1: \documentstyle[emulateapj,psfig]{article}
2:
3: \makeatletter
4: \newenvironment{inlinetable}{%
5: \def\@captype{table}%
6: \noindent\begin{minipage}{0.999\linewidth}\begin{center}\footnotesize}
7: {\end{center}\end{minipage}\smallskip}
8:
9: \newenvironment{inlinefigure}{%
10: \def\@captype{figure}%
11: \noindent\begin{minipage}{0.999\linewidth}\begin{center}}
12: {\end{center}\end{minipage}\smallskip}
13: \makeatother
14:
15: %\lefthead{Cowie et al.}
16:
17: \begin{document}
18: \title{The Evolution of the Ultraluminous Infrared Galaxy
19: Population from Redshift 0 to 1.5\altaffilmark{1,2}}
20: \author{
21: L.~L.~Cowie,$\!$\altaffilmark{3}
22: A.~J.~Barger,$\!$\altaffilmark{4,5,3}
23: E.~B.~Fomalont,$\!$\altaffilmark{6}
24: P.~Capak$\!$\altaffilmark{3}
25: }
26:
27: \altaffiltext{1}{Based in part on data obtained at the W. M. Keck
28: Observatory, which is operated as a scientific partnership among the
29: the California Institute of Technology, the University of
30: California, and NASA and was made possible by the generous financial
31: support of the W. M. Keck Foundation.}
32: \altaffiltext{2}{Based in part on data obtained at the VLA, which is
33: a facility of the National Radio Astronomy Observatory (NRAO).
34: The NRAO is a facility of the National Science Foundation operated
35: under cooperative agreement by Associated Universities, Inc.}
36: \altaffiltext{3}{Institute for Astronomy, University of Hawaii,
37: 2680 Woodlawn Drive, Honolulu, HI 96822.}
38: \altaffiltext{4}{Department of Astronomy, University of
39: Wisconsin-Madison, 475 North Charter Street, Madison, WI 53706.}
40: \altaffiltext{5}{Department of Physics and Astronomy,
41: University of Hawaii, 2505 Correa Road, Honolulu, HI 96822.}
42: \altaffiltext{6}{National Radio Astronomy Observatory,
43: Charlottesville, VA 22903.}
44:
45: \slugcomment{Accepted by The Astrophysical Journal Letters}
46:
47: \begin{abstract}
48: We use redshift observations of two deep 1.4~GHz fields
49: to probe the evolution of the bright end of the radio galaxy luminosity
50: function to $z=1.5$. We show that the number of galaxies
51: with radio power that would correspond locally to an ultraluminous
52: infrared galaxy (ULIG) evolves rapidly over this redshift
53: range. The optical spectra and X-ray luminosities are used to refine this
54: result by separating the sources with clear active galactic nucleus (AGN)
55: signatures from those that may be dominated by star formation. Both
56: populations show extremely rapid evolution over this redshift range. We
57: find that the number of sources with ULIG radio power and no clear AGN
58: signatures evolves as $(1+z)^{7}$.
59: \end{abstract}
60:
61: \keywords{cosmology: observations --- galaxies: distances and
62: redshifts --- galaxies: active --- X-rays: galaxies ---
63: galaxies: formation --- galaxies: evolution}
64:
65: \section{Introduction}
66: \label{secintro}
67:
68: Star-forming ultraluminous infrared galaxies, or ULIGs
69: ($L_{FIR}>10^{12}~L_\odot$ from
70: \markcite{sanders96}Sanders \& Mirabel 1996;
71: here we mean galaxies without clear signatures of active galactic nucleus
72: [AGN] activity in their spectra, even if there may be some AGN
73: contribution to their luminosities), represent the most luminous tip of
74: the star-forming galaxy population. Locally, ULIGs are very scarce
75: (\markcite{kim98}Kim \& Sanders 1998), but if, as is widely believed,
76: submillimeter galaxies are the high redshift analogs, then by
77: $z=1$ they are the dominant contributors to the
78: star formation history (e.g., \markcite{lilly99}Lilly et al.\ 1999;
79: \markcite{bcr00}Barger, Cowie, \& Richards 2000;
80: \markcite{gispert00}Gispert, Lagache, \& Puget 2000;
81: \markcite{chapman03}Chapman et al.\ 2003).
82: It is extremely hard to trace the evolution of the ULIG population
83: at low redshifts ($z<1.5$) because of the difficulties of
84: mapping large areas with current submillimeter
85: detectors and of identifying the optical counterparts to low spatial
86: resolution submillimeter or far-infrared (FIR) observations.
87:
88: Ultradeep decimetric radio surveys can play a major, complementary
89: role. The well-known tight correlation between
90: FIR luminosity and radio power in local star-forming galaxies and
91: radio-quiet AGNs (\markcite{condon92}Condon 1992)
92: means that we can identify ULIGs out to substantial redshifts based
93: on their radio power.
94: For a 1.4~GHz sensitivity limit of 40~$\mu$Jy,
95: an $L_{FIR}=10^{12}~L_\odot$ source will be detected to $z\sim 1.5$.
96: Thus, sources with radio power that would correspond locally to a ULIG
97: may be seen directly to large redshifts. Since the ultradeep 1.4~GHz
98: samples cover large fields, do not suffer from extinction,
99: and provide subarcsecond positional accuracy, it is easy to develop
100: large samples with highly complete optical counterpart identifications.
101: By contrast, optically selected samples may omit dusty sources at
102: these redshifts, while X-ray samples will select AGN dominated sources.
103:
104: Current decimetric surveys do not probe the
105: normal star-forming galaxy populations at $z=1$. To the deepest
106: 1.4~GHz sensitivity limits of $\sim 20$~$\mu$Jy ($5\sigma$),
107: a source with radio power corresponding
108: to $L_{FIR}=10^{10}~L_\odot$ would only be seen to $z\sim 0.3$.
109: Thus, attempts to provide maps of the entire star formation
110: history over the $z=0-1$ range based on decimetric observations
111: (e.g., \markcite{cram98}Cram 1998;
112: \markcite{mobasher99}Mobasher et al.\ 1999;
113: \markcite{haarsma00}Haarsma et al.\ 2000, hereafter H00)
114: have to rely on model luminosity
115: functions to extrapolate the measured data and obtain star formation rates.
116: In these descriptions, much of the star formation lies in
117: sources that are not directly measured by the observations (e.g., H00).
118:
119: We use ultradeep 1.4~GHz observations of the
120: Hubble Deep Field-North (HDF-N; \markcite{richards00}Richards 2000)
121: and SSA13 (\markcite{fomalont04}Fomalont et al.\ 2004) fields
122: to analyze the evolution of the radio luminosity function (LF) over the
123: range $z=0-1.5$. Our sample of 346 radio sources represents
124: an increase of nearly a factor of five over H00 and enables
125: us to provide a much more accurate determination of the $z=1$ radio
126: LF. We use optical spectral classifications
127: and X-ray characteristics to separate AGN dominated sources
128: from those that may be star formation dominated and to estimate
129: the number density of ULIGs. An analysis of the submillimeter
130: and far infrared properties of the sample will appear in
131: Wang, Barger, and Cowie 2004 in preparation.
132: We assume $\Omega_M={1\over 3}$, $\Omega_\Lambda={2\over 3}$,
133: and $H_0=65$~km~s$^{-1}$~Mpc$^{-1}$.
134:
135: \section{Sample}
136: \label{secsample}
137:
138: \markcite{richards00}Richards (2000) presented a catalog of 1.4~GHz
139: sources detected in a Very Large Array (VLA) map centered on the HDF-N
140: that covers a $40'$ diameter region with an effective resolution of
141: $1.8''$. The $5\sigma$ completeness limit for compact sources in
142: the central region of the map is 40~$\mu$Jy.
143: \markcite{fomalont04}Fomalont et al.\ (2004) presented a
144: similar catalog within a $17'$ radius of
145: the 1.4~GHz VLA map ($5\sigma$ completeness of 25~$\mu$Jy)
146: of the SSA13 field (R.A.$=13^h 23^m$, decl.$=42^\circ 38''$).
147: In both cases, the completeness limits
148: for more extended sources are several times higher.
149: The absolute radio positions are known to
150: $0.1''-0.2''$ rms. Optical imaging data for the two fields
151: can be found in \markcite{capak03}Capak et al.\ (2003) and
152: \markcite{barger01}Barger et al.\ (2001), respectively.
153: \markcite{alex03}Alexander et al.\ (2003) presented
154: 2~Ms {\it Chandra} Deep Field-North (CDF-N) X-ray images that
155: cover about 460~arcmin$^{2}$ around the HDF-N.
156: Near the aim point, the data reach
157: limiting fluxes of $\approx 1.5\times 10^{-17}$ ($0.5-2$~keV)
158: and $\approx 10^{-16}$ ergs~cm$^{-2}$~s$^{-1}$ ($2-8$~keV).
159: \markcite{mushotzky00}Mushotzky et al.\ (2000) and
160: \markcite{barger01}Barger et al.\ (2001) presented
161: 100~ks {\it Chandra} exposures of the SSA13 field with
162: limiting fluxes 20 times higher than the CDF-N.
163:
164: For the HDF-N region, we use a $9.3'$ radius
165: area around the approximate {\it Chandra} image center
166: R.A.$=12^h~36^m~51.20^s$ and decl.$=62^\circ 13'~43.0''$ (J2000.0).
167: The $9.3'$ radius is the largest circle that can be drawn
168: within the {\it Chandra} area. 195 of the 372 radio sources
169: lie within this circle. For the SSA13 field, we use a
170: rectangular area of 175 arcmin$^2$ that contains 221 sources.
171: We have so far spectroscopically observed 154 of these
172: sources, chosen without regard to their radio or optical properties;
173: thus, the effective area is 120~arcmin$^2$. Together these two
174: restricted samples constitute our radio sample with 346 sources.
175: All lie well within the half power radius of the radio observations,
176: so the two fields should have relatively uniform flux selections.
177:
178: %
179: % Figure 1
180: %
181: \begin{inlinefigure}
182: \psfig{figure=f1.eps,angle=90,width=3.5in}
183: \vspace{6pt}
184: \figurenum{1}
185: \caption{
186: Redshift versus $R$ magnitude for our radio sample,
187: excluding the two spectroscopically identified stars
188: ({\it solid squares}---spectroscopic redshifts;
189: {\it open triangles}---photometric redshifts;
190: {\it large solid squares}---broad-line AGNs).
191: Dashed horizontal line shows the redshift ($z=1.05$) for which
192: a source with the median radio flux of 80~$\mu$Jy would have
193: a radio power that would correspond locally to a ULIG.
194: Solid histogram shows the spectroscopic completeness:
195: at $R<23$, $96\%$ (152 of the 162 sources)
196: are identified; at $R=23.5-24$, just over $50\%$;
197: at fainter magnitudes, only a handful with strong
198: emission lines. Dashed histogram shows the
199: completeness of the combined photometric and spectroscopic redshifts
200: in the HDF-N: 92\% of the $R<25$ sources have either a spectroscopic
201: or a photometric redshift.
202: \label{fig1}
203: }
204: \addtolength{\baselineskip}{10pt}
205: \end{inlinefigure}
206:
207: We observed our radio sample with the
208: Deep Extragalactic Imaging Multi-Object Spectrograph
209: (DEIMOS; \markcite{faber03}Faber et al.\ 2003)
210: on the Keck~II 10~m telescope and
211: the Low-Resolution Imaging Spectrograph
212: (LRIS; \markcite{oke95}Oke et al.\ 1995) on Keck~I.
213: We spectroscopically identified 192 of the 346 sources (55\%),
214: two of which are stars. Figure~\ref{fig1} shows spectroscopic redshift
215: versus $R$ magnitude ({\it solid symbols}) for our radio sample, as
216: well as the completeness of our spectroscopic identifications
217: versus $R$ magnitude ({\it solid histogram}). A source
218: with the median radio flux of 80~$\mu$Jy has a radio power corresponding
219: to a local ULIG at $z=1.05$ ({\it dashed horizontal line});
220: this corresponds to $R$ magnitudes in the range 22.5 to 25
221: (see Fig.~\ref{fig1}). Spectroscopic identifications
222: can be obtained for the majority of sources to $R=24$
223: and for some in the $R=24$ to $R=25$ range (see
224: solid histogram).
225:
226: We can also estimate redshifts from photometric colors. For the
227: HDF-N, we use the photometric redshifts of
228: \markcite{capak04}Capak et al.\ (2004), which are
229: based on ultradeep imaging in seven colors from $U$ to $HK'$
230: (\markcite{capak03}Capak et al.\ 2003) and were computed using
231: the Bayesian code of \markcite{benitez00}Ben\'{\i}tez (2000).
232: This code assigns odds that the redshift lies close to the estimated
233: value; we only use cases with odds above 90\%.
234: The imaging data saturate at bright magnitudes, so we restrict
235: to sources with $R>20$, leaving 161 photometric redshifts.
236: The photometric redshifts ({\it open triangles}) for sources
237: that do not also have spectroscopic redshifts are shown in
238: Figure~\ref{fig1}, as is the completeness of our HDF-N combined
239: photometric and spectroscopic redshift sample
240: versus $R$ magnitude ({\it dashed histogram}).
241: Eighty-three sources with $R>20$ have both spectroscopic
242: and photometric redshifts, and 92\% of these agree to within
243: a multiplicative factor of 1.25. This agreement is independent
244: of spectral type, except for the very small number of
245: broad-line AGNs, where the failure rate is higher. The success rate
246: of the photometric redshifts is consistent with that expected
247: from the assigned odds. Since we do not have photometric
248: redshifts for the SSA13 sources, and since photometric redshifts do
249: not contain spectral type information, we only use the photometric
250: redshifts to test the effects of completeness on the total LF.
251:
252: For the spectroscopically identified sources, we classified
253: the optical spectra into four spectral types,
254: roughly following the procedure used by
255: \markcite{sadler02}Sadler et al.\ (2002, hereafter S02)
256: to analyze low redshift 1.4~GHz samples. We classified sources
257: without strong emission lines (EW([OII])$<3$~\AA\ or EW(H$\alpha+$NII)
258: $<10$~\AA) as absorbers; sources with strong Balmer
259: lines and no broad or high-ionization lines as star formers;
260: sources with [NeV] or CIV lines or strong [OIII]
261: (EW([OIII]~5007~\AA$)>3$~EW(H$\beta)$) as Seyferts; and, finally,
262: sources with FWHM$>1000$~km~s$^{-1}$ as
263: broad-line AGNs. The breakdown by spectral type
264: in each field, and also by radio flux, is given in Table~1.
265: S02 argued that the absorbers must be AGN dominated since
266: star formation cannot produce the synchrotron emission in these galaxies.
267: They combined all but the star formers into a single AGN class.
268:
269: The deep CDF-N X-ray data provide a check and a refinement of the
270: classifications. We computed
271: the $2-8$~keV rest-frame luminosity, $L_x$, for each source with a
272: spectroscopic redshift. In Figure~\ref{fig2} we
273: plot $L_x$ versus radio power for the
274: AGNs (classes 3 and 4; {\it triangles}), the star formers ({\it diamonds}),
275: and the absorbers ({\it squares}).
276: The sources classed as AGNs are generally
277: X-ray luminous. Of the 40 AGNs, 29 are detected in the $2-8$~keV
278: band, 26 of which have $L_x>10^{42}$~ergs~s$^{-1}$.
279: Such luminosities would securely identify them as AGNs
280: (\markcite{zezas98}Zezas, Georgantopoulos, \& Ward 1998;
281: \markcite{moran99}Moran, Lehnert, \& Helfand 1999).
282: The remaining upper bounds are consistent
283: with all of the sources having $L_x>10^{41}$~ergs~s$^{-1}$.
284:
285: %
286: % Figure 2
287: %
288: \begin{inlinefigure}
289: \psfig{figure=f2.eps,angle=90,width=3.5in}
290: \vspace{6pt}
291: \figurenum{2}
292: \caption{
293: $2-8$~keV luminosity versus 1.4~GHz power for the HDF-N radio sample.
294: Triangles show sources with AGN optical spectra
295: ({\it large symbols}---broad-line AGNs; {\it small symbols}---Seyferts);
296: diamonds show star formers; squares show absorbers.
297: Arrows denote upper limits for sources not detected in the 2~Ms CDF-N.
298: Solid diagonal line is an approximate linear fit to the AGN population
299: and is just shown for guidance. Dashed horizontal line shows the
300: luminosity that would securely identify a source as an AGN.
301: \label{fig2}
302: }
303: \addtolength{\baselineskip}{10pt}
304: \end{inlinefigure}
305:
306: By contrast, nearly all of the star formers have low X-ray
307: luminosities. Only six of the 55 sources are detected in
308: the $2-8$~keV band, and only
309: four have $L_x>10^{42}$~ergs~s$^{-1}$. The absorbers are
310: the most unusual class. If they are AGN dominated, then they
311: are very underluminous in their rest-frame $2-8$~keV
312: luminosities relative to their radio power. This suggests
313: that they could be Compton-thick or near Compton-thick
314: sources. It may therefore be useful to maintain the distinction
315: between the AGNs with clear optical signatures and the
316: absorbers. However, to be consistent with S02,
317: we combine the two classes. The absorbers class has about half as many
318: sources as the spectroscopic AGN class and a similar distribution
319: of radio power. Thus, the AGN radio LF may be
320: roughly corrected to a spectroscopic AGN LF
321: by multiplying by $0.66$. Some sources that have
322: star-forming optical signatures may also correspond to this
323: type of source: X-ray quiet, radio detected AGN with no spectroscopic
324: AGN signatures. There may be ongoing star formation in the galaxy, but
325: the AGN dominates the radio power. Thus, the optically classified
326: star formers may only be an upper limit to the population
327: where star formation dominates the radio power.
328:
329: For our final classification, we classed all absorbers, star formers
330: known to have measured (not limits) $L_x>10^{42}$~ergs~s$^{-1}$,
331: and sources with AGN optical spectra as AGNs. We classed the remaining
332: sources as star formers. This enables straightforward comparisons
333: with the local results of \markcite{condon89}Condon (1989),
334: \markcite{condon02}Condon, Cotton, \& Broderick (2002),
335: \markcite{mac01}Machalski \& Godlowski (2001, hereafter MG01),
336: and S02.
337:
338: \section{Discussion}
339: \label{secop}
340:
341: We constructed the LFs for the 1.4~GHz power over the
342: range $z=0.5-1.5$ using the $1/V_a$ method
343: (e.g., \markcite{felten77}Felten 1977). This is shown for the
344: total spectroscopic sample (AGNs and star formers;
345: {\it triangles}) in Figure~\ref{fig3}a, where it is compared with
346: the local measurements of \markcite{condon89}Condon (1989;
347: {\it solid line}). To investigate the effects
348: of incompleteness, we computed the maximum total LF
349: ({\it squares}) for just the HDF-N using the much more complete
350: combined photometric and spectroscopic samples and placing
351: all sources without a photometric or a spectroscopic
352: redshift into the $z=0.5-1.5$ bin.
353: Many of these probably lie at higher redshifts,
354: so this is an upper bound. Finally, we assumed that the
355: HDF-N sample was only 50\% complete in the $40-80$~$\mu$Jy
356: range, which approximates the completeness corrections
357: computed by \markcite{richards00}Richards (2000). The maximum
358: correction for incompleteness is a factor of 2.6 at the lower
359: radio powers but is relatively small at the higher radio powers.
360:
361: In order to parameterize the evolution, we assumed
362: that the local portion of the LF allocated to star formers
363: in \markcite{condon89}Condon (1989) undergoes pure luminosity evolution
364: and that allocated to AGNs undergoes pure number density
365: evolution. We used these assumptions to calculate the expected LF
366: in the $z=0.5-1.5$ interval. (The use of the S02 or the MG01 measurements,
367: which were constructed using more similar
368: methodologies to ours for classifying the galaxies, does not change
369: the diagrams significantly.) This simple model provides a reasonable
370: fit to our measured LF, with the AGN number
371: density evolving as $(1+z)^{3}$, and the star-forming luminosities
372: evolving as either $(1+z)^{3}$ ({\it dotted line}), if we match the
373: observed LF, or as $(1+z)^{3.8}$ ({\it dashed line}), if we match
374: the maximum total LF. This systematic uncertainty is much larger than
375: the statistical errors and primarily affects the evolution of the
376: star-forming LF because the uncertainties are larger at the lower
377: luminosities where this term dominates.
378: Our results for the evolution of the star-forming LF are
379: consistent with those determined from local samples by MG01
380: and \markcite{condon02}Condon et al.\ (2002),
381: who found a similar evolution of $(1+z)^{3\pm1}$.
382: Our results are also broadly consistent with those of H00, who
383: found a best pure luminosity evolution model with $(1+z)^{2.9}$.
384:
385: In Figures~\ref{fig3}b and \ref{fig3}c we show the measured LFs
386: in two narrower intervals
387: ($z=0.6-0.9$ and $z=0.9-1.4$, respectively, chosen to
388: match H00) and by spectroscopic class ({\it diamonds}---star formers;
389: {\it squares}---AGNs). For
390: the star formers, we also show the results of H00 ({\it triangles}),
391: which, within the uncertainties, agree with the present measurements.
392: Our results are again broadly consistent with
393: the evolution described above (see Fig.~\ref{fig3}b caption for
394: description), although the AGN LF is slightly
395: high in the $z=0.6-0.9$ interval. This suggests that the
396: power law evolution model may be too simple and that the
397: evolution of the AGN LF is faster at low redshifts and slower at high
398: redshifts. The star formers begin to show a more extended high
399: luminosity tail in the $z=0.9-1.4$ interval.
400:
401: %
402: % Figure 3
403: %
404: \begin{inlinefigure}
405: \psfig{figure=f3.eps,width=3.5in}
406: \vspace{6pt}
407: \figurenum{3}
408: \caption{
409: (a) Radio LF at $z=0.5-1.5$ for all the spectroscopically
410: identified radio sources ({\it triangles}), compared with
411: the local determination from Condon (1989; {\it solid line}).
412: Log denotes the base 10 logarithm. The $1\sigma$ uncertainties are
413: Poissonian. Squares denote the maximum total LF.
414: Vertical dotted line shows the equivalent
415: radio power of an $L_{FIR}=10^{12}~L_\odot$ source.
416: Dotted curve shows the expected LF if the local star-forming LF
417: undergoes pure luminosity evolution ($(1+z)^3$) and the local AGN
418: LF undergoes pure number density evolution ($(1+z)^3$).
419: Uncertainties are primarily systematic, and luminosities
420: evolving as $(1+z)^{3.8}$ provide a better fit
421: ({\it dashed line}) to the maximum total LF.
422: (b) Radio LF at $z=0.6-0.9$ for the star formers ({\it diamonds})
423: and the AGNs ({\it squares}), to be compared with the local
424: star former ({\it steep solid curve}) and AGN
425: ({\it shallow solid curve}) determinations from Condon (1989).
426: Triangles show the measurements of H00 (with a small correction
427: to make the geometries consistent) for the star formers.
428: Dot-dashed curve shows the expected LF if the local star-forming
429: LF undergoes pure luminosity evolution ($(1+z)^{3.8}$).
430: Dashed curve shows the expected LF if the local AGN LF
431: undergoes pure number density evolution ($(1+z)^{3}$).
432: (c) As for (b), but for the $z=0.9-1.4$ redshift interval.
433: \label{fig3}
434: }
435: \addtolength{\baselineskip}{10pt}
436: \end{inlinefigure}
437:
438: The rapid evolution of the radio LF implies
439: that the number density of very luminous sources rises
440: rapidly in the $z=0-1$ interval for both the
441: star former and AGN classes though it may
442: be begining to slow at redshifts $z>1$.
443: In Figure~4 we show the number
444: density of star formers with ULIG radio power ({\it solid squares})
445: in three redshift intervals.
446: The value from the local star-forming radio LF is given by the
447: open square. The evolution is reasonably well described by a $(1+z)^7$
448: evolution ({\it dashed line}). This evolution is similar
449: to that inferred by \markcite{bcr00}Barger et al.\ (2000)
450: from a comparison of the local number density
451: of ULIGs and near-ULIGs to the $z=1-3$ number density of $>6$~mJy
452: radio/submillimeter sources ({\it solid circle}).
453: If most of the submillimeter sources are dominated by star
454: formation (\markcite{bcr00}Barger et al.\ 2000;
455: \markcite{chapman03}Chapman et al.\ 2003), the
456: match to the radio evolution is reasonable, if the radio
457: galaxies classified as star formers are indeed dominated by
458: star formation and the FIR-radio correlation holds to
459: these redshifts as suggested by recent results (Garrett 2002).
460:
461: %
462: % Figure 4
463: %
464: \begin{inlinefigure}
465: \psfig{figure=f4.eps,angle=90,width=3.5in}
466: \vspace{6pt}
467: \figurenum{4}
468: \caption{
469: Number density of star formers with ULIG radio power
470: in the redshift intervals $z=0.1-0.6$, $z=0.6-1.0$, and $z=1.0-1.4$
471: ({\it solid squares}). Open square denotes the value from the
472: local star-forming radio LF.
473: Solid circles denote the local number density of ULIGs and near-ULIGs
474: and the $z=1-3$ number density of $>6$~mJy radio/submillimeter
475: sources (from Barger et al.\ 2000).
476: Dashed line shows a $(1+z)^7$ evolution.
477: \label{fig4}
478: }
479: \addtolength{\baselineskip}{10pt}
480: \end{inlinefigure}
481:
482: \section{Summary}
483: \label{secsummary}
484:
485: We have given a precise determination of
486: the high power end of the radio LF at high redshifts
487: ($z\sim 1$). The LF was shown to evolve rapidly with
488: redshift, both for galaxies with AGN spectra and for those
489: with only star-forming signatures. This result is consistent
490: with model expectations based on the local LF and the radio
491: number counts (e.g., \markcite{condon89}Condon 1989).
492: The number of sources with radio luminosities that would
493: correspond to ULIGs (based on local normalizations) matches
494: the observed evolution from the local ULIG population to the
495: radio/submillimeter sources at $z>1$.
496:
497: \acknowledgements
498: We thank the referee for helpful comments that improved the manuscript.
499: L.~L.~C. gratefully acknowledges support from NASA grants GO2-3187B
500: and HST-GO-09425.03-A and NSF grant AST-0084816, and
501: A.~J.~B. from NASA grant HST-GO-09425.30-A,
502: NSF grant AST-0084847, NSF CAREER award AST-0239425,
503: the UW Research Committee, the Alfred P. Sloan
504: Foundation, and the David and Lucile Packard Foundation.
505:
506: \newpage
507:
508: \begin{references}
509:
510: \reference{alex03}
511: Alexander, D. M., et al.\ 2003, \aj, 126, 539
512:
513: \reference{barger01}
514: Barger, A. J., Cowie, L. L., Mushotzky, R. F.,
515: \& Richards, E. A. 2001, \aj, 121, 662
516:
517: \reference{bcr00}
518: Barger, A. J., Cowie, L. L., \& Richards, E. A. 2000,
519: \aj, 119, 2092
520:
521: \reference{benitez00}
522: Ben\'{\i}tez, N.\ 2000, 536, 571
523:
524: \reference{capak03}
525: Capak, P., et al.\ 2003, \aj, 127, 180
526:
527: \reference{capak04}
528: Capak, P., et al.\ 2004, in preparation
529:
530: \reference{chapman03}
531: Chapman, S. C., Blain, A. W., Ivison, R. J., \& Smail, I.\ 2003,
532: \nat, 422, 695 %redshift paper
533:
534: \reference{condon89}
535: Condon, J. J.\ 1989, \apj, 338, 13
536:
537: \reference{condon92}
538: Condon, J. J.\ 1992, ARA\&A, 30, 575
539:
540: \reference{condon02}
541: Condon, J. J., Cotton, W. D., \& Broderick, J. J.\ 2002,
542: \aj, 124, 675
543:
544: \reference{cram98}
545: Cram, L. E.\ 1998, \apj, 506, L85
546:
547: \reference{faber03}
548: Faber, S. M., et al.\ 2003, Proc. SPIE, 4841, 1657
549:
550: \reference{felten77}
551: Felten, J. E.\ 1977, \aj, 82, 861
552:
553: \reference{fomalont04}
554: Fomalont, E. B., Kellermann, K. I., Cowie, L. L., Barger, A. J.,
555: Capak, P., Partridge, R. B., \& Windhorst, R. A.\ 2004,
556: \apj, in preparation
557:
558: \reference{garrett02}
559: Garrett, M.A. \ 2002, A\&A, 384, L19
560:
561: \reference{gispert00}
562: Gispert, R., Lagache, G., \& Puget, J. L.\ 2000, A\&A, 360, 1
563:
564: \reference{haarsma00}
565: Haarsma, D. B., Partridge, R. B., Windhorst, R. A., \&
566: Richards, E. A. \ 2000, \apj, 544, 641 (H00)
567:
568: \reference{kim98}
569: Kim, D.-C., \& Sanders, D. B.\ 1998, \apjs, 119, 41
570:
571: \reference{lilly99}
572: Lilly, S. J., et al.\ 1999, \apj, 518, 641
573:
574: \reference{mac01}
575: Machalski, J., \& Godlowski, W.\ 2001, A\&A, 370, 923 (MG01)
576:
577: \reference{mobasher99}
578: Mobasher, B., Cram, L., Georgakakis, A., \& Hopkins, A.\ 1999, \mnras,308, 45
579:
580: \reference{moran99}
581: Moran, E. C., Lehnert, M. D., \& Helfand, D. J.\ 1999, \apj, 526
582:
583: \reference{mush00}
584: Mushotzky, R. F., Cowie, L. L., Barger, A. J., \& Arnaud, K. A.\ 2000,
585: \nat, 404, 459
586:
587: \reference{oke95}
588: Oke, J. B., et al.\ 1995, \pasp, 107, 375
589:
590: \reference{richards00}
591: Richards, E. A.\ 2000, \apj, 533, 611
592:
593: \reference{sadler02}
594: Sadler, E. M., et al.\ 2002, \mnras, 329, 227 (S02)
595:
596: \reference{sanders96}
597: Sanders, D. B., \& Mirabel, I. F.\ 1996, ARA\&A, 34, 749
598:
599: \reference{zezas98}
600: Zezas, A., Georgantopoulos, I., \& Ward, M. J.\ 1998, \mnras, 30
601:
602: \end{references}
603:
604: \begin{deluxetable}{rrrrrrrr}
605: \renewcommand\baselinestretch{1.0}
606: \tablenum{1}
607: \tablewidth{0pt}
608: \tablecaption{Breakdown of spectral type by field and radio flux ($\mu$Jy)}
609: \small
610: \tablehead{CATEGORY & HDF & SSA13 & ALL & 25-50 & 50-100 & 100-200 & $>$~200
611: }
612: \startdata
613: total & 195 & 151 & 346 & 37 & 186 & 77 & 46 \cr
614: identified & 115 & 82 & 197 & 20 & 98 & 51 & 28 \cr
615: stars & 0 & 2 & 2 & 0 & 2 & 0 & 0 \cr
616: absorbers & 15 & 14 & 29 & 1 & 16 & 4 & 8 \cr
617: star formers & 57 & 49 & 106 & 10 & 51 & 33 & 12 \cr
618: Seyferts & 36 & 13 & 49 & 8 & 23 & 12 & 6 \cr
619: broad-line & 5 & 3 & 8 & 0 & 5 & 1 & 2 \cr
620: \enddata
621: \label{tab1}
622: \end{deluxetable}
623:
624: \end{document}
625: