1: % Beisbart, Buchert, Wagner:

2: % Quermass--Vectors in Cosmology

3: %                     Version of 6.7.2000

4: 

5: \documentclass[11pt]{article}  

6: \usepackage{amsmath,amssymb,epsfig,array,calc,multirow,amsfonts}

7: \def\refitem{\par\noindent\hangindent\parindent\hangafter1}

8: \def\p{{\mathbf p}}

9: \def\t{{\mathbf t}}

10: \def\mpc{h^{-1}{\rm Mpc}}

11: \def\x{{\mathbf x}}

12: \def\V{{\mathbf V}}

13: \def\Y{{\mathbf Y}}

14: \def\P{{\mathbb P}}

15: \def\tv{{\mathbf t}}

16: \def\Kc{{\mathbb K}}

17: \def\E{{\mathbb E}}

18: \def\Po{{\mathbb P}}

19: \def\phiv{{\vec{\varphi}}}

20: \def\cv{{\mathbf c}}

21: \def\phiv{{\vec{\varphi}}}

22: \def\CI{{\mathcal I}}

23: \def\K{{\mathcal K}}

24: \def\d{{\rm d}}

25: \def\CD{{\cal D}}

26: \def\dr{{\rm d}}

27: \newcommand{\daverage}[1]{\left\langle #1 \right\rangle_{\rm u}}

28: \hoffset=-1.5cm

29: \voffset=-2.4cm

30: \textwidth=16cm

31: \textheight=23cm

32: \begin{document}

33: \normalsize

34: \newcounter{cureqno}%

35: \newenvironment{mathletters}{%

36:  \refstepcounter{equation}%

37:  \setcounter{cureqno}{\value{equation}}%

38:  \edef\@tempa{\theequation}%

39:  \expandafter\def

40:  \expandafter\theequation

41:  \expandafter{\@tempa\alph{equation}}%

42:  \setcounter{equation}{0}%

43: }{%

44:  \let\theequation\

45:  \setcounter{equation}{\value{cureqno}}%

46: }%

47: \newcommand\eqnum[1]{%

48:  \def\theequation{#1}%

49:  \let\@currentlabel\theequation

50:  \addtocounter{equation}{\m@ne}%

51: }%

52: \centerline{\Large{\bf Morphometry of Spatial Patterns}}

53: 

54: \bigskip\medskip 

55: 

56: 

57: \centerline{Claus Beisbart$^1$, Thomas Buchert$^{2}$ and Herbert Wagner$^1$}

58: \centerline{\bf 7/2000}

59: 

60: \bigskip\bigskip

61: \centerline{$^1$Theoretische Physik, Ludwig--Maximilians--Universit\"at}

62: \smallskip

63: \centerline{Theresienstr. 37, D--80333 M\"unchen, Germany}

64: \smallskip

65: \centerline{$^2$Theoretical Astrophysics Division, National Astronomical Observatory}

66: \smallskip

67: \centerline{ 2--21--1 Osawa Mitaka Tokyo 181--8588, Japan }

68: \smallskip

69: \normalsize

70: 

71: \vspace{5mm}

72: \bigskip

73: \begin{abstract}

74:  

75:   Minkowski functionals constitute a  family of order parameters which

76:   discriminate   spatial  patterns  according   to  size,   shape  and

77:   connectivity. Here  we point out, that these  scalar descriptors can

78:   be complemented  by vector--valued curvature measures  also known as

79:   Querma{\ss}  vectors.    Using  examples  of   galaxy  clusters,  we

80:   demonstrate   that  the   Querma{\ss}  vectors   provide  additional

81:   morphological  information on  directional  features and  symmetries

82:   displayed by spatial data.

83: \end{abstract}

84: 

85: 

86: \centerline{\it Physica A }

87: 

88: \bigskip\medskip

89: 

90: PACS: 02.40.Ft (conv. sets and geometrical inequalities), 02.40.Vh (global analysis and analysis of manifolds), 98.65.Cw  (clusters of galaxies)

91: \bigskip\medskip

92: 

93: Key words: methods: morphology, integral geometry, cluster substructure, large--scale structure

94: 

95: \bigskip\medskip

96: \newpage

97: 

98: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

99: % Introduction

100: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

101: 

102: 

103: \section{Introduction}

104: \label{sec:in}

105: Spatial  patterns  originating  from  the polymorphic  aggregation  of

106: matter  in Nature occur  on vastly  different length--scales  and with

107: unlimited variety.  The pattern shown in Figure \ref{fig:gal} displays

108: an  isodensity contour  of  the smoothed  galaxy  distribution in  our

109: Universe, but it could  also represent a biological structure obtained

110: with  X--ray tomography, for  instance.  \\  The visual  impression of

111: such patterns  may already convey  valuable insights on  the formation

112: processes  or  the functionality  of  those  structures. However,  for

113: unbiased inferences as well  as for comparison with model simulations,

114: for  example,  one  needs   objective  and  quantitative  measures  to

115: characterize the  geometry and topology of  typical structural motifs.

116: \begin{figure}

117: \begin{minipage}[t]{.99\linewidth}

118: \centering

119: \epsfig{file=q1.eps,height=9cm}

120: \end{minipage}\hfill

121: \caption{ An isodensity contour  of the galaxy distribution within the

122: PSCz      survey       (from      the      PSCz       homepage      at

123: http://star-www.dur.ac.uk/cosmology/theory/pscz.html).

