1: \documentclass[aps,amsfonts,amsmath,prd,preprint,nofootinbib,tightenlines]{revtex4}
2: \usepackage{epsf}
3:
4:
5: \newcommand{\beq}{\begin{equation}}
6: \newcommand{\eeq}{\end{equation}}
7:
8: \def\be{\begin{equation}}
9: \def\ee{\end{equation}}
10: \def\beqa{\begin{eqnarray}}
11: \def\eeqa{\end{eqnarray}}
12: \def\p{\partial}
13: \def\mp{m_{\rm pl}}
14: \def\nn{\nonumber}
15: \def\r{\right.}
16: \def\l{\left.}
17: \def\q{\quad}
18: \begin{document}
19:
20: \title{Probabilities in the Bousso-Polchinski multiverse}
21: \author{Delia Schwartz-Perlov and Alexander
22: Vilenkin}
23: \affiliation{Institute of Cosmology, Department of Physics and
24: Astronomy\\
25: Tufts University, Medford, MA 02155, USA }
26:
27: \begin{abstract}
28: Using the recently introduced method to calculate bubble
29: abundances in an eternally inflating spacetime, we investigate the
30: volume distribution for the cosmological constant $\Lambda$ in the
31: context of the Bousso-Polchinski landscape model. We find that the
32: resulting distribution has a staggered appearance which is in sharp
33: contrast to the heuristically expected flat distribution. Previous
34: successful predictions for the observed value of $\Lambda$ have
35: hinged on the assumption of a flat volume distribution. To
36: reconcile our staggered distribution with observations for
37: $\Lambda$, the BP model would have to produce a huge number of vacua
38: in the anthropic range $\Delta\Lambda_A$ of $\Lambda$, so that the
39: distribution could conceivably become smooth after averaging over
40: some suitable scale $\delta\Lambda\ll\Delta\Lambda_A$.
41: %We
42: %conjecture that the staggered character of our distribution will
43: %arise in any landscape model which generates a dense spectrum of
44: %low-energy constants from a wide distribution of states in the
45: %parameter space of the fundamental theory.
46:
47: %If the BP model does not produce this huge number of vacua in the
48: %anthropic range, we would expect $\Delta\Lambda_A$ to be dominated
49: %by one or a few values of $\Lambda_j$ which have the highest volume
50: %fractions. These would have to be "accidentally" close to
51: %$\Lambda=0$ weakening the anthropic approach. Furthermore, if we
52: %look beyond the anthropic range, we would expect to find ourselves
53: %in a lone galaxy in an almost empty hubble volume, since larger
54: %values of $\Lambda>>\Delta\Lambda_A$ which lead to a suppression in
55: %galaxy formation rates would be compensated for by enhanced volume
56: %fractions.
57:
58:
59: \end{abstract}
60: \maketitle
61:
62: \section{Introduction}
63:
64: The cosmological constant problem is one of the most intriguing
65: mysteries that we now face in theoretical physics. The observed
66: value of the cosmological constant $\Lambda$ is many orders of
67: magnitude smaller than theoretical expectations and is surprisingly
68: close to the present matter density of the universe,\footnote{Here
69: and below we use reduced Planck units, $M_p^2/8\pi =1$, where $M_p$
70: is the Planck mass.} \beq \Lambda_0\sim\rho_{m0}\sim 10^{-120}.
71: \label{Lambdarho} \eeq As of now, the only plausible explanation for
72: these enigmatic facts has been given in terms of the multiverse
73: picture, which assumes that $\Lambda$ is a variable parameter taking
74: different values in different parts of the universe
75: \cite{Weinberg87,Linde87,AV95,Efstathiou,MSW,GLV,Bludman,AV05}. The
76: probability for a randomly picked observer to measure a given value
77: of $\Lambda$ can then be expressed as \cite{AV95} \be
78: P_{obs}(\Lambda)\propto P(\Lambda)n_{obs}(\Lambda), \label{Pobs} \ee
79: where $P(\Lambda)$ has the meaning of the volume fraction of the
80: regions with a given value of $\Lambda$ and $n_{obs}(\Lambda)$ is
81: the number of observers per unit volume.\footnote{$P(\Lambda)$ is
82: often called the prior probability. Here we avoid this terminology,
83: since it is usually used to characterize one's ignorance or
84: prejudice, while the volume factor $P(\Lambda)$ should be
85: calculable, at least in principle.} Disregarding the possible
86: variation of other ``constants'' and assuming that the density of
87: observers is roughly proportional to the fraction of matter
88: clustered in large galaxies, \be n_{obs}(\Lambda) \propto
89: f_G(\Lambda), \label{nobs} \ee one finds that the function
90: $n_{obs}(\Lambda)$ is narrowly peaked around $\Lambda =0$, with a
91: width
92: %comparable to $\rho_{m0}$.
93: \beq \Delta\Lambda_A\sim 100\Lambda_0 \sim 10^{-118}.
94: \label{DeltaLambdaA} \eeq
95:
96: In general, the volume factor $P(\Lambda)$ depends on the unknown
97: details of the fundamental theory and on the dynamics of eternal
98: inflation. However, it has been argued \cite{AV96,Weinberg96} that
99: it should be well approximated by a flat distribution,
100: \be
101: P(\Lambda)\approx {\rm const}.
102: \label{flat}
103: \ee
104: The reason is that the anthropic range (\ref{DeltaLambdaA}), where the
105: function $n_{obs}(\Lambda)$ is substantially different from zero, is
106: much narrower than the full range of variation of $\Lambda$, which is
107: expected to be set by the Planck scale. A smooth function varying on
108: this large characteristic scale will be nearly constant within the
109: tiny anthropic interval.
110:
111: Combination of Eqs.~(\ref{Pobs})-(\ref{flat}) yields the
112: distribution \be P_{obs}(\Lambda)\propto f_G(\Lambda), \label{PnG}
113: \ee which can be readily calculated using the Press-Schechter
114: approximation for $f_G$. The observed value of $\Lambda$ is within
115: the $2\sigma$ range of this distribution -- an impressive success of
116: the multiverse paradigm. One should keep in mind, however, that the
117: successful prediction for $\Lambda$ hinges on the assumption of a
118: flat volume distribution (\ref{flat}). We emphasize that the form
119: of the volume distribution is important. If, for example, one uses
120: $P(\Lambda) \propto\Lambda$ instead of (\ref{flat}), the $2\sigma$
121: prediction would be $10 <\Lambda/\Lambda_0 <500$ and the observed
122: value of $\Lambda$ would be ruled out at a 99.9\% confidence level
123: \cite{Pogosian}. The heuristic argument for a flat distribution
124: (\ref{flat}) sounds plausible, but it needs to be verified in
125: specific models.
126:
127: The simplest model with a variable effective cosmological constant is
128: that of a scalar field $\phi$ with a very slowly varying potential
129: $V(\phi)$ \cite{Banks84,Linde87}. In such models, $\Lambda$ takes a
130: continuum range of values. It has been verified that the resulting
131: volume distribution for $\Lambda$ is indeed flat for a wide range of
132: potentials \cite{Weinberg00,GV00,GV03}. The main challenge one has to face
133: in this type of model is to justify the exceedingly flat potential
134: which is required to keep the field $\phi$ from rolling downhill on
135: the present Hubble time scale.
136:
137: A model with a discrete spectrum of $\Lambda$ was first suggested by
138: Abbott \cite{Abbott}. He considered a scalar field with a
139: ``washboard'' potential, having many local minima separated by
140: barriers. Transitions between the minima could occur through bubble
141: nucleation. An alternative discrete model, first introduced by Brown
142: and Teitelboim \cite{BT}, assumes that the cosmological constant can
143: be expressed as
144: \be
145: \Lambda = \Lambda_{bare} +F^2/2.
146: \label{BT}
147: \ee
148: Here, $\Lambda_{bare}$ is the bare cosmological constant, which is
149: assumed to be large and negative, and $F$ is the magnitude of a
150: four-form field, which can change its value through the nucleation of
151: branes. The change of the field strength across the brane is
152: \be
153: \Delta F =\pm q,
154: \label{DeltaF}
155: \ee
156: where the ``charge'' $q$ is a constant fixed by the model.
157:
158: In order to explain observations, the discrete spectrum of $\Lambda$
159: has to be very dense, with separation between adjacent values
160: \beq
161: \Delta\Lambda\lesssim \Lambda_0,
162: \label{dense}
163: \eeq
164: which in turn requires that the charge $q$ has to be very small. If
165: this is satisfied, analysis shows that the flat volume distribution
166: (\ref{flat}) is quite generic \cite{GV01}. But once again, the
167: exceedingly small charge $q$ required by the model appears to be
168: unnatural.\footnote{Some ideas on how such small parameters could
169: arise in particle physics have been suggested in
170: \cite{Weinberg00,Donoghue00,FMSW,BDM,DV01}.}
171:
172: In an effort to remedy this problem, Bousso and Polchinski (hereafter
173: BP) extended the Brown-Teitelboim approach to include multiple
174: four-form fluxes \cite{BP}. They considered a model in which $J$
175: different fluxes give rise to a J-dimensional grid of vacua, each
176: labeled by a set of integers $n_a$. Each point in the grid corresponds
177: to a vacuum with the flux values $F_a = n_a q_a$ and a
178: cosmological constant
179: \be
180: \label{totalLambda}
181: \Lambda=\Lambda_{bare}+\frac{1}{2}\sum_{a=1}^{J}F_a^2=\Lambda_{bare}
182: +\frac{1}{2}\sum_{a=1}^{J}n_a^2q_a^2.
183: \ee
184: This model is particularly interesting because multiple fluxes
185: generally arise in string theory compactifications. The model can thus
186: be regarded as a toy model of the string theory landscape.
187: %The number of flux units in each direction of the grid is not expected
188: %to be very large, since large fluxes are likely to destabilize the
189: %vacuum. If we assume that $|n_a|\leq N$, with $N$ the same in all
190: %directions, then the total number of vacua in the grid is $(2N+1)^J$.
191: BP showed that with $J\sim 100$, the spectrum of allowed values of
192: $\Lambda$ can be sufficiently dense, even in the absence of very small
193: parameters, e.g., with $|\Lambda_{bare}|\sim 1$, $q_a\sim 0.1$.
194:
195: In the cosmological context, high-energy vacua of the BP grid will
196: drive exponential inflationary expansion. The flux configuration in
197: the inflating region can change from one point on the grid to the next
198: through bubble nucleation. Bubbles thus nucleate within bubbles, and
199: each time this happens the cosmological constant either increases or
200: decreases by a discrete amount. This mechanism allows the universe to
201: start off with an arbitrary large cosmological constant, and then to
202: diffuse through the BP grid of possible vacua, to generate regions
203: with each and every possible cosmological constant, including that
204: which we inhabit.
205:
206: Our goal in this paper is to study the volume distribution for
207: $\Lambda$ in the BP model. In particular, we would like to check
208: whether or not this distribution is approximately flat, as suggested
209: by the heuristic argument of \cite{AV96,Weinberg96}. Until recently,
210: such an analysis would have been rather problematic, since the
211: calculation of the volume fractions in an eternally inflating
212: universe is notoriously ambiguous. The volume of each type of vacuum
213: diverges in the limit $t\to\infty$, so in order to calculate
214: probabilities, one has to impose some kind of a cutoff. The answer,
215: however, turns out to be rather sensitive to the choice of the
216: cutoff procedure. If, for example, the cutoff is imposed on a
217: constant time surface, one gets very different distributions
218: depending on one's choice of the time variable $t$ \cite{LLM}. (For
219: more recent discussions, see \cite{VVW,Guth00,Tegmark}.)
220:
221: Fortunately, a fully gauge-invariant prescription for calculating
222: probabilities has been recently introduced in \cite{GSPVW}.
223: %It is called the comoving horizon cutoff method (CHC).
224: It has been tried on some simple models and seems to give reasonable
225: results. Here we shall apply it to the BP model.
226:
227: As BP themselves recognized, their model does not give an accurate
228: quantitative description of the string theory landscape. In
229: particular, it does not explain how the sizes of compact dimensions
230: get stabilized. This issue was later addressed by Kachru, Kallosh,
231: Linde and Trivedi (KKLT) \cite{KKLT}, who provided the first example
232: of a metastable string theory vacuum with a positive cosmological
233: constant. Apart from the flux contributions in (\ref{totalLambda}),
234: the vacuum energy in KKLT-type vacua gets contributions from
235: non-perturbative moduli potentials and from branes. The $4D$
236: Newton's constant in these vacua depends on the volume of extra
237: dimensions and changes from one vacuum to another. Douglas and
238: collaborators \cite{Douglas,AshokDouglas,DenefDouglas} studied the
239: statistics of KKLT-type vacua. Their aim was to find the number of
240: vacua with given properties (e.g., with a given value of $\Lambda$)
241: in the landscape. Our goal here is more ambitions: we want to find
242: the probability for a given $\Lambda$ to be observed.
243:
244: In our analysis, we shall disregard all the complications of the
245: KKLT vacua, with the hope that the simple BP model captures some of
246: the essential features of the landscape. At the end of the paper we
247: shall discuss which aspects of our results can be expected to
248: survive in more realistic models.
249:
250: We begin in the next section by summarizing the prescription of
251: Ref.~\cite{GSPVW} for calculating probabilities. We shall see that
252: the problem reduces to finding the smallest eigenvalue and the
253: corresponding eigenvector of a large matrix, whose matrix elements
254: are proportional to the transition rates between different vacua.
255: The calculation of the transition rates for the BP model is reviewed
256: in Section III. Some of these rates are extremely small, since the
257: upward transitions with an increase of $\Lambda$ are exponentially
258: suppressed relative to the downward transitions. In Section IV we
259: develop a perturbative method for solving our eigenvalue problem,
260: using the upward transition rates as small parameters. This method
261: is applied to the BP model in Section V. We find that the resulting
262: probability distribution has a very irregular, staggered appearance
263: and is very different from the flat distribution (\ref{flat}). The
264: implications of our results for the string theory landscape are
265: discussed in Section VI.
266:
267:
268: \section{Prescription for probabilities}
269:
270: %SO far this is a copy of Sec III a of CHC paper
271:
272: Here we summarize the prescription for calculating the volume
273: distribution proposed in Ref.~\cite{GSPVW}. Suppose we have a theory
274: with a discrete set of vacua, labeled by index $j$. The cosmological
275: constants $\Lambda_j$ can be positive, negative, or zero. Transitions
276: between the vacua can occur through bubble nucleation. The proposal of
277: Ref.~\cite{GSPVW} is that the volume distribution is given by
278: \be
279: P_j \propto p_j Z_j^3,
280: \label{PpZ}
281: \ee
282: where $p_j$ is the relative abundance of bubbles of type $j$ and $Z_j$
283: is (roughly) the amount of slow-roll inflationary expansion inside the
284: bubble after nucleation (so that $Z_j^3$ is the volume slow-roll
285: expansion factor).
286:
287: The total number of nucleated bubbles of any kind in an eternally
288: inflating universe is known to grow without bound, even in a region
289: of finite comoving size. We thus need to cut off our count. The
290: proposal of \cite{GSPVW} is that the counting should be done at the
291: future boundary of spacetime and should include only bubbles with
292: radii greater than some tiny co-moving size $\epsilon$. The limit
293: $\epsilon\rightarrow 0$ should then be taken at the end. (An
294: equivalent method for calculating $p_j$ was suggested in
295: \cite{ELM}.) It was shown in \cite{GSPVW} that this prescription is
296: insensitive to the choice of the time coordinate and to coordinate
297: transformations at future infinity.
298:
299: The bubble abundances $p_j$ can be related to the functions $f_j(t)$
300: expressing the fraction of comoving volume occupied by vacuum $j$ at
301: time $t$. These functions obey the evolution equation \cite{recycling}
302: \beq
303: {df_j\over{dt}}=\sum_i (-\kappa_{ij}f_j + \kappa_{ji}f_i),
304: \label{dfdt}
305: \eeq
306: where the first term on the right-hand side accounts for loss of
307: comoving volume due to bubbles of type $i$ nucleating within those of
308: type $j$, and the second term reflects the increase of comoving volume
309: due to nucleation of type-$j$ bubbles within type-$i$ bubbles.
310:
311: The transition rate $\kappa_{ij}$ is defined as the probability per
312: unit time for an observer who is currently in vacuum $j$ to find
313: herself in vacuum $i$. Its magnitude depends on the choice of the time
314: variable $t$. The most convenient choice for our purposes is to use
315: the logarithm of the scale factor as the time variable; this is the
316: so-called scale-factor time. With this choice,
317: \beq
318: \kappa_{ij}=\Gamma_{ij}{4\pi\over{3}}H_j^{-4},
319: \label{kappa}
320: \eeq
321: where $\Gamma_{ij}$ is the bubble nucleation rate per unit
322: physical spacetime volume (same as $\lambda_{ij}$ in \cite{GSPVW})
323: and
324: \be
325: H_j = (\Lambda_j/3)^{1/2}
326: \label{Hj}
327: \ee
328: is the expansion rate in vacuum $j$.
329: %Here and below we use the reduced
330: %Planck unts,
331: %\be
332: %M_p^2/8\pi = 1,
333: %\ee
334: %where $M_p$ is the Planck mass.
335:
336: We distinguish between the recyclable, non-terminal vacua, with
337: $\Lambda_j>0$, and the non-recyclable, "terminal vacua", for which
338: $\Lambda_j\leq 0$. Transitions from either a flat spacetime
339: ($\Lambda_j=0$), or a negative $\Lambda$ FRW spacetime
340: ($\Lambda_j<0$), which increase $\Lambda$ have a zero probability of
341: occurring.\footnote{This is because the volume of the instanton is
342: compact whilst the volume of the Euclideanized background spacetime
343: is infinite, so that the difference in their actions is infinite.}
344: Transitions from $\Lambda_j = 0$ and from small negative $\Lambda_j$
345: to even more negative $\Lambda$ are possible, but these $\Lambda$'s
346: are likely to be large and negative and are therefore of no
347: anthropic interest. We will label the recyclable, non-terminal vacua
348: by Greek letters, and for non-recyclable, terminal vacua, we will
349: reserve the indices $m$ and $n$. Then, by definition, \beq
350: \Gamma_{\alpha m}=\Gamma_{mn}=0. \label{lambda=0} \eeq Latin letters
351: other than $m,n$ will be used to label arbitrary vacua, both
352: recyclable and terminal, with the exception of letters $a,b$, which
353: we use to label the fluxes.
354:
355: Eq.~(\ref{dfdt}) can be written in a vector form,
356: \beq
357: {d{\bf f}\over{dt}}={\mathbf M}{\bf f},
358: \label{matrixf}
359: \eeq
360: where ${\bf f(t)}\equiv \{ f_j(t)\}$ and
361: \beq
362: M_{ij}=\kappa_{ij}-\delta_{ij}\sum_r \kappa_{ri}.
363: \label{Mij}
364: \eeq
365:
366: The asymptotic solution of (\ref{matrixf}) at large $t$ has the
367: form
368: \beq
369: {\bf f}(t)={\bf f^{(0)}}+{\bf s}e^{-q t}+ ...
370: \label{asympt}
371: \eeq
372: Here, ${\bf f}^{(0)}$ is a constant vector
373: which has nonzero components only in terminal vacua,
374: \beq
375: f_\alpha^{(0)}=0,
376: \label{falpha=0}
377: \eeq
378: while the values of $f_n^{(0)}$ depend on the choice of initial
379: conditions. It is clear from Eq.~(\ref{lambda=0}) that any such
380: vector is an eigenvector of the matrix ${\mathbf M}$ with zero
381: eigenvalue, \beq {\mathbf M}{\bf f_0}=0. \eeq As shown in
382: \cite{GSPVW}, all other eigenvalues of $\mathbf{M}$ have a negative
383: real part, so the solution approaches a constant at late times. We
384: have denoted by $-q$ the eigenvalue with the smallest (by magnitude)
385: negative real part and by $\mathbf{s}$ the corresponding eigenvector.
386:
387: It has been shown in \cite{GSPVW} that the bubble abundances $p_j$ can
388: be expressed in terms of the asymptotic solution (\ref{asympt}). The
389: resulting expression is
390: \beq
391: p_j\propto \sum_\alpha H_\alpha^q \kappa_{j\alpha}
392: s_\alpha.
393: \label{pJaume}
394: \eeq
395: where the summation is over all recyclable vacua which can directly
396: tunnel to $j$.
397:
398: Note that the calculation of $p_j$ requires only knowledge of the
399: components $s_\alpha$ for the recyclable vacua. The evolution of the
400: comoving volume fraction in these vacua is independent of that in the
401: terminal vacua. Formally, this can be seen from the fact that the
402: transition matrix $\mathbf M$ in (\ref{matrixf}) has the form
403: \be \mathbf{M}=
404: \left(%
405: \begin{array}{cc}
406: \mathbf{R} & 0 \\
407: \mathbf{T} & 0 \\
408: \end{array}%
409: \right) \ee Here, $\mathbf{R}$ is a square matrix with matrix
410: elements between recyclable vacua, while the matrix elements of
411: $\mathbf{T}$ correspond to transitions from recyclable vacua to
412: terminal ones. Eigenvectors of $\mathbf{M}$ are of the form ${\bf
413: f}=(\bf s,\bf t)$, where $s$ is an eigenvector of ${\mathbf R}$,
414: \beq \mathbf{R}{\bf s}=-q{\bf s}, \label{Rs} \eeq and ${\bf t}$ is
415: arbitrary. Eigenvalues of $\mathbf{M}$ are the same as those of
416: $\mathbf{R}$, except for some additional zero eigenvalues with
417: eigenvectors which only have nonzero entries for terminal vacua.
418:
419: The problem of calculating $p_j$ has thus been reduced to finding the
420: dominant eigenvalue $q$ and the corresponding eigenvector ${\bf s}$ of
421: the recyclable transition matrix $\mathbf{R}$. In the following
422: sections we shall apply this prescription to the BP model.
423:
424:
425: \section{Nucleation rates in the BP model}
426:
427: In the BP model, we have a $J$-dimensional grid of vacua characterized
428: by the fluxes $F_a=n_a q_a$ and vacuum energy densities given by
429: Eq. (\ref{totalLambda}). BP emphasized that $q_a$ need not be very
430: small, $q_a/|\Lambda_{bare}|\sim 0.1~-~1$. So the model does not have
431: any small parameters, except perhaps the values of $\Lambda_j$ in some
432: vacua, where the contribution of fluxes is nearly balanced by
433: $\Lambda_{bare}$.
434:
435: Transitions between the neighboring vacua, which change one of the
436: integers $n_a$ by $\pm 1$ can occur through bubble
437: nucleation. The
438: bubbles are bounded by thin branes, whose tension $\tau_a$ is related
439: to their charge $q_a$ as
440: \beq
441: \tau_a^2 =q_a^2/2.
442: \label{tauj}
443: \eeq
444: The latter relation is suggested by string theory
445: \cite{BP,FMSW}. It applies only in the supersymmetric limit, but
446: we shall neglect possible corrections due to supersymmetry
447: breaking. Transitions with multiple brane nucleation, in
448: which $|\Delta n_a|>1$ or several $n_a$ are changed at once, are
449: likely to be strongly suppressed \cite{Megevand}, and we shall
450: disregard them here.
451:
452: The bubble nucleation rate $\Gamma_{ij}$ per unit spacetime volume can
453: be expressed as \cite{CdL}
454: \be
455: \Gamma_{ij}=A_{ij} \exp^{-B_{ij}}
456: \label{Gamma}
457: \ee
458: with
459: \beq
460: B_{ij}=I_{ij}-S_j
461: \label{Bij}
462: \eeq
463: Here, $I_{ij}$ is the Coleman-DeLuccia instanton action and
464: \beq
465: S_j=-{8\pi^2\over{H_j^2}}
466: \label{Sj}
467: \eeq
468: is the background Euclidean action of de Sitter space.
469:
470: In the case of a thin-wall bubble, which is appropriate for our
471: problem, the instanton action $I_{ij}$ has been calculated in
472: Refs.~\cite{CdL,BT}. It depends on the values of $\Lambda$ inside
473: and outside the bubble and on the brane tension $\tau$.
474:
475: Let us first consider a bubble which changes the flux $a$ from $n_a$
476: to $n_a-1$ ($n_a>0$). The resulting change in the cosmological constant is
477: given by
478: \be
479: |\Delta\Lambda_a|=(n_a-1/2)q_a^2,
480: \label{DeltaLambda}
481: \ee
482: and the exponent in the tunneling rate (\ref{Gamma}) can be
483: expressed as
484: \be
485: B_{a\downarrow} = B_{a\downarrow}^{flatspace} r(x,y).
486: \label{Bdown}
487: \ee
488: Here, $B_{a\downarrow}^{flatspace}$ is the flat space
489: bounce action,
490: \be
491: B_{a\downarrow}^{flatspace}= \frac{27
492: \pi^2}{2}\frac{\tau_a^4}{|\Delta \Lambda_a|^3}.
493: \ee
494: With the aid of
495: Eqs.~(\ref{tauj}),(\ref{DeltaLambda}) it can be expressed as
496: \be
497: B_{a\downarrow}^{flatspace}= \frac{27
498: \pi^2}{8}\frac{1}{(n_a-1/2)^3q_a^2}
499: \label{Bflat}
500: \ee
501:
502: The gravitational correction factor $r(x,y)$ is given by
503: \cite{Parke}
504: \be
505: r(x,y) = \frac{2[(1+x
506: y)-(1+2xy+x^2)^{\frac{1}{2}}]}{x^2(y^2-1)(1+2xy+x^2)^{\frac{1}{2}}}
507: \label{gravfactor}
508: \ee
509: with the dimensionless parameters
510: \be
511: x\equiv
512: \frac{3q_a^2}{8|\Delta\Lambda_a|}=\frac{3}{8(n_a-1/2)}
513: \ee
514: and
515: \be
516: y\equiv \frac{2\Lambda}{|\Delta\Lambda_a|}-1, \label{y} \ee where
517: $\Lambda$ is the background value prior to nucleation.
518: %\frac{\Lambda_{n_j}+\Lambda_{n_j-1}}{\Lambda_{n_j}-\Lambda_{n_j-1}}=
519: %\frac{\Lambda_{n_j}+\Lambda_{n_j-1}}{\Delta \Lambda}.
520: %\ee
521:
522: The prefactors $A_{ij}$ in (\ref{Gamma}) can be estimated as
523: \beq
524: A_{ij} \sim 1.
525: \label{Aij}
526: \eeq
527: This is an obvious guess for nucleation out of vacua with
528: $\Lambda_j\sim 1$. (This guess is supported by the detailed analysis
529: in Ref.~\cite{Jaume}.) For $\Lambda_j \ll 1$, we still expect
530: Eq.(\ref{Aij}) to hold, since we know that the tunneling rate remains
531: finite in the limit $\Lambda_j \to 0$, $|\Delta\Lambda_a|\sim 1$.
532:
533: If the vacuum $n_a-1$ still has a positive energy density, then an
534: upward transition from $n_a -1$ to $n_a$ is also possible. The
535: corresponding transition rate is characterized by the same instanton
536: action and the same prefactor \cite{EWeinberg}, \beq I_{ij}=I_{ji},
537: ~~~~~~~ A_{ij}=A_{ji}, \label{ijji} \eeq and it follows from
538: Eqs.(\ref{Gamma}), (\ref{Bij}) and (\ref{Hj}) that the upward and
539: downward nucleation rates are related by \be \Gamma_{ji} =
540: \Gamma_{ij} \exp\left[24 \pi^2
541: \left(\frac{1}{\Lambda_{j}}-\frac{1}{\Lambda_{i}}\right)\right].
542: \label{updown} \ee The exponential factor on the right-hand side of
543: (\ref{updown}) depends very strongly on the value of $\Lambda_{j}$.
544: The closer we are to $\Lambda_j=0$, the more suppressed are the
545: upward transitions $j\to i$ relative to the downward ones.
546:
547: Eq.~(\ref{updown}) shows that the transition rate from $n_a$ up to
548: $n_{a+1}$ is suppressed relative to that from $n_{a+1}$ down to
549: $n_a$. It can also be shown that upward transitions from $n_a$ to
550: $n_{a+1}$ are similarly suppressed relative to the downward
551: transitions from $n_a$ to $n_{a-1}$. Using
552: Eqs.~(\ref{DeltaLambda})-(\ref{y}), the ratio of the corresponding
553: rates can be expressed as \beq
554: \ln(\Gamma_{\downarrow}/\Gamma_{\uparrow}) =
555: \Lambda^{-1}f(\Lambda/q_a^2,n_a). \label{Gammaf} \eeq The factor
556: $f(\Lambda/q_a^2,n_a)$ is plotted in Fig. \ref{7Dgammadownuplam} as
557: a function of $\Lambda/q_a^2$ for $n_a=1$ and $n_a=2$. The plot
558: shows that upward transitions are strongly suppressed, unless
559: $\Lambda/q_a^2$ is very large. The factor $\Lambda^{-1}$ in
560: Eq.~(\ref{Gammaf}) results in even stronger suppression when
561: $\Lambda$ is well below the Planck scale.
562:
563:
564:
565: %\begin{figure}
566: %\begin{center}
567: %\leavevmode\epsfxsize=5in\epsfbox{7Dgammadownup.eps}
568: %\end{center}
569: %\caption{The factor $f(\Lambda/q_a^2,n_a)$ as a function of
570: %$\Lambda/q_a^2$ for $n_a=1$ (solid line), $n_a=2$ (dashed line), and
571: %$n_a=10$ (dotted line).} \label{7Dgammadownup}
572: %\end{figure}
573:
574:
575: \begin{figure}
576: \begin{center}
577: \leavevmode\epsfxsize=5in\epsfbox{7Dgammadownuplam.eps}
578: \end{center}
579: \caption{The factor $f(\Lambda/q_a^2,n_a)$ as a function of
580: $\Lambda/q_a^2$ for $n_a=1$ (solid line), $n_a=2$ (dashed line), and
581: $n_a=10$ (dotted line).} \label{7Dgammadownuplam}
582: \end{figure}
583:
584: Transition rates from a given vacuum $j$ to different states $i$ are
585: related by
586: \beq
587: \Gamma_{ij}\propto \exp(-I_{ij}).
588: \label{GammaI}
589: \eeq
590: As a rule of thumb,
591: \beq
592: I_{ij}\sim -\Lambda_{max}^{-1},
593: \label{ILambda}
594: \eeq
595: where $\Lambda_{max}$ is the larger of $\Lambda_i$ and $\Lambda_j$. It
596: follows from (\ref{GammaI}),(\ref{ILambda}) that upward transitions
597: from a given site are more probable to the lower-energy vacua.
598:
599: To develop some intuition for the dependence of the tunneling
600: exponent $B_{a\downarrow}$ on the parameters of the model, we shall
601: consider the limits of small and large $\Lambda$. For $\Lambda\ll
602: |\Delta\Lambda_a|$, we have $y\approx -1$, and Eq.
603: (\ref{gravfactor}) gives \beq r(y\to -1)=(1-x)^{-2} >1. \eeq Hence,
604: for low-energy vacua the tunneling exponent is increased over its
605: flat-space value, resulting in a suppression of the nucleation rate.
606: (For small values of $x$, $r$ is increased only by a small fraction,
607: but the factor $B_{a\downarrow}^{flatspace}$ that it multiplies in
608: the tunneling exponent is typically rather large, so the resulting
609: suppression can still be significant.)
610:
611: In the opposite limit, $\Lambda\gg|\Delta\Lambda_a|$, $y\gg 1$, \beq
612: r(y\gg 1)\approx \sqrt{2}(xy)^{-3/2}, \eeq and \beq
613: B_{a\downarrow}\approx \frac{27\pi^2}{2} q_a
614: \left(\frac{2}{3\Lambda}\right)^{3/2}. \label{LargeLambda} \eeq
615:
616: The gravitational factor $r$ is plotted in Fig. \ref{7Drvsz}
617: as a function of $\Lambda/|\Delta\Lambda_a|$ for $n_a=1$, $n_a=2$ and
618: $n_a=10$ (corresponding to $x=3/4,~1/4$ and $0.04$, respectively). We
619: see that for large values of $\Lambda$, $r\ll 1$, so the nucleation
620: rate is enhanced. In order for our model to be viable, we must ensure
621: that the tunneling action is large enough to justify the use of the
622: semi-classical approximation: $B_{a\downarrow}\gg 1$, or
623: $\Lambda/q_a^2\ll 20 q_a^{-4/3}$.
624:
625:
626: \begin{figure}
627: \begin{center}
628: \leavevmode\epsfxsize=5in\epsfbox{7Drvsz.eps}
629: \end{center}
630: \caption{The gravitational factor $r$ as a function of
631: $\Lambda/|\Delta\Lambda_a|$ for $n_a=1$ (solid line), $n_a=2$
632: (dashed line), and $n_a=10$ (dotted line).} \label{7Drvsz}
633: \end{figure}
634:
635: %With $q_a\lesssim 1$, this also ensures
636: %that $\Lambda_j $ are small compared to the Planck density,
637: %$\rho_{Pl}=64\pi^2$.
638:
639: If $q_a$ and $\Lambda$ are changed simultaneously, keeping the
640: ratios $\Lambda/q_a^2$ fixed, then $x$ and $y$ do not change, and it
641: is clear from Eqs.~(\ref{Bdown}),(\ref{Bflat}),(\ref{gravfactor})
642: that the nucleation exponents scale as $B_{ij}\propto \Lambda^{-1}$.
643: This shows that bubble nucleation rates are strongly suppressed when
644: the energy scales of $q_a$ and $\Lambda$ are well below the Planck
645: scale.
646:
647:
648: \section{Perturbation theory}
649:
650: \subsection{Degeneracy factors}
651:
652: We shall assume for simplicity that the integers $n_a$ take values in
653: the range $|n_a|\leq N$, where $N$ is independent of $a$. The number
654: of vacua in the grid is then $(2N+1)^J$.
655:
656: To maximize computational abilities, we used the symmetry $n_a\to
657: -n_a$ and restricted the analysis to the sector $0\leq\{n_a\}\leq
658: N$. We took into account the degeneracies in $\Lambda$ that would
659: occur if we allowed negative values of $n_a$ by assigning
660: appropriate degeneracy factors to the probabilities that we
661: calculated. For example, if we have a two-dimensional grid, $J=2$,
662: and only consider the quadrant $n_a\geq 0$, then any point that lies
663: on one of the two axes will be doubly degenerate (configuration
664: $\{0,1\}$ has the same $\Lambda$ as $\{0,-1\}$), whilst a point that
665: lies in the interior of the quadrant will have a four-fold
666: degeneracy (configuration $\{1,2\}$ has the same $\Lambda$ as
667: $\{-1,2\}$, $\{-1,-2\}$, and $\{1,-2\}$).
668:
669: In general, the degeneracy of each site can be calculated
670: using the following formula:
671: \be
672: %degeneracy(n_1,~n_2,...,~n_J)
673: {\cal D}\{n_a\}= 2^{k\{n_a\}},
674: \ee
675: where
676: \beq
677: k\{n_a\} = J - (\delta_{0 n_1} + \delta_{0 n_2} +...+ \delta_{0
678: n_J})
679: \ee
680: %\textbf{ where the Kronecker delta function
681: %$\delta_{0 n}=1~~iff~~n=0$, else $\delta_{0 n}=0$.
682: So points which have no zero coordinates for a J=7 model have
683: ${\cal D} = 2^7 =128$. A point with one zero coordinate has
684: ${\cal D} = 2^6 =64$ etc.
685:
686: When we use Eq. (\ref{pJaume}) to calculate the
687: probabilities, we multiply the RHS by the appropriate degeneracy.
688:
689: Diffusion from a grid point for which $n =0$ to $n_a =-1$,
690: is equivalent to the diffusion from $n =0$ to $n_a =+1$. Also,
691: diffusion from $n_a =-1$ to $n =0$ is equivalent to $n_a =1$ to $n
692: =0$. Thus we were able to take into account these processes in the
693: transition matrix by double counting the positive $n_a$ to or from
694: $n=0$ transition rates. In summary, we implemented boundary
695: conditions such that our process is equivalent to diffusion through
696: a J-dimensional grid, with $-N\leq n_a \leq N$.
697:
698: As an illustrative example, we show in Fig. \ref{7Dspectrumrough} a
699: histogram of the number of vacua vs. $\Lambda$ for a model with
700: $J=7$ and $N=4$, which has $\sim 10^7$ vacua. The parameter values
701: used in this model are \beq {q_a}= {0.5308,~0.3909,~0.5175,~0.4722,~
702: 0.5103,~0.4036,~0.4541}; ~~~~~~~\Lambda_{bare}=-0.702. \label{7D}
703: \eeq The sharp spikes and dips in the histogram are due partly to
704: the non-uniform distribution of the vacua along the $\Lambda$-axis
705: and partly to the difference in degeneracy factors for different
706: vacua. The spikes disappear when the histogram is plotted with a
707: larger bin size, as shown in Fig.\ref{7Dspectrum}.
708:
709:
710: \begin{figure}
711: \begin{center}
712: \leavevmode\epsfxsize=5in\epsfbox{7Dspectrumrough.eps}
713: \end{center}
714: \caption{The spectrum of vacua for a $J=7$, $N=4$ BP grid with
715: parameters given in (\ref{7D}).}
716: \label{7Dspectrumrough}
717: \end{figure}
718:
719:
720: \begin{figure}
721: \begin{center}
722: \leavevmode\epsfxsize=5in\epsfbox{7Dspectrum.eps}
723: \end{center}
724: \caption{The smoothed spectrum for the above model.}
725: \label{7Dspectrum}
726: \end{figure}
727:
728:
729:
730: \subsection{Zeroth order}
731:
732: As outlined in Section II, the calculation of probabilities reduces
733: to finding the smallest eigenvalue and the corresponding eigenvector
734: for a huge ${\cal N}\times {\cal N}$ (recycling) transition matrix
735: $\mathbf{R}$. Here, ${\cal N}$ is the number of recyclable vacua,
736: which we expect to be comparable to the total number of vacua. In
737: our numerical example ${\cal N}\sim 10^7$, while for a realistic
738: string theory landscape it can be as large as $10^{500}$
739: \cite{Susskind,Douglas,AshokDouglas,DenefDouglas}. Matters are
740: further complicated by the fact that some of the elements of
741: $\mathbf{R}$ are exceedingly small. For example, it follows from
742: Eq.~(\ref{updown}) that upward transitions from low-energy vacua
743: with $\Lambda_j\ll 1$ are very strongly suppressed. Matrix
744: diagonalization programs like Mathematica are not well suited for
745: dealing with such matrices. We shall see, however, that the
746: smallness of the upward transition rates can be used to solve our
747: eigenvalue problem via perturbation theory, with the upward
748: transition rates playing the role of small expansion parameters.
749:
750: We represent our transition matrix as a sum of an unperturbed matrix
751: and a small correction, \be \mathbf{R}=\mathbf{R^{(0)}}+
752: \mathbf{R^{(1)}}, \ee where $\mathbf{R^{(0)}}$ contains all the
753: downward transition rates and $\mathbf{R^{(1)}}$ contains all the
754: upward transition rates. We will solve for the zero'th order
755: dominant eigensystem $\{q^{(0)},\mathbf{s^{(0)}}\}$ from
756: $\mathbf{R^{(0)}}$ and then find the first order corrections by
757: including contributions from $\mathbf{R^{(1)}}$. Note that the
758: eigenvalue correction $q^{(1)}$ is not needed for the calculation of
759: bubble abundances (\ref{pJaume}) to the lowest nonzero order. One
760: only needs to calculate the eigenvector correction
761: $\mathbf{s^{(1)}}$ (since the zeroth-order components
762: $s_\alpha^{(0)}$ vanish for recyclable vacua).
763:
764: %We note that some of the upward transition rates from high-energy
765: %vacua, which are included in $\mathbf{R^{(1)}}$, may actually be
766: %larger than some of the downward rates which are kept in
767: %$\mathbf{R^{(0)}}$. We shall see that this does not cause any
768: %problems. {\bf (We wish!)}
769:
770: If the vacua are arranged in the order of increasing $\Lambda$, so
771: that \beq \Lambda_1 \leq \Lambda_2 \leq ... \leq \Lambda_{\cal N},
772: \eeq then $\mathbf{R^{(0)}}$ is a triangular matrix, with all matrix
773: elements below the diagonal equal to zero. Its eigenvalues are
774: simply equal to its diagonal elements, \beq R^{(0)}_{\alpha\alpha}
775: =-\sum_{j<\alpha}\kappa_{j\alpha} \equiv -D_\alpha.
776: \label{Ralpha}\eeq Hence, the magnitude of the smallest zeroth-order
777: eigenvalue is \beq q^{(0)}=D_{{\alpha_*}} \equiv {\rm
778: min}\{D_\alpha\}. \label{q0} \eeq
779:
780: Up to the coefficient $(4\pi/3)H_\alpha^{-4}$, $D_\alpha$ is the total
781: decay rate of vacuum $\alpha$. As we discussed in Section III, bubble
782: nucleation rates are exponentially suppressed in low-energy vacua with
783: $\Lambda_j\ll 1$. We therefore expect that the vacuum ${\alpha_*}$
784: corresponding to the smallest eigenvalue $q^{(0)}$ is one of the
785: low-energy vacua.
786:
787: With $\Lambda_{\alpha_*}\ll 1$ and $q_a$ not very small,
788: Eq.(\ref{DeltaLambda}) suggests that downward transitions from
789: ${\alpha_*}$ will bring us to the negative-$\Lambda$ territory of
790: terminal vacua. Terminal vacua do not belong in the matrix
791: $\mathbf{R}$; hence, $R_{\beta{\alpha_*}}=0$ for
792: $\beta\neq{\alpha_*}$, and it is easy to see that our zeroth order
793: eigenvector has a single nonzero component, \beq s^{(0)}_\alpha=
794: \delta_{\alpha{\alpha_*}}. \label{s0} \eeq Eq. (\ref{pJaume}) then
795: implies that the only vacua with nonzero probabilities at zeroth order
796: are the negative-$\Lambda$ descendants which can be reached by a
797: single downward jump from the dominant vacuum ${\alpha_*}$. (Note
798: that the vacuum $\alpha_*$ itself has zero probability at this order.)
799: %\footnote{\bf If some of the $q_a$ are small, the
800: %descendants of the vacuum $\alpha_*$ may include some recyclable
801: %vacua. The zeroth order eigenvector may have will then have nonzero
802: %components in the corresponding vacua. Our scheme can be easily
803: %generalized to this case.}
804:
805:
806: \subsection{First order}
807:
808: The full eigenvalue equation can be written as
809: \beq
810: ({\mathbf R^{(0)}}+{\mathbf R^{(1)}})({\bf s^{(0)}} + {\bf
811: s^{(1)}}) = -(q^{(0)}+q^{(1)}) ({\bf s^{(0)}} + {\bf
812: s^{(1)}}).
813: \eeq
814: Neglecting second-order terms and using the zeroth-order relation
815: \beq
816: {\mathbf R^{(0)}}{\bf s^{(0)}} = - q^{(0)}{\bf s^{(0)}},
817: \eeq
818: we obtain an equation for the first-order corrections,
819: \beq
820: ({\mathbf R^{(0)}}+q^{(0)}{\mathbf I}){\bf s^{(1)}} =
821: -({\mathbf R^{(1)}}+q^{(1)}{\mathbf I}){\bf s^{(0)}},
822: \label{firstorder}
823: \eeq
824: where ${\mathbf I}$ is the unit matrix.
825:
826: Eq. (\ref{firstorder}) is a system of ${\cal N}$ linear equations
827: for the ${\cal N}$ components of ${\bf s^{(1)}}$. Note, however,
828: that the triangular matrix multiplying ${\bf s^{(1)}}$ on the
829: left-hand side has a zero diagonal element, \beq ({\mathbf
830: R^{(0)}}+q^{(0)}{\mathbf I})_{{\alpha_*}{\alpha_*}} = 0, \eeq which
831: means that the determinant of this matrix vanishes, so it cannot be
832: inverted. In other words, the equations in (\ref{firstorder}) are
833: not all linearly independent.
834:
835: This problem can be cured by dropping the ${\alpha_*}$-th equation
836: in (\ref{firstorder}) and replacing it by a constraint equation, which
837: we choose to enforce the orthogonality of ${\bf s^{(1)}}$ and ${\bf
838: s^{(0)}}$,
839: \beq
840: ({\bf s^{(0)}},{\bf s^{(1)}}) =0.
841: \label{constraint}
842: \eeq
843: Note that the ${\alpha_*}$-th equation is the only equation in
844: (\ref{firstorder}) that involves the eigenvalue correction
845: $q^{(1)}$. Now $q^{(1)}$ has dropped out of our modified system, and
846: we can straightforwardly solve for $s^{(1)}_\alpha$. We did this
847: numerically for a $J=7$ model; the results will be presented in
848: the following subsection. A $J=2$ analytic toy model is worked out in
849: the Appendix.
850:
851: %The calculation of eigenvalue corrections and a more systematic
852: %development of perturbation theory for triangular matrices are
853: %discussed in Appendix A.
854:
855:
856: \section{Bubble abundances in the BP model}
857:
858: We found in the preceding section that the zeroth order of
859: perturbation theory picks the vacuum $\alpha_*$ which decays the
860: slowest (we call it the dominant vacuum), and assigns non-zero
861: probabilities to it's offspring only - all other probabilities are
862: zero. In the first order of perturbation theory, all vacua connected
863: to the dominant vacuum via one upward jump, and any vacua connected to
864: these via a series of downward transitions, also acquire non-zero
865: probabilities.
866:
867: The bubble abundance factors $p_j$ for the 7-dimensional toy model
868: (\ref{7D}) are shown in Fig. \ref{7Dprobabilities}. We plot
869: $\log_{10}(1/p_j)$ vs. $\Lambda_j$, so higher points in the figure
870: correspond to smaller bubble abundances. The first thing one
871: notices is that there are several groups of points, marked by
872: triangles, boxes, etc.
873: %diamonds, crosses, \textbf{points}, and a
874: The star marks the dominant vacuum $\alpha_*$.
875:
876:
877: \begin{figure}
878: \begin{center}
879: \leavevmode\epsfxsize=5in\epsfbox{7Dprobabilities.eps}
880: \end{center}
881: \caption{ Plot of $\log_{10}(1/p_j)$ vs. $\Lambda_j$ for the BP
882: model with parameters given in (\ref{7D}). The star marks the
883: dominant vacuum $\alpha_*$. Triangles represent vacua in group 1,
884: squares in group 2, diamonds in groups 3 and 6, crosses in groups 4
885: and 7, and points in groups 5 and 8.} \label{7Dprobabilities}
886: \end{figure}
887:
888: In this particular example, the dominant site has coordinates
889: $(1,1,1,1,1,1,1)$. There are $J=7$ ways to jump up one unit from this
890: site, to arrive at the seven points indicated by black triangles,
891: which we shall call group 1. The coordinates of these points are
892: $(2,1,1,1,1,1,1),~ (1,2,1,1,1,1,1),~..., ~(1,1,1,1,1,1,2)$.
893: Lower-energy vacua in this group have higher bubble abundances, in
894: accordance with Eqs.~(\ref{GammaI}),(\ref{ILambda}) of Section III.
895:
896: The next group of states results from downward jumps out of vacua in
897: group 1 in all possible directions, excluding the jumps back to the
898: dominant site $\alpha_*$. We call it group 2. The number of states
899: in this group is $J(J-1)=42$. Consider, for example, the subgroup
900: of states in group 2 coming from the downward transitions out of the
901: state $(2,1,1,1,1,1,1)$. These states have coordinates
902: $(2,0,1,1,1,1,1),~...,~(2,1,1,1,1,1,0)$. Since they originate from
903: the same single parent, the difference in their bubble abundances
904: comes from the difference in the instanton actions $I_{ij}$. This
905: effect is much milder for downward transitions than it is for the
906: upward ones. That is why the spread in bubble abundances within the
907: subgroups of group 2 is much smaller than it is in group 1.
908:
909: Further downward jumps replacing one of the $J-2=5$ remaining 1's by
910: a 0 give rise to group 3, consisting of $J(J-1)(J-2)/{2!}=105$
911: states having flux configurations with one $n=2$, four $n=1$ and two
912: $n=0$. Similarly, group 4 includes $J(J-1)(J-2)(J-3)/{3!}= 140$
913: states with one $n=2$, three $n=1$ and three $n=0$, and group 5
914: includes $J(J-1)(J-2)(J-3)(J-4)/4!= 105$ states with one $n=2$, two
915: $n=1$ and four $n=0$. The factorial factors are included to avoid
916: double counting. For example, the site $(2,0,1,1,0,1,1)$ can be
917: reached by downward jumps from either $(2,1,1,1,0,1,1)$ or
918: $(2,0,1,1,1,1,1)$ and would be counted twice if we did not divide by
919: $2!$.
920: %Some states in groups 4 and 5 are
921: %terminal vacua.
922:
923: If a vacuum in group 2 has a coordinate jump from $n=2$ to $n=1$,
924: the resulting site is one of the daughter sites which can also be
925: reached by downward jumps from the dominant site. These
926: negative-$\Lambda$ vacua have non-zero probabilities already at the
927: zeroth-order level and are not represented in the figure.
928:
929: If a vacuum in group 3, 4 or 5 has a coordinate jump from $n=2$ to
930: $n=1$, the resulting sites are all terminal vacua (groups 6, 7 and
931: 8, respectively).
932:
933: We note that although the dominant vacuum $\alpha_*$ is one of the
934: low-energy vacua, there are many other recyclable vacua which have
935: lower $\Lambda$. Recall that the dominant vacuum has the smallest,
936: in magnitude, sum of its transition rates down in each possible
937: direction (see Eq.'s (\ref{Ralpha}) and (\ref{q0})). Each
938: transition rate depends exponentially on the value of $q_a$ and the
939: factor $r(x,y)/(n_a-1/2)^3$. From this factor and Fig.(\ref{7Drvsz})
940: we see that for $\Lambda/\Delta\Lambda<1$ (this is the case for
941: $\alpha_*$) any jump from an $n=2$ flux quanta will be less
942: suppressed than a jump in the same direction from an $n=1$ flux
943: quanta. Thus we are not surprised that states which contain a flux
944: quanta of $n=2$ are \emph{not} dominant sites despite having smaller
945: $\Lambda$ than $\alpha_*$. Since each transition rate is
946: exponentially dependent on the tunneling exponent, typically the
947: largest (in magnitude) transition rate will dominate the sum in
948: Eq.(\ref{Ralpha}). Thus, essentially for a vacuum to be the dominant
949: state its largest (in magnitude) transition rate should be smaller
950: that the largest transition rate of any other vacuum.
951:
952: %Thus even if a given vacuum can only jump down in a small number of
953: %directions (thus having few terms in the sum), so long as one of
954: %those transitions is relatively unsuppressed, the vacuum could still
955: %have a larger (in magnitude) value for $D_\alpha$ than some other
956: %vacuum which might be able to jump down in more directions.
957:
958: The distribution in Fig. \ref{7Dprobabilities} was obtained in the
959: first order of perturbation theory, which includes only the vacua
960: which can be reached by a single upward jump from the dominant site
961: $\alpha_*$, followed by some downward jumps. If higher orders were
962: included, we would see additional groups of vacua, reachable only
963: with two or more upward jumps. These vacua would have much smaller
964: bubble abundances than those already in the figure.
965:
966: The distribution $p_j$ in Fig. \ref{7Dprobabilities} spans more than
967: 300 orders of magnitude. It differs dramatically from the flat
968: distribution (\ref{flat}) suggested by the heuristic argument in the
969: Introduction. Many vacua with close values of $\Lambda_j$ have very
970: different abundances $p_j$. The reason is that despite their
971: closeness in $\Lambda$, such vacua are located far away from one
972: another in the BP grid, and the paths leading to them from the
973: dominant vacuum $\alpha_*$ are characterized by exponentially
974: different transition rates. Even the vacua resulting from tunneling
975: out of the same site typically have very different abundances, due
976: to the exponential dependence of the tunneling rates on $q_a$.
977:
978: Fig. \ref{7Dprobset2} shows the distribution of bubble abundances
979: for another $J=7$ model, with a different set of parameters: \beq
980: {q_i}= {0.6175,~0.3909,~0.6472,~0.5508,~0.5103,~0.7036,
981: 0.4541};~~~~~~~\Lambda_{bare}=-1.033. \label{7D'} \eeq In this
982: case, there is more scatter in the values of $q_a$, and the groups
983: of vacua are somewhat less pronounced. However, the staggered nature
984: of the distribution is still apparent.
985:
986:
987: \begin{figure}
988: \begin{center}
989: \leavevmode\epsfxsize=5in\epsfbox{7Dprobset2.eps}
990: \end{center}
991: \caption{ Plot of $\log_{10}(1/p_j)$ vs. $\Lambda_j$ for the BP
992: model with parameters (\ref{7D'}). The star marks the dominant
993: vacuum $\alpha_*$. Different groups of vacua are represented by the
994: same symbols as in Fig.5.} \label{7Dprobset2}
995: \end{figure}
996:
997:
998:
999: \section{Discussion}
1000:
1001: In this paper we have used the prescription of Refs.
1002: \cite{GSPVW,ELM} to determine the bubble abundances $p_j$ in the BP
1003: model. We found that the resulting distribution is very irregular,
1004: with values of $p_j$ soaring and plummeting wildly as $\Lambda_j$
1005: changes from one value to the next. This distribution is drastically
1006: different from the flat distribution (\ref{flat}) which was used as
1007: a basis for the successful anthropic prediction for $\Lambda$.
1008:
1009: Apart from the bubble abundance factor $p_j$, the volume distribution
1010: (\ref{PpZ}) includes the slow-roll expansion factor $Z_j$. In any
1011: realistic model, bubble nucleation should be followed by a period of
1012: slow-roll inflation, at least in some bubble types. The expansion
1013: factor $Z_j$ is, of course, model-dependent, but there is no reason to
1014: expect that it will somehow compensate for the wild swings in the
1015: values of $p_j$ as we go from one value of $\Lambda_j$ to the next.
1016:
1017: Another point to keep in mind is that, in a realistic setting, vacua
1018: with different values of the fluxes $F_a$ may have different
1019: low-energy physics, so the density of observers $n_{obs}(\Lambda)$
1020: would also be very different. We should therefore focus on the
1021: subset of vacua in the BP grid which differ only by the value of
1022: $\Lambda$ and have essentially identical low-energy constants. Once
1023: again, there seems to be no reason to expect any correlation between
1024: these constants and the up and down swings in the bubble abundances.
1025: We conclude that the staggered character of the distribution $P_j
1026: \equiv P(\Lambda_j)$ is expected to persist, even in more realistic
1027: versions of the model.
1028:
1029: This conclusion is not limited to the BP model. It is likely to
1030: arise in any landscape scenario, where a dense spectrum of
1031: low-energy constants is generated from a wide distribution of states
1032: in the parameter space of the fundamental theory. Vacua with nearly
1033: identical values of $\Lambda$ may then come from widely separated
1034: parts of the landscape and may have very different bubble abundances
1035: and volume fractions.
1036:
1037: Given the staggered character of the volume distribution, what kind
1038: of prediction can we expect for the observed value of $\Lambda$? The
1039: answer depends on the number ${\cal N}_A$ of possible vacua with
1040: $\Lambda_j$ within the anthropic range (\ref{DeltaLambdaA}),
1041: $\Delta\Lambda_A \sim 10^{-118}$. (We count only vacua in which all
1042: low-energy constants other than $\Lambda$ have nearly the same
1043: values as in our vacuum.)
1044:
1045: Suppose the volume factors in the distribution $P_j$ span $K$ orders
1046: of magnitude. ($K\sim 300$ in our numerical example in Section V.)
1047: We can divide all vacua into, say, $10K$ bins, such that the values
1048: of $P_j$ in each bin differ by no more than $10\%$. Suppose now that
1049: there are ${\cal N}_A \ll 10K$ vacua in the anthropic range
1050: $\Delta\Lambda_A$. We can then expect that most of these vacua will
1051: be characterized by vastly different volume factors $P_j$, so that
1052: the entire range will be dominated by one or few values of
1053: $\Lambda_j$ having much higher volume fractions than the rest.
1054:
1055: Moreover, there is a high likelihood of finding still greater volume
1056: fractions if we go somewhat beyond the anthropic range - simply
1057: because we would then search in a wider interval of $\Lambda$. We
1058: could, for example, find that a vacuum with $\Lambda_1 \sim
1059: 10^{-114}\sim 10^6\Lambda_0$ has a volume fraction 200 orders of
1060: magnitude greater than all other vacua in the range
1061: $0<\Lambda\lesssim \Lambda_1$. Galaxy formation is strongly
1062: suppressed in this vacuum: the fraction of matter that ends up in
1063: galaxies is only $f_G(\Lambda_1)\sim 10^{-110}$. However, this
1064: suppression is more than compensated for by the enhancement in the
1065: volume fraction.
1066:
1067: If this were the typical situation, most observers would find
1068: themselves in rare, isolated galaxies, surrounded by nearly empty
1069: space, all the way to the horizon. This is clearly not what we
1070: observe. The dominant value could by chance be very close to
1071: $\Lambda=0$, but if such an ``accident'' is required to explain the
1072: data, the anthropic model loses much of its appeal.
1073:
1074: In the opposite limit, ${\cal N}_A\gg 10K$, the number of vacua in the
1075: anthropic interval $\Delta\Lambda_A$ is so large that they may scan the
1076: entire range of $P_j$ many times. Then, it is conceivable that the
1077: distribution will become smooth after averaging over some suitable
1078: scale $\delta\Lambda$. If $\delta\Lambda$ can be chosen much smaller
1079: than $\Delta\Lambda_A$, then it is possible that the effective,
1080: averaged distribution will be flat, as suggested by the heuristic
1081: argument in the Introduction. The successful prediction for $\Lambda$
1082: would then be unaffected.\footnote{Joe Polchinski has informed us that
1083: a similar argument, indicating that the anthropic explanation for the
1084: observed $\Lambda$ requires a large number of vacua in the anthropic
1085: range, was suggested to him by Paul Steinhardt.}
1086:
1087: The above argument is somewhat simplistic, as it assumes that the
1088: vacua in the BP grid are more or less randomly distributed between the
1089: $10K$ bins, with roughly the same number of vacua in each bin. Such an
1090: ``equipartition'' is not likely to apply to the most abundant vacua,
1091: but it may hold for the vacua in the mid-range of $P_j$. Finding the
1092: conditions under which equipartition applies would require a
1093: statistical analysis that goes beyond the scope of the present paper.
1094:
1095: In summary, it appears that the staggered volume distribution
1096: resulting from the BP model is in conflict with observations, unless
1097: it yields a huge number of vacua in the anthropic range of
1098: $\Lambda$. Counting only vacua which have nearly the same low-energy
1099: physics as ours, we should have much more than $10K$; hence, the
1100: total number of vacua should be many orders of magnitude greater.
1101: The large number of vacua in the anthropic range is only a necessary
1102: condition for the distribution $P_j$ to average out to the flat
1103: distribution (\ref{flat}). Further analysis will be needed to find
1104: whether or not this actually happens, and if so, then under what
1105: conditions. It would also be interesting to analyze other simple
1106: models of the landscape, such as the ``predictive landscape'' of
1107: Arkani-Hamed, Dimopoulos and Kachru \cite{AHDK}, and see what
1108: similarities and differences they have compared to the BP model.
1109:
1110: Throughout this paper we assumed that the brane charges $q_a$ are
1111: not particularly small. This assumption may be violated in certain
1112: parts of the landscape, e.g., in the vicinity of conifold points,
1113: resulting in a much denser spectrum of vacua
1114: \cite{FMSW,AshokDouglas,DenefDouglas,Giryavets}. Infinite
1115: accumulations of vacua may occur near certain attractor points
1116: \cite{DVattractor,Gia}. Implications of these effects for the
1117: probability distribution on the landscape remain to be explored.
1118:
1119: \section{Acknowledgements}
1120:
1121: We are grateful to Michael Douglas, Gia Dvali, Jaume Garriga, Ken
1122: Olum and Joseph Polchinski for useful comments and discussions. This
1123: work was supported in part by the National Science Foundation.
1124:
1125:
1126: \section{Appendix}
1127:
1128: We will illustrate our perturbative method of calculation on a very
1129: basic BP model, which can be solved analytically. We consider $9$
1130: vacua arranged in a 2-D grid and labeled as indicated in Fig.
1131: \ref{2DanalyticBP}. There are three terminal vacua labeled 1, 2, 4,
1132: and six non-terminal vacua, 3, 5, 6, 7, 8, 9 in this model. We allow
1133: upward and downward transitions between nearest neighbor pairs, with
1134: transitions from non-terminal to terminal states allowed, but no
1135: transitions may take place from a terminal state. For simplicity,
1136: we disregard the vacua in the quadrants where $n_1<0$ and/or $n_2<0$
1137: and assume that the set of vacua in Fig.(\ref{2DanalyticBP}) is all
1138: there is.
1139:
1140: \begin{figure}
1141: \begin{center}
1142: \leavevmode\epsfxsize=5in\epsfbox{2DanalyticBP.eps}
1143: \end{center}
1144: \caption{The arrangement of vacua for a $J=2$, $N=2$ BP
1145: grid } \label{2DanalyticBP}
1146: \end{figure}
1147:
1148:
1149: The evolution equations for the set of \emph{non-terminal} vacua is
1150: \be
1151: \frac{d\mathbf{f}}{dt}=\mathbf{R}\mathbf{f},
1152: \ee
1153: where the vector $\mathbf{f} \equiv \{f_3, f_5, f_6, f_7, f_8, f_9\}$.
1154: Assuming that upward transition rates are far more suppressed
1155: than downward transition rates, we represent the transition matrix as
1156: \be
1157: \mathbf{R}=\mathbf{R^{(0)}}+
1158: \mathbf{R^{(1)}},
1159: \ee
1160: where
1161: \be
1162: \mathbf{R^{(0)}}=
1163: \left(%
1164: \begin{array}{cccccc}
1165: -D_3 & 0 & \kappa_{36} & 0 & 0 & 0 \\
1166: 0 & -D_5& \kappa_{56} & 0 & \kappa_{58}& 0 \\
1167: 0& 0& -D_6& 0 & 0& \kappa_{69} \\
1168: 0 & 0 & 0& -D_7 & \kappa_{78}& 0 \\
1169: 0 & 0 & 0& 0& -D_8 & \kappa_{89}\\
1170: 0 & 0 & 0& 0& 0&-D_9 \\
1171: \end{array}%
1172: \right)
1173: \label{R0}
1174: \ee
1175: and
1176: \be
1177: \mathbf{R^{(1)}}=
1178: \left(%
1179: \begin{array}{cccccc}
1180: -U_3 & 0 & 0& 0 & 0 & 0 \\
1181: 0 & -U_5& 0 & 0 & 0& 0 \\
1182: \kappa_{63}& \kappa_{65}& -U_6& 0 & 0& 0 \\
1183: 0 & 0 & 0& -U_7 & 0& 0 \\
1184: 0 & \kappa_{85} & 0& \kappa_{87}& -U_8 & 0\\
1185: 0 & 0 & \kappa_{96}& 0& \kappa_{98}&-U_9 \\
1186: \end{array}%
1187: \right)
1188: \label{R1}
1189: \ee
1190: and we have defined
1191: \beqa
1192: D_\alpha \equiv
1193: \sum_{j<\alpha}\kappa_{j\alpha},\\ U_\alpha \equiv
1194: \sum_{j>\alpha}\kappa_{j\alpha}.
1195: \eeqa
1196: $ D_\alpha$ and $U_\alpha$ represent, respectively, the sum of the
1197: downward and upward transition rates from vacuum $\alpha$.
1198:
1199: In our toy model we will assume that vacuum $5$ has the smallest (in
1200: magnitude) sum of downward transition rates, and therefore $-q^{(0)} =
1201: -D_5$ is the zero'th order dominant eigenvalue. By inspection, we see
1202: that the corresponding eigenvector is $\mathbf{s^{(0)}} \equiv \{0, 1, 0,
1203: 0, 0, 0\}$.
1204:
1205: We now need to calculate the first order correction to this
1206: eigenvector, $\mathbf{s^{(1)}}$. Substituting (\ref{R0}),(\ref{R1}) in
1207: Eq.~(\ref{firstorder}), we find
1208: \be
1209: \left(%
1210: \begin{array}{cccccc}
1211: q^{(0)}-D_3 & 0 & \kappa_{36} & 0 & 0 & 0 \\
1212: 0 & 0& \kappa_{56} & 0 & \kappa_{58}& 0 \\
1213: 0& 0& q^{(0)}-D_6& 0 & 0& \kappa_{69} \\
1214: 0 & 0 & 0&q^{(0)} -D_7 & \kappa_{78}& 0 \\
1215: 0 & 0 & 0& 0& q^{(0)}-D_8 & \kappa_{89}\\
1216: 0 & 0 & 0& 0& 0&q^{(0)}-D_9 \\
1217: \end{array}%
1218: \right)\mathbf{s^{(1)}} =\left(%
1219: \begin{array}{c}
1220: 0 \\
1221: -q^{(1)}+U_5 \\
1222: -\kappa_{65}\\
1223: 0 \\
1224: -\kappa_{85}\\
1225: 0 \\
1226: \end{array}%
1227: \right)
1228: \label{matrixinvert}
1229: \ee
1230: %Notice how the special form of $\mathbf{s^{(0)}}$ picks out only one
1231: %column of $(\mathbf{I}E_5^{(1)}-\mathbf{R^{(1)}})$.
1232: Note that the only equation in this set that depends on the first
1233: order correction to the eigenvalue is also the equation that needs to
1234: be dropped from our system, since it has a zero diagonal element -
1235: this causes the matrix on the right-hand side of
1236: (\ref{matrixinvert}) to have a zero determinant, which
1237: renders it uninvertible.
1238:
1239: This drop in the number of independent equations is replenished by
1240: including the constraint equation (\ref{constraint}); the resulting
1241: set of equations is
1242: \be \left(%
1243: \begin{array}{cccccc}
1244: q^{(0)}-D_3 & 0 & \kappa_{36} & 0 & 0 & 0 \\
1245: 0 & 1& 0 & 0 & 0& 0 \\
1246: 0& 0& q^{(0)}-D_6& 0 & 0& \kappa_{69} \\
1247: 0 & 0 & 0&q^{(0)} -D_7 & \kappa_{78}& 0 \\
1248: 0 & 0 & 0& 0& q^{(0)}-D_8 & \kappa_{89}\\
1249: 0 & 0 & 0& 0& 0&q^{(0)}-D_9 \\
1250: \end{array}%
1251: \right)\mathbf{s^{(1)}} =\left(%
1252: \begin{array}{c}
1253: 0 \\
1254: 0 \\
1255: -\kappa_{65}\\
1256: 0 \\
1257: -\kappa_{85}\\
1258: 0 \\
1259: \end{array}%
1260: \right)\ee
1261:
1262: The solution is readily determined, and we obtain
1263: \be
1264: \mathbf{s^{(1)}} = \left(%
1265: \begin{array}{c}
1266: \frac{\kappa_{36}}{D_3-D_5} \frac{\kappa_{65}} {D_6-D_5}\\
1267: 0 \\
1268: \frac{\kappa_{65}}{D_6- D_5}\\
1269: \frac{\kappa_{78}}{D_7-D_5} \frac{\kappa_{85}} {D_8-D_5} \\
1270: \frac{\kappa_{85}} {D_8-D_5}\\
1271: 0 \\
1272: \end{array}%
1273: \right) \ee This can now be used in Eq.(\ref{pJaume}) to determine
1274: the bubble abundances. For example, comparing the bubble abundances
1275: in vacua 3 and 7, we find \be \frac{p_3}{p_7}= {H_6^q
1276: \kappa_{36}\kappa_{65}(D_8 - D_5)
1277: \over{H_8^q \kappa_{78}\kappa_{85}(D_6 - D_5)}} \ee
1278:
1279: %\be \frac{p_3}{p_7}= \frac{H_6^{q} \kappa_{36}\frac{ \kappa_{65}}
1280: %{D_6-D_5}}{H_8^{q} \kappa_{78} \frac{\kappa_{85}} {D_8-D_5}} \ee
1281:
1282:
1283: \begin{thebibliography}{9}
1284:
1285:
1286: \bibitem{Weinberg87}
1287: S. Weinberg, Phys. Rev. Lett. {\bf 59}, 2607 (1987).
1288:
1289: \bibitem{Linde87}
1290: A.D. Linde, in {\it 300 Years of Gravitation}, ed. by S.W. Hawking and
1291: W. Israel, (Cambridge University Press, Cambridge, 1987).
1292:
1293: \bibitem{AV95}
1294: A. Vilenkin, Phys.\ Rev.\ Lett.\ {\bf 74}, 846 (1995).
1295:
1296: \bibitem{Efstathiou}
1297: G. Efstathiou, M.N.R.A.S. {\bf 274}, L73 (1995).
1298:
1299: \bibitem{MSW}
1300: H.~Martel, P.~R.\ Shapiro and S.~Weinberg, Ap.J.\ {\bf 492},
1301: 29 (1998).
1302:
1303: \bibitem{GLV} J. Garriga, M. Livio and A. Vilenkin, Phys. Rev. {\bf
1304: D61}, 023503 (2000).
1305:
1306: \bibitem{Bludman}
1307: S. Bludman, Nucl. Phys. {\bf A663-664},865 (2000).
1308:
1309: \bibitem{AV05}
1310: For a review, see, e.g., A. Vilenkin, ``Anthropic predictions: the
1311: case of the cosmological constant'', astro-ph/0407586.
1312:
1313: \bibitem{AV96}
1314: A. Vilenkin, in {\it Cosmological Constant and the Evolution of the
1315: Universe}, ed by K. Sato, T. Suginohara and N. Sugiyama (Universal
1316: Academy Press, Tokyo, 1996).
1317:
1318: \bibitem{Weinberg96}
1319: S. Weinberg, in {\it Critical Dialogues in Cosmology}, ed. by
1320: N.~G. Turok (World Scientific, Singapore, 1997).
1321:
1322: \bibitem{Pogosian}
1323: L. Pogosian, private communication.
1324:
1325: \bibitem{Banks84}
1326: T. Banks, Phys. Rev. Lett. {\bf 52}, 1461 (1984).
1327:
1328: \bibitem{Weinberg00}
1329: S. Weinberg, Phys. Rev. {\bf D61} 103505 (2000); astro-ph/0005265.
1330:
1331: \bibitem{GV00} J. Garriga and A. Vilenkin, Phys. Rev. {\bf D61},
1332: 083502 (2000).
1333:
1334: \bibitem{GV03}
1335: J. Garriga and A. Vilenkin, Phys. Rev. {\bf D67}, 043503 (2003)
1336: (corrected version at astro-ph/0210358).
1337:
1338: \bibitem{Abbott}
1339: L. F. Abbott, Phys. Lett. \textbf{B150}, 427 (1985).
1340:
1341: \bibitem{BT}
1342: J. D. Brown and C. Teitelboim, Phys. Lett. \textbf{B195}, 177 (1987);
1343: Nucl. Phys. \textbf{B297}, 787 (1988).
1344:
1345: \bibitem{GV01}
1346: J. Garriga and A. Vilenkin, Phys. Rev. {\bf D64}, 023517 (2001).
1347:
1348: \bibitem{Donoghue00}
1349: J. Donoghue, JHEP 0008:022 (2000).
1350:
1351: \bibitem{FMSW}
1352: J.L. Feng, J. March-Russell, S. Sethi and F. Wilczek, Nucl. Phys. {\bf
1353: B602}, 307 (2001).
1354:
1355: \bibitem{BDM}
1356: T. Banks, M. Dine and L. Motl, JHEP {\bf 0101}, 031 (2001).
1357:
1358: \bibitem{DV01}
1359: G. Dvali and A. Vilenkin, Phys. Rev. {\bf D64}, 063509 (2001).
1360:
1361: \bibitem{BP}
1362: R. Bousso and J. Polchinski, JHEP {\bf 0006}, 006 (2000).
1363:
1364: \bibitem{LLM}
1365: A.~D.~Linde, D.~A.~Linde, and A.~Mezhlumian, Phys.\ Rev.\ D {\bf 49},
1366: 1783 (1994).
1367:
1368: \bibitem{VVW}
1369: V.~Vanchurin, A.~Vilenkin, and S.~Winitzki, Phys.\ Rev.\ D {\bf 61},
1370: 083507 (2000).
1371:
1372: \bibitem{Guth00}
1373: A.~H.~Guth, Phys.\ Rept.\ {\bf 333}, 555 (2000).
1374:
1375: \bibitem{Tegmark}
1376: M.~Tegmark, JCAP {\bf 0504}, 001 (2005).
1377:
1378: \bibitem{GSPVW}
1379: J. Garriga, D. Schwartz-Perlov, A. Vilenkin and S. Winitzki,
1380: ``Probabilities in the inflationary multiverse'',
1381: hep-th/0509184.
1382:
1383: \bibitem{KKLT}
1384: S. Kachru, R. Kallosh, A. Linde and S. Trivedi, Phys. Rev. {\bf D68},
1385: 046005 (2003).
1386:
1387: \bibitem{Douglas}
1388: M.R. Douglas, JHEP {\bf 0305}, 046 (2003).
1389:
1390: \bibitem{AshokDouglas}
1391: S. Ashok and M.R. Douglas, JHEP {\bf 0401}, 060 (2004).
1392:
1393: \bibitem{DenefDouglas}
1394: F. Denef and M.R. Douglas, JHEP {\bf 0405}, 072 (2004).
1395:
1396: \bibitem{ELM}
1397: R. Easther, E.A. Lim and M.R. Martin, ``Counting pockets with world
1398: lines in eternal inflation'', arXiv:astro-ph/0511233.
1399:
1400: \bibitem{recycling}
1401: J. Garriga and A. Vilenkin, Phys. Rev. {\bf D57}, 2230 (1998).
1402:
1403: \bibitem{Megevand}
1404: J. Garriga and A. Megevand, Phys. Rev. {\bf D69}, 083510 (2004).
1405:
1406: \bibitem{CdL}
1407: S. Coleman and F. DeLuccia, Phys. Rev. {\bf D21}, 3305 (1980).
1408:
1409: \bibitem{Parke}
1410: S. Parke, ``Gravity and the decay of the false vacuum'',
1411: {Phys}.\ Letters\ B {\bf 121} (1983) 313.
1412:
1413: \bibitem{Jaume}
1414: J. Garriga, {Phys}.\ Rev.\ D {\bf 49} (1994) 6327.
1415:
1416: \bibitem{EWeinberg}
1417: K.~M.~Lee and E.~J.~Weinberg, Phys.\ Rev.\ D {\bf 36}, 1088 (1987).
1418:
1419: \bibitem{Susskind}
1420: L.~Susskind, ``The anthropic landscape of string theory,''
1421: arXiv:hep-th/0302219.
1422:
1423: \bibitem{Giryavets}
1424: A. Giryavets, S. Kachru and P.K. Tripathy, JHEP 0408, 002 (2004).
1425:
1426: \bibitem{DVattractor}
1427: G. Dvali and A. Vilenkin, Phys. Rev. {\bf D70}, 063501 (2004).
1428:
1429: \bibitem{Gia}
1430: G. Dvali, ``Large hierarchies from attractor vacua'', arXiv:hep-th/0410286.
1431:
1432: \bibitem{AHDK}
1433: N. Arkani-Hamed, S. Dimopoulos and S. Kachru, ``Predictive landscapes and new physics at a TeV'', arXiv:hep-th/0501082.
1434:
1435:
1436: \end{thebibliography}
1437:
1438: \end{document}
1439: