The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 doi:10.

1088/2041-8205/739/1/L33

C 2011. The American Astronomical Society. All rights reserved. Printed in the U.S.A.

EXPANDED VERY LARGE ARRAY OBSERVATIONS OF A PROTO-CLUSTER OF

MOLECULAR GAS-RICH GALAXIES AT z = 4.05∗
C. L. Carilli1 , J. Hodge2 , F. Walter2 , D. Riechers3,7 , E. Daddi4 , H. Dannerbauer2 , and G. E. Morrison5,6
1
National Radio Astronomy Observatory, P.O. Box 0, Socorro, NM 87801-0387, USA
2 Max-Planck Institute for Astronomy, Konigstuhl 17, 69117, Heidelberg, Germany
3 Department of Astronomy, California Institute of Technology, MC 249-17, 1200 East California Boulevard, Pasadena, CA 91125, USA
4 Laboratoire AIM, CEA/DSM-CNRS-University Paris Diderot, DAPNIA/Service Astrophysccique, CEA Saclay,
Orme des Merisiers, 91191 Gif-sur-Yvette, France
5 Institute for Astronomy, University of Hawaii, Honolulu, HI 96822, USA
6 Canada–France–Hawaii Telescope, Kamuela, HI 96743, USA
Received 2011 April 15; accepted 2011 May 17; published 2011 August 29

ABSTRACT
We present observations of the molecular gas in the GN20 proto-cluster of galaxies at z = 4.05 using the Expanded
Very Large Array (EVLA). This group of galaxies is the ideal laboratory for studying the formation of massive
galaxies via luminous, gas-rich starbursts within 1.6 Gyr of the big bang. We detect three galaxies in the proto-
cluster in CO 2–1 emission, with gas masses (H2 ) between 1010 and 1011 × (α/0.8) M . The emission from
the brightest source, GN20, is resolved with a size ∼2 and has a clear north–south velocity gradient, possibly
indicating ordered rotation. The gas mass in GN20 is comparable to the stellar mass (1.3 × 1011 × (α/0.8) M
and 2.3 × 1011 M , respectively), and the sum of gas plus stellar mass is comparable to the dynamical mass of the
system (∼3.4 × 1011 [sin(i)/ sin(45◦ )]−2 M ), within a 5 kpc radius. There is also evidence for a tidal tail extending
another 2 north of the galaxy with a narrow velocity dispersion. GN20 may be a massive, gas-rich disk that is
gravitationally disturbed, but not completely disrupted. There is one Lyman-break galaxy (BD29079) in the GN20
proto-cluster with an optical spectroscopic redshift within our search volume, and we set a 3σ limit to the molecular
gas mass of this galaxy of 1.1 × 1010 × (α/0.8) M .
Key words: galaxies: evolution – galaxies: ISM – galaxies: starburst – submillimeter: galaxies
Online-only material: color figure

1. INTRODUCTION of higher order CO emission from a sample of z ∼ 2 SMGs, that

SMGs are predominantly nuclear starbursts, with median sizes
Numerous lines of evidence support the hypothesis that mas- <0. 5 (<4 kpc), “representing extreme, short-lived, maximum
sive elliptical galaxies form the majority of their stars quickly star-forming events in highly dissipative mergers of gas-rich
at early epochs, including stellar population synthesis studies of galaxies.” This conclusion is supported by very long baseline
nearby elliptical galaxies (Renzini 2006; Collins et al. 2009), interferometry imaging of the star-forming regions in two SMGs
a faster decline with cosmic time in the star formation rate per (Momjian et al. 2005, 2010; Biggs et al. 2010). However, recent
unit stellar mass with increasing galaxy mass (Moresco et al. imaging of the lower order CO emission in a few SMGs (Ivison
2010), and the direct observation of “red and dead” ellipticals et al. 2010, 2011; Carilli et al. 2010; Riechers et al. 2011)
in clusters at high redshift (Kurk 2009; Doherty et al. 2009; suggests that the lower-excitation molecular gas reservoirs can
Andreon & Huertas-Company 2011). Plausible progenitors for be significantly more extended.
these evolved galaxies at high redshift are even higher redshift Daddi et al. (2009b) have identified a unique region in
submillimeter galaxies (SMGs), corresponding to dusty, lumi- GOODS-N, containing the three galaxies: GN20, GN20.2a,
nous starburst galaxies found in wide-field submillimeter sur- and GN20.2b, at z ∼ 4.05 ± 0.01, within 20 of each other
veys (Blain et al. 2002). These hyper-luminous high-z galaxies (projected physical separation = 140 kpc). GN20 is among
may trace high overdensities and are likely related to the forma- the brightest SMGs known, with a 350 GHz flux density
tion of clusters and large ellipticals (Stevens et al. 2003; Aravena of 23 mJy. The galaxies GN20.2a and b are separated by
et al. 2010; although cf. Chapman et al. 2009; Williams et al. only a few arcseconds, and hence are not spatially distinct in
2011). While the space density of SMGs peaks around z ∼ 2.3 typical submillimeter bolometer images. The two galaxies have
(Chapman et al. 2003), it is becoming clear that there is a sub- a summed flux density of S350 GHz ∼ 9.9 mJy. The implied
stantial high-redshift tail of the SMG population, extending to IR luminosities (8–1000 μm) are LTIR  1013 L , with star
z > 4 (Capak et al. 2011; Riechers et al. 2010; Wardlow et al. formation rates >1000 M yr−1 (Daddi et al. 2009b). High-
2011). Daddi et al. (2009b) conclude, based on SMG space order CO emission (4–3 and/or 6–5) has been detected from all
densities and duty cycles, that there are likely enough SMGs at three galaxies, although only marginally in the case of GN20.2b
z > 3.5 to account for the known populations of old massive (Daddi et al. 2009b).
galaxies at z ∼ 2–3. The GN20 field also contains numerous Lyman-break galax-
A key question for the SMGs is: what drives the prolific star ies (LBGs) at similar redshift and an overdensity of zphot > 3.5
formation? Tacconi et al. (2006, 2008) argue, based on imaging Infrared Array Camera selected galaxies (Daddi et al. 2009b).
∗ The Very Large Array of the National Radio Astronomy Observatory is a
Hence, this field offers an unprecedented opportunity to per-
form a definitive study of the gas distribution, kinematics, and
facility of the National Science Foundation operated under cooperative
agreement by Associated Universities, Inc. physical conditions in a proto-cluster of galaxies within 1.6 Gyr
7 Hubble Fellow. of the big bang. In this Letter, we present a survey with the

1
The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 Carilli et al.

Expanded Very Large Array (EVLA; Perley et al. 2011) of the galaxies. The total observing time was 28 hr, but only about 22
CO 2–1 emission from the GN20 proto-cluster. The low order to 24 antennas were available at a time due to ongoing EVLA
transitions are critical for determining the total gas mass. We work.
also search for CO emission from other galaxies in the cosmic Dynamic scheduling to ensure good weather is now standard
volume surveyed. at the EVLA. Fast switching phase calibration was employed
(Carilli & Holdaway 1999) on timescales between two and
2. THE GN20 PROTO-CLUSTER: A MASSIVE GALAXY three minutes using the VLA calibrator J1302+5748. Data were
FORMATION LABORATORY edited to remove time ranges of poor phase stability. The source
3C286 was used for flux density calibration. Standard EVLA
The Hubble Space Telescope (HST) I-band image of the bright data calibration and editing were performed using AIPS. After
SMG GN20 shows diffuse emission about 1. 5 in extent (Daddi calibration and data editing, images were synthesized using the
et al. 2009b), although offset from the radio and submillimeter robust weighting scheme of the uv data with R = 2 (Cornwell
emission by ∼1 , implying high obscuration of the most active et al. 1999).
star-forming regions in the galaxy. GN20 is detected at 1.4 GHz The final resolution for the images was 1. 9. Spectral line
and resolved on a scale of ∼1. 5, with a total flux density at cubes were generated at 12 MHz channel−1 (78 km s−1 ). The
72 ± 13 μJy (Morrison et al. 2010; Casey et al. 2009). High- rms noise per channel at 12 MHz resolution is 0.11 mJy beam−1 .
resolution imaging of the 850 μm emission shows resolved All images were corrected for the VLA primary beam response,
structure, with a north–south extension possibly as large as 1. 5 which has a FWHM ∼1 at 45 GHz. The field pointing center
(Younger et al. 2008; Iono et al. 2006). The 6.2 μm polycyclic was located 10 west of GN20.
aromatic hydrocarbon spectral feature has been detected in
GN20 using Spitzer (D. Riechers et al. 2011, in preparation). 4. RESULTS
The integrated CO 1–0 and 2–1 emission from GN20 has been
imaged at resolutions down to 0. 2 with the Very Large Array Figure 1 shows the CO 2–1 emission integrated over the full
(VLA; Carilli et al. 2010). The molecular gas is extended on frequency range covered by these observations for the GN20
a scale of at least 1. 5, and at high resolution forms a partial group. The crosses mark the 1.4 GHz continuum positions of
ring, or disk. Unfortunately, the old VLA correlator provided the three galaxies GN20, 20.2a,b (Morrison et al. 2010), and
no velocity information. The excitation of the integrated CO the optical position of one LBG with a spectroscopic redshift
emission from GN20 is higher than the Milky Way, but lower in the redshift range of the CO 2–1 observations (Section 5.2).
than high-redshift quasar hosts and the nuclear starburst regions We have added the previously published D array observa-
of IR-luminous galaxies. The CO 4–3 line strength in GN20 tions (Carilli et al. 2010) to improve the signal-to-noise ratio
is more than a factor of two lower than expected for thermal (S/N). Unfortunately, the previous data did not have the exact
excitation (Carilli et al. 2010). same velocity coverage, and hence the velocity-integrated flux
GN20.2a shows a complex and extended morphology in densities are not accurate. Hence, this image acts as a finding
the HST images (Daddi et al. 2009b), and it is a relatively chart for the CO-emitting regions, but accurate line fluxes will
strong radio source (S1.4 = 181 μJy), likely corresponding to be based on subsequent analysis of the new EVLA data itself.
a radio AGN. The spectral energy distribution (SED) in the We detect CO 2–1 emission from GN20, GN20.2a,
optical through near-IR is consistent with a star-forming galaxy. and GN20.2b. GN20 is extended on a scale ∼2 (see be-
GN20.2a has been detected in CO 4–3 emission by Daddi et al. low). GN20.2a is not spatially resolved by these observations.
(2009b). Daddi et al. derive an IR luminosity (8–1000 μm) of GN20.2b may be extended, although the S/N of these data is
LIR = 1.6 × 1013 L . not conclusive.
GN20.2b is faint and compact in the HST image. It is also Figure 2 shows spectra of the three sources. For GN20, we
detected in the radio continuum at S1.4 = 32 μJy, consistent integrate over an area of 2 . For the other two sources, we only
with star formation. CO 4–3 emission is marginally detected consider the spectra at the peak of the velocity-integrated CO
with the Plateau de Bure Interferometer (PdBI) and appears to image. Results for Gaussian fits to the lines are given in Table 1.
be broad (700 km s−1 ; Daddi et al. 2009b). The lines are broad for GN20 and GN20.2a (∼800 km s−1 ),
Daddi et al. (2009b) find 15 B-band dropout galaxies in a although not outside the range of the zero intensity line widths
25 radius centered on GN20, an overdensity of a factor of six seen in some nearby ULIRGs (Downes & Solomon 1998).
compared with the full GOODS-N area, which is significant at However, we note that due to the bandwidth limitation for
the 7σ level. A spike in the redshift distribution of galaxies the early EVLA the very lowest frequency (highest velocity)
at z = 4.06 ± 0.02 is observed in all of GOODS-N (13 emission for GN20 may be truncated by these observations,
spectroscopic redshifts in total at this redshift). Lastly, the SMG although the higher order CO transitions show that we expect
GN10 at z = 4.04 is located 9 from GN20 (Daddi et al. 2009a; little emission beyond our lowest spectral channel (Daddi et al.
Dannerbauer et al. 2008). Therefore, it appears that the GN20 2009b). GN20.2b shows a narrower profile at the peak position.
volume has a very significant overdensity, indicating a proto- However, the detection in the velocity-integrated image over the
cluster environment at z ∼ 4.05. Daddi et al. (2009b) estimate full frequency range (Figure 1) suggests that the emission may
a total mass for this structure of ∼1014 M . be extended both spatially and spectrally. A possibly broader
line is also suggested for the higher order CO emission, but
3. OBSERVATIONS again, the high-order CO detection is only marginal (Daddi et al.
2009b). We have investigated this possibility with the current
We observed the GN20 field with the EVLA in the D (1 km) data, and conclude that more sensitive observations are required
configuration in 2010 March and April. Observations were made to better characterize the CO emission from GN20.2b, and we
of the CO 2–1 line, using a total bandwidth of 246 MHz and in simply present the spectrum at the peak position herein.
each of two polarizations, centered at 45.655 GHz. The velocity Figure 3 shows the velocity channel images for the GN20
coverage is 1600 km s−1 , including the CO 2–1 lines in the three CO 2–1 emission at 78 km s−1 channel−1 and 1. 9 resolution.

2
The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 Carilli et al.

62 22 25

20

15

DECLINATION (J2000) 10

05

00

21 55

50

45
12 37 13 12 11 10 09 08 07
RIGHT ASCENSION (J2000)
Figure 1. CO 2–1 emission at 1. 9 resolution integrated over the full observed velocity range for the GN20 group of galaxies. In this case, no primary beam correction
is applied in order to ensure a flat noise across the image to search for sources. Crosses show the 1.4 GHz continuum positions of the three galaxies GN20, 20.2a,b
(Morrison et al. 2010), and the optical position of one LBG with a spectroscopic redshift. The contour levels are linear, in steps of 1σ (30 μJy beam−1 ), starting at
±2σ . Negative contours are dashed.
Table 1
Results from Gaussian Fitting to the CO 2–1 Line Profiles

Source Redshift Peak FWHM ICO Velocitya LCO M(H2 )

(mJy) (km s−1 ) (Jy km s−1 ) (km s−1 ) (K km s−1 pc2 ) (M )
GN20 4.0554 1.21 ± 0.11 679 ± 88 0.87 ± 0.088 −11 ± 32 1.6 × 1011 1.3 × 1011
GN20.2a 4.0508 0.54 ± 0.065 723 ± 110 0.41 ± 0.050 −283 ± 45 7.6 × 1010 6.1 × 1010
GN20.2b 4.0567 0.42 ± 0.083 290 ± 70 0.13 ± 0.025 67 ± 29 2.4 × 1010 1.9 × 1010

Note. a Velocity relative to z = 4.0556.

The emission clearly moves from the south to the north with an 5. ANALYSIS
increasing frequency. An interesting feature is seen in channel 6,
corresponding to a frequency of 45.598 GHz (or z = 4.05588). 5.1. Masses
The emission in channel 6 appears to be substantially more
extended than in other channels. Inspecting our bandpass We derive the mass in molecular gas (H2 ) from the ob-
calibration, we find no channel-dependent calibration errors that served CO 2–1 luminosities using the standard relationships
would lead to such a difference in morphology for this single in Solomon & Vanden Bout (2005). We extrapolate from CO
channel. In channel 6 the emission extends well to the north, 2–1 to 1–0 luminosity assuming constant brightness tempera-
with a total of extent of ∼4 . Admittedly, the S/N is not high, but ture, which is certainly correct to within 10% for these galaxies
this extended emission is suggestive of a gravitationally induced (Carilli et al. 2010). We adopt a CO luminosity to H2 con-
tidal feature. Such tidal tails in nearby galaxies often occur over version factor of α = 0.8 M K km s−1 pc2 , appropriate for
a narrow velocity range (Hibbard & Mihos 1995). nearby nuclear starburst galaxies (Downes & Solomon 1998).
Figure 4(a) shows the CO 2–1 emission from GN20 integrated The resulting values are listed in Table 1.
over the velocity range shown in the channel maps of Figure 3. Admittedly, the uncertainty on α is significant. A minimum
The emission appears extended, in particular to the southwest. gas mass can be derived assuming optically thin emission, and
A formal Gaussian fit to the emission yields a total flux density adopting a temperature and a CO to H2 abundance ratio. Ivison
of 0.69 ± .07 mJy, a peak surface brightness of 0.49 ± 0.03 et al. (2011) show that for reasonable assumptions (Galactic CO
Jy beam−1 , and a deconvolved source size of 1. 5 × 0. 9 with abundance in molecular clouds, and temperatures 40–50 K), the
major axis position angle (P.A.) = 69◦ . Figure 4(b) shows the lower limit to α ∼ 0.65 (see also Aalto et al. 1995).
iso-velocity contours (first moment) of the CO 2–1 emission, The gas mass in GN20 is 1.3 × 1011 × (α/0.8) M . Daddi
i.e., the weighted mean velocity derived after first blanking the et al. (2009b) derive a stellar mass of 2.3 × 1011 M from IR
channel images below 3σ channel−1 . There is a velocity gradient through optical SED fitting.
north–south, with a magnitude of ±250 km s−1 . Given the low We can obtain a very rough estimate of the gravitational
spatial resolution of these images, this corresponds to a lower mass of GN20 from the CO velocity field, although there are
limit to the true projected rotational velocity. substantial uncertainties due to the low spatial resolution of these

3
The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 Carilli et al.

for the CO ring seen in the high-resolution 2–1 observations

presented in Carilli et al. (2010), ∼5 kpc. The inclination angle
remains uncertain. The gravitational mass inside this radius
is ∼3.4 × 1011 [sin(i)/ sin(45◦ )]−2 M . This naively derived
dynamical mass is comparable to the sum of the stellar and
gas masses.

5.2. Lyman-break Galaxy Molecular Mass Limits

There is one LBG within our field with a spectro-
scopic redshift placing the CO 2–1 line within our band:
J123711.48+622155.8 at z = 4.058 (BD29079; Daddi et al.
2009b; Shim et al. 2011). We find no CO emission from this
galaxy, to a 1σ limit of 0.17 mJy beam−1 at 78 km s−1 channel−1 ,
after primary beam correction. Convolving to 312 km s−1 ,
we set a 3σ limit to the CO 1–0 luminosity of LCO 1–0 <
1.4 × 1010 K km s−1 pc2 , assuming constant brightness temper-
ature from CO 2–1 to 1–0. The implied molecular gas mass of
this galaxy is <1.1 × 1010 × (α/0.8) M . The stellar mass of
this galaxy is 2.6 × 1010 M (Daddi et al. 2009b), hence the
gas-to-stellar mass ratio is <0.53 × (α/0.8).
Daddi et al. (2009b) estimate a star formation rate of
150 M year−1 for BD29079, or an IR luminosity of 1.5 ×
1012 L . Daddi et al. (2010) have considered the “star for-
Figure 2. CO 2–1 spectra for GN20, GN20.2a, and GN20.2b at 78 km s−1
spectral resolution. Results for the Gaussian fitting are given in Table 1. mation law” in high-z galaxies, i.e., the relationship between
(A color version of this figure is available in the online journal.)
IR luminosity and CO luminosity, and find for galaxies of
this luminosity at z ∼ 2: LCO 1–0 ∼ 0.02LFIR K km s−1 pc2 .
Adopting this relation implies an expected CO luminosity of
observations. For the rotational velocity, we adopt the extremes LCO 1–0 ∼ 3 × 1010 L . Hence, BD29079 is underluminous in
of the velocity channels with significant emission in Figure 3, CO emission relative to the standard star formation law, although
or 390/ sin(i) km s−1 . For the radius, we adopt the value within the broad (factor three) scatter at any given luminosity.

62 22 16

14

12

10

08

06
DECLINATION (J2000)

62 22 16

14

12

10

08

06

62 22 16

14

12

10

08

06
12 37 12.5 12.0 11.5 12 37 12.5 12.0 11.5
RIGHT ASCENSION (J2000)
Figure 3. Channel images of the GN20 CO 2–1 spectrum at 1. 9 resolution and 78 km s−1 channel−1 . The contour levels are linear, in steps of 1σ (0.12 mJy beam−1 ),
starting at ±2σ . The channel labeled 3 is at 45.562 MHz, and each channel is 12 MHz wide. Negative contours are dashed.

4
The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 Carilli et al.

62 22 18

17
62 22 15
16

15 14
DECLINATION (J2000)

DECLINATION (J2000)
14

13
13

12
12
11

10
11

09

08 10

12 37 12.6 12.4 12.2 12.0 11.8 11.6 11.4 11.2 12 37 12.3 12.2 12.1 12.0 11.9 11.8 11.7 11.6 11.5
RIGHT ASCENSION (J2000) RIGHT ASCENSION (J2000)

Figure 4. Upper: total intensity CO 2–1 image of GN20, summing all the line channels (channels 1–12). The contour levels are linear, in steps of 1σ (30 μJy beam−1 ),
starting at ±2σ . Negative contours are dashed. Lower: velocity contours of the CO 2–1 emission from GN20 (first moment image). Contours range from −250 km s−1
to +250 km s−1 , in steps of 50 km s−1 , with zero velocity corresponding to z = 4.0556. Negative velocities are dashed.

There are 11 other LBGs within our field, and with photomet- its neighboring galaxies. This gravitational disturbance has not
ric redshifts that are consistent with the sampled range, although disrupted the disk, but has greatly enhanced star formation in the
the zphot error bars are such that the galaxies could be outside the disk, as well as generating the extended, tidal gas distribution.
velocity range of the CO observations. We have looked for CO Lastly, we have searched for CO emission from LBGs
emission at all these positions, and find no significant detections in the GN20 proto-cluster. We set an upper limit of 1.1 ×
to similar limits to those quoted above. 1010 × (α/0.8) M to the molecular gas mass in these galaxies,
We emphasize that there is a large uncertainty in the derived including one with an accurate spectroscopic redshift. Our limit
H2 mass for the LBGs, due to the unknown conversion factor, is below the expected value based on the standard star formation
α. It is possible that metal-poor star-forming galaxies have a law at this redshift, but within the broad scatter.
much larger value of α than ULIRGs or even the Milky Way We are obtaining high-resolution CO imaging observations
due to CO dissociation by the more pervasive UV radiation field of the GN20 proto-cluster with the EVLA and the PdBI to
(Madden et al. 1997; Papadopoulos & Pelupessy 2010). perform a more detailed dynamical analysis, and study the
spatial distribution of the CO excitation and gas-to-dust ratio
6. DISCUSSION of this forming cluster of massive galaxies at z = 4.05.
We have detected CO 2–1 emission from three galaxies in
the GN20 proto-cluster at z = 4.05 using the EVLA. The E.D. acknowledges the funding support of ERC-StG-
molecular gas masses range from 1.9 × 1010 × (α/0.8) M to UPGAL-240039 and ANR-08-JCJC-0008. D.R. acknowl-
1.3 × 1011 × (α/0.8) M . Hence, we are observing a group edges support from NASA through Hubble Fellowship grant
of molecular gas-rich galaxies undergoing extreme starbursts HST-HF-51235.01 awarded by the Space Telescope Science In-
within 1.6 Gyr of the big bang. stitute, which is operated by the Association of Universities for
GN20 presents a particularly interesting case, given its very Research in Astronomy, Inc., for NASA, under contract NAS
high luminosity and large spatial extent. We find that the sum 5-26555.
of the stellar and gas mass in GN20 is comparable to the
dynamical mass. While the assumptions involved in all three REFERENCES
mass calculations are uncertain, it appears that the baryons
likely dominate the mass content within 5 kpc radius of Aalto, S., Booth, R. S., Black, J. H., et al. 1995, A&A, 300, 369
Andreon, S., & Huertas-Company, M. 2011, A&A, 526, 11
GN20. Moreover, their is little room for a CO luminosity to Aravena, M., Bertoldi, F., Carilli, C., et al. 2010, ApJ, 708, L36
gas mass conversion factor substantially larger than 0.8 M Biggs, A., Younger, J., & Ivison, R. 2010, MNRAS, 408, 342
K km s−1 pc2 . High-resolution observations are required to Blain, A., Smail, I, Ivison, R., Kneib, J.-P., & Frayer, D. 2002, Phys. Rep., 369,
confirm the rotational aspect of the emission, and to obtain a 111
more accurate derivation of the dynamical mass. Capak, P., Riechers, D., Scoville, N. Z., et al. 2011, Nature, 470, 233
The morphology and velocity field for GN20 suggest a Carilli, C. L., Daddi, E., Riechers, D., et al. 2010, ApJ, 714, 1407
Carilli, C. L., & Holdaway, M. 1999, Radio Sci., 34, 817
rotating disk, with a possible tidal tail extending 10 kpc to the Casey, C. M., Chapman, S. C., Daddi, E., et al. 2009, MNRAS, 400, 670
north in a single 78 km s−1 channel. We can speculate that GN20 Chapman, S., Blain, A., Ibata, R., et al. 2009, ApJ, 691, 560
is a gas-rich disk galaxy that has been gravitationally torqued by Chapman, S., Blain, A., Ivison, R., & Smail, I. 2003, Nature, 422, 695

5
The Astrophysical Journal Letters, 739:L33 (6pp), 2011 September 20 Carilli et al.

Collins, D., Chris, A., Stott, John P., et al. 2009, Nature, 458, 603 Moresco, M., Pozzetti, L., Cimatti, A., et al. 2010, A&A, 524, 67
Cornwell, T., Braun, R., & Briggs, D. 1999, in ASP Conf. Ser. 180, Synthesis Morrison, G., Owen, F., Dickinson, M., Ivison, R., & Ibar, E. 2010, ApJS, 188,
Imaging in Radio Astronomy II, ed. G. B. Taylor, C. L. Carilli, & R. A. 178
Perley (San Francisco, CA: ASP), 151 Papadopoulos, P., & Pelupessy, F. 2010, ApJ, 717, 1037
Daddi, E., Dannerbauer, H., Krips, M., et al. 2009a, ApJ, 695, L176 Perley, R., Chandler, C., Butler, B., & Wrobel, J. 2011, ApJ, 739, L1
Daddi, E., Dannerbauer, H., Stern, D., et al. 2009b, ApJ, 694, 1517 Riechers, D. A., Hodge, J., Walter, F., Carilli, C. L., & Bertoldi, F. 2011, ApJ,
Daddi, E., Elbaz, D., Walter, F., et al. 2010, ApJ, 714, L118 739, L31
Dannerbauer, H., Walter, F., & Morrison, G. 2008, ApJ, 673, L127 Renzini, A. 2006, ARA&A, 44, 141
Doherty, M., Tanaka, M., De Breuck, C., et al. 2009, A&A, 509, 83 Riechers, D., Capak, P. L., Carilli, C. L., et al. 2010, ApJ, 720, L131
Downes, D., & Solomon, P. M. 1998, ApJ, 507, 615 Shim, H., Chary, R.-R., Dickison, M., et al. 2011, ApJ, submitted
Hibbard, J., & Mihos, C. 1995, AJ, 110, 140 (arXiv:1103.4124)
Iono, D., Peck, A. B., Pope, A., et al. 2006, ApJ, 640, L1 Solomon, P. M., & Vanden Bout, P. A. 2005, ARA&A, 43, 677
Ivison, R., Papadopoulos, P., Smail, I., et al. 2011, MNRAS, 412, 1913 Stevens, J., Ivison, R. J., Dunlop, J. S., et al. 2003, Nature, 425, 264
Ivison, R., Smail, I., Papadopoulos, P., et al. 2010, MNRAS, 404, 198 Tacconi, L., Genzel, R., Smail, I., et al. 2008, ApJ, 680, 246
Kurk, J. 2009, A&A, 504, 331 Tacconi, L., Neri, R., Chapman, S. C., et al. 2006, ApJ, 640, 228
Madden, S., Poglitsch, A., Geis, N., et al. 1997, ApJ, 483, 200 Wardlow, J., et al. 2011, MNRAS, in press
Momjian, E., Carilli, C., & Petric, A. 2005, AJ, 129, 1809 Williams, C. C., Giavalisco, M., Porciani, C., et al. 2011, ApJ, 733, 92
Momjian, E., Wang, W.-H., Knudsen, K. K., et al. 2010, AJ, 139, 1622 Younger, J., Fazio, G. G., Wilner, D. J., et al. 2008, ApJ, 688, 59