1: 

2: %preceeded by cat0.tex

3: \documentclass[12pt, reqno]{amsart}

4: \usepackage{graphicx}

5: %\usepackage[arrow,matrix,curve]{xy}

6: 

7: \newtheorem{theorem}{Theorem}[section]

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

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

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

11: \newtheorem{conjecture}[theorem]{Conjecture}

12: 

13: 

14: \theoremstyle{definition} 

15: \newtheorem{definition}[theorem]{Definition}

16: \newtheorem{example}[theorem]{Example}

17: \newtheorem{remark}[theorem]{Remark}

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

19: 

20: 

21: \newcommand{\C}{\mathbb C} \newcommand{\N}{\mathbb N}

22: \newcommand{\R}{\mathbb R} \newcommand{\T}{\mathbb T}

23: \newcommand{\Z}{\mathbb Z}

24: 

25: \DeclareMathOperator{\dist}{{\rm dist}}

26: \DeclareMathOperator{\sys}{{\rm sys}}

27: \DeclareMathOperator{\stsys}{{\rm stsys}}

28: \DeclareMathOperator{\confsys}{{\rm confsys}}

29: \DeclareMathOperator{\pisys}{\sys\!\pi} 

30: 

31: \def\phisys {\phi{\rm sys}}

32: \DeclareMathOperator{\vol}{{\rm vol}}

33: 

34: \DeclareMathOperator{\area}{{\rm area}}

35: \DeclareMathOperator{\length}{{\rm length}}

36: 

37: \DeclareMathOperator{\Aut}{{\rm Aut}}

38: 

39: \DeclareMathOperator{\tauzero}{{\tau_0}}

40: \DeclareMathOperator{\FillVol}{{\rm FillVol}}

41: \DeclareMathOperator{\gmetric}{{\mathcal G}}

42: \DeclareMathOperator{\Bolza}{{\mathcal B}}

43: 

44: \DeclareMathOperator{\CAT}{{\rm CAT}}

45: 

46: \def\SR{{\rm SR}}

47: 

48: 

49: 

50: \def\go {{\gmetric_{\mathcal O}}} \def\genus {{\it g}}

51: 

52: \def\ie {{\it i.e.\ }} 

53: \def\eg {{\it e.g.\ }} 

54: \def\cf {\hbox{\it cf.\ }}

55: \def\rat{{\angle\!^\mid \! \left(\frac{x}{2}, \frac{\pi}{8}\right)\ }}

56: 

57: \long\def\forget#1\forgotten{} %

58: 

59: \newcommand{\rp}{\mathbb R\mathbb P} 

60: 

61: \numberwithin{equation}{section} 

62: \numberwithin{figure}{section}

63: 

64: \begin{document}

65: 

66: 

67: \author[M.~Katz]{Mikhail G. Katz$^{*}$} \address{Department of

68: Mathematics, Bar Ilan University, Ramat Gan 52900 Israel}

69: \email{katzmik@math.biu.ac.il} \thanks{$^{*}$Supported by the Israel

70: Science Foundation (grants no.\ 620/00-10.0 and 84/03)}

71: 

72: \author[S.~Sabourau]{St\'ephane Sabourau} \address{Laboratoire de

73: Math\'ematiques et Physique Th\'eorique, Universit\'e de Tours, Parc

74: de Grandmont, 37400 Tours, France}

75: 

76: \address{Department of Mathematics, University of Pennsylvania,

77: 209 South 33rd Street, Philadelphia, PA 19104-6395, USA}

78: \email{sabourau@lmpt.math.univ-tours.fr}

79: 

80: \subjclass%[2000] 

81: {Primary 53C20, 53C23 %%Global topological methods (\`a la Gromov) 

82: }

83: 

84: \title[An optimal inequality for CAT(0) metrics in genus two]{An

85: optimal systolic inequality for CAT(0) metrics in genus two}

86: 

87: \keywords{Bolza surface, CAT(0) space, hyperelliptic surface, Voronoi

88: cell, Weierstrass point, systole}

89: 

90: \begin{abstract}

91: We prove an optimal systolic inequality for CAT(0) metrics on a

92: genus~2 surface.  We use a Voronoi cell technique, introduced by

93: C.~Bavard in the hyperbolic context.  The equality is saturated by a

94: flat singular metric in the conformal class defined by the smooth

95: completion of the curve $y^2=x^5-x$.  Thus, among all CAT(0) metrics,

96: the one with the best systolic ratio is composed of six flat regular

97: octagons centered at the Weierstrass points of the Bolza surface.

98: \end{abstract}

99: 

100: \maketitle

101: 

102: \tableofcontents

103: 

104: 

105: 

106: 

107: 

108: 

109: 

110: 

111: 

112: 

113: 

114: 

115: 

116: \section{Hyperelliptic surfaces of nonpositive curvature}

117: \label{Jone1}

118: 

119: 

120: Over half a century ago, a student of C. Loewner's named P. Pu

121: presented, in the pages of the Pacific Journal of Mathematics

122: \cite{Pu}, the first two optimal systolic inequalities, which came to

123: be known as the Loewner inequality for the torus, and Pu's

124: inequality~\eqref{J62} for the real projective plane.

125: 

126: 

127: The recent months have seen the discovery of a number of new systolic

128: inequalities \cite{Am, BK1, Sa1, BK2, IK, BCIK1, BCIK2, KL, Ka4, KS1,

129: KRS}, as well as near-optimal asymptotic bounds \cite{Ka3, KS2, Sa2,

130: KSV, Sa3, RS}.  A number of questions posed in \cite{CK} have thus

131: been answered.  A general framework for systolic geometry in a

132: topological context is proposed in~\cite{KR, KR2}.

133: 

134: The homotopy 1-systole, denoted $\pisys_1(X)$, of a compact metric

135: space $X$ is the least length of a noncontractible loop of $X$.  

136: 

137: 

138: 

139: 

140: 

141: 

142: Given a metric~$\gmetric$ on a surface, let~$\SR(\gmetric)$ denote its

143: systolic ratio

144: \[

145: \SR(\gmetric) =\frac{\pisys_1(\gmetric)^2}{\area(\gmetric)}.

146: \]

147: Given a compact Riemann surface~$\Sigma$, its {\em optimal systolic

148: ratio\/} is defined as~$\SR(\Sigma)= \sup_{\gmetric} \SR(\gmetric)$,

149: where the supremum is over all metrics in the conformal type

150: of~$\Sigma$.  Finally, given a smooth compact surface~$M$, its optimal

151: systolic ratio is defined by setting~$\SR(M)= \sup_\Sigma

152: \SR(\Sigma)$, where the supremum is over all conformal

153: structures~$\Sigma$ on~$M$.  The latter ratio is known for the Klein

154: bottle, \cf \eqref{Jcb}, in addition to the torus and real projective

155: plane already mentioned.

