1:  \documentstyle[psfig,12pt]{article}

2: \textheight=8.5truein

3: \textwidth=6.1truein

4: 

5: \overfullrule=0pt

6: \parskip=2pt

7: \parindent=12pt

8: \headheight=0in

9: \headsep=0in

10: \topmargin=0in

11: \oddsidemargin=0in

12: %--------+---------+---------+---------+---------+---------+---------+

13: \def\arccoth{\mop{arccoth}}

14: \def\href#1#2{#2}

15: \def\arcsinh{\mop{arcsinh}}

16: \def\arccosh{\mop{arccosh}}

17: \def\beq{\begin{equation}}

18: \def\eeq{\end{equation}}

19: \def\Re{\mop{Re}}

20: \def\Tr{\mop{Tr}}

21: \def\Im{\mop{Im}}

22: \def\W{{\cal A}}

23: \def\d{{\nabla}}

24: \def\comment#1{  \begin{raggedright}{\tt [#1]}\end{raggedright}}

25: %--------+---------+---------+---------+---------+---------+---------+

26: \begin{document}

27: \baselineskip=14.5pt

28: \pagestyle{plain}

29: \setcounter{page}{1}

30: \renewcommand{\thefootnote}{\fnsymbol{footnote}}

31: \newcommand{\da}{\dot{a}}

32: \newcommand{\db}{\dot{b}}

33: \newcommand{\dn}{\dot{n}}

34: \newcommand{\dda}{\ddot{a}}

35: \newcommand{\ddb}{\ddot{b}}

36: \newcommand{\ddn}{\ddot{n}}

37: \newcommand{\pa}{a^{\prime}}

38: \newcommand{\pb}{b^{\prime}}

39: \newcommand{\pn}{n^{\prime}}

40: \newcommand{\ppa}{a^{\prime \prime}}

41: \newcommand{\ppb}{b^{\prime \prime}}

42: \newcommand{\ppn}{n^{\prime \prime}}

43: \newcommand{\fda}{\frac{\da}{a}}

44: \newcommand{\fdb}{\frac{\db}{b}}

45: \newcommand{\fdn}{\frac{\dn}{n}}

46: \newcommand{\fdda}{\frac{\dda}{a}}

47: \newcommand{\fddb}{\frac{\ddb}{b}}

48: \newcommand{\fddn}{\frac{\ddn}{n}}

49: \newcommand{\fpa}{\frac{\pa}{a}}

50: \newcommand{\fpb}{\frac{\pb}{b}}

51: \newcommand{\fpn}{\frac{\pn}{n}}

52: \newcommand{\fppa}{\frac{\ppa}{a}}

53: \newcommand{\fppb}{\frac{\ppb}{b}}

54: \newcommand{\fppn}{\frac{\ppn}{n}}

55: \newcommand{\A}{A}

56: \newcommand{\B}{B}

57: \newcommand{\mmu}{\mu}

58: \newcommand{\mnu}{\nu}

59: \newcommand{\ii}{i}

60: \newcommand{\jj}{j}

61: \newcommand{\jl}{[}

62: \newcommand{\jr}{]}

63: \newcommand{\ml}{\sharp}

64: \newcommand{\mr}{\sharp}

65: 

66: %--------+---------+---------+---------+---------+---------+---------+

67: 

68: \begin{flushright}

69: UW/PT 04-24\\

70: {\tt hep-th/0412017}

71: \end{flushright}

72: 

73: \vskip 2cm

74: 

75: \begin{center}

76: {\Large \bf Hawking and Page on the brane} \\

77: \vskip 1cm

78: 

79: {\bf A.~Chamblin$^1$ and A.~Karch$^2$} \\

80: \vskip 0.5cm

81: {\it $^1$  Department of Physics, University of Louisville,

82: Louisville, KY 40292}\\

83: {\tt E-mail: chamblin@mit.edu} \\

84: \medskip

85: {\it $^2$ Department of Physics, University of Washington,\\

86: Seattle, WA 98195-1560} \\

87: {\tt E-mail: karch@phys.washington.edu}  \\

88: \end{center}

89: 

90: \vskip1cm

91: 

92: \begin{center}

93: {\bf Abstract}

94: \end{center}

95: \medskip

96: We show that the Hawking-Page phase transition of a CFT on AdS$_{d-1}$ weakly

97: coupled to gravity has a dual bulk description in terms of a phase

98: transition between a black string and a thermal gas on AdS$_{d}$.

99: At even lower temperatures the black string develops a Gregory

100: Laflamme instability, which is dual to black hole evaporation in the

101: boundary theory.

102: 

103: \newpage

104: 

105: 

106: \section{Introduction}

107: 

108: Gravitational theories on Anti de-Sitter space undergo an

109: interesting phase transition as the temperature is varied, the

110: well known Hawking-Page phase transition \cite{hp}. At high

111: temperatures a large black hole is formed, whereas at low

112: temperatures a thermal gas in the AdS$_d$ space with curvature

113: radius $L$ is the preferred configuration. A first order

114: transition occurs between the two geometries at a temperature

115: $T_{HP}=\frac{d-2}{2 \pi L}$. Using AdS/CFT the corresponding phase

116: transition gets mapped to a confinement/deconfinement transition

117: in the CFT on the boundary. As typical in a system with a first

118: order transition a second interesting temperature $T_S=

119: \frac{\sqrt{(d-2)^2-1}}{2 \pi L}<T_{HP}$ is associated with the

120: ``spinodal'', the point at which the false minimum disappears: In

121: the regime $T_S<T<T_{HP}$ both configurations are allowed, but the

122: thermal AdS is preferred. The black hole however still has a

123: positive specific heat, and like with water vapor lowered below

124: the boiling point, one can go to a supercooled quark gluon plasma

125: that is still stable against small thermal fluctuations and needs

126: to nucleate bubbles in order to transition to the global free

127: energy minimum. Below $T_S$ the specific heat of the black hole

128: becomes negative and the black hole configuration ceases to exist

129: even as a local minimum of the free energy.

130: 

131: Most of the discussion of the Hawking-Page phase transition in the

132: context of AdS/CFT focuses on the transition in the $d$

133: dimensional bulk and its $d-1$ dimensional field theory

134: interpretation. But AdS/CFT also allows us to study a conformal

135: field theory on AdS$_{d-1}$ as the {\it boundary geometry} \cite{kr}, weakly

136: coupled to gravity by introducing a UV cutoff in the form of an RS

137: Planck brane \cite{RS2}. In this setup the theory on the

138: boundary will undergo a Hawking-Page phase transition, and now one

139: can ask what the bulk description of both the phase transition and

140: the onset of the black hole instability at the spinodal point are

141: in terms of the bulk. In this paper we will show that former maps

142: to a novel Hawking-Page like transition in the bulk between a

143: black string and thermal AdS, while latter maps to classic

144: instability in the bulk of the Gregory Laflamme type \cite{GL}. A

145: similar treatment has appeared in \cite{emparan1,emparan2} and we

146: will discuss the relation between their findings and our results.

147: 

148: The main motivation is to test the gauge/gravity duality in the

149: context of cutoff AdS spaces like they appear in the RS scenario.

150: While AdS/CFT is well established in the conformal case, the

151: duality between a cutoff CFT coupled to gravity and the bulk

152: gravity-only theory has so far mostly been tested in the regime

153: where the localized graviton can be treated linearly. By studying

154: the Hawking-Page transition in this setup we can put this duality,

155: which is often invoked for model building purposes, on firmer

156: footing.

157: 

158: \section{The Hawking-Page Phase transition on the boundary}

159: 

160: \subsection{General Setup}

161: 

162: In order to establish the phase transition, one calculates the

163: difference of the on-shell gravitational action, \beq

164: \label{action} I = - M_{Pl}^{d-2} \int \sqrt{-g} d^dx (R -

165: 2\Lambda), \eeq evaluated on the two solutions, which gets

166: identified with the difference in $\beta F$, where $F$ is the free

167: energy and $\beta$ the inverse temperature. The two configurations

168: are thought of as two saddle points that dominate the euclidean

169: path integral. Note that the contribution of matter fields to the

170: action is ignored. This is reasonable as long as $M_{Pl}$ is

171: larger than the other scales, so that the gravitational

172: contribution to the action dominates. Note in particular that as

173: dynamical gravity is switched off by taking Newton's constant to

174: zero, that is $M_{pl}$ goes to infinity, the difference in free

175: energies becomes extremely large and the path integral over

176: metrics will still be dominated by one or the other configuration,

177: so the Hawking Page transition persists in that limit. Of course

178: the time it takes to actually transition from one geometry to the

179: other will go to infinity as dynamical gravity is switched off.

180: This is precisely the situation that one encounters when choosing

181: to study AdS$_{d}$/CFT$_{d-1}$ with an AdS$_{d-1}$ metric on the

182: boundary. Writing AdS$_d$ in AdS$_{d-1}$ slicing where the metric

183: $ds^2_{AdS_{d-1}}$ has curvature radius $L$ as well, \beq

184: \label{ads} ds^2 = \cosh^2 (\frac{r}{L}) \, ds^2_{AdS_{d-1}} +

185: dr^2 \eeq one obtains as a dual a CFT on AdS$_{d-1}$, or more

186: precisely two copies of AdS$_{d-1}$ at $r=\pm \infty$, which

187: communicate with each other via their common boundary \cite{kr} as

188: we will review in detail in the next subsection.

189: 

190: Introducing a UV brane at $r=r_0$ serves as a UV cutoff in the

191: field theory and in addition introduces dynamical gravity with an

192: almost massless graviton, where the mass can be understood as

193: being a CFT loop effect \cite{porrati}. The curvature radius on

194: the brane is $l=\cosh(\frac{r_0}{L}) L$. So at least for $r_0 >>

195: L$ we are precisely in the regime where the Hawking-Page analysis

196: applies. Even when we completely remove the UV brane and the $d-1$

197: dimensional Planck scale goes to infinity so that the path

198: integral over metrics is just equal to a sum over saddle points,

199: we still have a different configuration dominating at low and high

200: temperatures. To actually make a transition from one to the other

201: would however use an infinite amount of time. In this paper we

202: will precisely consider this $r_0>>L$ regime of a boundary CFT on

203: AdS$_{d-1}$ weakly coupled to gravity. As we discussed, going

204: beyond this regime will result in significant changes to the

205: Hawking Page picture due to the matter effects. While a detailed

206: discussion of those is clearly beyond the scope of the present

207: paper, we can identify how they manifest themselves on the bulk

208: side.

209: 

210: \subsection{Boundary Conditions: Interplay of two CFTs}

211: 

212: The bulk metric only defines the conformal structure at the

213: boundary. By picking a defining function suggested by the slicing,

214: in our case $\cosh(\frac{r}{L})$, one picks a particular metric on

215: the boundary. This way different coordinate systems in AdS give

216: different metrics on the boundary, but they all have the same

217: conformal structure. The difference becomes physically once

218: conformal invariance is broken, e.g. in our case by the cutoff

219: brane(s) and by the finite temperature.

220: 

221: To understand better the case of interest in this paper, where the

222: boundary consists of two copies of AdS$_{d-1}$, one can first of

223: all consider the conformal case of pure AdS$_d$ in the bulk, where

224: the physics is equivalent to the standard CFT on the $S^{d-2}

225: \times R$ boundary. Many of the issues discussed here can be found

226: also in \cite{kr,porrati}. The two copies of AdS$_{d-1}$ can be

227: conformally mapped to the southern and northern hemisphere of the

228: $S^{d-2} \times R$ boundary respectively. The conformal boundary

229: of AdS$_d$ gets realized as two disjoint AdS$_{d-1}$ spaces which

230: communicate via boundary conditions, so that together they have

231: the right topology of the $S^{d-2} \times R$ boundary to which

232: they are conformally related. The boundary conditions that need to

233: be imposed on the two copies of AdS$_{d-1}$ follow directly from

234: this conformal map: on the $S^{d-2}$ sphere the equator is not a

235: special locus at all, all fields have to be continuous and

236: differentiable across it. When mapped to two copies of AdS$_{d-1}$

237: this implies that the value of the field and its derivative close

238: to the boundary of one AdS$_{d-1}$ determine the same information

239: for the same field at the same point of the $S^{d-3} \times R$

240: boundary of the other AdS$_{d-1}$. For a lightray hitting the

241: boundary one imposes transparent boundary conditions instead of

242: the usual total reflecting ones, that is the signal leaves one

243: AdS$_{d-1}$ and enters the other AdS$_{d-1}$. The holographic dual

244: in this case is either a CFT living on $S^{d-2} \times R$ or

245: equivalently (simply related by a conformal transformation): two

246: copies of AdS$_{d-1}$, each with its own CFT, communicating with

247: each other via the transparent boundary conditions we just

248: described.

249: 

250: As soon as one introduces the cutoff conformal invariance is

251: broken, the boundary theory becomes sensitive to the actual

252: boundary metric and not just the conformal structure. In the bulk

253: the difference between distinct coordinate systems becomes

254: physical -- a brane at $r=const.$ in AdS$_{d-1}$ slicing is at a

255: different location than a brane in the standard Minkowski slicing.

256: To discuss the most general scenario let us introduce two cutoff

257: branes, one at a positive value of $r$, $r=r_0$, and one at a

258: negative value of $r$, $r=-r_1$. The holographic dual in this case

259: is hence again two copies of AdS$_{d-1}$, each with its own CFT

260: and dynamical graviton. $r_0$ and $r_1$ set the curvature radius

261: of the two AdS$_{d-1}$ spaces respectively. The two theories

262: communicate with each other via the transparent boundary

263: conditions for all fields including the graviton. Each AdS$_{d-1}$

264: is hence described by $d-1$ dimensional gravity coupled to a $d-1$

265: dimensional CFT and is expected to undergo its own Hawking Page

266: phase transition. The only departure from the standard analysis is

267: that instead of the reflecting boundary conditions here we have

268: the transparent boundary conditions that link the two AdS$_{d-1}$

269: spaces. Naively one would expect that this does not influence the

270: properties of the black hole: as long as the temperature in the

271: two AdS$_{d-1}$ spaces is the same, it should not matter if the

272: black hole is in a thermal bath of gravitons that get reflected

273: from the boundary or in a thermal bath of gravitons at the same

274: temperature that freely move from one AdS$_{d-1}$ to the other.

275: This expectation will be born out by computations in this paper.

276: 

277: One interesting special case is $r_1 \rightarrow \infty$, in which

278: case dynamical gravity switches off in one of the two AdS$_{d-1}$.

279: Another one is $r_1=r_0$. In this case one can impose an orbifold

280: projection $r \rightarrow -r$. In the orbifolded theory the

281: boundary is just one copy of AdS$_{d-1}$ and the transparent

282: boundary conditions turn into the regular reflecting boundary

283: conditions. Both those special cases are completely consistent

284: with the analysis in this paper.

285: 

286: Last but not least let us emphasize that even in the limit that

287: both cutoffs are taken to infinity one does not simply recover the

288: standard bulk Hawking Page physics. In the standard scenario one

289: demands the metric to take the form $S^{d-2}$ $\times$ $S^1$ close

290: to the boundary, corresponding to a CFT on the sphere at finite

291: temperature. Since we are interested in the scenario where gravity

292: on the AdS$_{d-1}$ slices is dynamical, we allow for non-trivial

293: geometry on those slices and only fix the boundary conditions on

294: the $S^{d-3}$ $\times$ $S^1$ along which the two AdS$_{d-1}$

295: components of the boundary communicate. While in the cutoff to

296: infinity limit dynamical gravity decouples, one of course only

297: obtains a smooth limit if one does not suddenly change the

298: boundary conditions. In particular the black string solution we

299: will introduce momentarily, which gives rise to two boundary

300: AdS$_{d-1}$ Schwarzschild black holes, would be allowed by our

301: boundary conditions and not by the standard bulk Hawking Page

302: analysis. As is apparent from formula (\ref{free}) the free energy

303: of the black string is negative and grows exponentially in the

304: cutoff $r_0$, whereas the free energy of the large AdS$_d$ black

305: hole is independent of the cutoff.  As soon as the boundary

306: conditions allow for the black string, it completely dominates the

307: thermal ensemble for large values of the cutoff and the physics is

308: genuinely different from the standard bulk Hawking Page analysis.

309: 

310: 

311: 

312: 

313: \subsection{The high temperature phase -- The black string solution}

314: 

315: The Hawking Page transition on the boundary occurs when the

316: inverse boundary temperature is $\beta^{bound}=\frac{2 \pi

317: l}{d-3}$. At higher temperatures the dominant configuration should

318: yield a $d-1$ dimensional AdS Schwarzschild (S-AdS) black hole as

319: the boundary metric. Or to be precise we get again two copies of

320: this spacetime at $r=\pm \infty$; for the reasons discussed in the

321: last subsection, we will focus on the physics associated with one

322: of those black holes.

323: 

324: One solution that realizes this boundary geometry is the black

325: string \cite{chamblin}

326: \begin{equation}

327: \label{sads}

328: ds^2 = \cosh^2 (\frac{r}{L}) ds^2_{S-AdS_{d-1}} + dr^2.

329: \end{equation}

330: That is one simply replaces the AdS$_{d-1}$ with a S-AdS$_{d-1}$

331: metric on every slice, as depicted in Figure 1. While for flat or

332: dS slices this string suffers from a Gregory Laflamme instability,

333: it has been argued in \cite{emparan1} based on thermodynamic

334: arguments that for sufficiently large horizon radius $r_H$ the

335: S-AdS black string is actually stable. This has been confirmed by

336: a numerical analysis for $d=5$ in \cite{numerics}. $T_{HP}$

337: corresponds to $r_H=L$. In order to establish that the black

338: string is the dominant configuration for $r_H>L$, one can use the

339: euclidean path integral as in the original Hawking Page paper to

340: establish that the difference in free energies associated with the

341: metrics (\ref{ads}) and (\ref{sads}) is given by the difference in

342: the value of on-shell actions (\ref{action}) \beq \label{free}

343: \beta  \Delta{F} = I^{Black-String} - I^{AdS_d} =

344: \int_{-r_1}^{r_0}dr \, \cosh^{d-3} (r/L) \left ( I^{S-AdS_{d-1}} -

345: I^{AdS_{d-1}} \right ). \eeq But $I^{S-AdS_{d-1}} - I^{AdS_{d-1}}$

346: is just the free energy difference that is governing the $d-1$

347: dimensional Hawking Page transition on the brane. It is negative

348: for $r_H>L$ and positive for $r_H<L$. To the extend that these are

349: the only two geometries that correspond to a boundary CFT on AdS

350: at a given temperature, this establishes that in the high

351: temperature regime the black string is the dominant configuration.

352: We will address the issue of possible other geometries later.

353: 

354: Note that the sign of this free energy difference is completely

355: independent of $r_1$. Both the black hole radius on the slice as

356: well as the AdS$_{d-1}$ curvature radius of the slice grow with

357: the same power of the warpfactor. The difference in free energies

358: is either positive on all slices or negative on all slices. For

359: generic values of $r_0$ and $r_1$ the phase transition occurs when

360: on every slice, including the two cutoff branes, the black hole

361: radius is equal to the AdS$_{d-1}$ curvature radius of the slice.

362: In the special case of $r_0=r_1$ with the orbifold projection

363: imposed, corresponding to the standard reflecting  boundary

364: conditions in AdS$_{d-1}$ as we explained above, we can take the

365: lower integration boundary to be $r=0$. This confirms our

366: expectation that the transparent boundary conditions do not alter

367: the $d-1$ dimensional Hawking Page physics at all.

368: 

369: \begin{figure}

370:    \centerline{\psfig{figure=blackstring.eps,width=1.5in}}

371:     \caption{The black string solution.}

372: \label{string}

373:  \end{figure}

374: 

375: \subsection{The low temperature phase -- Thermal AdS}

376: 

377: At temperatures below $T_{HP}$ the dominant boundary metric should be thermal AdS$_{d-1}$, that

378: is AdS$_{d-1}$ with periodic euclidean time. This is what

379: we get by simply replacing the bulk AdS$_d$ with thermal AdS$_d$. To compare temperatures,

380: we have to examine the change of variables between (\ref{ads}) and a more conventional way

381: of writing euclidean AdS$_d$, e.g.

382: \beq

383: \label{global}

384: ds^2 = \cosh^2(\frac{\rho}{L}) dt^2 + d \rho^2 + \sinh^2(\frac{\rho}{L}) d \Omega_{d-2}^2.

385: \eeq

386: The change of variables has for example been presented in

387: \cite{kr}. The important point is that $t=\tilde{t}$ and hence

388: $\cosh(\frac{r}{L}) \cosh(\frac{\tilde{\rho}}{L}) =

389: \cosh(\frac{\rho}{L})$. Tilde variables refer to coordinates of

390: the $d-1$ dimensional AdS on the boundary when written in a global

391: coordinate system analogous to (\ref{global}). In the context of

392: AdS/CFT the relevant notion of boundary temperature is that if the

393: metric in the bulk has the asymptotic form $f^2 ds^2_{bound.}$

394: where $ds^2_{bound.}$ is the boundary metric and $f$ a function of

395: the coordinates that diverges linearly as we approach the

396: boundary, the inverse temperature is the period of the time

397: variable with asymptotic metric $f^2 dt^2$. In our case the

398: temperature of the boundary CFT has to be measured at the $S^1

399: \times S^{d-3}$ that separates the two AdS$_{d-1}$ components of

400: the boundary of AdS$_d$. In the coordinates (\ref{global}) it is

401: time times the equator of the $S^{d-2}$.

402: 

403: We see that on the brane with curvature radius

404: $l=\cosh(\frac{r_0}{L}) L$ a period $\beta^{bulk}$ of the $t$

405: coordinate implies an inverse temperature \beq \beta^{bound} =

406: \cosh(\frac{r_0}{L}) \beta^{bulk} \eeq and correspondingly for

407: $r_1$. So at the phase transition point, the thermal boundary

408: AdS$_{d-1}$ with $\beta^{bound} =\frac{2 \pi l}{d-3}$ is what we

409: obtain from a thermal bulk AdS$_d$ with \beq \label{bulk}

410: \beta^{bulk} = \frac{2 \pi L}{d-3}. \eeq The Hawking Page phase

411: transition on the boundary maps to a very similar phase transition

412: in the bulk, where at high temperatures a black string dominates

413: the euclidean path integral, whereas at low temperatures thermal

414: AdS is the relevant configuration.

415: 

416: 

417: \subsection{The bulk black hole does not contribute}

418: 

419: In order to argue that this is the complete picture for the ensemble we study, we have to make

420: sure that there are no other configurations that yield the same $S^1 \times S^3$ geometry on

421: the boundary of the boundary and have an even lower free energy.

422: One obvious candidate is the bulk black hole, S-AdS$_d$. First note that the value

423: of $\beta$ at which the usual {\it bulk} Hawking Page transition happens, $\beta^{HP} = \frac{2 \pi L}{d-2}$,

424: is lower than the value relevant for the boundary Hawking page transition (\ref{bulk}),

425: and hence the corresponding temperature is higher.

426: So by the time the boundary made the transition to thermal AdS$_{d-1}$, in the bulk the S-AdS$_d$ black hole

427: is already a subdominant configuration -- actually the temperature is even

428: lower than $T_S$, so the bulk black hole solution doesn't even exist anymore.

429: 

430: S-AdS$_d$ could still be relevant for the high temperature regime.

431: For the case $r_0, r_1>>L$ we are discussing it is however clear

432: that the black string will always dominate over the black hole.

433: The difference in free energies between the string and the

434: background AdS according to (\ref{free}) grows exponentially with

435: $r_0$ and $r_1$, while the difference between the bulk S-AdS$_d$

436: and the background approaches a fixed value independent of $r_0$

437: and $r_1$. The same argument also suppresses other geometries that

438: would lead to induced metrics on the brane that differ from

439: thermal AdS and the S-AdS$_{d-1}$ black hole, like the bulk black

440: hole which gives rise to an FRW like cosmology on the brane. For

441: $r_0, r_1>>L$ only the same two geometries as in the original

442: Hawking-Page transition contribute on the brane, and the two bulk

443: geometries we considered are the bulk extensions corresponding to

444: those two boundary geometries. If we want to go beyond the $r_0,

445: r_1 >>L$ regime, the bulk black hole certainly will start

446: contributing. On the boundary side this maps to the strong

447: correction due to matter effects becoming important.

448: 

449: \section{Quantum Corrections}

450: 

451: The induced metric in the high temperature phase is exactly that

452: of a lower dimensional AdS Schwarzschild black hole. In

453: \cite{emparan2} however it was shown that a brane world black hole

454: can {\it not} be a solution to vacuum Einstein equations on the

455: brane. AdS/CFT predicts that classical gravity in the bulk is dual

456: to the quantum CFT including planar corrections on the boundary,

457: and coupled to classical gravity if the UV brane is

458: introduced. In particular $$M_{Pl,d-1}^{d-3} = (M_{Pl,d} L)

459: M_{Pl,d}^{d-3},$$ where $M_{Pl,d} L$ is a large number, e.g.

460: $N^{2/3}$ in Maldacena's $d=5$ example. So for

461: $M_{Pl,d}<<M<<M_{Pl,d-1}$ bulk Hawking radiation is negligible,

462: while on the brane Hawking radiation and its backreaction

463: dominate\footnote{The authors of \cite{emparan2} were mostly

464: interested in $d=5$ black holes with $M> (M_5 L)^{\frac{3}{2}} M_4$,

465: which are those with a horizon larger than $L$ and have a right to

466: be called brane black holes and Hawking radiation only yields a

467: small correction in both 5 and 4 dimensional description. But the

468: argument that bulk Hawking radiation can be neglected did not rely

469: on this.}

470: 

471: {} From this point of view it is a puzzle why the black string

472: gives a metric on the brane that is not corrected at all due to

473: Hawking radiation. While this still remains a puzzle to us in

474: general, in the $r_0, r_1>>L$ regime we are analyzing, where the

475: UV brane is close to the boundary this is not an issue. The 4d

476: Planck scale is still $M_{Pl,4}^2=M_{Pl,5}^3 L$, whereas the 4d

477: curvature radius on the brane is $l = \cosh( \frac{r_0}{L})$, so

478: for $r_0>>L$ a black hole with horizon radius larger than $l$ is

479: exponentially heavier than the Planck scale and any backreaction

480: from Hawking radiation can be neglected. The same reasoning also

481: applies to $r_1$. In the general case the black string solution

482: suggests that the large S-AdS blackhole solution has all effects

483: of Hawking radiation included. This is also the picture one is led

484: to from the original Hawking-Page analysis, which suggests that

485: the large black hole is in thermal equilibrium with the heat bath

486: generated by its own Hawking radiation.

487: 

488: 

489: \section{The onset of the Gregory Laflamme instability}

490: 

491: Like the bulk, the boundary black hole also disappears at a

492: spinodal at temperature $T_S^{bound}$. While in the temperature range

493: $T_s^{bound}<T<T_{HP}^{bound}$ the thermal AdS is the dominant configuration

494: both in the bulk and on the boundary, the black hole solution on

495: the boundary is still a valid local minimum of the free energy.

496: But when the temperature is lowered below $T_S^{bound}$ the black hole

497: on the boundary has now a negative heat capacity and no longer

498: even represents a local minimum. At the same time the Gubser-Mitra

499: hypothesis \cite{GM}, which states that classical stability of a

500: black string should come hand in hand with local thermal

501: stability, would now say that the black string in the bulk should

502: develop a classical instability of the Gregory Laflamme type.

503: While the numerical analysis of \cite{numerics} is too crude to

504: give a decisive answer to this question about the precise value at

505: which the instability sets in, their result is consistent with

506: this interpretation.

507: 

508: On the boundary, the black hole will start

509: evaporating, and will eventually disappear (as long as we keep the

510: induced Planck scale large but finite). At least the initial stage

511: of this evaporation process is completely captured by the

512: classical evolution of the unstable mode in the black string

513: background. Once the black string becomes sufficiently hot, $d$

514: dimensional quantum effects will become important and classical

515: bulk gravity is no longer a good description, so the final stages

516: of the evaporation process go beyond classical gravity.

517: 

518: This is an explicit realization of the scenario first envisioned

519: in \cite{chamblin}, where the classical dynamics associated with

520: the black string in the bulk captures the physics of the black

521: hole on the brane. Additional supporting evidence for our picture

522: comes from the analysis in $d=4$, where the so called c-metric

523: yields an exact solution describing the snapping string

524: \cite{emparan1}. Based on thermodynamic arguments they find that

525: for small enough mass (which for the negative specific heat black

526: holes means large temperature) the black string becomes

527: disfavored. Instead there exist two classically stable black hole

528: configurations on the brane. These should be thought of as the

529: endproducts of the classical evolution of the black string, any

530: further evaporation process would be due to bulk quantum effects.

531: 

532: One can envision various scenarios for how precisely the black

533: string evolves\footnote{As pointed out to us by N. Kaloper and R.

534: Emparan, a snapping string does not seem to describe Hawking

535: radiation escaping to infinity -- either in the field theory the

536: radiation stays close to the black hole during the initial stages

537: of the evaporation until bulk quantum effects set in or the string

538: in addition to shrinking at its waist will slide of the brane

539: towards the IR in order to describe the expanding cloud of Hawking

540: radiation.}. One still somewhat controversial aspect is whether

541: the string snaps at all \cite{hm}. If indeed the final

542: configuration of the string is a stable non-uniform string, we

543: would get a stable remnant as the endresult of the evaporation on

544: the brane. In any case, the map to quantum black hole physics

545: really shows how important it is to understand what happens at the

546: Gregory Laflamme point. It is however safe to say that whatever

547: the string does --- that is precisely what the black hole is going

548: to do on the brane!

549: 

550: \section*{Acknowledgements}

551: 

552: We thank Roberto Emparan, Nemanja Kaloper and Toby Wiseman for

553: very helpful discussions about black holes on the brane and for comments on

554: the draft. The research of AK is supported in part by DOE contract

555: \#DE-FG03-96-ER40956 and AK is also supported by NSF Grant SBE-0123552

556: as an ADVANCE professor.

557: 

558: 

559: \bibliography{hp}

560: \bibliographystyle{ssg}

561: \end{document}

562: 

563: \end{document}

564: