 10 facets

740: and 4 vertices.  If the replaced vertex is red, the replacement can be induced

741: by capping either blue cross polytope or the green cross polytope and then

742: applying the $E$-construction; the position of the replacement differs between

743: the two cases. If we replace two different red vertices, one by capping the

744: green cross polytope and the other by capping the blue cross polytope, then

745: the resulting polytope is $2$-simple and $2$-simplicial but not an

746: $E$-polytope.  This construction has several variations: for example, we can

747: also replace three vertices, one of each color.

748: \end{remark}

749: 

750: \subsection{Compounds involving the $600$-cell}

751: \label{s:600}

752: 

753: If $Q$ is the $600$-cell, then $E$ is a $2$-simple, $2$-simplicial polytope

754: with $f$-vector $(720,3600,3600,720)$ and fatness exactly $5$.  Again, we can

755: glue simplices onto any subset of nonadjacent facets of the $600$-cell,

756: creating convex compounds with dihedral angle $14\pi/15$. We did not count the

757: (large) number of distinct ways of choosing such a subset, analogous to

758: \prop{p:crosspoly}. It is not possible to glue a cross polytope to a

759: $600$-cell, because that would create an $11\pi/10$ angle (\ie, a concave

760: dihedral of $9\pi/10$) which cannot be filled by additional simplices or cross

761: polytopes.

762: 

763: The large dihedral angles of the $600$-cell make it difficult to form

764: compounds from it, but we can modify it as follows to create smaller dihedrals.

765: Remove a vertex and form the convex hull of the remaining 119 vertices.

766: The resulting convex polytope has 580 of the $600$-cell's tetrahedral facets

767: and one icosahedral facet. The pentagonal edge link (\fig{f:cutpents}(a)) of

768: the edges bordering this new facet become modified in a similar way, by

769: removing one vertex and forming the convex hull of the remaining four vertices

770: (\fig{f:cutpents}(b)), which results in a trapezoid; thus, the hyperbolic

771: dihedrals at the ridges around the new facet are $2\pi/5$.

772: 

773: \begin{fullfigure}{f:cutpents}{Edge figures of (a) a $600$-cell, (b) a

774:     $600$-cell with one vertex removed, (c) a $600$-cell with two removed

775:     vertices, and (d) an icosahedral cap}

776: \pspicture(0,-1)(9,10)

777: \rput(2,8){\pspicture(-2,-2)(2,2)

778: \pspolygon(2;90)(2;162)(2;234)(2;306)(2;18)

779: \uput[270](2;90){$\frac{3\pi}{5}$} \psarc(2;90){1.1}{216}{324}

780: \endpspicture} \rput(2,5.5){\large (a)}

781: \rput(7,8){\pspicture(-2,-2)(1.25,2)

782: \pspolygon(2;90)(2;162)(2;234)(2;306)

783: \uput{.4}[253](2;90){$\frac{2\pi}{5}$} \psarc(2;90){1.25}{216}{288}

784: \endpspicture} \rput(7,5.5){\large (b)}

785: \rput(2,2){\pspicture(-1.25,-2)(1.25,2)

786: \pspolygon(2;90)(2;234)(2;306)

787: \uput{.65}[270](2;90){$\frac{\pi}{5}$} \psarc(2;90){1.5}{252}{288}

788: \endpspicture} \rput(2,-.5){\large (c)}

789: \rput(7,2){\pspicture(-2,-2)(.25,2)

790: \pspolygon(2;90)(2;162)(2;234)

791: \uput{.65}[234](2;90){$\frac{\pi}{5}$} \psarc(2;90){1.5}{216}{252}

792: \endpspicture} \rput(7,-.5){\large (d)}

793: \endpspicture \eatline \end{fullfigure}

794: 

795: This cut polytope is not simplicial, but we may glue two of these polytopes

796: together along their icosahedral facets, forming a simplicial polytope with

797: $4\pi/5$ dihedrals along the glued ridges.  This compound's new edge links are

798: hexagons formed by gluing pairs of trapezoids (\fig{f:link600}(a)).  The

799: same cutting and gluing process may be repeated to form a sequence or tree of

800: $600$-cells, connected along cuts that do not share a ridge.  For such a chain

801: or tree formed from $n$ cut $600$-cells, the $f$-vector may be computed as

802: \[

803: f(Q)\;=\;(106n+14,666n+54,1120n+80,560n+40),

804: \]

805: so the $E$-construction yields

806: \[

807: f(E)\;=\;(666n+54,3360n+240,3360n+240,666n+54)

808: \]

809: and thus a fatness of

810: \[

811: \phi(E)\;=\;\frac{3360n+240}{666n+54}\ \

812: \longrightarrow\ \ \frac{560}{111}\ \approx\ 5.045045.

813: \]

814: Thus the fatness of the $2$-simple, $2$-simplicial polytopes formed by such

815: compounds improves slightly on that formed from the $600$-cell alone.

816: 

817: \begin{fullfigure}{f:link600}{Edge links of compounds of cut $600$-cells}

818: \pspicture(0,-5)(13.5,4)

819: \rput(2.5,0){\pspicture(-2,-2.5)(2,2.5)

820: \pspolygon(1.176,2.236)(-1.176,2.236)(-1.902,0)

821:     (-1.176,-2.236)(1.176,-2.236)(1.902,0)

822: \psline(-1.902,0)(1.902,0)

823: \endpspicture}

824: \rput(2.5,-4.5){\large (a)}

825: \rput(9.5,0){\pspicture(-4,-4)(4,4)

826: \pspolygon(3.804;0)(3.804;36)(3.804;72)(3.804;108)(3.804;144)

827:     (3.804;180)(3.804;216)(3.804;252)(3.804;288)(3.804;324)

828: \psline(3.804;0)(3.804;180) \psline(3.804;36)(3.804;216)

829: \psline(3.804;72)(3.804;252) \psline(3.804;108)(3.804;288)

830: \psline(3.804;144)(3.804;324)

831: \endpspicture}

832: \rput(9.5,-4.5){\large (b)}

833: \endpspicture

834: \end{fullfigure}

835: 

836: It is also possible to form compounds involving $600$-cells which have been cut

837: by removing several vertices (as described above) so that two of the resulting

838: icosahedral facets meet at a ridge. Each edge link at this ridge is an

839: isosceles triangle formed by removing two vertices from a pentagon

840: (\fig{f:cutpents}(c)). Thus the dihedral angle of the triangular ridge between

841: the icosahedral facets is $\pi/5$. We can therefore form compounds in which ten

842: of these doubly-cut $600$-cells share a triangle, whose edges links are a

843: regular decagon cut into ten isosceles triangles (\fig{f:link600}(b)).

844: Yet other compounds of cut $600$-cells and simplices are possible,

845: although we do not need them here.

846: 

847: The cut $600$-cells also form more complicated compounds which require some

848: group-theoretic terminology to explain.  The vertices of a regular $600$-cell

849: form a 120-element group under (rescaled) quaternionic multiplication, the

850: binary icosahedral group.  This group has a 24-element subgroup, the binary

851: tetrahedral group, which also arises as the units of the Hurwitz integers (see

852: Conway and Sloane \cite[\S2.2.6,8.2.1]{CS:splag}).  Let $A$ be the convex hull

853: of the other 96 vertices of the $600$-cell;  \ie, $A$ is formed by cutting 24

854: vertices from $600$-cell in the above manner. The resulting polytope is the

855: ``snub$\{3,4,3\}$'' (snub 24-cell) of Coxeter

856: \cite[\S8.4,8.5]{Coxeter:regpoly}.  Its $f$-vector is $(96,432,480,144)$.

857: Every icosahedral facet of $A$ is adjacent to $8$ other icosahedron facets, as

858: well as to $12$ tetrahedra. Thus $A$ has $96$ icosahedron-icosahedron ridges.

859: 

860: We can build new hyperbolic, edge-tangent, simplicial polytopes by gluing

861: copies of $A$ along icosahedral faces and capping the remaining icosahedral

862: facets with pyramidal caps of the type that we had cut off to form $A$. (The

863: edge links of such a cap $C$ are given by \fig{f:cutpents}(d).) The resulting

864: polytope $Q$ will be convex if at each icosahedral-icosahedral ridge of a copy

865: of $A$, either $10$ copies of $A$ meet, or two caps and one or two copies of

866: $A$ do.  Also at each icosahedral-tetrahedral ridge of a copy of $A$, either

867: two copies of $A$ or one each of $A$ and a cap must meet. If two copies of $A$

868: meet (in an icosahedral facet $F$), then they differ by a reflection through

869: $F$. These reflections generate a discrete hyperbolic reflection group $\Gamma$

870: since the supporting hyperplanes of the icosahedral facets (the facets of a

871: hyperideal $24$-cell, whose ridges are also ridges of $A$!) satisfy the Coxeter

872: condition: When they meet, they meet at an angle of $\pi/5$, which divides

873: $\pi$. Thus the copies of $A$ used in $Q$ are a finite subset $\Sigma$ of the

874: orbit of $A$ under $\Gamma$.  The set $\Sigma$ determines $Q$.

875: 

876: \begin{remark} That it suffices to consider the dihedral angles of adjacent

877: facets follows from Poincar\'e's covering-space argument:  Let $P$ be a

878: spherical, Euclidean, or hyperbolic polytope whose dihedral angles divide

879: $\pi$.  Let $X$ denote the space in which it lives.  Let $Y$ be the disjoint

880: union of all copies of $P$ in $X$ in every position, and let $Z$ be the

881: quotient of $Y$ given by identifying two copies of $P$ along a shared facet.

882: The space $Z$ is constructed abstractly so that $P$ tiles it.

883: 

884: We claim that $Z$ is a covering space of $X$.  Each $p \in Z$ lies in the

885: interior of some face $F$ of a copy of $P$.  If $F$ is a copy of $P$ or a

886: facet, this is elementary; if $F$ is a ridge, it follows from the dihedral

887: angle condition.  Otherwise it follows by applying the covering-space argument

888: inductively, replacing $X$ by the link $S$ of $F$ and $P$ by $P \cap S$.

889: 

890: Since $Z$ is a covering space, a connected component of $Z$ is a tiling of $X$

891: by $P$. See Vinberg \cite{Vinberg:reflection} for a survey of hyperbolic

892: reflection groups.

893: \end{remark}

894: 

895: There is no one best choice for $\Sigma$, only a supremal limit.  One

896: reasonable choice for $\Sigma$ is the corona of a copy of $A$, \ie, $A$

897: together with the set of all images under $\Gamma$ that meet it, necessarily at

898: a facet or a ridge. The corona of $A$ is depicted by a simplified (and

899: therefore erroneous) schematic in \fig{f:corona}; the reader should imagine the

900: correct, more complicated version.  The schematic uses a chemistry notation in

901: which each copy of $A$ is represented as an \emph{atom}, each pair of copies

902: that shares a facet is represented as a \emph{bond}, and ``$7A$'' denotes a

903: chain of 7 atoms.  To extend the terminology, we call 10 copies of $A$ that

904: meet at a ridge a \emph{ring}. The schematic is simplified in that the central

905: copy actually has 24 bonds (not 6), each of the neighbors has 9 bonds (not 3),

906: and there are 96 rings (not 6).

907: 

908: \begin{fullfigure}{f:corona}{An oversimplified schematic of a corona of $A$}

909: \begin{large} \pspicture(-4,-4)(4,4) \psset{nodesep=3pt}

910: \rput(0,0){\rnode{a}{$A$}}

911: \rput(2.4;0){\rnode{b1}{$A$}} \rput(2.4;60){\rnode{b2}{$A$}}

912: \rput(2.4;120){\rnode{b3}{$A$}} \rput(2.4;180){\rnode{b4}{$A$}}

913: \rput(2.4;240){\rnode{b5}{$A$}} \rput(2.4;300){\rnode{b6}{$A$}}

914: \rput(3.6;30){\rnode{c1}{$7A$}} \rput(3.6;90){\rnode{c2}{$7A$}}

915: \rput(3.6;150){\rnode{c3}{$7A$}} \rput(3.6;210){\rnode{c4}{$7A$}}

916: \rput(3.6;270){\rnode{c5}{$7A$}} \rput(3.6;330){\rnode{c6}{$7A$}}

917: \ncline{a}{b1} \ncline{a}{b2} \ncline{a}{b3}

918: \ncline{a}{b4} \ncline{a}{b5} \ncline{a}{b6}

919: \ncline{b1}{c1} \ncline{b1}{c6} \ncline{b2}{c2} \ncline{b2}{c1}

920: \ncline{b3}{c3} \ncline{b3}{c2} \ncline{b4}{c4} \ncline{b4}{c3}

921: \ncline{b5}{c5} \ncline{b5}{c4} \ncline{b6}{c6} \ncline{b6}{c5}

922: \endpspicture \end{large} \eatline \end{fullfigure}

923: 

924: Let $Q$ be the union of these copies of $A$ with the remaining icosahedral

925: facets capped.  To compute the $f$-vector of $Q$ it is easier to view each atom

926: as a copy of a 600-cell $B$, minus two caps for each bond.  There are $1+24+96

927: \cdot 7 = 697$ atoms and $24+ 96 \cdot 8 = 792$ bonds in total. Thus $Q$ has

928: \[

929: f_3(Q) = 697 f_3(B) - 792 \cdot 2 f_3'(C) = 386520

930: \]

931: facets, where $f_3'(C)=30$ is the number of simplicial facets of a cap $C$.

932: 

933: Counting vertices is more complicated.  Let $I$ be an icosahedron

934: and let $T$ be a triangle.  Then

935: \begin{align*}

936: f_0(Q) &= 697 f_0(B) - 792 \cdot 2 f_0(C) + 792 f_0(I) + 96 f_0(T) \\

937: &= 72840,

938: \end{align*}

939: because after the vertices of the caps are subtracted, the vertices of each

940: icosahedral facet at a bond are undercounted once, and after these are restored

941: the vertices of each triangle at the center of a ring are undercounted once.

942: The rest of the $f$-vector of $Q$ follows from the Dehn-Sommerville equations:

943: \begin{align*}

944: f_2(Q) &= 2f_3(Q) = 773040, \\

945: f_1(Q) &= f_0(Q) + f_3(Q) = 459360.

946: \end{align*}

947: The polytope $Q$ yields an $E$-polytope with fatness

948: \[

949: \phi(E) = 6\,\frac{f_3(Q)}{f_1(Q)} = \frac{3221}{638} \approx 5.048589.

950: \]

951: Note that since this bound arises from a specific choice of $\Sigma$

952: rather than a supremal limit, this is not optimal as a lower

953: bound of supremal fatness.

954: 

955: \subsection{Kissing numbers}

956: \label{s:kissing}

957: 

958: As mentioned above, another use of ridge-tangent polytopes $P$ is the average

959: kissing number problem \cite{Kuperberg:kissing}.  Let $X$ be a finite packing

960: of (not necessarily congruent) spheres $S^3$, which is equivalent to a finite

961: sphere packing in $\R^3$ by stereographic projection. The question is to

962: maximize the average number of kissing points of the spheres in $X$.  If $P$ is

963: ridge-tangent, its facets intersect the unit sphere $S^3$ in a sphere packing

964: $X$, in which the spheres kiss at the tangency points of $P$. Thus the

965: ridge-facet ratio of $P$ is exactly half the average kissing number of $X$.

966: (Not all sphere packings come from ridge-tangent polytopes in this way.)

967: 

968: The sphere packings due to Kuperberg and Schramm \cite{Kuperberg:kissing} can

969: be viewed as coming from a compound consisting of a chain or tree of $n$ cut

970: $600$-cells (\ie, atoms in the sense of \fig{f:corona}).  Their average kissing

971: numbers are

972: \[

973: \kappa\; =\; 2\,\frac{666n+54}{106n+14}\ \

974: \longrightarrow\ \ \frac{666}{53}\ \approx\ 12.56603.

975: \]

976: By contrast if $Q$ is the compound formed from a corona of $A$ as in

977: \sec{s:600}, then the average kissing number of the corresponding sphere

978: packing is

979: \[

980: \kappa(Q) = 2\,\frac{f_1(Q)}{f_0(Q)} = \frac{7656}{607} \approx 12.61285.

981: \]

982: Like the bound on fatness, it is not optimal as a lower bound

983: on the supremal average kissing number.

984: 

985: Here we offer no improvement on the upper bound

986: \[

987: \kappa < 8+4\sqrt{3} \approx 14.92820

988: \]

989: from \cite{Kuperberg:kissing}, even though it cannot be optimal either.

990: 

991: \section{3-Spheres}

992: \label{s:spheres}

993: 

994: In this section we construct a family of strongly regular cellulations of the

995: $3$-sphere with unbounded fatness.  Indeed, we provide an efficient version of

996: the construction, in the sense that it requires only polynomially many cells to

997: achieve a given fatness.  (The construction is also a polynomially effective

998: randomized algorithm.)  Given $N$, we find a strongly regular cellulation of

999: $S^3$ with $O(N^{12})$ cells and fatness at least $N$.  Note that there are

1000: also power-law upper bounds on fatness:  An $O(N^{1/3})$ upper bound for the

1001: fatness of a convex $4$-polytope with $N$ vertices follows from work by

1002: Edelsbrunner and Sharir \cite{ES:hyperplane}.  The K\H{o}vari-S\'os-Tur\'an

1003: theorem \cite{KST:problem} (see also \cite[p.~1239]{Bollobas:extremal} and

1004: \cite[Thm.~9.6,p.~121]{PA:combin}) implies an $O(N^{2/3})$ upper bound on the

1005: fatness of strongly regular cellulations of $S^3$, since the vertex-facet

1006: (atom-coatom) incidence graph has no $K_{3,3}$-subgraph, and thus has at most

1007: $O\big(f_0(f_0+f_3)^{2/3}\big)$ edges. Our construction provides an

1008: $\Omega(N^{1/12})$ lower bound.

1009: 

1010: The inefficient construction is a simpler version which we describe first.  For

1011: every $g>0$, $S^3$ can be realized as

1012: \[

1013: H_1 \cup (S_g \times I) \cup H_2,

1014: \]

1015: a thickened surface of genus $g$ capped on both ends with handlebodies. (This

1016: is obtained from a neighborhood of the standard [unknotted] smooth embedding

1017: of $S_g$ into $S^3$.) If for some $g$ we can find a fat cellulation of $S_g$,

1018: we can realize $S^3$ as a ``fat sausage with lean ends,'' as shown in

1019: \fig{f:sausage}.  We cross the fat cellulation of $S_g$ with an interval

1020: divided into $N$ segments to produce a fat cellulation of $S_g \times I$.  Then

1021: we fix arbitrary strongly regular cellulations of the handlebodies $H_1$ and

1022: $H_2$.  If we make the sausage long enough, \ie, if we take $N \to \infty$, the

1023: fatness of the sausage converges to the fatness of its middle regardless of the

1024: structure of its ends.

1025: 

1026: \begin{fullfigure}{f:sausage}{$S^3$ as a fat sausage with lean ends}

1027: \begin{large} \psset{unit=.2in} \pspicture(-8,-2.5)(8,3)

1028: \multips(-6,0)(4,0){4}{

1029:     \psbezier(.7,-1)(.7,-1.5)(.35,-2)(0,-2)

1030:     \psbezier(.7,-1)(.7,-.5)(.35,-.5)(.35,0)

1031:     \psbezier(.35,0)(.35,.5)(.7,.5)(.7,1)

1032:     \psbezier(.7,1)(.7,1.5)(.35,2)(0,2)

1033:     \psbezier(-.7,-1)(-.7,-1.5)(-.35,-2)(0,-2)

1034:     \psbezier(-.7,-1)(-.7,-.5)(-.35,-.5)(-.35,0)

1035:     \psbezier(-.35,0)(-.35,.5)(-.7,.5)(-.7,1)

1036:     \psbezier(-.7,1)(-.7,1.5)(-.35,2)(0,2)

1037:     \psbezier(0,.65)(.25,.9)(.25,1.15)(0,1.4)

1038:     \psbezier(0,.65)(-.25,.9)(-.25,1.15)(0,1.4)

1039:     \psbezier(0,-.65)(.25,-.9)(.25,-1.15)(0,-1.4)

1040:     \psbezier(0,-.65)(-.25,-.9)(-.25,-1.15)(0,-1.4)}

1041: \psline(-6,2)(6,2)\psline(-6,-2)(6,-2)

1042: \psarc(6,0){2}{270}{90}\psarc(-6,0){2}{90}{270}

1043: \psline[linestyle=dashed](-6,1.4)(6,1.4)

1044: \psline[linestyle=dashed](-6,.65)(6,.65)

1045: \psline[linestyle=dashed](-6,-1.4)(6,-1.4)

1046: \psline[linestyle=dashed](-6,-.65)(6,-.65)

1047: \rput(-7.25,0){$H_1$}\rput(0,0){$S_g \times I$}\rput(7.25,0){$H_2$}

1048: \endpspicture \end{large} \eatline \end{fullfigure}

1049: 

1050: It remains only to show that there are strongly regular fat cellulations of

1051: surfaces.  The surface $S_g$ has perfect cellulations with $f$-vector

1052: $(1,2g,1)$.  Such a cellulation is obtained by gluing pairs of sides of a

1053: $4g$-gon in such a way that all vertices are identified. It has fatness $g$,

1054: and it exists for arbitrarily large $g$, but it is far from regular. However,

1055: its lift to the universal cover $\tilde{S}_g$ is strongly regular, since any

1056: such cellulation can be represented by a tiling of the hyperbolic plane by

1057: \emph{convex} polygons.  (Indeed, if we take the regular $4g$-gon with angles

1058: of $\pi/(2g)$, which is certainly convex, then its edges and angles are

1059: compatible with any perfect cellulation.)  Moreover, Mal'cev's theorem

1060: \cite{Malcev:faithful}, states that finitely generated matrix groups are

1061: residually finite; this implies that every closed hyperbolic manifold admits

1062: intermediate finite covers with arbitrarily large injectivity radius.  (See

1063: \cite[\S4]{Kuperberg:saturated} for a detailed exposition.) In particular $S_g$

1064: admits an intermediate cover $\hS_g$ whose injectivity radius exceeds the

1065: diameter of a $2$-cell.  The cellulation of $\hS_g$ is then strongly

1066: regular.  Its genus is much larger than $g$, but its fatness is still $g$.

1067: 

1068: The efficient construction is the same: It only requires careful choices for

1069: the finite cover $\hS_g$ and for the handlebodies $H_1$ and $H_2$. Among

1070: the perfect cellulations of $S_g$, a convenient one for us is a $4g$-gon with

1071: opposite edges identified.  We describe the fundamental group $\pi_1(S_g)$

1072: using this cellulation.  As shown in \fig{f:sg}, we number the edges

1073: $x_0,\dots,x_{4g-1}$ consecutively, so that

1074: \eq{e:ident}{x_i^{} = x_{2g+i}^{-1}}

1075: and

1076: \eq{e:cell}{x_0x_1\cdots x_{4g-1} = 1.}

1077: (We interpret the indices as elements of $\Z/(4g)$.)

1078: Equation~\eqref{e:ident} expresses the identifications, while

1079: equation~\eqref{e:cell} expresses the boundary of the $2$-cell.

1080: 

1081: \begin{fullfigure}{f:sg}{Labelling edges of $S_g$}

1082: \pspicture(-4.5,-4.5)(4,4.5) \sgpie

1083: \uput[90](3.435;90){$x_0$} \uput[120](3.435;120){$x_1$}

1084: \uput[150](3.435;150){$x_2$} \uput[240](3.435;240){$x_{2g-1}$}

1085: \uput[270](3.435;270){$x_{2g}$} \uput[60](3.435;60){$x_{4g-1}$}

1086: \eatline \endpspicture \end{fullfigure}

1087: 

1088: We construct $\hS_g$ as a tower of two abelian finite covers, which

1089: together form an irregular cover of $S_g$. (No abelian cover of $S_g$ is

1090: strongly regular.) The surface $S_g$ satisfies the usual requirements of

1091: covering-space

1092: theory (see Fulton \cite[\S\ 13b, 14a]{Fulton:gtm}): It is a connected,

1093: locally path-connected, and locally simply connected space. For any finite

1094: group $A$, the \emph{$A$-coverings} of $S_g$ are covering spaces of the form

1095: $Y$ with $Y/A=S_g$, where $A$ acts properly discontinuously on $Y$. These

1096: coverings, up to isomorphism, are classified by the set of group homomorphisms

1097: $\Hom(\pi_1(S_g,x),A)$ \cite[Thm.~14.a]{Fulton:gtm}. Furthermore, if $A$ is

1098: abelian, then every such homomorphism maps the commutators in $\pi_1(S_g,x)$ to

1099: zero, so the $A$-coverings are classified by $\Hom(H_1(S_g,\Z),A)$. In other

1100: words, if $A$ is any abelian group with $n$ elements, then every homomorphism

1101: $\sigma: H_1(S_g)\rightarrow A$ defines an $n$-fold abelian covering of $S_g$.

1102: (If the homomorphism is not surjective, then the covering space is not

1103: connected \cite[p.~193]{Fulton:gtm}. In this case we use a connected component

1104: of the covering space, which has the same fatness but smaller genus.)

1105: 

1106: Now assume that $q = 4g+1$ is a prime power and let $\alpha$ generate the

1107: cyclic group $\F_q^*$, where $\F_q$ is the field with $q$ elements.  We can

1108: fulfill the assumption by changing $g$ by a bounded factor.  (Most simply we

1109: can let $q = 5^k$.  Or we can let $q$ be prime, so that $\F_q=\Z/q$, by a form

1110: of Bertrand's postulate for primes in congruence classes.  This result dates to

1111: the 19th century; see Erd\H{o}s \cite{Erdos:primzahlen} for an elementary

1112: proof.) Define a homomorphism $\sigma:H_1(S_g) \to \F_q$ by $\sigma([x_i]) =

1113: \alpha^i$, where $[x_i]$ is the 1-cycle (or homology class) represented by the

1114: loop $x_i$.  Since $\alpha^{2g} = -1$, the definition of $\sigma$ is consistent

1115: with equation~\eqref{e:ident}. Consistency with \eqref{e:cell} is then

1116: automatic. Let $S_g'$ be the finite cover corresponding to $\sigma$.

1117: 

1118: To prepare for the subsequent analysis of $\hS_g$, we give an explicit

1119: combinatorial description of $S'_g$.  Let $F^0$ be a $2$-cell of $S'_g$

1120: and label its vertices

1121: $$v_0^0,v_1^0,\ldots,v_{4g-1}^0$$

1122: in cyclic order.  See \fig{f:sg1}.

1123: 

1124: \begin{fullfigure}{f:sg1}{Labelling the vertices of $F^0$}

1125: \pspicture(-5,-4.75)(5,4.75) \sgpie \rput(0,0){\large $F^0$}

1126: \uput[90](3.435;90){$+1$} \uput[120](3.435;120){$+\alpha$}

1127: \uput[150](3.435;150){$+\alpha^2$}

1128: \uput{10pt}[270](3.435;270){$+\alpha^{2g} = -1$}

1129: \uput[300](3.435;300){$+\alpha^{2g} = -\alpha$}

1130: \uput[45](4;45){$v^0_{4g-1}$} \uput[75](4;75){$v^0_0$}

1131: \uput[105](4;105){$v^0_1$}

1132: \eatline \endpspicture \end{fullfigure}

1133: 

1134: For each $s \in \F_q$, let $F^s$ and $v_k^s$ be the images of $F^0$ and $v_k^0$

1135: under the action of $s$. Since $S_g$ has only one vertex, the vertices of

1136: $S_g'$ may be identified with $\F_q$. Thus, if we identify $v_0^0$ with

1137: $0\in\F_q$, then the action of $\F_q$ will identify $v_0^s$ with $s\in\F_q$.

1138: The structure of $\sigma$ further implies that

1139: \[

1140: v_k^s\ =\ s+1+\alpha+\alpha^2+\ldots+\alpha^{k-1}\ =\

1141:     s+\frac{\alpha^k-1}{\alpha-1}

1142: \]

1143: for all $k\in\Z/(4g)$ and $s\in\F_q$.  See \fig{f:sg2}.

1144: 

1145: \begin{fullfigure}{f:sg2}{Labelling the vertices of $F^s$}

1146: \pspicture(-7,-4.75)(5.5,4.75) \sgpie \rput(0,0){\large $F^s$}

1147: \uput[75](4;75){$v^s_0 = s$}

1148: \uput[105](4;105){$v^s_1 = s+1$}

1149: \uput[120](4;135){$v^s_2 = s+1+\alpha$}

1150: \uput[135](4;165){$v^s_3 = s+\tfrac{\alpha^3-1}{\alpha-1}$}

1151: \eatline \endpspicture \end{fullfigure}

1152: 

1153: Using this explicit description, it is routine to verify the following

1154: (remarkable) properties of the surface $S'_g$.

1155: 

1156: \begin{lemma}

1157: The cellulation of the abelian cover $S'_g$ is regular and has $f$-vector

1158: \[

1159: f(S'_g) = (q,2gq,q)=(q,\binom{q}{2},q).

1160: \]

1161: Every facet has $q-1=4g$ vertices, while every vertex has degree $q-1=4g$. The

1162: graph (or $1$-skeleton) of $S_g'$ is the complete graph on $q+1$ vertices. The

1163: dual graph is also complete; any two facets share exactly one edge (as well as

1164: $q-4$ other vertices).

1165: \label{l:regular} \end{lemma}

1166: 

1167: \begin{proof}

1168: In view of the combinatorial description above (\fig{f:sg2}), all these

1169: facts follow from simple computations in the field $\F_q$:

1170: \begin{itemize}\setlength{\itemsep}{0pt}

1171: \item $S_g'$ is regular --- for each $s\in\F_q$, the vertex labels

1172: $s+\frac{\alpha^k-1}{\alpha-1}$ ($0\le k<4g$) are distinct.

1173: \item The $1$-skeleton of $S_g'$ is complete --- for $v,v'\in\F_q$, $v\neq v'$

1174: there is a unique $s\in\F_q$ and $k\in\Z/(4g)$ with

1175: \[

1176: v=s+\frac{\alpha^k-1}{\alpha-1}\qquad

1177:     v'=s+\frac{\alpha^{k+1}-1}{\alpha-1}.

1178: \]

1179: \item The dual graph of $S_g'$ is complete --- for $s,s'\in\F_q$, $s\neq s'$,

1180: there are unique $k,\ell\in\Z/(4g)$ such that

1181: \begin{align*}

1182: s+\frac{\alpha^k-1}{\alpha-1}&=s'+\frac{\alpha^{\ell+1}-1}{\alpha-1} \\

1183: s+\frac{\alpha^{k+1}-1}{\alpha-1}&=s'+\frac{\alpha^\ell-1}{\alpha-1}.

1184: \end{align*}

1185: \end{itemize}

1186: \end{proof}

1187: 

1188: \begin{theorem}

1189: Let $n \ge 128 g^4$, and let

1190: \[

1191: \rho:H_1(S_g')\ \to\ \Z/n

1192: \]

1193: be a randomly chosen homomorphism, and let $\hS_g$ be the finite cover of

1194: $S_g'$ corresponding to $\rho$. Then with probability more than $\frac12$, the

1195: cellulation of\/ $\hS_g$ is strongly regular.

1196: \label{th:prob}

1197: \end{theorem}

1198: 

1199: In order to prove \thm{th:prob}, we need to more explicitly describe the

1200: condition of strong regularity as it applies to $\hS_g$. Let $X$ be a regular

1201: cell complex and suppose that its universal cover $\tilde{X}$ is strongly

1202: regular.  Recall that the \emph{star} $\st(v)$ of a vertex $v$ in $X$ is the

1203: subcomplex generated by the cells that contain $v$. The complex $X$ is strongly

1204: regular if and only if the star of each vertex is.  Suppose that $\tilde{v} \in

1205: \tilde{X}$ projects to $v\in X$. Then the star $\st(\tilde{v})$, which is

1206: strongly regular, projects to the star $\st(v)$.  The latter is strongly

1207: regular if and only if the projection is injective.  In other words, $X$ is

1208: strongly regular if and only if the stars of $\tilde{X}$ embed in $X$. If $X$

1209: is not strongly regular, then there must be a path $\tilde{\ell}$ connecting

1210: distinct vertices of $\st(\tilde{v})$ which projects to a loop $\ell$ in

1211: $\st(v)$.  We say that such a loop \emph{obstructs strong regularity}.  We can

1212: assume that $\tilde{\ell}$ is a pair of segments properly embedded in distinct

1213: cells in $\st(\tilde v)$, with only the end-points of the segments on the

1214: boundary of the cells, which implies that $\ell$ is embedded if $X$ is regular.

1215: \fig{f:obstruct} gives an example of such a loop $\ell$ in a regular

1216: cellulation of a torus.

1217: 

1218: \begin{fullfigure}{f:obstruct}{A loop $\ell$ that obstructs strong regularity

1219:     in the torus $S_1'$}

1220: \pspicture(0,0)(6,4)

1221: \pspolygon[linestyle=none,fillstyle=solid,fillcolor=gray90]

1222:     (0,0)(0,2)(2,2)(2,4)(6,4)(6,0)

1223: \psline(0,0)(6,0) \psline(0,2)(6,2) \psline(2,4)(6,4) \psline(0,0)(0,2)

1224: \psline(2,0)(2,4) \psline(4,0)(4,4) \psline(6,0)(6,4)

1225: \multips(0,0)(2,0){4}{\qdisk(0,0){.1}} \multips(0,2)(2,0){4}{\qdisk(0,0){.1}}

1226: \multips(2,4)(2,0){3}{\qdisk(0,0){.1}}

1227: \rput(1,1){\large $F^1$} \rput(3,1){\large $F^0$} \rput(5,1){\large $F^4$}

1228: \rput(3,3){\large $F^3$} \rput(5,3){\large $F^2$}

1229: \uput[225](0,0){$4$} \uput[270](2,0){$3$} \uput[270](4,0){$2$}

1230: \uput[315](6,0){$1$}

1231: \uput[135](0,2){$2$} \uput[135](2,2){$1$} \uput[135](4,2){$0$}

1232: \uput[45](6,2){$4$}

1233: \uput[135](2,4){$4$} \uput[90](4,4){$3$} \uput[45](6,4){$2$}

1234: \pscurve[linestyle=dashed](2,0)(2.8,.4)(3.6,.8)(3.4,1.6)(4,2)(4.4,3)

1235:     (4.6,3.6)(4,4)

1236: \rput(3.7,.4){\large $\ell$}

1237: \endpspicture

1238: \end{fullfigure}

1239: 

1240: In our case, the surfaces $S_g'$ are regular, but they have many obstructing

1241: loops.  \thm{th:prob} asserts that, with non-zero probability, all such loops

1242: lengthen when lifted to $\hS_g$.

1243: 

1244: \begin{lemma} No loops in $S_g'$ that obstruct strong regularity are

1245: null-homologous.  Furthermore, all obstructing loops represent indivisible

1246: elements in $H_1(S_g')$.

1247: \label{l:notnull} \end{lemma}

1248: 

1249: \begin{proof} In brief, they are indivisible because they are embedded,

1250: and they are too short to be null-homologous.

1251: 

1252: By \lem{l:regular}, $S_g'$ is regular.  By the discussion after the statement

1253: of \thm{th:prob}, each obstructing loop $\ell$ is embedded. If $\ell$ separates

1254: $S_g'$, then it is null-homologous. If $\ell$ does not separate $S_g'$, then it

1255: is indivisible in homology.  (To show this, we can appeal to the classification

1256: of surfaces by cutting $S_g'$ along $\ell$.  The classification implies that

1257: all non-separating positions for $\ell$ are equivalent up to homeomorphism of

1258: $S_g'$.  It is easy to find a standard position for $\ell$ in which it is

1259: indivisible in homology.) Thus it remains to show that no obstructing loop is

1260: null-homologous.

1261: 

1262: First, we claim that any obstructing loop $\ell$ can be supported on fewer than

1263: $4g$ edges of the $1$-skeleton of $S_g'$.  We homotop the two segments of

1264: $\ell$ to the boundaries of the 2-cell containing them, giving them each at

1265: most $2g$ edges.  Thus $\ell$ is represented by a sequence of at most $4g$

1266: edges in $S_g'$. The case of exactly $4g$ edges does not occur, since the

1267: endpoints of the loop coincide, and no two vertices $v\neq v'$ of $S_g'$ are

1268: opposite vertices in two different facets $F^s$.  As in \lem{l:regular}, this

1269: follows from the fact that for $v,v'\in\F_g$, $v\neq v'$, there are unique

1270: $s\in\F_q$ and $k\in\Z/(4g)$ with

1271: \[

1272: v = s+\frac{\alpha^k-1}{\alpha-1}\qquad

1273: v' = s+\frac{\alpha^{k+2g}-1}{\alpha-1}.

1274: \]

1275: 

1276: Second,

1277: we claim that any null-homologous loop in the $1$-skeleton

1278: of $S_g'$ contains at least $4g$ edges.  In other we if $f$ is a $2$-chain on

1279: $S_g'$ and $\d f \ne 0$, then $|\d f| \ge 4g$. Since $S_g'$ is orientable, we

1280: can regard $f$ as a function on its $2$-cells.  Since $f$ is non-constant, it

1281: attains some value $t$ on $k$ $2$-cells with $0 < k < q$.  Since any two

1282: $2$-cells share an edge by \lem{l:regular}, these $2$-cells share

1283: \[

1284: k(q-k) \ge 4g

1285: \]

1286: edges with the complementary set of $2$-cells, of which there are $q-k$.  Since

1287: $\d f$ is non-zero on these edges, $|\d f| \ge 4g$, as desired.

1288: \end{proof}

1289: 

1290: \begin{proof}[Proof of \thm{th:prob}] In brief, $S_g'$ has fewer than $64g^4$

1291: obstructing loops $\ell$.  For each one,

1292: \[

1293: P\bigl[\rho([\ell]) = 0\bigr] = \frac1n.

1294: \]

1295: The expected number of obstructing loops that lift from $S_g'$ to $\hS_g$

1296: without lengthening is less than $64g^4/n \le \frac12$.  Thus there is a good

1297: chance that all obstructing loops lengthen.

1298: 

1299: The homology group $H_1(S'_g)$ is a finitely generated free abelian group: It

1300: is isomorphic to $\Z^d$, for $d=1+q(g-1)$. Thus it admits $n^d$ homomorphisms

1301: $\rho$ to $\Z/n$.  Since this is a finite number, choosing one uniformly at

1302: random is well-defined.  If $c$ is any indivisible vector in $\Z^d$, then it is

1303: contained in a basis, and thus $\rho(c)$ is equidistributed. In particular, if

1304: $\ell$ is an obstructing loop, then $[\ell]$ is indivisible by \lem{l:notnull},

1305: so $\rho([\ell])$ is equidistributed in $\Z/n$.

1306: 

1307: It remains only to bound the number of obstructing loops in $S_g'$. A star

1308: $\st(v)$ in $\tilde{S}_g$ has $4g(4g-2)$ points other than $v$ itself.

1309: Without loss of generality, $v$ projects to $0$ in $S_g'$.  In this case the

1310: other vertices are equidistributed among the $4g$ non-zero elements of

1311: $\F_q$.  Therefore $\st(v)$ has

1312: \[

1313: 4g\binom{4g-2}{2}\ =\ \frac{4g(4g-2)(4g-3)}{2}

1314: \]

1315: pairs of arcs connecting $v$ to two vertices that are the same in $S_g'$.

1316: These pairs represent all two-segment obstructing loops that pass through $0$

1317: and a nonzero vertex $v'$ (and some of these loops are homotopic).  If we count

1318: such pairs of arcs for any pair of distinct vertices $v,v'$

1319: of $S_g'$, then we find that the total number is not more than

1320: \[

1321: \binom{4g+1}{2}\binom{4g-2}{2}\ =\

1322: \frac{(4g+1)4g(4g-2)(4g-3)}{4} < 64g^4,

1323: \]

1324: as desired.

1325: \end{proof}

1326: 

1327: \begin{question} For each $g>1$, what is the maximum fatness of a strongly

1328: regular cellulation of a surface of genus $g$? Equivalently, how many edges are

1329: needed for a strongly regular cellulation of a surface of genus $g$?

1330: \end{question}

1331: 

1332: \begin{remark}

1333: One interesting alternative to the construction of $S_g'$ is to assume instead

1334: that $q = 4g-1$ is a prime power, and to let $\alpha$ be an element of order

1335: $4g$ in $\F_{q^2}$.  The resulting $q^2$-fold cover $S_g''$ is almost strongly

1336: regular: the only obstructing loops are those that are null homologous in

1337: $S_g$.  Another interesting surface is the modular curve $X(2p)$, where $p$ is

1338: a prime \cite[\S13]{Silverman:gtm}. The inclusion $\Gamma(2p) \subset

1339: \Gamma(2)$ of modular groups induces a projection from $X(2p)$ to the modular

1340: curve $X(2)$, which is a sphere with three cusp points. If we connect two of

1341: these points by an arc which avoids the third, it lifts to a cellulation of

1342: $X(2p)$ with $f$-vector $(p^2-1,\frac{p(p^2-1)}{2},\frac{p^2-1}{2})$. Like

1343: $S_g''$, it has a few obstructing loops. Unfortunately we do not know a way to

1344: use either $S_g''$ or $X(2p)$ to make fat surfaces of lower genus (or

1345: equivalently fewer cells) than $\hS_g$.

1346: \end{remark}

1347: 

1348: Since \thm{th:prob} provides us with efficient fat surfaces $\hS_g$, the

1349: construction of fat cellulations of $S^3$ only requires efficient cellulations

1350: of the handlebodies $H_1$ and $H_2$ and an efficient way to attach them to

1351: $\hS_g$. In our construction the handlebody cellulations are \emph{a priori}

1352: unrelated to the cellulation of $\hS_g$.  Rather they are transverse after

1353: attachment, and each point of intersection will become a new vertex.  Thus the

1354: question is to position the cellulations to minimize their intersection.

1355: 

1356: We describe the cellulations in three stages:  first, a

1357: dissection of $H_1$ and $H_2$ individually into $3$-cells; second, their

1358: relative position; and third, their position relative to the cellulation of

1359: $\hS_g$.

1360: 

1361: Let $\hg$ be the genus of $\hS_g$.  A handlebody $H$ of genus $\hg$ can

1362: be formed by identifying $\hg$ pairs of disks on the surface of a $3$-cell.

1363: The result is a dissection of $H$ into $\hg$ $2$-cells and one $3$-cell,

1364: although it is not a cell complex because there are no $1$-cells or $0$-cells.

1365: We can still ask whether such a dissection is regular or strongly regular; this

1366: one is neither.  However, if we replace each $2$-cell by 3 parallel $2$-cells,

1367: it becomes strongly regular.  An example of the resulting dissection $A$ is

1368: shown in \fig{f:tripled}.

1369: 

1370: \begin{fullfigure}{f:tripled}

1371:     {A strongly regular dissection $A$ of a handlebody of genus 2}

1372: \pspicture(-4,-2)(4,3)

1373: \psbezier(-2,1.4)(-3,1.4)(-4,0.7)(-4,0)

1374: \psbezier(-2,1.4)(-1,1.4)(-1,0.7)(0,0.7)

1375: \psbezier(0,0.7)(1,0.7)(1,1.4)(2,1.4)

1376: \psbezier(2,1.4)(3,1.4)(4,0.7)(4,0)

1377: \psbezier(-2,-1.4)(-3,-1.4)(-4,-0.7)(-4,0)

1378: \psbezier(-2,-1.4)(-1,-1.4)(-1,-0.7)(0,-0.7)

1379: \psbezier(0,-0.7)(1,-0.7)(1,-1.4)(2,-1.4)

1380: \psbezier(2,-1.4)(3,-1.4)(4,-0.7)(4,0)

1381: \psbezier(1.3,0)(1.8,0.5)(2.3,0.5)(2.8,0)

1382: \psbezier(1.3,0)(1.8,-0.5)(2.3,-0.5)(2.8,0)

1383: \psbezier(-1.3,0)(-1.8,0.5)(-2.3,0.5)(-2.8,0)

1384: \psbezier(-1.3,0)(-1.8,-0.5)(-2.3,-0.5)(-2.8,0)

1385: \psbezier(-2.025,.375)(-1.575,.375)(-1.575,1.4)(-2.025,1.4)

1386: \psbezier[linestyle=dashed](-2.025,.375)(-2.475,.375)(-2.475,1.4)(-2.025,1.4)

1387: \psbezier(-2.425,.3)(-1.975,.3)(-1.975,1.35)(-2.425,1.35)

1388: \psbezier[linestyle=dashed](-2.425,.3)(-2.875,.3)(-2.875,1.35)(-2.425,1.35)

1389: \psbezier(-1.625,.3)(-1.175,.3)(-1.175,1.35)(-1.625,1.35)

1390: \psbezier[linestyle=dashed](-1.625,.3)(-2.075,.3)(-2.075,1.35)(-1.625,1.35)

1391: \psbezier[linestyle=dashed](2.025,.375)(1.575,.375)(1.575,1.4)(2.025,1.4)

1392: \psbezier(2.025,.375)(2.475,.375)(2.475,1.4)(2.025,1.4)

1393: \psbezier[linestyle=dashed](2.425,.3)(1.975,.3)(1.975,1.35)(2.425,1.35)

1394: \psbezier(2.425,.3)(2.875,.3)(2.875,1.35)(2.425,1.35)

1395: \psbezier[linestyle=dashed](1.625,.3)(1.175,.3)(1.175,1.35)(1.625,1.35)

1396: \psbezier(1.625,.3)(2.075,.3)(2.075,1.35)(1.625,1.35)

1397: \endpspicture \end{fullfigure}

1398: 

1399: The surface $S_\hg$ (which for our choice of $\hg$ is isomorphic to

1400: $\hS_g$) has another standard perfect cellulation called a \emph{canonical

1401: schema} in the computer science literature \cite{VY:surfaces}.  Using the

1402: labelling in \fig{f:sg}, we identify $x_{4k}^{}$ with $x_{4k+2}^{-1}$ (the even

1403: loops), and $x_{4k+1}^{}$ with $x_{4k+3}^{-1}$ (the odd loops), for all $0 \le

1404: k \le \hg$. By rounding corners we can make each even loop intersect one odd

1405: loop once and eliminate all other intersections between loops. If we interpret

1406: this pattern of loops as the standard Heegaard diagram for $S^3$

1407: \cite{Stillwell:gtm}, then the even loops bound disks in $H_1$ and the odd

1408: loops bound disks in $H_2$.  We can then put in two copies $A_1$ and $A_2$ of

1409: the cell division $A$ so that its $2$-cells run parallel to these loops.

1410: 

1411: We would like to position $A_1$ and $A_2$ to minimize their intersection with

1412: the cellulation of $\hS_g$. (Note that $A_1$ and $A_2$ do not intersect each

1413: other since $\hS_g \times I$ lies in between.) To this end Vegter and Yap

1414: \cite{VY:surfaces} proved that any cellulation of $S_\hg$ with $n$ edges admits

1415: a position of the canonical schema cellulation with $O(n\hg)$ intersections.

1416: (Strictly speaking the theorem applies to triangulations, but any regular

1417: cellulation of a surface with $n$ edges can be refined to a triangulation with

1418: less than $3n$ edges.) In our case

1419: \[

1420: n = \Theta(\hg) = \Theta(g^6).

1421: \]

1422: Thus, by tripling the edges of the canonical schema in the Vegter-Yap

1423: construction, we can position $A_1$ and $A_2$ so that the lean ends in the

1424: sausage have $O(g^{12})$ vertices.  If we give the fat part of the sausage $N =

1425: g^{12}$ slices, the total $f$-vector of the cellulation of $S^3$ is then

1426: \[

1427: (\Theta(g^{12}),\Theta(g^{13}),\Theta(g^{13}),\Theta(g^{12})),

1428: \]

1429: and its fatness is $\Theta(g)$.  This completes the efficient construction

1430: with unbounded fatness.

1431: 

1432: % \bibliography{me,co,gt,books}

1433: 

1434: \providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}

1435: \begin{thebibliography}{10}

1436: 

1437: \bibitem{Barnette:inequalities}

1438: David Barnette, \emph{Inequalities for $f$-vectors of $4$-polytopes}, Israel J.

1439:   Math. \textbf{11} (1972), 284--291.

1440: 

1441: \bibitem{Bayer:extended}

1442: Margaret~M. Bayer, \emph{The extended $f$-vectors of $4$-polytopes}, J.

1443:   Combinatorial Theory Ser. A \textbf{44} (1987), no.~1, 141--151.

1444: 

1445: \bibitem{BJ:dehn}

1446: Margaret~M. Bayer and Louis~J. Billera, \emph{Generalized {Dehn-Sommerville}

1447:   relations for polytopes, spheres and {Eulerian} partially ordered sets},

1448:   Inventiones Math. \textbf{79} (1985), no.~1, 143--157.

1449: 

1450: \bibitem{BL:sufficiency}

1451: Louis~J. Billera and Carl~W. Lee, \emph{A proof of the sufficiency of

1452:   {M}c{M}ullen's conditions for $f$-vectors of simplicial convex polytopes}, J.

1453:   Combin. Theory Ser. A \textbf{31} (1981), no.~3, 237--255.

1454: 

1455: \bibitem{Bollobas:extremal}

1456: B\'ela Bollob\'as, \emph{Extremal graph theory}, Handbook of Combinatorics

1457:   (R.~Graham, M.~Gr\"otschel, and L.~Lov\'asz, eds.), North-Holland/Elsevier,

1458:   Amsterdam, 1995, pp.~1231--1292.

1459: 

1460: \bibitem{Braden:glued}

1461: Tom Braden, \emph{A glued hypersimplex}, 1997, Personal communication.

1462: 

1463: \bibitem{CS:splag}

1464: John~H. Conway and Neil J.~A. Sloane, \emph{Sphere packings, lattices and

1465:   groups}, 3rd ed., Grundlehren der mathematischen {Wissenschaften}, vol. 290,

1466:   Springer-Verlag, New York, 1993.

1467: 

1468: \bibitem{Coxeter:regpoly}

1469: Harold Scott~MacDonald Coxeter, \emph{Regular polytopes}, 2nd ed., Macmillan,

1470:   New York, 1963, Corrected reprint, Dover, New York 1973.

1471: 

1472: \bibitem{ES:hyperplane}

1473: Herbert Edelsbrunner and Micha Sharir, \emph{A hyperplane incidence problem

1474:   with applications to counting distances}, Applied geometry and discrete

1475:   mathematics: {The Victor Klee Festschrift} (P.~Gritzman and B.~Sturmfels,

1476:   eds.), DIMACS Series in Discrete Mathematics and Theoretical Computer

1477:   Science, vol.~4, Amer. Math. Soc., Providence, RI, 1991, pp.~253--263.

1478: 

1479: \bibitem{Erdos:primzahlen}

1480: Paul Erd\H{o}s, \emph{{\"Uber} die {Primzahlen} gewisser arithmetischer

1481:   {Reihen}}, Math. Z. \textbf{39} (1935), 473--491.

1482: 

1483: \bibitem{Kuperberg:saturated}

1484: G\'abor {Fejes T\'oth}, Greg Kuperberg, and W{\l}odzimierz Kuperberg,

1485:   \emph{Highly saturated packings and reduced coverings}, Monatsh. Math.

1486:   \textbf{125} (1998), no.~2, 127--145.

1487: 

1488: \bibitem{Fulton:gtm}

1489: William Fulton, \emph{Algebraic topology. {A} first course}, Graduate Texts in

1490:   Mathematics, vol. 153, Springer-Verlag, New York, 1995.

1491: 

1492: \bibitem{Grunbaum:convex}

1493: Branko Gr\"unbaum, \emph{Convex polytopes}, Interscience, London, 1967.

1494: 

1495: \bibitem{HZ:census}

1496: Andrea H\"oppner and G\"unter~M. Ziegler, \emph{A census of flag-vectors of

1497:   $4$-polytopes}, Polytopes --- combinatorics and computation (G.~Kalai and

1498:   G.M. Ziegler, eds.), DMV Seminars, vol.~29, Birkh\"auser-Verlag, Basel, 2000,

1499:   pp.~105--110.

1500: 

1501: \bibitem{Iverson:hyperbolic}

1502: Birger Iversen, \emph{Hyperbolic geometry}, London Math. Soc. Student Texts,

1503:   vol.~25, Cambridge University Press, Cambridge, 1992.

1504: 

1505: \bibitem{JZ:cubical}

1506: Michael Joswig and G\"unter~M. Ziegler, \emph{Neighborly cubical polytopes},

1507:   Discrete Comput. Geom. \textbf{24} (2000), no.~2-3, 325--344,

1508:   \mbox{arXiv:math.CO/9812033}.

1509: 

1510: \bibitem{KST:problem}

1511: Tam\'as K\H{o}v\'ari, Vera~T. S\'os, and P\'al Tur\'an, \emph{On a problem of

1512:   {K. Zarankiewicz}}, Colloq. Math. \textbf{3} (1954), 50--57.

1513: 

1514: \bibitem{Kuperberg:kissing}

1515: Greg Kuperberg and Oded Schramm, \emph{Average kissing numbers for

1516:   non-congruent sphere packings}, Math. Res. Lett. \textbf{1} (1994), no.~3,

1517:   339--344, \mbox{arXiv:math.MG/9405218}.

1518: 

1519: \bibitem{Malcev:faithful}

1520: A.~I. Mal'cev, \emph{On the faithful representation of infinite groups by

1521:   matrices}, Amer. Math. Soc. Transl. (2) \textbf{45} (1965), 1--18.

1522: 

1523: \bibitem{Malkevitch:tiling}

1524: Joseph Malkevitch, \emph{Tiling convex polygons with equilateral triangles and

1525:   squares}, Discrete Geometry and Convexity (J.~E. Goodman, E.~Lutwak,

1526:   J.~Malkevitch, and R.~Pollack, eds.), Ann. New York Acad. Sci., vol. 440, New

1527:   York Academy of Sciences, 1985, pp.~299--303.

1528: 

1529: \bibitem{McMullen:simplicial}

1530: Peter McMullen, \emph{The numbers of faces of simplicial polytopes}, Israel J.

1531:   Math. \textbf{9} (1971), 559--570.

1532: 

1533: \bibitem{Munkres:algtop}

1534: James~R. Munkres, \emph{Elements of algebraic topology}, Addison-Wesley, Menlo

1535:   Park, CA, 1984.

1536: 

1537: \bibitem{PA:combin}

1538: J\'anos Pach and Pankaj~K. Agarwal, \emph{Combinatorial geometry}, J. Wiley and

1539:   Sons, New York, 1995.

1540: 

1541: \bibitem{Schulte:steinitz}

1542: Egon Schulte, \emph{Analogues of {Steinitz's} theorem about

1543:   non-in\-scrib\-a\-ble polytopes}, Intuitive geometry (Si\'ofok 1985)

1544:   (Amsterdam), Colloquia Soc. J\'anos Bolyai, vol.~48, North Holland, 1987,

1545:   pp.~503--516.

1546: 

1547: \bibitem{Silverman:gtm}

1548: Joseph~H. Silverman, \emph{The arithmetic of elliptic curves}, Graduate Texts

1549:   in Mathematics, vol. 106, Springer-Verlag, New York, 1986.

1550: 

1551: \bibitem{Stanley:simplicial}

1552: Richard~P. Stanley, \emph{The number of faces of a simplicial convex polytope},

1553:   Adv. in Math. \textbf{35} (1980), no.~3, 236--238.

1554: 

1555: \bibitem{Stanley:eulerian}

1556: \bysame, \emph{A survey of {Eulerian} posets}, Polytopes: abstract, convex and

1557:   computational (Scarborough, ON, 1993), NATO Adv. Sci. Inst. Ser. C Math.

1558:   Phys. Sci., vol. 440, Kluwer Acad. Publ., Dordrecht, 1994, pp.~301--333.

1559: 

1560: \bibitem{Steinitz:eulerschen}

1561: Ernst Steinitz, \emph{{\"Uber} die {Eulerschen Polyederrelationen}}, Archiv

1562:   f\"ur Mathematik und Physik \textbf{11} (1906), 86--88.

1563: 

1564: \bibitem{Stillwell:gtm}

1565: John Stillwell, \emph{Classical topology and combinatorial group theory}, 2nd

1566:   ed., Graduate Texts in Mathematics, no.~72, Springer-Verlag, 1993.

1567: 

1568: \bibitem{Thurston:geometry}

1569: William~P. Thurston, \emph{Three-dimensional geometry and topology}, vol.~I,

1570:   Princeton University Press, Princeton, NJ, 1997.

1571: 

1572: \bibitem{VY:surfaces}

1573: Gert Vegter and Chee~K. Yap, \emph{Computational complexity of combinatorial

1574:   surfaces}, Proc. 6th ACM Symp. Comp. Geom., ACM Press, 1990, pp.~102--111.

1575: 

1576: \bibitem{Vinberg:reflection}

1577: Ernest~B. Vinberg, \emph{Hyperbolic reflection groups}, Russian Math. Surveys

1578:   \textbf{40} (1985), 31--75.

1579: 

1580: \bibitem{Waite:combining}

1581: William Waite, \emph{Combining squares and triangles}, Cubism for Fun

1582:   \textbf{53} (2000), 18--22.

1583: 

1584: \bibitem{Walkup:lower}

1585: David~W. Walkup, \emph{The lower bound conjecture for $3$- and $4$-manifolds},

1586:   Acta Math. \textbf{125} (1970), 75--107.

1587: 

1588: \bibitem{Ziegler:gtm}

1589: G\"unter~M. Ziegler, \emph{Lectures on polytopes}, Graduate Texts in

1590:   Mathematics, vol. 152, Springer-Verlag, New York, 1995, Revised edition,

1591:   1998.

1592: 

1593: \end{thebibliography}

1594: 

1595: 

1596: \end{document}

1597: