AN INTENSELY STAR-FORMING GALAXY AT z ∼ 7 WITH LOW DUST AND METAL CONTENT REVEALED BY DEEP ALMA AND HST OBSERVATIONS

To understand whether obscured star-formation is an important issue and to determine the metallicity of Himiko, a key source at high redshift, we carried out deep ALMA Band 6 observations in 2012 July 15, 18, 28, and 31 with a 16 12 m antenna array under the extended configuration of 36–400 m baseline. The precipitable water vapor ranged from 0.7 to 1.6 mm during the observations. We targeted Himiko's [C ii] rest-frame 1900.54 GHz (157.74 μm), which is redshifted to 250.24 GHz (1.198 mm) at a redshift of z_Lyα = 6.595. Since we expect a brighter dust continuum at a higher frequency in the 1.2 mm regime, we extended our upper sideband (USB) to the high-frequency side. Thus, we targeted the [C ii] line with the lowest spectral window (among four spectral windows) in the lower sideband (LSB) and set the central frequency of the four spectral windows to be 250.24 and 252.11 GHz in LSB, and 265.90 and 267.78 GHz in USB with a bandwidth of 1875 MHz. The two spectral windows and each sideband contiguously cover the frequency ranges. The total on-source integration time was 3.17 hr. We used 3c454.3 and J0423−013 for bandpass calibrators and J0217+017 for a phase calibrator. The absolute flux scale was established by observations of Neptune and Callisto. Our data were reduced with the Common Astronomy Software Applications package. We rebin our data to a resolution of 166 MHz (200 km⁻¹). The FWHM beam size of the final image is 082 × 058 with a position angle of 795. The 1σ noise of the continuum image is σ_cont = 17.4 μJy beam⁻¹ over the total bandwidth of 19.417 GHz, whose 7.5 GHz is sampled. The 1σ noise of the [C ii] line image is σ_line = 83.3 μJy beam⁻¹ at 250.239 GHz over a channel width of 200 km⁻¹.

Further details of the ALMA observations and sensitivities are summarized in Table 1.

Table 1. ALMA Observations and Sensitivities

νcont	νline	σcont	σline	fcont	fline	LFIR	$L_{{\rm [C\,\scriptsize{II}]}}$
(GHz)	(GHz)	(μJy beam−1)	(μJy beam−1)	(μJy)	(μJy)	(1010 L☉)	(107 L☉)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
259.007	250.239	17.4	83.3	<52.1	<250.0	<8.0	<5.4

Notes. Columns: (1) and (2) central frequencies of continuum and [C ii] line observations that correspond to 1.16 and 1.20 mm, respectively. (3) and (4) 1σ sensitivities for continuum and [C ii] line in a unit of μJy beam⁻¹. The continuum sensitivity given in the total bandwidth for the continuum measurement is 19.417 GHz or 86.894 μm, which is a sum of four spectral windows (see text). The line sensitivity is defined with a channel width of 200 km⁻¹. (5) and (6) 3σ upper limits of continuum and [C ii] line in a unit of μJy. (7) and (8) 3σ upper limits of far-infrared continuum luminosity (8–1000 μm) and [C ii] line luminosity in a unit of 10¹⁰ and 10⁷ solar luminosities, respectively. We estimate 3σ upper limits of far-infrared continuum luminosities at 40–500 μm and 42.5–122.5 μm to be <7.36 × 10¹⁰ and <6.09 × 10¹⁰ L_☉, respectively. These far-infrared luminosities are estimated with the assumptions of the graybody, β_d = 1.5, and a dust temperature of T_d = 40 K.

Download table as:  ASCIITypeset image

We averaged fluxes over the two spectral windows of LSB (249.30–253.05 GHz or 1.203–1.185 mm) and USB (264.96–268.71 GHz or 1.131–1.116 mm) in the range of frequency free from the [C ii] line. Figure 1 presents the resulting ALMA continuum data at the 259.01 GHz frequency (or 1.167 mm in wavelength) with a 1σ sensitivity of 17.4 μJy beam⁻¹. There is a ∼2σ flux peak in the beam size at the position of Himiko. However, there are a series of negative pixels nearby that correspond to the 2σ–3σ level per beam. We conclude therefore that Himiko remains undetected in the 1.2 mm continuum with a 3σ upper limit of <52.1 μJy beam⁻¹. We note that this sensitivity is two and one order(s) of magnitudes better than those previously obtained by deep SCUBA/SHADES and IRAM/PdBI observations (Ouchi et al. 2009b; Walter et al. 2012). This clearly indicates that Himiko has very weak millimeter emission. Table 1 summarizes the flux upper limits for the continuum and [C ii] line derived from our ALMA data.

Zoom In Zoom Out Reset image size

Figure 2 shows the ALMA [C ii] velocity channel maps for Himiko. We searched for a signal of [C ii] over 600 km⁻¹ around the frequency corresponding to z_Lyα = 6.595. In Figure 2, there are noise peaks barely reaching the 3σ level in 0 km s⁻¹ and −200 km s⁻¹ that are slightly north and south of Himiko's optical center position, respectively. However, neither of these sources is a reliable counterpart of Himiko. Although the ALMA data reach a 1σ noise of 83.3 μJy beam⁻¹, no [C ii] line is detected. The corresponding 3σ upper limit for [C ii] is 250.0 μJy in a channel width of 200 km s⁻¹. The associated luminosity limit is <5.4 × 10⁷ L_☉.

Zoom In Zoom Out Reset image size

The primary goal of the associated HST observations of Himiko is related to the morphological nature of this remarkable source. We carried out deep HST/WFC3-IR broad-band (J₁₂₅ and H₁₆₀)¹² and medium-band (F098M) observations for Himiko. The two broad-band filters J₁₂₅ and H₁₆₀ measure the rest-frame UV continuum fluxes and are free from contamination from Lyα emission, thus maximizing the amount of information on the stellar content for SED fitting (Section 3.3). The intermediate-band filter of F098M fortuitously includes the spectroscopically confirmed Lyα line of Himiko at 9233 Å (Ouchi et al. 2009b) with a system throughput of 40%, which is close to the peak throughput of this filter (∼45%). Thus, the F098M image is ideal for mapping the distribution of Lyα emitting gas.

Our observations were conducted in 2010 September 9, 12, 15–16, 18, and 26 with an orientation of 275°. Some observations were partially lost because HST went into "safe mode" on 2010 September 9, 22:30 during the execution of one visit. The total integration times for usable imaging data are 15670.5, 13245.5, 18064.6 s for F098M, J₁₂₅, and H₁₆₀, respectively. The various WFC3 images were reduced with the WFC3 and MULTIDRIZZLE packages on PyRAF. To optimize our analyses, in the multidrizzle processing we chose a final$\_$pixfrac =0.5 and pixel scale of 005132. We degraded images of F098M and J₁₂₅ to match the point spread functions (PSFs) of these images with that of H₁₆₀, which has the largest size among the HST images. We ensured that the final WFC3 images have a matched PSF size of 019 FWHM.

Figure 3 presents a color composite HST UV-continuum image of Himiko as well as a large ionized Lyα cloud identified by the Subaru observations (Ouchi et al. 2009b). This image reveals that the system comprises three bright clumps of starlight surrounded by a vast Lyα nebula ≳ 17 kpc across. We denote the three clumps as A, B, and C. Figure 4 shows the HST, Subaru, and Spitzer images separately. The F098M image in Figure 4 detects only marginal extended Lyα emission, because of the shallower surface brightness limit of the 2.4 m HST compared to the 8 m Subaru telescope. Nevertheless, we have found a possible bright extended component at position D in Figure 4. We perform 04 diameter aperture photometry for clumps A–C and location D as well as 2'' diameter aperture photometry, which we adopt as the total magnitude of the system. Tables 2 and 3 summarize the photometric properties. It should be noted that Himiko is not only identified as an LAE, but also would be regarded as an LBG or "dropout" galaxy. Using the optical photometry of Ouchi et al. (2009b; see also Table 3), we find no blue continuum fluxes for filters B through i' up to the relevant detection limits of 28–29 mag. The very red color of i' − z' > 2.1 meets typical dropout selection criteria (e.g., Bouwens et al. 2011). Because the z'-band photometry includes the Lyα emission line and an Lyα-continuum break, we can also estimate the continuum break color using our HST photometry of J₁₂₅ and H₁₆₀ and the optical i-band photometry. Assuming the continuum spectrum is flat (f_ν = const.), we obtain a continuum break color i' − J₁₂₅ > 3.0 or i' − H₁₆₀ > 3.0, further supporting Himiko's classification as an LBG. Importantly, these classifications also apply to clumps A–C, ruling out the possibility that some could be foreground sources.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 2. Properties of the Subcomponents

Component	x(pix)	y(pix)	NB	J125	H160	β	L(Lyα)	EW0	SFR(UV)	SFR(Lyα)
			(mag)	(mag)	(mag)		(1042 erg s−1)	(Å)	(M☉ yr−1)	(M☉ yr−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
Totala	2815.4	2790.4	23.55 ± 0.05	24.99 ± 0.08	24.99 ± 0.10	−2.00 ± 0.57	38.4 ± 1.5	$78^{+8}_{-6}$	30 ± 2	35 ± 1
(F098M)	⋅⋅⋅	⋅⋅⋅	(24.84 ± 0.08)	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	(30.5 ± 15.6)	($61^{+28}_{-23}$)	⋅⋅⋅	(28 ± 14)
A(clump)	2803.0	2789.0	26.36 ± 0.04	26.54 ± 0.04	26.73 ± 0.06	−2.84 ± 0.32	8.1 ± 1.9	$68^{+14}_{-13}$	7 ± 0	7 ± 2
B(clump)	2816.5	2790.5	27.19 ± 0.09	27.03 ± 0.07	27.04 ± 0.08	−2.04 ± 0.47	0.2 ± 1.9	$3^{20}_{-18}$	5 ± 0	0 ± 2
C(clump)	2826.3	2791.9	26.57 ± 0.05	26.43 ± 0.04	26.48 ± 0.05	−2.22 ± 0.28	0.8 ± 1.9	$6^{+12}_{-10}$	8 ± 0	1 ± 2
D	2821	2801	28.47 ± 0.29	>28.86	>28.70	⋅⋅⋅	>1.7	>123	<1	>2

Notes. Columns: (1) Name of the component. (2) and (3) Positions in pixels. (4)–(6) Aperture magnitudes in NB, J₁₂₅, and H₁₆₀. NB indicates a narrow/intermediate band magnitude determined with F098M for all lines except in Total(NB921). The upper limit corresponds to a 3σ limit. (7) UV-continuum slope. (8) Lyα luminosity. (9) Rest-frame apparent equivalent width of Lyα emission line in Å. (10) Star-formation rate estimated from the UV magnitude. (11) Star-formation rate estimated from the UV magnitude. ^aThe NB magnitude corresponds to NB921. The quantities of Columns 8, 9, and 11 are estimated from NB921 photometry.

Download table as:  ASCIITypeset image

Table 3. Total Magnitudes and Fluxes of Himiko

Band	Mag/Flux(Total)a
Bb	>28.7
Vb	>28.2
Rb	>28.1
i'b	>28.0
z'b	25.86 ± 0.20
NB921b	23.55 ± 0.05
F098M	24.84 ± 0.08
J125	24.99 ± 0.08
H160	24.99 ± 0.10
Jc	25.03 ± 0.25
Hc	26.67 ± 2.21
Kc	24.77 ± 0.29
m(3.6 μm)	23.69 ± 0.09
m(4.5 μm)	24.28 ± 0.19
m(5.8 μm)b	>22.0
m(8.0 μm)b	>21.8
m(24 μm)b	>19.8
S(1.2 mm)	<52.1 μJyd
$f({\rm [C\,\scriptsize{II}]})$	<250.0 μJyd

Notes. ^aIn units of AB magnitudes, if not specified. The upper limits are 2σ and 3σ magnitudes in BVRi' and 5.8–24 μm bands, respectively. ^bMeasurements obtained in Ouchi et al. (2009b). The continuum magnitudes from B through z' are defined with a 2'' diameter aperture photometry that matches to the photometry of the total magnitudes of NIR bands. ^cTotal magnitudes from UKIDSS/UDS DR8 data that are estimated with a 2'' diameter aperture photometry and the aperture correction in the same manner as Ono et al. (2010b). ^dThree sigma upper limits derived with our ALMA data.

Download table as:  ASCIITypeset image

The UV continuum magnitudes of clumps A–C range from 26.4 to 27.0 mag in J₁₂₅ and H₁₆₀. Each clump has a UV luminosity corresponding to the characteristic luminosity L* of a z ∼ 7 galaxy, m = 26.8 mag (Ouchi et al. 2009a; Bouwens et al. 2011). Moreover, the variation in luminosity across the components is small; there is no single dominant point source in this system, confirming earlier deductions that the system does not contain an active nucleus.

The F098M image shows that Lyα emission is not uniformly distributed across the three clumps. Clump A shows intense Lyα emission with a rest-frame equivalent width (EW₀) of $68^{+14}_{-13}$ Å, placing it in the category of a LAE, whereas clumps B and C have emission more typical of LBGs with a rest-frame Lyα equivalent width (EW₀) less than 20 Å given the measurement uncertainties.

In summary, the HST and Subaru data indicate that Himiko is a triple L* galaxy system comprising one LAE and two LBGs surrounded by an extensive 17 kpc diffuse Lyα halo. Importantly, from the above morphological studies, we can easily eliminate the possibility that Himiko is gravitationally lensed by a foreground concentration. Ouchi et al. (2009a) have already made a strong case against lensing given that Keck spectroscopy revealed a velocity gradient of 60 km s⁻¹ across the system. We can further reject this supposition given that there are clear asymmetries in the outermost images (one has strong Lyα emission and the other does not).

Although Spitzer cannot match the resolution of the above morphological data, we use the very deep Spitzer/IRAC SEDS data reaching 26 mag at the 3σ level (Ashby et al. 2013) to investigate the counterpart of the overall Himiko system at the 3.6 μm and 4.5 μm bands. To improve the relative astrometric accuracy, we have realigned the SEDS images to the HST images, referring to bright stellar objects commonly detected in the Spitzer and HST images. The relative astrometric errors are estimated to be ≃ 01 rms. We obtain total magnitudes for the Spitzer/IRAC images from a 3'' diameter aperture and use an aperture correction given in Yan et al. (2005). The total magnitudes are 23.69 ± 0.09 mag and 24.28 ± 0.19 mag at the 3.6 μm and 4.5 μm bands, respectively. Because the Spitzer/IRAC 5.8 μm and 8.0 μm and Spitzer/MIPS 24 μm band images are not available in the SEDS data set, we use the relatively shallow Spitzer/SpUDS (PI: J. Dunlop, 2007) photometry measurements presented in Ouchi et al. (2009b). Table 3 summarizes these total magnitudes and fluxes.

We investigate obscured star-formation and dust properties of Himiko from its FIR SED using the newly available ALMA 1.2 mm continuum data. The SED from the optical to millimeter wavelengths is shown in Figure 5, together with that of various local starburst templates. The figure demonstrates that Himiko's millimeter flux is significantly weaker than that of dusty starbursts in the local universe such as Arp220 and M82, as well as the spiral galaxy NGC 6946; it is more comparable to those of dwarf galaxies of much lower mass. Similarly, Himiko's rest-frame optical flux derived from the Spitzer/IRAC 3.6 and 4.5 μm photometry is significantly weaker than that of dusty starbursts and spiral galaxies. Given its intense rest-frame UV luminosity and moderately high stellar mass, Himiko's dust emission and evolved stellar flux are remarkably weak. Both properties imply a low extinction and a relatively young stellar age (Section 3.3). In this sense, Himiko may be similar to many luminous z ∼ 3 LBGs whose cold-dust continuum emission is also comparable to unreddened local starburst galaxies (Ouchi et al. 1999).

Zoom In Zoom Out Reset image size

We can estimate a FIR luminosity of Himiko from our 1.2 mm continuum limit. Assuming an optically thin graybody of modified blackbody radiation with a dust emissivity power-law spectral index of β_d = 1.5 and a dust temperature of T_d = 40 K (Eales et al. 1989; Klaas et al. 1997), we obtain a 3σ upper limit of L_FIR < 8.0 × 10¹⁰ L_☉ integrated over 8–1000 μm. We also estimate 3σ upper limits of <7.4 × 10¹⁰ and <6.1 × 10¹⁰ L_☉ at 40–500 μm and 42.5–122.5 μm, respectively. Note that these upper limits depend upon the assumed dust temperature and β_d. For T_d = 25 K and T_d = 60 K, the 3σ upper limit luminosities in 8–1000 μm are <2.7 × 10¹⁰ and <3.0 × 10¹¹ L_☉, respectively. Similarly, for β_d = 0 and β_d = 2, the 3σ upper limit luminosities in 8–1000 μm are <3.5 × 10¹⁰ and <1.2 × 10¹¹ L_☉, respectively.

The preceding upper luminosity limits do depend somewhat on dust temperature and spectral index. Based on the Herschel measurements, Lee et al. (2012) find that the average dust temperature is ∼30 K under β_d = 1.5 for relatively high redshift (z ∼ 4) LBGs with a luminosity of L ≳ 2 L* comparable to Himiko. In the local universe, the median dust temperatures are 33 K, 30 K, and 36 K, for E/S0, Sb-Sbc, and infrared bright galaxies, respectively (Sauvage & Thuan 1994; Young et al. 1989). Recent numerical simulations have claimed that LAEs may have a relatively high dust temperature, due to the proximity of dust to star-forming regions. However, even in this case, the maximum temperature reaches only T_d ≃ 40 K (Yajima et al. 2012b). On the other hand, Himiko's dust must be heated to some lower limit by the cosmic microwave background (CMB), whose blackbody temperature scales as $T^{z=0}_{\rm CMB} (1+z)$, where $T^{z=0}_{\rm CMB}$ is the temperature of the present-day CMB, $T^{z=0}_{\rm CMB}=2.73$ K. Assuming local thermal equilibrium between the ISM of Himiko and the CMB at z = 6.595 (da Cunha et al. 2013) yields a lower limit of T_d = 21 K. Thus, it is appropriate to consider a range of T_d ≃ 20–40 K with β_d ≃ 1.5. Because the larger assumed dust temperature T_d = 40 K with β_d = 1.5 provides a weaker upper limit, we adopt a conservative 3σ upper limit of L_FIR < 8.0 × 10¹⁰ L_☉ (8–1000 μm). Tables 1 and 3 present the 3σ luminosity upper limit.

We now turn to estimating the metallicity of the ISM of Himiko using [C ii] emission as a valuable tracer in star-forming regions. Despite our significant integration, no line is seen. Figure 6 (and Table 3) presents the upper limit to the [C ii] luminosity in the context of the correlation with the SFR (de Looze et al. 2011). In the case of Himiko, the SFR was obtained by SED fitting of the rest-frame UV to optical data, including a correction for dust extinction (Section 3.3). Himiko clearly departs significantly from the scaling relation; the deficit amounts to a factor of ≃ times 30. Given the SFRs of de Looze et al. (2011) for local galaxies are derived in a similar manner to that for Himiko, including contributions from dust-free and dusty starbursts with Galaxy Evolution Explorer's UV and Spitzer's infrared fluxes, respectively, it is difficult to believe that this deficiency is due to some form of bias arising from comparing different populations.

Zoom In Zoom Out Reset image size

Graciá-Carpio et al. (2011) and Diaz-Santos et al. (2013) present $L_{{\rm [C\,\scriptsize{II}]}}/L_{\rm FIR}$ ratios for local starbursts that depend on L_FIR and the FIR and mid-IR surface brightnesses. As a result, Diaz-Santos et al. (2013) argue that $L_{{\rm [C\,\scriptsize{II}]}}$ may not represent a particularly reliable indicator of SFR. However, FIR and mid-IR luminosities only trace dusty starbursts and typically exclude dust-free measures such as the UV luminosity. Because galaxies with fainter FIR/mid-IR luminosities have a larger ratio of $L_{{\rm [C\,\scriptsize{II}]}}/L_{\rm FIR}$ in the datasets probed by Graciá-Carpio et al. (2011) and Diaz-Santos et al. (2013), more dust-free star-formation is expected in such systems. In this sense, the analysis of de Looze et al. (2011) is perhaps more relevant as a prediction of what to expect for Himiko. Nonetheless, given the importance of using $L_{{\rm [C\,\scriptsize{II}]}}$ as a possible tracer and the discussion that follows below, independent studies of $L_{{\rm [C\,\scriptsize{II}]}}$ as a function of UV luminosity and L_FIR would be desirable. Figure 6 also shows that HFLS3 at z = 6.3 (Riechers et al. 2013) follows the local scaling relation. However, it should be noted that the SFR of HFLS3 is derived from the FIR luminosity, and thus any contribution from dust-free star-formation would be missing. In this sense, the SFR may be a lower limit, in which case HFLS3 may also depart somewhat from the local relation.

The absence of [C ii] emission in Himiko is perhaps the most surprising result from our ALMA campaign. The emission line is often assumed to be the most robust far-IR tracer of star formation in high redshift galaxies, such that it may replace optical lines such as Lyα in securing spectroscopic redshifts in the reionization era. Our failure to detect this line in one of the most spectacular z ≃ 7 galaxies has significant implications, which we discuss in Section 4.

Although we derived some constraints on the integrated properties of Himiko in our earlier work (Ouchi et al. 2009b), we did not obtain an E(B − V) estimate and only the lower limit of SFR with E(B − V) ⩾ 0 was obtained, due to the large uncertainties of photometric measurements. We now refine these estimates based on our significantly deeper HST and Spitzer data. Our near-IR SED is taken using total magnitudes from the HST images (Section 2.2), the Spitzer/IRAC SEDS images (Section 2.3), and JHK DR8 data from the UKIDSS/UDS survey. We tabulate these total magnitudes in Table 3 including ground-based optical data previously given in Ouchi et al. (2009b).

We present the SED of Himiko in Figure 7 and undertake the χ² fitting of a range of stellar synthesis models in the same manner as Ono et al. (2010b), using the stellar synthesis models of Bruzual & Charlot (2003) with the dust attenuation formulation given by Calzetti et al. (2000). We adopt a Salpeter initial mass function (IMF; Salpeter 1955) with lower and upper mass cutoffs of 0.1 and 100 M_☉, respectively. Applying models of constant and exponentially decaying star-formation histories with metallicities ranging from Z = 0.02–1.0 Z_☉, we search for the best-fit model in a parameter space of E(B − V) = 0–1 and age =1–810 Myr (where the latter upper limit corresponds to the cosmic age at z = 6.595). Nebular continuum and line emission, estimated from the ionizing photons from young stars, are optionally included following the metallicity-dependent prescriptions presented in Schaerer & de Barros (2009) and Ono et al. (2010b).

Zoom In Zoom Out Reset image size

For a constant SFR history with no nebular emission and a fixed metallicity of Z = 0.2 Z_☉, we find that our best-fit model has a stellar mass of $M_*=3.0^{+0.4}_{-0.6}\times 10^{10}\, M_\odot$, a stellar age of $3.6^{+0.4}_{-0.8}\times 10^{8}$ yr, a SFR of $98^{+2}_{-2}\, M_\odot$ yr⁻¹, and an extinction of E(B − V) = 0.15 with a reduced χ² of 3.1. This is a significant improvement over our much weaker earlier constraints which did not have the benefit of the HST/WFC3 or Spitzer/SEDS data(Ouchi et al. 2009b). The new infrared data play a critical role in determining the Balmer break, thereby resolving the degeneracy between extinction and age. On the other hand, the fit itself is not very satisfactory. The reduced χ² is large and there is a significant discrepancy at 3.6 μm. Since the 3.6 μm and 4.5 μm bands sample the strong nebular lines of Hβ+[Oiii] and Hα, respectively, at z = 6.595, this encourages us to include nebular emission in our fitting procedure. In fact, in Figure 4, we note that the IRAC 4.5 μm emission shows a positional offset with respect to that at 3.6 μm, suggesting the possibility of contamination by nebular emission.

Adding nebular emission to the stellar SED models given above, the best fit has a more satisfactory reduced χ² of 1.6, and we derive a reduced stellar mass of $M_*=1.5^{+0.2}_{-0.2}\times 10^{10}\, M_\odot$ and a younger stellar age of $1.8^{+0.2}_{-0.2}\times 10^{8}$ yr, but similar values for the SFR of $100^{+2}_{-2}\, M_\odot$ yr⁻¹ and extinction of E(B − V) = 0.15. Table 4 summarizes the results of our SED fitting with the pure stellar and stellar+nebular models. In the stellar+nebular models, we assume that all ionizing photons lead to nebular emission lines corresponding to an escape fraction f_esc = 0. If we allow f_esc to be a free parameter, then following Ono et al. (2012b) we find no change from the model above (i.e., f_esc = 0) and formally establish that f_esc < 0.2.

Table 4. Stellar Population of Himiko

Model	M*	E(B − V)*	Age	SFR	sSFR	χ2/dof
	(M☉)	(mag)	(Myr)	(M☉ yr−1)	(yr−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)
Stellar+nebular	$1.5^{+0.2}_{-0.2} \times 10^{10}$	0.15a	$182^{+22}_{-20}$	100 ± 2	6.7 ± 0.9 × 10−9	1.55
Pure stellar	$3.0^{+0.4}_{-0.6} \times 10^{10}$	0.15a	$363^{+44}_{-75}$	98 ± 2	3.3 ± 0.5 × 10−9	3.13

Notes. Columns: (1) models with or without nebular emission. (2) Stellar mass. (3) Color excess of dust extinction for stellar continua. (4) Stellar age. (5) Star-formation rate. (6) Specific star-formation rate. (7) Reduced χ². The degree of freedom (dof) is six. ^aThe uncertainty of color excess is smaller than our model-parameter grid of ΔE(B − V) = 0.01.

Download table as:  ASCIITypeset image

Labbé et al. (2010) and Finkelstein et al. (2010) have suggested from their pure stellar models that HST z = 7–8 dropout galaxies have modest stellar masses (10⁸–10⁹ M_☉) and are quite young (30–300 Myr), in contrast with Himiko's stellar mass (M_* ≃ 3.0 × 10¹⁰ M_☉) and age (360 Myr) estimated with our pure stellar models. Of course, Himiko is more massive and energetic than typical LBGs seen in the small area of the Hubble Ultra Deep Field. Its most notable feature is its high SFR of ≃ 100 M_☉ yr⁻¹, which is more than an order of magnitude larger than those of the HST LBGs at similar redshifts (≃ 1–10 M_☉ yr⁻¹; Labbé et al. 2010). Himiko's selective extinction, E(B − V) = 0.15, is also larger than that of HST dropouts, more than half of which are consistent with no extinction (Finkelstein et al. 2010). On the other hand, the stellar mass of Himiko is only about 1/10th that of many sub-millimeter galaxies (SMGs) at z ∼ 3 (Chapman et al. 2005). We estimate a specific star-formation rate (sSFR), defined by a ratio of SFR to stellar mass, to be sSFR = 3.3 ± 0.5 × 10⁻⁹ and sSFR = 6.7 ± 0.9 × 10⁻⁹ yr⁻¹, for the pure stellar and stellar+nebular cases, respectively. Even though the stellar masses are very different, Himiko, SMGs, and LBGs at z ∼ 3 share comparable sSFRs ∼10⁻⁹–10⁻⁸ yr⁻¹ (see Figure 12 of Ono et al. 2010a).

The new HST data give us the first reliable measurement of the UV continuum slope for each of the morphological components identified in Figure 4. The UV spectral slope provides a valuable indicator of the combination of dust extinction, metallicity, the upper IMF, and stellar age. We estimated the UV slope, β, from the J₁₂₅ and H₁₆₀ photometry, which samples the continua at the rest-frame wavelengths of ∼1600 and ∼2100 Å, neither of which is contaminated by either Lyα emission or the Lyα-continuum break.

We calculate β via

$$\begin{equation} \beta =-\frac{J_{125} - H_{160}}{2.5 \log \left(\lambda _{\rm c}^1 / \lambda _{\rm c}^2 \right)} - 2, \end{equation} \tag{ 1 }$$

where $\lambda _{\rm c}^1$ and $\lambda _{\rm c}^2$ are the central wavelengths of the J₁₂₅ and H₁₆₀ filters, respectively. The estimates for each component are summarized in Table 2. We obtain β = −2.00 ± 0.57 for the entire system of Himiko, which is comparable to the average UV slope of ≃ L* LBGs, β = −2.09 ± 0.22 (Bouwens et al. 2012, see also Dunlop et al. 2013). Figure 8 shows the UV-to-FIR luminosity ratio, log (L_FIR/L₁₆₀₀), and the UV-continuum slope, β, for the entire system of Himiko, and compares these estimates with the relation of local starbursts (Meurer et al. 1999). Figure 8 indicates that Himiko has log (L_FIR/L₁₆₀₀)-β values comparable with or smaller than those of local dust-poor starbursts. Since the Small Magellanic Cloud (SMC) extinction has a smaller log (L_FIR/L₁₆₀₀) value at a given β (see Figure 10 of Reddy et al. 2010) due to SMC's steeper extinction curve in A_λ/A_V-1/λ than that for local starbursts, it may be more appropriate for Himiko. Our result also suggests that Himiko is not associated with additional FIR sources that are invisible in the rest-frame UV. These implications are consistent with the conclusions of the UV-FIR luminosity ratio discussed in Figure 5.

Zoom In Zoom Out Reset image size

More interestingly, the UV slopes of the individual substructures provide valuable information on the nature of Himiko. Clumps B and C have β = −2.04 ± 0.47 and β = −2.22 ± 0.28, respectively, which are comparable to the average UV slope of ≃ L* LBGs. However, Clump A presents a very blue UV slope, β = −2.84 ± 0.32. Because this component is detected at the ∼20σ level in both J₁₂₅ and H₁₆₀, the UV slope is quite reliable. Bouwens et al. (2012) claim that selection and photometric biases lead to an error of only Δβ ≃ +0.1 for the brightest of their sources with ∼20σ photometry (see also Dunlop et al. 2013). Even including such a possible bias, Clump A remains significantly bluer than the average ≃ L* LBGs at the ≃ 2σ level.

As presented in Section 2.2, Clump A also shows Lyα emission. Together with the blue UV slope, this suggests a very young and/or metal poor component. However, the Lyα equivalent width is only ${\rm EW}_0=68^{+14}_{-13}$ Å. To understand the significance of this, in Figure 9, we compare β and EW₀ for the entire Himiko system and the various clumps with the stellar and nebular models of Raiter et al. (2010), where a Salpeter IMF is assumed. In Figure 9, the arrow size in β for the stellar extinction of E(B − V)_s = 0.1 is calculated using a combination of the empirical relation, A₁₆₀₀ = 4.43 + 1.99β (Meurer et al. 1999), and Calzetti extinction, A₁₆₀₀ = k₁₆₀₀E(B − V)_s, where k₁₆₀₀ is 10 (Ouchi et al. 2004). Similarly, the arrow size in EW₀ for E(B − V)_s = 0.1 is estimated from the relation given in Ono et al. (2010a) under the assumption of a f_ν flat continuum and the standard SFR relations of UV and Lyα luminosities in the case of B recombination. Figure 9 shows that the data points of Himiko fall on the tracks of star-formation photoionization models (Raiter et al. 2010) within the measurement errors and the dust-extinction correction uncertainties, and indicates that the Lyα emission of Himiko can be explained by photoionization by massive stars.

Zoom In Zoom Out Reset image size

We have shown (Figure 5) that Himiko's sub-millimeter emission is comparable with or weaker than that of local dwarf irregulars with far lower SFRs, indicating intensive star-formation in a dust-poor gaseous environment. In fact, assuming the local starburst SFR–L(FIR) relation of Kennicutt (1998) with Himiko's FIR upper limit luminosity of <8 × 10¹⁰ L_☉, we obtain SFR(FIR) < 14 M_☉ yr⁻¹, which is far smaller than not only our best optical-NIR estimate SFR of ≃ 100 M_☉ yr⁻¹, but also the UV-luminosity SFR of SFR(UV) = 30 ± 2 M_☉ yr⁻¹ with no dust extinction correction. This is also true under the assumption of the SFR–L(FIR) relation (Buat & Xu 1996) which is valid for local dust poor disk systems of Sb and later galaxies, which provides SFR(FIR) < 25 M_☉ yr⁻¹. In this way, Himiko does not follow the SFR–L(FIR) relation of typical local galaxies, indicating a dust-poor gaseous environment. This seems similar to observations that find extended Lyα emission in dust poor low-z galaxies (Hayes et al. 2013) and a high-z QSO (Willott et al. 2013). Based on numerical simulations, Dayal et al. (2010) find that z ∼ 6–7 LAEs are dust poor with a dust-to-gas mass ratio smaller than the Milky Way by a factor of 20. Dayal et al. (2010) predict a 1.4 mm continuum flux of ≃ 50 μJy for sources with L(Lyα) = 2–3 × 10⁴³ erg s⁻¹ at z = 6.6, a result comparable to our ALMA observations. Deeper ALMA observations could further test the model of Dayal et al. (2010) and place important constraints on the dust-to-gas mass ratio.

Similarly, our strong upper limit on the [C ii] 158 μm line (Figure 6) places it significantly below the scaling relation of $L_{{\rm [C\,\scriptsize{II}]}}$ and SFR that is obeyed by lower redshift galaxies. This discovery indicates the following four possibilities: Himiko has (1) a hard ionizing spectrum from an AGN, (2) a very high density of photo-dissociation regions (PDRs), (3) a low metallicity, and (4) a large column density of dust. In case (1), a hard ionizing spectrum from an AGN can produce little [C ii] luminosity relative to FIR luminosity, due to the intense ionization field (Stacey et al. 2010). As we discuss in (2) of Section 4.2, there are no AGN signatures; there are no detections of X-ray and high-ionization lines, as well as extended sources plus non-AGN-like Lyα profile+surface brightness. We can rule out possibility (1). In case (2), a very high density of PDRs provides more rapid collisional de-excitations for the forbidden line of [C ii], and quenches a [C ii] emission line. In case (3), the PDRs in Himiko are composed of metal poor gas that may be quite typical of normal galaxies observed at early epochs. De Looze (2012) has argued that offsets from the [C ii]–SFR relation can be explained in terms of metal abundance and this would imply a gas-phase metallicity of ≲ 0.03 Z_☉. Indeed, for our young mean stellar age of 160–410 Myr, standard ionization-photon bounded Hii regions with a local chemical abundance would yield [C ii] emission somewhat above the local scaling relation, due to the expected large PDRs. Moreover, the recent numerical simulations predict that a [C ii] flux drops as metallicity decreases (Vallini et al. 2013). Vallini et al. (2013) claim that Himiko's gas-phase metallicity is sub-solar on the basis of the comparison of their models with the previous IRAM [C ii] upper limit. Comparing these numerical models with our strong ALMA upper limit of [C ii] would place further constraints on the metallicity of Himiko. In case (4), the depth of C⁺ zones in PDRs is determined by dust extinction. Since the C⁺ zones extend over the dust extinction up to A_v ⩽ 4 (Malhotra et al. 2001), heavy dust extinction in the ISM does not allow the creation of a large C⁺ zones emitting [C ii]. However, from the no detection of a 1.2 mm dust continuum discussed above, the heavy dust extinction narrowing the PDRs is unlikely. As dust extinction and gas phase metallicity generally correlate closely (Storchi-Bergmann et al. 1994; see also Finlator et al. 2006), the weak dust emission also suggests a very low metallicity gas. Thus, case (3) is probably true, which contributes the weak [C ii] emission. Case (2) could also help to weaken the [C ii] emission. To summarize, faint [C ii] and weak dust emission can be explained in a self-consistent manner with a very low metallicity gas and little dust in a near-primordial system.

It is informative to compare the above conclusion with the only other well-studied galaxy at this redshift, HFLS3 at z = 6.34 (Riechers et al. 2013), recognizing that both this galaxy and Himiko were selected based on their extreme properties. Figure 10 presents the ratio of [C ii] to the FIR luminosity as a function of FIR luminosity. Although Himiko is significantly offset from the trend shown by AGN and local starbursts, this is not the case for HFLS3. Although Riechers et al. (2013) claim that HFLS3 is free from AGN activity based on the level of excitation for CO and H₂O, its small [C ii] to L_FIR ratio of $L_{{\rm [C\,\scriptsize{II}]}}/L_{\rm FIR}=5\times 10^{-4}$ suggests otherwise (Stacey et al. 2010; Sargsyan et al. 2012). On the other hand, the recent study of Diaz-Santos et al. (2013) finds that a luminous infrared galaxy (LIRG) with compact star-forming regions show a smaller $L_{{\rm [C\,\scriptsize{II}]}}/L_{\rm FIR}$ value, and that a $L_{{\rm [C\,\scriptsize{II}]}}/L_{\rm FIR}$ value of pure star-forming LIRG drops by an order of magnitude, from 10⁻² to 10⁻³. Thus, there is the other possibility that HFLS3 could be a pure star-forming source that is more compact than those of local pure star-forming LIRGs.

Zoom In Zoom Out Reset image size

In considering the origin of Himiko's extreme SFR and extensive Lyα halo, it is convenient to return to the various explanations originally proposed by Ouchi et al. (2009b) on the basis of the Subaru, UKIDSS, and shallow Spitzer data available at the time, taking into account the progress achieved with our new deep ALMA, HST, and Spitzer data.

• 1.  
  A Gravitationally Lensed Source. Ouchi et al. (2009b) discounted this possibility on the basis of the resolved kinematics of the extended Lyα halo. In Section 2.2, we have strengthened the objections to this hypothesis since our HST data reveal three L* sources whose morphological asymmetries are not consistent with gravitational lensing. Moreover, we can find no potential foreground lens in the vicinity of Himiko (Ouchi et al. 2009b). Such a lens would have to be one of the three clumps revealed in the HST images, each of which has a 0.9 μm-continuum break and a blue UV continuum consistent with being physically associated at z ≃ 6.6. Our deep IRAC data also show no potential lensing sources near Himiko, suggesting that there are no lensing objects with very red color, such as dusty starbursts at intermediate redshifts, which are invisible in the optical and NIR bands. Thus, we conclude that Himiko is not a gravitational lensed system.
• 2.  
  Halo gas ionized by a hidden AGN. Our HST images do not reveal any obvious point source that could represent an active nucleus (Figure 4). Moreover, as noted by Ouchi et al. (2009b), Himiko is undetected at X-ray wavelengths of 0.5–2 keV down to 6 × 10⁻¹⁶ erg s⁻¹ cm⁻², and the Keck optical spectrum does not reveal any high ionization features such as Nv. Finally, radiative transfer simulations by Baek & Ferrara (2013) show that the Lyα line profile and surface brightness of Himiko are inconsistent with heating from either a Compton-thick or Compton-thin AGN. Thus, we conclude the Lyα halo is unlikely to be heated by an AGN.
• 3.  
  Clouds ofH iiregions in a single virialized galaxy. Ouchi et al. (2009b) discussed the possibility that Himiko could be a single virialized system. Since the new HST data reveals three distinct UV luminous clumps, each comparable to the characteristic luminosity L*, we consider that Himiko is unlikely to be a single virialized system Although there are many reports of disk galaxies with prominent clumps at z ∼ 2–3 (e.g., Genzel et al. 2011), the absence of a stellar disk (Figure 4) distinguishes Himiko from a single galaxy with clumpy structures such as those found at lower redshifts.
• 4.  
  Cold gas accretion onto a massive dark halo producing a central starburst. Some theoretical studies have suggested that cold gas can efficiently penetrate into the central regions of a dark halo if that halo is more massive than the shock-heating scale of ∼4–7 × 10¹¹ M_☉ at z > 4 (Dekel et al. 2009; Ocvirk et al. 2008). Given Himiko's stellar mass (1.5–3 × 10¹⁰ M_☉, Section 3.3) and little or no evolution in the ratio of stellar mass, M_*, to halo mass inferred for 0 < z < 1 (M_*/M_DH ≲ 0.05, Leauthaud et al. (2012)), we expect a halo mass of M_DH ≳ 3–6 × 10¹¹ M_☉. Abundance matching considerations support this estimate. Ouchi et al. (2009b) calculated that there should be at least one halo of mass 10¹² M_☉ in the survey volume of 8 × 10⁵ comoving Mpc³ where Himiko was found (see also Behroozi et al. 2012).Although Himiko's halo mass does likely lie in the range where cold accretion could be possible, we note that some recent simulations have cast doubt on the efficiency of this mode of assembly (Nelson et al. 2013; Vogelsberger et al. 2013).
• 5.  
  Outflowing gas excited by shocks or UV radiation from starbursts and/or mergers. The extensive Lyα nebula may be powered by star formation itself, but the gas could also be shock heated by strong outflows driven by multiple supernova explosions in an intensive starburst (Mori et al. 2004). Figure 9 shows the relation between the UV slope β, which characterizes the stellar population and the Lyα equivalent width EW₀. This shows that the photoionization models of Raiter et al. (2010) whereby Lyα photons are scattered by the ISM and circum-galactic medium can explain the properties of Himiko, notwithstanding the uncertainties in β. The success of this model depends, of course, on the escape fraction of ionizing photons which should be moderately low (<50%) so that scattering is effective. However, the conclusion is robust even if we adopt a moderate dust extinction of E(B − V)_s = 0.15 (Section 3.3).Thus, within the uncertainties, the amount of star formation observed is sufficient to power the extended Lyα nebula; outflow and shocks are not required. This simple photoionization scenario is consistent with the negligible hidden star-formation suggested by the weak dust and carbon emission from our ALMA observations. As discussed in Ouchi et al. (2009b), the FWHM of the Lyα line is only v_FWHM = 251 ± 21 km s⁻¹, further indicating that powerful outflows are not present.
• 6.  
  Merging bright galaxies. Although this is not a separate hypothesis from (5) above, we can ask what triggers the intense star formation that likely powers the extended nebula. In Figure 4, we have identified three L* clumps that are highly suggestive of a rare triple merger. As Ouchi et al. (2009b) reported, Himiko presents a small velocity offset of Lyα emission across the nebula (Δv = 60 km s⁻¹) with a narrow line width (v_FWHM = 251 km s⁻¹; see Figure 7 of Ouchi et al. 2009a for the Lyα line velocities that are measured on the slit position shown with the red box in Figure 1 of Ouchi et al. 2009a). Thus, the merger would have to be largely confined to the direction perpendicular to the line of sight.

Although a triple major merger is a rare event, our data suggest that this explanation is the most plausible. Recent numerical simulations predict that some extended Lyα sources originate in mergers (Yajima et al. 2012a). One interesting feature of Himiko is that the brightest portion of the Lyα nebula does not coincide with the geometric center of the three clumps, but is located between the blue Lyα clump A and clump B. Given the discussion in (5) above, this indicates that Lyα photons are mainly produced by Clump A, a very young and metal poor component.