Far-Ultraviolet Imaging of the Hubble Deep Field North: Star Formation in Normal Galaxies at z<1

Abstract

We present far-ultraviolet (FUV) imaging of the Hubble Deep Field North (HDF-N) taken with the Solar Blind Channel of the Advanced Camera for Surveys (ACS/SBC) and the FUV MAMA detector of the Space Telescope Imaging Spectrograph (STIS) onboard the Hubble Space Telescope. The full WFPC2 deep field has been observed at 1600 Angstroms. We detect 134 galaxies and one star down to a limit of FUV_{AB} ~ 29. All sources have counterparts in the WFPC2 image. Redshifts (spectroscopic or photometric) for the detected sources are in the range 0<z<1. We find that the FUV galaxy number counts are higher than those reported by GALEX, which we attribute at least in part to cosmic variance in the small HDF-N field of view. Six of the 13 Chandra sources at z<0.85 in the HDF-N are detected in the FUV, and those are consistent with starbursts rather than AGN. Cross-correlating with Spitzer sources in the field, we find that the FUV detections show general agreement with the expected L_IR/L_UV vs. Beta relationship. We infer star formation rates (SFRs), corrected for extinction using the UV slope, and find a median value of 0.3 Msun/yr for FUV-detected galaxies, with 75% of detected sources have SFR<1 Msun/yr. Examining the morphological distribution of sources, we find that about half of all FUV-detected sources are identied as spiral galaxies. Half of morphologically-selected spheroids at z<0.85 are detected in the FUV, suggesting that such sources have significant ongoing star-formation in the epoch since z=1.

Full text

Page 1

arXiv:astro-ph/0606244v1 10 Jun 2006
Far-Ultraviolet Imaging of the Hubble Deep Field North1: Star
Formation in Normal Galaxies at z < 1
H. I. Teplitz2, B. Siana2, T. M. Brown3, R. Chary2, J. W. Colbert2, C. Conselice4, D. F. de
Mello5,6,7, M. Dickinson8, H. C. Ferguson3, Jonathan P. Gardner5F. Menanteau7
ABSTRACT
We present far-ultraviolet (FUV) imaging of the Hubble Deep Field North
(HDF-N) taken with the Solar Blind Channel of the Advanced Camera for Sur-
veys (ACS/SBC) and the FUV MAMA detector of the Space Telescope Imaging
Spectrograph (STIS) onboard the Hubble Space Telescope. The full WFPC2 deep
field has been observed at 1600 Angstroms. We detect 134 galaxies and one star
down to a limit of FUVAB∼ 29. All sources have counterparts in the WFPC2
image. Redshifts (spectroscopic or photometric) for the detected sources are in
the range 0 < z < 1. We find that the FUV galaxy number counts are higher than
those reported by GALEX, which we attribute at least in part to cosmic variance
in the small HDF-N field of view. Six of the 13 Chandra sources at z < 0.85 in
the HDF-N are detected in the FUV, and those are consistent with starbursts
rather than AGN. Cross-correlating with Spitzer sources in the field, we find that
the FUV detections show general agreement with the expected LIR/LUVvs. β
relationship. We infer star formation rates (SFRs), corrected for extinction using
1Based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the Space
Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy,
Inc., under NASA contract NAS 5-26555. These observations are associated with proposals 7410 and 9478.
2Spitzer Science Center, MS 220-6, Caltech, Pasadena, CA 91125. hit@ipac.caltech.edu
3Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218
4University of Nottingham, Nottingham, NG7 2RD, UK
5Exploration of the Universe Division, Observational Cosmology Laboratory, Code 665, Goddard Space
Flight Center, Code 681, Greenbelt, MD 20771
6Department of Physics, Catholic University of America, 620 Michigan Avenue, Washington, DC 20064
7Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore,
MD 21218
8NOAO, 950 N. Cherry Ave., Tucson AZ 85719

Page 2

– 2 –
the UV slope, and find a median value of 0.3 M⊙/yr for FUV-detected galaxies,
with 75% of detected sources have SFR< 1 M⊙/yr. Examining the morphological
distribution of sources, we find that about half of all FUV-detected sources are
identied as spiral galaxies. Half of morphologically-selected spheroids at z < 0.85
are detected in the FUV, suggesting that such sources have significant ongoing
star-formation in the epoch since z ∼ 1.
Subject headings: cosmology: observations — galaxies: evolution — ultraviolet:
galaxies
1.Introduction
The star formation rate density of the Universe, integrated over all galaxy populations,
shows a sharp decline since redshifts near unity (e.g., Madau et al. 1996, 1998). While
the precise shape of the decline with redshift is still uncertain (Lilly et al. 1996; Hogg et
al. 1998; Flores et al. 1999; Wilson et al. 2002), its existence points to a “downsizing” in
galaxies that host most of the star formation at z < 1 (Cowie et al. 1996). The characteristics
(morphology, mass, luminosity) of these low redshift starbursts may explain the global decline
in star formation. Wolf et al. (2005) suggest that the decline is dominated by decreasing star
formation in normal spiral galaxies rather than, for example, the decreasing rate of major
mergers.
Ultraviolet (UV) emission is an indication of recent star formation in a galaxy. Despite
absorption by dust, the rest-frame UV is strong enough in the majority of star-forming
galaxies to be detected in current surveys (Adelberger & Steidel 2000). UV detection can
distinguish star-forming from quiescent systems, and indicates the amount of recent star
formation, subject to the effects of extinction by dust (e.g., Calzetti et al. 1994; Fitzpatrick
1986; Meurer et al. 1999; Buat et al. 2005). Far-ultraviolet (FUV) surveys, in particular,
can provide direct evidence of recently formed, massive stars in the dominant populations
at z < 1 (Schiminovich et al. 2005).
Of particular interest is the the star-formation activity present in early type galaxies.
Recent studies have shown that some galaxies which appear morphologically to be spheroids
have, nonetheless, substantial ongoing star formation. Yi et al. (2005) find that at least 15%
of bright, local ellipticals show evidence of recent star formation, ruling out pure monolithic
collapse histories for at least that fraction of such sources. Similarly, studies of the internal
color variations in elliptical galaxies have shown almost one third of them show gradients
inconsistent with passive evolution (Abraham et al. 1999; Menanteau et al. 2001; Papovich

Page 3

– 3 –
et al. 2003). The formation of spheroid galaxies, then, is a crucial component of successful
hierarchical models. Conselice et al. (2005) find that the massive galaxies at z < 1, both
spirals and ellipticals, likely have major-merger progenitors at higher redshifts. Nonetheless,
a significant fraction of stellar mass must still have formed since z ∼ 1 (Bell et al. 2004;
Dickinson et al. 2003). The assembly of that additional stellar mass should be detectable in
UV surveys.
We present a far-ultraviolet (1600˚ A; FUV) imaging survey of the Hubble Deep Field
North (HDF-N; Williams et al. 1996). The data were taken in two surveys.
utilized the FUV camera of the Space Telescope Imaging Spectrograph (STIS; Kimble et
al. 1998; Woodgate et al. 1998) to survey a small part of the field. That portion of the
data set has been been used to measure galaxy number-magnitude counts (Gardner et al.
2000a) and the diffuse FUV background emission (Brown et al. 2000a). We surveyed the
remaining area with the Solar Blind Channel (SBC) of the Advanced Camera for Surveys
(Ford et al. 1998). We outline the survey and data reduction in Section 2 and present the
catalog in Section 3. In Section 4, we discuss the properies of FUV-detected sources at
other wavelengths, the inferred star formation rates, and the implications of the detection
of elliptical galaxies. Throughout, we assume a Λ-dominated flat universe, with H0= 70
km s−1Mpc−1, ΩΛ= 0.7,Ωm= 0.3. Photometry is presented with magnitudes on the AB
system which is defined by AB = −2.5logFν− 48.6, where Fνis given in units of ergs cm−2
s−1Hz−1(Oke 1971).
The first
2. Observations and Data Reduction
FUV imaging of the HDF-N was obtained in two HST General Observer programs (No.
7410 with STIS and No. 9478 with ACS/SBC). The ACS survey is composed of fourteen
2-orbit pointings covering 3.77 square arcminutes. Each pointing consisted of 16×640 sec-
onds exposures with dithers of ∼10 pixels. The field of view of the ACS/SBC detector is
34.6′′×30.8′′. The images were obtained with the long pass quartz filter (F150LP) with an
effective wavelength of 1614˚ A and FWHM = 177˚ A. The STIS survey covered 1.02 square
arcminutes in six pointings for a combined exposure time of 124,330 seconds. The field of
view of the STIS detector is 25′′×25′′. The images were obtained through the crystal quartz
filter (F25QTZ) with a central wavelength of 1595˚ A and FWHM = 193˚ A.
Both the STIS and ACS/SBC filters are long-pass filters with a short wavelength cutoff
at λ < 1480˚ A. The STIS and ACS/SBC detectors are both Multi-Anode Microchannel
Arrays (MAMAs) and have similar spectral response curves which fall off slowly from 1500
to 1800˚ A. Therefore, the combined filter+detector system response curves of the STIS and

Page 4

– 4 –
ACS/SBC FUV configurations is similar, with the only significant difference being that the
ACS/SBC throughput is non-zero at 1850˚ A < λ < 2000˚ A (Figure 1).
Data for both surveys were reduced following the procedure outlined in Gardner et al.
(2000b). Full details of the STIS data reduction are given in that paper, and ACS-specific
reductions are discussed here.
The MAMA detector has no read noise and is insensitive to cosmic rays. The primary
source of noise is dark current, which has two components. When the temperature of the
MAMA is below 20C, the dark current is fairly uniform with an average count rate of
∼ 8 × 10−6counts s−1pixel−1. However, as the SBC is being used, the MAMA warms
up and produces an additional temperature-dependent dark “glow” near the center of the
detector. Therefore, we subtract the dark current in two stages. First we subtract the
primary calibration dark which was made from darks collected at T<20C. We then make
a residual dark by summing up all of the initial dark subtracted frames and fitting a two
dimensional, fifth-order spline curve. The isophotal segmentation maps from the HDF-North
Wide-Field Planetary Camera 2 (WFPC2) V+I images (Williams et al. 1996) were used to
mask areas contaminated by known objects using the blot capability of the DRIZZLE package
in IRAF. We then subtracted the residual dark after scaling to the average of a region near
the peak of the dark “glow”. We find that < 2% of the secondary dark remains after this
second subtraction.
The dark rate of the secondary “glow” near the center of the chip is typically larger
than that of the initial (i.e., at T < 20C) dark rate and is a function of the temperature of
the MAMA tube. The MAMA tube gets warmer as the SBC is used, and we find that the
count rate increases linearly with time with the rate near the center of the chip increasing
by ∼ 2.5 × 10−5counts−1s−1pixel−1hour−1(see Figure 2). Equivalently, this amounts
to increasing the dark rate by an amount equal to the “cold” dark rate every 20 minutes.
Some of our scheduled visits were longer than 6 hours with entire pointings done at the
end of the visit. Therefore the dark rate for these pointings was a factor of 10-20 times
larger than in those taken at the beginning of a visit, resulting in a decrease in sensitivity of
∼1-2 magnitudes in the “glow” regions. Therefore, large SBC programs in the future would
greatly benefit from segmenting their observations into multiple visits.
Standard calibration files were used for flat fielding, geometric distortion correction and
photometric calibration. Individual reduced images were registered and summed using the
DRIZZLE package in IRAF1. Both ACS and STIS data were drizzled to the pixel scale of
the HDF-N WFPC2 data products (0.03985′′pixel−1), and matched to the WFPC2 pixel
1IRAF is distributed by NOAO, which is operated by AURA Inc., under contract to the NSF

Page 5

– 5 –
positions. Shifts between the 16 dithered positions of each pointing were assumed to be
exactly as commanded. This assumption is reasonable, given the small-offset accuracy of
HST (3-5 mas). The registration of the 14 pointings was done by matching the FUV sources
to the B450WFPC2 image and computing shifts, rotation, and scaling with the GEOMAP
routine in IRAF. For all 14 pointings, the rms fit was < 0.03′′.
As discussed by Gardner et al. (2000b) and Brown et al. (2000a), the dark current is
the principle source of noise in MAMA imaging. The individual frames were weighted by
the square of the exposure time, divided by the total dark (initial + secondary). As the dark
count scales with exposure time, these weight maps scale linearly with the ratio of exposure
time to dark rate. Therefore, the final weight maps are the square of the signal-to-noise ratio
for objects fainter than the background.
Photometry was performed by summing the pixel values within the source areas defined
by the 3.25σ isophotes (where σ is the rms background noise) in the V+I WFPC2 segmen-
tation map produced by SExtractor (Bertin & Arnouts 1996). The V+I image was used
because it is the most sensitive image, and therefore the 3.25σ isophotes are contiguous (i.e.
individual galaxies are not broken into several isophotes) and encompass the large majority
of the galactic light for all but the faintest galaxies. To validate this method we extracted
fluxes in the B450image with 3.25σ and 0.65σ V+I isophotes. The fluxes derived with the
smaller 3.25σ isophotes missed 10–15% of the flux within the larger 0.65σ isophote, but were
significantly less noisy. We therefore use the smaller isophotes for detection and the ratio of
the two fluxes (in the F450W image) as an aperture correction. Again, we verified that these
“aperture corrected” UV fluxes agreed well with the fluxes derived in the larger aperture,
but with smaller errors. For the few objects with aperture corrections larger than 40%, the
larger apertures were used.
3.Results
Figure 3 shows the fully reduced, registered FUV mosaic. We detect 128 sources above
a signal-to-noise ratio, SNR> 3.5. We add an additional 7 sources by hand because their
UV flux is more compact and the larger V+I segmentation map causes large errors in the
flux estimates. The FUV properties of the 135 sources are given in Table 1. The detection
limits vary significantly between pointings due to large variations in dark “glow”. In those
regions least affected by the dark “glow”, the 3.5σ limiting magnitudes are FUVAB= 29.2
in the STIS survey and FUVAB= 28.8 in the ACS Survey for a 1′′diameter aperture.
Published spectroscopic redshifts are available for 60 of the 135 FUV detected sources