1: % $Id: prl.tex,v 1.75 2001/12/07 12:49:39 em Exp $
2:
3: %\documentclass[a4,12pt]{article}
4: \documentstyle[twocolumn,aps,amsmath,amssymb,graphicx,psfrag,prl]{revtex}
5: \bibliographystyle{prsty}
6:
7: %\addtolength{\oddsidemargin}{-20mm}
8: %\addtolength{\textwidth}{35mm}
9: %\addtolength{\topmargin}{-25mm}
10: %\addtolength{\textheight}{55mm}
11: \setlength{\unitlength}{1ex} % unit length for picture environment
12: %\addtolength{\parskip}{1ex}
13:
14: %\usepackage{amsmath,amssymb}
15:
16: % rotate argument 90 degrees right/left
17: %\def\rotateright#1{{\setbox0=\hbox{#1}\rotr0}}
18: %\def\rotateleft#1{{\setbox0=\hbox{#1}\rotl0}}
19:
20: \def\cf{cf.\hbox{}}
21: \def\eg{e.g.\hbox{}}
22: \def\etal{{\it et~al.\/}}
23: \def\ie{i.e.\hbox{}}
24: \def\opcit{{\it op.~cit.\/}}
25: \def\qv{{\it q.v.\/}}
26: \def\vs{vs.\hbox{}}
27:
28: \def\ccbeta{\eta}
29: \def\ltsim{\lesssim}
30: \def\gtsim{\gtrsim}
31: \def\thalf{{\textstyle\frac{1}{2}}}
32:
33: \def\constant{\text{constant}}
34: \def\turn{{\text{turn}}}
35: \def\CSS{{\text{CSS}}}
36: \def\DSS{{\text{DSS}}}
37:
38: % people's names and other words with accents in them
39: \def\Bizon{Bizo\'{n}}
40: \def\Hamade{Hamad\'{e}}
41: \def\OMurchadha{\'{O}~Murchadha}
42: \def\Poincare{\text{Poincar\'{e}}}
43: \def\Sundstrom{Sundstr\"{o}m}
44:
45: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
46:
47: \begin{document}
48: \title{A New Transition between Discrete and Continuous Self-Similarity
49: in Critical Gravitational Collapse}
50: \author{Christiane Lechner$^{(1,2)}$,
51: Jonathan Thornburg$^{(1,2)}$,
52: Sascha Husa$^{(2)}$,
53: and Peter C.~Aichelburg$^{(1)}$}
54: \address{$^{(1)}$%%%
55: Institut f\"{u}r Theoretische Physik,
56: Universit\"{a}t Wien, \\
57: Boltzmanngasse~5, A-1090 Wien, Austria \\
58: $^{(2)}$%%%
59: Max-Planck-Institut f\"ur Gravitationsphysik,
60: Albert-Einstein-Institut\\
61: Am M\"uhlenberg 1, D-14476 Golm, Germany \\
62: }
63:
64: %
65: % We want CVS to expand the Id keyword on the next line, but we don't
66: % want TeX to go into math mode to typeset the expansion (because that
67: % wouldn't look as nice in the output), so we use the "$ $" construct
68: % to get TeX out of math mode again when typesetting the expansion.
69: %
70: %\date{$ $Id: prl.tex,v 1.75 2001/12/07 12:49:39 em Exp $ $}
71: \date{Dec.~7 2001}
72: \maketitle
73:
74: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
75:
76: % n.b. PRL limits abstract to 600 *characters*
77: \begin{abstract}
78: We analyze a bifurcation phenomenon associated with critical
79: gravitational collapse in a family of self-gravitating SU(2) $\sigma$-models.
80: As the dimensionless coupling constant decreases,
81: the critical solution changes from discretely self-similar (DSS)
82: to continuously self-similar (CSS).
83: Numerical results
84: provide evidence for a bifurcation which
85: is analogous to a heteroclinic loop bifurcation in dynamical systems,
86: where two fixed points (CSS) collide with a limit cycle (DSS) in phase
87: space as the coupling constant tends to a critical value.
88: \end{abstract}
89: \pacs{PACS
90: 04.25.Dm, % Numerical relativity
91: 64.60.Ht, % Dynamic critical phenomena
92: 04.70.Bw % Classical black holes
93: }
94:
95: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
96:
97: %%%p%%%\paragraph*{Introduction}
98:
99: Gravitational collapse at the threshold of black hole formation
100: ties together many of the fundamental issues of general relativity, such
101: as the global aspects of solutions, the structure of singularities arising
102: from regular initial data and the cosmic censorship hypothesis~%%%
103: \cite{Penrose-strong-cosmic-censorship}.
104: By fine-tuning initial data for a gravitating massless scalar field
105: to the boundary between eventual dispersal and complete collapse,
106: Choptuik~\cite{Choptuik-1993-self-similarity} found phenomena reminiscent
107: of criticality associated with phase transitions in statistical physics,
108: such as universality and scaling (e.g. of the black hole mass).
109: Considerable qualitative understanding has been gained by explaining
110: critical collapse in terms of a single unstable mode of the universal
111: critical solution. This solution is understood as an intermediate attractor
112: located in a codimension-one stable hypersurface in phase space,
113: separating data which do or do not form black holes.
114: Critical collapse has by now been studied in a number of matter models --
115: in all of these the physics of the threshold of black hole formation
116: was found to be governed by symmetry: the critical solution exhibits
117: either continuous or discrete self-similarity, or staticity or
118: periodicity in time~\cite{Gundlach-1999-critical-phenomena-review}.
119:
120: In this paper, we study numerically a simple model with a single scalar
121: field in spherical symmetry which exhibits CSS critical behavior at
122: small coupling constants, and DSS critical behavior at large ones.
123: At intermediate coupling constants we observe a
124: competition between CSS and DSS solutions giving rise to a new phenomenon:
125: within an approximately-DSS critical evolution we find several episodes of
126: approximate CSS. Our main focus here is on the interpretation of this
127: observation in terms of an analogy to a heteroclinic loop bifurcation
128: in finite dimensional dynamical systems, at which the limit cycle
129: (DSS) merges
130: with two fixed points (the CSS solution and its negative).
131: Apart from the bifurcation itself, the model also shows other interesting
132: features, such as the existence of a stable (with respect to linear spherical
133: perturbations) self-similar solution for some finite range of the coupling,
134: and a ``suppression'' effect for the CSS solution in critical searches
135: which we interpret as ``shielding'' by an apparent horizon.
136:
137: Our results are based on the direct numerical construction of the
138: CSS and DSS solutions, a linear perturbation analysis of the CSS
139: solutions, and comparison with critical evolutions. The latter are
140: defined by considering a 1-parameter family of initial data
141: $\phi = \phi_p(u_0,r)$, such that (say) for small values of~$p$
142: the field eventually disperses (as determined by a numerical
143: evolution~\cite{Husa-etal-2000-sigma-model-DSS-criticality}), while
144: for large values of $p$ it eventually forms a black hole (diagnosed by
145: the appearance of an apparent horizon). We use a binary search in~$p$
146: to numerically approximate the critical solution at the
147: threshold of black hole formation.
148: We refer to such fine-tuned numerical solutions as near-critical
149: evolutions, and our results are taken from initial data
150: which are fine-tuned to the same tolerance $\delta p/p < 10^{-14}$.
151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
152:
153: %%%p%%%\paragraph*{The Model}
154:
155: The self-gravitating SU(2) $\sigma$-model \cite{Misner-1978-harmonic-maps}
156: under investigation
157: is a wave map from spacetime to the target manifold $S^3$ with the
158: standard metric. The so called Hedgehog ansatz of spherical symmetry
159: leaves a single matter field $\phi(u,r)$ coupled to gravity:
160: \begin{equation}
161: \square\phi
162: =\frac{\sin (2\phi )}{r^{2}}
163: \,\text{,}
164: \label{eqn-phi}
165: \end{equation}
166: where $\square$ is the spacetime wave operator. Our geometric setup,
167: numerical evolution scheme, and convergence tests are described in a previous
168: paper~\cite{Husa-etal-2000-sigma-model-DSS-criticality}.
169: In particular, we use retarded Bondi-like coordinates $(u,r)$ with
170: metric functions $\beta (u,r)$ and $V(u,r)$.
171: Suitable combinations of Einstein's equations lead to
172: \begin{equation}
173: \beta' = \frac{\ccbeta }{2}r(\phi')^2
174: \text{,}
175: \quad
176: V' = e^{2\beta}(1-2\ccbeta \sin^2 \phi)
177: \,
178: \text{,}
179: \label{eqn-beta'-V'}
180: \end{equation}
181: where prime denotes the derivatives with respect to $r$,
182: and $\eta$ the dimensionless coupling constant.
183: The hypersurface equations~\eqref{eqn-beta'-V'}
184: and the matter field equation~\eqref{eqn-phi} suffice to evolve
185: all the dynamical fields $V$, $\beta$, and $\phi$.
186:
187: For vanishing coupling ($\eta = 0$), the theory describes a $\sigma$-model
188: on a fixed background. Taking this background as Minkowski space,
189: \Bizon{} \etal{}~\cite{Bizon-Chmaj-Tabor-1999-sigma-3+1-evolution} and
190: Liebling \etal{}~\cite{Liebling-Hirschmann-Isenberg-1999-sigma-critical}
191: find a CSS critical solution at the threshold of singularity formation.
192: \Bizon{}~\cite{Bizon-1999-existence-of-self-similar-sigma-CSS-solutions} and
193: \Bizon{} and Wasserman~\cite{Bizon-Wasserman-2000-CSS-exists-for-nonzero-beta}
194: have shown that for each $0 \leq \eta < 0.5$, a countably infinite family
195: of CSS solutions exists, indexed by the number of nodes in
196: $\phi(u=\constant,r) - \pi/2$.
197: In the limit $\eta \rightarrow \infty$
198: Liebling~\cite{Liebling-inside-global-monopoles} finds DSS critical
199: collapse at the threshold of black-hole formation.
200: In Ref.~\cite{Husa-etal-2000-sigma-model-DSS-criticality}
201: we find that for $\eta \gtsim 0.2$ the system shows ``exact'' DSS
202: critical collapse, but for $0.18 \le \eta \ltsim 0.2$ we see
203: only approximate DSS behavior; furthermore the period~$\Delta$ exhibits
204: a sharp rise as the coupling decreases from~$0.5$ to~$0.18$
205: (Fig.~\ref{fig:Delta(eta)}). These results suggest a transition
206: from CSS to DSS critical collapse somewhere in the range
207: $0 < \eta \ltsim 0.18$, which we identify and discuss in the present
208: paper.
209: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
210:
211: \begin{figure}[htb]
212: \begin{center}
213: \begin{psfrags}
214: \psfrag{Delta}[c][c][1][-90]{$\Delta$}
215: \psfrag{eta}[c][c]{$\eta$}
216: \includegraphics[width=40ex]{Delta_eta.eps}
217: \end{psfrags}
218: \end{center}
219: \caption[DSS echoing period $\Delta$]
220: {%%%
221: The DSS echoing period $\Delta$
222: is shown as a function of the coupling constant $\eta$,
223: computed from dynamical evolutions (dots) and from
224: direct construction (line).
225: Our dynamical evolution is only approximately DSS
226: at its lowest $\eta$ value ($\eta = 0.18$), so we
227: can only determine $\Delta$ approximately there.
228: }
229: \label{fig:Delta(eta)}
230: \end{figure}
231:
232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
233:
234:
235: %%%p%%%\paragraph*{Direct Construction of CSS and DSS Solutions}
236:
237: For $0 \leq \eta < 0.5$ we have numerically constructed the CSS
238: solutions discussed in
239: \cite{Bizon-Wasserman-2000-CSS-exists-for-nonzero-beta} as an
240: ODE-eigenvalue problem.
241: We have also performed a linear stability analysis of the CSS solutions;
242: we find that the first excitation (one node),
243: which we refer to below as ``the'' CSS solution, has precisely
244: one unstable mode. This CSS solution takes the form
245: $\phi = \pm \phi_\CSS(z;u^*_\CSS)$, where $z = r/(u^*_\CSS-u)$, with a single
246: free parameter $u^*_\CSS$ giving the retarded time of the accumulation
247: point. The sign ambiguity is a consequence of Einstein's equations and
248: the field equation~\eqref{eqn-phi} being invariant under $\phi \to -\phi$.
249:
250: For $\eta \geq 0.1726$ we have also explicitly constructed the
251: ``Choptuon'' DSS solution via a pseudospectral method following
252: the lines of Gundlach~\cite{Gundlach-1996-understanding-critical-collapse}
253: (see \cite{Lechner-DSS-forthcoming}).
254: The DSS solution takes the form
255: $\phi = \phi_\DSS(\tau,z;u^*) = \phi_\DSS(\tau+n\Delta,z;u^*)$,
256: where $n$ is any integer, $\Delta$ is the DSS period,
257: $z$ is again given by $r/(u^*-u)$, and $\tau = - \ln(u^*-u)$.
258: As $\eta$ decreases $\Delta$
259: rises sharply (Fig.~\ref{fig:Delta(eta)}).
260: Also, a rapidly increasing number of Fourier components is required
261: to accurately represent the Choptuon, and the construction algorithm
262: becomes increasingly ill-conditioned.
263: Below we will give further arguments suggesting that the
264: DSS Choptuon ceases to exist somewhat below the lower limit of our
265: numerical construction.
266:
267: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
268:
269: %%%p%%%\paragraph*{CSS and DSS regimes}
270:
271: For very small couplings $\eta \ltsim 0.1$ the stable CSS ground state
272: causes a generic class of initial data to collapse to naked singularities.
273: Here we focus on the transition from CSS to DSS in critical collapse
274: at the threshold of {\em black hole\/} formation: We therefore restrict
275: our attention to $\eta \geq 0.1$, where we find only dispersal and
276: black hole (apparent horizon) formation as generic end-states.
277:
278: For $0.1 \leq \eta \ltsim 0.14$ we find CSS critical collapse, while
279: for large couplings $\eta \gtsim 0.2$ we have previously found DSS
280: critical collapse~\cite{Husa-etal-2000-sigma-model-DSS-criticality}.
281: In both ranges we observe scaling of the black hole mass for
282: supercritical initial data and of the maximum central Ricci scalar
283: for subcritical initial data. In the CSS regime $0.1 \leq \eta \ltsim 0.14$
284: and in the DSS regime for $\eta \gtrsim 0.2$ the
285: critical exponents are approximately constant (within a few percent):
286: $\gamma_\CSS \approx 0.18$ and $\gamma_\DSS \approx 0.11$.
287:
288: %%%p%%%\paragraph*{Episodic CSS}
289:
290: In the transition regime $0.14 \ltsim \eta \ltsim 0.2$, we find
291: that critical solutions show a new phenomenon which we call
292: ``episodic self-similarity'': The field configuration closely
293: approximates CSS behavior on large parts of the slice for a finite time,
294: then departs and returns to CSS again.
295: This cycle repeats several times
296: before the evolution either leads to black hole formation or dispersal.
297: We find that $\phi \approx +\phi_\CSS$ and $\phi \approx -\phi_\CSS$
298: episodes always alternate. The accumulation times $u^*_\CSS$
299: increase from one CSS episode to the next.
300:
301: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
302:
303: \begin{figure}[htb]
304: \begin{center}
305: \begin{psfrags}
306: \psfrag{-log(u* - u)}[c][c]{$- \ln (u^* - u)$}
307: \psfrag{RMS distance in phi}[c][c]{RMS distance in $\phi$}
308: \psfrag{max 2m/r}[c][c]{max $2m/r$}
309: \includegraphics[width=87mm]{css-dice.eps}
310: \end{psfrags}
311: \end{center}
312: \caption[Figure showing CSS-DICE fits at various coupling constants]
313: {
314: This figure shows the distance ($r^2$-weighted RMS difference
315: in $\phi$, taken between the origin and the self-similarity
316: horizon, within each near-critical-evolution slice)
317: between the CSS solution and the critical solution, for
318: $\eta = 0.16$, $0.1726$, $0.1806$, and $0.2003$.
319: (The choice of $u^*$, and thus the horizontal coordinate,
320: is somewhat arbitrary for $\eta = 0.16$.)
321: }
322: \label{fig:css-dice}
323: \end{figure}
324:
325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
326: In order to study episodic self-similarity quantitatively, we have fitted
327: numerical near-critical evolutions against our explicitly-constructed CSS solutions
328: (fitting the CSS parameter $u^*_\CSS$ independently at each
329: near-critical-evolution
330: slice). Figure~\ref{fig:css-dice} shows these fits for a range of
331: coupling constants. The repeated close approaches of the near-critical
332: evolutions to the CSS solutions are clearly visible; the approaches
333: become closer and closer and the time spent in the neighborhood
334: (\ie{}~within a given distance) of the CSS solution increases
335: as $\eta$ is decreased.
336:
337: In the range where episodic CSS occurs we also observe approximate DSS
338: behavior. It is therefore interesting to compare the near-critical evolution
339: to the explicitly constructed DSS solution.
340:
341:
342: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
343: \begin{figure}[]
344: \begin{center}
345: \begin{psfrags}
346: \psfrag{DSS tau (= Delta * phase in cycles)}[c][c]{DSS $\tau$}
347: \psfrag{RMS distance in phi}[c][c]{RMS distance in $\phi$}
348: \psfrag{max 2m/r}[c][c]{max $2m/r$}
349: \includegraphics[width=87mm]{dss-dice.eps}
350: \end{psfrags}
351: \end{center}
352: \caption[Figure showing DSS-DICE fits at various coupling constants]
353: {
354: Best fits between the DSS Choptuon
355: and the critical solution are shown for $\eta = 0.1726$, $0.1806$,
356: and $0.2003$. The lower subplot shows the
357: distance (same definition as in figure~\protect\ref{fig:css-dice})
358: between the DSS Choptuon and the corresponding best-fitting
359: slices of the near-critical evolution, as a function of
360: $\tau$. The upper subplot shows the maximum of
361: $2m/r$ within the same slices of the critical evolution.
362: $\tau$ is only defined up to an arbitrary integer multiple
363: of $\Delta$ at each coupling constant.
364: % but is consistent between the subplots.
365: }
366: \label{fig:dss-dice}
367: \end{figure}
368:
369: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
370:
371: Figure~\ref{fig:dss-dice} shows fits of numerical near-critical evolutions
372: against the explicitly-constructed DSS solutions (finding best-fitting
373: pairs of slices between the near-critical evolutions and the DSS solutions)
374: for several coupling constants where DSS exists. Notice that as $\eta$
375: decreases, the near-critical solution's approach to the DSS solution
376: becomes slower, and the closest approach becomes less close.
377: The time intervals in $\tau$ from approaching the Choptuon within an RMS error
378: of $\sim 0.1$ (which is where the curves in Fig.~\ref{fig:dss-dice} starts)
379: to the start of the departures are however roughly equal.
380: This and the slow approach account for the (only) approximate DSS behavior
381: of near-critical evolutions already observed in
382: \cite{Husa-etal-2000-sigma-model-DSS-criticality} for $0.18 \le \eta
383: \ltsim 0.2$.
384:
385: Comparing Figs. \ref{fig:css-dice} and \ref{fig:dss-dice} one infers that
386: as the critical evolution is attracted to the DSS solution it comes
387: periodically close to the CSS solution, which implies that the DSS and CSS
388: solutions must themselves be close.
389: Figure~\ref{fig:css-dss} shows fits between the CSS solutions and
390: the explicitly-constructed DSS solutions (again fitting the CSS
391: parameter $u^*_\CSS$ independently at each slice). There are two close
392: approaches
393: within each DSS cycle, corresponding to the two sign choices
394: $\phi = \pm \phi_\CSS$. Note that the close approaches become closer as
395: $\eta$ decreases.
396:
397:
398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
399:
400: \begin{figure}[]
401: \begin{center}
402: \begin{psfrags}
403: \psfrag{DSS phase (cycles)}[c][c]{DSS phase (cycles)}
404: \psfrag{RMS distance in phi}[c][c]{RMS distance in $\phi$}
405: \includegraphics[width=87mm]{css-dss.eps}
406: \end{psfrags}
407: \end{center}
408: \caption[Figure showing CSS-DSS fits at various coupling constants]
409: {
410: The distance (same definition as in figure~\protect\ref{fig:css-dice})
411: between the CSS and DSS solutions is shown as a function of DSS
412: phase (measured in orbits around the DSS Choptuon), for
413: $\eta = 0.1726$, $0.1806$, and $0.2003$. The origin of
414: the DSS phase scale is arbitrary at each coupling constant.
415: }
416: \label{fig:css-dss}
417: \end{figure}
418:
419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
420:
421:
422:
423:
424: Combining our results, we conjecture the following bifurcation
425: scenario in the language of dynamical systems: for large couplings
426: $0.2 \ltsim \eta < 0.5$, the limit cycle representing $\phi_\DSS$
427: in phase space and the two fixed points $\pm \phi_\CSS$ lie far apart
428: (in some suitable norm). As $\eta$ is decreased, the limit cycle and
429: the CSS points move closer and finally merge at some $\eta_c \approx 0.17$.
430: In this limit the DSS solution becomes a heteroclinic orbit connecting
431: the CSS fixed points and the period of the limit cycle tends to infinity
432: (see below).
433: For $\eta < \eta_c$, DSS ceases to exist.
434:
435: We conjecture that the DSS solution still plays the role
436: of a critical intermediate attractor even at coupling constants
437: just slightly larger than $\eta_c$, where the CSS solutions lie very close to
438: the DSS cycle.
439: An evolution, which is tuned to evolve towards the DSS-CSS region is carried
440: along by the flow of the DSS cycle to periodically come close to the
441: CSS fixed points. We have numerical evidence that the periodical turning away
442: from CSS is dominated by the unstable mode of CSS.
443:
444:
445: %%%p%%%\paragraph*{DSS as Perturbations of CSS}
446: When the DSS evolution
447: is close to one of the CSS fixed points in phase space we can expand
448: the field in terms of linear perturbations around the CSS solution.
449: The departure from CSS must thus happen via the unstable mode of CSS.
450: The amplitude for this mode grows from an initial amplitude $A_0$
451: to some fixed amplitude (still in the linear regime) in a
452: time $T = - (1/\lambda) \ln (A_0) + \constant$, where $\lambda$ denotes the
453: eigenvalue of the unstable mode of the CSS solution.
454: Since this happens twice
455: in a DSS cycle, we can write the total duration of the DSS cycle as
456: $\Delta = 2T + T_{\turn}$, where $T_\turn$ denotes the time spent in
457: the (nonlinear) turnover from one of $\pm \phi_\CSS$ to the other.
458: From our linear perturbation analysis of the CSS solution we find that
459: $\lambda \approx 5.14$ is only slowly varying for $\eta$ near $\eta_c$
460: \cite{Lechner-PhD}. If we assume
461: that $A_0 \sim \eta - \eta_c$ for $\eta$ near $\eta_c$
462: and that
463: $T_\turn$ is roughly constant, we have
464: $\Delta = - (2/\lambda) \, \ln (\eta - \eta_c) + \textrm{const}$.
465: To test this prediction, we have fitted the $\Delta(\eta)$ values
466: shown in Fig.~\ref{fig:Delta(eta)} to the 3-parameter functional
467: form $f(\eta) = -a \ln(\eta - \eta_c) + b$ in the range
468: $\eta \in [0.1726,0.195]$.
469: The fit is very good, with a maximum relative error of $0.3\%$.
470: We obtain
471: $\eta_c \simeq 0.17$, which is consistent with what we expect from the
472: raise in the number of relevant Fourier coefficients,
473: and $a \sim 2/\lambda$ with a relative error of $\sim 7 \%$. Given
474: the fact that we neglected higher order terms and the variation of $\lambda$,
475: the fitted value for $a$ is remarkably close to the theoretically predicted
476: one.
477:
478: Figure~\ref{fig:behavior-overview} gives a schematic overview of all our
479: observations.
480: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
481: \begin{figure}[htb]
482: \begin{center}
483: \begin{psfrags}
484: \psfrag{eta}[c][c]{$\eta$}
485: %\psfrag{CSS}[c][c]{{\footnotesize{CSS}}}
486: \includegraphics[width=87mm]{eta.xfigC.eps}
487: \end{psfrags}
488: \end{center}
489: \caption[Overview of phenomenology at various coupling constants]
490: {%%%
491: This figure shows the different phenomenology we observe
492: at various couplings $\eta$. The dashed line denotes
493: CSS solutions, that contain marginally trapped surfaces.
494: }
495: \label{fig:behavior-overview}
496: \end{figure}
497:
498: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
499:
500: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
501:
502: %%%p%%%\paragraph*{Discussion}
503:
504: We also observe episodic CSS at couplings $0.14 \lesssim \eta < \eta_c$,
505: where we believe the DSS solution ceases to exist.
506: What is tuned out in such a critical search? In the language of dynamical
507: systems the answer could be that below the critical coupling the cycle
508: of DSS is broken \ie{} it does not close, but the flow still defines locally
509: an invariant manifold of codimension one. It is this manifold which
510: separates black hole formation from dispersal.
511: The mass scaling from supercritical searches in the intermediate
512: regime $0.14 \ltsim \eta < 0.2$ has not been conclusive so far.
513: Observed deviations from a simple scaling law
514: require further investigation.
515:
516: Another important question is: Why does the CSS solution cease to
517: be a critical solution for black hole formation for larger couplings?
518: Our stability analysis shows that CSS has a single
519: unstable mode up to $\eta = 0.5$, which in principle could be tuned out in
520: a critical search.
521: We believe that the answer to this is related to the observation by
522: \Bizon{} and Wasserman
523: \cite{Bizon-Wasserman-2000-CSS-exists-for-nonzero-beta}, that this
524: solution contains a spacelike hypersurface of marginally trapped surfaces
525: outside the backwards light cone of the culmination point for $\eta > 0.152$.
526: We find that numerical evolutions for $\eta = 0.2$
527: with initial data that are close to the CSS solution
528: inside the backwards lightcone and are asymptotically flat outside, very
529: quickly develop an apparent horizon and thus become a black hole. If this is
530: the generic behavior, then the CSS solution can not lie on the boundary of
531: black hole formation.
532:
533: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
534:
535: %%%p%%%\paragraph*{Conclusions}
536: Summing up, the SU(2) $\sigma$-model shows CSS critical
537: behavior for small and DSS for large values of the coupling constant.
538: In the transition region
539: we observe
540: episodic CSS behavior. We have strong evidence that the CSS/DSS
541: transition is the infinite dimensional analog of a global heteroclinic
542: bifurcation, which is quite different
543: from previously reported bifurcations in self-similar critical collapse, which
544: were found to be characterized by a change of stability (see \eg{}
545: \cite{Liebling-Choptuik-CSS-DSS}).
546: In particular, analogies to finite dimensional dynamical systems
547: pictures have proven essential in interpreting critical collapse
548: (see e.g. Ref.~\cite{Gundlach-1999-critical-phenomena-review}
549: for an overview), and we believe that the bifurcation picture discussed here
550: will stimulate further insights into critical gravitational collapse.
551: It would be interesting to see whether episodic CSS occurs in the
552: critical collapse of different matter models, or also in completely
553: different physical systems.
554: Details of our methods and results, some of which could only be mentioned
555: here
556: will be published in a forthcoming paper.
557:
558: %%%p%%%\paragraph*{Acknowledgments}
559: This work has been supported by
560: the Austrian
561: %Fonds zur F\"{o}rderung der wissenschaftlichen Forschung
562: FWF
563: (project P12754-PHY), the Fundacion Federico,
564: the Alexander von Humboldt Foundation,
565: and G.~Rodgers and J.~Thorn [J.T.].
566: We thank Michael P\"urrer for his contributions to our numerical
567: evolution code~\cite{Husa-etal-2000-sigma-model-DSS-criticality},
568: Piotr \Bizon{} for many useful discussions and for sharing research
569: results in advance of publication,
570: and Carsten Gundlach and Jos\'e M.~Mart\'\i{}n-Garc\'\i{}a
571: for stimulating discussions
572: and for drawing our attention to the analysis leading to the logarithmic
573: divergence of $\Delta$.
574:
575: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
576:
577: % make LaTeX read in sigma_crit.bbl produced by bibtex using jt.bib
578: \bibliography{jt}
579:
580: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
581: \end{document}
582: