1: \documentclass[prd,twocolumn,showpacs,floatfix]{revtex4}
2: \usepackage{bm}
3: \usepackage{amssymb}
4: \begin{document}
5: \title{Survival of the black hole's Cauchy horizon under non-compact
6: perturbations}
7: \author{Lior M.\ Burko}
8: \affiliation{Department of Physics, University of Utah, Salt Lake City,
9: Utah 84112}
10: \date{June 4, 2002}
11: \begin{abstract}
12: We study numerically the evolution of spactime, and in particular of a
13: spacetime singularity, inside a black hole under a class of perturbations
14: of non-compact support. We use a very simplified toy model of a spherical
15: charged black hole which is perturbed nonlinearly by a self-gravitating,
16: spherical scalar field. The latter grows logarithmically with
17: advanced time along an outgoing characteristic hypersurface. We find that
18: for that class of perturbations a portion of the Cauchy horizon survives
19: as a non-central, null singularity.
20: \end{abstract}
21: \pacs{04.70.Bw, 04.20.Dw}
22: \maketitle
23:
24:
25: {\it Introduction and summary}: The geometrical and physical properties of
26: the Cauchy horizon singularity inside black holes have received much
27: attention \cite{review}. That singularity was shown to be null,
28: non-central, and weak. The weak nature of the Cauchy horizon singularity
29: has far-reaching implications. In particular, it leaves open the
30: possibility that physical objects which fall into a black hole may
31: traverse the Cauchy horizon singularity only mildly affected, and
32: re-emerge in another universe.
33:
34: The evolution of spacetime geometry into a (weak) curvature singularity at
35: the Cauchy horizon has been studied both numerically and analytically
36: \cite{poisson-israel90,ori91,ori92,brady-smith95,burko97,burko-ori98,
37: burko99a,remark}. In all these studies the black hole was taken to be
38: isolated, and the source of the perturbations was taken to be the
39: perturbations which result from the evolution of nonvanishing multipole
40: moments during the collapse. These perturbations are inherent to any
41: nonspherical gravitational collapse, and result from the backscattering of
42: waves, which are created during the collapse, off the curvature of
43: spacetime \cite{price}. These perturbations have a compact support at some
44: initial time.
45:
46: It is interesting to ask whether the evolution of a null and weak
47: singularity at the Cauchy horizon is just an artifact of the assumption of
48: compactness. That is, will any dominating perturbation field which has
49: non-compact
50: support on the initial time slice lead to the full destruction of the null
51: singularity, and its replacement by a spacelike one? In fact, the
52: perturbations due to the collapse and the resulting tails can be thought
53: of as a lower bound on the perturbation field. It is intersting to
54: ask what happens to the Cauchy horizon if perturbations which are stronger
55: than that lower bound are present.
56:
57: This question is interesting not just from the mathematical
58: viewpoint: indeed, a generic class of perturbations exists, where the
59: perturbation field has non-compact support. These are the perturbations
60: which arise from the capture of photons which originate from the relic
61: cosmic background radiation (CBR). Even if removed from any conceivable
62: astrophysical object, any black hole is still perturbed by the CBR. Unlike
63: the perturbations due to the collapse, the CBR perturbations are
64: non-compact. Because the perturbation field is greater than the lower
65: bound set by the perturbations due to the tails, one may suspect that the
66: evolution of spacetime, and in particular of the singularity, be
67: dominated by the non-compact perturbations, rather than by the compact
68: ones.
69:
70: If this is indeed the case, the non-compact perturbations threaten
71: to change our notions of the causal structure inside black holes. When the
72: perturbations due to the tails are considered, it is found that the
73: weakness of the Cauchy horizon singularity is a rather delicate issue: it
74: depends on certain integrals being bounded. Specifically, assume that
75: the field is due to the tails. Then, on
76: the event horizon the scalar field $\Phi$ behaves like $\Phi=(\kappa
77: v)^{-n}$, where $\kappa$ is a constant, $v$ is advanced time,
78: and $n$ is a positive integer which is related to the multipole
79: moment of the perturbation field. Denoting
80: schematically by ${\cal R}$ the fastest growing components of the
81: Riemann-Christoffel tensor approaching the Cauchy horizon, the curvature
82: at that limit behaves like ${\cal R}\approx
83: \tau^{-2}[-\ln(-\tau)]^{2n+2}$, where $\tau$ is proper time along a
84: timelike geodesic which is set equal to zero on the Cauchy horizon. The
85: Cauchy
86: horizon singularity is weak, if ${\cal R}$ is twice-integrable. (This last
87: statement can be made precise \cite{clarke-krolak}.) For positive value of
88: $n$ this is indeed the case. However, small changes in ${\cal R}$, e.g.,
89: another factor of $\tau^{-\epsilon}$ for any small and positive $\epsilon$,
90: would change the picture entirely, as ${\cal R}$ would no longer be twice
91: integrable. That is, the twice integrability of ${\cal R}$ is strongly
92: dependent on the form of the field at the event horizon. The twice
93: integrability of ${\cal R}$ (and consequently the weakness of the Cauchy
94: horizon singularity) depends then on the assumption that the scalar field
95: has a compact support (as this condition leads, through Price's analysis
96: \cite{price}, to the tail form for the field on the event horizon).
97: Dominating non-compact perturbations threaten to change ${\cal R}$ in a
98: significant way, such that it would no longer be twice integrable.
99: One can ask then the following question: Is the Cauchy horizon necessarily
100: utterly destroyed and replaced by a spacelike singularity when
101: perturbations with non-compact support are present, or can it still
102: survive (as a null, weakly singular hypersurface) also when perturbations
103: of non-compact support are present?
104:
105: In this paper we shall answer the
106: latter question in the affirmative. We show that a certain class of
107: non-compact perturbations still preserves the null, non-central nature of
108: the Cauchy horizon singularity, even though the evolution of geometry and
109: of the singularity is indeed dominated by the non-compact
110: perturbations. We emphasize that any perturbation field with non-compact
111: support (whose dynamics dominates deep inside the black hole) is
112: appropriate, as it serves as a counter-example for the claim that only
113: perturbations with compact support can evolve into a null and weak
114: singularity. The survival of the Cauchy horizon as a null, non-central
115: singularity can occur also when perturbations of non-compact support
116: (of certain classes) are present.
117:
118: {\it Model}: In this paper we study the evolution of spacetime curvature
119: inside a black hole in the presence of perturbations which have
120: non-compact support under a very simplified toy model. For simplicity, we
121: take the black hole to be spherically symmetric, and to have a fixed
122: electric charge $Q$. This is a useful toy model for a spinning black hole,
123: because the unperturbed spacetimes, namely the Reissner-Nordstr\"{o}m and
124: Kerr spacetimes, respectively, have very similar causal structures, which
125: lead to similar blue-sheet effects near their inner horizons. In
126: fact, much of the understanding we currently have about black hole
127: interiors have been obtained through the study of spherical charged
128: models. (One important difference is that the null singularity inside a
129: spherical charged black hole is monotonic, whereas the one inside a
130: spinning black hole is oscillatory \cite{ori99}. This difference is not
131: crucial for our purposes here. Another difference is related to the
132: question of the occurrence of a spacelike singularity inside black
133: holes. A spacelike singularity to the future of the Cauchy horizon
134: singularity was found in spherically-symmetric, charged models. It has
135: been argued that no corresponding spacelike singularity is likely to occur
136: inside a rotating black hole \cite{ori99}. Others have argued, that a
137: spacelike singularity, possibly of the Belinskii-Khalatnikov-Lifshitz
138: type, is a possible outcome. While this open question is extremely
139: important, it is unrelated to the nature of the null singularity which
140: precedes the spacelike one, if such a spacelike singularity exists.)
141:
142:
143: We write the spherically-symmetric metric in double-null
144: coordinates in the form
145: \begin{equation}
146: \,ds^2=-2e^{2\sigma (u,v)}\,du\,dv+r^2(u,v)\,d\Omega^2 \,
147: \label{metric}
148: \end{equation}
149: where $\,d\Omega^2$ is the line element on the unit two-sphere.
