astro-ph0110280/ms.tex
1: %% Following produces a manuscript format
2: % \documentclass{aastex}
3: 
4: %% preprint produces a one-column, single-spaced document:
5: \documentclass[12pt,preprint]{aastex}
6: 
7: %% preprint2 produces a double-column, single-spaced document:
8: % \documentclass[preprint2]{aastex}
9: 
10: %% Following would do an AASTeX 4 version if needed
11: %\documentstyle[11pt,aaspp4,tighten,flushrt]{article}  
12: \begin{document}
13: 
14: %\input def.tex
15: % Macro definitions for the z=5.7 candidates paper for LALA
16: %
17: \newcommand{\lya}{Lyman-$\alpha$}
18: \newcommand{\eqw}{\hbox{EW}}
19: \def\erg{\hbox{erg}}
20: \def\cm{\hbox{cm}}
21: \def\sec{\hbox{s}}
22: \def\f17{f_{17}}
23: \def\Mpc{\hbox{Mpc}}
24: \def\nm{\hbox{nm}}
25: \def\km{\hbox{km}}
26: \def\kms{\hbox{km s$^{-1}$}}
27: \def\year{\hbox{yr}}
28: \def\deg{\hbox{deg}}
29: \def\arcsec{\hbox{arcsec}}
30: \def\microJy{\mu\hbox{Jy}}
31: \def\zre{z_r}
32: \def\fesc{f_{\rm esc}}
33: 
34: \def\ergcm2s{\ifmmode {\rm\,erg\,cm^{-2}\,s^{-1}}\else
35:                 ${\rm\,ergs\,cm^{-2}\,s^{-1}}$\fi}
36: \def\ergsec{\ifmmode {\rm\,erg\,s^{-1}}\else
37:                 ${\rm\,ergs\,s^{-1}}$\fi}
38: \def\kmsMpc{\ifmmode {\rm\,km\,s^{-1}\,Mpc^{-1}}\else
39:                 ${\rm\,km\,s^{-1}\,Mpc^{-1}}$\fi}
40: \def\nv{\ion{N}{5} $\lambda$1240}
41: \def\civ{\ion{C}{4} $\lambda$1549}
42: \def\oii{[\ion{O}{2}] $\lambda$3727}
43: \def\oiipair{[\ion{O}{2}] $\lambda \lambda$3726,3729}
44: \def\oiii{[\ion{O}{3}] $\lambda$5007}
45: \def\oiiipair{[\ion{O}{3}] $\lambda \lambda$4959,5007}
46: 
47: 
48: \title{Lyman-$\alpha$ Emitters at Redshift z=5.7}
49: 
50: \author{James E. Rhoads \altaffilmark{1,4} \and
51: Sangeeta Malhotra \altaffilmark{2,3,4}}
52: 
53: \begin{abstract}
54: \lya\ galaxies at high redshifts offer a powerful probe of both the
55: formation of galaxies and the reionization of the intergalactic
56: medium.  \lya\ line emission is an efficient tool for identifying
57: young galaxies at high redshift, because it is strong in systems with
58: young stars and little or no dust --- properties expected in galaxies
59: undergoing their first burst of star-formation.
60: \lya\ galaxies also provide a robust test of the reionization epoch that
61: is independent of Gunn-Peterson trough observations in quasar spectra
62: and is better able to distinguish line center optical depths $\tau
63: \sim 5$ from $\tau \sim 10^5$.  This is because neutral gas
64: scatters \lya\ photons, dramatically ``blurring'' images of \lya\ galaxies
65: embedded in a neutral intergalactic medium and rendering them undetectable.
66: 
67: We present a photometrically selected sample of redshift $z \approx
68: 5.7$ \lya\ emitters derived from the Large Area Lyman Alpha survey.
69: The presence of these low-luminosity \lya\ sources immediately implies
70: that the reionization redshift $\zre > 5.7$.  Comparing these objects
71: to our earlier $z\approx 4.5$ sample, we find that the number of $z
72: \approx 5.7$ emitters at fixed line luminosity marginally exceeds the
73: no-evolution expectation, but falls well short of published model
74: predictions.  The equivalent width distribution is similar at the two
75: redshifts.  The large equivalent widths of the \lya\ line indicate
76: young galaxies undergoing their first star formation.
77: \end{abstract}
78: 
79: \altaffiltext{1}{Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218} 
80: \altaffiltext{2}{Johns Hopkins University, Charles and 34th Street, Bloomberg center, Baltimore, MD 21218}
81: \altaffiltext{3}{Hubble Fellow} 
82: \altaffiltext{4}{Visiting Astronomer, Kitt Peak National Observatory,
83: National Optical Astronomy Observatory, which is operated by the
84: Association of Universities for Research in Astronomy, Inc. (AURA)
85: under cooperative agreement with the National Science Foundation.}
86: 
87: \keywords{galaxies: general --- galaxies: evolution --- galaxies: formation 
88: --- galaxies: statistics --- cosmology: observations --- early universe}
89: 
90: \section{Introduction}
91: 
92: %\input intro.tex
93: Recent discoveries of high redshift objects are steadily taking us
94: closer to determining the epoch of galaxy formation (Dey et al. 1998;
95: Spinrad et al. 1998; Weyman et al. 1998; Hu, McMahon, \& Cowie 1999;
96: van Breugel et al. 1999; Fan et al. 2001), if such a well-defined time
97: period exists. Observations of neutral hydrogen absorption troughs
98: (the Gunn-Peterson effect) seem to indicate that intergalactic
99: hydrogen was reionized and hence that the ionizing UV ``turned on''
100: between $z\approx 6.3$ and $z\approx 5.7$ (Becker et al. 2001,
101: Djgorvski et al. 2001).
102: %  The difference between reionization redshifts
103: %  for hydrogen ($\zre \sim 6$) and for HeII ($\zre \sim 3$) links the
104: %  hydrogen reionization to galaxy formation: If hydrogen were reionized
105: %  by quasars, which have harder spectra, HeII would be ionized to HeIII
106: %  at about the same time (e.g., Madau \& Meiksin 1994; Kriss et al 2001).
107: 
108: A more direct probe of galaxy formation is to determine the
109: number counts and redshift evolution of nascent galaxies.
110: The Large Area Lyman Alpha (LALA) survey (Rhoads et
111: al. 2000) has been designed and executed to detect large enough
112: samples of high-redshift emitters to be statistically useful. Strong
113: \lya\ emission, shown by large equivalent widths in the
114: line, indicates that these are young galaxies (Malhotra et
115: al. 2001). Since large scale structure can influence
116: galaxy counts even at high redshifts (Steidel et al. 1998, Adelberger
117: et al. 1998), large area surveys are vital to study redshift evolution.
118: Narrow-band surveys maximize sensitivity and solid angle,
119: and allow us to push for higher redshift objects by designing
120: narrow-bands to avoid bright sky lines.
121: 
122: \lya\ galaxies also offer a particularly direct and robust test of
123: the reionization epoch (Haiman \& Spaans 1999).  Before reionization,
124: \lya\ emitting galaxies of normal luminosity are effectively hidden
125: from view by the scattering ``fog'' of the neutral IGM
126: (Miralda-Escud\'{e} 1998; Rybicki \& Loeb 1999).  Thus, the
127: reionization redshift should be marked by a sharp decrease in the
128: number counts of faint \lya\ emitters.  This test is independent of
129: the Gunn-Peterson trough observations in quasar spectra.  Moreover,
130: the Gunn-Peterson test has difficulty distinguishing opaque but modest
131: optical depths $\tau \sim 5$ from the much larger obscuration $\tau
132: \sim 10^5$ expected in a neutral IGM.
133: \lya\ galaxy counts can resolve this ambiguity.  The red wing of the
134: \lya\ line from a star forming galaxy sees only the red damping wing
135: of intergalactic \lya\ absorption, and \lya\ galaxies therefore remain
136: detectable up to line-center optical depths $\tau \sim 10^4$.
137: 
138: In this paper we report on a population of candidate $z\approx 5.7$
139: \lya\ emitters discovered in the second phase of LALA survey and
140: compare their properties to the z=4.5 sample (Rhoads et al. 2001).
141: Section~\ref{obs} describes our observations and data reduction. In
142: section 3 we present the candidates and their properties. Section 4
143: explores the implications of these discoveries for the epoch of galaxy
144: formation.
145: 
146: 
147: \section{Observations and data analysis}  \label{obs}
148: %\input obs.tex
149: 
150:  The Large Area Lyman Alpha survey began in spring 1998 as a
151: narrowband search for $z \approx 4.5$ \lya\ emitting galaxies.  The
152: project was designed to exploit the high survey efficiency offered by
153: the 36' field of the new CCD Mosaic Camera at the Kitt Peak National
154: Observatory 4m Mayall Telescope.  The core of the LALA survey is $0.72
155: \,\Box^\circ$ in two fields centered at 14:25:57 +35:32 (2000.0) (the
156: Bo\"{o}tes field) and 02:05:20 -04:55 (2000.0) (the Cetus field).
157: 
158: In April 2000, we extended the Bo\"{o}tes field survey to $z=5.7$
159: using custom-built narrowband filters having central wavelengths
160: $\lambda_c = 815, 823\nm$, full width at half maximum (FWHM)
161: transmission $7.5\nm$, and peak throughput $\sim 90\%$.
162: The wavelengths were chosen to fall in a gap in the night sky
163: emission line spectrum, thus minimizing sky noise.
164: The filter bandpass deteriorated below these specifications outside a
165: central circle of $\sim 30'$ diameter.  This resulted in substantially
166: degraded throughput in the outer parts of our images, which we therefore
167: excluded from our analysis.
168: 
169: The $z=5.7$ observing run was four nights, beginning with the night of
170: 10--11 April 2000 (i.e., 2000 April 11 UT).  The weather was generally
171: clear.  Photometric stability was checked by measuring
172: the fluxes of selected stars on all science exposures.  This showed $<
173: 11\%$ variation in throughput for the entire run, mostly due to
174: airmass-dependent extinction.
175: % The variation is 6% for the 823 filter.
176: % it is 11% for 815.
177: 
178: We obtained 9.85 hours of data in 40 exposures through the
179: 815 nm filter, and 9 hours in 36 exposures through the 823 nm filter.
180: Exposures were spatially dithered with characteristic (RMS) offsets of
181: $\sim 50''$.  Seeing ranged from $0.75''$ to $2.25''$ (median
182: $0.93''$) in the 815 nm filter, and $0.78''$ to $1.25''$ (median
183: $1.01''$) in the 823 nm filter.  Median narrowband sky brightnesses
184: were $27 \microJy \, \arcsec^{-2}$ at 815 nm and
185: $21 \microJy \, \arcsec^{-2}$ at 823 nm. We also obtained a deep (4
186: hour) image in a Sloan Digital Sky Survey z' filter during this run.
187: 
188: Data reduction was done in IRAF following the methods used for
189: the $z=4.5$ sources (Rhoads et al 2000).
190: %First, crosstalk between
191: %chip pairs that share readout electronics was removed.  Next, overscan
192: %corrections, bias frames, and flatfield corrections were applied.
193: %Dome flats are sufficient to correct cosmetically for spatial
194: %variations in the filter throughput (see above).  A ghost image of the
195: %telescope pupil was then modeled and subtracted from each frame.
196: %Finally, second order flat fielding problems were removed using a
197: %heavily smoothed super-sky flat constructed from the object frames,
198: %and a polynomial surface was fitted to blank sky regions on each chip
199: %and subtracted in order to remove any remaining sky gradients.  Cosmic
200: %rays in individual exposures were identified and flagged using the
201: %algorithm of Rhoads (2000). Each image was then placed on a world
202: %coordinate system using USNO-A catalog stars in the field.  Satellite
203: %trails and a few internal reflections due to bright stars were flagged
204: %by hand for exclusion from the final stacks.  Finally, the images were
205: %combined using IRAF MSCRED tasks (Valdes \& Tody 1998; Valdes 1998) to
206: %correct geometric distortions in the images.  Weights for the final
207: %stack were determined using the ATTWEIGHT code of Fischer \& Kochanski
208: %(1994), which accounts for sky level, transparency, and seeing to
209: %optimize the signal to noise level for compact sources in the final
210: %image.
211: %
212: The final stacks have point spread functions (PSF) with directly
213: measured full widths at half maximum of $0.88''$ ($815\nm$) and
214: $0.97''$ ($823\nm$).  The PSF wings are higher than Gaussians of the
215: same FWHM, since the images are sums of many frames with varying
216: seeing. Catalogs were generated using SExtractor (Bertin
217: \& Arnouts 1996).  Fluxes were measured in $2.32''$ (9 pixel) diameter
218: apertures, and colors were obtained using matched $2.32''$ apertures
219: in registered images.  We also measured fluxes in earlier
220: broad band images from the NOAO Deep Wide-Field Survey (B$_W$, R, and
221: I bands) (Jannuzi \& Dey 1999) and broad and narrow band images from
222: previous LALA runs (V band and four $\lambda \approx 6600$\AA\ narrow
223: bands).
224: 
225: We select $z \approx 5.7$ candidates using the following criteria:
226: (1) Secure detection in a narrowband filter ($> 5 \sigma$).
227: (2) A strong narrowband excess (narrow $-$ broad color $< -0.75$ magnitude,
228: corresponding to a factor of $\approx 2$ in flux density)
229: that is securely measured (the flux density in the narrow band should
230: exceed that in the broad band at the $4 \sigma$ level).
231: (3) No flux at wavelengths shorter than the expected Lyman break
232: wavelength (the object should remain undetected in B$_W$ and V band fluxes
233: at the $2 \sigma$ level or below).
234: The first two criteria are designed to ensure that real emission 
235: line objects are selected, while the third rejects most low redshift
236: emission line galaxies, using the blue bands as ``veto'' filters.
237: % Magnitudes and colors were computed using ``asinh'' magnitudes
238: % (Lupton, Gunn, \& Szalay 1999), which account for upper limits more
239: % gracefully than do traditional (purely logarithmic) magnitude scales.
240: 
241: 
242: 
243: \section{Survey Results}  \label{results}
244: 
245: In total, 18 objects in the two filters (11 in the $815\nm$ filter and
246: 7 in the $823\nm$ filter) passed all selection criteria and are
247: considered good $z \approx 5.7$ candidates.  Photometric
248: ``mini-spectra'' for these candidates are shown in figure~\ref{minispec}.
249: 
250: \begin{figure}
251: \epsscale{0.99}
252: % \plotone{z5.7_fig_v3.ps}
253: \plotone{f1.eps}
254: \caption{
255: Photometric ``mini-spectra'' for the 18 $z \approx 5.7$ 
256: \lya\ emitter candidates selected using the criteria discussed in
257: section~\ref{obs}.
258: All photometry was done through a $2.32''$ diameter aperture
259: centered on the object position derived from the narrowband filter
260: containing the excess emission.  
261: The candidates are assigned ranked ID numbers based on 
262: the ratio of broad to narrow band
263: flux density, which is a measure of emission line strength.
264: This rank increases from left to right and from top to bottom.
265: The last five candidates (LALA5 1-14 to LALA5 1-18) have observer frame
266: equivalent widths below $80$\AA\ if we correct the continuum flux
267: for the intergalactic hydrogen absorption expected at $z\approx 5.7$.
268: \label{minispec}}
269: \end{figure}
270: 
271: Intergalactic hydrogen absorption is expected to attenuate the
272: measured I band flux used for candidate selection.  
273: The formulation of Madau (1995) gives
274: correction factors of $2$ for $z=5.70$ and $2.2$ for $z=5.77$.
275: We apply these corrections to the I band flux before calculating the
276: equivalent widths of the emission lines from our broad and narrow band
277: photometry.  These corrections are comparable to the factor of 2
278: difference in flux density required for selection.  We therefore apply
279: another cut to our candidate lists, removing the five candidates whose
280: observer frame equivalent widths fall below $80$\AA\ after the
281: correction for hydrogen absorption. 
282: We expect these objects to be either high redshift {\it
283: or\/} strong emission line objects, but they need not be both.
284: Because intergalactic hydrogen absorption can also remove the bluest
285: $\sim 50\%$ of the \lya\ line flux, our equivalent width corrections
286: and the resulting candidate list cut are both quite conservative.
287: 
288: 
289: 
290: \section{Discussion and Conclusions} \label{discuss}
291: %\input discuss.tex
292: 
293: We now turn to the implications of our survey for galaxy formation and
294: cosmology. First, let us compare the statistics of $z\approx 5.7$
295: sources to the $z\approx 4.5$ \lya\ emitters.  There are 13 sources at
296: $z\approx 5.7$ and 156 at $z\approx 4.5$.  However, a minority of our
297: $z\approx 4.5$ are bright enough to be detectable at $z\approx 5.7$.
298: 
299: Using a $\Omega_m = 0.3$, $\Omega_\Lambda = 0.7$ cosmology and accounting
300: for the difference in solid angle between the two
301: redshifts\footnote{The solid angle for the higher redshift is approximately
302: a factor of two smaller due to the degradation of the bandpass near the
303: filter edges; see section~\ref{obs}.}, the comoving survey volume at
304: $z\approx 5.7$ is $31\%$ that at $z\approx 4.5$ ($16\%$ in each filter).
305: The $5\sigma$ flux thresholds used to identify candidates at $z\approx 5.7$
306: were $0.455 \microJy$ at $815\nm$ and $0.405 \microJy$ at $823\nm$.  The 
307: distance-corrected equivalent thresholds at $z\approx 4.5$
308: will be larger by a factor of $1.34^2 = 1.8$ for the above cosmology. 
309: There are then $15$ $z \approx 4.5$
310: candidates that could be detected in our $815\nm$ data, and $26$
311: that could be detected in our $823\nm$ data.  Thus, based on the
312: $z\approx 4.5$ survey and a no-evolution assumption, we would expect
313: to find $2.3$ sources in the $815\nm$ filter, and $4.1$ in the
314: $823\nm$ filter.  We actually identify $7$ and $6$ respectively.
315: 
316: This excess in the observed $z\approx 5.7$ candidate counts is
317: moderately significant (at the 99\% level assuming Poisson source
318: counts).  The important difference is in the $815\nm$ counts;
319: the $823\nm$ filter candidate counts are consistent with the
320: na\"{\i}ve scaling from $z\approx 4.5$.  We now examine several
321: possible explanations for this effect.
322: 
323: 
324: % Evolution
325: First, \lya\ emitters at $z\approx 5.7$ may be more numerous and/or more
326: luminous than those at $z\approx 4.5$.  Such effects are reasonable:
327: The \lya\ line is resonantly scattered and can be strongly quenched by
328: dust absorption.  Since dust is formed from stellar nucleosynthesis
329: products, we expect galaxies at earlier epochs to be (on average) less
330: chemically evolved and less dusty.  More luminous \lya\ lines may be a
331: natural byproduct, provided that the effects of chemical evolution are
332: not overwhelmed by the trend of decreasing halo mass at higher
333: redshift. Indeed, under the \lya\ emitter population models of Haiman
334: \& Spaans (1999), evolutionary effects are so strong that the
335: total surface density per unit redshift above a fixed line flux $\sim
336: 1.5 \times 10^{-17} \ergcm2s$ is approximately independent of redshift
337: for $3 \la z \la 6$.  Our results show a factor $\sim 7$ decrease in
338: the surface density of \lya\ emitter candidates from $z=4.5$ to
339: $z=5.7$ (in contrast to an expected factor $< 2$ from the Haiman \&
340: Spaans models).  The LALA data is thus more nearly consistent with a
341: no-evolution model than with the baseline model from Haiman \& Spaans
342: (1999).
343: 
344: % Large scale structure...
345: In addition, large scale structure can significantly affect the number
346: counts of objects in single fields.  Our $36'$ field makes LALA results
347: comparatively robust to such effects, but
348: a factor of $2$ difference in source density may still be consistent
349: with cosmic variance.
350: 
351: Finally, our candidate sample will contain some residual contamination
352: by foreground objects.  The numbers of such objects should be
353: comparatively small, given the high-redshift consistency checks
354: described in sections~\ref{obs} and~\ref{results}. 
355: Foreground emission line galaxies are the most serious concern.
356: Our primary defense against such galaxies is the
357: requirement that a good candidate be undetected in deep broadband
358: images with $\lambda < (1+z) \times 912$\AA.  This imposes an
359: approximate minimum observer frame equivalent width for any
360: interloper, which we estimate using the minimum narrow band flux $5
361: \sigma_n$ and maximum broad band flux $2 \sigma_b$.  (Here $\sigma_n$
362: and $\sigma_b$ are the photometric errors in flux density units for
363: the narrow and broad blue filters.)  The corresponding equivalent
364: width is then $\approx (f_n / f_b - 1) \times \Delta\lambda_n = \left[
365: 5 \sigma_n / (2 \sigma_b) - 1 \right] \times \Delta\lambda_n \sim
366: 1100$\AA, where we assume that the line lies outside the broad band
367: filter.  The redshifts of likely foreground objects are relatively low
368: ($0.24 \la z \la 1.21$), and the corresponding emission frame
369: equivalent widths therefore remain large ($\ga 500$\AA).  While
370: compact, narrow emission line galaxies can attain these equivalent
371: widths, they are very rare (e.g., Hogg et al 1998).
372: 
373: Cool stars can also enter narrowband emission line samples because
374: molecular absorption in their atmospheres produces spectral
375: features with widths of a few $\nm$ and amplitudes several tens of
376: percent.  However, in the absence of true emission lines, cool stars
377: will show detectable continuum at $\ga 1/2$ of the
378: narrowband flux density, which is also expected to be quite red, so
379: that our SDSS z' filter should clearly show such stars.  Most of our
380: candidates having strong z' detections were discarded from our sample
381: by the last cut applied (see section~\ref{results}).  Only two of the
382: 13 remaining sources, LALA5 1-08 and LALA5 1-10, have ``minispectra''
383: that plausibly resemble cool stars.
384: 
385: Because our narrow and broad band filters were obtained at different
386: times, variable objects may also enter the sample at a low level.  The
387: relevant objects would have to have $I \la 24.3$ during the
388: narrow band observations (2000 April 11--14 UT), and $I>25$ during
389: the I band observation (1999 March 27 UT).  (In
390: fact, our requirement on the blue flux imposes additional limits.)
391: The largest known source of such variable objects would be high
392: redshift supernovae.  However, these would show roughly
393: equal brightness in both $z\approx 5.7$ narrow bands and the SDSS z'
394: image, since their characteristic variability timescale is several
395: days.  Again, only LALA5 1-08 and LALA5 1-10 are
396: consistent with this expectation.  Sources with faster variability,
397: such as orphan GRB afterglows (Rhoads 1997), are very rare and we
398: would expect $\ll 1$ such source in our sample (e.g., Dalal et al 2001).
399: 
400: We compared the equivalent width distribution of the $z=4.5$ and $5.7$
401: samples. Very low and high equivalent width objects are missing in
402: z=5.7 sample compared to z=4.5 sample. 
403: Using a Kolmogorov-Smirnov test we find
404: a 20\% chance that the two samples are
405: drawn from the same parent distribution. This test is only
406: marginally conclusive, mostly because of the small number of $z=5.7$
407: objects.  The consistency of the present samples is nonetheless
408: further evidence that we are seeing the same class of object at
409: $z=5.7$ as at $z=4.5$.
410: 
411: %\begin{figure}
412: %% \plotone{ks.ps}
413: %\plotone{f2.eps}
414: %\caption{
415: %Cumulative equivalent width histograms for $z=4.5$ (solid line) and
416: %$z=5.7$ (dotted line) candidate \lya\ emitters from the LALA survey.
417: %The last bin includes all sources for which the data yield lower
418: %limits to the equivalent width.  A Kolmogorov-Smirnov test shows
419: %that the two distributions are consistent (see text).
420: %\label{ks_fig}}
421: %\end{figure}
422: 
423: In summary, we have a sample of $13$ $z\approx 5.7$ \lya\
424: emitting galaxy candidates.  Only two of these have the
425: apparent spectral energy distributions expected of stars or variable
426: objects.  The upper limits on blue flux rule out all but the highest
427: equivalent width foreground sources, which are too rare to account
428: for the entire sample.  The equivalent width distributions are 
429: consistent at $z\approx 5.7$ and $4.5$.  Thus, the simplest
430: explanation of the data is that we are really seeing $z\approx 5.7$
431: \lya\ emitters.
432: 
433: % Expectations under assumption that reionization occurred at z=5.x
434: We now draw one very robust and important conclusion from our sample.
435: The epoch of reionization is a major landmark in the evolution of
436: the universe.  Recently, the Gunn-Peterson trough has been reported
437: in spectra of two high redshift quasars: SDSSpJ 103027.10+052455.0
438: at $z = 6.28$ (Becker et al 2001), and SDSS 1044-0125 at $z = 5.73$
439: (Djorgovski et al 2001).  These authors have interpreted their spectra
440: as evidence that reionization was incomplete at these redshifts.
441: 
442: If the universe was neutral at redshift $\zre < 5.67$, \lya\ photons from
443: sources at higher redshift would be resonantly scattered in the
444: neutral intergalactic medium (IGM).  This effectively erases the \lya\
445: line from view for a survey like LALA. Thus, our $z\approx 5.7$ sample
446: implies that the reionization redshift lies at $\zre>5.8$ along the line
447: of sight to the LALA Bo\"{o}tes field.  This argument relies merely on
448: the {\it presence\/} of low-luminosity \lya\ sources, and is
449: independent of their exact number.
450: 
451: Two possible loopholes exist, but both
452: are readily closed.  First, in the absence of dust, the \lya\ photons
453: eventually scatter far enough into the line wing to propagate freely.
454: However, this process spreads the line emission greatly in both area
455: and frequency, resulting in a very large ``photosphere'' and a
456: correspondingly low surface brightness (Loeb \& Rybicki 1999).  All of
457: the sources reported in this paper are compact in the narrow band
458: image (i.e., essentially unresolved at our resolution limit of $\la
459: 1''$) and so morphologically inconsistent with a source embedded in a
460: neutral IGM.
461: 
462: Second, a source with a sufficient ionizing flux will generate
463: its own Stromgren sphere in the IGM.  If this sphere is large enough,
464: \lya\ photons will redshift by $\ga 1000 \kms$ before reaching the
465: surrounding neutral gas, and will thus avoid resonant scattering.
466: However, this effect requires sources either more luminous or older than
467: typical LALA objects.  We assume that a fraction $\fesc$ of ionizing
468: photons escape a galaxy to ionize its Stromgren sphere, and that $2/3$
469: of the remainder result in \lya\ photons, and that $H_0 = 70 \kmsMpc$,
470: $\Omega_m=0.3$, $\Omega_\Lambda = 0.7$, and $\Omega_b =0.05$.  We then
471: require an ionized bubble radius $r_s \ga 1.2 \Mpc$, which gives a
472: Gunn-Peterson effect optical depth $\tau \le 1$ at emitted line center
473: (see also Loeb \& Rybicki 1999).  Ignoring recombinations, $r_s \ga
474: 1.2 \Mpc$ requires a source with $L_{43} t_8 \fesc/(1-\fesc) \ga 5$,
475: where $L_{Ly-\alpha} = 10^{43} L_{43} \ergsec$ and the source is $10^8
476: t_8$ years old.  LALA $z\approx 5.7$ sources have $L_{43} \approx
477: 0.7$.  Their median rest frame equivalent width of $\ge 80$\AA\ 
478: requires $\fesc \la 0.2$ for any reasonable initial mass
479: function\footnote{Continuous star formation with IMF slope
480: $\alpha=2.35$ (i.e., Salpeter), upper mass cutoff $120 M_\odot$,
481: metallicity $Z = Z_\odot/20$, and $t_8 \ga 0.3$ gives $EW \approx 95
482: (1-\fesc)$\AA.  Substantially larger EW requires $\alpha \ll 2$.}, unless
483: $t_8 \la 0.3$ (Malhotra et al 2001; Kudritzki et al 1999).  The age of
484: the universe at $z=5.7$ requires $t_8 < 10$.  Thus, $L_{43} t_8
485: \fesc/(1-\fesc) \la 2$, and \lya\ radiation from these sources would
486: indeed suffer resonant scattering if the IGM were neutral.
487: 
488: We are thus confident that the compact, low-luminosity \lya\ emitters
489: we observe at $z \approx 5.7$ are in a mostly ionized universe.
490: This method offers a test for reionization 
491: independent of the spectroscopic search for Gunn-Peterson troughs.
492: Moreover, it is better able to distinguish line center optical depths
493: $\tau \sim 10^5$ (indicating a neutral IGM) from much smaller but still
494: essentially opaque optical depths $\tau \sim 10$ (indicating \lya\
495: forest absorption), because the \lya\ forest will in general absorb
496: only the blue side of the \lya\ emission line, while a neutral IGM
497: will absorb the entire line.
498: For an intrinsic line width of $\sim 100 \Delta v_{100} \kms$, we 
499: would still expect to see some \lya\ flux through an IGM with a
500: (homogeneously mixed) neutral fraction of $\la 0.1 \Delta v_{100}
501: \left((1+z)/6.7\right)^{-3/2}$, while the Gunn-Peterson trough is
502: optically thick ($\tau \ge 5$) for a neutral fraction as small as
503: $\sim 4 \times 10^{-5} \left((1+z)/6.7\right)^{-3/2}$.
504: 
505: To conclude, we have demonstrated that \lya\ emitters are found at
506: $z\approx 5.7$ in densities comparable to those at $z\approx 4.5$ at
507: similar line luminosities.  It follows that the reionization redshift
508: was $\zre > 5.8$, at least along this line of sight.  Tentatively, it
509: appears that the \lya\ source density at equal luminosity rises
510: from $z\approx 4.5$ to $z\approx 5.7$. We have also obtained pilot
511: observations in a $z\approx 6.6$ narrowband filter.  This redshift
512: lies beyond the recently reported Gunn-Peterson effect quasars, and
513: will ultimately allow us to bracket the reionization epoch through
514: \lya\ source counts.
515: 
516: \acknowledgements 
517: This work made use of images provided by the NOAO Deep Wide-Field
518: Survey (NDWFS; Jannuzi and Dey 1999), which is supported by the National
519: Optical Astronomy Observatory (NOAO).  NOAO is operated by AURA,
520: Inc., under a cooperative agreement with the National Science
521: Foundation.
522: %
523: We thank Buell Jannuzi, Arjun Dey, and the rest of the NDWFS team for
524: making their images public; and Richard Green and Jim De Veny for
525: their support of the LALA survey.
526: %
527: JER's research is supported by an Institute Fellowship at The Space
528: Telescope Science Institute (STScI).
529: %
530: SM's research funding is provided by NASA through Hubble Fellowship
531: grant \# HF-01111.01-98A from STScI.
532: STScI is operated by the Association of Universities for Research in
533: Astronomy, Inc., under NASA contract NAS5-26555.
534: 
535: %\input refs.tex
536: \begin{references}
537: \reference{} Adelberger, K. L., Steidel, C. C., Giavalisco, M., 
538:   Dickinson, M., Pettini, M., \& Kellogg, M. 1992, \apj\ 506, 18
539:   % counts-in-cells clustering analysis of Lyman break galaxies 
540: \reference{} Becker, R. H., et al. (SDSS consortium), 2001, astro-ph/0108097
541:  % reionization paper
542: \reference{} Bertin, E. and Arnouts, S., 1996, \aaps\ 117,393
543:         % SExtractor
544: \reference{} Dalal, N., Griest, K., \& Pruet, J. 2001, astro-ph/0103436
545:   % orphan afterglows are faint and/or rare
546: \reference{} Dey, A., Spinrad, H., Stern, D., Graham, J. R.,  Chaffee, F. H.,
547:   1998, ApJ, 498, L93 % z=5.34 source
548: \reference{} Djorgovski, S. G., Castro, S. M., Stern, D., \&
549:   Mahabal, A. 2001, astro-ph/0108069 % reionization paper
550: \reference{} Fan, X., et al. (SDSS consortium), 2001, astro-ph/0108063
551:  % big-z quasar.
552: %\reference{} Fischer, P., \& Kochanski, G. P. 1994, \aj\ 107, 802
553: %        % optimal coaddition...
554: \reference{} Haiman, Z., \& Spaans, M. 1999, \apj\ 518, 138
555:         % Models for Dusty Lyalpha Emitters at High Redshift
556: \reference{} Hogg, D.W., Cohen, J.G., Blandford, R. \& Pahre, M.A. 1998, ApJ,
557:   504, 622
558:   % OII emitter equivalent width distribution
559: \reference{} Hu, E. M., McMahon, R. G., Cowie, L. L., 1999, \apj\ 522, L9 
560:   % a z>5 Lyman alpha emitter
561: \reference{} Jannuzi, B. T. \& Dey, A., 1999, in ``Photometric Redshifts and
562:   the Detection of High Redshift Galaxies,'' ASP Conference Series
563:   Vol. 191, Edited by R. Weymann, L. Storrie-Lombardi, M. Sawicki
564:   and R. Brunner. ISBN: 158381-017-X, p. 111                              
565: %\reference{} Kriss, G. A., et al 2001, astro-ph/0108169, 
566: %  to appear in {\it Science}
567:   % He II Gunn-Peterson effect with FUSE
568: \reference{} Kudritzki, R.-P., Mendez, R. H., Feldmeier, J. J.,
569:         Ciardullo, R., Jacoby, G. H., Freeman, K. C., Arnaboldi, M.,
570:         Capaccioli, M., Gerhard, O., \& Ford, H. C. 2000, \apj\ 536, 19
571:         %  z=3.1 emitters from Virgo PNe surveys
572: \reference{} Loeb, A. \& Rybicki, G. B. 1999, \apj\ 524, L527
573:   % radiative xfer of LyA line before recombination.
574: % \reference{} Lupton, R. H., Gunn, J. E., \& Szalay, A. S. 1999, \aj\
575: %         118, 406.   % asinh mags
576: \reference{} Madau, P. 1995, \apj\ 441, 18
577:    % Radiative transfer in a clumpy universe
578: \reference{} Malhotra, S., et al 2001, 
579:   % Rhoads, J. E., Stern, D., Jannuzi, B., Dey, A., \& 
580:   % Spinrad, H. 2001, in some permutation, 
581:   to be submitted to \apj\ Letters.
582:   % The Large Equivalent Width paper
583: \reference{} Miralda-Escud\'{e}, J. 1998, \apj\ 501, 15
584:   % the red wing of the LyA line and IGM absorption.
585: \reference{} Rhoads, J. E. 1997, \apjl\ 487, L1  % Jet/balloon paper
586: \reference{} Rhoads, J. E., Malhotra, S., Dey, A., Stern, D., Spinrad,
587:   H., \& Jannuzi, B. T. 2000, \apjl\ 545, L85  % LALA1 letter
588: %\reference{} Rhoads, J. E. 2000, \pasp\ 112, 703
589:         % CR rejection paper
590: \reference{} Spinrad, H., Stern, D., Bunker, A., J., Dey, A., 
591:   Lanzetta, K., Yahil, A., Pascarelle, S., \& Fern\'{a}ndez-Soto, A.
592:   1998, AJ, 116, 2617 % z>5 galaxy
593: \reference{} Steidel, C. C., Adelberger, K. L., Dickinson., M., Giavalisco,
594:   M., Pettini, M., \& Kellogg, M. 1998, ApJ 492, 428
595:   % Large scale structure at z=3.1
596: %\reference{} Valdes, F., \& Tody, G. 1998, Proceedings
597: %        of SPIE Vol. 3355  % MSCRed + Mosaic data handling
598: %\reference{} Valdes, F. 1998, ADASS VII, ASP Conf. Series 145,
599: %        ed. R. Albrect  % Frank's latest on MSCRed
600: \reference{} van Breugel, W., De Breuck, C., Stanford, S. A., Stern, D., 
601:     Rottgering, H., \& Miley, G. 1999, \apj\ 518, L61
602:     % z = 5.19 radio galaxy
603: \reference{} Weymann, R. J., Stern, D., Bunker, A., Spinrad, H., 
604:     Chaffee, F. H., Thompson, R. I., \& Storrie-Lombardi, L. J. 1998,
605:     \apj\ 505, L95
606:   % a z = 5.60 galaxy
607: \end{references}
608: 
609: \end{document}
610: 
611: 
612: