arXiv:1011.3556v1 [astro-ph.CO] 16 Nov 2010
PAH Emission Within Lyman Alpha Blobs
James W. Colbert1, Claudia Scarlata1, Harry Teplitz1, Paul Francis2, Povilas Palunas3,
Gerard M. Williger4, Bruce Woodgate5
We present Spitzer observations of Lyα Blobs (LAB) at z=2.38–3.09. The
mid-infrared ratios (4.5/8µm and 8/24µm) indicate that ∼60% of LAB infrared
counterparts are cool, consistent with their infrared output being dominated by
star formation and not active galactic nuclei (AGN). The rest have a substantial
hot dust component that one would expect from an AGN or an extreme starburst.
Comparing the mid-infrared to submillimeter fluxes (∼850µm or rest frame far
infrared) also indicates a large percentage (∼2/3) of the LAB counterparts have
total bolometric energy output dominated by star formation, although the num-
ber of sources with sub-mm detections or meaningful upper limits remains small
(∼10). We obtained Infrared Spectrograph (IRS) spectra of 6 infrared-bright
sources associated with LABs. Four of these sources have measurable polycyclic
aromatic hydrocarbon (PAH) emission features, indicative of significant star for-
mation, while the remaining two show a featureless continuum, indicative of
infrared energy output completely dominated by an AGN. Two of the counter-
parts with PAHs are mixed sources, with PAH line-to-continuum ratios and PAH
equivalent widths indicative of large energy contributions from both star forma-
tion and AGN. Most of the LAB infrared counterparts have large stellar masses,
around 1011M⊙. There is a weak trend of mass upper limit with the Lyα lu-
minosity of the host blob, particularly after the most likely AGN contaminants
are removed. The range in likely energy sources for the LABs found in this and
previous studies suggests that there is no single source of power that is producing
all the known LABs.
Subject headings: galaxies: evolution, galaxies: high-redshift, infrared: galaxies
1Spitzer Science Center, California Institute of Technology, Pasadena, CA 91125
2Research School of Astronomy and Astrophysics, The Australian National University, Canberra, ACT
3Las Campanas Observatory, La Serena, Chile
4Dept. of Physics & Astronomy, University of Louisville, Louisville, KY 40292
6NASA Goddard Space Flight Center, Greenbelt, MD 20771
– 2 –
The Lyα blob (LAB) remains one of the great mysteries of the high redshift universe.
While these extended Lyα nebulae are similar in extent (5-20′′or ∼50-150 kpc) and Lyα flux
(∼1043−44ergs s−1) to high-redshift radio galaxies, blobs are radio quiet and are therefore
unlikely to arise from interaction with jets. They are found almost exclusively within high-
redshift galaxy over-densities (Matsuda et al. 2009; Prescott et al. 2008; Palunas et al. 2004;
Steidel et al. 2000) with none found so far at even moderate redshift (z<0.8; Keel et al.
2009), suggesting strong evolution. After more than a decade of searching, there are still
only a handful of the truly giant (>50 arcsec2, >5×1043ergs s−1) LABs known. However,
deep searches (Matsuda et al. 2004; Saito et al. 2006) show that the blobs are part of a
continuous size distribution of resolved Lyα emitters.
With most lying at the density peak of high redshift structures (Matsuda et al. 2009,
2004; Palunas et al. 2004) and with number densities comparable to galaxy clusters in the
nearby and high-z universe (10−5–10−6Mpc−3; Yang et al. 2009), it seems likely that the
giant LABs are at the very least signposts for regions of massive galaxy assembly, if not
the progenitors of the massive elliptical galaxies themselves. The limited HST imaging of
these objects to date shows some evidence for interaction and merger of multiple compact
objects (Chapman et al. 2004; Francis et al. 2001). Major mergers of high-mass galaxies,
like those predicted to build giant ellipticals, must be occurring at higher redshift (i.e.,
Narayanan et al. 2010; Stewart et al. 2009). The LAB may represent an opportunity for the
study of the merger formation of the most massive galaxies.
One of the biggest unknowns of the blobs is the source of their energy. Matsuda et al.
(2004) found that for at least a third of LABs, including all the biggest and brightest of
them, the galaxies within the blob do not emit enough rest-wavelength UV light to excite
such a vast quantity of hydrogen gas. The exciting ultraviolet illumination could be escaping
along different lines of sight from an obscured AGN (Basu-Zych & Scharf 2004). Many LABs
contain luminous X-ray counterparts (Geach et al. 2009; Yang et al. 2009), while others show
powerful AGN emission lines (Scarlata et al. 2009; Dey et al. 2005; Pascarelle et al. 1996).
Alternatively, outflows could be driving great plumes of gas into the surrounding ambient
medium and producing shocks. While AGNs are known to drive some outflows, supernova-
driven superwinds are also a viable model (Taniguchi & Shioya 2000; Ohyama & Taniguchi
2004). Such outflows would likely take the form of immense bubbles or shells expanding
outwards from the galaxy, some evidence of which has been claimed in the brightest LABs
(Mori & Umemura 2007).
Integral field spectroscopy of the LABs show the large Lyα velocty widths and structure
consistent with superwind outflows (Bower et al. 2004; Wilman et al. 2005; Weijmans et al.
– 3 –
2010), although the systems are complicated enough that other velocity models, like rotation,
can not be completely ruled out. Cooling flows have also been suggested as a possible source
of the extended Lyα emission (Haiman et al. 2000; Francis et al. 2001; Dijkstra et al. 2006).
The primary evidence to support the cooling model are LABs with no apparent internal
power source, even out to the mid-infrared (Nilsson et al. 2006; Smith et al. 2008), and
sources with He II emission with weak to no measurable CIV emission (Prescott et al. 2009;
Scarlata et al. 2009), indicative of the lower temperature gas emission that one would expect
from a cooling flow (Bertone & Schaye 2010).
Mid-infrared and submillimeter imaging show that it is very common that the extended
nebulae of LABs contain sources of extreme infrared luminosity. Powerful Spitzer 24µm
sources have been found within roughly ten LABs, and almost all the most luminous ones,
(Webb et al. 2009; Geach et al. 2007; Colbert et al. 2006; Dey et al. 2005), with fluxes of
0.05-0.86 mJy. Submillimeter flux has been measured for a similar number (Chapman et al.
2004; Smail et al. 2003; Geach et al. 2005; Beelen et al. 2008). Mid-infrared colors suggest
significant quantities of hot dust for these infrared-bright sources (Webb et al. 2009), al-
though whether from AGN or extreme star formation remains unclear.
In this paper we discuss mid-infrared and submillimeter observations of mid-infrared
sources identified within LABs from four different fields, with z=2.38-3.09. We examine both
their Spitzer mid-infrared flux ratios and mid-infrared to sub-mm flux ratios and compare
them to models in order to identify their likely power source, AGN or star formation. We
then look at Spitzer IRS spectra of 6 sources for an even more definitive AGN/star formation
separation. Finally, we look at the possible masses of these infrared sources from within the
LABs. We assume an ΩM=0.3, ΩΛ=0.7 universe with Ho=70 km s−1Mpc−1.
2. Data Acquisition and Reduction
2.1. LAB Nomenclature
Because of their rarity, most LABs are discovered in small numbers, often one or two
at a time. This has led to a naming system where the LABs in question are often just
referred to as Blob 1 or Blob 2, if they are given any specific name at all. That creates a
problem for a work such as this one, where multiple LABs are being discussed from multiple
fields, all known as Blob 1 or LAB1 or B1 or something similar. Since LABs are large, they
occasionally have multiple counterparts associated with them, which can make it even more
difficult to identify the correct object under discussion.
To address this issue we will use the following naming system in this paper. All LABs
– 4 –
will be referred to as:
LAB[number][letter] J[coordinate of associated field center]
The [number] used is that which has been associated with the LAB in previous publi-
cations, if there has been any. If no previous number has been assigned we will begin the
labeling at LAB1 and count upwards from there. In many cases, the original numbering
included all the detected Lyα emitting sources in the field so just because there is a LAB6
does not indicate there must be a LAB5. While these ”missing” numbers might create some
mild confusion, we found this system to be superior to the certain confusion that would
result from changing the numbers that have already been used to identify these sources from
one paper to the next.
The [letter] refers to the counterpart of the LAB being discussed. Most blob counterparts
have only one associated counterpart and will therefore be given no letter identifier, so the
existence of a letter implies that at some point in the literature, multiple components have
been assigned to the LAB. Discussions of the LABs themselves will never be given a letter.
The [coordinate of the associated field center] is an eight digit code giving the right
ascension and declination (J2000) of the field with which the LAB is associated. Most of
these fields have been observed multiple times with slightly different centers, but for this
paper we have chosen J2217+0017, J1714+5014, J1434+3317, and J2143-4423 to represent
the SSA22, 53w002, NDWFS, and J2143-4423 fields in this study (see below). We found
this method of only providing a field coordinate superior to a full 12-14 digit coordinate
designation as it makes the discussion of object names significantly less cumbersome and
keeps a clear connection with the field and/or structure they have been found within.
Table 1 provides the list of all the LAB infrared counterparts included in this paper,
providing both the new name as well as any previous names by which the LABs have been
2.2. IRAC and MIPS data
The LABs presented in this paper are spread across four different fields: J2143-4423,
SSA22, 53w002, and the NOAO Deep Wide-Field Survey. We assembled all available Spitzer
mid-infrared imaging data, which includes both IRAC and MIPS imaging.
The data for the J2143-4423 field (LAB1 J2143-4423, LAB6 J2143-4423, LAB7 J2143-
4423) come from GO-3699 (PI: Colbert) and were done in a 3 × 5 raster map covering 15′×
25′, centered at α= 21h42m35s, δ = -44◦27′(J2000.0). The total integration times per pixel
– 5 –
were 1800 s for IRAC channels 1-4 and 1818 s for MIPS 24µm. This reached 3 σ depths of
1, 7, and 40 µJy for 4.5, 8, and 24 µm respectively.
For the SSA22 field (see Steidel et al. 2000; Matsuda et al. 2004) we examine the 16
LABs with both an isophotal area greater than 20 square arcseconds and submillimeter data
coverage, excluding only LAB4 J2217+0017 which lies too close to a bright object for an
uncontaminated analysis. We assembled the Spitzer imaging from multiple programs taken
at different epochs, centered roughly at α= 22h17m40s, δ = +00◦17′(J2000.0). The IRAC
and MIPS data come from four programs: GO 30600 (PI: Colbert), GO 3473 (PI: Blain),
and GTO 64 & GTO 30328 (both PI: Fazio). The IRAC data was previously presented
in Webb et al. (2009), but we re-extracted the photometry to ensure uniformity among our
several fields and also extracted to slightly deeper (3σ) depths. The MIPS data presented
here includes new data (GO 30600, see below) not presented in Webb et al. (2009), covering
more sources and greatly increasing the 24 µm depth.
The SSA22 IRAC data from GTO 64 was a deep pair of single pointings covering most of
the known SSA22 blobs, GTO 30328 was a 3 × 4 raster map, while GO 3473 was a smaller 2
× 3 raster map that lies to the east of the majority of the known LAB positions. Altogether
the combined IRAC image maps out a region roughly 20 × 26 arcminutes, ranging in depth
from 1500 to 15,000 seconds. Most of the LABs lie at depths of 4000 s or greater, roughly
corresponding to 3 σ depths of 0.5 and 5 µJy at 4.5 and 8µm respectively. Only one of the
sub-mm detected blobs (LAB10 J2217+0017) completely falls off the area covered by IRAC.
The SSA22 MIPS data from GTO 64 is a single pointing (1120 s), GTO 30328 is a 3
× 3 map built from a set of cluster offsets (1200 s per pixel), and GO 3473 was a set of
pointings targeting known sub-mm galaxies (1200 s). GO 30600 consisted of two parts. The
first was a set of three pointings targeting the sub-mm detected blobs on the outskirts of the
previously taken 24µm data (2700 s each). The second part was an extremely deep pointing
on the central portion of the field (10,800 s) where the biggest LABs (LAB1 J2217+0017
& LAB2 J2217+0017; Steidel et al. 2000) are located. Assembled altogether the MIPS data
covers an area roughly 15 × 26 arcminutes, with the regions covering the LAB locations
ranging from 1200 - 13,000 seconds, with all sub-mm detected LABs observed for at least
3300 seconds, which provides a 3 σ depth of ∼30 µJy.
The 53w002 field, centered at α = 17h14m20s, δ = +50◦15′(J2000.0), contains two LABs
(Keel et al. 1999). The IRAC data come from GTO 211 (PI: Fazio) and were obtained in a
single pointing of 3300 s in all four IRAC bands. The MIPS data are a combination of GO
3329 (PI: Stern), a single 500s pointing, and GO 20253 (PI: Im), a much deeper exposure
of 18720 seconds. The 3 σ depths obtained are 0.6, 5, and 12 µJy for 4.5, 8, and 24 µm
respectively, not accounting for the effect of confusion noise which becomes significant in
– 6 –
such a deep 24 µm pointing. Both blobs in the field have mid-infrared counterparts that are
bright enough that confusion is not an issue.
The NOAO Deep Wide-Field Survey (NDWFS) contains one reported bright blob
(Dey et al. 2005), located at α = 14h34m11.0s, δ = +33◦17′31′′(J2000.0). We do not attempt
to re-reduce or extract new photometry for this source and use the IRAC and MIPS fluxes
already reported in Dey et al. (2005).
Both the IRAC and MIPS data were combined using the MOPEX package available
from the Spitzer Science Center (SSC). For the IRAC data we started with the artifact-
mitigated (.cbcd) files where available (GTO 30328). For the programs that had not been
yet been reprocessed above pipeline S16 (the first .cbcd files begin at S17, released Jan. 08),
we used the basic bcd and applied the IRAC Artifact Mitigation code (available from the
SSC) ourselves. For our extragalactic fields, which contain very few bright stars or extended
sources, we saw no significant difference between the two methods. We then removed a
residual offset from each frame by subtracting the median of each frame itself (i.e., removing
a constant sky), medianing all the subtracted frames, and then subtracting that resulting
offset frame from all the original frames. This prevents a gradient from appearing in the
final images. This is the data product we place into MOPEX to produce the final, combined
For the MIPS data, a median flat was created from all available data taken near in
time, normalized, and divided into each image. After the Overlap Correction module was
executed by MOPEX we found that we still had some trouble getting all the various sky levels
to match up, so an additional sky constant was also subtracted before final combination of
all the MIPS frames.
2.3. Optical and Near-Infrared Data
The optical and NIR data for the J2143-4423 field (U, B, G, R, J and H) were acquired
at the Cerro Tololo Inter-American Observatory (CTIO) 4 meter telescope, using the Mosaic
II imager, and the near Infrared Side Port Imager camera (ISPI). The seeing ranged from
0.9–1.4′′, reaching depths of ∼26 for U and B, ∼24 in R, and ∼23 in J and H. For further
details see Scarlata et al. (2009). All magnitudes are AB magnitudes here and throughout
– 7 –
2.4. IRS Spectra
As part of the Spitzer GO 30600, we targeted all known bright (> 200 µJy) 24µm sources
associated with LABs, including the those within the J2143-4423 110 Mpc filament structure
at z=2.38 (Palunas et al. 2004; Colbert et al. 2006) and one within the possible proto-galaxy
structure around radio galaxy 53w002 at z=2.39 (Pascarelle et al. 1996). Altogether we
targeted five mid-infrared sources associated with four Lyα blobs. LAB6 J2143-4423 contains
two powerful MIPS 24µm sources (Colbert et al. 2006), both of which have been confirmed to
be at the redshift of the blob (Scarlata et al. 2009). At the time of observation there was one
more known mid-infrared bright (0.86 mJy at 24µm ) LAB1 J1434+3317 discovered in the
NOAO Deep Wide-Field Survey by Dey et al. (2005) for which IRS spectra had already been
taken. We include its IRS data as part of our analysis. The complete list with coordinates
is presented in Table 2.
There are no > 200 µJy 24µm-emitting LABs in the most well studied LAB field of
all, the z=3.09 SSA22 field (Steidel et al. 2000), despite many of them having significant
sub-mm detections. This is likely the result of the strong 7.7µm feature having shifted out
of the MIPS 24µm filter. For these sources we only examine their mid-IR to sub-mm ratios.
The mid-infrared spectra were all taken with the Spitzer Infrared Spectrograph (IRS)
using the 1st order of the Long-Low module (LL1), which is sensitive from 19.5 - 38.0 µm.
The LL1 module has a spatial resolution of 5.1/arcsec per pixel and a wavelength resolution
of R=58-116, with the ∆λ an approximately constant 0.17µm per pixel. The aperture of the
LL1 long slit is 10.7× 168 arcseconds.
The IRS data were acquired from the Spitzer cycle 3 GO 30600 program, mostly taken
June 13-21, 2007, although one spectrum (LAB18 J1714+5015) was taken September 16,
2006. These spectra were acquired using the IRS Spectral Mapping Astronomical Observa-
tion Template (AOT), placing the source at six separate positions along the length of the slit,
each separated by 20′′. A high accuracy peak-up observation on a nearby 2MASS star was
done for each observation. We used the 120 second ramp exposure throughout, producing
total exposure times per source of 84 – 528 minutes.
We began our data reduction with the Basic Calibrated Data (BCDs) produced by
the Spitzer pipeline S16.1. A significant latent charge builds up on the detector over long
observations that varies depending on wavelength. To correct this we measure a median
background value for each wavelength row and then fit a slope to the change in background
with time. Using the slope at each wavelength we derive the corresponding latent build-up
for each frame and subtracted it.
We then averaged together all the exposures taken at the same position within each
– 8 –
Astronomical Observation Request (AOR), producing six two-dimensional spectra per AOR.
We also produced a separate sky frame for each position using all the images taken at
the other five positions medianed together. We subtracted each sky frame from the its
corresponding image and then ran the program IRSCLEAN, provided by the Spitzer Science
Center (SSC), in order to remove all remaining rogue pixels or cosmic ray strikes. We then
extracted the one-dimensional spectra for all six positions using the optimal extraction option
within the SSC tool SPICE (v.2.1.2). We used a smaller than standard aperture (5 pixels at
27µm, width scaling with wavelength), in order to avoid flux from two nearby bright sources
and to reduce the inclusion of noise from blank sky. This non-standard aperture alters the
SPICE generated calibration, but we use the known 24µm fluxes for final calibration so this
was not an issue.
The initial sky subtraction does not always produce a perfect zero background and can
not account for contamination from brighter, nearby sources. To address this we extracted
one dimensional sky spectra using the same method and aperture as that for the sources,
only offset by 30 arcseconds. Two sky spectra, one from each side, are extracted, except for
the spectra closest to the edge of the slit, where only one is extracted. We median all the
sky spectra together and apply a 1.7µm wide boxcar smoothing, before subtracting this one
dimensional sky from the extracted source spectra. This secondary sky subtraction adds ∼2-
5% to the noise of the the extracted source spectra, but can make significant corrections to
the spectral shape, particularly in the two cases (LAB1 J2143-4423 and LAB7 J2143-4423)
with bright nearby sources.
Finally, we average the spectra from all the positions into a final spectrum using a 2
sigma clipping. In the case of LAB1 J2143-4423 the data was taken in two AORs with slightly
different roll angles. This minor difference in position angles produced a large difference in
the contamination from a nearby quasar, producing significantly more noise (60% higher).
Therefore we applied a variance weighting to the combination the LAB1 J2143-4423 spectra.
In all the other cases the noise was constant across spectra and no weighting was required.
As part of our analysis we also include one IRS spectrum (LAB1 J1434+3317) taken as
part as part of the GTO 15 program in February and June of 2005. This spectrum was taken
using the IRS Staring AOT, which takes exposures at two separate positions along the slit
length, separated by ∼55 arcseconds. Because of the smaller number of positions, instead of
subtracting a two-dimensional medianed sky from each position we simply subtracted one
nod from the other. We also note the BCDs were produced by the S15.3 pipeline. Otherwise
the data reduction is identical to those spectra taken in IRS Spectral Mapping mode.
– 9 –
The LABs are large and commonly cover multiple IRAC sources, so we used the positions
of the brightest IRAC 8µm sources as the likely source of any 24 µm or 850µm emission.
This approach produced a good match between the centers of the associated 8 and 24µm
sources in all cases where both are strongly detected. For sources with no 8µm detections
within the blob (5 sources from SSA22), the position of the 3.6µm source within the LAB is
used. None of the sources without 8µm detections had 24µm or 850µm detections, meaning
they are not included in most of the following analysis of flux density ratios.
Two sources had multiple strong 8µm sources, each with measurable 24µm fluxes, within
the LAB: the LAB6 J2143-4423 (Colbert et al. 2006; Scarlata et al. 2009) and LAB1 J2217+0017(Geach et
2007; Webb et al. 2009). For the case of LAB6 J2143-4423, an examination of the full
SED and IRS spectra (see Scarlata et al. 2009, and below) reveals that the counterpart
LAB6a J2143-4423 is powered by an AGN while LAB6b J2143-4423 is powered by star for-
mation, making it almost a certainty that the sub-mm flux identified at that position is
coming from LAB6b J2143-4423 and all further analysis assumes that. The two sources in
LAB1 J2217+0017, on the other hand, have very similar mid-IR colors, so it is not possi-
ble to discriminate from which 8µm source the submillimeter might have originated. Using
both OVRO and VLA 21 cm observations, Chapman et al. (2004) found emission near the
most northern of the two 8µm objects (LAB1a J2217+0017; Geach et al. 2007). However,
there is evidence that the 850µm flux for this source is likely distributed across multiple
sources(Matsuda et al. 2007), indicating that the second 8µm source (LAB1b J2217+0017)
may also be a significant contributor to the total submillimeter flux. For our analysis of
24/850µm we therefore combined the two sources together (see more below).
MIPS 24µm photometry was derived from PRF-fitting using APEX, part of the Spitzer
MOPEX software tool Makovoz & Marleau (2005). In cases of 24µm non-detections we
extracted limits from the data using the known 8µm positions. The rest of the photometry
was done using apertures. For the U, B, R, J, & H bands, we derived the flux densities
using apertures of radius ∼1.5×FWHM (1.72′′, 2.10′′, 1.95′′, 1.35′′, 1.35′′). For the IRAC
channels 3 and 4, we used a 3.6′′aperture, but for channels 1 and 2 we used a 2.4′′aperture to
avoid crowding issues. All the above apertures only required moderate aperture corrections
Substantial submillimeter imaging has been published on many of the major fields known
to contain LABs, including the z=3.09 SSA22 field (Geach et al. 2005; Chapman et al. 2004),
the z=2.38 J2142-4423 cluster (Beelen et al. 2008), and the z=2.39 53w002 field (Smail et al.
– 10 –
The SSA22 field was imaged at 850µm using the Submillimetre Common User Bolome-
ter Array (SCUBA) on the James Clerk Maxwell Telescope reaching 1σ noise limits of 1.5
mJy using the pointed photometry mode and 5.3 mJy from a more shallow scan-map. Five
blobs were detected with fluxes ranging from 4.9-16.8 mJy (Geach et al. 2005). All but the
brightest one were found in the deeper pointed observations. There are also Submillime-
ter Array (SMA; Matsuda et al. 2007) and Atacama Submillimeter Telescope Experiment
(ASTE; Kohno et al. 2008) observations of the brightest blob (LAB1 2217+0017) in the
The J2143-4423cluster was imaged at 870µm (345 GHz) using the LABoCA (Siringo et al.
2007, 2008) instrument installed on the Atacama Pathfinder Experiment (APEX, G¨ usten et al.
2006) reaching a 1σ point source sensitivity of 1.4–7 mJy, depending on distance from center
of the field. LAB7 J2143-4423 is robustly detected (8.4±1.0 mJy), while LAB6 J2143-4423
is only marginally detected (4.9 ±2.0 mJy).
The SCUBA 850µm observations of the 53w002 field reach 1-1.5 mJy noise levels and
detect one of the two known blobs there (LAB18 J1714+5015; 5.6±0.9 mJy).
For ease of discussion, the rest of the paper will use 850µm to describe either the 850µm
SCUBA data or the 870µm LABoCA data. It is assumed that at the level of accuracy of the
reported sub-mm fluxes there are no measurable differences between the two sub-mm bands.
We present the flux densities of all the LAB counterparts presented in this paper in
4.1. Spitzer Flux Density Ratios
The mid-infrared properties of warm ULIRGs, those dominated by AGN radiation, are
significantly different than those of cool ULIRGs powered by star formation. The SEDs
of warm ULIRGs rise rapidly and steadily through the mid-infrared, roughly obeying a
power law, while the cool ULIRG SEDs do not rise steeply until far into the mid-infrared
(> 10µm). The other significant difference is the presence of strong polycyclic aromatic
hydrocarbon (PAH) features in cool ULIRGs, which can be compared to the continuum to
determine the fraction of infrared luminosity generated by star formation (Lutz et al. 1998;
Rigopoulou et al. 1999).
Because of these mid-infrared color differences, plots of Spitzer mid-infrared IRAC colors
can be powerful tools for identifying AGN and separting them from starbursts at low redshifts
– 11 –
(Stern et al. 2007, 2005; Lacy et al. 2004). By including MIPS 24µm one can then continue
this analysis to much higher redshift (i.e., Webb et al. 2009; Pope et al. 2008). We plot all
the mid-infrared sources associated with LABs that have 8µm detections in Figure 1, which
is a comparison of the 24/8µm flux density ratio to the 8/4.5µm flux density ratio.
Both Webb et al. (2009) and Pope et al. (2008) created just such a mid-infrared color-
color plot for their respective samples of LABs and submillimeter galaxies. The rectangular
box represents the region Pope et al. (2008) identified as the location of sources starburst-
dominated in the mid-infrared. The higher 8/4.5µm flux ratios are generally only obtainable
by AGN, as only their SEDs should be that steep at these high redshifts (z>2). Submillimeter
sources with similar redshifts to the LABs in this study, taken from Pope et al. (2006), have
very similar 24/8 and 8/4.5µm flux ratios, with mean colors that are the same, to within
one standard deviation, as that of the LAB counterparts presented in Figure 1.
For comparison we over-plot a series of models derived from Chary & Elbaz (2001,
hereafter CE01). As CE01 is based on low redshift infrared galaxies, we applied an evolution
in specific star formation to account for the expected build-up of stellar mass with time, a
factor of (1+z)2. This redshift evolution roughly matches the evolution in integrated specific
star formation found for the massive galaxies studied by Papovich et al. (2006), erring on the
side of more conservative (i.e. less) evolution. Stronger evolution in specific star formation
would produce larger 24µm to 8µm ratios, moving the star formation ends of all the models
upwards on the plot.
The CE01 models cover a range in luminosity, so we selected three that are representative
of that entire range. Changing exactly which models are plotted makes no difference to the
results or discussion we present here. Finally we mixed the star forming CE01 SED models
with the SED of an AGN (Mrk 231) to produce a range of ratios from AGN to star formation-
dominated. On the plot the models start to the left and/or high above as pure star formation
and converge on pure Mrk 231 around 8/4.5µm ∼2. The z=2.4 models (solid lines) cover
a much larger range in 24/8µmratio than the z=3.1 models (dotted lines) because of the
possible presence of the large 7.7µmPAH feature at that redshift.
Nearly two thirds (59%; 10 of 17) of the LAB-associated sources fall within the star-
formation rectangle, 29% have mid-infrared colors clearly indicative of AGN, while the two
remaining sources (12%) are borderline, but likely also contain significant AGN contribu-
tion. On the surface this would indicate that the majority of sources associated with LABs
are powered by starbursts. However, even an AGN dominant in the infrared can still have
significant contribution from stars at rest wavelength 1µm which has the effect of flatten-
ing its slope considerably, i.e. it decreases the 8/4.5µm ratio. For instance, the Mrk 231
ratios presented here (the converging end point of the models we plot in Figure 1) differ
– 12 –
somewhat from the Mrk 231 model ratios presented in Pope et al. (2008), particularly at
z∼3. This is likely a result of their use of a Mrk 231 model based only on mid-infrared
data (a Rigopoulou et al. (1999) mid-infrared spectrum combined with a fit to Infared As-
tronomical Satellite photometry). The Mrk 231 model we use comes from a multi-component
fit (Armus et al. 2007) that includes the critical stellar component fit to near-infared data
points, which has a strong effect on the resulting 8/4.5µm ratio at z>2.
In addition, at these redshifts these ratios are extremely sensitive to hot dust, which
one would expect to be dominated by any AGN, even if the AGN was not a significant
contributor to the total infrared luminosity. Alternatively, an extreme starburst can also
produce a large amount of emission from hot dust, which could produce similar IRAC colors
to an AGN (Yun et al. 2008). This diagnostic plot makes no direct measurement of the far
infrared luminosity, the source of which we are trying to determine.
4.2. Mid-IR to Sub-mm Ratios
At the redshifts of LABs with sub-mm imaging (z=2.4-3.1), 850µm is sensitive to rest-
wavelength 200-250µm, past the peak of far infrared cold dust emission, where flux density
is beginning to fall like a Rayleigh-Jeans law (Fν∝ ν2). Spitzer MIPS 24µm observations
are sensitive to the rest wavelength 6-7µm mid-infrared, blueward of the far-infrared peak
and very sensitive to both hot dust and the presence of PAHs (although less so for PAHs at
the higher redshift, see discussion below). Spitzer IRAC observations at these redshifts are
not really sensitive to dust emission but instead can be dominated by stellar emission. In
particular, 8µm is roughly equivalent to rest wavelength K-band, an excellent tracer of total
The combination of Spitzer mid-infrared with submillimetre flux densities, each being
sensitive to a different temperature regime (stars, hot dust, cold dust), can do a great deal to
reveal the likely power sources of the infrared-bright galaxies within the blobs. In particular,
a simple color-color plot (ratio of MIPS 24µm to 850µm sub-mm versus ratio of MIPS 24µm
to IRAC 8µm) can be a powerful tool for discriminating AGN from star formation, as well
as getting a picture of the specific star formation (SFR per unit mass).
In Figure 2 we plot the ratio of 24µm to 850µm sub-mm versus ratio of 24µm to 8µm
for the three fields listed above that have both blobs and deep sub-mm coverage. Any source
without a 24µm detection is not plotted. In the case of LAB6 J2143-4423 there are two
MIPS 24µm sources with confirmed z=2.38 redshifts (Scarlata et al. 2009)), so both sources
(LAB6a and LAB6b) are included on the plot, but the submm detection has been assigned
– 13 –
to LAB6b J2143-4423 (as described above). Similarly, LAB1 J2217+0017 is associated with
two equally bright 8µm sources (LAB1a and LAB1b). Due to the large sub-mm beamsize it
is not possible to deduce which source is likely associated with the 850µm flux and, of course,
both may be. We have therefore combined the 8µm and 24µm fluxes of LAB1a J2217+0017
and LAB1b J2217+0017 into a single source for the purpose of this plot. If all the sub-mm
flux were to originate in the slightly fainter 24µm source (LAB1a J2217+0017) the 24/850
µm ratio would go down by an additional factor of 3, with very little change in 24/8µm.
We overplot the same CE01/Mrk231 mixed models as used in Figure 1. We marked the
ratio of star formation luminosity to AGN luminosity for one of the models, running from
1.0 (100% starburst) upwards to 0.0 (100% AGN) contribution from star formation to the
total bolometric luminosity.
All eight of the sub-mm detected components for LABs have mid-IR/sub-mm ratios
consistent with star formation. An additional six of the Lyα blob components are only up-
per limits, but only two of those (LAB6a J2143-4423 and LAB19 J1714+5015) are clearly
outside and above the locus of star formation, consistent with IRS spectrum of LAB6a J2143-
4423 (see below, Scarlata et al. 2009) and the broad line nature of LAB19 J1714+5015
(Pascarelle et al. 1996; Colbert et al. 2006). This submillimetre evidence for significant quan-
tities of cool dust in LABs was previously noted by Webb et al. (2009), who found that all
but one of the SSA22 8µm sources they studied were detected at 850µm as well. That strong
trend does not hold true for our sample, where 6 of 14 of our LAB counterparts are detected
at 8µm but not at 850µm but that could be at least partially attributable to the slightly
different depths and redshifts in our study.
This plot strongly indicates that the majority of these infrared-bright components of
LABs are not powered by AGN alone, but there are two important caveats. First, the
present sensitivity limits of submillimeter surveys are just barely deep enough to conduct
this experiment. Not only are nearly half the sources not detected at all, but none of the
actual detections are that strong, ranging from 2.5-8 σ, with half of 4σ significance or less.
Fortunately the expected range of 24/850µm ratio is large, especially at z=3.1, allowing for
a strong discrimination even with very large sub-mm errors.
Secondly, while the correction for mass evolution is likely generally correct, it is almost
certain that there will be real variations in specific star formation history from galaxy to
galaxy. On the plot this will move galaxies back and forth horizontally, which will change
what models and AGN/star formation percentage one would associate with it. This is not
an important effect at z=3.1, where the models are less degenerative and there is large
separation in 24/850µm ratio as the AGN contribution is increased, but it can be a concern
at z=2.4. The mid-IR to sub-mm ratios at the z=2.4 are not as strong for discriminating
– 14 –
the energy source because of the presence of the powerful PAHs in the 24µm filter, which
can somewhat mimic the rising power law slope of an AGN.
For instance, LAB18 J1714+5015 (labeled in Figure 2) appears to lie roughly in the
neighborhood of the 93-98% star formation models. However, a change in the specific star
formation of the nearby models by 30% would shift it towards the AGN models such that it
would only take a factor of ∼2-3 change in the 24/850µm ratio (either the model or through
a measurement error) to make this source AGN-dominated. In fact, its IRS spectrum (see
below) suggests that it is completely AGN-dominated. To similarly shift z=3.1 sources would
require changes of factors of 10-20 in the 24/850µm ratio.
For further comparison we also plot sub-mm galaxies spectroscopically confirmed at
similar (±5% in 1+z) redshifts from Pope et al. (2006). The mid-infrared colors and mid-
infrared to sub-mm ratios for the Pope et al. (2006) sub-mm galaxies are very similar to
those from the LAB sample, with the exception of one z∼2.4 sub-mm source (GN22) with
a 24/850µm color so cool it can not be easily explained by our simple set of models. Three
of these z∼2.4 sub-mm galaxies also have IRS mid-infrared spectroscopy (GN04, GN05,
and GN19; Pope et al. 2008). All three have significant PAH features, despite the fact
that two of the three are detected in the hard X-rays. Looking at the line-to-continuum
ratios, Pope et al. (2008) found that only GN04 (the source with the largest 24/850µm ratio
included from their study) had >50% contribution to the mid-infrared energy output and
they classified it as a mixed (AGN + starburst) source.
We note that one blob in our sample falls on the low end of the models. This is the
z=3.09 Steidel et al. (2000) LAB1 J2217+0017, which has a large reported sub-mm flux
(16.8±2.9; Chapman et al. 2004), but is quite faint at 24µm with a flux of only 70±10 µJy.
This is actually the combined 24µm flux of the two bright 8µm sources identified at this
location (Webb et al. 2009; Geach et al. 2007). If split back into their two components the
24µm fluxes are 25 and 47 µJy, so if the submillimeter flux came from just one of them the
24/850µm ratio would be even more extreme. LAB1 J2217+0017 was undetected in high
spatial resolution observations (∼2′′) taken with the Submillimeter Array (Matsuda et al.
2007), indicating that the 850µm flux must be widely spread out or split into multiple sources.
Alternatively, the original measurement might be a several sigma deviation. ASTE-AzTEC
1.1 mm observations failed to detect it down to 10 mJy (Kohno et al. 2008). However, even
a factor of 10 less flux would still place LAB1 J2217+0017 in the region dominated by star
– 15 – Download full-text
4.3. PAH Features in LAB ULIRGs
Of the six sources associated with LABs targeted with IRS spectroscopy, four show
significant PAH features. In order to measure individual PAH emission lines, we used the
IDL PAHFIT software package (Smith et al. 2007), which fits all PAH emission lines and
continuum simultaneously. Given the limited coverage of the data, we only include the
PAH lines in the fit that cover the wavelength range of interest (6-9µm) and a single red
continuum. We chose this over the more direct method of isolating the region directly around
each line of interest and fitting a line and continuum there (i.e., Pope et al. 2008), as that
can underestimate the equivalent widths by factors of up to 4. Fits to all six spectra are
shown in Figure 3 and all fluxes, equivalent widths, and line-to-continuum ratios are listed
in Table 3.
Two of the spectra show no apparent PAH features, with all line fits producing fluxes
below the 1-σ uncertainty. These featureless sources are the counterpart LAB6a J2143-4423
(previously reported in Scarlata et al. 2009) and LAB18 J1714+5015 (Keel et al. 1999).
The remaining four sources all show the 7-9µm PAH complex (∼7.4-8.7µm) at the ex-
pected location for their redshifts. The weakest PAH detection is that of source LAB1 J2143-
4423, with a 7.7µm feature (defined as the combination of the 7.60 and 7.85µm PAH lines)
detection of only 2σ significance. However, the whole 7-9µm complex for the LAB1 J2143-
4423 source is detected at 4σ significance. Combined with their discovery at the appropriate
wavelengths there is little doubt the PAH features for LAB1 J2143-4423 are real.
We measured a 7.7µm line-to-continuum (L/C) ratio for each source in order to test the
contribution of any possible AGN to the total infrared energy output. For ease of comparison
we use the same method of Rigopoulou et al. (1999), taking the average intensity of the line
and continuum over 7.57-7.94µm. The error in line intensity is just a quadratic sum of
uncertainty in the data, but the error in continuum is dominated by the quality of the
continuum fit, which has only two parameters, temperature and a normalization factor, for
each of which the fitting routine returns an uncertainty. Using a Monte Carlo method we
take the two derived probability distributions of temperature and normalization factors to
create a large distribution of average continuum intensities from which we measure a final
rms. This continuum error is consistently large, often dominating the error in L/C.
Using their L/C definition, Rigopoulou et al. (1999) found that AGN have L/C≃ 0.2,
pure starbursts have L/C≃ 3, and the typical local ULIRGs have L/C≃ 2, indicating they
are almost completely star formation dominated. The two sources from our suvey with
no detected PAH features have L/C ratios < 0.2-0.3 (2σ limits), clearly AGN-dominated
sources. The two with the most powerful PAHs have L/C=5-6. Even accounting for large