150: As the source term for the Einstein equations, we take the contributions
151: of both the scalar field $\Phi$ and the (sourceless) spherical electric
152: field (see \cite{burko-ori97} for details). The dynamical equations are
153: the scalar field equation $\nabla_{\mu}\nabla^{\mu}\Phi=0$ and the
154: Einstein equations, which
155: reduce to
156: \begin{eqnarray}
157: \Phi_{,uv}+\frac{1}{r}\left(r_{,u}\Phi_{,v}+r_{,v}\Phi_{,u}\right)=0
158: \label{KGEQ}
159: \end{eqnarray}
160: \begin{eqnarray}
161: r_{,uv}+\frac{r_{,u}r_{,v}}{r}+\frac{e^{2\sigma}}{2r}\left(1-\frac{Q^{2}}
162: {r^{2}}\right)=0
163: \label{EEQ1}
164: \end{eqnarray}
165: and
166: \begin{eqnarray}
167: \sigma_{,uv}-\frac{r_{,u}r_{,v}}{r^2}-\frac{e^{2\sigma}}{2r^2}
168: \left(1-2\frac{Q^{2}}{r^{2}}\right)+
169: \Phi_{,u}\Phi_{,v}=0 \; .
170: \label{EEQ2}
171: \end{eqnarray}
172: These equations are
173: supplemented by the two constraint equations
174: \begin{eqnarray}
175: r_{,uu}-2\sigma_{,u}r_{,u}+r(\Phi_{,u})^{2}=0
176: \label{con1}
177: \end{eqnarray}
178: \begin{eqnarray}
179: r_{,vv}-2\sigma_{,v}r_{,v}+r(\Phi_{,v})^{2}=0.
180: \label{con2}
181: \end{eqnarray}
182: Although similar, there is an important difference between the numerical
183: evolution of a code based on Eqs.~(\ref{KGEQ})-(\ref{EEQ2}) and a code
184: which is based on the dynamical equations used in
185: Ref.~\cite{burko-ori97}: In the latter case, the wave equation
186: for the $g_{uv}$ metric function becomes a free wave equation
187: asymptotically close to the Cauchy horizon. (Notice that $g_{uv}$ vanishes
188: exponentially in $v$ near the Cauchy horizon, whereas
189: $r_{,u},r_{,v},\Phi_{,u}$ and $\Phi_{,v}$ decay like inverse powers of
190: $v$, where $v$ is proportional to advanced time -- see below.) This
191: implies that the numerical
192: integration becomes inaccurate near the Cauchy horizon because
193: dynamically-important terms become negligible. When the field equations
194: are written as Eqs.~(\ref{KGEQ})-(\ref{EEQ2}) this problem does not occur.
195:
196:
197: {\it Initial value problem}: From the pure initial-value viewpoint, we
198: need to specify three initial functions on each segment of the initial
199: surface: $r$, $\sigma$, and $\Phi$. The constraint equations reduce this
200: number: Eqs.~(\ref{con1}) and (\ref{con2}) impose one constraint each on
201: the initial data at $u=u_i$ and $v=v_i$, respectively. The remaining two
202: initial functions, however, represent only one physical degree of
203: freedom: The other degree of freedom expresses nothing but the gauge
204: freedom associated with the arbitrary coordinate transformation $u\to
205: \tilde u(u)\;,\;v\to \tilde v(v)$. In what follows we shall use a standard
206: gauge, in which $r$ is linear with $v$ or $u$, correspondingly, on the two
207: initial null segments. On the outgoing segment we take
208: $r_{,v}=1$. (Notice, that this implies that this $v$ is twice advanced
209: time at late times.) On the
210: ingoing segment, we take $r_{,u}={\rm const}\equiv r_{u0}$. (Notice that
211: $r_{u0}<0$.) The initial values of $r$ are thus uniquely determined by
212: the parameter $r_0\equiv r(u_i,v_i)$. We choose $u_{i}=0$ and
213: $v_{i}=r_{0}$, and thus we find: $r_v(v)=v$, and $r_u(u)=r_0+u
214: r_{u0}$. [Hereafter, we denote the initial values of the three fields
215: $r,\sigma,\Phi$ on the two segments of the characteristic hypersurface by
216: $r_v(v),\sigma_v(v),\Phi_v(v)$ and $r_u(u),\sigma_u(u),\Phi_u(u)$,
217: correspondingly.] Then, we can freely specify $\Phi_u(u)$ and $\Phi_v(v)$
218: (this choice represents a true physical degree of freedom). The initial
219: value of $\sigma$ is now determined from the constraint equations, namely
220: \begin{eqnarray}
221: \sigma_{u,u}=r_u(\Phi_{u,u})^{2}/(2r_{u0})\;\;,\;\;
222: \sigma_{v,v}=r_v(\Phi_{v,v})^{2}/2\;,
223: \label{initf}
224: \end{eqnarray}
225: together with the choice $\sigma(u_i,v_i)=-(1/2)\ln 2$. Thus, in the gauge
226: we use, we need to specify two functions of one variable [$\Phi_v(v)$ and
227: $\Phi_u(u)$] and two parameters ($r_0$ and $r_{u0}$) (in addition to
228: the charge $Q$) for the characteristic initial value problem.
229:
230: {\it Determination of characteristic data}: We take the characteristic
231: data to satisfy $\Phi_v(v)=(A/\sqrt{-2r_{u0}})\ln (v/v_i)$ along $u=u_i$.
232: Here, $A$ is a real constant, which is related to the amplitude of the
233: perturbation field.
234: This choice for the scalar field is clearly of noncompact
235: support, as the field grows logarithmically in advanced time. (Notice,
236: that this implies that spacetime is not asymptotically flat.)
237: In addition,
238: we require that $\Phi_{u,u}(u)=0$ on $v=v_i$, such that the field does not
239: propagate outside the event horizon on the ingoing segment of the
240: characteristic hypersurface. We also require that
241: $\Phi$ is continuous at $(u_i,v_i)$.
242:
243: We note that is it unimportant what the field is along $v=v_i$ and on
244: $u=u_i$ for $v_i<v<v_f$ for any finite $v_f$: Any perturbation
245: field with compact support leads to power-law tails at late times
246: regardless of the specific shape of the initial data. It is only the
247: contributions of the characteristic initial data from late advanced times
248: which are important. That is, one can approximate the characteristic
249: hypersurface by dividing it into two parts: a compact part which is
250: extended from $(u_i,v_i)$ to a point $(u_i,v_f)$ (with $v_f>v_i$),
251: and a non-compact part which extends from $(u_i,v_f)$ forward, i.e., the
252: points
253: $(u_i,v>v_f)$. The specific form of the characteristic initial data on
254: $(u_i,v<v_f)$ is unimportant: it is only the contribution of the initial
255: data at $(u_i,v>v_f)$ which is important. (Similarly, also the initial
256: data along $v=v_i$ is unimportant.) Consequently, we can determine
257: arbitrary initial data at early times.
258:
259: The solution of the characteristic initial value problem then is
260: given by $\Phi_v(v)=(A/\sqrt{-2r_{u0}})\ln (v/v_i)$, $\Phi_u(u)=0$,
261: $\sigma_v(v)=-\ln [2(v/v_0)^{A^2/(2r_{u0})}]/2$, and
262: $\sigma_u(u)=-(1/2)\ln 2$.
263:
264: {\it Numerical simulations}: Our numerical code is a free evolution code
265: in $(1+1)$-D in double-null
266: coordinates with an adaptive mesh refinement \cite{burko-ori97}. We tested
267: the code and found that it is stable, and converges with second
268: order. In the following we present results with the following choice of
269: parameters, unless stated otherwise: $M_{\rm initial}=1$, $Q=0.95$,
270: $r_0=5$ (these uniquely determine the value of $r_{u0}$), $A=0.3$, and
271: $N=10$. Here, $M_{\rm initial}$ is the initial mass of the black hole, and
272: $N$ is defined as the number density of grid points on the characteristic
273: hypersurface in both $u$ and $v$ directions. We find similar qualitative
274: results also for other choices of the parameters.
275: The stability and second-order convergence are demonstrated in
276: Fig.~\ref{fig1}, which displays $\Phi$ and $v$ as functions of $r$ along
277: an outgoing null ray deep inside the black hole for various values of the
278: grid parameter $N$.
279: \begin{figure}
280: \input epsf
281: %\epsfxsize=8.5cm
282: \centerline{ \epsfxsize 8.5cm
283: \epsfbox{f1.eps}}
284: \caption{Behavior of $\Phi$ and $r$ along an outgoing ray (at fixed
285: $u$) for various values of $N$. Upper panel (A): $v$ as a function of $r$.
286: Lower panel (B): $\Phi$ as a function of $r$.
287: In both panels dotted lines correspond to $N=10$,
288: dashed lines to $N=20$, and solid lines to $N=40$, and the data are taken
289: for $u=21.9$.}
290: \label{fig1}
291: \end{figure}
292:
293: Figure~\ref{fig2} displays equi-spaced (in $u$) outgoing null rays (with
294: constant values of $u$) in the $rv$-plane. The strong nonlinear
295: dynamics is demonstrated by the rapid increase in the apparent
296: horizon. All the rays which are not outside the event horizon and escape
297: to
298: infinity, either terminate at $r=0$ within a finite lapse of advanced time
299: $v$ (type I), or approach a finite limiting value of $r$ as $v\to\infty$
300: of $r$ at large values of $v$ (type II).
301:
302: {\it The null portion of the singularity}: Figure~\ref{grads}(A) shows
303: the behavior of $r_{,v}$ along a type-II outgoing ray. At late times
304: $r_{,v}\propto v^{-2}$. Figure~\ref{grads}(B) shows
305: the behavior of $\Phi_{,v}$ along the same outgoing ray. At late times
306: $\Phi_{,v}\propto v^{-1}$. This implies that along type-II rays $r$ indeed
307: approaches a non-zero finite value as $v\to\infty$, but $\Phi$ diverges
308: logarithmically in the same limit. This behavior is in sharp contrast with
309: the behavior of $\Phi$ in the case of perturbations with compact support,
310: where $\Phi$, too, approaches a non-zero finite value. We next check the
311: detailed behavior of the fields along type-II rays. We find that type-II
312: rays terminate (in the infinite future as $v\to\infty$) at a curvature
313: singularity. This is demonstrated by Fig.~\ref{grads}(C), which shows the
314: exponential increase in
315: $R\equiv (R_{\alpha\beta}R^{\alpha\beta})^{1/2}$ along the same
316: outgoing
317: ray, $R_{\alpha\beta}$ being the Ricci tensor. (Notice that here $R$ is
318: not the Ricci curvature scalar.)
319: The finiteness of $r$ at
320: the singular hypersurface suggests that the non-central portion
321: of the singularity is deformationally weak.
322:
323: Note that the late-time behavior of $r_{,v}$ and $\Phi_{,v}$, as is clear
324: from Fig.~\ref{grads}, starts to dominate much earlier than in the case of
325: perturbations of compact support. In particular, no quasi-normal modes
326: (QNM) are visible. The reason for that is that the gradients of $\Phi$
327: decays here much slower than the tails in the case of perturbations of
328: compact support. Specifically, $\Phi_{,v}$ decays here according to an
329: inverse power-law with a smaller index than in the case of perturbations
330: with compact support. (In the latter case the index is $4$ for spherical
331: perturbations, whereas here we have an index of $1$.) Because the field is
332: stronger, it starts dominating earlier, and overwhelms the
333: rapidly-decaying oscillations of the QNM.
334:
335: \begin{figure}
336: \input epsf
337: %\epsfxsize=8.5cm
338: \centerline{ \epsfxsize 8.5cm
339: \epsfbox{f2.eps}}
340: \caption{Outgoing null rays (with fixed values of $u$) in the
341: $rv$-plane.
342: The solid lines correspond to different fixed values of $u$,
343: and the
344: dotted line describes the apparent horizon, which approaches
345: $u=21.07185$ at late values of advanced time $v$. The values of $u$
346: for which rays are shown (from right to left) are from $u=20.8$ to
347: $22.7$ in equal increments of $\Delta u=0.1$.}
348: \label{fig2}
349: \end{figure}
350:
351: \begin{figure}
352: \input epsf
353: %\epsfxsize=8.5cm
354: \centerline{ \epsfxsize 8.5cm
355: \epsfbox{f3.eps}}
356: \caption{Behavior of the fields and the curvature of spacetime
357: along a type-II outgoing null ray. Upper panel (A): $r_{,v}$ as a function
358: of $v$. Middle panel (B): $\Phi_{,v}$ as a function of $v$. Lower panel
359: (C): $R$ as a function of $v$.
360: The data are shown along $u=21.1$.}
361: \label{grads}
362: \end{figure}
363:
364: {\it The spacelike portion of the singularity}: Type-I rays terminate at a
365: spacelike singularity.
366: The spacelike singularity inside a spherical charged black hole which is
367: perturbed by a scalar field was studied within a simplified homogeneous
368: model in Ref.~\cite{burko99}, where the pointwise behavior of the geometry
369: and the field was found. It was also shown in \cite{burko99} that
370: approaching the spacetime singularity, the fully nonlinear and
371: inhomogeneous numerical solution (where the perturbation field had a
372: compact support on the characteristic hypersurface) was in full agreement
373: with the pointwise behavior. The study of the singularity in
374: Ref.~\cite{burko99} was local: no assumptions were made regarding the form
375: of the perturbation field on the characteristic hypersurface. We thus
376: expect that type-I rays terminate at a spacelike singularity whose
377: pointwise behavior is well
378: described by the singularity of Ref.~\cite{burko99}. Assuming
379: homogeneity, one finds approaching the spacelike singularity inside a
380: spherical charged black hole with a scalar field, that
381: \begin{equation}
382: \Phi(r)=\sqrt{\beta +1}\ln r +O(r^{\beta})\, .
383: \label{phi}
384: \end{equation}
385: Here, $\beta >0 $ is a constant (which numerically can be found to depend
386: on $u$). Also, along an outgoing null ray one can show that, to the
387: leading order in $v_*-v$,
388: \begin{equation}
389: r(u_0,v)=[d^2/(\beta +1)]^{1/4}(v_*-v)^{1/2}\, ,
390: \label{r}
391: \end{equation}
392: where $d$ is a gauge-dependent quatity (which depends on the scaling of
393: the temporal coordinate), and $v_*=v(r=0)$ along that null ray.
394:
395: The agreement of our results with Eqs.~(\ref{phi}) and (\ref{r}) is
396: already apparent from Fig.~\ref{fig1}. Next, we check this agreement in
397: greater detail. In Fig.~\ref{fig3}(B) we show the
398: behavior of the scalar field $\Phi$ as a function of $r(v)$ along a type-I
399: outgoing null ray. This logarithmic behavior is consistent with
400: Eq.~(\ref{phi}). Along all type-I rays we find the same
401: logarithmic divergence of $\Phi$, including along rays which initially
402: are outside or inside the apparent horizon. (The only difference between
403: different rays is that the slope of the graph, i.e., the value of the
404: parameter $\beta$, changes from one ray to another.)
405: \begin{figure}
406: \input epsf
407: %\epsfxsize=8.5cm
408: \centerline{ \epsfxsize 8.5cm
409: \epsfbox{f4.eps}}
410: \caption{The scalar field $\Phi$ and the radial coordinate $r$ approaching
411: the singularity along a type-I null ray. Upper panel (A): $v_*-v$ as a
412: function of $r$. Lower panel (B): $\Phi$ as a function of $r$. The data
413: are shown along the outgoing null ray at $u=21.9$.}
414: \label{fig3}
415: \end{figure}
416: Figure \ref{fig3}(A) displays $v_*-v$ as a function of $r$ along
417: the same outgoing null ray. The asymptotic behavior approaching the
418: singularity agrees very nicely with Eq.~(\ref{r}).
419:
420: We conclude that the pointwise behavior at the
421: singularity which we find
422: in our simulations is well described by the singularity described in
423: Ref.~\cite{burko99}. Notice that this singularity is different from the
424: Schwarzschild singularity: The former has $\beta>0$, and the latter has
425: $\beta=-1$. This portion of the singularity then is scalar curvature,
426: spacelike, and deformationally strong.
427:
428: {\it Conclusions}: We studied the evolution of spacetime, and specifically
429: the formation of curvature singularities, for a very simplied toy model of
430: a spherical charged black hole, which is perturbed nonlinearly by a
431: self-gravitating, spherical scalar field, which has non-compact support on
432: the characteristic initial hypersurface. Although these perturbations are
433: stronger than those which result from an initial profile with compact
434: support (the gradient of the scalar field $\Phi$ decays at late times as
435: $v^{-1}$ in our case, and as $v^{-4}$ in the case of perturbations with
436: compact support) and consequently the evolution of spacetime and in
437: particular of spacetime curvature is indeed dominated by the non-compact
438: perturbations rather than by the perturbations due to the collapse, we
439: find that a portion of the Cauchy horizon still survives as a non-central,
440: null singularity, rather than being utterly destroyed and replaced by a
441: central, spacelike singularity. The null generators of the Cauchy horizon
442: contract with retarded time $u$, and eventually arrive at $r=0$, where the
443: causal structure and the strength the singularity change: the central
444: singularity is spacelike and deformationally strong. This situation and
445: the global causal structure is therefore very similar to that of a black
446: hole perturbed by a perturbation field with compact support, despite the
447: different details of the dynamics.
448:
449: The reason why the Cauchy horizon survived the introduction of
450: non-compact perturbations as a null, non-central singularity is the
451: following: We chose the characteristic field to be such, that although it
452: is non-compact and does not belong to the same class of behavior on the
453: event horizon at late advanced time as the tails, its gradient
454: does. Specifically, along the event horizon $\Phi$ is logarithmic in
455: advanced time. This certainly does not belong to the class of the tails,
456: which decay as an inverse integral power of advanced time on the event
457: horizon. However, $\phi_{,v}$ decays like $v^{-1}$ along the event horizon
458: at late advanced time, such that it does belong to the same class as the
459: gradients of the tails. (Note, that no tails would ever produce $n=1$, as
460: $n$ is at least $3$ for all tails.) It is, in fact, $\phi_{,v}$ which is
461: the important quantity, as curvature depends on the gradients of $\Phi$,
462: rather than on $\Phi$ itself. This particular form of $\phi_{,v}$
463: implies that ${\cal R}$ can still be twice integrable approaching the
464: Cauchy horizon.
465:
466: We therefore conclude that by themselves, perturbations of non-compact
467: support, even when they dominate the dynamics, are not sufficient to
468: obliterate the null, non-central singularity at the Cauchy horizon. It
469: remains an open question, however, whether other classes of non-compact
470: perturbations behave similarly. The CBR perturbations,
471: which are a generic source of perturbations for realistic black holes, are
472: of particular interest. The energy influx of the CBR decays only
473: on very long time scales due to the expansion of the universe (in a
474: matter-dominated universe). (In a dark-energy dominated universe the
475: influx of CBR energy decays faster.) It is interesting to investigate how
476: the Cauchy horizon singularity is affected by such perturbation fields,
477: and also to investigate which families of perturbing fields may destoy
478: the null, non-central singularity.
479:
480: I thank Karel Kucha\v{r} and Richard Price for discussions, and an
481: anonymous referee for useful comments. This research
482: was supported by the National Science Foundation through grant No.\
483: PHY-9734871.
484:
485:
486: \begin{thebibliography}{}
487: \bibitem{review} For recent reviews see L.~M.~Burko and A.~Ori, in {\it
488: Internal structure of black holes and spacetime singularities}, edited by
489: L.~M.~Burko and A.~Ori (Institute of Physics, Bristol, 1997); P.~R.~Brady,
490: Prog.~Theor.~Phys.~Suppl.~{\bf 136}, 29 (1999).
491: \bibitem{poisson-israel90} E.~Poisson and W.~Israel, Phys.~Rev.~D {\bf
492: 41}, 1796 (1990).
493: \bibitem{ori91} A.~Ori, Phys.~Rev.~Lett.~{\bf 67}, 789 (1991).
494: \bibitem{ori92} A.~Ori, Phys.~Rev.~Lett.~{\bf 68}, 2117 (1992).
495: \bibitem{brady-smith95} P.~R.~Brady and J.~D.~Smith, Phys.~Rev.~Lett.~{\bf
496: 75}, 1256 (1995).
497: \bibitem{burko97} L.~M.~Burko, Phys.~Rev.~Lett.~{\bf 79}, 4958 (1997).
498: \bibitem{burko-ori98} L.~M.~Burko and A.~Ori, Phys.~Rev.~D {\bf 57}, R7084
499: (1998).
500: \bibitem{burko99a} L.~M.~Burko, Phys.~Rev.~D {\bf 60}, 104033 (1999).
501: \bibitem{remark} Notice that there is a misprint in Eq.~(4) of
502: L.~M.~Burko, Mod.~Phys.~Lett.~A {\bf 14}, 1015 (1999). The correct
503: relation is given in Eq.~(14) of Ref.~\cite{burko99a}.
504: \bibitem{price} R.~H.~Price, Phys.~Rev.~D {\bf 5}, 2419 (1972).
505: \bibitem{clarke-krolak} C.~J.~S.~Clarke and A.~Kr\'{o}lak,
506: J.~Geom. Phys.~{\bf 2}, 127 (1985).
507: \bibitem{burko-qed} L.~M.~Burko, Phys.~Rev.~D {\bf 55}, 2105 (1997).
508: \bibitem{ori99} A.~Ori, Phys.~Rev.~Lett.~{\bf 83}, 5423 (1999).
509: \bibitem{burko-ori97} L.~M.~Burko and A.~Ori, Phys.~Rev.~D {\bf 56}, 7820
510: (1997).
511: \bibitem{burko99} L.~M.~Burko, Phys.~Rev.~D {\bf 59}, 024011 (1999).
512: \end{thebibliography}
513:
514: \end{document}
515: