betaFeedback

WorksPhysics and Astronomy

The ultraviolet continuum slopes of galaxies at z≳12\mathbf{z \gtrsim 12} selected from 0.6 deg2{\mathbf{0.6 \, deg}^2} of JWST imaging

Abstract

We present an analysis of the ultraviolet (UV) continuum slopes (β)(β) of 5555 galaxies at z≳12z\gtrsim 12, selected from a wide-area (>0.6>0.6 sq. deg) search across publicly available JWST/NIRCam imaging. Our sample spans the absolute UV magnitude (MUVM_{\mathrm{UV}}) range −21.6≤MUV≤−17.3-21.6 \leq M_{\mathrm{UV}} \leq -17.3 and we find an inverse variance-weighted mean ⟨β⟩=−2.57±0.04\langleβ\rangle = -2.57\pm0.04 indicating that, on average, the galaxy population is extremely blue at z≳12z\gtrsim 12. The average ββ at these redshifts is consistent with the intrinsic stellar plus nebular continuum limit in the absence of UV dust-attenuation. Consistent with earlier JWST works, our analysis suggests that the moderately dust-attenuated galaxies at z≲10z \lesssim 10 become essentially attenuation-free by z≃11z \simeq 11. We find no evidence for a β−MUVβ-M_{\mathrm{UV}} relation in our z≳12z\gtrsim 12 sample, indicating that the attenuation-free interpretation applies across all observed UV magnitudes. By augmenting our sample with previous JWST constraints down to z=9z=9, and incorporating MUVM_{\mathrm{UV}}-matched HST studies down to z=4z=4, we find a smooth evolution in β(z)β(z) over a wide range in redshift. The average value of ββ becomes progressively bluer at z>4z>4, with dβ/dz≃−0.1\mathrm{d}β/\mathrm{d}z \simeq -0.1. Finally, although we find that the typical galaxy at z≳12z \gtrsim 12 is consistent with the attenuation-free limit, the scatter in our individual ββ measurements is consistent with the existence of a small population of dust-reddened galaxies with β≳−2β\gtrsim -2.

Introduction

With the launch of JWST, the redshift frontier of galaxy evolution has been pushed out to z≃15z \simeq15 (e.g. Carniani et al. 2024; Naidu et al. 2026; Donnan et al. 2026b), and the detection of reliable z>10z > 10 galaxy candidates is becoming commonplace thanks to the increasing area covered by deep JWST/NIRCam imaging (Donnan et al. 2024; McLeod et al. 2024). The large sample of known z>10z > 10 galaxies is yielding crucial insights into the evolution of the population demographics via measurements of the ultraviolet (UV) luminosity function (LF) (see, e.g. Finkelstein et al. 2024; Weibel et al. 2026; McLeod et al. 2026).

Beyond the UVLF, deep JWST/NIRCam photometry can also be used to determine the shape of the UV continuum in these ultra-high-redshift sources, characterized by the UV continuum slope, β\beta, where fλ∝λβf_\lambda\propto\lambda^\beta for λrest≃0.12−0.3 μm\lambda_{\mathrm{rest}} \simeq0.12 - 0.3\,\mu\mathrm{m} (e.g. Calzetti et al. 1994; Leitherer et al. 1999). The UV continuum shape encodes information about the intrinsic emission from massive stars (dependent on their age and metallicity), the nebular continuum emission from star-forming regions, and the degree of UV dust attenuation (e.g. Topping et al. 2022; Cullen et al. 2024).

Stellar population models predict that the youngest, most metal-poor stellar populations set the so-called ‘blue limit’ for the intrinsic stellar UV continuum slope at β≃−3.2\beta\simeq-3.2 (e.g. Stanway et al. 2016; Topping et al. 2022). Nebular continuum emission, dust attenuation, and increases in stellar age or metallicity all act to redden the UV slope (i.e. drive β\beta to less negative values). At the highest redshifts, the effects of age and metallicity become subdominant (Tachella et al. 2022), and the UV continuum is, to a good approximation, sensitive only to dust reddening and the strength of nebular continuum emission (e.g. Cullen et al. 2024; Katz et al. 2025). Many studies have now shown that for single-age stellar populations in the absence of dust, the blue limit when including the nebular continuum contribution is β≃−2.6\beta\simeq-2.6 (Topping et al. 2022; Cullen et al. 2024)1. Galaxies with β\beta at or bluer than this limit can be considered to be free of UV dust attenuation (Cullen et al. 2024). Furthermore, galaxies with β\beta approaching the intrinsic stellar limit (i.e. β<−2.8\beta< -2.8) must necessarily also have a high ionising-photon escape fraction (fescf_{\mathrm{esc}}) to account for the absence of nebular continuum emission (e.g. Chisholm et al. 2022; Saxena et al. 2026; although see Katz et al. 2026 for another possible physical interpretation).

The picture is somewhat complicated by the existence of an extreme class of nebular-continuum-dominated galaxies that can exhibit red UV slopes (β≃−2\beta\simeq-2) and mimic dust-reddened galaxies despite being dust-poor (e.g. Cameron et al. 2024). In these systems, the nebular continuum is so strong that the UV spectrum is dominated by two-photon emission from a highly-ionized interstellar medium. Although the existence of such galaxies has been convincingly demonstrated (e.g. Katz et al. 2025), their frequency among the overall galaxy population remains uncertain. Fortunately, since these nebular-dominated galaxies exhibit red β\beta values, they are easily distinguishable from extremely dust-poor galaxies with β≲−2.6\beta\lesssim-2.6 that are the primary focus of this work.

Prior to the launch of JWST, significant effort was made to establish the evolution of typical β\beta values in the star-forming galaxy population from z≃2z \simeq2 to z≃8z \simeq8. At all redshifts, a correlation was found between β\beta and absolute UV magnitude (MUVM_{\mathrm{UV}}) such that galaxies were observed to become younger and more dust-poor towards fainter magnitudes (e.g. Rogers et al. 2014). Simultaneously, at fixed MUVM_{\mathrm{UV}}, galaxies became increasingly blue towards higher redshifts, reaching average UV continuum slopes of ⟨β⟩≃−2.1\langle\beta\rangle\simeq-2.1 at z=8z = 8 for galaxies with MUV≃−18M_{\mathrm{UV}} \simeq-18 (e.g. Finkelstein et al. 2012; Wilkins et al. 2012; Dunlop et al. 2013; Bouwens et al. 2014). Since the launch of JWST, the evolution of the UV continuum slope has now been studied out to z≃12z \simeq12 using large photometric samples (e.g. Topping et al. 2022, 2024; Cullen et al. 2023, 2024; Morales et al. 2024; Austin et al. 2025, 2026; Felicioni et al. 2026; Hainline et al. 2026). Spectroscopic samples are now also starting to be amassed beyond z>10z > 10 (including 10 at z>12z > 12), making it possible to correlate β\beta to other physical properties (e.g. Roberts-Borsani et al. 2024; Saxena et al. 2026; Pollock et al. 2026; Roberts-Borsani et al. 2026; Tang et al. 2026). The consensus of these studies is that ⟨β⟩\langle\beta\rangle continues to become bluer at fixed magnitude at z>8z > 8, continuing the trend observed at lower redshifts. The evolution of the form of the β−MUV\beta- M_{\mathrm{UV}} relation with redshift is less certain, but with growing evidence for a flattening (i.e. no relation) at z>10z > 10 (Saxena et al. 2026; Topping et al. 2024).

One important result that has emerged from these JWST studies is that by z≃11z \simeq11 the average UV continuum slope of the observed galaxy population with MUV≲−18M_{\mathrm{UV}} \lesssim-18 is consistent with the theoretical attenuation-free stellar plus nebular limit of β=−2.6\beta= -2.6 (e.g. Cullen et al. 2024; Topping et al. 2024). This compares to typical values of ⟨β⟩≃−2.2\langle\beta\rangle\simeq-2.2 to −2.3-2.3 at z≃9z \simeq9, just ≃150 Myr\simeq150\,\mathrm{Myr} later (Cullen et al. 2024; Topping et al. 2024).

A clear transition to a bluer population on average at z>10z > 10 is also evident in spectroscopic samples, although recently several redder objects have also been reported (e.g. Mitsuhashi et al. 2026; Rodighiero et al. 2026). The typical UV slopes of spectroscopic samples also tend to be slightly redder than those derived from photometry (e.g. Saxena et al. 2026; Roberts-Borsani et al. 2026).

An obvious implication of these results is that at z>10z > 10 a significant fraction of the galaxy population is unattenuated, which, for now, is also corroborated by the fact that there are still no confirmed dust continuum detections at z>9z > 9 (e.g. [8] (Kaasinen et al. 2023; Popping 2023; Yoon et al. 2023; Schouws et al. 2025)), whereas multiple detections have been achieved at z=8−9z = 8-9 (e.g. [7, 9] (Laporte et al. 2017)). The lack of significant dust attenuation at these redshifts is obviously relevant to the debate concerning the number of UV-bright galaxies being observed at z>10z > 10 with JWST, and has implications for the number of ionising photons escaping into the intergalactic medium (IGM) at the beginning of the reionization era (e.g. Muñoz et al. 2024).

The physical explanation for a lack of dust attenuation in these ultra-blue high-redshift systems is still being debated, but some proposed ideas include dust being efficiently removed from early compact galaxies via supernova-driven winds (e.g. Ferrara et al. 2023; Fiore et al. 2023; Ziparo et al. 2023; Martínez-González et al. 2026a), an evolution in the physical properties of dust grains (specifically on average larger grains at high redshift; Narayanan et al. 2025, 2026), a porous ISM (Sommovigo et al. 2026) and dust destruction (Esmerian & Gn edin 2024; Garaldi et al. 2026). It has also been suggested that the average stellar mass of the population simply moves below the mass threshold at which dust attenuation is observed to become negligible at lower redshifts (Donnan et al. 2025a). The debate remains unsettled, but it is nevertheless clear that constraints on the evolution of β\beta contain important information relating to the build up of dust and metals at early times.

Until now, most studies of UV continuum slopes have been limited to galaxies at z<12z < 12 (although see Hainline et al. 2026; Roberts-Borsani et al. 2026). This is primarily due to the larger areas needed to compensate for the decreasing number density of sources with increasing redshift. However, the observable galaxy population is known to extend to at least z≃14.5z \simeq14.5 (Naidu et al. 2026), and it is therefore important to investigate the evolution of β\beta out to this redshift frontier. Fortunately, with the amount of deep JWST/NIRCam imaging continually increasing, this analysis is now possible. In this paper, we examine the UV continuum slopes of 55 galaxies at z≳12z \gtrsim12 selected from >0.6 deg2>0.6\,\mathrm{deg}^{2} of JWST imaging. The sample analysed here was first presented in McLeod et al. 2026 and used to constrain the evolution of the UV luminosity function. Here, our aim is to use this sample to extend the analysis in Cullen et al. 2024 and investigate whether UV slopes continue to plateau at the β=−2.6\beta= -2.6 limit set by nebular physics.

The paper is structured as follows. In § 2 we describe our selection procedure and the properties of the final sample. In § 3 we present our main results, specifically the evolution of β\beta with redshift and MUVM_{\mathrm{UV}}. Finally, we discuss the implications of our results and present comparison with other literature analyses in § 4 and summarise our main results in § 5. Throughout we use the AB magnitude system (Oke 1974; Oke & Gunn 1983), and assume a standard cosmological model with H0=70 km s−1 Mpc−1H_{0} = 70\,\mathrm{km\,s^{-1}\,Mpc^{-1}}, Ωm=0.3\Omega_{\mathrm{m}} = 0.3 and ΩΛ=0.7\Omega_{\Lambda} = 0.7.

Sample Selection and Properties

The present study focuses on a sample of z≳12z \gtrsim12 galaxies that was first presented in McLeod et al. 2026. These galaxies were selected over an extensive, wide-area search, encompassing >0.6 deg2>0.6\,\mathrm{deg}^{2} and over ≃150\simeq150 sight-lines, enabling us to produce a large and robust z≳12z \gtrsim12 galaxy sample while mitigating the effects of cosmic variance. We refer interested readers to McLeod et al. (2026) for a complete description of the catalogue construction and high-redshift galaxy selection, but give a summary of the most relevant details below.

The z ≳ 12 sample

Across each of the fields, multi-wavelength photometry catalogues were produced with Source Extractor (Bertin & Arnouts (1996)) in dual-image mode. These catalogues were constructed using three different detection images: F200WF200W, F277WF277W and F356WF356W. The initial catalogues were culled such that we only considered F150WF150W dropout objects—i.e. with a signal-to-noise ratio (SNR) of SNR ≤2\leq2 for F150WF150W and all bluer filters—which also had SNR ≥8\geq8 in at least one of either the F200WF200W or F277WF277W images. As an additional step to exclude obvious artefacts, we also required SNR ≥3\geq3 in at least one other band redward of the Lyman break. Aperture photometry was performed in 0.2′′0.2^{\prime\prime}-diameter apertures on images that were homogenized to match the resolution of F444WF444W. The photometric uncertainties were calculated source-by-source as an aperture-to-aperture root mean square of the nearest 200 blank-sky apertures, as determined via a dilated segmentation map (see, e.g., Begley et al. (2025) for a full description of this method).

Photometric redshifts were estimated with the spectral energy distribution (SED) fitting code LePhare (Arnouts & Ilbert (2011)), using the BC03 template library (Bruzual & Charlot (2003)), assuming a Chabrier (2003) initial mass function and a Calzetti et al. (1994) attenuation law. Objects with zphot≥11.5z_{\mathrm{phot}} \geq11.5 and a Δχ2≥4\Delta\chi^{2} \geq4 between the primary high-redshift solution and the secondary low-redshift solution were retained, and subject to rigorous visual inspection to remove any remaining artefacts or lower redshift interlopers. The result of this process was a final sample of 55 galaxies, within the redshift range 11.5≤zphot≤15.211.5 \leq z_{\mathrm{phot}} \leq15.2 and absolute UV magnitude range −21.6≤MUV≤−17.3-21.6 \leq M_{\mathrm{UV}} \leq-17.3 (Figure 1; McLeod et al. (2026)).

Distribution of UV absolute magnitude and photometric redshift for the high-redshift galaxy sample

Figure 1. The distribution of UV absolute magnitude MUVM_{\mathrm{UV}} and photometric redshift zz for our high redshift galaxy sample. The purple diamonds show our new sample of 55 galaxy candidates at z≳12z \gtrsim12 selected from >0.6> 0.6 deg2^{2} of deep JWST/NIRCam imaging (see McLeod et al. (2026) for a full description of this sample). The grey triangles show a complementary sample of galaxies at 9<z<129 < z < 12 taken from our previous analysis (Cullen et al. (2024)). Our new z≳12z \gtrsim12 sample spans the redshift range 11.5≤z≤15.211.5 \leq z \leq15.2 and absolute UV magnitude range −21.6≤MUV≤−17.3-21.6 \leq M_{\mathrm{UV}} \leq-17.3.

Ancillary 9 < z < 12 sample

In addition to our new z≳12z \gtrsim12 sample, we also make use of the complementary z≳9z \gtrsim9 galaxy sample presented in our previous work that was selected following a very similar methodology (Cullen et al. (2024)). This sample was selected from a search for z>9z > 9 galaxies across ≃320\simeq320 sq. arcmin and fifteen independent datasets and was an extension of the z>9z > 9 galaxy sample first presented in McLeod et al. (2024). Photometry was measured in 0.35′′0.35^{\prime\prime}-diameter apertures on imaging that was PSF-homogenized to match the resolution of F444WF444W. Similar to the present study, we adopted a stringent SNR ≥8\geq8 criterion in at least one of the F150W, F200W or F277W images, with a requirement SNR ≤2\leq2 at bluer wavelengths. The SED fitting setup with LePhare and candidate selection procedure was identical to the one described above. This sample consists of 121 sources selected with zphot≥9.0z_{\mathrm{phot}} \geq9.0. Barring a small number of z>12z > 12 candidates (see Figure 1), this sample is practically confined to the redshift range 9<z<129 < z < 12 and absolute UV magnitude range −20.5<MUV<−17.0-20.5 < M_{\mathrm{UV}} < -17.0. We refer the interested reader to Cullen et al. (2024) for a full description of the data set and sample selection.

We remove two galaxies from the Cullen et al. (2024) sample that, when including new data, are no longer consistent with being z>9z > 9 galaxy candidates (IDs JADES-1058823 and NGDEEP-1003756 from table 1 of Cullen et al. (2024)). We also remove a further four galaxies that are duplicated in our new selection (IDs GLASS-9974, GLASS-28072, CEERS-37967, and JADES-9996) from table 1 of Cullen et al. (2024). This leaves a total of 170 sources at z≥9z \geq9; the full distribution in zz and MUVM_{\mathrm{UV}} is shown in Figure 1. It can be seen that with these new z≳12z \gtrsim12 candidates, we can naturally extend our previous analysis of UV continuum slopes at 9<z<129 < z < 12 (Cullen et al. (2024)). Both samples span a very similar range in MUVM_{\mathrm{UV}}, although the factor ≃7\simeq7 increase in area means that we have been able to find a large number of bright candidates (MUV≲−21M_{\mathrm{UV}} \lesssim-21) in the new z≳12z \gtrsim12 search.

Measuring the UV continuum slope

To fit the UV continuum slopes for our sample we follow the method outlined in Cullen et al. (2024) with minor adjustments. Briefly, the UV continuum is modelled as a power law following fλ∝λβf_{\lambda} \propto\lambda^{\beta} for λrest≤3000\lambda_{\mathrm{rest}} \leq3000 Å (Calzetti et al. (1994)). The effect of the intergalactic medium (IGM) is accounted for by setting all fλf_{\lambda} at λrest≤1216\lambda_{\mathrm{rest}} \leq1216 Å to zero and incorporating the Lyα\alpha damping wing using equation 2 of Miralda-Escudé (1998). Given the extreme redshifts of our objects (z≳12z \gtrsim12), we choose to fix the fraction of neutral hydrogen in the surrounding IGM (xHIx_{\mathrm{HI}}) to 1.

The model therefore has two free parameters: the UV continuum slope (β\beta) and the redshift (zz). To fit this model to the photometric data we use the nested sampling algorithm dynesty (Speagle (2020); Koposov & Speagle (2026)). For the UV continuum slope, we assume a flat prior in the range −10≤β≤10-10 \leq\beta\leq10. For the redshift prior, we sample from the individual redshift probability distributions, p(z)p(z), returned from fitting to the full SED. Since LePhare does not include a Lyα\alpha damping prescription, to obtain a more accurate p(z)p(z) we refit the SEDs using a custom fitting code, described in McLure et al. (2011), that has been modified to account for Lyα\alpha damping. The best-fitting parameter values returned by the dynesty fits are used throughout our analysis unless otherwise specified. The reported uncertainties are estimated using the 16th and 84th percentiles of the posterior probability distribution for each variable. We also adopt a minimum error of σβ=0.25\sigma_{\beta} = 0.25 for the UV continuum slope, following the discussion in Cullen et al. (2024). Some examples of fits to photometric data can be found in Figure 2 (purple dashed lines).

Photometric and spectroscopic UV continuum model fits for six galaxies and comparison of spectroscopic and photometric UV slopes

Figure 2. The first six panels show examples of our photometric (βphot\beta_{\mathrm{phot}}) and spectroscopic (βspec\beta_{\mathrm{spec}}) UV continuum model fits to the six galaxies in our sample with available JWST/NIRSpec PRISM spectra at z>12z > 12 (see § 2.5). The black lines show the JWST/NIRSpec PRISM spectra with the grey shading illustrating the 1σ1\sigma uncertainties on the flux densities. The white diamonds show the photometry for each object obtained from JWST/NIRCam imaging data. Our best-fitting models obtained when fitting to the spectroscopic data are shown by the solid teal lines, with the corresponding best-fitting models fitted to the photometric data shown as dashed purple lines. The best-fitting values of βspec\beta_{\mathrm{spec}} and βphot\beta_{\mathrm{phot}} are given in the legend of each panel. The final panel (bottom-right) shows a plot of βspec\beta_{\mathrm{spec}} versus βphot\beta_{\mathrm{phot}} with the z≳12z \gtrsim12 galaxies shown as dark-grey diamonds. The light grey circles show an ancillary sample of z≥9z \geq9 galaxies from Cullen et al. (2024). The pink square is the inverse variance-weighted average across all of these objects. Overall, we find good agreement between the averages with ⟨βspec⟩=−2.32±0.05\langle\beta_{\mathrm{spec}}\rangle= -2.32 \pm0.05 and ⟨βphot⟩=−2.41±0.08\langle\beta_{\mathrm{phot}}\rangle= -2.41 \pm0.08.

It is worth noting that we do not fix the redshift to the photometric redshift from the full SED fitting; the redshift values and uncertainties used throughout this paper are the ones obtained with our own model fits. We do this in order to marginalise over the redshift uncertainties consistent with our assumed model. However, we have verified that fixing the redshift to the best-fitting SED value does not change our results (see Figure A1). The median difference in β\beta between our method (allowing zz to vary) and a fixed-zz approach is Δβ=−0.06\Delta\beta= -0.06, with the median shifted difference being Δz=zdnesty−zSED=0.06\Delta z = z_{\mathrm{d}n\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{y}} - z_{\mathrm{S}\mathrm{E}\mathrm{D}} = 0.06.

UV continuum slope as a function of redshift, showing This Work, Cullen et al. (2024), Weighted Average, and the Stellar + Nebular Limit

Figure A1. The same as Figure 3 but with redshifts fixed to the redshifts obtained with the SED code in McLure et al. (2011). The trend seen in Figure 3 remains the same, with galaxies on average becoming blue (β∼−2.6\beta\sim-2.6) at z≳11z \gtrsim11.

Finally, although a number of sources in our sample have confirmed spectroscopic redshifts (see § 2.5) we do not fix the redshift for these sources either, so as not to bias our weighted means to galaxies with the most accurate constraints. As we will discuss in § 2.5 below, we show via a comparison between the photometric and spectroscopic β\beta measurements that, although on a source-by-source basis the photometric estimates do not recover the correct UV slope, the weighted mean of a sample of individual measurements, ⟨βphot⟩\langle\beta_{\mathrm{phot}}\rangle, should be fully consistent with ⟨βspec⟩\langle\beta_{\mathrm{spec}}\rangle.

Absolute UV Magnitudes

To determine absolute UV magnitudes and their corresponding uncertainties, we sample models from the full posterior distribution and integrate each through a top-hat filter centred on λrest=1500\lambda_{\mathrm{rest}} = 1500 Å with a width of 100 Å. The uncertainties are estimated using the 16th and 84th percentiles of the resulting MUVM_{\mathrm{UV}} distribution. The MUVM_{\mathrm{UV}} values are corrected to total by scaling to the FLUX_AUTO measurement from Source Extractor with an additional 10 per cent correction, to account for light outside the Kron aperture (Kron (1980)). For any object in a lensed field, we further correct by the magnification factor (μ\mu) using the relevant lensing map as referenced in Table C. Only a small number of our objects are lensed: five have a modest magnification (μ<2\mu< 2) and one is moderately lensed (μ≃3.2\mu\simeq3.2).

Spectroscopic control sample

The results presented in this study are based on an analysis of broad-band and medium-band photometry of photometrically-selected galaxy candidates at z>9z > 9. The main advantage of this approach is that it enables the analysis of a large galaxy sample with a well-defined selection function. However, one obvious drawback of a photometric analysis is that individual UV continuum slope measurements (βphot\beta_{\mathrm{phot}}) are much less accurate than can be achieved with spectroscopy (βspec\beta_{\mathrm{spec}}) due to the significantly decreased spectral resolution of photometric data. Fortunately, a small subset of our photometric sample has JWST/NIRSpec PRISM data available, which allows us to assess the accuracy of the photometric measurements. Specifically, we can check that, despite the large uncertainties of individual βphot\beta_{\mathrm{phot}} measurements, they can still accurately recover the average β\beta values of a given galaxy sample.

The basic methodology for measuring βspec\beta_{\mathrm{spec}} from JWST/NIRSpec PRISM is the same as for measuring βphot\beta_{\mathrm{phot}}. For fitting to the spectroscopic data the model is smoothed and resampled to match the resolution and wavelength sampling of the NIRSpec/PRISM data and compared on a pixel-by-pixel basis (see Cullen et al. (2024)). To minimize the effect of emission lines on our βspec\beta_{\mathrm{spec}} measurement, we mask the regions contaminated by strong emission or absorption line features using the Calzetti et al. (1994) windows. For one galaxy (JADES-GS-z13-1-LA; Witstok et al. (2025)) we additionally mask out the strong Lyα\alpha line that is present in its spectrum.

Out of the 55 objects in our z≳12z \gtrsim12 sample, 6 have available JWST NIRSpec/PRISM spectra: GHZ2/GLASS-z12 (Castellano et al. 2024), JADES-GS-z12-0 (Curtis-Lake et al. 2023), JADES-GS-z13-1-LA (Witstok et al. 2025), PAN-z14-1 (Donnan et al. 2026b), JADES-GS-z14-1 (Carniani et al. 2024), and MoM-z14 (Naidu et al. 2026). All of these, except PAN-z14-1, are publicly available and fully reduced on the DAWN JWST Archive (DJA; see Heintz et al. 2024; de Graaff et al. 2025, for descriptions of the reduction process). PAN-z14-1 was reduced as described in Arrabal Haro et al. (2023) with modifications as described in Donnan et al. (2026b). The model fits to these six objects are shown in Figure 2 as teal lines.

We note that there are two other known zspec>12z_{\rm spec} > 12 galaxies that are not in our sample: JADES-GS-z14-0 (Carniani et al. 2024) and JADES-GS-z13-0 (Curtis-Lake et al. 2023). The former is not selected due to its close proximity to a nearby low redshift foreground galaxy, while the latter is too faint to pass our photometric selection criteria, specifically the requirement of SNR ≥8\ge8 in at least one of either the F200WF200W or F277WF277W images.

To expand the βphot\beta_{\rm phot}–βspec\beta_{\rm spec} comparison we also incorporate objects from the ancillary Cullen et al. (2024) sample. Of the 115 objects in this sample, 22 have JWST NIRSpec/PRISM spectra publicly available on the DJA. For three of these spectra, we were unable to obtain a successful fit and they were removed from the sample. In total, therefore, we have a sample of N=25/170N = 25/170 galaxies with measurements of both βphot\beta_{\rm phot} and βspec\beta_{\rm spec}, spanning the full redshift range of our analysis (9<z<169 < z < 16).

Figure 2 shows our model fits to both the photometric data (dashed purple line) and spectroscopic data (solid teal line) for the 6 objects with JWST/NIRSpec PRISM spectra at z>12z > 12. The bottom right-hand panel of Figure 2 shows the comparison between βspec\beta_{\rm spec} and βphot\beta_{\rm phot} for all 25 galaxies with spectroscopic measurements, with the pink square showing the inverse variance-weighted average of all of the individual measurements. Overall, we find good agreement between βspec\beta_{\rm spec} and βphot\beta_{\rm phot} with 54 per cent of the values being consistent within the 1σ1\sigma uncertainties, and all values within 2σ2\sigma uncertainties.

The two weighted averages are also consistent at the ≃1σ\simeq1\sigma level, with ⟨βspec⟩=−2.32±0.05\langle\beta_{\rm spec}\rangle= -2.32 \pm0.05 and ⟨βphot⟩=−2.41±0.08\langle\beta_{\rm phot}\rangle= -2.41 \pm0.08, indicating that our photometric model is a good predictor of the UV continuum slope, especially when averaging over a large sample size. If we impose a minimum uncertainty floor of σβ=0.1\sigma_\beta= 0.1 on the spectroscopic measurements to avoid the average being strongly biased by higher SNR spectra, the weighted average moves closer to the photometric value, ⟨βspec⟩=−2.35±0.03\langle\beta_{\rm spec}\rangle= -2.35 \pm0.03. In our previous work (Cullen et al. 2024) we also found an offset of 0.06 between ⟨βspec⟩\langle\beta_{\rm spec}\rangle and ⟨βphot⟩\langle\beta_{\rm phot}\rangle in the same direction (i.e. ⟨βphot⟩<⟨βspec⟩\langle\beta_{\rm phot}\rangle< \langle\beta_{\rm spec}\rangle). Taken together, these comparisons suggest that our βphot\beta_{\rm phot} averages are likely to be biased slightly blue. We will return to this issue as part of a potential bias in our βphot\beta_{\rm phot} estimates later (Section 4.1).

Results

In this section we first discuss the range of measured β\beta values within our new z≳12z \gtrsim12 sample, and assess evidence for exotic ultra-blue individual sources. We then investigate the redshift evolution of ⟨β⟩\langle\beta\rangle both within our sample as well as more broadly across the redshift range z≃15z \simeq15 to z≃4z \simeq4 by incorporating the results from previous works. Finally, we discuss evidence for a β\beta–MUVM_{\rm UV} relation within our z≳12z \gtrsim12 sample.

UV continuum slopes at z 12

In Figure 3 we show βphot\beta_{\rm phot} as a function of redshift (zz) for our primary z≳12z \gtrsim12 sample (purple diamonds) and 9<z<129 < z < 12 sample (grey triangles; Cullen et al. 2024). The blue shaded region spans −2.6≤β≤−2.4-2.6 \le\beta\le-2.4, with β=−2.6\beta= -2.6 representing the minimum intrinsic β\beta value of a stellar or nebular continuum SED with fesc=0f_{\rm esc} = 0. However, it is worth noting that this is a minimum value across all stellar population models, stellar ages and metallicities. For reasonable assumptions regarding the stellar populations of z>9z > 9 star-forming galaxies, the true intrinsic UV slopes are likely between β≃−2.6\beta\simeq-2.6 and β≃−2.4\beta\simeq-2.4 (e.g. see fig. 2 of Topping et al. (2022) or fig. 9 of Cullen et al. 2024 and the corresponding discussions). UV slopes at or below this intrinsic limit must necessarily be free of dust attenuation. The primary result of Cullen et al. (2024) was that by z≃11z \simeq11 the typical UV slope of the galaxy population (with −20≲MUV≲−17-20 \lesssim M_{\rm UV} \lesssim-17) had reached this blue-limit, with the inverse variance-weighted mean of ⟨β⟩=−2.59±0.06\langle\beta\rangle= -2.59 \pm0.06 at z=11.5z = 11.5. The primary motivation of this work is to determine whether this result holds with the addition of our new z≳12z \gtrsim12 sample.

Figure 3. The UV continuum slope (β\beta) plotted against redshift (zz) for photometrically-selected galaxy candidates at z≥9z \geq9. The purple diamonds show our new z≳12z \gtrsim12 sample, with the unfilled points indicating z>14z > 14 galaxies with more tentative measurements that are not included when fitting β−z\beta-z relations. The grey triangles show the ancillary z≥9z \geq9 sample from Cullen et al. (2024). The region between β=−2.4\beta=-2.4 to β=−2.6\beta=-2.6 has been highlighted with a light blue shading to show the theoretical range of β\beta that represents the intrinsic stellar plus nebular continuum limit. At these values of β\beta it is likely that galaxies are free of dust attenuation in the rest-frame UV (see text for discussion). The inverse variance-weighted averages in bins of redshifts are shown as pink squares and reach values of ⟨β⟩≃−2.6\langle\beta\rangle\simeq-2.6 at z≳11z \gtrsim11. These results indicate that — although individual reddened galaxies can exist — the galaxy population on average is consistent with the intrinsic limit and suffers negligible UV dust attenuation.

Looking first at the individual β\beta measurements, it is clear that we observe a significant scatter, with values ranging between β=−1.76\beta= -1.76 and β=−3.72\beta= -3.72. This is in large part due to the significant uncertainties associated with individual photometric estimates, with the median uncertainties of the sample being ⟨σβ,upper⟩=0.31\langle\sigma_{\beta,\mathrm{upper}}\rangle= 0.31 and ⟨σβ,lower⟩=0.34\langle\sigma_{\beta,\mathrm{lower}}\rangle= 0.34. In fact, taking the large typical uncertainties at face value, it is likely that the sample is drawn from a population with a relatively narrow range of β\beta. However, larger spectroscopic samples will be needed to definitively confirm the true distribution of β\beta values at these redshifts.

Of the 55 galaxies in the primary sample, 11 have extremely blue UV slopes below or at the canonical ‘stellar-limit’ of β=−3.2\beta= -3.2, with the bluest observed value being β=−3.72−0.48+0.40\beta= -3.72^{+0.40}_{-0.48}. Definitive confirmation of sources with β<−3\beta< -3 would be very interesting, as these are likely extremely young, perhaps even metal-free, stellar populations with extremely high escape fractions (e.g. Schaerer 2002). Their existence would have important implications for our understanding of early star-formation and the process of reionization. However, from our ultra-blue sub-sample, we find that 6/11 are consistent within 1σ1\sigma of β=−3.00\beta= -3.00 and all are within 2σ2\sigma of this value. Moreover, all eleven sources are also consistent within 3σ3\sigma of the β=−2.6\beta= -2.6 attenuation-free limit. In other words, although they almost certainly have very steep UV slopes, none can be considered robustly β≲−3\beta\lesssim-3 sources.

Despite this, the fact that 20 per cent of our sample have measured values of β≤−3\beta\le-3 is one clear indication that the galaxy population at z≳12z \gtrsim12 must be, on average, very blue. Indeed, we find that the inverse variance-weighted average for our full z≳12z \gtrsim12 sample is ⟨β⟩=−2.57±0.04\langle\beta\rangle= -2.57 \pm0.04. This is consistent with the intrinsic β\beta limit, indicating that the typical galaxy at z≳12z \gtrsim12 is essentially free of UV dust attenuation. In the following sections, we consider the evolution of ⟨β⟩\langle\beta\rangle with redshift and MUVM_{\rm UV} in more detail.

The redshift evolution of the UV continuum slope

An evolution towards bluer β\beta is expected at higher redshifts as galaxies will have progressively younger stellar populations, lower metallicities and consequently lower dust attenuations (e.g. Robertson et al. 2010; Tacchella et al. 2022. One important result from early JWST studies has been that by z≃11z \simeq11 the average value of β\beta has reached the intrinsic limit of β≃−2.6\beta\simeq-2.6 (e.g. Cullen et al. 2024; Topping et al. 2024). Whether or not this remains true for the galaxy population at z≳12z \gtrsim12 remains an open question. Although current spectroscopic samples at z>12z > 12 suggest the galaxy population does remain ultra-blue (e.g., Pollock et al. 2026; Roberts-Borsani et al. 2026; Tang et al. 2026), this has yet to be tested with larger photometric samples. Figure 3 shows the β−z\beta-z relation combining our new z≳12z \gtrsim12 sample (purple diamonds) with the Cullen et al. (2024) sample at 9<z<129 < z < 12 (grey triangles). Weighted averages in five bins of redshift are shown by the pink squares, and the corresponding values are given in Table 1. In the redshift range 9<z<129 < z < 12 we reproduce the result reported in Cullen et al. (2024) — despite the minor changes to the sample in this redshift range. The average UV continuum slope evolves from ⟨β⟩=−2.19±0.06\langle\beta\rangle=-2.19\pm0.06 at z≃9.3z\simeq9.3 to ⟨β⟩=−2.55±0.05\langle\beta\rangle=-2.55\pm0.05 at z≃11.5z\simeq11.5. As discussed in Cullen et al. (2024), the straightforward interpretation of this result is that, while the typical galaxy at z<10z<10 is consistent with moderate amounts of UV attenuation (i.e. having β>−2.4\beta>-2.4), by z≃11z\simeq11 the population average is consistent with the expected intrinsic attenuation-free value of β\beta.

Redshift RangeNumber⟨z⟩\langle z\rangle⟨β⟩\langle\beta\rangle
z<10z < 10379.27±0.049.27 \pm0.04−2.19±0.06-2.19 \pm0.06
10≤z<1110 \leq z < 114110.42±0.0310.42 \pm0.03−2.36±0.05-2.36 \pm0.05
11≤z<1211 \leq z < 123711.52±0.0411.52 \pm0.04−2.55±0.05-2.55 \pm0.05
12≤z<1312 \leq z < 133712.30±0.0412.30 \pm0.04−2.50±0.05-2.50 \pm0.05
13≤z<1413 \leq z < 141313.28±0.1913.28 \pm0.19−2.58±0.11-2.58 \pm0.11

Table 1. The inverse variance-weighted redshift ⟨z⟩\langle z\rangle and UV continuum slope ⟨β⟩\langle\beta\rangle for our combined sample of 170 galaxies (z>9z>9 galaxies from Cullen et al. (2024) and z≳12z\gtrsim12 galaxies from McLeod et al. (2026)). The five bins are presented in Figure 3 and Figure 4.

In the additional redshift bins that we can now access with our z≳12z\gtrsim12 sample we find ⟨β⟩=−2.50±0.05\langle\beta\rangle=-2.50\pm0.05 at z≃12.3z\simeq12.3 and ⟨β⟩=−2.58±0.11\langle\beta\rangle=-2.58\pm0.11 at z≃13.3z\simeq13.3, suggesting that the general trend is for the galaxy population to remain consistent with the intrinsic limit at z>12z>12. At z≥14z\geq14, our sample contains very few galaxies (N = 5) and therefore do not consider the sample statistics sufficient to draw robust conclusions. Given the lack of robust photometric candidates at z>14z>14 to date, we treat these sources as tentative, presenting them as unfilled datapoints in Figure 3. However, it is interesting to note that all of the sources at these redshifts fall below β=−2.6\beta=-2.6, with the weighted average being ⟨β⟩=−2.99±0.15\langle\beta\rangle=-2.99\pm0.15. There are two spectroscopically-confirmed z>14z>14 galaxies of which one, MoM-z14 (or PRIME-COS-Z32181 in Table C, Naidu et al. (2022)), makes it into our sample selection (see § 2.5 for details on why the galaxy discovered in Carniani et al. (2024) is not included). We fit a spectroscopic UV slope of βspec=−2.50−0.18+0.07\beta_{\rm spec}=-2.50^{+0.07}_{-0.18} for MoM-z14, which is redder than our photometric constraint of βphot=−3.08−0.38+0.38\beta_{\rm phot}=-3.08^{+0.38}_{-0.38} (see Table C). The other confirmed z>14z>14 galaxy has a spectroscopic UV slope of βspec=−2.20±0.07\beta_{\rm spec}=-2.20\pm0.07, as reported in Carniani et al. (2024). It is therefore possible that a stronger bias affects photometric constraints at the highest redshifts. Indeed, our β\beta recovery simulations do show an increasing blue bias at z>14z>14 (see Section 4.1). Ultimately however, more z>14z>14 sources (including robust photometric candidates) are needed to make progress here. For the analysis below, we do not include the poorly-constrained z>14z>14 bin in our β−z\beta-z fitting analysis, and note that excluding these sources does not significantly change our z>12z>12 weighted average with the value going from −2.57±0.04-2.57\pm0.04 to −2.52±0.05-2.52\pm0.05.

The transition to an attenuation free population

Given the indication in Figure 3 that the typical UV continuum slope reaches the intrinsic limit at z≳10z\gtrsim10, we consider the redshift above which the galaxy population can be considered to be — on average — UV attenuation-free. To do this, we follow a similar approach to Cullen et al. (2024), who found based on their 9<z<129<z<12 sample that a transition to ⟨β⟩≃−2.6\langle\beta\rangle\simeq-2.6 occurred at z>10.5z>10.5. We fit two toy models to the individual β\beta measurements across both samples: (i) a step function and (ii) a piecewise function consisting of a linear relation (with slope mm) transitioning to a constant value. The step function has the following form:

βstep(z)={β0if z<zt,step,β1if z≥zt,step,(1)\beta_{\mathrm{step}}(z)= \begin{cases} \beta_0 & \text{if } z<z_{\mathrm{t,step}},\\ \beta_1 & \text{if } z\geq z_{\mathrm{t,step}}, \end{cases} \tag*{(1)}

and the piecewise function takes the form:

βpw(z)={mz+β0if z<zt,pw,β1if z≥zt,pw.(2)\beta_{\mathrm{pw}}(z)= \begin{cases} mz+\beta_0 & \text{if } z<z_{\mathrm{t,pw}},\\ \beta_1 & \text{if } z\geq z_{\mathrm{t,pw}}. \end{cases} \tag*{(2)}

Although neither model is strongly physically motivated, the fits nevertheless return a useful indication of a transition redshift (ztz_{\mathrm{t}}) above which the average galaxy population reaches its bluest value (β1\beta_1). The two transition redshifts we find are zt,step=10.56−0.05+0.03z_{\mathrm{t,step}}=10.56^{+0.03}_{-0.05} and zt,pw=10.56−0.02+0.05z_{\mathrm{t,pw}}=10.56^{+0.05}_{-0.02}, respectively. At redshifts above this threshold, the plateau value is β1=−2.54−0.03+0.05\beta_1=-2.54^{+0.05}_{-0.03}. In both cases these are consistent with findings in Cullen et al. (2024) and again highlight the fact that the average UV continuum slopes at z≳11z\gtrsim11 become consistent with the intrinsic stellar plus nebular limit. From these fits, we can say that z≃11z\simeq11 roughly marks a transition epoch above which the UV SED of the typical galaxy is not experiencing dust attenuation. We note that this is a conservative upper limit on the redshift, since β=−2.6\beta=-2.6 is the hard limit for an attenuation free stellar population but any galaxy with β≤−2.4\beta\leq-2.4 could plausibly be unattenuated.

Extending the β−z\beta-z relation to z=4z=4

We now consider the evolution of ⟨β⟩\langle\beta\rangle with redshift across a larger redshift baseline by comparison to a variety of previous literature measurements. The left-hand panel of Figure 4 shows a comparison with other photometric analyses, while the right-hand panel shows a comparison to studies based primarily on spectroscopy. We crucially extend the redshift evolution of ⟨β⟩\langle\beta\rangle down to z=4z=4 using the HSTHST study of Bouwens et al. (2014) (green diamonds). To do this, we take the weighted average of the values in table 2 of Bouwens et al. (2014) considering only MUVM_{\mathrm{UV}} values consistent with our data (i.e. −21.25<MUV<−17.25-21.25<M_{\mathrm{UV}}<-17.25). It is immediately clear by-eye that these measurements, combined with our data, form a natural trend. Motivated by the smooth evolution we see in comparison to the Bouwens et al. (2014) measurements, and in order to make a direct comparison with other literature studies, we fit a simple linear β−z\beta-z relation to our data (pink squares), finding

Figure 4. The left-hand panel shows a β−z\beta-z comparison between the results presented here and other photometric analyses from the literature ((Cullen et al. 2024; Topping et al. 2024; Austin et al. 2025; Hainline et al. 2026)). The pink line, which has a gradient of dβ/dz=−0.11±0.02\mathrm{d}\beta/\mathrm{d}z=-0.11\pm0.02 is a linear fit to our inverse variance-weighted averages at z≥9z\geq9 (pink squares), with the shaded region showing the 1σ1\sigma uncertainty bounds. We combine our measurements with an MUVM_{\mathrm{UV}}-matched sample from Bouwens et al. (2014) that extends the relation down to z=4z=4 (green diamonds). It is clear that these data closely follow the extrapolation of our fit at z≥9z\geq9. A combined fit to our measurements and the Bouwens et al. (2014) points yields a slope of dβ/dz=−0.097±0.004\mathrm{d}\beta/\mathrm{d}z=-0.097\pm0.004, which is consistent within the 1σ1\sigma uncertainty. The right-hand panel shows the same β−z\beta-z plane, but now comparing our data (and Bouwens et al. (2014)) to the spectroscopic constraints derived from stacked JWST/NIRSpec PRISM data presented in Roberts-Borsani et al. (2024), Roberts-Borsani et al. (2026), Saxena et al. (2026), and Tang et al. (2026). Although significant variance exists among the current spectroscopic measurements, the level of agreement at z>10z>10 is generally good.

β=(−0.11±0.02)z−(1.2±0.2).(3)\beta=(-0.11\pm0.02)z-(1.2\pm0.2). \tag*{(3)}

This relation is shown by the pink line in Figure 4.

It can be seen that extrapolating this relation to z<9z<9 yields a very good match to the Bouwens et al. (2014) data. If we fit a combined β−z\beta-z relation to our data and the Bouwens et al. (2014) points, we find a fully consistent slope:

β=(−0.097±0.004)z−(1.34±0.02).(4)\beta=(-0.097\pm0.004)z-(1.34\pm0.02). \tag*{(4)}

This relation is shown as a green line in Figure 4. We therefore find that, for galaxies in the range −21.25<MUV<−17.25-21.25<M_{\mathrm{UV}}<-17.25 the average UV continuum slope evolves smoothly to progressively redder values between z≃14z\simeq14 and z≃4z\simeq4 with a gradient of dβ/dz≃−0.1\mathrm{d}\beta/\mathrm{d}z\simeq-0.1. Whether this smooth trend continues beyond the β=−2.6\beta=-2.6 limit — or stabilises at this value — remains unclear. We also note that, while our z>9z>9 sample and the Bouwens et al. (2014) 4<z<94<z<9 sample cover the same range of MUVM_{\mathrm{UV}} (and the sample selections are subject to similar limiting magnitudes in the relevant selection bands for their respective redshift bins), this simple comparison does not account for the fact that the galaxies will be at different evolutionary stages at z≃4z \simeq4 and z≃12z \simeq12.

The redshift evolution we find is consistent with previous literature studies including Topping et al. (2024) who report dβ/dz=−0.07\mathrm{d}\beta/\mathrm{d}z = -0.07 for 8<z<128 < z < 12, Saxena et al. (2026) who report dβ/dz=−0.08\mathrm{d}\beta/\mathrm{d}z = -0.08 for z≥9.5z \geq9.5 and Austin et al. (2025) who report dβ/dz=−0.10\mathrm{d}\beta/\mathrm{d}z = -0.10 for 6.5≤z≤13.56.5 \leq z \leq13.5. Moreover, it can be seen from the left panel of Figure 4 that, at z>9z > 9, the normalization of our β−z\beta- z relation is also consistent with the measurements of Hainline et al. (2026) and Topping et al. (2024) (as well as, unsurprisingly, our previous work; Cullen et al. (2024)). At z<8z < 8 the Topping et al. (2024) constraints fall below our best-fitting β−z\beta- z relation, but this discrepancy is readily explained by the fact that the Topping et al. (2024) sample is on average fainter (MUV(median)=−18.6M_{\mathrm{UV}}(\mathrm{median}) = -18.6) in this redshift range than our sample (MUV(median)=−19.4M_{\mathrm{UV}}(\mathrm{median}) = -19.4), and consequently fainter than the MUVM_{\mathrm{UV}}-matched Bouwens et al. (2014) data at the same redshifts. Similarly, Austin et al. (2025) find bluer β\beta values on average at fixed zz (Figure 4). However, the difference in this case cannot be easily explained by a difference in MUVM_{\mathrm{UV}} as the Austin et al. (2025) sample is actually slightly brighter than ours (MUV(median)=−19.8M_{\mathrm{UV}}(\mathrm{median}) = -19.8) and should therefore be, if anything, redder. The discrepancy in this case most likely stems from differences in sample selection and methodology. However, it is worth noting that the difference is only in normalization, and that Austin et al. (2025) do find a dβ/dz\mathrm{d}\beta/\mathrm{d}z slope consistent with our estimate.

We note that the gradient we find here is somewhat shallower than our original estimate of dβ/dz=−0.28\mathrm{d}\beta/\mathrm{d}z = -0.28 based on the 9<z<129 < z < 12 sample in Cullen et al. (2024). It can be seen from Figure 4 that this was due to the narrow redshift range used in Cullen et al. (2024). Overall, across the numerous photometric analyses now available, there seems to be a growing consensus for a slope of dβ/dz≃−0.1\mathrm{d}\beta/\mathrm{d}z \simeq-0.1 in MUVM_{\mathrm{UV}}-matched samples.

Comparing with spectroscopic constraints

When looking at the right-hand panel of Figure 4, it can be seen that our results, and indeed our combined best-fitting β−z\beta- z relation, are generally in good agreement with the values reported in spectroscopic analyses, although with some exceptions. The largest difference is seen with respect to Roberts-Borsani et al. (2024), who report UV slopes measured from stacked JWST/NIRSpec PRISM spectra of star-forming galaxies at 5<z<115 < z < 11 and report extremely blue average UV slopes at all redshifts, even down to z≃6z \simeq6. Essentially all of these constraints fall below the photometric estimates of ⟨β⟩\langle\beta\rangle at the same redshifts. However, when taking weighted averages of the individual βspec\beta_{\mathrm{spec}} measurements from a more recent analysis presented in Roberts-Borsani et al. (2026), we find better agreement. Comparing to these latest constraints, our photometric results are in decent agreement at z≳12z \gtrsim12, but are slightly bluer in the redshift range 10<z<1210 < z < 12. Similarly, Saxena et al. (2026) find average spectroscopic β\beta values in line with our photometric estimate at z≃11z \simeq11, but also find bluer values than our relation would predict at z<10z < 10. Finally, we find excellent consistency with the values presented in Tang et al. (2026) which fall along our best-fitting relation at z=9−11z = 9 - 11.

Although a significant variance remains between independent spectroscopic samples, overall it seems reasonable to conclude from Figure 4 that the values derived from spectroscopic and photometric analyses are converging. Indeed, given the small sample sizes and inherent incompleteness of spectroscopic datasets, it makes sense — and is encouraging — that as the sample size has grown these spec-

troscopic constraints are becoming increasingly more aligned with the larger and more complete photometric samples. Nevertheless, it is clear that more work is needed in this area. Ultimately, large and complete spectroscopic samples are required to definitively prove that the galaxy population reaches the intrinsic attenuation-free limit at z>10z > 10, and to determine the intrinsic scatter of individual UV slopes at these early cosmic epochs.

The β−MUV\beta- M_{\mathrm{UV}} relation

It is well established that at z<10z < 10 there exists a relationship between absolute UV magnitude and β\beta (the so-called colour-magnitude, or β−MUV\beta- M_{\mathrm{UV}} relation; Bouwens et al. 2014, Rogers et al. 2014, Cullen et al. 2024, Topping et al. 2024). Due to sample sizes that lacked a sufficient dynamic range in MUVM_{\mathrm{UV}}, studying the β−MUV\beta- M_{\mathrm{UV}} relation at z>10z > 10 initially proved difficult (Cullen et al. 2024; Topping et al. 2024). However, the first constraints at these redshifts have recently been reported by Hainline et al. (2026). With our new sample of N = 55 galaxy candidates at z≳12z \gtrsim12, covering an impressive range in $M_{\mathrm{UV}} ({xref}Fig. 1`), we are in a position to provide an independent assessment at these redshifts.

Figure 5 shows the β−MUV\beta- M_{\mathrm{UV}} relation for our z≳12z \gtrsim12 sample compared with a number of previous literature estimates. As in Figure 3 and Figure 4, we show the inverse variance-weighted mean values for our z≳12z \gtrsim12 sample as pink squares (values given in Table 2). For clarity, we do not show the individual constraints in Figure 5 (see Figure 3 for a useful indication of the typical scatter in the individual β\beta constraints). We find that at all magnitudes ⟨β⟩\langle\beta\rangle is close to the minimum intrinsic stellar plus nebular value. In some sense this is unsurprising, given our general finding that by z≳11z \gtrsim11 the typical galaxy is consistent with this attenuation-free limit. We do not attempt to fit a relation to our data, as visually it can clearly be seen that these values are consistent with scatter about the sample-average value of ⟨β⟩=−2.57±0.04\langle\beta\rangle= -2.57 \pm0.04. This finding of a flat β−MUV\beta- M_{\mathrm{UV}} relation (i.e., no β−MUV\beta- M_{\mathrm{UV}} relation) is in clear contrast to the situation at lower redshifts, where a slope of −0.2<dβ/dMUV<−0.1-0.2 < d\beta/dM_{\mathrm{UV}} < -0.1 is commonly reported (see e.g. fig. 8 of Cullen et al. (2024)).

Figure 5. Plot of UV continuum slopes as a function of absolute UV magnitude (MUVM_{\mathrm{UV}}). The pink squares show the inverse variance-weighted average UV continuum slopes for our z≳12z \gtrsim12 sample. We find no evidence for a β−MUV\beta- M_{\mathrm{UV}} relation in our sample: the average UV slope is consistent with the intrinsic limit (indicated by the horizontal blue band) at all magnitudes. Various literature relations are also shown, including Cullen et al. (2024) (grey line), Topping et al. (2024) (green dashed line), and Hainline et al. (2026) (purple dotted line). The redshift ranges for these various relations are indicated in the legend. The purple crosses show the average values from Hainline et al. (2026). Although we see good agreement at the faint end, at the bright end there is a clear discrepancy between our data and the Hainline et al. (2026) z≥11z \geq11 sample. We discuss the possible reasons for this offset in Section 4.2.

MUVM_{\mathrm{UV}} RangeNumber⟨MUV⟩\langle M_{\mathrm{UV}}\rangle⟨β⟩\langle\beta\rangle
MUV<−21M_{\mathrm{UV}} < -216−21.23±0.03-21.23 \pm0.03−2.64±0.12-2.64 \pm0.12
−21≤MUV<−20-21 \le M_{\mathrm{UV}} < -207−20.40±0.04-20.40 \pm0.04−2.62±0.12-2.62 \pm0.12
−20≤MUV<−19-20 \le M_{\mathrm{UV}} < -1927−19.44±0.02-19.44 \pm0.02−2.67±0.07-2.67 \pm0.07
−19≤MUV<−18-19 \le M_{\mathrm{UV}} < -1810−18.60±0.03-18.60 \pm0.03−2.27±0.10-2.27 \pm0.10
−18≤MUV<−17-18 \le M_{\mathrm{UV}} < -175−17.59±0.03-17.59 \pm0.03−2.57±0.15-2.57 \pm0.15

Table 2. The inverse variance-weighted absolute UV magnitude ⟨MUV⟩\langle M_{\mathrm{UV}}\rangle and UV continuum slope ⟨β⟩\langle\beta\rangle for our z≳12z \gtrsim12 galaxy sample.

Comparing to other literature estimates, we find some interesting differences. Focusing first on the analyses of Cullen et al. (2024) and Topping et al. (2024) we find a general consistency in the approximate normalisation, although both of these studies report β−MUV\beta- M_{\mathrm{UV}} gradients. However, in both these cases—due to a lack of dynamic range in the data—these gradients are fixed to values measured at lower redshifts and only the normalisation is fitted (to illustrate this, the single data point from Topping et al. (2024) is shown in Figure 5). In light of this we consider our results fully consistent with these early investigations.

In contrast, Hainline et al. (2026) find dβ/dMUV=−0.22±0.05d\beta/dM_{\mathrm{UV}} = -0.22 \pm0.05 for their z≥11z \ge11 sample clearly contrasting with our findings. Although our data are fully consistent with Hainline et al. (2026) at MUV>−19M_{\mathrm{UV}} > -19, there is a clear divergence at the bright end. The reason for this discrepancy is due to significant differences in our sample selection techniques, specifically in the criteria applied to consider a high-redshift candidate robust. We return to a detailed discussion of this issue in § 4 below.

Discussion

The primary finding of this work is that at z≳12z \gtrsim12 we find, on average, an extremely blue galaxy population with the typical UV slope at these redshifts being consistent with the intrinsic, attenuation-free, limit of β≃−2.6\beta\simeq-2.6. We find that the evolution of the typical UV slopes of galaxies with −21≲MUV≲−17-21 \lesssim M_{\mathrm{UV}} \lesssim-17 is consistent with a smooth evolution to bluer values between z≃4z \simeq4 to z≃15z \simeq15 (Figure 4), with z≳11z \gtrsim11 marking the epoch at which the average value has reached the intrinsic limit. We also find a flat β−MUV\beta- M_{\mathrm{UV}} relation in our new z≳12z \gtrsim12 sample indicating that—on average—all galaxies are attenuation-free regardless of their observed brightness.

In light of some recent results reporting individual galaxies with relatively red UV slopes z≳10z \gtrsim10 (e.g. Mitsuhashi et al. 2026; Donnan et al. 2025b; Hainline et al. 2026; Rodighiero et al. 2026), below we investigate whether there is any evidence to suggest that the blue UV slope values we find are a result of a bias in our selection criteria or measurement methodology. We then discuss the existence of individual red sources within our z>12z > 12 sample and consider whether (or not) these are consistent with the general evolutionary trend towards an attenuation-free population.

Is our sample biased towards blue β\beta values?

Because the average UV continuum slopes we measure are extremely blue, it is important to determine whether our sample could be inherently biased towards bluer sources (as a consequence of some combination of sample selection effects and measurement bias). To do this, we generate galaxy SEDs across a wide range of β\beta and produce simulated photometry of these objects assuming typical depths consistent with our observations. We consider fixed β\beta values between −1.5-1.5 and −2.8-2.8 (with a step size of 0.10.1), and for each β\beta value simulate 10 200 SEDs within the redshift range 10.0≤z≤17.010.0 \le z \le17.0 and UV magnitude range −21.5≤MUV≤−17.0-21.5 \le M_{\mathrm{UV}} \le-17.0. To reflect the varying depths of the surveys used to construct our sample, we run separate simulations for two different depth tiers: an ultra-deep tier, based on the depths in JADES GOODS-South region, and a deep tier, based on median depths of the CEERS and PRIMER fields (which are very similar to the median depths of the pure parallel fields included in this study). The exact depths used for these two tiers are calculated using table A1 in McLeod et al. (2026) and are used to perturb the model photometry and generate realistic observed data. For each depth set, we use a filter-set consistent with the filters used in the representative programs. Overall, this results in simulated photometry for ≃288 000\simeq288\,000 objects.

To mimic our selection methodology, all of the simulated objects are fit with LePHARE to obtain photometric redshift estimates. We then apply the selection criteria as described in § 2 to retain only what we would consider robust high-redshift sources. This process left ≃145 000\simeq145\,000 selected sources. Assessing the β\beta distribution of the recovered sources, we find that our selection criteria do not preferentially select for bluer β\beta values. The fraction of sources recovered is essentially flat as a function of input β\beta, with a slight drop of ≃10\simeq10 per cent at the very red end (β>−1.7\beta> -1.7). This is unsurprising, since our selection criteria allow for an SNR >8> 8 detection in either of the F200WF200W or F277WF277W filters, which means we can in principle select objects with red slopes that are fainter at ≃0.15 μm\simeq0.15\,\mu\mathrm{m} and brighter at ≃0.2 μm\simeq0.2\,\mu\mathrm{m} (rest-frame). Indeed, we reached a similar conclusion in our previous work (see appendix C of Cullen et al. (2024)). The slight drop-off at the reddest values of β\beta reflects the fact that these objects are more likely to be confused with low-redshift quiescent sources. However, they are only marginally less likely to be selected: if they existed in large numbers, they would be present in our sample.

A separate question is whether our methodology for fitting β\beta results in a blue bias. For the remaining ≃145 000\simeq145\,000 objects, we measure β\beta using the methodology described in § 2.3. In Figure 6 we show the difference between the input β\beta and the recovered β\beta (inverse variance-weighted) as a function of the best-fitting redshift obtained via our fitting method. The panels show the two different depth tiers as indicated in their titles. The black line in each panel shows the running median value of this difference (Δβ\Delta\beta) across all values of input β\beta, and the various coloured lines break this down into six β\beta ranges as indicated in the legend. From Figure 6 it can be seen that, in general, we do see a small blue bias across all UV continuum slope values and redshifts, at least in the ‘Deep’ tier. This bias is consistently at the level of −0.05≲Δβ≲0.05-0.05 \lesssim\Delta\beta\lesssim0.05, with on average a slight, but essentially negligible blue bias (with a mild dependence on input β\beta). The fact that we find a slight blue bias in our photometric constraints is consistent with the small offset discussed with respect to the spectroscopic measurements in Figure 2. The bias does clearly worsen at z>14z > 14, although not at a level sufficient to explain ⟨β⟩≃−3\langle\beta\rangle\simeq-3 in our observed z>14z > 14 candidates. Overall however, the small bias we find here does not change any of the primary conclusions of this work.

Difference between recovered $\beta$ and input $\beta$ for simulated galaxies

Figure 6. The difference between recovered β\beta and input β\beta for a suite of simulated galaxies as a function of their recovered redshifts. The panels showcase two depth tiers (see Section 4.1 and panel headings). The black line shows the trend across all of the input β\beta values, and the coloured lines show the trend in bins of βinput\beta_{\mathrm{input}} as indicated in the legend. For the deep tier, we find a small bias of ∣Δβ∣≃0.05|\Delta\beta| \simeq0.05 that is only mildly dependent on the redshift and input β\beta. This bias is even lower for the ultra-deep tier.

As a final test, we investigate whether the bias is UV magnitude-dependent. In the pre-JWST era, it was common for fainter objects to be biased blue at the redshift-limit of HST studies (e.g. Rogers et al. 2014). Separating our simulated sample into bright (MUV≤−20M_{\mathrm{UV}} \leq-20) and a faint (MUV>−20M_{\mathrm{UV} > -20}) subsamples, we find no evidence for a larger blue bias in either, with the trend remaining the same as for the full sample. Again, this lack of MUVM_{\mathrm{UV}} dependence is unsurprising since the SEDs of all sources are well-sampled by the available photometry redward of the Lyman break, and our strict SNR >8> 8 criteria ensures that all sources are well-detected regardless of their intrinsic magnitude.

Evidence for red, dusty galaxies at z12

Recently, a number of works have reported instances of ‘red’ galaxies at z>10z > 10, including sources with β>−2\beta> -2 (e.g. Mitsuhashi et al. 2026; Donnan et al. 2025b; Hainline et al. 2026; Rodighiero et al. 2026). It is important to consider these sources in the context of our claim that the typical UV slope of the galaxy population is consistent with the intrinsic attenuation-free limit at z≳11z \gtrsim11.

Of these, the most robust is EGS-z11-R0, a galaxy at zspec=11.452±0.021z_{\mathrm{spec}} = 11.452 \pm0.021 that has a NIRSpec/PRISM spectrum taken as part of JWST GO program 4106 (PI: E. Nelson). Rodighiero et al. (2026) have recently presented an analysis of this source, reporting a remarkably red spectroscopic slope of βspec=−0.68±0.3\beta_{\mathrm{spec}} = -0.68 \pm0.3. This source was not selected in our initial Cullen et al. (2024) F115W-dropout sample as it fell on the edge of the short-wavelength imaging of our reduction. The same edge effect is present in the v1.0 CEERS data release (Bagley et al. 2023; Finkelstein et al. 2025), and so the object does not feature in our new sample either. From the publicly-available DJA spectrum, we find a somewhat bluer spectroscopic UV slope of βspec=−1.51−0.16+0.15\beta_{\mathrm{spec}} = -1.51^{+0.15}_{-0.16}, which is more consistent with the photometric measurement from Rodighiero et al. (2026). Despite the difference in our spectroscopic β\beta estimates, it nevertheless appears that this source is robustly β>−2\beta> -2.

However, it is clear from Figure 3 that the existence of individual objects with such red β\beta values is perfectly consistent with our analysis. For example, in our z>12z > 12 sample N=4/55N = 4/55 (≃10\simeq10 per cent) have β≳−2\beta\gtrsim-2. Similarly red sources are also clearly present at z>10z > 10 in the Cullen et al. (2024) sample. Within the context of their ‘attenuation-free model’, Ferrara et al. (2026) interpret these redder sources as representing a ‘red monster’ evolutionary phase, during which galaxies form stars at a moderate rate and build up ISM dust (see also Martínez-González et al. 2026b). This phase is followed by a powerful starburst that ultimately generates outflows to efficiently remove the dust and produces the more commonly-observed ‘blue monsters’ (e.g. Ziparo et al. 2023). One simple interpretation of the observations is therefore that the majority of z>10z > 10 galaxies are undergoing an extreme starburst. The quieter phase of dust build-up is either very transitory, or remains mostly hidden at the depth of JWST observations. The latter scenario seems unlikely as it would imply further boosting the surprisingly large abundance of MUVM_{\mathrm{UV}}-bright galaxies at these redshifts (e.g. Ferrara et al. 2023; McLeod et al. 2024).

Beyond the spectroscopic confirmation of the red z≃11z \simeq11 galaxy EGS-z11-R0, some authors have suggested that the inclusion of JWST/NIRCam medium band photometry increases the likelihood of detecting redder sources at the highest redshifts (e.g. Asada et al. 2026). The implication here is that earlier studies relying mostly on the JWST/NIRCam wide bands (e.g. Topping et al. 2024; Cullen et al. 2024) are biased against red galaxies. Recently Mitsuhashi et al. (2026), leveraging the additional medium band NIRCam filter coverage of Abell 2744 from the MegaScience survey (GO 4111; PI: Suess), identified two red z≃12z \simeq12 candidates with β=−1.33−0.08+0.09\beta= -1.33^{+0.09}_{-0.08} and β=−1.54−0.08+0.07\beta= -1.54^{+0.07}_{-0.08}. The authors specifically make the claim that there might be a bias towards selecting bluer galaxies when using broadband photometry only. However, both of these sources were selected in our previous broadband-selected samples with UV continuum slopes of β=−1.49±0.25\beta= -1.49 \pm0.25 and −1.65±0.25-1.65 \pm0.25 (McLeod et al. 2024; Cullen et al. 2024). As discussed in Section 4.1 above, we do not find any evidence for a bias towards preferentially recovering blue objects with our selection function (see also appendix C of Cullen et al. 2024). The existence of objects with β>−2\beta> -2 in our current sample is also a testament to the fact that broadband selection can – and does – select red objects. Nevertheless, the inclusion of medium bands is obviously beneficial in providing greater spectral sampling and thus tighter constraints on individual β\beta measurements.

Perhaps the most significant discrepancy between our results and results in the literature is with the z≥11z \ge11 β−MUV\beta- M_{\mathrm{UV}} relation reported by Hainline et al. (2026) (see Figure 5). Specifically, while we find a flat relation (i.e. galaxies being consistent with the intrinsic dust-free limit at all MUVM_{\mathrm{UV}}), Hainline et al. (2026) instead find a strong β−MUV\beta- M_{\mathrm{UV}} relation with the brightest sources (MUV≲−20M_{\mathrm{UV}} \lesssim-20) having very red UV slopes on average (⟨β⟩>−2\langle\beta\rangle> -2). To investigate this discrepancy we use the publicly-available photometry catalogue provided by Hainline et al. (2026). Focusing on the 2082 sources at z>8z > 8, we first remove N=683/2082N = 683/2082 that do not meet our SNR selection criteria (see § 2). We fit the remaining sources with LePhare and discard a further N=161/1399N = 161/1399 galaxies that either have a primary zphot<8z_{\mathrm{phot}} < 8, χreduced2>10\chi_{\mathrm{reduced}}^{2} > 10 or have a low redshift solution within Δχ2=4\Delta\chi^{2} = 4 of the high redshift solution. This results in a final sample of 1238 sources that mimics our selection criteria, highlighting the most obvious difference between our analyses: our stricter 8σ8\sigma selection criteria means that we would only consider ≃60\simeq60 per cent of the Hainline et al. (2026) sources to be robust z>8z > 8 galaxies.

Fitting β\beta and MUVM_{\mathrm{UV}} for the remaining 1238 sources following the methodology described in § 2.3, we find a flat β−MUV\beta- M_{\mathrm{UV}} relation that is much more consistent with the result we derive from our z≳12z \gtrsim12 sample. Of the three z>11z > 11 galaxies with MUV<−20M_{\mathrm{UV}} < -20 in the Hainline et al. (2026) sample with ⟨β⟩=−1.70±0.25\langle\beta\rangle= -1.70 \pm0.25 (see Figure 5), we find that only one source meets our selection criteria. The retained source is the bluest of the three (β=−2.33±0.05\beta= -2.33 \pm0.05 in Hainline et al. (2026); β=−2.17−0.25+0.25\beta= -2.17^{+0.25}_{-0.25} following our methodology). The removed sources are both very red (β=−1.70±0.20\beta= -1.70 \pm0.20 and β=−1.51±0.17\beta= -1.51 \pm0.17 in Hainline et al. (2026)) and are responsible for the red average value. One of these sources fails because of a statistically unacceptable fit (χν2=13\chi_{\nu}^{2} = 13) and the other for having a lower redshift solution within Δχ2<4\Delta\chi^{2} < 4 of the z>11z > 11 solution. We show our version of the β−MUV\beta- M_{\mathrm{UV}} relation derived from the Hainline et al. (2026) sample in Appendix B. The comparison highlights how crucial the details of the sample selection can be in studies such as these. Any impure selection is highly likely to bias samples to redder β\beta, since the majority of interlopers whose photometry can mimic a z>9z > 9 galaxy are by definition red sources. Ultimately, we still await larger z>10z > 10 spectroscopic samples to properly quantify the purity of the various selection techniques (e.g., Donnan et al. 2026a).

To summarise, while our analysis is fully consistent with the existence of individual red and dust-reddened sources, the fraction of such objects is low (a crude estimate being ≃10\simeq10 per cent with β>−2\beta> -2; Figure 3). Our main finding remains that the typical galaxy at these redshifts has a UV slope between β≃−2.6\beta\simeq-2.6 and β≃−2.4\beta\simeq-2.4, consistent with the intrinsic attenuation-free limit. We do not find any evidence to suggest that this is a result of a bias in our selection process.

Conclusions

We have presented new measurements of the photometric UV continuum slopes (β\beta) of the 55 galaxy candidates at z≳12z \gtrsim12 from McLeod et al. (2026). These galaxies were found through a search of >0.6 deg2> 0.6\,\mathrm{deg}^{2} of JWST/NIRCam imaging, spanning a redshift range of 11.5≤z≤15.211.5 \leq z \leq15.2 and an absolute UV magnitude range of −21.6≤MUV≤−17.3-21.6 \leq M_{\mathrm{UV}} \leq-17.3 (Figure 1). To increase the dynamic range in redshift, we supplemented the z≳12z \gtrsim12 sample with our previous z≳9z \gtrsim9 sample presented in Cullen et al. (2024). We obtain UV continuum slopes by fitting a power-law of the form fλ∝λβf_{\lambda} \propto\lambda^{\beta} to the photometric data, and have validated our photometric measurements by comparing to available spectroscopic measurements (Figure 2). In our analysis we focus mainly on the average values of the UV continuum slope ⟨β⟩\langle\beta\rangle to characterize typical galaxy properties at z>12z > 12. Our main findings are:

  1. We find a large range in individual β\beta measurements within the new z≳12z \gtrsim12 sample (−3.72<β<−1.76-3.72 < \beta< -1.76; Figure 3) but this is primarily a result of the significant uncertainties on the individual β\beta measurements (the median uncertainty is σβ≃0.3\sigma_{\beta} \simeq0.3). Despite our sample containing many β<−3\beta< -3 sources, which would be potentially indicative of exotic stellar populations, we find no individual sources that are robustly this blue within the uncertainties.

  1. We do, however, find that on average the galaxy population at z≳12z \gtrsim12 is extremely blue, with an inverse variance-weighted average of ⟨β⟩=−2.57±0.04\langle\beta\rangle= -2.57 \pm0.04. This value is consistent with the intrinsic β\beta limit for a stellar plus nebular continuum in the absence of dust-attenuation (−2.6≲β≲−2.4-2.6 \lesssim\beta\lesssim-2.4). Our results suggest that z≳12z \gtrsim12 galaxies are consistent with being, on average, dust-attenuation free.

  1. Our weighted averages at z>10z > 10 (Figure 3 and Table 1) are within the expected intrinsic β\beta range, forming a similar plateau of values as seen in Cullen et al. (2024). However, we find an evolution to redder values at z≤10z \leq10 (⟨β⟩=−2.19±0.06\langle\beta\rangle= -2.19 \pm0.06), suggesting that z≃10−11z \simeq10 - 11 marks the approximate epoch (ztz_{\mathrm{t}}) at which galaxies become on average dust-attenuation free. Fitting two value models to approximate ztz_{\mathrm{t}}, we find zt≃10.5z_{\mathrm{t}} \simeq10.5, consistent with the value reported in Cullen et al. (2024).

  1. Fitting a linear β−z\beta- z relation to our data we find β=(−0.11±0.02)z−(1.2±0.2)\beta= (-0.11 \pm0.02)z - (1.2 \pm0.2) over the redshift range 9≲z≲149 \lesssim z \lesssim14. Comparing this fit to an MUVM_{\mathrm{UV}}-matched sample of galaxies down to z≃4z \simeq4 from Bouwens et al. (2014), we find that the extrapolation of this relation is in good agreement with the data at lower redshifts (Figure 4). Fitting to our data combined with Bouwens et al. (2014), over the full range 4<z<144 < z < 14, we find β=(−0.097±0.004)z−(1.34±0.02)\beta= (-0.097 \pm0.004)z - (1.34 \pm0.02). Our results suggest that the galaxy population with −21≲MUV≲−17-21 \lesssim M_{\mathrm{UV}} \lesssim-17 becomes progressively bluer towards high redshifts with dβ/dz≃−0.1d\beta/dz \simeq-0.1.

  1. We find no evidence for a β−MUV\beta- M_{\mathrm{UV}} relation within our z≳12z \gtrsim12 sample (Figure 5). This result is in stark contrast to the situation at lower redshifts (z≲10z \lesssim10), in which a clear β−MUV\beta- M_{\mathrm{UV}} relation is consistently observed with a gradient in the range −0.2≲dβ/dMUV≲−0.1-0.2 \lesssim d\beta/dM_{\mathrm{UV}} \lesssim-0.1. Instead, we find that the z≳12z \gtrsim12 population is uniformly blue at all magnitudes and that the full galaxy population is therefore, on average, attenuation free.

  1. The lack of a β−MUV\beta- M_{\mathrm{UV}} relation in our sample contrasts with the results at z≥11z \geq11 presented in Hainline et al. (2026), who find that galaxies at the bright end (MUV<−20M_{\mathrm{UV}} < -20) are significantly reddened (β≃−1.7\beta\simeq-1.7; Figure 5). The reason for this discrepancy is the different selection criteria between the two studies. Specifically, when applying our selection cuts to the Hainline et al. (2026) sample, we find that only ≃60\simeq60 per cent of their sources are considered robust high-redshift sources based on our selection criteria. If we only consider these sources that would be recovered in our selection criteria, a similarly flat and blue β−MUV\beta- M_{\mathrm{UV}} relation is recovered (see discussion in § 4 and Appendix B).

  1. Running a suite of β\beta recovery simulations, we find that our sample selection criteria are not biased against redder galaxies, with the source recovery fraction dropping by only ≃10\simeq10 per cent for β=−1.7\beta= -1.7. If bright dust-reddened sources with robust photometric redshifts at z>11z > 11 existed they would be included in our sample. Moreover, we find that, at the typical depth of the surveys used, and for the filter sets available, the photometric β\beta measurements have only a small average bias of ∣Δβ∣≃0.05|\Delta\beta| \simeq0.05. Applying a correction of this magnitude would not affect any of our main results.

  1. Finally, we discuss our results in the context of individual red sources with β≳−2\beta\gtrsim-2 that have been discussed at z≃12z \simeq12 and

Mitsuhashi et al. 2026; Rodighiero et al. 2026. The distribution of individual β\beta measurements is fully consistent with the existence of individual red sources (Figure 3) although we find that the fraction of such sources is low (≃10\simeq10 per cent). Future follow up and confirmation of these early red sources could potentially provide crucial constraints on the nature of dust enrichment at the earliest cosmic epochs.

In summary, our main result is that galaxies at z≳12z \gtrsim12 are on average very blue, and consistent with the intrinsic dust-attenuation free β\beta limit. Overall, this is in good agreement with the trend discovered in early studies at slightly lower redshifts, and suggests that the galaxy population reaches — and remains at — the intrinsic β\beta limit at z≳10z \gtrsim10 (e.g. Cullen et al. 2024; Topping et al. 2024; Austin et al. 2025). Although we have not considered them in detail here, there exist numerous theoretical explanations to explain the uniformly blue UV colours of these sources, including a lack of ISM dust either due to dust destruction (Esmerian & Gnedin 2024), strong outflows (Ferrara et al. 2023; Fiore et al. 2023; Ziparo et al. 2023), a change in dust grain properties (specifically a top-heavy grain distribution; Narayanan et al. 2026, 2025; Sommovigo et al. 2026) which could result in a grey (i.e. flat) attenuation curve (McKinney et al. 2026; Nakazato et al. 2026; Zimmerman et al. 2026). In order to further understand the physical mechanism that leads to the very blue galaxy population observed at z≳10z \gtrsim10, we require continued spectroscopic follow-up of the current z>10z > 10 galaxy candidates with JWST and, ultimately, improved constraints on their rest-frame IR emission, which will hopefully become possible with the upcoming ALMA wide-band sensitivity upgrade.

Acknowledgements

We want to thank everyone who has worked to enable the many JWST surveys we use in this work. We also warmly thank Guido Roberts-Borsani for valuable insights and for providing data from Roberts-Borsani et al. (2026). DJM and JSD acknowledge the support of the Royal Society through the award of a Royal Society University Research Professorship to JSD. FC acknowledges support from a UKRI Frontier Research Guarantee Grant (PI Cullen; grant reference: EP/X021025/1). ACC, H-HL, ET, and SDS acknowledge support from a UKRI Frontier Research Guarantee Grant (PI Carnall; grant reference EP/Y037065/1). Some of the data products presented herein were retrieved from the Dawn JWST Archive (DJA). DJA is an initiative of the Cosmic Dawn Center (DAWN), which is funded by the Danish National Research Foundation under grant DNRF140.

This work is based primarily on observations made with the NASA/ESA/CSA James Webb Space Telescope. The data were obtained from the Mikulski Archive for Space Telescopes at the Space Telescope Science Institute, operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-03127 for JWST. The observations used in this study are associated with programs 1063, 1176, 1180, 1181, 1199, 1208, 1210, 1283, 1286, 1287, 1324, 1345, 1355, 1433, 1727, 1810, 1837, 1840, 1895, 1963, 1967, 2079, 2234, 2282, 2514, 2555, 2561, 2566, 2727, 2732, 2736, 2738, 2739, 2750, 2756, 2767, 2883, 3073, 3215, 3293, 3362, 3516, 3538, 3859, 3990, 4043, 4111, 4125, 4212, 4287, 4598, 4625, 4744, 4446, 5224, 5398, 5893, 6368, 6434, 6511, 6541, 6549, 6556, 6564, 6882, 6954, 8060. The authors acknowledge the associated teams for developing their observing program and in particular, we thank the programs that were developed and executed with a zero-exclusive-access period.

This work is also based (in part) on observations taken by the RELICS Treasury Program (GO 14096) with the NASA/ESA Hubble Space Telescope, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.

Some of the lensing clusters analysed come from the CLASH program with the NASA/ESA Hubble Space Telescope (GO 12065). The mass models, constructed in Zitrin et al. (2009, 2015), were obtained through the Hubble Space Telescope Archive, as a high-end science product of the CLASH program (Postman et al. 2012).

This work utilizes gravitational lensing models produced by PIs Bradač, Natarajan & Kneib (CATS), Merten & Zitrin, Sharon, Williams, Keeton, Bernstein and Diego, and the GLAFIC group. This lens modelling was partially funded by the HST Frontier Fields program conducted by STScI. STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. The lens models were obtained from the Mikulski Archive for Space Telescopes (MAST).

This research has made use of the SVO Filter Profile Service (http://svo2.cab.inta-csic.es/theory/fps/) supported from the Spanish MINECO through grant AY2017-84809 (Rodrigo et al. 2012; Rodrigo & Solano 2020).

This research has made use of NASA’s Astrophysics Data System Bibliographic Services.

Data Availability

At the time of writing, the photometric data sets used in this manuscript were all publicly available from the JWST archive on MAST. For JADES and CEERS, these are available as part of their respective public data releases. All the spectra, except PAN-z14-1, are available on the DJA, with PAN-z14-1 becoming available in mid-November.

References

Redshift evolution of the UV continuum slope with fixed redshift

Here we investigate whether fixing the redshift to the photometric redshift obtained from fitting the full rest-frame UV to optical SED affects the β−z\beta-z relation. In Figure A1 it can be seen that the general trend remains as the trend we find in Figure 3, with the values of the UV continuum slope being consistent with the intrinsic limit at z≳11z \gtrsim11. The median differences between our adopted method (in which we use the photometric redshift as a prior rather than fixing it; see § 2.3) and this fixed redshift approach is Δβ=−0.06\Delta\beta=-0.06 and Δz=0.06\Delta z=0.06.

The MUV- relation of the Hainline et al. sample

Here we show the impact on the MUV−βM_{\rm UV}-\beta relation derived from the Hainline et al. (2026) sample if we apply our selection criteria (which removes ≃40\simeq40 per cent of their candidates; see § 4.2). In Figure B1 it can be seen that applying our selection criteria results in an MUV−βM_{\rm UV}-\beta relation much more consistent with the one presented in this paper (i.e. consistent with a constant value of β\beta at all MUVM_{\rm UV}). The difference between the relation derived after our selection criteria are applied, compared to the relation presented in Hainline et al. (2026), is especially stark at the bright end. We find that ⟨β⟩\langle\beta\rangle in the −21<MUV<−20-21<M_{\rm UV}<-20 and −20<MUV<−19-20<M_{\rm UV}<-19 bins shifts from ⟨β⟩=−1.7±0.25\langle\beta\rangle=-1.7\pm0.25 and ⟨β⟩=−2.17±0.11\langle\beta\rangle=-2.17\pm0.11 (as reported in Hainline et al. (2026)) to ⟨β⟩=−2.25±0.18\langle\beta\rangle=-2.25\pm0.18 and ⟨β⟩=−2.45±0.18\langle\beta\rangle=-2.45\pm0.18, respectively.

Figure B1. A plot showing the effect of applying our selection criteria to the MUV−βM_{\rm UV}-\beta relation of Hainline et al. (2026). The purple crosses show the original values reported in Hainline et al. (2026) while the light purple diamonds show how these values change once our selection criteria are applied to their sample (which removes ≃40\simeq40 per cent of their sources; see § 4.2 for a full description). It is clear that, when considering only the sources that we would consider robustly selected, that the MUV−βM_{\rm UV}-\beta relation becomes essentially flat and much more consistent with our finding (pink squares).

Tabulated for the full sample

In Table C we present the properties for the individual galaxies in our sample.

This paper has been typeset from a TEX/LATEX file prepared by the author.

Table C1. The properties of the individual sources in our z>12z > 12 sample. The redshift (zDynestyz_{\mathrm{Dynesty}}) and UV continuum slope (β\beta) presented are from our fit with Dynesty. The absolute UV magnitude (MUVM_{\mathrm{UV}}) is calculated at 1500Å and is corrected to total and corrected for de-lensing (using magnification μ\mu). Where applicable we also include the spectroscopic redshift (zspecz_{\mathrm{spec}}) and UV continuum slope (βspec\beta_{\mathrm{spec}}). Note that for objects with μ=1.2\mu= 1.2 this implies that the candidate is in a lensed field, but is outside the lensing map coverage.

IDRADeczDynestyz_{\mathrm{Dynesty}}β\betaMUVM_{\mathrm{UV}}μ\muzspecz_{\mathrm{spec}}βspec\beta_{\mathrm{spec}}Ref.
2236+0032-202765222:36:44.57+00:32:18.5711.5−0.8+0.911.5^{+0.9}_{-0.8}−2.25−0.39+0.35-2.25^{+0.35}_{-0.39}−20.0−0.1+0.1-20.0^{+0.1}_{-0.1}----
PAN-1707+5852-201555717:07:15.91+58:51:44.1011.6−0.4+0.411.6^{+0.4}_{-0.4}−2.35−0.34+0.32-2.35^{+0.32}_{-0.34}−20.4−0.2+0.2-20.4^{+0.2}_{-0.2}----
PAN-0134-1531-202320801:34:43.79−15-15:31:27.3911.6−0.5+0.711.6^{+0.7}_{-0.5}−2.35−0.4+0.38-2.35^{+0.38}_{-0.4}−20.5−0.2+0.2-20.5^{+0.2}_{-0.2}----
PAN-0104-5508-202629601:04:17.50−55-55:07:35.5511.7−0.4+0.411.7^{+0.4}_{-0.4}−2.23−0.27+0.28-2.23^{+0.28}_{-0.27}−20.6−0.1+0.1-20.6^{+0.1}_{-0.1}----
UNCOVER-207909000:14:29.18−30-30:27:22.8411.8−0.3+0.411.8^{+0.4}_{-0.3}−1.94−0.28+0.25-1.94^{+0.25}_{-0.28}−19.2−0.1+0.1-19.2^{+0.1}_{-0.1}1.2--μ\mu: Furtak et al. (2023)
PAN-1007+2109-204086110:07:44.02+21:11:00.9811.9−0.3+0.311.9^{+0.3}_{-0.3}−2.83−0.28+0.28-2.83^{+0.28}_{-0.28}−20.7−0.1+0.1-20.7^{+0.1}_{-0.1}----
JADES-GS-219919603:32:39.92−27-27:49:17.6212.0−0.1+0.112.0^{+0.1}_{-0.1}−2.22−0.25+0.25-2.22^{+0.25}_{-0.25}−19.0−0.1+0.1-19.0^{+0.1}_{-0.1}-12.63−2.07−0.09+0.09-2.07^{+0.09}_{-0.09}Curtis-Lake et al. (2023)
JADES-GS-203735803:32:06.88−27-27:53:34.8512.1−0.2+0.312.1^{+0.3}_{-0.2}−2.23−0.25+0.25-2.23^{+0.25}_{-0.25}−18.5−0.1+0.1-18.5^{+0.1}_{-0.1}----
JADES-GS-210497603:32:25.12−27-27:51:42.7412.1−0.3+0.812.1^{+0.8}_{-0.3}−2.38−0.35+0.32-2.38^{+0.32}_{-0.35}−18.0−0.1+0.1-18.0^{+0.1}_{-0.1}----
CEERS-227413614:19:12.35+52:51:41.8112.1−0.4+0.612.1^{+0.6}_{-0.4}−3.62−0.7+0.57-3.62^{+0.57}_{-0.7}−18.9−0.0+0.1-18.9^{+0.1}_{-0.0}----
PAN-1343+5549-200035013:43:39.91+55:46:45.7612.1−0.4+1.112.1^{+1.1}_{-0.4}−2.77−0.36+0.35-2.77^{+0.35}_{-0.36}−21.3−0.1+0.1-21.3^{+0.1}_{-0.1}----
PAN-1707+5852-201006817:07:17.36+58:51:11.0612.1−0.4+1.312.1^{+1.3}_{-0.4}−2.28−0.3+0.28-2.28^{+0.28}_{-0.3}−21.6−0.2+0.1-21.6^{+0.1}_{-0.2}----
GLASS-203905200:13:59.75−30-30:19:29.1412.2−0.1+0.012.2^{+0.0}_{-0.1}−2.53−0.25+0.25-2.53^{+0.25}_{-0.25}−21.3−0.0+0.0-21.3^{+0.0}_{-0.0}1.212.34−2.78−0.06+0.06-2.78^{+0.06}_{-0.06}Castellano et al. (2024), μ\mu: Furtak et al. (2023)
JADES-GN-314571312:37:30.39+62:15:15.6812.2−0.2+0.712.2^{+0.7}_{-0.2}−2.57−0.32+0.28-2.57^{+0.28}_{-0.32}−19.9−0.1+0.1-19.9^{+0.1}_{-0.1}----
CEERS-East-116041114:20:03.15+52:52:30.4912.2−0.4+0.712.2^{+0.7}_{-0.4}−2.41−0.32+0.32-2.41^{+0.32}_{-0.32}−19.2−0.1+0.1-19.2^{+0.1}_{-0.1}----
CEERS-226417114:19:27.31+52:51:29.2012.3−0.4+0.812.3^{+0.8}_{-0.4}−2.28−0.35+0.33-2.28^{+0.33}_{-0.35}−19.3−0.1+0.1-19.3^{+0.1}_{-0.1}----
BEA-0217-0503-205273902:17:13.11−05-05:01:50.5412.3−0.4+0.812.3^{+0.8}_{-0.4}−2.77−0.56+0.44-2.77^{+0.44}_{-0.56}−19.0−0.1+0.1-19.0^{+0.1}_{-0.1}----
CEERS-East-108801814:20:12.28+52:48:32.6612.3−0.5+0.912.3^{+0.9}_{-0.5}−2.85−0.48+0.4-2.85^{+0.4}_{-0.48}−18.4−0.1+0.1-18.4^{+0.1}_{-0.1}----
PAN-1237+6210-101612512:37:46.16+62:11:49.7212.3−0.5+1.412.3^{+1.4}_{-0.5}−1.96−0.39+0.35-1.96^{+0.35}_{-0.39}−19.5−0.3+0.2-19.5^{+0.2}_{-0.3}----
J1235-205131612:35:53.39+04:55:53.0412.3−0.6+1.112.3^{+1.1}_{-0.6}−2.36−0.37+0.42-2.36^{+0.42}_{-0.37}−18.8−0.2+0.2-18.8^{+0.2}_{-0.2}----
PAN-0144+1714-204210501:44:22.44+17:15:31.3612.4−0.3+0.712.4^{+0.7}_{-0.3}−3.05−0.36+0.28-3.05^{+0.28}_{-0.36}−19.1−0.1+0.1-19.1^{+0.1}_{-0.1}----
PAN-1000+0207-203275110:00:24.17+02:08:31.0112.5−0.5+0.912.5^{+0.9}_{-0.5}−2.48−0.4+0.35-2.48^{+0.35}_{-0.4}−20.1−0.1+0.1-20.1^{+0.1}_{-0.1}----
CEERS-East-302442314:19:49.00+52:44:06.5712.5−0.5+1.412.5^{+1.4}_{-0.5}−1.89−0.4+0.25-1.89^{+0.25}_{-0.4}−19.5−0.1+0.4-19.5^{+0.4}_{-0.1}----
NGDEEP-MIDIS-204371803:32:58.43−27-27:51:02.9012.6−0.4+0.412.6^{+0.4}_{-0.4}−2.72−0.41+0.32-2.72^{+0.32}_{-0.41}−17.6−0.1+0.1-17.6^{+0.1}_{-0.1}----
SPT0615-102023606:15:41.93−57-57:43:59.7912.6−0.5+0.512.6^{+0.5}_{-0.5}−2.56−0.33+0.32-2.56^{+0.32}_{-0.33}−19.8−0.1+0.1-19.8^{+0.1}_{-0.1}1.2--μ\mu: Paterno-Mahler et al. (2018), Zitrin et al. (2015)
BEA-2303-6250-203980223:03:34.42−62-62:50:11.2612.6−0.5+0.712.6^{+0.7}_{-0.5}−2.61−0.25+0.25-2.61^{+0.25}_{-0.25}−21.1−0.1+0.1-21.1^{+0.1}_{-0.1}----
PRIMER-UDS-256439602:17:10.08−05-05:10:26.8912.7−0.2+0.312.7^{+0.3}_{-0.2}−2.85−0.25+0.26-2.85^{+0.26}_{-0.25}−20.1−0.1+0.1-20.1^{+0.1}_{-0.1}----
NGDEEP-MIDIS-205194003:33:02.18−27-27:50:20.1212.7−0.4+0.312.7^{+0.3}_{-0.4}−2.39−0.33+0.33-2.39^{+0.33}_{-0.33}−17.3−0.0+0.1-17.3^{+0.1}_{-0.0}----
NGDEEP-MIDIS-205128003:33:00.74−27-27:50:21.4112.7−0.5+0.412.7^{+0.4}_{-0.5}−2.35−0.25+0.25-2.35^{+0.25}_{-0.25}−18.0−0.1+0.1-18.0^{+0.1}_{-0.1}----
NGDEEP-MIDIS-204180903:32:56.97−27-27:51:10.6512.8−0.3+0.312.8^{+0.3}_{-0.3}−2.27−0.26+0.25-2.27^{+0.25}_{-0.26}−17.5−0.1+0.1-17.5^{+0.1}_{-0.1}----
MACS0416-100391404:16:16.35−24-24:07:41.7212.8−0.4+0.412.8^{+0.4}_{-0.4}−2.59−0.25+0.25-2.59^{+0.25}_{-0.25}−19.3−0.1+0.1-19.3^{+0.1}_{-0.1}----
BEA-2303-6250-100677923:03:57.79−62-62:52:17.4812.8−0.6+0.812.8^{+0.8}_{-0.6}−2.79−0.49+0.47-2.79^{+0.47}_{-0.49}−19.3−0.2+0.1-19.3^{+0.1}_{-0.2}----
CEERS-East-309230914:20:06.05+52:48:15.1812.9−0.3+0.312.9^{+0.3}_{-0.3}−3.12−0.28+0.25-3.12^{+0.25}_{-0.28}−19.3−0.0+0.0-19.3^{+0.0}_{-0.0}----
JADES-GS-204298603:32:15.54−27-27:53:24.8612.9−0.4+0.412.9^{+0.4}_{-0.4}−2.37−0.25+0.25-2.37^{+0.25}_{-0.25}−18.4−0.1+0.1-18.4^{+0.1}_{-0.1}-13.01−3.31−0.58+0.45-3.31^{+0.45}_{-0.58}Witstok et al. (2025)
NGDEEP-MIDIS-204862003:32:58.20−27-27:50:38.7212.9−0.5+0.412.9^{+0.4}_{-0.5}−3.43−0.37+0.37-3.43^{+0.37}_{-0.37}−17.7−0.0+0.0-17.7^{+0.0}_{-0.0}----
UNCOVER-219291800:14:12.63−30-30:21:00.7612.9−0.5+0.512.9^{+0.5}_{-0.5}−2.45−0.31+0.28-2.45^{+0.28}_{-0.31}−18.7−0.1+0.1-18.7^{+0.1}_{-0.1}1.9--μ\mu: Furtak et al. (2023)
BEA-0446-2636-203480504:46:42.76−26-26:34:42.7512.9−0.6+0.512.9^{+0.5}_{-0.6}−2.05−0.29+0.28-2.05^{+0.28}_{-0.29}−19.9−0.2+0.1-19.9^{+0.1}_{-0.2}----
NEP-TDF-203494517:22:57.48+65:46:29.0912.9−0.8+0.712.9^{+0.7}_{-0.8}−2.93−0.33+0.26-2.93^{+0.26}_{-0.33}−19.6−0.1+0.1-19.6^{+0.1}_{-0.1}----
PRIMER-UDS-213885402:17:01.80−05-05:17:18.6713.0−0.4+0.413.0^{+0.4}_{-0.4}−3.72−0.48+0.4-3.72^{+0.4}_{-0.48}−19.8−0.0+0.0-19.8^{+0.0}_{-0.0}----
BEA-2303-6250-202727723:03:47.47−62-62:51:07.3413.0−0.6+0.513.0^{+0.5}_{-0.6}−3.22−0.37+0.32-3.22^{+0.32}_{-0.37}−20.0−0.1+0.1-20.0^{+0.1}_{-0.1}----
JADES-GS-208460003:32:11.05−27-27:52:10.6613.0−0.7+0.613.0^{+0.6}_{-0.7}−1.76−0.25+0.25-1.76^{+0.25}_{-0.25}−18.8−0.2+0.2-18.8^{+0.2}_{-0.2}----
BEA-2058-4247-206717520:58:33.74−42-42:44:08.6113.1−0.4+0.513.1^{+0.5}_{-0.4}−3.31−0.5+0.48-3.31^{+0.48}_{-0.5}−21.2−0.1+0.1-21.2^{+0.1}_{-0.1}----
PLCKG165-204750511:27:13.98+42:28:56.1613.1−0.6+0.413.1^{+0.4}_{-0.6}−3.03−0.3+0.25-3.03^{+0.25}_{-0.3}−20.0−0.1+0.1-20.0^{+0.1}_{-0.1}3.2--μ\mu: Kamieneski et al. (2024)
MACS0416-100391204:16:16.33−24-24:07:41.7213.2−0.5+0.313.2^{+0.3}_{-0.5}−3.56−0.58+0.55-3.56^{+0.55}_{-0.58}−19.0−0.1+0.1-19.0^{+0.1}_{-0.1}----
IDRADeczDynestyz_{\mathrm{Dynesty}}β\betaMUVM_{\mathrm{UV}}μ\muzspecz_{\mathrm{spec}}βspec\beta_{\mathrm{spec}}Ref.
JADES-GS-107487303:32:05.31−27-27:51:26.0513.2−0.8+0.613.2^{+0.6}_{-0.8}−2.5−0.39+0.31-2.5^{+0.31}_{-0.39}−19.4−0.2+0.1-19.4^{+0.1}_{-0.2}----
GLASS-201420400:14:03.26−30-30:21:24.4013.3−0.8+0.513.3^{+0.5}_{-0.8}−2.68−0.34+0.28-2.68^{+0.28}_{-0.34}−19.9−0.1+0.1-19.9^{+0.1}_{-0.1}1.7--μ\mu: Furtak et al. (2023)
JADES-GS-106138603:32:20.33−27-27:52:00.0213.3−1.3+0.713.3^{+0.7}_{-1.3}−2.05−0.4+0.37-2.05^{+0.37}_{-0.4}−18.2−0.3+0.3-18.2^{+0.3}_{-0.3}----
CEERS-250758414:20:17.83+52+52:57:04.1213.4−0.7+0.513.4^{+0.5}_{-0.7}−2.72−0.41+0.34-2.72^{+0.34}_{-0.41}−19.2−0.2+0.1-19.2^{+0.1}_{-0.2}----
J1235-206231412:35:55.12+04+04:56:51.5013.4−0.8+0.513.4^{+0.5}_{-0.8}−2.78−0.4+0.34-2.78^{+0.34}_{-0.4}−19.4−0.2+0.1-19.4^{+0.1}_{-0.2}----
JADES-GS-205108803:32:17.83−27-27:53:09.3013.9−0.6+0.313.9^{+0.3}_{-0.6}−2.62−0.38+0.36-2.62^{+0.36}_{-0.38}−19.1−0.1+0.2-19.1^{+0.2}_{-0.1}-13.90−2.95−0.19+0.18-2.95^{+0.18}_{-0.19}Carniani et al. (2024)
PAN-2216+0024-300365822:17:00.12+00+00:22:44.6914.3−0.2+0.114.3^{+0.1}_{-0.2}−2.82−0.25+0.25-2.82^{+0.25}_{-0.25}−21.2−0.1+0.1-21.2^{+0.1}_{-0.1}-13.53−2.29−0.1+0.1-2.29^{+0.1}_{-0.1}Donnan et al. (2026b)
PRIMER-COS-233218110:00:22.40+02+02:16:23.3514.5−0.3+0.314.5^{+0.3}_{-0.3}−3.08−0.33+0.31-3.08^{+0.31}_{-0.33}−20.0−0.1+0.1-20.0^{+0.1}_{-0.1}-14.44−2.50−0.18+0.19-2.50^{+0.19}_{-0.18}Naidu et al. (2026)
JADES-GN-228504712:36:40.16+62+62:18:36.9414.8−0.3+0.214.8^{+0.2}_{-0.3}−2.9−0.34+0.3-2.9^{+0.3}_{-0.34}−20.0−0.1+0.1-20.0^{+0.1}_{-0.1}----
PAN-2345-4235-302223523:45:15.45−42-42:34:57.0715.2−0.3+0.215.2^{+0.2}_{-0.3}−2.94−0.29+0.28-2.94^{+0.28}_{-0.29}−20.9−0.1+0.1-20.9^{+0.1}_{-0.1}----
PAN-0332-2754-203053803:32:41.21−27-27:53:45.0015.2−0.3+0.315.2^{+0.3}_{-0.3}−3.44−0.37+0.38-3.44^{+0.38}_{-0.37}−19.2−0.1+0.1-19.2^{+0.1}_{-0.1}----

Table C1.

References

  1. [7]Bakx T. J. L. C., et al., 2020, MNRAS, 493, 4294
  2. [8]Bakx T. J. L. C., et al., 2023, MNRAS, 519, 5076
  3. [9]Bakx T. J. L. C., et al., 2025, MNRAS, 544, 1502

Paper details

vlnc:W_475y447u
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents