THE ASTRONOMICAL JOURNAL, 116 :1643È1649, 1998 October1998. The American Astronomical Society. All rights reserved. Printed in U.S.A.(

SURFACE BRIGHTNESS OF STARBURSTS AT LOW AND HIGH REDSHIFTS

DANIEL W. AND JEFFREY B.WEEDMAN WOLOVITZ

Department of Astronomy and Astrophysics, Pennsylvania State University, 525 Davey Laboratory, University Park, PA 16802 ;weedman=astro.psu.edu, jbw122=psu.edu

MATTHEW A. BERSHADY

Department of Astronomy, University of Wisconsin, 475 North Charter Street, Madison, WI 53706 ; mab=astro.wisc.edu

AND

DONALD P. SCHNEIDER

Department of Astronomy and Astrophysics, Pennsylvania State University, 525 Davey Laboratory, University Park, PA 16802 ; dps=astro.psu.eduReceived 1998 April 4

ABSTRACTObservations in the rest-frame ultraviolet from various space missions are used to deÐne the nearby

starburst regions having the highest surface brightness on scales of several hundred parsecs. The brightlimit is found to be 6 ] 10~16 ergs cm~2 s~1 arcsec~2 for rest-frame wavelength of 1830 SurfaceA� ~1 A� .brightness in the brightest pixel is measured for 18 galaxies in the Hubble Deep Field having z[ 2.2.After correcting for cosmological dimming, we Ðnd that the high-redshift starbursts have intrinsic ultra-violet surface brightness that is typically 4 times brighter than low-redshift starbursts. It is not possibleto conclude whether this di†erence is caused by decreased dust obscuration in the high-redshift starburstregions or by intrinsically more intense star formation. Surface brightness enhancement of starburstregions may be the primary factor for explaining the observed increase with redshift of the ultravioletluminosity arising from star formation.Key words : galaxies : distances and redshifts È galaxies : evolution È galaxies : starburst

1. INTRODUCTION

Many new observational results make possible investiga-tions of the evolution of star formation within galaxies,extending to redshifts approaching 4 et al.(Madau 1996 ;

et al. et al. To date, the essenceLilly 1996 ; Connolly 1997).of these results is that the global star-forming rate, deducedfrom the ultraviolet luminosity integrated over all observ-able galaxies in comoving volumes of the universe, was sig-niÐcantly greater at earlier epochs than it is today. Theseresults describe the overall rate of star formation. This starformation is in ““ starbursts ÏÏ for those galaxies whoseappearance is dominated by short-lived populations ofyoung stars. While counts of such star-forming galaxies candetermine global rates, these results do not address thequestion of whether the local characteristics of the star for-mation process within individual galaxies also depends onepoch in the universe. Understanding these characteristicsof star formation is crucial to explaining the changing rateof star formation with epoch.

Recent dramatic progress in identifying starburst galaxiesat redshifts above 2 in the Hubble Deep Field (HDF;

et al. et al. has produced aSteidel 1996 ; Lowenthal 1997)sample of such galaxies that can be examined with sufficientspatial resolution that individual starburst regions can beexamined within the galaxies. For the Ðrst time, it becomespossible to compare such regions with those of similar sizein nearby galaxies and to make this comparison in the rest-frame ultraviolet luminosity, which originates completelyfrom the starburst. This allows an initial comparison of thestarburst process between galaxies separated in time byamounts exceeding two-thirds the age of the universe.

2. SURFACE BRIGHTNESSES OF STARBURST REGIONS

Our objective in the present analysis is a direct compari-

son between the surface brightness of starburst regions atlow redshift and at high redshift, as observed in the rest-frame ultraviolet where the luminosity is completely domi-nated by the recently formed stars. If the starburst region isresolved, surface brightness is a very useful parameter forcomparing regions of greatly di†ering redshift ; surfacebrightness, unlike luminosity or diameter, does not dependon cosmological parameters, i.e., or This makesH0, )0, "0.possible a comparison of local physical conditions that isindependent of cosmological assumptions.

In Euclidean space, surface brightness is independent ofdistance. For signiÐcant redshifts, surface brightnesschanges dramatically but only in terms involving 1] z.When measured in units at a given rest-frame wavelength,fjdistant objects of the same intrinsic surface brightnessshould have their observed surface brightness fade by(1 ] z)~5 for Friedmann cosmologies. There are Ðve factorsof 1 ] z when observations are in units : two factorsfjbecause of the change in unit area observed from the galaxy,one for change in unit time, one for change in photonenergy, and one for change in unit wavelength bandpass (ifobservations are in bolometric quantities, such as orjfj lflunits, only the Ðrst four factors of 1 ] z enter. This is equiv-alent to observing in a bandpass Ðxed in either the observedor rest frame. If counting photons, there is still one lessfactor of 1] z).

Achieving a surface brightness measurement requires thatthe starburst regions be resolved. We refer to these resolvedsources as ““ extended starburst regions.ÏÏ We avoid star-bursts associated with galactic nuclei because the starburstluminosity in such cases may be confused with that of anunresolved active galactic nucleus. An empirical upper limitto the rest-frame ultraviolet surface brightness of suchregions within nearby galaxies is deduced by searching forthose nearby extended starburst regions with the highestsurface brightness. Observations in the ultraviolet from a

1643

1644 WEEDMAN ET AL. Vol. 116

variety of space missions are the source of the relevant data.These results are summarized in The galaxiesTable 1.observed with the various missions were chosen as sourcesexpected to be ultraviolet-bright. While it is possible thateven brighter extended starburst regions have been over-looked, the similarity in the results from independent selec-tions of galaxies gives reasonable conÐdence that ameaningful upper limit for surface brightness of extendedstarburst regions in nearby galaxies can be determined fromthese various observations.

An extensive compilation of observations of star-forminggalaxies with the International Ultraviolet Explorer (IUE) isgiven by et al. For many such observations,Kinney (1993).the starburst regions are signiÐcantly smaller than the10A ] 20A IUE aperture, so surface brightnesses within thisaperture are underestimates (ultraviolet images of somenearby galaxies from the Hubble Space Telescope [HST ]Faint Object Camera [FOC] are in et al. NotMaoz 1996).surprisingly, the brightest such galaxies within an aperturethis large are generally also the closest for which the star-bursts would be resolved by the IUE aperture. ExcludingNGC 1068, the Ðve galaxies with brightest total Ñux at 1900

(their 1863È1963 bandpass) are NGC 1705, 3310, 4449,A� A�5236, and 5253. The average surface brightness of these Ðve,with the IUE aperture taken as 200 arcsec2, is 5] 10~16ergs cm~2 s~1 arcsec~2. DeÐning the scale size by theA� ~120A length of the aperture, these brightnesses arise fromregions of average size 700 pc. While it is arbitrary torestrict our selection to the Ðve brightest, it can be seen from

that any other entries would be more than a factorTable 1of 2 fainter than the brightest of all (NGC 5236).

The galactic disk with highest observed ultravioletsurface brightness is the inner disk of NGC 1068, character-ized by extensive star formation throughout the 3 kpc diam-eter disk centered on the active nucleus of this prototypeSeyfert 2 galaxy. Numerous individual bright regions wereimaged with the Ultraviolet Imaging Telescope (UIT) onthe Astro 1 mission et al. The brightest of these(Ne† 1994).is their knot complex labeled ““ region J,ÏÏ stated to have Ñuxat 1500 of 2 ] 10~14 ergs cm~2 s~1 From theirA� A� ~1.published image, this region appears dominated by brightknots covering an area of about 30 arcsec2, which yields asurface brightness of 7 ] 10~16 ergs cm~2 s~1 A� ~1arcsec~2 within a characteristic diameter of about 350 pc(5A at an assumed distance of 15 Mpc for km s~1H0\ 75Mpc~1 ; this value of is assumed throughout).H0

TABLE 1

SURFACE BRIGHTNESSES OF BRIGHTEST LOW-REDSHIFT EXTENDED

STARBURST REGIONS

Wavelength Surface SizeObject (A� ) Brightnessa (pc) Referenceb

NGC 1705 . . . . . . . . . . 1900 4.2 (4.4) 770 1NGC 3310 . . . . . . . . . . 1900 3.8 (3.9) 1270 1NGC 4449 . . . . . . . . . . 1900 3.6 (3.7) 270 1NGC 5236 . . . . . . . . . . 1900 8.6 (8.9) 670 1NGC 5253 . . . . . . . . . . 1900 5.3 (5.5) 520 1NGC 3690-BC . . . . . . 2200 5.0 (6.0) 1300 2NGC 1068-J . . . . . . . . 1500 7.0 (5.7) 350 3

a Units of 10~16 ergs cm~2 s~1 arcsec~2 at the wavelength of theA� ~1observations ; values in parentheses give surface brightness at the normal-ized wavelength of 1830 assuming that all objects have ultravioletA� ,spectra of shape et al.fjP j~1 (Kinney 1993).

b (1) IUE et al. (2) HST FOC et al. (3)(Kinney 1993) ; (Meurer 1995) ;UIT et al.(Ne† 1994).

Several starburst galaxies were imaged at 2200 with theA�HST FOC et al. Among these is the highly(Meurer 1995).luminous and well-studied starburst galaxy NGC 3690(Mkn 171). The brightest UV knot, NGC 3690-BC, hassurface brightness of 5 ] 10~16 ergs cm~2 s~1 A� ~1arcsec~2 within diameter of (1360 pc at distance 446A.4Mpc).

For comparisons to follow with starburst regions at highredshift, we desire a normalizing wavelength of 1830 A� .Nominal corrections to the observed values in areTable 1applied by assuming a common spectral shape for allobjects observed. Taking the median starburst ultravioletspectral shape from et al. of theKinney (1993) fjP j~1,rest-frame surface brightnesses at 1830 are also tabulated.A�The results are adequately consistent to deÐne a meaningfulmeasure of the surface brightness for the brightest localstarbursts. Using these results, we adopt a value of6 ] 10~16 ergs cm~2 s~1 arcsec~2 for the surfaceA� ~1brightness of the ““ brightest ÏÏ starburst regions withinnearby starbursts when observed at a rest-frame wavelengthof 1830 We will use this number with which to compareA� .the observed surface brightness of systems at high redshift.

The image quality of HST allows star-forming regions tobe resolved and surface brightnesses to be measured, evento the highest observable redshifts. In the Hubble DeepField et al. the ““ drizzled add ÏÏ pixels are(Williams 1996),

in size. Such a pixel corresponds to a physical size of0A.04about 300 pc at 2 \ z\ 3 (for angular sizes)0\ 0.2 ;change little with redshift for this and This is)0 "0 \ 0).sufficiently small compared with the size of luminous star-burst regions in nearby galaxies that it is meaning-(Table 1)ful to compare surface brightnesses of distant starbursts inthe HDF with those in nearby galaxies. Although spatialresolution in the HDF as deÐned by the point-spread func-tion is larger than the size of a single pixel, the surfacebrightness measures from a single pixel are accurate mea-sures if starburst regions are fully resolved in the HDF. Ifthe starburst regions are unresolved, a measure of thesurface brightness in the brightest pixel provides a lowerlimit to the actual surface brightness of the source within anarea of 0.016 arcsec2. The surface brightness would behigher if the source is smaller than a single pixel and isunresolved. In such cases, however, the starburst regions athigh redshift would be signiÐcantly smaller than most of thenearby regions in Because of smearing by theTable 1.point-spread function, measuring surface brightness in thebrightest pixel does not yield a value that is signiÐcantlygreater than an average taken over a few pixels. The bright-est pixel is typically found to be a factor of 1.09 times bright-er than the average of the four brightest pixels (600 pc at2 \ z\ 3) and 1.22 times brighter than the average of thenine brightest (900 pc at 2 \ z\ 3). These apertures arecomparable to the IUE, HST FOC, and the UIT aperturesused to measure the surface brightness of the local star-bursts listed in Table 1.

There are 18 galaxies in the HDF with known redshiftssufficiently high that rest-frame Ñux measurements areavailable for ultraviolet wavelengths comparable to those atwhich surface brightnesses of local starburst regions havebeen measured. For these galaxies in the HDF, the ultra-violet surface brightness at 1830 rest-frame wavelength isA�given in Objects in this table have redshifts deter-Table 2.mined by the authors referenced using the Keck Telescope.Images of these objects from the HDF are shown inFigure 1.

No. 4, 1998 SURFACE BRIGHTNESS OF STARBURSTS 1645

TABLE 2

HUBBLE DEEP FIELD HIGH-REDSHIFT GALAXIES

BRIGHTEST

PIXEL

NUMBER CHIP x y z Sj(1830) REFERENCE

(1) (2) (3) (4) (5) (6) (7)

1 . . . . . . . . 2 647 1740 2.233 4.1 L2 . . . . . . . . 2 1952 209 2.267 4.4 L3 . . . . . . . . 4 742 960 2.268 4.9 C4 . . . . . . . . 2 546 1610 2.419 1.65 L5 . . . . . . . . 2 279 1321 2.591 3.8 S6 . . . . . . . . 3 360 1206 2.775 1.61 S7 . . . . . . . . 4 1586 1172 2.803 4.2 S8 . . . . . . . . 2 1805 960 2.845 9.1 S9 . . . . . . . . 4 289 188 2.931 1.24 L10 . . . . . . 4 1952 773 2.980 1.18 L11 . . . . . . 2 356 1299 2.991 2.24 L12 . . . . . . 2 1332 181 3.160 0.82 L13 . . . . . . 2 1281 1738 3.181 1.94 L14 . . . . . . 4 1070 1762 3.226 4.1 S, Z15 . . . . . . 3 330 606 3.233 2.1 L16 . . . . . . 3 506 1279 3.360 0.4 Z17 . . . . . . 2 627 1288 3.368 2.6 L18 . . . . . . 2 620 1219 3.430 1.12 L

NOTES.È(1) Identifying number in (2) Wide Field PlanetaryFig. 1.Camera 2 chip. (3) The x-location on HDF image (vers. 2) of brightestpixel. (4) The y-location on HDF image (vers. 2) of brightest pixel. (5)Spectroscopic redshift. (6) Surface brightness in brightest pixel for rest-frame wavelength of 1830 in units of 10~18 ergs cm~2 s~1 arcsec~2.A� A� ~1(7) References for spectroscopic redshifts : (L) et al. (C)Lowenthal 1997 ;

et al. (S) et al. (Z) Moustakas & DavisCohen 1996 ; Steidel 1996 ; Zepf,1997.

Surface brightness is measured from the single brightestpixel using a linear interpolation between the e†ectivewavelengths of the two Ðlters that Ñank the redshifted 1830

wavelength in the observerÏs frame. The brightest pixel isA�chosen from the F814W image, and the same pixel is used inthe F606W or other images to determine the interpolation.The choice of 1830 for the comparison wavelength isA�made so that the object of highest redshift in the analysishas a rest wavelength corresponding to an observed wave-length no longer than the e†ective wavelength of theF814W image.

3. COMPARING LOW-REDSHIFT AND HIGH-REDSHIFT

STARBURST REGIONS

Ultraviolet observations of nearby starburst regions sum-marized in indicate that the maximum surface brightness° 2of such regions is about 6 ] 10~16 ergs cm~2 s~1 A� ~1arcsec~2 when observed at a wavelength of 1830 DistantA� .objects of the same intrinsic surface brightness should havetheir observed surface brightness, at the same rest-framewavelength, reduced to 6] 10~16(1 ] z)~5 ergs cm~2 s~1

arcsec~2.A� ~1displays the observed surface brightnesses at theFigure 2

rest-frame wavelength of 1830 [observerÏs wavelength ofA�1830(1 ] z) from the HDF starburst regions inA� ] Table 2.These are compared with the expected surface brightnessfrom the scaling from low-redshift starbursts. As expectedfrom the cosmological fading, surface brightnesses of high-redshift starbursts are dramatically less than for starburstregions at low redshifts. Nevertheless, objects in the HDFwith high redshifts have rest-frame surface brightnesses sig-niÐcantly greater than expected using the scaling from localstarbursts. From the median excess is a factor of 4,Figure 2,

and the brightest starburst in the Ðgure exceeds theexpected surface brightness by a factor of 11. Keeping inmind that these values are only lower limits if starburstregions are smaller than a single pixel, this result clearlydemonstrates that galaxies at high redshift detected in theHDF have higher ultraviolet luminosity per unit area thanthe brightest known examples of local starbursts.

This is a di†erent conclusion from that of et al.Meurerand et al. These authors conclude from(1997) Pettini (1998).

ultraviolet surface brightnesses that starburst regions havesimilar star formation intensity per unit area regardless ofredshift. Pettini et al. state that starburst regions at highredshift are ““ spatially more extended versions of the localstarburst phenomenon.ÏÏ Di†erences between our conclu-sions and theirs may arise primarily because we measuresurface brightnesses within the smallest observable spatialscales, using the closest starburst regions for comparisonsand using only starbursts from the HDF to deduce thesurface brightness of the starburst regions at high redshift.Their results refer to spatial scales at high redshift averagedover half-light radii of about 2000 pc, which are signiÐcantlylarger scales than those within which we measure surfacebrightnesses.

The di†erence in our comparison between low-redshiftand high-redshift starburst regions is not explainable simplyas a selection e†ect arising from comparing signiÐcantlydi†erent volumes of space. If a much larger comovingvolume of the universe were observed to locate starbursts inthe HDF compared with the volume of space used to locatethe low-redshift comparison starbursts, it might be expectedthat rarer, brighter regions would deÐne the brightest exam-ples in the larger volume. Consider the volume examined toÐnd the high-redshift starbursts in the HDF. There are 18galaxies in distributed roughly evenly in redshift forTable 2,2.2\ z\ 3.4, and 17 of these galaxies exceed the surfacebrightness of the local starburst galaxies. The three wide-Ðeld CCDs of the HDF examined to Ðnd these galaxiescover 1.5] 10~3 deg2. Within this area of the sky, the co-moving volume increment in this redshift range is 15] 104Mpc3 (for or 5] 104 Mpc3 (for These)0\ 0.2) )0\ 1).alternative volumes yield volume densities of the brightesthigh-redshift starbursts between 11 ] 10~5 Mpc~3 and34 ] 10~5 Mpc~3.

It is more difficult to estimate precisely the low-redshiftvolume that has been included to deÐne the low-redshiftsurface brightness, but that volume can be bracketed rea-sonably well. The most distant galaxy used to deÐne themaximum local starburst surface brightness is NGC 3690,early identiÐed as the most dramatic example of anextended starburst in the Markarian sample of galaxies

Sramek, & Weedman Taking the Markarian(Gehrz, 1983).survey as covering 0.25 of the sky and taking the distance ofNGC 3690 as 42 Mpc, the volume of the universe examinedto Ðnd NGC 3690 is 7.5 ] 104 Mpc3. The next closestobject in is NGC 1068, at 15 Mpc distance. TheTable 1entire universe has been examined for bright galaxies withinthat distance, so 1.4 ] 104 Mpc3 has been searched. It isreasonable to conclude that these limits (1.4 ] 104 Mpc3 to7.5] 104 Mpc3) bracket the local volumes searched todeÐne the maximum surface brightness of nearby extendedstarburst regions as summarized in Based on the 18Table 1.sources found in the HDF, we would then expect 7~6`12sources of comparable surface brightness as the HDF high-redshift starbursts in the surveyed local volume. In other

FIG

.1.

ÈIm

ages

ofth

ega

laxi

esin

from

the

F81

4Wim

ages

ofth

eH

ubbl

eD

eep

Fie

ld.E

ach

box

isea

chtick

mar

kco

rres

pond

sto

roug

hly

1200

pckm

s~1

Mpc

~1an

dT

able

13A.

32]

3A.32

;(H

0\75

)0\

0.2)

.

SURFACE BRIGHTNESS OF STARBURSTS 1647

FIG. 2.ÈSurface brightness (units of 10~18 ergs cm~2 s~1 A� ~1arcsec~2) at 1830 compared with 1 ] z. The triangles represent observedA�values for galaxies in The curve represents expected values forTable 2.brightest local starburst regions (6 ] 10~16 ergs cm~2 s~1 arcsec~2 ;A� ~1see text) faded by a cosmological factor of (1] z)~5.

words, local surveys do cover sufficient volume to Ðnd thetypes of high surface brightness starbursts found in theHDF if they are at comparable space densities locally.Using only the seven objects in to deÐne the densityTable 1of locally brightest starbursts, the local density of these star-bursts is between 9 ] 10~5 Mpc~3 and 50 ] 10~5 Mpc~3.Because this density range is very similar to that of the(much more luminous) high-redshift sample, we also canconclude that extended starburst regions of similar co-moving density have much higher surface brightness at highredshift compared with low redshift.

The result that starburst regions are signiÐcantly brighterper unit area at high redshift compared with low redshift isempirical and is our primary conclusion. Understandingwhy this is the case will require knowing more about thespectral characteristics of starburst regions at high redshift,especially including infrared luminosities. There are twopossibilities for explaining this result. The Ðrst is that star-bursts at high redshift are intrinsically more intense per unitarea than are nearby starbursts. The second possibility isthat high-redshift starbursts are less absorbed by dust, sothat a higher proportion of the rest-frame ultravioletescapes and the starbursts simply appear brighter whenobserved in the rest-frame ultraviolet.

et al. and et al. conclude thatMeurer (1997) Pettini (1998)high-redshift starburst regions are signiÐcantly absorbed,although they di†er in conclusions regarding the amount ofabsorption. Neither suggests a systematic di†erencebetween low-redshift and high-redshift objects, whichwould discount di†ering obscuration as the explanation forthe di†erential brightening that we measure. Their e†orts tounderstand the dust absorption in high-redshift starburstswere motivated by the importance of understanding theoverall rate of star formation at high redshift. This rate canbe signiÐcantly underestimated if the ultraviolet light ofyoung stars is heavily obscured. Both analyses estimate dustobscuration by the reddening imposed on the ultravioletspectrum. A difficulty in using only ultraviolet spectra to

determine dust obscuration is that all stars that contributeto the ultraviolet luminosity must be assumed to bereddened by an amount small enough that they remainvisible but large enough that the e†ects of reddening arenoticeable in the spectrum. Stars that are so obscured thatnegligible ultraviolet luminosity escapes would not beaccounted for. Because of this e†ect, estimates of obscur-ation derived strictly from ultraviolet spectra provide onlyupper limits for the escaping fraction of ultraviolet lumi-nosity. Accounting for completely obscured stars requiresobservations in the rest-frame infrared, where the radiationemerges as reradiation from the obscuring dust (see Rowan-

et al. and Herter, & HaynesRobinson 1997 Smith, 1998).Such observations are not available for high-redshift star-bursts.

The results from Meurer et al. yield a distribution of dustobscuration in high-redshift starburst regions (their Fig. 8),a distribution that can be expressed in terms of the fractionof intrinsic ultraviolet luminosity that escapes. For thepresent analysis, we wish to know only if this distribution issystematically di†erent compared with low-redshift star-burst regions. For the low-redshift starbursts, we canexamine the quantitative obscuration by comparing ultra-violet and infrared luminosities, which together shouldaccount for all of the starburst, regardless of the degree ofobscuration. This analysis is similar to that in Weedman

but with much improved data. If an initial mass func-(1991)tion is assumed for a starburst and used with stellar atmo-spheric models describing the ultraviolet luminosity andbolometric luminosity of stars of various masses, the ratioof intrinsic ultraviolet to bolometric luminosity is knownfor the starburst. For the present analysis, we use the modeldescribed in et al. which normalizes suchMeurer (1997),that fj2200 \ 1.5] 10~4fbol.Because of the availability of far-infrared Ñuxes fromIRAS and ultraviolet Ñuxes from IUE, a substantialnumber of starburst galaxies have measured bolometricÑuxes et al. As these authors comment, there(Schmitt 1997).are uncertainties in comparing and because of di†er-fUV fboling aperture sizes in the observations, but the IUE apertureis large enough that it generally includes all of the relevantstarburst, and the IRAS Ñuxes, though having poorerspatial resolution, are generally dominated by the samestarburst regions et al. In are listed(Calzetti 1995). Table 3all of the starburst galaxies in et al. includingSchmitt (1997),those which they classify as both low reddening and highreddening, listing the as read from their plots. Also givenfbolare the Ñuxes as deduced from the summary of IUEfj2200observations in et al. (the are deducedKinney (1993) fj2200by interpolating between the entries with central wave-lengths of 1913 and 2373 The intrinsic deducedA� ). fj2200from using the Meurer result thatfbol fj2200 \ 1.5

is also listed in Comparing observed] 10~4fbol Table 3.and intrinsic yields the fraction of ultraviolet lumi-fj2200nosity that escapes, which is listed in the table.

In the cumulative distributions of escaping frac-Figure 3,tion for the ultraviolet luminosity are shown. For themedian in the distribution of low-redshift starbursts from

only 10% of the intrinsic 2200 luminosityTable 3, A�escapes. This is similar to the results deduced by etMeureral. using only the shape of the ultraviolet continuum(1997)as an indicator of dust absorption. For their high-redshiftsample of galaxies, the cumulative distribution is alsoshown in Comparison of the distributions gives noFigure 3.

1648 WEEDMAN ET AL. Vol. 116

TABLE 3

LOCAL STARBURST GALAXIES

fj2200 ESCAPE

OBJECTa fbolb Observedc Intrinsicd FRACTIONe

Low reddening :NGC 5236 . . . . . . 145 16.7 220 0.076NGC 5253 . . . . . . 38.9 9.3 58 0.16NGC 1140 . . . . . . 6.0 2.6 9.0 0.29NGC 7250 . . . . . . 4.5 0.81 6.8 0.12Mrk 960 . . . . . . . . 2.95 1.16 4.4 0.26NGC 3049 . . . . . . 4.17 0.86 6.3 0.13Mrk 542 . . . . . . . . 2.09 0.39 3.1 0.12Mrk 357 . . . . . . . . 1.74 0.89 2.6 0.34UGC 9560 . . . . . . 1.58 1.54 2.4 0.65NGC 6052 . . . . . . 7.94 0.92 11.9 0.076

High reddening :NGC 1097 . . . . . . 57.5 1.74 86 0.02NGC 3256 . . . . . . 102 1.14 153 0.007NGC 1672 . . . . . . 50.1 2.8 75 0.037NGC 7552 . . . . . . 83.2 1.92 125 0.015NGC 6217 . . . . . . 16.6 1.36 25 0.054NGC 7714 . . . . . . 14.1 2.14 21.1 0.10NGC 4385 . . . . . . 6.76 0.97 10.1 0.095NGC 6090 . . . . . . 8.91 0.77 13.4 0.058IC 214 . . . . . . . . . . 7.08 0.39 10.6 0.036NGC 7673 . . . . . . 6.61 1.22 9.9 0.123NGC 5860 . . . . . . 3.24 0.42 4.9 0.085NGC 5996 . . . . . . 0.89 0.96 1.34 0.707IC 1586 . . . . . . . . . 2.34 0.42 3.5 0.117NGC 7793 . . . . . . 12.3 0.70 18.5 0.038

a For convenience, objects are grouped according to the starburst cate-gories as in et al. which they describe as ““ low reddening ÏÏSchmitt 1997,and ““ high reddening ÏÏ based on the shapes of the ultraviolet spectra.

b Units of 10~10 ergs cm~2 s~1.c Units of 10~14 ergs cm~2 s~1 A� ~1.d Units of 10~14 ergs cm~2 s~1 Intrinsic isA� ~1. fj2200 \ 1.5] 10~4fbolfrom et al.Meurer 1997.e Escaping fraction\ observed / intrinsicfj2200 fj2200.

FIG. 3.ÈNormalized cumulative distribution of ultraviolet luminositythat escapes starburst region without being absorbed by dust. The solidline represents low-redshift starburst regions from (24 objectsTable 3total). The dashed line represents high-redshift starburst regions from

et al. 23 objects total).Meurer (1997 ;

indication that a systematically larger fraction of ultravioletluminosity emerges unobscured from the high-redshift star-burst regions.

Recall that the obscuration deduced for the high-redshiftsample is based only on the ultraviolet spectra and couldnot account for completely obscured stars. The distributionfor high redshift, therefore, actually shows upper limits forthe escaping fraction of ultraviolet luminosity ; the real frac-tion could be smaller if there are undetected stars because ofobscuration. Even so, these upper limits are comparable tothe values for the low-redshift sample, which do account forcompletely obscured stars. Any systematic di†erencebetween the samples arising from the di†erent methodsused to estimate obscuration would have underestimatedobscuration for the high-redshift starburst regions. This iscounter to the possible explanation being considered, whichrequires that high-redshift starburst regions have less, notmore, obscuration compared with low-redshift regions.

The conclusions above indicate that di†ering dustobscuration does not account for the systematically di†er-ent surface brightnesses between low-redshift and high-redshift starburst regions. This is not, however, fullyconsistent with results reported by et al.Pettini (1998).Although they do not present an analysis of individualspectra, Pettini et al. conclude that a higher fraction ofultraviolet luminosity escapes in the high-redshift galaxiesthan was concluded by et al. They report aMeurer (1997).median dust correction to the ultraviolet luminosity of afactor of 3, although they state it could range from 2 to 6.This dust correction is substantially less than the factor ofapproximately 10 shown by the distributions in IfFigure 3.it is correct that as much as one-third of intrinsic ultravioletluminosity characteristically escapes from high-redshiftstarburst regions, this would enhance their surface bright-ness almost enough to account for our observed di†erencebetween surface brightnesses of low-redshift and high-redshift regions (factor of 4 in the median). We note,however, that the Pettini et al. results also can only describean upper limit for escaping ultraviolet luminosity(equivalent to a lower limit to the obscuration factor)because of the absence of any information on bolometricluminosities.

4. SUMMARY

By examining available ultraviolet observations of low-redshift starburst regions, we deÐne an upper limit to thesurface brightness of such starburst regions as observed onscales of several hundred parsecs. This limit is 6] 10~16ergs cm~2 s~1 arcsec~2 for rest-frame wavelength ofA� ~11830 Ultraviolet surface brightnesses at comparableA� .scales for high-redshift starburst regions (z[ 2.2) are mea-sured from galaxies in the Hubble Deep Field. After correct-ing for cosmological e†ects, we Ðnd that high-redshiftstarbursts have intrinsic ultraviolet surface brightness thatis typically 4 times brighter than for low-redshift starbursts.

Because of di†ering conclusions about the amount ofobscuration in high-redshift starburst regions, we are leftwith an ambiguous interpretation as to whether the unex-pectedly high surface brightness of starburst regions at highredshift is caused by diminished dust obscuration. If not,some e†ect is required to make high-redshift star formationintrinsically more intense per unit area than star formationin the nearby universe. Either way, our results indicate thatsomething was di†erent on a local scale within starburst

No. 4, 1998 SURFACE BRIGHTNESS OF STARBURSTS 1649

regions at redshifts above 2 compared with those in nearbygalaxies, making star formation appear brighter per unitarea of a starburst region.

This increased surface brightness may be a signiÐcantfactor in explaining the increasing luminosity density withredshift observed for ultraviolet-bright galaxies. The dis-tribution with redshift of star-forming galaxies observed inground-based surveys and in the HDF indicates that thestar formation rate per comoving volume peaks at redshiftabout unity and remains 5È10 times greater than in thenearby universe for z[ 2 et al. et al.(Madau 1996 ; Lilly

et al. If there is no change in the size1996 ; Connolly 1997).or comoving density of starburst regions, but the star for-mation intensity per region increases by a factor of 4, thiswould be a major factor in explaining the observed di†er-

ences between the nearby universe and the high-redshiftuniverse. In this scenario, the early universe did not requiresigniÐcantly more starburst regions than are observed in thenearby universe, but each region had much more intense (ormuch less obscured) star formation. Our result is not ade-quate to prove such a scenario, because this would requirethat the surface brightness enhancement apply to all star-burst regions, whereas our result is able to show thisenhancement for only the brightest regions in both nearbyand high-redshift samples.

D. P. S. acknowledges support from NSF grant AST95-09919. M. A. B. acknowledges support from NASA grantNAG 5-6032.

REFERENCESD., Bohlin, R. C., Kinney, A. L., Storchi-Bergmann, T., &Calzetti,

Heckman, T. M. 1995, ApJ, 443, 136J. G., Cowie, L. L., Hogg, D. W., Songaila, A., Blandford, R., Hu,Cohen,

E. M., & Shopbell, P. 1996, ApJ, 471, L5A. J., Szalay, A. S., Dickinson, M., SubbaRao, M. U., & Brunner,Connolly,

R. J. 1997, ApJ, 486, L11R. D., Sramek, R. A., & Weedman, D. W. 1983, ApJ, 267,Gehrz, 551A. L., Bohlin, R. C., Calzetti, D., Panagia, N., & Wyse, R. F. G.Kinney,

1993, ApJS, 86, 5S. J., Le Fevre, O., Hammer, F., & Crampton, D. 1996, ApJ, 460,Lilly, L1

J. D., et al. 1997, ApJ, 481,Lowenthal, 673P., Ferguson, H. C., Dickinson, M. E., Giavalisco, M., Steidel,Madau,

C. C., & Fruchter, A. 1996, MNRAS, 283, 1388D., Filippenko, A. V., Ho, L. C., Macchetto, F. D., Rix, H.-W., &Maoz,

Schneider, D. P. 1996, ApJS, 107, 215G. R., Heckman, T. M., Leitherer, C., Kinney, A., Robert, C., &Meurer,

Garnett, D. R. 1995, AJ, 110, 2665G., Heckman, T., Leitherer, C., Lowenthal, J., & Lehnert, M. 1997,Meurer,

AJ, 114, 54

S. G., Fanelli, M. N., Roberts, L. J., OÏConnell, R. W., Bohlin, R.,Ne†,Roberts, M. S., Smith, A. M., & Stecher, T. P. 1994, ApJ, 430, 545

M., Steidel, C. C., Dickinson, M., Kellogg, M., Giavalisco, M., &Pettini,Adelberger, K. L. 1998, in AIP Conf. Proc. 408, Ultraviolet Universe atLow and High Redshift, ed. W. Waller (New York : AIP), 279

M., et al. 1997, MNRAS, 289,Rowan-Robinson, 490H. R., Kinney, A. L., Calzetti, D., & Storchi-Bergmann, T. 1997,Schmitt,

AJ, 114, 592D. A., Herter, T., & Haynes, M. P. 1998, ApJ, 494,Smith, 150C. C., Giavalisco, M., Dickinson, M., & Adelberger, K. L. 1996, AJ,Steidel,

112, 352D. W. 1991, in Massive Stars in Starbursts, ed. C. Leitherer,Weedman,

N. R. Walborn, T. M. Heckman, & C. A. Norman (Cambridge : Cam-bridge Univ. Press), 317

R. E., et al. 1996, AJ, 112,Williams, 1335S. E., Moustakas, L. A., & Davis, M. 1997, ApJ, 474,Zepf, L1