124: \label{fig:gal}}

125: \end{figure}

126: \\  In the present  paper we  discuss a  family of  suitable measures,

127: which  include,  besides   the  scalar  Minkowski  functionals,  their

128: vectorial  relatives  --   the  curvature  centroids,  or  Querma{\ss}

129: vectors.  Taken  together, these measures constitute  a complete class

130: of morphometric  order--parameters for  spatial patterns formed  -- or

131: approximated -- by the set  union and intersection of an arbitrary but

132: finite number  of convex ``pixels'' or ``voxels''.   \\ The usefulness

133: of Minkowski functionals has already been demonstrated by the analysis

134: of    galaxy    catalogues    and    simulations    of    large--scale

135: structure~\cite{kerscher:abell,kerscher:fluctuations,sahni:shapefinders,schmalzing:webI},

136: as well  as of the temperature  distribution in maps  of the microwave

137: background  radiation~\cite{schmalzing:cmb}.   Furthermore, they  were

138: used     to     characterize     the    morphology     of     spinodal

139: decomposition~\cite{mecke:spinodal}  and  of  patterns arising  within

140: reaction--diffusion  systems~\cite{mecke:reaction}.   Here,  we  focus

141: attention  on the  less  familiar Querma{\ss}  vectors and  illustrate

142: their  versatility  with the  examples  of  galaxy  clusters and  with

143: patterns generated by lattice automata.

144: 

145: 

146: 

147: 

148: 

149: 

150: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

151: % Minkowski functionals and Querma\ss\;vectors

152: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

153: 

154: 

155: \section{Morphometric order parameters}

156: \label{sec:quer}

157: In this section we recall the definitions of Minkowski functionals and

158: Querma{\ss}  vectors. The  presentation will  be brief  and  informal. 

159: Readers,  who  are interested  in  mathematical  details, may  consult

160: refs.~\cite{hadwiger:vekt,schneider:schwerpunkte,schneider:schwerpunkte2,hadwiger:vect2,mecke:robust}.

161: \\

162: Consider  a   compact  convex  point  set  ({\em   body})

163: $K\subset\E^d$, in Euclidean d--dimensional space with volume

164: \begin{equation}

165: V_0 (K)  = \int_K \dr^d \x > 0

166: \end{equation}

167: and with center of mass at the position:

168: \begin{equation}

169: \p_0 \equiv \V_0/V_0\quad,\quad \V_0 (K) =  \int_K\x \dr^d \x\;\;\;.

170: \end{equation}

171: Now, we  let the body  grow to form  a parallel body  $K_\epsilon$,

172: $\epsilon > 0$, by including  all points $\x$ within a distance

173: $d(K,\x)\leq\epsilon$ to $K$. From Steiner's theorem~\cite{mcmullen:valuations} we know  that the volume

174: $V_0(K_\epsilon)$ of  $K_\epsilon$ is a  finite polynomial in

175: $\epsilon$,

176: \begin{equation}

177: V_0 (K_\epsilon) = \sum_{i=0}^{d} \binom{d}{i} \omega_{i}V_{i}(K) \epsilon^i\;\;\;,

178: \end{equation}

179: where $\omega_i= \pi^{i/2}/\Gamma (1 + i/2)$ denotes the volume of the

180: $i$--dimensional unit ball.  The coefficients $V_i(K)$ define the {\it

181: Minkowski   functionals}  of   the   body  $K$.    Since  the   volume

182: $V_0(K_\epsilon)$  depends  on  the  shape  of  $K$,  the  functionals

183: $V_i(K)$ contain morphological information on the original body.  \\ A

184: corresponding      expansion     holds      for      the     vectorial

185: $\V_0(K)$~\cite{hadwiger:vekt},

186: \begin{equation}

187: \V_0 (K_\epsilon) = \sum_{i=0}^{d} \binom{d}{i}  \omega_{i} \V_{i}(K) \epsilon^i\;\;\;,

188: \end{equation}

189: which defines the {\em Querma{\ss}  vectors} $\V_i(K), \;0\leq i\leq d$. 

190: Under  translations  (t)  and  rotations  (r) of  the  body  $K$,  the

191: Minkowski functionals are {\em motion--invariant} scalars, whereas the

192: Querma{\ss} vectors are {\em motion--equivariant}:

193: \begin{mathletters}

194: \begin{eqnarray}

195: \V_i (t K) =& \V_i (K) + \tv V_i(K),      \\ 

196: \V_i (r K) = &R(r)\V_i (K),              

197: \end{eqnarray}

198: \end{mathletters}

199: \setcounter{equation}{5} 

200: \hspace{-5.5mm} where  $R(r)$ denotes an  orthogonal rotation

201: matrix and $\tv$ a translation  vector. Both the scalars $V_i$ and the

202: vectors $\V_i$ are {\em  additive} functionals in the following sense:

203: If  the  point  set  constituting  the body  $K\subset  \E^d$  is  the

204: set--union of two bodies $K_1$ and $K_2$, then

205: \begin{equation}

206: \phiv_i (K_1 \cup K_2) = \phiv_i (K_1) + \phiv_i (K_2) - \phiv_i(K_1 \cap K_2)\;\;\;,

207: \label{eq:add}

208: \end{equation}

209: \begin{figure}

210: \begin{minipage}[t]{.99\linewidth}

211: \centering

212: \epsfig{file=q2a.eps,height=4cm}

213: \epsfig{file=q2b.eps,height=4cm}

214: \epsfig{file=q2c.eps,height=4cm}

215: \epsfig{file=q2d.eps,height=4cm}

216: \epsfig{file=q2e.eps,height=4cm}

217: \epsfig{file=q2f.eps,height=4cm}

218: \end{minipage}\hfill

219: \caption{Examples of  centroids for a series of  simple patterns.  The

220: panels (1)--(4) show how  the curvature centroids reflect the symmetry

221: of a ``cluster''.  For patterns  consisting of more than one body, the

222: centroids weight  the components  according to their  scalar Minkowski

223: functionals.  Note that  the  patterns in  panel  (5) and  (6) can  be

224: discriminated  using the centroids,  but not  by merely  employing the

225: scalar Minkowski functionals.\label{fig:clusters}}

226: \end{figure}

227: where  we use  the compact  notation $\phiv  \equiv  \{V_i,\V_i\}$. If

228: $K_1$ and  $K_2$ are arbitrary  compact convex bodies, then  the union

229: set  $K_1\cup  K_2$  is  generally  no  longer  convex.  However,  the

230: right--hand--side  of Equation  \eqref{eq:add} is  still well--defined

231: and   may  be   taken   as  the   definition   of  $\phiv_i   (K_1\cup

232: K_2)$. Consequently, additivity allows us to extend the application of

233: the  functionals $\phiv$  to patterns  formed by  an  arbitrary finite

234: union of  convex bodies via iteration of  Equation \eqref{eq:add}.  \\

235: The functionals $\phiv_i$, for $1\leq i\leq d$, depend on the shape of

236: the body $K$; if its  surface, $\partial K$, is smooth, with principal

237: curvatures $c_j  (\x)$, $1\leq j\leq d-1$, then  these functionals may

238: be expressed  in terms of  surface integrals. For instance,  in $d=3$,

239: with  the  surface area  element  $\dr  S(\x)$,  and $\vec{z}  \equiv

240: (1,\x)$ we have:

241: \begin{gather}

242: \phiv_1 (K) = \frac{1}{6} \int_{\partial K}  \vec{z} \dr S(\x) \;\;\;,\;\;\;  \phiv_2 (K) = \frac{1}{3\pi} \int_{\partial K}  \frac{1}{2} (c_1(\x) +c_2(\x)) \vec{z} \dr S(\x)\;\;\;, \\

243: \phiv_3 (K) = \frac{1}{4\pi} \int_{\partial K}  c_1(\x)c_2(\x)  \vec{z} \dr S(\x) \nonumber\;\;\;.

244: \end{gather}

245: The vector $\V_3  (K)$ is known as the Steiner point  of the body $K$.

246: -- We list the  meanings of the Minkowski functionals and  the corresponding vectors in

247: two  dimensions  in   Table~\ref{tab:meaning}.\\  The  most  prominent

248: feature  of the  family $\{  \phiv_i( \cdot  ) |  0\leq i\leq  d\}$ is

249: expressed  by   Hadwiger's  characterization  theorem:   Consider  the

250: polyconvex ring  $\Po^d$ of spatial patterns, defined  to comprise the

251: collection  of  arbitrary finite  unions  of  convex  bodies, $\K_N  =

252: \bigcup_{\alpha=1}^N  K_\alpha \in  \Po^d$, $N<\infty$,  $0  \leq \dim

253: K_\alpha \leq  d$. According to  the theorem, any additive,  equi-- or

254: invariant   and   conditionally  continuous\footnote{The   conditional

255: continuity requires  that a functional  is continuous with  respect to

256: the Blaschke--Haussdorff metric on  the subset $\Kc^d\subset \Po^d$ of

257: convex  bodies.}   functional  $\vec{\Phi}$  on  $\Po^d$  is  uniquely

258: determined by the linear combination:

259: \begin{equation}

260: \vec{\Phi} (\K_N) = \sum_{i = 0}^{d} a_i \phiv_i  (\K_N),

261: \end{equation}

262: \begin{figure}

263: \centering

264: \begin{minipage}[t]{4.7cm}

265: \epsfig{file=q3a.eps,height=4.7cm}

266: \end{minipage}

267: \begin{minipage}[t]{0.5cm}

268: \centering

269: \vspace{-2.3cm}$\leftarrow$

270: \end{minipage}

271: \begin{minipage}[t]{4.7cm}

272: \epsfig{file=q3b.eps,height=4.7cm}

273: \end{minipage}

274: \begin{minipage}[t]{0.5cm}

275: \centering

276: \vspace{-2.3cm}$\rightarrow$ 

277: \end{minipage}

278: \begin{minipage}[t]{4.7cm}

279: \epsfig{file=q3c.eps,height=4.7cm}

280: \end{minipage}

281: \caption{ Two initial configurations  for the ``Game of Life'' (middle

282: panel)  and the  results  for  the 188th  generation  (right and  left

283: panel).  In  the initial state, the  ``thunderbirdfuse'' (black cells)

284: is  perturbed with  the grey  cell marked  by an  arrow.   Whereas the

285: ``thunderbirdfuse''  ends   with  a  ``blinker''   (left  panel),  the

286: ``perturbed  thunderbirdfuse''  evolves into  a  pattern ejecting  two

287: gliders (right panel).\label{fig:gol}}

288: \end{figure}

289: with the real coefficients $a_i$ independent of the pattern $\K_N$. In

290: this  sense,  the  family  $\{\phiv_i\}$  forms a  complete  basis  of

291: morphometric descriptors.   \\ For  the practical analysis  of cluster

292: structures, for  instance, it is convenient to  employ the (curvature)

293: centroids  $\p_i =\left(  p_i^x,p_i^y,...\right)=  \V_i/V_i$, provided

294: $V_i\neq  0$,   $i=0,..,d$.   \\  Let   us  consider  the   series  of

295: 2d--clusters shown in Figure \ref{fig:clusters} to gain some intuition

296: on the kind of information  supplied by the functionals $\phiv_i$, and

297: the centroids,  in particular.  \\  The clusters (1)--(4)  display the

298: effect of  symmetry reduction.  In  the case of  the disk (1)  and the

299: ellipse (2), the centroids  coincide at the symmetry center.  However,

300: \begin{figure}

301: \begin{minipage}[t]{.99\linewidth}

302: \centering

303: \epsfig{file=q4a.eps,height=5cm}

304: \epsfig{file=q4b.eps,height=5cm}

305: \epsfig{file=q4c.eps,height=5cm}

306: \end{minipage}\hfill

307: \caption{The Minkowski functionals for the two ``Game of Life'' series

308: depicted  in  Figure~\ref{fig:gol}.  We  compare  the  length  of  the

309: circumference (left  panel) and  both components of  the corresponding

310: centroid                (middle                and               right

311: panel)\label{fig:golmink}.     ``thunderbirdfuse'':     solid    line,

312: ``perturbed thunderbirdfuse'': dashed line.}

313: \end{figure}

314: \begin{figure}

315: \begin{minipage}[t]{.99\linewidth}

316: \centering

317: \epsfig{file=q5a.eps,height=6cm}

318: \epsfig{file=q5b.eps,height=6cm}

319: \end{minipage}\hfill

320: \caption{The  interplay  of all  curvature  centroids  for the  second

321: ``Game of Life'' series  depicted in Figure~\ref{fig:gol} allows us to

322: answer  questions  on  the  symmetry  and  movement  of  the  patterns

323: arising.\label{fig:golcc}}

324: \end{figure}

325: $C_1$  may   be  distinguished  from   $C_2$  by  the  value   of  the

326: isoperimetric ratio, $I(C_1)  =1$ (which holds only for  the disk) and

327: $I(C_2)>1$, where $I(K) \equiv  4 V_1^2(K)/[\pi V_0(K) V_2(K)] \geq 1$

328: for  convex  bodies.  The  centroids  of  $C_3$  and $C_4$  no  longer

329: coincide; in  $C_3$ they  fan out along  the single  mirror reflection

330: line, whereas  in $C_4$ they form  a triangle. The  clusters $C_5$ and

331: $C_6$ have the same three components, but in different configurations.

332: The scalar Minkowski  functionals of both $C_5$ and  $C_6$ differ from

333: the  ones  of  $C_1$--$C_4$  at  least  by  the  value  of  the  Euler

334: characteristic  $V_2$,  which  counts   here  the  number  of  cluster

335: components.  But the  scalar Minkowski  functionals are  unsuitable to

336: distinguish  $C_5$  from $C_6$  --  additivity  implies  $V_i (C_5)  =

337: V_i(C_6)$  --,  whereas the  centroids  discriminate  clearly.  \\  In

338: Figure  \ref{fig:gol}  we  illustrate  the order  parameters  using  a

339: dynamical   cellular   automaton    generated   by   the   ``Game   of

340: Life''~\cite{berlekamp:ways,gardner:life}.  We  consider two series of

341: patterns, the  ``thunderbirdfuse''~\footnote{available, e.g.  from the

342: library     of     ``Game     of     Life''     patterns     ``xlife''

343: (http://www.mindspring.com/\-$\sim$alanh/  life/).   }   and a  series

344: starting  with  a  slightly  perturbed initial  configuration  of  the

345: ``thunderbirdfuse'',   with   one   point   added  (left   column   of

346: Figure~\ref{fig:gol}).   The ``thunderbirdfuse''  lights consecutively

347: the ``blinkers''  visible as bars.   The temporal sequence as  well as

348: the  spatial course  of the  ignitions is  reflected by  the Minkowski

349: functionals  (Figure~\ref{fig:golmink}) and  the  curvature centroids.

350: The  evolution  terminates with  ``traffic  lights'' oscillating  with

351: period  two   and  constant  values   of  the  scalar   and  vectorial

352: functionals.    \\   The   evolution   of   the   slightly   perturbed

353: \begin{table}

354: \centering

355: \setlength{\extrarowheight}{1pt}

356: \begin{tabular}{|l|l|l|}

357: \hline Minkowski functional & meaning & corresponding vector \\ \hline

358:  $V_0$                      & surface content & $\int_K \d^2 V \x$    \\ \hline

359:  $V_1$                      & one fourth of the length of perimeter & $\frac{1}{4}\int_{\partial K} \d^1 S \x$    \\ \hline

360:  $V_2$                      & Euler characteristic & $\frac{1}{2\pi}\int_{\partial K} \d^1 S\; c(\x) \x$    \\ \hline 

361: \end{tabular} 

362: \caption{The meaning  of the Minkowski functionals  in two dimensions,

363: $c(\x)$ is the local curvature. \label{tab:meaning}}

364: \end{table}

365: ``thunderbirdfuse'' proceeds  rather differently.  The  early patterns

366: are  still comparable;  at  the  time, when  the  pattern reaches  the

367: modified pixel, the perturbed  ``thunderbirdfuse'' grows and attains a

368: constant  surface content from  generation 188  on.  However,  as seen

369: from  $\p_1$,  the pattern  keeps  changing  by  ejecting two  gliders

370: visible  in the  lower right  panel.  The  behaviour of  the curvature

371: centroids  is shown  in Figure~\ref{fig:golcc}:  The asymmetry  of the

372: pattern and its  temporal variation (the ``gliders'' move  one cell in

373: diagonal direction within four generations) is clearly recognizable.

374: 

375: 

376: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

377: % Applications

378: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

379: 

380: \section{Applications to galaxy clusters}

381: \label{sec:app}

382: \begin{figure}

383: \begin{minipage}[t]{.33\linewidth}

384: \centering

385: \epsfig{file=q6.eps,height=5cm}

386: \end{minipage}\hfill

387: \begin{minipage}[t]{.33\linewidth}

388: \centering

389: \epsfig{file=q7.eps,height=5cm}

390: \end{minipage}\hfill

391: \begin{minipage}[t]{.33\linewidth}

392: \centering

393: \epsfig{file=q8.eps,height=5cm}

394: \end{minipage}\hfill

395: \caption{ Example for the excursion  set approach.  The data points of

396: the cluster  $\tau$CDM (middle panel)  are smoothed using  a smoothing

397: length of $0.2\mpc$.  In the right  panel we show the contours and the

398: curvature centroids $\p_i$ of the excursion set at a density threshold

399: of $u  = 0.275  N/R_m^2$.  Here, $N$  denotes the number  of particles

400: within the cluster, $R_m$ quantifies the cluster scale via the maximum

401: distance     of     cluster     particles     from     the     cluster

402: center.\label{fig:constr} --  $h$ accounts for the  uncertainty in the

403: determination   of  the   Hubble   constant   $H$:  $   H   =  100   h

404: \frac{km/s}{Mpc}$.   --  The  left  panel  shows  the  counterpart  of

405: $\tau$CDM,  OCDM,   which  is  embedded  into   an  open  cosmological

406: model.}

407: \end{figure}

408: Evidently, observational  data of  galaxies within clusters  or X--ray

409: photon  maps  do not  provide  from the  outset  the  kind of  spatial

410: patterns required  for the application  of our morphometric  tools. To

411: start  with,  these  patterns  must  first  be  constructed  from  the

412: observational data; but  there is no canonical way  to proceed.  Here,

413: we employ two procedures, the  {\it excursion set method} and the {\it

414: Boolean grain method}  in order to further illustrate  our approach on

415: the basis of simulated and real cluster data.

416: \begin{figure}

417: \begin{minipage}[t]{.99\linewidth}

418: \centering

419: \epsfig{file=q9.eps,height=10cm}

420: \end{minipage}\hfill

421: \caption{

422:   First row: the scalar  Minkowski functionals of the cluster depicted

423:   in Figure \ref{fig:constr} vs. the density threshold $u$ ($u$ is

424:   given in  units of the mean  density within the cluster,  $N$ is the

425:   number of  cluster particles, $R_m$ the maximum  distance of cluster

426:   particles from the center of  mass of all cluster particles). Second

427:   row: one component  of the curvature centroids (left  panel) and the

428:   area  of the  triangle  spanned by  the  curvature centroids  (right

429:   panel).\label{fig:bsp}}

430: \end{figure}

431: 

432: \subsection{The excursion set method}

433: 

434: We smooth  the projected galaxy positions  or the pixels  of the X-ray

435: photon maps  with a Gaussian kernel.  The  smoothing length determines

436: the  scale of  interest.  Then  we  construct the  excursion sets  and

437: investigate their topology and  geometry using the Querma{\ss} vectors

438: and   the  Minkowski   functionals~\cite{schmalzing:beyond}.    \\  We

439: illustrate this procedure by comparing two simulated clusters (part of

440: the              GIF--simulations,              {\em              cf.}

441: \cite{bartelmann:arcIV,beisbart:app});  they   start  with  comparable

442: initial conditions and evolve within different Friedmann--Lema\^\i tre

443: models as  cosmological background,  thus exemplifying the  imprint of

444: the background.  We demonstrate  our method in Figure \ref{fig:constr}

445: using   the   cluster    $\tau$CDM   from   an   Einstein--de   Sitter

446: model~\footnote{A $\tau$CDM model  is a variety of a  Cold Dark Matter

447: structure  formation  scenario embedded  into  an Einstein--de  Sitter

448: background.}  and a smoothing length  of $0.2\mpc$. The results of the

449: Minkowski analysis are depicted  in Figure \ref{fig:bsp}.  We plot the

450: scalar Minkowski  functionals vs.  the density  threshold defining the

451: excursion sets: an averaged and smoothed density profile is encoded in

452: the first  Minkowski functional $V_0$;  a comparison of the  square of

453: $V_1$ and  $V_0$ quantifies how  crooked the isodensity  contours are.

454: The Euler characteristic $V_2$ counts the components (minus the number

455: of  eventual  holes)  of  this  substructure--rich  cluster.   \\  The

456: components $p_i^x$ of the curvature  centroids are shown in the fourth

457: panel of  Figure~\ref{fig:bsp}.  They wander  in space if  the density

458: threshold is  varied.  This indicates that  morphological features are

459: shifting.  Another way  to make the centroids more  illustrative is to

460: consider  the triangle  spanned by  the centroids  and to  compute its

461: volume  and  the  length  of  its perimeter  (fifth  panel  of  Figure

462: \ref{fig:bsp}).  These quantities tell us how symmetric the isodensity

463: contours are.\\ To compare this  cluster to OCDM, its counterpart in a

464: low--density   background   model,   we  condense   the   morphometric

465: information  present in  the Minkowski  functionals and  the centroids

466: into  a few dimensionless  descriptors, by  taking into  account, that

467: both  clusters have  different scales  and numbers  of  particles.  We

468: define for a function $f$ an average over density thresholds via:

469: \begin{equation}

470: \daverage{f} = \frac{1}{u_{max}-u_{min}}\int_{u_{min}}^{u_{max}} \d u f(u)\;\;\;

471: \end{equation}

472: \begin{table}

473: \centering

474: \setlength{\extrarowheight}{1pt}

475: \begin{tabular}{|l|l|l|l|l|}

476: \hline

477: $\sigma/\mpc$ & parameter& definition & $\tau$CDM  & OCDM  \\\hline\hline

478: \multirow{4}{10mm}{0.2}   &  c               &  $\sqrt{\daverage{(\chi-1)^2}}$  &   $3.20$        &    $2.54$      \\ \cline{2-5}

479:                            & $s_0/R_{cl}^2$   &  $\daverage{4V_0(\triangle(\p))}/R_{cl}^2$      &   $8.78\times 10^{-4}$    &    $6.31\times 10^{-4}$        \\ \cline{2-5}

480:                            & ${\rm shift}_0/R_{cl}$ &         $\int_{u_{min}}^{u_{max}} \left| \frac{\d  \p_i}{\d u} \right|\d u /R_{cl}$       \;\;\;.&  $3.85\times 10^{-1}$   &    $4.54\times 10^{-1}$        \\ \hline\hline

481: \end{tabular} 

482: \caption{  The  definitions of  the  condensed  descriptors and  their

483: values  for  the  clusters  $\tau$CDM  and  OCDM  depicted  in  Figure

484:  6. $4V_0(\triangle(\p))$  is the circumference  of the

485: triangle          spanned           by          the          curvature

486: centroids\label{tab:tauopen}.  $R_{cl}$ is the  scale of  the cluster,

487: determined via the area of the  convex hull of the cluster $A_h$: $\pi

488: R_{cl}^2= A_h$. $\sigma$ denotes the smoothing length.}

489: \end{table}

490: and consider  the {\it clumpiness},  the {\it symmetry  parameter} and

491: the {\it  shift of morphological  properties} as defined in  the third

492: column   of   Table~\ref{tab:tauopen}~\footnote{Here,  $u_{min}$   and

493: $u_{max}$ denote the maximum  density found within cluster and outside

494: the cluster, respectively, where ``outside the cluster'' means farther

495: away  from  the  cluster   center  than  any  cluster  point}.   These

496: parameters display the imprint of subclumps, the symmetry and shift of

497: the curvature centroid $\p_2$ and are the larger the more substructure

498: the  cluster exhibits.  As  visible from  Table~\ref{tab:tauopen}, the

499: cluster $\tau$CDM  owns more  substructure with respect  to clumpiness

500: and  symmetry.   A statistical  comparison  of  these descriptors  for

501: larger  numbers of  clusters simulated  in different  models  yields a

502: typical  cluster morphology  and  may help  to  constrain the  present

503: values of the cosmological parameters~\cite{beisbart:app}.

504: \subsection{The Boolean grain method}

505: \begin{figure}

506: \begin{minipage}[t]{.49\linewidth}

507: \centering

508: \epsfig{file=q10a.eps,height=7cm}

509: \end{minipage}\hfill

510: \begin{minipage}[t]{.49\linewidth}

511: \centering

512: \epsfig{file=q10b.eps,height=7cm}

513: \end{minipage}\hfill

514: \caption{

515:   The  cluster  population  of   Cl  0016+161  (left  panel)  and  the

516:   associated Boolean grain model  with circular windows (right panel). 

517:   The radius of the Boolean grains is $15$ arcseconds.

518:   The units are arcseconds.\label{fig:method}}

519: \end{figure}

520: \begin{figure}

521: \begin{minipage}[t]{.33\linewidth}

522: \centering

523: \epsfig{file=q11a.eps,height=5cm}

524: \end{minipage}\hfill

525: \begin{minipage}[t]{.33\linewidth}

526: \centering

527: \epsfig{file=q11b.eps,height=5cm}

528: \end{minipage}\hfill

529: \begin{minipage}[t]{.33\linewidth}

530: \centering

531: \epsfig{file=q11c.eps,height=5cm}

532: \end{minipage}\hfill

533: \caption{

534:   The scalar Minkowski functionals for the cluster Cl 0016+161 vs.

535:   the radius $R_D$ (in arcseconds)  of the sampling window. The radius

536:   of the grains is fixed ($15''$), the Minkowski functionals $V_i$ are

537:   given in units of arcsec$^{2-i}$.\label{fig:rc_box}}

538: \end{figure}

539: The Boolean  grain method decorates each  point $\x_i$ of  a point set

540: with a ball $B_{\x_i}(r)$ of radius  $r$. The union set of these balls

541: is diagnosed  by computing the Minkowski functionals  and centroids as

542: functions  of the  radius. To  probe the  morphology of  an individual

543: cluster locally,  we place a window  $\CD$ on the  cluster center (the

544: center  of  the point  set)  and  study  the centroids  of  $\bigcup_i

545: B_{\x_i}(r)\cap \CD$. Inflation of the sampling window $\CD$ allows us

546: to explore different regions of the cluster. We illustrate this method

547: in  Figure  \ref{fig:method}. Note,  that  also  galaxies outside  the

548: window   contribute~\footnote{Computational   details  are   described

549: in~\cite{beisbart:quer}.}.   \\  To  provide  a concrete  example,  we

550: investigate  the  cluster Cl  0016+161  at  $z\sim  0.54$ observed  by

551: Belloni   and   R{\"o}ser~\cite{belloni:cluster}   with  the   angular

552: positions  and   spectral  properties  of  galaxies  in   a  field  of

553: $3.5'\times 5'$  and brighter than  $R=23.5 {\rm mag}$.   Galaxies are

554: \begin{figure}

555: \begin{minipage}[t]{.99\linewidth}

556: \centering

557: \epsfig{file=q12.eps,height=7cm}

558: \end{minipage}\hfill

559: \caption{

560:   The  behaviour of the  curvature centroids  ($p_i^x$--component) for

561:   the  Boolean  grain model  in  Figure  \ref{fig:method} vs.  the

562:   radius of  the window  (in units of  arcseconds). The  straight line

563:   indicates the cluster center.\label{fig:rc_box_p}}

564: \end{figure}

565: considered  as  cluster  members,  if  their redshifts  lie  within  a

566: predetermined       range\footnote{More      precisely,      following

567: \cite{belloni:cluster} elliptical  and E+A galaxies  are considered as

568: cluster  members, if  for  their redshift  $z$: $0.525<z<0.575$.   For

569: spiral and  irregular galaxies the redshift  range is $0.515<z<0.585$.

570: We  consider only  galaxies  whose morphological  type was  determined

571: (Table 5 in \cite{belloni:cluster}).}.  We are left with $53$ galaxies

572: in  this  field,  as  shown  in Figure  \ref{fig:method}.\\  We  place

573: spherical windows $\CD= B_{R_D}(\cv)$  of different sizes $R_D$ on the

574: center of  mass $\cv$  of the  point set.  The  radius of  the Boolean

575: grains is  held fixed at  $15''$.  In Figure \ref{fig:rc_box}  we show

576: the behaviour  of the scalar  Minkowski functionals.  The area  of the

577: union set of  balls inside the window $\CD$  increases almost linearly

578: with $R_D$.  A homogenous  galaxy distribution would yield a parabolic

579: dependence, but here  the galaxy density drops at  larger scales.  The

580: small peaks in the Euler characteristic typically arise when two parts

581: of  a  connected subcluster  enter  the  window simultaneously.\\  The

582: curvature  centroids are more  sensitive to  geometrical anisotropies.

583: The components $p_i^x$ are depicted in Figure \ref{fig:rc_box_p}.  Two

584: features  are clearly visible:  the centroids  vary when  the sampling

585: window  grows  and,  at  a  fixed scale  of  the  window,  significant

586: differences between  the curvature centroids arise.   The variation of

587: the  centroids is  clearly  correlated with  subclusters entering  the

588: window,  e.g.   the wandering  of  $\p_1$  and  $\p_0$ towards  higher

589: $x$--values   reflects  the  large   cluster  component   in  positive

590: $x$--direction of  the center.  \\  To analyze the morphology  of this

591: cluster more in detail, we  simulate a reference model for comparison.

592: The simplest way to generate  a reference cluster is an inhomogeneous,

593: but              spherically             symmetric             Poisson

594: process~\cite{daley:point-processes}.   We   determine  the  projected

595: density profile  of the cluster non--parametrically  around the center

596: of mass  using a  binning, and simulate  $100$ Poisson  clusters using

597: this profile (the method  for simulating inhomogeneous point processes

598: is  described e.g.  in  ~\cite{stoyan:fractals}).  The  mean Minkowski

599: functionals and the centroids for the model clusters with their r.m.s.

600: fluctuations as well as the  results for the real cluster are depicted

601: in  Figure \ref{fig:poissonvgl}.\\  The  scalar Minkowski  functionals

602: deviate only slightly from the  Poisson model.  On the other hand, the

603: curvature  centroids  reveal  significant  deviations (second  row  of

604: Figure~\ref{fig:poissonvgl});  in  particular,  the $y$--component  of

605: $\p_0$ (right panel) is shifted away  from the true center of mass for

606: relatively small values of the window scale, reflecting the fact, that

607: there are more galaxies above than beneath the cluster center.

608: %   The centroids  show that  this generates  a significant

609: %deviation from a  Poisson model.

610: 

611: \begin{figure}

612: \begin{minipage}[t]{.33\linewidth}

613: \centering

614: \epsfig{file=q13a.eps,height=5cm}

615: \end{minipage}\hfill

616: \begin{minipage}[t]{.33\linewidth}

617: \centering

618: \epsfig{file=q13b.eps,height=5cm}

619: \end{minipage}\hfill

620: \begin{minipage}[t]{.33\linewidth}

621: \centering

622: \epsfig{file=q13c.eps,height=5cm}

623: \end{minipage}\hfill

624: \begin{minipage}[t]{.33\linewidth}

625: \centering

626: \epsfig{file=q13d.eps,height=5cm}

627: \end{minipage}\hfill

628: \begin{minipage}[t]{.33\linewidth}

629: \centering

630: \epsfig{file=q13e.eps,height=5cm}

631: \end{minipage}\hfill

632: \begin{minipage}[t]{.33\linewidth}

633: \centering

634: \epsfig{file=q13f.eps,height=5cm}

635: \end{minipage}\hfill

636: \caption{  The Minkowski  functionals  (first row)  and components  of

637: different curvature centroids (second row) for the cluster Cl 0016+161

638: (solid line) and for a set of Poissonian random clusters following the

639: same density profile  as Cl 0016+161 (dashed line)  vs.  the radius of

640: the Boolean grains  in units of arcseconds.  The  radius of the window

641: is    fixed   at    $69''$.    The    shaded   areas    indicate   the

642: one--$\sigma$--range  for  the random  clusters.   Whereas the  scalar

643: Minkowski  functionals  are consistent  with  the  Poisson model,  the

644: curvature  centroids,   especially  their  $y$--component,   reveal  a

645: significant deviation from this model.

646: \label{fig:poissonvgl}}

647: \end{figure}

648: -- To  strengthen  our  conclusions  we also  simulated  the  Poisson

649: clusters  with   a  different   binning,  but  the   results  remained

650: stable. The  results demonstrate that the  cluster Cl0016+161 exhibits

651: evidential subclustering in comparison with a Poissonian model.

652: 

653: 

654: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

655: % Con

656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

657: 

658: \section{Conclusions}\label{sec:con}

659: The examples discussed in this  note support our claim that the method

660: of  cluster analysis  based  on the  combination  of scalar  Minkowski

661: functionals  with vectorial  centroids  furnishes a  versatile set  of

662: order parameters  to sort out essential aspects  of cluster morphology

663: such as symmetry, clumpiness,  global shape and topology.  Our cluster

664: morphometry rests  on a  solid mathematical basis  derived from  a few

665: reasonable  requirements.   In this  sense,  it  allows  for a  unique

666: morphological description. The construction of patterns from empirical

667: data   introduces  additional   parameters  which   may   be  employed

668: advantageously  for scale--specific  diagnosis.  No  tacit statistical

669: assumptions are involved.

670: %Results from an application  of these tools

671: %in an investigation  of a large--scale galaxy survey  will be reported

672: %in~\cite{beisbart:quer}.  Of course,  our method  is also  feasible to

673: %quantify  the  different  aspects  of substructure  arising  in  other

674: %branches of physics.

675: \\ Finally, we note that  these families of morphological measures may

676: be extended  further to include  tensor--valued Minkowski functionals,

677: which generalize the concept of inertia tensors.

678: %We currently  study how  these descriptors

679: %feature additional  information on both clusters  and the large--scale

680: %structure.

681: \\

682: The  code to  compute the  Minkowski functionals  and  the Querma{\ss}

683: vectors    is    available    on    request    from    the    authors.

684: 

685: 

686: 

687: 

688: 

689: \section*{Acknowledgements}

690: We thank J. Schmalzing for providing his 3d--code for computing scalar

691: Minkowski  functionals  and  for  useful  comments,  J.   Colberg  for

692: providing the GIF-clusters and  M.  Kerscher for valuable discussions.

693: This work was supported  by the ``Sonderforschungsbereich 375-95 f\"ur

694: Astro-Teilchenphysik''                  der                  Deutschen

695: Forschungsgemeinschaft.   T.B.  acknowledges   generous   support  and

696: hospitality by the National Astronomical Observatory in Tokyo, as well

697: as hospitality at Tohoku University in Sendai, Japan.

698: 

699: 

700: \bibliographystyle{abbrv}

701: \begin{thebibliography}{10}

702: 

703: \bibitem{bartelmann:arcIV}

704: M.~{Bartelmann}, A.~{Huss}, J.~M. {Colberg}, A.~{Jenkins}, and F.~R. {Pearce}.

705: \newblock Arc statistics with realistic cluster potentials. {IV}. {C}lusters in

706:   different cosmologies.

707: \newblock {\em Astron. Astrophys.}, 330:1--9, 1998.

708: 

709: \bibitem{beisbart:app}

710: C.~Beisbart, T.~Buchert, M.~Bartelmann, J.~C. Colberg, and H.~Wagner.

711: \newblock {A}pplication of {Q}uerma{\ss} vectors to morphological evolution.

712: \newblock 2000.

713: \newblock In preparation.

714: 

715: \bibitem{beisbart:quer}

716: C.~Beisbart, T.~Buchert, and H.~Wagner.

717: \newblock {Q}uerma{\ss} vectors and {C}urvature {C}entroids: compact morpometry

718:   of cosmic {S}tructure.

719: \newblock 2000.

720: \newblock In preparation.

721: 

722: \bibitem{belloni:cluster}

723: P.~Belloni and H.-J. R{\"o}ser.

724: \newblock Galaxy population in distant clusters. {I}. {C}l0939+472 (z=0.41) and

725:   {C}l0016+161 (z=0.54).

726: \newblock {\em Ap. J. Suppl.}, 118:65, 1996.

727: 

728: \bibitem{berlekamp:ways}

729: E.~{B}erlekamp, J.~{C}onway, and R.~{G}uy.

730: \newblock {\em Winning {W}ays}.

731: \newblock Academic Press, 1982.

732: 

733: \bibitem{daley:point-processes}

734: D.~J. Daley and D.~Vere-Jones.

735: \newblock {\em An introduction to the Theory of Point Processes}.

736: \newblock Springer Verlag, Berlin, 1988.

737: 

738: \bibitem{gardner:life}

739: M.~Gardner.

740: \newblock The fantastic combinations of {J}ohn {C}onway's new solitaire game

741:   "{L}ife".

742: \newblock {\em Scientific American}, page 120, Oct. 1970.

743: 

744: \bibitem{hadwiger:vect2}

745: H.~Hadwiger and C.~Meier.

746: \newblock Studien zur vektoriellen {I}ntegralgeometrie.

747: \newblock {\em Math.\ Nachr.}, 56:361--368, 1974.

748: 

749: \bibitem{hadwiger:vekt}

750: H.~Hadwiger and R.~Schneider.

751: \newblock Vektorielle {I}ntegralgeometrie.

752: \newblock {\em Elemente der Mathematik}, 26:49--72, 1971.

753: 

754: \bibitem{kerscher:fluctuations}

755: M.~Kerscher, J.~Schmalzing, T.~Buchert, and H.~Wagner.

756: \newblock Fluctuations in the 1.2 {J}y galaxy catalogue.

757: \newblock {\em Astron. Astrophys.}, 333:1--12, 1998.

758: 

759: \bibitem{kerscher:abell}

760: M.~Kerscher, J.~Schmalzing, J.~Retzlaff, S.~Borgani, T.~Buchert,

761:   S.~Gottl{\"o}ber, V.~M{\"u}ller, M.~Plionis, and H.~Wagner.

762: \newblock {M}inkowski functionals of {A}bell/{ACO} clusters.

763: \newblock {\em Mon. Not. R. Astron. Soc.}, 284:73--84, 1997.

764: 

765: \bibitem{mcmullen:valuations}

766: P.~McMullen and R.~Schneider.

767: \newblock Valuations on convex bodies.

768: \newblock In P.~M. Gruber and J.~M. Wills, editors, {\em Convexity and its

769:   applications}, pages 170--247. Birkh{\"a}user, Basel, 1983.

770: 

771: \bibitem{mecke:reaction}

772: K.~Mecke.

773: \newblock Morphological characterization of patterns in reaction--diffusion

774:   systems.

775: \newblock {\em Phys. Rev. D}, 53:4794--4800, 1996.

776: 

777: \bibitem{mecke:robust}

778: K.~Mecke, T.~Buchert, and H.~Wagner.

779: \newblock Robust morphological measures for large-scale structure in the

780:   {U}niverse.

781: \newblock {\em Astron. Astrophys.}, 288:697, 1994.

782: 

783: \bibitem{mecke:spinodal}

784: K.~Mecke and V.~Sofonea.

785: \newblock Morphology of spinodal decomposition.

786: \newblock {\em Phys. Rev. D}, 56:R3761--R3764, 1997.

787: 

788: \bibitem{sahni:shapefinders}

789: V.~Sahni, B.~S. Sathyaprakash, and S.~F. Shandarin.

790: \newblock Shapefinders: a new shape diagnostic for large-scale structure.

791: \newblock {\em Ap. J. Lett.}, 459:5, 1998.

792: 

793: \bibitem{schmalzing:beyond}

794: J.~Schmalzing and T.~Buchert.

795: \newblock Beyond genus statistics: a unifying approach to the morphology of

796:   cosmic structure.

797: \newblock {\em Ap. J. Lett.}, 482:L1, 1997.

798: 

799: \bibitem{schmalzing:webI}

800: J.~{Schmalzing}, T.~{Buchert}, A.~L. {Melott}, V.~{Sahni}, B.~S.

801:   {Sathyaprakash}, and S.~F. {Shandarin}.

802: \newblock "disentangling the cosmic web. i. morphology of isodensity contours".

803: \newblock {\em Ap. J.}, 526:568--578, 1999.

804: 

805: \bibitem{schmalzing:cmb}

806: J.~Schmalzing and K.~Gorski.

807: \newblock Minkowski functionals used in the morphological analysis of cosmic

808:   microwave background anisotropy maps.

809: \newblock {\em Mon. Not. R. Astron. Soc.}, 297:355, 1998.

810: 

811: \bibitem{schneider:schwerpunkte}

812: R.~Schneider.

813: \newblock Kr\"ummungsschwerpunkte konvexer {K}\"orper (i).

814: \newblock {\em Abh. Math. Sem. Univ. Hamburg}, 37:112--132, 1972.

815: 

816: \bibitem{schneider:schwerpunkte2}

817: R.~Schneider.

818: \newblock Kr\"ummungsschwerpunkte konvexer {K}\"orper (ii).

819: \newblock {\em Abh. Math. Sem. Univ. Hamburg}, 37:202--217, 1972.

820: 

821: \bibitem{stoyan:fractals}

822: D.~Stoyan and H.~Stoyan.

823: \newblock {\em Fractals, Random Shapes and Point Fields}.

824: \newblock John Wiley \& Sons, Chichester, 1994.

825: 

826: \end{thebibliography}

827: 

828: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

829: %         Figures

830: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

831: 

832: 

833: \end{document}

834: 

835: