Detections of CO Molecular Gas in 24um-Bright ULIRGs at z~2 in the Spitzer First Look Survey
Lin Yan, L.J. Tacconi, N. Fiolet, A. Sajina, A. Omont, D. Lutz, M. Zamojski, R. Neri, P. Cox, K.M. Dasyra
TTo be submitted to the Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 11/10/09
DETECTIONS OF CO MOLECULAR GAS IN 24 MICRON-BRIGHT ULIRGS AT Z ∼ Spitzer
FIRSTLOOK SURVEY
Lin Yan, L. J. Tacconi, , N. Fiolet, A. Sajina, A. Omont, D. Lutz, M. Zamojski, R. Neri, P. Cox, K. M.Dasyra, Spitzer
Science Center, California Institute of Technology, MS 220-6, Pasadena, CA 91125, USA Max-Planck Institut f¨ur Extraterrestrische Physik (MPE),Giessenbachstrasse, D-85741 Garching, Germany UPMC University of Paris 06, UMR7095, Institut d’Astrophysique de Paris, F-75014, Paris, France Haverford College, Haverford, PA 19041, USA Institut de Radio Astronomie Millimetrique (IRAM), St. Martin dHeres, France and NASA Herschel Science Center, California Institute of Technology, MS 220-6, Pasadena, CA 91125, USA
To be submitted to the Astrophysical Journal
ABSTRACTWe present CO observations of nine ULIRGs at z ∼ f ν (24 µm ) ∼ > Spitzer
First Look Survey. All targets are required to have accurateredshifts from Keck/GEMINI near-IR spectra. Using the Plateau de Bure millimeter-wave Interfer-ometer (PdBI) at IRAM, we detect CO J(3-2) [7 objects] or J(2-1) [1 object] line emission from eightsources with integrated intensities I c ∼ (5 – 9) σ . The CO detected sources have a variety of mid-IRspectra, including strong PAH, deep silicate absorption and power-law continuum, implying that thesemolecular gas rich objects at z ∼ (cid:48) CO(3 − is (1.28 – 3.77) × K km/s pc . The averaged molecular gas mass M H is1 . × M (cid:12) , assuming CO-to-H conversion factor of 0.8 M (cid:12) /[K km/s pc ]. Three sources (33%) –MIPS506, MIPS16144 & MIPS8342 – have double peak velocity profiles. The CO double peaks inMIPS506 and MIPS16144 show spatial separations of 45 kpc and 10.9 kpc, allowing the estimates ofthe dynamical masses of 3.2 × sin − (i) M (cid:12) and 5.4 × sin − (i) M (cid:12) respectively. The impliedgas fraction, M gas / M dyn , is 3% and 4%, assuming an average inclination angle. Finally, the analysisof the HST/NICMOS images, mid-IR spectra and IR SED revealed that most of our sources aremergers, containing dust obscured AGNs dominating the luminosities at (3 – 6) µ m. Together, theseresults provide some evidence suggesting SMGs, bright 24 µ m ULIRGs and QSOs could representthree different stages of a single evolutionary sequence, however, a complete physical model wouldrequire much more data, especially high spatial resolution spectroscopy. Subject headings: galaxies: infrared luminous – galaxies: starburst – galaxies: high-redshifts – galaxies:evolution INTRODUCTION
Molecular gas holds one of the keys to fundamentalquestions about the formation and evolution of galax-ies. Particularly, at high redshift, significant galaxygrowth in its stellar population and black hole occursduring dusty, gas rich phases. The importance of infrared(IR) luminous, gas rich populations has been high-lightedby observations from the
Infrared Space Observatory ( ISO ), ground-based (sub)millimeter cameras, and morerecently, the
Spitzer Space Telescope (Genzel & Cesarsky2000; Blain et al. 2002; Lagache et al. 2005; Soifer et al.2008). Particularly,
Spitzer dramatically improved ourability to probe high redshift ( z ∼ > L IR ∼ > L (cid:12) , increasing the number of detectionsof such sources by orders of magnitude in comparisonwith ∼ (200 – 300) z ∼ Spitzer
InfraRed Spec-trograph (IRS; Houck et al. 2004) have confirmed withmid-IR spectra a population of 24 µ m-selected, Ultra-luminous infrared galaxies (ULIRGs) at z ∼ [email protected] the samples are selected to have 24 µ m fluxes brighterthan ∼ z ∼ L IR ∼ . − . L (cid:12) (Sajinaet al. 2007, 2008). Other gas rich galaxies with compa-rable or more L IR at z ∼ > Spitzer bright 24 µ m ULIRGs are selected in themid-IR, probing smaller and hotter dust grains.SMGs, Spitzer bright 24 µ m ULIRGs and gas richQSOs are all high- z systems with extremely high bolo-metric luminosities, leading to the important questionof how these systems are related or different. Studiesof the gas content, particularly CO gas, of SMGs havemade significant progress in recent years (Neri et al. 2003;Greve et al. 2005; Tacconi et al. 2006, 2008). The gen-eral consensus for this population is that SMGs containa reservoir of 10 − M (cid:12) of molecular gas, distributedover a small area of R / ∼ < ∼ M (cid:12) /yr, typically over several dynamical timescales of ∼ yrs. Recent several studies have also ex-amined the mid-IR spectral properties of SMGs (Lutz etal. 2005; Valiante et al. 2007; Pope et al. 2008; Men´endez- a r X i v : . [ a s t r o - ph . C O ] M a r Lin Yan et al.
TABLE 1Sample Targets
ID RA DEC z(Opt) a z(CO)deg degMIPS429 259.04930 59.20366 2.213 2.2010MIPS506 257.91080 58.64405 2.470 2.4704MIPS8196 258.79282 60.16533 2.586 .... b MIPS8327 258.89908 60.47375 2.441 2.4421MIPS8342 258.54813 60.18591 1.562 1.5619MIPS15949 260.28842 60.25035 2.122 2.1194MIPS16059 261.11850 60.25922 2.326 2.3256MIPS16080 259.68655 60.02107 2.007 2.0063MIPS16144 261.09207 59.53077 2.131 2.1280 a The optical spectroscopic redshifts are taken fromSajina et al. (2008). b MIPS8196 has no significant CO line detection, thus,its CO redshift is not determined.
Delmestre et al. 2007, 2009). The one clear difference isthat the observed 24 µ m fluxes of SMG are on averageseveral 100 µ Jy, or a factor of (2 – 3) smaller than their
Spitzer -selected counterparts. This is consistent withthe conclusion that bright 24 µ m z ∼ µ m than that of SMGs (Sajinaet al. 2008; Polletta et al. 2008). In contrast, for themajority of SMGs, starburst dominates the IR lumi-nosity, while AGN contributes only a small fraction ofL IR (Valiante et al. 2007; Pope et al. 2008; Alexander etal. 2008; Men´endez-Delmestre et al. 2009). In addition,abundant CO molecular gas has been detected amongoptically selected, high- z QSOs and radio galaxies, espe-cially gravitationally lensed systems. Solomon & VandenBout (2005) reviewed and compiled a list of 23 QSOs andradio galaxies which have at least low resolution CO lineobservations (also Greve et al. (2005) for a similar list).In this paper, we will present the measurement of coldmolecular gas masses among bright 24 µ m ULIRGs at z ∼
2, determine their gas dynamics, and understand howthey differ from SMGs and gas rich QSOs & radio galax-ies. To directly address these questions, we obtained COinterferometric observations for a sample of bright 24 µ mULIRGs at z ∼ § § µ m ULIRGs differsfrom SMGs and high- z QSOs, we make comparisons tothe QSOs and radio galaxy sample compiled by Solomon& Vanden Bout (2005) and also to a small subset of QSOswhich have reliable dynamic masses from high resolutionCO observations and robust black hole mass estimatesfrom UV spectroscopy. § M = 0.27, Ω Λ = 0.73 and H = 71 km s − Mpc − cos-mology (Spergel et al. 2003). OBSERVATIONAL DATA
The ground transition temperature for CO J(1-0) isonly 5.5 K, in contrast for H
500 K. Therefore, CO ro- tational emission lines are easily produced by collisionalexcitation. It has been shown that CO flux I CO linearlycorrelates with the column density of molecular hydro-gen H , providing a sensitive tracer of bulk of the coldmolecular gas in the Universe (Young & Scoville 1982;Dickman et al. 1986; Solomon et al. 1987). Besides themolecular gas content of a galaxy, CO interferometric ob-servations can also probe the spatial distribution as wellas velocity field of molecular gas. Therefore, we obtainednew CO emission line observations for nine bright 24 µ mULIRGs at z ∼ The sample and ancillary data
The nine targets observed by the PdBI were selectedfrom a large sample of z ∼ Spitzer mid-IR spectra published in Yan et al. (2007). This par-ent sample consists of 52 sources initially selected with24 µ m flux density brighter than 0.9 mJy, and very red24-to-8 µ m and 24-to-R colors from the 4 square degree Spitzer
Extragalactic First Look Survey (XFLS) . Thesubsequently obtained Spitzer mid-IR spectra coveringthe rest-frame 4 – 20 µ m determined that 74% of the sam-ple is at 1.5 < z < − µ m ) are in the range of 10 − L (cid:12) .The complete mid-IR spectral and SED analysis showsthat at least ∼
75% of the sample contain dust obscuredAGNs, which dominate the mid-IR (3 – 6) µ m luminosi-ties but star formation still contribute most of the far-IR luminosities (Sajina et al. 2007, 2008). Of the 52sources, 44 galaxies were observed at 1.2 mm with theIRAM 30 meter telescope using the 117 element versionof the MAMBO array (Kreysa et al. 1998). Of these,7 are detected at ∼ > σ (14 at ∼ > σ ), with an averagerms ∼ . z ∼ Spitzer
ULIRGs, we obtained CO interferometric obser-vations for sources with a broad range of mid-IR spectralproperties, including sources with strong PAH emission,deep silicate absorption, and mid-IR power-law continua.The primary selection of our CO targets was the avail-ability of accurate spectroscopic redshifts from Keck andGEMINI. This criterion has limited our targets to a smallsubset of the full sample. All of our targets have 1.2 mmobservations from MAMBO. Of the nine sources, onlyone have 1.2 mm SNR > σ , seven with SNR ∼ < > σ source, westacked the remaining 8 sources, yielding an averaged1.2 mm flux of 1.06 ± ± Spitzer sample (52objects) (Sajina et al. 2008). This implies that our COtargets systematically have more far-IR emission, in com-parison with the full
Spitzer mid-IR spectroscopic sampleat z ∼ R (24 , ≡ log ( νf ν (24 µm ) /νf ν (8 µm ) ∼ > R (24 , . ≡ log ( νf ν (24 µm ) /νf ν (0 . µm ) ∼ > etections of CO Molecular Gas in Spitzer z ∼ TABLE 2CO observations
ID Lines ν obs Config. Beam a PA Time b Noise/chan. Chan.width ∆ c RMS d GHz hours mJy/chan km/s km/s mJy/beamMIPS429 (3-2) 107.642 C 4 . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦
10 0.7 72 360 0.35MIPS16144 (3-2) 110.407 C 2 . (cid:48)(cid:48) × . (cid:48)(cid:48) ◦ a Beam is defined as the semi-major axis times the semi-minor axis in arcseconds. b This is the total on-target integration time, calculated equivalent to 6 antenna. c The summed CO maps are integrated over a velocity width ∆ with the center on the peak of the CO emission line. d This noise is RMS for the summed CO map made with the listed parameters.
Table 1 and Table 2 list the coordinates, redshifts (fromboth CO and near-IR spectra) and the CO observationalparameters for the 9 sources observed by PdBI. Table 3shows the ancillary observations from optical, near-IR,far-IR and 20 cm photometry (see Sajina et al. 2008, fordetails). The near-IR spectra were obtained using NIR-SPEC at the Keck and NIRI at the GEMINI. We havehigh resolution H-band images from
HST
NICMOS. Thisdataset has been published in Dasyra et al. (2008) and isincluded in the analysis of a larger sample of z ∼ µ m ULIRGs in (Zamojski et al. 2009). CO emission line observations
The observations were carried out in 2007-2008 withthe PdBI in D (2 objects) and C (7 objects) configu-ration in order to maximize the sensitivity. The 3-mmreceivers were tuned to the central observed frequencyaccording to the Near-IR spectroscopic redshifts. Of thetotal nine sources, we used the PdBI to observe the COJ(3-2) transition ( ν rest = 345.796 GHz) for eight objectsand the J(2-1) transition ( ν rest = 230.538 GHz) for oneobject at redshift of 1 . .Care is given to the phase, flux and amplitude calibra-tions by rejecting any calibration anomalies due to me-teorological conditions or electronics. After calibration,CLIC generates the visibility data, i.e. “uv” tables. Sec-ond, based on these uv tables, we extract CO maps andspectra using the IRAM MAPPING software . The syn-thesized, clean beam size is typically elongated, roughly(2.4 (cid:48)(cid:48) × (cid:48)(cid:48) ) to (5.6 (cid:48)(cid:48) × (cid:48)(cid:48) ) for the C configuration and(5.02 (cid:48)(cid:48) × (cid:48)(cid:48) ) to (6.2 (cid:48)(cid:48) × (cid:48)(cid:48) ) for the D configuration. The corresponding spatial resolution in linear size rangesfrom 20.5 to 53 kpc. For comparison, for the
Spitzer µ m band, the Full Width Half Maximum (FWHM)of a point source is 6 (cid:48)(cid:48) and the NICMOS H-band PSF is0 . (cid:48)(cid:48) .Figure 1 shows the integrated CO maps, with the smallinsert at the left bottom of each panel indicating thebeam size and positional angle. Before producing thesummed CO maps, we rebinned the data with an orig-inal 10 MHz channel width by a factor of (3 – 4), yield-ing smoothed spectral data cubes with a channel widthroughly (54 – 83) km/s. The CO maps shown in Fig-ure 1 were made by summing over a number of veloc-ity channels, ranging from 5 – 14 (equivalently, velocitywidth ∆ ∼
360 – 810 km/s in Table 2), centered on thepeak CO line emission. The final maps have 1 σ noisevalues ranging from 0.1 to 0.35 mJy/beam. Table 2 liststhe noise parameters and the parameters used for mak-ing the summed CO maps. In Figure 1, the first COcontour starts roughly at 1 σ with a step size of 1 σ for allpanels. The summed CO line emissions are detected at(4.7 – 9.7) σ for 8 sources. Table 4 lists the integrated COline fluxes, velocity width over which the CO line inte-gration is done, and derived CO luminosity and H (+He)molecular masses.The CO line luminosity can be derived from the lineintensity, following Solomon et al. (1997): L CO [ L (cid:12) ] = 1 . × − (cid:16) I CO Jy km s − (cid:17)(cid:16) ν obs GHz (cid:17)(cid:16) D L Mpc (cid:17) (1)and L (cid:48) CO [K km / s pc ] = 3 . × I CO ν obs − D L2 (1 + z) − , (2)Here L (cid:48) CO is the CO line luminosity in units ofK km/s pc , I CO is the integrated line intensity inJy km/s, D L is the luminosity distance in units of Mpc, ν obs is the observed frequency in GHz. The H (+He)molecular gas mass can be estimated using the followingequation: Lin Yan et al. MIPS429 MIPS16144 MIPS8342 MIPS8327 MIPS16059 MIPS506 MIPS15949 MIPS16080
18” 20” Jy km/s
Fig. 1.—
The integrated CO maps for the 8 detected sources. The small insert in each panel shows the size and shape of the beam. Thecross marks the center of 24 µm source. The first contour starts at 1.17 σ , 1.15 σ , 1.05 σ , 1.05 σ , 1.04 σ , 1.18 σ , 1 σ , and 1.08 σ respectivelyfor MIPS429, MIPS506, MIPS8327, MIPS8342, MIPS15949, MIPS16059, MIPS16080 and MIPS16144. The contour step size is 1 σ for allmaps. The rms σ (mJy/beam), the channel width and the number of channels over which the CO maps are integrated, are given in Table 2. M gas M (cid:12) = α (cid:16) T CO (3 − /T CO (1 − (cid:17) L (cid:48) CO (3)Here α is the CO J(1-0) luminosity to molecular gasmass conversion factor in units of M (cid:12) [K km / s pc ] − and T CO is the effective brightness temperature of a CO tran-sition. Here we assume T CO (3 − /T CO (1 − = 1 and α = 0 . (cid:12) (K km / s pc ) − (Solomon et al. 1997; Downes& Solomon 1998), the same assumption used for SMGsand QSOs(Greve et al. 2005; Tacconi et al. 2006, 2008). Although a more realistic T CO (3 − /T CO (1 − ratio isprobably between (0.5 – 0.7) (Wilson et al. 2008), weadopt our assumption in order to make direct compari-son without additional corrections to the published SMGand QSO data. Modeling the IR SED
To derive the total and far-IR luminosities(L IR = L − µ m , L FIR = L − µ m ), we fit the spectralenergy distribution of our sources from near-IR to far-IR(see Table 3). The longest wavelength data available isetections of CO Molecular Gas in Spitzer z ∼ σ or less. For these 8 sources,the sigma weighted, averaged flux is 1.06 ± σ ), providing constraining power in the SED fittingin Figure 2 (cyan, open stars). Our SED fitting usesthree components: pure starburst template (the aver-aged SMG SED from Pope et al. (2006, 2008), greendot-dashed line), reddened type-I quasar SED (Richardset al. (2006), blue dashed line), and a 2 Gyr old SingleStellar Population (SSP) model from Maraston (2005) toaccount for the near-IR emission (purple dot-dot-dashedline). The solid red line is the sum of these threecomponents. The type I quasar SED (blue) has to bereddened substantially ( A V ∼ >
5) assuming a MW-typeextinction (Draine 2003). For most sources this leadsto a reasonably good fit to the mid-IR continuumand silicate absorption feature. However, two of oursources (MIPS8342 and MIPS15949) show red mid-IRcontinua but no significant silicate absorption. Theabove approach cannot fit this type of spectra. For thesesources, we revert to our power-law continuum approachas used in Sajina et al. (2008).At the far-IR, four sources are detected in the 70 µ mband, and only one at 160 µ m. With so few far-IRdetections, is the starburst-component really required?The only sources for which the answer is obviouslyyes are MIPS8342 (based on its 160 µ m detection) andMIPS16144 (as it is a strong-PAH/mm-bright source).For MIPS16080, its 70 µ m detection is difficult to explainwithout such a component, unless we assume a somewhatcooler AGN template. In particular, the stacked 1.2 mmdetection (1.06 ± µ m and 1.2 mmfluxes, the starburst template is required. Sd templatescan explain the stacked flux at 1.2 mm, but fail signifi-cantly with 70 µ m detections and mid-IR spectra. Thestacked limit at 70 µ m is (2 ± FIR is dominated by starburst component for all sources inthis study. RESULTS
Detections of CO Line Emission
Of the nine sources observed, eight yielded signif-icant detections in the CO J(3-2) or J(2-1) transi-tions. The CO J(3-2) spectrum of MIPS16144 alsoshows a weak continuum at the level of 0.5 ± λ obs = 2.717 mm). This observed continuum flux is con-sistent with the full-IR SED (see the open triangle sym-bol in Figure 2). Figure 3 presents the CO maps,HST/NICMOS H-band images and mid-IR spectra, one row per source. The first column shows the CO contoursoverlaid on the 24 µ m image (red, FWHM ∼ (cid:48)(cid:48) ) and the HST
NICMOS H-band image (dark, FWHM ∼ (cid:48)(cid:48) ).All CO contours start at 1 σ with an increment contourstep size of 1 σ . MIPS429 has no NICMOS H-band data,and we used the available WFPC2 F814W image instead.It has bright 24 µ m flux but no optical counterpart inboth ground-based R image and the HST/WFPC2 im-age. The second column shows the HST/NICMOS H-band images of our sources, providing morphologies atthe rest-frame ∼ Spitzer observation aiming to detectwater ice and hydrocarbons among a sample of deeplyembedded z ∼ µ m hotdust continua are indicators of star formation and AGNrespectively. One interesting question is whether themid-IR star formation indicator is correlated with colddust and molecular gas. Figure 3 shows one obvious re-sult, that is the detectability of CO gas is not stronglycorrelated with the mid-IR spectral properties, with thecaveat that this finding is based on a small number ofsources. Of the nine sources, seven have mid-IR spectrawhose prominent features is the broad emission bump at(7-8) µ m and/or fully or partially observed silicate ab-sorption trough. This type of spectra is very characteris-tic for highly obscured 24 µ m sources at z ∼
2, as shownby several recent
Spitzer observations (Houck et al. 2005;Yan et al. 2005, 2007). Local examples with this type ofspectrum are IRAS F00183-7111 and NGC4418 (Spoonet al. 2004a,b, 2001). One characteristics of these deeplyembedded local ULIRGs is their multi-temperature ISM,ranging from hot gas near the central nuclei to cold gas toouter region (Spoon et al. 2004b). Our CO observationsshow that six of these seven sources with deep silicate ab-sorption have cold molecular gas, one (MIPS8196) doesnot have CO detection at I CO ∼ < σ ∼ µ m ULIRGs at z ∼
2, a good fraction of them, including SB/AGN com-posites, deep silicate absorption systems, and AGN dom-inated galaxies, could have abundant cold molecular gas.For these 24 µ m bright ULIRGs with substantial blackholes in their centers, one important yet unsolved ques-tion is how cold molecular gas is exactly turned into starsand feeds the growth of black holes. The answers willcome from future high resolution studies with ALMA. Cold molecular gas fraction and far-IR emission
Table 4 lists the integrated CO line intensity I co in Jy km/s and luminosity L (cid:48) CO for J(3-2) or J(2-1)transition. We emphasize again that to derive coldmolecular gas mass, we assume T co (3 − /T co (1 − = 1, T co (3 − /T co (1 − = 1 and CO J(1-0) luminosity to molec-ular gas mass conversion factor α = 0.8 M (cid:12) /[K km/s pc ].The eight sources have M gas ∼ (1.02 – 3.02) × M (cid:12) ,with a median value of 1.7 × M (cid:12) . In order to com-pare our sample with other galaxy populations at z ∼ > TABLE 3Ancillary data and derived parameters for the targets
ID R H Flux Flux a Flux a Flux Flux Fluxvega AB 24 µm µm µm < < <
30 1.03 ± < < <
30 1.37 ± . ± .
03 ...MIPS8196 22.45 18.68 1.50 5 . ± . <
30 0.99 ± < < <
30 1.03 ± . ± .
06 3 . ± . . ± . ± b ± . ± .
03 0 . ± . . ± <
30 1.24 ± . ± .
02 ...MIPS16059 23.48 20.49 1.29 < . <
30 1.20 ± . ± .
03 ...MIPS16080 22.86 20.31 1.10 5 . ± . <
30 0.69 ± . ± .
03 ...MIPS16144 23.10 20.85 1.12 < . <
30 2.93 ± . ± .
03 ... a For none-detections at 70 µ m and 160 µ m, the fluxes are listed as 3 σ . b MIPS8342 70 µ m position corresponds to multiple 24 µ m sources. Here we deblended the 70 µ mblob, and list only the partial flux belonging to MIPS8342 with the flux error from the total. SeeSajina et al. (2008) for detail. TABLE 4Derived CO parameters ID a FWHM1 b FWHM2 c I co L co L (cid:48) co (10 ) d Mass(H ) d km/s km/s Jy km/s 10 L (cid:12) K km / s pc M (cid:12) MIPS429 565 ±
80 777 1 . ± .
17 3.56 2.9 2.32MIPS506 300 ±
80 330 0 . ± .
11 2.16 1.63 1.30MIPS506 dbl ± ±
40 150,87 0 . ± .
11 2.16 1.63 1.30MIPS8196 ... ... < < < < ±
50 303 0 . ± .
07 1.69 1.28 1.02MIPS8342 325 ±
70 406 0 . ± .
07 0.88 2.24 1.79MIPS8342 dbl ± ±
40 128,130 0 . ± .
07 0.88 2.24 1.79MIPS15949 500 ±
117 129 1 . ± .
11 3.59 2.72 2.18MIPS16059 471 ±
54 426 0 . ± .
11 2.33 1.76 1.41MIPS16080 130 ±
60 159 0 . ± .
15 2.69 2.03 1.62MIPS16144 744 ±
217 824 1 . ± .
27 4.99 3.77 3.02MIPS16144 dbl ± ±
70 405,334 1 . ± .
27 4.99 3.77 3.02 a The sources with double peak CO spectra have two entries per object in this table. The entrymarked with dbl has the velocity FWHM fitted with double gaussian profiles, thus the secondcolumn FWHM1 has two values per row. b FWHM1 is from the fitting with one or two gaussians, see above note. c FWHM2 is directly measured from the data as the full width at the half peak intensity, withoutassuming the gaussian profile. For the double peak CO spectra, FWHM2 are measured for bothpeaks separately. d Here most CO line intensity and luminosity are for J(3-2) transition, except MIPS8342, whichis J(2-1). To derive H mass, we assume T co (3 − /T co (1 − = 1 and also T co (2 − /T co (1 − = 1. α is CO J(1-0) luminosity to H (+He) molecular mass conversion factor. We adopt α = 0.8 M (cid:12) /[K km/s pc ] for this paper, which is suitable for SMGs and local ULIRGs. TABLE 5Derived parameters from the ancillary data for the targets
ID Log [L aIR ] Log [L aFIR ] R eff EW(7.7PAH) rest τ . µm Log [L b1 . ]L (cid:12) L (cid:12) kpc µm W Hz − MIPS429 12.72 12.50 ± .
15 ... 0 . ± . > < . ± .
09 1.97 0 . ± . > . ± . ± .
15 3.31 0 . ± .
08 1 . ± . < . ± .
21 2.07 < . . ± . . ± . ± .
30 2.28 0 . ± .
14 0 . ± . . ± . ± .
18 2.36 0 . ± .
06 0 . ± . . ± . ± .
15 2.82 0 . ± .
07 2 . ± . . ± . ± .
15 2.43 0 . ± .
05 2 . ± . . ± . ± .
12 3.24 2 . ± .
26 2 . ± . . ± . a L IR is integrated over 8 – 1000 µm , L FIR is over 40-120 µm . b L . is the monochromatic luminosity at the rest-frame 1.4GHz, derived based on the observedradio fluxes at 1.4 GHz and 650 MHz. etections of CO Molecular Gas in Spitzer z ∼ ! F ! MIPS429 MIPS8196 0.0010.0100.1001.000 ! F ! MIPS16080 MIPS5061 10 100 " [ ! m]0.0010.0100.1001.000 ! F ! MIPS8327 1 10 100 " [ ! m] MIPS160591 10 100 " [ ! m]0.0010.0100.1001.000 ! F ! MIPS8342 1 10 100 " [ ! m] MIPS159491 10 100 " [ ! m]0.0010.0100.1001.000 MIPS16144 Silicate absorption sourcesPower law sourcesStrong-PAH source
Total galaxyAGNstarburststarsNot included in fits:Sd templatenormalized to the 1.2mmstacked MAMBO point
Fig. 2.—
Near-to-far-IR spectral energy distributions for the 9 targets. The figure legend shows the starburst, AGN and stellar components.The large cyan stars at the observed 1.2 mm mark the averaged flux value of 1.06 ± Lin Yan et al. H H ”
5” 12 ” ” H ” . ” ” N o H - b a nd d a t a M I PS M I PS M I PS M I PS Fig. 3.—
The first column shows the the 24 µ m image in pink, HST/NICMOS H-band image in black in its original spatial resolution,and CO image in contours. In all panels, the CO contours start at 1 σ with a step size of 1 σ . For σ values, see the text for detail. Theimage size is 12 (cid:48)(cid:48) × (cid:48)(cid:48) . The second column shows the H-band morphology. The third and fourth column show the CO and mid-IR spectra.The MIPS16144 CO spectrum shows a detection of continuum, marked by the horizontal dashed line. with CO detection, we compiled a list of SMGs based on the published data (Greve et al. 2005; Tacconi etetections of CO Molecular Gas in Spitzer z ∼ H H H H ” . ” M I PS M I PS M I PS M I PS Fig. 3.—
Continue. The first column image size is 10 (cid:48)(cid:48) × (cid:48)(cid:48) , and the second column image size is 1.4 (cid:48)(cid:48) × (cid:48)(cid:48) . al. 2006, 2008), shown in Table 6. This sample of 14SMGs includes only sources with significant detections,and their median gas mass M gas is 3.05 × M (cid:12) . In addition, a sample of 19 QSOs and radio galaxies withlow resolution CO observations (Solomon & Vanden Bout2005) is also used as a QSO comparison sample. As de-0 Lin Yan et al. H ” ” MIPS8196
Fig. 3.—
Of 9 objects observed, this is the only source with no significant detection. scribed in § L (cid:48) CO ver-sus redshift for the 24 µ m ULIRGs and the comparisonsamples of SMGs, QSOs and radio galaxies. The rightpanel in Figure 4 presents the histogram of L (cid:48) CO withthe dashed lines marking the median values of the threepopulations. The average CO mass of our 24 µ m ULIRGsample is about a factor of 2 less than the average val-ues calculated from the specific comparison samples ofSMGs and gas rich QSOs currently available in the lit-erature. Kolmogorov-Smirnov tests on 24 µ m ULIRGsversus SMGs and 24 µ m ULIRGs versus QSOs derivedthe probabilities of two datasets drawn from the sameparent sample of 0.19 and 0.189 respectively. We em-phasize that the statistics is still too small to draw anydefinitive conclusions for the overall populations of thesethree types of sources. Fig. 4.—
This compares the CO integrated line luminosity as afunction of redshift for various sources at high- z . The SMG dataare from Greve et al. (2005) and Tacconi et al. (2006). The QSO& radio galaxy data are from the compilation of Solomon & vandenBout (2005). The dashed lines in the L (cid:48) CO distribution plot (rightpanel) mark the median values for the three galaxy populations.The comparison reveals tentative evidence that on average 24 µ mULIRGs may have less cold molecular gas than those of SMGs andQSOs. If far-IR emission is considered mostly from massive young stars born in recent starbursts, L
FIR can be di-rectly related to star formation rate (SFR). Therefore,the ratio of L
FIR / L (cid:48) CO can be interpreted as describe(1) gas depletion time scale, τ SF = M gas / SFR or (2) starformation efficiency, SFE = L
FIR / M gas , i.e. how muchluminosity can be produced by 1 M (cid:12) of gas. If wetake the SFR[M (cid:12) / yr] = c ∗ (cid:0) L FIR L (cid:12) (cid:1) , where c = 0.8 – 3,depending on the definition of L FIR (Kennicutt 1998;Meurer et al. 1997). We adopt c = 1.5 because we useL
FIR = L − µ m and the commonly adopted relationby Kennicutt (1998) has c = 1.72 with L FIR = L − µ m .The L FIR is measured from the full SED fitting describedin § (cid:12) yr − , withthe mean value of 585 M (cid:12) yr − . These values are sub-stantially larger than normal star forming galaxies, buta factor of 2 – 3 less than what has been observed forSMGs at the similar redshifts. Furthermore, if we as-sume that 70% of L FIR from SMGs is due to star for-mation, as suggested by some recent studies (Alexanderet al. 2008), the averaged SFR for SGMs would still behigher than that of 24 µ m sources by a factor of (1.5 – 2).The inferred values for SFE and gas depletion time scale τ SF range from 69 – 388 L (cid:12) M − (cid:12) ( (cid:104) SF E (cid:105) = 330 L (cid:12) M − (cid:12) )and 17 – 96 Myr ( (cid:104) τ SF (cid:105) = 38 Myrs) respectively. For com-parison, star formation efficiency (cid:104) SF E (cid:105) is typically(180 ± L (cid:12) M − (cid:12) for local ULRIGs (Solomon et al.1997) and 450 ± L (cid:12) M − (cid:12) for SMGs (Greve et al. 2005;Tacconi et al. 2006). And the gas depletion time scale τ SF ranges from 10 – 100 Myrs for SMG, and a factor of10 longer for local LIRGs and ULIRGs (Solomon & Van-den Bout 2005). Figure 5 displays the L FIR / L (cid:48) CO ratioversus L (cid:48) CO (or equivalently cold molecular gas), visu-ally illustrating our conclusion that these bright 24 µ mselected z ∼ CO velocity widths and dynamical mass estimates
The CO velocity width (FWHM) can be measured byusing either a single or double gaussian fit or directlyfrom the data (full width at the half peak intensity).Table 4 shows the results using both methods. Assum-ing a single component and without profile fitting (Ta-ble 4 FWHM2), we have velocity widths ranging from128 km/s to 824 km/s, with a median value of 406 km/s.If we use two velocity components for sources with dou-ble peak profiles, the averaged velocity width reducesto 275 km/s. With a single component fitting, Figure 6compares the CO velocity widths of our sources with thatof SMGs and QSOs. Here the SMG comparison sample islisted in Table 6 for 14 objects compiled from Greve et al.etections of CO Molecular Gas in Spitzer z ∼ TABLE 6Comparison Sample I – SMG data ID z co L FIR L (cid:48) CO (3 − a V bFWHM double Ref. c L (cid:12) K km/s pc km/sSMMJ023956 − ±
60 yes (1)SMMJ023951 − ±
50 yes (1)SMMJ044307+0210 2.5094 3.04e+12 1.1e+10 350 ±
60 yes (1)SMMJ094303+4700 3.3460 1.67e+13 2.7e+10 420 ±
50 yes (1)SMMJ123549+6215(HDF76) 2.2021 8.90e+12 5.0e+10 600 ±
50 yes (2)SMMJ123707+6214(HDF242) 2.490 4.00e+12 2.0e+10 430 ±
60 yes (2)SMMJ131201+4242 3.408 1.18e+13 5.3e+10 530 ±
50 no (1)SMMJ140103+0252 2.5653 5.51e+12 1.7e+10 200 ±
40 no (1)SMMJ163554+6612 2.5168 1.33e+12 3.7e+9 500 ±
100 yes (1)SMMJ163650+4057(N2850.4) 2.3853 1.56e+13 7.0e+10 840 ±
110 yes (2)SMMJ163658+4105(N2850.2) 2.4500 2.03e+13 5.7e+10 870 ±
80 yes (2)SMMJ163706+4053 2.3800 2.00e+13 3.0e+10 830 ±
130 no (1)EROJ164502+4626 1.4950 9.69e+12 4.1e+10 400 ±
20 no (1)SMMJ221735+0015 3.0990 1.20e+13 3.8e+10 780 ±
100 no (1) a All Greve et al. (2005); Tacconi et al. (2006, 2008) assume that T co (3 − /T co (1 − = 1 and α = 0.8 M (cid:12) /[K,km/s pc ] for the molecular gas mass calculation. b The line velocity width is from a single gaussian fit to the spectral data. c References: (1) Greve et al. 2005, MNRAS, 359, 1165. This paper gives the original referencesfor some of the sources which were not observed by Greve et al. 2005. (2) Tacconi et al. 2008,ApJ, 680, 246
Fig. 5.—
The plot shows the ratio of L
FIR [L (cid:12) ] andL (cid:48) CO [K km / s pc ] versus L (cid:48) CO . The black, open circles indicate localinfrared and ultra-luminous galaxies from Solomon et al. (1997),the red, open triangles are our data points, the blue, open squaresthe SMGs the published data from Greve et al. (2005); Tacconi etal. (2006, 2008), and the green crosses the QSO & radio galaxydata compiled by Solomon & Vanden Bout (2005). (2005); Tacconi et al. (2006, 2008). The QSO comparisonsample is compiled by Solomon & Vanden Bout (2005)(also see Greve et al. (2005)). The right panel presentsthe velocity distributions, with dashed lines marking themedian values of 406, 635 and 350 km/s for Spitzer z ∼ z ∼ µ m ULIRGs versus SMGsand 24 µ m ULIRGs versus QSOs drawn from the same parent distributions.Recent N-body/Smoothed-Particle Hydrodynamic(SPH) numerical simulation combined with 3D poly-chromatic radiative transfer by Narayanan et al. (2009)show that SMGs are mostly produced by major mergersof gas rich, star forming dominated progenitors, whereas24 µ m bright ULIRGs at z ∼ µ m ULIRGs spread over(100-1500) km/s, overlapping with that observed amongSMGs, but the median value is smaller than that ofSMGs.Spatial distribution of molecular gas holds theimprints of kinematic state of a galaxy. Al-though the majority of our sample is unresolved,two targets, MIPS506 and MIPS16144, show spa-tially extended CO emission across regions of ∼ (cid:48)(cid:48) × (cid:48)(cid:48) ∼
41 kpc ×
36 kpc and2.44 (cid:48)(cid:48) × (cid:48)(cid:48) ∼
20 kpc ×
18 kpc respectively. MIPS506shows two distinct components with a separation of45 kpc. The CO spectra in Figure 3 show that MIPS506,MIPS16144, and MIPS8342 have double peaked veloc-ity profiles, with single and double gaussian fit (solidand dashed lines respectively) to the observed data (his-togram). A double peak CO velocity spectrum is in-dicative of either a rotating disks or of two galaxies inan ongoing merger. Our CO data measures 3 out of 9(33 ± ±
21% (9/14) having double ormultiple velocity components. If we consider pure Pois-son statistical errors, these two fractions are consistentwithin 1 σ .One important question is whether the two velocitycomponents in each of the three double peak sourcesare spatially separated. Figure 7 shows the CO con-tours overlaid on the 24 µ m (pink) and HST
H-band(black) images, with solid and dashed contours indicat-2 Lin Yan et al.
Fig. 6.—
The plot compares the CO velocity width (FWHM) ver-sus molecular gas masses of our sources with that of high-z QSO &radio galaxies and SMGs. All velocity widths are based on a singlecomponent fitting. The SMG comparison sample is listed in Ta-ble 6, compiled from Greve et al. (2005); Tacconi et al. (2006), andthe QSO & radio galaxy data are from the compilation of Solomon& Vanden Bout (2005); Greve et al. (2005). The right panel showsthe velocity distributions of three populations with the dashed linesmarking the median values. The velocity width distributions showtentative evidence that 24 µ m ULIRGs are similar to QSOs, havingsmaller CO velocity width than the averaged value for SMGs. ing two separate velocity peaks. Here, the integratedCO maps are summed over (4 – 5) velocity channels (seeTable 2 for the channel width), covering the respectivevelocity peak. For example, the integrated CO mapsfor MIPS16144 shown in Figure 7 are summed from -730 km/s to -330 km/s for the velocity peak at -605 km/sand from -330 km/s to -70 /km/s for the velocity peak at-138 km/s. The similar maps are made for MIPS506 andMIPS8342, and shown in Figure 7. The rms σ values usedfor the contours are in Table 2. The two velocity peaks inMIPS16144 have a clear spatial separation of 1 . (cid:48)(cid:48) , whichcorresponds to 10.9 kpc. The two velocity components inMIPS8342 are spatially too close to be resolved. In thecase of MIPS506, the separation is ∼ . (cid:48)(cid:48) , correspond-ing to 45 kpc. Such a large distance, particularly for coldmolecular gas, suggests that MIPS506 is probably a pairof merging galaxies, with substantial dust extinction inthe rest-frame optical band. This explains the very faintmultiple blobs (2.8 (cid:48)(cid:48) ∼
24 kpc separation) in the H-bandimage (6 (cid:48)(cid:48) × (cid:48)(cid:48) ) shown in Figure 3. We have examinedour shallow 3.5 – 8 µ m IRAC images to see if the galaxypair detected in the CO map is also detected in otherbands, only an unresolved object is detected at 3.5 and4.8 µ m, and no detections at other two IRAC bands. Thephysical model for MIPS16144 and MIPS8342 is prob-ably also merger, although rotating disk model is alsoconsistent. Additional supporting evidence for Spitzer
ULIRGs being mergers comes from the HST/NICMOSH-band images, with the detailed discussion in § dyn = R*v / G = R*∆V / (sin (i)G), where sin(i) is TABLE 7Comparison Sample II – QSO data a ID z co M BH M gas M dyn incl. angle10 M (cid:12) M (cid:12) M (cid:12) APM08279+5255 3.91 23 13 22 25degPSSJ2322+1944 4.12 1.5 1.7 4.4 no corr.BRI1335-0417 4.41 6 9.2 10 no corr.SDSSJ1148+5251 6.42 3 2.2 5.5 65deg a The black hole masses are derived from UV spectroscopy, thegas and dynamic masses from the high spatial resolution COobservations. References: Walter et al. (2004); Riechers et al.(2008a,b, 2009). the inclination angle, R is the separation between twomerging objects, and ∆V is the observed CO peak ve-locity difference. R and ∆V are (10.9 kpc, 463 km/s)and (45 kpc,174 km/s) for MIPS16144 and MIPS506 re-spectively. We derived M dyn = 5.4 × sin − (i) and3.2 × sin − (i)M (cid:12) for MIPS16144 and MIPS506. Ifassuming an averaged value of (cid:104) sin (i) (cid:105) = 2 /
3, the gasfraction, M gas / M dyn , is 4% and 3% for MIPS16144 andMIPS506 respectively. Although several recent papershave published gas fractions for various types of galax-ies, we caution that the equation used for calculatingthe dynamical masses based on observed emission lineFWHM could differ by a factor of (2 – 5), see Neri et al.(2003); Erb et al. (2006); Tacconi et al. (2006); Riecherset al. (2009).The stellar half-light radii have been measured fromthe HST/NICMOS H-band (1.6 µ m) images for our sam-ple (Dasyra et al. 2008; Zamojski et al. 2009). Us-ing M dyn = R / *v / G = R / *(∆V COFWHM ) / (sin (i)G)and assuming that cold molecular gas disks are smallerthan stellar half-light radii, we can also set limits on thedynamical masses of additional 5 sources in our sam-ple. This assumption is usually correct for nearby galax-ies (Sanders et al. 1991; Solomon et al. 1997). We notethat our sources probably have fairly high dust extinctionat the observed H-band (rest-frame 5000˚A), which willcause systematic under-estimates of the intrinsic stellarsizes. Using the half-light radii are listed in Table 5,we obtained M dyn ∼ (0.2 – 2.2) × M (cid:12) for MIPS8327,MIPS8432, MIPS15949, MIPS16059 and MIPS16080, us-ing an averaged inclination angle. These rough estimatesare on the same order of magnitude in comparison withthe averaged dynamical mass of SMGs, 1.2 × M (cid:12) (Greve et al. 2005; Tacconi et al. 2006). Our targets are likely mergers
The CO images and spectra in Figure 1 and Fig-ure 3 have revealed that at least one source, MIPS506,is a merger involving two galaxies with substantialcold molecular gas. In addition, MIPS8327 andMIPS16059 are likely interacting galaxies because theirHST/NICMOS H-band images show close companionswithin separations of 3.3 and 4 kpc (0 . (cid:48)(cid:48) & 0 . (cid:48)(cid:48) ) re-spectively. Zamojski et al. (2009) performed detailed2-dimensional light profile fitting to the HST images, as-etections of CO Molecular Gas in Spitzer z ∼ MIPS506 MIPS8342 MIPS16144
Fig. 7.—
Overlaid on the 24 µ m (pink) and NICMOS H-band(black) images are the CO contours of the integrated maps overthe positive (solid lines) and negative (dashed lines) velocity peaksfor the 3 sources with double peak profiles. The solid and dashedcontours are for the CO velocity peaks at (-138, -605) km/s,(160,-15) km/s and (-250,-520) km/s for MIPS16144,MIPS506 andMIPS8342 respectively. The contour starts at 1 σ with a step of 1 σ for all three panels. For the detail, see the text in § suming a sersic profile model . Figure 8 compares theoriginal H-band images with the model subtracted, resid-ual images (reversed intensity scale) for 5 of our sourceswith CO detections, MIPS506, MIPS16144, MIPS8342,MIPS16080 and MIPS15949. Multiple blobs an cleartidal tails in Figure 8 suggests that MIPS506, MIPS16144and MIPS16080 are probably the products of galaxyinteractions. Although the original H-band images ofMIPS8342 and MIPS15949 show isolated early typegalaxies, but the faint disks/rings in the residual imagesindicate that they could be late stage mergers. Majoritysources in our CO sample have merger signatures eitherin their CO maps or in the H-band images. Consider-ing the published studies of SMGs based on HST im-ages and integral spectroscopy (Chapman et al. 2003;Conselice et al. 2003; Swinbank et al. 2006), we spec-ulate that SMGs, bright 24 µ m ULIRGs, and gas richQSOs are connected by an evolutionary model involvingmergers, with SMGs being in an early stage of merg-ers of two gas rich, star forming progenitors, and bright24 µ m ULIRGs with substantial obscured AGN in a latermerger stage, while QSOs are fully merged with massiveblack holes in place, producing powerful feedback clear-ing away dust obscuration. This simple picture does notexplain how SMGs and gas rich QSOs could have such adifferent history of star formation and black hole growth.Specifically, most SMGs have very weak AGNs, and theirestimated black hole to stellar mass ratio is roughly onthe order of 1:10 (Borys et al. 2005; Alexander et al.2008), whereas, recent studies (see Table 7) found the I ( r ) = I e *exp( − κ ∗ [(r / r e ) / n − κ is a function of sersicindex n . Sersic index n is 4 for de Vaucouleurs profile and 1 forthe exponential profile. same ratio for these z ∼ > H assuminga M stars / L H ∼ BH / M stars is on the order of1:10 . Again, this suggests that Spitzer
ULIRGs may bea transitional type of sources between SMGs and QSOs.To really understand the physical relation among thesethree types of sources, we will need better UV opticalspectra, complete SEDs, and high resolution CO datafor
Spitzer
ULIRGs. SUMMARY AND DISCUSSIONS
We report CO interferometric observations of nine z ∼ S µm ∼ > Spitzer mid-IR spectra, pho-tometry and the CO data, we find the following results: • Of the nine sources observed at PdBI, eightsources have significant detections of the CO J(3-2) or J(2-1) transition with intensity I c ∼ (0.5 –1.5) Jy km s − at (4.7 – 9.7) σ . CO line emissionis detected from sources with a variety of mid-IRspectral type, including strong PAH, deep silicateabsorptions and mid-IR power-law. The observedCO J(3-2) [or (2-1)] line luminosities L (cid:48) CO are (1.3 –5.5) × K km/s pc , yielding cold molecular gasmasses of (1 – 4.4) × M (cid:12) , based on the assump-tion of T CO (3 − /T CO (1 − = 1 and the CO (1-0)luminosity to H gas mass conversion factor of0.8 M (cid:12) /[K Km/s pc ]. Based on small statisticalsamples, we find tentative evidence suggesting that Spitzer
ULIRGs have a factor of 2 less cold molec-ular gas than what observed among SMGs and gasrich QSOs. • The CO velocity width distribution of
Spitzer
ULIRGs is similar to that of QSOs with CO de-tections, but on average a factor of 1.5 smallerthan that of SMGs, with caveat that the avail-able CO statistics is small. Two objects, MIPS506and MIPS16144, have spatially resolved CO emis-sion across regions of 100 and 46 kpc respectively,significantly extended after taking into account oftheir beam sizes and shapes. Three of our nine tar-gets (33%) have double peaked CO spectra, in com-parison with 64% observed among SMGs. Two ofthese three, MIPS506 and MIPS16144, have theirtwo velocity peaks spatially separated by ∼
45 kpcand 10.9 kpc respectively. Such a large spatial sep-aration of two distinct CO knots suggests thatMIPS506 is a pair of merging galaxies. The in-ferred dynamical masses are 5.4 × sin − (i) and3.2 × sin − (i)M (cid:12) for MIPS16144 and MIPS506respectively, yielding the gas fraction (M gas / M dyn )of 8% and 6% if using an averaged inclination an-gle. • Together with spatially resolved CO data, ourHST/NICMOS H-band images suggest that ma-jority of our sample have signatures of mergers at4 Lin Yan et al.
Original Residual Original Residual
Fig. 8.—
The HST H-band images and their residual images with a sersic profile model subtracted for the 5 sources which do not showdistinct galaxy pairs. The intensity scale is reversed for the residual images in order to bring out the faint features. These images revealmorphological signatures suggesting dynamical interactions. various stages of dynamical interactions, includ-ing close companions, tidal tails, and rings in theresidual images. We hypothesis that SMGs, bright24 µ m ULIRGs and gas rich QSOs may be con-nected by the same evolutionary model in whichthe formation mechanism is primarily mergers,with SMGs mostly at an early passage of two gasrich, star forming galaxies, bright 24 µ m ULIRGs ata later stage of dynamical interaction, with gas anddust surrounding the central black holes, resultingsmaller observed CO velocity width, and finally,the gas rich QSOs representing the final mergedstage.Sylvain Veilleux is thanked for providing the electronicversion of the mid-IR spectra for local IRAS z QSOs with high resolution CO observations. This work isbased on observations with 30m telescope of the
Institutefor Radioastronomy at Millimeter Wavelengths (IRAM),which is funded by the German Max Planck Society, the French CNRS and the Spanish National GeographicalInstitute. We thank the staff of the IRAM Observa-tory for their support of this program. Also based onobservations taken with
Spitzer Space Telescope , whichis funded by NASA and operated by JPL/Caltech. H-band imaging data are from observations made with theNASA/ESA Hubble Space Telescope and obtained atSpace Telescope Science Institute, operated by the Asso-ciation of Universities for Research in Astronomy, Inc.,under NASA contract NAS5-26555. We thank helpfuldiscussions with Nick Scoville and Arjun Dey.etections of CO Molecular Gas in Spitzer z ∼2 ULIRGs 15