A Size Estimate for Galaxy GN-z11

First identified as GNS-JD2 in the Hubble Space Telescope (HST)/NICMOS data, GN-z11 was initially considered a possible dropout candidate due to its detection being limited to the 1.6 μm wavelength (Bouwens et al. 2010). Its proximity to another source made further conclusions difficult until 2014 when Oesch et al. (2014) classified it (under the designation GN-z10-1) as a z ∼ 9–10 galaxy, and later determined its redshift, through grism spectroscopy, to be $z={11.09}_{-0.12}^{+0.08}$ (Oesch et al. 2016), a measurement more accurately constrained with more recent NIRSpec data to be z = 10.603 (Bunker et al. 2023). One of the most distant galaxies observed with HST, GN-z11 is located in the CANDELS DEEP area in GOODS-North at (R.A., decl.) = (12:36:25.46, +62:14:31.4).

Holwerda et al. (2015) measured the structure of a sample of z ∼ 9–10 galaxies, including GN-z11 (designated GN-z10-1 in their paper), using the 2D fitting algorithm galfit (Peng et al. 2002, 2010). The two main drawbacks of galfit when determining the physical properties of galaxies at the distance and luminosity of GN-z11 are (a) the lack of posterior distributions complicate the interpretation of the uncertainty and (b) the software is not optimized for faint, distant objects. Holwerda et al. (2015) report a size for GN-z11 of 0.6 ± 0.3 kpc. The uncertainty reported is statistical, however, and based on an estimate of models with similar estimated structural properties (see Ono et al. 2013 for a description of galfit uncertainties and biases). Further, their measurement was obtained by fixing the Sérsic index, n, and the axis-ratio, q, to limit the statistical uncertainty. These fixed parameter models have allowed for initial estimations of the size of GN-z11, but may suffer from systematic uncertainties from modeling choices. Here, we build on previous work with galfit using a new Bayesian size-fitting software, forcepho, which is optimized for faint, blended objects and provides robust uncertainties (Johnson et al. 2024, in preparation).

GN-z11 was imaged using the WFC3/IR camera aboard HST. Existing imaging of the GOODS-N field from the CANDELS survey consisted of ∼5 orbits covering the central ∼65 arcmin² of the GOODS-N field (Oesch et al. 2014). 7 additional orbits in F160W were obtained as a result of a mid-cycle proposal for a total of 12 orbits, reaching 27.7 mag (5σ) (PI:Oesch, PID: 15977). See Oesch et al. (2014) and Illingworth et al. (2013) for a description of the data reduction.

We fit the structural parameters of GN-z11 using forcepho, which infers the fluxes and morphological parameters of galaxies via a Bayesian analysis (Johnson et al. 2024, in preparation). Models are generated for our sources which are compared to the observations via a likelihood function as the posterior is sampled. Crucially, forcepho is differentiable, meaning that gradients of the likelihood function can be taken, allowing for significant optimization in the posterior sampling. Additionally, forcepho utilizes Gaussians to approximate the point spread functions (PSFs)—constructed empirically from a stack of point sources in the HLF imaging—and the Sérsic profiles. The sums of these Gaussian parameters are used for convolution, which greatly increases efficiency. We use forcepho to measure the size and other structural parameters of GN-z11 in the F160W band. In addition to the half-light radius, we also infer Sérsic index and the axis-ratio, not setting them to fixed values as in Holwerda et al. (2015). The major source of uncertainty in our measurements come from the similarity of the galaxy size to the size of the PSF.

We measure a half-light radius of 0036 ± 0006 which corresponds to a physical size of 0.15 ± 0.025 kpc, assuming a redshift of z = 10.603 as measured by Bunker et al. (2023). In Figure 1 (top), we show the image, model, and residual (data—model) showing that the forcepho model is a good match to the data. We infer a flux in F160W of 139.16 ± 8.98 nJy, less than the initial Oesch et al. (2016) finding of 152 ± 10 nJy, but within the range of uncertainty. These measurements differ significantly from those obtained by Holwerda et al. (2015) (0.6 ± 0.3 kpc), finding of a much smaller half-light radius suggestive of a point source. The mean Sérsic index was measured to be 2.42 ± 1.12, and the axis-ratio was measured to have a mean of 0.91 ± 0.07. These measurements are higher than those found by more recent analysis which finds (n = 0.9 ± 0.1, q = 0.67 ± 0.05) using forcepho on 0.9–4.4 μm JWST/NIRCam imaging (Tacchella et al. 2023). This could be due to the increased spatial resolution and signal-to-noise ratio of the JWST observations or an intrinsically different shape in the longer wavelength bands. Figure 1 (bottom) shows the posteriors from the Markov Chain Monte Carlo sampling, including the mean values for each of these values.

Zoom In Zoom Out Reset image size

In summary, we performed a new size-measurement for GN-z11, the highest redshift galaxy spectroscopically confirmed by HST, using the Bayesian fitting software forcepho. Our physical size of 0.15 ± 0.1 kpc is much smaller than previous measurements with galfit, but consistent with measurements based on newer JWST/NIRCam data (Tacchella et al. 2023). The size of GN-z11, along with its redshift and luminosity, present interesting possibilities for the physical processes at play in such an early galaxy (Bunker et al. 2023; Tacchella et al. 2023). Our measurement of a very compact size in GN-z11 is consistent with WST/NIRSpec observations (Maiolino et al. 2023) suggesting that at least some of the remarkable luminosity of this object may be driven by an active galactic nucleus.

This work is based on observations taken by the 3D-HST Treasury Program (GO 12177 and 12328) with the NASA/ESA HST, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support from HST-GO-15977, HST-GO-13871, and the CANDELS survey (Faber 2011) survey is gratefully acknowledged.