156: 

157: 

158: In the class of all metrics without any curvature restrictions, no

159: singular flat metric on a surface of genus~2 can give the optimal

160: systolic ratio in this genus \cite{Sa1}.  The best available upper

161: bound for the systolic ratio of an arbitrary genus two surface

162: is~$\gamma_2 \simeq 1.1547$, \cf \cite{KS1}.

163: 

164: The precise value of~$\SR$ for the genus~2 surface has so far eluded

165: researchers, \cf~\cite{Ca, Bry}.  We propose an answer in the

166: framework of negatively curved, or more generally,~$\CAT(0)$ metrics.

167: 

168: 

169: The term ``$\CAT(0)$ space'' evokes an extension of the notion of a

170: manifold of nonpositive curvature, to encompass singular spaces.  We

171: will use the term to refer to surfaces with metrics with only mild

172: quotient singularities, defined below.  Here the condition of

173: nonpositive curvature translates into a lower bound of~$2\pi$ for the

174: total angle at the singularity.  We need such an extension so as to

175: encompass the metric which saturates our optimal

176: inequality~\eqref{J11}.

177: 

178: A mild quotient singularity is defined as follows.  Consider a smooth

179: metric on~$\R^2$.  Let~$q\geq 1$ be an integer.  Consider the~$q$-fold

180: cover~$X_q$ of~$\R^2\setminus \{0\}$ with the induced metric.  We

181: compactify~$X_q$ in the neighborhood of the origin to obtain a

182: complete metric space~$X_q^c = X_q \cup \{0\}$.  

183: 

184: \begin{definition}

185: \label{911}

186: Suppose~$X_q^c$ admits an isometric action of~$\Z_p$ fixing the

187: origin.  Then we can form the orbit space~$Y_{p,q} = X_q^c/\Z_p$.  The

188: space~$Y_{p,q}$ is then called mildly singular at the origin.  

189: \end{definition}

190: 

191: The total angle at the singularity is then $\tfrac {2\pi q}{p}$, and

192: the CAT(0) condition is $\tfrac q p \geq 1$.

193: 

194: \begin{remark}

195: \label{tro}

196: Alternatively, a point is singular of total angle $2\pi(1+\beta)$ if

197: the metric is of the form $e^{h(z)} |z|^{2\beta} |dz|^2$ in its

198: neighborhood, where $|dz|^2=dx^2 + dy^2$, \cf

199: M.~Troyanov~\cite[p.~915]{Tro}.

200: \end{remark}

201: 

202: \begin{theorem} 

203: \label{Jtheo:nonpositive}

204: Every~$\CAT(0)$ metric~$\gmetric$ on a surface~$\Sigma_2$ of genus

205: two, satisfies the following optimal inequality:

206: \begin{equation}

207: \label{J11}

208: \SR(\Sigma_2,\gmetric) \leq \tfrac{1}{3} \cot\tfrac{\pi}{8} =

209: \tfrac{1}{3}\left( \sqrt{2} + 1 \right) = 0.8047\ldots

210: \end{equation}

211: The inequality is saturated by a singular flat metric, with~$16$

212: conical singularities, in the conformal class of the Bolza

213: surface.

214: \end{theorem}

215: 

216: The Bolza surface is described in Section~\ref{J22}.  The optimal

217: metric is described in more detail in Section~\ref{Jgenus2}.

218: Theorem~\ref {Jtheo:nonpositive} is proved in Section~\ref{Jproof12}

219: based on the octahedral triangulation of~$S^2$.

220: 

221: \begin{remark}

222: A similar optimal inequality can be proved for hyperelliptic surfaces

223: of genus~5 based on the icosahedral triangulation, \cf~\cite{Bav1}.

224: \end{remark}

225: 

226: 

227: 

228: 

229: 

230: 

231: \section{Distinguishing $16$ points on the Bolza surface}

232: \label{J22}

233: 

234: The Bolza surface~$\Bolza$ is the smooth completion of the affine

235: algebraic curve

236: \begin{equation}

237: \label{J71}

238: y^2= x^5-x.

239: \end{equation}

240: It is the unique Riemann surface of genus two with a group of

241: holomorphic automorphisms of order 48.  (A way of passing from an

242: affine hyperelliptic surface to its smooth completion is described in

243: \cite[p.~60-61]{Mi}.)

244: 

245: \begin{definition}

246: A conformal involution~$J$ of a compact Riemann surface~$\Sigma$ of

247: genus $\genus$ is called {\em hyperelliptic\/} if~$J$ has

248: precisely~$2\genus + 2$ fixed points.  The fixed points of~$J$ are

249: called the Weierstrass points of~$\Sigma$.

250: \end{definition}

251: 

252: The quotient Riemann surface~$\Sigma/J$ is then necessarily the

253: Riemann sphere, denoted henceforth~$S^2$.  Let~$Q: \Sigma \to S^2$ be

254: the conformal ramified double cover, with~$2\genus+2$ branch points.

255: Thus,~$J$ acts on~$\Sigma$ by sheet interchange.  Recall that every

256: surface of genus~2 is hyperelliptic, \ie admits a hyperelliptic

257: involution \cite[Proposition III.7.2]{FK}.

258: 

259: 

260: We make note of $16$ special points on~$\Bolza$.  We refer to a point

261: as being special if it is a fixed point of an order 3 automorphism

262: of~$\Bolza$.

263: 

264: Consider the regular octahedral triangulation of~$S^2=\C\cup \infty$.

265: Its set of vertices is conformal to the set of roots of the polynomial

266: $x^5-x$ of formula~\eqref{J71}, together with the unique point at

267: infinity.  Thus the six points in question can be thought of as the

268: ramification points of the ramified conformal double cover~$Q:{\Bolza}

269: \to S^2$, while the $16$ special points of~$\Bolza$ project to the 8

270: vertices of the cubical subdivision dual to the octahedral

271: triangulation.

272: 

273: In other words, the~$x$-coordinates of the ramification points are

274: \[

275: \left\{ 0, \infty, 1, -1, i, -i \right\},

276: \]

277: which stereographically correspond to the vertices of a regular

278: inscribed octahedron.  The conformal type therefore admits the

279: symmetries of the cube.  If one includes both the hyperelliptic

280: involution, and the real (antiholomorphic) involution of~$\Bolza$

281: corresponding to the complex conjugation~$(x,y)\to (\bar{x}, \bar{y})$

282: of~$\C ^2$, one obtains the full symmetry group~$\Aut(\Bolza)$, of

283: order

284: \begin{equation}

285: \label{J51}

286: |\Aut(\Bolza)| = 96,

287: \end{equation}

288: see \cite[p.~404]{KuN} for more details.

289: 

290: \begin{lemma}

291: \label{J21}

292: The hyperbolic metric of~$\Bolza$ admits~$12$~systolic loops.

293: The~$12$ loops are in one to one correspondence with the edges of the

294: octahedral decomposition of~$S^{2}$.  The correspondence is given by

295: taking the inverse image under~$Q$ of an edge.  The~$12$~systolic

296: loops cut the surface up into~$16$~hyperbolic triangles.  The centers

297: of the triangles are the~$16$~special points.

298: \end{lemma}

299: 

300: See \cite[\S~5]{Sc1} for further details.  The Bolza surface is

301: extremal for two distinct problems:

302: \begin{itemize}

303: \item

304: systole of hyperbolic surfaces \cite{Bav2}, \cite[Theorem 5.2]{Sc1};

305: \item

306: conformal systole of Riemann surfaces \cite{BS}.

307: \end{itemize}

308: The square of the conformal systole of a Riemann surface is also known

309: as its Seshadri constant \cite{Kon}.  The Bolza surface is also

310: conjectured to be extremal for the first eigenvalue of the Laplacian.

311: Such extremality has been verified numerically \cite{JP}.  The above

312: evidence suggests that the systolically extremal surface may lie in

313: the conformal class of~$\Bolza$, as well.  Meanwhile, we have the

314: following result, proved in Section~\ref{Jfive}.

315: 

316: \begin{theorem}

317: \label{J13}

318: The Bolza surface~$\Bolza$ satisfies~$\SR(\Bolza) \leq \frac{\pi}{3}$.

319: \end{theorem}

320: Note that Theorems~\ref{J13} and \ref{Jtheo:nonpositive} imply that

321: $\SR(\Bolza) \in [0.8, 1.05]$.

322: 

323: \section{A flat singular metric in genus~two}

324: \label{Jgenus2}

325: 

326: 

327: The optimal systolic ratio of a genus~2 surface~$(\Sigma_2, \gmetric)$

328: is unknown, but it satisfies the Loewner inequality, \cf \cite{KS1}.

329: Here we discuss a {\em lower\/} bound for the optimal systolic ratio

330: in genus~2, briefly described in~\cite{CK}.

331: 

332: 

333: The example of M. Berger (see \cite [Example~5.6.B$'$]{Gr1}) in

334: genus~2 is a singular flat metric with conical singularities.  It has

335: systolic ratio~$\SR=0.6666$.  This ratio was improved by F. Jenni

336: \cite{Je}, who identified the hyperbolic genus~2 surface with the

337: optimal systolic ratio among all hyperbolic genus~2 surfaces (see also

338: C.~Bavard~\cite{Bav2} and P.~Schmutz \cite[Theorem~5.2]{Sc1}).  The

339: surface in question is a triangle surface~(2,3,8).  Its conformal

340: class is that of the Bolza surface, \cf Section~\ref{J22}.  It admits

341: a regular hyperbolic octagon as a fundamental domain, and has

342: 12~systolic loops of length~$2x$, where~$x=\cosh^{-1} (1+\sqrt{2})$.

343: It has

344: \[

345: \pisys_1= 2\log\left( 1+\sqrt{2}+ \sqrt{2+2\sqrt{2}} \right),

346: \]

347: area~${4\pi}$, and systolic ratio~$\SR\simeq 0.7437$.  This ratio can

348: be improved as follows (see the table of Figure~\ref{Jfig2}).

349: 

350: \begin{figure}

351: \renewcommand{\arraystretch}{1.8} 

352: \[

353: \begin{tabular}[t]{|

354: @{\hspace{3pt}}p{1.5in}|| @{\hspace{3pt}}p{.8in}|

355: @{\hspace{3pt}}p{.8in}| @{\hspace{3pt}}p{.8in}| } \hline & Berger &

356: Jenni & metric $\go$ on Bolza \\ \hline\hline $\displaystyle

357: \SR(\gmetric) =\frac{\pisys_1(\gmetric)^2}{\area(\gmetric)}$ & 0.6666

358: & 0.7437 & 0.8047 \\ \hline

359: \end{tabular}

360: \]

361: \renewcommand{\arraystretch}{1}

362: \caption{\textsf{Three systolic ratios in genus~2}}

363: \label{Jfig2}

364: \end{figure}

365: 

366: 

367: \begin{proposition}

368: The conformal class of the Bolza surface~$\Bolza$ admits a metric,

369: denoted~$\go$, with the following properties:

370: \begin{enumerate}

371: \item

372: the metric is singular flat, with conical singularities precisely at

373: the $16$ special points of Section~$\ref{J22}$;

374: \item

375: each singularity is of total angle~$\frac{9\pi}{4}$, so that the

376: metric~$\go$ is $\CAT(0)$;

377: \item

378: the metric is glued from six flat regular octagons, centered on the

379: Weierstrass points, while the 1-skeleton projects under~$Q:{\Bolza}

380: \to S^2$ to that of the dual cube in~$S^2$;

381: \item

382: the systolic ratio equals~$\SR(\go)= \tfrac{1}{3}\left( \sqrt{2} + 1

383: \right) > \frac{4}{5}$.

384: \end{enumerate}

385: \end{proposition}

386: 

387: 

388: 

389: 

390: 

391: \begin{proof} 

392: The octahedral triangulation of the sphere, \cf Section~\ref{J22},

393: lifts to a triangulation of~$\Bolza$ consisting of $16$ triangles,

394: which we think of as being ``equilateral''.  Here 8 equilateral

395: triangles are connected cyclically around each of the 6 Weierstrass

396: vertices of the triangulation of~$\Bolza$.

397: 

398: We further subdivide each equilateral triangle into 3 isosceles

399: triangles, with a common vertex at the center of the equilateral

400: triangle.  We equip each of the 48 isosceles triangles with a flat

401: metric with obtuse angle $\tfrac{3\pi}{4}$.

402: 

403: 

404: 

405: Each of the 6 Weierstrass vertices of the original triangulation is a

406: smooth point, since the total angle is ~$8\left(\tfrac{\pi}{4} \right)

407: =2\pi$.  Each equilateral triangle possesses a singularity at the

408: center with total angle~$\tfrac{9\pi}{4}=2\pi\left( 1+\tfrac{1}{8}

409: \right)> 2\pi$.  Alternatively, we can apply the Gauss-Bonnet formula

410: $\sum_{\sigma} \alpha (\sigma) = 2\genus-2$, in genus 2 with $16$

411: isometric singularities.  Here the sum is over all

412: singularities~$\sigma$ of a singular flat metric on a surface of

413: genus~$\genus$, where the cone angle at singularity~$\sigma$

414: is~$2\pi(1+\alpha(\sigma))$.  Since the metric~$\go$ is smooth at a

415: Weierstrass point of~$\Bolza$, the metric has

416: only $16$ singularities, precisely at the special points of

417: Section~\ref{J22}, proving items 1 and 2 of the proposition.

418: 

419: Let $x$ denote the sidelength of the equilateral triangle.  The

420: barycentric subdivision of each equilateral triangle consists of 6

421: copies of a flat right angle triangle, denoted $\rat$, with

422: side~$\frac{x}{2}$ and adjacent angle~$\frac{\pi}{8}$.  We thus obtain

423: a decomposition of the metric~$\go$ into 96 copies of the

424: triangle~$\rat$, which can be thought of as a fundamental domain for

425: the action of~$\Aut(\Bolza)$, \cf \eqref{J51}.

426: 

427: 

428: 

429: 

430: 

431: 

432: %The systolic ratio is therefore calculated as follows:

433: %\[

434: %\begin{array}{rcl}

435: %\label{J37}

436: %\displaystyle \frac{\left( \pisys_1(\go)\right)^2} {\area (\go)} & = &

437: %\displaystyle \frac{(2x)^2}{96\;\area\left(\rat \right)}

438: %_{\phantom{a\choose b}} = \displaystyle \frac {x^2}

439: %{24\left(\frac{1}{2} \left( \frac{x} {2} \right)^2 \tan\frac

440: %{\pi}{8}\right)}_{\phantom{a\choose b}} \\ & = & \displaystyle

441: %\left(3\sqrt{3-2\sqrt{2}}\right)^{-1^{\phantom{a}}} \simeq 0.8047,

442: %\end{array}

443: %\]

444: 

445: We have $\pisys_1(\go)= 2x$ by Lemma~\ref{J31}, proving item 4 of the

446: proposition.  To prove item 3, note that the union of the $16$

447: triangles~$\rat$ with a common Weierstrass vertex is a flat regular

448: octagon.  The latter is represented in Figure~\ref{Jfigure1}, together

449: with the systolic loops passing through~it.

450: \end{proof}

451: 

452: \begin{figure}

453: \setlength\unitlength{1pt}

454: \noindent \begin{picture}

455: (400,150)(0,0)

456: \put(120,15){\includegraphics[angle=0,height=120pt]{octagon.eps}}

457: %\put(150,5){{\scshape Figures 9-10}}

458: \end{picture}

459: \caption{\textsf{Flat regular octagon obtained as the union of $16$

460: triangles $\rat$.  The shaded interior octagon represents the region

461: with 4 geodesic loops through every point}}

462: \label{Jfigure1}

463: \end{figure}

464: 

465: 

466: \begin{lemma}

467: \label{J31}

468: The systole of the singular flat $\CAT(0)$ metric on the Bolza

469: surface equals twice the distance between a pair

470: of adjacent Weierstrass points.

471: \end{lemma}

472: 

473: \begin{proof}

474: Consider the smooth closed geodesic $\gamma \subset \Bolza$ which is

475: the inverse image under the map $Q: \Bolza \to S^2$ of an edge of the

476: octahedron, \cf Lemma~\ref{J21}.  Let~$x$ be the distance between a

477: pair of opposite sides of the regular flat octagon in $\Bolza$, or

478: equivalently, the distance between a pair of adjacent Weierstrass

479: points.  Thus, $\length(\gamma)=2x$.  Consider a loop~$\alpha \subset

480: \Bolza$ whose length satisfies

481: \begin{equation}

482: \label{941}

483: \length(\alpha) < 2x.

484: \end{equation}

485: We will prove that there are two possibilities for $\alpha$: it is

486: either contractible, or freely homotopic to one of the 12 geodesics

487: $\gamma$ of the type described above.  On the other hand, $\gamma$ is

488: necessarily length minimizing in its free homotopy class, by the

489: $\CAT(0)$ property of the metric \cite[Theorem~6.8]{BH}.  This will

490: rule out the second possibility, and prove the lemma.

491: 

492: Denote by~$\Bolza^{(1)}$ the graph on~$\Bolza$ given by the inverse

493: image under~$Q$ of the 1-skeleton of the cubical subdivision of~$S^2$.

494: The graph~$\Bolza^{(1)}$ partitions the surface into six regular

495: octagons, denoted $\Omega_k$:

496: \begin{equation}

497: \label{942}

498: \Bolza= \cup_{k=1}^6 \Omega_k.

499: \end{equation}

500: We will deform $\alpha$ to a loop~$\beta \subset \Bolza^{(1)}$, as

501: follows.  The partition \eqref{942} induces a partition of the

502: loop~$\alpha$ into arcs~$\alpha_i$, each lying in its respective

503: octagon $\Omega_{k_i}$.  We deform each~$\alpha_i$, without increasing

504: length, to the straight line segment $[p_i, q_i]\subset \Omega_{k_i}$.

505: The boundary points $p_i, q_i$ of~$\alpha_i$ split the boundary closed

506: curve $\partial \Omega_{k_i} \subset \Bolza^{(1)}$ into a pair of

507: paths.  Let~$\beta_i \subset \partial \Omega_{k_i}$ be the shorter of

508: the two paths.  Denote by~$y$ the distance between adjacent vertices

509: of the octagon.  Then clearly

510: \begin{equation}

511: \label{J32}

512: \length (\beta_i) \leq \frac{4y}{x} \length(\alpha_i).

513: \end{equation}

514: We first deform the loop $\alpha$ into the graph~$\Bolza^{(1)}$.  The

515: deformation fixes the intersection points~$\alpha \cap \Bolza^{(1)}$.

516: Inside $\Omega_{k_i}$, we deform the arc~$\alpha_i$ to the

517: path~$\beta_i$.

518: The length of the resulting loop is at most

519: \[

520: \frac{4y}{x} \length(\alpha) < 8 y

521: \]

522: by \eqref{941} and \eqref{J32}.  Therefore, its homotopy class

523: in~$\Bolza^{(1)}$ can be represented by an imbedded loop $\beta\subset

524: \Bolza^{(1)}$ of length at most~$8y$.  Thus,~$\beta$ contains fewer

525: than 8 edges of $\Bolza^{(1)}$.  Since the number of edges must be

526: even, its image under~$Q$ must retract to a circuit with at most six

527: edges in the 1-skeleton of the cubical subdivision of~$S^2$.  If the

528: number is 4, then the circuit lies in the boundary of a square face of

529: the cube in $S^2$.  But the boundary of a face does not lift

530: to~$\Bolza$, since it surrounds a single ramification point, namely

531: the center of the square face.

532: 

533: Hence there must be 6 edges in the circuit.  There are two types of

534: circuits with six edges in the 1-skeleton of the cubical subdivision

535: of~$S^2$:

536: \begin{enumerate}

537: \item[(a)]

538: the boundary of the union of a pair of adjacent squares; 

539: \item[(b)] a path consisting of the edges meeting a suitable great

540: circle.

541: \end{enumerate}

542: However, a path of type (b) surrounds an odd number, namely 3, of

543: ramification points, and hence does not lift to the genus 2 surface.

544: Meanwhile, a path of type (a) surrounds two ramification points, and

545: hence does lift to the surface.  Such a path is freely homotopic

546: in~$\Bolza$ to one of the 12 geodesics of type

547: $\gamma$ (\cf Lemma~\ref{J21}), completing the proof.

548: \end{proof}

549: 

550: 

551: 

552: 

553: 

554: 

555: 

556: 

557: 

558: \section{Voronoi cells and Euler characteristic}

559: \label{Jproof12}

560: 

561: The following proposition provides a preliminary lower bound on the

562: area of hyperelliptic surfaces with nonpositive curvature.

563: 

564: \begin{proposition} 

565: \label{Jprop:nonpositive}

566: Every~$J$-invariant~$\CAT(0)$ metric~$\gmetric$ on a closed

567: hyperelliptic surface~$\Sigma_\genus$ of genus~$\genus$, satisfies the

568: bound 

569: \[

570: \SR(\Sigma_\genus, \gmetric) \leq {8} ({(\genus+1)\pi})^{-1} .

571: \]

572: \end{proposition}

573: 

574: \begin{proof}

575: To prove this scale-invariant inequality, we normalize the metric

576: on~$\Sigma=\Sigma_\genus$ to unit systole,

577: \ie~$\pisys_1(\Sigma,\gmetric)=1$.  The preimage by~$Q:\Sigma \to S^2$

578: of an arc of~$S^2$ joining two distinct branch points forms a

579: noncontractible loop on~$\Sigma$.  Therefore, the distance between two

580: Weierstrass points is at least~$\frac{1}{2}

581: \pisys_1(\Sigma,\gmetric) = \frac{1}{2}$. Thus, we obtain $2\genus+2$

582: disjoint disks of radius~$\frac{1}{4}$, centered at the Weierstrass

583: points.  Since the metric is~$\CAT(0)$, the area of each disk is at

584: least~$\frac{\pi}{16}$.  Thus,~$\area(\Sigma,\gmetric) \geq

585: \frac{\genus+1}{8} \pi$.

586: \end{proof}

587: 

588: An {\em optimal\/} lower bound requires a more precise estimate on the

589: area of the Voronoi cells.  The idea is to replace area of balls by

590: area of polygons, where control over the number of sides is provided

591: by the Euler characteristic, \cf \cite{Bav2}.

592: 

593: Denote by~$u: \tilde{\Sigma} \rightarrow \Sigma$ its universal cover.  Let

594: $\{x_i \mid i \in \N \}$ be an enumeration of the lifts of Weierstrass

595: points on~$\tilde{\Sigma}$.  The Voronoi cell~$V_i \subset \tilde \Sigma$

596: centered at~$x_i$ is defined as the set of points closer to~$x_i$ than

597: to any other lift of a Weierstrass point.  In formulas,

598: \begin{equation}

599: \label{J}

600: V_i = \left\{ \left. x \in \tilde{\Sigma} \right| d(x,x_i) \leq d(x,x_j)

601: \text{ for every } j \neq i \right\}.

602: \end{equation}

603: The Voronoi cells on~$\tilde{\Sigma}$ are polygons whose edges are arcs of

604: the equidistant curves between a pair of lifts of Weierstrass

605: points. Note that these edges are not

606: necessarily geodesics.  The Voronoi cells on~$\tilde{\Sigma}$ are

607: topological disks, while their projections~$u(V_i)\subset \Sigma$ may have

608: more complicated topology.  Thus, the surface~$\Sigma$ decomposes

609: into~$2\genus+2$ images of Voronoi cells, centered at the~$2\genus+2$

610: Weierstrass points.  By the number of sides of~$u(V)$ we will mean the

611: number of sides of the polygon~${V}$.

612: 

613: 

614: 

615: \begin{lemma}

616: Let $\gmetric_{1}$ and $\gmetric_{2}$ be two $\CAT(0)$ metrics lying

617: in the same conformal class.  Then, the averaged metric

618: $\gmetric=\tfrac{1}{2} (\gmetric_{1}+\gmetric_{2})$ is $\CAT(0)$, as

619: well.

620: \end{lemma}

621: 

622: \begin{proof}

623: Choose a point~$x \in \Sigma$ and a metric~$\gmetric_{0}$ in its

624: conformal class, which is flat in a neighborhood of~$x$.  Every metric

625: $\gmetric = H \gmetric_{0}$ conformal to it satisfies

626: \[

627: K_{\gmetric} H = K_{\gmetric_{0}} - \tfrac{1}{2} \Delta \log H,

628: \]

629: \cf \cite[p.~265]{GHL}, where $K_{\gmetric}$ and $K_{\gmetric_{0}}$

630: are the Gaussian curvatures of $\gmetric$ and~$\gmetric_{0}$, and

631: $\Delta$ is the Laplacian of~$\gmetric_{0}$ with $\Delta f =

632: \mbox{div} \nabla f$.  Thus, the metrics $\gmetric_{i}$ can be written

633: as $\gmetric_{i} = e^{h_{i}} \gmetric_{0}$ where $h_{i}$ is

634: subharmonic in the neighborhood of~$x$, \ie $\Delta h_{i} \geq 0$.  A

635: simple computation shows that

636: \[

637: \Delta \log H \geq \frac{e^{h_{1}} \Delta h_{1} + e^{h_{2}} \Delta

638: h_{2}}{2H} \geq 0

639: \]

640: where $H = \tfrac{1}{2} (e^{h_{1}} + e^{h_{2}})$, proving the lemma if

641: both points are regular.  For singular points with positive angle

642: excess, the CAT(0) property for the averaged metric is immediate from

643: Remark~\ref{tro}.

644: \end{proof}

645: 

646: \begin{proof}[Proof of Theorem~\ref{Jtheo:nonpositive}]

647: Since averaging by $J$ can only improve the systolic ratio, we may

648: assume without loss of generality that our metric is already

649: $J$-invariant.

650: 

651: \forget

652: Consider the exponential map~$\exp_A$ at a Weierstrass point~$A\in \Sigma$.

653: If~$A$ is a singular point, we lift to the appropriate smooth finite

654: cover, and use its exponential map, \cf Definition~\ref{911}.

655: 

656: Let~$B\in \Sigma$ be another Weierstrass point.  Consider its lift~$B_0 =

657: \exp_A^{-1}(B)$ to the tangent plane~$T_A$.  Consider the equidistant

658: line

659: \[ 

660: L_{O, B_0} \subset T_A

661: \]

662: between the origin~$O\in T_A$ and the point~$B_0$.

663: 

664: By the Toponogov theorem applied to the hinge at~$O$ defined by the

665: point~$B_0$ and an arbitrary point~$X\in L_{O, B_0}$ on the

666: equidistant line, we have

667: \begin{equation}

668: \label{J44b}

669: \dist_\Sigma(A, \exp_A(X)) \leq \dist_\Sigma(B, \exp_A(X)) .

670: \end{equation}

671: \forgotten

672: 

673: 

674: 

675: There exists an extension of the notions of tangent plane and

676: exponential map, to surfaces with singularities.  Namely, let $A\in

677: \Sigma$.  There exists a CAT(0) piecewise flat plane~$T_{A}$ with

678: conical singularities and a covering $\exp_{A}:T_{A} \to \Sigma$ with

679: the following properties:

680: \begin{enumerate}

681: \item 

682: $\exp_A$ sends the origin $O$ of $T_{A}$ to~$A$;

683: 

684: \item 

685: $\exp_A$ takes the conical singularities of~$T_{A}$ to the

686: singularities of~$\Sigma$;

687: 

688: \item 

689: $\exp_A$ sends every pair of geodesic arcs issuing from the origin $O

690: \in T_{A}$ and forming an angle~$\theta$, to a pair of geodesic arcs

691: of the same lengths, and forming the same angle, at their

692: basepoint~$A$.

693: \end{enumerate}

694: 

695: By the Rauch Comparison Theorem, the exponential map $\exp_{A}$ does

696: not decrease distances.

697: 

698: Now assume $A\in \Sigma$ is a Weierstrass point, and let~$B\in \Sigma$

699: be another Weierstrass point.  Fix a lift~$B_0$ of~$B$ to the tangent

700: plane~$T_A$, along a minimizing arc.  Consider the equidistant line

701: \[ 

702: L_{O, B_0} \subset T_A

703: \]

704: between the origin~$O\in T_A$ and the point~$B_0$.  Consider a

705: point~$X_{0}\in L_{O, B_0}$.  Let $X=\exp_{A}(X_{0})$.  Since the

706: exponential map does not decrease distances, we have

707: \begin{equation}

708: \label{J44b}

709: \dist_\Sigma(A, X) = \dist_{T_{A}}(O,X_{0}) = \dist_{T_{A}} (B_{0},

710: X_{0}) \leq \dist_\Sigma(B, X) .

711: \end{equation}

712: 

713: Now consider the polygon in the tangent plane~$T_{A}$, obtained as the

714: intersection of the halfspaces containing the origin, defined by the

715: lines~$L_{O, B_0}$, as~$B$ runs over all Weierstrass points.  It

716: follows from~\eqref{J44b} that the exponential image of this polygon

717: is contained in the Voronoi cell of~$A$.  Since the exponential map

718: does not decrease distances, the area of the polygon is a lower bound

719: for the area of the Voronoi cell.  If~$k$ is the number of sides

720: of~$V$, then~$V$ is partitioned into~$k$ triangles with

721: angle~$\theta_i$ at~$O$, whose area is bounded below by

722: \begin{equation}

723: \label{J43}

724: \left( \tfrac{\pisys_1}{4} \right)^2 \tan\tfrac{\theta_i}{2}

725: \end{equation}

726: since $\dist(A,B) \geq \frac{\pisys_{1}}{2}$.

727: 

728: Consider the graph on~$S^2$ defined by the projections of the Voronoi

729: cells to the sphere.  Thus we have~$f=6$ faces.  Applying the formula

730: $v-e+f=2$ and the well-known fact that~$3v \leq {2} e$, we obtain

731: \[

732: {e} \leq 3f-6 =12.

733: \]

734: Hence the spherical graph has at most~12 edges.  Note that the maximum

735: is attained by the 1-skeleton of the cubical subdivision.

736: 

737: The area of a flat isosceles triangle with third angle~$\theta$, and

738: with unit altitude from the third vertex, is~$\tan \tfrac {\theta}

739: {2}$.  This formula provides a lower bound for the area of the Voronoi

740: cells as in \eqref{J43}.  The proof is completed by Jensen's

741: inequality applied to the convex function~$\tan (\tfrac {x} {2})$

742: when~$0 < x < \pi$.  In the boundary case of equality, we have~$e=12$,

743: all angles~$\theta_i$ as in~\eqref{J43} must be equal, curvature must

744: be zero because of equality in the Rauch Comparison Theorem, and we easily deduce that

745: each Voronoi cell is a regular octagon.  To minimize the area of the

746: octagon, we must choose~$\theta$ as small as possible.  The~$\CAT(0)$

747: hypothesis at the center of the octagon imposes a lower bound~$\theta

748: \geq \tfrac \pi{4}$.  Hence the optimal systolic ratio is achieved for

749: the regular flat octagon with a smooth point at the center.

750: \end{proof}

751: 

752: 

753: \section{Arbitrary metrics on the Bolza surface}

754: \label{Jfive}

755: 

756: The conformal class of the Bolza surface~$\Bolza$, \cf

757: Section~\ref{J22}, is likely to contain a systolically optimal surface

758: in genus~2, as discussed in Section~\ref{Jone1}.

759: 

760: 

761: 

762: 

763: 

764: \begin{theorem} \label{theo:conf}

765: Every metric~$\gmetric$ in the conformal class of the Bolza surface

766: satisfies the bound

767: \begin{equation}

768: \label{J12}

769: \SR(\gmetric) \leq \tfrac{\pi}{3} = 1.0471\ldots

770: \end{equation}

771: \end{theorem}

772: 

773: \begin{remark}

774: In particular, every metric~$\gmetric$ in the conformal class of the

775: Bolza surface satisfies Bavard's inequality

776: \begin{equation}

777: \label{Jcb}

778: \SR(\gmetric) \leq \frac{\pi}{2^{3/2}} \simeq 1.1107

779: \end{equation}

780: for the Klein bottle \cite{Bav1}.  This suggests a possible

781: monotonicity of~$\chi(\Sigma)$ as a function of~$\SR(\Sigma)$.

782: \end{remark}

783: 

784: \begin{lemma}

785: \label{J54}

786: Let~$\gmetric$ be an~$\Aut(\Bolza)$-invariant metric on~$\Bolza$.

787: Consider the displacement~$\delta(J)$ of~$J$ on the

788: 1-skeleton~$\Bolza^{(1)}$ of the Voronoi subdivision of~$\Bolza$,

789: namely

790: \[

791: \delta(J) = \min_{x\in \Bolza^{(1)}} \dist(x, J(x)).

792: \]

793: Then~$\area(\Bolza,\gmetric) \geq 6 \left( \tfrac{2}{\pi} \right)

794: \delta(J)^2$.

795: \end{lemma}

796: 

797: 

798: 

799: \begin{proof}

800: Consider the Voronoi subdivision with respect to the set of six

801: Weierstrass points on~$\Bolza$.  Since each Voronoi

802: cell~$\Omega\subset \Bolza$ is~$J$-invariant, we can identify all

803: pairs of opposite points of the boundary~$\partial \Omega$, to obtain

804: a projective plane

805: \begin{equation}

806: \label{J72} 

807: \rp^2 = \Omega/\sim.

808: \end{equation}

809: We now apply Pu's inequality \cite{Pu} to each of the six Voronoi

810: cells, to obtain

811: \begin{equation}

812: \label{J62}

813: \area(\rp ^2)\geq \tfrac{2}{\pi} \pisys_1(\rp^2)^2 .

814: \end{equation}

815: The lemma now follows from the bound~$\pisys_1(\rp^2) \geq \delta(J)$.

816: \end{proof}

817: 

818: \begin{lemma}

819: \label{J74}

820: Every~$\Aut(\Bolza)$-invariant metric~$\gmetric$ on~$\Bolza$ satisfies

821: the bound $2 \delta(J) \geq \pisys_1(\Bolza, \gmetric)$.

822: \end{lemma}

823: 

824: \begin{proof}

825: Let~$p, J(p)\in \partial \Omega$ satisfy~$\dist(p, J(p)) = \delta(J)$.

826: Let~$\alpha \subset \Omega$ be a minimizing path joining~$p$

827: to~$J(p)$.  Let~$\Omega'\subset \Bolza$ be the adjacent Voronoi cell

828: containing this pair of boundary points, and~$r: \Bolza \to \Bolza$

829: the anticonformal involution which switches~$\Omega$ and~$\Omega'$,

830: and fixes their common boundary.  The loop~$\alpha \cup r(\alpha)$

831: belongs to the free homotopy class of the noncontractible loop~$\gamma

832: \subset \Bolza$ obtained as the inverse image under~$Q: \Bolza \to

833: S^2$ of the edge of the octahedral decomposition of~$S^{2}$ joining

834: the images of the centers of~$\Omega$ and~$\Omega'$, \cf

835: Lemma~\ref{J21}.  Since~$\length(\alpha \cup r(\alpha)) = 2

836: \delta(J)$, the lemma follows.

837: \end{proof}

838: 

839: 

840: \begin{proof}[Proof of Theorem~\ref{theo:conf}]

841: We may assume that the metric on~$\Bolza$ is~$\Aut(\Bolza)$-invariant,

842: since averaging the metric by a finite group of holomorphic and

843: antiholomorphic diffeomorphisms can only improve the systolic ratio.

844: We combine the inequalities of Lemmas~\ref{J74} and \ref{J54} to prove

845: the theorem.

846: \end{proof}

847: 

848: 

849: 

850: %\vfill\eject

851: \begin{thebibliography}{ABCDe}

852: 

853: \bibitem[Am04]{Am} Ammann, B.: Dirac eigenvalue estimates on two-tori.

854: {\em J. Geom. Phys.} \textbf{51} (2004), no. 3, 372--386.

855: 

856: 

857: \bibitem[BCIK05]{BCIK1} Bangert, V; Croke, C.; Ivanov, S.; Katz, M.:

858: Filling area conjecture and ovalless real hyperelliptic surfaces, {\em

859: Geometric and Functional Analysis (GAFA)\/} \textbf{25} (2005), no.~3.

860: 577-597.  See \texttt{arXiv:math.DG/0405583}

861: 

862: \bibitem[BCIK06]{BCIK2} Bangert, V; Croke, C.; Ivanov, S.; Katz, M.:

863: Boundary case of equality in optimal Loewner-type inequalities, {\em

864: Trans. Amer. Math. Soc.} \textbf{358} (2006).  See

865: \texttt{arXiv:math.DG/0406008}

866: 

867: 

868: 

869: \bibitem[BK03]{BK1} Bangert, V.; Katz, M.: Stable systolic

870: inequalities and cohomology products, {\em Comm. Pure Appl. Math.}

871: \textbf{56} (2003), 979--997.  Available at

872: \texttt{arXiv:math.DG/0204181}

873: 

874: \bibitem[BK04]{BK2} Bangert, V; Katz, M.: An optimal Loewner-type

875: systolic inequality and harmonic one-forms of constant norm.  {\em

876: Comm. Anal. Geom.}  \textbf{12} (2004), number 3, 703-732.  See

877: \texttt{arXiv:math.DG/0304494}

878: 

879: \bibitem[Bav86]{Bav1} Bavard, C.: In\'egalit\'e isosystolique pour la

880: bouteille de Klein.  {\em Math. Ann.} \textbf{274} (1986), no. 3,

881: 439--441.

882: 

883: 

884: \bibitem[Bav92]{Bav2} Bavard, C.: La systole des surfaces

885: hyperelliptiques, {\em Prepubl. Ec. Norm. Sup. Lyon\/} \textbf{71}

886: (1992).

887: 

888: 

889: 

890: \bibitem[BH99]{BH} Bridson, M.; Haefliger, A.: Metric spaces of

891: non-positive curvature.  {\em Grundlehren der Mathematischen

892: Wissenschaften}, \textbf{319}.  Springer-Verlag, Berlin, 1999.

893: 

894: 

895: 

896: 

897: \bibitem[Br96]{Bry} Bryant, R.: On extremals with prescribed

898: Lagrangian densities, Manifolds and geometry (Pisa, 1993), 86--111,

899: {\em Sympos. Math.}, \textbf{36}, Cambridge Univ. Press, Cambridge,

900: 1996.

901: 

902: \bibitem[BS94]{BS} Buser, P.; Sarnak, P.: On the period matrix of a

903: Riemann surface of large genus.  With an appendix by J. H. Conway and

904: N. J. A. Sloane.  {\em Invent. Math.} \textbf{117} (1994), no.~1,

905: 27--56.

906: 

907: \bibitem[Ca96]{Ca} Calabi, E.: Extremal isosystolic metrics for

908: compact surfaces, Actes de la table ronde de geometrie differentielle,

909: {\em Sem. Congr.}  \textbf{1} (1996), Soc. Math. France, 167--204.

910: 

911: 

912: \bibitem[CK03]{CK} Croke, C.; Katz, M.: Universal volume bounds in

913: Riemannian mani\-folds, {\it Surveys in Differential Geometry\/}

914: \textbf{8}, Lectures on Geometry and Topology held in honor of Calabi,

915: Lawson, Siu, and Uhlenbeck at Harvard University, May 3- 5, 2002,

916: edited by S.T. Yau (Somerville, MA: International Press, 2003.)

917: pp. 109 - 137.  See \texttt{arXiv:math.DG/0302248}

918: 

919: 

920: \bibitem[FK92]{FK} Farkas, H.; Kra, I.: Riemann surfaces.  Second

921: edition. Graduate Texts in Mathematics, 71. Springer-Verlag, New York,

922: 1992.

923: 

924: \bibitem[GHL90]{GHL} Gallot, S.; Hulin, D.; Lafontaine, J.: Riemannian

925: Geometry. Springer-Verlag, Berlin, 1990.

926: 

927: \bibitem[Gr83]{Gr1} Gromov, M.: Filling Riemannian manifolds, {\em

928: J. Diff. Geom.\/} \textbf{18} (1983), 1-147.

929: 

930: \bibitem[Gr96]{Gr2} Gromov, M.: Systoles and intersystolic

931: inequalities.  Actes de la Table Ronde de G\'{e}om\'{e}trie

932: Diff\'{e}rentielle (Luminy, 1992), 291--362, {\em S\'{e}min. Congr.}

933: \textbf{1}, Soc. Math. France, Paris, 1996.

934: \newline\noindent

935: \texttt{www.emis.de/journals/SC/1996/1/ps/smf\_sem-cong\_1\_291-362.ps.gz}

936: 

937: \bibitem[Gr99]{Gr3} Gromov, M.: Metric structures for Riemannian and

938: non-Riemannian spaces.  {\em Progr. in Mathematics}, \textbf{152},

939: Birkh\"{a}user, Boston, 1999.

940: 

941: \bibitem[IK04]{IK} Ivanov, S.; Katz, M.: Generalized degree and

942: optimal Loewner-type inequalities, {\em Israel J. Math.}  \textbf{141}

943: (2004), 221-233.  \texttt{arXiv:math.DG/0405019}

944: 

945: 

946: \bibitem[JP04]{JP} Jakobson, D.; Levitin, M.; Nadirashvili, N.;

947: Polterovich, I.: Spectral problems with mixed Dirichlet-Neumann

948: boundary conditions: isospectrality and beyond, preprint.  See

949: \texttt{arXiv:math.SP/0409154}

950: 

951: \bibitem[Je84]{Je} Jenni, F.: \"Uber den ersten Eigenwert des

952: Laplace-Operators auf ausgew\"{a}hlten Beispielen kompakter

953: Riemannscher Fl\"{a}chen. [On the first eigenvalue of the Laplace

954: operator on selected examples of compact Riemann surfaces] {\em

955: Comment. Math. Helv.} \textbf{59} (1984), no. 2, 193--203.

956: 

957: 

958: \bibitem[Ka03]{Ka3} Katz, M.: Four-manifold systoles and surjectivity

959: of period map, {\em Comment. Math. Helv.} \textbf{78} (2003), 772-786.

960: \texttt{arXiv:math.DG/0302306}

961: 

962: \bibitem[Ka06]{Ka4} Katz, M.: Systolic inequalities and Massey

963: products in simply-connected manifolds.  See

964: \texttt{arXiv:math.DG/0604012}

965: 

966: 

967: \bibitem[KL05]{KL} Katz, M.; Lescop, C.: Filling area conjecture,

968: optimal systolic inequalities, and the fiber class in abelian covers.

969: Geometry, spectral theory, groups, and dynamics, 181--200, {\em

970: Contemp. Math.} \textbf{387}, Amer. Math. Soc., Providence, RI, 2005.

971: See \texttt{arXiv:math.DG/0412011}

972: 

973: 

974: \bibitem[KR06]{KR} Katz, M.; Rudyak, Y.: Lusternik-Schnirelmann

975: category and systolic category of low dimensional manifolds.  {\em

976: Communications on Pure and Applied Mathematics}, \textbf{59} (2006).

977: See \texttt{arXiv:math.DG/0410456}

978: 

979: \bibitem[KR07]{KR2} Katz, M.; Rudyak, Y.: Bounding volume by systoles

980: of 3-manifolds.  See \texttt{arXiv:math.DG/0504008}

981: 

982: 

983: \bibitem[KRS07]{KRS} Katz, M.; Rudyak, Y.: Sabourau, S.: Systoles of

984: 2-complexes, Reeb graph, and Grushko decomposition.  See

985: \texttt{arXiv:math.DG/0602009}

986: 

987: 

988: \bibitem[KS05]{KS2} Katz, M.; Sabourau, S.: Entropy of systolically

989: extremal surfaces and asymptotic bounds, {\em Ergodic Theory and

990: Dynamical Systems\/} \textbf{25} (2005), no.~4, 1209-1220.  Available

991: at \texttt{arXiv:math.DG/0410312}

992: 

993: \bibitem[KS06]{KS1} Katz, M.; Sabourau, S.: Hyperelliptic surfaces are

994: Loewner, {\em Proc. Amer. Math. Soc.} \textbf{134} (2006), no.~4,

995: 1189-1195.  See \texttt{arXiv:math.DG/0407009}

996: 

997: \bibitem[KSV06]{KSV} Katz, M.; Schaps, M.; Vishne, U.: Logarithmic

998: growth of systole of arithmetic Riemann surfaces along congruence

999: subgroups.  Available at \texttt{arXiv:math.DG/0505007}

1000: 

1001: \bibitem[Kon03]{Kon} Kong, J.: Seshadri constants on Jacobian of

1002: curves.  {\em Trans. Amer. Math. Soc.} \textbf{355} (2003), no. 8,

1003: 3175--3180.

1004: 

1005: \bibitem[KuN95]{KuN} Kuusalo, T.; N\"a\"at\"anen, M.: Geometric

1006: uniformization in genus~$2$.  {\em Ann. Acad. Sci. Fenn. Ser. A Math.}

1007: \textbf{20} (1995), no. 2, 401--418.

1008: 

1009: \bibitem[Mi95]{Mi} Miranda, R.: Algebraic curves and Riemann

1010: surfaces. {\em Graduate Studies in Mathematics}, {\textbf 5}.

1011: Amer. Math. Soc., Providence, RI, 1995.

1012: 

1013: \bibitem[Pu52]{Pu} Pu, P.M.: Some inequalities in certain

1014: nonorientable Riemannian manifolds, {\it Pacific J. Math.\/}

1015: \textbf{2} (1952), 55--71.

1016: 

1017: 

1018: \bibitem[RS07]{RS} Rudyak, Y.: Sabourau, S.: Systolic invariants of

1019: groups and $2$-complexes via Grushko decomposition.

1020: 

1021: \bibitem[Sa04]{Sa1} Sabourau, S.: Systoles des surfaces plates

1022: singuli\`eres de genre deux, {\em Math. Zeitschrift\/} \textbf{247}

1023: (2004), no. 4, 693--709.

1024: 

1025: \bibitem[Sab06]{Sa2} Sabourau, S.: Systolic volume and minimal entropy

1026: of aspherical manifolds, {\em J. Differential Geom.}, to appear.  See

1027: \texttt{arXiv:math.DG/0603695}

1028: 

1029: \bibitem[Sa06b]{Sa3} Sabourau, S.: Asymptotic bounds for separating

1030: systoles on surfaces, preprint.

1031: 

1032: \bibitem[Sc93]{Sc1} Schmutz, P.: Riemann surfaces with shortest

1033: geodesic of maximal length.  {\em Geom. Funct. Anal.} \textbf{3}

1034: (1993), no. 6, 564--631.

1035: 

1036: \bibitem[Tr90]{Tro} Troyanov, M.: Coordonn\'ees polaires sur les

1037: surfaces riemanniennes singuli\`eres.  {\em Ann. Inst. Fourier

1038: (Grenoble)} \textbf{40} (1990), no. 4, 913--937 (1991).

1039: 

1040: \end{thebibliography}

1041: 

1042: 

1043: 

1044: 

1045: \end{document}

1046: 

1047: 

1048: 

1049: