Star Formation at the Epoch of Reionization with CANUCS: The Ages of Stellar Populations in MACS1149-JD1

We present measurements of stellar populations properties of a z = 9.1 gravitationally lensed galaxy MACS1149-JD1 using deep James Webb Space Telescope NIRISS slitless spectroscopy as well as NIRISS and NIRCam imaging from the CAnadian NIRISS Unbiased Cluster Survey (CANUCS). The galaxy is split into four components. Three magnified (μ ∼ 11) star-forming components are unresolved, giving intrinsic sizes of <25 pc. In addition, the underlying extended component contains the bulk of the stellar mass, formed the majority of its stars ∼50 Myr earlier than the other three components, and is not the site of the most active star formation currently. The NIRISS and NIRCam resolved photometry does not confirm a strong Balmer break previously seen in Spitzer. The NIRISS grism spectrum has been extracted for the entire galaxy and shows a clear continuum and Lyman break, with no Lyα detected.


INTRODUCTION
Tracing star formation to the earliest times has been a long-standing goal of extragalactic astronomy.In particular, studying the onset of star formation is of importance not only for galaxy formation models but also for studies of the early universe.Spitzer and the Hubble Space Telescope (HST ) played a unique role in determining the onset of star formation of galaxies at redshift z ≳ 6 (e.g., Bradač 2020 for a review).
The James Webb Space Telescope (JWST , Gardner et al. 2023) is revolutionizing studies of the early onset of star formation in high-redshift galaxies.With the expanded sensitivity, filter set and wavelength cover-age compared to Spitzer, JWST can trace the full spectral energy distribution, and in some cases distinguish strong line emission from breaks due to evolved stars (e.g., Laporte et al. 2023).Early results from JWST appear to show a higher than expected ultraviolet luminosity density at z > 10 ( Harikane et al. 2023;Donnan et al. 2023).There have also been claims for the presence at 7 < z < 9 of massive galaxies with strong Balmer breaks in the CEERS survey (Labbé et al. 2023;Lovell et al. 2023;Boylan-Kolchin 2023).However, studies from other JWST surveys with comparable or larger volumes such as JADES, EPOCHS and CANUCS do not find such a high density of massive galaxies with strong arXiv:2308.13288v1[astro-ph.GA] 25 Aug 2023 Balmer breaks at these redshifts (Endsley et al. 2023b;Trussler et al. 2023;Desprez et al. 2023, in prep.).
One of the most intriguing objects showing a potential Balmer break from previous HST and Spitzer studies is the z = 9.1 galaxy MACS1149-JD1 behind the cluster MACS J1149.5+2223.MACS1149-JD1 was originally discovered in HST and shallow Spitzer data in Zheng et al. (2012).It was later detected in both channel 1 and channel 2 Spitzer bands using deeper data (Bradač et al. 2014;Huang et al. 2016;Zheng et al. 2017;Hoag et al. 2018) and its redshift was spectroscopically measured with the [O III] 88 µm line using ALMA by Hashimoto et al. (2018).With early data, it was concluded that the nebular emission lines are redshifted out of both Spitzer bands (at z > 9), yet the galaxy showed a strong color excess.It was therefore highly likely that old (∼ 300Myr) stellar populations are causing the red rest-frame optical colors (Hashimoto et al. 2018;Hoag et al. 2018;Huang et al. 2016).This was surprising, given the galaxy would need to start forming a significant amount of stars shortly after the Big Bang (∼ 250Myr).In addition, the cold dust content of the galaxy was constrained to be modest from observations taken with ALMA, making dust an unlikely cause of red Spitzer color (Hashimoto et al. 2018).
MACS1149-JD1 was recently observed as part of the CAnadian NIRISS Unbiased Cluster Survey (CANUCS, Willott et al. 2022) with the NIRCam and NIRISS instruments onboard JWST .The data provides superior photometry compared to what was possible with Spitzer .In addition, NIRISS spectra with its coverage from 1 to 2.5 µm allow us to investigate the rest frame UV spectrum, including searching for the presence of potential Lyman-α line (previously mentioned in Hashimoto et al. 2018).Here we describe these data and analysis of the stellar properties of MACS1149-JD1.
The paper is structured as follows.In Section 2 we present the data used in this paper and in Section 3 we describe the analysis of the photometric and spectroscopic data.In Section 4 we present the main science results.We summarize in Section 5 and give photometry and SED fitting results in the tables in Appendix A.
To reduce the imaging data we use the photometric pipeline that will be presented in Brammer et al. (in prep.), which also provides a compilation of the JWST ERO photometric data released to date.Briefly, the raw data has been reduced using the public Grism redshift & line analysis software Grizli (Brammer 2023a,b), which masks imaging artifacts, provides astrometric calibrations based on the Gaia Data Release 3 catalog, and shifts images using Astrodrizzle.The method closely follows the one outlined in Valentino et al. (2023).We show the cutouts of MACS1149-JD1 in Fig. 1.
Observations also consist of two NIRISS pointings, one centred on the cluster centre containing MACS1149-JD1 and the other coincident with a flanking field.Each pointing is observed with the GR150R and GR150C grisms through the F115W, F150W and F200W filters.Exposure times for the cluster field are 19240 seconds in each of the three filters.We also process all the NIRISS imaging and slitless spectroscopy with Grizli.Grizli performs full end-to-end processing of space-based slitless spectroscopic datasets.For full details see e.g., Matharu et al. (2021); Noirot et al. (2023); Matharu et al. (2023).In summary, raw data is downloaded from the Mikulski Archive for Space Telescopes (MAST) and pre-processed for cosmic rays, flat-fielding, sky subtraction, astrometric corrections and alignment.Contamination models (which correct for overlapping spectra from nearby sources) for each pointing are then generated and subtracted for each grism spectrum of interest.From these images we extract the spectrum of MACS1149-JD1.

Photometry and SED Fitting
The photometry is derived using the updated zeropoints, and corrected for Milky Way extinction.We use F150W and F200W NIRISS filters as well as F150W, F200W, F277W, F356W, F410M, and F444W NIR-Cam filters (in other JWST filters MACS1149-JD1 is not/barely detected) and HST upper limits for the entire source.Since the object is resolved into three distinct clumps and a smooth galaxy component, we perform a photometric fit using Galfit (Peng et al. 2010).We forward model the source assuming four components (three point sources for the clumps and a sersic profile for the diffused light), convolve them with the PSF and determine their parameters to minimise the residuals.We measure empirical PSF determined from the stars.
The resulting models and residuals are shown in Fig. 1.Residuals from the fits are negligible, confirming the original visual impression that the three compact sources are unresolved and an additional smooth component is present.The agreement between NIRISS and NIRCam fluxes in the two overlapping filters is a confirmation of the robustness of photometry.Resolved photometry is necessary, as global spectral energy distribution (SED) fitting can bias stellar masses when young stellar population outshine the first episodes of star formation (e.g., Sorba & Sawicki 2018;Giménez-Arteaga et al. 2023;Narayanan et al. 2023).Photometric properties are given in Table 1.
SEDs derived from our photometry were analysed using the DENSE BASIS method (Iyer & Gawiser 2017;Iyer et al. 2019) to determine nonparametric star formation histories (SFHs), masses, and ages for our sources in MACS1149-JD1.We adopt the Calzetti attenuation law (Calzetti 2001) and a Chabrier IMF (Chabrier 2003).We fix the redshift to that found by the [O III] 88 µm line in Hashimoto et al. (2018), 9.1096 ± 0.0006.All other parameters are left free.The primary advantage of using DENSE BASIS with nonparametric SFHs is that they allow us to account for flexible stellar populations.Both photometry and SED fit are shown in Fig. 2.

Grism spectroscopy
To extract the NIRISS spectrum of the source we also use the Grizli package.The Grizli reduction steps 1 https://archive.stsci.edu/prepds/frontier/lensmodels/Arnouts 1996) python wrapper sep (Barbary 2016), using the default detection parameters implemented in Grizli (a detection threshold 'threshold' of 1.8σ above the global background RMS, a minimum source area in pixels 'minarea' of 9, and deblending parameters 'deblend_cont' and 'deblend_ntresh' of 0.001 and 32, respectively).Matched aperture photometry on the available NIRISS filters is performed at the same stage.From this NIRISS imaging catalogue, the position of sources that contaminate the spectrum of MACS1149-JD1 are used to locate spectral traces in the grism data.The spectral continua of the sources are modelled using an iterative polynomial fitting of the data for contamination estimate and removal.The 2D and extracted 1D MACS1149-JD1 spectra with contamination removal and modelled spectrum are shown in Fig. 3.

Spatially Resolved Star Formation History
Using our photometry (Table 1) we now determine the stellar properties of each individual component.By determining β UV slopes based on NIRCam F150W, F200W, and F227W fluxes, we see that the three clumps have different properties than the underlying galaxy component.The three clumps have β UV,phot measured between −2.5 and, −2.8, whereas the galaxy itself is redder with β UV,phot = −1.9± 0.2 (Table 2).The values are consistent with with other observations from JWST (e.g., Topping et al. 2023;Endsley et al. 2023a;Bouwens et al. 2023;Franco et al. 2023).
Using DENSE BASIS we also perform the SED fit and determine nonparametric star formation histories.All four components have intrinsic (corrected for magnification) stellar masses between 5×10 6 and 10 8 M ⊙ and star formation rates (SFR) between 0.2 − 1M ⊙ yr −1 .While the errorbars are large and there is still a possibility that all components have the same star formation histories, there is nevertheless a hint that the galaxy itself (G) has started to form the bulk of the stars earlier (Fig. 4).This component also has the highest stellar mass (Fig. 5).In Fig. 6 we also show the mass fraction of stars formed as a function of lookback time.Once again, the error-bars are large, but there is an indication that the underlying galaxy has formed the bulk of its stellar masses earlier than the clumps, which are still actively star-forming.
The galaxy formed 50% of its total mass at t 50 = 134 +89 −82 Myr.In Hashimoto et al. (2018), the bulk of the stellar population was determined to have formed at a look-back time of ∼ 250Myr.The main reason is that the relative flux measured red-ward of ∼ 4000Å has decreased, making the potential Balmer-break less pronounced.We measure the Balmer break of galaxy G (based on fluxes in F444W and F277W, the latter being mostly emission line free) of ∆mag AB = 0.3 ± 0.2mag (F ν (F444W)/F ν (F277W) = 1.4 ± 0.2).This is lower compared to Spitzer measurements from Kokorev et al. (2022) of ∆mag AB = 0.5 ± 0.2, from Zheng et al. (2017) (also used in Hashimoto et al. 2018) ∆mag AB > 1.3 (1-σ) , from ASTRODEEP (Di Criscienzo et al. 2017) ∆mag AB > 0.7 (3-σ) and from (Huang et al. 2016) ∆mag AB = 0.8 ± 0.4.All Spitzer measurements are the average of the older G component and all the clumps, though the former dominates the flux.This discrepancy is unlikely caused by emission lines, as both NIRCam F444W and Spitzer Channel 2 have similar throughputs at the red-end, hence entering [O iii] λ4959 emission line (where both instruments have a throughput of 20%) could not play a role (for [O iii] λ5007 both are similarly at 1%).We think the most likely source of discrepancy is the contamination modelling which in the case of Spitzer 's large PSF is difficult.

Grism Spectrum
We clearly detect the continuum and Lyman-break in the NIRISS spectrum at the expected redshift.In Fig. 3 we show 1D spectral extraction with a fitted model The region where contamination subtraction failed is marked in red and the region between half power wavelengths at which the transmission in each filter falls below 50% of its peak value is marked in grey.
at the redshift determined by Hashimoto et al. (2018).However, even if we let the redshift be determined by the NIRISS data alone, we still recover the same redshift (z = 9.2 ± 0.1).We have also searched for the Lymanα emission line that was indicated in Hashimoto et al. (2018) at 12, 267.4Å with an integrated (lensed) flux of 4.3±1.1×10−18 erg s −1 cm −2 .We do not detect any lines at that wavelength.From the sensitivity of our observations, such a line would have been detected.We do, however, detect a line in one orientation PA = 212 deg at 17, 700Å which corresponds to N iii] λ1747,1749, with the flux of 4.6 ± 0.6 × 10 −18 erg s −1 cm −2 .Unfortunately, the other orientation is contaminated and furthermore, the spectrum is located towards the edge of the detector.Hence, we consider this line tentative.
The combined UV beta slope measured from the NIRISS spectrum between rest-frame wavelengths of 1400 − 1600Å and 1800 − 2000Å (we assume a simi-lar spectral range as used for photometry, excluding the part of the spectrum in the detector gap) is β UV,spec = −2.3±0.5.This is consistent with the average photometric measurements done for individual clumps (Table .2).
The spectrum also shows a softening of the Lymanbreak in the vicinity of Lyman-α, very likely caused by a largely neutral IGM (Mason & Gronke 2020;Curtis-Lake et al. 2023;Heintz et al. 2023).Unfortunately, the break falls at the gap between the two filters, hence we cannot characterize it fully.

CONCLUSIONS
The gravitationally lensed galaxy MACS1149-JD1 at z = 9.1096 ± 0.0006 has been well studied in the past.Spitzer data were showing what seemed to be a strong Balmer break, meaning that the dominant stellar component formed about 290Myr earlier (or around 240Myr   2).
New JWST observations with NIRISS and NIRCam reveal that the galaxy consists of three unresolved (with intrinsic sizes < 50pc) star-forming clumps and an underlying extended galaxy component.We individually perform SED fitting of all four components (Fig. 2, Table 2).The galaxy component (G) is showing somewhat older stellar population, albeit with large errobars.This component (i) contains the bulk of the stellar mass, (ii) likely formed the majority of its stars ∼ 50Myr earlier than the other components and (iii) is not the site of the most recent star formation.NIRISS spectrum of MACS1149-JD1 shows a clear detection of the continuum and Lyman-break.However, we do not detect the Lyman-α line previously reported in Hashimoto et al. (2018).Given that NIRISS spectra have low spectral resolution Lyman-α could still be present, though at a lower flux than previously reported.
In conclusion, MACS1149-JD1 is a highly magnified, intrinsically faint galaxy at z = 9.1.It shows properties that are consistent with other galaxies detected with JWST (e.g., Bunker et al. 2023;Topping et al. 2023); however, its true nature was only revealed through resolved SED fitting.While strong Balmer breaks can be present at high redshift, they are rare (Laporte et al.

APPENDIX
A. PHOTOMETRY AND SED FITTING In Table 1 we list photometry and in Table 2 derived quantities and results of SED fitting of MACS1149-JD1.All the procedures are described in the main text.

Figure 1 .
Figure 1.Images of the three clumps and underlying galaxy component in MACS1149-JD1.Shown are different filters, Galfit models and residuals (top panels), RGB (always using F150W, F277W, and F444W filters) model of the clumps and the galaxy (middle panels) and RGB model and RGB NIRCam image (bottom panels).The galaxy consists of at least 3 clumps (all marked in the middle panel), all of which are unresolved and an underlying smooth component.An upper limit to the magnified size (FWHM) of the clumps is 0.05 ′′ .The intrinsic (demagnified) size upper limit is < 50pc.

Figure 2 .
Figure 2. Results of the SED fitting for the three clumps (labelled 1-3) and the smooth light component (G).Shown are measured fluxes (i.e., we do not correct them for magnification) for both NIRCam and NIRISS imaging in red and SED predicted fluxes in open circles in units of µJy.Derived stellar properties are given in the inset.

Figure 3 .
Figure 3. NIRISS grism spectrum of MACS1149-JD1.Top: 2D spectrum of MACS1149-JD1 is shown in two orientations (PA = 212 deg top and 302 deg middle) and three filters.Direct images are also shown.The bottom row shows a combined spectrum with continuum emission from the object subtracted.All images are contamination subtracted, the residual contamination is from objects below the imaging detection threshold and objects outside the FoV of the direct imaging.Bottom: 1D spectrum extracted (black line with uncertainties in blue) and modelled given fixed redshift (red line).The positions of potential lines are marked.Only N iii] λ1747,1749 line is possibly detected in PA = 212 deg; the other orientation is contaminated and the spectrum falls on the edge of the detector (see top).The region where contamination subtraction failed is marked in red and the region between half power wavelengths at which the transmission in each filter falls below 50% of its peak value is marked in grey.

Figure 4 .
Figure 4. Star formation histories for the four components.While the three star-forming clumps have similar star-formation histories, the underlying galaxy component is different.

Figure 5 .
Figure 5. Specific star formation rate (sSFR) vs. stellar mass (M * ) plot for the four components.All three unresolved components show similar stellar ages, while the underlying galaxy component shows an older stellar population (albeit with large errorbars, see also Table2).

Figure 6 .
Figure 6.Mass fraction of stars formed as a function of look-back time for all four components (1-blue, 2-orange, 3green, galaxy-red).In Hashimoto et al. 2018 the authors conclude that the bulk of the stellar mass was produced within a short period corresponding to the redshift interval 12 < z < 16, with a dominant stellar component that formed at the look-back time of ∼ 290Myr.These new measurements show somewhat younger ages, with the oldest component (G) forming 50% of its total mass at t50 = 134 +89 −82 .

Table 2 .
Global properties of MACS1149-JD1 and stellar properties of individual clumps.