A massive interacting galaxy 510 million years after the Big Bang

A massive interacting galaxy 510 million years after the Big Bang

Abstract

JWST observations spectroscopically confirmed the existence of galaxies as early as 300 million years after the Big Bang.

It had a higher number density than what was expected based on:

• galaxy formation models
• Hubble Space Telescope observations.

Yet, the majority of sources confirmed spectroscopically so far in the first 500 million years have rest-frame UV-luminosities below the characteristic luminosity (M∗ UV), limiting the signal to noise ratio for investigating substructure.

We present a high-resolution spectroscopic and spatially resolved study of a bright (MUV = −21.66 ± 0.03, ∼ 2M∗ UV) galaxy at a redshift z = 9.3127 ± 0.0002 (510 million years after the Big Bang) with an estimated stellar mass of (1.6+0.5−0.4) × 109 M⊙, forming 19+5−6 Solar masses per year and with a metallicity of about one tenth of Solar.

The system has a morphology typically associated to two interacting galaxies, with a two-component main clump of very young stars (age less than 10 million years) surrounded by an extended stellar population (120 ± 20 million years old, identified from modeling of the NIRSpec spectrum) and an elongated clumpy tidal tail.

The observations acquired at high spectral resolution identify oxygen, neon and hydrogen emission lines, as well as the Lyman break, where there is evidence of substantial absorption of Lyα.

The [O ii] doublet is resolved spectrally, enabling an estimate of the electron number density and ionization parameter of the interstellar medium and showing higher densities and ionization than in analogs at lower redshifts. We identify evidence of absorption lines (silicon, carbon and iron), with low confidence individual detections but signal-to-noise ratio larger than 6 when stacked.

These absorption features suggest that Lyα is damped by the interstellar and circumgalactic medium. Our observations provide evidence of rapid and efficient build up of mass and metals in the immediate aftermath of the Big Bang through mergers, demonstrating that massive galaxies with several billion stars are in place at early times.

1 Main

The first generations of stars and galaxies in the Universe formed in physical conditions different to those of the modern Universe. In fact, gas is expected to have nearly primordial composition, with low levels of chemical enrichment and dust content (e.g., [1]). Gas cooling is further limited by the higher Cosmic Microwave Background radiation, possibly altering the characteristic fragmentation mass of protostellar clouds [2]. In addition, early forming galaxies are expected to experience an elevated merger rate [3], affecting their morphology and stellar populations.

Hubble Space Telescope observations have been able to identify galaxy candidates at redshift z ∼ 8−11 (∼ 600−400 million years after the Big Bang; e.g. see [4]), and follow-up with the Spitzer Space Telescope provided evidence of relatively old stellar populations, suggesting that star formation started at z > 15 [5, 6]. Yet, comparison to theoretical and numerical modeling has been restricted to number counts and luminosity functions by the limited angular resolution and dearth of spectroscopic data.

With JWST commencing science operations in July 2022, progress has been rapid and transformational. Already in the first Cycle of science programs, JWST has built a convincing sample of z > 10 candidates based on NIRCam photometry [7–13], and a growing number of sources at z ≳ 8 are being confirmed spectroscopically with NIRSpec [14–27]. Surprisingly, the number density of (spectroscopically confirmed) high-redshift galaxies has been higher than expected by most models, in particular at the bright end of the luminosity function, possibly suggesting that we are missing key physical processes connected to the formation of first galaxies [28, 29], though observations do not suggest inconsistency with ΛCDM [27, 30, 31]. The superior spectral and spatial resolution of JWST in the near-infrared offers a path to investigate these initial findings through detailed stellar population studies in early galaxies. An initial spectroscopic census of JWST galaxies above z > 7 paints a picture of elevated ionization parameters, low metallicities, low dust content, high specific star formation rates (sSFR), and potentially higher ionizing radiation escape fraction (fesc) than seen locally (e.g., [18, 22, 32–36]). Many of these galaxies show resemblance to the extreme interstellar medium (ISM) conditions observed in metal-poor actively star forming galaxies, both local (blueberries, greenpeas [37–40]) and at moderate redshifts (z ∼ 2, [41–43]). The spatial resolution of JWST also means the resolved stellar populations of galaxies in the epoch of reionization can be studied as well, revealing that these systems, once considered compact with HST, exhibit spatial variations in their physical properties [44–46].

In this work we extend the frontier of detailed investigations of individual galaxy properties at very high redshift by reporting on imaging and spectroscopic observations of one of the brightest among the galaxy candidates at z ≳ 9 observed with JWST, with flux ∼ 0.35 µJy in F444W (corresponding to mAB ∼ 25.0). These observations include 6-band NIRCam imaging (F115W, F150W, F200W, F277W, F356W, F444W, presented in Extended Data Table 1,) and NIRSpec high resolution (R ∼ 2700) Multi-object Spectroscopy (see 2.2). The galaxy - which we call here Gz9p3 - was initially identified as a potential F105W or F125W dropout based on HST observations, and then confirmed as a NIRCam F115W drop-out with a probable z ∼ 9.45 redshift [47].

The NIRSpec data we use are part of the GLASS-JWST program ERS-1324 [48] centered on the foreground galaxy cluster Abell 2744, which is gravitationally lensing the high-redshift background sources. Gz9p3 is located in the outskirts of the cluster and relatively far away from high-magnification regions, with a best estimate of the lensing magnification µ = 1.66±0.02 based on [49] (see 2.3).

The JWST photometry for Gz9p3 is shown in the top panel of Figure 1. The source has a magnification-corrected apparent AB magnitude mAB = 25.56±0.06 in F444W, which for the cosmology adopted in this paper (see 2.1) corresponds to MAB = −21.77 ± 0.06 – approximately 50% brighter than the characteristic luminosity (M∗) of galaxies at this time [4]. The figure also includes a 1-D extraction of the NIRSpec spectrum in the middle panel, with spectral coverage in the [1.1 : 4.5] µm range (with some gaps due to the target location relative to the edge of the instrument field of view as discussed in 2.2.1). We present the full 2D spectrum in Extended Data Figure 1. The NIRSpec Multi-shutter Assembly (MSA) was configured based on HST imaging and as such the shutter covers the main body of the galaxy as shown in the bottom left panel of the figure. The spectrum shows a clear continuum detection and four emission lines are detected at [3.8 : 4.1] µm, which we identify as ([O ii], [Ne iii] λ3870, and [Ne iii] λ3969 blended with Hϵ) for a source located at a redshift zspec = 9.3127 ± 0.0002 (see 2.5). In addition, the spectrum shows a Lyman break, and by modeling the stellar continuum as a step function around the break, we determine zbreak = 9.35+0.01−0.05, consistent with the redshift measurement from emission lines (shown in Extended Data Figure 2, with further details in 2.5.1).

Figure 1: JWST NIRCam and NIRSpec observations of Galaxy GLASS ID:Gz9p3.
Top row: NIRCam direct imaging in broad-band filters. Second row: 1D standard extracted spectrum in units of fλ = 10−19 erg s−1 cm−2 Å−1 (with Npix=20 binning) from NIRSpec F100LP/G170H, F170LP/G235H & F290LP/G395H filter-dispersor configurations (from left-to-right in blue, orange and green). The observed frame location of the Lyman-break (λrest = 1215.67) and the [O ii], [Ne iii] λ3870 & [Ne iii] λ3969+Hϵ emission lines are overlaid as vertical dashed red lines for a zspec = 9.313. The Solid grey regions mask the location of contaminating emission lines from the dispersed spectrum of a galaxy falling in an open shutter of a separate quadrant of the NIRSpec MSA. Bottom left panel shows the color composite from NIRCam with NIRSpec slit positioning overlayed. The other three bottom panels are zoom in on a ±400 Å region centered on each emission line complexes ([O ii], [Ne iii] λ3870, [Ne iii] λ3969+Hϵ) with the continuum subtracted and the best-fit profile overlaid (again in units of 10−19 erg s−1 cm−2 Å−1).

The galaxy properties are derived from Spectral Energy Distribution (SED) modeling using both broad-band photometry and the 1-D spectrum as input (see Extended Data Figure 3, with further details in 2.8). The results are summarized in Table 1. Based on photometry of the whole galaxy (Kron fluxes, see [50] and 2.8), the modeling returns a magnification-corrected stellar mass of log10(M∗/M⊙) = 9.2+0.1−0.2 and MUV = −21.66 ± 0.03. (These are statistical uncertainties from photometric errors only). This makes Gz9p3 one of the most massive and intrinsically brightest galaxies confirmed in the epoch of reionization, and the brightest and most massive at z > 9 (see Figure 2 for a census for high-redshift spectroscopically confirmed galaxies). Even when compared against photometric galaxy candidates [51], Gz9p3 has one of largest masses known within the first 750 Myr since the Big Bang. The SED modeling from the spectrum is restricted to the main region of the galaxy where the shutter was placed, and returns a stellar mass consistent with the photometric estimate (see 2.7). Both SED modeling approaches (photometry and spectrum fitting) also identify substantial ongoing star formation (9−19 M⊙ yr−1), and limited evidence of dust due to the blue spectral slope β in the UV (−2.2 ≲ β ≲ −1.9), with robust results over a range of assumed star formation histories. Interestingly, the spectrum-based modeling shows evidence for older stellar populations in the central region of the galaxy (age 120±20 Myr), indicating that star formation started as early as z ≳ 15 to produce the average age observed (see 2.7). In contrast, modeling based on photometry infers younger ages, with integrated-light fits giving an age of 25+15−12 Myr and spatially resolved analysis identifying regions with ages < 10 Myr (see 2.9.1 and Extended Data Table 2).

Figure 2: Census of MUV and stellar mass in high-redshift galaxies.
Left: Absolute UV magnitude of galaxies spectroscopically confirmed with JWST, corrected for lensing magnification where appropriate. Right: Stellar Mass (log10(M∗/M⊙)) distribution of confirmed galaxies, corrected for magnification where appropriate. Error bars derive from the measured mass or MUV of individual galaxies. In both panels, the point shape represents the resolution mode of the NIRSpec spectroscopy (low-resolution prism: circles; Medium resolution: squares; High resolution: diamonds and star). A black central dot marker indicates the detection of emission lines. Our target (red star with central dot) is one of the intrinsically brightest and most massive galaxies in the epoch of reionization among the current JWST samples from [15–27], and the highest and most massive at z > 9. It is also one of the highest redshift galaxies with emission line detections and the highest redshift one observed in the high resolution mode. Stellar masses were taken as quoted from each study. We note that we do not include MACS1149-JD1 [135] as a spectroscopically confirmed galaxy (at z=9.11) due to significant uncertainty on its lensing magnification, and hence also on its intrinsic MUV and Stellar Mass. Additionally, we note that only a subset of the [18] sample have their masses reported, and masses aren’t provided for [22]. We label all galaxies at z > 9. In these labels A23 refers to galaxy ID: 10058975 for [22] and Wi22 refers to [15].

The detection of rest-optical emission lines provides a window into the interstellar medium conditions in the galaxy by resolving the [O ii] doublet thanks to our high spectral resolution (Figure 1), measuring a line ratio of 0.94+0.14−0.18 (see 2.7). The relative strength of these low-ionization lines is sensitive to the electron number density ne [52], leading to a measurement of ne = 590+570−250 cm−3. This is marginally higher (at ≳ 1σ) than the median values of ne = 225 cm−3 seen in galaxies at z = 2.3, and ne = 26 cm−3 seen in local galaxies [52], qualitatively following the trend of ne increasing with redshift as reported in that study. From the spectrum we determine a Ne3O2 ([Ne iii] λ3870/[O ii]) ratio of 0.81 ± 0.09. As these two lines are close in wavelength, the ratio is insensitive to dust reddening. The measurement is higher than what is typically seen at lower redshift, indicating a high ionization parameter of log U = −2.13 ± 0.05 based on [53], and a low metallicity of 12 + log(O/H) = 7.6 ± 0.5 (depending on which one among the low-z Ne3O2 calibrations is used and including systematic uncertainties [54–57]; see 2.7.2 for further details). Together, these conditions indicate a sub-solar (Z ≲ 0.1 Z⊙) metal-poor interstellar medium, with a high electron density and ionization parameter, exhibiting similar properties to other galaxies spectroscopically confirmed at zspec > 8 [19, 20, 22, 32–34, 36]. The ISM conditions are consistent with expectations from the young stellar ages derived from SED fitting, providing a self-consistent picture of the stellar populations and their surrounding gas.

In Figure 3, the large stellar mass and low oxygen abundance place Gz9p3 below the mass-metallicity relations for z = 4 − 9 derived by [58], and marginally below the relation for z = 2 − 4 galaxies from [56, 59], even though systematic uncertainty may affect the robustness of these conclusions (see 2.7.2). The offset suggests Gz9p3 has a high gas fraction and potentially high accretion rates of pristine gas. This is qualitatively consistent with expectations from theoretical and numerical modeling of galaxies at these early times, given the short assembly times of their dark-matter halos [60].

Figure 3: Location of Gz9p3 on the Mass metallicity relation. Gz9p3 is shown in black, where the error bar presents the random and systematic uncertainty in the metallicity, based on Ne3O2 diagnostic calibrations from [54–57, 90]. The figure includes a comparison to mass metallicity relations covering 4 redshift epochs: z ∼ 0 (blue) from [56, 115], z ∼ 2.2 and z ∼ 3.5 (green and red) from [56, 59] and z = 4 − 10 (purple) from [58]. We additionally show the JWST z = 4 − 10 galaxies from [58] in purple, where the error bars derive from the mass and metallicity calculation for each individual galaxy.

Interestingly, not only is there no evidence of Lyα emission, with a stringent limit on the equivalent width from Table 1, but also the stellar continuum at 1216 Å < λrest < 1240 Å (redward of the Lyman break) shows a deficit. Measurements over ∆λrest = 4 Å and 24 Å windows show a 80% and 40% deficit at a 5.7σ and 6.5σ significance, respectively (see 2.7.6). The softening of the spectral break supports the presence of absorption from Lyα damping wings (seen in the spectra of many z > 9 galaxies, [24, 63, 64]). One interpretation is that the damping is due to absorption by the intergalactic medium [65], which would indicate that the galaxy does not reside within a large ionized bubble. Such a scenario falls in line with the expected transmission due to damping wing in a neutral IGM from [66], with a predicted flux of ∼ 0.3× continuum and ∼ 0.8× continuum at 1000 km s−1 (∼ 4 Å) and 6000 km s−1 (∼ 24 Å) respectively. However, it is difficult to reconcile the lack of a large ionizing bubble with the high stellar mass and presence of relatively old (> 100 Myr) stars, especially because Lyα emission is detected in galaxies at higher redshift with lower star formation rates and stellar masses such as GN-z11 [23]. An alternative interpretation is that the interstellar and circumgalactic medium in Gz9p3 is primarily responsible for the lack of Lyα emission in the spectrum, irrespective of the IGM conditions. This scenario is supported by the detection of the SiIIλ1260 and CIIλ1335 absorption features with a rest-frame equivalent width of (−3.7 ± 0.8) Å. In fact, it has been shown that their strength correlates with damping of Lyα and results in Lyα absorption when the equivalent width of low-ionization interstellar metal lines is ≲ −2 Å [67–69]. Also, the presence of strong low-ionization ISM line absorption and the stringent upper limit on CIII] emission suggest that the galaxy is unlikely to be a Lyman continuum leaker [70].

In addition to providing detailed spectral insight on the stellar populations within the shutter aperture, the NIRSpec observations allow us to fix the redshift of Gz9p3 for photometric pixel-by-pixel modeling. This allows us to investigate spatial variations across the galaxy, following an approach similar to the one adopted by [46] at lower redshift, to create a 2D distribution of the galaxy’s physical properties (see 2.9). The analysis is carried out with BAGPIPES [71] to determine the following maps of resolved properties, shown in Figure 4: stellar mass surface density (SMD); 100 Myr-time averaged star formation rate surface density (SFRD); mass-weighted stellar age; visual extinction (Av); and UV β slope (where fλ ∝ λ−β). The native pixel resolution in F444W is 0.031”/pixel (with a FWHM F444W ∼ 4.5 native pixels, or ∼ 0.14”), but for our analysis we bin pixels 2 × 2 to improve the signal-to-noise ratio per pixel, generating the maps at a 0.27 kpc/pixel resolution in the observed frame. This corresponds to 0.21 kpc/pixel in the image plane after accounting for gravitational magnification. The galaxy shows a morphology comprised of an elongated tail and a central body, which in F444W appears as a single core. The peak of the star forming activity within the stellar population is situated in the central core and this traces the distribution of stellar mass. This central region of active star formation exhibits young stellar population ages (< 50 Myr), as does the clump within the tail, whilst the surrounding regions on the outskirts of the central body show older populations, albeit with larger uncertainties (50 ≲ tage/Myr ≲ 125). The galaxy shows blue β slopes throughout and relatively low visual extinction, which implies low dust content (see 2.9).

Figure 4: 2D color and physical parameter distribution of Gz9p3. Properties inferred from photometric spectral energy distribution fitting (NIRCam pixels matched to F444W and binned 2x2), with an observed frame resolution of 0.27 kpc/pixel (0.21 kpc/pixel in the image plane). From left to right: the top and second row present the star formation rate surface density, stellar age, and stellar mass surface density. The third and bottom row present the color, visual extinction and UV β slope. (Upper panels: Median value, Lower panels: Uncertainty based on 16th and 84th percentiles). The F150W 10σ contour is presented in orange in the F150W-F444W panel and the FWHM and pixel-scale are shown in the F356W-F444W panel.

The spatially resolved modeling is suggestive of an interacting system undergoing (or having recently undergone) a major merger. To further investigate this scenario, we analyze photometry at rest-frame UV wavelengths, using a combined F150W+F200W image drizzled at 20 mas/pixel, presented in Figure 5. The data clearly show two distinct cores in the main region of the galaxy, and several components in the tail identified thorough a clump-finding algorithm (see 2.10). The morphology of Gz9p3 is described by morphological parameters that indicate the galaxy is a merger (Gini=0.61, M20=-1.29, A=0.35; see [72]).

Figure 5: Morphology of Gz9p3. F150W+F200W direct image at a 20 mas/px resolution in both panels. Left: direct imaging shows a double core within the central region and an elongated clumpy structure. Right: Overlaid Clump-map, showing 4 clumps detected within the tail of the system with a clumpiness parameter c = 0.56.

Informed by the morphological analysis, we repeat the spatially resolved SED modeling by placing apertures over different stellar populations shown in Extended Data Figure 5 (innermost region, a surrounding annulus and the tail), clearly seeing a distinction between active regions of star formation and an underlying older stellar population, as expected from the merging scenario (see 2.9.1).

The combined spectroscopic and imaging data paint a picture of a very bright and relatively massive interacting system just 0.5 Gyr after the Big Bang, raising the question of how likely such JWST observations should be. Figures 2-3 hint that the system could be an outlier. To quantify expectations we consider both analytical modeling of early galaxy formation and comparison to cosmological hydrodynamical simulations. We find that while the likelihood of capturing an interacting system is relatively high, i.e. ∼ 20% for a major merger, the stellar mass of Gz9p3 is higher than expected (see 2.11-2.12). This would indicate either the system is hosted in a very rare dark matter halo for that epoch, that we serendipitously observed, or more likely that the current recipes for star formation are missing some key ingredients at early times [73]. The latter interpretation would be consistent with the excess of sources identified by JWST at z > 10 through imaging programs [47] and with the high numbers of massive red galaxy candidates found at z ∼ 7.5 − 9 [51]. All these aspects make Gz9p3 an excellent target for further spectroscopic investigations, in particular through the Integral Field Unit mode on NIRSpec, that would shed further light on the kinematics of the system and on the complex interplay between assembly of dark matter halos, star formation and physical conditions in the interstellar, circumgalactic and intergalactic media at very early times.

Table 1: Physical properties for galaxy Gz9p3.

†1σ upper limit. Line fluxes are not corrected for dust extinction or slit loss.
b Photometry-only SED properties for the full system, corrected where appropriate for magnification.
c Spectrum+Photometry SED properties for our main region aperture (see Extended Data Figure 5), corrected for magnification and adopting slit-losses based on aperture photometry.
† The SFR is taken as the 100 Myr average from the BAGPIPES SFH model.