A Galactic Wind at z = 5.190

12 Pages·0.52 MB·

by  Steve Dawson

Acceptedforpublicationin theAstrophysical Journal
PreprinttypesetusingLATEXstyleemulateapjv.14/09/00
A GALACTIC WIND AT Z =5.1901
Steve Dawson2, Hyron Spinrad2, Daniel Stern3, Arjun Dey4, Wil van Breugel5, Wim de
Vries5, and Michiel Reuland5
Accepted for publication inthe Astrophysical Journal
ABSTRACT
We reportthe serendipitous detectionin high–resolutionopticalspectroscopyof a strong,asymmetric
Lyα emission line at z = 5.190. The detection was made in a 2.25 hour exposure with the Echelle
Spectrograph and Imager on the Keck II telescope through a spectroscopic slit of dimensions 1′′ × 20′′.
2 The progenitor of the emission line, J123649.2+621539 (hereafter ES1), lies in the Hubble Deep Field
0 North West Flanking Field where it appears faint and compact, subtending just 0′.′3 (FWHM) with
0 I =25.4. TheES1Lyαlinefluxof3.0×10−17ergscm−2 s−1 correspondstoaluminosityof9.0×1042
2 AB
ergs s−1, and the line profile shows the sharp blue cut–off and broad red wing commonly observed in

n star–forming systems and expected for radiative transfer in an expanding envelope. We find that the
a Lyα profile is consistent with a galaxy–scale outflow with a velocity of v > 300 km s−1. This value is
J consistent with wind speeds observed in powerful local starbursts (typically 102 to 103 km s−1), and

4 compares favorably to simulations of the late–stage evolution of Lyα emission in star–forming systems.
1
We discuss the implications of this high–redshift galactic wind for the early history of the evolution of

galaxies and the intergalactic medium, and for the origin of the UV backgroundat z >3.
1
v Subject headings: cosmology: observations — early universe — galaxies: high–redshift — galaxies:
4 individual (ES1) — galaxies: starburst
9
1
1. introduction et al. 2000, and references therein).
1
0 Followingthe epochofrecombination,the Universeset- We present a direct observation of such an out-
2 flow at z = 5.190 in high–resolution optical spec-
tled into the comparatively dormant dark ages, during
0 troscopy of the serendipitously detected star–forming
which the primordialglow had begun to fade but the first
/ galaxy J123649.2+621539 (hereafter ES1, for Echelle
h present–day astronomical objects had yet to form. This
Spectrograph and Imager serendipitous detection number
p tranquil epoch proved short–lived,however,as the forma-
- tion of the first stars and quasars ushered in the first era one). The sharp blue cut–off and broad red wing of the
o ES1 Lyα emission line are consistent with the profile ex-
of cosmological heating and enrichment at z < 20 (e.g.
r pectedforthetransferoflineradiationinanexpandingen-
t Gnedin & Ostriker 1997; Haiman & Loeb 1997,1998; Os-
s velope (e.g. Surdej 1979). The suggested outflow velocity
a triker & Gnedin 1996; Valageas & Silk 1999). Evidence
of v >300km s−1 is in broadagreementwith simulations
: of these processes in the form of galaxy–scale outflows is
v of the late evolution of Lyα emission and absorption in
abundant in spectroscopy of the high–redshift Universe.
i star–forminggalaxies(e.g.Tenorio–Tagleetal.1999),and
X Both optical/IR spectra of the z ∼ 3 Lyman–break pop-
is consistent with observations of powerful nearby star-
r ulation (e.g. Pettini et al. 2001) and optical spectra of
a lensed Lyα–emitting galaxies at z >4 (Frye, Broadhurst, bursts (e.g. Heckman, Armus, & Miley 1990). The spec-
trum of ES1 therefore presents evidence for both a high
& Benitez 2001) show metal absorption lines which are
star–formation rate and a high–redshift, starburst–driven
blueshiftedbyhundredsofkms−1 withrespecttothestel-
galactic wind, fitting well within the expectations of cur-
lar rest frame of the galaxy,and Lyα emissionlines which
rent models for the early history of galaxy formation.
are shifted similarly redward. These observations, as well
In §2 we discuss the detection of ES1 and we give a de-
as the characteristic P–Cygni profile of the Lyα emission
scription of the spectrum and the available archivalimag-
lines (e.g. Bunker, Moustakas, & Davis 2000; Dey et al.
ing. In §3 we detail the properties of the ES1 Lyα emis-
1997, 1998; Dickinson 1998; Ellis et al. 2001, Lowenthal
sion line, and we present a model for the emission line
et al. 1997; Weymann et al. 1998) paint a coherent pic-
profile consistent with the expanding shell scenario intro-
tureofopticallythickexpandingregionssurroundingstar–
duced above. We conclude in §4 with a discussion of the
forming galaxies, most naturally driven by the starbursts
implications of the evidence for a strong outflow in ES1
that render them visible in the first place (e.g. Heckman
1 Based onobservations made at the W.M.Keck Observatory, which isoperated as ascientific partnership amongthe CaliforniaInstitute of
Technology, theUniversityofCaliforniaandtheNationalAeronauticsandSpaceAdministration. TheObservatorywasmadepossiblebythe
generous financialsupportoftheW.M.KeckFoundation.
2 Astronomy Department, University of California at Berkeley, Mail Code 3411, Berkeley, CA 94720 USA, email: (sdawson, spin-
rad)@astro.berkeley.edu
3 Jet Propulsion Laboratory, California Institute of Technology, Mail Stop 169–327, Pasadena, CA 91109 USA, email:
[email protected]
4 KPNO/NOAO,850N.CherryAve.,P.O.Box26732,Tucson,AZ85726USA,email: [email protected]
5 Institute ofGeophysics andPlanetaryPhysics,LawrenceLivermoreNational Laboratory,L–413, P.O.Box808, Livermore,CA94550 USA,
email: (wil,vries,mreuland)@igpp.ucllnl.org
1
2 A Galactic Wind at z =5.190
for both the evolution of galaxies and the intergalactic larabsorptionlines. Thenon–detectionofhigh–ionization
medium (IGM) at high redshift, and for the origin of the stateemissionlinestypicallyobservedinAGNspectra,e.g.
UV background at z > 3. Throughout this paper we NV λ1240, CIV λ1549, or HeII λ1640, strongly suggests
adopt the currently favored Λ–cosmology of Ω = 0.35 thatthe LyαemissioninES1is due to the Lymancontin-
M
and Ω = 0.65, with H = 65 km s−1 Mpc−1 (e.g. Riess uum flux of OB stars, rather than the hard UV spectrum
Λ 0
et al. 2001). At z = 5.190, such a universe is only 1.10 ofanAGN.Forz =5.190,thenebularemissionlinestypi-
Gyr old — corresponding to a look–back time of 92.1% callyassociatedwithstar–forminggalaxiesareinaccessible
of the age of the Universe — and an angular size of 1′.′0 to optical spectroscopy.
corresponds to 6.31 kpc. WewerefortunatethatES1islocatedintheHDFNorth
West Flanking Field. Figure 2 displays a portion of the
2. observation and data reduction single–orbit Hubble Space Telescope (hereafter HST) I814
imageofthatregionwithaprojectionofthespectroscopic
ES1 was detected in a 2.25 hour exposure made with
slit. ES1 is faint and compact, typical for a very distant
the Echelle Spectrograph and Imager (ESI; Sheinis et al.
galaxy(Steideletal.1996). Byrunningthesourceextrac-
2000) at the Cassegrain focus of the Keck II telescope on
tionalgorithmSExtractor(Bertin&Arnouts1996)onthe
UT 2001 February 25. The instrument was configured in
Flanking Field image and employing the conversion from
itsmedium–resolutionechelletemodewithaspectroscopic
data number to AB magnitude given in Williams et al.
slit of dimensions 1′′ × 20′′, yielding a spectral resolution
(1996), we determine an isophotal magnitude for ES1 of
of ∼ 2 ˚A (78 km s−1) at 7500 ˚A. The 2.25 hour expo-
I = 25.4±0.2. This is challengingly faint, comparable
AB
surewasbrokenintofourintegrationsof1800secondsand
to the first detected galaxies at z > 5, e.g. HDF 3–951
one integration of 900 seconds; we performed 3′′ spatial
with I = 25.6 (Spinrad et al. 1998). Moreover, as an
814
offsets between each integration to facilitate the removal
appreciablefractionoftheI–bandradiationforthesource
of fringing at long wavelengths. We used IRAF6 (Tody
is from the Lyα emission line, the true continuum magni-
1993) to process the echellogram, following standard slit
tude must be still fainter than that determined from the
spectroscopy procedures (e.g. Massey, Valdes, & Barnes
source extraction. ES1 appears marginally resolved, sub-
1993)7. Some aspects of treating the ten individual or- tending 0′.′3 (FWHM) on the Flanking Field image, while
dersofthespectrumwerefacilitatedbythesoftwarepack- stars on the same image have a FWHM near 0′.′2. Thus,
age BOGUS8, createdby D. Stern, A.J. Bunker, andS.A.
the physicalsize ofthe emitting regionappearsto be only
Stanford. Wavelength calibrations were performed in the
∼4 kpc in diameter.
standard fashion using Xe, HgNe, and CuAr arc lamps;
We present ground–based V, R, I, and z images of
we employed telluric sky lines to adjust the wavelength
ES1 in Figure 3 and we give the ground–based photom-
zero–point. The nightwas nearphotometric with 0′.′6 see-
etryinTable2. TheV andI imagesarefromthe Canada
ing, and we performed flux calibrations with observations
FranceHawaiiTelescopeimagingcampaignofBargeretal.
of standard stars from Massey & Gronwall (1990) taken
(1999); the R and z images are from our own Keck imag-
with the instrument in the same configuration as the tar-
ingcampaignoftheHDFanditsenvirons(seeSternetal.
get observation. A portion of one order of the discovery
2000). The fact that ES1 is not detectable in the V or
spectrum, centered on the ES1 emission line, is shown in
R bands is characteristic of high–redshift galaxies, where
Figure 1.
interveningneutralhydrogen(the Lymanforests)severely
The target object for this observation was a Lyman–
attenuates the continuum signalbluewardofLyα (Madau
break galaxy at z = 3.125 (D16; Steidel 2001, private
1995;Steidel et al.1996;Stern & Spinrad1999). The fact
communication), located in the Hubble Deep Field North
that ES1 is not detectable in the z band is due only to
WestFlankingField(Williamsetal.1996). ES1wasfortu-
the comparative shallowness of the z band image. Based
itouslyplacedonthespectroscopicslitroughly4′′ southof
on the model spectrum described in section § 3.2 and as-
D16, along the parallactic angle of 150◦ (Figure 2). Upon
suming that ES1 has a flatspectrum in f atwavelengths
ν
making the observation,we immediately noticeda strong,
longerthantheemissionline,weexpectaVega–basedcon-
serendipitously detected emission line near 7527 ˚A whose tinuum magnitude in the z band of z>27.0. This value is
∼
asymmetric profile suggestedredshifted Lyα atz =5.190.
almost five times as faint as the 3σ limiting magnitude in
As the spectrograph configuration for the discovery spec-
that image (see Table 2).
trum coveredonly 20 arcsec2 — suggesting a surface den-
sityofhighredshiftLyα–emittersroughly30to90timesin 3. properties of the Lyα emission line
excessofcurrentestimates(e.g.Cowie&Hu1998;Dawson
3.1. The Emission Line Luminosity & Equivalent Width
et al. 2001; Stern & Spinrad 1999; Thompson, Weymann,
& Storrie–Lombardi2001)—we consider the detectionof InthefashionablecosmologyΩM=0.35,ΩΛ =0.65,the
ES1 to be highly providential. ES1Lyα flux of3.0×10−17 ergscm−2 s−1 correspondsto
Careful inspection of the reduced spectrum failed to a line luminosity of 9.0×1042 ergs s−1. For the prescrip-
reveal emission lines at other wavelengths. Though the tiongiveninDeyetal.(1998)assumingnodustaborption
galaxy continuum is faintly detected redwardof the emis- andnegligibleextinction, this luminositycorrespondsto a
sionline,the observationdidnotachievesufficientsignal– star formation rate (SFR) of ∼ 10 M yr−1. As is evi-
⊙
to–noise to confirm the presence or absence of interstel- dent in Figure 4, these values are typical of Lyα–emitting
6 IRAFisdistributedbytheNationalOpticalAstronomyObservatories,whichareoperatedbytheAssociationofUniversitiesforResearchin
Astronomy,Inc.,undercooperativeagreementwiththeNationalScienceFoundation.
7 A User’sGuide to Reducing Slit Spectra with IRAF,availableonlineathttp://iraf.noao.edu/iraf/web/docs/spectra.html
8 BOGUSisavailableonlineathttp://zwolfkinder.jpl.nasa.gov/∼stern/homepage/bogus.html.
Dawson et al. 3
galaxiesathighredshift(z >3)discoveredserendipitously onthe emissionline profile. The net effect is to create the
(e.g. Dawson et al. 2001; Dey et al. 1998; Manning et al. P–Cygni profile ubiquitous in observations of expanding
2000; Spinrad et al. 1999; Stern & Spinrad 1999) or in shells.
narrow band surveys (e.g. Hu et al. 1999; Rhoads et al. The comparativelyhighsignal–to–noiseofthe ES1Lyα
2000). Of course, such galaxies may represent rare sys- detection created the opportunity for probing this emis-
tems at the high–luminosity tip of an unexplored under- sion line structure with a simple model. In accordance
lying population, and the observed Lyα luminosities may withthescenariodescribedabove,wefittheLyαemission
accordingly be governed by a selection effect. This point feature with three components: (1) a comparatively large
is borne out by the faint system discovered in the blind amplitude, narrowGaussianintendedto model line radia-
spectroscopic survey of well–constrained lensing clusters tion generated by recombination in the ionized hydrogen;
(Ellisetal.2001),whichisatahigherredshift(z =5.576) (2) a small amplitude, broad Gaussian intended to model
and is an order of magnitude less luminous in Lyα than theredwingoflinephotonsback–scatteredoffthefarside
ES1. of the expanding shell; and (3) a Voigt absorption profile
Weestimatearest–frameequivalentwidthfortheemis- intended to model the blue decrement caused by the ab-
sion line of W = 120±40 ˚A based on the emission sorption of line photons by the near side of the shell. We
λ,rest
line model described in the following section. This re- modeled the weak continuum with a constant (f ∝ λ0)
λ
sult should be treated with a degree of circumspection, baseline.
however. The foremost caveat is that both the total line To accountfor the continuum decrementcausedby line
flux and the continuum level which enter the calculation blanketing in the Lyα forest, we attenuated the model
wereobtainedfromthemodelspectrumbeforeattenuation spectrum by a transmission profile adopted from Madau
by neutral hydrogen absorption. In this manner we at- (1995),wheretheopticaldepthduetothecombinedeffect
temptedtocircumventtheambiguityinderivinganequiv- of many Lyα absorption lines is given as
alent width from a strongly P–Cygni line profile coupled 3.46
λ
with a pronounced continuum break, essentially arriving τ =0.0036 , (1)
(cid:18)λ (cid:19)
at a theoretical equivalent width from the spectrum as α
it would appear if not truncated by foreground absorp- with λα =1216˚A. At high redshifts (z >4.5), absorption
tion. As a second caveat, the continuum redward of the by metal systems makes a non–negligible contribution to
emission line is not well–detected, with a significance of cosmic opacity. Hence, we accounted for the combined ef-
far less than 1σ. Together, we expect these two effects fect of metal lines with the additional factor also given in
to cause an over-estimation of W . However, this ten- Madau (1995),
λ,rest
dency toward over–estimation may be offset by the fact 1.68
λ
that this is a serendipitous detection, so ES1 was not well τ =0.0017 . (2)
(cid:18)λ (cid:19)
centered in the spectroscopic slit (see Figure 2). As such, α
we have been conservative in our estimate of the uncer- Neither Lyman series line blanketing nor continuum ab-
taintyinW ,whichincludesskynoise,the uncertainty sorption from neutral hydrogen were included, as both
λ,rest
in the continuum level redwardof the emission line in the these effects fall shortward of the wavelength range of in-
extracted spectrum, and the fit errors in the model spec- terest.
trum. Even so, with W = 120±40 ˚A, ES1 figures in Figure 5 shows the minimum–χ2 model resulting when
λ,rest
the top 2% of Lyman–break galaxies with Lyα in emis- thecentroids,amplitudes,andwidthsofeachofthemodel
sion in the near–complete continuum–selected sample of components are left unconstrained. We weighted the χ2–
Steidel et al. (2000). fit by the errorspectrumshowninthe figure;this hadthe
effect of diminishing the contribution to the fit by pixels
which fall on OH and O night sky emission lines (which
3.2. Modeling the Emission Line 2
are the dominant source of error in low signal–to–noise
TheasymmetricLyαemissionlinescommonlyobserved spectraoverthewavelengthsofinterest). Wenotethatthe
in high–redshift starburst galaxies are generally ascribed residualsaredistributedevenlyaboutzero,demonstrating
to the interactionofLymancontinuumphotonsgenerated anencouraginglackofsystematicerrorinourmodel. Fur-
by newborn OB associations with a galaxy–scale expand- thermore, when just the components intended to model
ing shell of neutral hydrogen. For a sufficiently massive the expanding shell were examined (the red, broad Gaus-
starburst,thehotionizedgascreatedinthevicinityofthe sian and the blue Voigt–profile absorption), we found the
stars vents into the halo of the galaxy,where it sweeps up resulting profile to be in excellent qualitative agreement
neutralhydrogenintoanopticallythickshell. Recombina- with the P–Cygniline profiles expected for the transferof
tionintheionizedgasconvertsLymancontinuumphotons line photons in expanding envelopes (e.g. see Surdej 1979,
escaping from the surface of the hot stars into line pho- and references therein).
tons. Then,fromthevantageofanobserver,thenearside Table 1 summarizes the minimum–χ2 model parame-
of the expandingshell absorbsphotons onthe blue side of ters. This best–fit model yields a redshift for the central
the resonant Lyα emission line, causing a flux decrement component of the ES1 emission line of z = 5.190±0.001.
onwhatwouldotherwisebethebluewingoftheLyαemis- Most strikingly, to fit the red wing of the emission line,
sion. The far side of the shell back–scatters Lyα photons themodeldemandsthatthebroademissioncomponentbe
intotheobserver’sline–of–sight;asthesephotonsareoffset displaced by 320 km s−1 from the central emission com-
redward by hundreds of km s−1 from both the rest frame ponent. Owing to the strong signal of the Lyα forest at
ofthegalaxyandtheapproachingsideoftheneutralshell, such a high redshift, only a small blue absorption com-
theyescapethegalaxyandimposeapronouncedredwing ponent is required. Still, though of minor amplitude, the
4 A Galactic Wind at z =5.190
absorption component is displaced by fully 360 km s−1 likely culprit (e.g. Renzini et al. 1993). Finally, galactic
fromthecentralemissioncomponent. Thesedisplacement winds have proved important in reproducing the faint–
velocities are in broad agreement with those predicted by end slope of the observed field galaxy luminosity function
Tenorio–Tagle et al. (1999) for the late stages of the evo- insemi–analyticmodelsofgalaxyformation. Outflowsare
lution of the Lyα profile of a star–forming galaxy. More- invokedtosuppressstar–formationinlow–massdarkmat-
over, though these values somewhat exceed the ∼102 km terhalos,eitherviathedirectescapeofgas–phasebaryons
s−1 winds typically observedin nearbystarburstingdwarf in the outflow itself (e.g. Somerville & Primack 1999), or
galaxies (e.g. Martin 1998), they compare favorably with byrampressurestrippingofthegas–phasebaryonsbyen-
the 102 – 103 km s−1 outflows observed in more powerful ergetic winds from neighboring galaxies (Scannapieco &
local starbursts (Heckman et al. 1990). Broadhurst2001).
Weexploredavarietyofalternativekinematicscenarios As a somewhatspeculative conclusion,we nowconsider
for the ES1 emission, most notably by fixing the centroid the expected correlation between strong galactic outflows
for the central emission component and arbitrarily slid- and the escape of Lyman continuum radiation from star–
ingtheseparationbetweenthecentralcomponentandthe forming galaxies. This correlation bears directly on the
broadcomponentoverarangeofvalues. Thesemodelssuf- much–debated physical nature and relative contributions
fered from worsenedχ2, with broader emission demanded of the sources which comprise the UV background, as a
as the displacement velocity was diminished. Figure 6 il- significant contribution by sources other than QSOs is re-
lustrates this trend, implying that even if the minimum– quired at high redshift, owing to the rapid decline in the
χ2 model in Figure 5 over–estimates the separation be- space density of optical and radio–loud quasars at z > 3
tween the central component and the broad component (Bianchi, Cristiani, & Kim 2001; Madau, Haardt, & Rees
of the Lyα emission, the broad component itself can only 1999).
get broader. That is, no matter what combination of fit– It is likely that star–forming galaxies fill this niche.
parameters one chooses, there is no escaping a very ener- From the theoretical standpoint, mechanical energy de-
getic component to the Lyα emission of ES1. In a similar positioninthe formof supernovaeandstellarwinds is ex-
vein,itisalsoworthyofnotethatmodelswhichdonotin- pected to result in an over–pressuredcavity of hot gas in-
cludeahigh–velocitycompenent(e.g.modelswithasingle side star–forminggalaxies. In galaxies for which the star–
Gaussianemissioncompenent)alsosufferfromaworsened formation rate per unit area Σ ≥ 10−1M yr−1 kpc−2,
∗ ⊙
χ2 compared to that of the model in Figure 5. the superbubble ultimately expands and bursts out into
the galaxy halo,allowing for the escape of hot gas and fa-
4. discussion and conclusion
cilitatingtheleakofLymancontinuumphotons(Heckman
ThespectralprofileoftheLyαemissionofES1presents et al. 2000; Tenorio–Tagle et al. 1999). Of course, as the
evidence for a galaxy–scale outflow with a velocity of superbubble will expand in the direction of the vertical
v >300km s−1. Ofcourse,as Heckmanetal. (2000)cau- pressure gradient, the burst is expected to take the form
tion,theoutflowrateofagalacticwindcannotnecessarily of a weakly collimated, bipolar wind. Hence, the leak of
be interpretedasthe rateatwhichmassorenergyescapes UV radiation may depend sensitively on not only the dis-
into the IGM, since the observable manifestation of an tributionofneutralgasanddustinthe galaxyinterstellar
outflow may be produced by material still deep inside the medium, but on the inclination of the system. Nonethe-
gravitational potential well of the galaxy halo. Nonethe- less,fromtheobservationalstandpoint,Steidel,Pettini,&
less, the outflow velocity estimated for ES1 far exceeds Adelberger (2001, hereafter SPA01) report the detection
the escape speed of a nominally low mass (M <1010M ) of significant Lyman continuum emission in a composite
⊙
pregalactic fragment, consistent with the generalobserva- spectrumof29Lyman–breakgalaxiesathzi=3.40±0.09,
tion that hot gas can readily escape from dwarf galaxies, suggesting an escape fraction9 of UV ionizing photons of
thoughperhapsnotfrommoremassivesystems(Heckman fesc>∼0.5.
et al. 2000;Heckman 2000; Martin 1999). GiventheevidenceforastrongoutflowinES1,itwould
This conclusion bears on a host of cosmological issues beintriguingtomeasurethefluxofphotonsbelowλ=912
surroundingtheevolutionofgalaxiesandtheIGMathigh ˚A.However,ES1isveryfaintevenaboveLyα;weestimate
redshift. Foremost, it suggests that processed material that it fades to IAB > 29 below the Lyman limit. We can
from ES1 will become available to the IGM, potentially do only slightly better at high redshifts with more acces-
providing the enrichment necessary to account for the sible spectra: even in a composite of four of the highest
amount of metals there observed. Indeed, recent observa- signal–to–noise Keck spectra of galaxies at z > 4.5 col-
tionsofCIVabsorptionsystemsalongthelines–of–sightto lected by Spinrad and collaborators (Figure 7), our mea-
lensed QSOs call for enrichment at increasingly high red- surement of the flux of 900 ˚A photons is consistent with
shift, beyond even z >5 (e.g. Aguirre et al. 2001; Rauch, zero at the 1σ level. This result translates to the coarsely
Sargent, & Barlow 2001). Additionally, both detailed ob- constrainedfluxratiof (1100 ˚A)/f (900 ˚A)=16.7±51.9
ν ν
servations andcareful theoreticalstudies demand a mech- (1σ uncertainty). To convert this value to the more
anism for pre–heating the material out of which galaxy useful f (1500 ˚A)/f (900 ˚A) ratio, we adopt an empir-
ν ν
clustersultimatelycollapseandbecomebound(e.g.Kaiser ical correction factor based on the set of fluxes given
1991; Mushotzky & Scharf 1997, and references therein). in SPA01, yielding an effective f (1500 ˚A)/f (900 ˚A) =
ν ν
Here again, galaxy–scale outflows at high redshift are the
9Here,fescisthefractionofemitted900˚Aphotonsthatescapesthegalaxywithoutbeingabsorbedbyinterstellarmaterial,normalizedbythe
fractionofemitted1500˚Aphotonswhichsimilarlyescapes. AsSPA01pointout,thisdefinitiondiffersfromdefinitionsencounteredelsewhere,
which typically consider only the fraction of emitted 900 ˚A photons which escapes (e.g. Bianchi et al. 2001; Heckman et al. 2001; Hurwitz,
Jelinsky,&Dixon1997;Leithereretal.1995).
Dawson et al. 5
31.4 ± 98.2. Finally, for the same intrinsic Lyman dis- the stewardofthe HDF andforprovidingthe NorthWest
continuity of 3 adopted by SPA1, we find an escape frac- FlankingFieldmosaicincludedinthiswork;totheanony-
tion of f >0.1±0.3. As we did not correct our initial mousrefereeforprovidingcareful,usefulcommentary;and
esc∼
f (1100 ˚A)/f (900 ˚A) ratio for the opacity of the IGM, to the expert staff of the Keck Observatory for their in-
ν ν
this value represents a lower limit. Evidently, to satisfac- valuable assistance in making the observation. The work
torily constrain the correlation between outflow dynam- of SD was supported by IGPP–LLNL University Collab-
ics and the escape of UV ionizing photons in an individ- orative Research Program grant #02–AP–015. HS grate-
ual high redshift galaxy, we will require spectroscopy of fully acknowledgesNSF grantAST 95–28536for support-
a lensed, blue candidate system viewed along the outflow ing much of the research presented herein. AD acknowl-
direction. edges partial support from NASA HF–01089.01–97A and
from NOAO. NOAO is operated by AURA, Inc., under
We are humbly indebted to E. Scannapiecoand J.Wal- cooperativeagreementwiththeNSF.TheworkofDSwas
ters for making generous and substantial contributions to carried out at the Jet Propulsion Laboratory, California
this work, and for providing prodigious comic relief. We Institute of Technology, under contract with NASA. The
are grateful to A. Barger for making public optical im- work of SD, WvB, WdV, and MR was performed under
ages ofthe HDF and its Flanking Fields ofwhich we have the auspices of the U.S. Department of Energy, National
made extensive use; to C. Steidel for providing the tar- NuclearSecurityAdministrationbythe UniversityofCal-
getwhichresultedinthisfortuitousdetection;toM.Hunt ifornia, Lawrence Livermore National Laboratory under
for unwittingly providing software which aided in the re- contract No. W–7405–Eng–48. This work made use of
duction of the echelle data; to M. Dickinson for acting as NASA’s Astrophysics Data System Abstract Service.
REFERENCES
Aguirre, A., Hernquist, L., Schaye, J., Weinberg, D., & Katz, N. Massey, P., Valdes, R., & Barnes, J. 1993, A
2001,ApJ,inpress[astro–ph/0006345] User’s Guide to Reducing Slit Spectra with IRAF,
Barger, A. J.,Cowie, L.L., Trentham, N.,Fulton, E.,Hu, E. M., http://iraf.noao.edu/iraf/web/docs/spectra.html
Songaila,A.,&Hall,D.1999, AJ,117,102 Mushotzky,R.F.&Scharf,C.A.1997, ApJ,482,L13
Bertin,E.&Arnouts,S.1996,A&AS,117,393 Ostriker,J.P.&Gnedin,N.Y.1996, ApJ,472,L63
Bianchi,S.,Cristiani,S.,&Kim,T.S.2001,A&A,376,1 Pettini,M.,Shapley, A.E.,Steidel,C.C.,Cuby,J.,Dickinson,M.,
Bunker,A.J.,Moustakas,L.A.,&Davis,M.2000,ApJ,531,95 Moorwood, A.F.M.,Adelberger, K.L.,&Giavalisco, M.2001,
Cowie,L.L.&Hu,E.M.1998,AJ,115,1319 ApJ,554,981
Dawson,S.,Stern,D.,Bunker,A.J.,Spinrad,H.,&Dey,A.2001, Rauch,M.,Sargent,W.L.W.,&Barlow,T.A.2001,ApJ,554,823
AJ,122,598 Renzini, A., Ciotti, L., D’Ercole, A., & Pellegrini, S. 1993, ApJ,
Dey, A.,Spinrad, H.,Stern, D.,Graham, J. R., & Chaffee, F. H. 419,52
1998,ApJ,498,L93 Rhoads, J. E., Malhotra, S., Dey, A., Stern, D., Spinrad, H., &
Dey, A., van Breugel, W., Vacca, W. D., & Antonucci, R. 1997, Jannuzi, B.T.2000,ApJ,545,L85
ApJ,490,698 Riess,A.etal.2001,ApJ,inpress[astro–ph/0104455]
Dickinson, M. 1998, in STScI Symp. Ser. 11, The Hubble Deep Scannapieco, E.&Broadhurst,T.2001, ApJ,549,28
Field,ed.M.Livio,S.M.Fall,P.Madau(Cambridge:Cambridge Sheinis, A. I., Miller, J. S., Bolte, M., & Sutin, B. M. 2000,
UniversityPress),219 in Proc. SPIE Vol. 4008, 522–533, Optical and IR Telescope
Ellis,R.,Santos, M.R.,Kneib, J.,&Kuijken,K.,2001, ApJ,560, Instrumentation and Detectors, ed. M. Iye & A. F. Moorwood
L119 (Bellingham:SPIE),522
Frye, B., Broadhurst, T., & Benitez, N., 2001, ApJ, in press Somerville,R.S.&Primack,J.R.1999,MNRAS,310,1087
[astro–ph/0112095] Spinrad, H., Dey, A., Stern, D., & Bunker, A. 1999, in Proc.
Gnedin,N.Y.&Ostriker,J.P.1997,ApJ,486,581 KNAW Colloq., 257, The Most Distant Radio Galaxies, ed. H.
Heckman, T. M. 2000, in ASP Conf. Ser. 240, Gas & Galaxy Rottgering, N.Best,M.Lenhert(Utrecht: KNAW),257
Evolution, ed. J. Hibbard, M. Rupen, J. van Gorkom (San Spinrad, H., Stern, D., Bunker, A.,Dey, A., Lanzetta, K.,Yahil,
Francisco:ASP),availableNovember 2001[astro–ph/0009075] A.,Pascarelle,S.,&Ferna´ndez–Soto, A.1998,AJ,116,2617
Haiman,Z.&Loeb,A.1997,ApJ,483,21 Steidel, C. C., Adelberger, K. L., Shapley, A. E., Pettini, M.,
Haiman,Z.&Loeb,A.1998,ApJ,503,505 Dickinson,M.,&Giavalisco,M.2000,ApJ,532,170
Heckman, T.M.,Armus,L.,&Miley,G.K.1990,ApJS,74,833 Steidel, C. C., Giavalisco, M., Pettini, M., Dickinson, M., &
Heckman, T. M.,Lehnert, M.D., Strickland, D. K.,& Armus,L. Adelberger,K.L.1996,ApJ,462,L17
2000,ApJS,129,493 Steidel,C.C.,Pettini,M.,&Adelberger,K.L.2001,ApJ,546,665
Heckman, T. M., Sembach, K. R., Meurer, G. R., Leitherer, C., (SPA01)
Calzetti,D.,&Martin,C.L.2001, ApJ,558,56 Stern,D.,Eisenhardt,P.,Spinrad,H.,Dawson,S.,vanBreugel,W.,
Hu,E.M.,McMahon,R.G.,&Cowie,L.L.1999,ApJ,522,L9 Dey,A.,deVries,W.,&Stanford,S.A.2000,Nature,408,560
Hurwitz,M.,Jelinsky,P.,&Dixon,W.V.D.1997,ApJ,481,L31 Stern,D.&Spinrad,H.1999, PASP,111,1475
Kaiser,N.1991,ApJ,383,104 Surdej,J.1979, A&A,73,1
Leitherer,C.,Ferguson,H.C.,Heckman,T.M.,&Lowenthal,J.D. Tenorio–Tagle, G., Silich, S. A., Kunth, D., Terlevich, E., &
1995,ApJ,454,L19 Terlevich,R.1999,MNRAS,309,332
Lowenthal, J.D.,etal.1997, ApJ,481,673 Thompson,R.I.,Weymann, R.J.,&Storrie–Lombardi,L.J.2001,
Madau,P.1995, ApJ,441,18 ApJ,546,694
Madau,P.,Haardt,F.,&Rees,M.J.1999,ApJ,514,648 Tody, D.1993, inASPConf. Ser.52, AstronomicalData Analysis
Manning, C., Stern, D., Spinrad, H., & Bunker, A. J. 2000, ApJ, SoftwareandSystemsII,ed.R.Hanisch,R.Brissenden,J.Barnes
537,65 (SanFrancisco:ASP),173
Martin,C.L.1998, ApJ,506,222 Valageas,P.&Silk,J.1999,A&A,347,1
—.1999,ApJ,513,156 Weymann,R.J.,Stern,D.,Bunker,A.,Spinrad,H.,Chaffee,F.H.,
Massey,P.&Gronwall,C.1990, ApJ,358,344 Thompson,R.I.,&Storrie–Lombardi,L.J.1998,ApJ,505,L95
Williams,R.E.,etal.1996,AJ,112,1335
6 A Galactic Wind at z =5.190

c) 8
e
s
c
r
a 4
(

t
i
l
S
g 0
n
o
l
A
−4
n
o
i
t
i
s
o−8
P
7490 7500 7510 7520 7530 7540 7550 7560
Observed Wavelength (Å)
Fig. 1.—AportionofoneorderoftheES1discoveryspectrum. Thedispersionaxisishorizontal;thespatialaxisisvertical. Thevertical
featuresflankingtheemissionlineareremnantsoftheOHandO2night–skyemissionlinesat7524˚Aand7531˚A,respectively. Thecontinuum
ofthetargetgalaxy,D16,isbarelyvisibleneartheslitposition−4′′. See§2foradescriptionoftheobservation.
Table 1
ES1 Lyα Line Model Parameters
Component λ Peak Amplitude FWHM ∆v†
(˚A) (10−18 ergs s−1 cm−2 ˚A−1) (˚A / km s−1) (km s−1)
Narrow Component 7527 3.16 7 / 280 ···
Broad Component 7535 0.62 14 / 560 320
Absorption Component 7518 −0.06 20 / 800 −360
† Displacement velocity relative to the central, narrow emission component. Positive
velocity is a redshift; negative velocity is a blueshift.
Dawson et al. 7

D16
N
ES1
E
5"
Fig. 2.—CentralregionoftheHSTI814 mosaicoftheHubbleDeepFieldNorthWestFlankingFieldwithaprojectionofthespectroscopic
slit. Thetargetgalaxy(D16)atα=12h36m49′.′0,δ=+62◦15′43′′(J2000)andtheserendipitouslydetectedgalaxy(ES1)atα=12h36m49′.′2,
δ=+62◦15′39′′ (J2000)areindicated. Thepanelmeasures20′′ squareandtheslitdimensionsare1′′ ×20′′. Themosaicandtheastrometry
thereinwereprovidedbyDickinson(2001,privatecommunication). Theslitpositionwasdeterminedbyoffsettingfromtheset–upstarused
intheoriginalobservation(notshown),therebynullifyingerrorsinabsoluteastrometry.
Table 2
ES1 Ground–Based Photometry
Band Magnitude
(Vega–based, 1′.′5 aperture)
V >26.1†
R >27.5†
I 25.1±0.1‡
z >25.2†
†3σ limiting magnitude.
‡Add ∼ 0.4 magnitudes to this
Vega–basedmagnitudeforcomparison
to the AB isophotal magnitude cited
in § 2.
8 A Galactic Wind at z =5.190
Fig. 3.—Ground–based supportingimagingforES1. TheV andI imagesarefromBargeretal.(1999); the Randz imagesarefromour
ownimagingcampaignoftheHDFanditsenvirons(seeSternetal.2000). Thefieldsare1′square;NorthisupandEastistotheleft. Notice
that ES1is notdetectable intheV orRbands butisseenintheI band. Thisischaracteristic ofhigh–redshiftgalaxies, whereintervening
neutral hydrogen (theLymanforests)severelyattenuates thecontinuum signalbluewardofLyα. Thefactthat ES1isnotdetectable inthe
z band is due only to the shallowness of that image. We expect the z band continuum magnitude for ES1 to be z >27.0; the 3σ limiting
magnitudeofthez bandimageisz=25.2. SeeTable2forasummaryoftheground–basedphotometry.
Dawson et al. 9
Fig. 4.—Lyαemissionlineluminosityvs.redshiftfor13Lyα–emitting galaxies inanΩM =0.35, ΩΛ =0.65cosmology. ES1isindicated
withastar. TheSFRscalehasbeenadoptedfromDeyetal.(1998). ThesampleofgalaxiesrepresentedbycircleswascompiledfromDawson
etal.(2001), Deyetal.(1998), Hu,McMahon,&Cowie(1999), Manningetal.(2000), Rhoadsetal.(2000), Spinradetal.(1999), Stanford
(2001, privatecommunication), and Stern& Spinrad(1999). The galaxyrepresented bythe triangleisfromElliset al.(2001). Thedashed
lineindicatesthelimitingsensitivitytolinefluxintheLargeAreaLymanAlphaSurvey(∼2×10−17 ergcm−2 s−1;Rhoadsetal.2000).
10 A Galactic Wind at z =5.190
Fig. 5.—(a)Theminimum–χ2 fittotheES1Lyαemissionline. Thelineprofileisthesumof(1)anarrow(280kms−1 FWHM)central
Gaussianintendedtomodelrecombinationinthehotionizedgasofthestarburst,(2)abroad(560kms−1 FWHM)Gaussianredshiftedby
320kms−1 fromthecentralcomponentintendedtomodelback–scattering offofthefarsideofanexpandingshell,and(3)abroad(800km
s−1 FWHM)Voigtabsorptioncomponent blueshiftedby360kms−1 fromthecentral component intended tomodel absorptionbythenear
sideofthe expanding shell. Themodel spectrum wasattenuated bythe model ofthe Lyαforestpresented byMadau (1995). (b)Theerror
per pixel inthe same flux units and over the same wavelength range. For background–limited observations of faint objects inthis region of
wavelength space, night–sky emission lines are the dominant source of noise. (c) The model–fit minus the data in the same flux units and
overthesamewavelengthrange. Theevendistributionoftheresidualsdemonstrates alackofsystematicerrorinthemodel.