DYNAMICAL STRUCTURE OF THE MOLECULAR INTERSTELLAR MEDIUM IN AN EXTREMELY BRIGHT, MULTIPLY LENSED z ≃ 3 SUBMILLIMETER GALAXY DISCOVERED WITH HERSCHEL

We observed the CO(J = 5 → 4) transition line (ν_rest = 576.2679305 GHz, redshifted to 145.633 GHz, or 2.06 mm) toward HLSW-01 (z = 2.957), using the PdBI. We used the WideX correlator with a total bandwidth of 3.6 GHz (∼7400 km s⁻¹, at 1.95 MHz resolution) to cover the CO(J = 5 → 4) line and the underlying 2 mm (rest-frame 520 μm) continuum emission. Observations were carried out under good 2 mm weather conditions for two tracks in 5D configuration on 2010 June 14 and 17. This resulted in 1.0 hr of six antenna-equivalent on-source time after discarding unusable visibility data. The nearby (106 distant) source B0954+658 (J0958+655; 0.98 Jy) was observed every 20 minutes for pointing, amplitude, and phase calibration. MWC 349 (1.50 Jy) and 3C273 (10.5 Jy) were observed for flux and bandpass calibration, yielding ∼10% calibration accuracy.

The IRAM GILDAS package was used for data reduction and analysis. All data were mapped using the CLEAN algorithm with "natural" weighting, resulting in a synthesized beam of 46 × 28. The final rms is 1.2 mJy beam⁻¹ over 250 MHz (corresponding to 515 km s⁻¹), 3.4 mJy beam⁻¹ over 31.25 MHz (64 km s⁻¹), and 6.0 mJy beam⁻¹ over 10 MHz (21 km s⁻¹).

We observed the CO(J = 3 → 2) transition line (ν_rest = 345.7959899 GHz, redshifted to 87.388 GHz, or 3.43 mm), using CARMA. A total bandwidth of 3.7 GHz (∼12,700 km s⁻¹, at 5.208 MHz resolution) was used to cover the CO(J = 3 → 2) line and the underlying 3.4 mm (rest-frame 870 μm) continuum emission. Observations were carried out under good 3 mm weather conditions (typically at moderate elevations after transit) for six tracks in D configuration between 2010 September 01 and 17. This resulted in 6.6 hr of 15 antenna-equivalent on-source time after discarding unusable visibility data. The nearby source J0958+655 (0.86 Jy) was observed every 15 minutes for pointing, amplitude, and phase calibration. Fluxes were bootstrapped relative to Mars (based on a brightness temperature model of the planet). Several nearby calibrators (3C 84, 3C 273, 3C 345, J0927+390, J2015+372) were observed for bandpass calibration, yielding ∼15% calibration accuracy.

The MIRIAD package was used for data reduction and analysis. All data were mapped using the CLEAN algorithm with "natural" weighting, resulting in a synthesized beam of 77 × 50. The final rms is 0.48 mJy beam⁻¹ over 226 MHz (corresponding to 777 km s⁻¹), 1.6 mJy beam⁻¹ over 20.8 MHz (71 km s⁻¹), and 2.3 mJy beam⁻¹ over 10.4 MHz (36 km s⁻¹).

We observed the CO(J = 1 → 0) transition line (ν_rest = 115.2712018 GHz, redshifted to 29.131 GHz, or 10.29 mm), using the Zpectrometer analog lag cross-correlation spectrometer (Harris et al. 2007) attached to the GBT Ka band receiver. This yields an instantaneous bandwidth of 10.5 GHz (∼108,000 km s⁻¹) at a beam size of 27''–19'' (25.6–36.1 GHz). Observations were carried out under good 1 cm weather conditions on 2010 October 2. This resulted in 1.5 hr on-source time after discarding unusable data. Subreflector beam switching was used every 10 s to observe the source alternately with the receiver's two beams, with the off-source beam monitoring the sky background in parallel. A nearby second, fainter source was observed with the same pattern, switching between targets with 8 minute cycles. Residual structure from optical beam imbalance in the "source–sky" difference spectra of each target was then removed by differencing both sources. A spectral ripple with well-defined period generated in the receiver was removed with narrowband Fourier filtering. This strategy yields a flat baseline (offset from zero flux by to the difference of the two faint source continua) without standard polynomial baseline removal. The nearby quasar 1143+4931 was observed every ∼1 hr to monitor telescope pointing and gain stability. Passband gains and absolute fluxes were determined from spectra of 3C286 (2.04 Jy at 32 GHz; Ott et al. 1994), yielding 15% calibration accuracy.³⁰

The lag data were processed using the Zpectrometer's data reduction pipeline and standard recipes, yielding a frequency-calibrated spectrum at 8 MHz resolution (see Harris et al. 2010 for details). The instrumental spectral response is nearly a sinc function with an FWHM of 20 MHz, i.e., individual 8 MHz spectral channels are not statistically independent. However, the linewidth correction for the instrumental response for spectral lines with intrinsic Gaussian FWHM of >30 MHz (∼300 km s⁻¹) is minor.

We have detected spatially resolved CO(J = 3 → 2) and CO(J = 5 → 4) (Figure 1) and unresolved CO(J = 1 → 0) line emission toward HLSW-01 at high significance. From Gaussian fitting to the (spatially integrated) line profiles (Figure 2), we obtain CO(J = 1 → 0), CO(J = 3 → 2), and CO(J = 5 → 4) line peak strengths of 3.09 ± 0.37, 26.4 ± 1.5, and 64.0 ± 4.4 mJy at a median FWHM of 348 ± 18 km s⁻¹ (350 ± 55, 350 ± 25, and 346 ± 29 km s⁻¹ for the CO J = 1 → 0, 3 → 2, and 5 → 4 lines). We marginally detect 3.43 mm continuum emission under the CO(J = 3 → 2) line at 0.61 ± 0.19 mJy strength, and do not detect 2.06 mm continuum emission down to a 2σ limit of 1.4 mJy. By combining all lines, we find a median redshift of z = 2.9574 ± 0.0001. The line parameters correspond to velocity-integrated emission line strengths of 1.14 ± 0.11, 9.74 ± 0.49, and 23.6 ± 1.4 Jy km s⁻¹, and line luminosities of L'_CO(1–0) = (4.17 ± 0.41), L'_CO(3–2) = (3.96 ± 0.20), and L'_CO(5–4) = (3.45 ± 0.20) × 10¹⁰ (μ_L/10.9)⁻¹ K km s⁻¹ pc², respectively. This corresponds to line brightness temperature ratios of r₃₁ = 0.95 ± 0.10, r₅₃ = 0.87 ± 0.06, and r₅₁ = 0.83 ± 0.09. This suggests that at least the CO(J = 5 → 4) line is not thermalized (i.e., not in radiative equilibrium with the lower-J lines; r < 1; see also S11). We do not find evidence for a substantial low-excitation gas component in HLSW-01 (i.e., gas with r₃₁ ≪ 1), contrary to what is found in some other z > 2 SMGs (e.g., Carilli et al. 2010; Riechers et al. 2010; Harris et al. 2010; Ivison et al. 2010a). We estimate the total molecular gas mass assuming a conversion factor of α_CO = 0.8 M_☉ (K km s⁻¹ pc²)⁻¹ from L'_CO(1–0) to M_gas, as appropriate for ultra-luminous infrared galaxies (ULIRGs; Downes & Solomon 1998) and commonly used for SMGs. This yields M_gas = 3.3×10¹⁰ (α_CO/0.8) (μ_L/10.9)⁻¹ M_☉.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Several lines of dense molecular gas tracers fall within the spectral range covered by the CO(J = 3 → 2) and CO(J = 5 → 4) observations. We obtain 3σ upper limits of <1.6 Jy km s⁻¹ on the integrated HCN(J = 4 → 3), CS(J = 7 → 6), SiO(J = 8 → 7), and HC₃N(J = 39  →  38) emission line strengths, and <3.2 Jy km s⁻¹ for HC₃N(J = 63  →  62) and HC₃N(J = 64  →  63) (averaged over 348 km s⁻¹). This corresponds to line luminosity limits of <0.61, <0.66, <0.64, <0.61, <0.48, and <0.46 × 10¹⁰ (μ_L/10.9)⁻¹ K km s⁻¹ pc² for the HCN, CS, SiO, HC₃N(J = 39 → 38), HC₃N(J = 63 → 62), and HC₃N(J = 64 → 63) lines, respectively. This also corresponds to line brightness temperature ratio limits of <0.15, <0.16, <0.15, <0.15, <0.11, and <0.11 relative to CO(J = 1  →  0). Nearby starburst galaxies and ULIRGs like NGC 253, NGC 6240, and Arp 220 typically have HCN J = 4 → 3/J = 1 → 0 line ratios of 0.45–0.8 (Knudsen et al. 2007; Greve et al. 2009). If this ratio is similar in HLSW-01, it suggests an HCN/CO ratio of <0.18–0.33 in the respective J = 1 → 0 transitions, consistent with what is found in nearby ULIRGs and other high-z gas-rich galaxies (median ratios of 0.17; Gao & Solomon 2004a, 2004b; Gao et al. 2007; Riechers et al. 2007).

We spatially resolve the (lensed) molecular gas reservoir in HLSW-01 in the CO(J = 3  →  2) and CO(J = 5  →  4) emission lines. The characteristic effective resolution provided by the lensing effect is ∼22 (i.e., regions of ∼8.5 kpc radius) in CO(J = 5  →  4), but substantially higher in the brighter images and close to critical lines (G11). Overlays of the CO emission with 890 μm (rest-frame 225 μm) continuum emission are shown in Figure 3. The rest-frame far-infrared (FIR) continuum emission dominantly traces the regions of most active star formation in this system (star formation rate: 2460 ± 160  (μ_L/10.9)⁻¹ M_☉ yr⁻¹; C11). The CO(J = 5 → 4) emission shows discernible peaks at the positions of lens images A, B, and D, with an extension of the emission associated with image B toward image C. Accounting for beam convolution, there is no evidence for differences in the spatial distribution and/or line shapes of the CO(J = 3 → 2) and CO(J = 5 → 4) emission, indicating that both lines trace the same material. The small offsets between the CO and FIR peaks (typically <10% of the relative beam sizes) are a combination of beam convolution (given the different beam sizes and orientations) and velocity-averaged dynamical structure in the integrated molecular line maps. The molecular gas reservoir may extend beyond the FIR continuum peaks, but maps at higher spatial resolution and sensitivity are required to investigate this in more detail. The overall striking match between the gas and (FIR) dust emission suggests that, indeed, the bulk of the FIR light is associated with the gas-rich, star-forming regions.

Zoom In Zoom Out Reset image size

An overlay of the CO(J = 5  →  4) emission with 2.2 μm (rest-frame 560 nm; smoothed with a 01 Gaussian kernel) continuum emission is shown in Figure 4. The CO emission is clearly associated with the faint optical lensed images of the SMG, which appear to consist of multiple clumps. However, the brightest CO peaks are slightly offset from the peaks of the rest-frame optical emission. This offset translates to ∼2.4 kpc in the source plane after correcting for gravitational magnification (G11). This may indicate that the most gas-rich regions are optically obscured and/or the presence of multiple galaxy components, as commonly observed in SMGs. Also, the FIR-brightest lens image A is the optically faintest image of the galaxy, which lends further support to this picture.

Zoom In Zoom Out Reset image size

A first moment map of the CO(J = 5  →  4) emission is shown in Figure 5 (clipped below 2σ in each velocity channel, smoothed before detection). There is a clear velocity gradient between images A/B and C/D. The peak of image D is blueshifted by up to ∼380 km s⁻¹ (relative to zero velocity in Figure 2). The peak of image C is blended with image B, but is consistent with being redshifted by up to ∼150 km s⁻¹. Images A and B peak on the line center. Image A shows a smooth north–south velocity gradient from red to blue, and image B shows a gradient in the opposite direction. This gradient is also seen in CO(J = 3 → 2) emission, indicating that these are likely dynamically resolved lensed images of the same region in the galaxy. Images C and D are consistent with being unresolved, (in their peaks) kinematically distinct regions in the same galaxy or a companion system. The complexity of the velocity structure is unlikely to be due to distortion of the velocity field by gravitational lensing alone, but would be consistent with the dynamical structure of a major, "wet" (i.e., gas-rich) merger in progress (e.g., Tacconi et al. 2008; Engel et al. 2010).

Zoom In Zoom Out Reset image size

Based on the velocity-integrated CO(J = 5  →  4) map shown in Figure 1, G11 find a half-flux radius of 1.1 ± 0.5 kpc for the CO reservoir in HLSW-01, after correcting for gravitational magnification. However, the structure seen in the CO velocity map indicates that this size estimate may be a lower limit. Thus, different parts of the CO emission may be differentially magnified. However, the lens model by G11 suggests that differential lensing effects are only moderate (≲25%), even out to a radius of 10 kpc (their Figure 3), and, thus, have a relatively minor effect on the integrated gas properties.

The minimum time for which the starburst in HLSW-01 can be maintained at its current rate is given by the gas depletion timescale³¹τ^0.8_dep = M_gas/SFR ∼ 14 ± 2 Myr, which is short but within the range of values found for "typical" SMGs (e.g., Greve et al. 2005).

The ratio between L_IR (∝SFR) and L'_CO (∝M_gas) can be used as a measure of the star formation efficiency. Using L_IR = (1.43 ± 0.09)×10¹³ (μ_L/10.9)⁻¹ L_☉ (C11), we find a ratio of ∼340 ± 40, comparable to what is found in "typical" SMGs (e.g., Tacconi et al. 2006; Riechers et al. 2010).

HLSW-01 has a dust mass in the range of M_dust = (1.0–5.2)×10⁸ (μ_L/10.9)⁻¹ M_☉ (C11; S11). This corresponds to a gas-to-dust mass ratio of f^0.8_gd = M_gas/M_dust = 60–330 and is consistent with the range of values found in other SMGs (e.g., Michalowski et al. 2010).

HLSW-01 has a stellar mass of M_⋆ = (6.3 ± 3.4)×10¹⁰ (μ_L/10.9)⁻¹ M_☉ (C11), yielding a gas mass fraction of f^0.8_gas = M_gas/M_⋆ = 0.53 ± 0.29, and a baryonic gas mass fraction of f^{g, 0.8}_bary = M_gas/(M_gas+M_⋆)=0.35 ± 0.13. This also yields a specific star formation rate (SSFR, i.e., SFR/M_⋆) of 39 ± 21 Gyr⁻¹, and a stellar mass doubling timescale (M_⋆/SFR) of 26 ± 14 Myr. These values are consistent with what is found for other SMGs as well (e.g., Daddi et al. 2009a; Tacconi et al. 2006, 2008).