1: \documentclass{emulateapj}
2: \usepackage{epsf}
3: \usepackage{apjfonts}
4: \usepackage{xspace}
5: \usepackage{amsmath}
6: \bibliographystyle{apj}
7:
8:
9: \def\figdir{.}
10:
11:
12: \def\HI{{\rm H\,I}}
13: \def\HII{{\rm H\,II}}
14: \def\GI{{\rm He\,I}}
15: \def\GII{{\rm He\,II}}
16: \def\GIII{{\rm He\,III}}
17:
18: \def\Lsun{\, L_{\odot}}
19: \def\Msun{\, M_{\odot}}
20: \def\chimp{h^{-1}\dim{Mpc}}
21: \def\kms{\dim{km/s}}
22: \def\fesc{f_{\rm esc}}
23: \def\fescg{f_{\rm esc, \HI}}
24:
25: \def\dim#1{\mbox{\,#1}}
26: %
27: % Use the first version of this macro for preprint, the second for
28: % ApJ submission
29: %
30: \def\figname#1#2{\figdir/#1}
31: %\def\figname#1#2{#2}
32: %
33: \def\note#1{}
34: \def\hide#1{}
35: \def\new#1{{\bf #1}}
36:
37:
38: \begin{document}
39:
40: %=================================================
41: \title{Escape of Ionizing Radiation from High Redshift Galaxies}
42: %=================================================
43:
44: \author{Nickolay Y.\ Gnedin\altaffilmark{1,2,3}, Andrey
45: V. Kravtsov\altaffilmark{2,3,4}, and Hsiao-Wen Chen\altaffilmark{3}}
46: \altaffiltext{1}{Particle Astrophysics Center, Fermi National
47: Accelerator Laboratory, Batavia, IL 60510, USA; gnedin@fnal.gov}
48: \altaffiltext{2}{Kavli Institute for Cosmological Physics, The
49: University of Chicago, Chicago, IL 60637, USA}
50: \altaffiltext{3}{Department of Astronomy \& Astrophysics, The
51: University of Chicago, Chicago, IL 60637 USA}
52: \altaffiltext{4}{Enrico Fermi Institute, The University of Chicago,
53: Chicago, IL 60637, USA}
54:
55: \begin{abstract}
56: We model the escape of ionizing radiation from high-redshift galaxies
57: using high-resolution Adaptive Mesh Refinement
58: $N$-body$+$hydrodynamics simulations. Our simulations include
59: time-dependent and spatially-resolved transfer of ionizing radiation
60: in three dimensions, including effects of dust absorption. For
61: galaxies of total mass $M\gtrsim 10^{11}{\ \rm M_{\odot}}$ and star
62: formation rates ${\rm SFR}\approx 1-5{\rm\ M_{\odot}\,yr^{-1}}$, we
63: find angular averaged escape fractions of $1-3$\% over the entire
64: redshift interval studied ($3<z<9$). In addition, we find that the
65: escape fraction varies by more than an order of magnitude along
66: different lines-of-sight within individual galaxies, from the largest
67: values near galactic poles to the smallest along the galactic disk.
68: The escape fraction declines steeply at lower masses and SFR. We show
69: that the low values of escape fractions are due to a small fraction of
70: young stars located just outside the edge of HI disk. This fraction,
71: and hence the escape fraction, is progressively smaller in disks of
72: smaller galaxies because their HI disks are thicker and more extended
73: relative to the distribution of young stars compared to massive
74: galaxies. Our results suggest that high-redshift galaxies are
75: inefficient in releasing ionizing radiation produced by young stars
76: into the intergalactic medium. We compare our predicted escape
77: fraction of ionizing photons with previous results, and find a general
78: agreement with both other simulation results and available direct
79: detection measurements at $z\sim3$. We also compare our simulations
80: with a novel method to estimate the escape fraction in galaxies from
81: the observed distribution of neutral hydrogen column densities along
82: the lines of sights to long duration $\gamma$-ray bursts. Using this
83: method we find escape fractions of the GRB host galaxies of
84: $2-3$\%, consistent with our theoretical predictions.
85: \end{abstract}
86:
87: \keywords{cosmology: theory - galaxies: dwarf - galaxies: evolution -
88: galaxies: formation - stars: formation - methods: numerical}
89:
90:
91:
92: %----------------------
93: \section{Introduction}
94: \label{sec:intro}
95: %----------------------
96:
97: Star forming galaxies have a number of important effects on the
98: surrounding intergalactic medium (IGM) and subsequent gas accretion.
99: The ionizing radiation from galaxies is thought to be responsible for the
100: re-ionization of the universe \citep[e.g.,][]{mhr99,bhvs05}, altering
101: the thermal state of the IGM
102: \citep{gh98,g00c,rgs00,stre00,mmrs01,fr07}, reducing gas accretion
103: onto small dwarf galaxies \citep{e92,tw96,dhrw04}, and evaporating the
104: existing gas in small halos \citep{bl99,shaviv_dekel03,sir04}.
105:
106: The amount of radiation emitted by galaxies into the IGM depends not
107: only on the abundance of hot, young stars, but also on the spatial
108: distribution of absorbing gas and
109: dust in individual galaxies and their immediate surroundings. The
110: escape of ionizing radiation from galaxies is therefore the focus of a
111: number of observational and theoretical studies. Nevertheless, the
112: ``escape fraction'', $\fesc$, that characterizes the fraction of total
113: ionizing radiation released into the IGM from individual galaxies
114: remains poorly constrained.
115:
116: Recent empirical measurements of the escape fraction from normal
117: galaxies in the local universe \citep{lfhl95,hsml01,bzaa06} and at
118: high redshifts \citep{gcdf02,flc03,sspa06}\footnote{Note that both
119: \citet{gcdf02} and \citet{sspa06} quote larger values, up to
120: 15\%. However, the quoted numbers are for {\it relative\/} escape
121: fractions between the $1500\AA$ and the Lyman limit. The {\it
122: absolute\/} escape fractions are actually about 3\%, as we discuss in
123: more detail in \S \ref{sec:discussion}.} have generally produced
124: modest values in the range of a few percent. In contrast, theoretical
125: studies of the escape of ionizing radiation from high-redshift
126: galaxies have largely been inconclusive. Many of the previous studies
127: have applied simplified analytic models
128: \citep{ds94,hl97,dsf00,rs00,wl00,co02,fmma03} that predicted a wide
129: range of values for $\fesc$.
130:
131: A more reliable estimate can be derived from
132: self-consistent cosmological numerical simulations of galaxy formation
133: that include three-dimensional radiative transfer and model the
134: three-dimensional distribution of absorbing gas in and around
135: individual galaxies. This approach is computationally challenging and
136: has been attempted only recently at a range of redshifts, from
137: $z\sim20$ \citep{abs06} to more modest redshifts $z\ga2$
138: \citep{rs06,rs07}, where a comparison with the observations is
139: possible.
140:
141: In this paper, we continue this line of work using fully
142: self-consistent cosmological simulations that include radiative
143: transfer and resolve the interstellar medium in modest-sized galaxies
144: at high redshifts. Our work is similar but complementary to the work
145: of \citet{rs06,rs07}. In both approaches, self-consistent cosmological
146: simulations of normal galaxies are used. However, while
147: \citet{rs06,rs07} use the Smooth Particle Hydrodynamics (SPH) for
148: modeling galaxies with modest numbers of stellar particles (up to
149: several thousand), we apply Adaptive Mesh Refinement (AMR) method in
150: this study to reach a larger spatial dynamic range. \citet{rs06,rs07}
151: use the exact ray-tracing method for modeling the radiative transfer
152: of ionizing radiation from simulated galaxies without taking into
153: account effects of radiative transfer during simulation, while we use
154: an approximate Optically Thin Variable Eddington Tensor (OTVET) method
155: of \citet{ga01} with radiative transfer calculations performed during
156: the course of simulation self-consistently.
157:
158: Most importantly, in this work we explicitly include absorption by
159: dust, which allows a direct comparison with
160: observational estimates of the escape fraction. Given the differences
161: between \citet{rs06,rs07} and our approaches, a comparison
162: of the results
163: is useful for estimating the range of systematic numerical
164: uncertainties in the problem, where a rigorous convergence study is
165: not yet feasible.
166:
167: The paper is organized as follows. In \S~\ref{sec:sims} we describe
168: the simulations and numerical techniques used in our study.
169: In \S~\ref{sec:method} we provide operational definition
170: of the escape fraction and in \S~\ref{sec:dust} we describe our
171: model for dust absorption of the ionizing radiation. We present
172: the results of the study in \S~\ref{sec:results} and discuss
173: their implications in \S~\ref{sec:discussion}.
174:
175:
176: %---------------------
177: \section{Simulation}
178: \label{sec:sims}
179: %---------------------
180:
181: The simulation we use in this paper is described in detail by
182: \citet[][where it is labeled as FNEQ-RT]{tkg07}. Here we briefly
183: summarize its setup and numerical parameters.
184:
185: The simulation is performed using the Eulerian, gas dynamics +
186: $N$-body Adaptive Refinement Tree (ART) code \citep{kkk97,kkh02}. A
187: large dynamic range is achieved through the use of Adaptive Mesh
188: Refinement (AMR) in both the gas dynamics and gravity calculations.
189: The calculation is started from a random realization of a Gaussian density
190: field at $z=50$ in a periodic box of $6h^{-1}\ \rm Mpc$ in a flat
191: $\Lambda$CDM model ($\Omega_m=1-\Omega_{\Lambda}=0.3$,
192: $\Omega_b=0.043$, $h=0.7$, $n_s=1$, and $\sigma_8=0.9$). A Lagrangian
193: region corresponding to five virial radii of a Milky Way sized galaxy
194: at $z=0$ is identified and evolved with $2.6\times 10^6$ dark matter
195: particles with masses of $9.2\times 10^5h^{-1}\Msun$.
196:
197: The code employs a uniform $64^3$ grid to cover the entire computational
198: box. The Lagrangian region is, however, always unconditionally refined
199: to the third refinement level, corresponding to the effective grid
200: size of $512^3$. As the matter distribution evolves, the code
201: adaptively and recursively refines the mesh in high-density regions
202: beyond the third level up to the maximum allowed 9th refinement level,
203: which corresponds to the comoving spatial resolution of $260\dim{pc}$,
204: or physical resolution of $65 (50) \dim{pc}$ at $z=3 (4)$.
205:
206: The ART code computes metallicity-dependent, non-equilibrium gas
207: cooling ``on the fly'' based on the abundances of five atomic species
208: and molecular hydrogen. Star formation and feedback (both radiative
209: and thermal via supernova explosions and stellar winds) are included,
210: as described in \citet{tkg07}.
211:
212: The self-consistent 3D radiative transfer of UV radiation from
213: individual stellar particles is followed with the OTVET algorithm
214: \citep{ga01}. The details of our implementation of the OTVET algorithm
215: on adaptively refined meshes will be described elsewhere (Gnedin \&
216: Kravtsov 2007, in preparation). A comparison of our radiative transfer
217: scheme with several other existing numerical implementations for the
218: time-dependent and spatially-inhomogeneous radiative transfer is
219: presented in \citet{icam06} and in the follow-up paper (Iliev et al.\
220: 2007, in preparation).
221:
222: In addition to the main progenitor of the Milky Way type galaxy, the
223: fully refined Lagrangian region contains several dozens of smaller
224: galaxies spanning a wide range of masses. We will use all these
225: systems to estimate the escape fraction of ionizing radiation from
226: star forming regions for galaxies of different masses and star
227: formation rate (SFR).
228:
229:
230:
231: %---------------------
232: \section{Method}
233: \label{sec:method}
234: %---------------------
235:
236: Because our simulation includes an approximate treatment of the
237: radiative transfer self-consistently, the non-equilibrium abundances
238: of all ions for each cell in the AMR grid hierarchy are available
239: throughout the calculation. We can thus measure the opacity at a given
240: frequency along a given direction from any point in the simulation
241: volume to any other point.
242:
243: Since the opacity can only increase along a given line of sight, the
244: escape fraction is, in general, a function of the distance from the
245: source. We choose to measure this distance in units of the virial
246: radius of a given galaxy and we measure the escape fraction at 0.5, 1,
247: and 2 virial radii from the center of the galaxy (defined as the peak
248: of the dark matter density). Unless specified otherwise, hereafter we
249: use the escape fraction at one virial radius as our fiducial working
250: definition of the escape fraction.
251:
252: In computing the escape fraction, it is important to separate
253: satellite galaxies from isolated ones. When a small satellite galaxy
254: is located within a larger galaxy, the concept of virial radius and
255: escape fraction becomes ambiguous. Therefore, we only consider
256: isolated galaxies and only include sources that actually reside in
257: (i.e.\ gravitationally bound to ) the main galaxy, thus excluding the
258: satellites. In computing the continuum
259: absorption by ionic species we include the detailed velocity structure
260: of the galactic and infalling gas, although we find that ignoring the
261: gas velocity field makes no measurable difference on the escape
262: fraction.
263:
264: Operationally, the escape fraction $\fesc$ of a specific galaxy at a given
265: redshift is then a function of two variables: the frequency and the
266: direction of propagation of escaping radiation:
267: \begin{equation}
268: \fesc \equiv \fesc(\nu,\vec{\theta}).
269: \end{equation}
270: For studies of reionization of the universe and the Lyman-alpha forest,
271: the most important quantity is the escape fraction of
272: ionizing radiation,
273: \begin{equation}
274: \fesc^{j}(\vec{\theta}) \equiv
275: \frac{\int d\nu\,\fesc(\nu,\vec{\theta})\sigma^{j}_\nu S_\nu}
276: {\int d\nu\,\sigma^{j}_\nu S_\nu},
277: \label{eq:fion}
278: \end{equation}
279: where $\sigma^{j}_\nu$ is the photo-ionization cross-section for the
280: specie $j$ ($j=\HI,\GI,\GII$) and $S_\nu$ is the spectrum of
281: the sources of radiation. In our simulation, we include only stellar
282: sources and compute the ionizing radiation spectra using the
283: Starburst99 package \citep{lsgd99}. We assume continuous star
284: formation, 0.04 solar metallicity (typical of galaxies in our
285: simulation), and a Salpeter initial mass function over a mass range
286: from 1 to 100 M$_\odot$. The spectral shape is shown in Figure~4 of
287: \citet{rgs02a}. Note that the Starburst99 spectral shape is computed
288: for the unobscured stellar population, which is the appropriate
289: spectral shape to use in equation (\ref{eq:fion}), which explicitly
290: includes the $\fesc$ factor.
291:
292: However, observationally the escape fraction of ionizing radiation is
293: difficult to determine, since it requires measuring the whole ionizing
294: spectrum. Instead, the escape fraction at the hydrogen ionization edge
295: (Lyman limit), $\fesc(\nu_0)$, is usually measured in observational
296: studies. We present the relationship between these two quantities in
297: the appendix.
298:
299:
300:
301: %---------------------
302: \section{Absorption by Dust}
303: \label{sec:dust}
304: %---------------------
305:
306: Incorporating dust absorption into the simulations for calculating the
307: escape fraction is critical, because dust may contribute significantly
308: to the absorption of ionizing radiation
309: \cite[e.g.][]{wd01}. Unfortunately, properties of dust in
310: high-redshift galaxies are not known well. In particular, dust
311: absorption cross section depends on the dust composition and grain
312: size distribution, which is measured only in nearby galaxies. In our
313: analysis, we adopt the dust extinction curves for Large and
314: Small Magellanic Clouds (LMC and SMC respectively) from \citet{wd01}
315: as representative of high redshift galaxies,
316: because low metallicities of these galaxies are closer to typical
317: metallicities of high-redshift galaxies in our simulation.
318:
319: A convenient parameterization for the dust extinction law in
320: LMC and SMC is given by \citet{p92}, based on the earlier data,
321: \begin{equation}
322: \sigma^{\rm d}_\nu = \sigma_0 \sum_{i=1}^7
323: f(\lambda/\lambda_i,a_i,b_i,p_i,q_i),
324: \label{eq:pei}
325: \end{equation}
326: where the fitting function $f$ has the form
327: \begin{equation}
328: f(x,a,b,p,q) = \frac{a}{x^p + x^{-q} + b}.
329: \label{eq:fit}
330: \end{equation}
331:
332: \begin{deluxetable}{lccccc}
333: \tablecaption{Parameters for the dust extinction law fits\label{tab:pei}}
334: \tablewidth{\columnwidth}
335: \tablehead{
336: \colhead{Term} & \colhead{$\Lambda_i$ ($\mu$m)} &
337: \colhead{$a_i$} & \colhead{$b_i$} & \colhead{$p_i$} & \colhead{$q_i$}
338: }
339: \startdata\\
340: \cutinhead{SMC dust model}
341: 1 & 0.042 & 185 & 90 & 2 & 2\\
342: 2 & 0.08 & 27 & {\bf 15.5}\tablenotemark{a} & 4 & 4\\
343: 3 & 0.22 & 0.005 & -1.95 & 2 & 2\\
344: 4 & 9.7 & 0.010 & -1.95 & 2 & 2\\
345: 5 & 18 & 0.012 & -1.8 & 2 & 2\\
346: 6 & 25 & 0.030 & 0 & 2 & 2\\
347: 7 & {\bf 0.067} & {\bf 10} & {\bf 1.9} & {\bf 4} & {\bf 15}\\
348: \cutinhead{LMC dust model}
349: 1 & 0.046 & {\bf 90} & 90 & 2 & 2\\
350: 2 & 0.08 & 19 & {\bf 21} & 4.5 & 4.5\\
351: 3 & 0.22 & 0.023 & -1.95 & 2 & 2\\
352: 4 & 9.7 & 0.005 & -1.95 & 2 & 2\\
353: 5 & 18 & 0.006 & -1.8 & 2 & 2\\
354: 6 & 25 & 0.02 & 0 & 2 & 2\\
355: 7 & {\bf 0.067} & {\bf 10} & {\bf 1.9} & {\bf 4} & {\bf 15}\\
356: \enddata
357: \tablenotetext{a}{Values changed from \citet{p92} are shown in bold.}
358: \end{deluxetable}
359:
360: We correct the fits of \citet{p92} using the newer data from
361: \citet{wd01} by adding the seventh term in equation (\ref{eq:pei}) in
362: order to account for the narrow and asymmetric shape of the FUV peak
363: in the dust extinction law. We also change values for some parameters from
364: those adopted by \citet{p92}. The values for the parameters we use are
365: listed in Table \ref{tab:pei}, and we show in bold changes from
366: \citet{p92}.
367:
368: The overall normalization for the dust cross section is determined by
369: parameters $\sigma_{0, {\rm LMC}}=5.6\times10^{-22}\dim{cm}^2$ and
370: $\sigma_{0, {\rm SMC}}=1.1\times10^{-22}\dim{cm}^2$. With these
371: functional forms, the dust cross sections for both LMC and SMC fit the
372: plotted curves of \citet{wd01} to a few percent accuracy.
373:
374: In order to limit the range of possible dust effects, we consider three
375: extreme dust models: 1) a model with no dust at all, 2) a
376: model in which we assume that the dust column density scales with
377: neutral (atomic and molecular) gas column density,
378: \begin{equation}
379: N_{\rm dust} = \frac{Z}{Z_0} \times (N_{\HI}+2N_{\rm H_2})
380: \label{eq:insub}
381: \end{equation}
382: (where $Z$ is the gas metallicity), and 3) a model where the dust
383: column density scales with the total gas column density (both neutral
384: and ionized),
385: \begin{equation}
386: N_{\rm dust} = \frac{Z}{Z_0} \times N_{\rm H},
387: \label{eq:nosub}
388: \end{equation}
389: where $Z_0=0.32$ ($0.2$) is the gas-phase metallicity of the LMC
390: \citep[or SMC,][]{wlbh97,wfsy99}. Note that emission line studies
391: \citep{ppr00,kw06} usually indicate somewhat larger values for both metallicities
392: than the absorption line metallicities that we adopt.
393: If we adopted larger values for the LMC and SMC
394: metallicities, the dust effect on our measured escape fractions would
395: be even {\it smaller\/}.
396:
397: Physically, equation (\ref{eq:insub}) implies a complete instant
398: sublimation of dust in the ionized gas, while equation
399: (\ref{eq:nosub}) implies no dust sublimation at all. Of course, the
400: truth lies somewhere in between, but these two extreme cases bracket
401: the range of possibilities.
402:
403: \begin{figure}[t]
404: \plotone{\figname{figFN.ps}{f1.eps}}
405: \caption{The frequency dependence of the angular averaged escape
406: fraction for the central galaxy at $z=3$ for several dust models: no
407: dust (solid line), SMC dust with instant sublimation (eq.\
408: [\ref{eq:insub}], dotted line), LMC dust with instant sublimation (eq.\
409: [\ref{eq:insub}], short-dashed line), and SMC dust with no
410: sublimation (eq.\ [\ref{eq:nosub}], long-dashed line). In addition,
411: we also show with a short-long-dashed line the SMC no
412: sublimation model with the dust cross section increased by a factor
413: of 3, as a certain {\it overestimate\/} of the dust absorption. The
414: black point with error-bars shows the extinction at $1600\AA$ from
415: \citet[][derived as $A_{1600}=4.43+1.99\beta$ with
416: $\beta=-1.46\pm0.48$]{as00}.}
417: \label{figFN}
418: \end{figure}
419: Figure~\ref{figFN} shows the angular averaged escape fraction for the
420: Milky Way progenitor galaxy at $z=3$ as a function of frequency. We
421: show several dust models, together with typical reddening correction
422: for Lyman Break Galaxies from \citet{as00}. We also show the SMC no
423: sublimation dust model with dust cross section arbitrarily increased
424: by a factor of 3. As we have mentioned above, the no sublimation model
425: is likely to overestimate the effect of dust absorption. Since it is
426: highly unlikely that uncertainties in our adopted value for the SMC
427: metallicity and the dust extinction curve are as large as a factor of
428: 3, this latter model serves as an absolute (and, likely, implausibly
429: high) upper limit for the dust absorption.
430:
431: \begin{figure}[t]
432: \plotone{\figname{figLS.ps}{f2.eps}}
433: \caption{The luminosity weighted distribution of the escape fraction
434: at the Lyman limit in a random direction on the sky from all
435: positions within the central galaxy for the SMC instant sublimation
436: dust model (our fiducial one). Three separate distributions are
437: shown: absorption by gas only (dotted line), absorption by dust only
438: (dashed line), and total absorption (dust + gas, solid line). Notice
439: that the gas-only distribution is essentially identical to the total
440: distribution, because $Z\sigma^{\rm d} \ll \sigma^{\HI}$ at the
441: Lyman limit; so the dotted line coincides with the solid
442: one. The histograms are normalized to add up to 1, so the spike at
443: $\fesc\approx0$ contains most of the data points.\newline}
444: \label{figLS}
445: \end{figure}
446: A rather unexpected feature of Figure~\ref{figFN} is that
447: the escape fraction above the Lyman limit is almost independent of the
448: dust absorption. In order to understand this phenomenon, we show in
449: Figure~\ref{figLS} the distributions of escape fractions from all
450: positions within the central galaxy in a fixed, randomly selected
451: direction on the sky. The
452: distributions have similar shapes: they are weak
453: functions of $\fesc$ for $0<\fesc<1$ and exhibit a primary peak at
454: $\fesc\approx0$ and a secondary peak at $\fesc\approx1$. Thus, the
455: escaping radiation is produced {\it not by sources in a semi-opaque
456: medium,\/} with each source attenuated by a similar amount, but {\it
457: by a small fraction of essentially unobscured sources\/}. We discuss
458: this point in detail in \S~\ref{sec:results} below (see
459: Fig.~\ref{figDK}).
460:
461:
462: For a distribution like those in Figure~\ref{figLS}, the effects of
463: dust absorption can be easily understood, if we
464: ignore the ``translucent points'' at $0<\fesc<1$ and use a toy-model
465: distribution for the gas-only $\fescg$,
466: \begin{equation}
467: p(\fescg) = \alpha\delta(1-\fescg) + (1-\alpha)\delta(\fescg),
468: \label{eq:fdist}
469: \end{equation}
470: where $\alpha\ll1$ is constant and $\delta(x)$ is a Dirac delta
471: function. The average escape fraction in this model is simply
472: \begin{equation}
473: \langle\fescg\rangle = \int_0^1 \fescg p(\fescg) d\fescg = \alpha.
474: \label{eq:favg}
475: \end{equation}
476: If the dust absorption is included, then the escape
477: fraction at each location is changed by a factor that
478: depends on the dust opacity,
479: $$
480: \fesc = \fescg e^{-\tau_{\rm d}} \equiv e^{-\tau_{\HI}-\tau_{\rm d}}.
481: $$
482: At the locations where $\fescg\approx1$ ($\tau_{\HI}\ll 1$),
483: the effect of dust is negligible (since $\tau_{\rm d}\ll\tau_{\HI}$).
484: At locations where $\tau_{\rm d}\ga 1$, the hydrogen opacity is
485: already so large that no radiation escapes from this position,
486: irrespective of how much dust is mixed with the gas. As the result,
487: the escape fraction does not change at all.
488:
489: In reality, ``translucent points'' with $0<\fesc<1$ are affected by
490: dust, but their integrated contribution to the average escape fraction
491: remains small. In the no sublimation dust model, the situation
492: may be more complex, since there is dust absorption in the ionized
493: gas. However, Figure~\ref{figFN} demonstrates that this is not a large
494: effect.
495:
496: At frequencies below the Lyman limit the distribution of $\fesc$ from
497: dust is similar to Figure~\ref{figLS}, but there is no gas
498: absorption. Because dust absorption is weaker than gas absorption,
499: the average escape fraction at these frequencies is much larger and
500: increases to unity as the frequency decreases down to infra-red.
501:
502: The important result of this section is that, while dust absorption is
503: the dominant effect for UV radiation below the Lyman Limit, it does
504: not substantially affect the escape fraction of ionizing radiation
505: in our model galaxies, as Figure~\ref{figFN} demonstrates.
506:
507: In the rest of this paper, we adopt the SMC instant sublimation model
508: as our fiducial dust model, but we note that our final results are not
509: sensitive to this particular choice.
510:
511:
512:
513: %-------------------
514: \section{Results}
515: \label{sec:results}
516: %-------------------
517:
518: \begin{figure}[t]
519: \plotone{\figname{fesc-z5.2.ps}{f3a.eps}}
520: \plotone{\figname{fesc-z4.2.ps}{f3b.eps}}
521: \plotone{\figname{fesc-z3.2.ps}{f3c.eps}}
522: \caption{The escape fraction of $\HI$ ionizing radiation $\fesc^{\HI}$
523: as a function of direction, as visible from the center of the Milky
524: Way type galaxy, at the virial radius of the galaxy at $z=5$, $4$,
525: and $3$. The sky is oriented in coordinates aligned with galactic
526: gas disk such that the disk is horizontal in the middle of the plot. Note a
527: different color scale for $z=5$ image.}
528: \label{figIM}
529: \end{figure}
530:
531: Figure~\ref{figIM} shows the escape fraction of hydrogen ionizing
532: radiation $\fesc^{\HI}$ (eq.\ \ref{eq:fion}) as seen from the center
533: of the central galaxy (the Milky Way progenitor) at three different
534: redshifts. The celestial coordinates in Figure\ \ref{figIM} are
535: aligned with the principal axes of the galaxy (``galactic''
536: coordinates), with the plane of the disk of the galaxy crossing the middle of
537: the plot. Typically, the escape fraction is close to zero along the
538: plane of the disk and approaches maximum values near the poles,
539: although there are significant small-scale variations at all redshifts
540: that underscore the complex, perturbed nature of high redshift disk
541: galaxies. At $z=5$ the galaxy is experiencing a substantial major
542: merger (and a lesser one at $z=3$), so the angular distribution of
543: escape fractions is more irregular at these epochs.
544:
545: Another interesting feature of the $\fesc$ distribution shown in
546: Figure~\ref{figIM} is a few small very opaque (dark blue) clouds of
547: gas that block ionizing radiation at larger distances. These clouds
548: can be counterparts of the Lyman Limit absorbers observed in the
549: spectra of distant quasars. Previous studies, based on lower
550: resolution simulations, have indicated that Lyman Limit systems tend
551: to cluster around large galaxies \citep{kg07}. Our simulations
552: confirm that such clouds do exist around high-redshift galaxies.
553:
554: We have checked that, as we integrate further in distance, more of
555: Lyman Limit systems fall inside the radius of integration, and the
556: ``sky'', as seen from the center of the galaxy, appears progressively
557: more opaque. The sky should become completely opaque at a distance of
558: a few mean free paths for ionizing radiation. At $z=4$ the mean free
559: path for ionizing radiation is about $85\,h^{-1}$ co-moving Mpc
560: \citep{m03}, much larger than the size of our computational box. Thus,
561: we cannot actually reach the full opacity with our current
562: simulation. However, the important point to make is that the mean free
563: path is much larger than a virial radius of any galaxy at these
564: redshifts, so the concept of the ``escape fraction'' is well defined
565: and robust at intermediate redshifts.
566:
567: \begin{figure}[t]
568: \plotone{\figname{figHA.ps}{f4.eps}}
569: \caption{The probability per unit log and per unit solid angle
570: for having a specific
571: value of the escape fraction of $\HI$ ionizing radiation for the
572: central galaxy at $z=4$ for 3 distances from the galaxy center (as
573: measured in units of the galaxy virial radius). The shaded region
574: around the middle curve shows the Poisson errors due to the finite
575: number of pixels in the angular maps.}
576: \label{figHA}
577: \end{figure}
578:
579: Figure~\ref{figHA} illustrates this point in a quantitative way. It
580: shows the probability density for having a specific value of the
581: escape fraction on a celestial sphere at different distances from the
582: center of the main progenitor of the Milky Way-sized galaxy. As the
583: distance form the galaxy increases, the tail of the distribution at
584: low escape fractions grows, as Lyman limit systems cover a
585: progressively larger fraction of the sky. However, at distance
586: comparable to the virial radius of the galaxy the effect of Lyman
587: limit systems is small, as can be seen from the plot: the fraction of
588: the sky below $\fesc^{\HI}=0.003$ increases from 3\% to only 6\% as
589: the distance increases from $0.5R_{\rm VIR}$ to $2R_{\rm VIR}$.
590:
591: Poisson errors on the probability density shown in Figure\ \ref{figHA}
592: remain small for escape fractions well below $10^{-3}$, indicating
593: that angular maps from Figure\ \ref{figIM} resolve the structure in the
594: gas down to that level. Of course, the small-scale structure in the
595: gas is limited by the finite resolution of our simulation.
596:
597: \begin{figure}[t]
598: \plotone{\figname{figFM.ps}{f5a.eps}}
599: \plotone{\figname{figFS.ps}{f5b.eps}}
600: \caption{The angular averaged escape fraction of $\HI$ ionizing
601: radiation as a function of galaxy mass (top) and star
602: formation rate
603: (bottom) at three redshifts (squares). The error-bars show the
604: 10\%--90\% range for all possible directions.}
605: \label{figFM}
606: \end{figure}
607:
608: Figure~\ref{figFM} presents the main result of this paper: the angular
609: averaged escape fraction for hydrogen ionizing radiation as a function
610: of galaxy mass or star formation rate at a range of redshifts. We only
611: show results for the galaxies which are resolved down to the maximum
612: ninth level of refinement. This resolution criterion corresponds
613: approximately to a minimum mass of $10^{10}\Msun$ (or $\gtrsim 10^4$
614: dark matter particles). A general trend of increasing escape fraction
615: with increasing galaxy mass and SFR is clearly observed: the escape
616: fraction changes by approximately two orders of magnitude from
617: $\approx 10^{-4}$ to $\approx 0.02-0.05$ for total masses between
618: $10^{10}$ to $4\times 10^{11}\Msun$ (or SFR from $0.1$ to
619: $\approx 7\,\Msun\dim{yr}^{-1}$). At the same time, the figure also
620: shows that there is little change with redshift, from $z=5$ to $z=3$,
621: either in the values of the escape fraction or in the dominant trend
622: with mass or SFR.
623:
624: \begin{figure}[t]
625: \plotone{\figname{figFZ.ps}{f6.eps}}
626: \caption{The evolution of the angular averaged escape fraction of
627: $\HI$ ionizing radiation for the most galaxy ($M_{\rm
628: tot}=4\times10^{11}\Msun$ at $z=3$). The shaded band shows the
629: 10\%--90\% range for all possible directions.}
630: \label{figFZ}
631: \end{figure}
632:
633: Figure~\ref{figFZ} further illustrates the lack of redshift evolution in
634: the escape fraction found in our simulations. The figure shows the
635: average escape fraction of the most massive Milky Way progenitor in the
636: simulation at different redshifts, along with the variation in escape fraction
637: in different directions. While the escape fraction fluctuates with
638: redshift, a spread in the escape fraction in different directions at a
639: given redshift is much larger than the variation in the average escape
640: fraction with redshifts at $z<7$. A similarly weak trend is found
641: by most well resolved galaxies from our simulation.
642:
643: \begin{figure}[t]
644: \plotone{\figname{figFI.ps}{f7.eps}}
645: \caption{The mass dependence of the escape fraction of ionized
646: radiation for three ionic species at $z=3$ (similar to the top panel
647: of Figure\ \ref{figFM}). We show only well-resolved galaxies and omit
648: error-bars for clarity. The escape fractions are clumped at the
649: level of $10^{-6}$ to allow all points to be visible on the graph.}
650: \label{figFI}
651: \end{figure}
652:
653: There is clearly a significant drop in the escape fraction at $\GII$
654: ionization threshold. This is further illustrated in Figure\
655: \ref{figFI}, which shows all three escape fractions for ionizing
656: radiation as a function of the total mass for model galaxies. We note
657: that $\HI$ and $\GI$ are invariably close to each other, while $\GII$
658: escape fractions are systematically much lower. This behavior is again
659: completely expected, as only active galactic nuclei are thought to
660: fully (i.e.\ doubly) ionize helium.
661:
662: \begin{figure}[t]
663: \plotone{\figname{disk.ps}{f8.eps}}
664: \caption{The pseudo-color composite edge-on view of the main
665: galaxy. The top half shows the gas density, while the bottom half
666: depicts the neutral hydrogen fraction. Each image is composed of 3
667: snapshots from the simulation taken $10\dim{Myr}$ apart at $z\sim4$
668: and combined as red, green, and blue channels of a color image. The
669: predominance of purple (red and blue) color on top and green color
670: on bottom indicates that the galactic disk oscillates globally on a
671: time scale of about $20\dim{Myr}$. The oscillation pattern is much
672: more chaotic in $\HI$ image outside of about $1\dim{kpc}$. (This
673: figure should be viewed in color.)}
674: \label{figDK}
675: \end{figure}
676:
677: We also study the mechanism for the escape of ionizing radiation from
678: galactic disks in our simulations and for the predicted small values
679: of $\fesc$. As Figure~\ref{figLS} demonstrates, the ionizing radiation
680: that escapes is emitted preferentially by unobscured sources, with
681: only a small fraction of all sources being unobscured at
682: any given time (rather than being emitted by the majority of sources,
683: which are only partially obscured). In order to visualize the regions
684: of a galaxy where these unobscured sources reside,
685: Figure~\ref{figDK} shows a pseudo-color composite image of three
686: different snapshots from the simulation closely spaced in time.
687:
688: The edge-on disk of the main galaxy is shown in the total gas density
689: (top) and in the $\HI$ fraction (bottom). Three snapshots
690: $10\dim{Myr}$ apart at $z\sim4$ are overlaid together so that the
691: first snapshot is shown in red, the second in green, and the last one
692: in blue. If the gas distribution did not change between the snapshots, then the
693: image would appear as an equal combination of red, green, and blue,
694: i.e.\ as a pure grayscale. In reality, however, the upper surface of
695: the disk appears purple (red plus blue), while the bottom surface is
696: pure green. This means that the galactic disk oscillates with a period
697: of about $20\dim{Myr}$: the disk bended downwards between the first
698: (red) snapshot and the second (green) snapshot, but returned to the
699: first configuration in the third (blue) snapshot, so that the first
700: and third snapshots look almost identical, blending into a single
701: purple color. Oscillating modes with shorter wavelengths are also
702: visible in the image.
703:
704: This behavior is clearly visible in the total gas density image and in
705: the $\HI$ fraction image within the inner $0.5-0.7\dim{kpc}$; the
706: $\HI$ disk oscillates much more violently at larger radii. This is not
707: surprising, because the oscillations of the $\HI$ edge are subject to
708: the positive feedback. As the $\HI$ edge oscillates and uncovers
709: ionizing sources, the ionizing intensity in a given point is likely to
710: increase (similarly to how the ionizing intensity increases when
711: isolated $\HII$ regions merge during reionization).
712:
713: Figure~\ref{figDK} thus shows why ionizing radiation of some young
714: stars can leave the galaxy with virtually no attenuation. The cold
715: $\HI$ disk, where young stars form is very thin ($\sim
716: 100-200\dim{pc}$). Some of the stars may have sufficient velocity to move
717: outside the edge of the $\HI$ disk while they are still young, bright
718: emitters of UV photons. For example, a star traveling at
719: $10\dim{km}\dim{s}^{-1}$ will move by $\approx 100\dim{pc}$ in $10^7$ years.
720: In addition, the outer edge of the $\HI$ disk is changing significantly
721: on the same time scale. These changes can expose some of the young
722: stars and clear the way for the ionizing radiation to leave the
723: system. {\em The escaped ionizing radiation is thus mostly due to a
724: small fraction of stars in a thin shell surrounding the $\HI$ disk.}
725:
726: The relative constancy of the escape fraction with redshift or SFR
727: that we find is then a simple consequence of the galactic disks
728: oscillating by a comparable amount on $10\dim{Myr}$ time scale. Note
729: that this time scale is close to the local dynamical time scale of the
730: dense high-redshift disks and lifetime of massive, UV-bright
731: stars. Such oscillations are thus likely dynamical in origin: the
732: disk can be perturbed by mergers and interactions with satellites, which
733: are only weakly correlated with SFR or redshift (unless a disk goes
734: into a starburst phase).
735:
736: This picture also explains why the escape fraction decreases steeply
737: for lower mass galaxies. Gas distribution in dwarf galaxies is
738: relatively more extended in comparison to the stellar distribution in
739: our simulations, due to lower efficiency of star formation in dwarfs
740: relative to massive galaxies \citep[see][for detailed
741: discussion]{tkg07}. Stars are thus concentrated towards central
742: regions of the disk and a smaller fraction of them can reach the edge
743: of $\HI$ disk or get exposed by fluctuations of its boundary.
744: Note, however, that resolution of dwarf galaxies in our simulations
745: is considerably worse than for the massive galaxies. This explanation
746: and the trend with mass will thus need to be
747: verified in the future by higher resolution simulations.
748:
749:
750: %-----------------------------------
751: \section{Discussion and Conclusions}
752: \label{sec:discussion}
753: %-----------------------------------
754:
755: \begin{figure}[t]
756: \plotone{\figname{figFO.ps}{f9.eps}}
757: \caption{A comparison of the angular averaged escape fraction at
758: hydrogen ionization edge at $z=3$ from various
759: determinations. Filled red/gray upward triangle and circles show the
760: observational measurements from \citet{sspa06} and \citet{flc03}
761: respectively, presented as average values for their data
762: samples. Two filled red/gray downward triangles show the upper limits from
763: \citet{gcdf02}. Both \citet{sspa06} are \citet{gcdf02} points
764: are corrected
765: from the measured relative escape fraction to the absolute escape
766: fraction, as explained in the text (open symbols show
767: the original uncorrected measurements). The filled blue/black diamond
768: is the simulation results of \citet{rs06} for the single
769: galaxy they report the star formation rate for. Filled blue/black
770: squares with error-bars show our results, similar to
771: Figure~\ref{figFM}.}
772: \label{figFO}
773: \end{figure}
774:
775: To summarize the current understanding of the escape fraction of
776: photons at the ionizing threshold of hydrogen, we show in
777: Figure~\ref{figFO} both simulation predictions and observational
778: constraints available at $z=3$. The simulation predictions are taken
779: from \citet{rs06} and our work. Observational constraints are from
780: \citet[][upper limits only]{gcdf02}, \citet{flc03}, and
781: \citet{sspa06}. The measurements of \citet{gcdf02} are for individual
782: galaxies with SFR taken from \citet{pksd98}. The measurement of
783: \citet{sspa06} is an average of 14 star-forming galaxies at $z\sim 3$.
784: The SFR is estimated based on the observed UV flux and corrected for
785: dust extinction for $\langle E(B-V)\rangle=0.11$. Both \citet{gcdf02}
786: and \citet{sspa06} measurements are corrected from the relative
787: measurements of the ratio $\fesc(912\AA)/\fesc(1500\AA)$ they report
788: to the absolute measurement of $\fesc(912\AA)$ by adopting a value of
789: $\fesc(1500\AA)=0.2$ from \citet{as00}. This correction factor is also
790: consistent with other observed estimates of reddening at $1500\AA$
791: \citep{pksd98,spa01,sspa06}. The measurements of \citet{flc03} are
792: averaged over 14(13) high(low) luminosity galaxies found at $1.9 < z <
793: 3.5$ in the Hubble Deep Field (the second one is actually an upper
794: limit). The star formation rates are conversions of the observed UV
795: flux using the scaling relation from \citet{mpd98}, and are corrected
796: for dust extinction assuming $E(B-V)=0.1$ and the extinction law from
797: \citet{cabk00}. \citet{flc03} measure the absolute escape fraction, so
798: no correction is applied to their points.
799:
800: There still exists a gap between simulations and observations in the
801: luminosity and SFR parameter space. While our simulations model
802: modest-mass galaxies, observational measurements are reported mostly
803: for the brightest, $\ga L_{\ast}$ galaxies. Nevertheless, the
804: simulations are tantalizingly close to reaching a similar range of
805: luminosity and SFR covered by the observations. Both the observations
806: and our simulations indicate little (if any) trend in the {\it
807: absolute\/} escape fraction with the galaxy mass or star formation
808: rate, except for a rapid drop in $\fesc$ for smaller mass
809: galaxies. Our simulations indicate that this characteristic scale
810: corresponds to about SFR$\sim 1\,\Msun\,{\rm yr^{-1}}$ or $M_{\rm
811: tot}\sim\,10^{11}\Msun$, although an upper limit of $\fesc<0.4\%$ from
812: \citet{flc03} at SFR$\sim10\Msun\,{\rm yr^{-1}}$ may indicate that
813: simulations overestimate escape fractions in galaxies with
814: SFR$\la10\Msun\,{\rm yr^{-1}}$. The level of the disagreement (if any)
815: cannot yet be accurately deduced from our simulations, and, formally,
816: the \citet{flc03} upper limit is fully consistent with our results.
817: Note also that \citet{siana_etal07} in their recent study of the Lyman
818: continuum escape fractions in galaxies at $z\sim 1$ also find small
819: values ($\fesc\lesssim 0.1$), which is consistent with weak evolution
820: of escape fractions we find in our simulations. The general agreement
821: between our results and observations is definitely encouraging and may
822: indicate that relative distribution of the young stars and UV
823: absorbing gas is modeled faithfully in our high-resolution AMR
824: simulations.
825:
826: \begin{figure}[t]
827: \plotone{\figname{figGB.ps}{f10.eps}}
828: \caption{A comparison of the cumulative distribution of $\HI$ column
829: densities toward ionizing sources from our simulation at $z=3$ (dashed
830: line) with that
831: measured in the host galaxies of long duration $\gamma$-ray bursts at
832: $z>2$. The black solid line shows the observational data compiled by
833: \citet{velw04} and \citet{jflv06}, as well as the data for two
834: additional sources from \citet{sbpf06} and \citet{rstm07}. In two
835: cases (GRB 060124 and GRB 060607) only upper limits have been
836: measured. The dotted line is a power-law fit to the observational data
837: for the range of $\HI$ column densities between $10^{19.5}\dim{cm}^2$
838: and $10^{21.5}\dim{cm}^2$.}
839: \label{figGB}
840: \end{figure}
841: An interesting alternative way of constraining the escape fraction
842: observationally, independent of the observed UV light, is offered by
843: the observed distribution of $\HI$ column densities in the host
844: galaxies of long duration $\gamma$-ray bursts (GRBs; Chen, Prochaska,
845: \& Gnedin 2007, in preparation). Assuming that GRBs sample the
846: distribution of young stars in an unbiased fashion, the fraction of
847: young stars located in ionized regions with small or negligible
848: attenuation of UV radiation should correspond to the fraction of GRBs
849: with $\HI$ column densities lower than some threshold value corresponding
850: to the transition between neutral and ionized hydrogen.
851:
852: Figure~\ref{figGB} shows the observed distribution of $\HI$ column
853: densities in spectra of GRBs with confirmed redshifts of $z>2$,
854: compared to the corresponding distribution in our simulation at
855: $z=3$. The observed distribution is remarkably close to a power-law
856: for $10^{19.5}\dim{cm}^2 < N_{\HI} <
857: 10^{21.5}\dim{cm}^2$. Extrapolating this power-law to lower column
858: densities of $\approx (5-8)\times10^{17}\dim{cm}^2$
859: (or $\tau_{LL}=3-5$, appropriate for the boundary between the ionized and
860: neutral gas\footnote{The specific location of the ionization edge of a
861: galaxy depends on a number of factors and does not correspond to a
862: particular value of $\tau_{LL}$. A value of $\tau_{LL}=1$, which
863: corresponds to the suppression of ionizing background by only a factor
864: of 3, is certainly not enough to make the gas substantially neutral. A
865: value of $\tau_{LL}=10$ is, on the other hand, enough to make the gas
866: even at the mean cosmic density largely neutral. Thus, a value of
867: $\tau_{LL}$ around 3 to 5 should correspond to the approximate
868: location for the galaxy ionization edge.})
869: implies that about $2-3$\% of all GRBs are located in ionized regions
870: in which the UV radiation of their progenitors should have escaped
871: freely into the IGM, so that
872: $$
873: \fesc \approx f_{\rm GRB}(N_{\HI} < (5-8)\times10^{17}\dim{cm}^2) = 0.02-0.03.
874: $$
875: This value
876: is remarkably close to the escape fractions of $\approx 1-3\%$
877: measured in our simulations. This test also provides strong support
878: for our interpretation of why the escape fractions are so low.
879:
880: Note that the $\HI$ column density distribution toward ionizing
881: sources (i.e. column densities from source locations to one virial
882: radius, as explained in \S 3) from our fiducial
883: simulation at $z=3$ is reasonably close to the observed distribution
884: of $N_{\HI}$ from GRBs for $N_{\HI}<3\times10^{21}\dim{cm}^2$. At
885: higher column density we do not expect a good agreement between our
886: simulation and the data, because the simulation does not incorporate
887: the physics of formation and self-shielding of molecular hydrogen
888: correctly, which becomes important in that regime. Thus, the
889: simulation is expected to overpredict $N_{\HI}$ column densities,
890: because in reality some of this $\HI$ is in molecular form.
891:
892: While higher resolution simulations and deeper observations are needed
893: to bridge the remaining small gap in the probed star formation regime
894: between the data and the theory, our results suggest that average
895: escape fractions from bright galaxies at intermediate redshifts do not
896: depend strongly on galaxy properties or redshift. This is interesting
897: because this implies that porosity of the interstellar medium in
898: galaxies, and hence its opacity to ionizing radiation, does not
899: dramatically change at higher star formation rates, as expected in
900: some of the theoretical models \citep[e.g.,][]{co02}. Note that
901: observed escape fractions at higher star formation rates are
902: consistent with being simply an extrapolation of the trend seen in
903: simulations at lower rates. If this is indeed the case, this implies
904: absence of a well-defined critical SFR above which the escape fraction
905: sharply increases to unity.
906:
907: At the same time, the low values of escape fractions found
908: in our simulations suggest that high-redshift galaxies are quite
909: inefficient in emitting ionizing radiation produced by their young
910: stars into the intergalactic medium, with escape fractions decreasing
911: sharply for galaxies of $M_{\rm tot}\lesssim 10^{11}\Msun$ (or
912: SFR$\lesssim 1\Msun\dim{yr}^{-1}$). This conclusion potentially has
913: important implications for the contribution from normal galaxies to
914: the early reionization of the universe and for the relative role of
915: galaxies and quasars in keeping the universe ionized at intermediate
916: redshifts.
917:
918:
919: \acknowledgements
920:
921: We thank Douglas Rudd for constructive and useful comments on the
922: draft of this paper. This work was supported in part by the DOE and
923: the NASA grant NAG 5-10842 at Fermilab, by the NSF grants AST-0206216,
924: AST-0239759, and AST-0507596, and by the Kavli Institute for
925: Cosmological Physics at the University of Chicago. Supercomputer
926: simulations were run on the IBM P690 array at the National Center for
927: Supercomputing Applications and San Diego Supercomputing Center (under
928: grant AST-020018N) and the Sanssouci computing cluster at the
929: Astrophysikalisches Institut Potsdam. This work made extensive use of
930: the NASA Astrophysics Data System, {\tt arXiv.org} preprint server,
931: and the HEALPix\footnote{\tt http://healpix.jpl.nasa.gov}
932: \citep{ghbw05} package.
933:
934:
935:
936: \appendix
937:
938: \section{Escape fraction for ionizing radiation and the escape
939: fraction at the ionization edge}
940:
941: \begin{figure}[t]
942: \plotone{\figname{figFF.ps}{f11.eps}}
943: \caption{A comparison of the angular averaged escape fraction for
944: $\HI$ ionizing radiation (eq.\ \ref{eq:fion}) with the angular
945: averaged escape fraction at the ionization edge (Lyman limit, the quantity
946: usually measured in observational work), $h\nu_0=13.6\dim{eV}$.
947: Solid and dotted lines show the $\fesc^{\HI}=\fesc(\nu_0)$ and
948: $\fesc^{\HI}=1.25\fesc(\nu_0)$ relations respectively.}
949: \label{figFF}
950: \end{figure}
951:
952: In order to facilitate comparison between the theoretical and
953: observational results, in Figure\ \ref{figFF} we show a relationship
954: between the theoretically relevant escape fraction of ionizing
955: radiation (i.e.\ an integral quantity, as defined in eq.\
956: \ref{eq:fion}, since it determines the ionization state of the IGM)
957: and the escape fraction at the hydrogen ionization edge (Lyman limit),
958: which is usually measured in observational studies. We find a tight
959: correlation between the two quantities in the form
960: \begin{equation}
961: \fesc^{\HI}=1.25\fesc(\nu_0),
962: \label{eq:ff}
963: \end{equation}
964: which is helpful for relating observationally measured and
965: theoretically relevant quantities. We use this relation in Fig.\
966: \ref{figFO} when comparing observational and theoretical values on the
967: same plot.
968:
969:
970: \bibliographystyle{apj}
971: \bibliography{igm,gnedin,misc,dsh,grb}
972:
973: \end{document}
974:
975:
976: