0808.0880/ms.tex
1: % last modified by vh Aug. 5 08
2: 
3: %\documentclass[nofootinbib,floatfix,prd,preprintnumbers,twocolumn]{revtex4}
4: \documentclass[10pt]{iopart}
5: \usepackage{graphicx}
6: \usepackage{iopams}
7: \usepackage{bm}
8: 
9: %\usepackage{hyperref}
10: %\usepackage[margin=0.75in,a4paper]{geometry}
11: 
12: \begin{document}
13: 
14: \renewcommand*{\max}{{\mathrm{max}}}
15: \newcommand*{\eff}{\mathrm{eff}}
16: \newcommand*{\h}{\mathrm{h}}
17: \renewcommand*{\b}{\mathrm{b}}
18: \newcommand*{\cosmo}{\mathrm{m}}
19: \newcommand*{\di}{\partial}
20: \newcommand*{\approaches}[2]{\xrightarrow[#2]{\,\,\,{#1}\,\,\,}}
21: 
22: \title{Ricci flows, wormholes and critical phenomena}
23: 
24: \bibliographystyle{iopart-num}
25: 
26: \author{Viqar Husain and Sanjeev S.~Seahra}
27: 
28: \address{Department of Mathematics \& Statistics, University of
29: New Brunswick, Fredericton, NB E3B 5A3, Canada}
30: 
31: \begin{abstract}
32: We study the evolution of wormhole geometries under Ricci flow using
33: numerical methods. Depending on values of initial data parameters,
34: wormhole throats either pinch off or evolve to a monotonically
35: growing state. The transition between these two behaviors exhibits a
36: from of critical phenomena reminiscent of that observed in
37: gravitational collapse. Similar results are obtained for initial
38: data that describe space bubbles attached to asymptotically flat
39: regions. Our numerical methods are applicable to ``matter-coupled''
40: Ricci flows derived from conformal invariance in string theory.
41: \end{abstract}
42: 
43: \section{Introduction}
44: 
45: An analog of the diffusion equation for geometries is the so-called
46: Ricci flow. This is a type of parabolic partial differential
47: equation that homogenizes geometry in a manner similar to the
48: homogenization of density or heat produced by the diffusion
49: equation.
50: 
51: The Ricci flow equation was first studied in the early 1980s.  It
52: arose nearly simultaneously in mathematics \cite{hamilt}, in the
53: study of the classification of 3-geometries, and in string theory
54: physics \cite{friedan,sen} from the requirement that a quantized
55: string  retain conformal invariance. Certain modifications of the
56: purely geometric flow equation by ``matter terms'' motivated by
57: string theory were central to Perelman's \cite{perel} work on the
58: Poincare conjecture.
59: 
60: A goal of analytic work on the flow equations is to study the
61: approach to singularities, and to understand the  so-called
62: ``ancient solutions,''  which are the non-linear analogs of the heat
63: kernel of the diffusion equation. Much of the work in this area has
64: focused on the flow of compact geometries \cite{cao-chow}, with
65: relatively little on non-compact cases such as the asymptotically
66: flat or anti-deSitter geometries. In the asymptotically flat case
67: for dimension $n\ge3$, it has been demonstrated recently that the
68: flow exists and remains asymptotically flat for an interval of time,
69: with the mass remaining constant \cite{ow-asympflat,dai-ma}. For the
70: rotationally symmetric case, Ref. \cite{ow-asympflat} also shows
71: that the flow is eternal for data containing no minimal surfaces.
72: 
73: The purpose of this paper is to investigate numerically some
74: examples of  the flow of asymptotically flat geometries, and to
75: describe techniques that are easy to generalize to cases where
76: matter flow equations are included.
77: 
78: Previous numerical work on Ricci  flows focused on the (compact)
79: case of a 3-sphere that is corseted (i.e., a surface formed by
80: uniformly shrinking the equator of a 3-sphere), where it was found
81: that the manifold flows to a round 3-sphere or a neck pinch
82: depending on the degree of corseting \cite{gi}. The flow of horizon
83: area and Hawking mass has also been studied \cite{sam} for initial
84: data that is viewed as time symmetric data for the Einstein
85: constraint equations. Numerical work pertaining to string theory
86: aims at obtaining new black hole solutions as fixed points
87: \cite{hw}.
88: 
89: The focus of the present work is the geometric flow of wormhole
90: geometries. We study several  classes of initial data  and find two
91: possible outcomes: the wormhole ``pinches off'' leaving two disjoint
92: asymptotically flat regions, or it expands indefinitely. We also
93: find that there is a type of critical behaviour that separates these
94: extreme cases. The numerical methods we utilize are stable for long
95: time flows, and have potential applicability to the largely
96: unexplored geometry-matter cases.
97: 
98: \section{Flow equations}
99: 
100: The Hamilton-DeTurck Ricci flow is defined by the equation
101: \cite{hamilt,deturck}
102: \begin{equation}\label{eq:flow equation 1}
103:     \di_t g_{ab} = -2R_{ab} + 2 \nabla_{(a} V_{b)}.
104: \end{equation}
105: For our work, we take $g_{ab} = g_{ab}(t,x^c)$ to be the 3-metric on
106: a manifold $\Sigma_t$ with Euclidean signature and $R_{ab}$ is the
107: associated Ricci tensor. The parameter $t$ labels different
108: 3-geometries along the flow and should not be confused with the
109: intrinsic 3-dimensional coordinates $x^c$. Finally, $V^a$ is a
110: DeTurck vector field that generates diffeomorphisms along the flow.
111: The DeTurck field essentially represents the freedom to change
112: 3-dimensional coordinates as the flow progresses.  It is useful to
113: note that $t$ has dimensions of $[\mathrm{length}]^{2}$ and $V^a$
114: has dimensions of $[\mathrm{length}]^{-1}$.
115: 
116: We assume the 3-manifolds $\Sigma_t$ are spherically symmetric, for
117: which  a general metric \emph{ansatz}  is
118: \begin{equation}\label{eq:metric ansatz}
119:     ds^2 = e^{2X(t,\rho)}[ d\rho^2 + R^2(t,\rho)d\Omega^2 ].
120: \end{equation}
121: Spherical symmetry implies the following form for the DeTurck field:
122: \begin{equation}\label{eq:vector ansatz}
123:     V^a = V(t,\rho) \di_\rho.
124: \end{equation}
125: When (\ref{eq:metric ansatz}) and (\ref{eq:vector ansatz}) are put
126: into (\ref{eq:flow equation 1}) we obtain two independent flow
127: equations for $\di_t X(t,\rho)$ and $\di_t R(t,\rho)$.  There is no
128: dynamical equation for $V(t,\rho)$, so the system is
129: underdetermined, which reflects the coordinate freedom embodied by
130: the DeTurck field.  To close the system, we need to fix coordinate
131: gauges. There are three choices we have investigated:
132: \begin{enumerate}
133:     \item\label{case:1} $V(t,\rho) \equiv 0$;
134:     \item\label{case:2} $R(t,\rho) \equiv \rho$ (conformally flat gauge); or
135:     \item\label{case:3} $X(t,\rho) \equiv 0$ (areal radius gauge).
136: \end{enumerate}
137: Enforcement of condition (\ref{case:2}) or (\ref{case:3}) in the
138: flow equations (\ref{eq:flow equation 1}) results in a single
139: dynamical equation for $X(t,\rho)$ or $R(t,\rho)$, respectively, and
140: an equation of constraint for $V(t,\rho)$.  The choice
141: (\ref{case:1}) gives two dynamical equations for both metric
142: functions.
143: 
144: For our purposes, a wormhole geometry is defined by a metric
145: (\ref{eq:metric ansatz}) that has two asymptotically flat regions as
146: $\rho \rightarrow \pm \infty$.  The combination $e^{X(t,\rho)}
147: R(t,\rho)$ should be non-zero for all $\rho$, and the wormhole
148: throat (or throats) are located at local minima $\rho =
149: \rho_\mathrm{th}$ of this function\footnote{If we interpret the
150: 3-geometries $\Sigma_t$ as being embedded in 4-dimensional
151: Lorentzian spacetimes as moments of time symmetry, each wormhole
152: throat represents an apparent horizon.  However, if the $\Sigma_t$
153: are embedded with non-zero extrinsic curvature, such an
154: interpretation is not possible.}
155: \begin{equation}
156:     \di_\rho(e^{X} R)\big|_{\rho = \rho_\mathrm{th}} = 0,
157:     \quad \di^2_\rho (e^{X} R)\big|_{\rho = \rho_\mathrm{th}} > 0.
158: \end{equation}
159:  The definition of the wormhole throat suggests a convenient
160: coordinate gauge choice is given by case (\ref{case:3}) above, with
161: line element
162: \begin{equation}\label{eq:areal radius gauge}
163:     ds^2 = d\rho^2 + R^2(t,\rho) \, d\Omega^2.
164: \end{equation}
165: In this gauge the area of 2-spheres of constant $\rho$ is $4\pi
166: R^2(t,\rho)$, hence the name ``areal radius'' gauge. The condition
167: $X \equiv 0$ reduces the flow equations (\ref{eq:flow equation 1})
168: to
169: %
170: \numparts\label{eq:main equations}
171: \begin{eqnarray}\label{eq:flow equation 2}
172:     \di_t R = \di^2_\rho R + \frac{(\di_\rho R)^2}{R} - \frac{1}{R}
173:     + V \di_\rho R, \\ \label{eq:constraint equation}
174:     \di_\rho V = -2\frac{\di_\rho^2 R}{R}.
175: \end{eqnarray}
176: \endnumparts
177: As mentioned above, the latter is a constraint equation. These
178: equations must be supplemented by boundary conditions and initial
179: data.  In the areal radius gauge, asymptotic flatness requires that
180: \begin{equation}\label{eq:asymptotic BC 1}
181:     R \sim |\rho| + \mathcal{O}\left(l_1 \ln \left(|\rho|/l_2\right) \right),
182: \end{equation}
183: where $l_1$ and $l_2$ are two length scales. Putting this asymptotic
184: limit into (\ref{eq:flow equation 2}) and (\ref{eq:constraint
185: equation}), we obtain
186: \begin{equation}\label{eq:asymptotic BC 2}
187:     V \sim \mathcal{O}\left( |\rho|^{-2} \right).
188: \end{equation}
189: The asymptotic behaviour of $R$ suggests that we search for
190: solutions that are  even functions of $\rho$.  We then only need to
191: consider the interval $\rho \in [0,\infty)$, and we have the
192: additional boundary condition
193: %
194: \begin{equation}\label{eq:Neumann}
195:     \di_\rho R \big|_{\rho = 0} = 0.
196: \end{equation}
197: %
198: The boundary conditions imply that there is always a local extrema
199: of $R$ at $\rho = 0$. If this is a minima, then there is a throat at
200: $\rho = 0$.  To complete the specification of the flow, we need to
201: give initial data for $R$ at $t = 0$.  The various classes of
202: initial data considered in this work are described in
203: \S\ref{sec:Morris-Thorne} and \S\ref{sec:bubbles}.
204: 
205: Finally, we note that it is possible to embed the 3-geometries
206: (\ref{eq:areal radius gauge}) in 4-dimensional flat space
207: \begin{equation}
208:     ds^2 = dR^2 + dZ^2 + R^2 d\Omega^2,
209: \end{equation}
210: via the parametric equations
211: \begin{equation}
212:     R = R(t,\rho), \quad Z = Z(t,\rho), \quad \di_\rho Z =
213:     \sqrt{1-(\di_\rho R)^2}.
214: \end{equation}
215: These embeddings are only possible for areal radius functions
216: satisfying $|\di_\rho R| < 1$.  We use these as a tool for the
217: visualization of Ricci flow initial data in Figs. \ref{fig:embedding
218: 1} and \ref{fig:embedding 2}.
219: 
220: \section{Numerical method}\label{sec:numerical}
221: 
222: We solve the equations (\ref{eq:flow equation 2}) and
223: (\ref{eq:constraint equation}) numerically using finite difference
224: methods.  The computational domain is defined by $\rho \in
225: [0,\rho_\mathrm{max}]$ and $t \in [0,t_\mathrm{max}]$.  The spatial
226: and temporal intervals are discretized into elements of size
227: $\delta\rho$ and $\delta t$, respectively.  The notation $A_i^j$
228: refers to the value of the quantity $A(t,\rho)$ at the
229: $i^\mathrm{th}$ spatial and $j^\mathrm{th}$  temporal node.  The
230: Neumann boundary condition (\ref{eq:Neumann}) is imposed explicitly,
231: while the asymptotic boundary conditions (\ref{eq:asymptotic BC 1})
232: and (\ref{eq:asymptotic BC 2}) are replaced with the Dirichlet
233: conditions
234: \begin{equation}
235:     R(t,\rho_\mathrm{max}) = R_\mathrm{max}, \quad
236:     V(t,\rho_\mathrm{max}) = 0.
237: \end{equation}
238: $R_\mathrm{max}$ is fixed by the \emph{asymptotically flat} initial
239: data. With this, the former condition effectively gives $dR/d\rho
240: \sim 1$ for large $\rho$ in our simulations, while the latter leads
241: to a non-zero value of  $V$ at the origin $\rho=0$.\footnote{This
242: means that data symmetric about the origin evolves asymmetrically.
243: Since our computational domain is $\rho \in [0,\rho_\mathrm{max}]$,
244: the other side of the wormhole (ie. $\rho \in
245: [-\rho_\mathrm{max},0])$ can be glued continuously using the
246: junction conditions with a jump in $V$ at $\rho=0$, which represents
247: the freedom to choose different coordinates on each side of the
248: join.}  For $\rho_\mathrm{max}$ sufficiently large compared to
249: $t_\mathrm{max}$, the simulation results are insensitive to the
250: position of the outer boundary.  In practice, our simulations are
251: performed with $R_\mathrm{max} \sim \rho_\mathrm{max} \ge 200$.
252: 
253: The two types of finite differencing schemes used to approximate
254: derivatives are summarized in Table \ref{tab:schemes}. When applied
255: to linear parabolic systems, the simple Euler (SE) method is known
256: to be conditionally stable when $\delta t / \delta \rho^2 \lesssim
257: 1$.  Conversely, the modified Dufort-Frankel (MDF) method is known
258: to be unconditionally stable.  Via direct experimentation, we have
259: found that these conclusions also hold for the nonlinear problem
260: given by equations (\ref{eq:flow equation 2}) and
261: (\ref{eq:constraint equation}), which would na\"{\i}vely suggest
262: that the MDF method is vastly superior to the SE method due to its
263: excellent stability.
264: %
265: \begin{table}
266: \begin{tabular}{c|cc}
267:     \hline  & Simple Euler (SE) & Modified Dufort-Frankel (MDF) \\
268:     \hline $\di_\rho A$ & $(A^j_{i+1}-A^j_{i-1})/(2\,\delta \rho)$ &
269:     $(A^j_{i+1}-A^j_{i-1})/(2\,\delta\rho)$ \\ $\di^2_\rho R$ in eq.~(\ref{eq:flow equation 2}) & $(R^j_{i+1} -
270:     2R^j_i + R^j_{i-1})/{\delta \rho^2}$ & $ (R^j_{i+1} - R^{j+1}_i -
271:     R^{j-1}_i + R^j_{i-1})/{\delta \rho^2}$
272:     \\ $\di^2_\rho R$ in eq.~(\ref{eq:constraint equation}) & $(R^j_{i+1} -
273:     2R^j_i + R^j_{i-1})/{\delta \rho^2}$ & $(R^j_{i+1} -
274:     2R^j_i + R^j_{i-1})/{\delta \rho^2}$ \\ $\di_t R$ & $(R^{j+1}_{i}-R^j_{i})/\delta t$ &
275:     $(R^{j+1}_{i}-R^{j-1}_{i})/(2\,\delta t)$ \\ accuracy &
276:     $\mathcal{O}(\delta t, \delta\rho^2)$ & $\mathcal{O}(\delta t^2,
277:     \delta \rho^2)$  \\ \hline
278: \end{tabular}
279: \caption{Summary of the simple Euler (SE) and modified
280: Dufort-Frankel (MDF) finite difference derivative approximations. In
281: the top line, $A$ stands for either $R$ or $V$.  Notice that for the
282: MDF method, the second-order spatial derivative of $R$ is handled
283: differently in the flow (\ref{eq:flow equation 2}) and constraint
284: (\ref{eq:constraint equation}) equations.}\label{tab:schemes}
285: \end{table}
286: %
287: However, there is a non-trivial price to be paid for this attractive
288: feature. One can show that the numerical results obtained with the
289: MDF approximation actually converge to solutions of
290: \begin{equation}\label{eq:flow equation 3}
291:     \di_t R = \di^2_\rho R + \frac{(\di_\rho R)^2}{R} - \frac{1}{R}
292:     + V \di_\rho R - \left(\frac{\delta t}{\delta \rho} \right)^2 \di_t^2
293:     R,
294: \end{equation}
295: rather than the solutions of the original flow equation
296: (\ref{eq:flow equation 2}).\footnote{The origin of the anomalous
297: time derivative on the r.h.s. of (\ref{eq:flow equation 3}) is the
298: MDF approximation of $\di_\rho^2 R$ in (\ref{eq:flow equation 2}).
299: Essentially, the MDF prescription approximates $R_i^j$ by its
300: \emph{temporal} average $(R^{j+1}_i + R^{j-1}_i)/2$.  The use of
301: nonlocal time data to resolve the spatial derivative gives rise to
302: the extra $\di_t^2 R$ term.}  In other words, the MDF method solves
303: a modified version of the original system of PDEs. In order to
304: minimize the discrepancy between MDF results and the ``true''
305: solutions of (\ref{eq:flow equation 2}), one must make the ratio
306: $\epsilon \equiv \delta t/ \delta \rho$ as small as is
307: computationally feasible.
308: 
309: The numerical results presented in this paper are obtained using the
310: MDF method with $\epsilon = 10^{-4}$.  For several individual cases,
311: we have checked that the simulation results are insensitive to the
312: particular choice of $\delta t$ and $\delta \rho$ provided that
313: $\epsilon \lesssim 10^{-3}$, and that the MDF and SE schemes give
314: virtually identical answers for $\epsilon = 10^{-4}$.
315: 
316: \section{Evolution of Morris-Thorne wormhole
317: geometries}\label{sec:Morris-Thorne}
318: 
319: The Morris-Thorne wormholes \cite{mt-worms} have spatial sections
320: \begin{equation}
321:     ds^2 = \frac{1}{f(r)} dr^2 + r^2 d\Omega^2, \quad f(r) = 1 -
322:     \frac{b(r)}{r}.
323: \end{equation}
324: In this section, we will use these metrics as initial data for the
325: Ricci flow.  At the initial time $t = 0$, this 3-metric can be
326: transformed into the areal radius gauge (\ref{eq:areal radius
327: gauge}) by the coordinate transformation $r = R(0,\rho)$ with
328: \begin{equation}
329:     \di_\rho R(0,\rho) = \sqrt{f(R(0,\rho))},
330: \end{equation}
331: where $R(0,\rho)$ is the initial profile for the areal radius metric
332: function. We will concentrate on a subset of this class defined by
333: \begin{equation}
334:     b(r) = \frac{r_0^\alpha}{r^{\alpha - 1}}, \quad f(r) = 1 -
335:     \left( \frac{r_0}{r} \right)^\alpha,
336: \end{equation}
337: where $r_0$ is a length scale and $\alpha$ is a dimensionless
338: parameter (a similar family of wormholes was defined in
339: \cite{Lobo:2005us}). The spatial sections of the Schwarzschild
340: metric (known as the Einstein-Rosen bridge) correspond to $\alpha =
341: 1$. Another special case is the Ellis wormhole with $\alpha = 2$
342: \cite{ellis}, which has $R(0,\rho) = \sqrt{\rho^2 + r_0^2}$.
343: Embedding diagrams of both types of wormholes appear in Figure
344: \ref{fig:embedding 1}. For Ricci flow of these wormholes, it is
345: useful to work with dimensionless quantities defined by
346: \begin{equation}\label{eq:dimensionless}
347:     \hat\rho = \rho/r_0, \quad \hat{t} = t/r_0^2, \quad \hat{R}
348:     \rightarrow R/r_0, \quad \hat{V} \rightarrow r_0 V.
349: \end{equation}
350: When the hatted-variables are substituted into the flow equations
351: the $r_0$ length scale drops out of the problem; i.e., our
352: simulation results are essentially independent of $r_0$ up to
353: trivial scalings.
354: \begin{figure}
355:     \begin{center}
356:     \includegraphics[scale=0.775]{fig1}
357:     \end{center}
358:     \caption{Embedding diagrams of the Einstein-Rosen and
359:     Ellis wormholes in flat space (we have set $\theta = \pi/2$ to obtain a
360:     3-dimensional plot).}\label{fig:embedding 1}
361: \end{figure}
362: \begin{figure}
363:     \begin{center}
364:     \includegraphics[scale=0.775]{fig2}
365:     \end{center}
366:     \caption{Evolution of the areal radius function $R(t,\rho)$
367:     under the Ricci flow for Morris-Thorne wormhole initial data
368:     corresponding to $\alpha = 1$ and $\alpha = 2$.  Note that our
369:     boundary conditions ensure that the wormhole throat is always at $\rho = 0$
370:     for these simulations.}\label{fig:explicit evolution}
371: \end{figure}
372: 
373: In Figure \ref{fig:explicit evolution}, we show the results of our
374: simulations for the evolution of the areal radius function
375: $R(t,\rho)$ for the Einstein-Rosen and Ellis cases. One can see in
376: Figure \ref{fig:embedding 1} that the initial data for the two
377: scenarios are visually quite similar. However, the evolution of $R$
378: is entirely different: in the Ellis case the areal radius tends to
379: diverge for all $\rho$ as $t \rightarrow \infty$, while for the
380: Einstein-Rosen bridge $R$ is everywhere decreasing and the wormhole
381: throat pinches off in finite time.  Due to computing limitations, we
382: cannot determine the ultimate fate of the Ellis initial data, but it
383: appears to be a warped cylinder with a monotonically increasing
384: radius.
385: 
386: In Figure \ref{fig:throat evolution}, we plot the evolution of the
387: throat radius $R_\mathrm{th}(t) = R(t,0)$ for various values of
388: $\alpha$.  We find that the wormhole throat pinches off for initial
389: data with $\alpha$ less than a critical value
390: $\alpha_\mathrm{crit}$.  For cases with $\alpha >
391: \alpha_\mathrm{crit}$, the throat approaches an ever-expanding state
392: after a brief initial period of contraction. Numerically, we have
393: determined the value of the critical parameter to be approximately
394: 1.259.  For $\alpha > \alpha_\mathrm{crit}$, the throat radius
395: appears to grow like $\sqrt{t}$ at late times.
396: \begin{figure}
397:     \begin{center}
398:     \includegraphics[scale=0.775]{fig3}
399:     \end{center}
400:     \caption{Ricci flow evolution of the throat radius $R_\mathrm{th}(t) = R(t,0)$
401:     for various Morris-Thorne wormhole initial data.}
402:     \label{fig:throat evolution}
403: \end{figure}
404: 
405: \section{Evolution of ``bubble'' geometries}\label{sec:bubbles}
406: 
407: A geometrically interesting question is: What does the Ricci flow
408: evolution of a bubble attached to asymptotically flat space look
409: like? In the areal radius gauge, initial data for this kind of
410: problem would roughly be
411: \begin{equation}\label{eq:bubble data 1}
412:     R(0,\rho) \sim r_0
413:     \cases{ \sin \rho / r_0 & for $0 \le \rho \lesssim \rho_0$ \\
414:     \rho/r_0 & for $\rho \gtrsim \rho_0$ },
415: \end{equation}
416: where $\rho_0 < \pi r_0$ is some transition radius.  Notice that
417: this initial data has $R = 0$ and $\di_\rho R \ne 0$ at $\rho = 0$,
418: so it violates the boundary conditions that our code was designed to
419: deal with.  We therefore consider the related problem of a bubble
420: attached to two asymptotically flat regions, and leave initial data
421: of the form (\ref{eq:bubble data 1}) for future work. We take
422: initial data for the current problem to be
423: \begin{equation}
424:     R(0,\rho) =
425:     \cases{ r_0 (1-\rho^2/r_0^2) & for $|\rho| < \rho_0$ \\
426:     \sqrt{(|\rho| - \rho_1)^2 + \rho_2^2}  & for $|\rho| \ge \rho_0$
427:     }.
428: \end{equation}
429: Here, $\rho_1$ and $\rho_2$ are selected such that both $R$ and
430: $\di_\rho R$ are continuous at the transition points $\rho = \pm
431: \rho_0$.  Defining the dimensionless parameter $\beta \equiv
432: \rho_0/r_0$, we show embedding diagrams for two examples of this
433: type of initial data in Figure \ref{fig:embedding 2}.  Note that
434: there are two wormhole throats (local minima of $R$) in this case at
435: $\rho = \pm \rho_1$, while there is a local maxima of $R$ at $\rho =
436: 0$. This maxima corresponds to the equator of our bubble, and it is
437: easy to show that the ratio of the throat radii to the equator
438: radius is
439: \begin{equation}
440:     {R_\mathrm{th}}/{R_\mathrm{eq}} = \rho_2/r_0 = (1-\beta^2)
441:     \sqrt{1-4\beta^2}.
442: \end{equation}
443: Hence, in order to have sensible initial data with $R > 0$ we are
444: obliged to take $0 \le \beta \le 1/2$.
445: \begin{figure}
446:     \begin{center}
447:     \includegraphics[scale=0.775]{fig4}
448:     \end{center}
449:     \caption{Embedding diagrams of 3-geometries representing a bubble
450:     connecting two asymptotically flat regions}\label{fig:embedding 2}
451: \end{figure}
452: 
453: Making use of the dimensionless quantities (\ref{eq:dimensionless}),
454: we have simulated the Ricci flow evolution of various types of
455: bubble geometries.  As in \S\ref{sec:Morris-Thorne} above, we take
456: $R$ to be an even function of $\rho$ and impose the boundary
457: condition that $\di_\rho R = 0$ for $\rho = 0$.  We find that for
458: $R_\mathrm{th}/R_\mathrm{eq}$ initially small, the bubbles disappear
459: and the geometry ``pinches-off'' into a pair of disjoint
460: asymptotically flat regions. For $R_\mathrm{th}/R_\mathrm{eq}$
461: initially close to unity, the bubbles also disappear and the areal
462: radius blows up, somewhat similar to the $\alpha >
463: \alpha_\mathrm{crit}$ cases in \S\ref{sec:Morris-Thorne}. Explicit
464: examples of these behaviours are shown in Figure \ref{fig:bubble
465: evolution}.
466: \begin{figure}
467:     \begin{center}
468:     \includegraphics[scale=0.775]{fig5}
469:     \end{center}
470:     \caption{The evolution of two bubble geometries under the Ricci flow.}
471:     \label{fig:bubble evolution}
472: \end{figure}
473: 
474: \section{Conclusions}
475: 
476: We have observed that the Ricci flow of certain  families  of
477: rotationally symmetric wormhole geometries exhibit a form of
478: critical behaviour:  wormhole throats pinch off or expand forever
479: depending on the values of initial data parameters. The expanding
480: solutions are perhaps surprising at first sight, but are
481: nevertheless understandable in that they appear to converge to an
482: infinite volume cylindrical attractor.
483: 
484: This work  complements both the analytic work on the flow of
485: asymptotically flat geometries \cite{ow-asympflat} and the numerical
486: work on corsetted  3--spheres  \cite{gi}.  It would be of interest
487: to see if the neck pinching seen in our  simulations of wormhole
488: throats may be modeled by the Bryant steady solitons, as has been
489: observed for rotationally symmetric geometries on $S^3$ \cite{gi2}.
490: 
491: The methods we have used have applicability to a variety of
492: situations for relevance to string theory, where the renormalization
493: group flows give matter coupled modifications of Ricci flow. The
494: interest here is in discovering new fixed points of physical
495: relevance. For example, in special cases with compact target spaces
496: and  a tachyon field, there is evidence that properties of the
497: tachyon potential are linked to the existence of non-trivial fixed
498: points \cite{gs-tachyon}. It  is of interest to extend such cases to
499: non-compact geometries with gauge or 2-form fields, and consider
500: questions such as the possibility of stabilization of  wormholes
501: geometries.
502: \medskip
503: 
504: \noindent{\bf Acknowledgements} We would like to thank Eric Woolgar
505: and Jack Gegenberg for discussions, David Hobill for suggesting that
506: we try the Dufort-Frankel method, and David Garfinkle for comments
507: on the manuscript. This work was supported by NSERC.
508: 
509: \section*{References}
510: \begin{thebibliography}{10}
511: %
512: \bibitem{hamilt}  R. S. Hamilton, {\it Three manifolds of Positive Ricci curvature}, J. Diff. Geom. {\bf 17} 255-306 (1982).
513: 
514: %
515: \bibitem{friedan}  Friedan D 1980 {\it Phys. Rev. Lett}. {\bf  45}  1057
516: 
517: %
518: \bibitem{sen}  Sen Ashoke 1985 {\it Phys. Rev.}  {\bf D32}  2102;  1985 {\it Phys. Rev. Lett. }{\bf 55} 1846.
519: 
520: %
521: \bibitem{perel}  Perelman G 2003 math/0211159; math/0307245.
522: 
523: %
524: \bibitem{cao-chow} Cao H D and Chow B 1999 {\it Recent developments on Ricci flow} Bull. Am. Math. Soc. {\bf 36} 59-74.
525: 
526: %
527: \bibitem{ow-asympflat} Oliynyk T and  Woolgar E, {\it Asymptotically Flat Ricci Flows},
528: Commun. in Analysis and Geometry 15 (2007) 535--568;
529: arXiv:math/0607438v2 [math.DG].
530: 
531: %
532: \bibitem{dai-ma} Dai X and Ma L {\it Mass under the Ricci flow}, arXiv:math/0510083v1 [math.DG].
533: 
534: %
535: \bibitem{gi} Garfinkle D and Isenberg J 2002 {\it Critical behaviour in Ricci flow}
536:     arXiv: math/0306129v1 [math.DG].
537: 
538: %
539: \bibitem{sam} Samuel J and Chowdhry S R  2008 {\it Class. Quant. Grav. } {\bf 25} 035012.
540: 
541: %
542: \bibitem {hw} Hedrick M and Wiseman T 2006 {\it Class. Quant. Grav.}  {\bf 23} 6683.
543: 
544: \bibitem{deturck} DeTurck D 1983 {\it J. Differential Geom.} {\bf 18} 157.
545: 
546: %
547: \bibitem{mt-worms}  Morris M and Thorne K S ,  Am. J. Phys. {\bf 56} 395-412 (1988).
548: 
549: \bibitem{Lobo:2005us}
550:   Lobo F (2005) Phys.\ Rev.\ D {\bf 71} 084011
551: 
552: \bibitem{ellis} Ellis  H G (1973) \emph{J. Math. Phys.} {\bf 14} 104
553:  %
554: \bibitem{gi2} Garfinkle D and Isenberg J {\it Modelling of Degenerate Neck Pinch Singularities in Ricci Flow by Bryant Solitons},  arXiv: 0709.0514.
555: 
556: %
557: \bibitem{gs-tachyon} Gegenberg J and Suneeta V,  JHEP 0609: 045, 2006; arXiv: hep-th/0605230.
558: 
559: 
560: 
561: 
562: \end{thebibliography}
563: \end{document}
564: