ULTRALUMINOUS STAR-FORMING GALAXIES AND EXTREMELY LUMINOUS WARM MOLECULAR HYDROGEN EMISSION AT z = 2.16 IN THE PKS 1138−26 RADIO GALAXY PROTOCLUSTER

Galaxy clusters are the largest virialized structures in the universe. Massive, X-ray-selected clusters are well established at redshift z = 1 (Finoguenov et al. 2007), and their brightest galaxies have colors consistent with passively evolving ellipticals (e.g., McCarthy et al. 2007). The progenitors of such clusters must have formed at an earlier epoch, around the highest density peaks in the cosmos. Such protoclusters have been found around z = 2–5 radio galaxies (Pentericci et al. 2000; Venemans et al. 2007; Hatch et al. 2011; Kuiper et al. 2011; Tanaka et al. 2011), which appear to be construction zones of cluster ellipticals and their supermassive black holes (Lacy et al. 2000).

The baryonic mass of z < 1 galaxy clusters is dominated by primordial gas that is virialized and heated to 10⁷ K in the potential wells of dark matter halos (Andreon 2010; Giodini et al. 2009). A reduction of the gas fraction relative to the cosmic baryon fraction can be explained by a combination of star formation, supernova feedback, and active galactic nucleus (AGN) feedback (Bode et al. 2009). There is evidence that z = 2 protoclusters are X-ray deficient (L_X < 10⁴⁴ erg s⁻¹) compared to clusters at lower redshift (Overzier et al. 2005), supporting the idea that they are in the process of formation and not yet virialized.

Chandra X-ray imaging of the Spiderweb radio galaxy (PKS 1138−26, z = 2.156) shows extended X-ray emission aligned with the radio lobes (Carilli et al. 2002), which indicates that the X-ray emission may come from gas that is shock heated by the radio jet rather than virialized in the cluster potential. Large polarization rotation measures are found for PKS 1138−26 and other high-redshift radio galaxy lobes, indicating the presence of large column densities of ionized, magnetized gas (Carilli et al. 1997; Athreya et al. 1998). This gas may have its origin in radio jet-driven outflows (Nesvadba et al. 2006).

Large quantities of shock-heated warm molecular gas are found by Spitzer in 30% of 3C radio galaxies at z < 0.2 (Ogle et al. 2010, 2007; Guillard et al. 2012). The heretofore most luminous and distant H₂ emission is seen in the z = 0.29 Zw 3146 brightest cluster galaxy, with a single H₂ line luminosity L(H₂ 0–0 S(3)) = 6.1 × 10⁴² erg s⁻¹ (Egami et al. 2006). In radio galaxies, the H₂ is most likely powered by kinetic energy from the relativistic jet that is dissipated in the host galaxy interstellar medium (ISM). The large ratio of H₂ to polycyclic aromatic hydrocarbon (PAH) luminosity observed in these and other H₂ emission galaxies (MOHEGs) cannot be generated in stellar photodissociation regions (PDRs). X-ray heating by relatively low luminosity AGNs is also insufficient to power the observed H₂ rotational line luminosities (Ogle et al. 2010). Star formation appears to be suppressed in radio-loud MOHEGs (Nesvadba et al. 2010), falling a factor of ∼50 below the Kennicutt–Schmidt relation between specific star formation rate and gas surface density (Kennicutt 1998).

A spectacular protocluster surrounds the Spiderweb radio galaxy (Kurk et al. 2000). As many as 121 candidate cluster members within a 45 (2.2 Mpc) radius have been identified as possible Lyα-emitters (37 LAEs), Hα-emitters (40 HAEs), or extremely red objects (44 EROs) via filter photometry (Kurk et al. 2000, 2004b). A large overdensity of sources with red IRAC [3.6]–[4.5] colors or bright 24 μm emission are found within a radius of ∼1'–2' (Galametz et al. 2012; Mayo et al. 2012). At least 24 cluster members have been spectroscopically detected in Lyα (15) or Hα emission (9) and have redshifts close to that of the Spiderweb (Pentericci et al. 2000; Kurk et al. 2004a; Venemans et al. 2007; Doherty et al. 2010). The cluster consists of two merging sub-clumps, with velocity dispersions of 520 and 280 km s⁻¹ for the associated LAEs and a combined virial mass of ∼1 × 10¹⁴ M_☉ (Pentericci et al. 2000). Several HAEs are bound to the Spiderweb Galaxy with a velocity dispersion of 360 km s⁻¹, indicating a virial mass of 8 × 10¹³ M_☉(Kurk et al. 2004a).

Rest UV to optical continuum imaging of the cluster core with the Hubble Space Telescope (HST) shows multiple satellite galaxies with peculiar morphologies within 150 kpc of the Spiderweb Galaxy (Pentericci et al. 1998; Miley et al. 2006; Hatch et al. 2009). An enormous Lyα halo imaged by the Very Large Telescope surrounds the galaxy out to a radius of >40 kpc, indicating a large reservoir of atomic gas (Kurk et al. 2002; Hatch et al. 2008).

We made a deep Spitzer infrared spectroscopic map of the central region of the PKS 1138−26 protocluster, centered on the Spiderweb radio galaxy, in order to explore mid-IR indicators of star formation and shocked gas. Here, we present mid-IR spectra of the Spiderweb and two other protocluster galaxies which are bright in the MIPS 24 μm band. Spectra of the remaining sources will be presented in a future paper. Distance-dependent quantities in this paper are calculated using cosmological parameters H₀ = 70 km s⁻¹, Ω_Λ = 0.7, and Ω_m = 0.3.

We used the Spitzer Infrared Spectrograph (IRS; Houck et al. 2004) to map a 32 × 11 strip of the PKS 1138−26 protocluster (Figure 1). The LL1 (20.5–38.5 μm) low-resolution slit (168'' × 107) was stepped perpendicular to the slit in 11 × 50 steps and parallel to the slit in 5 × 50 steps to create a map centered on the Spiderweb radio galaxy. Exposures were 2 × 120 s per pointing. The map was repeated eight times over five days, giving a total exposure time of 29.3 hr and a mean exposure time per pixel of 4.2 hr. Rotation of the spacecraft and spectrograph slit (P.A. = 1246–1263) over this time period improved the spatial sampling of the map. Parallel data were also recorded in the LL2 (14–21 μm) slit, which will be reported elsewhere.

Zoom In Zoom Out Reset image size

Spectra (Figure 2) were reduced from SSC Pipeline 18.18 basic calibrated data (BCDs) using pipeline uncertainties and custom background subtraction. The background was estimated by stacking the BCD frames in a time series, masking sources, and fitting each pixel time series with a robust, low-order polynomial. The LL1 spectral cube was built using CUBISM (Smith et al. 2007a), with a spatial oversampling factor of three, and cleaned using automatic 10σ and interactive bad-pixel identification and removal. A noise cube was constructed from the difference of the first and second exposures for each pointing, and its pixel-to-pixel standard deviation was found to agree closely with the pipeline-derived uncertainties as a function of wavelength. For most purposes, we confidently use the pipeline uncertainties in our data analysis. However, these uncertainties appear to significantly underestimate the noise at observed wavelengths >33 μm (rest frame 10.4 μm). We attribute this to systematically deviant pixels that we were unable to identify and remove in this noisier wavelength range. Feature fluxes measured in this wavelength range, such as those of [S iv] and the 11.3 μm PAH, should be interpreted with caution.

Zoom In Zoom Out Reset image size

Sources were identified and their positions measured in the collapsed IRS cube weighted by the MIPS 24 μm band spectral response (Figure 1). Spectral extractions were performed using a three-dimensional optimal extraction algorithm that we adapted from SPICE two-dimensional optimal extraction (Narron et al. 2007; Horne 1986). We used a wavelength smoothed and interpolated three-dimensional point-spread function template derived from the Spiderweb itself for optimal extraction. The flux from this source is dominated by the spatially unresolved AGN. Spectra were optimally extracted (using signal-to-noise weighting) within an aperture that expands linearly with wavelength and has a diameter of eight pixels (2.7 native LL1 pixels or 14'') at 27 μm. Such a small aperture is utilized to minimize contamination from nearby sources. A wavelength-independent flux calibration factor equal to the pixel solid angle (5.624 × 10⁻¹¹ sr pixel⁻¹) was derived from the ratio of SPICE over CUBISM extractions of the Spiderweb, in BCDs where it was centered on the slit. We applied a correction for detector fringing to the Spiderweb spectrum, reducing ∼3% oscillations in the observed 21–25 μm region down to the 1% level.

We used PAHFIT (Smith et al. 2007b) to obtain a least-squares fit to the continuum, emission lines, and PAH features. The continuum was fitted with the sum of blackbody spectra covering a range of dust temperatures. No extinction was included in the fits since it did not improve them. The emission lines were fitted with unresolved Gaussian line profiles. In the case of emission line blends, the fluxes were summed to give the total flux of the blend, rather than report individual line fluxes. The PAH features are fit by Drude profiles, with rest wavelengths and widths fixed to values appropriate for star-forming low-redshift galaxies, but with relative intensities allowed to vary. While it is not clear a priori that the mean PAH composition and feature shapes should be the same at high redshift, this provides a good fit for most features, except for the 8.6 μm PAH which appears to be shifted to shorter wavelength (Section 3).

We present the stacked IRS cube in Figure 1, weighted by the MIPS 24 μm band spectral response. We detect 13 sources with peak flux >0.2 mJy that yield high-quality spectra with peak S/N > 10 per 0.18 μm bin in the observed 21–35 μm band. In this paper, we concentrate on the three brightest 24 μm sources matched to spectroscopically confirmed cluster members—Spiderweb, PKS 1138−26:[KPR2004] HAE 0229, and PKS 1138−26:[KPR2004] HAE 0131 (hereafter HAE 229 and HAE 131).

The IRS spectrum of the Spiderweb is dominated by hot dust emission from the central AGN (Figures 2(a) and (b)). A silicate emission bump observed at 9.3–11 μm (rest) is fit by a 120 K blackbody multiplied by the silicate emissivity function. We detect PAH emission at 7.7 μm, 8.6 μm, and 11.2 μm. The 11.33 μm PAH feature falls in a noisy region of the spectrum, so it is formally not detected. However, the majority of the reported 11.23 μm flux (Table 1) is likely contributed by the blue wing of the 11.33 μm PAH feature.

We detect a number unresolved emission lines (Table 2) in the spectrum of the Spiderweb, including [Ne vi], [Ar iii], [S iv], and H₂ 0–0 S(3). The H₂ 0–0 S(4) line may be detected, but could be confused with substructure near the peak of the 7.7 μm PAH feature. A blend of [Ar ii] and H₂ 0–0 S(5) is detected at 6.9 μm. We confirm the H₂ 0–0 S(3) line detection with a continuum-subtracted map of the line at z = 2.156 (Figure 1, bottom). This line is only detected at the location of the Spiderweb Galaxy, with a peak significance of >6σ in the map and an integrated line detection of 14σ in the spectrum.

HAE 229 and HAE 131 both have PAH-rich spectra (Figures 2(c) and (d), and Table 1). We detect PAH blends at 7.7 μm, 8.6 μm, and 11.3 μm (rest) in both galaxies. The [Ar ii] line is detected in HAE 229 and marginally detected in HAE 131. The [Ne vi] and H₂ 0–0 S(4) lines may be detected in HAE 131, however, it is difficult to clearly distinguish these lines from possible substructure in the 7.7 μm PAH feature. While the H₂ 0–0 S(3) line is formally detected in HAE 131 (0.11 ± 0.03 × 10⁻¹⁴ erg s⁻¹ cm⁻²), the peak of the line is not at the expected wavelength and may be contaminated by a noise spike, so we treat this as an upper limit. This line is clearly not detected in the continuum-subtracted H₂ S(3) map (Figure 1). Hot dust continuum underlies the PAH spectrum in HAE 229 and weaker continuum emission is also visible at the short wavelength end of the HAE 131 spectrum.

Shifting the wavelength of the 8.6 μm PAH feature from the standard wavelength of 8.61 μm to 8.57 μm (corresponding to a Doppler shift of −1400 km s⁻¹) gave a better fit to the spectra of all three sources. We believe this indicates a shift in the intrinsic wavelength rather than a Doppler shift, because the emission lines and 7.7 and 11.3 μm PAH features have wavelengths consistent with the optical–NIR emission line redshifts. The 8.6/7.7  μm PAH ratios (0.21–0.38) seen in the three galaxies are somewhat greater than the range of PAH ratios (0.11–0.21) seen in low-redshift star-forming galaxies (Smith et al. 2007b). These differences in 8.6 μm PAH wavelength and PAH ratios may indicate a difference in chemical composition or excitation of the PAH molecules compared to low-redshift galaxies.

4.1. Spiderweb Galaxy

The Spiderweb Galaxy is surrounded by a large number of smaller satellite galaxies, seen in HST ACS images (Pentericci et al. 1998; Miley et al. 2006; Hatch et al. 2009). Eleven cluster members or candidates have been identified within our 14'' diameter spectral extraction aperture, according to NED (Figure 1). These include the Spiderweb, X-ray source CXOGBA J114047.9-262906, four LAE candidates, three HAE candidates, and two EROs. Except for the Spiderweb, none have spectroscopic redshifts available. From the strong hot dust emission and low equivalent widths of the PAH features in the Spiderweb spectrum, it is clear that most of the observed 24 μm emission comes from the AGN. The strong [Ne vi] and [S iv] emission lines are additional indicators of a dominant AGN contribution to the spectrum.

The Spiderweb has a 7.7 μm PAH luminosity of 1.7 × 10⁴⁵ erg s⁻¹ (4.4 × 10¹¹ L_☉, Table 3). In comparison, the brightest ULIRG in the Elbaz et al. (2011) sample has a PAH-dominated 8 μm luminosity of 2 × 10¹¹ L_☉, a star-forming IR (SFIR) luminosity of L_SFIR ∼ 3.3 × 10¹² L_☉, and a star formation rate of SFR = 500 M_☉ yr⁻¹. The corresponding conversions among 8 μm luminosity, total SFIR luminosity, and star formation rate are L_SFIR/L₈ = 17 and L₈/SFR = 4 × 10⁸ L_☉ M⁻¹_☉ yr. Lacking rest-frame far-IR fluxes, we adopt these as representative conversion factors for the ULIRGs in our sample. This gives an approximate PAH-based star-forming FIR luminosity for the Spiderweb of L_SFIR ∼ 7.4 × 10¹² L_☉ and a star formation rate of ∼1100 M_☉ yr⁻¹. Equally large PAH luminosities and PAH-based star formation rates have been reported in the z = 3.53 radio galaxy 6C J1908+7220 (Seymour et al. 2008; 4000 M_☉ yr⁻¹), the Cloverleaf Quasar (Lutz et al. 2007; ∼1000 M_☉ yr⁻¹), and other high-redshift mm-bright QSOs (Lutz et al. 2008).

Table 3. Luminosities and SFRs

Source	L(PAH7.7)a	L(Hα)b	LSFIRc	SFR(PAH)d	SFR(Hα)e
Spiderweb	4.4 (0.3)	38 (3)	7.4 (0.5)	1100 (60)	1200 (100)
HAE 229	3.5 (0.5)	0.64 (0.17)	6.0 (0.9)	880 (130)	20 (5)
HAE 131	1.9 (0.2)	1.1 (0.2)	3.2 (0.3)	470 (30)	34 (6)

Notes. ^aLuminosity (10¹¹ L_☉) of 7.7 μm PAH blend, including contributions from the 7.42 μm, 7.60 μm, and 7.85 μm features. ^bHα luminosity (10⁹ L_☉). Hα fluxes for HAE 229 and HAE 131 are from Kurk et al. (2004a). Spiderweb extended narrow Hα flux has AGN contribution subtracted (Nesvadba et al. 2006). ^cEstimated star-forming total IR luminosity L_SFIR(10¹² L_☉) = 17 × L(PAH7.7) (Elbaz et al. 2011). ^dPAH-based star formation rate SFR(PAH, M_☉ yr⁻¹) = L(PAH7.7)/4 × 10⁸ (see the text). ^eSFR(Hα, M_☉ yr⁻¹) = 3.1 × 10⁻⁸L(Hα) (Kennicutt et al. 1994).

Download table as:  ASCIITypeset image

We estimate a similar star formation rate of ∼1200 M_☉ yr⁻¹ from the extended Hα luminosity of the Spiderweb (Nesvadba et al. 2006). Here, we use SFR(Hα, M_☉ yr⁻¹) = 3.1 × 10⁻⁸L(Hα) (Kennicutt et al. 1994) and assume no extinction. The extended Hα luminosity was measured after subtracting the point-like AGN contribution (Nesvadba et al. 2006), but the AGN might still contribute to the extended Hα emission via photoionization or radio-jet driven shocks.

A much lower star formation rate of 57 ± 8 M_☉ yr⁻¹ is estimated from the diffuse 32 kpc radius UV halo (Hatch et al. 2008). This rate increases to 142 M_☉ yr⁻¹, correcting for screen extinction (E(B − V) ∼ 0.1 from rest UV color), and 325 M_☉ yr⁻¹ if all of the diffuse and compact UV sources are included in the estimate (Hatch et al. 2008). This is still far below the PAH-based star formation rate, indicating that the assumption of a uniform dust screen is too simplistic. Instead, much of the star formation in the Spiderweb must be highly obscured (>2.2 mag) at rest-UV wavelengths.

The Spiderweb is the brightest and most massive galaxy in the PKS 1138−26 cluster, with a stellar mass of 1.1 × 10¹² M_☉ estimated from the spectral energy distribution (SED) of the central radio galaxy (Hatch et al. 2009; Seymour et al. 2007). Adding the masses of the satellite galaxies estimated by Hatch et al. (2009) increases the total stellar mass to 1.2 × 10¹² M_☉ within our 7'' (60 kpc) radius spectral extraction aperture. The stellar mass doubling timescale at the PAH-inferred star formation rate is therefore ∼1.1 Gyr. It appears that the Spiderweb is in the process of building the bulk of its stellar mass at the observed epoch. The Spiderweb may gain additional stellar mass (a factor of 1.1–2.2) by accreting known satellite galaxies within a 150 kpc radius (Hatch et al. 2009).

We measure an H₂ 0–0 S(3) luminosity of 1.4 × 10⁴⁴ erg s⁻¹ (3.7 × 10¹⁰ L_☉). We note that the H₂ 0–0 S(2), S(1), and S(0) lines may be similarly luminous, but are redshifted beyond the longest wavelength detectable with the IRS. Taken at face value, the H₂ 0–0 S(3) and S(5) line fluxes fit by PAHFIT would indicate a temperature of ∼920 K and an H₂ mass of ∼1 × 10⁷ M_☉, assuming a single-temperature gas in equilibrium. However, the H₂ 0–0 S(5) line is blended with [Ar ii]. We estimate a more conservative upper limit on the H₂ temperature and lower limit on the H₂ mass by assuming that the [Ar ii] + H₂ 0–0 S(5) emission line fit gives an upper limit on the H₂ 0–0 S(5) line flux. This yields an H₂ temperature (for gas contributing to the S(3) and S(5) lines) of 300 K < T < 1200 K and a warm molecular hydrogen mass of >9 × 10⁶–1 × 10⁹ M_☉ (Figure 3).

Zoom In Zoom Out Reset image size

In the same temperature range, 3C radio galaxies at z < 0.22 (Ogle et al. 2010) have warm H₂ masses more than eight times smaller than the Spiderweb (Figure 3). In the lowest temperature range probed by Spitzer, their H₂ 0–0 S(0)–S(2) lines indicate warm (T = 100–300 K) H₂ masses >8 × 10⁶–2 × 10¹⁰ M_☉. It is likely that the Spiderweb has a much larger total warm H₂ mass at T = 100–300 K than what we have detected with the 0–0 S(3) and S(5) lines at T > 300 K. If the ratio of warm to total H₂ mass is similar to that of low-redshift radio galaxies, then the extremely luminous H₂ emission from the Spiderweb could be explained by an extremely large molecular gas mass of ∼2 × 10¹¹ M_☉. This would give a reasonable H₂ to stellar mass ratio of ∼0.2. A similarly large molecular gas mass of 1.4 × 10¹¹ M_☉ has been inferred from CO (1–0) emission from the z = 2.8 lensed quasar SMM J041327.2+102743 (Riechers et al. 2011).

We measure a lower limit (from S(3) only) for the ratio of the summed first four H₂ 0–0 pure rotational lines to 7.7 μm PAH of L(H₂ 0–0 S(0)–S(3))/L(PAH7.7) > 0.08, more than a factor of two greater than the empirical and theoretical ratio of <0.04 for UV heating in a PDR. This makes the Spiderweb Galaxy the most luminous known radio MOHEG (Ogle et al. 2007, 2010; Guillard et al. 2012), a class of galaxies where the molecular gas is heated by dissipation of mechanical energy. By analogy to low-redshift radio MOHEGs, the presence of a large mass of warm molecular hydrogen indicates that a significant fraction of the molecular gas content of the Spiderweb has been shock-heated by the radio jet, in turn suppressing star formation (Ogle et al. 2010; Nesvadba et al. 2010). Nesvadba et al. (2006) estimate a radio jet mechanical power, based on the radio luminosity of PKS 1138−26, of ∼10⁴⁶ erg s⁻¹. Only ∼1% of this would be enough to power the observed H₂ 0–0 S(3) line. On the other hand, the absorption-corrected (intrinsic N_H = 2.6 × 10²² cm⁻²) 2–10 keV X-ray luminosity of the AGN is only 4 × 10⁴⁵ erg s⁻¹ (Carilli et al. 2002), which is insufficient to power the H₂ emission. The observed ratio L(H₂ 0–0 S(0)–S(3))/L_X (2–10 keV) > 0.035 is more than five times greater than the theoretical maximum ratio for AGN X-ray heating of <0.007 (Ogle et al. 2010; Guillard et al. 2012).

From our Spitzer spectrum, the monochromatic mid-IR luminosity of the dust-obscured Spiderweb AGN is νL_ν(11 μm) = 1.9 × 10⁴⁶ erg s⁻¹. Using an IR/UV flux conversion factor of f_UV = 0.36 and bolometric IR correction factor of f_bol = 1.7 for the AGN dusty torus, both derived from z < 1 radio galaxies and quasars (Ogle et al. 2006), we estimate an AGN bolometric luminosity of L_bol = (f_bol/f_UV)νL_ν(11 μm) = 9.5 × 10⁴⁶ erg s⁻¹, ∼3 times the total IR luminosity from star formation. This is a factor of ∼2 larger than the bolometric luminosity estimate of Nesvadba et al. (2010), which is based on the absorption-corrected AGN X-ray luminosity (Carilli et al. 2002). Given the uncertainties in the IR and X-ray bolometric corrections, we consider the two estimates to be in reasonable agreement. Using our mid-IR based estimate of the hidden quasar luminosity and an assumed mass–energy conversion efficiency of 0.1, we estimate a mass accretion rate of ∼16 M_☉ yr⁻¹ for the supermassive black hole that powers the Spiderweb AGN. This is ∼1% of the PAH-based instantaneous star formation rate integrated over the entire galaxy.

4.2. HAE 229 and HAE 131

We have found PAH emission indicating star formation rates of 500–1100 M_☉ yr⁻¹ in three galaxies in the PKS 1138−26 radio galaxy protocluster, including the Spiderweb central radio galaxy, HAE 229, and HAE 131. The corresponding stellar mass doubling timescales in the Spiderweb and HAE 229 are 1.1 Gyr and 600 Myr, respectively. This indicates that these protocluster galaxies are undergoing a period of rapid growth and that the PKS 1138−26 protocluster is a massive galaxy construction zone. Similarly high star formation rates have been observed in other high-redshift radio galaxies (e.g., Seymour et al. 2008), in contrast to low-redshift (z < 0.2) radio galaxies where star formation rates are generally lower (<5 M_☉ yr⁻¹), in spite of the presence of up to 10¹⁰ M_☉ of molecular gas in the central few kpc (Shi et al. 2007; Ogle et al. 2010; Seymour et al. 2011). Low-redshift radio jets appear to be more effective at suppressing star formation in their host galaxies. However, the dynamical state and gas physical properties of low-redshift radio galaxies are also much more evolved, as indicated by the presence of virialized hot gas in their halos and surrounding galaxy clusters.

Star formation rates estimated from the 7.7 μm PAH feature are a factor of 14–40 greater than estimates based on Hα, for HAE 229 and HAE 131. This indicates that star formation is heavily obscured (by >2.9–4.1 mag) at Hα and shorter wavelengths. This is in contrast to the Spiderweb, where the PAH-based and Hα-based star formation rate estimates are comparable. This may suggest a different mode of more-extended, less obscured star formation in the Spiderweb. This in turn may be consistent with the observed extended UV continuum emission (Hatch et al. 2008) if the UV extinction is more complicated than a simple obscuring screen.

The extended mode of star formation in the Spiderweb could possibly be a result of negative feedback quenching star formation in the nucleus, coupled with accretion of many satellite galaxies enhancing star formation in the outer regions of the galaxy. High luminosity H₂ 0–0 S(3) emission from the Spiderweb indicates a large mass of warm (T > 300 K) molecular hydrogen, most likely heated by radio jet feedback from the AGN, which may have already quenched nuclear star formation and may eventually quench quasar activity. In other words, the high star formation rates observed in high-redshift radio galaxies may be attributed to the accretion of gas and gas-rich satellite galaxies rather than nuclear star formation. At low redshift, the gas has been virialized and heated to 10⁷ K and the satellites have been incorporated into the central radio galaxy, such that there is no longer a steady gas supply for star formation. The remaining gas that is able to cool and make its way into the central kpc-scale disk is unable to form stars efficiently, because it is either continuously or episodically heated by the radio jet.

Herschel FIR-sub-mm photometry will determine the total IR luminosities of the galaxies in the PKS 1138−26 protocluster, yielding more accurate star formation rates and dust mass estimates. Herschel PACS spectroscopy may be sensitive enough to detect the H₂ S(0) line and thereby measure the total mass of warm molecular gas in the Spiderweb. The extreme luminosity of H₂ rotational line emission in the Spiderweb provides a strong incentive for launching a cooled, large aperture FIR space telescope to detect these lines and use them to measure the impact of kinetic energy dissipation on molecular gas and star formation in even higher-redshift galaxies (Appleton et al. 2009).

This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under NASA contract 1407. This work benefited greatly from the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. N.S. is a recipient of an Australian Research Council Future Fellowship.