JADES Imaging of GN-z11: Revealing the Morphology and Environment of a Luminous Galaxy 430 Myr after the Big Bang

The NIRCam observations presented here come from the JWST Advanced Deep Extragalactic Survey (JADES), which is conducting deep JWST imaging and spectroscopy of the GOODS-S and GOODS-N fields. The nine-band imaging of GN-z11 results from the combination of two NIRCam pointings, observations 2 and 3, from program 1181 (PI: Eisenstein) taken on UT2023-02-03. Each pointing is a six-point dither, conducted with a two-point subdither with the MIRI F1800W pattern ²⁹ in a three-part IntramoduleX dither. The two pointings intentionally overlap on GN-z11, but in opposite portions of NIRCam module B (detectors B3 and B2). Hence, in most filters, GN-z11 was placed on 12 distinct and well-separated pixels. Both pointings include eight filters: F090W, F115W, F150W, F200W, F277W, F356W, F410M, and F444W. One pointing additionally includes F335M, paired with an extra six exposures of F115W. The exposure times are listed in Table 1.

The details of the data reduction of the NIRCam data will be presented as part of the JADES program in S. Tacchella et al. (2023, in preparation). We give here a brief overview of the main steps. We use the JWST Calibration Pipeline v1.9.2 with the CRDS pipeline mapping (pmap) context 1039. We process the raw images (uncal frames) with the JWST Stage 1, which performs detector-level corrections and produces count-rate images (rate frames). We run this step with the default parameters, including masking and correction of "snowballs" in the images caused by charge deposition following cosmic ray hits.

Stage 2 of the JWST pipeline performs the flat-fielding and applies the flux calibration (conversion from counts per second to megajanskys per steradian; Boyer et al. 2022). For the long-wavelength (LW) filters, we find that the current pipeline flats (ground flat corrected for in-flight performance) introduce artifacts in the background that become visible in the final mosaics. We therefore construct sky-flats for the Stage 2 step by stacking, separately for each LW filter and module, 80–200 uncal frames from PID 1180, 1210 1286, and the public program JEMS (JWST Extragalactic Medium-band Survey; PID 1963; Williams et al. 2023). Since we do not have enough exposures to construct a robust sky-flat for the F335M and F410M filters, we have effectively interpolated those flat fields via a linear combination of flat field "components" determined from a non-negative factorization of the wide-band sky-flats (i.e., F227W, F356W, and F444W). The rest of Stage 2 is run with the default values.

Following Stage 2, we perform several custom corrections to account for several features in the NIRCam images (Rigby et al. 2022), including the 1/f noise (Schlawin et al. 2020), scattered-light effects ("wisps"), and the large-scale background. Since all of those effects are additive, we fit and subtract them. We assume a parametric model for the 1/f noise, a scaled wisp template (only for the SW channel detectors A3, A4, B3, and B4), and a constant, homogeneous large-scale background. We have constructed wisp templates by stacking all images from our JADES (PID 1180, 1210, 1286) program and several other programs that are publicly available (PIDs 1063, 1345, 1837, 2738).

The final mosaics are constructed using Stage 3 of the JWST Pipeline. Before combining the individual exposures into a mosaic, we perform an astrometric alignment using a custom version of JWST TweakReg. We calculate both the relative and absolute astrometric correction for images grouped by visit and band by matching sources to a reference catalog constructed from HST F814W and F160W mosaics in the GOODS-N field with astrometry tied to Gaia Early Data Release 3 (Gaia Collaboration et al. 2021; G. Brammer 2023, private communication). We then run Stage 3 of the JWST pipeline, combining all exposures of a given filter and a given visit. We choose a pixel scale of 003 pixel⁻¹ and drizzle parameter of pixfrac = 1 for the SW and LW images. ³⁰ A careful analysis of the astrometric quality reveals an overall good alignment with relative offsets between bands of less than 0.1 short-wavelength pixel (<3 mas).

The resulting thumbnail images together with the SED are shown in Figure 1. These thumbnails show a compact source at α = 189.106042°, δ = + 62.242042° that is an obvious F115W dropout, with a faint haze to the northeast. The haze is seen in several bands, including in both the SW and LW detectors, establishing it as a true on-sky signal.

Zoom In Zoom Out Reset image size

Using an inverse-variance-weighted stack of the F277W, F335M, F356W, F410M, and F444W images, a signal-to-noise ratio (S/N) image is constructed to provide a detection image. Contiguous regions of greater than five pixels with signal-to-noise ratio (S/N) ≥ 3 were selected as potential sources. For every source location, forced photometry was performed in 01 and 035 radius circular apertures on the JADES/NIRCam and the 30 mas pixel scale HST Legacy Fields mosaics (Illingworth et al. 2013; Whitaker et al. 2019) for the ACS F435W, F606W, F775W, F814W, F850LP and WFC3/IR F105W, F125W, F140W, and F160W filters. We used photutils to perform the force photometry measurements (Bradley et al. 2022). An annular aperture of width Δr = 01 and inner radius r = 04 about each source is used to measure and remove the local background. The subtracted background fluxes 1–9 nJy, i.e., roughly 3%–5% of the source flux (maximum of 8% in F150W). No PSF matching was performed on the HST and JWST bands for the forced photometry, but we perform a point-source (PS) aperture correction. The 01 (035) aperture corrections for F090W, F115W, F150W, F200W, F277W, F335M, F356W, F410M, and F444W amount to 1.37 (1.11), 1.31 (1.10), 1.33 (1.11), 1.38 (1.10), 1.63 (1.15), 1.74 (1.14), 1.79 (1.15), 1.88 (1.13), and 2.07 (1.17), respectively.

We note that in the JWST SW images, GN-z11 lies directly on a diffraction spike from an F115W∼18.4 AB star at [α, δ] ≃ [189.10568, 62.2458] that necessitates the local background correction. We report the aperture photometry in Table 1, which includes both sky and source photon contributions. The uncertainties are computed by adding in quadrature the contribution from the sky measured by placing random apertures on the images and the Poisson uncertainty from the source counts. For the long-wavelength bands, these uncertainties are larger than those measured at the source locations in the JWST pipeline ERR mosaics, which also include both sky and source photon contributions, because the random aperture sky uncertainties include the effects of correlated noise from resampling the mosaics. Our quoted uncertainties do not include any contribution from photometric zero-point uncertainties or from large-scale gradients in the instrument flat fields. We include an error floor of 5% in our SED and photometric redshift fits to hedge against these.

The SED of GN-z11 is displayed in Figure 1 and shows the classic shape of a high-redshift Lyα dropout. The continuum is strong and blue, roughly zero color in the AB system, before plummeting shortward of the break. Bunker et al. (2023) provides a robust determination of z = 10.60 through the detection of many well-detected narrow lines. The photometric measurements are wholly consistent with this redshift. Lyα is shifted to 1.41 μm, lying in the F150W filter. We observe a rest-UV flux density of 144.4 ± 2.7 nJy in F200W. The F115W filter lies entirely shortward of the Lyα wavelength and is indeed observed as a complete dropout, with a flux of 1.2 ± 3.1 nJy. This is a very strong suppression of a factor of at least ∼20 (95% confidence), nearly three magnitudes.

The last nine bands of Table 1 are the ACS and WFC3 07 diameter aperture photometry from the HST Hubble Legacy Field v2.5 images of GOODS-N. Oesch et al. (2014) reported −7 ± 9 nJy, 11 ± 8 nJy, 64 ± 13 nJy and 152 ± 10 nJy for HST WFC3 F105W, F125W, F140W, and F160W, respectively. Our measurements are in agreement with those of Oesch et al. (2014), excepting F160W for which we measure a flux lower by about 25% (3σlower). The Spitzer IRAC photometry of 139 ± 20 and 122 ± 21 nJy at 3.6 and 4.5 μm (Ashby et al. 2013; Oesch et al. 2014) are also in good agreement.

Longward of Lyα, the ratio between the fluxes measured in F200W and F356W indicates a rest-UV slope of $\beta =-{2.41}_{-0.07}^{+0.06}$, where the flux density scales as f_ν ∝ λ^β+2. This is consistent with the value determined from the prism spectrum of β = −2.36 ± 0.10 (Bunker et al. 2023). At the longest NIRCam wavelengths, the F410M filter corresponds to 3360–3700 Å in the rest frame, just blueward of the Balmer jump, while the F444W filter extends to 4300 Å. Hence, the comparison of these two filters can be sensitive to the presence of older stars.

To investigate the morphology of GN-z11, we need to assess the PSF and then fit models to our multiband images. We do this in several ways.

We perform fits using ForcePho (B. D. Johnson et al. 2023, in preparation), which fits multiple Sérsic components to our individual exposures across all filters, producing a Markov Chain of joint parameter fits. This method was used and is described further in Tacchella et al. (2022a), Robertson et al. (2023a), and Robertson et al. (2023a). ForcePho uses Gaussian mixtures and graphics processing units to accelerate the convolution of Sérsic models with the point-spread function, taken from WebbPSF (Perrin et al. 2014). We note that the Gaussian mixture model of the PSF is designed to reproduce the inner core and smoothed light profile, but does not have enough flexibility to capture the Airy rings of the PSF. In our analysis, we require that each component have the same light profile in all bands; only the amplitude changes per filter. However, because all components are fitted simultaneously in the Markov Chain, ForcePho captures the covariance between components, e.g,. how the flux might be differently assigned to different components. Multiple components of the same object can have their flux co-added in the Markov Chain outputs to yield total photometry that is more stable than each component separately. Because ForcePho runs on the individual exposures, it avoids issues of how mosaicking smooths the PSF or introduces covariance between pixels.

We also analyze the mosaics. For this, we need an estimate of the PSF after image combination. We build the empirical PSF (ePSF) from bright stars in the NIRCam mosaic with the Photutils package (Bradley et al. 2022). We first select bright stars from the public CANDELS catalog in the GOODS-N field (Barro et al. 2019), restricting to those with CLASS_STAR > 0.75 and FLAGS = 0. Through visual inspection, we end up with 10 and 9 unsaturated stars that have the highest S/N in the individual SW (F090W, F115W, F150W, and F200W) and LW (F277W, F335M, F356W, F410M, and F444W) mosaics, respectively. The HST H-band magnitudes for these stars range from 19.2 to 22.4 AB mag. We use the EPSFBuilder module in Photutils to model ePSF from the stars, which follows the prescription of Anderson & King (2000) and Anderson (2016). As expected, the resulting ePSF is mildly broader (see Figure 2) than the WebbPSF, which is oversampled and un-mosaicked. We then fit both single Sérsic and PS models to the combined mosaicked data using ProFit (Robotham et al. 2017) and Lenstronomy (Birrer & Amara 2018). Both ProFit and Lenstronomy are configurable codes that produce consistent results, similar to the commonly used GALFIT fitting software (e.g., Kawinwanichakij et al. 2021; Robertson et al. 2023b).

Zoom In Zoom Out Reset image size

The first JWST images revealed a diversity of morphological and structural properties of high-z galaxies (e.g., Ferreira et al. 2022; Jacobs et al. 2023; Suess et al. 2022; Robertson et al. 2023b; Kartaltepe et al. 2023; Nelson et al. 2023), including compact and clumpy structures (e.g., Tacchella et al. 2023; Chen et al. 2023). GN-z11 is not an exception and the images are morphologically complex because of the haze to the northeast. Furthermore, it quickly became clear that the core of GN-z11 is extremely compact. Fitting a single Sérsic component yielded very large Sérsic indices using Forcepho, ProFit, and Lenstronomy, raising concerns that a combination of diffuse light and a bright core might be at play.

In Figure 2, we show the encircled flux as a function of radius, both for full annuli and for wedges on and off the haze. The haze causes a clear excess at angular separations of 0.2–06. We therefore introduce an off-center Sérsic component to fit this source; we call this the "Haze." For the light near the core, the profile is mildly more extended than either PSF estimate. We opt to fit a central PS and a separate Sérsic component, the latter being named "Extended." We also include a Sérsic component for the brighter galaxy just over 1'' away to the northeast. All four components are varied simultaneously by ForcePho, resulting in a Markov Chain that incorporates the joint covariances of the fits to these components.

The model photometry results are presented in Table 2. The PS and Extended component are both well detected. As expected, the fluxes from these two very close components are anti-correlated, so that the sum, called "Galaxy," has substantially smaller errors than the quadrature sum. Both components show a sharp Lyα dropout. The summed Galaxy photometry is a close match to the 02 diameter aperture photometry. The model images and residuals for the best-fit model are shown in Figure 3, from which one can see that the three components substantially explain the images in all filters.

Zoom In Zoom Out Reset image size

The centroid of the Extended component is allowed to shift from the PS, and the best fit gives a small shift of 21 mas, less than one pixel. The best fit has a Sérsic index n = 0.9 ± 0.1 and half-light radius along the major axis of 49 ± 3 mas, corresponding to 200 pc at z = 10.6. The combined sizes (Point and Extended source) are 0016 ± 0005 (64 ± 20 pc), which we obtain by flux-weighting the sizes of both models and taking into account the 21 mas spatial offset between the PS and extended components. For the Extended component we find an axis ratio of q = 0.67 ± 0.05 and a position angle PA of 34° ± 5°. The two components have similar SEDs, but it is interesting to note the variations at F410M and F444W, which straddle the Balmer break. The Extended component shows a redder F410M–F444W color, while the PS is mildly blue. This will be discussed in more detail in the next section.

The Haze component also fits to n = 0.9 ± 0.1 and a half-light radius of 011 ± 001, offset 041 from the PS. Its axis ratio is q = 0.57 ± 0.15 at a PA of −55° ± 16°. The Haze clearly differs in its spectrum from the other two components. It is much redder in F200W–F277W color and is not detected in F150W. We also find notable drops in flux in the two medium-band filters, F335M and F410M, which might be indicative of strong emission lines in the SED. Photometric redshift fits with EAZY and Prospector substantially favor 4 < z < 5 solutions compared to those at z = 10.6. From this, we conclude that it is more likely that the Haze is a chance projection with a lower redshift low-surface-brightness galaxy. The nature of the 4 < z < 5 solution is probably that of a young star-bursting galaxy with emission lines and a high dust attenuation, similar to the recently discussed high-z interlopers (Naidu et al. 2022b; Zavala et al. 2023). That said, given the low S/N ratio, there remains some chance that the Haze could be associated with GN-z11, e.g., as the tidal spray from a merger, with a cessation of star formation dropping the far-UV emission.

To investigate whether the second Extended component is indeed present, we plot in Figure 4 the ForcePho posterior distribution of the fraction of F200W flux in the Extended component as a function of its half-light size (i.e., major axis radii). There is a covariance between this fraction and the size: the larger the Extended component, the less flux it contributes. Importantly, there is no tail toward zero size or zero flux, emphasizing that the Extended component is clearly detected and indeed extended. Based upon a Bayesian Information Criterion model selection, the two-component fit is preferred.

Zoom In Zoom Out Reset image size

We confirm the compact size of GN-z11 by fitting both single Sérsic and PS models to the mosaic using Lenstronomy (Birrer & Amara 2018). To ensure that we account for the extended background from the bright neighbor ∼1'' to the NE, we fit GN-z11 and the bright neighbor simultaneously using single Sérsic profiles for both galaxies. In all bands, the best-fit model for GN-z11 has an intrinsic half-light radius < 1 pixel and a Sérsic index of >7.8, which indicates a source that is not significantly resolved with respect to the ePSF. Fitting GN-z11 with a PS model instead of a Sérsic profile yields slightly lower residuals, further indicating the compact nature of the source.

Using ProFit, we analyze the structure of GN-z11 in the F277W mosaic using a multicomponent fit including the central PS, the central extended source, and the Haze. Given the number of components, a nominal ProFit model would involve optimizing 24 free parameters. Unlike our ForcePho method, in using ProFit we are limited to the information provided by a single filter in the mosaic (see the discussion above), and hence we restrict the number of free parameters to the centroids of each component, their brightnesses, the Sérsic indices, and the effective radii. The axis ratios and position angles are kept fixed at the value found in our ForcePho models, and we set the isophotal boxiness to be negligible in each case. Importantly, we find a better fit including a three component model (PS, extended component, and Haze) than for a PS + Haze two-component model. We see no evidence that the PSF adopted by ForcePho is artificially inflating the size of Extended component or the Haze because we can find independently a good quality fit for the extended component and the Haze with ProFit with an effective radius of 004 ± 001 and 011 ± 006, respectively. The Sersic indexes of both the extended component and the Haze are consistent with n = 1. Similarly, the relative brightnesses of the Extended component and the Haze to the PS are very similar to the ForcePho fit (about two magnitudes for the latter). In summary, ProFit allows us to confirm the presence of the Extended component, albeit with less constraining power than what we can do with ForcePho. This is expected given the limited amount of data used in the fitting (single band and mosaic).

We note that our modeled size of GN-z11 is noticeably smaller than the 0.6 ± 0.3 kpc half-light radius reported in Holwerda et al. (2015) fitting HST WFC3 data using a Sérsic index n = 1.5. Indeed, even at JWST resolution, the unresolved component has over half of the total light (Figure 4). We do not think that the difference can be due to the Haze because that component is very weak in F150W. Clearly the angular resolution of JWST will be very important in probing the physical sizes of these very small high-redshift galaxies (e.g., Wu et al. 2020; Ono et al. 2023; Suess et al. 2022; Tacchella et al. 2023; Costantin et al. 2023).

We analyze the SEDs with three different spectral synthesis codes, fixing the redshift to the spectroscopic redshift of 10.60 throughout. The first is Prospector (Johnson et al. 2021), which we use for the figures in this paper. Prospector computes stellar population synthesis combined with a model of nebular line and continuum emission, as well as dust attenuation, comparing to the data with a Bayesian formalism and using Markov Chain Monte Carlo to quantify the posterior. Here, we utilize a similar setup as in Tacchella et al. (2022, see also Tacchella et al. 2023; Whitler et al. 2023). Specifically, we assume a nonparametric star formation history with six time bins and a bursty continuity prior. We put them at 0–5 Myr and 5–10 Myr, and the four bins are logarithmically spaced up to z = 20. We adopt a single metallicity for both stars and gas, assuming a truncated log-normal centered on $\mathrm{log}(Z/{Z}_{\odot })=-1.5$ with width of 0.5, minimum of −2.0, and maximum of 0.0. We model dust attenuation using a two-component dust attenuation model with a flexible attenuation curve. The first component is a birth-cloud component in our model that attenuates nebular emission and stellar emission only from stars formed in the last 10 Myr (the attenuation law is a power law with a slope of −1). The second component is a diffuse component that has a variable attenuation curve and attenuates all stellar and nebular emission from the galaxy. The variation in the attenuation law is modeled as a multiplicative factor of the Calzetti et al. (2007) law to account for uncertainties in the geometry of dust extinction. For the stellar population synthesis, we adopt the MIST isochrones (Choi et al. 2016) that include effects of stellar rotation but not binaries, and assume a Chabrier (2003) initial mass function (IMF) between 0.08 and 120 M_⊙. A No Lyα emission line is added to the model to account for resonant absorption effects. This assumption might be too simplistic given the Lyα detection in the spectrum (Bunker et al. 2023). However, we estimate that this will not affect our SED fits significantly because this line contributes only at the 0.05 mag level, while we put an error floor of 5% on the photometry. The rest of the nebular emission (emission lines and continuum) is self-consistently modeled (Byler et al. 2017) with two parameters, the gas-phase metallicity (tied to the stellar metallicity), and the ionization parameter (uniform prior in $-4\lt \mathrm{log}(U)\lt -1$).

Since these data for GN-z11 fall mostly in the rest-UV, the broadband photometry is less sensitive to the presence of very strong emission lines. In particular, the strong [O iii], Hβ, and Hα lines fall redward of our wavelength range. However, some contribution from bluer lines in the F444W band is expected; this will be quantified in Bunker et al. (2023).

A key component that is not included in this modeling is the possibility of luminosity from an AGN. Clearly, the very compact morphology of GN-z11, with about two-thirds of the emission coming from an unresolved nucleus, permits this. Since the SED from an AGN component is highly flexible, we opt here to present results based on the null hypothesis that the light is dominated by stars. Clearly, a luminous AGN would decrease the inferred star formation rate and stellar mass. Whether or not GN-z11 contains a luminous AGN is a question that will be investigated with the spectroscopy Bunker et al. (2023) and Maiolino et al. (2023).

In addition to the effects of metallicity, dust attenuation, and star formation history on the stellar mass-to-light ratio, several other model assumptions may affect the stellar mass inferred from the SED (e.g., Conroy 2013). We highlight some of these, while noting that the effects on the inferred masses can be complex because other inferred parameters may compensate to predict the same SED with a similar stellar mass under different assumptions. If the IMF had fewer very low mass stars than we assume, then the inferred stellar masses would be overestimated (e.g., Steinhardt et al. 2022). The slope of the upper IMF affects the UV-optical color and mass-to-light ratio for a given star formation history. The evolution of high-mass stars at low metallicity is important to the UV and ionizing continuum of galaxies, but is not well constrained from observations in the local universe (Eldridge & Stanway 2022). Binary stellar evolution and stellar rotation may increase the lifetimes of massive UV bright stars, leading to lower mass-to-light ratios for a given star formation history (e.g., Choi et al. 2017). Furthermore, non-solar abundance ratios may be common in the early universe (e.g., Steidel et al. 2016), leading to changes in stellar evolutionary tracks and stellar SEDs at a variety of ages. Our nebular emission models may not capture the effects of complex geometries on the emergent nebular continuum and nebular line ratios (e.g., Jin et al. 2022), which is important for the shape of the Balmer break and hence mass-to-light ratios (e.g., Papovich et al. 2023). We do not include potential contributions from Population III stars.

We compared the Prospector results to those from BEAGLE (Chevallard & Charlot 2016) and BAGPIPES (Carnall et al. 2018), which provide similar functionality but differ in numerous modeling aspects. We fit the observed ForcePho flux with both BEAGLE and BAGPIPES using a delayed exponential star formation history, fixing the redshift to the spectroscopic value. For both BEAGLE and BAGPIPES, we use the Bruzual & Charlot (2003, with the 2016 update) stellar templates and Kroupa's (2001) IMF. We assume the Charlot & Fall (2000) and the Calzetti et al. (2000) dust attenuation law for BEAGLE and BAGPIPES, respectively.

We first focus on the combined photometry of the PS and the Extended Component, which we refer to as "Galaxy" GN-z11 (Table 2). We show the resulting posterior distributions of several key stellar population parameters in Figure 5, including the stellar mass, specific SFR, stellar age, dust attenuation, and stellar metallicity (see also Table 3). We find a formed stellar mass of 10^9.1±0.3 M_⊙. We find that GN-z11 is actively forming stars with ${\mathrm{SFR}}_{30\mathrm{Myr}}={21}_{-10}^{+22}\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$ (${\mathrm{SFR}}_{10\mathrm{Myr}}\,={12}_{-3}^{+10}\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$) and a specific SFR of sSFR_30Myr = 10^−7.7±0.3 yr⁻¹, indicating that this galaxy is doubling its stellar mass roughly every ∼50 Myr. (s)SFR_30Myr refers to the (specific) SFR averaged over the past 30 Myr. Consistent with this, we find that the galaxy has a stellar age (half-mass time: lookback time at which 50% of the stellar mass formed) of $\sim {24}_{-10}^{+20}$ Myr. The inset on the top right-hand side of Figure 5 shows the posterior of the star formation history. We find that the SFR has increased ∼60 Myr ago (z ≈ 12), peaked at a lookback time of 10–20 Myr, and has slightly decreased in the recent 10 Myr.

Zoom In Zoom Out Reset image size

Table 3. Properties of GN-z11

Property	Galaxy	PS	Extended
Stellar mass $\mathrm{log}({M}_{\star }/{M}_{\odot })$	${9.1}_{-0.4}^{+0.3}$	${8.4}_{-0.3}^{+0.3}$	${8.9}_{-0.3}^{+0.2}$
Observed MUV (AB mag)	$-{21.58}_{-0.04}^{+0.03}$	$-{21.10}_{-0.03}^{+0.04}$	$-{20.48}_{-0.05}^{+0.05}$
Intrinsic MUV (AB mag)	$-{21.79}_{-0.47}^{+0.24}$	$-{21.08}_{-0.29}^{+0.23}$	$-{20.73}_{-0.45}^{+0.21}$
UV continuum slope β	$-{2.41}_{-0.07}^{+0.06}$	$-{2.48}_{-0.11}^{+0.08}$	$-{2.43}_{-0.09}^{+0.09}$
SFR (M⊙ yr–1)	${21}_{-10}^{+22}$	${6}_{-2}^{+3}$	${9}_{-4}^{+7}$
$\mathrm{log}(\mathrm{sSFR})$ (yr−1)	$-{7.7}_{-0.3}^{+0.2}$	$-{7.5}_{-0.4}^{+0.1}$	$-{7.9}_{-0.3}^{+0.3}$
Stellar age t⋆ (Myr)	${24}_{-10}^{+20}$	${11}_{-7}^{+58}$	${35}_{-19}^{+15}$
Attenuation AV (mag)	${0.08}_{-0.06}^{+0.23}$	${0.05}_{-0.05}^{+0.08}$	${0.09}_{-0.07}^{+0.26}$
Half-light size Re (arcsec)	0.016 ± 0.005	Point-like	0.049 ± 0.003
Half-light size Re (pc)	64 ± 20	Point-like	196 ± 12

Notes. The stellar population properties quoted here (median and 16th–84th percentiles) come from the SED modeling with Prospector of the combined photometry of the PS and Extended component (i.e., "Galaxy" in Table 2), the PS photometry and the Extended component photometry. The average half-light size (and its standard deviation) for the combined GN-z11 is derived by flux-weighting the sizes of both the PS and Extended models.

Download table as:  ASCIITypeset image

These Prospector-inferred parameters are in overall good agreement with the parameters inferred with BEAGLE and BAGPIPES. Specifically, using BEAGLE with a parametric star formation history, we find log(M_⋆,BEAGLE/M_⊙) = 8.9 ± 0.1, SFR_10Myr= 22 ± 5 M_⊙ yr^–1, and attenuation A_V = 0.09 ± 0.08 mag. With BAGPIPES, we obtain log(M ${}_{\star ,\mathrm{BAGPIPES}}/{M}_{\odot })={9.3}_{-0.8}^{+0.1}$, SFR_10Myr= 16 ± 6 M_⊙ yr^–1, and attenuation ${A}_{{\rm{V}}}={0.14}_{-0.07}^{+0.03}$ mag. All inferred parameters are consistent with previous inferences in Jiang et al. (2021), who used a single stellar population to model the broadband HST+Spitzer photometry. Our results are also within ∼1σ of those obtained in Bunker et al. (2023) by fitting the NIRSpec spectroscopy with BEAGLE. We consider this to be a good agreement given the differences in the parameterization of the star formation histories and associated priors (e.g., Leja et al. 2019; Whitler et al. 2023).

Figure 6 shows the Prospector results for the two components of GN-z11 (PS and Extended) individually. The top panels are the results for the PS, while the bottom panels are for the Extended Source. The individual component (PS and Extended) stellar population parameters agree between Prospector, BEAGLE, and BAGPIPES.

Zoom In Zoom Out Reset image size

The SEDs of those two components show an important difference around the observed wavelength of 4 μm, which is around the rest-frame Balmer jump. The PS shows a blue F410M–F444W color and a red F356W–F410M color, indicating strong nebular emission. In contrast, the Extended component shows a red F410M–F444W and blue F356W–F410M color, consistent with less strong nebular emission and weak Balmer/4000 Å break. From the ForcePho Markov Chains, we find that the F410M–F444W color of the Extended component is redder than that of the PS in 95% of cases, with a median color difference of 0.7 mag. Not surprisingly, Prospector prefers a younger age and an increasing star formation history for the central PS, while the Extended Source is consistent with a more extended star formation history (even decreasing in the recent 10 Myr). The Extended Source is slightly more massive than the PS (10^8.9±0.3 M_⊙ versus 10^8.4±0.3 M_⊙), even though this component is only one-third of the combined rest-UV flux.

In summary, the SED modeling presented here does not reveal a strong support for the AGN scenario. We find that that the central emission can be powered by intense and compact star formation, thereby outshining the underlying extended emission of the galaxy. However, we cannot rule out the presence of a luminous AGN, and we expect that detailed analysis of the NIRSpec spectroscopy (Bunker et al. 2023 and Maiolino et al. 2023) will be needed to fully explore this scenario.