1: %\documentclass[manuscript]{aastex}

2: % \documentclass[12pt,preprint]{aastex}  % This is needed for e-submission

3: 

4: % preprint format

5: \documentclass{article}

6: \usepackage{emulateapj}

7: \usepackage{apjfonts}

8: \usepackage{graphics}

9: 

10: \newenvironment{inlinefigure}{%

11: \def\@captype{figure}%

12: \noindent\begin{minipage}{0.999\linewidth}\begin{center}}

13: {\end{center}\end{minipage}\smallskip}

14: 

15: \newlength{\colwidth}

16: \setlength{\colwidth}{\textwidth}

17: \addtolength{\colwidth}{-\columnsep}

18: \setlength{\colwidth}{0.5\colwidth}

19: 

20: %some notations

21: \newcommand{\E}[1]{{\times 10^{#1}}}

22: \newcommand{\Mpc}{{\rm Mpc}}

23: \newcommand{\grad}{\nabla}

24: \newcommand{\apm}{{\sc {APMSPH }}}

25: \newcommand{\hydra}{{\sc{HYDRA }}}

26: \newcommand{\tree}{{\sc{TREESPH }}}

27: \newcommand{\cmbfast}{{\sc {CMBFAST }}}

28: \newcommand{\lya}{Ly$\alpha$~}

29: % ions

30: \renewcommand{\H}{\ion{H}{1}}

31: \newcommand{\Hp}{\ion{H}{2}}

32: \newcommand{\He}{\ion{He}{1}}

33: \newcommand{\Hep}{\ion{He}{2}}

34: \newcommand{\Hepp}{\ion{He}{3}}

35: \newcommand{\h}{{\rm H\,{\sc I}}}

36: \newcommand{\hp}{{\rm H\,{\sc II}}}

37: \newcommand{\he}{{\rm He\,{\sc I}}}

38: \newcommand{\hep}{{\rm He\,{\sc II}}}

39: \newcommand{\hepp}{{\rm He\,{\sc III}}}

40: %

41: %

42: %	some backspaces

43: \newcommand{\bbbb}{{\!\!\!\!}}

44: \newcommand{\s}{{\rm s}}

45: \newcommand{\erg}{{\rm erg}}

46: \newcommand{\g}{{\rm g}}

47: \newcommand{\cm}{{\rm cm}}

48: \newcommand{\sr}{{\rm sr}}

49: \newcommand{\Hz}{{\rm Hz}}

50: \newcommand{\K}{{\rm K}}

51: \newcommand{\kms}{{\rm km}\,{\rm s}^{-1}}

52: 

53: \lefthead{Theuns et al.}

54: \righthead{Constraints on reionization}

55: 

56: \begin{document}

57: 

58: \submitted{Accepted for publication in the Astrophysical Journal Letters}

59: 

60: \title{Constraints on reionization from the thermal history\\ of the

61: intergalactic medium\altaffilmark{1,2}}

62: \author{Tom Theuns\altaffilmark{3}, Joop Schaye\altaffilmark{4},

63: Saleem Zaroubi\altaffilmark{5}, Tae-Sun

64: Kim\altaffilmark{6}, Panayiotis Tzanavaris\altaffilmark{3}\\ and Bob

65: Carswell\altaffilmark{3}}

66: \altaffiltext{1}{Based on observations made at the W.M. Keck Observatory

67: which is operated as a scientific partnership between the California

68: Institute of Technology and the University of California; it was made

69: possible by the generous support of the W.M. Keck Foundation. }

70: \altaffiltext{2}{Based on public

71: data released from the VLT/UVES Commissioning and Science

72: Verification and from the OPC program 65.O-296A (PI S.-D'Odorico) at

73: the VLT/Kueyen telescope, ESO, Paranal, Chile.}

74: \altaffiltext{3} {Institute of Astronomy, Madingley Road, Cambridge CB3

75: 0HA, UK}

76: \altaffiltext{4} {School of Natural Sciences, Institute for Advanced

77: Study, Einstein Drive, Princeton NJ 08540}

78: \altaffiltext{5} {Max-Planck Institut f\"ur Astrophysik, Postfach 123,

79: D-85740 Garching bei M\"unchen, Germany}

80: \altaffiltext{6} {European Southern Observatory, Karl-Schwarzschild-Stra\ss e 2, D-85748 Garching bei

81: M\"unchen, Germany}

82: 

83: \begin{abstract}

84: The temperature of the diffuse, photo-heated intergalactic medium (IGM)

85: depends on its reionization history because the thermal time scales are

86: long. The widths of the hydrogen \lya absorption lines seen in the

87: spectra of distant quasars that arise in the IGM can be used to

88: determine its temperature. We use a wavelet analysis of the \lya forest

89: region of quasar spectra to demonstrate that there is a relatively

90: sudden increase in the line widths between redshifts $z\approx 3.5$ and

91: 3.0, which we associate with entropy injection resulting from the

92: reionization of \Hep. The subsequent fall-off in temperature after

93: $z\approx 3.5$, is consistent with a thermal evolution dominated by

94: adiabatic expansion. If, as expected, the temperature also drops

95: rapidly after hydrogen reionization, then the high temperatures

96: inferred from the line widths before \Hep\ reionization imply that

97: hydrogen reionization occurred below redshift $z=9$.

98: \end{abstract}

99: 

100: \keywords {cosmology: observations --- cosmology: theory ---

101: galaxies: formation --- intergalactic medium --- quasars: absorption

102: lines} 

103: 

104: 

105: \section{Introduction}

106: Neutral hydrogen in the intergalactic medium (IGM) along the line of

107: sight to quasars at redshifts $z\le 6$ produces hundreds of \lya

108: absorption lines.  The fact that not all flux is absorbed (i.e., the

109: absence of a \lq Gunn-Peterson\rq~ trough, Gunn \& Peterson 1965)

110: requires that the universe be ionized to a level far higher than can be

111: attributed to residual ionization from recombination. At lower

112: redshifts, $z\la 3$, {\em observed} stars and quasars produce enough

113: ionizing photons to explain the high levels of ionization, but the

114: nature of the sources responsible for converting most of the IGM from

115: neutral to ionized remain uncertain, as does the epoch of reionization

116: (e.g., Barkana \& Loeb 2001).

117: 

118: The observed mean flux decrement $D_A$ blueward of the quasar's \lya

119: emission line increases with the redshift of the quasar, both because

120: the intensity of the ionizing background radiation decreases above

121: $z=4$ (e.g., McDonald \& Miralda-Escud\'e 2001) and because the mean

122: density $\bar\rho$ of the universe -- and hence the neutral fraction

123: $x\propto \bar\rho T^{-0.7}/\Gamma_\h$ for fixed values of the

124: photo-ionization rate $\Gamma_\h$ and temperature $T$ --

125: increases. Recently, Becker et al.\ (2001) and Djorgovski et al.\

126: (2001) observed a sudden increase in $D_A$ in the spectra of redshift

127: $z\sim 6$ quasars discovered by the {\sc SLOAN} digital sky

128: survey. Such a sharp rise has been predicted to mark the transition

129: associated with a sudden epoch of reionization (e.g., Cen \& Ostriker

130: 1993; Gnedin 2000). Similarly, a sudden increase in the \Hep\ opacity

131: has been detected around $z\sim 3$ (Reimers et al.\ 1997; Heap et al.\

132: 2000; Kriss et al.\ 2001), associated with helium reionization.

133: 

134: Another way to study the reionization history of the IGM is to

135: investigate its thermal evolution. Because its cooling time is long,

136: the low-density IGM retains some memory of when and how it was

137: reionized (e.g., Miralda-Escud\'e \& Rees 1994; Hui \& Gnedin 1997;

138: Haehnelt \& Steinmetz 1998).  The combined effects of photo-ionization

139: heating and adiabatic expansion introduce a tight temperature-density

140: relation in the unshocked IGM, which can be approximated by a power-law

141: $T=T_0 (\rho/\bar\rho)^{\gamma-1}$ for densities around the cosmic mean

142: (Hui \& Gnedin 1997). A change in these parameters influences the

143: shapes of the \lya lines, because thermal broadening and Jeans

144: smoothing determine the line widths (e.g., Theuns, Schaye \& Haehnelt

145: 2000). Schaye et al.\ (1999) used hydrodynamical simulations to

146: demonstrate that one can accurately calibrate the relation between the

147: minimum line width ($b$) as a function of column density ($N_\h$) on the one

148: hand, and the underlying $T-\rho$ relation on the other. Schaye et al.\

149: (2000) applied the method to observations in the redshift range 2.0-4.5

150: and found that $T_0$ peaks at $z\sim 3$, which they interpreted as

151: evidence for the reionization of \Hep. Ricotti, Gnedin \& Shull (2000)

152: used pseudo hydrodynamical simulations and found a similar temperature

153: increase, albeit only at the 0.5$\sigma$ level. McDonald et

154: al.~(2001) found no evidence for temperature evolution, but their

155: analysis neglected the important temperature dependence of Jeans smoothing. Finally,

156: the analysis of Zaldarriaga et al.\ (2001) neglected hydrodynamical

157: effects all together.

158: 

159: Here, we provide new evidence for a relatively sudden increase in $T_0$

160: between redshifts $z\approx 3.5$ and 3.0, using a new method based on a

161: wavelet decomposition of the absorption spectrum. We then use the

162: measured values of $T_0$ at higher redshift to constrain the epoch of

163: hydrogen reionization $z_{\rm H}$ and find that the data require

164: $z_{\rm H} \le 9$ for any reasonable value ($T_0\la 6\times 10^4~\K$)

165: of the hydrogen reionization temperature.

166: 

167: \newpage

168: \section{Helium reionization}

169: \subsection{Wavelet analysis}

170: Our analysis uses wavelets to characterize line-widths. By picking an

171: appropriate wavelet scale (we used $\sim 15$ km s$^{-1}$), we find that the

172: amplitude $A$ of the wavelet anti-correlates with the widths of the

173: lines -- and hence the temperature $T_0$ of the absorbing gas, $\langle

174: A\rangle\propto T_0^{-1}$. By examining the statistics of $A$ along

175: the spectrum, we can look for changes in temperature in an objective

176: way, given that the wavelet decomposition is unique. Full details can

177: be found in Theuns \& Zaroubi (2000) and Theuns et. al\ (2002), here

178: we give only a brief summary of the underlying reasoning.

179: 

180: To investigate whether a region of given size $V$ of the QSO spectrum

181: has an unusual temperature, we compare the cumulative probability

182: distribution ${\cal C}_V(A)$ of the wavelet amplitudes in that region,

183: with ${\cal C}(A)$ for the spectrum as a whole. If the region is

184: unusually hot, it will tend to have very few large wavelet amplitudes,

185: and hence the maximum difference $\Delta(V)={\cal C}_V-{\cal C}$ will

186: be large.  (Note that by construction, $|\Delta|\le 1$, and it is

187: defined for a given {\em region}).  Conversely, cold regions will have

188: large, negative $\Delta(V)$. In panel (a) of Fig.~1, we have plotted

189: $\Delta$ for regions of size 5000\ km s$^{-1}$ along a mock spectrum of

190: a simulation\footnote{The cosmological parameters for this

191: vacuum-energy dominated, flat, cold dark matter, smoothed particle

192: hydrodynamics (SPH) simulation are $(\Omega_m,\Omega_\Lambda,\Omega_b

193: h^2,h,\sigma_8)=(0.3,0.7,0.019,0.65,0.9)$. The simulation box is

194: $12\,h^{-1}$~Mpc on a side, and gas and dark matter are represented

195: with 256$^3$ particles each.}. The mock spectrum has a jump in $T_0$

196: from $T_0=2.2\times 10^4$\ K below redshift $z=3.3$ to $T_0=1.5\times

197: 10^4$\ K above it. As expected, there is a corresponding jump in

198: $\Delta$ at $z=3.3$, from positive values at low $z$ to negative ones

199: at higher $z$.

200: 

201: To determine whether a high value of $|\Delta|$ is statistically

202: significant -- and hence whether we have identified a region with an

203: unusual temperature -- we proceed as follows. We repeat the procedure

204: with spectra where the absorption lines are randomly scrambled (with

205: replacement, see Theuns et al.\ 2002), in order to wash-out any

206: correlations in the wavelet amplitudes, resulting from intrinsic

207: temperature fluctuations. Using these randomized spectra, we can

208: construct the statistical probability $P(\Delta)$ from the fraction of

209: regions in the randomized spectra, that have a given value of $\Delta$.

210: Given $P(\Delta)$, we can determine how likely a value of $\Delta$ --

211: and hence of a temperature deviation -- in the original spectrum is.

212: Performing this analysis, we find that the detected change in $\Delta$

213: in the mock spectrum has a statistical significance of more than 99.5

214: per cent, shown as the full line in panel (a) (positive values refer to

215: regions hotter than average, and vice versa for negative values). This

216: means that in only one of out 200 random realisation, do we, by chance,

217: get values of $|\Delta| \ga 0.2$. Note that we only use the simulation

218: to generate the spectrum, and not to assign the statistical

219: significance of a change in $\Delta$. In the preparation of the mock

220: spectrum, we have imposed the same biases as are present in the real

221: data, by adding noise and instrumental broadening to the simulated

222: lines, and by scaling the mean absorption to the observed value. We

223: therefore believe that the method can be applied to real data with

224: confidence.

225: 

226: We have applied the same wavelet analysis to high-resolution echelle

227: spectra of quasars Q0055-2169 (emission redshift $z_{\rm em}=3.6$, Kim

228: et al 2001), the combined spectrum of Q0302-003 ($z=[3,3.27]$, Kim et

229: al 2001) and APM~0827+5255 ($z=[3.27,3.7]$, Ellison et al 1999) and

230: Q1422+231 ($z_{\rm em}=3.6$, Rauch et al 1997) (panels b, c and d of

231: Fig.~1 respectively). In each spectrum, we find a cold region at high

232: redshift, and a hot region at lower redshift, each significant at the

233: more than 99 percent level when compared to randomized spectra. Panels

234: (b) and (c) appear very similar to the mock spectrum of panel (a),

235: where we had imposed a sudden temperature increase below $z=3.3$. Note

236: that the implied temperature evolution is exactly opposite of what one

237: would expect from photo-heating in the optically thin limit, in which

238: case the IGM will gradually cool down. We take this as strong evidence

239: that a large fraction of \Hep\ is reionized around redshift $z \approx

240: 3.3$.

241: 

242: % preprint format

243: \begin{inlinefigure}

244: \centerline{\resizebox{0.96\colwidth}{!}{\includegraphics{f1.eps}}}

245: \vspace{-1.5cm}

246: \figcaption[fig:jump]{

247: % submission format

248: % \begin{figure}

249: % \plotone{f1.eps}

250: % \caption[f1.eps]{

251: Temperature statistic $\Delta$ (right-hand scale, dot-dashed line) and

252: statistical significance for $\Delta$ in per cent (left-hand scale,

253: full line) as a function of redshift for the mock spectrum, Q0055-2169,

254: the combined spectrum of Q0302-003 ($z=[3,3.27]$) and APM~0827+5255

255: ($z=[3.27,3.7]$), and Q1422+231, panels a-d respectively. Negative

256: values for $\Delta$ and the significance denote cold regions, and vice

257: versa for hot regions. The mock spectrum has a jump in $T_0$ from

258: $T_0=2.2\times 10^4$~K below $z=3.3$ to $T_0=1.5\times 10^4$~K above

259: $z=3.3$. The statistical significance of the regions with a large value

260: of $|\Delta|$ at high and low $z$, when compared to randomized spectra

261: made from this spectrum, but without a sudden temperature change, is

262: more than 99 percent. The observed data exhibit a similarly significant

263: jump, in the sense that the higher redshift halves are colder than the

264: lower redshift halves.  }

265: \end{inlinefigure}

266: %\end{figure}

267: 

268: The mean wavelet amplitude scales approximately inversely with

269: temperature: $\langle A\rangle\propto T_0^{-1}$ (Theuns et al.\

270: 2002). We have used hydrodynamical simulations to calibrate the

271: proportionality constant, and investigate its dependence on

272: $\gamma$. Applying the calibration to the above QSO sample, we obtain

273: values for $T_0$ in good agreement with those obtained by Schaye et al

274: (2000) from the cutoff in the line width -- column density relation,

275: i.e., $T_0\approx 10^{4.1}$ for $z\ge 3.4$, and $T_0\approx 10^{4.3}$

276: at $z\approx 3.0$. This assumes the value of $\gamma$ as determined by

277: Schaye et al., but reasonable changes of $\gamma$ do not change $T_0$

278: by more than $\sim 10$ per cent (Theuns et al. 2002). Simulations that

279: include radiative transfer are required to investigate whether such a

280: temperature change is consistent with \Hep\ reionization.

281: 

282: Other heating mechanisms, for example shock heating by galactic winds,

283: do not have a major influence on the value of $T_0$ deduced from

284: fitting the cutoff in the $b-N_\h$ diagram, at least as long as the

285: volume fraction of shocked gas remains small. This is because the

286: method is based on identifying the {\em narrowest} lines in a region,

287: irrespective of whether there is also a set of much broader lines. In

288: contrast, the wavelet method used here examines {\em all} lines in a

289: stretch of spectrum. So the fact that both methods find similar values

290: for $T_0$ suggests that photo-heating is indeed the dominant heating

291: mechanism, and that the volume 

292: % % preprint format

293: \begin{inlinefigure}

294: \centerline{\resizebox{0.96\colwidth}{!}{\includegraphics{f2.eps}}}

295: \figcaption[fig:tversusz_low]{

296: % submission format

297: % \begin{figure}

298: % \plotone{f2.eps}

299: % \caption[f2.eps]{

300: Temperature evolution of the IGM after $z=3.4$. The points with error

301: bars are determined from fitting the cutoff in the $b-N_\h$ relation

302: (Schaye et al\ 2000). The curves emanating from the star all assume

303: that $T_0=10^{4.4}~\K$ at $z=3.4$ as a result of \hep\ reionization,

304: and differ in the assumed UV-background at lower redshifts. The solid

305: line includes photo-heating by \h, \he\ and \hep. The short dashed

306: curve neglects heating from \hep, the long dashed curve neglects

307: heating from both \h\ and \he, and finally the dot-dashed curve ignores

308: photo-heating all together. All models fit the inferred temperature

309: evolution extremely well.}

310: \end{inlinefigure}

311: %\end{figure}

312: 

313: filling factor of gas that has been shocked by winds is small.

314: 

315: \subsection{Thermal evolution}

316: After reionization, the evolution of $T_0$ is given by

317: \begin{equation}

318: {1 \over T_0}{dT_0 \over dt} - {1\over\mu} {d\mu\over dt} = - 2H + {\mu \Delta_\epsilon \over {3

319: \over 2} k_B T_0},

320: \label{eq:temp}

321: \end{equation}

322: where $H$ is the Hubble parameter, $k_B$ Boltzmann's constant, $\mu$

323: the mean molecular weight, and $\Delta_\epsilon$ is the effective

324: radiative cooling rate [in $\erg\,\g^{-1} \s^{-1}$]. $\Delta_\epsilon$

325: is negative (positive) for net cooling (heating) and includes

326: photo-electric heating and cooling via recombination, excitation,

327: inverse Compton scattering, collisional ionization, and bremsstrahlung

328: (we use the rates listed in Appendix B of Theuns et al.\ 1998). For gas

329: around the mean density, the dominant cooling process is the adiabatic

330: expansion of the universe (the first term on the right-hand side of

331: equation \ref{eq:temp}), except at $z>7$ where inverse Compton cooling

332: off the cosmic microwave background is more efficient.  The radiative

333: heating and cooling rates depend on the ionization balance of the gas,

334: which itself depends on the temperature. By coupling equation

335: \ref{eq:temp} to the differential equations for the ionization balance

336: (listed in Appendix B of Theuns et al.~1998), we can thus solve for the

337: evolution of $T_0$ given a model for the ionizing background, an

338: initial value for $T_0$, and the initial ionization state. We have

339: tested this procedure using full hydrodynamic simulations and find that

340: the evolution of $T_0$ is reproduced very well.

341: 

342: We assume the universe to be permeated by a uniform UV-background with

343: a power-law spectral shape,

344: \begin{equation}

345: J = \left \{ \nonumber

346: \begin{array}{ll} 

347: J_{\h} \left ({\nu \over \nu_{\h}} \right

348: )^{-1.8} & \nu < \nu_{\hep} \nonumber \\

349: J_{\hep} \left ({\nu \over \nu_{\h}}\right )^{-1.8} & \nu \ge

350: \nu_{\hep} \nonumber \\ 

351: \end{array}

352: \right. ~\erg\,\s^{-1}\,\cm^{-2}\,\sr^{-1}\,\Hz^{-1} \nonumber,

353: \end{equation}

354: where $\nu_\h$ is the \H\ ionization threshold.  (Note that we allow

355: for a step between the intensities of \H\ and \Hep\ ionizing photons,

356: but normalize $J_{\hep}$ at the hydrogen ionization edge.) We set

357: $J_{\h} = 4 \times 10^{-23}$, which yields a photo-ionization rate of

358: $\Gamma_{\h} = 10^{-13}~\s^{-1}$ and vary $J_{\hep}$ (and the

359: corresponding ionization rate $\Gamma_{\hep}$) as described below.

360: Because the photo-heating rate is independent of the amplitude of the

361: UV-background as long as the gas is highly ionized, the exact value of

362: $J_{\h}$ after hydrogen reionization is unimportant. Since \Hep\ is

363: not always highly ionized, the thermal evolution does depend on

364: $J_{\hep}$.

365: 

366: The predicted evolution of $T_0$ is compared with the data from Schaye

367: et al (2000) in Figure~2. We assume the IGM to be in ionization

368: equilibrium at temperature $T_0=10^{4.4}~\K$ at redshift $z=3.4$ as a

369: result of \Hep\ reionization. The curves emanating from the big star

370: then show the subsequent evolution of $T_0$, for the following imposed

371: UV-backgrounds. The solid curve is for a constant ionizing rate of

372: $\Gamma_\h=\Gamma_\hep=10^{-13}~\s^{-1}$, and the corresponding

373: photo-heating from all species. Increasing or decreasing $\Gamma_\h$ by

374: a factor 10 does not change the evolution appreciably. The short and

375: long dashed curves ignore photo-electric heating of \Hep\ and of both

376: \H\ and \He\, respectively. Finally, the dot-dashed line ignores

377: photo-heating all together. The measurements of Schaye et al.\ (2000)

378: below $z\sim 3$ are clearly consistent with a thermal evolution

379: dominated by adiabatic cooling.

380: 

381: Prior to \Hep\ reionization, some regions will already be ionized in

382: \Hep\ by local sources. If most of the universe is reionized

383: significantly later, then such differences in reionization epoch will

384: lead to spatial variations in $T_0$. The wavelet analysis by Theuns et

385: al. (2002) can detect variations in $T_0$ of order 50 per cent, over a

386: region of size 5000 km s$^{-1}$, yet no such fluctuations were found in

387: the data. The wavelet analysis by Zaldarriage (2001) also failed to

388: detect any such temperature fluctuations. This suggests that the

389: temperature increase is the result of the overall increase in the far

390: UV-background following the percolation of \Hepp\ regions, which prior

391: to reionization are too small to be detected by current methods.

392: 

393: \section{Hydrogen reionization}

394: Given the success in reproducing the $z<3.4$ temperature evolution, we

395: now turn to higher redshifts.  Abel \& Haehnelt (1999) performed

396: radiative transfer calculations of the expansion of an ionization

397: bubble around a QSO and found typical post-reionization temperatures of

398: $4\times 10^4$~K, {\em including} \Hep\ reionization. To put a

399: conservative upper limit on the hydrogen reionization redshift

400: $z_{\h}$, we assume a higher value of $T_0=6\times 10^4~\K$. Assuming

401: the gas to be in ionization equilibrium just after reionization, we

402: compute $T_0(z)$ for a given $z_{\h}$ as before. In Fig.~3 the

403: predicted thermal evolution for various values of $z_{\h}$ is compared

404: with the data.

405: 

406: The temperature decreases rapidly with decreasing redshift because at

407: these temperatures photo-heating cannot compensate for the rapid

408: expansion cooling. Consequently, {\em the temperature drops quickly

409: below the values measured at $z\sim 4$, unless hydrogen reionization

410: occurred relatively recently}. The solid curves are for

411: $\Gamma_\h=10^{-13}$ s$^{-1}$ and a range of reionization redshifts, as

412: indicated in the figure. For $z_{\rm H}=6.2$, we show

413: $\Gamma_\h=10^{-14}~\s^{-1}$ for comparison. The models with late \H\

414: reionization, $z_{\rm H}\le 8$, fit the data best. Models with $z_{\rm

415: H}\ge 9$ have $z\sim 4$ temperatures that are significantly below the

416: measured values. For example, the $z_{\rm H}= 9.2$ curve has a reduced

417: $\chi^2$ of 4.5 for the 5 $z> 3.4$ data points.

418: 

419: % % preprint form

420: \begin{inlinefigure}

421: \centerline{\resizebox{0.96\colwidth}{!}{\includegraphics{f3.eps}}}

422: \figcaption[fig:tversusz_high]{

423: % submission format

424: % \begin{figure}

425: % \plotone{f3.eps}

426: % \caption[f3.eps]{

427: The temperature evolution of the IGM above redshift 3.4. The solid

428: curves indicate the evolution of the temperature at the mean density

429: for various \h\ reionization redshifts $z_{\rm H}$, as indicated. The

430: post-hydrogen reionization temperature is assumed to be $T_0=6\times

431: 10^4~\K$ and the hydrogen photo-ionization rate is $\Gamma_\h=10^{-13}~

432: \s^{-1}$ (the short dashed line has $\Gamma_\h=10^{-14}~ \s^{-1}$). The

433: \hep\ photo-ionization rate is adjusted so that the \hepp\ abundance

434: $x_\hepp\approx 0.1$ at $z=3.5$. The solid line connecting filled

435: squares is for $z_{\rm H}=10.2$, and a higher \hep\ photo-ionization

436: rate, $x_\hepp(z=3.5)=0.6$. Finally, the long dashed line has $z_{\rm

437: H}=20$, but a still higher \hep\ photo-ionization rate,

438: $x_\hepp(z=3.5)=0.95$.  If He is mostly singly ionized at $z\ga 3.5$,

439: then the rapid decrease in $T_0$ after reionization places an upper

440: limit of $z_{\rm H}< 9$ on the redshift of hydrogen reionization.}

441: \end{inlinefigure}

442: %\end{figure}

443: 

444: The contribution of \Hep\ photo-heating at $z\ge 5$ is uncertain.  It

445: is likely that stars dominate the \H\ ionizing background at such high

446: redshifts (e.g., Madau, Haardt, \& Rees 1999), and stars emit very few

447: \Hep\ ionizing photons. However, very massive and extremely metal poor

448: stars could lead to a non-negligible \Hep\ ionizing background

449: (Tumlinson \& Shull 2000). The QSO contribution is also uncertain,

450: although the paucity of faint point sources in the Hubble Deep Field

451: does provide some constraints (Haiman, Madau \& Loeb 1999). If \Hep\

452: reionized at $z\sim 3$, as argued in the previous sections, then the

453: \Hepp\ abundance, $x_\hepp$, should be small at higher $z$. We therefore set

454: $\Gamma_\hep=10^{-15}~\s^{-1}$ so that $x_\hepp \approx 0.1$ at

455: redshift $z=3.5$. This limits the plausible contribution from \Hep\

456: photo-heating, and allows us to put a conservative upper limit on the

457: redshift of \H\ reionization, $z_{\rm H}< 9$.

458: 

459: The importance of \Hep\ heating is illustrated by the solid curve

460: connecting the filled squares. This model has $z_{\rm H}=10.2$, but the

461: \Hep\ ionizing background is increased such that $x_\hepp\approx 0.6$

462: at $z=3.5$. This increases $T_0(z=4)$ significantly, although it still

463: falls below the measured values. Finally, the long dashed line is for a

464: model with $z_{\rm H}=20$ and $J_\hep$ further increased so that

465: $x_\hepp\approx 0.96$ at $z=3.5$. The temperature of this model is

466: consistent with the data, yet \Hep\ is ionized at the more than 90

467: percent level as early as $z=5$. Such a high level of ionization

468: conflicts with the evidence that \Hep\ reionizes at $z\sim 3$ from the

469: observed \Hep\ opacities and the associated increase in

470: $T_0$.

471: 

472: All this leads us to the following conclusions. Two independent methods

473: consistently find a rather sudden increase in the temperature of the

474: IGM over the range $z\sim$ 3.5--3.0, which we associate with \Hep\

475: reionization. If this interpretation is correct, then the \hepp\

476: fraction must be low at higher redshifts. Therefore, above redshifts

477: 3.5, \Hep\ photo-heating cannot be significant and the IGM cools

478: rapidly following \H\ reionization. The high values of the IGM

479: temperature at $z\sim 4$ then require that \H\ reionization occurred

480: late as well, $z_{\rm H}< 9$, for any reasonable value $T_0\la 6\times

481: 10^4$~K for the \h\ reionization temperature.  More plausible

482: reionization temperatures of $T_0\sim 4\times 10^4$~K and $2\times

483: 10^4$~K would constrain the hydrogen reionization redshift further to

484: $z_{\rm H}\la 8$ and $z_{\rm H}\la 7$, respectively.\\

485: 

486: 

487: {\em Acknowledgments} TT thanks PPARC for the award of an Advanced

488: Fellowship. JS is supported by a grant from the W.M.~Keck Foundation.

489: We acknowledge support from the \lq Physics of the Intergalactic

490: Medium\rq\ network set up by the European Commission. Research was

491: conducted in cooperation with Silicon Graphics/Cray Research utilizing

492: the Origin 2000 super computer at DAMTP, Cambridge.

493: 

494: \begin{thebibliography}{}

495: \bibitem[Abel \& Haehnelt(1999)]{1999ApJ...520L..13A} Abel, T.~\& Haehnelt, 

496: M.~G.\ 1999, \apjl, 520, L13 

497: \bibitem[Barkana \& Loeb(2001)]{2001PhR...349..125B} Barkana, R.~\& Loeb, 

498: A.\ 2001, \physrep, 349, 125 

499: \bibitem[]{} Becker, R. H., et al 2001, preprint (astro-ph/0108431)

500: \bibitem[]{} Cen, R.~\& Ostriker, J. P., 1993, ApJ, 417, 404

501: \bibitem[]{} Djorgovski, S. G., Castro, S. M., Stern, D., \& Mahabel, A.A.,

502: 2001, preprint (astro-ph/0108069)

503: \bibitem[Ellison et al.(1999)]{1999PASP..111..946E} Ellison, S.~L., Lewis, 

504: G.~F., Pettini, M., Sargent, W.~L.~W., Chaffee, F.~H., Foltz, C.~B., Rauch, 

505: M., \& Irwin, M.~J.\ 1999, \pasp, 111, 946 

506: \bibitem[]{} Gnedin, N. Y., 2000, ApJ, 535, 530

507: \bibitem[Gunn \& Peterson(1965)]{1965ApJ...142.1633G} Gunn, J.~E.~\& 

508: Peterson, B.~A.\ 1965, \apj, 142, 1633 

509: \bibitem[Haehnelt \& Steinmetz(1998)]{1998MNRAS.298L..21H} Haehnelt, 

510: M.~G.~\& Steinmetz, M.\ 1998, \mnras, 298, L21 

511: \bibitem[Haiman, Madau, \& Loeb(1999)]{1999ApJ...514..535H} Haiman, Z.~;., 

512: Madau, P., \& Loeb, A.\ 1999, \apj, 514, 535 

513: \bibitem[Heap et al.(2000)]{2000ApJ...534...69H} Heap, S.~R., Williger, 

514: G.~M., Smette, A., Hubeny, I., Sahu, M.~S., Jenkins, E.~B., Tripp, T.~M., 

515: \& Winkler, J.~N.\ 2000, \apj, 534, 69 

516: \bibitem[Hui \& Gnedin(1997)]{1997MNRAS.292...27H} Hui, L.~\& Gnedin, 

517: N.~Y.\ 1997, \mnras, 292, 27 

518: \bibitem[]{}Kim, T.-S., et al.~2001, submitted to MNRAS

519: \bibitem[Kriss et al.(2001)]{2001Sci...293.1112K} Kriss, G.~A.~et al.\ 

520: 2001, Science, 293, 1112 

521: \bibitem[Madau, Haardt, \& Rees(1999)]{1999ApJ...514..648M} Madau, P., 

522: Haardt, F., \& Rees, M.~J.\ 1999, \apj, 514, 648 

523: \bibitem[McDonald \& Miralda-Escud{\' e}(2001)]{2001ApJ...549L..11M} 

524: McDonald, P.~\& Miralda-Escud{\' e}, J.\ 2001, \apjl, 549, L11 

525: \bibitem[McDonald et al.(2001)]{}

526: McDonald, P., Miralda-Escud\'e, J., Rauch, M., Sargent, W. L. W.,

527: Barlow, T. A., \& Cen, R. 2001, preprint (astro-ph/0005553)

528: \bibitem[Miralda-Escude \& Rees(1994)]{1994MNRAS.266..343M} Miralda-Escude, 

529: J.~\& Rees, M.~J.\ 1994, \mnras, 266, 343 

530: \bibitem[Rauch et al.(1997)]{1997ApJ...489....7R} Rauch, M.~et al.\ 1997, 

531: \apj, 489, 7 

532: \bibitem[]{} Reimers, D., Kohler, S., Wisotzki, L., Groote, D.,

533: Rodeiguez-Pascual, P., \& Warmsteker, W., 1997, A\&A, 327, 890.

534: \bibitem[Ricotti, Gnedin, \& Shull(2000)]{2000ApJ...534...41R} Ricotti, M., 

535: Gnedin, N.~Y., \& Shull, J.~M.\ 2000, \apj, 534, 41 

536: \bibitem[Schaye, Theuns, Leonard, \& Efstathiou(1999)]{1999MNRAS.310...57S} 

537: Schaye, J., Theuns, T., Leonard, A., \& Efstathiou, G.\ 1999, \mnras, 310, 

538: 57 

539: \bibitem[Schaye et al.(2000)]{2000MNRAS.318..817S} Schaye, J., Theuns, T., 

540: Rauch, M., Efstathiou, G., \& Sargent, W.~L.~W.\ 2000, \mnras, 318, 817 

541: \bibitem[Theuns \& Zaroubi(2000)]{2000MNRAS.317..989T} Theuns, T.~\& 

542: Zaroubi, S.\ 2000, \mnras, 317, 989 

543: \bibitem[Theuns, Schaye, \& Haehnelt(2000)]{2000MNRAS.315..600T} Theuns, 

544: T., Schaye, J., \& Haehnelt, M.~G.\ 2000, \mnras, 315, 600 

545: \bibitem[Theuns et al.(1998)]{1998MNRAS.301..478T} Theuns, T., Leonard, A., 

546: Efstathiou, G., Pearce, F.~R., \& Thomas, P.~A.\ 1998, \mnras, 301, 478 

547: \bibitem[]{} Theuns, T., Zaroubi, S., Kim, T.-S., Tzanavaris, P.,

548: Carswell, R.F., 2002, MNRAS in press, (astro-ph/0110600)

549: \bibitem[Tumlinson \& Shull(2000)]{2000ApJ...528L..65T} Tumlinson, J.~\& 

550: Shull, J.~M.\ 2000, \apjl, 528, L65 

551: \bibitem[Zaldarriaga, Hui, \& Tegmark(2001)]{2001ApJ...557..519Z} Zaldarriaga, M., Hui, L., \& Tegmark, M.\ 2001, \apj, 557, 519

552: \bibitem[]{} Zaldarriaga, M., 2001, preprint (astro-ph/0102205)

553: \end{thebibliography}{}

554: 

555: \end{document}

556: