0709.0520/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: 
3: %===========================================================================
4: % NPH defined commands
5: \newcommand{\ii}{$i'$}
6: \newcommand{\zz}{$z'$}
7: \newcommand{\uu}{$U$}
8: \newcommand{\bb}{$B$}
9: \newcommand{\vv}{$V$}
10: \newcommand{\jj}{$J$}
11: \newcommand{\hh}{$H$}
12: \newcommand{\gw}{$G$}
13: \newcommand{\rw}{$R$}
14: \newcommand{\iw}{$I$}
15: \newcommand{\ts}{\thinspace}
16: \newcommand{\iz}{\ensuremath{(i'\!-\!z')}}
17: \newcommand{\gi}{\ensuremath{(g\!-\!i')}}
18: \newcommand{\zj}{\ensuremath{(z'-J)}}
19: \newcommand{\etal}{{et\thinspace al.}}
20: \newcommand{\cg}{c.\thinspace g.}
21: \newcommand{\Ang}{\AA\thinspace}
22: \newcommand{\SN}{$S/N$}
23: \newcommand{\Ho}{$H_{0}$}
24: \newcommand{\super}[1]{$^{#1}$}
25: \newcommand{\sub}[1]{$_{#1}$}
26: \newcommand{\tabref}[1]{Table~\ref{#1}}
27: \newcommand{\figref}[1]{Figure~\ref{#1}}
28: \newcommand{\secref}[1]{\S~\ref{#1}}
29: 
30: %===========================================================================
31: 
32: \begin{document}
33: 
34: \title{Starburst Intensity Limit of Galaxies at $z\!\simeq\!5\!-\!6$}
35: 
36: \shorttitle{Starburst Intensity Limit at $z\!\simeq\!5\!-\!6$}
37: 
38: \author{N. P. Hathi\altaffilmark{1}, S. Malhotra\altaffilmark{1,2} and J. E. Rhoads\altaffilmark{1,2}}
39: 
40: \altaffiltext{1}{Department of Physics, Arizona State University, Tempe, AZ 85287-1504, USA}
41: 
42: \altaffiltext{2}{School of Earth and Space Exploration, Arizona State University, Tempe, AZ 85287-1404, USA}
43: 
44: \email{Nimish.Hathi@asu.edu}
45: \shortauthors{Hathi et al}
46: 
47: %--------------------------------------------------
48: 
49: \begin{abstract}
50: The peak star formation intensity  in starburst galaxies does not vary
51: significantly from  the local  universe to redshift  $z\!\sim\!6$.  We
52: arrive at this conclusion  through new surface brightness measurements
53: of  47  starburst   galaxies  at  $z\!\simeq\!5\!-\!6$,  doubling  the
54: redshift   range   for   such   observations.   These   galaxies   are
55: spectroscopically  confirmed in  the  Hubble Ultra  Deep Field  (HUDF)
56: through  the  GRism ACS  program  for  Extragalactic Science  (GRAPES)
57: project.    The   starburst    intensity   limit   for   galaxies   at
58: $z\!\simeq\!5\!-\!6$  agree  with  those at  $z\!\simeq\!3\!-\!4$  and
59: $z\!\simeq\!0$  to within  a factor  of  a few,  after correcting  for
60: cosmological  surface  brightness  dimming  and for  dust.   The  most
61: natural interpretation of this constancy  over cosmic time is that the
62: same physical mechanisms limit starburst intensity at all redshifts up
63: to $z\!\simeq\!6$ (be  they galactic winds, gravitational instability,
64: or something else).  We do see two trends with redshift: First, the UV
65: spectral slope ($\beta$) of  galaxies at $z\!\simeq\!5\!-\!6$ is bluer
66: than that  of $z\!\simeq\!3$ galaxies, suggesting an  increase in dust
67: content over  time.  Second, the  galaxy sizes from  $z\!\simeq\!3$ to
68: $z\!\simeq\!6$   scale   approximately   as   the   Hubble   parameter
69: $H^{-1}(z)$.   Thus,  galaxies  at  $z\!\simeq\!6$ are  high  redshift
70: starbursts, much  like their local  analogs except for  slightly bluer
71: colors,  smaller  physical sizes,  and  correspondingly lower  overall
72: luminosities.   If we  now assume  a constant  maximum  star formation
73: intensity,  the  differences in  observed  surface brightness  between
74: $z\!\simeq\!0$  and   $z\!\simeq\!6$  are  consistent   with  standard
75: expanding cosmology and strongly inconsistent with tired light model.
76: %
77: \end{abstract}
78: 
79: \keywords{galaxies: high redshift --- galaxies: starburst}
80: 
81: %--------------------------------------------------
82: 
83: \section{Introduction}\label{introduction}
84: 
85: Star formation on galactic scales is a key ingredient in understanding
86: galaxy evolution.  We  cannot compare structure formation calculations
87: to  observed  galaxy  populations  without  some model  for  how  star
88: formation proceeds.  Such models are based on detailed observations in
89: the nearby  universe, combined  with physically motivated  scaling for
90: differing conditions elsewhere in  the universe.  To test the validity
91: of such scaling, it is  valuable to directly measure the properties of
92: star  formation events  in  the  distant universe,  and  see how  they
93: compare with their nearby counterparts.
94: 
95: Starbursts  are regions  of intense  massive star  formation  that can
96: dominate a galaxy's integrated  spectrum.  By comparing the properties
97: of starbursts over a wide range  of redshifts, we can test whether the
98: most  intense  star formation  events  look  the  same throughout  the
99: observable  history  of  the  universe.  High  redshift  galaxies  are
100: expected, on average, to be  less massive and lower in metal abundance
101: than their present-day counterparts.  Either effect could in principle
102: change the maximum intensity of  star formation that such galaxies can
103: sustain.
104:    
105: \citet{meur97}   (hereafter  M97)   measured  the   effective  surface
106: brightness, i.e.,  the average  surface brightness within  an aperture
107: that encompasses half  of the total light, for  various samples.  They
108: conclude that  the maximum  effective surface brightness  of starburst
109: galaxies  is unchanged to  better than  an order  of magnitude  out to
110: redshifts  $z\!\simeq\!3$.   \citet{weed98}  (hereafter W98)  measured
111: observed  surface  brightness  from  the single  brightest  pixel  and
112: concluded that  high-redshift (2.2$\la\!z\!\la$3.5) starburst galaxies
113: have  intrinsic  ultra-violet   (UV)  surface  brightnesses  that  are
114: typically  4 times  higher than  for low-redshift  starburst galaxies.
115: Both   M97   and   W98   measured   their   surface   brightness   for
116: spectroscopically  confirmed   galaxies  in  the   Hubble  Deep  Field
117: \citep[HDF;][]{will96}.
118: 
119: We  have  measured the  surface  brightness  of  starburst regions  at
120: $3\!\la\!z\!\la\!6$, using photometry from the Hubble Ultra Deep Field
121: (HUDF) images \citep{beck06} and  redshifts from the GRism ACS Program
122: for    Extragalactic   Science    (GRAPES)   project    \citep[PI   S.
123: Malhotra;][]{pirz04}.  We combine these with earlier published results
124: comparing  $z\!\simeq\!3$ and  $z\!\simeq\!0$  starbursts (M97,  W98).
125: The    starburst   intensity   limit    of   starburst    regions   at
126: $5\!\la\!z\!\la\!6$ sample  is consistent with  that at $z\!\simeq\!3$
127: and  $z\!\simeq\!0$ to  within the  uncertainties, which  are  about a
128: factor of three.  \emph {These  high redshift star forming regions are
129:   thus starbursts,  with a star  formation intensity similar  to their
130:   local  counterparts  despite any  effects  of differing  metallicity
131:   and/or  galaxy  size.}  The  starbursts  should  then  be a  set  of
132: standard surface brightness objects, and can be used to apply Tolman's
133: test  for  expansion of  the  universe  \citep{tolm30,tolm34} over  an
134: unprecedented   redshift  range   ($0\!<\!z\!\la\!6$).    Our  surface
135: brightness  observations  fully  support standard  expanding  universe
136: models.  This result is robust to even rather large systematic errors,
137: thanks to the wide redshift range spanned by the data.
138: 
139: This paper is organized as  follows: In \secref{data} we summarize the
140: HUDF and the GRAPES observations, and we present details of the sample
141: selection.  In  \secref{m97} we describe our data  analysis to measure
142: the  UV  spectral  slope  and  the  effective  surface  brightness  of
143: starburst galaxies  following the method used by  M97, in \secref{w98}
144: we  apply the  pixel  based method  of  W98 to  estimate the  limiting
145: surface brightness  for our  galaxies, in \secref{results}  we discuss
146: measurement   biases   that  will   affect   our  surface   brightness
147: measurements, and  our results of  the starburst intensity  limit, the
148: size evolution  \& the change in  the UV spectral  slope.  Finally, in
149: \secref{summary} we summarize our results.
150: 
151: Throughout this paper we denote the \emph{HST}/ACS F435W, F606W, F775W
152: and F850LP  filters as \bb,  \vv, \ii, \zz, \emph{HST}/WFPC2  F814W as
153: \iw\ts and  \emph{HST}/NIC3 F110W and  F160W as \jj\ts  and \hh-bands,
154: respectively.  We assume a \emph{Wilkinson Microwave Anisotropy Probe}
155: (WMAP)  cosmology   of  $\Omega_m$=0.24,  $\Omega_{\Lambda}$=0.76  and
156: \Ho=73~km~s$^{-1}$~Mpc$^{-1}$, in  accord with the recent  3 year WMAP
157: estimates of \citet{sper07}.  This implies  an age for the Universe of
158: 13.7~Gyr.  Magnitudes are given in the AB system \citep{oke83}.
159: 
160: %--------------------------------------------------
161: 
162: \section{Observational  Data and Sample Selection}\label{data}
163: 
164: The HUDF is a 400 orbit survey of a $3.4'\times3.4'$ field carried out
165: with   the  ACS   in  the   \bb,  \vv,   \ii\ts  and   \zz\ts  filters
166: \citep[see][for  further details]{beck06}.  We  have carried  out deep
167: unbiased slitless  spectroscopy of  this field with  the ACS  grism as
168: part of  the GRAPES  project, which was  awarded 40 HST  orbits during
169: Cycle  12 (ID  9793; PI  S.  Malhotra).   The grism  observations were
170: taken at four different orientations  in order to minimize the spectra
171: contamination and overlapping from  nearby sources.  We have extracted
172: useful low resolution spectra ($R\simeq100$) from 5900\Ang to 9500\Ang
173: for many sources in the HUDF  down to a limiting magnitude of $z'_{\rm
174: AB}\!\simeq$27.5 in the AB  system.  Details of the observations, data
175: reduction  and  final GRAPES  catalog  are  described  in a  paper  by
176: \citet{pirz04}.
177: 
178: We identify  high-redshift galaxies  on the basis  of their  ACS grism
179: spectra.  This  identification was based on detecting  the Lyman break
180: in the  continuum or the  Ly$\alpha$ emission for these  sources. With
181: the ACS  grism low-resolution  spectra, we are  able to  determine the
182: redshifts  to an  accuracy  of $\Delta  z\!\approx$0.15  even for  the
183: faintest   detectable   Lyman    Break   Galaxies   (LBGs)   ($z'_{\rm
184:   AB}\!\simeq$27.5).  Details  of the selection  process are described
185: in   \citet{malh05}.    There   are   47  star-forming   galaxies   at
186: $z\!\simeq\!5\!-\!6$ in  the GRAPES/HUDF with  confirmed spectroscopic
187: redshifts.  These redshifts are sufficiently high that rest-frame flux
188: measurements are  available for UV wavelengths comparable  to those at
189: which M97 and W98 measured the surface brightness of starburst regions
190: at   $z\!\lesssim\!3$.   We   used  the   following  spectroscopically
191: confirmed samples for our analysis.
192: \begin{itemize}
193: 
194: \item   The   $z\!\simeq\!3$  LBGs   from   the   Hubble  Deep   Field
195: \citep[HDF,][]{giav96,stei96a,stei96b}   were  used  to   compare  our
196: measurements with  M97.  This sample of 10  galaxies at $z\!\simeq\!3$
197: forms  the subset  of UV  samples  used by  M97.  We  measured the  UV
198: spectral  slopes ($\beta$)  using observed  (\vv--\iw)  and ($G$--$R$)
199: colors,  and  rest-frame UV  fluxes  were  derived  from the  observed
200: $R$-band   (combined   \vv\ts    and   \iw\ts\ts   light)   magnitudes
201: \citep{giav96,stei96a,stei96b}.
202: 
203: \item The $z\!\simeq\!4$ \bb-band  dropout galaxies were selected from
204: the VLT redshift  catalog of \citet{vanz06}.  This sample  is small (4
205: galaxies),  but  is  useful  for  comparing  $z\!\simeq\!3\!-\!4$  and
206: $z\!\simeq\!5\!-\!6$ galaxy samples.   We used the observed (\ii--\zz)
207: color to estimate $\beta$, and  rest-frame UV fluxes were derived from
208: the  observed \zz-band  magnitudes.   The observed  \ii- and  \zz-band
209: magnitudes  for $z\!\simeq\!4$  galaxies were  obtained from  the HUDF
210: catalogs of \citet{beck06}.
211: 
212: \item We  use a sample  of 47 $z\!\simeq\!5\!-\!6$  starburst galaxies
213: having GRAPES redshifts (Malhotra et  al. 2005; Rhoads et al. 2007, in
214: preparation) in the HUDF. \tabref{tab1} shows properties (coordinates,
215: magnitudes,  sizes and redshifts)  for these  47 galaxies.  We further
216: selected a  subset of 19  galaxies covered by the  \citet{thom05} HUDF
217: NICMOS  images.   We estimate  $\beta$  from  the observed  (\jj--\hh)
218: color.  The rest-frame UV fluxes were derived from the observed \jj\ts
219: and \hh-band magnitudes for 19  galaxies, while the average $\beta$ is
220: used to  predict \jj\ts and  \hh-band magnitudes for the  remaining 28
221: galaxies.
222: \end{itemize}
223: %--------------------------------------------------
224: 
225: \section{Starburst Intensity Limit Using M97 Approach}\label{m97}
226: 
227: %--------------------------------------------------
228: 
229: \subsection{Magnitudes and Color Measurements}\label{mags}
230: 
231: All measurements for galaxies at $z\!\simeq\!4\!-\!6$ were done on the
232: HST   ACS   HUDF  images   \citep{beck06}   and   HST  NICMOS   images
233: \citep{thom05}.  The  HST NICMOS images  were reprocessed by  L. Eddie
234: Bergeron  (private  communication).    The  HST  NICMOS  images  cover
235: $\sim$50\% of the HUDF ACS field. Therefore, only half of our galaxies
236: have \jj- and \hh-band imaging.   We need \jj\ts and \hh\ts magnitudes
237: for galaxies at $z\!\simeq\!5\!-\!6$ to measure the UV spectral slopes
238: and  luminosities at  $\sim$2200\Ang\ rest-frame.   We  use rest-frame
239: $\sim$2200\Ang\ts  to  measure the  effective  surface brightness  for
240: consistency  with  M97.  All  broad-band  magnitudes  are measured  as
241: \texttt{SExtractor}   \citep{bert96}   \texttt{MAG\_AUTO}  magnitudes,
242: using dual-image  mode to generate aperture matched  catalogs.  We use
243: the  \zz-band  image   as  the  detection  image.   \texttt{MAG\_AUTO}
244: apertures  are Kron-like  \citep{kron80} flexible  apertures  and they
245: enclose most of  the flux for an object.  These  apertures are same in
246: all filters for a  given object.  We also measured \texttt{SExtractor}
247: \citet{petr76}   magnitudes  (\texttt{MAG\_PETRO}),   with  $\eta$=0.2
248: \citep{holw05}, and isophotal magnitudes (\texttt{MAG\_ISO}).  We find
249: that   the   average   difference   between   \texttt{MAG\_AUTO}   and
250: \texttt{MAG\_PETRO}  is $\sim$0.1  mag, while  the  average difference
251: between  \texttt{MAG\_AUTO} and  \texttt{MAG\_ISO}  is $\sim$0.2  mag.
252: The  effect of this  magnitude uncertainty  on the  surface brightness
253: measurements  is  very small  ($<$0.1  dex).   Our  method to  measure
254: magnitudes is  the same for all galaxies  at $z\!\simeq\!4\!-\!6$, and
255: is consistent with the  curve-of-growth method used for our comparison
256: sample of  $z\!\simeq\!3$ galaxies \citep{giav96,stei96a,stei96b}.  We
257: use \texttt{MAG\_AUTO}  magnitudes to calculate the  UV spectral slope
258: ($\beta$) and the effective surface brightness.
259: 
260: \subsection{The UV Spectral Slope (\boldmath{$\beta$})}\label{beta}
261: 
262: The UV  spectral slope ($\beta$) is  determined from a  power-law fit to
263: the UV continuum spectrum \citep{calz94},
264: \[ 
265: f_{\lambda} \varpropto \lambda^{\beta} ~~,
266: \]
267: where  $f_{\lambda}$ is  the flux  density per  unit  wavelength (ergs
268: s$^{-1}$ cm$^{-2}$ \Ang$^{-1}$).  Converting this into magnitude units
269: yields   a   linear   relationship   between   $\beta$   and   colors.
270: \figref{fig1}  shows the  average values  of the  UV  spectral slopes,
271: $\beta$,     for      galaxies     at     $z\!\simeq\!3\!-\!4$     and
272: $z\!\simeq\!5\!-\!6$.   The mean  values are  plotted with  error bars
273: indicating  the  standard deviation  of  the  mean  (i.e.  the  sample
274: standard deviation ($\sigma$) divided by the square root of the sample
275: size  ($N$)).  The  UV slopes  for 10  galaxies at  $z\!\simeq\!3$ are
276: obtained  using the  following two  equations (M97)  for  two slightly
277: different samples:
278: \[ 
279: \beta = 2.55 \cdot (G - R) - 2  \; \; {\rm and} \; \; \beta = 3.23 \cdot (V - I) - 2  ~~.
280: \]
281: Here $G$ and $R$ filters  are defined in \citet{stei93}.  We also plot
282: the average $\beta$ measured for 4 galaxies at $z\!\simeq\!4$, using:
283: \[
284: \beta =  5.65 \cdot (i' - z') - 2
285: \]
286: where  we   have  used  pivot  wavelengths  for   \ii-  and  \zz-band,
287: $\lambda_i$=7693\Ang  and  $\lambda_{z'}$=9055\Ang,  respectively,  to
288: obtain  the  slope  of  5.65 in  $\beta$--color  linear  relationship.
289: \figref{fig1}  also  shows the  average  $\beta$  for  19 galaxies  at
290: $z\!\simeq\!5\!-\!6$ obtained using
291: \[ 
292: \beta = 2.56 \cdot (J - H) - 2 
293: \]
294: where  we   have  used  pivot  wavelengths  for   \jj-  and  \hh-band,
295: $\lambda_J$=11200\Ang\ts  and $\lambda_H$=16040\Ang,  respectively, to
296: obtain the slope of 2.56 in $\beta$--color linear relationship.  We do
297: not  use  (\zz--\jj)  color   to  estimate  $\beta$  for  galaxies  at
298: $z\!\simeq\!5\!-\!6$ because, the (\zz--\jj) colors can be insensitive
299: to rest-frame UV colors due to the shorter color baseline and are also
300: more sensitive to uncertainties in the optical to infrared zero points
301: \citep{bouw06}.   Therefore for comparison,  we also  measured $\beta$
302: for galaxies at $z\!\simeq\!5$  using (\ii--\zz) colors and found that
303: the  average $\beta$ is  --1.53$\pm$0.38 compared  to --1.65$\pm$0.21,
304: the average  $\beta$ at $z\!\simeq\!5$ using  (\jj--\hh) colors.  This
305: comparison  implies  that  small   variations  in  the  UV  rest-frame
306: wavelength and  observed colors does  not affect the slope  within the
307: quoted uncertainties.
308: 
309: \figref{fig1}   shows  that   the  average   $\beta$   decreases  from
310: --1.13$\pm$0.17    at    $z\!\simeq\!3$    to    --1.74$\pm$0.35    at
311: $z\!\simeq\!6$.   The  change in  the  average  slope  shows that  the
312: galaxies   at   $z\!\simeq\!5\!-\!6$   are   bluer   than   those   at
313: $z\!\simeq\!3$,   but  redder   than   the  flat   slope  in   $f_\nu$
314: ($f_{\lambda}  \propto \lambda^{-2}$),  that would  be expected  for a
315: dust-free   starburst   galaxy. 
316: 
317: %--------------------------------------------------
318: 
319: \subsection{Half-Light Radius (\boldmath{$r_e$}) Measurements}
320: 
321: The half-light radius is defined  as the radius containing 50\% of the
322: total  flux  of an  object.   The  half-light  radii for  galaxies  at
323: $z\!\simeq\!3$  are derived by  \citet{giav96} \&  \citet{stei96b} and
324: are measured such  that half of the total  emission from the starburst
325: is   enclosed    within   a   circular   aperture.     We   used   the
326: \texttt{SExtractor} half-light radii (R50, radii enclosing 50\% of the
327: flux  within a  circular  aperture) obtained  from  the HUDF  \zz-band
328: catalog  \citep{beck06}  for  galaxies at  $z\!\simeq\!4\!-\!6$.   All
329: radii are converted from arcsecs to  kpc using a WMAP cosmology in the
330: cosmological calculator  by \citet{wrig06}.  The  half-light radii for
331: galaxies  at  $z\!\simeq\!4\!-\!6$   are  measured  at  rest-frame  UV
332: wavelengths  $\lambda$=1500$\pm$300\Ang.   The  sizes  do  not  change
333: appreciably when measured within this wavelength range.  \figref{fig2}
334: shows  the average  values of  the half-light  radii for  10 starburst
335: galaxies  at $z\!\simeq\!3$, 4  starburst galaxies  at $z\!\simeq\!4$,
336: and  47 starburst galaxies  at $z\!\simeq\!5\!-\!6$.   Mean half-light
337: radii are plotted with error bars indicating the standard deviation of
338: the mean.  The solid and dashed  curves show the trend if sizes evolve
339: as  $H^{-1}(z)$ or  $H^{-2/3}(z)$, respectively,  where $H(z)$  is the
340: Hubble parameter  at redshift $z$.   The curves are normalized  to the
341: mean  size  we  measure  at  $z\!\simeq\!4$  ($\sim$0.21\arcsec\ts  or
342: $\sim$1.5  kpc).  Comparison  between galaxies  at  $z\!\simeq\!3$ and
343: $z\!\simeq\!5\!-\!6$  shows that the  galaxy sizes  increase as  we go
344: from $z\!\simeq\!6$ to $z\!\simeq\!3$.
345: 
346: We independently measured various flavors of \texttt{SExtractor} radii
347: (half-light,     \citet{petr76},     Kron)     for     galaxies     at
348: $z\!\simeq\!5\!-\!6$ to assess  the differences in these measurements.
349: The Petrosian  radius is  defined as the  radius at which  the surface
350: brightness  is   certain  factor  ($\eta$)  of   the  average  surface
351: brightness within this  isophote, while the Kron radius  is defined as
352: the typical  size of the  flexible aperture computed from  the moments
353: \citep[see     \texttt{SExtractor}      manual     by][for     further
354: details]{holw05}.   The  difference  in  the  average  values  of  the
355: half-light    and   the    other   two    radii    was   approximately
356: $\pm$0.04\arcsec. Therefore,  we expect about 30\%  uncertainty in the
357: measurements    of   the    half-light   radii    for    galaxies   at
358: $z\!\simeq\!5\!-\!6$.
359: 
360: %--------------------------------------------------
361: 
362: \subsection{Calculation of Surface Brightness for Starburst Galaxies}\label{calculations}
363: 
364: We measure the effective  surface brightness for 14 starburst galaxies
365: at    $z\!\simeq\!3\!-\!4$    and    47    starburst    galaxies    at
366: $z\!\simeq\!5\!-\!6$ by  adopting the method  used by M97.   First, we
367: need to  estimate dust  extinction A$_{\rm 1600}$  and $k$-corrections
368: from the  UV spectral slope  ($\beta$).  Dust extinction  is estimated
369: using the linear empirical relation between A$_{\rm 1600}$ and $\beta$
370: \citep{meur99}, which is given by following equation:
371: \[	
372: {\rm A}_{\rm 1600}=4.43+1.99 \cdot (\beta) 
373: \]
374: where A$_{\rm  1600}$ is the net  absorption in magnitudes  by dust at
375: 1600\Ang.
376: 
377: The next step in correcting apparent flux for corresponding rest-frame
378: UV flux  is to  apply, where appropriate,  the $k$-correction.  We use
379: following equation to estimate $k$-correction (M97):
380: \[ 
381: k = \frac{f_{\rm 2320}}{f_{\lambda_c/1+z}} = \left [\frac{(1+z) \cdot 2320 {\rm \Ang}}{\lambda_c} \right ]^\beta 
382: \]
383: where $\lambda_c$ corresponds to the central wavelength of the filters
384: used for the observed flux.  Here we reference all observations to the
385: observations in M97, which has  UV central wavelength of 2320\Ang.  We
386: apply above mentioned dust  and $k$-corrections to apparent magnitudes
387: to estimate absolute magnitudes  and intrinsic UV luminosities for our
388: samples of  $z\!\simeq\!3\!-\!6$ galaxies.  The ratio  of intrinsic UV
389: to bolometric  luminosity can be calculated for  young starbursts.  By
390: using the stellar population  models of \citet{bruz03} to convert from
391: intrinsic F220W flux/luminosity to bolometric luminosity, we find that
392: the ratio of the UV to bolometric luminosity changes due to variations
393: in  the metallicity and  the dust  attenuation.  The  luminosity ratio
394: spans a range  from $\sim$0.2 to $\sim$0.5.  The  adopted ratio of the
395: intrinsic UV to bolometric luminosity is:
396: \[
397: \frac{L}{L_{\rm bol}} \simeq 0.33
398: \]	
399: We are using this value for  the UV to bolometric luminosity ratio for
400: two  reasons: (1)  this bolometric  correction  is very  close to  the
401: average value we  get from our stellar population  models and (2) this
402: correction factor is also predicted by the models used by M97.
403: 
404: From  $L_{\rm bol}$  and the  half-light radii  ($r_e$),  we calculate
405: effective surface brightness (S$_{e}$) using following relation:
406: \[
407: {\rm S}_e = \frac{L_{\rm bol}}{2 \pi r_e^2}  \; \; \;  \left (\frac{L_{\sun}}{{\rm kpc}^2} \right ) ~~.
408: \]	
409: Here $r_e$ is measured in  kpc and $L_{\rm bol}$ in solar luminosities
410: ($L_{\sun}$).
411: 
412: To  characterize the  S$_{e}$  distribution for  all  galaxies in  our
413: sample  we consider  the median  and 90$^{th}$  percentiles,  which we
414: denote as S$_{e,\rm 50}$  and S$_{e,\rm 90}$, respectively.  The upper
415: limit  to  the  surface  brightness  (starburst  intensity  limit)  of
416: starbursts is  traced by  S$_{e,\rm 90}$.  Here  we have  used average
417: $\beta$ (from  \figref{fig1}) for each sample to  estimate the surface
418: brightness.  \figref{fig3} shows $L_{\rm bol}$ and S$_e$ as a function
419: of   $r_e$   for  the   $z\!\simeq\!0\!-\!6$   galaxy  samples.    The
420: $z\!\simeq\!0$  and $z\!\simeq\!0.4$  surface  brightness measurements
421: are  taken from  M97.  The  $z\!\simeq\!0$  data point  is the  median
422: measurement for 11 nearby  galaxies.  The S$_{e,\rm 50}$ and S$_{e,\rm
423:   90}$ surface brightness levels of the combined sample are plotted as
424: dashed and dotted lines respectively.
425: 
426: The  top  panel  of   the  \figref{fig3}  shows  that  the  bolometric
427: luminosities for galaxies at $z\!\simeq\!5\!-\!6$ are smaller than the
428: luminosities  of galaxies  at $z\!\simeq\!3$.   Using a  two-sided K-S
429: test on these luminosity  distributions, we reject the hypothesis that
430: the  $z\!\simeq\!3$ and  $z\!\simeq\!5\!-\!6$  luminosities are  drawn
431: from  the same  population at  $>$99\% probability.   From  the bottom
432: panel of \figref{fig3},  it is apparent that S$_e$  shows little or no
433: dependence on $r_e$  over about one order of  magnitude in size; hence
434: there  is no  dependence on  $L_{\rm bol}$  over about  two  orders of
435: magnitude  in luminosity.  \figref{fig4}  shows the  effective surface
436: brightness (S$_{e}$)  as a  function of redshift.   The $z\!\simeq\!0$
437: and  $z\!\simeq\!0.4$ surface brightness  measurements are  taken from
438: M97.   From \figref{fig3}  and \figref{fig4},  we find  that S$_{e,\rm
439:   90}$ of the starbursts  remains constant (within uncertainties) with
440: redshift.
441: 
442: %--------------------------------------------------
443: 
444: \section{Starburst Intensity Limit Using W98 Approach}\label{w98}
445: 
446: We  also studied the  surface brightnesses  using the  brightest pixel
447: approach pioneered by \citet{weed98}. \citet{weed98} measures observed
448: surface   brightness  of   the   brightest  pixel   for  galaxies   at
449: $2.2\!\la\!z\!\la\!3.5$  in  the  HDF   and  compared  with  the  local
450: starbursts by  fading their observed  ($f_\lambda$) surface brightness
451: by  (1+$z$)$^{-5}$.    We  use   this  approach  to   compare  surface
452: brightnesses of the brightest pixel for galaxies at $z\simeq\!3\!-\!6$.
453: 
454: A   postage   stamp  (51$\times$51   pixels)   for   each  galaxy   at
455: $z\simeq5\!-\!6$  was excised  from  the \zz-band  HUDF image.   Using
456: \zz-band  segmentation maps,  only object  pixels were  selected.  For
457: each  object pixel  in the  postage  stamp, we  estimate the  apparent
458: magnitude.  For  the brightest  pixel in each  galaxy, the  average UV
459: spectral slope $\beta$  was used to predict \jj-band  (for galaxies at
460: $z\!\simeq\!5$)   and  \hh-band   (for  galaxies   at  $z\!\simeq\!6$)
461: magnitudes from \zz-band  apparent magnitudes.  Using similar approach
462: as discussed in \secref{calculations},  we calculated the intrinsic UV
463: luminosity  ($L_{\sun}^{\rm  UV}$) for  the  brightest  pixel in  each
464: galaxy at $z\!\simeq\!5\!-\!6$.   For surface brightness measurements,
465: we divide  the intrinsic  UV luminosity  by the area  of one  pixel in
466: kpc$^2$.
467: 
468: \figref{fig5}     shows     surface     brightness     ($L_{\sun}^{\rm
469: UV}$~kpc$^{-2}$) for the brightest pixel in each of the 47 galaxies at
470: $z\!\simeq\!5\!-\!6$ and  4 galaxies at $z\!\simeq\!4$.   We have also
471: plotted 18 galaxies at $z\!\simeq\!3$  from W98.  The W98 galaxies has
472: observed surface brightnesses for  the brightest pixels and hence, for
473: proper comparison, we converted  observed surface brightnesses for W98
474: galaxies to their corresponding rest-frame surface brightnesses.  Here
475: we have used  average $\beta$ for $z\!\simeq\!3$ galaxies  as shown in
476: \figref{fig1}, to estimate  extinction and $k$-correction.  The median
477: (S$_{bp,\rm 50}$)  and 90$^{th}$ percentile  (S$_{bp,\rm 90}$) surface
478: brightness levels  of the  combined sample are  plotted as  dashed and
479: dotted lines,  respectively.  Here,  the starburst intensity  limit of
480: starbursts  is traced  by  S$_{bp,\rm 90}$.   The  S$_{bp,\rm 90}$  of
481: galaxies at $z\!\simeq\!3$ and  $z\!\simeq\!5\!-\!6$ is same to within
482: a factor of $\sim$0.10 dex.
483: 
484: We cannot  properly compare  the brightest pixel  surface brightnesses
485: between   the   W98  sample   at   $z\!\simeq\!0$   and  galaxies   at
486: $z\!\simeq\!3\!-\!6$ for  two reasons.  First, the aperture  sizes for
487: W98  galaxies at  $z\!\simeq\!0$ are  larger than  the  physical sizes
488: corresponding to one pixel  at $z\!\simeq\!3\!-\!6$, and the sizes for
489: W98  galaxies at  $z\!\simeq\!0$ are  also larger  than  the effective
490: radii  measured by \citet{meur95}  for some  of these  local galaxies.
491: Second,  the discrepancy  between the  W98 adopted  UV  spectral slope
492: ($\beta$) and $\beta$  from \citet{meur95} is large for  some of these
493: galaxies, which  affects the  applied extinction for  their luminosity
494: measurements.
495: 
496: The effective  surface brightness  approach of M97  (\secref{m97}) and
497: the  brightest  pixel approach  of  W98  suggests  that the  starburst
498: intensity limit, as  defined by S$_{e,\rm 90}$ or  S$_{bp,\rm 90}$, of
499: the    starbursts   is    unchanged   (within    uncertainties)   from
500: $z\!\simeq\!5\!-\!6$ down to $z\!\simeq\!3$.
501: 
502: %--------------------------------------------------
503: 
504: \section{Results and Discussion}\label{results}
505: 
506: \subsection{Selection and Measurement Effects}\label{bias}
507: 
508: \emph{Different radii-measurements.}  --- The results shown in Figures
509: 2  and   4  use   \texttt{SExtractor}  half-light  radii   derived  by
510: \citet{beck06}.   Here  we  discuss  the uncertainty  in  the  surface
511: brightness  measurements  due to  radii  measurements.   We use  three
512: different  flavors of  radii measured  using  \texttt{SExtractor}.  We
513: measured the  half-light radii, the \citet{petr76} radii  and the Kron
514: radii for galaxies at $z\!\simeq\!5\!-\!6$.  We found that the average
515: difference  between the  half-light  radius and  other  two radii  was
516: approximately $\pm$0.04\arcsec.   This difference does  not affect our
517: conclusion that $z\!\simeq\!5\!-\!6$  galaxies are smaller compared to
518: $z\!\simeq\!3$ galaxies.  As shown in \figref{fig2}, the average value
519: of  the  half-light  radii  for galaxies  at  $z\!\simeq\!5\!-\!6$  is
520: 0.14\arcsec\ts compared  to 0.32\arcsec\ts which is  the average value
521: for galaxies  at $z\!\simeq\!3$.   The uncertainty in  three different
522: radii  measurements  is   $\pm$0.04\arcsec.   Therefore  for  a  given
523: luminosity, we  expect that this difference in  radii measurement will
524: cause $\sim$0.2 dex difference in S$_{e,\rm 90}$ estimate.
525:  
526: \emph{Surface  brightness   selection.}   ---  The   limiting  surface
527: brightness  in  the  HUDF  samples  is S$_{e}  \sim$  5.0  x  10$^{9}$
528: $L_{\sun}$ kpc$^{-2}$  at $z\!\simeq\!5\!-\!6$, which  is much fainter
529: than the observed maximum surface brightness (S$_{e,\rm 90} \simeq 2.3
530: {\rm \;  x \;} 10^{11}$  $L_{\sun}$ kpc$^{-2}$) for all  galaxies.  At
531: low surface brightness limit our sample could be incomplete but at the
532: maximum surface brightness level our sample is complete.
533: 
534: \emph{Size  selection effect.}   --- The  $z\!\simeq\!3$  galaxies are
535: spectroscopically  confirmed  by   Keck  observations.   The  limiting
536: magnitude for Keck spectroscopy  is $R \lesssim 25.3$ \citep{stei96b}.
537: Using  S$_{e,\rm 90}$  and  the limiting  magnitude,  we estimate  the
538: minimum  observable   size  for  this   sample  as  $\sim$700   pc  or
539: $\sim$0.1\arcsec.    Therefore,    this   sample   of    galaxies   at
540: $z\!\simeq\!3$ is not biased against smaller sizes.
541: 
542: \emph{Flux uncertainty.}   --- Here we discuss  three possible sources
543: of uncertainties in the magnitudes (for a given size) that will affect
544: our  surface brightness  measurements. They  are as  follows:  (1) The
545: average \texttt{SExtractor}  uncertainties in $J$,  $H$ magnitudes are
546: $\sim$0.2  mag.  The  largest  magnitude uncertainties  in the  sample
547: range up  to $\sim$0.5  mag for three  objects.  Even this  worst case
548: magnitude uncertainty  affects S$_{e,\rm  90}$ by only  $\sim$0.2 dex.
549: The  uncertainty  in S$_{e,\rm  90}$  due  to  the average  difference
550: between  various  magnitudes (\texttt{MAG\_AUTO},  \texttt{MAG\_PETRO}
551: and  \texttt{MAG\_ISO})  is very  small  ($<$0.1  dex)  and hence  the
552: magnitude  uncertainty dominates  the uncertainty  in  S$_{e,\rm 90}$.
553: (2) The  ratio  of  the  UV  to bolometric  luminosity  for  starburst
554: galaxies is based on the stellar population models.  We find that this
555: ratio is affected  by the change in metallicity  or dust extinction in
556: the stellar population models. The uncertainty in this ratio can be as
557: large as  a factor of  $\sim$2.0.  This uncertainty in  the bolometric
558: correction will  change S$_{e,\rm 90}$  estimate by $\sim$0.3  dex but
559: this is systematic uncertainty and will affect all starburst galaxies.
560: (3) We  have used  linear fit  to  the observed  relation between  the
561: F$_{\rm FIR}$/F$_{\rm  1600}$ and the  UV spectral slope  ($\beta$) to
562: estimate  A$_{\rm 1600}$ \citep[Fig.1  in][]{meur99}.  The  scatter in
563: this plot can cause an uncertainty in A$_{\rm 1600}$ of $\sim$0.4 mag.
564: This uncertainty in A$_{\rm 1600}$ will change S$_{e,\rm 90}$ estimate
565: by  $\sim$0.2 dex.   This uncertainty  will systematically  affect all
566: galaxies in this study.
567: 
568: Therefore, if  these logarithmic  uncertainties add in  quadrature, we
569: would expect the  total uncertainty in S$_{e,\rm 90}$  to be $\sim$0.5
570: dex.
571: 
572: \subsection{The Starburst Intensity Limit}
573: 
574: We measure the effective  surface brightness (S$_e$) and the brightest
575: pixel  surface brightness  (S$_{bp}$) of  the \emph{spectroscopically}
576: confirmed  galaxies at  $z\!\simeq\!5\!-\!6$ in  the HUDF  and compare
577: with the spectroscopically confirmed galaxies at $z\!\simeq\!3$ in the
578: HDF.  We  conclude that to  better than a  factor of 3,  the starburst
579: intensity    limit    of     starbursts    at    $z\!\simeq\!3$    and
580: $z\!\simeq\!5\!-\!6$  are  the same.   Using  the  K-S  test on  these
581: distributions, the resulting probabilities ($\gtrsim$20\%) support the
582: hypothesis  that   these  distributions   are  drawn  from   the  same
583: population.   By  combining the  samples  at $z\!\simeq\!3\!-\!4$  and
584: $z\!\simeq\!5\!-\!6$, we find a mean  S$_{e,\rm 90} \simeq 2.3 {\rm \;
585:   x  \;}  10^{11}$  $L_{\sun}$  kpc$^{-2}$  (with a  factor  of  3  or
586: $\sim$0.5 dex uncertainty).  We quantify  the scatter in the S$_e$ and
587: S$_{bp}$  distributions  (Figure  5   and  6)  by  measuring  standard
588: deviation ($\sigma$)  from a Gaussian fit to  these distributions.  We
589: find    that     the    scatter    in     the    S$_e$    distribution
590: ($\sigma_{log(S_e)}\!\simeq\!0.37$) is higher  than the scatter in the
591: S$_{bp}$                                                   distribution
592: ($\sigma_{log(S_{bp})}\!\simeq\!0.27$). Therefore, the brightest pixel
593: method  could be  very useful  in estimating  the  starburst intensity
594: limit of starburst galaxies.
595: 
596: The  approximate   constancy  of  peak  starburst   intensity  can  be
597: interpreted  as the evidence  that the  interstellar medium  (ISM) can
598: only  support  some  maximum  pressure.  \citet{heck90}  used  [S  II]
599: emission  line ratios  to determine  the  ISM pressure  (P$_0$) for  a
600: sample of  galaxies (mostly  starbursts) undergoing a  strong galactic
601: wind.   The star  formation  intensity required  to  produce P$_0$  is
602: S$_{e} \simeq 1.7 {\rm \;  x \;} 10^{11}$ $L_{\sun}$ kpc$^{-2}$ (M97).
603: This value  agrees closely with  our S$_{e,\rm 90}$ but  derived using
604: different method.   The physical process can be  explained by assuming
605: that the  ISM pressure provides  a ``thermostat'' for  star formation,
606: such  that strong  outflows will  result whenever  the  pressure rises
607: above P$_0$  and shut  down further star  formation for a  time.  This
608: results  in  a characteristic  peak  starburst  intensity.  We  convert
609: S$_{e,\rm 90}$ to an equivalent star formation intensity ($M_{\sun} \;
610: {\rm yr}^{-1} \; {\rm  kpc}^{-2}$) by using conversion factors between
611: UV luminosity and star formation  rate from \citet{kenn98} and M97. We
612: get star formation intensity in  the range of 30--50 $M_{\sun} \; {\rm
613: yr}^{-1} \; {\rm kpc}^{-2}$ depending on the conversion factor.
614: %
615: This peak intensity is  physically distinct from the minimum intensity
616: required  to produce  a galactic  wind, which  is orders  of magnitude
617: lower, at  $\sim 0.1 \; M_{\sun}  \; {\rm yr}^{-1}  \; {\rm kpc}^{-2}$
618: \citep{lehn96,heck01}.   Thus, {\em  every}  galaxy in  our sample  is
619: expected to have a wind, as \citet{lehn07} remark for a similar sample
620: of $z\ga 3$ Lyman break  galaxies.  However, only in galaxies near the
621: starburst intensity limit does the ``thermostat'' become active.
622: 
623: While the  peak starburst intensity stays constant  with redshift, the
624: dust  optical depth  that we  infer from  the observed  spectral slope
625: $\beta$  decreases systematically with  redshift.  This  suggests that
626: the star formation intensity limit does not depend on dust content, at
627: least over the range of dust optical depths $\tau_{dust} \ga 1$ probed
628: by  our  sample.   At  yet  higher  redshifts  we  might  expect  that
629: $\tau_{dust} <  1$, and  that the starburst  radiation field  would no
630: longer couple  efficiently to  the interstellar medium.   Whether this
631: would substantially change the  starburst intensity limit would depend
632: on the  fraction of  starburst wind driving  that is due  to radiation
633: pressure  rather  than  stellar   winds  and  mechanical  energy  from
634: supernova explosions.
635: 
636: The  constancy  of   maximum  star-formation  surface  intensity  with
637: redshift provides a basis for a strong test of the expanding Universe.
638: Standard  cosmologies predict that  the bolometric  surface brightness
639: should  vary  with  redshift  as  $(1+z)^{-4}$.   Our  results  (using
640: standard cosmology)  combined with M97  results at low  redshifts show
641: that  the maximum  star formation  intensity remains  constant (within
642: uncertainties) from $z\!\simeq\!0$ to $z\!\simeq\!6$.  This conclusion
643: depends  critically on  the use  of a  standard cosmology  to  go from
644: observed flux and radius  to inferred star formation intensity.  Thus,
645: if the  peak star formation rate  per unit area is  controlled by some
646: physical limit  that is based on  local, redshift-independent physics,
647: our  observations  essentially  require  standard  surface  brightness
648: dimming.  While other evolutionary  effects could become important, we
649: have minimized these by measuring surface brightnesses at nearly fixed
650: rest wavelength.  Had we instead  taken a ``tired light'' model, where
651: bolometric   surface  brightness   falls  of   as   $(1+z)^{-1}$,  our
652: observations would require the true star formation intensity to be
dramatically lower at  high redshift--- by  a factor of  order $(1+z)^3
653: \sim  7^3  \sim  300$.   This  factor greatly  exceeds  the  estimated
654: uncertainties in our analysis.  Hence,  the wide redshift range of our
655: sample yields  a strong application  of Tolman's \citep{tolm30,tolm34}
656: test, and we derive strong evidence in favor of the expanding Universe
657: and against any alternative ``tired light'' models.
658: %We don't expect that any
659: %improvement in  our analysis  would bring this  factor into  line with
660: %unity.   Thus, if  there is  a characteristic  maximum  star formation
661: %intensity,  our observations  do  allow a  strong  application of  the
662: %Tolman test \citep{tolm30,tolm34}.
663: 
664: \subsection{Size and Luminosity Evolution}
665: 
666: An average galaxy at $z\!\simeq\!5\!-\!6$ in our sample has half-light
667: radii   of   $\sim$0.8  kpc   or   $\sim$0.14\arcsec,   as  shown   in
668: \figref{fig2},  which is  in good  agreement with  a number  of recent
669: studies  \citep{bouw04,bouw06,pirz07,dow07}.   \citet{ferg04} compares
670: sizes  of galaxies  at $z\!\simeq\!1$--$5$  within a  fixed luminosity
671: range.  Our results agree with \cite{ferg04} for $z\!\simeq\!3\!-\!4$.
672: The UV intrinsic luminosities for  our sample of galaxies are brighter
673: or equal to $L^{*}$ (i.e. $L\!\gtrsim\!L^{*}$) at respective redshifts
674: but  we  do  not  require  any particular  minimum  luminosity,  while
675: \citet{ferg04} do  require a  minimum luminosity.  Given  that surface
676: brightness is near-constant,  a minimum luminosity immediately implies
677: some  minimum  radius  for  inclusion in  the  \citet{ferg04}  sample.
678: \figref{fig2} also shows comparison  of our size measurements with the
679: \citet{ferg04} and  the \citet{bouw06}  results. The solid  and dashed
680: curves  in  the  \figref{fig2}  show  the trend  if  sizes  evolve  as
681: $H^{-1}(z)$ or $H^{-2/3}(z)$, respectively, where $H(z)$ is the Hubble
682: parameter at redshift $z$.  The curves are normalized to the mean size
683: we measure at  $z\!\simeq\!4$ ($\sim$0.21\arcsec\ts or $\sim$1.5 kpc).
684: The   galaxy  sizes  from   $z\!\simeq\!3$  to   $z\!\simeq\!6$  scale
685: approximately  as  the  Hubble  parameter $H^{-1}(z)$,  though  it  is
686: difficult  to  conclude  with  certainty  how  sizes  scale  from  the
687: measurements of high redshift ($z\!\gtrsim\!3$) galaxies only, because
688: two trends diverge significantly at lower redshifts ($z\!<\!3$).
689: 
690: We  also find  that  the  bolometric or  intrinsic  UV luminosity  for
691: galaxies  at  $z\!\simeq\!5\!-\!6$  is  lower  than  for  galaxies  at
692: $z\!\simeq\!3\!-\!4$.    The  galaxies  at   $z\!\simeq\!5\!-\!6$  are
693: smaller  compared  to   $z\!\simeq\!3$  galaxies,  while  the  surface
694: brightness as measured  in \secref{calculations} remains approximately
695: constant.   Therefore, the luminosity  evolution could  be due  to the
696: constant  upper limit  on  the  surface brightness  as  a function  of
697: redshift.
698: 
699: \subsection{Evolution in UV Spectral Slope \boldmath{($\beta$)}}\label{betainfo}
700: 
701: Our $z\!\simeq\!3\!-\!6$ sample also allows us to place constraints on
702: the rest-frame  UV slope.  We obtained these  constraints by measuring
703: rest-frame UV colors  of $z\!\simeq\!3\!-\!4$ and $z\!\simeq\!5\!-\!6$
704: galaxies  as shown  in \figref{fig1}.   A comparison  of  our measured
705: colors    with    those    obtained    in   two    previous    studies
706: \citep{stan05,bouw06}    show   agreement    within    our   1$\sigma$
707: uncertainties.   The  mean  $\beta$   inferred  from  this  study  for
708: $z\!\simeq\!6$  galaxies is  $\beta$=--1.74$\pm$0.35, which  is redder
709: than the  $\beta$=--2.0$\pm$0.3 inferred in the  \citet{bouw06} or the
710: $\beta$=--2.2$\pm$0.2  inferred in the  \citet{stan05}.  We  also find
711: that the  mean $\beta$ value  for galaxies at  $z\!\simeq\!5\!-\!6$ is
712: bluer than  the $\beta$=--1.1$\pm$0.2 we measure  at $z\!\simeq\!3$ or
713: the    $\beta$=--1.5$\pm$0.4   observed    by    the   \citet{adel00}.
714: Irrespective  of  the exact  $\beta$,  the  mean  rest-frame UV  slope
715: observed  at  $z\!\simeq\!5\!-\!6$  is  bluer than  that  observed  at
716: $z\!\simeq\!3$.  This evolution  is consistent  with number  of recent
717: studies \citep{stan05, yan05,bouw06}.
718: 
719: To   understand   the  evolution   in   the   $\beta$,   we  use   the
720: \texttt{STARBURST99}   stellar  synthesis   code   \emph{version  5.1}
721: \citep{leit99,vazq05}  to  investigate the  variations  in $\beta$  as
722: function  of IMF,  metallicity  and the  star  formation history.   We
723: assume  two  different  metallicities  ($Z$  = 0.004  and  0.02),  two
724: different  star formation histories  (constant and  instantaneous) and
725: two different versions of  the \citet{salp55} IMFs ($\alpha$=2.35 with
726: M$_{up}$=100   M$_{\odot}$    and   $\alpha$=2.35   with   M$_{up}$=30
727: M$_{\odot}$).   We measure  $\beta$ for  various models  by  fitting a
728: first-order  polynomial  to  the  UV spectra  through  the  wavelength
729: interval 1250--1850\Ang.  We find  that the changes in metallicity and
730: the IMF have much smaller effect  on the rest-frame UV slope for young
731: ($\lesssim$20   Myr)   starbursts.   Our   results   agree  with   the
732: \citet{leit99}  models  showing  that   the  UV  spectral  slopes  are
733: independent of  evolution and IMF effects.   \citet{leit99} shows that
734: for young  starbursts, the UV spectral slope  ($\beta$) at 2500\Ang\ts
735: also does not change very much  as a function of model parameters ($Z$
736: and  IMF).    Therefore  the   observed  evolution  in   $\beta$  from
737: $z\!\simeq\!5\!-\!6$  to  $z\!\simeq\!3$ is  more  likely  due to  the
738: change in the dust content.
739: 
740: %--------------------------------------------------
741: \section{Summary}\label{summary}
742: 
743: We    have    measured    the    starburst   intensity    limit    for
744: \emph{spectroscopically  confirmed}  galaxies at  $z\!\simeq\!5\!-\!6$
745: from  ACS grism  survey GRAPES  in the  HUDF.  We  find that  there is
746: little  variation in  the  surface brightness  from $z\!\simeq\!3$  to
747: $z\!\simeq\!6$ and the starburst intensity limit is within a factor of
748: 3 when compared with the sample of $z\!\simeq\!3\!-\!4$ galaxies.  The
749: constancy  of  starburst intensity  limit  for  starburst galaxies  at
750: $z\!\simeq\!5\!-\!6$  combined with  the results  obtained by  M97 for
751: starbursts  at $z\!\lesssim\!3$  implies that  the  physical processes
752: limiting starburst  intensity at lower  redshifts also apply  to these
753: high  redshift galaxies.  We  find that  the high  redshift starbursts
754: have   a  smaller   characteristic  linear   size  than   their  local
755: counterparts,  and  a correspondingly  lower  luminosity (since  their
756: surface  brightnesses  are  similar,  and their  sizes  smaller).   We
757: observe   the   galaxy   size   evolution   from   $z\!\simeq\!3$   to
758: $z\!\simeq\!6$  and find  that the  sizes scale  approximately  as the
759: Hubble parameter $H^{-1}(z)$.  Finally,  using rest-frame UV colors we
760: conclude   that  the  evolution   in  the   UV  spectral   slope  from
761: $z\!\simeq\!3$ to  $z\!\simeq\!6$ reinforces the  dust evolution which
762: leads to  bluer galaxies at $z\!\simeq\!5\!-\!6$  compared to galaxies
763: at  $z\!\simeq\!3$. This  implies that  starbursts were  less obscured
764: when the universe was younger  and had lower heavy element abundances.
765: Any future  search for galaxies at  higher redshifts \citep[e.g.\emph{
766:   James Webb  Space Telescope  (JWST)};][]{wind07} needs to  take into
767: account  the  size  evolution  \&  constancy  of  surface  brightness;
768: therefore the decrease in characteristic luminosity with redshift.
769: 
770: 
771: %--------------------------------------------------
772: 
773: \acknowledgments 
774: We would  like to thank Rogier  Windhorst, Rolf Jansen  and Seth Cohen
775: for providing  useful comments on  earlier drafts of this  paper. This
776: work  was supported  by grants  GO 9793  and GO  10530 from  the Space
777: Telescope  Science Institute,  which is  operated by  AURA  under NASA
778: contract NAS5-26555.  We also  thank the anonymous referee for helpful
779: suggestions that improved the paper.
780: 
781: %Facilities: \facility{HST(ACS)}
782: %--------------------------------------------------
783: %\pagebreak
784: 
785: \begin{thebibliography}{}
786: \bibitem[Adelberger  \& Steidel(2000)]{adel00}  Adelberger, K.  L., \&
787:   Steidel, C. C. 2000, ApJ, 544, 218
788: \bibitem[Beckwith \etal(2006)]{beck06} Beckwith,  S., et al. 2006, AJ,
789:   132, 1729
790: \bibitem[Bertin  \&  Arnouts(1996)]{bert96}  Bertin, E.,  \&  Arnouts,
791:   S. 1996, A\&AS, 117, 393
792: \bibitem[Bouwens  \etal(2004)]{bouw04}  Bouwens,  R. J.,  Illingworth,
793:   G. D., Blakeslee, J. P., Broadhurst,  T. J., \& Franx, M. 2004, ApJ,
794:   611, 1
795: \bibitem[Bouwens  \etal(2006)]{bouw06}  Bouwens,  R. J.,  Illingworth,
796:   G. D., Blakeslee, J. P., \& Franx, M. 2006, ApJ, 653, 53
797: \bibitem[Bruzual  \& Charlot(2003)]{bruz03}  Bruzual, G.,  \& Charlot,
798:   S. 2003, MNRAS, 344, 1000
799: \bibitem[Calzetti \etal(1994)]{calz94} Calzetti, D., Kinney, A. L., \&
800:   Storchi-Bergmann, T. 1994, ApJ, 429, 582
801: \bibitem[Dow-Hygelund  \etal(2007)]{dow07}  Dow-Hygelund,  C.  C.,  et
802:   al. 2007, ApJ, 660, 47
803: \bibitem[Eyles  \etal(2005)]{eyle05} Eyles,  L.  P.,  Bunker,  A.  J.,
804:   Stanway, E.  R., Lacy, M., Ellis R. S., \& Doherty, M.  2005, MNRAS,
805:   364, 443
806: \bibitem[Ferguson \etal(2004)]{ferg04}  Ferguson, H. C.,  et al. 2004,
807:   ApJ, 600, L107
808: \bibitem[Giavalisco  \etal(1996)]{giav96}   Giavalisco,  M.,  Steidel,
809:   C. C., \& Macchetto, F. D. 1996, ApJ, 470, 189
810: \bibitem[Heckman  \etal(1990)]{heck90} Heckman, T.  M., Armus,  L., \&
811:   Miley, G. K. 1990, ApJS, 74, 833
812: \bibitem[Heckman \etal(2000)]{heck00} Heckman,  T. M., Lehnert, M. D.,
813:   Strickland, D. K., \& Armus, L. 2000, ApJS, 129, 493
814: \bibitem[Heckman(2001)]{heck01}   Heckman,   T.   M.  2001,   in   ASP
815:   Conf.  Ser.  240, Gas  and  Galaxy  Evolution,  ed. J.  E.  Hibbard,
816:   M. Rupen, \& J. H. van Gorkom (San Francisco: ASP), 345
817: \bibitem[Holwerda(2005)]{holw05} Holwerda, B. W. 2005, (astro-ph/0512139)
818: \bibitem[Kennicutt(1998)]{kenn98} Kennicutt, R.  C., Jr. 1998, ARA\&A,
819:   36, 189
820: \bibitem[Kron(1980)]{kron80} Kron, R. G. 1980, ApJS, 43, 305
821: \bibitem[Lehnert \& Heckman (1996)]{lehn96} Lehnert, M. D., \& Heckman, 
822:   T. 1996, ApJ  472, 546
823: \bibitem[Lehnert  \etal(2007)]{lehn07}  Lehnert,  M. D.,  Bremer,  M.,
824:   Verma,  A., Douglas,  L.,  \&  Forster Schreiber,  N.  2007, in  ASP
825:   Conf.    Ser.,    Pathways    Through    an    Eclectic    Universe,
826:   ed. J. H. Knappen, T. J. Mahoney, \& A. Vazedekis (astro-ph/0708.3000)
827: \bibitem[Leitherer  \etal(1999)]{leit99} Leitherer,  C., et  al. 1999,
828:   ApJS, 123, 3
829: \bibitem[Malhotra \etal(2005)]{malh05} Malhotra, S., et al. 2005, ApJ,
830:   626, 666
831: \bibitem[Meurer  \etal(1995)]{meur95} Meurer, G.  R., Heckman,  T. M.,
832:   Leitherer, C., Kinney,  A., Robert, A., \& Garnett,  D. R. 1995, AJ,
833:   110, 2665
834: \bibitem[Meurer  \etal(1997)]{meur97} Meurer, G.  R., Heckman,  T. M.,
835:   Lehnert, M.  D., Leitherer, C., \&  Lowenthal, J. 1997,  AJ, 114, 54
836:   (M97)
837: \bibitem[Meurer \etal(1999)]{meur99} Meurer, G. R., Heckman, T. M., \&
838:   Calzetti, D. 1999, ApJ, 521, 64
839: \bibitem[Oke  \& Gunn(1983)]{oke83} Oke,  J. B.,  \& Gunn,  J.E. 1983,
840:   ApJ, 266, 713
841: \bibitem[Petrosian(1976)]{petr76} Petrosian, V. 1976, ApJ, 209, L1
842: \bibitem[Pirzkal \etal(2004)]{pirz04} Pirzkal,  N., et al. 2004, ApJS,
843:   154, 501
844: \bibitem[Pirzkal  \etal(2007)]{pirz07}   Pirzkal,  N.,  Malhotra,  S.,
845:   Rhoads, J., \& Xu, C. 2007, ApJ, in press (astro-ph/0612513)
846: \bibitem[Salpeter(1955)]{salp55} Salpeter, E. E. 1955, ApJ, 121, 161
847: \bibitem[Spergel  \etal(2007)]{sper07} Spergel,  D. N.,  et  al. 2007,
848:   ApJS, 170, 377
849: \bibitem[Stanway \etal(2005)]{stan05} Stanway,  E. R., McMahon, R. G.,
850:   \& Bunker, A. J. 2005, MNRAS, 359, 1184
851: \bibitem[Steidel  \&   Hamilton(1993)]{stei93}  Steidel,  C.   C.,  \&
852:   Hamilton, D. 1993, AJ, 105, 2017
853: \bibitem[Steidel  \etal(1996a)]{stei96a} Steidel,  C.  C., Giavalisco,
854:   M., Pettini,  M., Dickinson,  M., \& Adelberger,  K. L.  1996a, ApJ,
855:   462, L17
856: \bibitem[Steidel  \etal(1996b)]{stei96b} Steidel,  C.  C., Giavalisco,
857:   M., Dickinson, M., \& Adelberger, K. L. 1996b, AJ, 112, 352
858: \bibitem[Thompson \etal(2005)]{thom05}  Thompson, R. I.,  et al. 2005,
859:   AJ, 130, 1
860: \bibitem[Tolman(1930)]{tolm30} Tolman, R. C. 1930, PNAS, 16, 511
861: \bibitem[Tolman(1934)]{tolm34}  Tolman, R. C.  1934, \emph{Relativity,
862:     Thermodynamics, \& Cosmology}, (Oxford: Oxford Univ. Press), 467
863: \bibitem[Vanzella  \etal(2006)]{vanz06}  Vanzella,  E., et  al.  2006,
864:   A\&A, 454, 423
865: \bibitem[V\'{a}zquez  \& Leitherer(2005)]{vazq05} V\'{a}zquez,  G. A.,
866:   \& Leitherer, C. 2005, ApJ, 621, 695
867: \bibitem[Weedman \etal(1998)]{weed98} Weedman, D. W., Wolovitz, J. B.,
868:   Bershady, M. A., \& Schneider, D. P. 1998, AJ, 116, 1643 (W98)
869: \bibitem[Williams \etal(1996)]{will96}  Williams, R. E.,  Blacker, B.,
870:   Dickinson, M., et al. 1996, AJ, 112, 1335
871: \bibitem[Windhorst  \etal(2007)]{wind07}   Windhorst,  R.  A.,  Hathi,
872:   N. P., Cohen,  S.H., \& Jansen, R. A.  2007, \emph{Advances in Space
873:     Research}, in press (astro-ph/0703171)
874: \bibitem[Wright(2006)]{wrig06} Wright, E. L. 2006, PASP, 118, 1711
875: \bibitem[Yan \etal(2005)]{yan05} Yan, H., et al. 2005, ApJ, 634, 109
876: \end{thebibliography}
877: 
878: %--------------------------------------------------
879: \clearpage
880: 
881: \input tab1.tex
882: 
883: %--------------------------------------------------
884: \clearpage
885: 
886: \begin{figure}
887: \epsscale{0.7}
888: \plotone{f1.eps}
889: \caption{UV  spectral  slopes ($\beta$)  v/s  redshift relation.   Mean
890:   $\beta$  are  plotted  with   error  bars  indicating  the  standard
891:   deviation of the  mean (i.e.  $\sigma / \sqrt  N$).  We have plotted
892:   the  data  points  at  $z\!\simeq\!3$  from  \citet{adel00}  and  at
893:   $z\!\simeq\!6$ from \citet{bouw06} for comparison.}\label{fig1}
894: \end{figure}
895: 
896: %--------------------------------------------------
897: \clearpage
898: 
899: \begin{figure}
900: \epsscale{0.7}
901: \plotone{f2.eps}
902: \caption{Size v/s redshift relation. Mean half-light radii are plotted
903:   with error bars indicating the  standard deviation of the mean (i.e.
904:   $\sigma / \sqrt N$).   The solid and dashed curves  shows the trend
905:   if  sizes  evolve as  $H^{-1}(z)$  and $H^{-2/3}(z)$,  respectively.
906:   Both  curves  are normalized  to  the  mean  size at  $z\!\simeq\!4$
907:   ($\sim$0.21\arcsec\ts or  $\sim$1.5 kpc).  We have  plotted the data
908:   points  at $z\!\simeq\!3$ from  \citet{ferg04} and  at $z\!\simeq\!6$
909:   from \citet{bouw06} for comparison.}\label{fig2}
910: \end{figure}
911: 
912: %--------------------------------------------------
913: \clearpage
914: 
915: \begin{figure}
916: \epsscale{0.7}
917: \plotone{f3.eps} 
918: \caption{Bolometric  luminosity ($L_{\sun}^{\rm  bol}$)  and effective
919:   surface   brightness   ($L_{\sun}^{\rm   bol}$~kpc$^{-2}$)   against
920:   effective radii for starburst galaxies.  The filled squares, circles
921:   and  triangles  are  measurements  for galaxies  at  $z\!\simeq\!3$,
922:   $z\!\simeq\!4$  and  $z\!\simeq\!5\!-\!6$,  respectively.  The  open
923:   square  ($z\!\simeq\!0$)  is the  median  measurement  of 11  nearby
924:   galaxies from M97.  The open diamonds ($z\!\simeq\!0.4$) are S$_{e}$
925:   measurements from  M97.  The dotted  and dashed lines  correspond to
926:   S$_{e,\rm   90}$  and   S$_{e,\rm  50}$   of  the   combine  sample.
927:   Uncertainties in  $z\!\simeq\!3\!-\!6$ surface brightness  and radii
928:   measurements are shown  in lower right corner, details  are given in
929:   \secref{bias}.}\label{fig3}
930: \end{figure}
931: 
932: %--------------------------------------------------
933: \clearpage
934: 
935: \begin{figure}
936: \epsscale{0.7}
937: \plotone{f4.eps}
938: \caption{Bolometric   effective  surface   brightness  ($L_{\sun}^{\rm
939:     bol}$~kpc$^{-2}$)  as a  function  of redshift.   The open  square
940:     ($z\!\simeq\!0$) is  the median measurement of  11 nearby galaxies
941:     from  M97.   The  filled  squares ($z\!\simeq\!0.4$)  are  S$_{e}$
942:     measurements  from M97.  The  filled squares  ($z\!\simeq\!3$) are
943:     the galaxies from the sample  of M97 for which we measured surface
944:     brightnesses.   The  circles  are   the  galaxies  in  our  sample
945:     ($z\!\simeq\!4\!-\!6$).  The dotted and dashed lines correspond to
946:     S$_{e,\rm  90}$   and  S$_{e,\rm  50}$  of   the  combine  sample.
947:     Uncertainties in  $z\!\simeq\!3\!-\!6$ surface brightness  (due to
948:     radii  and flux  uncertainties)  is shown  in  lower left  corner,
949:     details are given in \secref{bias}.}\label{fig4}
950: \end{figure}
951: 
952: %--------------------------------------------------
953: \clearpage
954: 
955: \begin{figure}
956: \epsscale{0.7}
957: \plotone{f5.eps}
958: \caption{Surface Brightness  ($L_{\sun}^{\rm UV}$~kpc$^{-2}$) obtained
959:   from the brightest pixel of 47 galaxy images at $z\!\simeq\!5\!-\!6$
960:   and  4 galaxy images  at $z\!\simeq\!4$  .  The  squares are  18 W98
961:   galaxies.  The  W98 galaxies  had observed surface  brightnesses and
962:   hence,  for   proper  comparison,  we   converted  observed  surface
963:   brightnesses to corresponding  rest-frame surface brightnesses.  The
964:   dotted and dashed lines correspond to S$_{bp,\rm 90}$ and S$_{bp,\rm
965:     50}$ of  the combine sample.  Uncertainties  in surface brightness
966:   (due to  flux uncertainties  only) is shown  in lower  right corner,
967:   details are given in \secref{bias}.}\label{fig5}
968: \end{figure}
969: 
970: %--------------------------------------------------
971: 
972: 
973: \end{document}
974: